Loading...
1/*
2 * fs/f2fs/segment.h
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11#include <linux/blkdev.h>
12
13/* constant macro */
14#define NULL_SEGNO ((unsigned int)(~0))
15#define NULL_SECNO ((unsigned int)(~0))
16
17#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
18
19/* L: Logical segment # in volume, R: Relative segment # in main area */
20#define GET_L2R_SEGNO(free_i, segno) (segno - free_i->start_segno)
21#define GET_R2L_SEGNO(free_i, segno) (segno + free_i->start_segno)
22
23#define IS_DATASEG(t) (t <= CURSEG_COLD_DATA)
24#define IS_NODESEG(t) (t >= CURSEG_HOT_NODE)
25
26#define IS_CURSEG(sbi, seg) \
27 ((seg == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
28 (seg == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
29 (seg == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
30 (seg == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
31 (seg == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
32 (seg == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
33
34#define IS_CURSEC(sbi, secno) \
35 ((secno == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
36 sbi->segs_per_sec) || \
37 (secno == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
38 sbi->segs_per_sec) || \
39 (secno == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
40 sbi->segs_per_sec) || \
41 (secno == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
42 sbi->segs_per_sec) || \
43 (secno == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
44 sbi->segs_per_sec) || \
45 (secno == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
46 sbi->segs_per_sec)) \
47
48#define START_BLOCK(sbi, segno) \
49 (SM_I(sbi)->seg0_blkaddr + \
50 (GET_R2L_SEGNO(FREE_I(sbi), segno) << sbi->log_blocks_per_seg))
51#define NEXT_FREE_BLKADDR(sbi, curseg) \
52 (START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff)
53
54#define MAIN_BASE_BLOCK(sbi) (SM_I(sbi)->main_blkaddr)
55
56#define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) \
57 ((blk_addr) - SM_I(sbi)->seg0_blkaddr)
58#define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
59 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> sbi->log_blocks_per_seg)
60#define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
61 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (sbi->blocks_per_seg - 1))
62
63#define GET_SEGNO(sbi, blk_addr) \
64 (((blk_addr == NULL_ADDR) || (blk_addr == NEW_ADDR)) ? \
65 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
66 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
67#define GET_SECNO(sbi, segno) \
68 ((segno) / sbi->segs_per_sec)
69#define GET_ZONENO_FROM_SEGNO(sbi, segno) \
70 ((segno / sbi->segs_per_sec) / sbi->secs_per_zone)
71
72#define GET_SUM_BLOCK(sbi, segno) \
73 ((sbi->sm_info->ssa_blkaddr) + segno)
74
75#define GET_SUM_TYPE(footer) ((footer)->entry_type)
76#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = type)
77
78#define SIT_ENTRY_OFFSET(sit_i, segno) \
79 (segno % sit_i->sents_per_block)
80#define SIT_BLOCK_OFFSET(sit_i, segno) \
81 (segno / SIT_ENTRY_PER_BLOCK)
82#define START_SEGNO(sit_i, segno) \
83 (SIT_BLOCK_OFFSET(sit_i, segno) * SIT_ENTRY_PER_BLOCK)
84#define SIT_BLK_CNT(sbi) \
85 ((TOTAL_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
86#define f2fs_bitmap_size(nr) \
87 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
88#define TOTAL_SEGS(sbi) (SM_I(sbi)->main_segments)
89#define TOTAL_SECS(sbi) (sbi->total_sections)
90
91#define SECTOR_FROM_BLOCK(sbi, blk_addr) \
92 (((sector_t)blk_addr) << (sbi)->log_sectors_per_block)
93#define SECTOR_TO_BLOCK(sbi, sectors) \
94 (sectors >> (sbi)->log_sectors_per_block)
95#define MAX_BIO_BLOCKS(max_hw_blocks) \
96 (min((int)max_hw_blocks, BIO_MAX_PAGES))
97
98/*
99 * indicate a block allocation direction: RIGHT and LEFT.
100 * RIGHT means allocating new sections towards the end of volume.
101 * LEFT means the opposite direction.
102 */
103enum {
104 ALLOC_RIGHT = 0,
105 ALLOC_LEFT
106};
107
108/*
109 * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
110 * LFS writes data sequentially with cleaning operations.
111 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
112 */
113enum {
114 LFS = 0,
115 SSR
116};
117
118/*
119 * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
120 * GC_CB is based on cost-benefit algorithm.
121 * GC_GREEDY is based on greedy algorithm.
122 */
123enum {
124 GC_CB = 0,
125 GC_GREEDY
126};
127
128/*
129 * BG_GC means the background cleaning job.
130 * FG_GC means the on-demand cleaning job.
131 */
132enum {
133 BG_GC = 0,
134 FG_GC
135};
136
137/* for a function parameter to select a victim segment */
138struct victim_sel_policy {
139 int alloc_mode; /* LFS or SSR */
140 int gc_mode; /* GC_CB or GC_GREEDY */
141 unsigned long *dirty_segmap; /* dirty segment bitmap */
142 unsigned int max_search; /* maximum # of segments to search */
143 unsigned int offset; /* last scanned bitmap offset */
144 unsigned int ofs_unit; /* bitmap search unit */
145 unsigned int min_cost; /* minimum cost */
146 unsigned int min_segno; /* segment # having min. cost */
147};
148
149struct seg_entry {
150 unsigned short valid_blocks; /* # of valid blocks */
151 unsigned char *cur_valid_map; /* validity bitmap of blocks */
152 /*
153 * # of valid blocks and the validity bitmap stored in the the last
154 * checkpoint pack. This information is used by the SSR mode.
