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)						\
57 	(SM_I(sbi) ? SM_I(sbi)->main_blkaddr : 				\
58 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
59 #define SEG0_BLKADDR(sbi)						\
60 	(SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : 				\
61 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
62 
63 #define MAIN_SEGS(sbi)	(SM_I(sbi)->main_segments)
64 #define MAIN_SECS(sbi)	((sbi)->total_sections)
65 
66 #define TOTAL_SEGS(sbi)							\
67 	(SM_I(sbi) ? SM_I(sbi)->segment_count : 				\
68 		le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
69 #define TOTAL_BLKS(sbi)	(TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
70 
71 #define MAX_BLKADDR(sbi)	(SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
72 #define SEGMENT_SIZE(sbi)	(1ULL << ((sbi)->log_blocksize +	\
73 					(sbi)->log_blocks_per_seg))
74 
75 #define START_BLOCK(sbi, segno)	(SEG0_BLKADDR(sbi) +			\
76 	 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
77 
78 #define NEXT_FREE_BLKADDR(sbi, curseg)					\
79 	(START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
80 
81 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr)	((blk_addr) - SEG0_BLKADDR(sbi))
82 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr)				\
83 	(GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
84 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr)				\
85 	(GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
86 
87 #define GET_SEGNO(sbi, blk_addr)					\
88 	((!is_valid_data_blkaddr(sbi, blk_addr)) ?			\
89 	NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi),			\
90 		GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
91 #define BLKS_PER_SEC(sbi)					\
92 	((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
93 #define GET_SEC_FROM_SEG(sbi, segno)				\
94 	(((segno) == -1) ? -1: (segno) / (sbi)->segs_per_sec)
95 #define GET_SEG_FROM_SEC(sbi, secno)				\
96 	((secno) * (sbi)->segs_per_sec)
97 #define GET_ZONE_FROM_SEC(sbi, secno)				\
98 	(((secno) == -1) ? -1: (secno) / (sbi)->secs_per_zone)
99 #define GET_ZONE_FROM_SEG(sbi, segno)				\
100 	GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
101 
102 #define GET_SUM_BLOCK(sbi, segno)				\
103 	((sbi)->sm_info->ssa_blkaddr + (segno))
104 
105 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
106 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
107 
108 #define SIT_ENTRY_OFFSET(sit_i, segno)					\
109 	((segno) % (sit_i)->sents_per_block)
110 #define SIT_BLOCK_OFFSET(segno)					\
111 	((segno) / SIT_ENTRY_PER_BLOCK)
112 #define	START_SEGNO(segno)		\
113 	(SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
114 #define SIT_BLK_CNT(sbi)			\
115 	((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
116 #define f2fs_bitmap_size(nr)			\
117 	(BITS_TO_LONGS(nr) * sizeof(unsigned long))
118 
119 #define SECTOR_FROM_BLOCK(blk_addr)					\
120 	(((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
121 #define SECTOR_TO_BLOCK(sectors)					\
122 	((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
123 
124 /*
125  * indicate a block allocation direction: RIGHT and LEFT.
126  * RIGHT means allocating new sections towards the end of volume.
127  * LEFT means the opposite direction.
128  */
129 enum {
130 	ALLOC_RIGHT = 0,
131 	ALLOC_LEFT
132 };
133 
134 /*
135  * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
136  * LFS writes data sequentially with cleaning operations.
137  * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
138  */
139 enum {
140 	LFS = 0,
141 	SSR
142 };
143 
144 /*
145  * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
146  * GC_CB is based on cost-benefit algorithm.
147  * GC_GREEDY is based on greedy algorithm.
148  */
149 enum {
150 	GC_CB = 0,
151 	GC_GREEDY,
152 	ALLOC_NEXT,
153 	FLUSH_DEVICE,
154 	MAX_GC_POLICY,
155 };
156 
157 /*
158  * BG_GC means the background cleaning job.
