]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/f2fs/segment.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: introduce __wait_discard_cmd
[mirror_ubuntu-artful-kernel.git] / fs / f2fs / segment.c
CommitLineData
0a8165d7 1/*
351df4b2
JK		 2 * fs/f2fs/segment.c
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/fs.h>
12#include <linux/f2fs_fs.h>
13#include <linux/bio.h>
14#include <linux/blkdev.h>
690e4a3e 15#include <linux/prefetch.h>
6b4afdd7 16#include <linux/kthread.h>
74de593a 17#include <linux/swap.h>
60b99b48 18#include <linux/timer.h>
351df4b2
JK		 19
20#include "f2fs.h"
21#include "segment.h"
22#include "node.h"
9e4ded3f 23#include "trace.h"
6ec178da 24#include <trace/events/f2fs.h>
351df4b2 25
9a7f143a
CL		 26#define __reverse_ffz(x) __reverse_ffs(~(x))
27
7fd9e544 28static struct kmem_cache *discard_entry_slab;
b01a9201 29static struct kmem_cache *discard_cmd_slab;
184a5cd2 30static struct kmem_cache *sit_entry_set_slab;
88b88a66 31static struct kmem_cache *inmem_entry_slab;
7fd9e544 32
f96999c3
JK		 33static unsigned long __reverse_ulong(unsigned char *str)
34{
35 unsigned long tmp = 0;
36 int shift = 24, idx = 0;
37
38#if BITS_PER_LONG == 64
39 shift = 56;
40#endif
41 while (shift >= 0) {
42 tmp |= (unsigned long)str[idx++] << shift;
43 shift -= BITS_PER_BYTE;
44 }
45 return tmp;
46}
47
9a7f143a
CL		 48/*
49 * __reverse_ffs is copied from include/asm-generic/bitops/__ffs.h since
50 * MSB and LSB are reversed in a byte by f2fs_set_bit.
51 */
52static inline unsigned long __reverse_ffs(unsigned long word)
53{
54 int num = 0;
55
56#if BITS_PER_LONG == 64
f96999c3 57 if ((word & 0xffffffff00000000UL) == 0)
9a7f143a 58 num += 32;
f96999c3 59 else
9a7f143a 60 word >>= 32;
9a7f143a 61#endif
f96999c3 62 if ((word & 0xffff0000) == 0)
9a7f143a 63 num += 16;
f96999c3 64 else
9a7f143a 65 word >>= 16;
f96999c3
JK		 66
67 if ((word & 0xff00) == 0)
9a7f143a 68 num += 8;
f96999c3 69 else
9a7f143a 70 word >>= 8;
f96999c3 71
9a7f143a
CL		 72 if ((word & 0xf0) == 0)
73 num += 4;
74 else
75 word >>= 4;
f96999c3 76
9a7f143a
CL		 77 if ((word & 0xc) == 0)
78 num += 2;
79 else
80 word >>= 2;
f96999c3 81
9a7f143a
CL		 82 if ((word & 0x2) == 0)
83 num += 1;
84 return num;
85}
86
87/*
e1c42045 88 * __find_rev_next(_zero)_bit is copied from lib/find_next_bit.c because
9a7f143a 89 * f2fs_set_bit makes MSB and LSB reversed in a byte.
692223d1 90 * @size must be integral times of unsigned long.
9a7f143a 91 * Example:
f96999c3
JK		 92 * MSB <--> LSB
93 * f2fs_set_bit(0, bitmap) => 1000 0000
94 * f2fs_set_bit(7, bitmap) => 0000 0001
9a7f143a
CL		 95 */
96static unsigned long __find_rev_next_bit(const unsigned long *addr,
97 unsigned long size, unsigned long offset)
98{
99 const unsigned long *p = addr + BIT_WORD(offset);
692223d1 100 unsigned long result = size;
9a7f143a 101 unsigned long tmp;
9a7f143a
CL		 102
103 if (offset >= size)
104 return size;
105
692223d1 106 size -= (offset & ~(BITS_PER_LONG - 1));
9a7f143a 107 offset %= BITS_PER_LONG;
f96999c3 108
692223d1
FL		 109 while (1) {
110 if (*p == 0)
111 goto pass;
9a7f143a 112
f96999c3 113 tmp = __reverse_ulong((unsigned char *)p);
692223d1
FL		 114
115 tmp &= ~0UL >> offset;
116 if (size < BITS_PER_LONG)
117 tmp &= (~0UL << (BITS_PER_LONG - size));
9a7f143a 118 if (tmp)
692223d1
FL		 119 goto found;
120pass:
121 if (size <= BITS_PER_LONG)
122 break;
9a7f143a 123 size -= BITS_PER_LONG;
692223d1 124 offset = 0;
f96999c3 125 p++;
9a7f143a 126 }
692223d1
FL		 127 return result;
128found:
129 return result - size + __reverse_ffs(tmp);
9a7f143a
CL		 130}
131
132static unsigned long __find_rev_next_zero_bit(const unsigned long *addr,
133 unsigned long size, unsigned long offset)
134{
135 const unsigned long *p = addr + BIT_WORD(offset);
80609448 136 unsigned long result = size;
9a7f143a 137 unsigned long tmp;
9a7f143a
CL		 138
139 if (offset >= size)
140 return size;
141
80609448 142 size -= (offset & ~(BITS_PER_LONG - 1));
9a7f143a 143 offset %= BITS_PER_LONG;
80609448
JK		 144
145 while (1) {
146 if (*p == ~0UL)
147 goto pass;
148
f96999c3 149 tmp = __reverse_ulong((unsigned char *)p);
80609448
JK		 150
151 if (offset)
152 tmp |= ~0UL << (BITS_PER_LONG - offset);
153 if (size < BITS_PER_LONG)
154 tmp |= ~0UL >> size;
f96999c3 155 if (tmp != ~0UL)
80609448
JK		 156 goto found;
157pass:
158 if (size <= BITS_PER_LONG)
159 break;
9a7f143a 160 size -= BITS_PER_LONG;
80609448 161 offset = 0;
f96999c3 162 p++;
9a7f143a 163 }
80609448
JK		 164 return result;
165found:
166 return result - size + __reverse_ffz(tmp);
9a7f143a
CL		 167}
168
88b88a66
JK		 169void register_inmem_page(struct inode *inode, struct page *page)
170{
171 struct f2fs_inode_info *fi = F2FS_I(inode);
172 struct inmem_pages *new;
9be32d72 173
9e4ded3f 174 f2fs_trace_pid(page);
0722b101 175
decd36b6
CY		 176 set_page_private(page, (unsigned long)ATOMIC_WRITTEN_PAGE);
177 SetPagePrivate(page);
178
88b88a66
JK		 179 new = f2fs_kmem_cache_alloc(inmem_entry_slab, GFP_NOFS);
180
181 /* add atomic page indices to the list */
182 new->page = page;
183 INIT_LIST_HEAD(&new->list);
decd36b6 184
88b88a66
JK		 185 /* increase reference count with clean state */
186 mutex_lock(&fi->inmem_lock);
187 get_page(page);
188 list_add_tail(&new->list, &fi->inmem_pages);
8dcf2ff7 189 inc_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
88b88a66 190 mutex_unlock(&fi->inmem_lock);
8ce67cb0
JK		 191
192 trace_f2fs_register_inmem_page(page, INMEM);
88b88a66
JK		 193}
194
28bc106b
CY		 195static int __revoke_inmem_pages(struct inode *inode,
196 struct list_head *head, bool drop, bool recover)
29b96b54 197{
28bc106b 198 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
