xref: /openbmc/linux/fs/f2fs/segment.h (revision 60772e48)
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 	rs->mtime = cpu_to_le64(se->mtime);
359 }
360 
361 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
362 				struct page *page, unsigned int start)
363 {
364 	struct f2fs_sit_block *raw_sit;
365 	struct seg_entry *se;
366 	struct f2fs_sit_entry *rs;
367 	unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
368 					(unsigned long)MAIN_SEGS(sbi));
369 	int i;
370 
371 	raw_sit = (struct f2fs_sit_block *)page_address(page);
372 	for (i = 0; i < end - start; i++) {
373 		rs = &raw_sit->entries[i];
374 		se = get_seg_entry(sbi, start + i);
375 		__seg_info_to_raw_sit(se, rs);
376 	}
377 }
378 
379 static inline void seg_info_to_raw_sit(struct seg_entry *se,
380 					struct f2fs_sit_entry *rs)
381 {
382 	__seg_info_to_raw_sit(se, rs);
383 
384 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
385 	se->ckpt_valid_blocks = se->valid_blocks;
386 }
387 
388 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
389 		unsigned int max, unsigned int segno)
390 {
391 	unsigned int ret;
392 	spin_lock(&free_i->segmap_lock);
393 	ret = find_next_bit(free_i->free_segmap, max, segno);
394 	spin_unlock(&free_i->segmap_lock);
395 	return ret;
396 }
397 
398 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
399 {
400 	struct free_segmap_info *free_i = FREE_I(sbi);
401 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
402 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
403 	unsigned int next;
404 
405 	spin_lock(&free_i->segmap_lock);
406 	clear_bit(segno, free_i->free_segmap);
407 	free_i->free_segments++;
408 
409 	next = find_next_bit(free_i->free_segmap,
410 			start_segno + sbi->segs_per_sec, start_segno);
411 	if (next >= start_segno + sbi->segs_per_sec) {
412 		clear_bit(secno, free_i->free_secmap);
413 		free_i->free_sections++;
414 	}
415 	spin_unlock(&free_i->segmap_lock);
416 }
417 
418 static inline void __set_inuse(struct f2fs_sb_info *sbi,
419 		unsigned int segno)
420 {
421 	struct free_segmap_info *free_i = FREE_I(sbi);
422 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
423 
424 	set_bit(segno, free_i->free_segmap);
425 	free_i->free_segments--;
426 	if (!test_and_set_bit(secno, free_i->free_secmap))
427 		free_i->free_sections--;
428 }
429 
430 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
431 		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 
438 	spin_lock(&free_i->segmap_lock);
439 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
440 		free_i->free_segments++;
441 
442 		next = find_next_bit(free_i->free_segmap,
443 				start_segno + sbi->segs_per_sec, start_segno);
444 		if (next >= start_segno + sbi->segs_per_sec) {
445 			if (test_and_clear_bit(secno, free_i->free_secmap))
446 				free_i->free_sections++;
447 		}
448 	}
449 	spin_unlock(&free_i->segmap_lock);
450 }
451 
452 static inline void __set_test_and_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 	spin_lock(&free_i->segmap_lock);
459 	if (!test_and_set_bit(segno, free_i->free_segmap)) {
460 		free_i->free_segments--;
461 		if (!test_and_set_bit(secno, free_i->free_secmap))
462 			free_i->free_sections--;
463 	}
464 	spin_unlock(&free_i->segmap_lock);
465 }
466 
467 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
468 		void *dst_addr)
469 {
470 	struct sit_info *sit_i = SIT_I(sbi);
471 
472 #ifdef CONFIG_F2FS_CHECK_FS
473 	if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
474 						sit_i->bitmap_size))
475 		f2fs_bug_on(sbi, 1);
476 #endif
477 	memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
478 }
479 
480 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
481 {
482 	return SIT_I(sbi)->written_valid_blocks;
483 }
484 
485 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
486 {
487 	return FREE_I(sbi)->free_segments;
488 }
489 
490 static inline int reserved_segments(struct f2fs_sb_info *sbi)
491 {
492 	return SM_I(sbi)->reserved_segments;
493 }
494 
495 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
496 {
497 	return FREE_I(sbi)->free_sections;
498 }
499 
500 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
501 {
502 	return DIRTY_I(sbi)->nr_dirty[PRE];
503 }
504 
505 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
506 {
507 	return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
508 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
509 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
510 		DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
511 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
512 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
513 }
514 
515 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
516 {
517 	return SM_I(sbi)->ovp_segments;
518 }
519 
