Commit | Line | Data |
---|---|---|
7c1a000d | 1 | // SPDX-License-Identifier: GPL-2.0 |
0a8165d7 | 2 | /* |
e05df3b1 JK |
3 | * fs/f2fs/node.c |
4 | * | |
5 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
6 | * http://www.samsung.com/ | |
e05df3b1 JK |
7 | */ |
8 | #include <linux/fs.h> | |
9 | #include <linux/f2fs_fs.h> | |
10 | #include <linux/mpage.h> | |
11 | #include <linux/backing-dev.h> | |
12 | #include <linux/blkdev.h> | |
13 | #include <linux/pagevec.h> | |
14 | #include <linux/swap.h> | |
15 | ||
16 | #include "f2fs.h" | |
17 | #include "node.h" | |
18 | #include "segment.h" | |
87905682 | 19 | #include "xattr.h" |
9e4ded3f | 20 | #include "trace.h" |
51dd6249 | 21 | #include <trace/events/f2fs.h> |
e05df3b1 | 22 | |
4d57b86d | 23 | #define on_f2fs_build_free_nids(nmi) mutex_is_locked(&(nm_i)->build_lock) |
f978f5a0 | 24 | |
e05df3b1 JK |
25 | static struct kmem_cache *nat_entry_slab; |
26 | static struct kmem_cache *free_nid_slab; | |
aec71382 | 27 | static struct kmem_cache *nat_entry_set_slab; |
50fa53ec | 28 | static struct kmem_cache *fsync_node_entry_slab; |
e05df3b1 | 29 | |
a4f843bd JK |
30 | /* |
31 | * Check whether the given nid is within node id range. | |
32 | */ | |
4d57b86d | 33 | int f2fs_check_nid_range(struct f2fs_sb_info *sbi, nid_t nid) |
a4f843bd JK |
34 | { |
35 | if (unlikely(nid < F2FS_ROOT_INO(sbi) || nid >= NM_I(sbi)->max_nid)) { | |
36 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
37 | f2fs_msg(sbi->sb, KERN_WARNING, | |
38 | "%s: out-of-range nid=%x, run fsck to fix.", | |
39 | __func__, nid); | |
40 | return -EINVAL; | |
41 | } | |
42 | return 0; | |
43 | } | |
44 | ||
4d57b86d | 45 | bool f2fs_available_free_memory(struct f2fs_sb_info *sbi, int type) |
cdfc41c1 | 46 | { |
6fb03f3a | 47 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
cdfc41c1 | 48 | struct sysinfo val; |
e5e7ea3c | 49 | unsigned long avail_ram; |
cdfc41c1 | 50 | unsigned long mem_size = 0; |
6fb03f3a | 51 | bool res = false; |
cdfc41c1 JK |
52 | |
53 | si_meminfo(&val); | |
e5e7ea3c JK |
54 | |
55 | /* only uses low memory */ | |
56 | avail_ram = val.totalram - val.totalhigh; | |
57 | ||
429511cd CY |
58 | /* |
59 | * give 25%, 25%, 50%, 50%, 50% memory for each components respectively | |
60 | */ | |
6fb03f3a | 61 | if (type == FREE_NIDS) { |
9a4ffdf5 | 62 | mem_size = (nm_i->nid_cnt[FREE_NID] * |
b8559dc2 | 63 | sizeof(struct free_nid)) >> PAGE_SHIFT; |
e5e7ea3c | 64 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); |
6fb03f3a | 65 | } else if (type == NAT_ENTRIES) { |
e5e7ea3c | 66 | mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >> |
09cbfeaf | 67 | PAGE_SHIFT; |
e5e7ea3c | 68 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); |
e589c2c4 JK |
69 | if (excess_cached_nats(sbi)) |
70 | res = false; | |
a1257023 JK |
71 | } else if (type == DIRTY_DENTS) { |
72 | if (sbi->sb->s_bdi->wb.dirty_exceeded) | |
73 | return false; | |
74 | mem_size = get_pages(sbi, F2FS_DIRTY_DENTS); | |
75 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); | |
e5e7ea3c JK |
76 | } else if (type == INO_ENTRIES) { |
77 | int i; | |
78 | ||
39d787be | 79 | for (i = 0; i < MAX_INO_ENTRY; i++) |
8f73cbb7 KM |
80 | mem_size += sbi->im[i].ino_num * |
81 | sizeof(struct ino_entry); | |
82 | mem_size >>= PAGE_SHIFT; | |
e5e7ea3c | 83 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
429511cd | 84 | } else if (type == EXTENT_CACHE) { |
7441ccef JK |
85 | mem_size = (atomic_read(&sbi->total_ext_tree) * |
86 | sizeof(struct extent_tree) + | |
429511cd | 87 | atomic_read(&sbi->total_ext_node) * |
09cbfeaf | 88 | sizeof(struct extent_node)) >> PAGE_SHIFT; |
429511cd | 89 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
57864ae5 JK |
90 | } else if (type == INMEM_PAGES) { |
91 | /* it allows 20% / total_ram for inmemory pages */ | |
92 | mem_size = get_pages(sbi, F2FS_INMEM_PAGES); | |
93 | res = mem_size < (val.totalram / 5); | |
1e84371f | 94 | } else { |
1663cae4 JK |
95 | if (!sbi->sb->s_bdi->wb.dirty_exceeded) |
96 | return true; | |
6fb03f3a JK |
97 | } |
98 | return res; | |
cdfc41c1 JK |
99 | } |
100 | ||
e05df3b1 JK |
101 | static void clear_node_page_dirty(struct page *page) |
102 | { | |
e05df3b1 | 103 | if (PageDirty(page)) { |
4d57b86d | 104 | f2fs_clear_radix_tree_dirty_tag(page); |
e05df3b1 | 105 | clear_page_dirty_for_io(page); |
aec2f729 | 106 | dec_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 JK |
107 | } |
108 | ClearPageUptodate(page); | |
109 | } | |
110 | ||
111 | static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
112 | { | |
80551d17 | 113 | return f2fs_get_meta_page_nofail(sbi, current_nat_addr(sbi, nid)); |
e05df3b1 JK |
114 | } |
115 | ||
116 | static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
117 | { | |
118 | struct page *src_page; | |
119 | struct page *dst_page; | |
e05df3b1 JK |
120 | pgoff_t dst_off; |
121 | void *src_addr; | |
122 | void *dst_addr; | |
123 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
124 | ||
80551d17 | 125 | dst_off = next_nat_addr(sbi, current_nat_addr(sbi, nid)); |
e05df3b1 JK |
126 | |
127 | /* get current nat block page with lock */ | |
80551d17 | 128 | src_page = get_current_nat_page(sbi, nid); |
4d57b86d | 129 | dst_page = f2fs_grab_meta_page(sbi, dst_off); |
9850cf4a | 130 | f2fs_bug_on(sbi, PageDirty(src_page)); |
e05df3b1 JK |
131 | |
132 | src_addr = page_address(src_page); | |
133 | dst_addr = page_address(dst_page); | |
09cbfeaf | 134 | memcpy(dst_addr, src_addr, PAGE_SIZE); |
e05df3b1 JK |
135 | set_page_dirty(dst_page); |
136 | f2fs_put_page(src_page, 1); | |
137 | ||
138 | set_to_next_nat(nm_i, nid); | |
139 | ||
140 | return dst_page; | |
141 | } | |
142 | ||
12f9ef37 YH |
143 | static struct nat_entry *__alloc_nat_entry(nid_t nid, bool no_fail) |
144 | { | |
145 | struct nat_entry *new; | |
146 | ||
147 | if (no_fail) | |
2882d343 | 148 | new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO); |
12f9ef37 | 149 | else |
2882d343 | 150 | new = kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO); |
12f9ef37 YH |
151 | if (new) { |
152 | nat_set_nid(new, nid); | |
153 | nat_reset_flag(new); | |
154 | } | |
155 | return new; | |
156 | } | |
157 | ||
158 | static void __free_nat_entry(struct nat_entry *e) | |
159 | { | |
160 | kmem_cache_free(nat_entry_slab, e); | |
161 | } | |
162 | ||
163 | /* must be locked by nat_tree_lock */ | |
164 | static struct nat_entry *__init_nat_entry(struct f2fs_nm_info *nm_i, | |
165 | struct nat_entry *ne, struct f2fs_nat_entry *raw_ne, bool no_fail) | |
166 | { | |
167 | if (no_fail) | |
168 | f2fs_radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne); | |
169 | else if (radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne)) | |
170 | return NULL; | |
171 | ||
172 | if (raw_ne) | |
173 | node_info_from_raw_nat(&ne->ni, raw_ne); | |
22969158 CY |
174 | |
175 | spin_lock(&nm_i->nat_list_lock); | |
12f9ef37 | 176 | list_add_tail(&ne->list, &nm_i->nat_entries); |
22969158 CY |
177 | spin_unlock(&nm_i->nat_list_lock); |
178 | ||
12f9ef37 YH |
179 | nm_i->nat_cnt++; |
180 | return ne; | |
181 | } | |
182 | ||
e05df3b1 JK |
183 | static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) |
184 | { | |
22969158 CY |
185 | struct nat_entry *ne; |
186 | ||
187 | ne = radix_tree_lookup(&nm_i->nat_root, n); | |
188 | ||
189 | /* for recent accessed nat entry, move it to tail of lru list */ | |
190 | if (ne && !get_nat_flag(ne, IS_DIRTY)) { | |
191 | spin_lock(&nm_i->nat_list_lock); | |
192 | if (!list_empty(&ne->list)) | |
193 | list_move_tail(&ne->list, &nm_i->nat_entries); | |
194 | spin_unlock(&nm_i->nat_list_lock); | |
195 | } | |
196 | ||
197 | return ne; | |
e05df3b1 JK |
198 | } |
199 | ||
200 | static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, | |
201 | nid_t start, unsigned int nr, struct nat_entry **ep) | |
202 | { | |
203 | return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); | |
204 | } | |
205 | ||
206 | static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) | |
207 | { | |
e05df3b1 JK |
208 | radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); |
209 | nm_i->nat_cnt--; | |
12f9ef37 | 210 | __free_nat_entry(e); |
e05df3b1 JK |
211 | } |
212 | ||
780de47c CY |
213 | static struct nat_entry_set *__grab_nat_entry_set(struct f2fs_nm_info *nm_i, |
214 | struct nat_entry *ne) | |
309cc2b6 JK |
215 | { |
216 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); | |
217 | struct nat_entry_set *head; | |
218 | ||
309cc2b6 JK |
219 | head = radix_tree_lookup(&nm_i->nat_set_root, set); |
220 | if (!head) { | |
80c54505 | 221 | head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS); |
309cc2b6 JK |
222 | |
223 | INIT_LIST_HEAD(&head->entry_list); | |
224 | INIT_LIST_HEAD(&head->set_list); | |
225 | head->set = set; | |
226 | head->entry_cnt = 0; | |
9be32d72 | 227 | f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head); |
309cc2b6 | 228 | } |
780de47c CY |
229 | return head; |
230 | } | |
231 | ||
232 | static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
233 | struct nat_entry *ne) | |
234 | { | |
235 | struct nat_entry_set *head; | |
236 | bool new_ne = nat_get_blkaddr(ne) == NEW_ADDR; | |
237 | ||
238 | if (!new_ne) | |
239 | head = __grab_nat_entry_set(nm_i, ne); | |
240 | ||
241 | /* | |
242 | * update entry_cnt in below condition: | |
243 | * 1. update NEW_ADDR to valid block address; | |
244 | * 2. update old block address to new one; | |
245 | */ | |
246 | if (!new_ne && (get_nat_flag(ne, IS_PREALLOC) || | |
247 | !get_nat_flag(ne, IS_DIRTY))) | |
248 | head->entry_cnt++; | |
249 | ||
250 | set_nat_flag(ne, IS_PREALLOC, new_ne); | |
febeca6d CY |
251 | |
252 | if (get_nat_flag(ne, IS_DIRTY)) | |
253 | goto refresh_list; | |
254 | ||
309cc2b6 | 255 | nm_i->dirty_nat_cnt++; |
309cc2b6 | 256 | set_nat_flag(ne, IS_DIRTY, true); |
febeca6d | 257 | refresh_list: |
22969158 | 258 | spin_lock(&nm_i->nat_list_lock); |
780de47c | 259 | if (new_ne) |
febeca6d CY |
260 | list_del_init(&ne->list); |
261 | else | |
262 | list_move_tail(&ne->list, &head->entry_list); | |
22969158 | 263 | spin_unlock(&nm_i->nat_list_lock); |
309cc2b6 JK |
264 | } |
265 | ||
266 | static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
0b28b71e | 267 | struct nat_entry_set *set, struct nat_entry *ne) |
309cc2b6 | 268 | { |
22969158 | 269 | spin_lock(&nm_i->nat_list_lock); |
0b28b71e | 270 | list_move_tail(&ne->list, &nm_i->nat_entries); |
22969158 CY |
271 | spin_unlock(&nm_i->nat_list_lock); |
272 | ||
0b28b71e KM |
273 | set_nat_flag(ne, IS_DIRTY, false); |
274 | set->entry_cnt--; | |
275 | nm_i->dirty_nat_cnt--; | |
309cc2b6 JK |
276 | } |
277 | ||
278 | static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, | |
279 | nid_t start, unsigned int nr, struct nat_entry_set **ep) | |
280 | { | |
281 | return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep, | |
282 | start, nr); | |
283 | } | |
284 | ||
50fa53ec CY |
285 | bool f2fs_in_warm_node_list(struct f2fs_sb_info *sbi, struct page *page) |
286 | { | |
287 | return NODE_MAPPING(sbi) == page->mapping && | |
288 | IS_DNODE(page) && is_cold_node(page); | |
289 | } | |
290 | ||
291 | void f2fs_init_fsync_node_info(struct f2fs_sb_info *sbi) | |
292 | { | |
293 | spin_lock_init(&sbi->fsync_node_lock); | |
294 | INIT_LIST_HEAD(&sbi->fsync_node_list); | |
295 | sbi->fsync_seg_id = 0; | |
296 | sbi->fsync_node_num = 0; | |
297 | } | |
298 | ||
299 | static unsigned int f2fs_add_fsync_node_entry(struct f2fs_sb_info *sbi, | |
300 | struct page *page) | |
301 | { | |
302 | struct fsync_node_entry *fn; | |
303 | unsigned long flags; | |
304 | unsigned int seq_id; | |
305 | ||
306 | fn = f2fs_kmem_cache_alloc(fsync_node_entry_slab, GFP_NOFS); | |
307 | ||
308 | get_page(page); | |
309 | fn->page = page; | |
310 | INIT_LIST_HEAD(&fn->list); | |
311 | ||
312 | spin_lock_irqsave(&sbi->fsync_node_lock, flags); | |
313 | list_add_tail(&fn->list, &sbi->fsync_node_list); | |
314 | fn->seq_id = sbi->fsync_seg_id++; | |
315 | seq_id = fn->seq_id; | |
316 | sbi->fsync_node_num++; | |
317 | spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); | |
318 | ||
319 | return seq_id; | |
320 | } | |
321 | ||
322 | void f2fs_del_fsync_node_entry(struct f2fs_sb_info *sbi, struct page *page) | |
323 | { | |
324 | struct fsync_node_entry *fn; | |
325 | unsigned long flags; | |
326 | ||
327 | spin_lock_irqsave(&sbi->fsync_node_lock, flags); | |
328 | list_for_each_entry(fn, &sbi->fsync_node_list, list) { | |
329 | if (fn->page == page) { | |
330 | list_del(&fn->list); | |
331 | sbi->fsync_node_num--; | |
332 | spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); | |
333 | kmem_cache_free(fsync_node_entry_slab, fn); | |
334 | put_page(page); | |
335 | return; | |
336 | } | |
337 | } | |
338 | spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); | |
339 | f2fs_bug_on(sbi, 1); | |
340 | } | |
341 | ||
342 | void f2fs_reset_fsync_node_info(struct f2fs_sb_info *sbi) | |
343 | { | |
344 | unsigned long flags; | |
345 | ||
346 | spin_lock_irqsave(&sbi->fsync_node_lock, flags); | |
347 | sbi->fsync_seg_id = 0; | |
348 | spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); | |
349 | } | |
350 | ||
4d57b86d | 351 | int f2fs_need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
352 | { |
353 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
354 | struct nat_entry *e; | |
2dcf51ab | 355 | bool need = false; |
e05df3b1 | 356 | |
b873b798 | 357 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 | 358 | e = __lookup_nat_cache(nm_i, nid); |
2dcf51ab JK |
359 | if (e) { |
360 | if (!get_nat_flag(e, IS_CHECKPOINTED) && | |
361 | !get_nat_flag(e, HAS_FSYNCED_INODE)) | |
362 | need = true; | |
363 | } | |
b873b798 | 364 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 365 | return need; |
e05df3b1 JK |
366 | } |
367 | ||
4d57b86d | 368 | bool f2fs_is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) |
479f40c4 JK |
369 | { |
370 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
371 | struct nat_entry *e; | |
2dcf51ab | 372 | bool is_cp = true; |
479f40c4 | 373 | |
b873b798 | 374 | down_read(&nm_i->nat_tree_lock); |
2dcf51ab JK |
375 | e = __lookup_nat_cache(nm_i, nid); |
376 | if (e && !get_nat_flag(e, IS_CHECKPOINTED)) | |
377 | is_cp = false; | |
b873b798 | 378 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 379 | return is_cp; |
479f40c4 JK |
380 | } |
381 | ||
4d57b86d | 382 | bool f2fs_need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino) |
b6fe5873 JK |
383 | { |
384 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
385 | struct nat_entry *e; | |
88bd02c9 | 386 | bool need_update = true; |
b6fe5873 | 387 | |
b873b798 | 388 | down_read(&nm_i->nat_tree_lock); |
88bd02c9 JK |
389 | e = __lookup_nat_cache(nm_i, ino); |
390 | if (e && get_nat_flag(e, HAS_LAST_FSYNC) && | |
391 | (get_nat_flag(e, IS_CHECKPOINTED) || | |
