• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 
sanity_check_seg_type(struct f2fs_sb_info * sbi,unsigned short seg_type)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 #ifdef CONFIG_F2FS_GRADING_SSR
134 #define KBS_PER_SEGMENT 2048
135 #define SSR_MIN_BLKS_LIMIT (16 << 18)	/* 16G */
136 #define SSR_CONTIG_DIRTY_NUMS	32	/* Dirty pages for LFS alloction in grading ssr. */
137 #define SSR_CONTIG_LARGE	256	/* Larege files */
138 #endif
139 
140 enum {
141 	SEQ_NONE,
142 	SEQ_32BLKS,
143 	SEQ_256BLKS
144 };
145 /*
146  * indicate a block allocation direction: RIGHT and LEFT.
147  * RIGHT means allocating new sections towards the end of volume.
148  * LEFT means the opposite direction.
149  */
150 enum {
151 	ALLOC_RIGHT = 0,
152 	ALLOC_LEFT
153 };
154 
155 /*
156  * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
157  * LFS writes data sequentially with cleaning operations.
158  * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
159  * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
160  * fragmented segment which has similar aging degree.
161  */
162 enum {
163 	LFS = 0,
164 	SSR,
165 	AT_SSR,
166 };
167 
168 /*
169  * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
170  * GC_CB is based on cost-benefit algorithm.
171  * GC_GREEDY is based on greedy algorithm.
172  * GC_AT is based on age-threshold algorithm.
173  */
174 enum {
175 	GC_CB = 0,
176 	GC_GREEDY,
177 	GC_AT,
178 	ALLOC_NEXT,
179 	FLUSH_DEVICE,
180 	MAX_GC_POLICY,
181 };
182 
183 /*
184  * BG_GC means the background cleaning job.
185  * FG_GC means the on-demand cleaning job.
186  * FORCE_FG_GC means on-demand cleaning job in background.
187  */
188 enum {
189 	BG_GC = 0,
190 	FG_GC,
191 	FORCE_FG_GC,
192 };
193 
194 #ifdef CONFIG_F2FS_GRADING_SSR
195 enum {
196 	GRADING_SSR_OFF = 0,
197 	GRADING_SSR_ON
198 };
199 #endif
200 
201 /* for a function parameter to select a victim segment */
202 struct victim_sel_policy {
203 	int alloc_mode;			/* LFS or SSR */
204 	int gc_mode;			/* GC_CB or GC_GREEDY */
205 	unsigned long *dirty_bitmap;	/* dirty segment/section bitmap */
206 	unsigned int max_search;	/*
207 					 * maximum # of segments/sections
208 					 * to search
209 					 */
210 	unsigned int offset;		/* last scanned bitmap offset */
211 	unsigned int ofs_unit;		/* bitmap search unit */
212 	unsigned int min_cost;		/* minimum cost */
213 	unsigned long long oldest_age;	/* oldest age of segments having the same min cost */
214 	unsigned int min_segno;		/* segment # having min. cost */
215 	unsigned long long age;		/* mtime of GCed section*/
216 	unsigned long long age_threshold;/* age threshold */
217 };
218 
219 struct seg_entry {
220 	unsigned int type:6;		/* segment type like CURSEG_XXX_TYPE */
221 	unsigned int valid_blocks:10;	/* # of valid blocks */
222 	unsigned int ckpt_valid_blocks:10;	/* # of valid blocks last cp */
223 	unsigned int padding:6;		/* padding */
224 	unsigned char *cur_valid_map;	/* validity bitmap of blocks */
225 #ifdef CONFIG_F2FS_CHECK_FS
226 	unsigned char *cur_valid_map_mir;	/* mirror of current valid bitmap */
227 #endif
228 	/*
229 	 * # of valid blocks and the validity bitmap stored in the last
230 	 * checkpoint pack. This information is used by the SSR mode.