159  * FG_GC means the on-demand cleaning job.
160  * FORCE_FG_GC means on-demand cleaning job in background.
161  */
162 enum {
163 	BG_GC = 0,
164 	FG_GC,
165 	FORCE_FG_GC,
166 };
167 
168 /* for a function parameter to select a victim segment */
169 struct victim_sel_policy {
170 	int alloc_mode;			/* LFS or SSR */
171 	int gc_mode;			/* GC_CB or GC_GREEDY */
172 	unsigned long *dirty_segmap;	/* dirty segment bitmap */
173 	unsigned int max_search;	/* maximum # of segments to search */
174 	unsigned int offset;		/* last scanned bitmap offset */
175 	unsigned int ofs_unit;		/* bitmap search unit */
176 	unsigned int min_cost;		/* minimum cost */
177 	unsigned int min_segno;		/* segment # having min. cost */
178 };
179 
180 struct seg_entry {
181 	unsigned int type:6;		/* segment type like CURSEG_XXX_TYPE */
182 	unsigned int valid_blocks:10;	/* # of valid blocks */
183 	unsigned int ckpt_valid_blocks:10;	/* # of valid blocks last cp */
184 	unsigned int padding:6;		/* padding */
185 	unsigned char *cur_valid_map;	/* validity bitmap of blocks */
186 #ifdef CONFIG_F2FS_CHECK_FS
187 	unsigned char *cur_valid_map_mir;	/* mirror of current valid bitmap */
188 #endif
189 	/*
190 	 * # of valid blocks and the validity bitmap stored in the the last
191 	 * checkpoint pack. This information is used by the SSR mode.
192 	 */
193 	unsigned char *ckpt_valid_map;	/* validity bitmap of blocks last cp */
194 	unsigned char *discard_map;
195 	unsigned long long mtime;	/* modification time of the segment */
196 };
197 
198 struct sec_entry {
199 	unsigned int valid_blocks;	/* # of valid blocks in a section */
200 };
201 
202 struct segment_allocation {
203 	void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
204 };
205 
206 /*
207  * this value is set in page as a private data which indicate that
208  * the page is atomically written, and it is in inmem_pages list.
209  */
210 #define ATOMIC_WRITTEN_PAGE		((unsigned long)-1)
211 #define DUMMY_WRITTEN_PAGE		((unsigned long)-2)
212 
213 #define IS_ATOMIC_WRITTEN_PAGE(page)			\
214 		(page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
215 #define IS_DUMMY_WRITTEN_PAGE(page)			\
216 		(page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
217 
218 #define MAX_SKIP_GC_COUNT			16
219 
220 struct inmem_pages {
221 	struct list_head list;
222 	struct page *page;
223 	block_t old_addr;		/* for revoking when fail to commit */
224 };
225 
226 struct sit_info {
227 	const struct segment_allocation *s_ops;
228 
229 	block_t sit_base_addr;		/* start block address of SIT area */
230 	block_t sit_blocks;		/* # of blocks used by SIT area */
231 	block_t written_valid_blocks;	/* # of valid blocks in main area */
232 	char *sit_bitmap;		/* SIT bitmap pointer */
233 #ifdef CONFIG_F2FS_CHECK_FS
234 	char *sit_bitmap_mir;		/* SIT bitmap mirror */
235 #endif
236 	unsigned int bitmap_size;	/* SIT bitmap size */
237 
238 	unsigned long *tmp_map;			/* bitmap for temporal use */
239 	unsigned long *dirty_sentries_bitmap;	/* bitmap for dirty sentries */
240 	unsigned int dirty_sentries;		/* # of dirty sentries */
241 	unsigned int sents_per_block;		/* # of SIT entries per block */
242 	struct rw_semaphore sentry_lock;	/* to protect SIT cache */
243 	struct seg_entry *sentries;		/* SIT segment-level cache */
244 	struct sec_entry *sec_entries;		/* SIT section-level cache */
245 
246 	/* for cost-benefit algorithm in cleaning procedure */
247 	unsigned long long elapsed_time;	/* elapsed time after mount */
248 	unsigned long long mounted_time;	/* mount time */
249 	unsigned long long min_mtime;		/* min. modification time */