29b96b54 199 struct inmem_pages *cur, *tmp;
28bc106b 200 int err = 0;
29b96b54
CY		 201
202 list_for_each_entry_safe(cur, tmp, head, list) {
28bc106b
CY		 203 struct page *page = cur->page;
204
205 if (drop)
206 trace_f2fs_commit_inmem_page(page, INMEM_DROP);
207
208 lock_page(page);
29b96b54 209
28bc106b
CY		 210 if (recover) {
211 struct dnode_of_data dn;
212 struct node_info ni;
213
214 trace_f2fs_commit_inmem_page(page, INMEM_REVOKE);
215
216 set_new_dnode(&dn, inode, NULL, NULL, 0);
217 if (get_dnode_of_data(&dn, page->index, LOOKUP_NODE)) {
218 err = -EAGAIN;
219 goto next;
220 }
221 get_node_info(sbi, dn.nid, &ni);
222 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
223 cur->old_addr, ni.version, true, true);
224 f2fs_put_dnode(&dn);
225 }
226next:
63c52d78
JK		 227 /* we don't need to invalidate this in the sccessful status */
228 if (drop || recover)
229 ClearPageUptodate(page);
28bc106b 230 set_page_private(page, 0);
c81ced05 231 ClearPagePrivate(page);
28bc106b 232 f2fs_put_page(page, 1);
29b96b54
CY		 233
234 list_del(&cur->list);
235 kmem_cache_free(inmem_entry_slab, cur);
236 dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
237 }
28bc106b 238 return err;
29b96b54
CY		 239}
240
241void drop_inmem_pages(struct inode *inode)
242{
243 struct f2fs_inode_info *fi = F2FS_I(inode);
244
245 mutex_lock(&fi->inmem_lock);
28bc106b 246 __revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
29b96b54 247 mutex_unlock(&fi->inmem_lock);
5fe45743
CY		 248
249 clear_inode_flag(inode, FI_ATOMIC_FILE);
250 stat_dec_atomic_write(inode);
29b96b54
CY		 251}
252
8c242db9
JK		 253void drop_inmem_page(struct inode *inode, struct page *page)
254{
255 struct f2fs_inode_info *fi = F2FS_I(inode);
256 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
257 struct list_head *head = &fi->inmem_pages;
258 struct inmem_pages *cur = NULL;
259
260 f2fs_bug_on(sbi, !IS_ATOMIC_WRITTEN_PAGE(page));
261
262 mutex_lock(&fi->inmem_lock);
263 list_for_each_entry(cur, head, list) {
264 if (cur->page == page)
265 break;
266 }
267
268 f2fs_bug_on(sbi, !cur || cur->page != page);
269 list_del(&cur->list);
270 mutex_unlock(&fi->inmem_lock);
271
272 dec_page_count(sbi, F2FS_INMEM_PAGES);
273 kmem_cache_free(inmem_entry_slab, cur);
274
275 ClearPageUptodate(page);
276 set_page_private(page, 0);
277 ClearPagePrivate(page);
278 f2fs_put_page(page, 0);
279
280 trace_f2fs_commit_inmem_page(page, INMEM_INVALIDATE);
281}
282
28bc106b
CY		 283static int __commit_inmem_pages(struct inode *inode,
284 struct list_head *revoke_list)
88b88a66
JK		 285{
286 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
287 struct f2fs_inode_info *fi = F2FS_I(inode);
288 struct inmem_pages *cur, *tmp;
88b88a66 289 struct f2fs_io_info fio = {
05ca3632 290 .sbi = sbi,
88b88a66 291 .type = DATA,
04d328de 292 .op = REQ_OP_WRITE,
70fd7614 293 .op_flags = REQ_SYNC | REQ_PRIO,
4375a336 294 .encrypted_page = NULL,
88b88a66 295 };
942fd319 296 pgoff_t last_idx = ULONG_MAX;
edb27dee 297 int err = 0;
88b88a66 298
88b88a66 299 list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
28bc106b
CY		 300 struct page *page = cur->page;
301
302 lock_page(page);
303 if (page->mapping == inode->i_mapping) {
304 trace_f2fs_commit_inmem_page(page, INMEM);
305
306 set_page_dirty(page);
307 f2fs_wait_on_page_writeback(page, DATA, true);
933439c8 308 if (clear_page_dirty_for_io(page)) {
29b96b54 309 inode_dec_dirty_pages(inode);
933439c8
CY		 310 remove_dirty_inode(inode);
311 }
28bc106b
CY		 312
313 fio.page = page;
29b96b54
CY		 314 err = do_write_data_page(&fio);
315 if (err) {
28bc106b 316 unlock_page(page);
29b96b54 317 break;
70c640b1 318 }
29b96b54 319
28bc106b
CY		 320 /* record old blkaddr for revoking */
321 cur->old_addr = fio.old_blkaddr;
942fd319 322 last_idx = page->index;
28bc106b
CY		 323 }
324 unlock_page(page);
325 list_move_tail(&cur->list, revoke_list);
88b88a66 326 }
29b96b54 327
942fd319
JK		 328 if (last_idx != ULONG_MAX)
329 f2fs_submit_merged_bio_cond(sbi, inode, 0, last_idx,
330 DATA, WRITE);
28bc106b
CY		 331
332 if (!err)
333 __revoke_inmem_pages(inode, revoke_list, false, false);
334
29b96b54
CY		 335 return err;
336}
337
338int commit_inmem_pages(struct inode *inode)
339{
340 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
341 struct f2fs_inode_info *fi = F2FS_I(inode);
28bc106b
CY		 342 struct list_head revoke_list;
343 int err;
29b96b54 344
28bc106b 345 INIT_LIST_HEAD(&revoke_list);
29b96b54
CY		 346 f2fs_balance_fs(sbi, true);
347 f2fs_lock_op(sbi);
348
5fe45743
CY		 349 set_inode_flag(inode, FI_ATOMIC_COMMIT);
350
29b96b54 351 mutex_lock(&fi->inmem_lock);
28bc106b
CY		 352 err = __commit_inmem_pages(inode, &revoke_list);
353 if (err) {
354 int ret;
355 /*
356 * try to revoke all committed pages, but still we could fail
357 * due to no memory or other reason, if that happened, EAGAIN
358 * will be returned, which means in such case, transaction is
359 * already not integrity, caller should use journal to do the
360 * recovery or rewrite & commit last transaction. For other
361 * error number, revoking was done by filesystem itself.
362 */
363 ret = __revoke_inmem_pages(inode, &revoke_list, false, true);
364 if (ret)
365 err = ret;
366
367 /* drop all uncommitted pages */
368 __revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
369 }
88b88a66
JK		 370 mutex_unlock(&fi->inmem_lock);
371
5fe45743
CY		 372 clear_inode_flag(inode, FI_ATOMIC_COMMIT);
373
29b96b54 374 f2fs_unlock_op(sbi);
edb27dee 375 return err;
88b88a66
JK		 376}
377
0a8165d7 378/*
351df4b2
JK		 379 * This function balances dirty node and dentry pages.
380 * In addition, it controls garbage collection.
