xref: /openbmc/linux/fs/f2fs/segment.h (revision 8781e5df)
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 
20 /* L: Logical segment # in volume, R: Relative segment # in main area */
21 #define GET_L2R_SEGNO(free_i, segno)	((segno) - (free_i)->start_segno)
22 #define GET_R2L_SEGNO(free_i, segno)	((segno) + (free_i)->start_segno)
23 
24 #define IS_DATASEG(t)	((t) <= CURSEG_COLD_DATA)
25 #define IS_NODESEG(t)	((t) >= CURSEG_HOT_NODE)
26 
27 #define IS_HOT(t)	((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
28 #define IS_WARM(t)	((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
29 #define IS_COLD(t)	((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
30 
31 #define IS_CURSEG(sbi, seg)						\
32 	(((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) ||	\
33 	 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) ||	\
34 	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) ||	\
35 	 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) ||	\
36 	 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) ||	\
37 	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
38 
39 #define IS_CURSEC(sbi, secno)						\
40 	(((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno /		\
41 	  (sbi)->segs_per_sec) ||	\
42 	 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno /		\
43 	  (sbi)->segs_per_sec) ||	\
44 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno /		\
45 	  (sbi)->segs_per_sec) ||	\
46 	 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno /		\
47 	  (sbi)->segs_per_sec) ||	\
48 	 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno /		\
49 	  (sbi)->segs_per_sec) ||	\
50 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno /		\
51 	  (sbi)->segs_per_sec))	\
52 
53 #define MAIN_BLKADDR(sbi)						\
54 	(SM_I(sbi) ? SM_I(sbi)->main_blkaddr : 				\
55 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
56 #define SEG0_BLKADDR(sbi)						\
57 	(SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : 				\
58 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
59 
60 #define MAIN_SEGS(sbi)	(SM_I(sbi)->main_segments)
61 #define MAIN_SECS(sbi)	((sbi)->total_sections)
62 
63 #define TOTAL_SEGS(sbi)							\
64 	(SM_I(sbi) ? SM_I(sbi)->segment_count : 				\
65 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
66 #define TOTAL_BLKS(sbi)	(TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
67 
68 #define MAX_BLKADDR(sbi)	(SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
69 #define SEGMENT_SIZE(sbi)	(1ULL << ((sbi)->log_blocksize +	\
70 					(sbi)->log_blocks_per_seg))
71 
72 #define START_BLOCK(sbi, segno)	(SEG0_BLKADDR(sbi) +			\
73 	 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
74 
75 #define NEXT_FREE_BLKADDR(sbi, curseg)					\
76 	(START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
77 
78 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr)	((blk_addr) - SEG0_BLKADDR(sbi))
79 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr)				\
80 	(GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
81 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr)				\
82 	(GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
83 
84 #define GET_SEGNO(sbi, blk_addr)					\
85 	((!__is_valid_data_blkaddr(blk_addr)) ?			\
86 	NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi),			\
87 		GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
88 #define BLKS_PER_SEC(sbi)					\
89 	((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
90 #define GET_SEC_FROM_SEG(sbi, segno)				\
91 	((segno) / (sbi)->segs_per_sec)
92 #define GET_SEG_FROM_SEC(sbi, secno)				\
93 	((secno) * (sbi)->segs_per_sec)
94 #define GET_ZONE_FROM_SEC(sbi, secno)				\
95 	((secno) / (sbi)->secs_per_zone)
96 #define GET_ZONE_FROM_SEG(sbi, segno)				\
97 	GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
98 
99 #define GET_SUM_BLOCK(sbi, segno)				\
100 	((sbi)->sm_info->ssa_blkaddr + (segno))
101 
102 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
103 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
104 
105 #define SIT_ENTRY_OFFSET(sit_i, segno)					\
106 	((segno) % (sit_i)->sents_per_block)
107 #define SIT_BLOCK_OFFSET(segno)					\
108 	((segno) / SIT_ENTRY_PER_BLOCK)
109 #define	START_SEGNO(segno)		\
110 	(SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
111 #define SIT_BLK_CNT(sbi)			\
112 	DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
113 #define f2fs_bitmap_size(nr)			\
114 	(BITS_TO_LONGS(nr) * sizeof(unsigned long))
115 
116 #define SECTOR_FROM_BLOCK(blk_addr)					\
117 	(((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
118 #define SECTOR_TO_BLOCK(sectors)					\
119 	((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
120 
121 /*
122  * indicate a block allocation direction: RIGHT and LEFT.