250 	unsigned long long max_mtime;		/* max. modification time */
251 
252 	unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
253 };
254 
255 struct free_segmap_info {
256 	unsigned int start_segno;	/* start segment number logically */
257 	unsigned int free_segments;	/* # of free segments */
258 	unsigned int free_sections;	/* # of free sections */
259 	spinlock_t segmap_lock;		/* free segmap lock */
260 	unsigned long *free_segmap;	/* free segment bitmap */
261 	unsigned long *free_secmap;	/* free section bitmap */
262 };
263 
264 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
265 enum dirty_type {
266 	DIRTY_HOT_DATA,		/* dirty segments assigned as hot data logs */
267 	DIRTY_WARM_DATA,	/* dirty segments assigned as warm data logs */
268 	DIRTY_COLD_DATA,	/* dirty segments assigned as cold data logs */
269 	DIRTY_HOT_NODE,		/* dirty segments assigned as hot node logs */
270 	DIRTY_WARM_NODE,	/* dirty segments assigned as warm node logs */
271 	DIRTY_COLD_NODE,	/* dirty segments assigned as cold node logs */
272 	DIRTY,			/* to count # of dirty segments */
273 	PRE,			/* to count # of entirely obsolete segments */
274 	NR_DIRTY_TYPE
275 };
276 
277 struct dirty_seglist_info {
278 	const struct victim_selection *v_ops;	/* victim selction operation */
279 	unsigned long *dirty_segmap[NR_DIRTY_TYPE];
280 	struct mutex seglist_lock;		/* lock for segment bitmaps */
281 	int nr_dirty[NR_DIRTY_TYPE];		/* # of dirty segments */
282 	unsigned long *victim_secmap;		/* background GC victims */
283 };
284 
285 /* victim selection function for cleaning and SSR */
286 struct victim_selection {
287 	int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
288 							int, int, char);
289 };
290 
291 /* for active log information */
292 struct curseg_info {
293 	struct mutex curseg_mutex;		/* lock for consistency */
294 	struct f2fs_summary_block *sum_blk;	/* cached summary block */
295 	struct rw_semaphore journal_rwsem;	/* protect journal area */
296 	struct f2fs_journal *journal;		/* cached journal info */
297 	unsigned char alloc_type;		/* current allocation type */
298 	unsigned int segno;			/* current segment number */
299 	unsigned short next_blkoff;		/* next block offset to write */
300 	unsigned int zone;			/* current zone number */
301 	unsigned int next_segno;		/* preallocated segment */
302 };
303 
304 struct sit_entry_set {
305 	struct list_head set_list;	/* link with all sit sets */
306 	unsigned int start_segno;	/* start segno of sits in set */
307 	unsigned int entry_cnt;		/* the # of sit entries in set */
308 };
309 
310 /*
311  * inline functions
312  */
CURSEG_I(struct f2fs_sb_info * sbi,int type)313 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
314 {
315 	return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
316 }
317 
get_seg_entry(struct f2fs_sb_info * sbi,unsigned int segno)318 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
319 						unsigned int segno)
320 {
321 	struct sit_info *sit_i = SIT_I(sbi);
322 	return &sit_i->sentries[segno];
323 }
324 
get_sec_entry(struct f2fs_sb_info * sbi,unsigned int segno)325 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
326 						unsigned int segno)
327 {
328 	struct sit_info *sit_i = SIT_I(sbi);
329 	return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
330 }
331 
get_valid_blocks(struct f2fs_sb_info * sbi,unsigned int segno,bool use_section)332 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
333 				unsigned int segno, bool use_section)
334 {
335 	/*
336 	 * In order to get # of valid blocks in a section instantly from many
337 	 * segments, f2fs manages two counting structures separately.
