xref: /openbmc/linux/fs/f2fs/segment.h (revision 4a3fad70)
1 /*
2  * fs/f2fs/segment.h
3  *
4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5  *             http://www.samsung.com/
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  */
11 #include <linux/blkdev.h>
12 #include <linux/backing-dev.h>
13 
14 /* constant macro */
15 #define NULL_SEGNO			((unsigned int)(~0))
16 #define NULL_SECNO			((unsigned int)(~0))
17 
18 #define DEF_RECLAIM_PREFREE_SEGMENTS	5	/* 5% over total segments */
19 #define DEF_MAX_RECLAIM_PREFREE_SEGMENTS	4096	/* 8GB in maximum */
20 
21 #define F2FS_MIN_SEGMENTS	9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
22 
23 /* L: Logical segment # in volume, R: Relative segment # in main area */
24 #define GET_L2R_SEGNO(free_i, segno)	((segno) - (free_i)->start_segno)
25 #define GET_R2L_SEGNO(free_i, segno)	((segno) + (free_i)->start_segno)
26 
27 #define IS_DATASEG(t)	((t) <= CURSEG_COLD_DATA)
28 #define IS_NODESEG(t)	((t) >= CURSEG_HOT_NODE)
29 
30 #define IS_HOT(t)	((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
31 #define IS_WARM(t)	((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
32 #define IS_COLD(t)	((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
33 
34 #define IS_CURSEG(sbi, seg)						\
35 	(((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) ||	\
36 	 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) ||	\
37 	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) ||	\
38 	 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) ||	\
39 	 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) ||	\
40 	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
41 
42 #define IS_CURSEC(sbi, secno)						\
43 	(((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno /		\
44 	  (sbi)->segs_per_sec) ||	\
45 	 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno /		\
46 	  (sbi)->segs_per_sec) ||	\
47 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno /		\
48 	  (sbi)->segs_per_sec) ||	\
49 	 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno /		\
50 	  (sbi)->segs_per_sec) ||	\
51 	 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno /		\
52 	  (sbi)->segs_per_sec) ||	\
53 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno /		\
54 	  (sbi)->segs_per_sec))	\
55 
56 #define MAIN_BLKADDR(sbi)	(SM_I(sbi)->main_blkaddr)
57 #define SEG0_BLKADDR(sbi)	(SM_I(sbi)->seg0_blkaddr)
58 
59 #define MAIN_SEGS(sbi)	(SM_I(sbi)->main_segments)
60 #define MAIN_SECS(sbi)	((sbi)->total_sections)
61 
62 #define TOTAL_SEGS(sbi)	(SM_I(sbi)->segment_count)
63 #define TOTAL_BLKS(sbi)	(TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
64 
65 #define MAX_BLKADDR(sbi)	(SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
66 #define SEGMENT_SIZE(sbi)	(1ULL << ((sbi)->log_blocksize +	\
67 					(sbi)->log_blocks_per_seg))
68 
69 #define START_BLOCK(sbi, segno)	(SEG0_BLKADDR(sbi) +			\
70 	 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
71 
72 #define NEXT_FREE_BLKADDR(sbi, curseg)					\
73 	(START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
74 
75 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr)	((blk_addr) - SEG0_BLKADDR(sbi))
76 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr)				\
77 	(GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
78 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr)				\
79 	(GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
80 
81 #define GET_SEGNO(sbi, blk_addr)					\
82 	((((blk_addr) == NULL_ADDR) || ((blk_addr) == NEW_ADDR)) ?	\
83 	NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi),			\
84 		GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
85 #define BLKS_PER_SEC(sbi)					\
86 	((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
87 #define GET_SEC_FROM_SEG(sbi, segno)				\
88 	((segno) / (sbi)->segs_per_sec)
89 #define GET_SEG_FROM_SEC(sbi, secno)				\
90 	((secno) * (sbi)->segs_per_sec)
91 #define GET_ZONE_FROM_SEC(sbi, secno)				\
92 	((secno) / (sbi)->secs_per_zone)
93 #define GET_ZONE_FROM_SEG(sbi, segno)				\
94 	GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
95 
96 #define GET_SUM_BLOCK(sbi, segno)				\
97 	((sbi)->sm_info->ssa_blkaddr + (segno))
98 
99 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
100 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
101 
102 #define SIT_ENTRY_OFFSET(sit_i, segno)					\
103 	((segno) % (sit_i)->sents_per_block)
104 #define SIT_BLOCK_OFFSET(segno)					\
105 	((segno) / SIT_ENTRY_PER_BLOCK)
106 #define	START_SEGNO(segno)		\
107 	(SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
108 #define SIT_BLK_CNT(sbi)			\
109 	((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
110 #define f2fs_bitmap_size(nr)			\
111 	(BITS_TO_LONGS(nr) * sizeof(unsigned long))
112 
113 #define SECTOR_FROM_BLOCK(blk_addr)					\
114 	(((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
115 #define SECTOR_TO_BLOCK(sectors)					\
116 	((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
117 
118 /*
119  * indicate a block allocation direction: RIGHT and LEFT.