123  * RIGHT means allocating new sections towards the end of volume.
124  * LEFT means the opposite direction.
125  */
126 enum {
127 	ALLOC_RIGHT = 0,
128 	ALLOC_LEFT
129 };
130 
131 /*
132  * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
133  * LFS writes data sequentially with cleaning operations.
134  * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
135  */
136 enum {
137 	LFS = 0,
138 	SSR
139 };
140 
141 /*
142  * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
143  * GC_CB is based on cost-benefit algorithm.
144  * GC_GREEDY is based on greedy algorithm.
145  */
146 enum {
147 	GC_CB = 0,
148 	GC_GREEDY,
149 	ALLOC_NEXT,
150 	FLUSH_DEVICE,
151 	MAX_GC_POLICY,
152 };
153 
154 /*
155  * BG_GC means the background cleaning job.
156  * FG_GC means the on-demand cleaning job.
157  * FORCE_FG_GC means on-demand cleaning job in background.
158  */
159 enum {
160 	BG_GC = 0,
161 	FG_GC,
162 	FORCE_FG_GC,
163 };
164 
165 /* for a function parameter to select a victim segment */
166 struct victim_sel_policy {
167 	int alloc_mode;			/* LFS or SSR */
168 	int gc_mode;			/* GC_CB or GC_GREEDY */
169 	unsigned long *dirty_segmap;	/* dirty segment bitmap */
170 	unsigned int max_search;	/* maximum # of segments to search */
171 	unsigned int offset;		/* last scanned bitmap offset */
172 	unsigned int ofs_unit;		/* bitmap search unit */
173 	unsigned int min_cost;		/* minimum cost */
174 	unsigned int min_segno;		/* segment # having min. cost */
175 };
176 
177 struct seg_entry {
178 	unsigned int type:6;		/* segment type like CURSEG_XXX_TYPE */
179 	unsigned int valid_blocks:10;	/* # of valid blocks */
180 	unsigned int ckpt_valid_blocks:10;	/* # of valid blocks last cp */
181 	unsigned int padding:6;		/* padding */
182 	unsigned char *cur_valid_map;	/* validity bitmap of blocks */
183 #ifdef CONFIG_F2FS_CHECK_FS
184 	unsigned char *cur_valid_map_mir;	/* mirror of current valid bitmap */
185 #endif
186 	/*
187 	 * # of valid blocks and the validity bitmap stored in the the last
188 	 * checkpoint pack. This information is used by the SSR mode.
189 	 */
190 	unsigned char *ckpt_valid_map;	/* validity bitmap of blocks last cp */
191 	unsigned char *discard_map;
192 	unsigned long long mtime;	/* modification time of the segment */
193 };
194 
195 struct sec_entry {
196 	unsigned int valid_blocks;	/* # of valid blocks in a section */
197 };
198 
199 struct segment_allocation {
200 	void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
201 };
202 
203 /*
204  * this value is set in page as a private data which indicate that
205  * the page is atomically written, and it is in inmem_pages list.
