xref: /openbmc/linux/fs/f2fs/segment.h (revision e1181b5b)
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 two 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 two 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  * FORCE_FG_GC means on-demand cleaning job in background.
176  */
177 enum {
178 	BG_GC = 0,
179 	FG_GC,
180 	FORCE_FG_GC,
181 };
182 
183 /* for a function parameter to select a victim segment */
184 struct victim_sel_policy {
185 	int alloc_mode;			/* LFS or SSR */
186 	int gc_mode;			/* GC_CB or GC_GREEDY */
187 	unsigned long *dirty_bitmap;	/* dirty segment/section bitmap */
188 	unsigned int max_search;	/*
189 					 * maximum # of segments/sections
190 					 * to search
191 					 */
192 	unsigned int offset;		/* last scanned bitmap offset */
193 	unsigned int ofs_unit;		/* bitmap search unit */
194 	unsigned int min_cost;		/* minimum cost */
195 	unsigned long long oldest_age;	/* oldest age of segments having the same min cost */
196 	unsigned int min_segno;		/* segment # having min. cost */
197 	unsigned long long age;		/* mtime of GCed section*/
198 	unsigned long long age_threshold;/* age threshold */
199 };
200 
201 struct seg_entry {
202 	unsigned int type:6;		/* segment type like CURSEG_XXX_TYPE */
203 	unsigned int valid_blocks:10;	/* # of valid blocks */
204 	unsigned int ckpt_valid_blocks:10;	/* # of valid blocks last cp */
205 	unsigned int padding:6;		/* padding */
206 	unsigned char *cur_valid_map;	/* validity bitmap of blocks */
207 #ifdef CONFIG_F2FS_CHECK_FS
208 	unsigned char *cur_valid_map_mir;	/* mirror of current valid bitmap */
209 #endif
210 	/*
211 	 * # of valid blocks and the validity bitmap stored in the last
212 	 * checkpoint pack. This information is used by the SSR mode.
213 	 */
214 	unsigned char *ckpt_valid_map;	/* validity bitmap of blocks last cp */
215 	unsigned char *discard_map;
216 	unsigned long long mtime;	/* modification time of the segment */
217 };
218 
219 struct sec_entry {
220 	unsigned int valid_blocks;	/* # of valid blocks in a section */
221 };
222 
223 struct segment_allocation {
224 	void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
225 };
226 
227 #define MAX_SKIP_GC_COUNT			16
228 
229 struct inmem_pages {
230 	struct list_head list;
231 	struct page *page;
232 	block_t old_addr;		/* for revoking when fail to commit */
233 };
234 
235 struct sit_info {
236 	const struct segment_allocation *s_ops;
237 
238 	block_t sit_base_addr;		/* start block address of SIT area */
239 	block_t sit_blocks;		/* # of blocks used by SIT area */
240 	block_t written_valid_blocks;	/* # of valid blocks in main area */
241 	char *bitmap;			/* all bitmaps pointer */
242 	char *sit_bitmap;		/* SIT bitmap pointer */
243 #ifdef CONFIG_F2FS_CHECK_FS
244 	char *sit_bitmap_mir;		/* SIT bitmap mirror */
245 
246 	/* bitmap of segments to be ignored by GC in case of errors */
247 	unsigned long *invalid_segmap;
248 #endif
249 	unsigned int bitmap_size;	/* SIT bitmap size */
250 
251 	unsigned long *tmp_map;			/* bitmap for temporal use */
252 	unsigned long *dirty_sentries_bitmap;	/* bitmap for dirty sentries */
253 	unsigned int dirty_sentries;		/* # of dirty sentries */
254 	unsigned int sents_per_block;		/* # of SIT entries per block */
255 	struct rw_semaphore sentry_lock;	/* to protect SIT cache */
256 	struct seg_entry *sentries;		/* SIT segment-level cache */
257 	struct sec_entry *sec_entries;		/* SIT section-level cache */
258 
259 	/* for cost-benefit algorithm in cleaning procedure */
260 	unsigned long long elapsed_time;	/* elapsed time after mount */
261 	unsigned long long mounted_time;	/* mount time */
262 	unsigned long long min_mtime;		/* min. modification time */