120  * RIGHT means allocating new sections towards the end of volume.
121  * LEFT means the opposite direction.
122  */
123 enum {
124 	ALLOC_RIGHT = 0,
125 	ALLOC_LEFT
126 };
127 
128 /*
129  * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
130  * LFS writes data sequentially with cleaning operations.
131  * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
132  */
133 enum {
134 	LFS = 0,
135 	SSR
136 };
137 
138 /*
139  * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
140  * GC_CB is based on cost-benefit algorithm.
141  * GC_GREEDY is based on greedy algorithm.
142  */
143 enum {
144 	GC_CB = 0,
145 	GC_GREEDY,
146 	ALLOC_NEXT,
147 	FLUSH_DEVICE,
148 	MAX_GC_POLICY,
149 };
150 
151 /*
152  * BG_GC means the background cleaning job.
153  * FG_GC means the on-demand cleaning job.
154  * FORCE_FG_GC means on-demand cleaning job in background.
155  */
156 enum {
157 	BG_GC = 0,
158 	FG_GC,
159 	FORCE_FG_GC,
160 };
161 
162 /* for a function parameter to select a victim segment */
163 struct victim_sel_policy {
164 	int alloc_mode;			/* LFS or SSR */
165 	int gc_mode;			/* GC_CB or GC_GREEDY */
166 	unsigned long *dirty_segmap;	/* dirty segment bitmap */
167 	unsigned int max_search;	/* maximum # of segments to search */
168 	unsigned int offset;		/* last scanned bitmap offset */
169 	unsigned int ofs_unit;		/* bitmap search unit */
170 	unsigned int min_cost;		/* minimum cost */
171 	unsigned int min_segno;		/* segment # having min. cost */
172 };
173 
174 struct seg_entry {
175 	unsigned int type:6;		/* segment type like CURSEG_XXX_TYPE */
176 	unsigned int valid_blocks:10;	/* # of valid blocks */
177 	unsigned int ckpt_valid_blocks:10;	/* # of valid blocks last cp */
178 	unsigned int padding:6;		/* padding */
179 	unsigned char *cur_valid_map;	/* validity bitmap of blocks */
180 #ifdef CONFIG_F2FS_CHECK_FS
181 	unsigned char *cur_valid_map_mir;	/* mirror of current valid bitmap */
182 #endif
183 	/*
184 	 * # of valid blocks and the validity bitmap stored in the the last
185 	 * checkpoint pack. This information is used by the SSR mode.
186 	 */
187 	unsigned char *ckpt_valid_map;	/* validity bitmap of blocks last cp */
188 	unsigned char *discard_map;
189 	unsigned long long mtime;	/* modification time of the segment */
190 };
191 
192 struct sec_entry {
193 	unsigned int valid_blocks;	/* # of valid blocks in a section */
194 };
195 
196 struct segment_allocation {
197 	void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
198 };
199 
200 /*
201  * this value is set in page as a private data which indicate that
202  * the page is atomically written, and it is in inmem_pages list.