338 	 */
339 	if (use_section && sbi->segs_per_sec > 1)
340 		return get_sec_entry(sbi, segno)->valid_blocks;
341 	else
342 		return get_seg_entry(sbi, segno)->valid_blocks;
343 }
344 
seg_info_from_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)345 static inline void seg_info_from_raw_sit(struct seg_entry *se,
346 					struct f2fs_sit_entry *rs)
347 {
348 	se->valid_blocks = GET_SIT_VBLOCKS(rs);
349 	se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
350 	memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
351 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
352 #ifdef CONFIG_F2FS_CHECK_FS
353 	memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
354 #endif
355 	se->type = GET_SIT_TYPE(rs);
356 	se->mtime = le64_to_cpu(rs->mtime);
357 }
358 
__seg_info_to_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)359 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
360 					struct f2fs_sit_entry *rs)
361 {
362 	unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
363 					se->valid_blocks;
364 	rs->vblocks = cpu_to_le16(raw_vblocks);
365 	memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
366 	rs->mtime = cpu_to_le64(se->mtime);
367 }
368 
seg_info_to_sit_page(struct f2fs_sb_info * sbi,struct page * page,unsigned int start)369 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
370 				struct page *page, unsigned int start)
371 {
372 	struct f2fs_sit_block *raw_sit;
373 	struct seg_entry *se;
374 	struct f2fs_sit_entry *rs;
375 	unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
376 					(unsigned long)MAIN_SEGS(sbi));
377 	int i;
378 
379 	raw_sit = (struct f2fs_sit_block *)page_address(page);
380 	memset(raw_sit, 0, PAGE_SIZE);
381 	for (i = 0; i < end - start; i++) {
382 		rs = &raw_sit->entries[i];
383 		se = get_seg_entry(sbi, start + i);
384 		__seg_info_to_raw_sit(se, rs);
385 	}
386 }
387 
seg_info_to_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)388 static inline void seg_info_to_raw_sit(struct seg_entry *se,
389 					struct f2fs_sit_entry *rs)
390 {
391 	__seg_info_to_raw_sit(se, rs);
392 
393 	memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
394 	se->ckpt_valid_blocks = se->valid_blocks;
395 }
396 
find_next_inuse(struct free_segmap_info * free_i,unsigned int max,unsigned int segno)397 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
398 		unsigned int max, unsigned int segno)
399 {
400 	unsigned int ret;
401 	spin_lock(&free_i->segmap_lock);
402 	ret = find_next_bit(free_i->free_segmap, max, segno);
403 	spin_unlock(&free_i->segmap_lock);
404 	return ret;
405 }
406 
__set_free(struct f2fs_sb_info * sbi,unsigned int segno)407 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
408 {
409 	struct free_segmap_info *free_i = FREE_I(sbi);
410 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
411 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
412 	unsigned int next;
413 
414 	spin_lock(&free_i->segmap_lock);
415 	clear_bit(segno, free_i->free_segmap);
416 	free_i->free_segments++;
417 
418 	next = find_next_bit(free_i->free_segmap,
419 			start_segno + sbi->segs_per_sec, start_segno);
420 	if (next >= start_segno + sbi->segs_per_sec) {
421 		clear_bit(secno, free_i->free_secmap);
422 		free_i->free_sections++;
423 	}
424 	spin_unlock(&free_i->segmap_lock);
425 }
426 
__set_inuse(struct f2fs_sb_info * sbi,unsigned int segno)427 static inline void __set_inuse(struct f2fs_sb_info *sbi,
428 		unsigned int segno)
429 {
430 	struct free_segmap_info *free_i = FREE_I(sbi);
431 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
432 
433 	set_bit(segno, free_i->free_segmap);
434 	free_i->free_segments--;
435 	if (!test_and_set_bit(secno, free_i->free_secmap))