263 	unsigned long long max_mtime;		/* max. modification time */
264 	unsigned long long dirty_min_mtime;	/* rerange candidates in GC_AT */
265 	unsigned long long dirty_max_mtime;	/* rerange candidates in GC_AT */
266 
267 	unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
268 };
269 
270 struct free_segmap_info {
271 	unsigned int start_segno;	/* start segment number logically */
272 	unsigned int free_segments;	/* # of free segments */
273 	unsigned int free_sections;	/* # of free sections */
274 	spinlock_t segmap_lock;		/* free segmap lock */
275 	unsigned long *free_segmap;	/* free segment bitmap */
276 	unsigned long *free_secmap;	/* free section bitmap */
277 };
278 
279 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
280 enum dirty_type {
281 	DIRTY_HOT_DATA,		/* dirty segments assigned as hot data logs */
282 	DIRTY_WARM_DATA,	/* dirty segments assigned as warm data logs */
283 	DIRTY_COLD_DATA,	/* dirty segments assigned as cold data logs */
284 	DIRTY_HOT_NODE,		/* dirty segments assigned as hot node logs */
285 	DIRTY_WARM_NODE,	/* dirty segments assigned as warm node logs */
286 	DIRTY_COLD_NODE,	/* dirty segments assigned as cold node logs */
287 	DIRTY,			/* to count # of dirty segments */
288 	PRE,			/* to count # of entirely obsolete segments */
289 	NR_DIRTY_TYPE
290 };
291 
292 struct dirty_seglist_info {
293 	const struct victim_selection *v_ops;	/* victim selction operation */
294 	unsigned long *dirty_segmap[NR_DIRTY_TYPE];
295 	unsigned long *dirty_secmap;
296 	struct mutex seglist_lock;		/* lock for segment bitmaps */
297 	int nr_dirty[NR_DIRTY_TYPE];		/* # of dirty segments */
298 	unsigned long *victim_secmap;		/* background GC victims */
299 };
300 
301 /* victim selection function for cleaning and SSR */
302 struct victim_selection {
303 	int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
304 					int, int, char, unsigned long long);
305 };
306 
307 /* for active log information */
308 struct curseg_info {
309 	struct mutex curseg_mutex;		/* lock for consistency */
310 	struct f2fs_summary_block *sum_blk;	/* cached summary block */
311 	struct rw_semaphore journal_rwsem;	/* protect journal area */
312 	struct f2fs_journal *journal;		/* cached journal info */
313 	unsigned char alloc_type;		/* current allocation type */
314 	unsigned short seg_type;		/* segment type like CURSEG_XXX_TYPE */
315 	unsigned int segno;			/* current segment number */
316 	unsigned short next_blkoff;		/* next block offset to write */
317 	unsigned int zone;			/* current zone number */
318 	unsigned int next_segno;		/* preallocated segment */
319 	bool inited;				/* indicate inmem log is inited */
320 };
321 
322 struct sit_entry_set {
323 	struct list_head set_list;	/* link with all sit sets */
324 	unsigned int start_segno;	/* start segno of sits in set */
325 	unsigned int entry_cnt;		/* the # of sit entries in set */
326 };
327 
328 /*
329  * inline functions
330  */
331 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
332 {
333 	return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
334 }
335 
336 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
337 						unsigned int segno)
338 {
339 	struct sit_info *sit_i = SIT_I(sbi);
340 	return &sit_i->sentries[segno];
341 }
342 
343 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
344 						unsigned int segno)
345 {
346 	struct sit_info *sit_i = SIT_I(sbi);
347 	return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
348 }
349 
350 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
351 				unsigned int segno, bool use_section)
352 {
353 	/*
354 	 * In order to get # of valid blocks in a section instantly from many
355 	 * segments, f2fs manages two counting structures separately.
