1
2
3
4
5
6
7
8
9
10
11#include <linux/fs.h>
12#include <linux/f2fs_fs.h>
13#include <linux/mpage.h>
14#include <linux/backing-dev.h>
15#include <linux/blkdev.h>
16#include <linux/pagevec.h>
17#include <linux/swap.h>
18
19#include "f2fs.h"
20#include "node.h"
21#include "segment.h"
22#include "xattr.h"
23#include "trace.h"
24#include <trace/events/f2fs.h>
25
26#define on_f2fs_build_free_nids(nmi) mutex_is_locked(&(nm_i)->build_lock)
27
28static struct kmem_cache *nat_entry_slab;
29static struct kmem_cache *free_nid_slab;
30static struct kmem_cache *nat_entry_set_slab;
31
32
33
34
35int f2fs_check_nid_range(struct f2fs_sb_info *sbi, nid_t nid)
36{
37 if (unlikely(nid < F2FS_ROOT_INO(sbi) || nid >= NM_I(sbi)->max_nid)) {
38 set_sbi_flag(sbi, SBI_NEED_FSCK);
39 f2fs_msg(sbi->sb, KERN_WARNING,
40 "%s: out-of-range nid=%x, run fsck to fix.",
41 __func__, nid);
42 return -EINVAL;
43 }
44 return 0;
45}
46
47bool f2fs_available_free_memory(struct f2fs_sb_info *sbi, int type)
48{
49 struct f2fs_nm_info *nm_i = NM_I(sbi);
50 struct sysinfo val;
51 unsigned long avail_ram;
52 unsigned long mem_size = 0;
53 bool res = false;
54
55 si_meminfo(&val);
56
57
58 avail_ram = val.totalram - val.totalhigh;
59
60
61
62
63 if (type == FREE_NIDS) {
64 mem_size = (nm_i->nid_cnt[FREE_NID] *
65 sizeof(struct free_nid)) >> PAGE_SHIFT;
66 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
67 } else if (type == NAT_ENTRIES) {
68 mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >>
69 PAGE_SHIFT;
70 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
71 if (excess_cached_nats(sbi))
72 res = false;
73 } else if (type == DIRTY_DENTS) {
74 if (sbi->sb->s_bdi->wb.dirty_exceeded)
75 return false;
76 mem_size = get_pages(sbi, F2FS_DIRTY_DENTS);
77 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
78 } else if (type == INO_ENTRIES) {
79 int i;
80
81 for (i = 0; i < MAX_INO_ENTRY; i++)
82 mem_size += sbi->im[i].ino_num *
83 sizeof(struct ino_entry);
84 mem_size >>= PAGE_SHIFT;
85 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
86 } else if (type == EXTENT_CACHE) {
87 mem_size = (atomic_read(&sbi->total_ext_tree) *
88 sizeof(struct extent_tree) +
89 atomic_read(&sbi->total_ext_node) *
90 sizeof(struct extent_node)) >> PAGE_SHIFT;
91 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
92 } else if (type == INMEM_PAGES) {
93
94 mem_size = get_pages(sbi, F2FS_INMEM_PAGES);
95 res = mem_size < (val.totalram / 5);
96 } else {
97 if (!sbi->sb->s_bdi->wb.dirty_exceeded)
98 return true;
99 }
100 return res;
101}
102
103static void clear_node_page_dirty(struct page *page)
104{
105 if (PageDirty(page)) {
106 f2fs_clear_radix_tree_dirty_tag(page);
107 clear_page_dirty_for_io(page);
108 dec_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES);
109 }
110 ClearPageUptodate(page);
111}
112
113static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
114{
115 pgoff_t index = current_nat_addr(sbi, nid);
116 return f2fs_get_meta_page(sbi, index);
117}
118
119static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
120{
121 struct page *src_page;
122 struct page *dst_page;
123 pgoff_t src_off;
124 pgoff_t dst_off;
125 void *src_addr;
126 void *dst_addr;
127 struct f2fs_nm_info *nm_i = NM_I(sbi);
128
129 src_off = current_nat_addr(sbi, nid);
130 dst_off = next_nat_addr(sbi, src_off);
131
132
133 src_page = f2fs_get_meta_page(sbi, src_off);
134 dst_page = f2fs_grab_meta_page(sbi, dst_off);
135 f2fs_bug_on(sbi, PageDirty(src_page));
136
137 src_addr = page_address(src_page);
138 dst_addr = page_address(dst_page);
139 memcpy(dst_addr, src_addr, PAGE_SIZE);
140 set_page_dirty(dst_page);
141 f2fs_put_page(src_page, 1);
142
143 set_to_next_nat(nm_i, nid);
144
145 return dst_page;
146}
147
148static struct nat_entry *__alloc_nat_entry(nid_t nid, bool no_fail)
149{
150 struct nat_entry *new;
151
152 if (no_fail)
153 new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO);
154 else
155 new = kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO);
156 if (new) {
157 nat_set_nid(new, nid);
158 nat_reset_flag(new);
159 }
160 return new;
161}
162
163static void __free_nat_entry(struct nat_entry *e)
164{
165 kmem_cache_free(nat_entry_slab, e);
166}
167
168
169static struct nat_entry *__init_nat_entry(struct f2fs_nm_info *nm_i,
170 struct nat_entry *ne, struct f2fs_nat_entry *raw_ne, bool no_fail)
171{
172 if (no_fail)
173 f2fs_radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne);
174 else if (radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne))
175 return NULL;
176
177 if (raw_ne)
178 node_info_from_raw_nat(&ne->ni, raw_ne);
179 list_add_tail(&ne->list, &nm_i->nat_entries);
180 nm_i->nat_cnt++;
181 return ne;
182}
183
184static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n)
185{
186 return radix_tree_lookup(&nm_i->nat_root, n);
187}
188
189static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i,
190 nid_t start, unsigned int nr, struct nat_entry **ep)
191{
192 return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr);
193}
194
195static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e)
196{
197 list_del(&e->list);
198 radix_tree_delete(&nm_i->nat_root, nat_get_nid(e));
199 nm_i->nat_cnt--;
200 __free_nat_entry(e);
201}
202
203static struct nat_entry_set *__grab_nat_entry_set(struct f2fs_nm_info *nm_i,
204 struct nat_entry *ne)
205{
206 nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid);
207 struct nat_entry_set *head;
208
209 head = radix_tree_lookup(&nm_i->nat_set_root, set);
210 if (!head) {
211 head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS);
212
213 INIT_LIST_HEAD(&head->entry_list);
214 INIT_LIST_HEAD(&head->set_list);
215 head->set = set;
216 head->entry_cnt = 0;
217 f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head);
218 }
219 return head;
220}
221
222static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i,
223 struct nat_entry *ne)
224{
225 struct nat_entry_set *head;
226 bool new_ne = nat_get_blkaddr(ne) == NEW_ADDR;
227
228 if (!new_ne)
229 head = __grab_nat_entry_set(nm_i, ne);
230
231
232
233
234
235
236 if (!new_ne && (get_nat_flag(ne, IS_PREALLOC) ||
237 !get_nat_flag(ne, IS_DIRTY)))
238 head->entry_cnt++;
239
240 set_nat_flag(ne, IS_PREALLOC, new_ne);
241
242 if (get_nat_flag(ne, IS_DIRTY))
243 goto refresh_list;
244
245 nm_i->dirty_nat_cnt++;
246 set_nat_flag(ne, IS_DIRTY, true);
247refresh_list:
248 if (new_ne)
249 list_del_init(&ne->list);
250 else
251 list_move_tail(&ne->list, &head->entry_list);
252}
253
254static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i,
255 struct nat_entry_set *set, struct nat_entry *ne)
256{
257 list_move_tail(&ne->list, &nm_i->nat_entries);
258 set_nat_flag(ne, IS_DIRTY, false);
259 set->entry_cnt--;
260 nm_i->dirty_nat_cnt--;
261}
262
263static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i,
264 nid_t start, unsigned int nr, struct nat_entry_set **ep)
265{
266 return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep,
267 start, nr);
268}
269
270int f2fs_need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid)
271{
272 struct f2fs_nm_info *nm_i = NM_I(sbi);
273 struct nat_entry *e;
274 bool need = false;
275
276 down_read(&nm_i->nat_tree_lock);
277 e = __lookup_nat_cache(nm_i, nid);
278 if (e) {
279 if (!get_nat_flag(e, IS_CHECKPOINTED) &&
280 !get_nat_flag(e, HAS_FSYNCED_INODE))
281 need = true;
282 }
283 up_read(&nm_i->nat_tree_lock);
284 return need;
285}
286
287bool f2fs_is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
288{
289 struct f2fs_nm_info *nm_i = NM_I(sbi);
290 struct nat_entry *e;
291 bool is_cp = true;
292
293 down_read(&nm_i->nat_tree_lock);
294 e = __lookup_nat_cache(nm_i, nid);
295 if (e && !get_nat_flag(e, IS_CHECKPOINTED))
296 is_cp = false;
297 up_read(&nm_i->nat_tree_lock);
298 return is_cp;
299}
300
301bool f2fs_need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino)
302{
303 struct f2fs_nm_info *nm_i = NM_I(sbi);
304 struct nat_entry *e;
305 bool need_update = true;
306
307 down_read(&nm_i->nat_tree_lock);
308 e = __lookup_nat_cache(nm_i, ino);
309 if (e && get_nat_flag(e, HAS_LAST_FSYNC) &&
310 (get_nat_flag(e, IS_CHECKPOINTED) ||
311 get_nat_flag(e, HAS_FSYNCED_INODE)))
312 need_update = false;
313 up_read(&nm_i->nat_tree_lock);
314 return need_update;
315}
316
317
318static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid,
319 struct f2fs_nat_entry *ne)
320{
321 struct f2fs_nm_info *nm_i = NM_I(sbi);
322 struct nat_entry *new, *e;
323
324 new = __alloc_nat_entry(nid, false);
325 if (!new)
326 return;
327
328 down_write(&nm_i->nat_tree_lock);
329 e = __lookup_nat_cache(nm_i, nid);
330 if (!e)
331 e = __init_nat_entry(nm_i, new, ne, false);
332 else
333 f2fs_bug_on(sbi, nat_get_ino(e) != le32_to_cpu(ne->ino) ||
334 nat_get_blkaddr(e) !=
335 le32_to_cpu(ne->block_addr) ||
336 nat_get_version(e) != ne->version);
337 up_write(&nm_i->nat_tree_lock);
338 if (e != new)
339 __free_nat_entry(new);
340}
341
342static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
343 block_t new_blkaddr, bool fsync_done)
344{
345 struct f2fs_nm_info *nm_i = NM_I(sbi);
346 struct nat_entry *e;
347 struct nat_entry *new = __alloc_nat_entry(ni->nid, true);
348
349 down_write(&nm_i->nat_tree_lock);
350 e = __lookup_nat_cache(nm_i, ni->nid);
351 if (!e) {