436 		free_i->free_sections--;
437 }
438 
__set_test_and_free(struct f2fs_sb_info * sbi,unsigned int segno)439 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
440 		unsigned int segno)
441 {
442 	struct free_segmap_info *free_i = FREE_I(sbi);
443 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
444 	unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
445 	unsigned int next;
446 
447 	spin_lock(&free_i->segmap_lock);
448 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
449 		free_i->free_segments++;
450 
451 		if (IS_CURSEC(sbi, secno))
452 			goto skip_free;
453 		next = find_next_bit(free_i->free_segmap,
454 				start_segno + sbi->segs_per_sec, start_segno);
455 		if (next >= start_segno + sbi->segs_per_sec) {
456 			if (test_and_clear_bit(secno, free_i->free_secmap))
457 				free_i->free_sections++;
458 		}
459 	}
460 skip_free:
461 	spin_unlock(&free_i->segmap_lock);
462 }
463 
__set_test_and_inuse(struct f2fs_sb_info * sbi,unsigned int segno)464 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
465 		unsigned int segno)
466 {
467 	struct free_segmap_info *free_i = FREE_I(sbi);
468 	unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
469 
470 	spin_lock(&free_i->segmap_lock);
471 	if (!test_and_set_bit(segno, free_i->free_segmap)) {
472 		free_i->free_segments--;
473 		if (!test_and_set_bit(secno, free_i->free_secmap))
474 			free_i->free_sections--;
475 	}
476 	spin_unlock(&free_i->segmap_lock);
477 }
478 
get_sit_bitmap(struct f2fs_sb_info * sbi,void * dst_addr)479 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
480 		void *dst_addr)
481 {
482 	struct sit_info *sit_i = SIT_I(sbi);
483 
484 #ifdef CONFIG_F2FS_CHECK_FS
485 	if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
486 						sit_i->bitmap_size))
487 		f2fs_bug_on(sbi, 1);
488 #endif
489 	memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
490 }
491 
written_block_count(struct f2fs_sb_info * sbi)492 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
493 {
494 	return SIT_I(sbi)->written_valid_blocks;
495 }
496 
free_segments(struct f2fs_sb_info * sbi)497 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
498 {
499 	return FREE_I(sbi)->free_segments;
500 }
501 
reserved_segments(struct f2fs_sb_info * sbi)502 static inline int reserved_segments(struct f2fs_sb_info *sbi)
503 {
504 	return SM_I(sbi)->reserved_segments;
505 }
506 
free_sections(struct f2fs_sb_info * sbi)507 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
508 {
509 	return FREE_I(sbi)->free_sections;
510 }
511 
prefree_segments(struct f2fs_sb_info * sbi)512 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
513 {
514 	return DIRTY_I(sbi)->nr_dirty[PRE];
515 }
516 
dirty_segments(struct f2fs_sb_info * sbi)517 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
518 {
519 	return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
520 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
521 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
522 		DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
523 		DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
524 		DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
525 }
526 
overprovision_segments(struct f2fs_sb_info * sbi)527 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
528 {
529 	return SM_I(sbi)->ovp_segments;
530 }
531 
reserved_sections(struct f2fs_sb_info * sbi)532 static inline int reserved_sections(struct f2fs_sb_info *sbi)
533 {
534 	return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
535 }
536 
has_curseg_enough_space(struct f2fs_sb_info * sbi,unsigned int node_blocks,unsigned int dent_blocks)537 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
538 			unsigned int node_blocks, unsigned int dent_blocks)