231 	 */
232 	unsigned char *ckpt_valid_map;	/* validity bitmap of blocks last cp */
233 	unsigned char *discard_map;
234 	unsigned long long mtime;	/* modification time of the segment */
235 };
236 
237 struct sec_entry {
238 	unsigned int valid_blocks;	/* # of valid blocks in a section */
239 };
240 
241 struct segment_allocation {
242 	void (*allocate_segment)(struct f2fs_sb_info *, int, bool, int);
243 };
244 
245 #define MAX_SKIP_GC_COUNT			16
246 
247 struct inmem_pages {
248 	struct list_head list;
249 	struct page *page;
250 	block_t old_addr;		/* for revoking when fail to commit */
251 };
252 
253 struct sit_info {
254 	const struct segment_allocation *s_ops;
255 
256 	block_t sit_base_addr;		/* start block address of SIT area */
257 	block_t sit_blocks;		/* # of blocks used by SIT area */
258 	block_t written_valid_blocks;	/* # of valid blocks in main area */
259 	char *bitmap;			/* all bitmaps pointer */
260 	char *sit_bitmap;		/* SIT bitmap pointer */
261 #ifdef CONFIG_F2FS_CHECK_FS
262 	char *sit_bitmap_mir;		/* SIT bitmap mirror */
263 
264 	/* bitmap of segments to be ignored by GC in case of errors */
265 	unsigned long *invalid_segmap;
266 #endif
267 	unsigned int bitmap_size;	/* SIT bitmap size */
268 
269 	unsigned long *tmp_map;			/* bitmap for temporal use */
270 	unsigned long *dirty_sentries_bitmap;	/* bitmap for dirty sentries */
271 	unsigned int dirty_sentries;		/* # of dirty sentries */
272 	unsigned int sents_per_block;		/* # of SIT entries per block */
273 	struct rw_semaphore sentry_lock;	/* to protect SIT cache */
274 	struct seg_entry *sentries;		/* SIT segment-level cache */
275 	struct sec_entry *sec_entries;		/* SIT section-level cache */
276 
277 	/* for cost-benefit algorithm in cleaning procedure */
278 	unsigned long long elapsed_time;	/* elapsed time after mount */
279 	unsigned long long mounted_time;	/* mount time */
280 	unsigned long long min_mtime;		/* min. modification time */
281 	unsigned long long max_mtime;		/* max. modification time */
282 	unsigned long long dirty_min_mtime;	/* rerange candidates in GC_AT */
283 	unsigned long long dirty_max_mtime;	/* rerange candidates in GC_AT */
284 
285 	unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
286 };
287 
288 struct free_segmap_info {
289 	unsigned int start_segno;	/* start segment number logically */
290 	unsigned int free_segments;	/* # of free segments */
291 	unsigned int free_sections;	/* # of free sections */
292 	spinlock_t segmap_lock;		/* free segmap lock */
293 	unsigned long *free_segmap;	/* free segment bitmap */
294 	unsigned long *free_secmap;	/* free section bitmap */
295 };
296 
297 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
298 enum dirty_type {
299 	DIRTY_HOT_DATA,		/* dirty segments assigned as hot data logs */
300 	DIRTY_WARM_DATA,	/* dirty segments assigned as warm data logs */
301 	DIRTY_COLD_DATA,	/* dirty segments assigned as cold data logs */
302 	DIRTY_HOT_NODE,		/* dirty segments assigned as hot node logs */
303 	DIRTY_WARM_NODE,	/* dirty segments assigned as warm node logs */
304 	DIRTY_COLD_NODE,	/* dirty segments assigned as cold node logs */
305 	DIRTY,			/* to count # of dirty segments */
306 	PRE,			/* to count # of entirely obsolete segments */
307 	NR_DIRTY_TYPE
308 };
309 
310 struct dirty_seglist_info {
311 	const struct victim_selection *v_ops;	/* victim selction operation */
312 	unsigned long *dirty_segmap[NR_DIRTY_TYPE];
313 	unsigned long *dirty_secmap;
314 	struct mutex seglist_lock;		/* lock for segment bitmaps */
315 	int nr_dirty[NR_DIRTY_TYPE];		/* # of dirty segments */
316 	unsigned long *victim_secmap;		/* background GC victims */
317 };
318 
319 /* victim selection function for cleaning and SSR */
320 struct victim_selection {
321 	int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
322 					int, int, char, unsigned long long);
323 };
324 
325 /* for active log information */
326 struct curseg_info {
327 	struct mutex curseg_mutex;		/* lock for consistency */
328 	struct f2fs_summary_block *sum_blk;	/* cached summary block */
329 	struct rw_semaphore journal_rwsem;	/* protect journal area */
330 	struct f2fs_journal *journal;		/* cached journal info */
331 	unsigned char alloc_type;		/* current allocation type */
