xref: /linux/fs/f2fs/segment.h (revision 65e9fb081877a18c432c6ff344937b7277c044b5)
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 
29 static 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 GET_SEC_FROM_SEG(sbi, segno)				\
108 	(((segno) == -1) ? -1: (segno) / (sbi)->segs_per_sec)
109 #define GET_SEG_FROM_SEC(sbi, secno)				\
110 	((secno) * (sbi)->segs_per_sec)
111 #define GET_ZONE_FROM_SEC(sbi, secno)				\
112 	(((secno) == -1) ? -1: (secno) / (sbi)->secs_per_zone)
113 #define GET_ZONE_FROM_SEG(sbi, segno)				\
114 	GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
115 
116 #define GET_SUM_BLOCK(sbi, segno)				\
117 	((sbi)->sm_info->ssa_blkaddr + (segno))
118 
119 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
120 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
121 
122 #define SIT_ENTRY_OFFSET(sit_i, segno)					\
123 	((segno) % (sit_i)->sents_per_block)
124 #define SIT_BLOCK_OFFSET(segno)					\
125 	((segno) / SIT_ENTRY_PER_BLOCK)
126 #define	START_SEGNO(segno)		\
127 	(SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
128 #define SIT_BLK_CNT(sbi)			\
129 	DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
130 #define f2fs_bitmap_size(nr)			\
131 	(BITS_TO_LONGS(nr) * sizeof(unsigned long))
132 
133 #define SECTOR_FROM_BLOCK(blk_addr)					\
134 	(((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
135 #define SECTOR_TO_BLOCK(sectors)					\
136 	((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
137 
138 /*
139  * indicate a block allocation direction: RIGHT and LEFT.
140  * RIGHT means allocating new sections towards the end of volume.
141  * LEFT means the opposite direction.
142  */
143 enum {
144 	ALLOC_RIGHT = 0,
145 	ALLOC_LEFT
146 };
147 
148 /*
149  * In the victim_sel_policy->alloc_mode, there are three block allocation modes.
150  * LFS writes data sequentially with cleaning operations.
151  * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
152  * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
153  * fragmented segment which has similar aging degree.
154  */
155 enum {
156 	LFS = 0,
157 	SSR,
158 	AT_SSR,
159 };
160 
161 /*
162  * In the victim_sel_policy->gc_mode, there are three gc, aka cleaning, modes.
163  * GC_CB is based on cost-benefit algorithm.
164  * GC_GREEDY is based on greedy algorithm.
165  * GC_AT is based on age-threshold algorithm.
166  */
167 enum {
168 	GC_CB = 0,
169 	GC_GREEDY,
170 	GC_AT,
171 	ALLOC_NEXT,
172 	FLUSH_DEVICE,
173 	MAX_GC_POLICY,
174 };
175 
176 /*
177  * BG_GC means the background cleaning job.
178  * FG_GC means the on-demand cleaning job.
179  */
180 enum {
181 	BG_GC = 0,
182 	FG_GC,
183 };
184 
185 /* for a function parameter to select a victim segment */
186 struct victim_sel_policy {
187 	int alloc_mode;			/* LFS or SSR */
188 	int gc_mode;			/* GC_CB or GC_GREEDY */
189 	unsigned long *dirty_bitmap;	/* dirty segment/section bitmap */
190 	unsigned int max_search;	/*
191 					 * maximum # of segments/sections
192 					 * to search
193 					 */
194 	unsigned int offset;		/* last scanned bitmap offset */
195 	unsigned int ofs_unit;		/* bitmap search unit */
196 	unsigned int min_cost;		/* minimum cost */
197 	unsigned long long oldest_age;	/* oldest age of segments having the same min cost */
198 	unsigned int min_segno;		/* segment # having min. cost */
199 	unsigned long long age;		/* mtime of GCed section*/
200 	unsigned long long age_threshold;/* age threshold */
201 };
202 
203 struct seg_entry {
204 	unsigned int type:6;		/* segment type like CURSEG_XXX_TYPE */
205 	unsigned int valid_blocks:10;	/* # of valid blocks */
206 	unsigned int ckpt_valid_blocks:10;	/* # of valid blocks last cp */
207 	unsigned int padding:6;		/* padding */
208 	unsigned char *cur_valid_map;	/* validity bitmap of blocks */
209 #ifdef CONFIG_F2FS_CHECK_FS
210 	unsigned char *cur_valid_map_mir;	/* mirror of current valid bitmap */
211 #endif
212 	/*
213 	 * # of valid blocks and the validity bitmap stored in the last
214 	 * checkpoint pack. This information is used by the SSR mode.