539 {
540 
541 	unsigned int segno, left_blocks;
542 	int i;
543 
544 	/* check current node segment */
545 	for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
546 		segno = CURSEG_I(sbi, i)->segno;
547 		left_blocks = sbi->blocks_per_seg -
548 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
549 
550 		if (node_blocks > left_blocks)
551 			return false;
552 	}
553 
554 	/* check current data segment */
555 	segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
556 	left_blocks = sbi->blocks_per_seg -
557 			get_seg_entry(sbi, segno)->ckpt_valid_blocks;
558 	if (dent_blocks > left_blocks)
559 		return false;
560 	return true;
561 }
562 
has_not_enough_free_secs(struct f2fs_sb_info * sbi,int freed,int needed)563 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
564 					int freed, int needed)
565 {
566 	unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
567 					get_pages(sbi, F2FS_DIRTY_DENTS) +
568 					get_pages(sbi, F2FS_DIRTY_IMETA);
569 	unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
570 	unsigned int node_secs = total_node_blocks / BLKS_PER_SEC(sbi);
571 	unsigned int dent_secs = total_dent_blocks / BLKS_PER_SEC(sbi);
572 	unsigned int node_blocks = total_node_blocks % BLKS_PER_SEC(sbi);
573 	unsigned int dent_blocks = total_dent_blocks % BLKS_PER_SEC(sbi);
574 	unsigned int free, need_lower, need_upper;
575 
576 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
577 		return false;
578 
579 	free = free_sections(sbi) + freed;
580 	need_lower = node_secs + dent_secs + reserved_sections(sbi) + needed;
581 	need_upper = need_lower + (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0);
582 
583 	if (free > need_upper)
584 		return false;
585 	else if (free <= need_lower)
586 		return true;
587 	return !has_curseg_enough_space(sbi, node_blocks, dent_blocks);
588 }
589 
excess_prefree_segs(struct f2fs_sb_info * sbi)590 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
591 {
592 	return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
593 }
594 
utilization(struct f2fs_sb_info * sbi)595 static inline int utilization(struct f2fs_sb_info *sbi)
596 {
597 	return div_u64((u64)valid_user_blocks(sbi) * 100,
598 					sbi->user_block_count);
599 }
600 
601 /*
602  * Sometimes f2fs may be better to drop out-of-place update policy.
603  * And, users can control the policy through sysfs entries.
604  * There are five policies with triggering conditions as follows.
605  * F2FS_IPU_FORCE - all the time,
606  * F2FS_IPU_SSR - if SSR mode is activated,
607  * F2FS_IPU_UTIL - if FS utilization is over threashold,
608  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
609  *                     threashold,
610  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
611  *                     storages. IPU will be triggered only if the # of dirty
612  *                     pages over min_fsync_blocks.
613  * F2FS_IPUT_DISABLE - disable IPU. (=default option)
614  */
615 #define DEF_MIN_IPU_UTIL	70
616 #define DEF_MIN_FSYNC_BLOCKS	8
617 #define DEF_MIN_HOT_BLOCKS	16
618 
619 #define SMALL_VOLUME_SEGMENTS	(16 * 512)	/* 16GB */
620 
621 enum {
622 	F2FS_IPU_FORCE,
623 	F2FS_IPU_SSR,
624 	F2FS_IPU_UTIL,
625 	F2FS_IPU_SSR_UTIL,
626 	F2FS_IPU_FSYNC,
627 	F2FS_IPU_ASYNC,
628 };
629 
curseg_segno(struct f2fs_sb_info * sbi,int type)630 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
631 		int type)
632 {
633 	struct curseg_info *curseg = CURSEG_I(sbi, type);
634 	return curseg->segno;
635 }
636 
curseg_alloc_type(struct f2fs_sb_info * sbi,int type)637 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
638 		int type)
639 {
640 	struct curseg_info *curseg = CURSEG_I(sbi, type);