332 	unsigned short seg_type;		/* segment type like CURSEG_XXX_TYPE */
333 	unsigned int segno;			/* current segment number */
334 	unsigned short next_blkoff;		/* next block offset to write */
335 	unsigned int zone;			/* current zone number */
336 	unsigned int next_segno;		/* preallocated segment */
337 	bool inited;				/* indicate inmem log is inited */
338 };
339 
340 struct sit_entry_set {
341 	struct list_head set_list;	/* link with all sit sets */
342 	unsigned int start_segno;	/* start segno of sits in set */
343 	unsigned int entry_cnt;		/* the # of sit entries in set */
344 };
345 
346 /*
347  * inline functions
348  */
CURSEG_I(struct f2fs_sb_info * sbi,int type)349 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
350 {
351 	return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
352 }
353 
get_seg_entry(struct f2fs_sb_info * sbi,unsigned int segno)354 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
355 						unsigned int segno)
356 {
357 	struct sit_info *sit_i = SIT_I(sbi);
358 	return &sit_i->sentries[segno];
359 }
360 
get_sec_entry(struct f2fs_sb_info * sbi,unsigned int segno)361 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
362 						unsigned int segno)
363 {
364 	struct sit_info *sit_i = SIT_I(sbi);
365 	return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
366 }
367 
get_valid_blocks(struct f2fs_sb_info * sbi,unsigned int segno,bool use_section)368 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
369 				unsigned int segno, bool use_section)
370 {
371 	/*
372 	 * In order to get # of valid blocks in a section instantly from many
373 	 * segments, f2fs manages two counting structures separately.
374 	 */
375 	if (use_section && __is_large_section(sbi))
376 		return get_sec_entry(sbi, segno)->valid_blocks;
377 	else
378 		return get_seg_entry(sbi, segno)->valid_blocks;
379 }
380 
get_ckpt_valid_blocks(struct f2fs_sb_info * sbi,unsigned int segno,bool use_section)381 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
382 				unsigned int segno, bool use_section)
383 {
384 	if (use_section && __is_large_section(sbi)) {
385 		unsigned int start_segno = START_SEGNO(segno);
386 		unsigned int blocks = 0;
387 		int i;
388 
389 		for (i = 0; i < sbi->segs_per_sec; i++, start_segno++) {
390 			struct seg_entry *se = get_seg_entry(sbi, start_segno);
391 
392 			blocks += se->ckpt_valid_blocks;
393 		}
394 		return blocks;
395 	}
396 	return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
397 }
398 
seg_info_from_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)399 static inline void seg_info_from_raw_sit(struct seg_entry *se,
400 					struct f2fs_sit_entry *rs)
401 {
402 	se->valid_blocks = GET_SIT_VBLOCKS(rs);
403 	se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
404 	memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
405 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
406 #ifdef CONFIG_F2FS_CHECK_FS
407 	memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
408 #endif
409 	se->type = GET_SIT_TYPE(rs);
410 	se->mtime = le64_to_cpu(rs->mtime);
411 }
412 
__seg_info_to_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)413 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
414 					struct f2fs_sit_entry *rs)
415 {
416 	unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
417 					se->valid_blocks;
418 	rs->vblocks = cpu_to_le16(raw_vblocks);
419 	memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
420 	rs->mtime = cpu_to_le64(se->mtime);
421 }
422 
seg_info_to_sit_page(struct f2fs_sb_info * sbi,struct page * page,unsigned int start)423 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
424 				struct page *page, unsigned int start)
425 {
426 	struct f2fs_sit_block *raw_sit;
427 	struct seg_entry *se;
428 	struct f2fs_sit_entry *rs;
429 	unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
430 					(unsigned long)MAIN_SEGS(sbi));
431 	int i;
432 
433 	raw_sit = (struct f2fs_sit_block *)page_address(page);
434 	memset(raw_sit, 0, PAGE_SIZE);
435 	for (i = 0; i < end - start; i++) {
436 		rs = &raw_sit->entries[i];
437 		se = get_seg_entry(sbi, start + i);
438 		__seg_info_to_raw_sit(se, rs);
439 	}
440 }
441 
seg_info_to_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)442 static inline void seg_info_to_raw_sit(struct seg_entry *se,