392 | get_nat_flag(e, HAS_FSYNCED_INODE))) | |
393 | need_update = false; | |
b873b798 | 394 | up_read(&nm_i->nat_tree_lock); |
88bd02c9 | 395 | return need_update; |
b6fe5873 JK |
396 | } |
397 | ||
12f9ef37 | 398 | /* must be locked by nat_tree_lock */ |
1515aef0 | 399 | static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid, |
e05df3b1 JK |
400 | struct f2fs_nat_entry *ne) |
401 | { | |
1515aef0 | 402 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
12f9ef37 | 403 | struct nat_entry *new, *e; |
9be32d72 | 404 | |
12f9ef37 YH |
405 | new = __alloc_nat_entry(nid, false); |
406 | if (!new) | |
407 | return; | |
408 | ||
409 | down_write(&nm_i->nat_tree_lock); | |
e05df3b1 | 410 | e = __lookup_nat_cache(nm_i, nid); |
12f9ef37 YH |
411 | if (!e) |
412 | e = __init_nat_entry(nm_i, new, ne, false); | |
413 | else | |
0c0b471e EB |
414 | f2fs_bug_on(sbi, nat_get_ino(e) != le32_to_cpu(ne->ino) || |
415 | nat_get_blkaddr(e) != | |
416 | le32_to_cpu(ne->block_addr) || | |
1515aef0 | 417 | nat_get_version(e) != ne->version); |
12f9ef37 YH |
418 | up_write(&nm_i->nat_tree_lock); |
419 | if (e != new) | |
420 | __free_nat_entry(new); | |
e05df3b1 JK |
421 | } |
422 | ||
423 | static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, | |
479f40c4 | 424 | block_t new_blkaddr, bool fsync_done) |
e05df3b1 JK |
425 | { |
426 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
427 | struct nat_entry *e; | |
12f9ef37 | 428 | struct nat_entry *new = __alloc_nat_entry(ni->nid, true); |
9be32d72 | 429 | |
b873b798 | 430 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
431 | e = __lookup_nat_cache(nm_i, ni->nid); |
432 | if (!e) { | |
12f9ef37 | 433 | e = __init_nat_entry(nm_i, new, NULL, true); |
5c27f4ee | 434 | copy_node_info(&e->ni, ni); |
9850cf4a | 435 | f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR); |
e05df3b1 JK |
436 | } else if (new_blkaddr == NEW_ADDR) { |
437 | /* | |
438 | * when nid is reallocated, | |
439 | * previous nat entry can be remained in nat cache. | |
440 | * So, reinitialize it with new information. | |
441 | */ | |
5c27f4ee | 442 | copy_node_info(&e->ni, ni); |
9850cf4a | 443 | f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR); |
e05df3b1 | 444 | } |
12f9ef37 YH |
445 | /* let's free early to reduce memory consumption */ |
446 | if (e != new) | |
447 | __free_nat_entry(new); | |
e05df3b1 | 448 | |
e05df3b1 | 449 | /* sanity check */ |
9850cf4a JK |
450 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr); |
451 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR && | |
e05df3b1 | 452 | new_blkaddr == NULL_ADDR); |
9850cf4a | 453 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR && |
e05df3b1 | 454 | new_blkaddr == NEW_ADDR); |
e1da7872 | 455 | f2fs_bug_on(sbi, is_valid_data_blkaddr(sbi, nat_get_blkaddr(e)) && |
e05df3b1 JK |
456 | new_blkaddr == NEW_ADDR); |
457 | ||
e1c42045 | 458 | /* increment version no as node is removed */ |
e05df3b1 JK |
459 | if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { |
460 | unsigned char version = nat_get_version(e); | |
461 | nat_set_version(e, inc_node_version(version)); | |
462 | } | |
463 | ||
464 | /* change address */ | |
465 | nat_set_blkaddr(e, new_blkaddr); | |
e1da7872 | 466 | if (!is_valid_data_blkaddr(sbi, new_blkaddr)) |
88bd02c9 | 467 | set_nat_flag(e, IS_CHECKPOINTED, false); |
e05df3b1 | 468 | __set_nat_cache_dirty(nm_i, e); |
479f40c4 JK |
469 | |
470 | /* update fsync_mark if its inode nat entry is still alive */ | |
d5b692b7 CY |
471 | if (ni->nid != ni->ino) |
472 | e = __lookup_nat_cache(nm_i, ni->ino); | |
88bd02c9 JK |
473 | if (e) { |
474 | if (fsync_done && ni->nid == ni->ino) | |
475 | set_nat_flag(e, HAS_FSYNCED_INODE, true); | |
476 | set_nat_flag(e, HAS_LAST_FSYNC, fsync_done); | |
477 | } | |
b873b798 | 478 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
479 | } |
480 | ||
4d57b86d | 481 | int f2fs_try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) |
e05df3b1 JK |
482 | { |
483 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1b38dc8e | 484 | int nr = nr_shrink; |
e05df3b1 | 485 | |
b873b798 JK |
486 | if (!down_write_trylock(&nm_i->nat_tree_lock)) |
487 | return 0; | |
e05df3b1 | 488 | |
22969158 CY |
489 | spin_lock(&nm_i->nat_list_lock); |
490 | while (nr_shrink) { | |
e05df3b1 | 491 | struct nat_entry *ne; |
22969158 CY |
492 | |
493 | if (list_empty(&nm_i->nat_entries)) | |
494 | break; | |
495 | ||
e05df3b1 JK |
496 | ne = list_first_entry(&nm_i->nat_entries, |
497 | struct nat_entry, list); | |
22969158 CY |
498 | list_del(&ne->list); |
499 | spin_unlock(&nm_i->nat_list_lock); | |
500 | ||
e05df3b1 JK |
501 | __del_from_nat_cache(nm_i, ne); |
502 | nr_shrink--; | |
22969158 CY |
503 | |
504 | spin_lock(&nm_i->nat_list_lock); | |
e05df3b1 | 505 | } |
22969158 CY |
506 | spin_unlock(&nm_i->nat_list_lock); |
507 | ||
b873b798 | 508 | up_write(&nm_i->nat_tree_lock); |
1b38dc8e | 509 | return nr - nr_shrink; |
e05df3b1 JK |
510 | } |
511 | ||
0a8165d7 | 512 | /* |
e1c42045 | 513 | * This function always returns success |
e05df3b1 | 514 | */ |
7735730d | 515 | int f2fs_get_node_info(struct f2fs_sb_info *sbi, nid_t nid, |
4d57b86d | 516 | struct node_info *ni) |
e05df3b1 JK |
517 | { |
518 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
519 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 520 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
521 | nid_t start_nid = START_NID(nid); |
522 | struct f2fs_nat_block *nat_blk; | |
523 | struct page *page = NULL; | |
524 | struct f2fs_nat_entry ne; | |
525 | struct nat_entry *e; | |
66a82d1f | 526 | pgoff_t index; |
e05df3b1 JK |
527 | int i; |
528 | ||
529 | ni->nid = nid; | |
530 | ||
531 | /* Check nat cache */ | |
b873b798 | 532 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
533 | e = __lookup_nat_cache(nm_i, nid); |
534 | if (e) { | |
535 | ni->ino = nat_get_ino(e); | |
536 | ni->blk_addr = nat_get_blkaddr(e); | |
537 | ni->version = nat_get_version(e); | |
b873b798 | 538 | up_read(&nm_i->nat_tree_lock); |
7735730d | 539 | return 0; |
1515aef0 | 540 | } |
e05df3b1 | 541 | |
3547ea96 JK |
542 | memset(&ne, 0, sizeof(struct f2fs_nat_entry)); |
543 | ||
e05df3b1 | 544 | /* Check current segment summary */ |
b7ad7512 | 545 | down_read(&curseg->journal_rwsem); |
4d57b86d | 546 | i = f2fs_lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0); |
e05df3b1 | 547 | if (i >= 0) { |
dfc08a12 | 548 | ne = nat_in_journal(journal, i); |
e05df3b1 JK |
549 | node_info_from_raw_nat(ni, &ne); |
550 | } | |
b7ad7512 | 551 | up_read(&curseg->journal_rwsem); |
66a82d1f YH |
552 | if (i >= 0) { |
553 | up_read(&nm_i->nat_tree_lock); | |
e05df3b1 | 554 | goto cache; |
66a82d1f | 555 | } |
e05df3b1 JK |
556 | |
557 | /* Fill node_info from nat page */ | |
66a82d1f YH |
558 | index = current_nat_addr(sbi, nid); |
559 | up_read(&nm_i->nat_tree_lock); | |
560 | ||
4d57b86d | 561 | page = f2fs_get_meta_page(sbi, index); |
7735730d CY |
562 | if (IS_ERR(page)) |
563 | return PTR_ERR(page); | |
564 | ||
e05df3b1 JK |
565 | nat_blk = (struct f2fs_nat_block *)page_address(page); |
566 | ne = nat_blk->entries[nid - start_nid]; | |
567 | node_info_from_raw_nat(ni, &ne); | |
568 | f2fs_put_page(page, 1); | |
569 | cache: | |
570 | /* cache nat entry */ | |
1515aef0 | 571 | cache_nat_entry(sbi, nid, &ne); |
7735730d | 572 | return 0; |
e05df3b1 JK |
573 | } |
574 | ||
79344efb JK |
575 | /* |
576 | * readahead MAX_RA_NODE number of node pages. | |
577 | */ | |
4d57b86d | 578 | static void f2fs_ra_node_pages(struct page *parent, int start, int n) |
79344efb JK |
579 | { |
580 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); | |
581 | struct blk_plug plug; | |
582 | int i, end; | |
583 | nid_t nid; | |
584 | ||
585 | blk_start_plug(&plug); | |
586 | ||
587 | /* Then, try readahead for siblings of the desired node */ | |
588 | end = start + n; | |
589 | end = min(end, NIDS_PER_BLOCK); | |
590 | for (i = start; i < end; i++) { | |
591 | nid = get_nid(parent, i, false); | |
4d57b86d | 592 | f2fs_ra_node_page(sbi, nid); |
79344efb JK |
593 | } |
594 | ||
595 | blk_finish_plug(&plug); | |
596 | } | |
597 | ||
4d57b86d | 598 | pgoff_t f2fs_get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs) |
3cf45747 CY |
599 | { |
600 | const long direct_index = ADDRS_PER_INODE(dn->inode); | |
601 | const long direct_blks = ADDRS_PER_BLOCK; | |
602 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
603 | unsigned int skipped_unit = ADDRS_PER_BLOCK; | |
604 | int cur_level = dn->cur_level; | |
605 | int max_level = dn->max_level; | |
606 | pgoff_t base = 0; | |
607 | ||
608 | if (!dn->max_level) | |
609 | return pgofs + 1; | |
610 | ||
611 | while (max_level-- > cur_level) | |
612 | skipped_unit *= NIDS_PER_BLOCK; | |
613 | ||
614 | switch (dn->max_level) { | |
615 | case 3: | |
616 | base += 2 * indirect_blks; | |
617 | case 2: | |
618 | base += 2 * direct_blks; | |
619 | case 1: | |
620 | base += direct_index; | |
621 | break; | |
622 | default: | |
623 | f2fs_bug_on(F2FS_I_SB(dn->inode), 1); | |
624 | } | |
625 | ||
626 | return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base; | |
627 | } | |
628 | ||
0a8165d7 | 629 | /* |
e05df3b1 JK |
630 | * The maximum depth is four. |
631 | * Offset[0] will have raw inode offset. | |
632 | */ | |
81ca7350 | 633 | static int get_node_path(struct inode *inode, long block, |
de93653f | 634 | int offset[4], unsigned int noffset[4]) |
e05df3b1 | 635 | { |
81ca7350 | 636 | const long direct_index = ADDRS_PER_INODE(inode); |
e05df3b1 JK |
637 | const long direct_blks = ADDRS_PER_BLOCK; |
638 | const long dptrs_per_blk = NIDS_PER_BLOCK; | |
639 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
640 | const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; | |
641 | int n = 0; | |
642 | int level = 0; | |
643 | ||
644 | noffset[0] = 0; | |
645 | ||
646 | if (block < direct_index) { | |
25c0a6e5 | 647 | offset[n] = block; |
e05df3b1 JK |
648 | goto got; |
649 | } | |
650 | block -= direct_index; | |
651 | if (block < direct_blks) { | |
652 | offset[n++] = NODE_DIR1_BLOCK; | |
653 | noffset[n] = 1; | |
25c0a6e5 | 654 | offset[n] = block; |
e05df3b1 JK |
655 | level = 1; |
656 | goto got; | |
657 | } | |
658 | block -= direct_blks; | |
659 | if (block < direct_blks) { | |
660 | offset[n++] = NODE_DIR2_BLOCK; | |
661 | noffset[n] = 2; | |
25c0a6e5 | 662 | offset[n] = block; |
e05df3b1 JK |
663 | level = 1; |
664 | goto got; | |
665 | } | |
666 | block -= direct_blks; | |
667 | if (block < indirect_blks) { | |
668 | offset[n++] = NODE_IND1_BLOCK; | |
669 | noffset[n] = 3; | |
670 | offset[n++] = block / direct_blks; | |
671 | noffset[n] = 4 + offset[n - 1]; | |
25c0a6e5 | 672 | offset[n] = block % direct_blks; |
e05df3b1 JK |
673 | level = 2; |
674 | goto got; | |
675 | } | |
676 | block -= indirect_blks; | |
677 | if (block < indirect_blks) { | |
678 | offset[n++] = NODE_IND2_BLOCK; | |
679 | noffset[n] = 4 + dptrs_per_blk; | |
680 | offset[n++] = block / direct_blks; | |
681 | noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; | |
25c0a6e5 | 682 | offset[n] = block % direct_blks; |
e05df3b1 JK |
683 | level = 2; |
684 | goto got; | |
685 | } | |
686 | block -= indirect_blks; | |
687 | if (block < dindirect_blks) { | |
688 | offset[n++] = NODE_DIND_BLOCK; | |
689 | noffset[n] = 5 + (dptrs_per_blk * 2); | |
690 | offset[n++] = block / indirect_blks; | |
691 | noffset[n] = 6 + (dptrs_per_blk * 2) + | |
692 | offset[n - 1] * (dptrs_per_blk + 1); | |
693 | offset[n++] = (block / direct_blks) % dptrs_per_blk; | |
694 | noffset[n] = 7 + (dptrs_per_blk * 2) + | |
695 | offset[n - 2] * (dptrs_per_blk + 1) + | |
696 | offset[n - 1]; | |
25c0a6e5 | 697 | offset[n] = block % direct_blks; |
e05df3b1 JK |
698 | level = 3; |
699 | goto got; | |
700 | } else { | |
adb6dc19 | 701 | return -E2BIG; |
e05df3b1 JK |
702 | } |
703 | got: | |
704 | return level; | |
705 | } | |
706 | ||
707 | /* | |
708 | * Caller should call f2fs_put_dnode(dn). | |
4f4124d0 CY |
709 | * Also, it should grab and release a rwsem by calling f2fs_lock_op() and |
710 | * f2fs_unlock_op() only if ro is not set RDONLY_NODE. | |
39936837 | 711 | * In the case of RDONLY_NODE, we don't need to care about mutex. |
e05df3b1 | 712 | */ |
4d57b86d | 713 | int f2fs_get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) |
e05df3b1 | 714 | { |
4081363f | 715 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 | 716 | struct page *npage[4]; |
f1a3b98e | 717 | struct page *parent = NULL; |
e05df3b1 JK |
718 | int offset[4]; |
719 | unsigned int noffset[4]; | |
720 | nid_t nids[4]; | |
3cf45747 | 721 | int level, i = 0; |
e05df3b1 JK |
722 | int err = 0; |
723 | ||
81ca7350 | 724 | level = get_node_path(dn->inode, index, offset, noffset); |
adb6dc19 JK |
725 | if (level < 0) |
726 | return level; | |
e05df3b1 JK |
727 | |
728 | nids[0] = dn->inode->i_ino; | |
1646cfac | 729 | npage[0] = dn->inode_page; |
e05df3b1 | 730 | |
1646cfac | 731 | if (!npage[0]) { |
4d57b86d | 732 | npage[0] = f2fs_get_node_page(sbi, nids[0]); |
1646cfac JK |
733 | if (IS_ERR(npage[0])) |
734 | return PTR_ERR(npage[0]); | |
735 | } | |
f1a3b98e JK |
736 | |
737 | /* if inline_data is set, should not report any block indices */ | |
738 | if (f2fs_has_inline_data(dn->inode) && index) { | |
76629165 | 739 | err = -ENOENT; |
f1a3b98e JK |
740 | f2fs_put_page(npage[0], 1); |
741 | goto release_out; | |
742 | } | |
743 | ||
e05df3b1 | 744 | parent = npage[0]; |
52c2db3f CL |
745 | if (level != 0) |
746 | nids[1] = get_nid(parent, offset[0], true); | |
e05df3b1 JK |
747 | dn->inode_page = npage[0]; |
748 | dn->inode_page_locked = true; | |
749 | ||
750 | /* get indirect or direct nodes */ | |
751 | for (i = 1; i <= level; i++) { | |
752 | bool done = false; | |
753 | ||
266e97a8 | 754 | if (!nids[i] && mode == ALLOC_NODE) { |
e05df3b1 | 755 | /* alloc new node */ |
4d57b86d | 756 | if (!f2fs_alloc_nid(sbi, &(nids[i]))) { |
e05df3b1 JK |
757 | err = -ENOSPC; |
758 | goto release_pages; | |
759 | } | |
760 | ||
761 | dn->nid = nids[i]; | |
4d57b86d | 762 | npage[i] = f2fs_new_node_page(dn, noffset[i]); |
e05df3b1 | 763 | if (IS_ERR(npage[i])) { |
4d57b86d | 764 | f2fs_alloc_nid_failed(sbi, nids[i]); |
e05df3b1 JK |
765 | err = PTR_ERR(npage[i]); |
766 | goto release_pages; | |
767 | } | |
768 | ||
769 | set_nid(parent, offset[i - 1], nids[i], i == 1); | |
4d57b86d | 770 | f2fs_alloc_nid_done(sbi, nids[i]); |
e05df3b1 | 771 | done = true; |
266e97a8 | 772 | } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { |
4d57b86d | 773 | npage[i] = f2fs_get_node_page_ra(parent, offset[i - 1]); |
e05df3b1 JK |
774 | if (IS_ERR(npage[i])) { |
775 | err = PTR_ERR(npage[i]); | |
776 | goto release_pages; | |
777 | } | |
778 | done = true; | |
779 | } | |
780 | if (i == 1) { | |
781 | dn->inode_page_locked = false; | |