352 e = __init_nat_entry(nm_i, new, NULL, true);
353 copy_node_info(&e->ni, ni);
354 f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR);
355 } else if (new_blkaddr == NEW_ADDR) {
356
357
358
359
360
361 copy_node_info(&e->ni, ni);
362 f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR);
363 }
364
365 if (e != new)
366 __free_nat_entry(new);
367
368
369 f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr);
370 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR &&
371 new_blkaddr == NULL_ADDR);
372 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR &&
373 new_blkaddr == NEW_ADDR);
374 f2fs_bug_on(sbi, is_valid_blkaddr(nat_get_blkaddr(e)) &&
375 new_blkaddr == NEW_ADDR);
376
377
378 if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) {
379 unsigned char version = nat_get_version(e);
380 nat_set_version(e, inc_node_version(version));
381 }
382
383
384 nat_set_blkaddr(e, new_blkaddr);
385 if (!is_valid_blkaddr(new_blkaddr))
386 set_nat_flag(e, IS_CHECKPOINTED, false);
387 __set_nat_cache_dirty(nm_i, e);
388
389
390 if (ni->nid != ni->ino)
391 e = __lookup_nat_cache(nm_i, ni->ino);
392 if (e) {
393 if (fsync_done && ni->nid == ni->ino)
394 set_nat_flag(e, HAS_FSYNCED_INODE, true);
395 set_nat_flag(e, HAS_LAST_FSYNC, fsync_done);
396 }
397 up_write(&nm_i->nat_tree_lock);
398}
399
400int f2fs_try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
401{
402 struct f2fs_nm_info *nm_i = NM_I(sbi);
403 int nr = nr_shrink;
404
405 if (!down_write_trylock(&nm_i->nat_tree_lock))
406 return 0;
407
408 while (nr_shrink && !list_empty(&nm_i->nat_entries)) {
409 struct nat_entry *ne;
410 ne = list_first_entry(&nm_i->nat_entries,
411 struct nat_entry, list);
412 __del_from_nat_cache(nm_i, ne);
413 nr_shrink--;
414 }
415 up_write(&nm_i->nat_tree_lock);
416 return nr - nr_shrink;
417}
418
419
420
421
422void f2fs_get_node_info(struct f2fs_sb_info *sbi, nid_t nid,
423 struct node_info *ni)
424{
425 struct f2fs_nm_info *nm_i = NM_I(sbi);
426 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
427 struct f2fs_journal *journal = curseg->journal;
428 nid_t start_nid = START_NID(nid);
429 struct f2fs_nat_block *nat_blk;
430 struct page *page = NULL;
431 struct f2fs_nat_entry ne;
432 struct nat_entry *e;
433 pgoff_t index;
434 int i;
435
436 ni->nid = nid;
437
438
439 down_read(&nm_i->nat_tree_lock);
440 e = __lookup_nat_cache(nm_i, nid);
441 if (e) {
442 ni->ino = nat_get_ino(e);
443 ni->blk_addr = nat_get_blkaddr(e);
444 ni->version = nat_get_version(e);
445 up_read(&nm_i->nat_tree_lock);
446 return;
447 }
448
449 memset(&ne, 0, sizeof(struct f2fs_nat_entry));
450
451
452 down_read(&curseg->journal_rwsem);
453 i = f2fs_lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0);
454 if (i >= 0) {
455 ne = nat_in_journal(journal, i);
456 node_info_from_raw_nat(ni, &ne);
457 }
458 up_read(&curseg->journal_rwsem);
459 if (i >= 0) {
460 up_read(&nm_i->nat_tree_lock);
461 goto cache;
462 }
463
464
465 index = current_nat_addr(sbi, nid);
466 up_read(&nm_i->nat_tree_lock);
467
468 page = f2fs_get_meta_page(sbi, index);
469 nat_blk = (struct f2fs_nat_block *)page_address(page);
470 ne = nat_blk->entries[nid - start_nid];
471 node_info_from_raw_nat(ni, &ne);
472 f2fs_put_page(page, 1);
473cache:
474
475 cache_nat_entry(sbi, nid, &ne);
476}
477
478
479
480
481static void f2fs_ra_node_pages(struct page *parent, int start, int n)
482{
483 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
484 struct blk_plug plug;
485 int i, end;
486 nid_t nid;
487
488 blk_start_plug(&plug);
489
490
491 end = start + n;
492 end = min(end, NIDS_PER_BLOCK);
493 for (i = start; i < end; i++) {
494 nid = get_nid(parent, i, false);
495 f2fs_ra_node_page(sbi, nid);
496 }
497
498 blk_finish_plug(&plug);
499}
500
501pgoff_t f2fs_get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs)
502{
503 const long direct_index = ADDRS_PER_INODE(dn->inode);
504 const long direct_blks = ADDRS_PER_BLOCK;
505 const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK;
506 unsigned int skipped_unit = ADDRS_PER_BLOCK;
507 int cur_level = dn->cur_level;
508 int max_level = dn->max_level;
509 pgoff_t base = 0;
510
511 if (!dn->max_level)
512 return pgofs + 1;
513
514 while (max_level-- > cur_level)
515 skipped_unit *= NIDS_PER_BLOCK;
516
517 switch (dn->max_level) {
518 case 3:
519 base += 2 * indirect_blks;
520
521 case 2:
522 base += 2 * direct_blks;
523
524 case 1:
525 base += direct_index;
526 break;
527 default:
528 f2fs_bug_on(F2FS_I_SB(dn->inode), 1);
529 }
530
531 return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base;
532}
533
534
535
536
537
538static int get_node_path(struct inode *inode, long block,
539 int offset[4], unsigned int noffset[4])
540{
541 const long direct_index = ADDRS_PER_INODE(inode);
542 const long direct_blks = ADDRS_PER_BLOCK;
543 const long dptrs_per_blk = NIDS_PER_BLOCK;
544 const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK;
545 const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK;
546 int n = 0;
547 int level = 0;
548
549 noffset[0] = 0;
550
551 if (block < direct_index) {
552 offset[n] = block;
553 goto got;
554 }
555 block -= direct_index;
556 if (block < direct_blks) {
557 offset[n++] = NODE_DIR1_BLOCK;
558 noffset[n] = 1;
559 offset[n] = block;
560 level = 1;
561 goto got;
562 }
563 block -= direct_blks;
564 if (block < direct_blks) {
565 offset[n++] = NODE_DIR2_BLOCK;
566 noffset[n] = 2;
567 offset[n] = block;
568 level = 1;
569 goto got;
570 }
571 block -= direct_blks;
572 if (block < indirect_blks) {
573 offset[n++] = NODE_IND1_BLOCK;
574 noffset[n] = 3;
575 offset[n++] = block / direct_blks;
576 noffset[n] = 4 + offset[n - 1];
577 offset[n] = block % direct_blks;
578 level = 2;
579 goto got;
580 }
581 block -= indirect_blks;
582 if (block < indirect_blks) {
583 offset[n++] = NODE_IND2_BLOCK;
584 noffset[n] = 4 + dptrs_per_blk;
585 offset[n++] = block / direct_blks;
586 noffset[n] = 5 + dptrs_per_blk + offset[n - 1];
587 offset[n] = block % direct_blks;
588 level = 2;
589 goto got;
590 }
591 block -= indirect_blks;
592 if (block < dindirect_blks) {
593 offset[n++] = NODE_DIND_BLOCK;
594 noffset[n] = 5 + (dptrs_per_blk * 2);
595 offset[n++] = block / indirect_blks;
596 noffset[n] = 6 + (dptrs_per_blk * 2) +
597 offset[n - 1] * (dptrs_per_blk + 1);
598 offset[n++] = (block / direct_blks) % dptrs_per_blk;
599 noffset[n] = 7 + (dptrs_per_blk * 2) +
600 offset[n - 2] * (dptrs_per_blk + 1) +
601 offset[n - 1];
602 offset[n] = block % direct_blks;
603 level = 3;
604 goto got;
605 } else {
606 return -E2BIG;
607 }
608got:
609 return level;
610}
611
612
613
614
615
616
617
618int f2fs_get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
619{
620 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
621 struct page *npage[4];
622 struct page *parent = NULL;
623 int offset[4];
624 unsigned int noffset[4];
625 nid_t nids[4];
626 int level, i = 0;
627 int err = 0;
628
629 level = get_node_path(dn->inode, index, offset, noffset);
630 if (level < 0)
631 return level;
632
633 nids[0] = dn->inode->i_ino;
634 npage[0] = dn->inode_page;
635
636 if (!npage[0]) {
637 npage[0] = f2fs_get_node_page(sbi, nids[0]);
638 if (IS_ERR(npage[0]))
639 return PTR_ERR(npage[0]);
640 }
641
642
643 if (f2fs_has_inline_data(dn->inode) && index) {
644 err = -ENOENT;
645 f2fs_put_page(npage[0], 1);
646 goto release_out;
647 }
648
649 parent = npage[0];
650 if (level != 0)
651 nids[1] = get_nid(parent, offset[0], true);
652 dn->inode_page = npage[0];
653 dn->inode_page_locked = true;
654
655
656 for (i = 1; i <= level; i++) {
657 bool done = false;
658
659 if (!nids[i] && mode == ALLOC_NODE) {
660
661 if (!f2fs_alloc_nid(sbi, &(nids[i]))) {
662 err = -ENOSPC;
663 goto release_pages;
664 }
665
666 dn->nid = nids[i];
667 npage[i] = f2fs_new_node_page(dn, noffset[i]);
668 if (IS_ERR(npage[i])) {
669 f2fs_alloc_nid_failed(sbi, nids[i]);
670 err = PTR_ERR(npage[i]);
671 goto release_pages;
672 }
673
674 set_nid(parent, offset[i - 1], nids[i], i == 1);
675 f2fs_alloc_nid_done(sbi, nids[i]);
676 done = true;
677 } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) {
678 npage[i] = f2fs_get_node_page_ra(parent, offset[i - 1]);
679 if (IS_ERR(npage[i])) {
680 err = PTR_ERR(npage[i]);
681 goto release_pages;
682 }
683 done = true;
684 }
685 if (i == 1) {
686 dn->inode_page_locked = false;
687 unlock_page(parent);
688 } else {
689 f2fs_put_page(parent, 1);
690 }
691
692 if (!done) {
693 npage[i] = f2fs_get_node_page(sbi, nids[i]);
694 if (IS_ERR(npage[i])) {
695 err = PTR_ERR(npage[i]);
696 f2fs_put_page(npage[0], 0);
697 goto release_out;
698 }
699 }
700 if (i < level) {
701 parent = npage[i];
702 nids[i + 1] = get_nid(parent, offset[i], false);
703 }
704 }
705 dn->nid = nids[level];
706 dn->ofs_in_node = offset[level];
707 dn->node_page = npage[level];
708 dn->data_blkaddr = datablock_addr(dn->inode,
709 dn->node_page, dn->ofs_in_node);
710 return 0;
711
712release_pages:
713 f2fs_put_page(parent, 1);
714 if (i > 1)
715 f2fs_put_page(npage[0], 0);
716release_out:
717 dn->inode_page = NULL;
718 dn->node_page = NULL;
719 if (err == -ENOENT) {
720 dn->cur_level = i;
721 dn->max_level = level;
722 dn->ofs_in_node = offset[level];
723 }
724 return err;
725}
726
727static void truncate_node(struct dnode_of_data *dn)
728{