203  */
204 #define ATOMIC_WRITTEN_PAGE		((unsigned long)-1)
205 #define DUMMY_WRITTEN_PAGE		((unsigned long)-2)
206 
207 #define IS_ATOMIC_WRITTEN_PAGE(page)			\
208 		(page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
209 #define IS_DUMMY_WRITTEN_PAGE(page)			\
210 		(page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
211 
212 struct inmem_pages {
213 	struct list_head list;
214 	struct page *page;
215 	block_t old_addr;		/* for revoking when fail to commit */
216 };
217 
218 struct sit_info {
219 	const struct segment_allocation *s_ops;
220 
221 	block_t sit_base_addr;		/* start block address of SIT area */
222 	block_t sit_blocks;		/* # of blocks used by SIT area */
223 	block_t written_valid_blocks;	/* # of valid blocks in main area */
224 	char *sit_bitmap;		/* SIT bitmap pointer */
225 #ifdef CONFIG_F2FS_CHECK_FS
226 	char *sit_bitmap_mir;		/* SIT bitmap mirror */
227 #endif
228 	unsigned int bitmap_size;	/* SIT bitmap size */
229 
230 	unsigned long *tmp_map;			/* bitmap for temporal use */
231 	unsigned long *dirty_sentries_bitmap;	/* bitmap for dirty sentries */
232 	unsigned int dirty_sentries;		/* # of dirty sentries */
233 	unsigned int sents_per_block;		/* # of SIT entries per block */
234 	struct rw_semaphore sentry_lock;	/* to protect SIT cache */
235 	struct seg_entry *sentries;		/* SIT segment-level cache */
236 	struct sec_entry *sec_entries;		/* SIT section-level cache */
237 
238 	/* for cost-benefit algorithm in cleaning procedure */
239 	unsigned long long elapsed_time;	/* elapsed time after mount */
240 	unsigned long long mounted_time;	/* mount time */
241 	unsigned long long min_mtime;		/* min. modification time */
242 	unsigned long long max_mtime;		/* max. modification time */
243 
244 	unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
245 };
246 
247 struct free_segmap_info {
248 	unsigned int start_segno;	/* start segment number logically */
249 	unsigned int free_segments;	/* # of free segments */
250 	unsigned int free_sections;	/* # of free sections */
251 	spinlock_t segmap_lock;		/* free segmap lock */
252 	unsigned long *free_segmap;	/* free segment bitmap */
253 	unsigned long *free_secmap;	/* free section bitmap */
254 };
255 
256 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
257 enum dirty_type {
258 	DIRTY_HOT_DATA,		/* dirty segments assigned as hot data logs */
259 	DIRTY_WARM_DATA,	/* dirty segments assigned as warm data logs */
260 	DIRTY_COLD_DATA,	/* dirty segments assigned as cold data logs */
261 	DIRTY_HOT_NODE,		/* dirty segments assigned as hot node logs */
262 	DIRTY_WARM_NODE,	/* dirty segments assigned as warm node logs */
263 	DIRTY_COLD_NODE,	/* dirty segments assigned as cold node logs */
264 	DIRTY,			/* to count # of dirty segments */
265 	PRE,			/* to count # of entirely obsolete segments */
266 	NR_DIRTY_TYPE
267 };
268 
269 struct dirty_seglist_info {
270 	const struct victim_selection *v_ops;	/* victim selction operation */
271 	unsigned long *dirty_segmap[NR_DIRTY_TYPE];
272 	struct mutex seglist_lock;		/* lock for segment bitmaps */
273 	int nr_dirty[NR_DIRTY_TYPE];		/* # of dirty segments */
274 	unsigned long *victim_secmap;		/* background GC victims */
275 };
276 
277 /* victim selection function for cleaning and SSR */
278 struct victim_selection {
279 	int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
280 							int, int, char);
281 };
282 
283 /* for active log information */
284 struct curseg_info {
285 	struct mutex curseg_mutex;		/* lock for consistency */
286 	struct f2fs_summary_block *sum_blk;	/* cached summary block */
287 	struct rw_semaphore journal_rwsem;	/* protect journal area */
288 	struct f2fs_journal *journal;		/* cached journal info */
289 	unsigned char alloc_type;		/* current allocation type */
290 	unsigned int segno;			/* current segment number */
291 	unsigned short next_blkoff;		/* next block offset to write */
292 	unsigned int zone;			/* current zone number */
293 	unsigned int next_segno;		/* preallocated segment */
294 };
295 
296 struct sit_entry_set {
297 	struct list_head set_list;	/* link with all sit sets */
298 	unsigned int start_segno;	/* start segno of sits in set */
299 	unsigned int entry_cnt;		/* the # of sit entries in set */
300 };
301 
302 /*
303  * inline functions
304  */
305 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
306 {
307 	return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
308 }
309 
310 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
311 						unsigned int segno)
312 {
313 	struct sit_info *sit_i = SIT_I(sbi);
314 	return &sit_i->sentries[segno];
315 }
316 
317 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
318 						unsigned int segno)
319 {
320 	struct sit_info *sit_i = SIT_I(sbi);
321 	return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
322 }
323 
324 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
325 				unsigned int segno, bool use_section)
326 {
327 	/*
328 	 * In order to get # of valid blocks in a section instantly from many
329 	 * segments, f2fs manages two counting structures separately.