443 					struct f2fs_sit_entry *rs)
444 {
445 	__seg_info_to_raw_sit(se, rs);
446 
447 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
448 	se->ckpt_valid_blocks = se->valid_blocks;
449 }
450 
find_next_inuse(struct free_segmap_info * free_i,unsigned int max,unsigned int segno)451 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
452 		unsigned int max, unsigned int segno)
453 {
454 	unsigned int ret;
455 	spin_lock(&free_i->segmap_lock);
456 	ret = find_next_bit(free_i->free_segmap, max, segno);
457 	spin_unlock(&free_i->segmap_lock);
458 	return ret;
459 }
460 
__set_free(struct f2fs_sb_info * sbi,unsigned int segno)461 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
462 {
463 	struct free_segmap_info *free_i = FREE_I(sbi);
464 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
465 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
466 	unsigned int next;
467 	unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
468 
469 	spin_lock(&free_i->segmap_lock);
470 	clear_bit(segno, free_i->free_segmap);
471 	free_i->free_segments++;
472 
473 	next = find_next_bit(free_i->free_segmap,
474 			start_segno + sbi->segs_per_sec, start_segno);
475 	if (next >= start_segno + usable_segs) {
476 		clear_bit(secno, free_i->free_secmap);
477 		free_i->free_sections++;
478 	}
479 	spin_unlock(&free_i->segmap_lock);
480 }
481 
__set_inuse(struct f2fs_sb_info * sbi,unsigned int segno)482 static inline void __set_inuse(struct f2fs_sb_info *sbi,
483 		unsigned int segno)
484 {
485 	struct free_segmap_info *free_i = FREE_I(sbi);
486 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
487 
488 	set_bit(segno, free_i->free_segmap);
489 	free_i->free_segments--;
490 	if (!test_and_set_bit(secno, free_i->free_secmap))
491 		free_i->free_sections--;
492 }
493 
__set_test_and_free(struct f2fs_sb_info * sbi,unsigned int segno,bool inmem)494 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
495 		unsigned int segno, bool inmem)
496 {
497 	struct free_segmap_info *free_i = FREE_I(sbi);
498 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
499 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
500 	unsigned int next;
501 	unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
502 
503 	spin_lock(&free_i->segmap_lock);
504 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
505 		free_i->free_segments++;
506 
507 		if (!inmem && IS_CURSEC(sbi, secno))
508 			goto skip_free;
509 		next = find_next_bit(free_i->free_segmap,
510 				start_segno + sbi->segs_per_sec, start_segno);
511 		if (next >= start_segno + usable_segs) {
512 			if (test_and_clear_bit(secno, free_i->free_secmap))
513 				free_i->free_sections++;
514 		}
515 	}
516 skip_free:
517 	spin_unlock(&free_i->segmap_lock);
518 }
519 
__set_test_and_inuse(struct f2fs_sb_info * sbi,unsigned int segno)520 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
521 		unsigned int segno)
522 {
523 	struct free_segmap_info *free_i = FREE_I(sbi);
524 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
525 
526 	spin_lock(&free_i->segmap_lock);
527 	if (!test_and_set_bit(segno, free_i->free_segmap)) {
528 		free_i->free_segments--;
529 		if (!test_and_set_bit(secno, free_i->free_secmap))
530 			free_i->free_sections--;
531 	}
532 	spin_unlock(&free_i->segmap_lock);
533 }
534 
get_sit_bitmap(struct f2fs_sb_info * sbi,void * dst_addr)535 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
536 		void *dst_addr)
537 {
538 	struct sit_info *sit_i = SIT_I(sbi);
539 
540 #ifdef CONFIG_F2FS_CHECK_FS
541 	if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
542 						sit_i->bitmap_size))
543 		f2fs_bug_on(sbi, 1);
544 #endif
545 	memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
546 }
547 
written_block_count(struct f2fs_sb_info * sbi)548 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
549 {
550 	return SIT_I(sbi)->written_valid_blocks;
551 }
552 
free_segments(struct f2fs_sb_info * sbi)553 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
554 {
555 	return FREE_I(sbi)->free_segments;
556 }
557 
reserved_segments(struct f2fs_sb_info * sbi)558 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
559 {
560 	return SM_I(sbi)->reserved_segments +
561 			SM_I(sbi)->additional_reserved_segments;
562 }
563 