356 	 */
357 	if (use_section && __is_large_section(sbi))
358 		return get_sec_entry(sbi, segno)->valid_blocks;
359 	else
360 		return get_seg_entry(sbi, segno)->valid_blocks;
361 }
362 
363 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
364 				unsigned int segno)
365 {
366 	return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
367 }
368 
369 static inline void seg_info_from_raw_sit(struct seg_entry *se,
370 					struct f2fs_sit_entry *rs)
371 {
372 	se->valid_blocks = GET_SIT_VBLOCKS(rs);
373 	se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
374 	memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
375 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
376 #ifdef CONFIG_F2FS_CHECK_FS
377 	memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
378 #endif
379 	se->type = GET_SIT_TYPE(rs);
380 	se->mtime = le64_to_cpu(rs->mtime);
381 }
382 
383 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
384 					struct f2fs_sit_entry *rs)
385 {
386 	unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
387 					se->valid_blocks;
388 	rs->vblocks = cpu_to_le16(raw_vblocks);
389 	memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
390 	rs->mtime = cpu_to_le64(se->mtime);
391 }
392 
393 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
394 				struct page *page, unsigned int start)
395 {
396 	struct f2fs_sit_block *raw_sit;
397 	struct seg_entry *se;
398 	struct f2fs_sit_entry *rs;
399 	unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
400 					(unsigned long)MAIN_SEGS(sbi));
401 	int i;
402 
403 	raw_sit = (struct f2fs_sit_block *)page_address(page);
404 	memset(raw_sit, 0, PAGE_SIZE);
405 	for (i = 0; i < end - start; i++) {
406 		rs = &raw_sit->entries[i];
407 		se = get_seg_entry(sbi, start + i);
408 		__seg_info_to_raw_sit(se, rs);
409 	}
410 }
411 
412 static inline void seg_info_to_raw_sit(struct seg_entry *se,
413 					struct f2fs_sit_entry *rs)
414 {
415 	__seg_info_to_raw_sit(se, rs);
416 
417 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
418 	se->ckpt_valid_blocks = se->valid_blocks;
419 }
420 
421 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
422 		unsigned int max, unsigned int segno)
423 {
424 	unsigned int ret;
425 	spin_lock(&free_i->segmap_lock);
426 	ret = find_next_bit(free_i->free_segmap, max, segno);
427 	spin_unlock(&free_i->segmap_lock);
428 	return ret;
429 }
430 
431 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
432 {
433 	struct free_segmap_info *free_i = FREE_I(sbi);
434 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
435 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
436 	unsigned int next;
437 	unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
438 
439 	spin_lock(&free_i->segmap_lock);
440 	clear_bit(segno, free_i->free_segmap);
441 	free_i->free_segments++;
442 
443 	next = find_next_bit(free_i->free_segmap,
444 			start_segno + sbi->segs_per_sec, start_segno);
445 	if (next >= start_segno + usable_segs) {
446 		clear_bit(secno, free_i->free_secmap);
447 		free_i->free_sections++;
448 	}
449 	spin_unlock(&free_i->segmap_lock);
450 }
451 
452 static inline void __set_inuse(struct f2fs_sb_info *sbi,
453 		unsigned int segno)
454 {
455 	struct free_segmap_info *free_i = FREE_I(sbi);
456 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
457 
458 	set_bit(segno, free_i->free_segmap);
459 	free_i->free_segments--;
460 	if (!test_and_set_bit(secno, free_i->free_secmap))
461 		free_i->free_sections--;
462 }
463 
464 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
465 		unsigned int segno, bool inmem)
466 {
467 	struct free_segmap_info *free_i = FREE_I(sbi);
468 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
469 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
470 	unsigned int next;
471 	unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
472 
473 	spin_lock(&free_i->segmap_lock);
474 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
475 		free_i->free_segments++;
476 
477 		if (!inmem && IS_CURSEC(sbi, secno))
478 			goto skip_free;
479 		next = find_next_bit(free_i->free_segmap,
480 				start_segno + sbi->segs_per_sec, start_segno);
481 		if (next >= start_segno + usable_segs) {
482 			if (test_and_clear_bit(secno, free_i->free_secmap))
483 				free_i->free_sections++;
484 		}
485 	}
486 skip_free:
487 	spin_unlock(&free_i->segmap_lock);
488 }
489 
490 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
491 		unsigned int segno)
492 {
493 	struct free_segmap_info *free_i = FREE_I(sbi);
494 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
495 
496 	spin_lock(&free_i->segmap_lock);
497 	if (!test_and_set_bit(segno, free_i->free_segmap)) {
498 		free_i->free_segments--;
