Merge tag 'configfs-for-4.7' of git://git.infradead.org/users/hch/configfs
[linux-2.6-block.git] / fs / f2fs / segment.h
CommitLineData
0a8165d7 1/*
39a53e0c
JK
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 */
ac5d156c 11#include <linux/blkdev.h>
66114cad 12#include <linux/backing-dev.h>
ac5d156c 13
39a53e0c
JK
14/* constant macro */
15#define NULL_SEGNO ((unsigned int)(~0))
5ec4e49f 16#define NULL_SECNO ((unsigned int)(~0))
39a53e0c 17
58c41035 18#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
81eb8d6e 19
6224da87 20/* L: Logical segment # in volume, R: Relative segment # in main area */
39a53e0c
JK
21#define GET_L2R_SEGNO(free_i, segno) (segno - free_i->start_segno)
22#define GET_R2L_SEGNO(free_i, segno) (segno + free_i->start_segno)
23
61ae45c8
CL
24#define IS_DATASEG(t) (t <= CURSEG_COLD_DATA)
25#define IS_NODESEG(t) (t >= CURSEG_HOT_NODE)
39a53e0c 26
5c773ba3
JK
27#define IS_CURSEG(sbi, seg) \
28 ((seg == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
29 (seg == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
30 (seg == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
31 (seg == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
32 (seg == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
33 (seg == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
39a53e0c
JK
34
35#define IS_CURSEC(sbi, secno) \
36 ((secno == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
37 sbi->segs_per_sec) || \
38 (secno == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
39 sbi->segs_per_sec) || \
40 (secno == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
41 sbi->segs_per_sec) || \
42 (secno == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
43 sbi->segs_per_sec) || \
44 (secno == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
45 sbi->segs_per_sec) || \
46 (secno == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
47 sbi->segs_per_sec)) \
48
7cd8558b
JK
49#define MAIN_BLKADDR(sbi) (SM_I(sbi)->main_blkaddr)
50#define SEG0_BLKADDR(sbi) (SM_I(sbi)->seg0_blkaddr)
51
52#define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
53#define MAIN_SECS(sbi) (sbi->total_sections)
54
55#define TOTAL_SEGS(sbi) (SM_I(sbi)->segment_count)
56#define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << sbi->log_blocks_per_seg)
57
58#define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
8a21984d 59#define SEGMENT_SIZE(sbi) (1ULL << (sbi->log_blocksize + \
7cd8558b
JK
60 sbi->log_blocks_per_seg))
61
62#define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
39a53e0c 63 (GET_R2L_SEGNO(FREE_I(sbi), segno) << sbi->log_blocks_per_seg))
7cd8558b 64
39a53e0c
JK
65#define NEXT_FREE_BLKADDR(sbi, curseg) \
66 (START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff)
67
7cd8558b 68#define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
39a53e0c
JK
69#define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
70 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> sbi->log_blocks_per_seg)
491c0854
JK
71#define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
72 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (sbi->blocks_per_seg - 1))
73
39a53e0c
JK
74#define GET_SEGNO(sbi, blk_addr) \
75 (((blk_addr == NULL_ADDR) || (blk_addr == NEW_ADDR)) ? \
76 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
77 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
78#define GET_SECNO(sbi, segno) \
79 ((segno) / sbi->segs_per_sec)
80#define GET_ZONENO_FROM_SEGNO(sbi, segno) \
81 ((segno / sbi->segs_per_sec) / sbi->secs_per_zone)
82
83#define GET_SUM_BLOCK(sbi, segno) \
84 ((sbi->sm_info->ssa_blkaddr) + segno)
85
86#define GET_SUM_TYPE(footer) ((footer)->entry_type)
87#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = type)
88
89#define SIT_ENTRY_OFFSET(sit_i, segno) \
90 (segno % sit_i->sents_per_block)
d3a14afd 91#define SIT_BLOCK_OFFSET(segno) \
39a53e0c 92 (segno / SIT_ENTRY_PER_BLOCK)
d3a14afd
CY
93#define START_SEGNO(segno) \
94 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
74de593a 95#define SIT_BLK_CNT(sbi) \
7cd8558b 96 ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
39a53e0c
JK
97#define f2fs_bitmap_size(nr) \
98 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
39a53e0c 99
55cf9cb6
CY
100#define SECTOR_FROM_BLOCK(blk_addr) \
101 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
102#define SECTOR_TO_BLOCK(sectors) \
103 (sectors >> F2FS_LOG_SECTORS_PER_BLOCK)
90a893c7
JK
104#define MAX_BIO_BLOCKS(sbi) \
105 ((int)min((int)max_hw_blocks(sbi), BIO_MAX_PAGES))
3cd8a239 106
39a53e0c
JK
107/*
108 * indicate a block allocation direction: RIGHT and LEFT.