729 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
730 struct node_info ni;
731
732 f2fs_get_node_info(sbi, dn->nid, &ni);
733
734
735 f2fs_invalidate_blocks(sbi, ni.blk_addr);
736 dec_valid_node_count(sbi, dn->inode, dn->nid == dn->inode->i_ino);
737 set_node_addr(sbi, &ni, NULL_ADDR, false);
738
739 if (dn->nid == dn->inode->i_ino) {
740 f2fs_remove_orphan_inode(sbi, dn->nid);
741 dec_valid_inode_count(sbi);
742 f2fs_inode_synced(dn->inode);
743 }
744
745 clear_node_page_dirty(dn->node_page);
746 set_sbi_flag(sbi, SBI_IS_DIRTY);
747
748 f2fs_put_page(dn->node_page, 1);
749
750 invalidate_mapping_pages(NODE_MAPPING(sbi),
751 dn->node_page->index, dn->node_page->index);
752
753 dn->node_page = NULL;
754 trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr);
755}
756
757static int truncate_dnode(struct dnode_of_data *dn)
758{
759 struct page *page;
760
761 if (dn->nid == 0)
762 return 1;
763
764
765 page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid);
766 if (IS_ERR(page) && PTR_ERR(page) == -ENOENT)
767 return 1;
768 else if (IS_ERR(page))
769 return PTR_ERR(page);
770
771
772 dn->node_page = page;
773 dn->ofs_in_node = 0;
774 f2fs_truncate_data_blocks(dn);
775 truncate_node(dn);
776 return 1;
777}
778
779static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs,
780 int ofs, int depth)
781{
782 struct dnode_of_data rdn = *dn;
783 struct page *page;
784 struct f2fs_node *rn;
785 nid_t child_nid;
786 unsigned int child_nofs;
787 int freed = 0;
788 int i, ret;
789
790 if (dn->nid == 0)
791 return NIDS_PER_BLOCK + 1;
792
793 trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr);
794
795 page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid);
796 if (IS_ERR(page)) {
797 trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page));
798 return PTR_ERR(page);
799 }
800
801 f2fs_ra_node_pages(page, ofs, NIDS_PER_BLOCK);
802
803 rn = F2FS_NODE(page);
804 if (depth < 3) {
805 for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) {
806 child_nid = le32_to_cpu(rn->in.nid[i]);
807 if (child_nid == 0)
808 continue;
809 rdn.nid = child_nid;
810 ret = truncate_dnode(&rdn);
811 if (ret < 0)
812 goto out_err;
813 if (set_nid(page, i, 0, false))
814 dn->node_changed = true;
815 }
816 } else {
817 child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1;
818 for (i = ofs; i < NIDS_PER_BLOCK; i++) {
819 child_nid = le32_to_cpu(rn->in.nid[i]);
820 if (child_nid == 0) {
821 child_nofs += NIDS_PER_BLOCK + 1;
822 continue;
823 }
824 rdn.nid = child_nid;
825 ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1);
826 if (ret == (NIDS_PER_BLOCK + 1)) {
827 if (set_nid(page, i, 0, false))
828 dn->node_changed = true;
829 child_nofs += ret;
830 } else if (ret < 0 && ret != -ENOENT) {
831 goto out_err;
832 }
833 }
834 freed = child_nofs;
835 }
836
837 if (!ofs) {
838
839 dn->node_page = page;
840 truncate_node(dn);
841 freed++;
842 } else {
843 f2fs_put_page(page, 1);
844 }
845 trace_f2fs_truncate_nodes_exit(dn->inode, freed);
846 return freed;
847
848out_err:
849 f2fs_put_page(page, 1);
850 trace_f2fs_truncate_nodes_exit(dn->inode, ret);
851 return ret;
852}
853
854static int truncate_partial_nodes(struct dnode_of_data *dn,
855 struct f2fs_inode *ri, int *offset, int depth)
856{
857 struct page *pages[2];
858 nid_t nid[3];
859 nid_t child_nid;
860 int err = 0;
861 int i;
862 int idx = depth - 2;
863
864 nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
865 if (!nid[0])
866 return 0;
867
868
869 for (i = 0; i < idx + 1; i++) {
870
871 pages[i] = f2fs_get_node_page(F2FS_I_SB(dn->inode), nid[i]);
872 if (IS_ERR(pages[i])) {
873 err = PTR_ERR(pages[i]);
874 idx = i - 1;
875 goto fail;
876 }
877 nid[i + 1] = get_nid(pages[i], offset[i + 1], false);
878 }
879
880 f2fs_ra_node_pages(pages[idx], offset[idx + 1], NIDS_PER_BLOCK);
881
882
883 for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) {
884 child_nid = get_nid(pages[idx], i, false);
885 if (!child_nid)
886 continue;
887 dn->nid = child_nid;
888 err = truncate_dnode(dn);
889 if (err < 0)
890 goto fail;
891 if (set_nid(pages[idx], i, 0, false))
892 dn->node_changed = true;
893 }
894
895 if (offset[idx + 1] == 0) {
896 dn->node_page = pages[idx];
897 dn->nid = nid[idx];
898 truncate_node(dn);
899 } else {
900 f2fs_put_page(pages[idx], 1);
901 }
902 offset[idx]++;
903 offset[idx + 1] = 0;
904 idx--;
905fail:
906 for (i = idx; i >= 0; i--)
907 f2fs_put_page(pages[i], 1);
908
909 trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err);
910
911 return err;
912}
913
914
915
916
917int f2fs_truncate_inode_blocks(struct inode *inode, pgoff_t from)
918{
919 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
920 int err = 0, cont = 1;
921 int level, offset[4], noffset[4];
922 unsigned int nofs = 0;
923 struct f2fs_inode *ri;
924 struct dnode_of_data dn;
925 struct page *page;
926
927 trace_f2fs_truncate_inode_blocks_enter(inode, from);
928
929 level = get_node_path(inode, from, offset, noffset);
930 if (level < 0)
931 return level;
932
933 page = f2fs_get_node_page(sbi, inode->i_ino);
934 if (IS_ERR(page)) {
935 trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page));
936 return PTR_ERR(page);
937 }
938
939 set_new_dnode(&dn, inode, page, NULL, 0);
940 unlock_page(page);
941
942 ri = F2FS_INODE(page);
943 switch (level) {
944 case 0:
945 case 1:
946 nofs = noffset[1];
947 break;
948 case 2:
949 nofs = noffset[1];
950 if (!offset[level - 1])
951 goto skip_partial;
952 err = truncate_partial_nodes(&dn, ri, offset, level);
953 if (err < 0 && err != -ENOENT)
954 goto fail;
955 nofs += 1 + NIDS_PER_BLOCK;
956 break;
957 case 3:
958 nofs = 5 + 2 * NIDS_PER_BLOCK;
959 if (!offset[level - 1])
960 goto skip_partial;
961 err = truncate_partial_nodes(&dn, ri, offset, level);
962 if (err < 0 && err != -ENOENT)
963 goto fail;
964 break;
965 default:
966 BUG();
967 }
968
969skip_partial:
970 while (cont) {
971 dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
972 switch (offset[0]) {
973 case NODE_DIR1_BLOCK:
974 case NODE_DIR2_BLOCK:
975 err = truncate_dnode(&dn);
976 break;
977
978 case NODE_IND1_BLOCK:
979 case NODE_IND2_BLOCK:
980 err = truncate_nodes(&dn, nofs, offset[1], 2);
981 break;
982
983 case NODE_DIND_BLOCK:
984 err = truncate_nodes(&dn, nofs, offset[1], 3);
985 cont = 0;
986 break;
987
988 default:
989 BUG();
990 }
991 if (err < 0 && err != -ENOENT)
992 goto fail;
993 if (offset[1] == 0 &&
994 ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) {
995 lock_page(page);
996 BUG_ON(page->mapping != NODE_MAPPING(sbi));
997 f2fs_wait_on_page_writeback(page, NODE, true);
998 ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0;
999 set_page_dirty(page);
1000 unlock_page(page);
1001 }
1002 offset[1] = 0;
1003 offset[0]++;
1004 nofs += err;
1005 }
1006fail:
1007 f2fs_put_page(page, 0);
1008 trace_f2fs_truncate_inode_blocks_exit(inode, err);
1009 return err > 0 ? 0 : err;
1010}
1011
1012
1013int f2fs_truncate_xattr_node(struct inode *inode)
1014{
1015 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1016 nid_t nid = F2FS_I(inode)->i_xattr_nid;
1017 struct dnode_of_data dn;
1018 struct page *npage;
1019
1020 if (!nid)
1021 return 0;
1022
1023 npage = f2fs_get_node_page(sbi, nid);
1024 if (IS_ERR(npage))
1025 return PTR_ERR(npage);
1026
1027 f2fs_i_xnid_write(inode, 0);
1028
1029 set_new_dnode(&dn, inode, NULL, npage, nid);
1030 truncate_node(&dn);
1031 return 0;
1032}
1033
1034
1035
1036
1037
1038int f2fs_remove_inode_page(struct inode *inode)
1039{
1040 struct dnode_of_data dn;
1041 int err;
1042
1043 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
1044 err = f2fs_get_dnode_of_data(&dn, 0, LOOKUP_NODE);
1045 if (err)
1046 return err;
1047
1048 err = f2fs_truncate_xattr_node(inode);
1049 if (err) {
1050 f2fs_put_dnode(&dn);
1051 return err;
1052 }
1053
1054
1055 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1056 S_ISLNK(inode->i_mode))
1057 f2fs_truncate_data_blocks_range(&dn, 1);
1058
1059
1060 f2fs_bug_on(F2FS_I_SB(inode),
1061 inode->i_blocks != 0 && inode->i_blocks != 8);
1062
1063
1064 truncate_node(&dn);
1065 return 0;
1066}
1067
1068struct page *f2fs_new_inode_page(struct inode *inode)
1069{
1070 struct dnode_of_data dn;
1071
1072
1073 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
1074
1075
1076 return f2fs_new_node_page(&dn, 0);
1077}
1078
1079struct page *f2fs_new_node_page(struct dnode_of_data *dn, unsigned int ofs)
1080{
1081 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1082 struct node_info new_ni;
1083 struct page *page;
1084 int err;
1085
1086 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
1087 return ERR_PTR(-EPERM);
1088
1089 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), dn->nid, false);
1090 if (!page)
1091 return ERR_PTR(-ENOMEM);
1092
1093 if (unlikely((err = inc_valid_node_count(sbi, dn->inode, !ofs))))
1094 goto fail;
1095
1096#ifdef CONFIG_F2FS_CHECK_FS
1097 f2fs_get_node_info(sbi, dn->nid, &new_ni);
1098 f2fs_bug_on(sbi, new_ni.blk_addr != NULL_ADDR);
1099#endif
1100 new_ni.nid = dn->nid;
1101 new_ni.ino = dn->inode->i_ino;
1102 new_ni.blk_addr = NULL_ADDR;
1103 new_ni.flag = 0;
1104 new_ni.version = 0;
1105 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
1106
1107 f2fs_wait_on_page_writeback(page, NODE, true);
1108 fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true);
