xref: /linux/fs/f2fs/segment.h (revision 3f0a50f345f78183f6e9b39c2f45ca5dcaa511ca)
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 
28 static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
29 						unsigned short seg_type)
30 {
31 	f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG);
32 }
33 
34 #define IS_HOT(t)	((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
35 #define IS_WARM(t)	((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
36 #define IS_COLD(t)	((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
37 
38 #define IS_CURSEG(sbi, seg)						\
39 	(((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) ||	\
40 	 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) ||	\
41 	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) ||	\
42 	 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) ||	\
43 	 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) ||	\
44 	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno) ||	\
45 	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno) ||	\
46 	 ((seg) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno))
47 
48 #define IS_CURSEC(sbi, secno)						\
49 	(((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno /		\
50 	  (sbi)->segs_per_sec) ||	\
51 	 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno /		\
52 	  (sbi)->segs_per_sec) ||	\
53 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno /		\
54 	  (sbi)->segs_per_sec) ||	\
55 	 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno /		\
56 	  (sbi)->segs_per_sec) ||	\
57 	 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno /		\
58 	  (sbi)->segs_per_sec) ||	\
59 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno /		\
60 	  (sbi)->segs_per_sec) ||	\
61 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno /	\
62 	  (sbi)->segs_per_sec) ||	\
63 	 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno /	\
64 	  (sbi)->segs_per_sec))
65 
66 #define MAIN_BLKADDR(sbi)						\
67 	(SM_I(sbi) ? SM_I(sbi)->main_blkaddr : 				\
68 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
69 #define SEG0_BLKADDR(sbi)						\
70 	(SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : 				\
71 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
72 
73 #define MAIN_SEGS(sbi)	(SM_I(sbi)->main_segments)
74 #define MAIN_SECS(sbi)	((sbi)->total_sections)
75 
76 #define TOTAL_SEGS(sbi)							\
77 	(SM_I(sbi) ? SM_I(sbi)->segment_count : 				\
78 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
79 #define TOTAL_BLKS(sbi)	(TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
80 
81 #define MAX_BLKADDR(sbi)	(SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
82 #define SEGMENT_SIZE(sbi)	(1ULL << ((sbi)->log_blocksize +	\
83 					(sbi)->log_blocks_per_seg))
84 
85 #define START_BLOCK(sbi, segno)	(SEG0_BLKADDR(sbi) +			\
86 	 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
87 
88 #define NEXT_FREE_BLKADDR(sbi, curseg)					\
89 	(START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
90 
91 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr)	((blk_addr) - SEG0_BLKADDR(sbi))
92 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr)				\
93 	(GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
94 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr)				\
95 	(GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
96 
97 #define GET_SEGNO(sbi, blk_addr)					\
98 	((!__is_valid_data_blkaddr(blk_addr)) ?			\
99 	NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi),			\
100 		GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
101 #define BLKS_PER_SEC(sbi)					\
102 	((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
103 #define GET_SEC_FROM_SEG(sbi, segno)				\
104 	(((segno) == -1) ? -1: (segno) / (sbi)->segs_per_sec)
105 #define GET_SEG_FROM_SEC(sbi, secno)				\
106 	((secno) * (sbi)->segs_per_sec)
107 #define GET_ZONE_FROM_SEC(sbi, secno)				\
108 	(((secno) == -1) ? -1: (secno) / (sbi)->secs_per_zone)
109 #define GET_ZONE_FROM_SEG(sbi, segno)				\
110 	GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
111 
112 #define GET_SUM_BLOCK(sbi, segno)				\
113 	((sbi)->sm_info->ssa_blkaddr + (segno))
114 
115 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
116 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
117 
118 #define SIT_ENTRY_OFFSET(sit_i, segno)					\
119 	((segno) % (sit_i)->sents_per_block)
120 #define SIT_BLOCK_OFFSET(segno)					\
121 	((segno) / SIT_ENTRY_PER_BLOCK)
122 #define	START_SEGNO(segno)		\
123 	(SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
124 #define SIT_BLK_CNT(sbi)			\
125 	DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
126 #define f2fs_bitmap_size(nr)			\
127 	(BITS_TO_LONGS(nr) * sizeof(unsigned long))
128 
129 #define SECTOR_FROM_BLOCK(blk_addr)					\
130 	(((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
131 #define SECTOR_TO_BLOCK(sectors)					\
132 	((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
133 
134 /*
135  * indicate a block allocation direction: RIGHT and LEFT.