109 * RIGHT means allocating new sections towards the end of volume.
110 * LEFT means the opposite direction.
111 */
112enum {
113 ALLOC_RIGHT = 0,
114 ALLOC_LEFT
115};
116
117/*
118 * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
119 * LFS writes data sequentially with cleaning operations.
120 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
121 */
122enum {
123 LFS = 0,
124 SSR
125};
126
127/*
128 * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
129 * GC_CB is based on cost-benefit algorithm.
130 * GC_GREEDY is based on greedy algorithm.
131 */
132enum {
133 GC_CB = 0,
134 GC_GREEDY
135};
136
137/*
138 * BG_GC means the background cleaning job.
139 * FG_GC means the on-demand cleaning job.
6aefd93b 140 * FORCE_FG_GC means on-demand cleaning job in background.
39a53e0c
JK
141 */
142enum {
143 BG_GC = 0,
6aefd93b
JK
144 FG_GC,
145 FORCE_FG_GC,
39a53e0c
JK
146};
147
148/* for a function parameter to select a victim segment */
149struct victim_sel_policy {
150 int alloc_mode; /* LFS or SSR */
151 int gc_mode; /* GC_CB or GC_GREEDY */
152 unsigned long *dirty_segmap; /* dirty segment bitmap */
a26b7c8a 153 unsigned int max_search; /* maximum # of segments to search */
39a53e0c
JK
154 unsigned int offset; /* last scanned bitmap offset */
155 unsigned int ofs_unit; /* bitmap search unit */
156 unsigned int min_cost; /* minimum cost */
157 unsigned int min_segno; /* segment # having min. cost */
158};
159
160struct seg_entry {
f51b4ce6
CY
161 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
162 unsigned int valid_blocks:10; /* # of valid blocks */
163 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
164 unsigned int padding:6; /* padding */
39a53e0c
JK
165 unsigned char *cur_valid_map; /* validity bitmap of blocks */
166 /*
167 * # of valid blocks and the validity bitmap stored in the the last
168 * checkpoint pack. This information is used by the SSR mode.
169 */
f51b4ce6 170 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
a66cdd98 171 unsigned char *discard_map;
39a53e0c
JK
172 unsigned long long mtime; /* modification time of the segment */
173};
174
175struct sec_entry {
176 unsigned int valid_blocks; /* # of valid blocks in a section */
177};
178
179struct segment_allocation {
180 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
181};
182
decd36b6
CY
183/*
184 * this value is set in page as a private data which indicate that
185 * the page is atomically written, and it is in inmem_pages list.
186 */
d48dfc20 187#define ATOMIC_WRITTEN_PAGE ((unsigned long)-1)
decd36b6
CY
188
189#define IS_ATOMIC_WRITTEN_PAGE(page) \
190 (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
191
88b88a66
JK
192struct inmem_pages {
193 struct list_head list;
194 struct page *page;
28bc106b 195 block_t old_addr; /* for revoking when fail to commit */
88b88a66
JK
196};
197
39a53e0c
JK
198struct sit_info {
199 const struct segment_allocation *s_ops;
200
201 block_t sit_base_addr; /* start block address of SIT area */
202 block_t sit_blocks; /* # of blocks used by SIT area */
203 block_t written_valid_blocks; /* # of valid blocks in main area */
204 char *sit_bitmap; /* SIT bitmap pointer */
205 unsigned int bitmap_size; /* SIT bitmap size */
206
60a3b782 207 unsigned long *tmp_map; /* bitmap for temporal use */
39a53e0c
JK
208 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
209 unsigned int dirty_sentries; /* # of dirty sentries */
210 unsigned int sents_per_block; /* # of SIT entries per block */
211 struct mutex sentry_lock; /* to protect SIT cache */
212 struct seg_entry *sentries; /* SIT segment-level cache */
213 struct sec_entry *sec_entries; /* SIT section-level cache */
214
215 /* for cost-benefit algorithm in cleaning procedure */
216 unsigned long long elapsed_time; /* elapsed time after mount */
217 unsigned long long mounted_time; /* mount time */
218 unsigned long long min_mtime; /* min. modification time */
219 unsigned long long max_mtime; /* max. modification time */
220};
221
222struct free_segmap_info {
223 unsigned int start_segno; /* start segment number logically */
224 unsigned int free_segments; /* # of free segments */
225 unsigned int free_sections; /* # of free sections */
1a118ccf 226 spinlock_t segmap_lock; /* free segmap lock */
39a53e0c
JK
227 unsigned long *free_segmap; /* free segment bitmap */
228 unsigned long *free_secmap; /* free section bitmap */
229};
230
231/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
232enum dirty_type {
233 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
234 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