1109 set_cold_node(page, S_ISDIR(dn->inode->i_mode));
1110 if (!PageUptodate(page))
1111 SetPageUptodate(page);
1112 if (set_page_dirty(page))
1113 dn->node_changed = true;
1114
1115 if (f2fs_has_xattr_block(ofs))
1116 f2fs_i_xnid_write(dn->inode, dn->nid);
1117
1118 if (ofs == 0)
1119 inc_valid_inode_count(sbi);
1120 return page;
1121
1122fail:
1123 clear_node_page_dirty(page);
1124 f2fs_put_page(page, 1);
1125 return ERR_PTR(err);
1126}
1127
1128
1129
1130
1131
1132
1133static int read_node_page(struct page *page, int op_flags)
1134{
1135 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1136 struct node_info ni;
1137 struct f2fs_io_info fio = {
1138 .sbi = sbi,
1139 .type = NODE,
1140 .op = REQ_OP_READ,
1141 .op_flags = op_flags,
1142 .page = page,
1143 .encrypted_page = NULL,
1144 };
1145
1146 if (PageUptodate(page))
1147 return LOCKED_PAGE;
1148
1149 f2fs_get_node_info(sbi, page->index, &ni);
1150
1151 if (unlikely(ni.blk_addr == NULL_ADDR)) {
1152 ClearPageUptodate(page);
1153 return -ENOENT;
1154 }
1155
1156 fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr;
1157 return f2fs_submit_page_bio(&fio);
1158}
1159
1160
1161
1162
1163void f2fs_ra_node_page(struct f2fs_sb_info *sbi, nid_t nid)
1164{
1165 struct page *apage;
1166 int err;
1167
1168 if (!nid)
1169 return;
1170 if (f2fs_check_nid_range(sbi, nid))
1171 return;
1172
1173 rcu_read_lock();
1174 apage = radix_tree_lookup(&NODE_MAPPING(sbi)->i_pages, nid);
1175 rcu_read_unlock();
1176 if (apage)
1177 return;
1178
1179 apage = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
1180 if (!apage)
1181 return;
1182
1183 err = read_node_page(apage, REQ_RAHEAD);
1184 f2fs_put_page(apage, err ? 1 : 0);
1185}
1186
1187static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid,
1188 struct page *parent, int start)
1189{
1190 struct page *page;
1191 int err;
1192
1193 if (!nid)
1194 return ERR_PTR(-ENOENT);
1195 if (f2fs_check_nid_range(sbi, nid))
1196 return ERR_PTR(-EINVAL);
1197repeat:
1198 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
1199 if (!page)
1200 return ERR_PTR(-ENOMEM);
1201
1202 err = read_node_page(page, 0);
1203 if (err < 0) {
1204 f2fs_put_page(page, 1);
1205 return ERR_PTR(err);
1206 } else if (err == LOCKED_PAGE) {
1207 err = 0;
1208 goto page_hit;
1209 }
1210
1211 if (parent)
1212 f2fs_ra_node_pages(parent, start + 1, MAX_RA_NODE);
1213
1214 lock_page(page);
1215
1216 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1217 f2fs_put_page(page, 1);
1218 goto repeat;
1219 }
1220
1221 if (unlikely(!PageUptodate(page))) {
1222 err = -EIO;
1223 goto out_err;
1224 }
1225
1226 if (!f2fs_inode_chksum_verify(sbi, page)) {
1227 err = -EBADMSG;
1228 goto out_err;
1229 }
1230page_hit:
1231 if(unlikely(nid != nid_of_node(page))) {
1232 f2fs_msg(sbi->sb, KERN_WARNING, "inconsistent node block, "
1233 "nid:%lu, node_footer[nid:%u,ino:%u,ofs:%u,cpver:%llu,blkaddr:%u]",
1234 nid, nid_of_node(page), ino_of_node(page),
1235 ofs_of_node(page), cpver_of_node(page),
1236 next_blkaddr_of_node(page));
1237 err = -EINVAL;
1238out_err:
1239 ClearPageUptodate(page);
1240 f2fs_put_page(page, 1);
1241 return ERR_PTR(err);
1242 }
1243 return page;
1244}
1245
1246struct page *f2fs_get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid)
1247{
1248 return __get_node_page(sbi, nid, NULL, 0);
1249}
1250
1251struct page *f2fs_get_node_page_ra(struct page *parent, int start)
1252{
1253 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
1254 nid_t nid = get_nid(parent, start, false);
1255
1256 return __get_node_page(sbi, nid, parent, start);
1257}
1258
1259static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino)
1260{
1261 struct inode *inode;
1262 struct page *page;
1263 int ret;
1264
1265
1266 inode = ilookup(sbi->sb, ino);
1267 if (!inode)
1268 return;
1269
1270 page = f2fs_pagecache_get_page(inode->i_mapping, 0,
1271 FGP_LOCK|FGP_NOWAIT, 0);
1272 if (!page)
1273 goto iput_out;
1274
1275 if (!PageUptodate(page))
1276 goto page_out;
1277
1278 if (!PageDirty(page))
1279 goto page_out;
1280
1281 if (!clear_page_dirty_for_io(page))
1282 goto page_out;
1283
1284 ret = f2fs_write_inline_data(inode, page);
1285 inode_dec_dirty_pages(inode);
1286 f2fs_remove_dirty_inode(inode);
1287 if (ret)
1288 set_page_dirty(page);
1289page_out:
1290 f2fs_put_page(page, 1);
1291iput_out:
1292 iput(inode);
1293}
1294
1295static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino)
1296{
1297 pgoff_t index;
1298 struct pagevec pvec;
1299 struct page *last_page = NULL;
1300 int nr_pages;
1301
1302 pagevec_init(&pvec);
1303 index = 0;
1304
1305 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
1306 PAGECACHE_TAG_DIRTY))) {
1307 int i;
1308
1309 for (i = 0; i < nr_pages; i++) {
1310 struct page *page = pvec.pages[i];
1311
1312 if (unlikely(f2fs_cp_error(sbi))) {
1313 f2fs_put_page(last_page, 0);
1314 pagevec_release(&pvec);
1315 return ERR_PTR(-EIO);
1316 }
1317
1318 if (!IS_DNODE(page) || !is_cold_node(page))
1319 continue;
1320 if (ino_of_node(page) != ino)
1321 continue;
1322
1323 lock_page(page);
1324
1325 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1326continue_unlock:
1327 unlock_page(page);
1328 continue;
1329 }
1330 if (ino_of_node(page) != ino)
1331 goto continue_unlock;
1332
1333 if (!PageDirty(page)) {
1334
1335 goto continue_unlock;
1336 }
1337
1338 if (last_page)
1339 f2fs_put_page(last_page, 0);
1340
1341 get_page(page);
1342 last_page = page;
1343 unlock_page(page);
1344 }
1345 pagevec_release(&pvec);
1346 cond_resched();
1347 }
1348 return last_page;
1349}
1350
1351static int __write_node_page(struct page *page, bool atomic, bool *submitted,
1352 struct writeback_control *wbc, bool do_balance,
1353 enum iostat_type io_type)
1354{
1355 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1356 nid_t nid;
1357 struct node_info ni;
1358 struct f2fs_io_info fio = {
1359 .sbi = sbi,
1360 .ino = ino_of_node(page),
1361 .type = NODE,
1362 .op = REQ_OP_WRITE,
1363 .op_flags = wbc_to_write_flags(wbc),
1364 .page = page,
1365 .encrypted_page = NULL,
1366 .submitted = false,
1367 .io_type = io_type,
1368 .io_wbc = wbc,
1369 };
1370
1371 trace_f2fs_writepage(page, NODE);
1372
1373 if (unlikely(f2fs_cp_error(sbi)))
1374 goto redirty_out;
1375
1376 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1377 goto redirty_out;
1378
1379
1380 nid = nid_of_node(page);
1381 f2fs_bug_on(sbi, page->index != nid);
1382
1383 if (wbc->for_reclaim) {
1384 if (!down_read_trylock(&sbi->node_write))
1385 goto redirty_out;
1386 } else {
1387 down_read(&sbi->node_write);
1388 }
1389
1390 f2fs_get_node_info(sbi, nid, &ni);
1391
1392
1393 if (unlikely(ni.blk_addr == NULL_ADDR)) {
1394 ClearPageUptodate(page);
1395 dec_page_count(sbi, F2FS_DIRTY_NODES);
1396 up_read(&sbi->node_write);
1397 unlock_page(page);
1398 return 0;
1399 }
1400
1401 if (atomic && !test_opt(sbi, NOBARRIER))
1402 fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
1403
1404 set_page_writeback(page);
1405 ClearPageError(page);
1406 fio.old_blkaddr = ni.blk_addr;
1407 f2fs_do_write_node_page(nid, &fio);
1408 set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page));
1409 dec_page_count(sbi, F2FS_DIRTY_NODES);
1410 up_read(&sbi->node_write);
1411
1412 if (wbc->for_reclaim) {
1413 f2fs_submit_merged_write_cond(sbi, page->mapping->host, 0,
1414 page->index, NODE);
1415 submitted = NULL;
1416 }
1417
1418 unlock_page(page);
1419
1420 if (unlikely(f2fs_cp_error(sbi))) {
1421 f2fs_submit_merged_write(sbi, NODE);
1422 submitted = NULL;
1423 }
1424 if (submitted)
1425 *submitted = fio.submitted;
1426
1427 if (do_balance)
1428 f2fs_balance_fs(sbi, false);
1429 return 0;
1430
1431redirty_out:
1432 redirty_page_for_writepage(wbc, page);
1433 return AOP_WRITEPAGE_ACTIVATE;
1434}
1435
1436void f2fs_move_node_page(struct page *node_page, int gc_type)
1437{
1438 if (gc_type == FG_GC) {
1439 struct writeback_control wbc = {
1440 .sync_mode = WB_SYNC_ALL,
1441 .nr_to_write = 1,
1442 .for_reclaim = 0,
1443 };
1444
1445 set_page_dirty(node_page);
1446 f2fs_wait_on_page_writeback(node_page, NODE, true);
1447
1448 f2fs_bug_on(F2FS_P_SB(node_page), PageWriteback(node_page));
1449 if (!clear_page_dirty_for_io(node_page))
1450 goto out_page;
1451
1452 if (__write_node_page(node_page, false, NULL,
1453 &wbc, false, FS_GC_NODE_IO))
1454 unlock_page(node_page);
1455 goto release_page;
1456 } else {
1457
1458 if (!PageWriteback(node_page))
1459 set_page_dirty(node_page);
1460 }
1461out_page:
1462 unlock_page(node_page);
1463release_page:
1464 f2fs_put_page(node_page, 0);
1465}
1466
1467static int f2fs_write_node_page(struct page *page,
1468 struct writeback_control *wbc)
1469{
1470 return __write_node_page(page, false, NULL, wbc, false, FS_NODE_IO);
1471}
1472
1473int f2fs_fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode,
1474 struct writeback_control *wbc, bool atomic)
1475{
1476 pgoff_t index;
1477 pgoff_t last_idx = ULONG_MAX;
1478 struct pagevec pvec;
1479 int ret = 0;
1480 struct page *last_page = NULL;
1481 bool marked = false;
1482 nid_t ino = inode->i_ino;
1483 int nr_pages;
1484
1485 if (atomic) {
1486 last_page = last_fsync_dnode(sbi, ino);
1487 if (IS_ERR_OR_NULL(last_page))
1488 return PTR_ERR_OR_ZERO(last_page);
1489 }
1490retry:
1491 pagevec_init(&pvec);
1492 index = 0;