499 		if (!test_and_set_bit(secno, free_i->free_secmap))
500 			free_i->free_sections--;
501 	}
502 	spin_unlock(&free_i->segmap_lock);
503 }
504 
505 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
506 		void *dst_addr)
507 {
508 	struct sit_info *sit_i = SIT_I(sbi);
509 
510 #ifdef CONFIG_F2FS_CHECK_FS
511 	if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
512 						sit_i->bitmap_size))
513 		f2fs_bug_on(sbi, 1);
514 #endif
515 	memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
516 }
517 
518 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
519 {
520 	return SIT_I(sbi)->written_valid_blocks;
521 }
522 
523 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
524 {
525 	return FREE_I(sbi)->free_segments;
526 }
527 
528 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
529 {
530 	return SM_I(sbi)->reserved_segments;
531 }
532 
533 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
534 {
535 	return FREE_I(sbi)->free_sections;
536 }
537 
538 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
539 {
540 	return DIRTY_I(sbi)->nr_dirty[PRE];
541 }
542 
543 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
544 {
545 	return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
546 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
547 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
548 		DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
549 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
550 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
551 }
552 
553 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
554 {
555 	return SM_I(sbi)->ovp_segments;
556 }
557 
558 static inline int reserved_sections(struct f2fs_sb_info *sbi)
559 {
560 	return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
561 }
562 
563 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
564 {
565 	unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
566 					get_pages(sbi, F2FS_DIRTY_DENTS);
567 	unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
568 	unsigned int segno, left_blocks;
569 	int i;
570 
571 	/* check current node segment */
572 	for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
573 		segno = CURSEG_I(sbi, i)->segno;
574 		left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
575 				get_seg_entry(sbi, segno)->ckpt_valid_blocks;
576 
577 		if (node_blocks > left_blocks)
578 			return false;
579 	}
580 
581 	/* check current data segment */
582 	segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
583 	left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
584 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
585 	if (dent_blocks > left_blocks)
586 		return false;
587 	return true;
588 }
589 
590 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
591 					int freed, int needed)
592 {
593 	int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
594 	int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
595 	int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
596 
597 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
598 		return false;
599 
600 	if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
601 			has_curseg_enough_space(sbi))
602 		return false;
603 	return (free_sections(sbi) + freed) <=
604 		(node_secs + 2 * dent_secs + imeta_secs +
605 		reserved_sections(sbi) + needed);
606 }
607 
608 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
609 {
610 	if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
611 		return true;
612 	if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
613 		return true;
614 	return false;
615 }
616 
617 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
618 {
619 	return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
620 }
621 
622 static inline int utilization(struct f2fs_sb_info *sbi)
623 {
624 	return div_u64((u64)valid_user_blocks(sbi) * 100,
625 					sbi->user_block_count);
626 }
627 
628 /*
629  * Sometimes f2fs may be better to drop out-of-place update policy.
630  * And, users can control the policy through sysfs entries.
631  * There are five policies with triggering conditions as follows.
632  * F2FS_IPU_FORCE - all the time,
633  * F2FS_IPU_SSR - if SSR mode is activated,
634  * F2FS_IPU_UTIL - if FS utilization is over threashold,
635  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
636  *                     threashold,
637  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
638  *                     storages. IPU will be triggered only if the # of dirty
639  *                     pages over min_fsync_blocks. (=default option)
640  * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
641  * F2FS_IPU_NOCACHE - disable IPU bio cache.