206  */
207 #define ATOMIC_WRITTEN_PAGE		((unsigned long)-1)
208 #define DUMMY_WRITTEN_PAGE		((unsigned long)-2)
209 
210 #define IS_ATOMIC_WRITTEN_PAGE(page)			\
211 		(page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
212 #define IS_DUMMY_WRITTEN_PAGE(page)			\
213 		(page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
214 
215 #define MAX_SKIP_GC_COUNT			16
216 
217 struct inmem_pages {
218 	struct list_head list;
219 	struct page *page;
220 	block_t old_addr;		/* for revoking when fail to commit */
221 };
222 
223 struct sit_info {
224 	const struct segment_allocation *s_ops;
225 
226 	block_t sit_base_addr;		/* start block address of SIT area */
227 	block_t sit_blocks;		/* # of blocks used by SIT area */
228 	block_t written_valid_blocks;	/* # of valid blocks in main area */
229 	char *bitmap;			/* all bitmaps pointer */
230 	char *sit_bitmap;		/* SIT bitmap pointer */
231 #ifdef CONFIG_F2FS_CHECK_FS
232 	char *sit_bitmap_mir;		/* SIT bitmap mirror */
233 
234 	/* bitmap of segments to be ignored by GC in case of errors */
235 	unsigned long *invalid_segmap;
236 #endif
237 	unsigned int bitmap_size;	/* SIT bitmap size */
238 
239 	unsigned long *tmp_map;			/* bitmap for temporal use */
240 	unsigned long *dirty_sentries_bitmap;	/* bitmap for dirty sentries */
241 	unsigned int dirty_sentries;		/* # of dirty sentries */
242 	unsigned int sents_per_block;		/* # of SIT entries per block */
243 	struct rw_semaphore sentry_lock;	/* to protect SIT cache */
244 	struct seg_entry *sentries;		/* SIT segment-level cache */
245 	struct sec_entry *sec_entries;		/* SIT section-level cache */
246 
247 	/* for cost-benefit algorithm in cleaning procedure */
248 	unsigned long long elapsed_time;	/* elapsed time after mount */
249 	unsigned long long mounted_time;	/* mount time */
250 	unsigned long long min_mtime;		/* min. modification time */
251 	unsigned long long max_mtime;		/* max. modification time */
252 
253 	unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
254 };
255 
256 struct free_segmap_info {
257 	unsigned int start_segno;	/* start segment number logically */
258 	unsigned int free_segments;	/* # of free segments */
259 	unsigned int free_sections;	/* # of free sections */
260 	spinlock_t segmap_lock;		/* free segmap lock */
261 	unsigned long *free_segmap;	/* free segment bitmap */
262 	unsigned long *free_secmap;	/* free section bitmap */
263 };
264 
265 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
266 enum dirty_type {
267 	DIRTY_HOT_DATA,		/* dirty segments assigned as hot data logs */
268 	DIRTY_WARM_DATA,	/* dirty segments assigned as warm data logs */
269 	DIRTY_COLD_DATA,	/* dirty segments assigned as cold data logs */
270 	DIRTY_HOT_NODE,		/* dirty segments assigned as hot node logs */
271 	DIRTY_WARM_NODE,	/* dirty segments assigned as warm node logs */
272 	DIRTY_COLD_NODE,	/* dirty segments assigned as cold node logs */
273 	DIRTY,			/* to count # of dirty segments */
274 	PRE,			/* to count # of entirely obsolete segments */
275 	NR_DIRTY_TYPE
276 };
277 
278 struct dirty_seglist_info {
279 	const struct victim_selection *v_ops;	/* victim selction operation */
280 	unsigned long *dirty_segmap[NR_DIRTY_TYPE];
281 	struct mutex seglist_lock;		/* lock for segment bitmaps */
282 	int nr_dirty[NR_DIRTY_TYPE];		/* # of dirty segments */
283 	unsigned long *victim_secmap;		/* background GC victims */
284 };
285 
286 /* victim selection function for cleaning and SSR */
287 struct victim_selection {
288 	int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
289 							int, int, char);
290 };
291 
292 /* for active log information */
293 struct curseg_info {
294 	struct mutex curseg_mutex;		/* lock for consistency */
295 	struct f2fs_summary_block *sum_blk;	/* cached summary block */
296 	struct rw_semaphore journal_rwsem;	/* protect journal area */
297 	struct f2fs_journal *journal;		/* cached journal info */
298 	unsigned char alloc_type;		/* current allocation type */
299 	unsigned int segno;			/* current segment number */
300 	unsigned short next_blkoff;		/* next block offset to write */
301 	unsigned int zone;			/* current zone number */
302 	unsigned int next_segno;		/* preallocated segment */
303 };
304 
305 struct sit_entry_set {
306 	struct list_head set_list;	/* link with all sit sets */
307 	unsigned int start_segno;	/* start segno of sits in set */
308 	unsigned int entry_cnt;		/* the # of sit entries in set */
309 };
310 
311 /*
312  * inline functions
313  */
314 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
315 {
316 	return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
317 }
318 
319 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
320 						unsigned int segno)
321 {
322 	struct sit_info *sit_i = SIT_I(sbi);
323 	return &sit_i->sentries[segno];
324 }
325 
326 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
327 						unsigned int segno)
328 {
329 	struct sit_info *sit_i = SIT_I(sbi);
330 	return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
331 }
332 
333 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
334 				unsigned int segno, bool use_section)
335 {
336 	/*
337 	 * In order to get # of valid blocks in a section instantly from many
338 	 * segments, f2fs manages two counting structures separately.