782 | unlock_page(parent); | |
783 | } else { | |
784 | f2fs_put_page(parent, 1); | |
785 | } | |
786 | ||
787 | if (!done) { | |
4d57b86d | 788 | npage[i] = f2fs_get_node_page(sbi, nids[i]); |
e05df3b1 JK |
789 | if (IS_ERR(npage[i])) { |
790 | err = PTR_ERR(npage[i]); | |
791 | f2fs_put_page(npage[0], 0); | |
792 | goto release_out; | |
793 | } | |
794 | } | |
795 | if (i < level) { | |
796 | parent = npage[i]; | |
797 | nids[i + 1] = get_nid(parent, offset[i], false); | |
798 | } | |
799 | } | |
800 | dn->nid = nids[level]; | |
801 | dn->ofs_in_node = offset[level]; | |
802 | dn->node_page = npage[level]; | |
7a2af766 CY |
803 | dn->data_blkaddr = datablock_addr(dn->inode, |
804 | dn->node_page, dn->ofs_in_node); | |
e05df3b1 JK |
805 | return 0; |
806 | ||
807 | release_pages: | |
808 | f2fs_put_page(parent, 1); | |
809 | if (i > 1) | |
810 | f2fs_put_page(npage[0], 0); | |
811 | release_out: | |
812 | dn->inode_page = NULL; | |
813 | dn->node_page = NULL; | |
3cf45747 CY |
814 | if (err == -ENOENT) { |
815 | dn->cur_level = i; | |
816 | dn->max_level = level; | |
0a2aa8fb | 817 | dn->ofs_in_node = offset[level]; |
3cf45747 | 818 | } |
e05df3b1 JK |
819 | return err; |
820 | } | |
821 | ||
7735730d | 822 | static int truncate_node(struct dnode_of_data *dn) |
e05df3b1 | 823 | { |
4081363f | 824 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 | 825 | struct node_info ni; |
7735730d | 826 | int err; |
e05df3b1 | 827 | |
7735730d CY |
828 | err = f2fs_get_node_info(sbi, dn->nid, &ni); |
829 | if (err) | |
830 | return err; | |
e05df3b1 | 831 | |
e05df3b1 | 832 | /* Deallocate node address */ |
4d57b86d | 833 | f2fs_invalidate_blocks(sbi, ni.blk_addr); |
000519f2 | 834 | dec_valid_node_count(sbi, dn->inode, dn->nid == dn->inode->i_ino); |
479f40c4 | 835 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
e05df3b1 JK |
836 | |
837 | if (dn->nid == dn->inode->i_ino) { | |
4d57b86d | 838 | f2fs_remove_orphan_inode(sbi, dn->nid); |
e05df3b1 | 839 | dec_valid_inode_count(sbi); |
0f18b462 | 840 | f2fs_inode_synced(dn->inode); |
e05df3b1 | 841 | } |
000519f2 | 842 | |
e05df3b1 | 843 | clear_node_page_dirty(dn->node_page); |
caf0047e | 844 | set_sbi_flag(sbi, SBI_IS_DIRTY); |
e05df3b1 JK |
845 | |
846 | f2fs_put_page(dn->node_page, 1); | |
bf39c00a JK |
847 | |
848 | invalidate_mapping_pages(NODE_MAPPING(sbi), | |
849 | dn->node_page->index, dn->node_page->index); | |
850 | ||
e05df3b1 | 851 | dn->node_page = NULL; |
51dd6249 | 852 | trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); |
7735730d CY |
853 | |
854 | return 0; | |
e05df3b1 JK |
855 | } |
856 | ||
857 | static int truncate_dnode(struct dnode_of_data *dn) | |
858 | { | |
e05df3b1 | 859 | struct page *page; |
7735730d | 860 | int err; |
e05df3b1 JK |
861 | |
862 | if (dn->nid == 0) | |
863 | return 1; | |
864 | ||
865 | /* get direct node */ | |
4d57b86d | 866 | page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
e05df3b1 JK |
867 | if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) |
868 | return 1; | |
869 | else if (IS_ERR(page)) | |
870 | return PTR_ERR(page); | |
871 | ||
872 | /* Make dnode_of_data for parameter */ | |
873 | dn->node_page = page; | |
874 | dn->ofs_in_node = 0; | |
4d57b86d | 875 | f2fs_truncate_data_blocks(dn); |
7735730d CY |
876 | err = truncate_node(dn); |
877 | if (err) | |
878 | return err; | |
879 | ||
e05df3b1 JK |
880 | return 1; |
881 | } | |
882 | ||
883 | static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, | |
884 | int ofs, int depth) | |
885 | { | |
e05df3b1 JK |
886 | struct dnode_of_data rdn = *dn; |
887 | struct page *page; | |
888 | struct f2fs_node *rn; | |
889 | nid_t child_nid; | |
890 | unsigned int child_nofs; | |
891 | int freed = 0; | |
892 | int i, ret; | |
893 | ||
894 | if (dn->nid == 0) | |
895 | return NIDS_PER_BLOCK + 1; | |
896 | ||
51dd6249 NJ |
897 | trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); |
898 | ||
4d57b86d | 899 | page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
51dd6249 NJ |
900 | if (IS_ERR(page)) { |
901 | trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); | |
e05df3b1 | 902 | return PTR_ERR(page); |
51dd6249 | 903 | } |
e05df3b1 | 904 | |
4d57b86d | 905 | f2fs_ra_node_pages(page, ofs, NIDS_PER_BLOCK); |
79344efb | 906 | |
45590710 | 907 | rn = F2FS_NODE(page); |
e05df3b1 JK |
908 | if (depth < 3) { |
909 | for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { | |
910 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
911 | if (child_nid == 0) | |
912 | continue; | |
913 | rdn.nid = child_nid; | |
914 | ret = truncate_dnode(&rdn); | |
915 | if (ret < 0) | |
916 | goto out_err; | |
12719ae1 JK |
917 | if (set_nid(page, i, 0, false)) |
918 | dn->node_changed = true; | |
e05df3b1 JK |
919 | } |
920 | } else { | |
921 | child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; | |
922 | for (i = ofs; i < NIDS_PER_BLOCK; i++) { | |
923 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
924 | if (child_nid == 0) { | |
925 | child_nofs += NIDS_PER_BLOCK + 1; | |
926 | continue; | |
927 | } | |
928 | rdn.nid = child_nid; | |
929 | ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); | |
930 | if (ret == (NIDS_PER_BLOCK + 1)) { | |
12719ae1 JK |
931 | if (set_nid(page, i, 0, false)) |
932 | dn->node_changed = true; | |
e05df3b1 JK |
933 | child_nofs += ret; |
934 | } else if (ret < 0 && ret != -ENOENT) { | |
935 | goto out_err; | |
936 | } | |
937 | } | |
938 | freed = child_nofs; | |
939 | } | |
940 | ||
941 | if (!ofs) { | |
942 | /* remove current indirect node */ | |
943 | dn->node_page = page; | |
7735730d CY |
944 | ret = truncate_node(dn); |
945 | if (ret) | |
946 | goto out_err; | |
e05df3b1 JK |
947 | freed++; |
948 | } else { | |
949 | f2fs_put_page(page, 1); | |
950 | } | |
51dd6249 | 951 | trace_f2fs_truncate_nodes_exit(dn->inode, freed); |
e05df3b1 JK |
952 | return freed; |
953 | ||
954 | out_err: | |
955 | f2fs_put_page(page, 1); | |
51dd6249 | 956 | trace_f2fs_truncate_nodes_exit(dn->inode, ret); |
e05df3b1 JK |
957 | return ret; |
958 | } | |
959 | ||
960 | static int truncate_partial_nodes(struct dnode_of_data *dn, | |
961 | struct f2fs_inode *ri, int *offset, int depth) | |
962 | { | |
e05df3b1 JK |
963 | struct page *pages[2]; |
964 | nid_t nid[3]; | |
965 | nid_t child_nid; | |
966 | int err = 0; | |
967 | int i; | |
968 | int idx = depth - 2; | |
969 | ||
970 | nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); | |
971 | if (!nid[0]) | |
972 | return 0; | |
973 | ||
974 | /* get indirect nodes in the path */ | |
a225dca3 | 975 | for (i = 0; i < idx + 1; i++) { |
e1c42045 | 976 | /* reference count'll be increased */ |
4d57b86d | 977 | pages[i] = f2fs_get_node_page(F2FS_I_SB(dn->inode), nid[i]); |
e05df3b1 | 978 | if (IS_ERR(pages[i])) { |
e05df3b1 | 979 | err = PTR_ERR(pages[i]); |
a225dca3 | 980 | idx = i - 1; |
e05df3b1 JK |
981 | goto fail; |
982 | } | |
983 | nid[i + 1] = get_nid(pages[i], offset[i + 1], false); | |
984 | } | |
985 | ||
4d57b86d | 986 | f2fs_ra_node_pages(pages[idx], offset[idx + 1], NIDS_PER_BLOCK); |
79344efb | 987 | |
e05df3b1 | 988 | /* free direct nodes linked to a partial indirect node */ |
a225dca3 | 989 | for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) { |
e05df3b1 JK |
990 | child_nid = get_nid(pages[idx], i, false); |
991 | if (!child_nid) | |
992 | continue; | |
993 | dn->nid = child_nid; | |
994 | err = truncate_dnode(dn); | |
995 | if (err < 0) | |
996 | goto fail; | |
12719ae1 JK |
997 | if (set_nid(pages[idx], i, 0, false)) |
998 | dn->node_changed = true; | |
e05df3b1 JK |
999 | } |
1000 | ||
a225dca3 | 1001 | if (offset[idx + 1] == 0) { |
e05df3b1 JK |
1002 | dn->node_page = pages[idx]; |
1003 | dn->nid = nid[idx]; | |
7735730d CY |
1004 | err = truncate_node(dn); |
1005 | if (err) | |
1006 | goto fail; | |
e05df3b1 JK |
1007 | } else { |
1008 | f2fs_put_page(pages[idx], 1); | |
1009 | } | |
1010 | offset[idx]++; | |
a225dca3 | 1011 | offset[idx + 1] = 0; |
1012 | idx--; | |
e05df3b1 | 1013 | fail: |
a225dca3 | 1014 | for (i = idx; i >= 0; i--) |
e05df3b1 | 1015 | f2fs_put_page(pages[i], 1); |
51dd6249 NJ |
1016 | |
1017 | trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); | |
1018 | ||
e05df3b1 JK |
1019 | return err; |
1020 | } | |
1021 | ||
0a8165d7 | 1022 | /* |
e05df3b1 JK |
1023 | * All the block addresses of data and nodes should be nullified. |
1024 | */ | |
4d57b86d | 1025 | int f2fs_truncate_inode_blocks(struct inode *inode, pgoff_t from) |
e05df3b1 | 1026 | { |
4081363f | 1027 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
e05df3b1 JK |
1028 | int err = 0, cont = 1; |
1029 | int level, offset[4], noffset[4]; | |
7dd690c8 | 1030 | unsigned int nofs = 0; |
58bfaf44 | 1031 | struct f2fs_inode *ri; |
e05df3b1 JK |
1032 | struct dnode_of_data dn; |
1033 | struct page *page; | |
1034 | ||
51dd6249 NJ |
1035 | trace_f2fs_truncate_inode_blocks_enter(inode, from); |
1036 | ||
81ca7350 | 1037 | level = get_node_path(inode, from, offset, noffset); |
adb6dc19 JK |
1038 | if (level < 0) |
1039 | return level; | |
ff373558 | 1040 | |
4d57b86d | 1041 | page = f2fs_get_node_page(sbi, inode->i_ino); |
51dd6249 NJ |
1042 | if (IS_ERR(page)) { |
1043 | trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); | |
e05df3b1 | 1044 | return PTR_ERR(page); |
51dd6249 | 1045 | } |
e05df3b1 JK |
1046 | |
1047 | set_new_dnode(&dn, inode, page, NULL, 0); | |
1048 | unlock_page(page); | |
1049 | ||
58bfaf44 | 1050 | ri = F2FS_INODE(page); |
e05df3b1 JK |
1051 | switch (level) { |
1052 | case 0: | |
1053 | case 1: | |
1054 | nofs = noffset[1]; | |
1055 | break; | |
1056 | case 2: | |
1057 | nofs = noffset[1]; | |
1058 | if (!offset[level - 1]) | |
1059 | goto skip_partial; | |
58bfaf44 | 1060 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
1061 | if (err < 0 && err != -ENOENT) |
1062 | goto fail; | |
1063 | nofs += 1 + NIDS_PER_BLOCK; | |
1064 | break; | |
1065 | case 3: | |
1066 | nofs = 5 + 2 * NIDS_PER_BLOCK; | |
1067 | if (!offset[level - 1]) | |
1068 | goto skip_partial; | |
58bfaf44 | 1069 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
1070 | if (err < 0 && err != -ENOENT) |
1071 | goto fail; | |
1072 | break; | |
1073 | default: | |
1074 | BUG(); | |
1075 | } | |
1076 | ||
1077 | skip_partial: | |
1078 | while (cont) { | |
58bfaf44 | 1079 | dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); |
e05df3b1 JK |
1080 | switch (offset[0]) { |
1081 | case NODE_DIR1_BLOCK: | |
1082 | case NODE_DIR2_BLOCK: | |
1083 | err = truncate_dnode(&dn); | |
1084 | break; | |
1085 | ||
1086 | case NODE_IND1_BLOCK: | |
1087 | case NODE_IND2_BLOCK: | |
1088 | err = truncate_nodes(&dn, nofs, offset[1], 2); | |
1089 | break; | |
1090 | ||
1091 | case NODE_DIND_BLOCK: | |
1092 | err = truncate_nodes(&dn, nofs, offset[1], 3); | |
1093 | cont = 0; | |
1094 | break; | |
1095 | ||
1096 | default: | |
1097 | BUG(); | |
1098 | } | |
1099 | if (err < 0 && err != -ENOENT) | |
1100 | goto fail; | |
1101 | if (offset[1] == 0 && | |
58bfaf44 | 1102 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) { |
e05df3b1 | 1103 | lock_page(page); |
ff373558 | 1104 | BUG_ON(page->mapping != NODE_MAPPING(sbi)); |
fec1d657 | 1105 | f2fs_wait_on_page_writeback(page, NODE, true); |
58bfaf44 | 1106 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; |
e05df3b1 JK |
1107 | set_page_dirty(page); |
1108 | unlock_page(page); | |
1109 | } | |
1110 | offset[1] = 0; | |
1111 | offset[0]++; | |
1112 | nofs += err; | |
1113 | } | |
1114 | fail: | |
1115 | f2fs_put_page(page, 0); | |
51dd6249 | 1116 | trace_f2fs_truncate_inode_blocks_exit(inode, err); |
e05df3b1 JK |
1117 | return err > 0 ? 0 : err; |
1118 | } | |
1119 | ||
9c77f754 | 1120 | /* caller must lock inode page */ |
4d57b86d | 1121 | int f2fs_truncate_xattr_node(struct inode *inode) |
4f16fb0f | 1122 | { |
4081363f | 1123 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
4f16fb0f JK |
1124 | nid_t nid = F2FS_I(inode)->i_xattr_nid; |
1125 | struct dnode_of_data dn; | |
1126 | struct page *npage; | |
7735730d | 1127 | int err; |
4f16fb0f JK |
1128 | |
1129 | if (!nid) | |
1130 | return 0; | |
1131 | ||
4d57b86d | 1132 | npage = f2fs_get_node_page(sbi, nid); |
4f16fb0f JK |
1133 | if (IS_ERR(npage)) |
1134 | return PTR_ERR(npage); | |
1135 | ||
7735730d CY |
1136 | set_new_dnode(&dn, inode, NULL, npage, nid); |
1137 | err = truncate_node(&dn); | |
1138 | if (err) { | |
1139 | f2fs_put_page(npage, 1); | |
1140 | return err; | |
1141 | } | |
1142 | ||
205b9822 | 1143 | f2fs_i_xnid_write(inode, 0); |
65985d93 | 1144 | |
4f16fb0f JK |
1145 | return 0; |
1146 | } | |
1147 | ||
39936837 | 1148 | /* |
4f4124d0 CY |
1149 | * Caller should grab and release a rwsem by calling f2fs_lock_op() and |
1150 | * f2fs_unlock_op(). | |
39936837 | 1151 | */ |
4d57b86d | 1152 | int f2fs_remove_inode_page(struct inode *inode) |
e05df3b1 | 1153 | { |
e05df3b1 | 1154 | struct dnode_of_data dn; |
13ec7297 | 1155 | int err; |
e05df3b1 | 1156 | |
c2e69583 | 1157 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); |
4d57b86d | 1158 | err = f2fs_get_dnode_of_data(&dn, 0, LOOKUP_NODE); |
13ec7297 CY |
1159 | if (err) |
1160 | return err; | |
e05df3b1 | 1161 | |
4d57b86d | 1162 | err = f2fs_truncate_xattr_node(inode); |
13ec7297 | 1163 | if (err) { |
c2e69583 | 1164 | f2fs_put_dnode(&dn); |
13ec7297 | 1165 | return err; |
e05df3b1 | 1166 | } |
c2e69583 JK |
1167 | |
1168 | /* remove potential inline_data blocks */ | |
1169 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
1170 | S_ISLNK(inode->i_mode)) | |
4d57b86d | 1171 | f2fs_truncate_data_blocks_range(&dn, 1); |
c2e69583 | 1172 | |
e1c42045 | 1173 | /* 0 is possible, after f2fs_new_inode() has failed */ |
8d714f8a JK |
1174 | if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { |
1175 | f2fs_put_dnode(&dn); | |
1176 | return -EIO; | |
1177 | } | |
9850cf4a | 1178 | f2fs_bug_on(F2FS_I_SB(inode), |
0eb0adad | 1179 | inode->i_blocks != 0 && inode->i_blocks != 8); |
c2e69583 JK |
1180 | |
1181 | /* will put inode & node pages */ | |
7735730d CY |
1182 | err = truncate_node(&dn); |
1183 | if (err) { | |
1184 | f2fs_put_dnode(&dn); | |
1185 | return err; | |
1186 | } | |
13ec7297 | 1187 | return 0; |
e05df3b1 JK |
1188 | } |
1189 | ||
4d57b86d | 1190 | struct page *f2fs_new_inode_page(struct inode *inode) |
e05df3b1 | 1191 | { |
e05df3b1 JK |
1192 | struct dnode_of_data dn; |
1193 | ||
1194 | /* allocate inode page for new inode */ | |
1195 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); | |
44a83ff6 JK |
1196 | |
1197 | /* caller should f2fs_put_page(page, 1); */ | |
4d57b86d | 1198 | return f2fs_new_node_page(&dn, 0); |
e05df3b1 JK |
1199 | } |
1200 | ||
4d57b86d | 1201 | struct page *f2fs_new_node_page(struct dnode_of_data *dn, unsigned int ofs) |
e05df3b1 | 1202 | { |
4081363f | 1203 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