642  * F2FS_IPUT_DISABLE - disable IPU. (=default option in LFS mode)
643  */
644 #define DEF_MIN_IPU_UTIL	70
645 #define DEF_MIN_FSYNC_BLOCKS	8
646 #define DEF_MIN_HOT_BLOCKS	16
647 
648 #define SMALL_VOLUME_SEGMENTS	(16 * 512)	/* 16GB */
649 
650 enum {
651 	F2FS_IPU_FORCE,
652 	F2FS_IPU_SSR,
653 	F2FS_IPU_UTIL,
654 	F2FS_IPU_SSR_UTIL,
655 	F2FS_IPU_FSYNC,
656 	F2FS_IPU_ASYNC,
657 	F2FS_IPU_NOCACHE,
658 };
659 
660 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
661 		int type)
662 {
663 	struct curseg_info *curseg = CURSEG_I(sbi, type);
664 	return curseg->segno;
665 }
666 
667 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
668 		int type)
669 {
670 	struct curseg_info *curseg = CURSEG_I(sbi, type);
671 	return curseg->alloc_type;
672 }
673 
674 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
675 {
676 	struct curseg_info *curseg = CURSEG_I(sbi, type);
677 	return curseg->next_blkoff;
678 }
679 
680 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
681 {
682 	f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
683 }
684 
685 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
686 {
687 	struct f2fs_sb_info *sbi = fio->sbi;
688 
689 	if (__is_valid_data_blkaddr(fio->old_blkaddr))
690 		verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
691 					META_GENERIC : DATA_GENERIC);
692 	verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
693 					META_GENERIC : DATA_GENERIC_ENHANCE);
694 }
695 
696 /*
697  * Summary block is always treated as an invalid block
698  */
699 static inline int check_block_count(struct f2fs_sb_info *sbi,
700 		int segno, struct f2fs_sit_entry *raw_sit)
701 {
702 	bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
703 	int valid_blocks = 0;
704 	int cur_pos = 0, next_pos;
705 	unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
706 
707 	/* check bitmap with valid block count */
708 	do {
709 		if (is_valid) {
710 			next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
711 					usable_blks_per_seg,
712 					cur_pos);
713 			valid_blocks += next_pos - cur_pos;
714 		} else
715 			next_pos = find_next_bit_le(&raw_sit->valid_map,
716 					usable_blks_per_seg,
717 					cur_pos);
718 		cur_pos = next_pos;
719 		is_valid = !is_valid;
720 	} while (cur_pos < usable_blks_per_seg);
721 
722 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
723 		f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
724 			 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
725 		set_sbi_flag(sbi, SBI_NEED_FSCK);
726 		return -EFSCORRUPTED;
727 	}
728 
729 	if (usable_blks_per_seg < sbi->blocks_per_seg)
730 		f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
731 				sbi->blocks_per_seg,
732 				usable_blks_per_seg) != sbi->blocks_per_seg);
733 
734 	/* check segment usage, and check boundary of a given segment number */
735 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
736 					|| segno > TOTAL_SEGS(sbi) - 1)) {
737 		f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
738 			 GET_SIT_VBLOCKS(raw_sit), segno);
739 		set_sbi_flag(sbi, SBI_NEED_FSCK);
740 		return -EFSCORRUPTED;
741 	}
742 	return 0;
743 }
744 
745 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
746 						unsigned int start)
747 {
748 	struct sit_info *sit_i = SIT_I(sbi);
749 	unsigned int offset = SIT_BLOCK_OFFSET(start);
750 	block_t blk_addr = sit_i->sit_base_addr + offset;
751 
752 	check_seg_range(sbi, start);
753 
754 #ifdef CONFIG_F2FS_CHECK_FS
755 	if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
756 			f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
757 		f2fs_bug_on(sbi, 1);
758 #endif
759 
760 	/* calculate sit block address */
761 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
762 		blk_addr += sit_i->sit_blocks;
763 
764 	return blk_addr;
765 }