381 */
2c4db1a6 382void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
351df4b2 383{
0f348028 384#ifdef CONFIG_F2FS_FAULT_INJECTION
55523519
CY		 385 if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
386 f2fs_show_injection_info(FAULT_CHECKPOINT);
0f348028 387 f2fs_stop_checkpoint(sbi, false);
55523519 388 }
0f348028
CY		 389#endif
390
e589c2c4 391 /* balance_fs_bg is able to be pending */
a7881893 392 if (need && excess_cached_nats(sbi))
e589c2c4
JK		 393 f2fs_balance_fs_bg(sbi);
394
351df4b2 395 /*
029cd28c
JK		 396 * We should do GC or end up with checkpoint, if there are so many dirty
397 * dir/node pages without enough free segments.
351df4b2 398 */
7f3037a5 399 if (has_not_enough_free_secs(sbi, 0, 0)) {
351df4b2 400 mutex_lock(&sbi->gc_mutex);
e066b83c 401 f2fs_gc(sbi, false, false, NULL_SEGNO);
351df4b2
JK		 402 }
403}
404
4660f9c0
JK		 405void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi)
406{
1dcc336b 407 /* try to shrink extent cache when there is no enough memory */
554df79e
JK		 408 if (!available_free_memory(sbi, EXTENT_CACHE))
409 f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER);
1dcc336b 410
1b38dc8e
JK		 411 /* check the # of cached NAT entries */
412 if (!available_free_memory(sbi, NAT_ENTRIES))
413 try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK);
414
31696580 415 if (!available_free_memory(sbi, FREE_NIDS))
ad4edb83
JK		 416 try_to_free_nids(sbi, MAX_FREE_NIDS);
417 else
22ad0b6a 418 build_free_nids(sbi, false, false);
31696580 419
f455c8a5
JK		 420 if (!is_idle(sbi))
421 return;
31696580 422
1b38dc8e
JK		 423 /* checkpoint is the only way to shrink partial cached entries */
424 if (!available_free_memory(sbi, NAT_ENTRIES) ||
60b99b48 425 !available_free_memory(sbi, INO_ENTRIES) ||
7d768d2c
CY		 426 excess_prefree_segs(sbi) ||
427 excess_dirty_nats(sbi) ||
f455c8a5 428 f2fs_time_over(sbi, CP_TIME)) {
e9f5b8b8
CY		 429 if (test_opt(sbi, DATA_FLUSH)) {
430 struct blk_plug plug;
431
432 blk_start_plug(&plug);
36b35a0d 433 sync_dirty_inodes(sbi, FILE_INODE);
e9f5b8b8
CY		 434 blk_finish_plug(&plug);
435 }
4660f9c0 436 f2fs_sync_fs(sbi->sb, true);
42190d2a 437 stat_inc_bg_cp_count(sbi->stat_info);
36b35a0d 438 }
4660f9c0
JK		 439}
440
20fda56b
KM		 441static int __submit_flush_wait(struct f2fs_sb_info *sbi,
442 struct block_device *bdev)
3c62be17
JK		 443{
444 struct bio *bio = f2fs_bio_alloc(0);
445 int ret;
446
09cb6464 447 bio->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
3c62be17
JK		 448 bio->bi_bdev = bdev;
449 ret = submit_bio_wait(bio);
450 bio_put(bio);
20fda56b
KM		 451
452 trace_f2fs_issue_flush(bdev, test_opt(sbi, NOBARRIER),
453 test_opt(sbi, FLUSH_MERGE), ret);
3c62be17
JK		 454 return ret;
455}
456
457static int submit_flush_wait(struct f2fs_sb_info *sbi)
458{
20fda56b 459 int ret = __submit_flush_wait(sbi, sbi->sb->s_bdev);
3c62be17
JK		 460 int i;
461
20fda56b
KM		 462 if (!sbi->s_ndevs || ret)
463 return ret;
464
465 for (i = 1; i < sbi->s_ndevs; i++) {
466 ret = __submit_flush_wait(sbi, FDEV(i).bdev);
467 if (ret)
468 break;
3c62be17
JK		 469 }
470 return ret;
471}
472
2163d198 473static int issue_flush_thread(void *data)
6b4afdd7
JK		 474{
475 struct f2fs_sb_info *sbi = data;
b01a9201 476 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
a688b9d9 477 wait_queue_head_t *q = &fcc->flush_wait_queue;
6b4afdd7
JK		 478repeat:
479 if (kthread_should_stop())
480 return 0;
481
721bd4d5 482 if (!llist_empty(&fcc->issue_list)) {
6b4afdd7
JK		 483 struct flush_cmd *cmd, *next;
484 int ret;
485
721bd4d5
GZ		 486 fcc->dispatch_list = llist_del_all(&fcc->issue_list);
487 fcc->dispatch_list = llist_reverse_order(fcc->dispatch_list);
488
3c62be17 489 ret = submit_flush_wait(sbi);
8b8dd65f
CY		 490 atomic_inc(&fcc->issued_flush);
491
721bd4d5
GZ		 492 llist_for_each_entry_safe(cmd, next,
493 fcc->dispatch_list, llnode) {
6b4afdd7 494 cmd->ret = ret;
6b4afdd7
JK		 495 complete(&cmd->wait);
496 }
a688b9d9 497 fcc->dispatch_list = NULL;
6b4afdd7
JK		 498 }
499
a688b9d9 500 wait_event_interruptible(*q,
721bd4d5 501 kthread_should_stop() || !llist_empty(&fcc->issue_list));
6b4afdd7
JK		 502 goto repeat;
503}
504
505int f2fs_issue_flush(struct f2fs_sb_info *sbi)
506{
b01a9201 507 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
adf8d90b 508 struct flush_cmd cmd;
8b8dd65f 509 int ret;
6b4afdd7 510
0f7b2abd
JK		 511 if (test_opt(sbi, NOBARRIER))
512 return 0;
513
8b8dd65f
CY		 514 if (!test_opt(sbi, FLUSH_MERGE)) {
515 ret = submit_flush_wait(sbi);
516 atomic_inc(&fcc->issued_flush);
517 return ret;
518 }
740432f8 519
8b8dd65f
CY		 520 if (!atomic_read(&fcc->issing_flush)) {
521 atomic_inc(&fcc->issing_flush);
3c62be17 522 ret = submit_flush_wait(sbi);
8b8dd65f
CY		 523 atomic_dec(&fcc->issing_flush);
524
525 atomic_inc(&fcc->issued_flush);
740432f8
JK		 526 return ret;
527 }
6b4afdd7 528
adf8d90b 529 init_completion(&cmd.wait);
6b4afdd7 530
8b8dd65f 531 atomic_inc(&fcc->issing_flush);
721bd4d5 532 llist_add(&cmd.llnode, &fcc->issue_list);
6b4afdd7 533
a688b9d9
GZ		 534 if (!fcc->dispatch_list)
535 wake_up(&fcc->flush_wait_queue);
6b4afdd7 536
5eba8c5d
JK		 537 if (fcc->f2fs_issue_flush) {
538 wait_for_completion(&cmd.wait);
8b8dd65f 539 atomic_dec(&fcc->issing_flush);
5eba8c5d
JK		 540 } else {
541 llist_del_all(&fcc->issue_list);
8b8dd65f 542 atomic_set(&fcc->issing_flush, 0);
5eba8c5d 543 }
adf8d90b
CY		 544
545 return cmd.ret;
6b4afdd7
JK		 546}
547
2163d198
GZ		 548int create_flush_cmd_control(struct f2fs_sb_info *sbi)
549{
550 dev_t dev = sbi->sb->s_bdev->bd_dev;