1493
1494 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
1495 PAGECACHE_TAG_DIRTY))) {
1496 int i;
1497
1498 for (i = 0; i < nr_pages; i++) {
1499 struct page *page = pvec.pages[i];
1500 bool submitted = false;
1501
1502 if (unlikely(f2fs_cp_error(sbi))) {
1503 f2fs_put_page(last_page, 0);
1504 pagevec_release(&pvec);
1505 ret = -EIO;
1506 goto out;
1507 }
1508
1509 if (!IS_DNODE(page) || !is_cold_node(page))
1510 continue;
1511 if (ino_of_node(page) != ino)
1512 continue;
1513
1514 lock_page(page);
1515
1516 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1517continue_unlock:
1518 unlock_page(page);
1519 continue;
1520 }
1521 if (ino_of_node(page) != ino)
1522 goto continue_unlock;
1523
1524 if (!PageDirty(page) && page != last_page) {
1525
1526 goto continue_unlock;
1527 }
1528
1529 f2fs_wait_on_page_writeback(page, NODE, true);
1530 BUG_ON(PageWriteback(page));
1531
1532 set_fsync_mark(page, 0);
1533 set_dentry_mark(page, 0);
1534
1535 if (!atomic || page == last_page) {
1536 set_fsync_mark(page, 1);
1537 if (IS_INODE(page)) {
1538 if (is_inode_flag_set(inode,
1539 FI_DIRTY_INODE))
1540 f2fs_update_inode(inode, page);
1541 set_dentry_mark(page,
1542 f2fs_need_dentry_mark(sbi, ino));
1543 }
1544
1545 if (!PageDirty(page))
1546 set_page_dirty(page);
1547 }
1548
1549 if (!clear_page_dirty_for_io(page))
1550 goto continue_unlock;
1551
1552 ret = __write_node_page(page, atomic &&
1553 page == last_page,
1554 &submitted, wbc, true,
1555 FS_NODE_IO);
1556 if (ret) {
1557 unlock_page(page);
1558 f2fs_put_page(last_page, 0);
1559 break;
1560 } else if (submitted) {
1561 last_idx = page->index;
1562 }
1563
1564 if (page == last_page) {
1565 f2fs_put_page(page, 0);
1566 marked = true;
1567 break;
1568 }
1569 }
1570 pagevec_release(&pvec);
1571 cond_resched();
1572
1573 if (ret || marked)
1574 break;
1575 }
1576 if (!ret && atomic && !marked) {
1577 f2fs_msg(sbi->sb, KERN_DEBUG,
1578 "Retry to write fsync mark: ino=%u, idx=%lx",
1579 ino, last_page->index);
1580 lock_page(last_page);
1581 f2fs_wait_on_page_writeback(last_page, NODE, true);
1582 set_page_dirty(last_page);
1583 unlock_page(last_page);
1584 goto retry;
1585 }
1586out:
1587 if (last_idx != ULONG_MAX)
1588 f2fs_submit_merged_write_cond(sbi, NULL, ino, last_idx, NODE);
1589 return ret ? -EIO: 0;
1590}
1591
1592int f2fs_sync_node_pages(struct f2fs_sb_info *sbi,
1593 struct writeback_control *wbc,
1594 bool do_balance, enum iostat_type io_type)
1595{
1596 pgoff_t index;
1597 struct pagevec pvec;
1598 int step = 0;
1599 int nwritten = 0;
1600 int ret = 0;
1601 int nr_pages, done = 0;
1602
1603 pagevec_init(&pvec);
1604
1605next_step:
1606 index = 0;
1607
1608 while (!done && (nr_pages = pagevec_lookup_tag(&pvec,
1609 NODE_MAPPING(sbi), &index, PAGECACHE_TAG_DIRTY))) {
1610 int i;
1611
1612 for (i = 0; i < nr_pages; i++) {
1613 struct page *page = pvec.pages[i];
1614 bool submitted = false;
1615
1616
1617 if (atomic_read(&sbi->wb_sync_req[NODE]) &&
1618 wbc->sync_mode == WB_SYNC_NONE) {
1619 done = 1;
1620 break;
1621 }
1622
1623
1624
1625
1626
1627
1628
1629 if (step == 0 && IS_DNODE(page))
1630 continue;
1631 if (step == 1 && (!IS_DNODE(page) ||
1632 is_cold_node(page)))
1633 continue;
1634 if (step == 2 && (!IS_DNODE(page) ||
1635 !is_cold_node(page)))
1636 continue;
1637lock_node:
1638 if (!trylock_page(page))
1639 continue;
1640
1641 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1642continue_unlock:
1643 unlock_page(page);
1644 continue;
1645 }
1646
1647 if (!PageDirty(page)) {
1648
1649 goto continue_unlock;
1650 }
1651
1652
1653 if (is_inline_node(page)) {
1654 clear_inline_node(page);
1655 unlock_page(page);
1656 flush_inline_data(sbi, ino_of_node(page));
1657 goto lock_node;
1658 }
1659
1660 f2fs_wait_on_page_writeback(page, NODE, true);
1661
1662 BUG_ON(PageWriteback(page));
1663 if (!clear_page_dirty_for_io(page))
1664 goto continue_unlock;
1665
1666 set_fsync_mark(page, 0);
1667 set_dentry_mark(page, 0);
1668
1669 ret = __write_node_page(page, false, &submitted,
1670 wbc, do_balance, io_type);
1671 if (ret)
1672 unlock_page(page);
1673 else if (submitted)
1674 nwritten++;
1675
1676 if (--wbc->nr_to_write == 0)
1677 break;
1678 }
1679 pagevec_release(&pvec);
1680 cond_resched();
1681
1682 if (wbc->nr_to_write == 0) {
1683 step = 2;
1684 break;
1685 }
1686 }
1687
1688 if (step < 2) {
1689 step++;
1690 goto next_step;
1691 }
1692
1693 if (nwritten)
1694 f2fs_submit_merged_write(sbi, NODE);
1695
1696 if (unlikely(f2fs_cp_error(sbi)))
1697 return -EIO;
1698 return ret;
1699}
1700
1701int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino)
1702{
1703 pgoff_t index = 0;
1704 struct pagevec pvec;
1705 int ret2, ret = 0;
1706 int nr_pages;
1707
1708 pagevec_init(&pvec);
1709
1710 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
1711 PAGECACHE_TAG_WRITEBACK))) {
1712 int i;
1713
1714 for (i = 0; i < nr_pages; i++) {
1715 struct page *page = pvec.pages[i];
1716
1717 if (ino && ino_of_node(page) == ino) {
1718 f2fs_wait_on_page_writeback(page, NODE, true);
1719 if (TestClearPageError(page))
1720 ret = -EIO;
1721 }
1722 }
1723 pagevec_release(&pvec);
1724 cond_resched();
1725 }
1726
1727 ret2 = filemap_check_errors(NODE_MAPPING(sbi));
1728 if (!ret)
1729 ret = ret2;
1730 return ret;
1731}
1732
1733static int f2fs_write_node_pages(struct address_space *mapping,
1734 struct writeback_control *wbc)
1735{
1736 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
1737 struct blk_plug plug;
1738 long diff;
1739
1740 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1741 goto skip_write;
1742
1743
1744 f2fs_balance_fs_bg(sbi);
1745
1746
1747 if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE))
1748 goto skip_write;
1749
1750 if (wbc->sync_mode == WB_SYNC_ALL)
1751 atomic_inc(&sbi->wb_sync_req[NODE]);
1752 else if (atomic_read(&sbi->wb_sync_req[NODE]))
1753 goto skip_write;
1754
1755 trace_f2fs_writepages(mapping->host, wbc, NODE);
1756
1757 diff = nr_pages_to_write(sbi, NODE, wbc);
1758 blk_start_plug(&plug);
1759 f2fs_sync_node_pages(sbi, wbc, true, FS_NODE_IO);
1760 blk_finish_plug(&plug);
1761 wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff);
1762
1763 if (wbc->sync_mode == WB_SYNC_ALL)
1764 atomic_dec(&sbi->wb_sync_req[NODE]);
1765 return 0;
1766
1767skip_write:
1768 wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES);
1769 trace_f2fs_writepages(mapping->host, wbc, NODE);
1770 return 0;
1771}
1772
1773static int f2fs_set_node_page_dirty(struct page *page)
1774{
1775 trace_f2fs_set_page_dirty(page, NODE);
1776
1777 if (!PageUptodate(page))
1778 SetPageUptodate(page);
1779 if (!PageDirty(page)) {
1780 __set_page_dirty_nobuffers(page);
1781 inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES);
1782 SetPagePrivate(page);
1783 f2fs_trace_pid(page);
1784 return 1;
1785 }
1786 return 0;
1787}
1788
1789
1790
1791
1792const struct address_space_operations f2fs_node_aops = {
1793 .writepage = f2fs_write_node_page,
1794 .writepages = f2fs_write_node_pages,
1795 .set_page_dirty = f2fs_set_node_page_dirty,
1796 .invalidatepage = f2fs_invalidate_page,
1797 .releasepage = f2fs_release_page,
1798#ifdef CONFIG_MIGRATION
1799 .migratepage = f2fs_migrate_page,
1800#endif
1801};
1802
1803static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i,
1804 nid_t n)
1805{
1806 return radix_tree_lookup(&nm_i->free_nid_root, n);
1807}
1808
1809static int __insert_free_nid(struct f2fs_sb_info *sbi,
1810 struct free_nid *i, enum nid_state state)
1811{
1812 struct f2fs_nm_info *nm_i = NM_I(sbi);
1813
1814 int err = radix_tree_insert(&nm_i->free_nid_root, i->nid, i);
1815 if (err)
1816 return err;
1817
1818 f2fs_bug_on(sbi, state != i->state);
1819 nm_i->nid_cnt[state]++;
1820 if (state == FREE_NID)
1821 list_add_tail(&i->list, &nm_i->free_nid_list);
1822 return 0;
1823}
1824
1825static void __remove_free_nid(struct f2fs_sb_info *sbi,
1826 struct free_nid *i, enum nid_state state)
1827{
1828 struct f2fs_nm_info *nm_i = NM_I(sbi);
1829
1830 f2fs_bug_on(sbi, state != i->state);
1831 nm_i->nid_cnt[state]--;
1832 if (state == FREE_NID)
1833 list_del(&i->list);
1834 radix_tree_delete(&nm_i->free_nid_root, i->nid);
1835}
1836
1837static void __move_free_nid(struct f2fs_sb_info *sbi, struct free_nid *i,
1838 enum nid_state org_state, enum nid_state dst_state)
1839{
1840 struct f2fs_nm_info *nm_i = NM_I(sbi);
1841
1842 f2fs_bug_on(sbi, org_state != i->state);
1843 i->state = dst_state;
1844 nm_i->nid_cnt[org_state]--;
1845 nm_i->nid_cnt[dst_state]++;
1846
1847 switch (dst_state) {
1848 case PREALLOC_NID:
1849 list_del(&i->list);
1850 break;
1851 case FREE_NID:
1852 list_add_tail(&i->list, &nm_i->free_nid_list);
1853 break;
1854 default:
1855 BUG_ON(1);
1856 }
1857}
1858
1859static void update_free_nid_bitmap(struct f2fs_sb_info *sbi, nid_t nid,
1860 bool set, bool build)
1861{
1862 struct f2fs_nm_info *nm_i = NM_I(sbi);
1863 unsigned int nat_ofs = NAT_BLOCK_OFFSET(nid);
1864 unsigned int nid_ofs = nid - START_NID(nid);
1865
1866 if (!test_bit_le(nat_ofs, nm_i->nat_block_bitmap))
1867 return;
1868
1869 if (set) {
1870 if (test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
1871 return;
1872 __set_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
1873 nm_i->free_nid_count[nat_ofs]++;
1874 } else {