235 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
236 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
237 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
238 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
239 DIRTY, /* to count # of dirty segments */
240 PRE, /* to count # of entirely obsolete segments */
241 NR_DIRTY_TYPE
242};
243
244struct dirty_seglist_info {
245 const struct victim_selection *v_ops; /* victim selction operation */
246 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
247 struct mutex seglist_lock; /* lock for segment bitmaps */
248 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
5ec4e49f 249 unsigned long *victim_secmap; /* background GC victims */
39a53e0c
JK
250};
251
252/* victim selection function for cleaning and SSR */
253struct victim_selection {
254 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
255 int, int, char);
256};
257
258/* for active log information */
259struct curseg_info {
260 struct mutex curseg_mutex; /* lock for consistency */
261 struct f2fs_summary_block *sum_blk; /* cached summary block */
b7ad7512
CY
262 struct rw_semaphore journal_rwsem; /* protect journal area */
263 struct f2fs_journal *journal; /* cached journal info */
39a53e0c
JK
264 unsigned char alloc_type; /* current allocation type */
265 unsigned int segno; /* current segment number */
266 unsigned short next_blkoff; /* next block offset to write */
267 unsigned int zone; /* current zone number */
268 unsigned int next_segno; /* preallocated segment */
269};
270
184a5cd2
CY
271struct sit_entry_set {
272 struct list_head set_list; /* link with all sit sets */
273 unsigned int start_segno; /* start segno of sits in set */
274 unsigned int entry_cnt; /* the # of sit entries in set */
275};
276
39a53e0c
JK
277/*
278 * inline functions
279 */
280static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
281{
282 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
283}
284
285static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
286 unsigned int segno)
287{
288 struct sit_info *sit_i = SIT_I(sbi);
289 return &sit_i->sentries[segno];
290}
291
292static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
293 unsigned int segno)
294{
295 struct sit_info *sit_i = SIT_I(sbi);
296 return &sit_i->sec_entries[GET_SECNO(sbi, segno)];
297}
298
299static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
300 unsigned int segno, int section)
301{
302 /*
303 * In order to get # of valid blocks in a section instantly from many
304 * segments, f2fs manages two counting structures separately.
305 */
306 if (section > 1)
307 return get_sec_entry(sbi, segno)->valid_blocks;
308 else
309 return get_seg_entry(sbi, segno)->valid_blocks;
310}
311
312static inline void seg_info_from_raw_sit(struct seg_entry *se,
313 struct f2fs_sit_entry *rs)
314{
315 se->valid_blocks = GET_SIT_VBLOCKS(rs);
316 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
317 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
318 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
319 se->type = GET_SIT_TYPE(rs);
320 se->mtime = le64_to_cpu(rs->mtime);
321}
322
323static inline void seg_info_to_raw_sit(struct seg_entry *se,
324 struct f2fs_sit_entry *rs)
325{
326 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
327 se->valid_blocks;
328 rs->vblocks = cpu_to_le16(raw_vblocks);
329 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
330 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
331 se->ckpt_valid_blocks = se->valid_blocks;
332 rs->mtime = cpu_to_le64(se->mtime);
333}
334
335static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
336 unsigned int max, unsigned int segno)
337{
338 unsigned int ret;
1a118ccf 339 spin_lock(&free_i->segmap_lock);
39a53e0c 340 ret = find_next_bit(free_i->free_segmap, max, segno);
1a118ccf 341 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK
342 return ret;
343}
344
345static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
346{
347 struct free_segmap_info *free_i = FREE_I(sbi);
348 unsigned int secno = segno / sbi->segs_per_sec;
349 unsigned int start_segno = secno * sbi->segs_per_sec;
350 unsigned int next;
351
1a118ccf 352 spin_lock(&free_i->segmap_lock);
39a53e0c
JK
353 clear_bit(segno, free_i->free_segmap);
354 free_i->free_segments++;
355
7fd97019
WL
356 next = find_next_bit(free_i->free_segmap,
357 start_segno + sbi->segs_per_sec, start_segno);
39a53e0c
JK
358 if (next >= start_segno + sbi->segs_per_sec) {
359 clear_bit(secno, free_i->free_secmap);
360 free_i->free_sections++;
361 }
1a118ccf 362 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK
363}
364
365static inline void __set_inuse(struct f2fs_sb_info *sbi,
366 unsigned int segno)
367{
368 struct free_segmap_info *free_i = FREE_I(sbi);
369 unsigned int secno = segno / sbi->segs_per_sec;
370 set_bit(segno, free_i->free_segmap);
371 free_i->free_segments--;
372 if (!test_and_set_bit(secno, free_i->free_secmap))
373 free_i->free_sections--;
374}
375
376static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
377 unsigned int segno)
378{
379 struct free_segmap_info *free_i = FREE_I(sbi);
380 unsigned int secno = segno / sbi->segs_per_sec;
381 unsigned int start_segno = secno * sbi->segs_per_sec;
382 unsigned int next;
383
1a118ccf 384 spin_lock(&free_i->segmap_lock);
39a53e0c
JK
385 if (test_and_clear_bit(segno, free_i->free_segmap)) {
386 free_i->free_segments++;
387
f1121ab0
CY
388 next = find_next_bit(free_i->free_segmap,
389 start_segno + sbi->segs_per_sec, start_segno);
39a53e0c
JK
390 if (next >= start_segno + sbi->segs_per_sec) {
391 if (test_and_clear_bit(secno, free_i->free_secmap))
392 free_i->free_sections++;
393 }
394 }
1a118ccf 395 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK
396}
397
398static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
399 unsigned int segno)
400{
401 struct free_segmap_info *free_i = FREE_I(sbi);
402 unsigned int secno = segno / sbi->segs_per_sec;
1a118ccf 403 spin_lock(&free_i->segmap_lock);
39a53e0c
JK
404 if (!test_and_set_bit(segno, free_i->free_segmap)) {
405 free_i->free_segments--;
406 if (!test_and_set_bit(secno, free_i->free_secmap))
407 free_i->free_sections--;
408 }
1a118ccf 409 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK
410}
411
412static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
413 void *dst_addr)
414{
415 struct sit_info *sit_i = SIT_I(sbi);
416 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
417}
418
419static inline block_t written_block_count(struct f2fs_sb_info *sbi)
420{
8b8343fa 421 return SIT_I(sbi)->written_valid_blocks;
39a53e0c
JK
422}
423
424static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
425{
8b8343fa 426 return FREE_I(sbi)->free_segments;
39a53e0c
JK
427}
428
429static inline int reserved_segments(struct f2fs_sb_info *sbi)
430{
431 return SM_I(sbi)->reserved_segments;
432}
433
434static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
435{
8b8343fa 436 return FREE_I(sbi)->free_sections;
39a53e0c
JK
437}
438
439static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
440{
441 return DIRTY_I(sbi)->nr_dirty[PRE];
442}
443
444static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
445{
446 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
447 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
448 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
449 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
450 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
451 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
452}
453
454static inline int overprovision_segments(struct f2fs_sb_info *sbi)
455{
456 return SM_I(sbi)->ovp_segments;
457}
458
459static inline int overprovision_sections(struct f2fs_sb_info *sbi)
460{
461 return ((unsigned int) overprovision_segments(sbi)) / sbi->segs_per_sec;
462}
463
464static inline int reserved_sections(struct f2fs_sb_info *sbi)
465{
466 return ((unsigned int) reserved_segments(sbi)) / sbi->segs_per_sec;
467}
468
469static inline bool need_SSR(struct f2fs_sb_info *sbi)
470{
95dd8973
JK
471 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
472 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
473 return free_sections(sbi) <= (node_secs + 2 * dent_secs +
474 reserved_sections(sbi) + 1);
39a53e0c
JK
475}
476
43727527 477static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, int freed)
39a53e0c 478{
5ac206cf
NJ
479 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
480 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
43727527 481
caf0047e 482 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
029cd28c
JK
483 return false;
484
6c311ec6
CF
485 return (free_sections(sbi) + freed) <= (node_secs + 2 * dent_secs +
486 reserved_sections(sbi));
39a53e0c
JK
487}
488
81eb8d6e
JK
489static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
490{
6c311ec6 491 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
81eb8d6e
JK
492}
493
39a53e0c
JK
494static inline int utilization(struct f2fs_sb_info *sbi)
495{
6c311ec6
CF
496 return div_u64((u64)valid_user_blocks(sbi) * 100,
497 sbi->user_block_count);
39a53e0c
JK
498}
499
500/*
501 * Sometimes f2fs may be better to drop out-of-place update policy.