136  * RIGHT means allocating new sections towards the end of volume.
137  * LEFT means the opposite direction.
138  */
139 enum {
140 	ALLOC_RIGHT = 0,
141 	ALLOC_LEFT
142 };
143 
144 /*
145  * In the victim_sel_policy->alloc_mode, there are three block allocation modes.
146  * LFS writes data sequentially with cleaning operations.
147  * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
148  * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
149  * fragmented segment which has similar aging degree.
150  */
151 enum {
152 	LFS = 0,
153 	SSR,
154 	AT_SSR,
155 };
156 
157 /*
158  * In the victim_sel_policy->gc_mode, there are three gc, aka cleaning, modes.
159  * GC_CB is based on cost-benefit algorithm.
160  * GC_GREEDY is based on greedy algorithm.
161  * GC_AT is based on age-threshold algorithm.
162  */
163 enum {
164 	GC_CB = 0,
165 	GC_GREEDY,
166 	GC_AT,
167 	ALLOC_NEXT,
168 	FLUSH_DEVICE,
169 	MAX_GC_POLICY,
170 };
171 
172 /*
173  * BG_GC means the background cleaning job.
174  * FG_GC means the on-demand cleaning job.
175  */
176 enum {
177 	BG_GC = 0,
178 	FG_GC,
179 };
180 
181 /* for a function parameter to select a victim segment */
182 struct victim_sel_policy {
183 	int alloc_mode;			/* LFS or SSR */
184 	int gc_mode;			/* GC_CB or GC_GREEDY */
185 	unsigned long *dirty_bitmap;	/* dirty segment/section bitmap */
186 	unsigned int max_search;	/*
187 					 * maximum # of segments/sections
188 					 * to search
189 					 */
190 	unsigned int offset;		/* last scanned bitmap offset */
191 	unsigned int ofs_unit;		/* bitmap search unit */
192 	unsigned int min_cost;		/* minimum cost */
193 	unsigned long long oldest_age;	/* oldest age of segments having the same min cost */
194 	unsigned int min_segno;		/* segment # having min. cost */
195 	unsigned long long age;		/* mtime of GCed section*/
196 	unsigned long long age_threshold;/* age threshold */
197 };
198 
199 struct seg_entry {
200 	unsigned int type:6;		/* segment type like CURSEG_XXX_TYPE */
201 	unsigned int valid_blocks:10;	/* # of valid blocks */
202 	unsigned int ckpt_valid_blocks:10;	/* # of valid blocks last cp */
203 	unsigned int padding:6;		/* padding */
204 	unsigned char *cur_valid_map;	/* validity bitmap of blocks */
205 #ifdef CONFIG_F2FS_CHECK_FS
206 	unsigned char *cur_valid_map_mir;	/* mirror of current valid bitmap */
207 #endif
208 	/*
209 	 * # of valid blocks and the validity bitmap stored in the last
210 	 * checkpoint pack. This information is used by the SSR mode.