520 static inline int reserved_sections(struct f2fs_sb_info *sbi)
521 {
522 	return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
523 }
524 
525 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
526 {
527 	unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
528 					get_pages(sbi, F2FS_DIRTY_DENTS);
529 	unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
530 	unsigned int segno, left_blocks;
531 	int i;
532 
533 	/* check current node segment */
534 	for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
535 		segno = CURSEG_I(sbi, i)->segno;
536 		left_blocks = sbi->blocks_per_seg -
537 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
538 
539 		if (node_blocks > left_blocks)
540 			return false;
541 	}
542 
543 	/* check current data segment */
544 	segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
545 	left_blocks = sbi->blocks_per_seg -
546 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
547 	if (dent_blocks > left_blocks)
548 		return false;
549 	return true;
550 }
551 
552 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
553 					int freed, int needed)
554 {
555 	int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
556 	int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
557 	int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
558 
559 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
560 		return false;
561 
562 	if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
563 			has_curseg_enough_space(sbi))
564 		return false;
565 	return (free_sections(sbi) + freed) <=
566 		(node_secs + 2 * dent_secs + imeta_secs +
567 		reserved_sections(sbi) + needed);
568 }
569 
570 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
571 {
572 	return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
573 }
574 
575 static inline int utilization(struct f2fs_sb_info *sbi)
576 {
577 	return div_u64((u64)valid_user_blocks(sbi) * 100,
578 					sbi->user_block_count);
579 }
580 
581 /*
582  * Sometimes f2fs may be better to drop out-of-place update policy.
583  * And, users can control the policy through sysfs entries.
584  * There are five policies with triggering conditions as follows.
585  * F2FS_IPU_FORCE - all the time,
586  * F2FS_IPU_SSR - if SSR mode is activated,
587  * F2FS_IPU_UTIL - if FS utilization is over threashold,
588  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
589  *                     threashold,
590  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
591  *                     storages. IPU will be triggered only if the # of dirty
592  *                     pages over min_fsync_blocks.
593  * F2FS_IPUT_DISABLE - disable IPU. (=default option)
594  */
595 #define DEF_MIN_IPU_UTIL	70
596 #define DEF_MIN_FSYNC_BLOCKS	8
597 #define DEF_MIN_HOT_BLOCKS	16
598 
599 enum {
600 	F2FS_IPU_FORCE,
601 	F2FS_IPU_SSR,
602 	F2FS_IPU_UTIL,
603 	F2FS_IPU_SSR_UTIL,
604 	F2FS_IPU_FSYNC,
605 	F2FS_IPU_ASYNC,
606 };
607 
608 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
609 		int type)
610 {
611 	struct curseg_info *curseg = CURSEG_I(sbi, type);
612 	return curseg->segno;
613 }
614 
615 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
616 		int type)
617 {
618 	struct curseg_info *curseg = CURSEG_I(sbi, type);
619 	return curseg->alloc_type;
620 }
621 
622 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
623 {
624 	struct curseg_info *curseg = CURSEG_I(sbi, type);
625 	return curseg->next_blkoff;
626 }
627 
628 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
629 {
630 	f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
631 }
632 
633 static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr)
634 {
635 	BUG_ON(blk_addr < SEG0_BLKADDR(sbi)
636 			|| blk_addr >= MAX_BLKADDR(sbi));
637 }
638 
639 /*
640  * Summary block is always treated as an invalid block
641  */
642 static inline int check_block_count(struct f2fs_sb_info *sbi,
643 		int segno, struct f2fs_sit_entry *raw_sit)
644 {
645 #ifdef CONFIG_F2FS_CHECK_FS
646 	bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
647 	int valid_blocks = 0;
648 	int cur_pos = 0, next_pos;
649 
650 	/* check bitmap with valid block count */
651 	do {
652 		if (is_valid) {
653 			next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
654 					sbi->blocks_per_seg,
655 					cur_pos);
656 			valid_blocks += next_pos - cur_pos;
657 		} else
658 			next_pos = find_next_bit_le(&raw_sit->valid_map,
659 					sbi->blocks_per_seg,
660 					cur_pos);
661 		cur_pos = next_pos;
662 		is_valid = !is_valid;
663 	} while (cur_pos < sbi->blocks_per_seg);
664 
665 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
666 		f2fs_msg(sbi->sb, KERN_ERR,
667 				"Mismatch valid blocks %d vs. %d",
668 					GET_SIT_VBLOCKS(raw_sit), valid_blocks);
669 		set_sbi_flag(sbi, SBI_NEED_FSCK);