330 	 */
331 	if (use_section && sbi->segs_per_sec > 1)
332 		return get_sec_entry(sbi, segno)->valid_blocks;
333 	else
334 		return get_seg_entry(sbi, segno)->valid_blocks;
335 }
336 
337 static inline void seg_info_from_raw_sit(struct seg_entry *se,
338 					struct f2fs_sit_entry *rs)
339 {
340 	se->valid_blocks = GET_SIT_VBLOCKS(rs);
341 	se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
342 	memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
343 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
344 #ifdef CONFIG_F2FS_CHECK_FS
345 	memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
346 #endif
347 	se->type = GET_SIT_TYPE(rs);
348 	se->mtime = le64_to_cpu(rs->mtime);
349 }
350 
351 static inline void seg_info_to_raw_sit(struct seg_entry *se,
352 					struct f2fs_sit_entry *rs)
353 {
354 	unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
355 					se->valid_blocks;
356 	rs->vblocks = cpu_to_le16(raw_vblocks);
357 	memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
358 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
359 	se->ckpt_valid_blocks = se->valid_blocks;
360 	rs->mtime = cpu_to_le64(se->mtime);
361 }
362 
363 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
364 		unsigned int max, unsigned int segno)
365 {
366 	unsigned int ret;
367 	spin_lock(&free_i->segmap_lock);
368 	ret = find_next_bit(free_i->free_segmap, max, segno);
369 	spin_unlock(&free_i->segmap_lock);
370 	return ret;
371 }
372 
373 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
374 {
375 	struct free_segmap_info *free_i = FREE_I(sbi);
376 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
377 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
378 	unsigned int next;
379 
380 	spin_lock(&free_i->segmap_lock);
381 	clear_bit(segno, free_i->free_segmap);
382 	free_i->free_segments++;
383 
384 	next = find_next_bit(free_i->free_segmap,
385 			start_segno + sbi->segs_per_sec, start_segno);
386 	if (next >= start_segno + sbi->segs_per_sec) {
387 		clear_bit(secno, free_i->free_secmap);
388 		free_i->free_sections++;
389 	}
390 	spin_unlock(&free_i->segmap_lock);
391 }
392 
393 static inline void __set_inuse(struct f2fs_sb_info *sbi,
394 		unsigned int segno)
395 {
396 	struct free_segmap_info *free_i = FREE_I(sbi);
397 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
398 
399 	set_bit(segno, free_i->free_segmap);
400 	free_i->free_segments--;
401 	if (!test_and_set_bit(secno, free_i->free_secmap))
402 		free_i->free_sections--;
403 }
404 
405 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
406 		unsigned int segno)
407 {
408 	struct free_segmap_info *free_i = FREE_I(sbi);
409 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
410 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
411 	unsigned int next;
412 
413 	spin_lock(&free_i->segmap_lock);
414 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
415 		free_i->free_segments++;
416 
417 		next = find_next_bit(free_i->free_segmap,
418 				start_segno + sbi->segs_per_sec, start_segno);
419 		if (next >= start_segno + sbi->segs_per_sec) {
420 			if (test_and_clear_bit(secno, free_i->free_secmap))
421 				free_i->free_sections++;
422 		}
423 	}
424 	spin_unlock(&free_i->segmap_lock);
425 }
426 
427 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
428 		unsigned int segno)
429 {
430 	struct free_segmap_info *free_i = FREE_I(sbi);
431 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
432 
433 	spin_lock(&free_i->segmap_lock);
434 	if (!test_and_set_bit(segno, free_i->free_segmap)) {
435 		free_i->free_segments--;
436 		if (!test_and_set_bit(secno, free_i->free_secmap))
437 			free_i->free_sections--;
438 	}
439 	spin_unlock(&free_i->segmap_lock);
440 }
441 
442 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
443 		void *dst_addr)
444 {
445 	struct sit_info *sit_i = SIT_I(sbi);
446 
447 #ifdef CONFIG_F2FS_CHECK_FS
448 	if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