155 */
156 unsigned short ckpt_valid_blocks;
157 unsigned char *ckpt_valid_map;
158 unsigned char type; /* segment type like CURSEG_XXX_TYPE */
159 unsigned long long mtime; /* modification time of the segment */
160};
161
162struct sec_entry {
163 unsigned int valid_blocks; /* # of valid blocks in a section */
164};
165
166struct segment_allocation {
167 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
168};
169
170struct sit_info {
171 const struct segment_allocation *s_ops;
172
173 block_t sit_base_addr; /* start block address of SIT area */
174 block_t sit_blocks; /* # of blocks used by SIT area */
175 block_t written_valid_blocks; /* # of valid blocks in main area */
176 char *sit_bitmap; /* SIT bitmap pointer */
177 unsigned int bitmap_size; /* SIT bitmap size */
178
179 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
180 unsigned int dirty_sentries; /* # of dirty sentries */
181 unsigned int sents_per_block; /* # of SIT entries per block */
182 struct mutex sentry_lock; /* to protect SIT cache */
183 struct seg_entry *sentries; /* SIT segment-level cache */
184 struct sec_entry *sec_entries; /* SIT section-level cache */
185
186 /* for cost-benefit algorithm in cleaning procedure */
187 unsigned long long elapsed_time; /* elapsed time after mount */
188 unsigned long long mounted_time; /* mount time */
189 unsigned long long min_mtime; /* min. modification time */
190 unsigned long long max_mtime; /* max. modification time */
191};
192
193struct free_segmap_info {
194 unsigned int start_segno; /* start segment number logically */
195 unsigned int free_segments; /* # of free segments */
196 unsigned int free_sections; /* # of free sections */
197 rwlock_t segmap_lock; /* free segmap lock */
198 unsigned long *free_segmap; /* free segment bitmap */
199 unsigned long *free_secmap; /* free section bitmap */
200};
201
202/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
203enum dirty_type {
204 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
205 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
206 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
207 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
208 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
209 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
210 DIRTY, /* to count # of dirty segments */
211 PRE, /* to count # of entirely obsolete segments */
212 NR_DIRTY_TYPE
213};
214
215struct dirty_seglist_info {
216 const struct victim_selection *v_ops; /* victim selction operation */
217 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
218 struct mutex seglist_lock; /* lock for segment bitmaps */
219 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
220 unsigned long *victim_secmap; /* background GC victims */
221};
222
223/* victim selection function for cleaning and SSR */
224struct victim_selection {
225 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
226 int, int, char);
227};
228
229/* for active log information */
230struct curseg_info {
231 struct mutex curseg_mutex; /* lock for consistency */
232 struct f2fs_summary_block *sum_blk; /* cached summary block */
233 unsigned char alloc_type; /* current allocation type */
234 unsigned int segno; /* current segment number */
235 unsigned short next_blkoff; /* next block offset to write */
236 unsigned int zone; /* current zone number */
237 unsigned int next_segno; /* preallocated segment */
238};
239
240/*
241 * inline functions
242 */
243static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
244{
245 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
246}
247
248static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
249 unsigned int segno)
250{
251 struct sit_info *sit_i = SIT_I(sbi);
252 return &sit_i->sentries[segno];
253}
254
255static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
256 unsigned int segno)
257{
258 struct sit_info *sit_i = SIT_I(sbi);
259 return &sit_i->sec_entries[GET_SECNO(sbi, segno)];
260}
261
262static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
263 unsigned int segno, int section)
264{
265 /*
266 * In order to get # of valid blocks in a section instantly from many
267 * segments, f2fs manages two counting structures separately.
268 */
269 if (section > 1)
270 return get_sec_entry(sbi, segno)->valid_blocks;
271 else
272 return get_seg_entry(sbi, segno)->valid_blocks;
273}
274
275static inline void seg_info_from_raw_sit(struct seg_entry *se,
276 struct f2fs_sit_entry *rs)
277{
278 se->valid_blocks = GET_SIT_VBLOCKS(rs);
279 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
280 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
281 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
282 se->type = GET_SIT_TYPE(rs);
283 se->mtime = le64_to_cpu(rs->mtime);
284}
285
286static inline void seg_info_to_raw_sit(struct seg_entry *se,
287 struct f2fs_sit_entry *rs)
288{
289 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
290 se->valid_blocks;
291 rs->vblocks = cpu_to_le16(raw_vblocks);
292 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
293 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
294 se->ckpt_valid_blocks = se->valid_blocks;
295 rs->mtime = cpu_to_le64(se->mtime);
296}
297
298static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
299 unsigned int max, unsigned int segno)
300{
301 unsigned int ret;
302 read_lock(&free_i->segmap_lock);
303 ret = find_next_bit(free_i->free_segmap, max, segno);
304 read_unlock(&free_i->segmap_lock);
305 return ret;
306}
307
308static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
309{
310 struct free_segmap_info *free_i = FREE_I(sbi);
311 unsigned int secno = segno / sbi->segs_per_sec;
312 unsigned int start_segno = secno * sbi->segs_per_sec;
313 unsigned int next;
314
315 write_lock(&free_i->segmap_lock);
316 clear_bit(segno, free_i->free_segmap);
317 free_i->free_segments++;
318
319 next = find_next_bit(free_i->free_segmap, TOTAL_SEGS(sbi), start_segno);
320 if (next >= start_segno + sbi->segs_per_sec) {
321 clear_bit(secno, free_i->free_secmap);
322 free_i->free_sections++;
323 }
324 write_unlock(&free_i->segmap_lock);
325}
326
327static inline void __set_inuse(struct f2fs_sb_info *sbi,
328 unsigned int segno)
329{
330 struct free_segmap_info *free_i = FREE_I(sbi);
331 unsigned int secno = segno / sbi->segs_per_sec;
332 set_bit(segno, free_i->free_segmap);
333 free_i->free_segments--;
334 if (!test_and_set_bit(secno, free_i->free_secmap))
335 free_i->free_sections--;
336}
337
338static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
339 unsigned int segno)
340{
341 struct free_segmap_info *free_i = FREE_I(sbi);
342 unsigned int secno = segno / sbi->segs_per_sec;
343 unsigned int start_segno = secno * sbi->segs_per_sec;
344 unsigned int next;
345
346 write_lock(&free_i->segmap_lock);
347 if (test_and_clear_bit(segno, free_i->free_segmap)) {
348 free_i->free_segments++;
349
350 next = find_next_bit(free_i->free_segmap, TOTAL_SEGS(sbi),
351 start_segno);
352 if (next >= start_segno + sbi->segs_per_sec) {
353 if (test_and_clear_bit(secno, free_i->free_secmap))
354 free_i->free_sections++;
355 }
356 }
357 write_unlock(&free_i->segmap_lock);
358}
359
360static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
361 unsigned int segno)
362{
363 struct free_segmap_info *free_i = FREE_I(sbi);
364 unsigned int secno = segno / sbi->segs_per_sec;
365 write_lock(&free_i->segmap_lock);
366 if (!test_and_set_bit(segno, free_i->free_segmap)) {
367 free_i->free_segments--;
368 if (!test_and_set_bit(secno, free_i->free_secmap))
369 free_i->free_sections--;
370 }
371 write_unlock(&free_i->segmap_lock);
372}
373
374static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