215 	 */
216 	unsigned char *ckpt_valid_map;	/* validity bitmap of blocks last cp */
217 	unsigned char *discard_map;
218 	unsigned long long mtime;	/* modification time of the segment */
219 };
220 
221 struct sec_entry {
222 	unsigned int valid_blocks;	/* # of valid blocks in a section */
223 };
224 
225 struct segment_allocation {
226 	void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
227 };
228 
229 #define MAX_SKIP_GC_COUNT			16
230 
231 struct revoke_entry {
232 	struct list_head list;
233 	block_t old_addr;		/* for revoking when fail to commit */
234 	pgoff_t index;
235 };
236 
237 struct sit_info {
238 	const struct segment_allocation *s_ops;
239 
240 	block_t sit_base_addr;		/* start block address of SIT area */
241 	block_t sit_blocks;		/* # of blocks used by SIT area */
242 	block_t written_valid_blocks;	/* # of valid blocks in main area */
243 	char *bitmap;			/* all bitmaps pointer */
244 	char *sit_bitmap;		/* SIT bitmap pointer */
245 #ifdef CONFIG_F2FS_CHECK_FS
246 	char *sit_bitmap_mir;		/* SIT bitmap mirror */
247 
248 	/* bitmap of segments to be ignored by GC in case of errors */
249 	unsigned long *invalid_segmap;
250 #endif
251 	unsigned int bitmap_size;	/* SIT bitmap size */
252 
253 	unsigned long *tmp_map;			/* bitmap for temporal use */
254 	unsigned long *dirty_sentries_bitmap;	/* bitmap for dirty sentries */
255 	unsigned int dirty_sentries;		/* # of dirty sentries */
256 	unsigned int sents_per_block;		/* # of SIT entries per block */
257 	struct rw_semaphore sentry_lock;	/* to protect SIT cache */
258 	struct seg_entry *sentries;		/* SIT segment-level cache */
259 	struct sec_entry *sec_entries;		/* SIT section-level cache */
260 
261 	/* for cost-benefit algorithm in cleaning procedure */
262 	unsigned long long elapsed_time;	/* elapsed time after mount */
263 	unsigned long long mounted_time;	/* mount time */
264 	unsigned long long min_mtime;		/* min. modification time */
265 	unsigned long long max_mtime;		/* max. modification time */
266 	unsigned long long dirty_min_mtime;	/* rerange candidates in GC_AT */
267 	unsigned long long dirty_max_mtime;	/* rerange candidates in GC_AT */
268 
269 	unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
270 };
271 
272 struct free_segmap_info {
273 	unsigned int start_segno;	/* start segment number logically */
274 	unsigned int free_segments;	/* # of free segments */
275 	unsigned int free_sections;	/* # of free sections */
276 	spinlock_t segmap_lock;		/* free segmap lock */
277 	unsigned long *free_segmap;	/* free segment bitmap */
278 	unsigned long *free_secmap;	/* free section bitmap */
279 };
280 
281 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
282 enum dirty_type {
283 	DIRTY_HOT_DATA,		/* dirty segments assigned as hot data logs */
284 	DIRTY_WARM_DATA,	/* dirty segments assigned as warm data logs */
285 	DIRTY_COLD_DATA,	/* dirty segments assigned as cold data logs */
286 	DIRTY_HOT_NODE,		/* dirty segments assigned as hot node logs */