449 						sit_i->bitmap_size))
450 		f2fs_bug_on(sbi, 1);
451 #endif
452 	memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
453 }
454 
455 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
456 {
457 	return SIT_I(sbi)->written_valid_blocks;
458 }
459 
460 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
461 {
462 	return FREE_I(sbi)->free_segments;
463 }
464 
465 static inline int reserved_segments(struct f2fs_sb_info *sbi)
466 {
467 	return SM_I(sbi)->reserved_segments;
468 }
469 
470 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
471 {
472 	return FREE_I(sbi)->free_sections;
473 }
474 
475 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
476 {
477 	return DIRTY_I(sbi)->nr_dirty[PRE];
478 }
479 
480 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
481 {
482 	return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
483 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
484 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
485 		DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
486 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
487 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
488 }
489 
490 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
491 {
492 	return SM_I(sbi)->ovp_segments;
493 }
494 
495 static inline int reserved_sections(struct f2fs_sb_info *sbi)
496 {
497 	return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
498 }
499 
500 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
501 {
502 	unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
503 					get_pages(sbi, F2FS_DIRTY_DENTS);
504 	unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
505 	unsigned int segno, left_blocks;
506 	int i;
507 
508 	/* check current node segment */
509 	for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
510 		segno = CURSEG_I(sbi, i)->segno;
511 		left_blocks = sbi->blocks_per_seg -
512 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
513 
514 		if (node_blocks > left_blocks)
515 			return false;
516 	}
517 
518 	/* check current data segment */
519 	segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
520 	left_blocks = sbi->blocks_per_seg -
521 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
522 	if (dent_blocks > left_blocks)
523 		return false;
524 	return true;
525 }
526 
527 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
528 					int freed, int needed)
529 {
530 	int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
531 	int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
532 	int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
533 
534 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
535 		return false;
536 
537 	if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
538 			has_curseg_enough_space(sbi))
539 		return false;
540 	return (free_sections(sbi) + freed) <=
541 		(node_secs + 2 * dent_secs + imeta_secs +
542 		reserved_sections(sbi) + needed);
543 }
544 
545 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
546 {
547 	return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
548 }
549 
550 static inline int utilization(struct f2fs_sb_info *sbi)
551 {
552 	return div_u64((u64)valid_user_blocks(sbi) * 100,
553 					sbi->user_block_count);
554 }
555 
556 /*
557  * Sometimes f2fs may be better to drop out-of-place update policy.
558  * And, users can control the policy through sysfs entries.
559  * There are five policies with triggering conditions as follows.
560  * F2FS_IPU_FORCE - all the time,
561  * F2FS_IPU_SSR - if SSR mode is activated,
562  * F2FS_IPU_UTIL - if FS utilization is over threashold,
563  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
564  *                     threashold,
565  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
566  *                     storages. IPU will be triggered only if the # of dirty
567  *                     pages over min_fsync_blocks.