551 struct flush_cmd_control *fcc;
552 int err = 0;
553
b01a9201
JK		 554 if (SM_I(sbi)->fcc_info) {
555 fcc = SM_I(sbi)->fcc_info;
5eba8c5d
JK		 556 goto init_thread;
557 }
558
2163d198
GZ		 559 fcc = kzalloc(sizeof(struct flush_cmd_control), GFP_KERNEL);
560 if (!fcc)
561 return -ENOMEM;
8b8dd65f
CY		 562 atomic_set(&fcc->issued_flush, 0);
563 atomic_set(&fcc->issing_flush, 0);
2163d198 564 init_waitqueue_head(&fcc->flush_wait_queue);
721bd4d5 565 init_llist_head(&fcc->issue_list);
b01a9201 566 SM_I(sbi)->fcc_info = fcc;
5eba8c5d 567init_thread:
2163d198
GZ		 568 fcc->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi,
569 "f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev));
570 if (IS_ERR(fcc->f2fs_issue_flush)) {
571 err = PTR_ERR(fcc->f2fs_issue_flush);
572 kfree(fcc);
b01a9201 573 SM_I(sbi)->fcc_info = NULL;
2163d198
GZ		 574 return err;
575 }
2163d198
GZ		 576
577 return err;
578}
579
5eba8c5d 580void destroy_flush_cmd_control(struct f2fs_sb_info *sbi, bool free)
2163d198 581{
b01a9201 582 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
2163d198 583
5eba8c5d
JK		 584 if (fcc && fcc->f2fs_issue_flush) {
585 struct task_struct *flush_thread = fcc->f2fs_issue_flush;
586
587 fcc->f2fs_issue_flush = NULL;
588 kthread_stop(flush_thread);
589 }
590 if (free) {
591 kfree(fcc);
b01a9201 592 SM_I(sbi)->fcc_info = NULL;
5eba8c5d 593 }
2163d198
GZ		 594}
595
351df4b2
JK		 596static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
597 enum dirty_type dirty_type)
598{
599 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
600
601 /* need not be added */
602 if (IS_CURSEG(sbi, segno))
603 return;
604
605 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[dirty_type]))
606 dirty_i->nr_dirty[dirty_type]++;
607
608 if (dirty_type == DIRTY) {
609 struct seg_entry *sentry = get_seg_entry(sbi, segno);
4625d6aa 610 enum dirty_type t = sentry->type;
b2f2c390 611
ec325b52
JK		 612 if (unlikely(t >= DIRTY)) {
613 f2fs_bug_on(sbi, 1);
614 return;
615 }
4625d6aa
CL		 616 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[t]))
617 dirty_i->nr_dirty[t]++;
351df4b2
JK		 618 }
619}
620
621static void __remove_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
622 enum dirty_type dirty_type)
623{
624 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
625
626 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[dirty_type]))
627 dirty_i->nr_dirty[dirty_type]--;
628
629 if (dirty_type == DIRTY) {
4625d6aa
CL		 630 struct seg_entry *sentry = get_seg_entry(sbi, segno);
631 enum dirty_type t = sentry->type;
632
633 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
634 dirty_i->nr_dirty[t]--;
b2f2c390 635
302bd348 636 if (get_valid_blocks(sbi, segno, true) == 0)
4ddb1a4d 637 clear_bit(GET_SEC_FROM_SEG(sbi, segno),
5ec4e49f 638 dirty_i->victim_secmap);
351df4b2
JK		 639 }
640}
641
0a8165d7 642/*
351df4b2
JK		 643 * Should not occur error such as -ENOMEM.
644 * Adding dirty entry into seglist is not critical operation.
645 * If a given segment is one of current working segments, it won't be added.
646 */
8d8451af 647static void locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 648{
649 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
650 unsigned short valid_blocks;
651
652 if (segno == NULL_SEGNO || IS_CURSEG(sbi, segno))
653 return;
654
655 mutex_lock(&dirty_i->seglist_lock);
656
302bd348 657 valid_blocks = get_valid_blocks(sbi, segno, false);
351df4b2
JK		 658
659 if (valid_blocks == 0) {
660 __locate_dirty_segment(sbi, segno, PRE);
661 __remove_dirty_segment(sbi, segno, DIRTY);
662 } else if (valid_blocks < sbi->blocks_per_seg) {
663 __locate_dirty_segment(sbi, segno, DIRTY);
664 } else {
665 /* Recovery routine with SSR needs this */
666 __remove_dirty_segment(sbi, segno, DIRTY);
667 }
668
669 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 670}
671
004b6862 672static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi,
c81abe34
JK		 673 struct block_device *bdev, block_t lstart,
674 block_t start, block_t len)
275b66b0 675{
0b54fb84 676 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
ba48a33e 677 struct list_head *pend_list;
b01a9201 678 struct discard_cmd *dc;
275b66b0 679
ba48a33e
CY		 680 f2fs_bug_on(sbi, !len);
681
682 pend_list = &dcc->pend_list[plist_idx(len)];
683
b01a9201
JK		 684 dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS);
685 INIT_LIST_HEAD(&dc->list);
c81abe34 686 dc->bdev = bdev;
b01a9201 687 dc->lstart = lstart;
c81abe34 688 dc->start = start;
b01a9201 689 dc->len = len;
15469963 690 dc->state = D_PREP;
c81abe34 691 dc->error = 0;
b01a9201 692 init_completion(&dc->wait);
22d375dd 693 list_add_tail(&dc->list, pend_list);
5f32366a 694 atomic_inc(&dcc->discard_cmd_cnt);
d84d1cbd 695 dcc->undiscard_blks += len;
004b6862
CY		 696
697 return dc;
698}
699
700static struct discard_cmd *__attach_discard_cmd(struct f2fs_sb_info *sbi,
701 struct block_device *bdev, block_t lstart,
702 block_t start, block_t len,
703 struct rb_node *parent, struct rb_node **p)
704{
705 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
706 struct discard_cmd *dc;
707
708 dc = __create_discard_cmd(sbi, bdev, lstart, start, len);
709
710 rb_link_node(&dc->rb_node, parent, p);
711 rb_insert_color(&dc->rb_node, &dcc->root);
712
713 return dc;
15469963
JK		 714}
715
004b6862
CY		 716static void __detach_discard_cmd(struct discard_cmd_control *dcc,
717 struct discard_cmd *dc)
15469963 718{
dcc9165d 719 if (dc->state == D_DONE)
004b6862
CY		 720 atomic_dec(&dcc->issing_discard);
721