287 	DIRTY_WARM_NODE,	/* dirty segments assigned as warm node logs */
288 	DIRTY_COLD_NODE,	/* dirty segments assigned as cold node logs */
289 	DIRTY,			/* to count # of dirty segments */
290 	PRE,			/* to count # of entirely obsolete segments */
291 	NR_DIRTY_TYPE
292 };
293 
294 struct dirty_seglist_info {
295 	const struct victim_selection *v_ops;	/* victim selction operation */
296 	unsigned long *dirty_segmap[NR_DIRTY_TYPE];
297 	unsigned long *dirty_secmap;
298 	struct mutex seglist_lock;		/* lock for segment bitmaps */
299 	int nr_dirty[NR_DIRTY_TYPE];		/* # of dirty segments */
300 	unsigned long *victim_secmap;		/* background GC victims */
301 	unsigned long *pinned_secmap;		/* pinned victims from foreground GC */
302 	unsigned int pinned_secmap_cnt;		/* count of victims which has pinned data */
303 	bool enable_pin_section;		/* enable pinning section */
304 };
305 
306 /* victim selection function for cleaning and SSR */
307 struct victim_selection {
308 	int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
309 					int, int, char, unsigned long long);
310 };
311 
312 /* for active log information */
313 struct curseg_info {
314 	struct mutex curseg_mutex;		/* lock for consistency */
315 	struct f2fs_summary_block *sum_blk;	/* cached summary block */
316 	struct rw_semaphore journal_rwsem;	/* protect journal area */
317 	struct f2fs_journal *journal;		/* cached journal info */
318 	unsigned char alloc_type;		/* current allocation type */
319 	unsigned short seg_type;		/* segment type like CURSEG_XXX_TYPE */
320 	unsigned int segno;			/* current segment number */
321 	unsigned short next_blkoff;		/* next block offset to write */
322 	unsigned int zone;			/* current zone number */
323 	unsigned int next_segno;		/* preallocated segment */
324 	int fragment_remained_chunk;		/* remained block size in a chunk for block fragmentation mode */
325 	bool inited;				/* indicate inmem log is inited */
326 };
327 
328 struct sit_entry_set {
329 	struct list_head set_list;	/* link with all sit sets */
330 	unsigned int start_segno;	/* start segno of sits in set */
331 	unsigned int entry_cnt;		/* the # of sit entries in set */
332 };
333 
334 /*
335  * inline functions
336  */
337 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
338 {
339 	return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
340 }
341 
342 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
343 						unsigned int segno)
344 {
345 	struct sit_info *sit_i = SIT_I(sbi);
346 	return &sit_i->sentries[segno];
347 }
348 
349 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
350 						unsigned int segno)
351 {
352 	struct sit_info *sit_i = SIT_I(sbi);
353 	return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
354 }
355 
356 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
357 				unsigned int segno, bool use_section)
358 {
359 	/*
360 	 * In order to get # of valid blocks in a section instantly from many
361 	 * segments, f2fs manages two counting structures separately.