free_sections(struct f2fs_sb_info * sbi)564 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
565 {
566 	return FREE_I(sbi)->free_sections;
567 }
568 
prefree_segments(struct f2fs_sb_info * sbi)569 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
570 {
571 	return DIRTY_I(sbi)->nr_dirty[PRE];
572 }
573 
dirty_segments(struct f2fs_sb_info * sbi)574 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
575 {
576 	return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
577 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
578 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
579 		DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
580 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
581 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
582 }
583 
overprovision_segments(struct f2fs_sb_info * sbi)584 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
585 {
586 	return SM_I(sbi)->ovp_segments;
587 }
588 
reserved_sections(struct f2fs_sb_info * sbi)589 static inline int reserved_sections(struct f2fs_sb_info *sbi)
590 {
591 	return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
592 }
593 
has_curseg_enough_space(struct f2fs_sb_info * sbi)594 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
595 {
596 	unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
597 					get_pages(sbi, F2FS_DIRTY_DENTS);
598 	unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
599 	unsigned int segno, left_blocks;
600 	int i;
601 
602 	/* check current node segment */
603 	for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
604 		segno = CURSEG_I(sbi, i)->segno;
605 		left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
606 				get_seg_entry(sbi, segno)->ckpt_valid_blocks;
607 
608 		if (node_blocks > left_blocks)
609 			return false;
610 	}
611 
612 	/* check current data segment */
613 	segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
614 	left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
615 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
616 	if (dent_blocks > left_blocks)
617 		return false;
618 	return true;
619 }
620 
has_not_enough_free_secs(struct f2fs_sb_info * sbi,int freed,int needed)621 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
622 					int freed, int needed)
623 {
624 	int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
625 	int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
626 	int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
627 
628 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
629 		return false;
630 
631 	if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
632 			has_curseg_enough_space(sbi))
633 		return false;
634 	return (free_sections(sbi) + freed) <=
635 		(node_secs + 2 * dent_secs + imeta_secs +
636 		reserved_sections(sbi) + needed);
637 }
638 
f2fs_is_checkpoint_ready(struct f2fs_sb_info * sbi)639 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
640 {
641 	if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
642 		return true;
643 	if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
644 		return true;
645 	return false;
646 }
647 
excess_prefree_segs(struct f2fs_sb_info * sbi)648 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
649 {
650 	return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
651 }
652 
utilization(struct f2fs_sb_info * sbi)653 static inline int utilization(struct f2fs_sb_info *sbi)
654 {
655 	return div_u64((u64)valid_user_blocks(sbi) * 100,
656 					sbi->user_block_count);
657 }
658 
659 /*
660  * Sometimes f2fs may be better to drop out-of-place update policy.
661  * And, users can control the policy through sysfs entries.
662  * There are five policies with triggering conditions as follows.
663  * F2FS_IPU_FORCE - all the time,
664  * F2FS_IPU_SSR - if SSR mode is activated,
665  * F2FS_IPU_UTIL - if FS utilization is over threashold,
666  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
667  *                     threashold,
668  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
669  *                     storages. IPU will be triggered only if the # of dirty
670  *                     pages over min_fsync_blocks. (=default option)
671  * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
672  * F2FS_IPU_NOCACHE - disable IPU bio cache.