1875 if (!test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
1876 return;
1877 __clear_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
1878 if (!build)
1879 nm_i->free_nid_count[nat_ofs]--;
1880 }
1881}
1882
1883
1884static bool add_free_nid(struct f2fs_sb_info *sbi,
1885 nid_t nid, bool build, bool update)
1886{
1887 struct f2fs_nm_info *nm_i = NM_I(sbi);
1888 struct free_nid *i, *e;
1889 struct nat_entry *ne;
1890 int err = -EINVAL;
1891 bool ret = false;
1892
1893
1894 if (unlikely(nid == 0))
1895 return false;
1896
1897 i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS);
1898 i->nid = nid;
1899 i->state = FREE_NID;
1900
1901 radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
1902
1903 spin_lock(&nm_i->nid_list_lock);
1904
1905 if (build) {
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927 ne = __lookup_nat_cache(nm_i, nid);
1928 if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) ||
1929 nat_get_blkaddr(ne) != NULL_ADDR))
1930 goto err_out;
1931
1932 e = __lookup_free_nid_list(nm_i, nid);
1933 if (e) {
1934 if (e->state == FREE_NID)
1935 ret = true;
1936 goto err_out;
1937 }
1938 }
1939 ret = true;
1940 err = __insert_free_nid(sbi, i, FREE_NID);
1941err_out:
1942 if (update) {
1943 update_free_nid_bitmap(sbi, nid, ret, build);
1944 if (!build)
1945 nm_i->available_nids++;
1946 }
1947 spin_unlock(&nm_i->nid_list_lock);
1948 radix_tree_preload_end();
1949
1950 if (err)
1951 kmem_cache_free(free_nid_slab, i);
1952 return ret;
1953}
1954
1955static void remove_free_nid(struct f2fs_sb_info *sbi, nid_t nid)
1956{
1957 struct f2fs_nm_info *nm_i = NM_I(sbi);
1958 struct free_nid *i;
1959 bool need_free = false;
1960
1961 spin_lock(&nm_i->nid_list_lock);
1962 i = __lookup_free_nid_list(nm_i, nid);
1963 if (i && i->state == FREE_NID) {
1964 __remove_free_nid(sbi, i, FREE_NID);
1965 need_free = true;
1966 }
1967 spin_unlock(&nm_i->nid_list_lock);
1968
1969 if (need_free)
1970 kmem_cache_free(free_nid_slab, i);
1971}
1972
1973static void scan_nat_page(struct f2fs_sb_info *sbi,
1974 struct page *nat_page, nid_t start_nid)
1975{
1976 struct f2fs_nm_info *nm_i = NM_I(sbi);
1977 struct f2fs_nat_block *nat_blk = page_address(nat_page);
1978 block_t blk_addr;
1979 unsigned int nat_ofs = NAT_BLOCK_OFFSET(start_nid);
1980 int i;
1981
1982 __set_bit_le(nat_ofs, nm_i->nat_block_bitmap);
1983
1984 i = start_nid % NAT_ENTRY_PER_BLOCK;
1985
1986 for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) {
1987 if (unlikely(start_nid >= nm_i->max_nid))
1988 break;
1989
1990 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr);
1991 f2fs_bug_on(sbi, blk_addr == NEW_ADDR);
1992 if (blk_addr == NULL_ADDR) {
1993 add_free_nid(sbi, start_nid, true, true);
1994 } else {
1995 spin_lock(&NM_I(sbi)->nid_list_lock);
1996 update_free_nid_bitmap(sbi, start_nid, false, true);
1997 spin_unlock(&NM_I(sbi)->nid_list_lock);
1998 }
1999 }
2000}
2001
2002static void scan_curseg_cache(struct f2fs_sb_info *sbi)
2003{
2004 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2005 struct f2fs_journal *journal = curseg->journal;
2006 int i;
2007
2008 down_read(&curseg->journal_rwsem);
2009 for (i = 0; i < nats_in_cursum(journal); i++) {
2010 block_t addr;
2011 nid_t nid;
2012
2013 addr = le32_to_cpu(nat_in_journal(journal, i).block_addr);
2014 nid = le32_to_cpu(nid_in_journal(journal, i));
2015 if (addr == NULL_ADDR)
2016 add_free_nid(sbi, nid, true, false);
2017 else
2018 remove_free_nid(sbi, nid);
2019 }
2020 up_read(&curseg->journal_rwsem);
2021}
2022
2023static void scan_free_nid_bits(struct f2fs_sb_info *sbi)
2024{
2025 struct f2fs_nm_info *nm_i = NM_I(sbi);
2026 unsigned int i, idx;
2027 nid_t nid;
2028
2029 down_read(&nm_i->nat_tree_lock);
2030
2031 for (i = 0; i < nm_i->nat_blocks; i++) {
2032 if (!test_bit_le(i, nm_i->nat_block_bitmap))
2033 continue;
2034 if (!nm_i->free_nid_count[i])
2035 continue;
2036 for (idx = 0; idx < NAT_ENTRY_PER_BLOCK; idx++) {
2037 idx = find_next_bit_le(nm_i->free_nid_bitmap[i],
2038 NAT_ENTRY_PER_BLOCK, idx);
2039 if (idx >= NAT_ENTRY_PER_BLOCK)
2040 break;
2041
2042 nid = i * NAT_ENTRY_PER_BLOCK + idx;
2043 add_free_nid(sbi, nid, true, false);
2044
2045 if (nm_i->nid_cnt[FREE_NID] >= MAX_FREE_NIDS)
2046 goto out;
2047 }
2048 }
2049out:
2050 scan_curseg_cache(sbi);
2051
2052 up_read(&nm_i->nat_tree_lock);
2053}
2054
2055static void __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
2056 bool sync, bool mount)
2057{
2058 struct f2fs_nm_info *nm_i = NM_I(sbi);
2059 int i = 0;
2060 nid_t nid = nm_i->next_scan_nid;
2061
2062 if (unlikely(nid >= nm_i->max_nid))
2063 nid = 0;
2064
2065
2066 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
2067 return;
2068
2069 if (!sync && !f2fs_available_free_memory(sbi, FREE_NIDS))
2070 return;
2071
2072 if (!mount) {
2073
2074 scan_free_nid_bits(sbi);
2075
2076 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
2077 return;
2078 }
2079
2080
2081 f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES,
2082 META_NAT, true);
2083
2084 down_read(&nm_i->nat_tree_lock);
2085
2086 while (1) {
2087 if (!test_bit_le(NAT_BLOCK_OFFSET(nid),
2088 nm_i->nat_block_bitmap)) {
2089 struct page *page = get_current_nat_page(sbi, nid);
2090
2091 scan_nat_page(sbi, page, nid);
2092 f2fs_put_page(page, 1);
2093 }
2094
2095 nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK));
2096 if (unlikely(nid >= nm_i->max_nid))
2097 nid = 0;
2098
2099 if (++i >= FREE_NID_PAGES)
2100 break;
2101 }
2102
2103
2104 nm_i->next_scan_nid = nid;
2105
2106
2107 scan_curseg_cache(sbi);
2108
2109 up_read(&nm_i->nat_tree_lock);
2110
2111 f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid),
2112 nm_i->ra_nid_pages, META_NAT, false);
2113}
2114
2115void f2fs_build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount)
2116{
2117 mutex_lock(&NM_I(sbi)->build_lock);
2118 __f2fs_build_free_nids(sbi, sync, mount);
2119 mutex_unlock(&NM_I(sbi)->build_lock);
2120}
2121
2122
2123
2124
2125
2126
2127bool f2fs_alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid)
2128{
2129 struct f2fs_nm_info *nm_i = NM_I(sbi);
2130 struct free_nid *i = NULL;
2131retry:
2132#ifdef CONFIG_F2FS_FAULT_INJECTION
2133 if (time_to_inject(sbi, FAULT_ALLOC_NID)) {
2134 f2fs_show_injection_info(FAULT_ALLOC_NID);
2135 return false;
2136 }
2137#endif
2138 spin_lock(&nm_i->nid_list_lock);
2139
2140 if (unlikely(nm_i->available_nids == 0)) {
2141 spin_unlock(&nm_i->nid_list_lock);
2142 return false;
2143 }
2144
2145
2146 if (nm_i->nid_cnt[FREE_NID] && !on_f2fs_build_free_nids(nm_i)) {
2147 f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list));
2148 i = list_first_entry(&nm_i->free_nid_list,
2149 struct free_nid, list);
2150 *nid = i->nid;
2151
2152 __move_free_nid(sbi, i, FREE_NID, PREALLOC_NID);
2153 nm_i->available_nids--;
2154
2155 update_free_nid_bitmap(sbi, *nid, false, false);
2156
2157 spin_unlock(&nm_i->nid_list_lock);
2158 return true;
2159 }
2160 spin_unlock(&nm_i->nid_list_lock);
2161
2162
2163 f2fs_build_free_nids(sbi, true, false);
2164 goto retry;
2165}
2166
2167
2168
2169
2170void f2fs_alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid)
2171{
2172 struct f2fs_nm_info *nm_i = NM_I(sbi);
2173 struct free_nid *i;
2174
2175 spin_lock(&nm_i->nid_list_lock);
2176 i = __lookup_free_nid_list(nm_i, nid);
2177 f2fs_bug_on(sbi, !i);
2178 __remove_free_nid(sbi, i, PREALLOC_NID);
2179 spin_unlock(&nm_i->nid_list_lock);
2180
2181 kmem_cache_free(free_nid_slab, i);
2182}
2183
2184
2185
2186
2187void f2fs_alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid)
2188{
2189 struct f2fs_nm_info *nm_i = NM_I(sbi);
2190 struct free_nid *i;
2191 bool need_free = false;
2192
2193 if (!nid)
2194 return;
2195
2196 spin_lock(&nm_i->nid_list_lock);
2197 i = __lookup_free_nid_list(nm_i, nid);
2198 f2fs_bug_on(sbi, !i);
2199
2200 if (!f2fs_available_free_memory(sbi, FREE_NIDS)) {
2201 __remove_free_nid(sbi, i, PREALLOC_NID);
2202 need_free = true;
2203 } else {
2204 __move_free_nid(sbi, i, PREALLOC_NID, FREE_NID);
2205 }
2206
2207 nm_i->available_nids++;
2208
2209 update_free_nid_bitmap(sbi, nid, true, false);
2210
2211 spin_unlock(&nm_i->nid_list_lock);
2212
2213 if (need_free)
2214 kmem_cache_free(free_nid_slab, i);
2215}
2216
2217int f2fs_try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink)
2218{
2219 struct f2fs_nm_info *nm_i = NM_I(sbi);
2220 struct free_nid *i, *next;
2221 int nr = nr_shrink;
2222
2223 if (nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
2224 return 0;
2225
2226 if (!mutex_trylock(&nm_i->build_lock))
2227 return 0;
2228
2229 spin_lock(&nm_i->nid_list_lock);
2230 list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) {
2231 if (nr_shrink <= 0 ||
2232 nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
2233 break;
2234
2235 __remove_free_nid(sbi, i, FREE_NID);
2236 kmem_cache_free(free_nid_slab, i);
2237 nr_shrink--;
2238 }
2239 spin_unlock(&nm_i->nid_list_lock);
2240 mutex_unlock(&nm_i->build_lock);
2241
2242 return nr - nr_shrink;
2243}
2244
2245void f2fs_recover_inline_xattr(struct inode *inode, struct page *page)
2246{
2247 void *src_addr, *dst_addr;
2248 size_t inline_size;
2249 struct page *ipage;
2250 struct f2fs_inode *ri;
2251
2252 ipage = f2fs_get_node_page(F2FS_I_SB(inode), inode->i_ino);
2253 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage));
