]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - fs/f2fs/segment.h
Merge tag 'pwm/for-5.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/thierry...
[mirror_ubuntu-hirsute-kernel.git] / fs / f2fs / segment.h
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
39a53e0c
JK
3 * fs/f2fs/segment.h
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
39a53e0c 7 */
ac5d156c 8#include <linux/blkdev.h>
66114cad 9#include <linux/backing-dev.h>
ac5d156c 10
39a53e0c
JK
11/* constant macro */
12#define NULL_SEGNO ((unsigned int)(~0))
5ec4e49f 13#define NULL_SECNO ((unsigned int)(~0))
39a53e0c 14
58c41035 15#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
44a83499 16#define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
81eb8d6e 17
2040fce8
JK
18#define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
19
6224da87 20/* L: Logical segment # in volume, R: Relative segment # in main area */
68afcf2d
TK
21#define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno)
22#define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno)
39a53e0c 23
68afcf2d
TK
24#define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
25#define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE)
39a53e0c 26
a912b54d
JK
27#define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
28#define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
29#define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
30
5c773ba3 31#define IS_CURSEG(sbi, seg) \
68afcf2d
TK
32 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
33 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
34 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
35 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
36 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
37 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
39a53e0c
JK
38
39#define IS_CURSEC(sbi, secno) \
68afcf2d
TK
40 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
41 (sbi)->segs_per_sec) || \
42 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
43 (sbi)->segs_per_sec) || \
44 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
45 (sbi)->segs_per_sec) || \
46 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
47 (sbi)->segs_per_sec) || \
48 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
49 (sbi)->segs_per_sec) || \
50 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
51 (sbi)->segs_per_sec)) \
39a53e0c 52
0833721e
YH
53#define MAIN_BLKADDR(sbi) \
54 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
55 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
56#define SEG0_BLKADDR(sbi) \
57 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
58 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
7cd8558b
JK
59
60#define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
68afcf2d 61#define MAIN_SECS(sbi) ((sbi)->total_sections)
7cd8558b 62
0833721e
YH
63#define TOTAL_SEGS(sbi) \
64 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
65 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
68afcf2d 66#define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
7cd8558b
JK
67
68#define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
68afcf2d
TK
69#define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
70 (sbi)->log_blocks_per_seg))
7cd8558b
JK
71
72#define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
68afcf2d 73 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
7cd8558b 74
39a53e0c 75#define NEXT_FREE_BLKADDR(sbi, curseg) \
68afcf2d 76 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
39a53e0c 77
7cd8558b 78#define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
39a53e0c 79#define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
68afcf2d 80 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
491c0854 81#define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
68afcf2d 82 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
491c0854 83
39a53e0c 84#define GET_SEGNO(sbi, blk_addr) \
93770ab7 85 ((!__is_valid_data_blkaddr(blk_addr)) ? \
39a53e0c
JK
86 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
87 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
4ddb1a4d
JK
88#define BLKS_PER_SEC(sbi) \
89 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
90#define GET_SEC_FROM_SEG(sbi, segno) \
68afcf2d 91 ((segno) / (sbi)->segs_per_sec)
4ddb1a4d 92#define GET_SEG_FROM_SEC(sbi, secno) \
63fcf8e8 93 ((secno) * (sbi)->segs_per_sec)
4ddb1a4d
JK
94#define GET_ZONE_FROM_SEC(sbi, secno) \
95 ((secno) / (sbi)->secs_per_zone)
96#define GET_ZONE_FROM_SEG(sbi, segno) \
97 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
39a53e0c
JK
98
99#define GET_SUM_BLOCK(sbi, segno) \
68afcf2d 100 ((sbi)->sm_info->ssa_blkaddr + (segno))
39a53e0c
JK
101
102#define GET_SUM_TYPE(footer) ((footer)->entry_type)
68afcf2d 103#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
39a53e0c
JK
104
105#define SIT_ENTRY_OFFSET(sit_i, segno) \
68afcf2d 106 ((segno) % (sit_i)->sents_per_block)
d3a14afd 107#define SIT_BLOCK_OFFSET(segno) \
68afcf2d 108 ((segno) / SIT_ENTRY_PER_BLOCK)
d3a14afd
CY
109#define START_SEGNO(segno) \
110 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
74de593a 111#define SIT_BLK_CNT(sbi) \
f91108b8 112 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
39a53e0c
JK
113#define f2fs_bitmap_size(nr) \
114 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
39a53e0c 115
55cf9cb6
CY
116#define SECTOR_FROM_BLOCK(blk_addr) \
117 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
118#define SECTOR_TO_BLOCK(sectors) \
68afcf2d 119 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
3cd8a239 120
39a53e0c
JK
121/*
122 * indicate a block allocation direction: RIGHT and LEFT.