211 	 */
212 	unsigned char *ckpt_valid_map;	/* validity bitmap of blocks last cp */
213 	unsigned char *discard_map;
214 	unsigned long long mtime;	/* modification time of the segment */
215 };
216 
217 struct sec_entry {
218 	unsigned int valid_blocks;	/* # of valid blocks in a section */
219 };
220 
221 struct segment_allocation {
222 	void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
223 };
224 
225 #define MAX_SKIP_GC_COUNT			16
226 
227 struct inmem_pages {
228 	struct list_head list;
229 	struct page *page;
230 	block_t old_addr;		/* for revoking when fail to commit */
231 };
232 
233 struct sit_info {
234 	const struct segment_allocation *s_ops;
235 
236 	block_t sit_base_addr;		/* start block address of SIT area */
237 	block_t sit_blocks;		/* # of blocks used by SIT area */
238 	block_t written_valid_blocks;	/* # of valid blocks in main area */
239 	char *bitmap;			/* all bitmaps pointer */
240 	char *sit_bitmap;		/* SIT bitmap pointer */
241 #ifdef CONFIG_F2FS_CHECK_FS
242 	char *sit_bitmap_mir;		/* SIT bitmap mirror */
243 
244 	/* bitmap of segments to be ignored by GC in case of errors */
245 	unsigned long *invalid_segmap;
246 #endif
247 	unsigned int bitmap_size;	/* SIT bitmap size */
248 
249 	unsigned long *tmp_map;			/* bitmap for temporal use */
250 	unsigned long *dirty_sentries_bitmap;	/* bitmap for dirty sentries */
251 	unsigned int dirty_sentries;		/* # of dirty sentries */
252 	unsigned int sents_per_block;		/* # of SIT entries per block */
253 	struct rw_semaphore sentry_lock;	/* to protect SIT cache */
254 	struct seg_entry *sentries;		/* SIT segment-level cache */
255 	struct sec_entry *sec_entries;		/* SIT section-level cache */
256 
257 	/* for cost-benefit algorithm in cleaning procedure */
258 	unsigned long long elapsed_time;	/* elapsed time after mount */
259 	unsigned long long mounted_time;	/* mount time */
260 	unsigned long long min_mtime;		/* min. modification time */
261 	unsigned long long max_mtime;		/* max. modification time */
262 	unsigned long long dirty_min_mtime;	/* rerange candidates in GC_AT */
263 	unsigned long long dirty_max_mtime;	/* rerange candidates in GC_AT */
264 
265 	unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
266 };
267 
268 struct free_segmap_info {
269 	unsigned int start_segno;	/* start segment number logically */
270 	unsigned int free_segments;	/* # of free segments */
271 	unsigned int free_sections;	/* # of free sections */
272 	spinlock_t segmap_lock;		/* free segmap lock */
273 	unsigned long *free_segmap;	/* free segment bitmap */
274 	unsigned long *free_secmap;	/* free section bitmap */
275 };
276 
277 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
278 enum dirty_type {
279 	DIRTY_HOT_DATA,		/* dirty segments assigned as hot data logs */
280 	DIRTY_WARM_DATA,	/* dirty segments assigned as warm data logs */
281 	DIRTY_COLD_DATA,	/* dirty segments assigned as cold data logs */
282 	DIRTY_HOT_NODE,		/* dirty segments assigned as hot node logs */
283 	DIRTY_WARM_NODE,	/* dirty segments assigned as warm node logs */
284 	DIRTY_COLD_NODE,	/* dirty segments assigned as cold node logs */
285 	DIRTY,			/* to count # of dirty segments */
286 	PRE,			/* to count # of entirely obsolete segments */
287 	NR_DIRTY_TYPE
288 };
289 
290 struct dirty_seglist_info {
291 	const struct victim_selection *v_ops;	/* victim selction operation */
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 };
298 
299 /* victim selection function for cleaning and SSR */
300 struct victim_selection {
301 	int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
302 					int, int, char, unsigned long long);
303 };
304 
305 /* for active log information */
306 struct curseg_info {
307 	struct mutex curseg_mutex;		/* lock for consistency */
308 	struct f2fs_summary_block *sum_blk;	/* cached summary block */
309 	struct rw_semaphore journal_rwsem;	/* protect journal area */
310 	struct f2fs_journal *journal;		/* cached journal info */
311 	unsigned char alloc_type;		/* current allocation type */
312 	unsigned short seg_type;		/* segment type like CURSEG_XXX_TYPE */
313 	unsigned int segno;			/* current segment number */
314 	unsigned short next_blkoff;		/* next block offset to write */
315 	unsigned int zone;			/* current zone number */
316 	unsigned int next_segno;		/* preallocated segment */
317 	int fragment_remained_chunk;		/* remained block size in a chunk for block fragmentation mode */
318 	bool inited;				/* indicate inmem log is inited */
319 };
320 
321 struct sit_entry_set {
322 	struct list_head set_list;	/* link with all sit sets */
323 	unsigned int start_segno;	/* start segno of sits in set */
324 	unsigned int entry_cnt;		/* the # of sit entries in set */
325 };
326 
327 /*
328  * inline functions
329  */
330 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
331 {
332 	return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
333 }
334 
335 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
336 						unsigned int segno)
337 {
338 	struct sit_info *sit_i = SIT_I(sbi);
339 	return &sit_i->sentries[segno];
340 }
341 
342 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
343 						unsigned int segno)
344 {
345 	struct sit_info *sit_i = SIT_I(sbi);
346 	return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
347 }
348 
349 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
350 				unsigned int segno, bool use_section)
351 {
352 	/*
353 	 * In order to get # of valid blocks in a section instantly from many
354 	 * segments, f2fs manages two counting structures separately.