2254
2255 ri = F2FS_INODE(page);
2256 if (ri->i_inline & F2FS_INLINE_XATTR) {
2257 set_inode_flag(inode, FI_INLINE_XATTR);
2258 } else {
2259 clear_inode_flag(inode, FI_INLINE_XATTR);
2260 goto update_inode;
2261 }
2262
2263 dst_addr = inline_xattr_addr(inode, ipage);
2264 src_addr = inline_xattr_addr(inode, page);
2265 inline_size = inline_xattr_size(inode);
2266
2267 f2fs_wait_on_page_writeback(ipage, NODE, true);
2268 memcpy(dst_addr, src_addr, inline_size);
2269update_inode:
2270 f2fs_update_inode(inode, ipage);
2271 f2fs_put_page(ipage, 1);
2272}
2273
2274int f2fs_recover_xattr_data(struct inode *inode, struct page *page)
2275{
2276 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2277 nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid;
2278 nid_t new_xnid;
2279 struct dnode_of_data dn;
2280 struct node_info ni;
2281 struct page *xpage;
2282
2283 if (!prev_xnid)
2284 goto recover_xnid;
2285
2286
2287 f2fs_get_node_info(sbi, prev_xnid, &ni);
2288 f2fs_invalidate_blocks(sbi, ni.blk_addr);
2289 dec_valid_node_count(sbi, inode, false);
2290 set_node_addr(sbi, &ni, NULL_ADDR, false);
2291
2292recover_xnid:
2293
2294 if (!f2fs_alloc_nid(sbi, &new_xnid))
2295 return -ENOSPC;
2296
2297 set_new_dnode(&dn, inode, NULL, NULL, new_xnid);
2298 xpage = f2fs_new_node_page(&dn, XATTR_NODE_OFFSET);
2299 if (IS_ERR(xpage)) {
2300 f2fs_alloc_nid_failed(sbi, new_xnid);
2301 return PTR_ERR(xpage);
2302 }
2303
2304 f2fs_alloc_nid_done(sbi, new_xnid);
2305 f2fs_update_inode_page(inode);
2306
2307
2308 memcpy(F2FS_NODE(xpage), F2FS_NODE(page), VALID_XATTR_BLOCK_SIZE);
2309
2310 set_page_dirty(xpage);
2311 f2fs_put_page(xpage, 1);
2312
2313 return 0;
2314}
2315
2316int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct page *page)
2317{
2318 struct f2fs_inode *src, *dst;
2319 nid_t ino = ino_of_node(page);
2320 struct node_info old_ni, new_ni;
2321 struct page *ipage;
2322
2323 f2fs_get_node_info(sbi, ino, &old_ni);
2324
2325 if (unlikely(old_ni.blk_addr != NULL_ADDR))
2326 return -EINVAL;
2327retry:
2328 ipage = f2fs_grab_cache_page(NODE_MAPPING(sbi), ino, false);
2329 if (!ipage) {
2330 congestion_wait(BLK_RW_ASYNC, HZ/50);
2331 goto retry;
2332 }
2333
2334
2335 remove_free_nid(sbi, ino);
2336
2337 if (!PageUptodate(ipage))
2338 SetPageUptodate(ipage);
2339 fill_node_footer(ipage, ino, ino, 0, true);
2340 set_cold_node(page, false);
2341
2342 src = F2FS_INODE(page);
2343 dst = F2FS_INODE(ipage);
2344
2345 memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src);
2346 dst->i_size = 0;
2347 dst->i_blocks = cpu_to_le64(1);
2348 dst->i_links = cpu_to_le32(1);
2349 dst->i_xattr_nid = 0;
2350 dst->i_inline = src->i_inline & (F2FS_INLINE_XATTR | F2FS_EXTRA_ATTR);
2351 if (dst->i_inline & F2FS_EXTRA_ATTR) {
2352 dst->i_extra_isize = src->i_extra_isize;
2353
2354 if (f2fs_sb_has_flexible_inline_xattr(sbi->sb) &&
2355 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2356 i_inline_xattr_size))
2357 dst->i_inline_xattr_size = src->i_inline_xattr_size;
2358
2359 if (f2fs_sb_has_project_quota(sbi->sb) &&
2360 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2361 i_projid))
2362 dst->i_projid = src->i_projid;
2363 }
2364
2365 new_ni = old_ni;
2366 new_ni.ino = ino;
2367
2368 if (unlikely(inc_valid_node_count(sbi, NULL, true)))
2369 WARN_ON(1);
2370 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
2371 inc_valid_inode_count(sbi);
2372 set_page_dirty(ipage);
2373 f2fs_put_page(ipage, 1);
2374 return 0;
2375}
2376
2377void f2fs_restore_node_summary(struct f2fs_sb_info *sbi,
2378 unsigned int segno, struct f2fs_summary_block *sum)
2379{
2380 struct f2fs_node *rn;
2381 struct f2fs_summary *sum_entry;
2382 block_t addr;
2383 int i, idx, last_offset, nrpages;
2384
2385
2386 last_offset = sbi->blocks_per_seg;
2387 addr = START_BLOCK(sbi, segno);
2388 sum_entry = &sum->entries[0];
2389
2390 for (i = 0; i < last_offset; i += nrpages, addr += nrpages) {
2391 nrpages = min(last_offset - i, BIO_MAX_PAGES);
2392
2393
2394 f2fs_ra_meta_pages(sbi, addr, nrpages, META_POR, true);
2395
2396 for (idx = addr; idx < addr + nrpages; idx++) {
2397 struct page *page = f2fs_get_tmp_page(sbi, idx);
2398
2399 rn = F2FS_NODE(page);
2400 sum_entry->nid = rn->footer.nid;
2401 sum_entry->version = 0;
2402 sum_entry->ofs_in_node = 0;
2403 sum_entry++;
2404 f2fs_put_page(page, 1);
2405 }
2406
2407 invalidate_mapping_pages(META_MAPPING(sbi), addr,
2408 addr + nrpages);
2409 }
2410}
2411
2412static void remove_nats_in_journal(struct f2fs_sb_info *sbi)
2413{
2414 struct f2fs_nm_info *nm_i = NM_I(sbi);
2415 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2416 struct f2fs_journal *journal = curseg->journal;
2417 int i;
2418
2419 down_write(&curseg->journal_rwsem);
2420 for (i = 0; i < nats_in_cursum(journal); i++) {
2421 struct nat_entry *ne;
2422 struct f2fs_nat_entry raw_ne;
2423 nid_t nid = le32_to_cpu(nid_in_journal(journal, i));
2424
2425 raw_ne = nat_in_journal(journal, i);
2426
2427 ne = __lookup_nat_cache(nm_i, nid);
2428 if (!ne) {
2429 ne = __alloc_nat_entry(nid, true);
2430 __init_nat_entry(nm_i, ne, &raw_ne, true);
2431 }
2432
2433
2434
2435
2436
2437
2438 if (!get_nat_flag(ne, IS_DIRTY) &&
2439 le32_to_cpu(raw_ne.block_addr) == NULL_ADDR) {
2440 spin_lock(&nm_i->nid_list_lock);
2441 nm_i->available_nids--;
2442 spin_unlock(&nm_i->nid_list_lock);
2443 }
2444
2445 __set_nat_cache_dirty(nm_i, ne);
2446 }
2447 update_nats_in_cursum(journal, -i);
2448 up_write(&curseg->journal_rwsem);
2449}
2450
2451static void __adjust_nat_entry_set(struct nat_entry_set *nes,
2452 struct list_head *head, int max)
2453{
2454 struct nat_entry_set *cur;
2455
2456 if (nes->entry_cnt >= max)
2457 goto add_out;
2458
2459 list_for_each_entry(cur, head, set_list) {
2460 if (cur->entry_cnt >= nes->entry_cnt) {
2461 list_add(&nes->set_list, cur->set_list.prev);
2462 return;
2463 }
2464 }
2465add_out:
2466 list_add_tail(&nes->set_list, head);
2467}
2468
2469static void __update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid,
2470 struct page *page)
2471{
2472 struct f2fs_nm_info *nm_i = NM_I(sbi);
2473 unsigned int nat_index = start_nid / NAT_ENTRY_PER_BLOCK;
2474 struct f2fs_nat_block *nat_blk = page_address(page);
2475 int valid = 0;
2476 int i = 0;
2477
2478 if (!enabled_nat_bits(sbi, NULL))
2479 return;
2480
2481 if (nat_index == 0) {
2482 valid = 1;
2483 i = 1;
2484 }
2485 for (; i < NAT_ENTRY_PER_BLOCK; i++) {
2486 if (nat_blk->entries[i].block_addr != NULL_ADDR)
2487 valid++;
2488 }
2489 if (valid == 0) {
2490 __set_bit_le(nat_index, nm_i->empty_nat_bits);
2491 __clear_bit_le(nat_index, nm_i->full_nat_bits);
2492 return;
2493 }
2494
2495 __clear_bit_le(nat_index, nm_i->empty_nat_bits);
2496 if (valid == NAT_ENTRY_PER_BLOCK)
2497 __set_bit_le(nat_index, nm_i->full_nat_bits);
2498 else
2499 __clear_bit_le(nat_index, nm_i->full_nat_bits);
2500}
2501
2502static void __flush_nat_entry_set(struct f2fs_sb_info *sbi,
2503 struct nat_entry_set *set, struct cp_control *cpc)
2504{
2505 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2506 struct f2fs_journal *journal = curseg->journal;
2507 nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK;
2508 bool to_journal = true;
2509 struct f2fs_nat_block *nat_blk;
2510 struct nat_entry *ne, *cur;
2511 struct page *page = NULL;
2512
2513
2514
2515
2516
2517
2518 if (enabled_nat_bits(sbi, cpc) ||
2519 !__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL))
2520 to_journal = false;
2521
2522 if (to_journal) {
2523 down_write(&curseg->journal_rwsem);
2524 } else {
2525 page = get_next_nat_page(sbi, start_nid);
2526 nat_blk = page_address(page);
2527 f2fs_bug_on(sbi, !nat_blk);
2528 }
2529
2530
2531 list_for_each_entry_safe(ne, cur, &set->entry_list, list) {
2532 struct f2fs_nat_entry *raw_ne;
2533 nid_t nid = nat_get_nid(ne);
2534 int offset;
2535
2536 f2fs_bug_on(sbi, nat_get_blkaddr(ne) == NEW_ADDR);
2537
2538 if (to_journal) {
2539 offset = f2fs_lookup_journal_in_cursum(journal,
2540 NAT_JOURNAL, nid, 1);
2541 f2fs_bug_on(sbi, offset < 0);
2542 raw_ne = &nat_in_journal(journal, offset);
2543 nid_in_journal(journal, offset) = cpu_to_le32(nid);
2544 } else {
2545 raw_ne = &nat_blk->entries[nid - start_nid];
2546 }
2547 raw_nat_from_node_info(raw_ne, &ne->ni);
2548 nat_reset_flag(ne);
2549 __clear_nat_cache_dirty(NM_I(sbi), set, ne);
2550 if (nat_get_blkaddr(ne) == NULL_ADDR) {
2551 add_free_nid(sbi, nid, false, true);
2552 } else {
2553 spin_lock(&NM_I(sbi)->nid_list_lock);
2554 update_free_nid_bitmap(sbi, nid, false, false);
2555 spin_unlock(&NM_I(sbi)->nid_list_lock);
2556 }
2557 }
2558
2559 if (to_journal) {
2560 up_write(&curseg->journal_rwsem);
2561 } else {
2562 __update_nat_bits(sbi, start_nid, page);
2563 f2fs_put_page(page, 1);
2564 }
2565
2566
2567 if (!set->entry_cnt) {
2568 radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set);
2569 kmem_cache_free(nat_entry_set_slab, set);
2570 }
2571}
2572
2573
2574
2575
2576void f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
2577{
2578 struct f2fs_nm_info *nm_i = NM_I(sbi);
2579 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2580 struct f2fs_journal *journal = curseg->journal;
2581 struct nat_entry_set *setvec[SETVEC_SIZE];
2582 struct nat_entry_set *set, *tmp;
2583 unsigned int found;
2584 nid_t set_idx = 0;