25cc5d3b | 1204 | struct node_info new_ni; |
e05df3b1 JK |
1205 | struct page *page; |
1206 | int err; | |
1207 | ||
91942321 | 1208 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
e05df3b1 JK |
1209 | return ERR_PTR(-EPERM); |
1210 | ||
300e129c | 1211 | page = f2fs_grab_cache_page(NODE_MAPPING(sbi), dn->nid, false); |
e05df3b1 JK |
1212 | if (!page) |
1213 | return ERR_PTR(-ENOMEM); | |
1214 | ||
0abd675e | 1215 | if (unlikely((err = inc_valid_node_count(sbi, dn->inode, !ofs)))) |
9c02740c | 1216 | goto fail; |
0abd675e | 1217 | |
25cc5d3b | 1218 | #ifdef CONFIG_F2FS_CHECK_FS |
7735730d CY |
1219 | err = f2fs_get_node_info(sbi, dn->nid, &new_ni); |
1220 | if (err) { | |
1221 | dec_valid_node_count(sbi, dn->inode, !ofs); | |
1222 | goto fail; | |
1223 | } | |
25cc5d3b JK |
1224 | f2fs_bug_on(sbi, new_ni.blk_addr != NULL_ADDR); |
1225 | #endif | |
1226 | new_ni.nid = dn->nid; | |
e05df3b1 | 1227 | new_ni.ino = dn->inode->i_ino; |
25cc5d3b JK |
1228 | new_ni.blk_addr = NULL_ADDR; |
1229 | new_ni.flag = 0; | |
1230 | new_ni.version = 0; | |
479f40c4 | 1231 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
9c02740c | 1232 | |
fec1d657 | 1233 | f2fs_wait_on_page_writeback(page, NODE, true); |
9c02740c | 1234 | fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); |
c5667575 | 1235 | set_cold_node(page, S_ISDIR(dn->inode->i_mode)); |
237c0790 JK |
1236 | if (!PageUptodate(page)) |
1237 | SetPageUptodate(page); | |
12719ae1 JK |
1238 | if (set_page_dirty(page)) |
1239 | dn->node_changed = true; | |
e05df3b1 | 1240 | |
4bc8e9bc | 1241 | if (f2fs_has_xattr_block(ofs)) |
205b9822 | 1242 | f2fs_i_xnid_write(dn->inode, dn->nid); |
479bd73a | 1243 | |
e05df3b1 JK |
1244 | if (ofs == 0) |
1245 | inc_valid_inode_count(sbi); | |
e05df3b1 JK |
1246 | return page; |
1247 | ||
1248 | fail: | |
71e9fec5 | 1249 | clear_node_page_dirty(page); |
e05df3b1 JK |
1250 | f2fs_put_page(page, 1); |
1251 | return ERR_PTR(err); | |
1252 | } | |
1253 | ||
56ae674c JK |
1254 | /* |
1255 | * Caller should do after getting the following values. | |
1256 | * 0: f2fs_put_page(page, 0) | |
86531d6b | 1257 | * LOCKED_PAGE or error: f2fs_put_page(page, 1) |
56ae674c | 1258 | */ |
04d328de | 1259 | static int read_node_page(struct page *page, int op_flags) |
e05df3b1 | 1260 | { |
4081363f | 1261 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1262 | struct node_info ni; |
cf04e8eb | 1263 | struct f2fs_io_info fio = { |
05ca3632 | 1264 | .sbi = sbi, |
cf04e8eb | 1265 | .type = NODE, |
04d328de MC |
1266 | .op = REQ_OP_READ, |
1267 | .op_flags = op_flags, | |
05ca3632 | 1268 | .page = page, |
4375a336 | 1269 | .encrypted_page = NULL, |
cf04e8eb | 1270 | }; |
7735730d | 1271 | int err; |
e05df3b1 | 1272 | |
54c55c4e WG |
1273 | if (PageUptodate(page)) { |
1274 | #ifdef CONFIG_F2FS_CHECK_FS | |
1275 | f2fs_bug_on(sbi, !f2fs_inode_chksum_verify(sbi, page)); | |
1276 | #endif | |
3bdad3c7 | 1277 | return LOCKED_PAGE; |
54c55c4e | 1278 | } |
3bdad3c7 | 1279 | |
7735730d CY |
1280 | err = f2fs_get_node_info(sbi, page->index, &ni); |
1281 | if (err) | |
1282 | return err; | |
e05df3b1 | 1283 | |
83a3bfdb JK |
1284 | if (unlikely(ni.blk_addr == NULL_ADDR) || |
1285 | is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) { | |
2bca1e23 | 1286 | ClearPageUptodate(page); |
e05df3b1 | 1287 | return -ENOENT; |
393ff91f JK |
1288 | } |
1289 | ||
7a9d7548 | 1290 | fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr; |
05ca3632 | 1291 | return f2fs_submit_page_bio(&fio); |
e05df3b1 JK |
1292 | } |
1293 | ||
0a8165d7 | 1294 | /* |
e05df3b1 JK |
1295 | * Readahead a node page |
1296 | */ | |
4d57b86d | 1297 | void f2fs_ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 | 1298 | { |
e05df3b1 | 1299 | struct page *apage; |
56ae674c | 1300 | int err; |
e05df3b1 | 1301 | |
e8458725 CY |
1302 | if (!nid) |
1303 | return; | |
4d57b86d | 1304 | if (f2fs_check_nid_range(sbi, nid)) |
a4f843bd | 1305 | return; |
e8458725 | 1306 | |
999270de | 1307 | rcu_read_lock(); |
b93b0163 | 1308 | apage = radix_tree_lookup(&NODE_MAPPING(sbi)->i_pages, nid); |
999270de FL |
1309 | rcu_read_unlock(); |
1310 | if (apage) | |
393ff91f | 1311 | return; |
e05df3b1 | 1312 | |
300e129c | 1313 | apage = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false); |
e05df3b1 JK |
1314 | if (!apage) |
1315 | return; | |
1316 | ||
70246286 | 1317 | err = read_node_page(apage, REQ_RAHEAD); |
86531d6b | 1318 | f2fs_put_page(apage, err ? 1 : 0); |
e05df3b1 JK |
1319 | } |
1320 | ||
17a0ee55 | 1321 | static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid, |
0e022ea8 | 1322 | struct page *parent, int start) |
e05df3b1 | 1323 | { |
e05df3b1 | 1324 | struct page *page; |
0e022ea8 | 1325 | int err; |
e05df3b1 | 1326 | |
e05df3b1 JK |
1327 | if (!nid) |
1328 | return ERR_PTR(-ENOENT); | |
4d57b86d | 1329 | if (f2fs_check_nid_range(sbi, nid)) |
a4f843bd | 1330 | return ERR_PTR(-EINVAL); |
afcb7ca0 | 1331 | repeat: |
300e129c | 1332 | page = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false); |
e05df3b1 JK |
1333 | if (!page) |
1334 | return ERR_PTR(-ENOMEM); | |
1335 | ||
70fd7614 | 1336 | err = read_node_page(page, 0); |
86531d6b JK |
1337 | if (err < 0) { |
1338 | f2fs_put_page(page, 1); | |
e05df3b1 | 1339 | return ERR_PTR(err); |
86531d6b | 1340 | } else if (err == LOCKED_PAGE) { |
1f258ec1 | 1341 | err = 0; |
56ae674c | 1342 | goto page_hit; |
86531d6b | 1343 | } |
e05df3b1 | 1344 | |
0e022ea8 | 1345 | if (parent) |
4d57b86d | 1346 | f2fs_ra_node_pages(parent, start + 1, MAX_RA_NODE); |
c718379b | 1347 | |
e05df3b1 | 1348 | lock_page(page); |
0e022ea8 | 1349 | |
4ef51a8f | 1350 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1351 | f2fs_put_page(page, 1); |
1352 | goto repeat; | |
1353 | } | |
1563ac75 | 1354 | |
1f258ec1 CY |
1355 | if (unlikely(!PageUptodate(page))) { |
1356 | err = -EIO; | |
1563ac75 | 1357 | goto out_err; |
1f258ec1 | 1358 | } |
704956ec CY |
1359 | |
1360 | if (!f2fs_inode_chksum_verify(sbi, page)) { | |
1361 | err = -EBADMSG; | |
1362 | goto out_err; | |
1363 | } | |
e0f56cb4 | 1364 | page_hit: |
0c9df7fb | 1365 | if(unlikely(nid != nid_of_node(page))) { |
1f258ec1 CY |
1366 | f2fs_msg(sbi->sb, KERN_WARNING, "inconsistent node block, " |
1367 | "nid:%lu, node_footer[nid:%u,ino:%u,ofs:%u,cpver:%llu,blkaddr:%u]", | |
1368 | nid, nid_of_node(page), ino_of_node(page), | |
1369 | ofs_of_node(page), cpver_of_node(page), | |
1370 | next_blkaddr_of_node(page)); | |
1f258ec1 | 1371 | err = -EINVAL; |
0c9df7fb | 1372 | out_err: |
ee605234 | 1373 | ClearPageUptodate(page); |
0c9df7fb | 1374 | f2fs_put_page(page, 1); |
1f258ec1 | 1375 | return ERR_PTR(err); |
0c9df7fb | 1376 | } |
e05df3b1 JK |
1377 | return page; |
1378 | } | |
1379 | ||
4d57b86d | 1380 | struct page *f2fs_get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) |
0e022ea8 CY |
1381 | { |
1382 | return __get_node_page(sbi, nid, NULL, 0); | |
1383 | } | |
1384 | ||
4d57b86d | 1385 | struct page *f2fs_get_node_page_ra(struct page *parent, int start) |
0e022ea8 CY |
1386 | { |
1387 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); | |
1388 | nid_t nid = get_nid(parent, start, false); | |
1389 | ||
1390 | return __get_node_page(sbi, nid, parent, start); | |
1391 | } | |
1392 | ||
2049d4fc JK |
1393 | static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino) |
1394 | { | |
1395 | struct inode *inode; | |
1396 | struct page *page; | |
0f3311a8 | 1397 | int ret; |
2049d4fc JK |
1398 | |
1399 | /* should flush inline_data before evict_inode */ | |
1400 | inode = ilookup(sbi->sb, ino); | |
1401 | if (!inode) | |
1402 | return; | |
1403 | ||
01eccef7 CY |
1404 | page = f2fs_pagecache_get_page(inode->i_mapping, 0, |
1405 | FGP_LOCK|FGP_NOWAIT, 0); | |
2049d4fc JK |
1406 | if (!page) |
1407 | goto iput_out; | |
1408 | ||
1409 | if (!PageUptodate(page)) | |
1410 | goto page_out; | |
1411 | ||
1412 | if (!PageDirty(page)) | |
1413 | goto page_out; | |
1414 | ||
1415 | if (!clear_page_dirty_for_io(page)) | |
1416 | goto page_out; | |
1417 | ||
0f3311a8 CY |
1418 | ret = f2fs_write_inline_data(inode, page); |
1419 | inode_dec_dirty_pages(inode); | |
4d57b86d | 1420 | f2fs_remove_dirty_inode(inode); |
0f3311a8 | 1421 | if (ret) |
2049d4fc JK |
1422 | set_page_dirty(page); |
1423 | page_out: | |
4a6de50d | 1424 | f2fs_put_page(page, 1); |
2049d4fc JK |
1425 | iput_out: |
1426 | iput(inode); | |
1427 | } | |
1428 | ||
608514de JK |
1429 | static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino) |
1430 | { | |
028a63a6 | 1431 | pgoff_t index; |
608514de JK |
1432 | struct pagevec pvec; |
1433 | struct page *last_page = NULL; | |
028a63a6 | 1434 | int nr_pages; |
608514de | 1435 | |
86679820 | 1436 | pagevec_init(&pvec); |
608514de | 1437 | index = 0; |
028a63a6 JK |
1438 | |
1439 | while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
67fd707f | 1440 | PAGECACHE_TAG_DIRTY))) { |
028a63a6 | 1441 | int i; |
608514de JK |
1442 | |
1443 | for (i = 0; i < nr_pages; i++) { | |
1444 | struct page *page = pvec.pages[i]; | |
1445 | ||
1446 | if (unlikely(f2fs_cp_error(sbi))) { | |
1447 | f2fs_put_page(last_page, 0); | |
1448 | pagevec_release(&pvec); | |
1449 | return ERR_PTR(-EIO); | |
1450 | } | |
1451 | ||
1452 | if (!IS_DNODE(page) || !is_cold_node(page)) | |
1453 | continue; | |
1454 | if (ino_of_node(page) != ino) | |
1455 | continue; | |
1456 | ||
1457 | lock_page(page); | |
1458 | ||
1459 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { | |
1460 | continue_unlock: | |
1461 | unlock_page(page); | |
1462 | continue; | |
1463 | } | |
1464 | if (ino_of_node(page) != ino) | |
1465 | goto continue_unlock; | |
1466 | ||
1467 | if (!PageDirty(page)) { | |
1468 | /* someone wrote it for us */ | |
1469 | goto continue_unlock; | |
1470 | } | |
1471 | ||
1472 | if (last_page) | |
1473 | f2fs_put_page(last_page, 0); | |
1474 | ||
1475 | get_page(page); | |
1476 | last_page = page; | |
1477 | unlock_page(page); | |
1478 | } | |
1479 | pagevec_release(&pvec); | |
1480 | cond_resched(); | |
1481 | } | |
1482 | return last_page; | |
1483 | } | |
1484 | ||
d68f735b | 1485 | static int __write_node_page(struct page *page, bool atomic, bool *submitted, |
b0af6d49 | 1486 | struct writeback_control *wbc, bool do_balance, |
50fa53ec | 1487 | enum iostat_type io_type, unsigned int *seq_id) |
faa24895 JK |
1488 | { |
1489 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); | |
1490 | nid_t nid; | |
1491 | struct node_info ni; | |
1492 | struct f2fs_io_info fio = { | |
1493 | .sbi = sbi, | |
39d787be | 1494 | .ino = ino_of_node(page), |
faa24895 JK |
1495 | .type = NODE, |
1496 | .op = REQ_OP_WRITE, | |
1497 | .op_flags = wbc_to_write_flags(wbc), | |
1498 | .page = page, | |
1499 | .encrypted_page = NULL, | |
d68f735b | 1500 | .submitted = false, |
b0af6d49 | 1501 | .io_type = io_type, |
578c6478 | 1502 | .io_wbc = wbc, |
faa24895 | 1503 | }; |
50fa53ec | 1504 | unsigned int seq; |
faa24895 JK |
1505 | |
1506 | trace_f2fs_writepage(page, NODE); | |
1507 | ||
868de613 JK |
1508 | if (unlikely(f2fs_cp_error(sbi))) |
1509 | goto redirty_out; | |
db198ae0 | 1510 | |
faa24895 JK |
1511 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
1512 | goto redirty_out; | |
faa24895 | 1513 | |
fd8c8caf CY |
1514 | if (wbc->sync_mode == WB_SYNC_NONE && |
1515 | IS_DNODE(page) && is_cold_node(page)) | |
1516 | goto redirty_out; | |
1517 | ||
faa24895 JK |
1518 | /* get old block addr of this node page */ |
1519 | nid = nid_of_node(page); | |
1520 | f2fs_bug_on(sbi, page->index != nid); | |
1521 | ||
7735730d CY |
1522 | if (f2fs_get_node_info(sbi, nid, &ni)) |
1523 | goto redirty_out; | |
1524 | ||
faa24895 JK |
1525 | if (wbc->for_reclaim) { |
1526 | if (!down_read_trylock(&sbi->node_write)) | |
1527 | goto redirty_out; | |
1528 | } else { | |
1529 | down_read(&sbi->node_write); | |
1530 | } | |
1531 | ||
faa24895 JK |
1532 | /* This page is already truncated */ |
1533 | if (unlikely(ni.blk_addr == NULL_ADDR)) { | |
1534 | ClearPageUptodate(page); | |
1535 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1536 | up_read(&sbi->node_write); | |
1537 | unlock_page(page); | |
1538 | return 0; | |
1539 | } | |
1540 | ||
c9b60788 CY |
1541 | if (__is_valid_data_blkaddr(ni.blk_addr) && |
1542 | !f2fs_is_valid_blkaddr(sbi, ni.blk_addr, DATA_GENERIC)) | |
1543 | goto redirty_out; | |
1544 | ||
e7c75ab0 JK |
1545 | if (atomic && !test_opt(sbi, NOBARRIER)) |
1546 | fio.op_flags |= REQ_PREFLUSH | REQ_FUA; | |
1547 | ||
faa24895 | 1548 | set_page_writeback(page); |
17c50035 | 1549 | ClearPageError(page); |
50fa53ec CY |
1550 | |
1551 | if (f2fs_in_warm_node_list(sbi, page)) { | |
1552 | seq = f2fs_add_fsync_node_entry(sbi, page); | |
1553 | if (seq_id) | |
1554 | *seq_id = seq; | |
1555 | } | |
1556 | ||
faa24895 | 1557 | fio.old_blkaddr = ni.blk_addr; |
4d57b86d | 1558 | f2fs_do_write_node_page(nid, &fio); |
faa24895 JK |
1559 | set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page)); |
1560 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1561 | up_read(&sbi->node_write); | |
1562 | ||
d68f735b | 1563 | if (wbc->for_reclaim) { |
b9109b0e JK |
1564 | f2fs_submit_merged_write_cond(sbi, page->mapping->host, 0, |
1565 | page->index, NODE); | |
d68f735b JK |
1566 | submitted = NULL; |
1567 | } | |
faa24895 JK |
1568 | |
1569 | unlock_page(page); | |
1570 | ||
d68f735b | 1571 | if (unlikely(f2fs_cp_error(sbi))) { |
b9109b0e | 1572 | f2fs_submit_merged_write(sbi, NODE); |
d68f735b JK |
1573 | submitted = NULL; |
1574 | } | |
1575 | if (submitted) | |
1576 | *submitted = fio.submitted; | |
faa24895 | 1577 | |
401db79f YS |
1578 | if (do_balance) |
1579 | f2fs_balance_fs(sbi, false); | |
faa24895 JK |
1580 | return 0; |
1581 | ||
1582 | redirty_out: | |
1583 | redirty_page_for_writepage(wbc, page); | |
1584 | return AOP_WRITEPAGE_ACTIVATE; | |
1585 | } | |
1586 | ||
4d57b86d | 1587 | void f2fs_move_node_page(struct page *node_page, int gc_type) |
f15194fc YH |
1588 | { |
1589 | if (gc_type == FG_GC) { | |
1590 | struct writeback_control wbc = { | |
1591 | .sync_mode = WB_SYNC_ALL, | |
1592 | .nr_to_write = 1, | |
1593 | .for_reclaim = 0, | |
1594 | }; | |
1595 | ||
1596 | set_page_dirty(node_page); | |
1597 | f2fs_wait_on_page_writeback(node_page, NODE, true); | |
1598 | ||
1599 | f2fs_bug_on(F2FS_P_SB(node_page), PageWriteback(node_page)); | |
1600 | if (!clear_page_dirty_for_io(node_page)) | |
1601 | goto out_page; | |
1602 | ||
1603 | if (__write_node_page(node_page, false, NULL, | |
50fa53ec | 1604 | &wbc, false, FS_GC_NODE_IO, NULL)) |
f15194fc YH |
1605 | unlock_page(node_page); |
1606 | goto release_page; | |
1607 | } else { | |
1608 | /* set page dirty and write it */ | |
1609 | if (!PageWriteback(node_page)) | |
1610 | set_page_dirty(node_page); | |
1611 | } | |
1612 | out_page: | |
1613 | unlock_page(node_page); | |
1614 | release_page: | |