123 * RIGHT means allocating new sections towards the end of volume.
124 * LEFT means the opposite direction.
125 */
126enum {
127 ALLOC_RIGHT = 0,
128 ALLOC_LEFT
129};
130
131/*
132 * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
133 * LFS writes data sequentially with cleaning operations.
134 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
135 */
136enum {
137 LFS = 0,
138 SSR
139};
140
141/*
142 * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
143 * GC_CB is based on cost-benefit algorithm.
144 * GC_GREEDY is based on greedy algorithm.
145 */
146enum {
147 GC_CB = 0,
e066b83c
JK
148 GC_GREEDY,
149 ALLOC_NEXT,
150 FLUSH_DEVICE,
151 MAX_GC_POLICY,
39a53e0c
JK
152};
153
154/*
155 * BG_GC means the background cleaning job.
156 * FG_GC means the on-demand cleaning job.
6aefd93b 157 * FORCE_FG_GC means on-demand cleaning job in background.
39a53e0c
JK
158 */
159enum {
160 BG_GC = 0,
6aefd93b
JK
161 FG_GC,
162 FORCE_FG_GC,
39a53e0c
JK
163};
164
165/* for a function parameter to select a victim segment */
166struct victim_sel_policy {
167 int alloc_mode; /* LFS or SSR */
168 int gc_mode; /* GC_CB or GC_GREEDY */
169 unsigned long *dirty_segmap; /* dirty segment bitmap */
a26b7c8a 170 unsigned int max_search; /* maximum # of segments to search */
39a53e0c
JK
171 unsigned int offset; /* last scanned bitmap offset */
172 unsigned int ofs_unit; /* bitmap search unit */
173 unsigned int min_cost; /* minimum cost */
174 unsigned int min_segno; /* segment # having min. cost */
175};
176
177struct seg_entry {
f51b4ce6
CY
178 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
179 unsigned int valid_blocks:10; /* # of valid blocks */
180 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
181 unsigned int padding:6; /* padding */
39a53e0c 182 unsigned char *cur_valid_map; /* validity bitmap of blocks */
355e7891
CY
183#ifdef CONFIG_F2FS_CHECK_FS
184 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
185#endif
39a53e0c
JK
186 /*
187 * # of valid blocks and the validity bitmap stored in the the last
188 * checkpoint pack. This information is used by the SSR mode.
189 */
f51b4ce6 190 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
a66cdd98 191 unsigned char *discard_map;
39a53e0c
JK
192 unsigned long long mtime; /* modification time of the segment */
193};
194
195struct sec_entry {
196 unsigned int valid_blocks; /* # of valid blocks in a section */
197};
198
199struct segment_allocation {
200 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
201};
202
6f8d4455 203#define MAX_SKIP_GC_COUNT 16
2ef79ecb 204
88b88a66
JK
205struct inmem_pages {
206 struct list_head list;
207 struct page *page;
28bc106b 208 block_t old_addr; /* for revoking when fail to commit */
88b88a66
JK
209};
210
39a53e0c
JK
211struct sit_info {
212 const struct segment_allocation *s_ops;
213
214 block_t sit_base_addr; /* start block address of SIT area */
215 block_t sit_blocks; /* # of blocks used by SIT area */
216 block_t written_valid_blocks; /* # of valid blocks in main area */
2fde3dd1 217 char *bitmap; /* all bitmaps pointer */
39a53e0c 218 char *sit_bitmap; /* SIT bitmap pointer */
ae27d62e
CY
219#ifdef CONFIG_F2FS_CHECK_FS
220 char *sit_bitmap_mir; /* SIT bitmap mirror */
bbf9f7d9
ST
221
222 /* bitmap of segments to be ignored by GC in case of errors */
223 unsigned long *invalid_segmap;
ae27d62e 224#endif
39a53e0c
JK
225 unsigned int bitmap_size; /* SIT bitmap size */
226
60a3b782 227 unsigned long *tmp_map; /* bitmap for temporal use */
39a53e0c
JK
228 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
229 unsigned int dirty_sentries; /* # of dirty sentries */
230 unsigned int sents_per_block; /* # of SIT entries per block */
3d26fa6b 231 struct rw_semaphore sentry_lock; /* to protect SIT cache */
39a53e0c
JK
232 struct seg_entry *sentries; /* SIT segment-level cache */