722 list_del(&dc->list);
723 rb_erase(&dc->rb_node, &dcc->root);
d84d1cbd 724 dcc->undiscard_blks -= dc->len;
004b6862
CY		 725
726 kmem_cache_free(discard_cmd_slab, dc);
727
728 atomic_dec(&dcc->discard_cmd_cnt);
729}
730
731static void __remove_discard_cmd(struct f2fs_sb_info *sbi,
732 struct discard_cmd *dc)
733{
734 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
dcc9165d 735
c81abe34
JK		 736 if (dc->error == -EOPNOTSUPP)
737 dc->error = 0;
15469963 738
c81abe34 739 if (dc->error)
15469963 740 f2fs_msg(sbi->sb, KERN_INFO,
c81abe34 741 "Issue discard failed, ret: %d", dc->error);
004b6862 742 __detach_discard_cmd(dcc, dc);
275b66b0
CY		 743}
744
c81abe34
JK		 745static void f2fs_submit_discard_endio(struct bio *bio)
746{
747 struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
748
c81abe34
JK		 749 dc->error = bio->bi_error;
750 dc->state = D_DONE;
fa64a003 751 complete(&dc->wait);
c81abe34
JK		 752 bio_put(bio);
753}
754
755/* this function is copied from blkdev_issue_discard from block/blk-lib.c */
756static void __submit_discard_cmd(struct f2fs_sb_info *sbi,
757 struct discard_cmd *dc)
758{
759 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
760 struct bio *bio = NULL;
761
762 if (dc->state != D_PREP)
763 return;
764
0243a5f9
CY		 765 trace_f2fs_issue_discard(dc->bdev, dc->start, dc->len);
766
c81abe34
JK		 767 dc->error = __blkdev_issue_discard(dc->bdev,
768 SECTOR_FROM_BLOCK(dc->start),
769 SECTOR_FROM_BLOCK(dc->len),
770 GFP_NOFS, 0, &bio);
771 if (!dc->error) {
772 /* should keep before submission to avoid D_DONE right away */
773 dc->state = D_SUBMIT;
8b8dd65f
CY		 774 atomic_inc(&dcc->issued_discard);
775 atomic_inc(&dcc->issing_discard);
c81abe34
JK		 776 if (bio) {
777 bio->bi_private = dc;
778 bio->bi_end_io = f2fs_submit_discard_endio;
779 bio->bi_opf |= REQ_SYNC;
780 submit_bio(bio);
46f84c2c 781 list_move_tail(&dc->list, &dcc->wait_list);
c81abe34
JK		 782 }
783 } else {
784 __remove_discard_cmd(sbi, dc);
785 }
786}
787
004b6862
CY		 788static struct discard_cmd *__insert_discard_tree(struct f2fs_sb_info *sbi,
789 struct block_device *bdev, block_t lstart,
790 block_t start, block_t len,
791 struct rb_node **insert_p,
792 struct rb_node *insert_parent)
c81abe34 793{
004b6862
CY		 794 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
795 struct rb_node **p = &dcc->root.rb_node;
796 struct rb_node *parent = NULL;
797 struct discard_cmd *dc = NULL;
798
799 if (insert_p && insert_parent) {
800 parent = insert_parent;
801 p = insert_p;
802 goto do_insert;
803 }
c81abe34 804
004b6862
CY		 805 p = __lookup_rb_tree_for_insert(sbi, &dcc->root, &parent, lstart);
806do_insert:
807 dc = __attach_discard_cmd(sbi, bdev, lstart, start, len, parent, p);
808 if (!dc)
809 return NULL;
c81abe34 810
004b6862 811 return dc;
c81abe34
JK		 812}
813
ba48a33e
CY		 814static void __relocate_discard_cmd(struct discard_cmd_control *dcc,
815 struct discard_cmd *dc)
816{
817 list_move_tail(&dc->list, &dcc->pend_list[plist_idx(dc->len)]);
818}
819
3d6a650f
YH		 820static void __punch_discard_cmd(struct f2fs_sb_info *sbi,
821 struct discard_cmd *dc, block_t blkaddr)
822{
ba48a33e 823 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 824 struct discard_info di = dc->di;
825 bool modified = false;
3d6a650f 826
004b6862 827 if (dc->state == D_DONE || dc->len == 1) {
3d6a650f
YH		 828 __remove_discard_cmd(sbi, dc);
829 return;
830 }
831
d84d1cbd
CY		 832 dcc->undiscard_blks -= di.len;
833
004b6862 834 if (blkaddr > di.lstart) {
3d6a650f 835 dc->len = blkaddr - dc->lstart;
d84d1cbd 836 dcc->undiscard_blks += dc->len;
ba48a33e 837 __relocate_discard_cmd(dcc, dc);
df0f6b44 838 f2fs_bug_on(sbi, !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862
CY		 839 modified = true;
840 }
841
842 if (blkaddr < di.lstart + di.len - 1) {
843 if (modified) {
844 __insert_discard_tree(sbi, dc->bdev, blkaddr + 1,
845 di.start + blkaddr + 1 - di.lstart,
846 di.lstart + di.len - 1 - blkaddr,
847 NULL, NULL);
df0f6b44
CY		 848 f2fs_bug_on(sbi,
849 !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862
CY		 850 } else {
851 dc->lstart++;
852 dc->len--;
853 dc->start++;
d84d1cbd 854 dcc->undiscard_blks += dc->len;
ba48a33e 855 __relocate_discard_cmd(dcc, dc);
df0f6b44
CY		 856 f2fs_bug_on(sbi,
857 !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862 858 }
3d6a650f
YH		 859 }
860}
861
004b6862
CY		 862static void __update_discard_tree_range(struct f2fs_sb_info *sbi,
863 struct block_device *bdev, block_t lstart,
864 block_t start, block_t len)
275b66b0 865{
0b54fb84 866 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 867 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
868 struct discard_cmd *dc;
869 struct discard_info di = {0};
870 struct rb_node **insert_p = NULL, *insert_parent = NULL;
871 block_t end = lstart + len;
275b66b0 872
15469963 873 mutex_lock(&dcc->cmd_lock);
40465257 874
004b6862
CY		 875 dc = (struct discard_cmd *)__lookup_rb_tree_ret(&dcc->root,
876 NULL, lstart,
877 (struct rb_entry **)&prev_dc,
878 (struct rb_entry **)&next_dc,
879 &insert_p, &insert_parent, true);
880 if (dc)
881 prev_dc = dc;
882
883 if (!prev_dc) {
884 di.lstart = lstart;
885 di.len = next_dc ? next_dc->lstart - lstart : len;
886 di.len = min(di.len, len);
887 di.start = start;
22d375dd 888 }
15469963 889
004b6862
CY		 890 while (1) {
891 struct rb_node *node;
892 bool merged = false;
893 struct discard_cmd *tdc = NULL;
894
895 if (prev_dc) {
896 di.lstart = prev_dc->lstart + prev_dc->len;
897 if (di.lstart < lstart)
898 di.lstart = lstart;