1615 | f2fs_put_page(node_page, 0); | |
1616 | } | |
1617 | ||
faa24895 JK |
1618 | static int f2fs_write_node_page(struct page *page, |
1619 | struct writeback_control *wbc) | |
1620 | { | |
50fa53ec CY |
1621 | return __write_node_page(page, false, NULL, wbc, false, |
1622 | FS_NODE_IO, NULL); | |
faa24895 JK |
1623 | } |
1624 | ||
4d57b86d | 1625 | int f2fs_fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode, |
50fa53ec CY |
1626 | struct writeback_control *wbc, bool atomic, |
1627 | unsigned int *seq_id) | |
e05df3b1 | 1628 | { |
028a63a6 | 1629 | pgoff_t index; |
942fd319 | 1630 | pgoff_t last_idx = ULONG_MAX; |
e05df3b1 | 1631 | struct pagevec pvec; |
c267ec15 | 1632 | int ret = 0; |
608514de JK |
1633 | struct page *last_page = NULL; |
1634 | bool marked = false; | |
26de9b11 | 1635 | nid_t ino = inode->i_ino; |
028a63a6 | 1636 | int nr_pages; |
52681375 | 1637 | |
608514de JK |
1638 | if (atomic) { |
1639 | last_page = last_fsync_dnode(sbi, ino); | |
1640 | if (IS_ERR_OR_NULL(last_page)) | |
1641 | return PTR_ERR_OR_ZERO(last_page); | |
1642 | } | |
1643 | retry: | |
86679820 | 1644 | pagevec_init(&pvec); |
52681375 | 1645 | index = 0; |
028a63a6 JK |
1646 | |
1647 | while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
67fd707f | 1648 | PAGECACHE_TAG_DIRTY))) { |
028a63a6 | 1649 | int i; |
52681375 JK |
1650 | |
1651 | for (i = 0; i < nr_pages; i++) { | |
1652 | struct page *page = pvec.pages[i]; | |
d68f735b | 1653 | bool submitted = false; |
52681375 JK |
1654 | |
1655 | if (unlikely(f2fs_cp_error(sbi))) { | |
608514de | 1656 | f2fs_put_page(last_page, 0); |
52681375 | 1657 | pagevec_release(&pvec); |
9de69279 CY |
1658 | ret = -EIO; |
1659 | goto out; | |
52681375 JK |
1660 | } |
1661 | ||
1662 | if (!IS_DNODE(page) || !is_cold_node(page)) | |
1663 | continue; | |
1664 | if (ino_of_node(page) != ino) | |
1665 | continue; | |
1666 | ||
1667 | lock_page(page); | |
1668 | ||
1669 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { | |
1670 | continue_unlock: | |
1671 | unlock_page(page); | |
1672 | continue; | |
1673 | } | |
1674 | if (ino_of_node(page) != ino) | |
1675 | goto continue_unlock; | |
1676 | ||
608514de | 1677 | if (!PageDirty(page) && page != last_page) { |
52681375 JK |
1678 | /* someone wrote it for us */ |
1679 | goto continue_unlock; | |
1680 | } | |
1681 | ||
1682 | f2fs_wait_on_page_writeback(page, NODE, true); | |
1683 | BUG_ON(PageWriteback(page)); | |
52681375 | 1684 | |
d29fd172 JK |
1685 | set_fsync_mark(page, 0); |
1686 | set_dentry_mark(page, 0); | |
1687 | ||
608514de JK |
1688 | if (!atomic || page == last_page) { |
1689 | set_fsync_mark(page, 1); | |
26de9b11 JK |
1690 | if (IS_INODE(page)) { |
1691 | if (is_inode_flag_set(inode, | |
1692 | FI_DIRTY_INODE)) | |
4d57b86d | 1693 | f2fs_update_inode(inode, page); |
608514de | 1694 | set_dentry_mark(page, |
4d57b86d | 1695 | f2fs_need_dentry_mark(sbi, ino)); |
26de9b11 | 1696 | } |
608514de JK |
1697 | /* may be written by other thread */ |
1698 | if (!PageDirty(page)) | |
1699 | set_page_dirty(page); | |
1700 | } | |
1701 | ||
1702 | if (!clear_page_dirty_for_io(page)) | |
1703 | goto continue_unlock; | |
52681375 | 1704 | |
e7c75ab0 | 1705 | ret = __write_node_page(page, atomic && |
d68f735b | 1706 | page == last_page, |
b0af6d49 | 1707 | &submitted, wbc, true, |
50fa53ec | 1708 | FS_NODE_IO, seq_id); |
c267ec15 | 1709 | if (ret) { |
52681375 | 1710 | unlock_page(page); |
608514de JK |
1711 | f2fs_put_page(last_page, 0); |
1712 | break; | |
d68f735b | 1713 | } else if (submitted) { |
942fd319 | 1714 | last_idx = page->index; |
608514de | 1715 | } |
3f5f4959 | 1716 | |
608514de JK |
1717 | if (page == last_page) { |
1718 | f2fs_put_page(page, 0); | |
1719 | marked = true; | |
52681375 | 1720 | break; |
c267ec15 | 1721 | } |
52681375 JK |
1722 | } |
1723 | pagevec_release(&pvec); | |
1724 | cond_resched(); | |
1725 | ||
608514de | 1726 | if (ret || marked) |
52681375 JK |
1727 | break; |
1728 | } | |
608514de JK |
1729 | if (!ret && atomic && !marked) { |
1730 | f2fs_msg(sbi->sb, KERN_DEBUG, | |
1731 | "Retry to write fsync mark: ino=%u, idx=%lx", | |
1732 | ino, last_page->index); | |
1733 | lock_page(last_page); | |
d40a43af | 1734 | f2fs_wait_on_page_writeback(last_page, NODE, true); |
608514de JK |
1735 | set_page_dirty(last_page); |
1736 | unlock_page(last_page); | |
1737 | goto retry; | |
1738 | } | |
9de69279 | 1739 | out: |
942fd319 | 1740 | if (last_idx != ULONG_MAX) |
b9109b0e | 1741 | f2fs_submit_merged_write_cond(sbi, NULL, ino, last_idx, NODE); |
c267ec15 | 1742 | return ret ? -EIO: 0; |
52681375 JK |
1743 | } |
1744 | ||
4d57b86d CY |
1745 | int f2fs_sync_node_pages(struct f2fs_sb_info *sbi, |
1746 | struct writeback_control *wbc, | |
b0af6d49 | 1747 | bool do_balance, enum iostat_type io_type) |
52681375 | 1748 | { |
028a63a6 | 1749 | pgoff_t index; |
52681375 JK |
1750 | struct pagevec pvec; |
1751 | int step = 0; | |
12bb0a8f | 1752 | int nwritten = 0; |
3f5f4959 | 1753 | int ret = 0; |
c29fd0c0 | 1754 | int nr_pages, done = 0; |
e05df3b1 | 1755 | |
86679820 | 1756 | pagevec_init(&pvec); |
e05df3b1 JK |
1757 | |
1758 | next_step: | |
1759 | index = 0; | |
028a63a6 | 1760 | |
c29fd0c0 CY |
1761 | while (!done && (nr_pages = pagevec_lookup_tag(&pvec, |
1762 | NODE_MAPPING(sbi), &index, PAGECACHE_TAG_DIRTY))) { | |
028a63a6 | 1763 | int i; |
e05df3b1 JK |
1764 | |
1765 | for (i = 0; i < nr_pages; i++) { | |
1766 | struct page *page = pvec.pages[i]; | |
d68f735b | 1767 | bool submitted = false; |
e05df3b1 | 1768 | |
c29fd0c0 CY |
1769 | /* give a priority to WB_SYNC threads */ |
1770 | if (atomic_read(&sbi->wb_sync_req[NODE]) && | |
1771 | wbc->sync_mode == WB_SYNC_NONE) { | |
1772 | done = 1; | |
1773 | break; | |
1774 | } | |
1775 | ||
e05df3b1 JK |
1776 | /* |
1777 | * flushing sequence with step: | |
1778 | * 0. indirect nodes | |
1779 | * 1. dentry dnodes | |
1780 | * 2. file dnodes | |
1781 | */ | |
1782 | if (step == 0 && IS_DNODE(page)) | |
1783 | continue; | |
1784 | if (step == 1 && (!IS_DNODE(page) || | |
1785 | is_cold_node(page))) | |
1786 | continue; | |
1787 | if (step == 2 && (!IS_DNODE(page) || | |
1788 | !is_cold_node(page))) | |
1789 | continue; | |
9a4cbc9e | 1790 | lock_node: |
4b270a8c CY |
1791 | if (wbc->sync_mode == WB_SYNC_ALL) |
1792 | lock_page(page); | |
1793 | else if (!trylock_page(page)) | |
e05df3b1 JK |
1794 | continue; |
1795 | ||
4ef51a8f | 1796 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1797 | continue_unlock: |
1798 | unlock_page(page); | |
1799 | continue; | |
1800 | } | |
e05df3b1 JK |
1801 | |
1802 | if (!PageDirty(page)) { | |
1803 | /* someone wrote it for us */ | |
1804 | goto continue_unlock; | |
1805 | } | |
1806 | ||
2049d4fc | 1807 | /* flush inline_data */ |
52681375 | 1808 | if (is_inline_node(page)) { |
2049d4fc JK |
1809 | clear_inline_node(page); |
1810 | unlock_page(page); | |
1811 | flush_inline_data(sbi, ino_of_node(page)); | |
9a4cbc9e | 1812 | goto lock_node; |
2049d4fc JK |
1813 | } |
1814 | ||
fa3d2bdf JK |
1815 | f2fs_wait_on_page_writeback(page, NODE, true); |
1816 | ||
1817 | BUG_ON(PageWriteback(page)); | |
e05df3b1 JK |
1818 | if (!clear_page_dirty_for_io(page)) |
1819 | goto continue_unlock; | |
1820 | ||
52681375 JK |
1821 | set_fsync_mark(page, 0); |
1822 | set_dentry_mark(page, 0); | |
52746519 | 1823 | |
401db79f | 1824 | ret = __write_node_page(page, false, &submitted, |
50fa53ec | 1825 | wbc, do_balance, io_type, NULL); |
d68f735b | 1826 | if (ret) |
52746519 | 1827 | unlock_page(page); |
d68f735b | 1828 | else if (submitted) |
3f5f4959 | 1829 | nwritten++; |
e05df3b1 JK |
1830 | |
1831 | if (--wbc->nr_to_write == 0) | |
1832 | break; | |
1833 | } | |
1834 | pagevec_release(&pvec); | |
1835 | cond_resched(); | |
1836 | ||
1837 | if (wbc->nr_to_write == 0) { | |
1838 | step = 2; | |
1839 | break; | |
1840 | } | |
1841 | } | |
1842 | ||
1843 | if (step < 2) { | |
fd8c8caf CY |
1844 | if (wbc->sync_mode == WB_SYNC_NONE && step == 1) |
1845 | goto out; | |
e05df3b1 JK |
1846 | step++; |
1847 | goto next_step; | |
1848 | } | |
fd8c8caf | 1849 | out: |
3f5f4959 | 1850 | if (nwritten) |
b9109b0e | 1851 | f2fs_submit_merged_write(sbi, NODE); |
db198ae0 CY |
1852 | |
1853 | if (unlikely(f2fs_cp_error(sbi))) | |
1854 | return -EIO; | |
3f5f4959 | 1855 | return ret; |
e05df3b1 JK |
1856 | } |
1857 | ||
50fa53ec CY |
1858 | int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, |
1859 | unsigned int seq_id) | |
cfe58f9d | 1860 | { |
50fa53ec CY |
1861 | struct fsync_node_entry *fn; |
1862 | struct page *page; | |
1863 | struct list_head *head = &sbi->fsync_node_list; | |
1864 | unsigned long flags; | |
1865 | unsigned int cur_seq_id = 0; | |
280db3c8 | 1866 | int ret2, ret = 0; |
cfe58f9d | 1867 | |
50fa53ec CY |
1868 | while (seq_id && cur_seq_id < seq_id) { |
1869 | spin_lock_irqsave(&sbi->fsync_node_lock, flags); | |
1870 | if (list_empty(head)) { | |
1871 | spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); | |
1872 | break; | |
1873 | } | |
1874 | fn = list_first_entry(head, struct fsync_node_entry, list); | |
1875 | if (fn->seq_id > seq_id) { | |
1876 | spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); | |
1877 | break; | |
1878 | } | |
1879 | cur_seq_id = fn->seq_id; | |
1880 | page = fn->page; | |
1881 | get_page(page); | |
1882 | spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); | |
4ef51a8f | 1883 | |
50fa53ec CY |
1884 | f2fs_wait_on_page_writeback(page, NODE, true); |
1885 | if (TestClearPageError(page)) | |
1886 | ret = -EIO; | |
cfe58f9d | 1887 | |
50fa53ec | 1888 | put_page(page); |
cfe58f9d | 1889 | |
50fa53ec CY |
1890 | if (ret) |
1891 | break; | |
cfe58f9d JK |
1892 | } |
1893 | ||
280db3c8 | 1894 | ret2 = filemap_check_errors(NODE_MAPPING(sbi)); |
cfe58f9d JK |
1895 | if (!ret) |
1896 | ret = ret2; | |
50fa53ec | 1897 | |
cfe58f9d JK |
1898 | return ret; |
1899 | } | |
1900 | ||
e05df3b1 JK |
1901 | static int f2fs_write_node_pages(struct address_space *mapping, |
1902 | struct writeback_control *wbc) | |
1903 | { | |
4081363f | 1904 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
9dfa1baf | 1905 | struct blk_plug plug; |
50c8cdb3 | 1906 | long diff; |
e05df3b1 | 1907 | |
0771fcc7 CY |
1908 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
1909 | goto skip_write; | |
1910 | ||
4660f9c0 JK |
1911 | /* balancing f2fs's metadata in background */ |
1912 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1913 | |
a7fdffbd | 1914 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1915 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1916 | goto skip_write; |
a7fdffbd | 1917 | |
c29fd0c0 CY |
1918 | if (wbc->sync_mode == WB_SYNC_ALL) |
1919 | atomic_inc(&sbi->wb_sync_req[NODE]); | |
1920 | else if (atomic_read(&sbi->wb_sync_req[NODE])) | |
1921 | goto skip_write; | |
1922 | ||
d31c7c3f YH |
1923 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1924 | ||
50c8cdb3 | 1925 | diff = nr_pages_to_write(sbi, NODE, wbc); |
9dfa1baf | 1926 | blk_start_plug(&plug); |
4d57b86d | 1927 | f2fs_sync_node_pages(sbi, wbc, true, FS_NODE_IO); |
9dfa1baf | 1928 | blk_finish_plug(&plug); |
50c8cdb3 | 1929 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
c29fd0c0 CY |
1930 | |
1931 | if (wbc->sync_mode == WB_SYNC_ALL) | |
1932 | atomic_dec(&sbi->wb_sync_req[NODE]); | |
e05df3b1 | 1933 | return 0; |
d3baf95d JK |
1934 | |
1935 | skip_write: | |
1936 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
d31c7c3f | 1937 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
d3baf95d | 1938 | return 0; |
e05df3b1 JK |
1939 | } |
1940 | ||
1941 | static int f2fs_set_node_page_dirty(struct page *page) | |
1942 | { | |
26c6b887 JK |
1943 | trace_f2fs_set_page_dirty(page, NODE); |
1944 | ||
237c0790 JK |
1945 | if (!PageUptodate(page)) |
1946 | SetPageUptodate(page); | |
54c55c4e WG |
1947 | #ifdef CONFIG_F2FS_CHECK_FS |
1948 | if (IS_INODE(page)) | |
1949 | f2fs_inode_chksum_set(F2FS_P_SB(page), page); | |
1950 | #endif | |
e05df3b1 | 1951 | if (!PageDirty(page)) { |
b87078ad | 1952 | __set_page_dirty_nobuffers(page); |
4081363f | 1953 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 | 1954 | SetPagePrivate(page); |
9e4ded3f | 1955 | f2fs_trace_pid(page); |
e05df3b1 JK |
1956 | return 1; |
1957 | } | |
1958 | return 0; | |
1959 | } | |
1960 | ||
0a8165d7 | 1961 | /* |
e05df3b1 JK |
1962 | * Structure of the f2fs node operations |
1963 | */ | |
1964 | const struct address_space_operations f2fs_node_aops = { | |
1965 | .writepage = f2fs_write_node_page, | |
1966 | .writepages = f2fs_write_node_pages, | |
1967 | .set_page_dirty = f2fs_set_node_page_dirty, | |
487261f3 CY |
1968 | .invalidatepage = f2fs_invalidate_page, |
1969 | .releasepage = f2fs_release_page, | |
5b7a487c WG |
1970 | #ifdef CONFIG_MIGRATION |
1971 | .migratepage = f2fs_migrate_page, | |
1972 | #endif | |
e05df3b1 JK |
1973 | }; |
1974 | ||
8a7ed66a JK |
1975 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1976 | nid_t n) | |
e05df3b1 | 1977 | { |
8a7ed66a | 1978 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1979 | } |
1980 | ||
9a4ffdf5 | 1981 | static int __insert_free_nid(struct f2fs_sb_info *sbi, |
a0761f63 | 1982 | struct free_nid *i, enum nid_state state) |
e05df3b1 | 1983 | { |
b8559dc2 CY |
1984 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1985 | ||
a0761f63 FL |
1986 | int err = radix_tree_insert(&nm_i->free_nid_root, i->nid, i); |
1987 | if (err) | |
1988 | return err; | |
eb0aa4b8 | 1989 | |
9a4ffdf5 CY |
1990 | f2fs_bug_on(sbi, state != i->state); |
1991 | nm_i->nid_cnt[state]++; | |
1992 | if (state == FREE_NID) | |
1993 | list_add_tail(&i->list, &nm_i->free_nid_list); | |
eb0aa4b8 | 1994 | return 0; |
b8559dc2 CY |
1995 | } |
1996 | ||
9a4ffdf5 | 1997 | static void __remove_free_nid(struct f2fs_sb_info *sbi, |
a0761f63 | 1998 | struct free_nid *i, enum nid_state state) |
b8559dc2 CY |
1999 | { |
2000 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2001 | ||
9a4ffdf5 CY |
2002 | f2fs_bug_on(sbi, state != i->state); |
2003 | nm_i->nid_cnt[state]--; | |
2004 | if (state == FREE_NID) | |
2005 | list_del(&i->list); | |
a0761f63 FL |
2006 | radix_tree_delete(&nm_i->free_nid_root, i->nid); |
2007 | } | |
2008 | ||
2009 | static void __move_free_nid(struct f2fs_sb_info *sbi, struct free_nid *i, | |
2010 | enum nid_state org_state, enum nid_state dst_state) | |
b8559dc2 CY |
2011 | { |
2012 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2013 | ||
a0761f63 FL |
2014 | f2fs_bug_on(sbi, org_state != i->state); |
2015 | i->state = dst_state; | |
2016 | nm_i->nid_cnt[org_state]--; | |
2017 | nm_i->nid_cnt[dst_state]++; | |
2018 | ||
2019 | switch (dst_state) { | |
2020 | case PREALLOC_NID: | |
2021 | list_del(&i->list); | |
2022 | break; | |
2023 | case FREE_NID: | |
2024 | list_add_tail(&i->list, &nm_i->free_nid_list); | |
2025 | break; | |