673  * F2FS_IPUT_DISABLE - disable IPU. (=default option in LFS mode)
674  */
675 #define DEF_MIN_IPU_UTIL	70
676 #define DEF_MIN_FSYNC_BLOCKS	8
677 #define DEF_MIN_HOT_BLOCKS	16
678 
679 #define SMALL_VOLUME_SEGMENTS	(16 * 512)	/* 16GB */
680 
681 enum {
682 	F2FS_IPU_FORCE,
683 	F2FS_IPU_SSR,
684 	F2FS_IPU_UTIL,
685 	F2FS_IPU_SSR_UTIL,
686 	F2FS_IPU_FSYNC,
687 	F2FS_IPU_ASYNC,
688 	F2FS_IPU_NOCACHE,
689 };
690 
curseg_segno(struct f2fs_sb_info * sbi,int type)691 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
692 		int type)
693 {
694 	struct curseg_info *curseg = CURSEG_I(sbi, type);
695 	return curseg->segno;
696 }
697 
curseg_alloc_type(struct f2fs_sb_info * sbi,int type)698 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
699 		int type)
700 {
701 	struct curseg_info *curseg = CURSEG_I(sbi, type);
702 	return curseg->alloc_type;
703 }
704 
curseg_blkoff(struct f2fs_sb_info * sbi,int type)705 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
706 {
707 	struct curseg_info *curseg = CURSEG_I(sbi, type);
708 	return curseg->next_blkoff;
709 }
710 
check_seg_range(struct f2fs_sb_info * sbi,unsigned int segno)711 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
712 {
713 	f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
714 }
715 
verify_fio_blkaddr(struct f2fs_io_info * fio)716 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
717 {
718 	struct f2fs_sb_info *sbi = fio->sbi;
719 
720 	if (__is_valid_data_blkaddr(fio->old_blkaddr))
721 		verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
722 					META_GENERIC : DATA_GENERIC);
723 	verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
724 					META_GENERIC : DATA_GENERIC_ENHANCE);
725 }
726 
727 /*
728  * Summary block is always treated as an invalid block
729  */
check_block_count(struct f2fs_sb_info * sbi,int segno,struct f2fs_sit_entry * raw_sit)730 static inline int check_block_count(struct f2fs_sb_info *sbi,
731 		int segno, struct f2fs_sit_entry *raw_sit)
732 {
733 	bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
734 	int valid_blocks = 0;
735 	int cur_pos = 0, next_pos;
736 	unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
737 
738 	/* check bitmap with valid block count */
739 	do {
740 		if (is_valid) {
741 			next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
742 					usable_blks_per_seg,
743 					cur_pos);
744 			valid_blocks += next_pos - cur_pos;
745 		} else
746 			next_pos = find_next_bit_le(&raw_sit->valid_map,
747 					usable_blks_per_seg,
748 					cur_pos);
749 		cur_pos = next_pos;
750 		is_valid = !is_valid;
751 	} while (cur_pos < usable_blks_per_seg);
752 
753 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
754 		f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
755 			 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
756 		set_sbi_flag(sbi, SBI_NEED_FSCK);
757 		return -EFSCORRUPTED;
758 	}
759 
760 	if (usable_blks_per_seg < sbi->blocks_per_seg)
761 		f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
762 				sbi->blocks_per_seg,
763 				usable_blks_per_seg) != sbi->blocks_per_seg);
764 
765 	/* check segment usage, and check boundary of a given segment number */
766 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
767 					|| segno > TOTAL_SEGS(sbi) - 1)) {
768 		f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
769 			 GET_SIT_VBLOCKS(raw_sit), segno);
770 		set_sbi_flag(sbi, SBI_NEED_FSCK);
771 		return -EFSCORRUPTED;
772 	}
773 	return 0;
774 }
775 
current_sit_addr(struct f2fs_sb_info * sbi,unsigned int start)776 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
777 						unsigned int start)
778 {
779 	struct sit_info *sit_i = SIT_I(sbi);
780 	unsigned int offset = SIT_BLOCK_OFFSET(start);
781 	block_t blk_addr = sit_i->sit_base_addr + offset;
782 
783 	check_seg_range(sbi, start);
784 
785 #ifdef CONFIG_F2FS_CHECK_FS
786 	if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
787 			f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
788 		f2fs_bug_on(sbi, 1);
789 #endif
790 
791 	/* calculate sit block address */
792 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
793 		blk_addr += sit_i->sit_blocks;
794 
795 	return blk_addr;
796 }
797 
next_sit_addr(struct f2fs_sb_info * sbi,pgoff_t block_addr)798 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
799 						pgoff_t block_addr)
800 {
801 	struct sit_info *sit_i = SIT_I(sbi);
802 	block_addr -= sit_i->sit_base_addr;