362 	 */
363 	if (use_section && __is_large_section(sbi))
364 		return get_sec_entry(sbi, segno)->valid_blocks;
365 	else
366 		return get_seg_entry(sbi, segno)->valid_blocks;
367 }
368 
369 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
370 				unsigned int segno, bool use_section)
371 {
372 	if (use_section && __is_large_section(sbi)) {
373 		unsigned int start_segno = START_SEGNO(segno);
374 		unsigned int blocks = 0;
375 		int i;
376 
377 		for (i = 0; i < sbi->segs_per_sec; i++, start_segno++) {
378 			struct seg_entry *se = get_seg_entry(sbi, start_segno);
379 
380 			blocks += se->ckpt_valid_blocks;
381 		}
382 		return blocks;
383 	}
384 	return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
385 }
386 
387 static inline void seg_info_from_raw_sit(struct seg_entry *se,
388 					struct f2fs_sit_entry *rs)
389 {
390 	se->valid_blocks = GET_SIT_VBLOCKS(rs);
391 	se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
392 	memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
393 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
394 #ifdef CONFIG_F2FS_CHECK_FS
395 	memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
396 #endif
397 	se->type = GET_SIT_TYPE(rs);
398 	se->mtime = le64_to_cpu(rs->mtime);
399 }
400 
401 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
402 					struct f2fs_sit_entry *rs)
403 {
404 	unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
405 					se->valid_blocks;
406 	rs->vblocks = cpu_to_le16(raw_vblocks);
407 	memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
408 	rs->mtime = cpu_to_le64(se->mtime);
409 }
410 
411 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
412 				struct page *page, unsigned int start)
413 {
414 	struct f2fs_sit_block *raw_sit;
415 	struct seg_entry *se;
416 	struct f2fs_sit_entry *rs;
417 	unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
418 					(unsigned long)MAIN_SEGS(sbi));
419 	int i;
420 
421 	raw_sit = (struct f2fs_sit_block *)page_address(page);
422 	memset(raw_sit, 0, PAGE_SIZE);
423 	for (i = 0; i < end - start; i++) {
424 		rs = &raw_sit->entries[i];
425 		se = get_seg_entry(sbi, start + i);
426 		__seg_info_to_raw_sit(se, rs);
427 	}
428 }
429 
430 static inline void seg_info_to_raw_sit(struct seg_entry *se,
431 					struct f2fs_sit_entry *rs)
432 {
433 	__seg_info_to_raw_sit(se, rs);
434 
435 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
436 	se->ckpt_valid_blocks = se->valid_blocks;
437 }
438 
439 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
440 		unsigned int max, unsigned int segno)
441 {
442 	unsigned int ret;
443 	spin_lock(&free_i->segmap_lock);
444 	ret = find_next_bit(free_i->free_segmap, max, segno);
445 	spin_unlock(&free_i->segmap_lock);
446 	return ret;
447 }
448 
449 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
450 {
451 	struct free_segmap_info *free_i = FREE_I(sbi);
452 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
453 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
454 	unsigned int next;
455 	unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
456 
457 	spin_lock(&free_i->segmap_lock);
458 	clear_bit(segno, free_i->free_segmap);
459 	free_i->free_segments++;
460 
461 	next = find_next_bit(free_i->free_segmap,
462 			start_segno + sbi->segs_per_sec, start_segno);
463 	if (next >= start_segno + usable_segs) {
464 		clear_bit(secno, free_i->free_secmap);
465 		free_i->free_sections++;
466 	}
467 	spin_unlock(&free_i->segmap_lock);
468 }
469 
470 static inline void __set_inuse(struct f2fs_sb_info *sbi,
471 		unsigned int segno)
472 {
473 	struct free_segmap_info *free_i = FREE_I(sbi);
474 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
475 
476 	set_bit(segno, free_i->free_segmap);
477 	free_i->free_segments--;
478 	if (!test_and_set_bit(secno, free_i->free_secmap))
479 		free_i->free_sections--;
480 }
481 
482 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
483 		unsigned int segno, bool inmem)
484 {
485 	struct free_segmap_info *free_i = FREE_I(sbi);
486 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
487 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
488 	unsigned int next;
489 	unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
490 
491 	spin_lock(&free_i->segmap_lock);
492 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
493 		free_i->free_segments++;
494 
495 		if (!inmem && IS_CURSEC(sbi, secno))
496 			goto skip_free;
497 		next = find_next_bit(free_i->free_segmap,
498 				start_segno + sbi->segs_per_sec, start_segno);
499 		if (next >= start_segno + usable_segs) {
500 			if (test_and_clear_bit(secno, free_i->free_secmap))
501 				free_i->free_sections++;
502 		}
503 	}
504 skip_free:
505 	spin_unlock(&free_i->segmap_lock);
506 }
507 
508 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
509 		unsigned int segno)
510 {
511 	struct free_segmap_info *free_i = FREE_I(sbi);
512 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
513 
514 	spin_lock(&free_i->segmap_lock);
515 	if (!test_and_set_bit(segno, free_i->free_segmap)) {
516 		free_i->free_segments--;