355 	 */
356 	if (use_section && __is_large_section(sbi))
357 		return get_sec_entry(sbi, segno)->valid_blocks;
358 	else
359 		return get_seg_entry(sbi, segno)->valid_blocks;
360 }
361 
362 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
363 				unsigned int segno, bool use_section)
364 {
365 	if (use_section && __is_large_section(sbi)) {
366 		unsigned int start_segno = START_SEGNO(segno);
367 		unsigned int blocks = 0;
368 		int i;
369 
370 		for (i = 0; i < sbi->segs_per_sec; i++, start_segno++) {
371 			struct seg_entry *se = get_seg_entry(sbi, start_segno);
372 
373 			blocks += se->ckpt_valid_blocks;
374 		}
375 		return blocks;
376 	}
377 	return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
378 }
379 
380 static inline void seg_info_from_raw_sit(struct seg_entry *se,
381 					struct f2fs_sit_entry *rs)
382 {
383 	se->valid_blocks = GET_SIT_VBLOCKS(rs);
384 	se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
385 	memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
386 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
387 #ifdef CONFIG_F2FS_CHECK_FS
388 	memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
389 #endif
390 	se->type = GET_SIT_TYPE(rs);
391 	se->mtime = le64_to_cpu(rs->mtime);
392 }
393 
394 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
395 					struct f2fs_sit_entry *rs)
396 {
397 	unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
398 					se->valid_blocks;
399 	rs->vblocks = cpu_to_le16(raw_vblocks);
400 	memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
401 	rs->mtime = cpu_to_le64(se->mtime);
402 }
403 
404 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
405 				struct page *page, unsigned int start)
406 {
407 	struct f2fs_sit_block *raw_sit;
408 	struct seg_entry *se;
409 	struct f2fs_sit_entry *rs;
410 	unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
411 					(unsigned long)MAIN_SEGS(sbi));
412 	int i;
413 
414 	raw_sit = (struct f2fs_sit_block *)page_address(page);
415 	memset(raw_sit, 0, PAGE_SIZE);
416 	for (i = 0; i < end - start; i++) {
417 		rs = &raw_sit->entries[i];
418 		se = get_seg_entry(sbi, start + i);
419 		__seg_info_to_raw_sit(se, rs);
420 	}
421 }
422 
423 static inline void seg_info_to_raw_sit(struct seg_entry *se,
424 					struct f2fs_sit_entry *rs)
425 {
426 	__seg_info_to_raw_sit(se, rs);
427 
428 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
429 	se->ckpt_valid_blocks = se->valid_blocks;
430 }
431 
432 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
433 		unsigned int max, unsigned int segno)
434 {
435 	unsigned int ret;
436 	spin_lock(&free_i->segmap_lock);
437 	ret = find_next_bit(free_i->free_segmap, max, segno);
438 	spin_unlock(&free_i->segmap_lock);
439 	return ret;
440 }
441 
442 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
443 {
444 	struct free_segmap_info *free_i = FREE_I(sbi);
445 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
446 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
447 	unsigned int next;
448 	unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
449 
450 	spin_lock(&free_i->segmap_lock);
451 	clear_bit(segno, free_i->free_segmap);
452 	free_i->free_segments++;
453 
454 	next = find_next_bit(free_i->free_segmap,
455 			start_segno + sbi->segs_per_sec, start_segno);
456 	if (next >= start_segno + usable_segs) {
457 		clear_bit(secno, free_i->free_secmap);
458 		free_i->free_sections++;
459 	}
460 	spin_unlock(&free_i->segmap_lock);
461 }
462 
463 static inline void __set_inuse(struct f2fs_sb_info *sbi,
464 		unsigned int segno)
465 {
466 	struct free_segmap_info *free_i = FREE_I(sbi);
467 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
468 
469 	set_bit(segno, free_i->free_segmap);
470 	free_i->free_segments--;
471 	if (!test_and_set_bit(secno, free_i->free_secmap))
472 		free_i->free_sections--;
473 }
474 
475 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
476 		unsigned int segno, bool inmem)
477 {
478 	struct free_segmap_info *free_i = FREE_I(sbi);
479 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
480 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
481 	unsigned int next;
482 	unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
483 
484 	spin_lock(&free_i->segmap_lock);
485 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
486 		free_i->free_segments++;
487 
488 		if (!inmem && IS_CURSEC(sbi, secno))
489 			goto skip_free;
490 		next = find_next_bit(free_i->free_segmap,
491 				start_segno + sbi->segs_per_sec, start_segno);
492 		if (next >= start_segno + usable_segs) {