2585 LIST_HEAD(sets);
2586
2587 if (!nm_i->dirty_nat_cnt)
2588 return;
2589
2590 down_write(&nm_i->nat_tree_lock);
2591
2592
2593
2594
2595
2596
2597 if (enabled_nat_bits(sbi, cpc) ||
2598 !__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL))
2599 remove_nats_in_journal(sbi);
2600
2601 while ((found = __gang_lookup_nat_set(nm_i,
2602 set_idx, SETVEC_SIZE, setvec))) {
2603 unsigned idx;
2604 set_idx = setvec[found - 1]->set + 1;
2605 for (idx = 0; idx < found; idx++)
2606 __adjust_nat_entry_set(setvec[idx], &sets,
2607 MAX_NAT_JENTRIES(journal));
2608 }
2609
2610
2611 list_for_each_entry_safe(set, tmp, &sets, set_list)
2612 __flush_nat_entry_set(sbi, set, cpc);
2613
2614 up_write(&nm_i->nat_tree_lock);
2615
2616}
2617
2618static int __get_nat_bitmaps(struct f2fs_sb_info *sbi)
2619{
2620 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2621 struct f2fs_nm_info *nm_i = NM_I(sbi);
2622 unsigned int nat_bits_bytes = nm_i->nat_blocks / BITS_PER_BYTE;
2623 unsigned int i;
2624 __u64 cp_ver = cur_cp_version(ckpt);
2625 block_t nat_bits_addr;
2626
2627 if (!enabled_nat_bits(sbi, NULL))
2628 return 0;
2629
2630 nm_i->nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
2631 nm_i->nat_bits = f2fs_kzalloc(sbi,
2632 nm_i->nat_bits_blocks << F2FS_BLKSIZE_BITS, GFP_KERNEL);
2633 if (!nm_i->nat_bits)
2634 return -ENOMEM;
2635
2636 nat_bits_addr = __start_cp_addr(sbi) + sbi->blocks_per_seg -
2637 nm_i->nat_bits_blocks;
2638 for (i = 0; i < nm_i->nat_bits_blocks; i++) {
2639 struct page *page = f2fs_get_meta_page(sbi, nat_bits_addr++);
2640
2641 memcpy(nm_i->nat_bits + (i << F2FS_BLKSIZE_BITS),
2642 page_address(page), F2FS_BLKSIZE);
2643 f2fs_put_page(page, 1);
2644 }
2645
2646 cp_ver |= (cur_cp_crc(ckpt) << 32);
2647 if (cpu_to_le64(cp_ver) != *(__le64 *)nm_i->nat_bits) {
2648 disable_nat_bits(sbi, true);
2649 return 0;
2650 }
2651
2652 nm_i->full_nat_bits = nm_i->nat_bits + 8;
2653 nm_i->empty_nat_bits = nm_i->full_nat_bits + nat_bits_bytes;
2654
2655 f2fs_msg(sbi->sb, KERN_NOTICE, "Found nat_bits in checkpoint");
2656 return 0;
2657}
2658
2659static inline void load_free_nid_bitmap(struct f2fs_sb_info *sbi)
2660{
2661 struct f2fs_nm_info *nm_i = NM_I(sbi);
2662 unsigned int i = 0;
2663 nid_t nid, last_nid;
2664
2665 if (!enabled_nat_bits(sbi, NULL))
2666 return;
2667
2668 for (i = 0; i < nm_i->nat_blocks; i++) {
2669 i = find_next_bit_le(nm_i->empty_nat_bits, nm_i->nat_blocks, i);
2670 if (i >= nm_i->nat_blocks)
2671 break;
2672
2673 __set_bit_le(i, nm_i->nat_block_bitmap);
2674
2675 nid = i * NAT_ENTRY_PER_BLOCK;
2676 last_nid = nid + NAT_ENTRY_PER_BLOCK;
2677
2678 spin_lock(&NM_I(sbi)->nid_list_lock);
2679 for (; nid < last_nid; nid++)
2680 update_free_nid_bitmap(sbi, nid, true, true);
2681 spin_unlock(&NM_I(sbi)->nid_list_lock);
2682 }
2683
2684 for (i = 0; i < nm_i->nat_blocks; i++) {
2685 i = find_next_bit_le(nm_i->full_nat_bits, nm_i->nat_blocks, i);
2686 if (i >= nm_i->nat_blocks)
2687 break;
2688
2689 __set_bit_le(i, nm_i->nat_block_bitmap);
2690 }
2691}
2692
2693static int init_node_manager(struct f2fs_sb_info *sbi)
2694{
2695 struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi);
2696 struct f2fs_nm_info *nm_i = NM_I(sbi);
2697 unsigned char *version_bitmap;
2698 unsigned int nat_segs;
2699 int err;
2700
2701 nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr);
2702
2703
2704 nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1;
2705 nm_i->nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg);
2706 nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks;
2707
2708
2709 nm_i->available_nids = nm_i->max_nid - sbi->total_valid_node_count -
2710 sbi->nquota_files - F2FS_RESERVED_NODE_NUM;
2711 nm_i->nid_cnt[FREE_NID] = 0;
2712 nm_i->nid_cnt[PREALLOC_NID] = 0;
2713 nm_i->nat_cnt = 0;
2714 nm_i->ram_thresh = DEF_RAM_THRESHOLD;
2715 nm_i->ra_nid_pages = DEF_RA_NID_PAGES;
2716 nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD;
2717
2718 INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC);
2719 INIT_LIST_HEAD(&nm_i->free_nid_list);
2720 INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO);
2721 INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO);
2722 INIT_LIST_HEAD(&nm_i->nat_entries);
2723
2724 mutex_init(&nm_i->build_lock);
2725 spin_lock_init(&nm_i->nid_list_lock);
2726 init_rwsem(&nm_i->nat_tree_lock);
2727
2728 nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid);
2729 nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
2730 version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP);
2731 if (!version_bitmap)
2732 return -EFAULT;
2733
2734 nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size,
2735 GFP_KERNEL);
2736 if (!nm_i->nat_bitmap)
2737 return -ENOMEM;
2738
2739 err = __get_nat_bitmaps(sbi);
2740 if (err)
2741 return err;
2742
2743#ifdef CONFIG_F2FS_CHECK_FS
2744 nm_i->nat_bitmap_mir = kmemdup(version_bitmap, nm_i->bitmap_size,
2745 GFP_KERNEL);
2746 if (!nm_i->nat_bitmap_mir)
2747 return -ENOMEM;
2748#endif
2749
2750 return 0;
2751}
2752
2753static int init_free_nid_cache(struct f2fs_sb_info *sbi)
2754{
2755 struct f2fs_nm_info *nm_i = NM_I(sbi);
2756 int i;
2757
2758 nm_i->free_nid_bitmap =
2759 f2fs_kzalloc(sbi, array_size(sizeof(unsigned char *),
2760 nm_i->nat_blocks),
2761 GFP_KERNEL);
2762 if (!nm_i->free_nid_bitmap)
2763 return -ENOMEM;
2764
2765 for (i = 0; i < nm_i->nat_blocks; i++) {
2766 nm_i->free_nid_bitmap[i] = f2fs_kvzalloc(sbi,
2767 NAT_ENTRY_BITMAP_SIZE_ALIGNED, GFP_KERNEL);
2768 if (!nm_i->free_nid_bitmap)
2769 return -ENOMEM;
2770 }
2771
2772 nm_i->nat_block_bitmap = f2fs_kvzalloc(sbi, nm_i->nat_blocks / 8,
2773 GFP_KERNEL);
2774 if (!nm_i->nat_block_bitmap)
2775 return -ENOMEM;
2776
2777 nm_i->free_nid_count =
2778 f2fs_kvzalloc(sbi, array_size(sizeof(unsigned short),
2779 nm_i->nat_blocks),
2780 GFP_KERNEL);
2781 if (!nm_i->free_nid_count)
2782 return -ENOMEM;
2783 return 0;
2784}
2785
2786int f2fs_build_node_manager(struct f2fs_sb_info *sbi)
2787{
2788 int err;
2789
2790 sbi->nm_info = f2fs_kzalloc(sbi, sizeof(struct f2fs_nm_info),
2791 GFP_KERNEL);
2792 if (!sbi->nm_info)
2793 return -ENOMEM;
2794
2795 err = init_node_manager(sbi);
2796 if (err)
2797 return err;
2798
2799 err = init_free_nid_cache(sbi);
2800 if (err)
2801 return err;
2802
2803
2804 load_free_nid_bitmap(sbi);
2805
2806 f2fs_build_free_nids(sbi, true, true);
2807 return 0;
2808}
2809
2810void f2fs_destroy_node_manager(struct f2fs_sb_info *sbi)
2811{
2812 struct f2fs_nm_info *nm_i = NM_I(sbi);
2813 struct free_nid *i, *next_i;
2814 struct nat_entry *natvec[NATVEC_SIZE];
2815 struct nat_entry_set *setvec[SETVEC_SIZE];
2816 nid_t nid = 0;
2817 unsigned int found;
2818
2819 if (!nm_i)
2820 return;
2821
2822
2823 spin_lock(&nm_i->nid_list_lock);
2824 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) {
2825 __remove_free_nid(sbi, i, FREE_NID);
2826 spin_unlock(&nm_i->nid_list_lock);
2827 kmem_cache_free(free_nid_slab, i);
2828 spin_lock(&nm_i->nid_list_lock);
2829 }
2830 f2fs_bug_on(sbi, nm_i->nid_cnt[FREE_NID]);
2831 f2fs_bug_on(sbi, nm_i->nid_cnt[PREALLOC_NID]);
2832 f2fs_bug_on(sbi, !list_empty(&nm_i->free_nid_list));
2833 spin_unlock(&nm_i->nid_list_lock);
2834
2835
2836 down_write(&nm_i->nat_tree_lock);
2837 while ((found = __gang_lookup_nat_cache(nm_i,
2838 nid, NATVEC_SIZE, natvec))) {
2839 unsigned idx;
2840
2841 nid = nat_get_nid(natvec[found - 1]) + 1;
2842 for (idx = 0; idx < found; idx++)
2843 __del_from_nat_cache(nm_i, natvec[idx]);
2844 }
2845 f2fs_bug_on(sbi, nm_i->nat_cnt);
2846
2847
2848 nid = 0;
2849 while ((found = __gang_lookup_nat_set(nm_i,
2850 nid, SETVEC_SIZE, setvec))) {
2851 unsigned idx;
2852
2853 nid = setvec[found - 1]->set + 1;
2854 for (idx = 0; idx < found; idx++) {
2855
2856 f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list));
2857 radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set);
2858 kmem_cache_free(nat_entry_set_slab, setvec[idx]);
2859 }
2860 }
2861 up_write(&nm_i->nat_tree_lock);
2862
2863 kvfree(nm_i->nat_block_bitmap);
2864 if (nm_i->free_nid_bitmap) {
2865 int i;
2866
2867 for (i = 0; i < nm_i->nat_blocks; i++)
2868 kvfree(nm_i->free_nid_bitmap[i]);
2869 kfree(nm_i->free_nid_bitmap);
2870 }
2871 kvfree(nm_i->free_nid_count);
2872
2873 kfree(nm_i->nat_bitmap);
2874 kfree(nm_i->nat_bits);
2875#ifdef CONFIG_F2FS_CHECK_FS
2876 kfree(nm_i->nat_bitmap_mir);
2877#endif
2878 sbi->nm_info = NULL;
2879 kfree(nm_i);
2880}
2881
2882int __init f2fs_create_node_manager_caches(void)
2883{
2884 nat_entry_slab = f2fs_kmem_cache_create("nat_entry",
2885 sizeof(struct nat_entry));
2886 if (!nat_entry_slab)
2887 goto fail;
2888
2889 free_nid_slab = f2fs_kmem_cache_create("free_nid",
2890 sizeof(struct free_nid));
2891 if (!free_nid_slab)
2892 goto destroy_nat_entry;
2893
2894 nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set",
2895 sizeof(struct nat_entry_set));
2896 if (!nat_entry_set_slab)
2897 goto destroy_free_nid;
2898 return 0;
2899
2900destroy_free_nid:
2901 kmem_cache_destroy(free_nid_slab);
2902destroy_nat_entry:
2903 kmem_cache_destroy(nat_entry_slab);
2904fail:
2905 return -ENOMEM;
2906}
2907
2908void f2fs_destroy_node_manager_caches(void)
2909{
2910 kmem_cache_destroy(nat_entry_set_slab);
2911 kmem_cache_destroy(free_nid_slab);
2912 kmem_cache_destroy(nat_entry_slab);
2913}
2914