2026 | default: | |
2027 | BUG_ON(1); | |
2028 | } | |
e05df3b1 JK |
2029 | } |
2030 | ||
5921aaa1 L |
2031 | static void update_free_nid_bitmap(struct f2fs_sb_info *sbi, nid_t nid, |
2032 | bool set, bool build) | |
2033 | { | |
2034 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2035 | unsigned int nat_ofs = NAT_BLOCK_OFFSET(nid); | |
2036 | unsigned int nid_ofs = nid - START_NID(nid); | |
2037 | ||
2038 | if (!test_bit_le(nat_ofs, nm_i->nat_block_bitmap)) | |
2039 | return; | |
2040 | ||
2041 | if (set) { | |
2042 | if (test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs])) | |
2043 | return; | |
2044 | __set_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]); | |
2045 | nm_i->free_nid_count[nat_ofs]++; | |
2046 | } else { | |
2047 | if (!test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs])) | |
2048 | return; | |
2049 | __clear_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]); | |
2050 | if (!build) | |
2051 | nm_i->free_nid_count[nat_ofs]--; | |
2052 | } | |
2053 | } | |
2054 | ||
4ac91242 | 2055 | /* return if the nid is recognized as free */ |
5921aaa1 L |
2056 | static bool add_free_nid(struct f2fs_sb_info *sbi, |
2057 | nid_t nid, bool build, bool update) | |
e05df3b1 | 2058 | { |
6fb03f3a | 2059 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
30a61ddf | 2060 | struct free_nid *i, *e; |
59bbd474 | 2061 | struct nat_entry *ne; |
30a61ddf CY |
2062 | int err = -EINVAL; |
2063 | bool ret = false; | |
9198aceb JK |
2064 | |
2065 | /* 0 nid should not be used */ | |
cfb271d4 | 2066 | if (unlikely(nid == 0)) |
4ac91242 | 2067 | return false; |
59bbd474 | 2068 | |
7bd59381 | 2069 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); |
e05df3b1 | 2070 | i->nid = nid; |
9a4ffdf5 | 2071 | i->state = FREE_NID; |
e05df3b1 | 2072 | |
5921aaa1 | 2073 | radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); |
769ec6e5 | 2074 | |
b8559dc2 | 2075 | spin_lock(&nm_i->nid_list_lock); |
30a61ddf CY |
2076 | |
2077 | if (build) { | |
2078 | /* | |
2079 | * Thread A Thread B | |
2080 | * - f2fs_create | |
2081 | * - f2fs_new_inode | |
4d57b86d | 2082 | * - f2fs_alloc_nid |
9a4ffdf5 | 2083 | * - __insert_nid_to_list(PREALLOC_NID) |
30a61ddf | 2084 | * - f2fs_balance_fs_bg |
4d57b86d CY |
2085 | * - f2fs_build_free_nids |
2086 | * - __f2fs_build_free_nids | |
30a61ddf CY |
2087 | * - scan_nat_page |
2088 | * - add_free_nid | |
2089 | * - __lookup_nat_cache | |
2090 | * - f2fs_add_link | |
4d57b86d CY |
2091 | * - f2fs_init_inode_metadata |
2092 | * - f2fs_new_inode_page | |
2093 | * - f2fs_new_node_page | |
30a61ddf | 2094 | * - set_node_addr |
4d57b86d | 2095 | * - f2fs_alloc_nid_done |
9a4ffdf5 CY |
2096 | * - __remove_nid_from_list(PREALLOC_NID) |
2097 | * - __insert_nid_to_list(FREE_NID) | |
30a61ddf CY |
2098 | */ |
2099 | ne = __lookup_nat_cache(nm_i, nid); | |
2100 | if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) || | |
2101 | nat_get_blkaddr(ne) != NULL_ADDR)) | |
2102 | goto err_out; | |
2103 | ||
2104 | e = __lookup_free_nid_list(nm_i, nid); | |
2105 | if (e) { | |
9a4ffdf5 | 2106 | if (e->state == FREE_NID) |
30a61ddf CY |
2107 | ret = true; |
2108 | goto err_out; | |
2109 | } | |
2110 | } | |
2111 | ret = true; | |
a0761f63 | 2112 | err = __insert_free_nid(sbi, i, FREE_NID); |
30a61ddf | 2113 | err_out: |
5921aaa1 L |
2114 | if (update) { |
2115 | update_free_nid_bitmap(sbi, nid, ret, build); | |
2116 | if (!build) | |
2117 | nm_i->available_nids++; | |
2118 | } | |
eb0aa4b8 JK |
2119 | spin_unlock(&nm_i->nid_list_lock); |
2120 | radix_tree_preload_end(); | |
5921aaa1 | 2121 | |
30a61ddf | 2122 | if (err) |
e05df3b1 | 2123 | kmem_cache_free(free_nid_slab, i); |
30a61ddf | 2124 | return ret; |
e05df3b1 JK |
2125 | } |
2126 | ||
b8559dc2 | 2127 | static void remove_free_nid(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 | 2128 | { |
b8559dc2 | 2129 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 2130 | struct free_nid *i; |
cf0ee0f0 CY |
2131 | bool need_free = false; |
2132 | ||
b8559dc2 | 2133 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 2134 | i = __lookup_free_nid_list(nm_i, nid); |
9a4ffdf5 | 2135 | if (i && i->state == FREE_NID) { |
a0761f63 | 2136 | __remove_free_nid(sbi, i, FREE_NID); |
cf0ee0f0 | 2137 | need_free = true; |
e05df3b1 | 2138 | } |
b8559dc2 | 2139 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
2140 | |
2141 | if (need_free) | |
2142 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
2143 | } |
2144 | ||
e2374015 | 2145 | static int scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
2146 | struct page *nat_page, nid_t start_nid) |
2147 | { | |
6fb03f3a | 2148 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
2149 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
2150 | block_t blk_addr; | |
4ac91242 | 2151 | unsigned int nat_ofs = NAT_BLOCK_OFFSET(start_nid); |
e05df3b1 JK |
2152 | int i; |
2153 | ||
23380b85 | 2154 | __set_bit_le(nat_ofs, nm_i->nat_block_bitmap); |
4ac91242 | 2155 | |
e05df3b1 JK |
2156 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
2157 | ||
2158 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
cfb271d4 | 2159 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 2160 | break; |
23d38844 HL |
2161 | |
2162 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
e2374015 CY |
2163 | |
2164 | if (blk_addr == NEW_ADDR) | |
2165 | return -EINVAL; | |
2166 | ||
5921aaa1 L |
2167 | if (blk_addr == NULL_ADDR) { |
2168 | add_free_nid(sbi, start_nid, true, true); | |
2169 | } else { | |
2170 | spin_lock(&NM_I(sbi)->nid_list_lock); | |
2171 | update_free_nid_bitmap(sbi, start_nid, false, true); | |
2172 | spin_unlock(&NM_I(sbi)->nid_list_lock); | |
2173 | } | |
4ac91242 | 2174 | } |
e2374015 CY |
2175 | |
2176 | return 0; | |
4ac91242 CY |
2177 | } |
2178 | ||
2fbaa25f | 2179 | static void scan_curseg_cache(struct f2fs_sb_info *sbi) |
4ac91242 | 2180 | { |
4ac91242 CY |
2181 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); |
2182 | struct f2fs_journal *journal = curseg->journal; | |
2fbaa25f CY |
2183 | int i; |
2184 | ||
2185 | down_read(&curseg->journal_rwsem); | |
2186 | for (i = 0; i < nats_in_cursum(journal); i++) { | |
2187 | block_t addr; | |
2188 | nid_t nid; | |
2189 | ||
2190 | addr = le32_to_cpu(nat_in_journal(journal, i).block_addr); | |
2191 | nid = le32_to_cpu(nid_in_journal(journal, i)); | |
2192 | if (addr == NULL_ADDR) | |
5921aaa1 | 2193 | add_free_nid(sbi, nid, true, false); |
2fbaa25f CY |
2194 | else |
2195 | remove_free_nid(sbi, nid); | |
2196 | } | |
2197 | up_read(&curseg->journal_rwsem); | |
2198 | } | |
2199 | ||
2200 | static void scan_free_nid_bits(struct f2fs_sb_info *sbi) | |
2201 | { | |
2202 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
4ac91242 | 2203 | unsigned int i, idx; |
97456574 | 2204 | nid_t nid; |
4ac91242 CY |
2205 | |
2206 | down_read(&nm_i->nat_tree_lock); | |
2207 | ||
2208 | for (i = 0; i < nm_i->nat_blocks; i++) { | |
2209 | if (!test_bit_le(i, nm_i->nat_block_bitmap)) | |
2210 | continue; | |
586d1492 CY |
2211 | if (!nm_i->free_nid_count[i]) |
2212 | continue; | |
4ac91242 | 2213 | for (idx = 0; idx < NAT_ENTRY_PER_BLOCK; idx++) { |
97456574 FL |
2214 | idx = find_next_bit_le(nm_i->free_nid_bitmap[i], |
2215 | NAT_ENTRY_PER_BLOCK, idx); | |
2216 | if (idx >= NAT_ENTRY_PER_BLOCK) | |
2217 | break; | |
4ac91242 CY |
2218 | |
2219 | nid = i * NAT_ENTRY_PER_BLOCK + idx; | |
5921aaa1 | 2220 | add_free_nid(sbi, nid, true, false); |
4ac91242 | 2221 | |
9a4ffdf5 | 2222 | if (nm_i->nid_cnt[FREE_NID] >= MAX_FREE_NIDS) |
4ac91242 CY |
2223 | goto out; |
2224 | } | |
2225 | } | |
2226 | out: | |
2fbaa25f | 2227 | scan_curseg_cache(sbi); |
4ac91242 | 2228 | |
4ac91242 | 2229 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2230 | } |
2231 | ||
e2374015 | 2232 | static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi, |
4d57b86d | 2233 | bool sync, bool mount) |
e05df3b1 | 2234 | { |
e05df3b1 | 2235 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e2374015 | 2236 | int i = 0, ret; |
55008d84 | 2237 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 2238 | |
e9cdd307 YH |
2239 | if (unlikely(nid >= nm_i->max_nid)) |
2240 | nid = 0; | |
2241 | ||
55008d84 | 2242 | /* Enough entries */ |
9a4ffdf5 | 2243 | if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK) |
e2374015 | 2244 | return 0; |
e05df3b1 | 2245 | |
4d57b86d | 2246 | if (!sync && !f2fs_available_free_memory(sbi, FREE_NIDS)) |
e2374015 | 2247 | return 0; |
e05df3b1 | 2248 | |
4ac91242 CY |
2249 | if (!mount) { |
2250 | /* try to find free nids in free_nid_bitmap */ | |
2251 | scan_free_nid_bits(sbi); | |
2252 | ||
74986213 | 2253 | if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK) |
e2374015 | 2254 | return 0; |
22ad0b6a JK |
2255 | } |
2256 | ||
55008d84 | 2257 | /* readahead nat pages to be scanned */ |
4d57b86d | 2258 | f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, |
26879fb1 | 2259 | META_NAT, true); |
e05df3b1 | 2260 | |
b873b798 | 2261 | down_read(&nm_i->nat_tree_lock); |
a5131193 | 2262 | |
e05df3b1 | 2263 | while (1) { |
66e83361 YH |
2264 | if (!test_bit_le(NAT_BLOCK_OFFSET(nid), |
2265 | nm_i->nat_block_bitmap)) { | |
2266 | struct page *page = get_current_nat_page(sbi, nid); | |
e05df3b1 | 2267 | |
e2374015 | 2268 | ret = scan_nat_page(sbi, page, nid); |
66e83361 | 2269 | f2fs_put_page(page, 1); |
e2374015 CY |
2270 | |
2271 | if (ret) { | |
2272 | up_read(&nm_i->nat_tree_lock); | |
2273 | f2fs_bug_on(sbi, !mount); | |
2274 | f2fs_msg(sbi->sb, KERN_ERR, | |
2275 | "NAT is corrupt, run fsck to fix it"); | |
2276 | return -EINVAL; | |
2277 | } | |
66e83361 | 2278 | } |
e05df3b1 JK |
2279 | |
2280 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 2281 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 2282 | nid = 0; |
55008d84 | 2283 | |
a6d494b6 | 2284 | if (++i >= FREE_NID_PAGES) |
e05df3b1 JK |
2285 | break; |
2286 | } | |
2287 | ||
55008d84 JK |
2288 | /* go to the next free nat pages to find free nids abundantly */ |
2289 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
2290 | |
2291 | /* find free nids from current sum_pages */ | |
2fbaa25f | 2292 | scan_curseg_cache(sbi); |
dfc08a12 | 2293 | |
b873b798 | 2294 | up_read(&nm_i->nat_tree_lock); |
2db2388f | 2295 | |
4d57b86d | 2296 | f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid), |
ea1a29a0 | 2297 | nm_i->ra_nid_pages, META_NAT, false); |
e2374015 CY |
2298 | |
2299 | return 0; | |
e05df3b1 JK |
2300 | } |
2301 | ||
e2374015 | 2302 | int f2fs_build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount) |
2411cf5b | 2303 | { |
e2374015 CY |
2304 | int ret; |
2305 | ||
2411cf5b | 2306 | mutex_lock(&NM_I(sbi)->build_lock); |
e2374015 | 2307 | ret = __f2fs_build_free_nids(sbi, sync, mount); |
2411cf5b | 2308 | mutex_unlock(&NM_I(sbi)->build_lock); |
e2374015 CY |
2309 | |
2310 | return ret; | |
2411cf5b CY |
2311 | } |
2312 | ||
e05df3b1 JK |
2313 | /* |
2314 | * If this function returns success, caller can obtain a new nid | |
2315 | * from second parameter of this function. | |
2316 | * The returned nid could be used ino as well as nid when inode is created. | |
2317 | */ | |
4d57b86d | 2318 | bool f2fs_alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) |
e05df3b1 JK |
2319 | { |
2320 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2321 | struct free_nid *i = NULL; | |
e05df3b1 | 2322 | retry: |
55523519 CY |
2323 | if (time_to_inject(sbi, FAULT_ALLOC_NID)) { |
2324 | f2fs_show_injection_info(FAULT_ALLOC_NID); | |
cb78942b | 2325 | return false; |
55523519 | 2326 | } |
7fa750a1 | 2327 | |
b8559dc2 | 2328 | spin_lock(&nm_i->nid_list_lock); |
e05df3b1 | 2329 | |
04d47e67 CY |
2330 | if (unlikely(nm_i->available_nids == 0)) { |
2331 | spin_unlock(&nm_i->nid_list_lock); | |
2332 | return false; | |
2333 | } | |
e05df3b1 | 2334 | |
4d57b86d CY |
2335 | /* We should not use stale free nids created by f2fs_build_free_nids */ |
2336 | if (nm_i->nid_cnt[FREE_NID] && !on_f2fs_build_free_nids(nm_i)) { | |
9a4ffdf5 CY |
2337 | f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list)); |
2338 | i = list_first_entry(&nm_i->free_nid_list, | |
b8559dc2 | 2339 | struct free_nid, list); |
55008d84 | 2340 | *nid = i->nid; |
b8559dc2 | 2341 | |
a0761f63 | 2342 | __move_free_nid(sbi, i, FREE_NID, PREALLOC_NID); |
04d47e67 | 2343 | nm_i->available_nids--; |
4ac91242 | 2344 | |
346fe752 | 2345 | update_free_nid_bitmap(sbi, *nid, false, false); |
4ac91242 | 2346 | |
b8559dc2 | 2347 | spin_unlock(&nm_i->nid_list_lock); |
55008d84 JK |
2348 | return true; |
2349 | } | |
b8559dc2 | 2350 | spin_unlock(&nm_i->nid_list_lock); |
55008d84 JK |
2351 | |
2352 | /* Let's scan nat pages and its caches to get free nids */ | |
4d57b86d | 2353 | f2fs_build_free_nids(sbi, true, false); |
55008d84 | 2354 | goto retry; |
e05df3b1 JK |
2355 | } |
2356 | ||
0a8165d7 | 2357 | /* |
4d57b86d | 2358 | * f2fs_alloc_nid() should be called prior to this function. |
e05df3b1 | 2359 | */ |
4d57b86d | 2360 | void f2fs_alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
2361 | { |
2362 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2363 | struct free_nid *i; | |
2364 | ||
b8559dc2 | 2365 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 2366 | i = __lookup_free_nid_list(nm_i, nid); |
b8559dc2 | 2367 | f2fs_bug_on(sbi, !i); |
a0761f63 | 2368 | __remove_free_nid(sbi, i, PREALLOC_NID); |
b8559dc2 | 2369 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
2370 | |
2371 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
2372 | } |
2373 | ||
0a8165d7 | 2374 | /* |
4d57b86d | 2375 | * f2fs_alloc_nid() should be called prior to this function. |
e05df3b1 | 2376 | */ |
4d57b86d | 2377 | void f2fs_alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 | 2378 | { |
49952fa1 JK |
2379 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
2380 | struct free_nid *i; | |
cf0ee0f0 | 2381 | bool need_free = false; |
49952fa1 | 2382 | |
65985d93 JK |
2383 | if (!nid) |
2384 | return; | |
2385 | ||
b8559dc2 | 2386 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 2387 | i = __lookup_free_nid_list(nm_i, nid); |
b8559dc2 CY |
2388 | f2fs_bug_on(sbi, !i); |
2389 | ||
4d57b86d | 2390 | if (!f2fs_available_free_memory(sbi, FREE_NIDS)) { |
a0761f63 | 2391 | __remove_free_nid(sbi, i, PREALLOC_NID); |
cf0ee0f0 | 2392 | need_free = true; |
95630cba | 2393 | } else { |
a0761f63 | 2394 | __move_free_nid(sbi, i, PREALLOC_NID, FREE_NID); |
95630cba | 2395 | } |
04d47e67 CY |
2396 | |
2397 | nm_i->available_nids++; | |
2398 | ||
346fe752 | 2399 | update_free_nid_bitmap(sbi, nid, true, false); |
4ac91242 | 2400 | |
b8559dc2 | 2401 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
2402 | |
2403 | if (need_free) | |
2404 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
2405 | } |
2406 | ||
4d57b86d | 2407 | int f2fs_try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink) |