493 			if (test_and_clear_bit(secno, free_i->free_secmap))
494 				free_i->free_sections++;
495 		}
496 	}
497 skip_free:
498 	spin_unlock(&free_i->segmap_lock);
499 }
500 
501 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
502 		unsigned int segno)
503 {
504 	struct free_segmap_info *free_i = FREE_I(sbi);
505 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
506 
507 	spin_lock(&free_i->segmap_lock);
508 	if (!test_and_set_bit(segno, free_i->free_segmap)) {
509 		free_i->free_segments--;
510 		if (!test_and_set_bit(secno, free_i->free_secmap))
511 			free_i->free_sections--;
512 	}
513 	spin_unlock(&free_i->segmap_lock);
514 }
515 
516 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
517 		void *dst_addr)
518 {
519 	struct sit_info *sit_i = SIT_I(sbi);
520 
521 #ifdef CONFIG_F2FS_CHECK_FS
522 	if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
523 						sit_i->bitmap_size))
524 		f2fs_bug_on(sbi, 1);
525 #endif
526 	memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
527 }
528 
529 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
530 {
531 	return SIT_I(sbi)->written_valid_blocks;
532 }
533 
534 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
535 {
536 	return FREE_I(sbi)->free_segments;
537 }
538 
539 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
540 {
541 	return SM_I(sbi)->reserved_segments +
542 			SM_I(sbi)->additional_reserved_segments;
543 }
544 
545 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
546 {
547 	return FREE_I(sbi)->free_sections;
548 }
549 
550 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
551 {
552 	return DIRTY_I(sbi)->nr_dirty[PRE];
553 }
554 
555 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
556 {
557 	return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
558 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
559 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
560 		DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
561 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
562 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
563 }
564 
565 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
566 {
567 	return SM_I(sbi)->ovp_segments;
568 }
569 
570 static inline int reserved_sections(struct f2fs_sb_info *sbi)
571 {
572 	return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
573 }
574 
575 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
576 {
577 	unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
578 					get_pages(sbi, F2FS_DIRTY_DENTS);
579 	unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
580 	unsigned int segno, left_blocks;
581 	int i;
582 
583 	/* check current node segment */
584 	for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
585 		segno = CURSEG_I(sbi, i)->segno;
586 		left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
587 				get_seg_entry(sbi, segno)->ckpt_valid_blocks;
588 
589 		if (node_blocks > left_blocks)
590 			return false;
591 	}
592 
593 	/* check current data segment */
594 	segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
595 	left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
596 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
597 	if (dent_blocks > left_blocks)
598 		return false;
599 	return true;
600 }
601 
602 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
603 					int freed, int needed)
604 {
605 	int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
606 	int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
607 	int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
608 
609 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
610 		return false;
611 
612 	if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
613 			has_curseg_enough_space(sbi))
614 		return false;
615 	return (free_sections(sbi) + freed) <=
616 		(node_secs + 2 * dent_secs + imeta_secs +
617 		reserved_sections(sbi) + needed);
618 }
619 
620 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
621 {
622 	if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
623 		return true;
624 	if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
625 		return true;
626 	return false;
627 }
628 
629 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
630 {
631 	return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
632 }
633 
634 static inline int utilization(struct f2fs_sb_info *sbi)
635 {
636 	return div_u64((u64)valid_user_blocks(sbi) * 100,
637 					sbi->user_block_count);
638 }
639 
640 /*
641  * Sometimes f2fs may be better to drop out-of-place update policy.
642  * And, users can control the policy through sysfs entries.