670 		return -EINVAL;
671 	}
672 #endif
673 	/* check segment usage, and check boundary of a given segment number */
674 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
675 					|| segno > TOTAL_SEGS(sbi) - 1)) {
676 		f2fs_msg(sbi->sb, KERN_ERR,
677 				"Wrong valid blocks %d or segno %u",
678 					GET_SIT_VBLOCKS(raw_sit), segno);
679 		set_sbi_flag(sbi, SBI_NEED_FSCK);
680 		return -EINVAL;
681 	}
682 	return 0;
683 }
684 
685 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
686 						unsigned int start)
687 {
688 	struct sit_info *sit_i = SIT_I(sbi);
689 	unsigned int offset = SIT_BLOCK_OFFSET(start);
690 	block_t blk_addr = sit_i->sit_base_addr + offset;
691 
692 	check_seg_range(sbi, start);
693 
694 #ifdef CONFIG_F2FS_CHECK_FS
695 	if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
696 			f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
697 		f2fs_bug_on(sbi, 1);
698 #endif
699 
700 	/* calculate sit block address */
701 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
702 		blk_addr += sit_i->sit_blocks;
703 
704 	return blk_addr;
705 }
706 
707 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
708 						pgoff_t block_addr)
709 {
710 	struct sit_info *sit_i = SIT_I(sbi);
711 	block_addr -= sit_i->sit_base_addr;
712 	if (block_addr < sit_i->sit_blocks)
713 		block_addr += sit_i->sit_blocks;
714 	else
715 		block_addr -= sit_i->sit_blocks;
716 
717 	return block_addr + sit_i->sit_base_addr;
718 }
719 
720 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
721 {
722 	unsigned int block_off = SIT_BLOCK_OFFSET(start);
723 
724 	f2fs_change_bit(block_off, sit_i->sit_bitmap);
725 #ifdef CONFIG_F2FS_CHECK_FS
726 	f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
727 #endif
728 }
729 
730 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
731 {
732 	struct sit_info *sit_i = SIT_I(sbi);
733 	time64_t now = ktime_get_real_seconds();
734 
735 	return sit_i->elapsed_time + now - sit_i->mounted_time;
736 }
737 
738 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
739 			unsigned int ofs_in_node, unsigned char version)
740 {
741 	sum->nid = cpu_to_le32(nid);
742 	sum->ofs_in_node = cpu_to_le16(ofs_in_node);
743 	sum->version = version;
744 }
745 
746 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
747 {
748 	return __start_cp_addr(sbi) +
749 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
750 }
751 
752 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
753 {
754 	return __start_cp_addr(sbi) +
755 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
756 				- (base + 1) + type;
757 }
758 
759 static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi,
760 						unsigned int secno)
761 {
762 	if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >
763 						sbi->fggc_threshold)
764 		return true;
765 	return false;
766 }
767 
768 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
769 {
770 	if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
771 		return true;
772 	return false;
773 }
774 
775 /*
776  * It is very important to gather dirty pages and write at once, so that we can
777  * submit a big bio without interfering other data writes.
778  * By default, 512 pages for directory data,
779  * 512 pages (2MB) * 8 for nodes, and
780  * 256 pages * 8 for meta are set.
781  */
782 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
783 {
784 	if (sbi->sb->s_bdi->wb.dirty_exceeded)
785 		return 0;
786 
787 	if (type == DATA)
788 		return sbi->blocks_per_seg;
789 	else if (type == NODE)
790 		return 8 * sbi->blocks_per_seg;
791 	else if (type == META)
792 		return 8 * BIO_MAX_PAGES;
793 	else
794 		return 0;
795 }
796 
797 /*
798  * When writing pages, it'd better align nr_to_write for segment size.
799  */
800 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
801 					struct writeback_control *wbc)
802 {
803 	long nr_to_write, desired;
804 
805 	if (wbc->sync_mode != WB_SYNC_NONE)
806 		return 0;
807 
808 	nr_to_write = wbc->nr_to_write;
809 	desired = BIO_MAX_PAGES;
810 	if (type == NODE)
811 		desired <<= 1;
812 
813 	wbc->nr_to_write = desired;
814 	return desired - nr_to_write;
815 }
816 
817 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
818 {
819 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
820 	bool wakeup = false;
821 	int i;
822 
823 	if (force)
824 		goto wake_up;
825 
826 	mutex_lock(&dcc->cmd_lock);
827 	for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
828 		if (i + 1 < dcc->discard_granularity)
829 			break;
830 		if (!list_empty(&dcc->pend_list[i])) {
831 			wakeup = true;
832 			break;
833 		}
834 	}
835 	mutex_unlock(&dcc->cmd_lock);
836 	if (!wakeup)
837 		return;
838 wake_up:
839 	dcc->discard_wake = 1;
840 	wake_up_interruptible_all(&dcc->discard_wait_queue);
841 }
842