568  * F2FS_IPUT_DISABLE - disable IPU. (=default option)
569  */
570 #define DEF_MIN_IPU_UTIL	70
571 #define DEF_MIN_FSYNC_BLOCKS	8
572 #define DEF_MIN_HOT_BLOCKS	16
573 
574 enum {
575 	F2FS_IPU_FORCE,
576 	F2FS_IPU_SSR,
577 	F2FS_IPU_UTIL,
578 	F2FS_IPU_SSR_UTIL,
579 	F2FS_IPU_FSYNC,
580 	F2FS_IPU_ASYNC,
581 };
582 
583 static inline bool need_inplace_update_policy(struct inode *inode,
584 				struct f2fs_io_info *fio)
585 {
586 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
587 	unsigned int policy = SM_I(sbi)->ipu_policy;
588 
589 	if (test_opt(sbi, LFS))
590 		return false;
591 
592 	/* if this is cold file, we should overwrite to avoid fragmentation */
593 	if (file_is_cold(inode))
594 		return true;
595 
596 	if (policy & (0x1 << F2FS_IPU_FORCE))
597 		return true;
598 	if (policy & (0x1 << F2FS_IPU_SSR) && need_SSR(sbi))
599 		return true;
600 	if (policy & (0x1 << F2FS_IPU_UTIL) &&
601 			utilization(sbi) > SM_I(sbi)->min_ipu_util)
602 		return true;
603 	if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && need_SSR(sbi) &&
604 			utilization(sbi) > SM_I(sbi)->min_ipu_util)
605 		return true;
606 
607 	/*
608 	 * IPU for rewrite async pages
609 	 */
610 	if (policy & (0x1 << F2FS_IPU_ASYNC) &&
611 			fio && fio->op == REQ_OP_WRITE &&
612 			!(fio->op_flags & REQ_SYNC) &&
613 			!f2fs_encrypted_inode(inode))
614 		return true;
615 
616 	/* this is only set during fdatasync */
617 	if (policy & (0x1 << F2FS_IPU_FSYNC) &&
618 			is_inode_flag_set(inode, FI_NEED_IPU))
619 		return true;
620 
621 	return false;
622 }
623 
624 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
625 		int type)
626 {
627 	struct curseg_info *curseg = CURSEG_I(sbi, type);
628 	return curseg->segno;
629 }
630 
631 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
632 		int type)
633 {
634 	struct curseg_info *curseg = CURSEG_I(sbi, type);
635 	return curseg->alloc_type;
636 }
637 
638 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
639 {
640 	struct curseg_info *curseg = CURSEG_I(sbi, type);
641 	return curseg->next_blkoff;
642 }
643 
644 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
645 {
646 	f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
647 }
648 
649 static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr)
650 {
651 	BUG_ON(blk_addr < SEG0_BLKADDR(sbi)
652 			|| blk_addr >= MAX_BLKADDR(sbi));
653 }
654 
655 /*
656  * Summary block is always treated as an invalid block
657  */
658 static inline void check_block_count(struct f2fs_sb_info *sbi,
659 		int segno, struct f2fs_sit_entry *raw_sit)
660 {
661 #ifdef CONFIG_F2FS_CHECK_FS
662 	bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
663 	int valid_blocks = 0;
664 	int cur_pos = 0, next_pos;
665 
666 	/* check bitmap with valid block count */
667 	do {
668 		if (is_valid) {
669 			next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
670 					sbi->blocks_per_seg,
671 					cur_pos);
672 			valid_blocks += next_pos - cur_pos;
673 		} else
674 			next_pos = find_next_bit_le(&raw_sit->valid_map,
675 					sbi->blocks_per_seg,
676 					cur_pos);
677 		cur_pos = next_pos;
678 		is_valid = !is_valid;
679 	} while (cur_pos < sbi->blocks_per_seg);
680 	BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks);
681 #endif
682 	/* check segment usage, and check boundary of a given segment number */
683 	f2fs_bug_on(sbi, GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
684 					|| segno > TOTAL_SEGS(sbi) - 1);
685 }
686 
687 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
688 						unsigned int start)
689 {
690 	struct sit_info *sit_i = SIT_I(sbi);
691 	unsigned int offset = SIT_BLOCK_OFFSET(start);
692 	block_t blk_addr = sit_i->sit_base_addr + offset;
693 
694 	check_seg_range(sbi, start);
695 
696 #ifdef CONFIG_F2FS_CHECK_FS
697 	if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
698 			f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