643  * There are five policies with triggering conditions as follows.
644  * F2FS_IPU_FORCE - all the time,
645  * F2FS_IPU_SSR - if SSR mode is activated,
646  * F2FS_IPU_UTIL - if FS utilization is over threashold,
647  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
648  *                     threashold,
649  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
650  *                     storages. IPU will be triggered only if the # of dirty
651  *                     pages over min_fsync_blocks. (=default option)
652  * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
653  * F2FS_IPU_NOCACHE - disable IPU bio cache.
654  * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has
655  *                            FI_OPU_WRITE flag.
656  * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode)
657  */
658 #define DEF_MIN_IPU_UTIL	70
659 #define DEF_MIN_FSYNC_BLOCKS	8
660 #define DEF_MIN_HOT_BLOCKS	16
661 
662 #define SMALL_VOLUME_SEGMENTS	(16 * 512)	/* 16GB */
663 
664 enum {
665 	F2FS_IPU_FORCE,
666 	F2FS_IPU_SSR,
667 	F2FS_IPU_UTIL,
668 	F2FS_IPU_SSR_UTIL,
669 	F2FS_IPU_FSYNC,
670 	F2FS_IPU_ASYNC,
671 	F2FS_IPU_NOCACHE,
672 	F2FS_IPU_HONOR_OPU_WRITE,
673 };
674 
675 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
676 		int type)
677 {
678 	struct curseg_info *curseg = CURSEG_I(sbi, type);
679 	return curseg->segno;
680 }
681 
682 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
683 		int type)
684 {
685 	struct curseg_info *curseg = CURSEG_I(sbi, type);
686 	return curseg->alloc_type;
687 }
688 
689 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
690 {
691 	struct curseg_info *curseg = CURSEG_I(sbi, type);
692 	return curseg->next_blkoff;
693 }
694 
695 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
696 {
697 	f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
698 }
699 
700 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
701 {
702 	struct f2fs_sb_info *sbi = fio->sbi;
703 
704 	if (__is_valid_data_blkaddr(fio->old_blkaddr))
705 		verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
706 					META_GENERIC : DATA_GENERIC);
707 	verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
708 					META_GENERIC : DATA_GENERIC_ENHANCE);
709 }
710 
711 /*
712  * Summary block is always treated as an invalid block
713  */
714 static inline int check_block_count(struct f2fs_sb_info *sbi,
715 		int segno, struct f2fs_sit_entry *raw_sit)
716 {
717 	bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
718 	int valid_blocks = 0;
719 	int cur_pos = 0, next_pos;
720 	unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
721 
722 	/* check bitmap with valid block count */
723 	do {
724 		if (is_valid) {
725 			next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
726 					usable_blks_per_seg,
727 					cur_pos);
728 			valid_blocks += next_pos - cur_pos;
729 		} else
730 			next_pos = find_next_bit_le(&raw_sit->valid_map,
731 					usable_blks_per_seg,
732 					cur_pos);
733 		cur_pos = next_pos;
734 		is_valid = !is_valid;
735 	} while (cur_pos < usable_blks_per_seg);
736 
737 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
738 		f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
739 			 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
740 		set_sbi_flag(sbi, SBI_NEED_FSCK);
741 		return -EFSCORRUPTED;
742 	}
743 
744 	if (usable_blks_per_seg < sbi->blocks_per_seg)
745 		f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
746 				sbi->blocks_per_seg,
747 				usable_blks_per_seg) != sbi->blocks_per_seg);
748 
749 	/* check segment usage, and check boundary of a given segment number */
750 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
751 					|| segno > TOTAL_SEGS(sbi) - 1)) {
752 		f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
753 			 GET_SIT_VBLOCKS(raw_sit), segno);
754 		set_sbi_flag(sbi, SBI_NEED_FSCK);
755 		return -EFSCORRUPTED;
756 	}
757 	return 0;
758 }
759 
760 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
761 						unsigned int start)
762 {
763 	struct sit_info *sit_i = SIT_I(sbi);
764 	unsigned int offset = SIT_BLOCK_OFFSET(start);
765 	block_t blk_addr = sit_i->sit_base_addr + offset;
766 