641 	return curseg->alloc_type;
642 }
643 
curseg_blkoff(struct f2fs_sb_info * sbi,int type)644 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
645 {
646 	struct curseg_info *curseg = CURSEG_I(sbi, type);
647 	return curseg->next_blkoff;
648 }
649 
check_seg_range(struct f2fs_sb_info * sbi,unsigned int segno)650 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
651 {
652 	f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
653 }
654 
verify_block_addr(struct f2fs_io_info * fio,block_t blk_addr)655 static inline void verify_block_addr(struct f2fs_io_info *fio, block_t blk_addr)
656 {
657 	struct f2fs_sb_info *sbi = fio->sbi;
658 
659 	if (__is_meta_io(fio))
660 		verify_blkaddr(sbi, blk_addr, META_GENERIC);
661 	else
662 		verify_blkaddr(sbi, blk_addr, DATA_GENERIC);
663 }
664 
665 /*
666  * Summary block is always treated as an invalid block
667  */
check_block_count(struct f2fs_sb_info * sbi,int segno,struct f2fs_sit_entry * raw_sit)668 static inline int check_block_count(struct f2fs_sb_info *sbi,
669 		int segno, struct f2fs_sit_entry *raw_sit)
670 {
671 	bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
672 	int valid_blocks = 0;
673 	int cur_pos = 0, next_pos;
674 
675 	/* check bitmap with valid block count */
676 	do {
677 		if (is_valid) {
678 			next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
679 					sbi->blocks_per_seg,
680 					cur_pos);
681 			valid_blocks += next_pos - cur_pos;
682 		} else
683 			next_pos = find_next_bit_le(&raw_sit->valid_map,
684 					sbi->blocks_per_seg,
685 					cur_pos);
686 		cur_pos = next_pos;
687 		is_valid = !is_valid;
688 	} while (cur_pos < sbi->blocks_per_seg);
689 
690 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
691 		f2fs_msg(sbi->sb, KERN_ERR,
692 				"Mismatch valid blocks %d vs. %d",
693 					GET_SIT_VBLOCKS(raw_sit), valid_blocks);
694 		set_sbi_flag(sbi, SBI_NEED_FSCK);
695 		return -EFSCORRUPTED;
696 	}
697 
698 	/* check segment usage, and check boundary of a given segment number */
699 	if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
700 					|| segno > TOTAL_SEGS(sbi) - 1)) {
701 		f2fs_msg(sbi->sb, KERN_ERR,
702 				"Wrong valid blocks %d or segno %u",
703 					GET_SIT_VBLOCKS(raw_sit), segno);
704 		set_sbi_flag(sbi, SBI_NEED_FSCK);
705 		return -EFSCORRUPTED;
706 	}
707 	return 0;
708 }
709 
current_sit_addr(struct f2fs_sb_info * sbi,unsigned int start)710 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
711 						unsigned int start)
712 {
713 	struct sit_info *sit_i = SIT_I(sbi);
714 	unsigned int offset = SIT_BLOCK_OFFSET(start);
715 	block_t blk_addr = sit_i->sit_base_addr + offset;
716 
717 	check_seg_range(sbi, start);
718 
719 #ifdef CONFIG_F2FS_CHECK_FS
720 	if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
721 			f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
722 		f2fs_bug_on(sbi, 1);
723 #endif
724 
725 	/* calculate sit block address */
726 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
727 		blk_addr += sit_i->sit_blocks;
728 
729 	return blk_addr;
730 }
731 
next_sit_addr(struct f2fs_sb_info * sbi,pgoff_t block_addr)732 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
733 						pgoff_t block_addr)
734 {
735 	struct sit_info *sit_i = SIT_I(sbi);
736 	block_addr -= sit_i->sit_base_addr;
737 	if (block_addr < sit_i->sit_blocks)
738 		block_addr += sit_i->sit_blocks;
739 	else
740 		block_addr -= sit_i->sit_blocks;
741 
742 	return block_addr + sit_i->sit_base_addr;
743 }
744 
set_to_next_sit(struct sit_info * sit_i,unsigned int start)745 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
746 {
747 	unsigned int block_off = SIT_BLOCK_OFFSET(start);
748 
749 	f2fs_change_bit(block_off, sit_i->sit_bitmap);
