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