899 if (di.lstart >= end)
900 break;
901
902 if (!next_dc || next_dc->lstart > end)
903 di.len = end - di.lstart;
904 else
905 di.len = next_dc->lstart - di.lstart;
906 di.start = start + di.lstart - lstart;
907 }
908
909 if (!di.len)
910 goto next;
911
912 if (prev_dc && prev_dc->state == D_PREP &&
913 prev_dc->bdev == bdev &&
914 __is_discard_back_mergeable(&di, &prev_dc->di)) {
915 prev_dc->di.len += di.len;
d84d1cbd 916 dcc->undiscard_blks += di.len;
ba48a33e 917 __relocate_discard_cmd(dcc, prev_dc);
df0f6b44
CY		 918 f2fs_bug_on(sbi,
919 !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862
CY		 920 di = prev_dc->di;
921 tdc = prev_dc;
922 merged = true;
923 }
924
925 if (next_dc && next_dc->state == D_PREP &&
926 next_dc->bdev == bdev &&
927 __is_discard_front_mergeable(&di, &next_dc->di)) {
928 next_dc->di.lstart = di.lstart;
929 next_dc->di.len += di.len;
930 next_dc->di.start = di.start;
d84d1cbd 931 dcc->undiscard_blks += di.len;
ba48a33e 932 __relocate_discard_cmd(dcc, next_dc);
004b6862
CY		 933 if (tdc)
934 __remove_discard_cmd(sbi, tdc);
df0f6b44
CY		 935 f2fs_bug_on(sbi,
936 !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862 937 merged = true;
4e6a8d9b 938 }
004b6862 939
df0f6b44 940 if (!merged) {
004b6862
CY		 941 __insert_discard_tree(sbi, bdev, di.lstart, di.start,
942 di.len, NULL, NULL);
df0f6b44
CY		 943 f2fs_bug_on(sbi,
944 !__check_rb_tree_consistence(sbi, &dcc->root));
945 }
004b6862
CY		 946 next:
947 prev_dc = next_dc;
948 if (!prev_dc)
949 break;
950
951 node = rb_next(&prev_dc->rb_node);
952 next_dc = rb_entry_safe(node, struct discard_cmd, rb_node);
953 }
954
955 mutex_unlock(&dcc->cmd_lock);
956}
957
958static int __queue_discard_cmd(struct f2fs_sb_info *sbi,
959 struct block_device *bdev, block_t blkstart, block_t blklen)
960{
961 block_t lblkstart = blkstart;
962
0243a5f9 963 trace_f2fs_queue_discard(bdev, blkstart, blklen);
004b6862
CY		 964
965 if (sbi->s_ndevs) {
966 int devi = f2fs_target_device_index(sbi, blkstart);
967
968 blkstart -= FDEV(devi).start_blk;
969 }
970 __update_discard_tree_range(sbi, bdev, lblkstart, blkstart, blklen);
004b6862
CY		 971 return 0;
972}
973
bd5b0738
CY		 974static void __issue_discard_cmd(struct f2fs_sb_info *sbi, bool issue_cond)
975{
976 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
977 struct list_head *pend_list;
978 struct discard_cmd *dc, *tmp;
979 struct blk_plug plug;
980 int i, iter = 0;
981
982 mutex_lock(&dcc->cmd_lock);
983 blk_start_plug(&plug);
984 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
985 pend_list = &dcc->pend_list[i];
986 list_for_each_entry_safe(dc, tmp, pend_list, list) {
987 f2fs_bug_on(sbi, dc->state != D_PREP);
988
989 if (!issue_cond || is_idle(sbi))
990 __submit_discard_cmd(sbi, dc);
991 if (issue_cond && iter++ > DISCARD_ISSUE_RATE)
992 goto out;
993 }
994 }
995out:
996 blk_finish_plug(&plug);
997 mutex_unlock(&dcc->cmd_lock);
998}
999
63a94fa1
CY		 1000static void __wait_discard_cmd(struct f2fs_sb_info *sbi, bool wait_cond)
1001{
1002 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1003 struct list_head *wait_list = &(dcc->wait_list);
1004 struct discard_cmd *dc, *tmp;
1005
1006 mutex_lock(&dcc->cmd_lock);
1007 list_for_each_entry_safe(dc, tmp, wait_list, list) {
1008 if (!wait_cond || dc->state == D_DONE) {
1009 wait_for_completion_io(&dc->wait);
1010 __remove_discard_cmd(sbi, dc);
1011 }
1012 }
1013 mutex_unlock(&dcc->cmd_lock);
1014}
1015
004b6862
CY		 1016/* This should be covered by global mutex, &sit_i->sentry_lock */
1017void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
1018{
1019 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1020 struct discard_cmd *dc;
1021
1022 mutex_lock(&dcc->cmd_lock);
1023
1024 dc = (struct discard_cmd *)__lookup_rb_tree(&dcc->root, NULL, blkaddr);
1025 if (dc) {
1026 if (dc->state != D_PREP)
1027 wait_for_completion_io(&dc->wait);
1028 __punch_discard_cmd(sbi, dc, blkaddr);
275b66b0 1029 }
40465257 1030
d431413f
CY		 1031 mutex_unlock(&dcc->cmd_lock);
1032}
1033
1034/* This comes from f2fs_put_super */
1035void f2fs_wait_discard_bios(struct f2fs_sb_info *sbi)
1036{
bd5b0738 1037 __issue_discard_cmd(sbi, false);
63a94fa1 1038 __wait_discard_cmd(sbi, false);
275b66b0
CY		 1039}
1040
15469963
JK		 1041static int issue_discard_thread(void *data)
1042{
1043 struct f2fs_sb_info *sbi = data;
1044 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1045 wait_queue_head_t *q = &dcc->discard_wait_queue;
15469963
JK		 1046repeat:
1047 if (kthread_should_stop())
1048 return 0;
1049
bd5b0738 1050 __issue_discard_cmd(sbi, true);
63a94fa1 1051 __wait_discard_cmd(sbi, true);
15469963 1052
15469963
JK		 1053 congestion_wait(BLK_RW_SYNC, HZ/50);
1054
22d375dd 1055 wait_event_interruptible(*q, kthread_should_stop() ||
ba48a33e 1056 atomic_read(&dcc->discard_cmd_cnt));
15469963
JK		 1057 goto repeat;
1058}
1059
f46e8809 1060#ifdef CONFIG_BLK_DEV_ZONED
3c62be17
JK		 1061static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
1062 struct block_device *bdev, block_t blkstart, block_t blklen)
f46e8809 1063{
92592285 1064 sector_t sector, nr_sects;
10a875f8 1065 block_t lblkstart = blkstart;
3c62be17
JK		 1066 int devi = 0;
1067
1068 if (sbi->s_ndevs) {
1069 devi = f2fs_target_device_index(sbi, blkstart);
1070 blkstart -= FDEV(devi).start_blk;
1071 }
f46e8809
DLM		 1072
1073 /*
1074 * We need to know the type of the zone: for conventional zones,
1075 * use regular discard if the drive supports it. For sequential
1076 * zones, reset the zone write pointer.