31696580 CY |
2408 | { |
2409 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2410 | struct free_nid *i, *next; | |
2411 | int nr = nr_shrink; | |
2412 | ||
9a4ffdf5 | 2413 | if (nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS) |
ad4edb83 JK |
2414 | return 0; |
2415 | ||
31696580 CY |
2416 | if (!mutex_trylock(&nm_i->build_lock)) |
2417 | return 0; | |
2418 | ||
b8559dc2 | 2419 | spin_lock(&nm_i->nid_list_lock); |
9a4ffdf5 | 2420 | list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) { |
b8559dc2 | 2421 | if (nr_shrink <= 0 || |
9a4ffdf5 | 2422 | nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS) |
31696580 | 2423 | break; |
b8559dc2 | 2424 | |
a0761f63 | 2425 | __remove_free_nid(sbi, i, FREE_NID); |
31696580 CY |
2426 | kmem_cache_free(free_nid_slab, i); |
2427 | nr_shrink--; | |
31696580 | 2428 | } |
b8559dc2 | 2429 | spin_unlock(&nm_i->nid_list_lock); |
31696580 CY |
2430 | mutex_unlock(&nm_i->build_lock); |
2431 | ||
2432 | return nr - nr_shrink; | |
2433 | } | |
2434 | ||
4d57b86d | 2435 | void f2fs_recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 2436 | { |
28cdce04 CY |
2437 | void *src_addr, *dst_addr; |
2438 | size_t inline_size; | |
2439 | struct page *ipage; | |
2440 | struct f2fs_inode *ri; | |
2441 | ||
4d57b86d | 2442 | ipage = f2fs_get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 2443 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 2444 | |
e3b4d43f | 2445 | ri = F2FS_INODE(page); |
1eca05aa YH |
2446 | if (ri->i_inline & F2FS_INLINE_XATTR) { |
2447 | set_inode_flag(inode, FI_INLINE_XATTR); | |
2448 | } else { | |
91942321 | 2449 | clear_inode_flag(inode, FI_INLINE_XATTR); |
e3b4d43f JK |
2450 | goto update_inode; |
2451 | } | |
2452 | ||
6afc662e CY |
2453 | dst_addr = inline_xattr_addr(inode, ipage); |
2454 | src_addr = inline_xattr_addr(inode, page); | |
28cdce04 CY |
2455 | inline_size = inline_xattr_size(inode); |
2456 | ||
fec1d657 | 2457 | f2fs_wait_on_page_writeback(ipage, NODE, true); |
28cdce04 | 2458 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 2459 | update_inode: |
4d57b86d | 2460 | f2fs_update_inode(inode, ipage); |
28cdce04 CY |
2461 | f2fs_put_page(ipage, 1); |
2462 | } | |
2463 | ||
4d57b86d | 2464 | int f2fs_recover_xattr_data(struct inode *inode, struct page *page) |
abb2366c | 2465 | { |
4081363f | 2466 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c | 2467 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
87905682 YH |
2468 | nid_t new_xnid; |
2469 | struct dnode_of_data dn; | |
abb2366c | 2470 | struct node_info ni; |
d260081c | 2471 | struct page *xpage; |
7735730d | 2472 | int err; |
abb2366c | 2473 | |
abb2366c JK |
2474 | if (!prev_xnid) |
2475 | goto recover_xnid; | |
2476 | ||
d260081c | 2477 | /* 1: invalidate the previous xattr nid */ |
7735730d CY |
2478 | err = f2fs_get_node_info(sbi, prev_xnid, &ni); |
2479 | if (err) | |
2480 | return err; | |
2481 | ||
4d57b86d | 2482 | f2fs_invalidate_blocks(sbi, ni.blk_addr); |
000519f2 | 2483 | dec_valid_node_count(sbi, inode, false); |
479f40c4 | 2484 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
2485 | |
2486 | recover_xnid: | |
d260081c | 2487 | /* 2: update xattr nid in inode */ |
4d57b86d | 2488 | if (!f2fs_alloc_nid(sbi, &new_xnid)) |
87905682 YH |
2489 | return -ENOSPC; |
2490 | ||
2491 | set_new_dnode(&dn, inode, NULL, NULL, new_xnid); | |
4d57b86d | 2492 | xpage = f2fs_new_node_page(&dn, XATTR_NODE_OFFSET); |
87905682 | 2493 | if (IS_ERR(xpage)) { |
4d57b86d | 2494 | f2fs_alloc_nid_failed(sbi, new_xnid); |
87905682 YH |
2495 | return PTR_ERR(xpage); |
2496 | } | |
2497 | ||
4d57b86d CY |
2498 | f2fs_alloc_nid_done(sbi, new_xnid); |
2499 | f2fs_update_inode_page(inode); | |
d260081c CY |
2500 | |
2501 | /* 3: update and set xattr node page dirty */ | |
87905682 | 2502 | memcpy(F2FS_NODE(xpage), F2FS_NODE(page), VALID_XATTR_BLOCK_SIZE); |
abb2366c | 2503 | |
d260081c CY |
2504 | set_page_dirty(xpage); |
2505 | f2fs_put_page(xpage, 1); | |
abb2366c | 2506 | |
d260081c | 2507 | return 0; |
abb2366c JK |
2508 | } |
2509 | ||
4d57b86d | 2510 | int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
e05df3b1 | 2511 | { |
58bfaf44 | 2512 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
2513 | nid_t ino = ino_of_node(page); |
2514 | struct node_info old_ni, new_ni; | |
2515 | struct page *ipage; | |
7735730d | 2516 | int err; |
e05df3b1 | 2517 | |
7735730d CY |
2518 | err = f2fs_get_node_info(sbi, ino, &old_ni); |
2519 | if (err) | |
2520 | return err; | |
e8271fa3 JK |
2521 | |
2522 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
2523 | return -EINVAL; | |
e8ea9b3d | 2524 | retry: |
300e129c | 2525 | ipage = f2fs_grab_cache_page(NODE_MAPPING(sbi), ino, false); |
e8ea9b3d JK |
2526 | if (!ipage) { |
2527 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
2528 | goto retry; | |
2529 | } | |
e05df3b1 | 2530 | |
e1c42045 | 2531 | /* Should not use this inode from free nid list */ |
b8559dc2 | 2532 | remove_free_nid(sbi, ino); |
e05df3b1 | 2533 | |
237c0790 JK |
2534 | if (!PageUptodate(ipage)) |
2535 | SetPageUptodate(ipage); | |
e05df3b1 | 2536 | fill_node_footer(ipage, ino, ino, 0, true); |
c5667575 | 2537 | set_cold_node(page, false); |
e05df3b1 | 2538 | |
58bfaf44 JK |
2539 | src = F2FS_INODE(page); |
2540 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 2541 | |
58bfaf44 JK |
2542 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
2543 | dst->i_size = 0; | |
2544 | dst->i_blocks = cpu_to_le64(1); | |
2545 | dst->i_links = cpu_to_le32(1); | |
2546 | dst->i_xattr_nid = 0; | |
7a2af766 | 2547 | dst->i_inline = src->i_inline & (F2FS_INLINE_XATTR | F2FS_EXTRA_ATTR); |
5c57132e | 2548 | if (dst->i_inline & F2FS_EXTRA_ATTR) { |
7a2af766 | 2549 | dst->i_extra_isize = src->i_extra_isize; |
6afc662e CY |
2550 | |
2551 | if (f2fs_sb_has_flexible_inline_xattr(sbi->sb) && | |
2552 | F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize), | |
2553 | i_inline_xattr_size)) | |
2554 | dst->i_inline_xattr_size = src->i_inline_xattr_size; | |
2555 | ||
5c57132e CY |
2556 | if (f2fs_sb_has_project_quota(sbi->sb) && |
2557 | F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize), | |
2558 | i_projid)) | |
2559 | dst->i_projid = src->i_projid; | |
2560 | } | |
e05df3b1 JK |
2561 | |
2562 | new_ni = old_ni; | |
2563 | new_ni.ino = ino; | |
2564 | ||
0abd675e | 2565 | if (unlikely(inc_valid_node_count(sbi, NULL, true))) |
65e5cd0a | 2566 | WARN_ON(1); |
479f40c4 | 2567 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 2568 | inc_valid_inode_count(sbi); |
617deb8c | 2569 | set_page_dirty(ipage); |
e05df3b1 JK |
2570 | f2fs_put_page(ipage, 1); |
2571 | return 0; | |
2572 | } | |
2573 | ||
7735730d | 2574 | int f2fs_restore_node_summary(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
2575 | unsigned int segno, struct f2fs_summary_block *sum) |
2576 | { | |
2577 | struct f2fs_node *rn; | |
2578 | struct f2fs_summary *sum_entry; | |
e05df3b1 | 2579 | block_t addr; |
9ecf4b80 | 2580 | int i, idx, last_offset, nrpages; |
e05df3b1 JK |
2581 | |
2582 | /* scan the node segment */ | |
2583 | last_offset = sbi->blocks_per_seg; | |
2584 | addr = START_BLOCK(sbi, segno); | |
2585 | sum_entry = &sum->entries[0]; | |
2586 | ||
9ecf4b80 | 2587 | for (i = 0; i < last_offset; i += nrpages, addr += nrpages) { |
664ba972 | 2588 | nrpages = min(last_offset - i, BIO_MAX_PAGES); |
393ff91f | 2589 | |
e1c42045 | 2590 | /* readahead node pages */ |
4d57b86d | 2591 | f2fs_ra_meta_pages(sbi, addr, nrpages, META_POR, true); |
e05df3b1 | 2592 | |
9ecf4b80 | 2593 | for (idx = addr; idx < addr + nrpages; idx++) { |
4d57b86d | 2594 | struct page *page = f2fs_get_tmp_page(sbi, idx); |
9af0ff1c | 2595 | |
7735730d CY |
2596 | if (IS_ERR(page)) |
2597 | return PTR_ERR(page); | |
2598 | ||
9ecf4b80 CY |
2599 | rn = F2FS_NODE(page); |
2600 | sum_entry->nid = rn->footer.nid; | |
2601 | sum_entry->version = 0; | |
2602 | sum_entry->ofs_in_node = 0; | |
2603 | sum_entry++; | |
2604 | f2fs_put_page(page, 1); | |
9af0ff1c | 2605 | } |
bac4eef6 | 2606 | |
9ecf4b80 | 2607 | invalidate_mapping_pages(META_MAPPING(sbi), addr, |
bac4eef6 | 2608 | addr + nrpages); |
e05df3b1 | 2609 | } |
7735730d | 2610 | return 0; |
e05df3b1 JK |
2611 | } |
2612 | ||
aec71382 | 2613 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
2614 | { |
2615 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2616 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2617 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
2618 | int i; |
2619 | ||
b7ad7512 | 2620 | down_write(&curseg->journal_rwsem); |
dfc08a12 | 2621 | for (i = 0; i < nats_in_cursum(journal); i++) { |
e05df3b1 JK |
2622 | struct nat_entry *ne; |
2623 | struct f2fs_nat_entry raw_ne; | |
dfc08a12 | 2624 | nid_t nid = le32_to_cpu(nid_in_journal(journal, i)); |
e05df3b1 | 2625 | |
dfc08a12 | 2626 | raw_ne = nat_in_journal(journal, i); |
9be32d72 | 2627 | |
e05df3b1 | 2628 | ne = __lookup_nat_cache(nm_i, nid); |
e05df3b1 | 2629 | if (!ne) { |
12f9ef37 YH |
2630 | ne = __alloc_nat_entry(nid, true); |
2631 | __init_nat_entry(nm_i, ne, &raw_ne, true); | |
e05df3b1 | 2632 | } |
04d47e67 CY |
2633 | |
2634 | /* | |
2635 | * if a free nat in journal has not been used after last | |
2636 | * checkpoint, we should remove it from available nids, | |
2637 | * since later we will add it again. | |
2638 | */ | |
2639 | if (!get_nat_flag(ne, IS_DIRTY) && | |
2640 | le32_to_cpu(raw_ne.block_addr) == NULL_ADDR) { | |
2641 | spin_lock(&nm_i->nid_list_lock); | |
2642 | nm_i->available_nids--; | |
2643 | spin_unlock(&nm_i->nid_list_lock); | |
2644 | } | |
2645 | ||
e05df3b1 | 2646 | __set_nat_cache_dirty(nm_i, ne); |
e05df3b1 | 2647 | } |
dfc08a12 | 2648 | update_nats_in_cursum(journal, -i); |
b7ad7512 | 2649 | up_write(&curseg->journal_rwsem); |
e05df3b1 JK |
2650 | } |
2651 | ||
309cc2b6 JK |
2652 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
2653 | struct list_head *head, int max) | |
e05df3b1 | 2654 | { |
309cc2b6 | 2655 | struct nat_entry_set *cur; |
e05df3b1 | 2656 | |
309cc2b6 JK |
2657 | if (nes->entry_cnt >= max) |
2658 | goto add_out; | |
e05df3b1 | 2659 | |
309cc2b6 JK |
2660 | list_for_each_entry(cur, head, set_list) { |
2661 | if (cur->entry_cnt >= nes->entry_cnt) { | |
2662 | list_add(&nes->set_list, cur->set_list.prev); | |
2663 | return; | |
2664 | } | |
aec71382 | 2665 | } |
309cc2b6 JK |
2666 | add_out: |
2667 | list_add_tail(&nes->set_list, head); | |
2668 | } | |
e05df3b1 | 2669 | |
9f7e4a2c | 2670 | static void __update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid, |
22ad0b6a JK |
2671 | struct page *page) |
2672 | { | |
2673 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2674 | unsigned int nat_index = start_nid / NAT_ENTRY_PER_BLOCK; | |
2675 | struct f2fs_nat_block *nat_blk = page_address(page); | |
2676 | int valid = 0; | |
37a0ab2a | 2677 | int i = 0; |
22ad0b6a JK |
2678 | |
2679 | if (!enabled_nat_bits(sbi, NULL)) | |
2680 | return; | |
2681 | ||
37a0ab2a FL |
2682 | if (nat_index == 0) { |
2683 | valid = 1; | |
2684 | i = 1; | |
2685 | } | |
2686 | for (; i < NAT_ENTRY_PER_BLOCK; i++) { | |
2687 | if (nat_blk->entries[i].block_addr != NULL_ADDR) | |
22ad0b6a JK |
2688 | valid++; |
2689 | } | |
2690 | if (valid == 0) { | |
23380b85 JK |
2691 | __set_bit_le(nat_index, nm_i->empty_nat_bits); |
2692 | __clear_bit_le(nat_index, nm_i->full_nat_bits); | |
22ad0b6a JK |
2693 | return; |
2694 | } | |
2695 | ||
23380b85 | 2696 | __clear_bit_le(nat_index, nm_i->empty_nat_bits); |
22ad0b6a | 2697 | if (valid == NAT_ENTRY_PER_BLOCK) |
23380b85 | 2698 | __set_bit_le(nat_index, nm_i->full_nat_bits); |
22ad0b6a | 2699 | else |
23380b85 | 2700 | __clear_bit_le(nat_index, nm_i->full_nat_bits); |
22ad0b6a JK |
2701 | } |
2702 | ||
309cc2b6 | 2703 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
22ad0b6a | 2704 | struct nat_entry_set *set, struct cp_control *cpc) |
309cc2b6 JK |
2705 | { |
2706 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2707 | struct f2fs_journal *journal = curseg->journal; |
309cc2b6 JK |
2708 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; |
2709 | bool to_journal = true; | |
2710 | struct f2fs_nat_block *nat_blk; | |
2711 | struct nat_entry *ne, *cur; | |
2712 | struct page *page = NULL; | |
e05df3b1 | 2713 | |
aec71382 CY |
2714 | /* |
2715 | * there are two steps to flush nat entries: | |
2716 | * #1, flush nat entries to journal in current hot data summary block. | |
2717 | * #2, flush nat entries to nat page. | |
2718 | */ | |
22ad0b6a JK |
2719 | if (enabled_nat_bits(sbi, cpc) || |
2720 | !__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL)) | |
309cc2b6 JK |
2721 | to_journal = false; |
2722 | ||
2723 | if (to_journal) { | |
b7ad7512 | 2724 | down_write(&curseg->journal_rwsem); |
309cc2b6 JK |
2725 | } else { |
2726 | page = get_next_nat_page(sbi, start_nid); | |
2727 | nat_blk = page_address(page); | |
2728 | f2fs_bug_on(sbi, !nat_blk); | |
2729 | } | |
aec71382 | 2730 | |
309cc2b6 JK |
2731 | /* flush dirty nats in nat entry set */ |
2732 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
2733 | struct f2fs_nat_entry *raw_ne; | |
2734 | nid_t nid = nat_get_nid(ne); | |
2735 | int offset; | |
2736 | ||
febeca6d | 2737 | f2fs_bug_on(sbi, nat_get_blkaddr(ne) == NEW_ADDR); |
aec71382 CY |
2738 | |
2739 | if (to_journal) { | |
4d57b86d | 2740 | offset = f2fs_lookup_journal_in_cursum(journal, |
309cc2b6 JK |
2741 | NAT_JOURNAL, nid, 1); |
2742 | f2fs_bug_on(sbi, offset < 0); | |
dfc08a12 CY |
2743 | raw_ne = &nat_in_journal(journal, offset); |
2744 | nid_in_journal(journal, offset) = cpu_to_le32(nid); | |
aec71382 | 2745 | } else { |
309cc2b6 | 2746 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 2747 | } |
309cc2b6 | 2748 | raw_nat_from_node_info(raw_ne, &ne->ni); |
309cc2b6 | 2749 | nat_reset_flag(ne); |
0b28b71e | 2750 | __clear_nat_cache_dirty(NM_I(sbi), set, ne); |
04d47e67 | 2751 | if (nat_get_blkaddr(ne) == NULL_ADDR) { |
5921aaa1 | 2752 | add_free_nid(sbi, nid, false, true); |
4ac91242 CY |
2753 | } else { |
2754 | spin_lock(&NM_I(sbi)->nid_list_lock); | |
346fe752 | 2755 | update_free_nid_bitmap(sbi, nid, false, false); |
04d47e67 CY |
2756 | spin_unlock(&NM_I(sbi)->nid_list_lock); |
2757 | } | |
309cc2b6 | 2758 | } |
e05df3b1 | 2759 | |
22ad0b6a | 2760 | if (to_journal) { |
b7ad7512 | 2761 | up_write(&curseg->journal_rwsem); |
22ad0b6a JK |
2762 | } else { |
2763 | __update_nat_bits(sbi, start_nid, page); | |
309cc2b6 | 2764 | f2fs_put_page(page, 1); |
22ad0b6a | 2765 | } |
aec71382 | 2766 | |
59c9081b YH |
2767 | /* Allow dirty nats by node block allocation in write_begin */ |
2768 | if (!set->entry_cnt) { | |
2769 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); | |
2770 | kmem_cache_free(nat_entry_set_slab, set); | |
2771 | } | |
309cc2b6 | 2772 | } |
aec71382 | 2773 | |