339 	 */
340 	if (use_section && __is_large_section(sbi))
341 		return get_sec_entry(sbi, segno)->valid_blocks;
342 	else
343 		return get_seg_entry(sbi, segno)->valid_blocks;
344 }
345 
346 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
347 				unsigned int segno)
348 {
349 	return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
350 }
351 
352 static inline void seg_info_from_raw_sit(struct seg_entry *se,
353 					struct f2fs_sit_entry *rs)
354 {
355 	se->valid_blocks = GET_SIT_VBLOCKS(rs);
356 	se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
357 	memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
358 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
359 #ifdef CONFIG_F2FS_CHECK_FS
360 	memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
361 #endif
362 	se->type = GET_SIT_TYPE(rs);
363 	se->mtime = le64_to_cpu(rs->mtime);
364 }
365 
366 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
367 					struct f2fs_sit_entry *rs)
368 {
369 	unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
370 					se->valid_blocks;
371 	rs->vblocks = cpu_to_le16(raw_vblocks);
372 	memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
373 	rs->mtime = cpu_to_le64(se->mtime);
374 }
375 
376 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
377 				struct page *page, unsigned int start)
378 {
379 	struct f2fs_sit_block *raw_sit;
380 	struct seg_entry *se;
381 	struct f2fs_sit_entry *rs;
382 	unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
383 					(unsigned long)MAIN_SEGS(sbi));
384 	int i;
385 
386 	raw_sit = (struct f2fs_sit_block *)page_address(page);
387 	memset(raw_sit, 0, PAGE_SIZE);
388 	for (i = 0; i < end - start; i++) {
389 		rs = &raw_sit->entries[i];
390 		se = get_seg_entry(sbi, start + i);
391 		__seg_info_to_raw_sit(se, rs);
392 	}
393 }
394 
395 static inline void seg_info_to_raw_sit(struct seg_entry *se,
396 					struct f2fs_sit_entry *rs)
397 {
398 	__seg_info_to_raw_sit(se, rs);
399 
400 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
401 	se->ckpt_valid_blocks = se->valid_blocks;
402 }
403 
404 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
405 		unsigned int max, unsigned int segno)
406 {
407 	unsigned int ret;
408 	spin_lock(&free_i->segmap_lock);
409 	ret = find_next_bit(free_i->free_segmap, max, segno);
410 	spin_unlock(&free_i->segmap_lock);
411 	return ret;
412 }
413 
414 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
415 {
416 	struct free_segmap_info *free_i = FREE_I(sbi);
417 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
418 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
419 	unsigned int next;
420 
421 	spin_lock(&free_i->segmap_lock);
422 	clear_bit(segno, free_i->free_segmap);
423 	free_i->free_segments++;
424 
425 	next = find_next_bit(free_i->free_segmap,
426 			start_segno + sbi->segs_per_sec, start_segno);
427 	if (next >= start_segno + sbi->segs_per_sec) {
428 		clear_bit(secno, free_i->free_secmap);
429 		free_i->free_sections++;
430 	}
431 	spin_unlock(&free_i->segmap_lock);
432 }
433 
434 static inline void __set_inuse(struct f2fs_sb_info *sbi,
435 		unsigned int segno)
436 {
437 	struct free_segmap_info *free_i = FREE_I(sbi);
438 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
439 
440 	set_bit(segno, free_i->free_segmap);
441 	free_i->free_segments--;
442 	if (!test_and_set_bit(secno, free_i->free_secmap))
443 		free_i->free_sections--;
444 }
445 
446 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
447 		unsigned int segno)
448 {
449 	struct free_segmap_info *free_i = FREE_I(sbi);
450 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
451 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
452 	unsigned int next;
453 
454 	spin_lock(&free_i->segmap_lock);
455 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
456 		free_i->free_segments++;
457 
458 		if (IS_CURSEC(sbi, secno))
459 			goto skip_free;
460 		next = find_next_bit(free_i->free_segmap,
461 				start_segno + sbi->segs_per_sec, start_segno);
462 		if (next >= start_segno + sbi->segs_per_sec) {
463 			if (test_and_clear_bit(secno, free_i->free_secmap))
464 				free_i->free_sections++;
465 		}
466 	}
467 skip_free:
468 	spin_unlock(&free_i->segmap_lock);
469 }
470 