1077 */
3c62be17 1078 switch (get_blkz_type(sbi, bdev, blkstart)) {
f46e8809
DLM		 1079
1080 case BLK_ZONE_TYPE_CONVENTIONAL:
1081 if (!blk_queue_discard(bdev_get_queue(bdev)))
1082 return 0;
c81abe34 1083 return __queue_discard_cmd(sbi, bdev, lblkstart, blklen);
f46e8809
DLM		 1084 case BLK_ZONE_TYPE_SEQWRITE_REQ:
1085 case BLK_ZONE_TYPE_SEQWRITE_PREF:
92592285
JK		 1086 sector = SECTOR_FROM_BLOCK(blkstart);
1087 nr_sects = SECTOR_FROM_BLOCK(blklen);
1088
1089 if (sector & (bdev_zone_sectors(bdev) - 1) ||
1090 nr_sects != bdev_zone_sectors(bdev)) {
1091 f2fs_msg(sbi->sb, KERN_INFO,
1092 "(%d) %s: Unaligned discard attempted (block %x + %x)",
1093 devi, sbi->s_ndevs ? FDEV(devi).path: "",
1094 blkstart, blklen);
1095 return -EIO;
1096 }
d50aaeec 1097 trace_f2fs_issue_reset_zone(bdev, blkstart);
f46e8809
DLM		 1098 return blkdev_reset_zones(bdev, sector,
1099 nr_sects, GFP_NOFS);
1100 default:
1101 /* Unknown zone type: broken device ? */
1102 return -EIO;
1103 }
1104}
1105#endif
1106
3c62be17
JK		 1107static int __issue_discard_async(struct f2fs_sb_info *sbi,
1108 struct block_device *bdev, block_t blkstart, block_t blklen)
1109{
1110#ifdef CONFIG_BLK_DEV_ZONED
1111 if (f2fs_sb_mounted_blkzoned(sbi->sb) &&
1112 bdev_zoned_model(bdev) != BLK_ZONED_NONE)
1113 return __f2fs_issue_discard_zone(sbi, bdev, blkstart, blklen);
1114#endif
c81abe34 1115 return __queue_discard_cmd(sbi, bdev, blkstart, blklen);
3c62be17
JK		 1116}
1117
1e87a78d 1118static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
37208879
JK		 1119 block_t blkstart, block_t blklen)
1120{
3c62be17
JK		 1121 sector_t start = blkstart, len = 0;
1122 struct block_device *bdev;
a66cdd98
JK		 1123 struct seg_entry *se;
1124 unsigned int offset;
1125 block_t i;
3c62be17
JK		 1126 int err = 0;
1127
1128 bdev = f2fs_target_device(sbi, blkstart, NULL);
1129
1130 for (i = blkstart; i < blkstart + blklen; i++, len++) {
1131 if (i != start) {
1132 struct block_device *bdev2 =
1133 f2fs_target_device(sbi, i, NULL);
1134
1135 if (bdev2 != bdev) {
1136 err = __issue_discard_async(sbi, bdev,
1137 start, len);
1138 if (err)
1139 return err;
1140 bdev = bdev2;
1141 start = i;
1142 len = 0;
1143 }
1144 }
a66cdd98 1145
a66cdd98
JK		 1146 se = get_seg_entry(sbi, GET_SEGNO(sbi, i));
1147 offset = GET_BLKOFF_FROM_SEG0(sbi, i);
1148
1149 if (!f2fs_test_and_set_bit(offset, se->discard_map))
1150 sbi->discard_blks--;
1151 }
f46e8809 1152
3c62be17
JK		 1153 if (len)
1154 err = __issue_discard_async(sbi, bdev, start, len);
1155 return err;
1e87a78d
JK		 1156}
1157
25290fa5
JK		 1158static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc,
1159 bool check_only)
adf4983b 1160{
b2955550
JK		 1161 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
1162 int max_blocks = sbi->blocks_per_seg;
4b2fecc8 1163 struct seg_entry *se = get_seg_entry(sbi, cpc->trim_start);
b2955550
JK		 1164 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1165 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
a66cdd98 1166 unsigned long *discard_map = (unsigned long *)se->discard_map;
60a3b782 1167 unsigned long *dmap = SIT_I(sbi)->tmp_map;
b2955550 1168 unsigned int start = 0, end = -1;
4b2fecc8 1169 bool force = (cpc->reason == CP_DISCARD);
a7eeb823 1170 struct discard_entry *de = NULL;
46f84c2c 1171 struct list_head *head = &SM_I(sbi)->dcc_info->entry_list;
b2955550
JK		 1172 int i;
1173
3e025740 1174 if (se->valid_blocks == max_blocks || !f2fs_discard_en(sbi))
25290fa5 1175 return false;
b2955550 1176
a66cdd98
JK		 1177 if (!force) {
1178 if (!test_opt(sbi, DISCARD) || !se->valid_blocks ||
0b54fb84
JK		 1179 SM_I(sbi)->dcc_info->nr_discards >=
1180 SM_I(sbi)->dcc_info->max_discards)
25290fa5 1181 return false;
4b2fecc8
JK		 1182 }
1183
b2955550
JK		 1184 /* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
1185 for (i = 0; i < entries; i++)
a66cdd98 1186 dmap[i] = force ? ~ckpt_map[i] & ~discard_map[i] :
d7bc2484 1187 (cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
b2955550 1188
0b54fb84
JK		 1189 while (force || SM_I(sbi)->dcc_info->nr_discards <=
1190 SM_I(sbi)->dcc_info->max_discards) {
b2955550
JK		 1191 start = __find_rev_next_bit(dmap, max_blocks, end + 1);
1192 if (start >= max_blocks)
1193 break;
1194
1195 end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1);
c7b41e16
YH		 1196 if (force && start && end != max_blocks
1197 && (end - start) < cpc->trim_minlen)
1198 continue;
1199
25290fa5
JK		 1200 if (check_only)
1201 return true;
1202
a7eeb823
CY		 1203 if (!de) {
1204 de = f2fs_kmem_cache_alloc(discard_entry_slab,
1205 GFP_F2FS_ZERO);
1206 de->start_blkaddr = START_BLOCK(sbi, cpc->trim_start);
1207 list_add_tail(&de->list, head);
1208 }
1209
1210 for (i = start; i < end; i++)
1211 __set_bit_le(i, (void *)de->discard_map);
1212
1213 SM_I(sbi)->dcc_info->nr_discards += end - start;
b2955550 1214 }
25290fa5 1215 return false;
b2955550
JK		 1216}
1217
4b2fecc8
JK		 1218void release_discard_addrs(struct f2fs_sb_info *sbi)
1219{
46f84c2c 1220 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
4b2fecc8
JK		 1221 struct discard_entry *entry, *this;
1222
1223 /* drop caches */
1224 list_for_each_entry_safe(entry, this, head, list) {
1225 list_del(&entry->list);
1226 kmem_cache_free(discard_entry_slab, entry);
1227 }
1228}
1229
0a8165d7 1230/*
351df4b2
JK		 1231 * Should call clear_prefree_segments after checkpoint is done.