699 		f2fs_bug_on(sbi, 1);
700 #endif
701 
702 	/* calculate sit block address */
703 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
704 		blk_addr += sit_i->sit_blocks;
705 
706 	return blk_addr;
707 }
708 
709 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
710 						pgoff_t block_addr)
711 {
712 	struct sit_info *sit_i = SIT_I(sbi);
713 	block_addr -= sit_i->sit_base_addr;
714 	if (block_addr < sit_i->sit_blocks)
715 		block_addr += sit_i->sit_blocks;
716 	else
717 		block_addr -= sit_i->sit_blocks;
718 
719 	return block_addr + sit_i->sit_base_addr;
720 }
721 
722 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
723 {
724 	unsigned int block_off = SIT_BLOCK_OFFSET(start);
725 
726 	f2fs_change_bit(block_off, sit_i->sit_bitmap);
727 #ifdef CONFIG_F2FS_CHECK_FS
728 	f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
729 #endif
730 }
731 
732 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
733 {
734 	struct sit_info *sit_i = SIT_I(sbi);
735 	time64_t now = ktime_get_real_seconds();
736 
737 	return sit_i->elapsed_time + now - sit_i->mounted_time;
738 }
739 
740 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
741 			unsigned int ofs_in_node, unsigned char version)
742 {
743 	sum->nid = cpu_to_le32(nid);
744 	sum->ofs_in_node = cpu_to_le16(ofs_in_node);
745 	sum->version = version;
746 }
747 
748 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
749 {
750 	return __start_cp_addr(sbi) +
751 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
752 }
753 
754 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
755 {
756 	return __start_cp_addr(sbi) +
757 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
758 				- (base + 1) + type;
759 }
760 
761 static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi,
762 						unsigned int secno)
763 {
764 	if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >
765 						sbi->fggc_threshold)
766 		return true;
767 	return false;
768 }
769 
770 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
771 {
772 	if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
773 		return true;
774 	return false;
775 }
776 
777 /*
778  * It is very important to gather dirty pages and write at once, so that we can
779  * submit a big bio without interfering other data writes.
780  * By default, 512 pages for directory data,
781  * 512 pages (2MB) * 8 for nodes, and
782  * 256 pages * 8 for meta are set.
783  */
784 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
785 {
786 	if (sbi->sb->s_bdi->wb.dirty_exceeded)
787 		return 0;
788 
789 	if (type == DATA)
790 		return sbi->blocks_per_seg;
791 	else if (type == NODE)
792 		return 8 * sbi->blocks_per_seg;
793 	else if (type == META)
794 		return 8 * BIO_MAX_PAGES;
795 	else
796 		return 0;
797 }
798 
799 /*
800  * When writing pages, it'd better align nr_to_write for segment size.
801  */
802 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
803 					struct writeback_control *wbc)
804 {
805 	long nr_to_write, desired;
806 
807 	if (wbc->sync_mode != WB_SYNC_NONE)
808 		return 0;
809 
810 	nr_to_write = wbc->nr_to_write;
811 	desired = BIO_MAX_PAGES;
812 	if (type == NODE)
813 		desired <<= 1;
814 
815 	wbc->nr_to_write = desired;
816 	return desired - nr_to_write;
817 }
818 
819 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
820 {
821 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
822 	bool wakeup = false;
823 	int i;
824 
825 	if (force)
826 		goto wake_up;
827 
828 	mutex_lock(&dcc->cmd_lock);
829 	for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
830 		if (i + 1 < dcc->discard_granularity)
831 			break;
832 		if (!list_empty(&dcc->pend_list[i])) {
833 			wakeup = true;
834 			break;
835 		}
836 	}
837 	mutex_unlock(&dcc->cmd_lock);
838 	if (!wakeup)
839 		return;
840 wake_up:
841 	dcc->discard_wake = 1;
842 	wake_up_interruptible_all(&dcc->discard_wait_queue);
843 }
844