309cc2b6 JK |
2774 | /* |
2775 | * This function is called during the checkpointing process. | |
2776 | */ | |
4d57b86d | 2777 | void f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
309cc2b6 JK |
2778 | { |
2779 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2780 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2781 | struct f2fs_journal *journal = curseg->journal; |
7aed0d45 | 2782 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
309cc2b6 JK |
2783 | struct nat_entry_set *set, *tmp; |
2784 | unsigned int found; | |
2785 | nid_t set_idx = 0; | |
2786 | LIST_HEAD(sets); | |
2787 | ||
7f2ecdd8 JK |
2788 | /* during unmount, let's flush nat_bits before checking dirty_nat_cnt */ |
2789 | if (enabled_nat_bits(sbi, cpc)) { | |
2790 | down_write(&nm_i->nat_tree_lock); | |
2791 | remove_nats_in_journal(sbi); | |
2792 | up_write(&nm_i->nat_tree_lock); | |
2793 | } | |
2794 | ||
20d047c8 CL |
2795 | if (!nm_i->dirty_nat_cnt) |
2796 | return; | |
a5131193 | 2797 | |
b873b798 | 2798 | down_write(&nm_i->nat_tree_lock); |
a5131193 | 2799 | |
309cc2b6 JK |
2800 | /* |
2801 | * if there are no enough space in journal to store dirty nat | |
2802 | * entries, remove all entries from journal and merge them | |
2803 | * into nat entry set. | |
2804 | */ | |
900f7362 | 2805 | if (enabled_nat_bits(sbi, cpc) || |
22ad0b6a | 2806 | !__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL)) |
309cc2b6 JK |
2807 | remove_nats_in_journal(sbi); |
2808 | ||
309cc2b6 | 2809 | while ((found = __gang_lookup_nat_set(nm_i, |
7aed0d45 | 2810 | set_idx, SETVEC_SIZE, setvec))) { |
309cc2b6 JK |
2811 | unsigned idx; |
2812 | set_idx = setvec[found - 1]->set + 1; | |
2813 | for (idx = 0; idx < found; idx++) | |
2814 | __adjust_nat_entry_set(setvec[idx], &sets, | |
dfc08a12 | 2815 | MAX_NAT_JENTRIES(journal)); |
e05df3b1 | 2816 | } |
aec71382 | 2817 | |
309cc2b6 JK |
2818 | /* flush dirty nats in nat entry set */ |
2819 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
22ad0b6a | 2820 | __flush_nat_entry_set(sbi, set, cpc); |
309cc2b6 | 2821 | |
b873b798 | 2822 | up_write(&nm_i->nat_tree_lock); |
59c9081b | 2823 | /* Allow dirty nats by node block allocation in write_begin */ |
e05df3b1 JK |
2824 | } |
2825 | ||
22ad0b6a JK |
2826 | static int __get_nat_bitmaps(struct f2fs_sb_info *sbi) |
2827 | { | |
2828 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
2829 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2830 | unsigned int nat_bits_bytes = nm_i->nat_blocks / BITS_PER_BYTE; | |
2831 | unsigned int i; | |
2832 | __u64 cp_ver = cur_cp_version(ckpt); | |
22ad0b6a JK |
2833 | block_t nat_bits_addr; |
2834 | ||
2835 | if (!enabled_nat_bits(sbi, NULL)) | |
2836 | return 0; | |
2837 | ||
df033caf | 2838 | nm_i->nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); |
acbf054d CY |
2839 | nm_i->nat_bits = f2fs_kzalloc(sbi, |
2840 | nm_i->nat_bits_blocks << F2FS_BLKSIZE_BITS, GFP_KERNEL); | |
22ad0b6a JK |
2841 | if (!nm_i->nat_bits) |
2842 | return -ENOMEM; | |
2843 | ||
2844 | nat_bits_addr = __start_cp_addr(sbi) + sbi->blocks_per_seg - | |
2845 | nm_i->nat_bits_blocks; | |
2846 | for (i = 0; i < nm_i->nat_bits_blocks; i++) { | |
7735730d CY |
2847 | struct page *page; |
2848 | ||
2849 | page = f2fs_get_meta_page(sbi, nat_bits_addr++); | |
2850 | if (IS_ERR(page)) { | |
2851 | disable_nat_bits(sbi, true); | |
2852 | return PTR_ERR(page); | |
2853 | } | |
22ad0b6a JK |
2854 | |
2855 | memcpy(nm_i->nat_bits + (i << F2FS_BLKSIZE_BITS), | |
2856 | page_address(page), F2FS_BLKSIZE); | |
2857 | f2fs_put_page(page, 1); | |
2858 | } | |
2859 | ||
ced2c7ea | 2860 | cp_ver |= (cur_cp_crc(ckpt) << 32); |
22ad0b6a JK |
2861 | if (cpu_to_le64(cp_ver) != *(__le64 *)nm_i->nat_bits) { |
2862 | disable_nat_bits(sbi, true); | |
2863 | return 0; | |
2864 | } | |
2865 | ||
2866 | nm_i->full_nat_bits = nm_i->nat_bits + 8; | |
2867 | nm_i->empty_nat_bits = nm_i->full_nat_bits + nat_bits_bytes; | |
2868 | ||
2869 | f2fs_msg(sbi->sb, KERN_NOTICE, "Found nat_bits in checkpoint"); | |
2870 | return 0; | |
2871 | } | |
2872 | ||
bd80a4b9 | 2873 | static inline void load_free_nid_bitmap(struct f2fs_sb_info *sbi) |
7041d5d2 CY |
2874 | { |
2875 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2876 | unsigned int i = 0; | |
2877 | nid_t nid, last_nid; | |
2878 | ||
2879 | if (!enabled_nat_bits(sbi, NULL)) | |
2880 | return; | |
2881 | ||
2882 | for (i = 0; i < nm_i->nat_blocks; i++) { | |
2883 | i = find_next_bit_le(nm_i->empty_nat_bits, nm_i->nat_blocks, i); | |
2884 | if (i >= nm_i->nat_blocks) | |
2885 | break; | |
2886 | ||
2887 | __set_bit_le(i, nm_i->nat_block_bitmap); | |
2888 | ||
2889 | nid = i * NAT_ENTRY_PER_BLOCK; | |
f6986ede | 2890 | last_nid = nid + NAT_ENTRY_PER_BLOCK; |
7041d5d2 | 2891 | |
346fe752 | 2892 | spin_lock(&NM_I(sbi)->nid_list_lock); |
7041d5d2 | 2893 | for (; nid < last_nid; nid++) |
346fe752 CY |
2894 | update_free_nid_bitmap(sbi, nid, true, true); |
2895 | spin_unlock(&NM_I(sbi)->nid_list_lock); | |
7041d5d2 CY |
2896 | } |
2897 | ||
2898 | for (i = 0; i < nm_i->nat_blocks; i++) { | |
2899 | i = find_next_bit_le(nm_i->full_nat_bits, nm_i->nat_blocks, i); | |
2900 | if (i >= nm_i->nat_blocks) | |
2901 | break; | |
2902 | ||
2903 | __set_bit_le(i, nm_i->nat_block_bitmap); | |
2904 | } | |
2905 | } | |
2906 | ||
e05df3b1 JK |
2907 | static int init_node_manager(struct f2fs_sb_info *sbi) |
2908 | { | |
2909 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
2910 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2911 | unsigned char *version_bitmap; | |
22ad0b6a JK |
2912 | unsigned int nat_segs; |
2913 | int err; | |
e05df3b1 JK |
2914 | |
2915 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
2916 | ||
2917 | /* segment_count_nat includes pair segment so divide to 2. */ | |
2918 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
22ad0b6a JK |
2919 | nm_i->nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); |
2920 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks; | |
7ee0eeab | 2921 | |
b63da15e | 2922 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
04d47e67 | 2923 | nm_i->available_nids = nm_i->max_nid - sbi->total_valid_node_count - |
292c196a | 2924 | sbi->nquota_files - F2FS_RESERVED_NODE_NUM; |
9a4ffdf5 CY |
2925 | nm_i->nid_cnt[FREE_NID] = 0; |
2926 | nm_i->nid_cnt[PREALLOC_NID] = 0; | |
e05df3b1 | 2927 | nm_i->nat_cnt = 0; |
cdfc41c1 | 2928 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
ea1a29a0 | 2929 | nm_i->ra_nid_pages = DEF_RA_NID_PAGES; |
2304cb0c | 2930 | nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD; |
e05df3b1 | 2931 | |
8a7ed66a | 2932 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
9a4ffdf5 | 2933 | INIT_LIST_HEAD(&nm_i->free_nid_list); |
769ec6e5 JK |
2934 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO); |
2935 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO); | |
e05df3b1 | 2936 | INIT_LIST_HEAD(&nm_i->nat_entries); |
22969158 | 2937 | spin_lock_init(&nm_i->nat_list_lock); |
e05df3b1 JK |
2938 | |
2939 | mutex_init(&nm_i->build_lock); | |
b8559dc2 | 2940 | spin_lock_init(&nm_i->nid_list_lock); |
b873b798 | 2941 | init_rwsem(&nm_i->nat_tree_lock); |
e05df3b1 | 2942 | |
e05df3b1 | 2943 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 2944 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
2945 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
2946 | if (!version_bitmap) | |
2947 | return -EFAULT; | |
2948 | ||
79b5793b AG |
2949 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
2950 | GFP_KERNEL); | |
2951 | if (!nm_i->nat_bitmap) | |
2952 | return -ENOMEM; | |
599a09b2 | 2953 | |
22ad0b6a JK |
2954 | err = __get_nat_bitmaps(sbi); |
2955 | if (err) | |
2956 | return err; | |
2957 | ||
599a09b2 CY |
2958 | #ifdef CONFIG_F2FS_CHECK_FS |
2959 | nm_i->nat_bitmap_mir = kmemdup(version_bitmap, nm_i->bitmap_size, | |
2960 | GFP_KERNEL); | |
2961 | if (!nm_i->nat_bitmap_mir) | |
2962 | return -ENOMEM; | |
2963 | #endif | |
2964 | ||
e05df3b1 JK |
2965 | return 0; |
2966 | } | |
2967 | ||
9f7e4a2c | 2968 | static int init_free_nid_cache(struct f2fs_sb_info *sbi) |
4ac91242 CY |
2969 | { |
2970 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
bb1105e4 | 2971 | int i; |
4ac91242 | 2972 | |
026f0507 KC |
2973 | nm_i->free_nid_bitmap = |
2974 | f2fs_kzalloc(sbi, array_size(sizeof(unsigned char *), | |
2975 | nm_i->nat_blocks), | |
2976 | GFP_KERNEL); | |
4ac91242 CY |
2977 | if (!nm_i->free_nid_bitmap) |
2978 | return -ENOMEM; | |
2979 | ||
bb1105e4 JK |
2980 | for (i = 0; i < nm_i->nat_blocks; i++) { |
2981 | nm_i->free_nid_bitmap[i] = f2fs_kvzalloc(sbi, | |
e15d54d5 | 2982 | f2fs_bitmap_size(NAT_ENTRY_PER_BLOCK), GFP_KERNEL); |
68c43a23 | 2983 | if (!nm_i->free_nid_bitmap[i]) |
bb1105e4 JK |
2984 | return -ENOMEM; |
2985 | } | |
2986 | ||
628b3d14 | 2987 | nm_i->nat_block_bitmap = f2fs_kvzalloc(sbi, nm_i->nat_blocks / 8, |
4ac91242 CY |
2988 | GFP_KERNEL); |
2989 | if (!nm_i->nat_block_bitmap) | |
2990 | return -ENOMEM; | |
586d1492 | 2991 | |
9d2a789c KC |
2992 | nm_i->free_nid_count = |
2993 | f2fs_kvzalloc(sbi, array_size(sizeof(unsigned short), | |
2994 | nm_i->nat_blocks), | |
2995 | GFP_KERNEL); | |
586d1492 CY |
2996 | if (!nm_i->free_nid_count) |
2997 | return -ENOMEM; | |
4ac91242 CY |
2998 | return 0; |
2999 | } | |
3000 | ||
4d57b86d | 3001 | int f2fs_build_node_manager(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
3002 | { |
3003 | int err; | |
3004 | ||
acbf054d CY |
3005 | sbi->nm_info = f2fs_kzalloc(sbi, sizeof(struct f2fs_nm_info), |
3006 | GFP_KERNEL); | |
e05df3b1 JK |
3007 | if (!sbi->nm_info) |
3008 | return -ENOMEM; | |
3009 | ||
3010 | err = init_node_manager(sbi); | |
3011 | if (err) | |
3012 | return err; | |
3013 | ||
4ac91242 CY |
3014 | err = init_free_nid_cache(sbi); |
3015 | if (err) | |
3016 | return err; | |
3017 | ||
7041d5d2 CY |
3018 | /* load free nid status from nat_bits table */ |
3019 | load_free_nid_bitmap(sbi); | |
3020 | ||
e2374015 | 3021 | return f2fs_build_free_nids(sbi, true, true); |
e05df3b1 JK |
3022 | } |
3023 | ||
4d57b86d | 3024 | void f2fs_destroy_node_manager(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
3025 | { |
3026 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
3027 | struct free_nid *i, *next_i; | |
3028 | struct nat_entry *natvec[NATVEC_SIZE]; | |
7aed0d45 | 3029 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
e05df3b1 JK |
3030 | nid_t nid = 0; |
3031 | unsigned int found; | |
3032 | ||
3033 | if (!nm_i) | |
3034 | return; | |
3035 | ||
3036 | /* destroy free nid list */ | |
b8559dc2 | 3037 | spin_lock(&nm_i->nid_list_lock); |
9a4ffdf5 | 3038 | list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { |
a0761f63 | 3039 | __remove_free_nid(sbi, i, FREE_NID); |
b8559dc2 | 3040 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 | 3041 | kmem_cache_free(free_nid_slab, i); |
b8559dc2 | 3042 | spin_lock(&nm_i->nid_list_lock); |
e05df3b1 | 3043 | } |
9a4ffdf5 CY |
3044 | f2fs_bug_on(sbi, nm_i->nid_cnt[FREE_NID]); |
3045 | f2fs_bug_on(sbi, nm_i->nid_cnt[PREALLOC_NID]); | |
3046 | f2fs_bug_on(sbi, !list_empty(&nm_i->free_nid_list)); | |
b8559dc2 | 3047 | spin_unlock(&nm_i->nid_list_lock); |
e05df3b1 JK |
3048 | |
3049 | /* destroy nat cache */ | |
b873b798 | 3050 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
3051 | while ((found = __gang_lookup_nat_cache(nm_i, |
3052 | nid, NATVEC_SIZE, natvec))) { | |
3053 | unsigned idx; | |
7aed0d45 | 3054 | |
b6ce391e | 3055 | nid = nat_get_nid(natvec[found - 1]) + 1; |
22969158 CY |
3056 | for (idx = 0; idx < found; idx++) { |
3057 | spin_lock(&nm_i->nat_list_lock); | |
3058 | list_del(&natvec[idx]->list); | |
3059 | spin_unlock(&nm_i->nat_list_lock); | |
3060 | ||
b6ce391e | 3061 | __del_from_nat_cache(nm_i, natvec[idx]); |
22969158 | 3062 | } |
e05df3b1 | 3063 | } |
9850cf4a | 3064 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
7aed0d45 JK |
3065 | |
3066 | /* destroy nat set cache */ | |
3067 | nid = 0; | |
3068 | while ((found = __gang_lookup_nat_set(nm_i, | |
3069 | nid, SETVEC_SIZE, setvec))) { | |
3070 | unsigned idx; | |
3071 | ||
3072 | nid = setvec[found - 1]->set + 1; | |
3073 | for (idx = 0; idx < found; idx++) { | |
3074 | /* entry_cnt is not zero, when cp_error was occurred */ | |
3075 | f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list)); | |
3076 | radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set); | |
3077 | kmem_cache_free(nat_entry_set_slab, setvec[idx]); | |
3078 | } | |
3079 | } | |
b873b798 | 3080 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 | 3081 | |
4ac91242 | 3082 | kvfree(nm_i->nat_block_bitmap); |
bb1105e4 JK |
3083 | if (nm_i->free_nid_bitmap) { |
3084 | int i; | |
3085 | ||
3086 | for (i = 0; i < nm_i->nat_blocks; i++) | |
3087 | kvfree(nm_i->free_nid_bitmap[i]); | |
3088 | kfree(nm_i->free_nid_bitmap); | |
3089 | } | |
586d1492 | 3090 | kvfree(nm_i->free_nid_count); |
4ac91242 | 3091 | |
e05df3b1 | 3092 | kfree(nm_i->nat_bitmap); |
22ad0b6a | 3093 | kfree(nm_i->nat_bits); |
599a09b2 CY |
3094 | #ifdef CONFIG_F2FS_CHECK_FS |
3095 | kfree(nm_i->nat_bitmap_mir); | |
3096 | #endif | |
e05df3b1 JK |
3097 | sbi->nm_info = NULL; |
3098 | kfree(nm_i); | |
3099 | } | |
3100 | ||
4d57b86d | 3101 | int __init f2fs_create_node_manager_caches(void) |
e05df3b1 JK |
3102 | { |
3103 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 3104 | sizeof(struct nat_entry)); |
e05df3b1 | 3105 | if (!nat_entry_slab) |
aec71382 | 3106 | goto fail; |
e05df3b1 JK |
3107 | |
3108 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 3109 | sizeof(struct free_nid)); |
aec71382 | 3110 | if (!free_nid_slab) |
ce3e6d25 | 3111 | goto destroy_nat_entry; |
aec71382 CY |
3112 | |
3113 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
3114 | sizeof(struct nat_entry_set)); | |
3115 | if (!nat_entry_set_slab) | |
ce3e6d25 | 3116 | goto destroy_free_nid; |
50fa53ec CY |
3117 | |
3118 | fsync_node_entry_slab = f2fs_kmem_cache_create("fsync_node_entry", | |
3119 | sizeof(struct fsync_node_entry)); | |
3120 | if (!fsync_node_entry_slab) | |
3121 | goto destroy_nat_entry_set; | |
e05df3b1 | 3122 | return 0; |
aec71382 | 3123 | |
50fa53ec CY |
3124 | destroy_nat_entry_set: |
3125 | kmem_cache_destroy(nat_entry_set_slab); | |
ce3e6d25 | 3126 | destroy_free_nid: |
aec71382 | 3127 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 3128 | destroy_nat_entry: |
aec71382 CY |
3129 | kmem_cache_destroy(nat_entry_slab); |
3130 | fail: | |
3131 | return -ENOMEM; | |
e05df3b1 JK |
3132 | } |
3133 | ||
4d57b86d | 3134 | void f2fs_destroy_node_manager_caches(void) |
e05df3b1 | 3135 | { |
50fa53ec | 3136 | kmem_cache_destroy(fsync_node_entry_slab); |
aec71382 | 3137 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
3138 | kmem_cache_destroy(free_nid_slab); |
3139 | kmem_cache_destroy(nat_entry_slab); | |
3140 | } |