233 struct sec_entry *sec_entries; /* SIT section-level cache */
234
235 /* for cost-benefit algorithm in cleaning procedure */
236 unsigned long long elapsed_time; /* elapsed time after mount */
237 unsigned long long mounted_time; /* mount time */
238 unsigned long long min_mtime; /* min. modification time */
239 unsigned long long max_mtime; /* max. modification time */
e066b83c
JK
240
241 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
39a53e0c
JK
242};
243
244struct free_segmap_info {
245 unsigned int start_segno; /* start segment number logically */
246 unsigned int free_segments; /* # of free segments */
247 unsigned int free_sections; /* # of free sections */
1a118ccf 248 spinlock_t segmap_lock; /* free segmap lock */
39a53e0c
JK
249 unsigned long *free_segmap; /* free segment bitmap */
250 unsigned long *free_secmap; /* free section bitmap */
251};
252
253/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
254enum dirty_type {
255 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
256 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
257 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
258 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
259 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
260 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
261 DIRTY, /* to count # of dirty segments */
262 PRE, /* to count # of entirely obsolete segments */
263 NR_DIRTY_TYPE
264};
265
266struct dirty_seglist_info {
267 const struct victim_selection *v_ops; /* victim selction operation */
268 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
269 struct mutex seglist_lock; /* lock for segment bitmaps */
270 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
5ec4e49f 271 unsigned long *victim_secmap; /* background GC victims */
39a53e0c
JK
272};
273
274/* victim selection function for cleaning and SSR */
275struct victim_selection {
276 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
277 int, int, char);
278};
279
280/* for active log information */
281struct curseg_info {
282 struct mutex curseg_mutex; /* lock for consistency */
283 struct f2fs_summary_block *sum_blk; /* cached summary block */
b7ad7512
CY
284 struct rw_semaphore journal_rwsem; /* protect journal area */
285 struct f2fs_journal *journal; /* cached journal info */
39a53e0c
JK
286 unsigned char alloc_type; /* current allocation type */
287 unsigned int segno; /* current segment number */
288 unsigned short next_blkoff; /* next block offset to write */
289 unsigned int zone; /* current zone number */
290 unsigned int next_segno; /* preallocated segment */
291};
292
184a5cd2
CY
293struct sit_entry_set {
294 struct list_head set_list; /* link with all sit sets */
295 unsigned int start_segno; /* start segno of sits in set */
296 unsigned int entry_cnt; /* the # of sit entries in set */
297};
298
39a53e0c
JK
299/*
300 * inline functions
301 */
302static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
303{
f5a53edc
JK
304 if (type == CURSEG_COLD_DATA_PINNED)
305 type = CURSEG_COLD_DATA;
39a53e0c
JK
306 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
307}
308
309static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
310 unsigned int segno)
311{
312 struct sit_info *sit_i = SIT_I(sbi);
313 return &sit_i->sentries[segno];
314}
315
316static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
317 unsigned int segno)
318{
319 struct sit_info *sit_i = SIT_I(sbi);
4ddb1a4d 320 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
39a53e0c
JK
321}
322
323static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
302bd348 324 unsigned int segno, bool use_section)
39a53e0c
JK
325{
326 /*
327 * In order to get # of valid blocks in a section instantly from many
328 * segments, f2fs manages two counting structures separately.