375 void *dst_addr)
376{
377 struct sit_info *sit_i = SIT_I(sbi);
378 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
379}
380
381static inline block_t written_block_count(struct f2fs_sb_info *sbi)
382{
383 return SIT_I(sbi)->written_valid_blocks;
384}
385
386static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
387{
388 return FREE_I(sbi)->free_segments;
389}
390
391static inline int reserved_segments(struct f2fs_sb_info *sbi)
392{
393 return SM_I(sbi)->reserved_segments;
394}
395
396static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
397{
398 return FREE_I(sbi)->free_sections;
399}
400
401static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
402{
403 return DIRTY_I(sbi)->nr_dirty[PRE];
404}
405
406static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
407{
408 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
409 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
410 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
411 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
412 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
413 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
414}
415
416static inline int overprovision_segments(struct f2fs_sb_info *sbi)
417{
418 return SM_I(sbi)->ovp_segments;
419}
420
421static inline int overprovision_sections(struct f2fs_sb_info *sbi)
422{
423 return ((unsigned int) overprovision_segments(sbi)) / sbi->segs_per_sec;
424}
425
426static inline int reserved_sections(struct f2fs_sb_info *sbi)
427{
428 return ((unsigned int) reserved_segments(sbi)) / sbi->segs_per_sec;
429}
430
431static inline bool need_SSR(struct f2fs_sb_info *sbi)
432{
433 return (prefree_segments(sbi) / sbi->segs_per_sec)
434 + free_sections(sbi) < overprovision_sections(sbi);
435}
436
437static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, int freed)
438{
439 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
440 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
441
442 if (unlikely(sbi->por_doing))
443 return false;
444
445 return (free_sections(sbi) + freed) <= (node_secs + 2 * dent_secs +
446 reserved_sections(sbi));
447}
448
449static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
450{
451 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
452}
453
454static inline int utilization(struct f2fs_sb_info *sbi)
455{
456 return div_u64((u64)valid_user_blocks(sbi) * 100,
457 sbi->user_block_count);
458}
459
460/*
461 * Sometimes f2fs may be better to drop out-of-place update policy.
462 * And, users can control the policy through sysfs entries.
463 * There are five policies with triggering conditions as follows.
464 * F2FS_IPU_FORCE - all the time,
465 * F2FS_IPU_SSR - if SSR mode is activated,
466 * F2FS_IPU_UTIL - if FS utilization is over threashold,
467 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
468 * threashold,
469 * F2FS_IPUT_DISABLE - disable IPU. (=default option)
470 */
471#define DEF_MIN_IPU_UTIL 70
472
473enum {
474 F2FS_IPU_FORCE,
475 F2FS_IPU_SSR,
476 F2FS_IPU_UTIL,
477 F2FS_IPU_SSR_UTIL,
478 F2FS_IPU_DISABLE,
479};
480
481static inline bool need_inplace_update(struct inode *inode)
482{
483 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
484
485 /* IPU can be done only for the user data */
486 if (S_ISDIR(inode->i_mode))
487 return false;
488
489 switch (SM_I(sbi)->ipu_policy) {
490 case F2FS_IPU_FORCE:
491 return true;
492 case F2FS_IPU_SSR:
493 if (need_SSR(sbi))
494 return true;
495 break;
496 case F2FS_IPU_UTIL:
497 if (utilization(sbi) > SM_I(sbi)->min_ipu_util)
498 return true;
499 break;
500 case F2FS_IPU_SSR_UTIL:
501 if (need_SSR(sbi) && utilization(sbi) > SM_I(sbi)->min_ipu_util)
502 return true;
503 break;
504 case F2FS_IPU_DISABLE:
505 break;
506 }
507 return false;
508}
509
510static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
511 int type)
512{
513 struct curseg_info *curseg = CURSEG_I(sbi, type);
514 return curseg->segno;
515}
516
517static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
518 int type)
519{
520 struct curseg_info *curseg = CURSEG_I(sbi, type);
521 return curseg->alloc_type;
522}
523
524static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
525{
526 struct curseg_info *curseg = CURSEG_I(sbi, type);
527 return curseg->next_blkoff;
528}
529
530#ifdef CONFIG_F2FS_CHECK_FS
531static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
532{
533 unsigned int end_segno = SM_I(sbi)->segment_count - 1;
534 BUG_ON(segno > end_segno);
535}
536
537static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr)
538{
539 struct f2fs_sm_info *sm_info = SM_I(sbi);
540 block_t total_blks = sm_info->segment_count << sbi->log_blocks_per_seg;
541 block_t start_addr = sm_info->seg0_blkaddr;
542 block_t end_addr = start_addr + total_blks - 1;
543 BUG_ON(blk_addr < start_addr);
544 BUG_ON(blk_addr > end_addr);
545}
546
547/*
548 * Summary block is always treated as invalid block
549 */
550static inline void check_block_count(struct f2fs_sb_info *sbi,
551 int segno, struct f2fs_sit_entry *raw_sit)
552{
553 struct f2fs_sm_info *sm_info = SM_I(sbi);
554 unsigned int end_segno = sm_info->segment_count - 1;
555 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
556 int valid_blocks = 0;
557 int cur_pos = 0, next_pos;
558
559 /* check segment usage */
560 BUG_ON(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg);
561
562 /* check boundary of a given segment number */
563 BUG_ON(segno > end_segno);
564
565 /* check bitmap with valid block count */
566 do {
567 if (is_valid) {
568 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
569 sbi->blocks_per_seg,
570 cur_pos);
571 valid_blocks += next_pos - cur_pos;
572 } else
573 next_pos = find_next_bit_le(&raw_sit->valid_map,
574 sbi->blocks_per_seg,
575 cur_pos);
576 cur_pos = next_pos;
577 is_valid = !is_valid;
578 } while (cur_pos < sbi->blocks_per_seg);
579 BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks);
580}
581#else
582#define check_seg_range(sbi, segno)
583#define verify_block_addr(sbi, blk_addr)
584#define check_block_count(sbi, segno, raw_sit)
585#endif
586
587static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
588 unsigned int start)
589{
590 struct sit_info *sit_i = SIT_I(sbi);
591 unsigned int offset = SIT_BLOCK_OFFSET(sit_i, start);
592 block_t blk_addr = sit_i->sit_base_addr + offset;
593
594 check_seg_range(sbi, start);
595
596 /* calculate sit block address */
597 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
598 blk_addr += sit_i->sit_blocks;
599
600 return blk_addr;
601}
602
603static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
604 pgoff_t block_addr)
605{
606 struct sit_info *sit_i = SIT_I(sbi);
607 block_addr -= sit_i->sit_base_addr;
608 if (block_addr < sit_i->sit_blocks)
609 block_addr += sit_i->sit_blocks;
610 else
611 block_addr -= sit_i->sit_blocks;
612
613 return block_addr + sit_i->sit_base_addr;
614}
615
616static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
617{
618 unsigned int block_off = SIT_BLOCK_OFFSET(sit_i, start);
619
620 if (f2fs_test_bit(block_off, sit_i->sit_bitmap))
621 f2fs_clear_bit(block_off, sit_i->sit_bitmap);
622 else
623 f2fs_set_bit(block_off, sit_i->sit_bitmap);
624}
625
626static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
627{
628 struct sit_info *sit_i = SIT_I(sbi);
629 return sit_i->elapsed_time + CURRENT_TIME_SEC.tv_sec -
630 sit_i->mounted_time;
631}
632
633static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
634 unsigned int ofs_in_node, unsigned char version)
635{
636 sum->nid = cpu_to_le32(nid);
637 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
638 sum->version = version;
639}
640