750 #ifdef CONFIG_F2FS_CHECK_FS
751 	f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
752 #endif
753 }
754 
get_mtime(struct f2fs_sb_info * sbi,bool base_time)755 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
756 						bool base_time)
757 {
758 	struct sit_info *sit_i = SIT_I(sbi);
759 	time64_t diff, now = ktime_get_real_seconds();
760 
761 	if (now >= sit_i->mounted_time)
762 		return sit_i->elapsed_time + now - sit_i->mounted_time;
763 
764 	/* system time is set to the past */
765 	if (!base_time) {
766 		diff = sit_i->mounted_time - now;
767 		if (sit_i->elapsed_time >= diff)
768 			return sit_i->elapsed_time - diff;
769 		return 0;
770 	}
771 	return sit_i->elapsed_time;
772 }
773 
set_summary(struct f2fs_summary * sum,nid_t nid,unsigned int ofs_in_node,unsigned char version)774 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
775 			unsigned int ofs_in_node, unsigned char version)
776 {
777 	sum->nid = cpu_to_le32(nid);
778 	sum->ofs_in_node = cpu_to_le16(ofs_in_node);
779 	sum->version = version;
780 }
781 
start_sum_block(struct f2fs_sb_info * sbi)782 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
783 {
784 	return __start_cp_addr(sbi) +
785 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
786 }
787 
sum_blk_addr(struct f2fs_sb_info * sbi,int base,int type)788 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
789 {
790 	return __start_cp_addr(sbi) +
791 		le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
792 				- (base + 1) + type;
793 }
794 
sec_usage_check(struct f2fs_sb_info * sbi,unsigned int secno)795 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
796 {
797 	if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
798 		return true;
799 	return false;
800 }
801 
802 /*
803  * It is very important to gather dirty pages and write at once, so that we can
804  * submit a big bio without interfering other data writes.
805  * By default, 512 pages for directory data,
806  * 512 pages (2MB) * 8 for nodes, and
807  * 256 pages * 8 for meta are set.
808  */
nr_pages_to_skip(struct f2fs_sb_info * sbi,int type)809 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
810 {
811 	if (sbi->sb->s_bdi->wb.dirty_exceeded)
812 		return 0;
813 
814 	if (type == DATA)
815 		return sbi->blocks_per_seg;
816 	else if (type == NODE)
817 		return 8 * sbi->blocks_per_seg;
818 	else if (type == META)
819 		return 8 * BIO_MAX_PAGES;
820 	else
821 		return 0;
822 }
823 
824 /*
825  * When writing pages, it'd better align nr_to_write for segment size.
826  */
nr_pages_to_write(struct f2fs_sb_info * sbi,int type,struct writeback_control * wbc)827 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
828 					struct writeback_control *wbc)
829 {
830 	long nr_to_write, desired;
831 
832 	if (wbc->sync_mode != WB_SYNC_NONE)
833 		return 0;
834 
835 	nr_to_write = wbc->nr_to_write;
836 	desired = BIO_MAX_PAGES;
837 	if (type == NODE)
838 		desired <<= 1;
839 
840 	wbc->nr_to_write = desired;
841 	return desired - nr_to_write;
842 }
843 
wake_up_discard_thread(struct f2fs_sb_info * sbi,bool force)844 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
845 {
846 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
847 	bool wakeup = false;
848 	int i;
849 
850 	if (force)
851 		goto wake_up;
852 
853 	mutex_lock(&dcc->cmd_lock);
854 	for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
855 		if (i + 1 < dcc->discard_granularity)
856 			break;
857 		if (!list_empty(&dcc->pend_list[i])) {
858 			wakeup = true;
859 			break;
860 		}
861 	}
862 	mutex_unlock(&dcc->cmd_lock);
863 	if (!wakeup)
864 		return;
865 wake_up:
866 	dcc->discard_wake = 1;
867 	wake_up_interruptible_all(&dcc->discard_wait_queue);
868 }
869