329 */
2c70c5e3 330 if (use_section && __is_large_section(sbi))
39a53e0c
JK
331 return get_sec_entry(sbi, segno)->valid_blocks;
332 else
333 return get_seg_entry(sbi, segno)->valid_blocks;
334}
335
4354994f
DR
336static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
337 unsigned int segno)
338{
339 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
340}
341
39a53e0c
JK
342static inline void seg_info_from_raw_sit(struct seg_entry *se,
343 struct f2fs_sit_entry *rs)
344{
345 se->valid_blocks = GET_SIT_VBLOCKS(rs);
346 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
347 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
348 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
355e7891
CY
349#ifdef CONFIG_F2FS_CHECK_FS
350 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
351#endif
39a53e0c
JK
352 se->type = GET_SIT_TYPE(rs);
353 se->mtime = le64_to_cpu(rs->mtime);
354}
355
068c3cd8 356static inline void __seg_info_to_raw_sit(struct seg_entry *se,
39a53e0c
JK
357 struct f2fs_sit_entry *rs)
358{
359 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
360 se->valid_blocks;
361 rs->vblocks = cpu_to_le16(raw_vblocks);
362 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
068c3cd8
YH
363 rs->mtime = cpu_to_le64(se->mtime);
364}
365
366static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
367 struct page *page, unsigned int start)
368{
369 struct f2fs_sit_block *raw_sit;
370 struct seg_entry *se;
371 struct f2fs_sit_entry *rs;
372 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
373 (unsigned long)MAIN_SEGS(sbi));
374 int i;
375
376 raw_sit = (struct f2fs_sit_block *)page_address(page);
81114baa 377 memset(raw_sit, 0, PAGE_SIZE);
068c3cd8
YH
378 for (i = 0; i < end - start; i++) {
379 rs = &raw_sit->entries[i];
380 se = get_seg_entry(sbi, start + i);
381 __seg_info_to_raw_sit(se, rs);
382 }
383}
384
385static inline void seg_info_to_raw_sit(struct seg_entry *se,
386 struct f2fs_sit_entry *rs)
387{
388 __seg_info_to_raw_sit(se, rs);
389
39a53e0c
JK
390 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
391 se->ckpt_valid_blocks = se->valid_blocks;
39a53e0c
JK
392}
393
394static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
395 unsigned int max, unsigned int segno)
396{
397 unsigned int ret;
1a118ccf 398 spin_lock(&free_i->segmap_lock);
39a53e0c 399 ret = find_next_bit(free_i->free_segmap, max, segno);
1a118ccf 400 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK
401 return ret;
402}
403
404static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
405{
406 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK
407 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
408 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
39a53e0c
JK
409 unsigned int next;
410
1a118ccf 411 spin_lock(&free_i->segmap_lock);
39a53e0c
JK
412 clear_bit(segno, free_i->free_segmap);
413 free_i->free_segments++;
414
7fd97019
WL
415 next = find_next_bit(free_i->free_segmap,
416 start_segno + sbi->segs_per_sec, start_segno);
39a53e0c
JK
417 if (next >= start_segno + sbi->segs_per_sec) {
418 clear_bit(secno, free_i->free_secmap);
419 free_i->free_sections++;
420 }
1a118ccf 421 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK
422}
423
424static inline void __set_inuse(struct f2fs_sb_info *sbi,
425 unsigned int segno)
426{
427 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK
428 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
429
39a53e0c
JK
430 set_bit(segno, free_i->free_segmap);
431 free_i->free_segments--;
432 if (!test_and_set_bit(secno, free_i->free_secmap))
433 free_i->free_sections--;
434}
435
436static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
437 unsigned int segno)
438{
439 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK
440 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
441 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
39a53e0c
JK
442 unsigned int next;
443
1a118ccf 444 spin_lock(&free_i->segmap_lock);
39a53e0c
JK
445 if (test_and_clear_bit(segno, free_i->free_segmap)) {
446 free_i->free_segments++;
447
3611ce99
YS
448 if (IS_CURSEC(sbi, secno))
449 goto skip_free;
f1121ab0
CY
450 next = find_next_bit(free_i->free_segmap,
451 start_segno + sbi->segs_per_sec, start_segno);
39a53e0c
JK
452 if (next >= start_segno + sbi->segs_per_sec) {
453 if (test_and_clear_bit(secno, free_i->free_secmap))
454 free_i->free_sections++;
455 }
456 }
3611ce99 457skip_free:
1a118ccf 458 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK
459}
460
461static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
462 unsigned int segno)
463{
464 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK
465 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
466
1a118ccf 467 spin_lock(&free_i->segmap_lock);
39a53e0c
JK
468 if (!test_and_set_bit(segno, free_i->free_segmap)) {
469 free_i->free_segments--;
470 if (!test_and_set_bit(secno, free_i->free_secmap))
471 free_i->free_sections--;
472 }
1a118ccf 473 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK
474}
475
476static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
477 void *dst_addr)
478{
479 struct sit_info *sit_i = SIT_I(sbi);
ae27d62e
CY
480
481#ifdef CONFIG_F2FS_CHECK_FS
482 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
483 sit_i->bitmap_size))
484 f2fs_bug_on(sbi, 1);
485#endif
39a53e0c
JK
486 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
487}
488
489static inline block_t written_block_count(struct f2fs_sb_info *sbi)
490{
8b8343fa 491 return SIT_I(sbi)->written_valid_blocks;
39a53e0c
JK
492}
493
494static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
495{
8b8343fa 496 return FREE_I(sbi)->free_segments;
39a53e0c
JK
497}
498
499static inline int reserved_segments(struct f2fs_sb_info *sbi)
500{
501 return SM_I(sbi)->reserved_segments;
502}
503
504static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
505{
8b8343fa 506 return FREE_I(sbi)->free_sections;
39a53e0c
JK
507}
508
509static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
510{
511 return DIRTY_I(sbi)->nr_dirty[PRE];
512}
513
514static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
515{
516 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
517 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
518 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
519 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
520 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
521 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
522}
523
524static inline int overprovision_segments(struct f2fs_sb_info *sbi)
525{
526 return SM_I(sbi)->ovp_segments;
527}
528
39a53e0c
JK
529static inline int reserved_sections(struct f2fs_sb_info *sbi)
530{
4ddb1a4d 531 return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
39a53e0c
JK
532}
533
bf34c93d
CY
534static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
535{
536 unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
537 get_pages(sbi, F2FS_DIRTY_DENTS);
538 unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
539 unsigned int segno, left_blocks;
540 int i;
541
542 /* check current node segment */
543 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
544 segno = CURSEG_I(sbi, i)->segno;
545 left_blocks = sbi->blocks_per_seg -
546 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
547
548 if (node_blocks > left_blocks)
549 return false;
550 }
551
552 /* check current data segment */
553 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
554 left_blocks = sbi->blocks_per_seg -
555 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
556 if (dent_blocks > left_blocks)
557 return false;
558 return true;
559}
560
7f3037a5
JK
561static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
562 int freed, int needed)
39a53e0c 563{
5ac206cf
NJ
564 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
565 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
b9610bdf 566 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
0f18b462 567
caf0047e 568 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
029cd28c
JK
569 return false;
570
bf34c93d
CY
571 if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
572 has_curseg_enough_space(sbi))
573 return false;
7f3037a5 574 return (free_sections(sbi) + freed) <=
b9610bdf
JK
575 (node_secs + 2 * dent_secs + imeta_secs +
576 reserved_sections(sbi) + needed);
39a53e0c
JK
577}
578
00e09c0b 579static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
4354994f
DR
580{
581 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
00e09c0b 582 return true;
4354994f 583 if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
00e09c0b
CY
584 return true;
585 return false;
4354994f
DR
586}
587
81eb8d6e
JK
588static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
589{
6c311ec6 590 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
81eb8d6e
JK
591}
592
39a53e0c
JK
593static inline int utilization(struct f2fs_sb_info *sbi)
594{
6c311ec6
CF
595 return div_u64((u64)valid_user_blocks(sbi) * 100,
596 sbi->user_block_count);
39a53e0c
JK
597}
598
599/*
600 * Sometimes f2fs may be better to drop out-of-place update policy.
216fbd64
JK
601 * And, users can control the policy through sysfs entries.
602 * There are five policies with triggering conditions as follows.