641static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
642{
643 return __start_cp_addr(sbi) +
644 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
645}
646
647static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
648{
649 return __start_cp_addr(sbi) +
650 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
651 - (base + 1) + type;
652}
653
654static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
655{
656 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
657 return true;
658 return false;
659}
660
661static inline unsigned int max_hw_blocks(struct f2fs_sb_info *sbi)
662{
663 struct block_device *bdev = sbi->sb->s_bdev;
664 struct request_queue *q = bdev_get_queue(bdev);
665 return SECTOR_TO_BLOCK(sbi, queue_max_sectors(q));
666}
667
668/*
669 * It is very important to gather dirty pages and write at once, so that we can
670 * submit a big bio without interfering other data writes.
671 * By default, 512 pages for directory data,
672 * 512 pages (2MB) * 3 for three types of nodes, and
673 * max_bio_blocks for meta are set.
674 */
675static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
676{
677 if (type == DATA)
678 return sbi->blocks_per_seg;
679 else if (type == NODE)
680 return 3 * sbi->blocks_per_seg;
681 else if (type == META)
682 return MAX_BIO_BLOCKS(max_hw_blocks(sbi));
683 else
684 return 0;
685}
686
687/*
688 * When writing pages, it'd better align nr_to_write for segment size.
689 */
690static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
691 struct writeback_control *wbc)
692{
693 long nr_to_write, desired;
694
695 if (wbc->sync_mode != WB_SYNC_NONE)
696 return 0;
697
698 nr_to_write = wbc->nr_to_write;
699
700 if (type == DATA)
701 desired = 4096;
702 else if (type == NODE)
703 desired = 3 * max_hw_blocks(sbi);
704 else
705 desired = MAX_BIO_BLOCKS(max_hw_blocks(sbi));
706
707 wbc->nr_to_write = desired;
708 return desired - nr_to_write;
709}
1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * fs/f2fs/segment.h
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8#include <linux/blkdev.h>
9#include <linux/backing-dev.h>
10
11/* constant macro */
12#define NULL_SEGNO ((unsigned int)(~0))
13#define NULL_SECNO ((unsigned int)(~0))
14
15#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
16#define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
17
18#define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
19#define F2FS_MIN_META_SEGMENTS 8 /* SB + 2 (CP + SIT + NAT) + SSA */
20
21/* L: Logical segment # in volume, R: Relative segment # in main area */
22#define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno)
23#define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno)
24
25#define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
26#define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE && (t) <= CURSEG_COLD_NODE)
27#define SE_PAGETYPE(se) ((IS_NODESEG((se)->type) ? NODE : DATA))
28
29static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
30 unsigned short seg_type)
31{
32 f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG);
33}
34
35#define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
36#define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
37#define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
38
39#define IS_CURSEG(sbi, seg) \
40 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
41 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
42 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
43 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
44 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
45 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno) || \
46 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno) || \
47 ((seg) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno))
48
49#define IS_CURSEC(sbi, secno) \
50 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
51 (sbi)->segs_per_sec) || \
52 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
53 (sbi)->segs_per_sec) || \
54 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
55 (sbi)->segs_per_sec) || \
56 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
57 (sbi)->segs_per_sec) || \
58 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
59 (sbi)->segs_per_sec) || \
60 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
61 (sbi)->segs_per_sec) || \
62 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno / \
63 (sbi)->segs_per_sec) || \
64 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno / \
65 (sbi)->segs_per_sec))
66
67#define MAIN_BLKADDR(sbi) \
68 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
69 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
70#define SEG0_BLKADDR(sbi) \
71 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
72 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
73
74#define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
75#define MAIN_SECS(sbi) ((sbi)->total_sections)
76
77#define TOTAL_SEGS(sbi) \
78 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
79 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
80#define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
81
82#define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
83#define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
84 (sbi)->log_blocks_per_seg))
85
86#define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
87 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
88
89#define NEXT_FREE_BLKADDR(sbi, curseg) \
90 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
91
92#define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
93#define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
94 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
95#define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
96 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
97
98#define GET_SEGNO(sbi, blk_addr) \
99 ((!__is_valid_data_blkaddr(blk_addr)) ? \
100 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
101 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
102#define BLKS_PER_SEC(sbi) \
103 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
104#define CAP_BLKS_PER_SEC(sbi) \
105 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg - \
106 (sbi)->unusable_blocks_per_sec)
107#define CAP_SEGS_PER_SEC(sbi) \
108 ((sbi)->segs_per_sec - ((sbi)->unusable_blocks_per_sec >>\
109 (sbi)->log_blocks_per_seg))
110#define GET_SEC_FROM_SEG(sbi, segno) \
111 (((segno) == -1) ? -1 : (segno) / (sbi)->segs_per_sec)
112#define GET_SEG_FROM_SEC(sbi, secno) \
113 ((secno) * (sbi)->segs_per_sec)
114#define GET_ZONE_FROM_SEC(sbi, secno) \
115 (((secno) == -1) ? -1 : (secno) / (sbi)->secs_per_zone)
116#define GET_ZONE_FROM_SEG(sbi, segno) \
117 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
118
119#define GET_SUM_BLOCK(sbi, segno) \
120 ((sbi)->sm_info->ssa_blkaddr + (segno))
121
122#define GET_SUM_TYPE(footer) ((footer)->entry_type)
123#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
124
125#define SIT_ENTRY_OFFSET(sit_i, segno) \
126 ((segno) % (sit_i)->sents_per_block)
127#define SIT_BLOCK_OFFSET(segno) \
128 ((segno) / SIT_ENTRY_PER_BLOCK)
129#define START_SEGNO(segno) \
130 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
131#define SIT_BLK_CNT(sbi) \
132 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
133#define f2fs_bitmap_size(nr) \
134 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
135
136#define SECTOR_FROM_BLOCK(blk_addr) \
137 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
138#define SECTOR_TO_BLOCK(sectors) \
139 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
140
141/*
142 * indicate a block allocation direction: RIGHT and LEFT.