803 	if (block_addr < sit_i->sit_blocks)
804 		block_addr += sit_i->sit_blocks;
805 	else
806 		block_addr -= sit_i->sit_blocks;
807 
808 	return block_addr + sit_i->sit_base_addr;
809 }
810 
set_to_next_sit(struct sit_info * sit_i,unsigned int start)811 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
812 {
813 	unsigned int block_off = SIT_BLOCK_OFFSET(start);
814 
815 	f2fs_change_bit(block_off, sit_i->sit_bitmap);
816 #ifdef CONFIG_F2FS_CHECK_FS
817 	f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
818 #endif
819 }
820 
get_mtime(struct f2fs_sb_info * sbi,bool base_time)821 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
822 						bool base_time)
823 {
824 	struct sit_info *sit_i = SIT_I(sbi);
825 	time64_t diff, now = ktime_get_boottime_seconds();
826 
827 	if (now >= sit_i->mounted_time)
828 		return sit_i->elapsed_time + now - sit_i->mounted_time;
829 
830 	/* system time is set to the past */
831 	if (!base_time) {
832 		diff = sit_i->mounted_time - now;
833 		if (sit_i->elapsed_time >= diff)
834 			return sit_i->elapsed_time - diff;
835 		return 0;
836 	}
837 	return sit_i->elapsed_time;
838 }
839 
set_summary(struct f2fs_summary * sum,nid_t nid,unsigned int ofs_in_node,unsigned char version)840 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
841 			unsigned int ofs_in_node, unsigned char version)
842 {
843 	sum->nid = cpu_to_le32(nid);
844 	sum->ofs_in_node = cpu_to_le16(ofs_in_node);
845 	sum->version = version;
846 }
847 
start_sum_block(struct f2fs_sb_info * sbi)848 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
849 {
850 	return __start_cp_addr(sbi) +
851 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
852 }
853 
sum_blk_addr(struct f2fs_sb_info * sbi,int base,int type)854 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
855 {
856 	return __start_cp_addr(sbi) +
857 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
858 				- (base + 1) + type;
859 }
860 
sec_usage_check(struct f2fs_sb_info * sbi,unsigned int secno)861 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
862 {
863 	if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
864 		return true;
865 	return false;
866 }
867 
868 /*
869  * It is very important to gather dirty pages and write at once, so that we can
870  * submit a big bio without interfering other data writes.
871  * By default, 512 pages for directory data,
872  * 512 pages (2MB) * 8 for nodes, and
873  * 256 pages * 8 for meta are set.
874  */
nr_pages_to_skip(struct f2fs_sb_info * sbi,int type)875 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
876 {
877 	if (sbi->sb->s_bdi->wb.dirty_exceeded)
878 		return 0;
879 
880 	if (type == DATA)
881 		return sbi->blocks_per_seg;
882 	else if (type == NODE)
883 		return 8 * sbi->blocks_per_seg;
884 	else if (type == META)
885 		return 8 * BIO_MAX_PAGES;
886 	else
887 		return 0;
888 }
889 
890 /*
891  * When writing pages, it'd better align nr_to_write for segment size.
892  */
nr_pages_to_write(struct f2fs_sb_info * sbi,int type,struct writeback_control * wbc)893 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
894 					struct writeback_control *wbc)
895 {
896 	long nr_to_write, desired;
897 
898 	if (wbc->sync_mode != WB_SYNC_NONE)
899 		return 0;
900 
901 	nr_to_write = wbc->nr_to_write;
902 	desired = BIO_MAX_PAGES;
903 	if (type == NODE)
904 		desired <<= 1;
905 
906 	wbc->nr_to_write = desired;
907 	return desired - nr_to_write;
908 }
909 
wake_up_discard_thread(struct f2fs_sb_info * sbi,bool force)910 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
911 {
912 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
913 	bool wakeup = false;
914 	int i;
915 
916 	if (force)
917 		goto wake_up;
918 
919 	mutex_lock(&dcc->cmd_lock);
920 	for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
921 		if (i + 1 < dcc->discard_granularity)
922 			break;
923 		if (!list_empty(&dcc->pend_list[i])) {
924 			wakeup = true;
925 			break;
926 		}
927 	}
928 	mutex_unlock(&dcc->cmd_lock);
929 	if (!wakeup || !is_idle(sbi, DISCARD_TIME))
930 		return;
931 wake_up:
932 	dcc->discard_wake = 1;
933 	wake_up_interruptible_all(&dcc->discard_wait_queue);
934 }
935 
936 #ifdef CONFIG_F2FS_GRADING_SSR
check_io_seq(int blks)937 static inline int check_io_seq(int blks)
938 {
939 	if (blks >= SSR_CONTIG_LARGE)
940 		return SEQ_256BLKS;
941 	if (blks >= SSR_CONTIG_DIRTY_NUMS)
942 		return SEQ_32BLKS;
943 	return SEQ_NONE;
944 }
945 #endif
946