766 
767 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
768 						pgoff_t block_addr)
769 {
770 	struct sit_info *sit_i = SIT_I(sbi);
771 	block_addr -= sit_i->sit_base_addr;
772 	if (block_addr < sit_i->sit_blocks)
773 		block_addr += sit_i->sit_blocks;
774 	else
775 		block_addr -= sit_i->sit_blocks;
776 
777 	return block_addr + sit_i->sit_base_addr;
778 }
779 
780 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
781 {
782 	unsigned int block_off = SIT_BLOCK_OFFSET(start);
783 
784 	f2fs_change_bit(block_off, sit_i->sit_bitmap);
785 #ifdef CONFIG_F2FS_CHECK_FS
786 	f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
787 #endif
788 }
789 
790 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
791 						bool base_time)
792 {
793 	struct sit_info *sit_i = SIT_I(sbi);
794 	time64_t diff, now = ktime_get_boottime_seconds();
795 
796 	if (now >= sit_i->mounted_time)
797 		return sit_i->elapsed_time + now - sit_i->mounted_time;
798 
799 	/* system time is set to the past */
800 	if (!base_time) {
801 		diff = sit_i->mounted_time - now;
802 		if (sit_i->elapsed_time >= diff)
803 			return sit_i->elapsed_time - diff;
804 		return 0;
805 	}
806 	return sit_i->elapsed_time;
807 }
808 
809 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
810 			unsigned int ofs_in_node, unsigned char version)
811 {
812 	sum->nid = cpu_to_le32(nid);
813 	sum->ofs_in_node = cpu_to_le16(ofs_in_node);
814 	sum->version = version;
815 }
816 
817 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
818 {
819 	return __start_cp_addr(sbi) +
820 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
821 }
822 
823 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
824 {
825 	return __start_cp_addr(sbi) +
826 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
827 				- (base + 1) + type;
828 }
829 
830 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
831 {
832 	if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
833 		return true;
834 	return false;
835 }
836 
837 /*
838  * It is very important to gather dirty pages and write at once, so that we can
839  * submit a big bio without interfering other data writes.
840  * By default, 512 pages for directory data,
841  * 512 pages (2MB) * 8 for nodes, and
842  * 256 pages * 8 for meta are set.
843  */
844 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
845 {
846 	if (sbi->sb->s_bdi->wb.dirty_exceeded)
847 		return 0;
848 
849 	if (type == DATA)
850 		return sbi->blocks_per_seg;
851 	else if (type == NODE)
852 		return 8 * sbi->blocks_per_seg;
853 	else if (type == META)
854 		return 8 * BIO_MAX_VECS;
855 	else
856 		return 0;
857 }
858 
859 /*
860  * When writing pages, it'd better align nr_to_write for segment size.
861  */
862 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
863 					struct writeback_control *wbc)
864 {
865 	long nr_to_write, desired;
866 
867 	if (wbc->sync_mode != WB_SYNC_NONE)
868 		return 0;
869 
870 	nr_to_write = wbc->nr_to_write;
871 	desired = BIO_MAX_VECS;
872 	if (type == NODE)
873 		desired <<= 1;
874 
875 	wbc->nr_to_write = desired;
876 	return desired - nr_to_write;
877 }
878 
879 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
880 {
881 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
882 	bool wakeup = false;
883 	int i;
884 
885 	if (force)
886 		goto wake_up;
887 
888 	mutex_lock(&dcc->cmd_lock);
889 	for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
890 		if (i + 1 < dcc->discard_granularity)
891 			break;
892 		if (!list_empty(&dcc->pend_list[i])) {
893 			wakeup = true;
894 			break;
895 		}
896 	}
897 	mutex_unlock(&dcc->cmd_lock);
898 	if (!wakeup || !is_idle(sbi, DISCARD_TIME))
899 		return;
900 wake_up:
901 	dcc->discard_wake = 1;
902 	wake_up_interruptible_all(&dcc->discard_wait_queue);
903 }
904