603 * F2FS_IPU_FORCE - all the time,
604 * F2FS_IPU_SSR - if SSR mode is activated,
605 * F2FS_IPU_UTIL - if FS utilization is over threashold,
606 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
607 * threashold,
c1ce1b02
JK
608 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
609 * storages. IPU will be triggered only if the # of dirty
d7b0a23d
JK
610 * pages over min_fsync_blocks. (=default option)
611 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
0e7f4197 612 * F2FS_IPU_NOCACHE - disable IPU bio cache.
d7b0a23d 613 * F2FS_IPUT_DISABLE - disable IPU. (=default option in LFS mode)
39a53e0c 614 */
216fbd64 615#define DEF_MIN_IPU_UTIL 70
c1ce1b02 616#define DEF_MIN_FSYNC_BLOCKS 8
ef095d19 617#define DEF_MIN_HOT_BLOCKS 16
216fbd64 618
84b89e5d
JK
619#define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
620
216fbd64
JK
621enum {
622 F2FS_IPU_FORCE,
623 F2FS_IPU_SSR,
624 F2FS_IPU_UTIL,
625 F2FS_IPU_SSR_UTIL,
c1ce1b02 626 F2FS_IPU_FSYNC,
04485987 627 F2FS_IPU_ASYNC,
0e7f4197 628 F2FS_IPU_NOCACHE,
216fbd64
JK
629};
630
39a53e0c
JK
631static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
632 int type)
633{
634 struct curseg_info *curseg = CURSEG_I(sbi, type);
635 return curseg->segno;
636}
637
638static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
639 int type)
640{
641 struct curseg_info *curseg = CURSEG_I(sbi, type);
642 return curseg->alloc_type;
643}
644
645static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
646{
647 struct curseg_info *curseg = CURSEG_I(sbi, type);
648 return curseg->next_blkoff;
649}
650
651static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
652{
7a04f64d 653 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
39a53e0c
JK
654}
655
93770ab7 656static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
39a53e0c 657{
0833721e
YH
658 struct f2fs_sb_info *sbi = fio->sbi;
659
93770ab7
CY
660 if (__is_valid_data_blkaddr(fio->old_blkaddr))
661 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
662 META_GENERIC : DATA_GENERIC);
663 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
664 META_GENERIC : DATA_GENERIC_ENHANCE);
39a53e0c
JK
665}
666
667/*
e1c42045 668 * Summary block is always treated as an invalid block
39a53e0c 669 */
c39a1b34 670static inline int check_block_count(struct f2fs_sb_info *sbi,
39a53e0c
JK
671 int segno, struct f2fs_sit_entry *raw_sit)
672{
44c60bf2 673 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
39a53e0c 674 int valid_blocks = 0;
44c60bf2 675 int cur_pos = 0, next_pos;
39a53e0c 676
39a53e0c 677 /* check bitmap with valid block count */
44c60bf2
CY
678 do {
679 if (is_valid) {
680 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
681 sbi->blocks_per_seg,
682 cur_pos);
683 valid_blocks += next_pos - cur_pos;
684 } else
685 next_pos = find_next_bit_le(&raw_sit->valid_map,
686 sbi->blocks_per_seg,
687 cur_pos);
688 cur_pos = next_pos;
689 is_valid = !is_valid;
690 } while (cur_pos < sbi->blocks_per_seg);
c39a1b34
JK
691
692 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
dcbb4c10
JP
693 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
694 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
c39a1b34 695 set_sbi_flag(sbi, SBI_NEED_FSCK);
10f966bb 696 return -EFSCORRUPTED;
c39a1b34 697 }
e95bcdb2 698
4c278394 699 /* check segment usage, and check boundary of a given segment number */
c39a1b34
JK
700 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
701 || segno > TOTAL_SEGS(sbi) - 1)) {
dcbb4c10
JP
702 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
703 GET_SIT_VBLOCKS(raw_sit), segno);
c39a1b34 704 set_sbi_flag(sbi, SBI_NEED_FSCK);
10f966bb 705 return -EFSCORRUPTED;
c39a1b34
JK
706 }
707 return 0;
7a04f64d 708}
39a53e0c
JK
709
710static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
711 unsigned int start)
712{
713 struct sit_info *sit_i = SIT_I(sbi);
d3a14afd 714 unsigned int offset = SIT_BLOCK_OFFSET(start);
39a53e0c
JK
715 block_t blk_addr = sit_i->sit_base_addr + offset;
716
717 check_seg_range(sbi, start);
718
ae27d62e
CY
719#ifdef CONFIG_F2FS_CHECK_FS
720 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
721 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
722 f2fs_bug_on(sbi, 1);