143 * RIGHT means allocating new sections towards the end of volume.
144 * LEFT means the opposite direction.
145 */
146enum {
147 ALLOC_RIGHT = 0,
148 ALLOC_LEFT
149};
150
151/*
152 * In the victim_sel_policy->alloc_mode, there are three block allocation modes.
153 * LFS writes data sequentially with cleaning operations.
154 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
155 * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
156 * fragmented segment which has similar aging degree.
157 */
158enum {
159 LFS = 0,
160 SSR,
161 AT_SSR,
162};
163
164/*
165 * In the victim_sel_policy->gc_mode, there are three gc, aka cleaning, modes.
166 * GC_CB is based on cost-benefit algorithm.
167 * GC_GREEDY is based on greedy algorithm.
168 * GC_AT is based on age-threshold algorithm.
169 */
170enum {
171 GC_CB = 0,
172 GC_GREEDY,
173 GC_AT,
174 ALLOC_NEXT,
175 FLUSH_DEVICE,
176 MAX_GC_POLICY,
177};
178
179/*
180 * BG_GC means the background cleaning job.
181 * FG_GC means the on-demand cleaning job.
182 */
183enum {
184 BG_GC = 0,
185 FG_GC,
186};
187
188/* for a function parameter to select a victim segment */
189struct victim_sel_policy {
190 int alloc_mode; /* LFS or SSR */
191 int gc_mode; /* GC_CB or GC_GREEDY */
192 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */
193 unsigned int max_search; /*
194 * maximum # of segments/sections
195 * to search
196 */
197 unsigned int offset; /* last scanned bitmap offset */
198 unsigned int ofs_unit; /* bitmap search unit */
199 unsigned int min_cost; /* minimum cost */
200 unsigned long long oldest_age; /* oldest age of segments having the same min cost */
201 unsigned int min_segno; /* segment # having min. cost */
202 unsigned long long age; /* mtime of GCed section*/
203 unsigned long long age_threshold;/* age threshold */
204};
205
206struct seg_entry {
207 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
208 unsigned int valid_blocks:10; /* # of valid blocks */
209 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
210 unsigned int padding:6; /* padding */
211 unsigned char *cur_valid_map; /* validity bitmap of blocks */
212#ifdef CONFIG_F2FS_CHECK_FS
213 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
214#endif
215 /*
216 * # of valid blocks and the validity bitmap stored in the last
217 * checkpoint pack. This information is used by the SSR mode.
218 */
219 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
220 unsigned char *discard_map;
221 unsigned long long mtime; /* modification time of the segment */
222};
223
224struct sec_entry {
225 unsigned int valid_blocks; /* # of valid blocks in a section */
226};
227
228#define MAX_SKIP_GC_COUNT 16
229
230struct revoke_entry {
231 struct list_head list;
232 block_t old_addr; /* for revoking when fail to commit */
233 pgoff_t index;
234};
235
236struct sit_info {
237 block_t sit_base_addr; /* start block address of SIT area */
238 block_t sit_blocks; /* # of blocks used by SIT area */
239 block_t written_valid_blocks; /* # of valid blocks in main area */
240 char *bitmap; /* all bitmaps pointer */
241 char *sit_bitmap; /* SIT bitmap pointer */
242#ifdef CONFIG_F2FS_CHECK_FS
243 char *sit_bitmap_mir; /* SIT bitmap mirror */
244
245 /* bitmap of segments to be ignored by GC in case of errors */
246 unsigned long *invalid_segmap;
247#endif
248 unsigned int bitmap_size; /* SIT bitmap size */
249
250 unsigned long *tmp_map; /* bitmap for temporal use */
251 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
252 unsigned int dirty_sentries; /* # of dirty sentries */
253 unsigned int sents_per_block; /* # of SIT entries per block */
254 struct rw_semaphore sentry_lock; /* to protect SIT cache */
255 struct seg_entry *sentries; /* SIT segment-level cache */
256 struct sec_entry *sec_entries; /* SIT section-level cache */
257
258 /* for cost-benefit algorithm in cleaning procedure */
259 unsigned long long elapsed_time; /* elapsed time after mount */
260 unsigned long long mounted_time; /* mount time */
261 unsigned long long min_mtime; /* min. modification time */
262 unsigned long long max_mtime; /* max. modification time */
263 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */
264 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */
265
266 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
267};
268
269struct free_segmap_info {
270 unsigned int start_segno; /* start segment number logically */
271 unsigned int free_segments; /* # of free segments */
272 unsigned int free_sections; /* # of free sections */
273 spinlock_t segmap_lock; /* free segmap lock */
274 unsigned long *free_segmap; /* free segment bitmap */
275 unsigned long *free_secmap; /* free section bitmap */
276};
277
278/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
279enum dirty_type {
280 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
281 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
282 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
283 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
284 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
285 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
286 DIRTY, /* to count # of dirty segments */
287 PRE, /* to count # of entirely obsolete segments */
288 NR_DIRTY_TYPE
289};
290
291struct dirty_seglist_info {
292 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
293 unsigned long *dirty_secmap;
294 struct mutex seglist_lock; /* lock for segment bitmaps */
295 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
296 unsigned long *victim_secmap; /* background GC victims */
297 unsigned long *pinned_secmap; /* pinned victims from foreground GC */
298 unsigned int pinned_secmap_cnt; /* count of victims which has pinned data */
299 bool enable_pin_section; /* enable pinning section */
300};
301
302/* for active log information */
303struct curseg_info {
304 struct mutex curseg_mutex; /* lock for consistency */
305 struct f2fs_summary_block *sum_blk; /* cached summary block */
306 struct rw_semaphore journal_rwsem; /* protect journal area */
307 struct f2fs_journal *journal; /* cached journal info */
308 unsigned char alloc_type; /* current allocation type */
309 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */
310 unsigned int segno; /* current segment number */
311 unsigned short next_blkoff; /* next block offset to write */
312 unsigned int zone; /* current zone number */
313 unsigned int next_segno; /* preallocated segment */
314 int fragment_remained_chunk; /* remained block size in a chunk for block fragmentation mode */
315 bool inited; /* indicate inmem log is inited */
316};
317
318struct sit_entry_set {
319 struct list_head set_list; /* link with all sit sets */
320 unsigned int start_segno; /* start segno of sits in set */
321 unsigned int entry_cnt; /* the # of sit entries in set */
322};
323
324/*
325 * inline functions
326 */
327static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
328{
329 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
330}
331
332static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
333 unsigned int segno)
334{
335 struct sit_info *sit_i = SIT_I(sbi);
336 return &sit_i->sentries[segno];
337}
338
339static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
340 unsigned int segno)
341{
342 struct sit_info *sit_i = SIT_I(sbi);
343 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
344}
345
346static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
347 unsigned int segno, bool use_section)
348{
349 /*
350 * In order to get # of valid blocks in a section instantly from many
351 * segments, f2fs manages two counting structures separately.