517 		if (!test_and_set_bit(secno, free_i->free_secmap))
518 			free_i->free_sections--;
519 	}
520 	spin_unlock(&free_i->segmap_lock);
521 }
522 
523 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
524 		void *dst_addr)
525 {
526 	struct sit_info *sit_i = SIT_I(sbi);
527 
528 #ifdef CONFIG_F2FS_CHECK_FS
529 	if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
530 						sit_i->bitmap_size))
531 		f2fs_bug_on(sbi, 1);
532 #endif
533 	memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
534 }
535 
536 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
537 {
538 	return SIT_I(sbi)->written_valid_blocks;
539 }
540 
541 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
542 {
543 	return FREE_I(sbi)->free_segments;
544 }
545 
546 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
547 {
548 	return SM_I(sbi)->reserved_segments +
549 			SM_I(sbi)->additional_reserved_segments;
550 }
551 
552 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
553 {
554 	return FREE_I(sbi)->free_sections;
555 }
556 
557 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
558 {
559 	return DIRTY_I(sbi)->nr_dirty[PRE];
560 }
561 
562 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
563 {
564 	return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
565 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
566 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
567 		DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
568 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
569 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
570 }
571 
572 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
573 {
574 	return SM_I(sbi)->ovp_segments;
575 }
576 
577 static inline int reserved_sections(struct f2fs_sb_info *sbi)
578 {
579 	return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
580 }
581 
582 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
583 			unsigned int node_blocks, unsigned int dent_blocks)
584 {
585 
586 	unsigned int segno, left_blocks;
587 	int i;
588 
589 	/* check current node segment */
590 	for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
591 		segno = CURSEG_I(sbi, i)->segno;
592 		left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
593 				get_seg_entry(sbi, segno)->ckpt_valid_blocks;
594 
595 		if (node_blocks > left_blocks)
596 			return false;
597 	}
598 
599 	/* check current data segment */
600 	segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
601 	left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
602 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
603 	if (dent_blocks > left_blocks)
604 		return false;
605 	return true;
606 }
607 
608 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
609 					int freed, int needed)
610 {
611 	unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
612 					get_pages(sbi, F2FS_DIRTY_DENTS) +
613 					get_pages(sbi, F2FS_DIRTY_IMETA);
614 	unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
615 	unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi);
616 	unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi);
617 	unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi);
618 	unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi);
619 	unsigned int free, need_lower, need_upper;
620 
621 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
622 		return false;
623 
624 	free = free_sections(sbi) + freed;
625 	need_lower = node_secs + dent_secs + reserved_sections(sbi) + needed;
626 	need_upper = need_lower + (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0);
627 
628 	if (free > need_upper)
629 		return false;
630 	else if (free <= need_lower)
631 		return true;
632 	return !has_curseg_enough_space(sbi, node_blocks, dent_blocks);
633 }
634 
635 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
636 {
637 	if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
638 		return true;
639 	if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
640 		return true;
641 	return false;
642 }
643 
644 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
645 {
646 	return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
647 }
648 
649 static inline int utilization(struct f2fs_sb_info *sbi)
650 {
651 	return div_u64((u64)valid_user_blocks(sbi) * 100,
652 					sbi->user_block_count);
653 }
654 
655 /*
656  * Sometimes f2fs may be better to drop out-of-place update policy.
657  * And, users can control the policy through sysfs entries.
658  * There are five policies with triggering conditions as follows.
659  * F2FS_IPU_FORCE - all the time,
660  * F2FS_IPU_SSR - if SSR mode is activated,
661  * F2FS_IPU_UTIL - if FS utilization is over threashold,
662  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
663  *                     threashold,
664  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
665  *                     storages. IPU will be triggered only if the # of dirty
666  *                     pages over min_fsync_blocks. (=default option)
667  * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
668  * F2FS_IPU_NOCACHE - disable IPU bio cache.