723#endif
724
39a53e0c
JK
725 /* calculate sit block address */
726 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
727 blk_addr += sit_i->sit_blocks;
728
729 return blk_addr;
730}
731
732static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
733 pgoff_t block_addr)
734{
735 struct sit_info *sit_i = SIT_I(sbi);
736 block_addr -= sit_i->sit_base_addr;
737 if (block_addr < sit_i->sit_blocks)
738 block_addr += sit_i->sit_blocks;
739 else
740 block_addr -= sit_i->sit_blocks;
741
742 return block_addr + sit_i->sit_base_addr;
743}
744
745static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
746{
d3a14afd 747 unsigned int block_off = SIT_BLOCK_OFFSET(start);
39a53e0c 748
c6ac4c0e 749 f2fs_change_bit(block_off, sit_i->sit_bitmap);
ae27d62e
CY
750#ifdef CONFIG_F2FS_CHECK_FS
751 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
752#endif
39a53e0c
JK
753}
754
a1f72ac2
CY
755static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
756 bool base_time)
39a53e0c
JK
757{
758 struct sit_info *sit_i = SIT_I(sbi);
a1f72ac2 759 time64_t diff, now = ktime_get_real_seconds();
48fbfe50 760
a1f72ac2
CY
761 if (now >= sit_i->mounted_time)
762 return sit_i->elapsed_time + now - sit_i->mounted_time;
763
764 /* system time is set to the past */
765 if (!base_time) {
766 diff = sit_i->mounted_time - now;
767 if (sit_i->elapsed_time >= diff)
768 return sit_i->elapsed_time - diff;
769 return 0;
770 }
771 return sit_i->elapsed_time;
39a53e0c
JK
772}
773
774static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
775 unsigned int ofs_in_node, unsigned char version)
776{
777 sum->nid = cpu_to_le32(nid);
778 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
779 sum->version = version;
780}
781
782static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
783{
784 return __start_cp_addr(sbi) +
785 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
786}
787
788static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
789{
790 return __start_cp_addr(sbi) +
791 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
792 - (base + 1) + type;
793}
5ec4e49f
JK
794
795static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
796{
797 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
798 return true;
799 return false;
800}
ac5d156c 801
87d6f890
JK
802/*
803 * It is very important to gather dirty pages and write at once, so that we can
804 * submit a big bio without interfering other data writes.
805 * By default, 512 pages for directory data,
727ebb09
KM
806 * 512 pages (2MB) * 8 for nodes, and
807 * 256 pages * 8 for meta are set.
87d6f890
JK
808 */
809static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
810{
a88a341a 811 if (sbi->sb->s_bdi->wb.dirty_exceeded)
510184c8
JK
812 return 0;
813
a1257023
JK
814 if (type == DATA)
815 return sbi->blocks_per_seg;
816 else if (type == NODE)
2c237eba 817 return 8 * sbi->blocks_per_seg;
87d6f890 818 else if (type == META)
664ba972 819 return 8 * BIO_MAX_PAGES;
87d6f890
JK
820 else
821 return 0;
822}
50c8cdb3
JK
823
824/*
825 * When writing pages, it'd better align nr_to_write for segment size.
826 */
827static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
828 struct writeback_control *wbc)
829{
830 long nr_to_write, desired;
831
832 if (wbc->sync_mode != WB_SYNC_NONE)
833 return 0;
834
835 nr_to_write = wbc->nr_to_write;
664ba972 836 desired = BIO_MAX_PAGES;
28ea6162 837 if (type == NODE)
664ba972 838 desired <<= 1;
50c8cdb3
JK
839
840 wbc->nr_to_write = desired;
841 return desired - nr_to_write;
842}
01983c71
JK
843
844static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
845{
846 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
847 bool wakeup = false;
848 int i;
849
850 if (force)
851 goto wake_up;
852
853 mutex_lock(&dcc->cmd_lock);
78997b56
CY
854 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
855 if (i + 1 < dcc->discard_granularity)
856 break;
01983c71
JK
857 if (!list_empty(&dcc->pend_list[i])) {
858 wakeup = true;
859 break;
860 }
861 }
862 mutex_unlock(&dcc->cmd_lock);
b460866d 863 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
01983c71
JK
864 return;
865wake_up:
866 dcc->discard_wake = 1;
867 wake_up_interruptible_all(&dcc->discard_wait_queue);
868}