352 */
353 if (use_section && __is_large_section(sbi))
354 return get_sec_entry(sbi, segno)->valid_blocks;
355 else
356 return get_seg_entry(sbi, segno)->valid_blocks;
357}
358
359static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
360 unsigned int segno, bool use_section)
361{
362 if (use_section && __is_large_section(sbi)) {
363 unsigned int start_segno = START_SEGNO(segno);
364 unsigned int blocks = 0;
365 int i;
366
367 for (i = 0; i < sbi->segs_per_sec; i++, start_segno++) {
368 struct seg_entry *se = get_seg_entry(sbi, start_segno);
369
370 blocks += se->ckpt_valid_blocks;
371 }
372 return blocks;
373 }
374 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
375}
376
377static inline void seg_info_from_raw_sit(struct seg_entry *se,
378 struct f2fs_sit_entry *rs)
379{
380 se->valid_blocks = GET_SIT_VBLOCKS(rs);
381 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
382 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
383 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
384#ifdef CONFIG_F2FS_CHECK_FS
385 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
386#endif
387 se->type = GET_SIT_TYPE(rs);
388 se->mtime = le64_to_cpu(rs->mtime);
389}
390
391static inline void __seg_info_to_raw_sit(struct seg_entry *se,
392 struct f2fs_sit_entry *rs)
393{
394 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
395 se->valid_blocks;
396 rs->vblocks = cpu_to_le16(raw_vblocks);
397 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
398 rs->mtime = cpu_to_le64(se->mtime);
399}
400
401static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
402 struct page *page, unsigned int start)
403{
404 struct f2fs_sit_block *raw_sit;
405 struct seg_entry *se;
406 struct f2fs_sit_entry *rs;
407 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
408 (unsigned long)MAIN_SEGS(sbi));
409 int i;
410
411 raw_sit = (struct f2fs_sit_block *)page_address(page);
412 memset(raw_sit, 0, PAGE_SIZE);
413 for (i = 0; i < end - start; i++) {
414 rs = &raw_sit->entries[i];
415 se = get_seg_entry(sbi, start + i);
416 __seg_info_to_raw_sit(se, rs);
417 }
418}
419
420static inline void seg_info_to_raw_sit(struct seg_entry *se,
421 struct f2fs_sit_entry *rs)
422{
423 __seg_info_to_raw_sit(se, rs);
424
425 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
426 se->ckpt_valid_blocks = se->valid_blocks;
427}
428
429static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
430 unsigned int max, unsigned int segno)
431{
432 unsigned int ret;
433 spin_lock(&free_i->segmap_lock);
434 ret = find_next_bit(free_i->free_segmap, max, segno);
435 spin_unlock(&free_i->segmap_lock);
436 return ret;
437}
438
439static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
440{
441 struct free_segmap_info *free_i = FREE_I(sbi);
442 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
443 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
444 unsigned int next;
445 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
446
447 spin_lock(&free_i->segmap_lock);
448 clear_bit(segno, free_i->free_segmap);
449 free_i->free_segments++;
450
451 next = find_next_bit(free_i->free_segmap,
452 start_segno + sbi->segs_per_sec, start_segno);
453 if (next >= start_segno + usable_segs) {
454 clear_bit(secno, free_i->free_secmap);
455 free_i->free_sections++;
456 }
457 spin_unlock(&free_i->segmap_lock);
458}
459
460static inline void __set_inuse(struct f2fs_sb_info *sbi,
461 unsigned int segno)
462{
463 struct free_segmap_info *free_i = FREE_I(sbi);
464 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
465
466 set_bit(segno, free_i->free_segmap);
467 free_i->free_segments--;
468 if (!test_and_set_bit(secno, free_i->free_secmap))
469 free_i->free_sections--;
470}
471
472static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
473 unsigned int segno, bool inmem)
474{
475 struct free_segmap_info *free_i = FREE_I(sbi);
476 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
477 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
478 unsigned int next;
479 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
480
481 spin_lock(&free_i->segmap_lock);
482 if (test_and_clear_bit(segno, free_i->free_segmap)) {
483 free_i->free_segments++;
484
485 if (!inmem && IS_CURSEC(sbi, secno))
486 goto skip_free;
487 next = find_next_bit(free_i->free_segmap,
488 start_segno + sbi->segs_per_sec, start_segno);
489 if (next >= start_segno + usable_segs) {
490 if (test_and_clear_bit(secno, free_i->free_secmap))
491 free_i->free_sections++;
492 }
493 }
494skip_free:
495 spin_unlock(&free_i->segmap_lock);
496}
497
498static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
499 unsigned int segno)
500{
501 struct free_segmap_info *free_i = FREE_I(sbi);
502 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
503
504 spin_lock(&free_i->segmap_lock);
505 if (!test_and_set_bit(segno, free_i->free_segmap)) {
506 free_i->free_segments--;
507 if (!test_and_set_bit(secno, free_i->free_secmap))
508 free_i->free_sections--;
509 }
510 spin_unlock(&free_i->segmap_lock);
511}
512
513static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
514 void *dst_addr)
515{
516 struct sit_info *sit_i = SIT_I(sbi);
517
518#ifdef CONFIG_F2FS_CHECK_FS
519 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
520 sit_i->bitmap_size))
521 f2fs_bug_on(sbi, 1);