471 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
472 		unsigned int segno)
473 {
474 	struct free_segmap_info *free_i = FREE_I(sbi);
475 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
476 
477 	spin_lock(&free_i->segmap_lock);
478 	if (!test_and_set_bit(segno, free_i->free_segmap)) {
479 		free_i->free_segments--;
480 		if (!test_and_set_bit(secno, free_i->free_secmap))
481 			free_i->free_sections--;
482 	}
483 	spin_unlock(&free_i->segmap_lock);
484 }
485 
486 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
487 		void *dst_addr)
488 {
489 	struct sit_info *sit_i = SIT_I(sbi);
490 
491 #ifdef CONFIG_F2FS_CHECK_FS
492 	if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
493 						sit_i->bitmap_size))
494 		f2fs_bug_on(sbi, 1);
495 #endif
496 	memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
497 }
498 
499 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
500 {
501 	return SIT_I(sbi)->written_valid_blocks;
502 }
503 
504 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
505 {
506 	return FREE_I(sbi)->free_segments;
507 }
508 
509 static inline int reserved_segments(struct f2fs_sb_info *sbi)
510 {
511 	return SM_I(sbi)->reserved_segments;
512 }
513 
514 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
515 {
516 	return FREE_I(sbi)->free_sections;
517 }
518 
519 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
520 {
521 	return DIRTY_I(sbi)->nr_dirty[PRE];
522 }
523 
524 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
525 {
526 	return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
527 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
528 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
529 		DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
530 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
531 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
532 }
533 
534 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
535 {
536 	return SM_I(sbi)->ovp_segments;
537 }
538 
539 static inline int reserved_sections(struct f2fs_sb_info *sbi)
540 {
541 	return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
542 }
543 
544 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
545 {
546 	unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
547 					get_pages(sbi, F2FS_DIRTY_DENTS);
548 	unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
549 	unsigned int segno, left_blocks;
550 	int i;
551 
552 	/* check current node segment */
553 	for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
554 		segno = CURSEG_I(sbi, i)->segno;
555 		left_blocks = sbi->blocks_per_seg -
556 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
557 
558 		if (node_blocks > left_blocks)
559 			return false;
560 	}
561 
562 	/* check current data segment */
563 	segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
564 	left_blocks = sbi->blocks_per_seg -
565 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
566 	if (dent_blocks > left_blocks)
567 		return false;
568 	return true;
569 }
570 
571 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
572 					int freed, int needed)
573 {
574 	int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
575 	int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
576 	int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
577 
578 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
579 		return false;
580 
581 	if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
582 			has_curseg_enough_space(sbi))
583 		return false;
584 	return (free_sections(sbi) + freed) <=
585 		(node_secs + 2 * dent_secs + imeta_secs +
586 		reserved_sections(sbi) + needed);
587 }
588 
589 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
590 {
591 	if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
592 		return true;
593 	if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
594 		return true;
595 	return false;
596 }
597 
598 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
599 {
600 	return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
601 }
602 
603 static inline int utilization(struct f2fs_sb_info *sbi)
604 {
605 	return div_u64((u64)valid_user_blocks(sbi) * 100,
606 					sbi->user_block_count);
607 }
608 
609 /*
610  * Sometimes f2fs may be better to drop out-of-place update policy.