767 	check_seg_range(sbi, start);
768 
769 #ifdef CONFIG_F2FS_CHECK_FS
770 	if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
771 			f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
772 		f2fs_bug_on(sbi, 1);
773 #endif
774 
775 	/* calculate sit block address */
776 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
777 		blk_addr += sit_i->sit_blocks;
778 
779 	return blk_addr;
780 }
781 
782 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
783 						pgoff_t block_addr)
784 {
785 	struct sit_info *sit_i = SIT_I(sbi);
786 	block_addr -= sit_i->sit_base_addr;
787 	if (block_addr < sit_i->sit_blocks)
788 		block_addr += sit_i->sit_blocks;
789 	else
790 		block_addr -= sit_i->sit_blocks;
791 
792 	return block_addr + sit_i->sit_base_addr;
793 }
794 
795 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
796 {
797 	unsigned int block_off = SIT_BLOCK_OFFSET(start);
798 
799 	f2fs_change_bit(block_off, sit_i->sit_bitmap);
800 #ifdef CONFIG_F2FS_CHECK_FS
801 	f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
802 #endif
803 }
804 
805 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
806 						bool base_time)
807 {
808 	struct sit_info *sit_i = SIT_I(sbi);
809 	time64_t diff, now = ktime_get_boottime_seconds();
810 
811 	if (now >= sit_i->mounted_time)
812 		return sit_i->elapsed_time + now - sit_i->mounted_time;
813 
814 	/* system time is set to the past */
815 	if (!base_time) {
816 		diff = sit_i->mounted_time - now;
817 		if (sit_i->elapsed_time >= diff)
818 			return sit_i->elapsed_time - diff;
819 		return 0;
820 	}
821 	return sit_i->elapsed_time;
822 }
823 
824 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
825 			unsigned int ofs_in_node, unsigned char version)
826 {
827 	sum->nid = cpu_to_le32(nid);
828 	sum->ofs_in_node = cpu_to_le16(ofs_in_node);
829 	sum->version = version;
830 }
831 
832 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
833 {
834 	return __start_cp_addr(sbi) +
835 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
836 }
837 
838 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
839 {
840 	return __start_cp_addr(sbi) +
841 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
842 				- (base + 1) + type;
843 }
844 
845 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
846 {
847 	if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
848 		return true;
849 	return false;
850 }
851 
852 /*
853  * It is very important to gather dirty pages and write at once, so that we can
854  * submit a big bio without interfering other data writes.
855  * By default, 512 pages for directory data,
856  * 512 pages (2MB) * 8 for nodes, and
857  * 256 pages * 8 for meta are set.
858  */
859 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
860 {
861 	if (sbi->sb->s_bdi->wb.dirty_exceeded)
862 		return 0;
863 
864 	if (type == DATA)
865 		return sbi->blocks_per_seg;
866 	else if (type == NODE)
867 		return 8 * sbi->blocks_per_seg;
868 	else if (type == META)
869 		return 8 * BIO_MAX_VECS;
870 	else
871 		return 0;
872 }
873 
874 /*
875  * When writing pages, it'd better align nr_to_write for segment size.
876  */
877 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
878 					struct writeback_control *wbc)
879 {
880 	long nr_to_write, desired;
881 
882 	if (wbc->sync_mode != WB_SYNC_NONE)
883 		return 0;
884 
885 	nr_to_write = wbc->nr_to_write;
886 	desired = BIO_MAX_VECS;
887 	if (type == NODE)
888 		desired <<= 1;
889 
890 	wbc->nr_to_write = desired;
891 	return desired - nr_to_write;
892 }
893 
894 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
895 {
896 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
897 	bool wakeup = false;
898 	int i;
899 
900 	if (force)
901 		goto wake_up;
902 
903 	mutex_lock(&dcc->cmd_lock);
904 	for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
905 		if (i + 1 < dcc->discard_granularity)
906 			break;
907 		if (!list_empty(&dcc->pend_list[i])) {
908 			wakeup = true;
909 			break;
910 		}
911 	}
912 	mutex_unlock(&dcc->cmd_lock);
913 	if (!wakeup || !is_idle(sbi, DISCARD_TIME))
914 		return;
915 wake_up:
916 	dcc->discard_wake = 1;
917 	wake_up_interruptible_all(&dcc->discard_wait_queue);
918 }
919