669  * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has
670  *                            FI_OPU_WRITE flag.
671  * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode)
672  */
673 #define DEF_MIN_IPU_UTIL	70
674 #define DEF_MIN_FSYNC_BLOCKS	8
675 #define DEF_MIN_HOT_BLOCKS	16
676 
677 #define SMALL_VOLUME_SEGMENTS	(16 * 512)	/* 16GB */
678 
679 enum {
680 	F2FS_IPU_FORCE,
681 	F2FS_IPU_SSR,
682 	F2FS_IPU_UTIL,
683 	F2FS_IPU_SSR_UTIL,
684 	F2FS_IPU_FSYNC,
685 	F2FS_IPU_ASYNC,
686 	F2FS_IPU_NOCACHE,
687 	F2FS_IPU_HONOR_OPU_WRITE,
688 };
689 
690 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
691 		int type)
692 {
693 	struct curseg_info *curseg = CURSEG_I(sbi, type);
694 	return curseg->segno;
695 }
696 
697 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
698 		int type)
699 {
700 	struct curseg_info *curseg = CURSEG_I(sbi, type);
701 	return curseg->alloc_type;
702 }
703 
704 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
705 {
706 	struct curseg_info *curseg = CURSEG_I(sbi, type);
707 	return curseg->next_blkoff;
708 }
709 
710 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
711 {
712 	f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
713 }
714 
715 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
716 {
717 	struct f2fs_sb_info *sbi = fio->sbi;
718 
719 	if (__is_valid_data_blkaddr(fio->old_blkaddr))
720 		verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
721 					META_GENERIC : DATA_GENERIC);
722 	verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
723 					META_GENERIC : DATA_GENERIC_ENHANCE);
724 }
725 
726 /*
727  * Summary block is always treated as an invalid block
728  */
729 static inline int check_block_count(struct f2fs_sb_info *sbi,
730 		int segno, struct f2fs_sit_entry *raw_sit)
731 {
732 	bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
733 	int valid_blocks = 0;
734 	int cur_pos = 0, next_pos;
735 	unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
736 
737 	/* check bitmap with valid block count */
738 	do {
739 		if (is_valid) {
740 			next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
741 					usable_blks_per_seg,
742 					cur_pos);
743 			valid_blocks += next_pos - cur_pos;
744 		} else
745 			next_pos = find_next_bit_le(&raw_sit->valid_map,
746 					usable_blks_per_seg,
747 					cur_pos);
748 		cur_pos = next_pos;
749 		is_valid = !is_valid;
750 	} while (cur_pos < usable_blks_per_seg);
751 
752 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
753 		f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
754 			 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
755 		set_sbi_flag(sbi, SBI_NEED_FSCK);
756 		f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
757 		return -EFSCORRUPTED;
758 	}
759 
760 	if (usable_blks_per_seg < sbi->blocks_per_seg)
761 		f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
762 				sbi->blocks_per_seg,
763 				usable_blks_per_seg) != sbi->blocks_per_seg);
764 
765 	/* check segment usage, and check boundary of a given segment number */
766 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
767 					|| segno > TOTAL_SEGS(sbi) - 1)) {
768 		f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
769 			 GET_SIT_VBLOCKS(raw_sit), segno);
770 		set_sbi_flag(sbi, SBI_NEED_FSCK);
771 		f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
772 		return -EFSCORRUPTED;
773 	}
774 	return 0;
775 }
776 
777 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
778 						unsigned int start)
779 {
780 	struct sit_info *sit_i = SIT_I(sbi);
781 	unsigned int offset = SIT_BLOCK_OFFSET(start);
782 	block_t blk_addr = sit_i->sit_base_addr + offset;
783 
784 	check_seg_range(sbi, start);
785 
786 #ifdef CONFIG_F2FS_CHECK_FS
787 	if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
788 			f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
789 		f2fs_bug_on(sbi, 1);
790 #endif
791 
792 	/* calculate sit block address */