611  * And, users can control the policy through sysfs entries.
612  * There are five policies with triggering conditions as follows.
613  * F2FS_IPU_FORCE - all the time,
614  * F2FS_IPU_SSR - if SSR mode is activated,
615  * F2FS_IPU_UTIL - if FS utilization is over threashold,
616  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
617  *                     threashold,
618  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
619  *                     storages. IPU will be triggered only if the # of dirty
620  *                     pages over min_fsync_blocks.
621  * F2FS_IPUT_DISABLE - disable IPU. (=default option)
622  */
623 #define DEF_MIN_IPU_UTIL	70
624 #define DEF_MIN_FSYNC_BLOCKS	8
625 #define DEF_MIN_HOT_BLOCKS	16
626 
627 #define SMALL_VOLUME_SEGMENTS	(16 * 512)	/* 16GB */
628 
629 enum {
630 	F2FS_IPU_FORCE,
631 	F2FS_IPU_SSR,
632 	F2FS_IPU_UTIL,
633 	F2FS_IPU_SSR_UTIL,
634 	F2FS_IPU_FSYNC,
635 	F2FS_IPU_ASYNC,
636 };
637 
638 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
639 		int type)
640 {
641 	struct curseg_info *curseg = CURSEG_I(sbi, type);
642 	return curseg->segno;
643 }
644 
645 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
646 		int type)
647 {
648 	struct curseg_info *curseg = CURSEG_I(sbi, type);
649 	return curseg->alloc_type;
650 }
651 
652 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
653 {
654 	struct curseg_info *curseg = CURSEG_I(sbi, type);
655 	return curseg->next_blkoff;
656 }
657 
658 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
659 {
660 	f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
661 }
662 
663 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
664 {
665 	struct f2fs_sb_info *sbi = fio->sbi;
666 
667 	if (__is_valid_data_blkaddr(fio->old_blkaddr))
668 		verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
669 					META_GENERIC : DATA_GENERIC);
670 	verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
671 					META_GENERIC : DATA_GENERIC_ENHANCE);
672 }
673 
674 /*
675  * Summary block is always treated as an invalid block
676  */
677 static inline int check_block_count(struct f2fs_sb_info *sbi,
678 		int segno, struct f2fs_sit_entry *raw_sit)
679 {
680 	bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
681 	int valid_blocks = 0;
682 	int cur_pos = 0, next_pos;
683 
684 	/* check bitmap with valid block count */
685 	do {
686 		if (is_valid) {
687 			next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
688 					sbi->blocks_per_seg,
689 					cur_pos);
690 			valid_blocks += next_pos - cur_pos;
691 		} else
692 			next_pos = find_next_bit_le(&raw_sit->valid_map,
693 					sbi->blocks_per_seg,
694 					cur_pos);
695 		cur_pos = next_pos;
696 		is_valid = !is_valid;
697 	} while (cur_pos < sbi->blocks_per_seg);
698 
699 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
700 		f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
701 			 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
702 		set_sbi_flag(sbi, SBI_NEED_FSCK);
703 		return -EFSCORRUPTED;
704 	}
705 
706 	/* check segment usage, and check boundary of a given segment number */
707 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
708 					|| segno > TOTAL_SEGS(sbi) - 1)) {
709 		f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
710 			 GET_SIT_VBLOCKS(raw_sit), segno);
711 		set_sbi_flag(sbi, SBI_NEED_FSCK);
712 		return -EFSCORRUPTED;
713 	}
714 	return 0;
715 }
716 
717 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
718 						unsigned int start)
719 {
720 	struct sit_info *sit_i = SIT_I(sbi);
721 	unsigned int offset = SIT_BLOCK_OFFSET(start);
722 	block_t blk_addr = sit_i->sit_base_addr + offset;
723 
724 	check_seg_range(sbi, start);
725 
726 #ifdef CONFIG_F2FS_CHECK_FS
727 	if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
728 			f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
729 		f2fs_bug_on(sbi, 1);
730 #endif
731 
732 	/* calculate sit block address */