216fbd64
JK
502 * And, users can control the policy through sysfs entries.
503 * There are five policies with triggering conditions as follows.
504 * F2FS_IPU_FORCE - all the time,
505 * F2FS_IPU_SSR - if SSR mode is activated,
506 * F2FS_IPU_UTIL - if FS utilization is over threashold,
507 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
508 * threashold,
c1ce1b02
JK
509 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
510 * storages. IPU will be triggered only if the # of dirty
511 * pages over min_fsync_blocks.
216fbd64 512 * F2FS_IPUT_DISABLE - disable IPU. (=default option)
39a53e0c 513 */
216fbd64 514#define DEF_MIN_IPU_UTIL 70
c1ce1b02 515#define DEF_MIN_FSYNC_BLOCKS 8
216fbd64
JK
516
517enum {
518 F2FS_IPU_FORCE,
519 F2FS_IPU_SSR,
520 F2FS_IPU_UTIL,
521 F2FS_IPU_SSR_UTIL,
c1ce1b02 522 F2FS_IPU_FSYNC,
216fbd64
JK
523};
524
39a53e0c
JK
525static inline bool need_inplace_update(struct inode *inode)
526{
4081363f 527 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9b5f136f 528 unsigned int policy = SM_I(sbi)->ipu_policy;
216fbd64
JK
529
530 /* IPU can be done only for the user data */
88b88a66 531 if (S_ISDIR(inode->i_mode) || f2fs_is_atomic_file(inode))
39a53e0c 532 return false;
216fbd64 533
9b5f136f 534 if (policy & (0x1 << F2FS_IPU_FORCE))
39a53e0c 535 return true;
9b5f136f
JK
536 if (policy & (0x1 << F2FS_IPU_SSR) && need_SSR(sbi))
537 return true;
538 if (policy & (0x1 << F2FS_IPU_UTIL) &&
539 utilization(sbi) > SM_I(sbi)->min_ipu_util)
540 return true;
541 if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && need_SSR(sbi) &&
542 utilization(sbi) > SM_I(sbi)->min_ipu_util)
543 return true;
544
545 /* this is only set during fdatasync */
546 if (policy & (0x1 << F2FS_IPU_FSYNC) &&
547 is_inode_flag_set(F2FS_I(inode), FI_NEED_IPU))
548 return true;
549
39a53e0c
JK
550 return false;
551}
552
553static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
554 int type)
555{
556 struct curseg_info *curseg = CURSEG_I(sbi, type);
557 return curseg->segno;
558}
559
560static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
561 int type)
562{
563 struct curseg_info *curseg = CURSEG_I(sbi, type);
564 return curseg->alloc_type;
565}
566
567static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
568{
569 struct curseg_info *curseg = CURSEG_I(sbi, type);
570 return curseg->next_blkoff;
571}
572
573static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
574{
7a04f64d 575 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
39a53e0c
JK
576}
577
39a53e0c
JK
578static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr)
579{
7a04f64d
LX
580 f2fs_bug_on(sbi, blk_addr < SEG0_BLKADDR(sbi)
581 || blk_addr >= MAX_BLKADDR(sbi));
39a53e0c
JK
582}
583
584/*
e1c42045 585 * Summary block is always treated as an invalid block
39a53e0c
JK
586 */
587static inline void check_block_count(struct f2fs_sb_info *sbi,
588 int segno, struct f2fs_sit_entry *raw_sit)
589{
4c278394 590#ifdef CONFIG_F2FS_CHECK_FS
44c60bf2 591 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
39a53e0c 592 int valid_blocks = 0;
44c60bf2 593 int cur_pos = 0, next_pos;
39a53e0c 594
39a53e0c 595 /* check bitmap with valid block count */
44c60bf2
CY
596 do {
597 if (is_valid) {
598 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
599 sbi->blocks_per_seg,
600 cur_pos);
601 valid_blocks += next_pos - cur_pos;
602 } else
603 next_pos = find_next_bit_le(&raw_sit->valid_map,
604 sbi->blocks_per_seg,
605 cur_pos);
606 cur_pos = next_pos;
607 is_valid = !is_valid;
608 } while (cur_pos < sbi->blocks_per_seg);
39a53e0c 609 BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks);
5d56b671 610#endif
4c278394
JK
611 /* check segment usage, and check boundary of a given segment number */