522#endif
523 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
524}
525
526static inline block_t written_block_count(struct f2fs_sb_info *sbi)
527{
528 return SIT_I(sbi)->written_valid_blocks;
529}
530
531static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
532{
533 return FREE_I(sbi)->free_segments;
534}
535
536static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
537{
538 return SM_I(sbi)->reserved_segments +
539 SM_I(sbi)->additional_reserved_segments;
540}
541
542static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
543{
544 return FREE_I(sbi)->free_sections;
545}
546
547static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
548{
549 return DIRTY_I(sbi)->nr_dirty[PRE];
550}
551
552static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
553{
554 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
555 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
556 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
557 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
558 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
559 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
560}
561
562static inline int overprovision_segments(struct f2fs_sb_info *sbi)
563{
564 return SM_I(sbi)->ovp_segments;
565}
566
567static inline int reserved_sections(struct f2fs_sb_info *sbi)
568{
569 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
570}
571
572static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
573 unsigned int node_blocks, unsigned int dent_blocks)
574{
575
576 unsigned int segno, left_blocks;
577 int i;
578
579 /* check current node segment */
580 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
581 segno = CURSEG_I(sbi, i)->segno;
582 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
583 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
584
585 if (node_blocks > left_blocks)
586 return false;
587 }
588
589 /* check current data segment */
590 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
591 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
592 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
593 if (dent_blocks > left_blocks)
594 return false;
595 return true;
596}
597
598/*
599 * calculate needed sections for dirty node/dentry
600 * and call has_curseg_enough_space
601 */
602static inline void __get_secs_required(struct f2fs_sb_info *sbi,
603 unsigned int *lower_p, unsigned int *upper_p, bool *curseg_p)
604{
605 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
606 get_pages(sbi, F2FS_DIRTY_DENTS) +
607 get_pages(sbi, F2FS_DIRTY_IMETA);
608 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
609 unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi);
610 unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi);
611 unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi);
612 unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi);
613
614 if (lower_p)
615 *lower_p = node_secs + dent_secs;
616 if (upper_p)
617 *upper_p = node_secs + dent_secs +
618 (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0);
619 if (curseg_p)
620 *curseg_p = has_curseg_enough_space(sbi,
621 node_blocks, dent_blocks);
622}
623
624static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
625 int freed, int needed)
626{
627 unsigned int free_secs, lower_secs, upper_secs;
628 bool curseg_space;
629
630 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
631 return false;
632
633 __get_secs_required(sbi, &lower_secs, &upper_secs, &curseg_space);
634
635 free_secs = free_sections(sbi) + freed;
636 lower_secs += needed + reserved_sections(sbi);
637 upper_secs += needed + reserved_sections(sbi);
638
639 if (free_secs > upper_secs)
640 return false;
641 else if (free_secs <= lower_secs)
642 return true;
643 return !curseg_space;
644}
645
646static inline bool has_enough_free_secs(struct f2fs_sb_info *sbi,
647 int freed, int needed)
648{
649 return !has_not_enough_free_secs(sbi, freed, needed);
650}
651
652static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
653{
654 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
655 return true;
656 if (likely(has_enough_free_secs(sbi, 0, 0)))
657 return true;
658 return false;
659}
660
661static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
662{
663 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
664}
665
666static inline int utilization(struct f2fs_sb_info *sbi)
667{
668 return div_u64((u64)valid_user_blocks(sbi) * 100,
669 sbi->user_block_count);
670}
671
672/*
673 * Sometimes f2fs may be better to drop out-of-place update policy.
674 * And, users can control the policy through sysfs entries.
675 * There are five policies with triggering conditions as follows.
676 * F2FS_IPU_FORCE - all the time,
677 * F2FS_IPU_SSR - if SSR mode is activated,
678 * F2FS_IPU_UTIL - if FS utilization is over threashold,
679 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
680 * threashold,
681 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
682 * storages. IPU will be triggered only if the # of dirty
683 * pages over min_fsync_blocks. (=default option)
684 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
685 * F2FS_IPU_NOCACHE - disable IPU bio cache.
686 * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has
687 * FI_OPU_WRITE flag.