733 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
734 		blk_addr += sit_i->sit_blocks;
735 
736 	return blk_addr;
737 }
738 
739 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
740 						pgoff_t block_addr)
741 {
742 	struct sit_info *sit_i = SIT_I(sbi);
743 	block_addr -= sit_i->sit_base_addr;
744 	if (block_addr < sit_i->sit_blocks)
745 		block_addr += sit_i->sit_blocks;
746 	else
747 		block_addr -= sit_i->sit_blocks;
748 
749 	return block_addr + sit_i->sit_base_addr;
750 }
751 
752 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
753 {
754 	unsigned int block_off = SIT_BLOCK_OFFSET(start);
755 
756 	f2fs_change_bit(block_off, sit_i->sit_bitmap);
757 #ifdef CONFIG_F2FS_CHECK_FS
758 	f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
759 #endif
760 }
761 
762 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
763 						bool base_time)
764 {
765 	struct sit_info *sit_i = SIT_I(sbi);
766 	time64_t diff, now = ktime_get_real_seconds();
767 
768 	if (now >= sit_i->mounted_time)
769 		return sit_i->elapsed_time + now - sit_i->mounted_time;
770 
771 	/* system time is set to the past */
772 	if (!base_time) {
773 		diff = sit_i->mounted_time - now;
774 		if (sit_i->elapsed_time >= diff)
775 			return sit_i->elapsed_time - diff;
776 		return 0;
777 	}
778 	return sit_i->elapsed_time;
779 }
780 
781 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
782 			unsigned int ofs_in_node, unsigned char version)
783 {
784 	sum->nid = cpu_to_le32(nid);
785 	sum->ofs_in_node = cpu_to_le16(ofs_in_node);
786 	sum->version = version;
787 }
788 
789 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
790 {
791 	return __start_cp_addr(sbi) +
792 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
793 }
794 
795 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
796 {
797 	return __start_cp_addr(sbi) +
798 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
799 				- (base + 1) + type;
800 }
801 
802 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
803 {
804 	if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
805 		return true;
806 	return false;
807 }
808 
809 /*
810  * It is very important to gather dirty pages and write at once, so that we can
811  * submit a big bio without interfering other data writes.
812  * By default, 512 pages for directory data,
813  * 512 pages (2MB) * 8 for nodes, and
814  * 256 pages * 8 for meta are set.
815  */
816 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
817 {
818 	if (sbi->sb->s_bdi->wb.dirty_exceeded)
819 		return 0;
820 
821 	if (type == DATA)
822 		return sbi->blocks_per_seg;
823 	else if (type == NODE)
824 		return 8 * sbi->blocks_per_seg;
825 	else if (type == META)
826 		return 8 * BIO_MAX_PAGES;
827 	else
828 		return 0;
829 }
830 
831 /*
832  * When writing pages, it'd better align nr_to_write for segment size.
833  */
834 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
835 					struct writeback_control *wbc)
836 {
837 	long nr_to_write, desired;
838 
839 	if (wbc->sync_mode != WB_SYNC_NONE)
840 		return 0;
841 
842 	nr_to_write = wbc->nr_to_write;
843 	desired = BIO_MAX_PAGES;
844 	if (type == NODE)
845 		desired <<= 1;
846 
847 	wbc->nr_to_write = desired;
848 	return desired - nr_to_write;
849 }
850 
851 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
852 {
853 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
854 	bool wakeup = false;
855 	int i;
856 
857 	if (force)
858 		goto wake_up;
859 
860 	mutex_lock(&dcc->cmd_lock);
861 	for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
862 		if (i + 1 < dcc->discard_granularity)
863 			break;
864 		if (!list_empty(&dcc->pend_list[i])) {
865 			wakeup = true;
866 			break;
867 		}
868 	}
869 	mutex_unlock(&dcc->cmd_lock);
870 	if (!wakeup || !is_idle(sbi, DISCARD_TIME))
871 		return;
872 wake_up:
873 	dcc->discard_wake = 1;
874 	wake_up_interruptible_all(&dcc->discard_wait_queue);
875 }
876