612 f2fs_bug_on(sbi, GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
613 || segno > TOTAL_SEGS(sbi) - 1);
7a04f64d 614}
39a53e0c
JK
615
616static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
617 unsigned int start)
618{
619 struct sit_info *sit_i = SIT_I(sbi);
d3a14afd 620 unsigned int offset = SIT_BLOCK_OFFSET(start);
39a53e0c
JK
621 block_t blk_addr = sit_i->sit_base_addr + offset;
622
623 check_seg_range(sbi, start);
624
625 /* calculate sit block address */
626 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
627 blk_addr += sit_i->sit_blocks;
628
629 return blk_addr;
630}
631
632static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
633 pgoff_t block_addr)
634{
635 struct sit_info *sit_i = SIT_I(sbi);
636 block_addr -= sit_i->sit_base_addr;
637 if (block_addr < sit_i->sit_blocks)
638 block_addr += sit_i->sit_blocks;
639 else
640 block_addr -= sit_i->sit_blocks;
641
642 return block_addr + sit_i->sit_base_addr;
643}
644
645static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
646{
d3a14afd 647 unsigned int block_off = SIT_BLOCK_OFFSET(start);
39a53e0c 648
c6ac4c0e 649 f2fs_change_bit(block_off, sit_i->sit_bitmap);
39a53e0c
JK
650}
651
652static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
653{
654 struct sit_info *sit_i = SIT_I(sbi);
655 return sit_i->elapsed_time + CURRENT_TIME_SEC.tv_sec -
656 sit_i->mounted_time;
657}
658
659static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
660 unsigned int ofs_in_node, unsigned char version)
661{
662 sum->nid = cpu_to_le32(nid);
663 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
664 sum->version = version;
665}
666
667static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
668{
669 return __start_cp_addr(sbi) +
670 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
671}
672
673static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
674{
675 return __start_cp_addr(sbi) +
676 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
677 - (base + 1) + type;
678}
5ec4e49f
JK
679
680static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
681{
682 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
683 return true;
684 return false;
685}
ac5d156c
JK
686
687static inline unsigned int max_hw_blocks(struct f2fs_sb_info *sbi)
688{
689 struct block_device *bdev = sbi->sb->s_bdev;
690 struct request_queue *q = bdev_get_queue(bdev);
55cf9cb6 691 return SECTOR_TO_BLOCK(queue_max_sectors(q));
ac5d156c 692}
87d6f890
JK
693
694/*
695 * It is very important to gather dirty pages and write at once, so that we can
696 * submit a big bio without interfering other data writes.
697 * By default, 512 pages for directory data,
698 * 512 pages (2MB) * 3 for three types of nodes, and
699 * max_bio_blocks for meta are set.
700 */
701static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
702{
a88a341a 703 if (sbi->sb->s_bdi->wb.dirty_exceeded)
510184c8
JK
704 return 0;
705
a1257023
JK
706 if (type == DATA)
707 return sbi->blocks_per_seg;
708 else if (type == NODE)
87d6f890
JK
709 return 3 * sbi->blocks_per_seg;
710 else if (type == META)
90a893c7 711 return MAX_BIO_BLOCKS(sbi);
87d6f890
JK
712 else
713 return 0;
714}
50c8cdb3
JK
715
716/*
717 * When writing pages, it'd better align nr_to_write for segment size.
718 */
719static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
720 struct writeback_control *wbc)
721{
722 long nr_to_write, desired;
723
724 if (wbc->sync_mode != WB_SYNC_NONE)
725 return 0;
726
727 nr_to_write = wbc->nr_to_write;
728
729 if (type == DATA)
730 desired = 4096;
731 else if (type == NODE)
732 desired = 3 * max_hw_blocks(sbi);
733 else
90a893c7 734 desired = MAX_BIO_BLOCKS(sbi);
50c8cdb3
JK
735
736 wbc->nr_to_write = desired;
737 return desired - nr_to_write;
738}