793 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
794 		blk_addr += sit_i->sit_blocks;
795 
796 	return blk_addr;
797 }
798 
799 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
800 						pgoff_t block_addr)
801 {
802 	struct sit_info *sit_i = SIT_I(sbi);
803 	block_addr -= sit_i->sit_base_addr;
804 	if (block_addr < sit_i->sit_blocks)
805 		block_addr += sit_i->sit_blocks;
806 	else
807 		block_addr -= sit_i->sit_blocks;
808 
809 	return block_addr + sit_i->sit_base_addr;
810 }
811 
812 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
813 {
814 	unsigned int block_off = SIT_BLOCK_OFFSET(start);
815 
816 	f2fs_change_bit(block_off, sit_i->sit_bitmap);
817 #ifdef CONFIG_F2FS_CHECK_FS
818 	f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
819 #endif
820 }
821 
822 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
823 						bool base_time)
824 {
825 	struct sit_info *sit_i = SIT_I(sbi);
826 	time64_t diff, now = ktime_get_boottime_seconds();
827 
828 	if (now >= sit_i->mounted_time)
829 		return sit_i->elapsed_time + now - sit_i->mounted_time;
830 
831 	/* system time is set to the past */
832 	if (!base_time) {
833 		diff = sit_i->mounted_time - now;
834 		if (sit_i->elapsed_time >= diff)
835 			return sit_i->elapsed_time - diff;
836 		return 0;
837 	}
838 	return sit_i->elapsed_time;
839 }
840 
841 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
842 			unsigned int ofs_in_node, unsigned char version)
843 {
844 	sum->nid = cpu_to_le32(nid);
845 	sum->ofs_in_node = cpu_to_le16(ofs_in_node);
846 	sum->version = version;
847 }
848 
849 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
850 {
851 	return __start_cp_addr(sbi) +
852 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
853 }
854 
855 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
856 {
857 	return __start_cp_addr(sbi) +
858 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
859 				- (base + 1) + type;
860 }
861 
862 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
863 {
864 	if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
865 		return true;
866 	return false;
867 }
868 
869 /*
870  * It is very important to gather dirty pages and write at once, so that we can
871  * submit a big bio without interfering other data writes.
872  * By default, 512 pages for directory data,
873  * 512 pages (2MB) * 8 for nodes, and
874  * 256 pages * 8 for meta are set.
875  */
876 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
877 {
878 	if (sbi->sb->s_bdi->wb.dirty_exceeded)
879 		return 0;
880 
881 	if (type == DATA)
882 		return sbi->blocks_per_seg;
883 	else if (type == NODE)
884 		return 8 * sbi->blocks_per_seg;
885 	else if (type == META)
886 		return 8 * BIO_MAX_VECS;
887 	else
888 		return 0;
889 }
890 
891 /*
892  * When writing pages, it'd better align nr_to_write for segment size.
893  */
894 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
895 					struct writeback_control *wbc)
896 {
897 	long nr_to_write, desired;
898 
899 	if (wbc->sync_mode != WB_SYNC_NONE)
900 		return 0;
901 
902 	nr_to_write = wbc->nr_to_write;
903 	desired = BIO_MAX_VECS;
904 	if (type == NODE)
905 		desired <<= 1;
906 
907 	wbc->nr_to_write = desired;
908 	return desired - nr_to_write;
909 }
910 
911 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
912 {
913 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
914 	bool wakeup = false;
915 	int i;
916 
917 	if (force)
918 		goto wake_up;
919 
920 	mutex_lock(&dcc->cmd_lock);
921 	for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
922 		if (i + 1 < dcc->discard_granularity)
923 			break;
924 		if (!list_empty(&dcc->pend_list[i])) {
925 			wakeup = true;
926 			break;
927 		}
928 	}
929 	mutex_unlock(&dcc->cmd_lock);
930 	if (!wakeup || !is_idle(sbi, DISCARD_TIME))
931 		return;
932 wake_up:
933 	dcc->discard_wake = 1;
934 	wake_up_interruptible_all(&dcc->discard_wait_queue);
935 }
936