688 * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode)
689 */
690#define DEF_MIN_IPU_UTIL 70
691#define DEF_MIN_FSYNC_BLOCKS 8
692#define DEF_MIN_HOT_BLOCKS 16
693
694#define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
695
696#define F2FS_IPU_DISABLE 0
697
698/* Modification on enum should be synchronized with ipu_mode_names array */
699enum {
700 F2FS_IPU_FORCE,
701 F2FS_IPU_SSR,
702 F2FS_IPU_UTIL,
703 F2FS_IPU_SSR_UTIL,
704 F2FS_IPU_FSYNC,
705 F2FS_IPU_ASYNC,
706 F2FS_IPU_NOCACHE,
707 F2FS_IPU_HONOR_OPU_WRITE,
708 F2FS_IPU_MAX,
709};
710
711static inline bool IS_F2FS_IPU_DISABLE(struct f2fs_sb_info *sbi)
712{
713 return SM_I(sbi)->ipu_policy == F2FS_IPU_DISABLE;
714}
715
716#define F2FS_IPU_POLICY(name) \
717static inline bool IS_##name(struct f2fs_sb_info *sbi) \
718{ \
719 return SM_I(sbi)->ipu_policy & BIT(name); \
720}
721
722F2FS_IPU_POLICY(F2FS_IPU_FORCE);
723F2FS_IPU_POLICY(F2FS_IPU_SSR);
724F2FS_IPU_POLICY(F2FS_IPU_UTIL);
725F2FS_IPU_POLICY(F2FS_IPU_SSR_UTIL);
726F2FS_IPU_POLICY(F2FS_IPU_FSYNC);
727F2FS_IPU_POLICY(F2FS_IPU_ASYNC);
728F2FS_IPU_POLICY(F2FS_IPU_NOCACHE);
729F2FS_IPU_POLICY(F2FS_IPU_HONOR_OPU_WRITE);
730
731static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
732 int type)
733{
734 struct curseg_info *curseg = CURSEG_I(sbi, type);
735 return curseg->segno;
736}
737
738static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
739 int type)
740{
741 struct curseg_info *curseg = CURSEG_I(sbi, type);
742 return curseg->alloc_type;
743}
744
745static inline bool valid_main_segno(struct f2fs_sb_info *sbi,
746 unsigned int segno)
747{
748 return segno <= (MAIN_SEGS(sbi) - 1);
749}
750
751static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
752{
753 struct f2fs_sb_info *sbi = fio->sbi;
754
755 if (__is_valid_data_blkaddr(fio->old_blkaddr))
756 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
757 META_GENERIC : DATA_GENERIC);
758 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
759 META_GENERIC : DATA_GENERIC_ENHANCE);
760}
761
762/*
763 * Summary block is always treated as an invalid block
764 */
765static inline int check_block_count(struct f2fs_sb_info *sbi,
766 int segno, struct f2fs_sit_entry *raw_sit)
767{
768 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
769 int valid_blocks = 0;
770 int cur_pos = 0, next_pos;
771 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
772
773 /* check bitmap with valid block count */
774 do {
775 if (is_valid) {
776 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
777 usable_blks_per_seg,
778 cur_pos);
779 valid_blocks += next_pos - cur_pos;
780 } else
781 next_pos = find_next_bit_le(&raw_sit->valid_map,
782 usable_blks_per_seg,
783 cur_pos);
784 cur_pos = next_pos;
785 is_valid = !is_valid;
786 } while (cur_pos < usable_blks_per_seg);
787
788 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
789 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
790 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
791 set_sbi_flag(sbi, SBI_NEED_FSCK);
792 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
793 return -EFSCORRUPTED;
794 }
795
796 if (usable_blks_per_seg < sbi->blocks_per_seg)
797 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
798 sbi->blocks_per_seg,
799 usable_blks_per_seg) != sbi->blocks_per_seg);
800
801 /* check segment usage, and check boundary of a given segment number */
802 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
803 || !valid_main_segno(sbi, segno))) {
804 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
805 GET_SIT_VBLOCKS(raw_sit), segno);
806 set_sbi_flag(sbi, SBI_NEED_FSCK);
807 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
808 return -EFSCORRUPTED;
809 }
810 return 0;
811}
812
813static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
814 unsigned int start)
815{
816 struct sit_info *sit_i = SIT_I(sbi);
817 unsigned int offset = SIT_BLOCK_OFFSET(start);
818 block_t blk_addr = sit_i->sit_base_addr + offset;
819
820 f2fs_bug_on(sbi, !valid_main_segno(sbi, start));
821
822#ifdef CONFIG_F2FS_CHECK_FS
823 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
824 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
825 f2fs_bug_on(sbi, 1);
826#endif
827
828 /* calculate sit block address */
829 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
830 blk_addr += sit_i->sit_blocks;
831
832 return blk_addr;
833}
834
835static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
836 pgoff_t block_addr)
837{
838 struct sit_info *sit_i = SIT_I(sbi);
839 block_addr -= sit_i->sit_base_addr;
840 if (block_addr < sit_i->sit_blocks)
841 block_addr += sit_i->sit_blocks;
842 else
843 block_addr -= sit_i->sit_blocks;
844
845 return block_addr + sit_i->sit_base_addr;
846}
847
848static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
849{
850 unsigned int block_off = SIT_BLOCK_OFFSET(start);
851
852 f2fs_change_bit(block_off, sit_i->sit_bitmap);
853#ifdef CONFIG_F2FS_CHECK_FS
854 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
855#endif
856}
857
858static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
859 bool base_time)
860{
861 struct sit_info *sit_i = SIT_I(sbi);
862 time64_t diff, now = ktime_get_boottime_seconds();
863
864 if (now >= sit_i->mounted_time)
865 return sit_i->elapsed_time + now - sit_i->mounted_time;
866
867 /* system time is set to the past */
868 if (!base_time) {
869 diff = sit_i->mounted_time - now;
870 if (sit_i->elapsed_time >= diff)
871 return sit_i->elapsed_time - diff;
872 return 0;
873 }
874 return sit_i->elapsed_time;
875}
876
877static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
878 unsigned int ofs_in_node, unsigned char version)
879{
880 sum->nid = cpu_to_le32(nid);
881 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
882 sum->version = version;
883}
884
885static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
886{
887 return __start_cp_addr(sbi) +
888 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
889}
890
891static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
892{
893 return __start_cp_addr(sbi) +
894 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
895 - (base + 1) + type;
896}
897
898static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
899{
900 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
901 return true;
902 return false;
903}
904
905/*
906 * It is very important to gather dirty pages and write at once, so that we can
907 * submit a big bio without interfering other data writes.
908 * By default, 512 pages for directory data,
909 * 512 pages (2MB) * 8 for nodes, and
910 * 256 pages * 8 for meta are set.
911 */
912static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
913{
914 if (sbi->sb->s_bdi->wb.dirty_exceeded)
915 return 0;
916
917 if (type == DATA)
918 return sbi->blocks_per_seg;
919 else if (type == NODE)
920 return 8 * sbi->blocks_per_seg;
921 else if (type == META)
922 return 8 * BIO_MAX_VECS;
923 else
924 return 0;
925}
926
927/*
928 * When writing pages, it'd better align nr_to_write for segment size.
929 */
930static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
931 struct writeback_control *wbc)
932{
933 long nr_to_write, desired;
934
935 if (wbc->sync_mode != WB_SYNC_NONE)
936 return 0;
937
938 nr_to_write = wbc->nr_to_write;
939 desired = BIO_MAX_VECS;
940 if (type == NODE)
941 desired <<= 1;
942
943 wbc->nr_to_write = desired;
944 return desired - nr_to_write;
945}
946
947static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
948{
949 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
950 bool wakeup = false;
951 int i;
952
953 if (force)
954 goto wake_up;
955
956 mutex_lock(&dcc->cmd_lock);
957 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
958 if (i + 1 < dcc->discard_granularity)
959 break;
960 if (!list_empty(&dcc->pend_list[i])) {
961 wakeup = true;
962 break;
963 }
964 }
965 mutex_unlock(&dcc->cmd_lock);
966 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
967 return;
968wake_up:
969 dcc->discard_wake = true;
970 wake_up_interruptible_all(&dcc->discard_wait_queue);
971}