1232 */
1233static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
1234{
1235 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
b65ee148 1236 unsigned int segno;
351df4b2
JK		 1237
1238 mutex_lock(&dirty_i->seglist_lock);
7cd8558b 1239 for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], MAIN_SEGS(sbi))
351df4b2 1240 __set_test_and_free(sbi, segno);
351df4b2
JK		 1241 mutex_unlock(&dirty_i->seglist_lock);
1242}
1243
836b5a63 1244void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2 1245{
46f84c2c 1246 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
2d7b822a 1247 struct discard_entry *entry, *this;
351df4b2 1248 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
29e59c14 1249 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
29e59c14 1250 unsigned int start = 0, end = -1;
36abef4e 1251 unsigned int secno, start_segno;
c24a0fd6 1252 bool force = (cpc->reason == CP_DISCARD);
351df4b2
JK		 1253
1254 mutex_lock(&dirty_i->seglist_lock);
29e59c14 1255
351df4b2 1256 while (1) {
29e59c14 1257 int i;
7cd8558b
JK		 1258 start = find_next_bit(prefree_map, MAIN_SEGS(sbi), end + 1);
1259 if (start >= MAIN_SEGS(sbi))
351df4b2 1260 break;
7cd8558b
JK		 1261 end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
1262 start + 1);
29e59c14
CL		 1263
1264 for (i = start; i < end; i++)
1265 clear_bit(i, prefree_map);
1266
1267 dirty_i->nr_dirty[PRE] -= end - start;
1268
650d3c4e 1269 if (!test_opt(sbi, DISCARD))
29e59c14 1270 continue;
351df4b2 1271
650d3c4e
YH		 1272 if (force && start >= cpc->trim_start &&
1273 (end - 1) <= cpc->trim_end)
1274 continue;
1275
36abef4e
JK		 1276 if (!test_opt(sbi, LFS) || sbi->segs_per_sec == 1) {
1277 f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
37208879 1278 (end - start) << sbi->log_blocks_per_seg);
36abef4e
JK		 1279 continue;
1280 }
1281next:
4ddb1a4d
JK		 1282 secno = GET_SEC_FROM_SEG(sbi, start);
1283 start_segno = GET_SEG_FROM_SEC(sbi, secno);
36abef4e 1284 if (!IS_CURSEC(sbi, secno) &&
302bd348 1285 !get_valid_blocks(sbi, start, true))
36abef4e
JK		 1286 f2fs_issue_discard(sbi, START_BLOCK(sbi, start_segno),
1287 sbi->segs_per_sec << sbi->log_blocks_per_seg);
1288
1289 start = start_segno + sbi->segs_per_sec;
1290 if (start < end)
1291 goto next;
8b107f5b
JK		 1292 else
1293 end = start - 1;
351df4b2
JK		 1294 }
1295 mutex_unlock(&dirty_i->seglist_lock);
b2955550
JK		 1296
1297 /* send small discards */
2d7b822a 1298 list_for_each_entry_safe(entry, this, head, list) {
a7eeb823
CY		 1299 unsigned int cur_pos = 0, next_pos, len, total_len = 0;
1300 bool is_valid = test_bit_le(0, entry->discard_map);
1301
1302find_next:
1303 if (is_valid) {
1304 next_pos = find_next_zero_bit_le(entry->discard_map,
1305 sbi->blocks_per_seg, cur_pos);
1306 len = next_pos - cur_pos;
1307
1308 if (force && len < cpc->trim_minlen)
1309 goto skip;
1310
1311 f2fs_issue_discard(sbi, entry->start_blkaddr + cur_pos,
1312 len);
1313 cpc->trimmed += len;
1314 total_len += len;
1315 } else {
1316 next_pos = find_next_bit_le(entry->discard_map,
1317 sbi->blocks_per_seg, cur_pos);
1318 }
836b5a63 1319skip:
a7eeb823
CY		 1320 cur_pos = next_pos;
1321 is_valid = !is_valid;
1322
1323 if (cur_pos < sbi->blocks_per_seg)
1324 goto find_next;
1325
b2955550 1326 list_del(&entry->list);
a7eeb823 1327 SM_I(sbi)->dcc_info->nr_discards -= total_len;
b2955550
JK		 1328 kmem_cache_free(discard_entry_slab, entry);
1329 }
34e159da
CY		 1330
1331 wake_up(&SM_I(sbi)->dcc_info->discard_wait_queue);
351df4b2
JK		 1332}
1333
8ed59745 1334static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84 1335{
15469963 1336 dev_t dev = sbi->sb->s_bdev->bd_dev;
0b54fb84 1337 struct discard_cmd_control *dcc;
ba48a33e 1338 int err = 0, i;
0b54fb84
JK		 1339
1340 if (SM_I(sbi)->dcc_info) {
1341 dcc = SM_I(sbi)->dcc_info;
1342 goto init_thread;
1343 }
1344
1345 dcc = kzalloc(sizeof(struct discard_cmd_control), GFP_KERNEL);
1346 if (!dcc)
1347 return -ENOMEM;
1348
46f84c2c 1349 INIT_LIST_HEAD(&dcc->entry_list);
ba48a33e
CY		 1350 for (i = 0; i < MAX_PLIST_NUM; i++)
1351 INIT_LIST_HEAD(&dcc->pend_list[i]);
46f84c2c 1352 INIT_LIST_HEAD(&dcc->wait_list);
15469963 1353 mutex_init(&dcc->cmd_lock);
8b8dd65f
CY		 1354 atomic_set(&dcc->issued_discard, 0);
1355 atomic_set(&dcc->issing_discard, 0);
5f32366a 1356 atomic_set(&dcc->discard_cmd_cnt, 0);
0b54fb84 1357 dcc->nr_discards = 0;
d618ebaf 1358 dcc->max_discards = MAIN_SEGS(sbi) << sbi->log_blocks_per_seg;
d84d1cbd 1359 dcc->undiscard_blks = 0;
004b6862 1360 dcc->root = RB_ROOT;
0b54fb84 1361
15469963 1362 init_waitqueue_head(&dcc->discard_wait_queue);
0b54fb84
JK		 1363 SM_I(sbi)->dcc_info = dcc;
1364init_thread:
15469963
JK		 1365 dcc->f2fs_issue_discard = kthread_run(issue_discard_thread, sbi,
1366 "f2fs_discard-%u:%u", MAJOR(dev), MINOR(dev));
1367 if (IS_ERR(dcc->f2fs_issue_discard)) {
1368 err = PTR_ERR(dcc->f2fs_issue_discard);
1369 kfree(dcc);
1370 SM_I(sbi)->dcc_info = NULL;
1371 return err;
1372 }
1373
0b54fb84
JK		 1374 return err;
1375}
1376
f099405f 1377static void destroy_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84
JK		 1378{
1379 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1380
f099405f
CY		 1381 if (!dcc)
1382 return;
1383
1384 if (dcc->f2fs_issue_discard) {
15469963
JK		 1385 struct task_struct *discard_thread = dcc->f2fs_issue_discard;
1386
1387 dcc->f2fs_issue_discard = NULL;
1388 kthread_stop(discard_thread);
1389 }
f099405f
CY		 1390
1391 kfree(dcc);
1392 SM_I(sbi)->dcc_info = NULL;
0b54fb84
JK		 1393}
1394
184a5cd2 1395static bool __mark_sit_entry_dirty(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 1396{
1397 struct sit_info *sit_i = SIT_I(sbi);
184a5cd2
CY		 1398
1399 if (!__test_and_set_bit(segno, sit_i->dirty_sentries_bitmap)) {
351df4b2 1400 sit_i->dirty_sentries++;
184a5cd2
CY		 1401 return false;
1402 }
1403
1404 return true;
351df4b2
JK		 1405}
1406
1407static void __set_sit_entry_type(struct f2fs_sb_info *sbi, int type,
1408 unsigned int segno, int modified)
1409{
1410 struct seg_entry *se = get_seg_entry(sbi, segno);
1411 se->type = type;
1412 if (modified)
1413 __mark_sit_entry_dirty(sbi, segno);
1414}
1415
1416static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
1417{
1418 struct seg_entry *se;
1419 unsigned int segno, offset;
1420 long int new_vblocks;
1421
1422 segno = GET_SEGNO(sbi, blkaddr);
1423
1424 se = get_seg_entry(sbi, segno);
1425 new_vblocks = se->valid_blocks + del;
491c0854 1426 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
351df4b2 1427
9850cf4a 1428 f2fs_bug_on(sbi, (new_vblocks >> (sizeof(unsigned short) << 3) ||
351df4b2
JK		 1429 (new_vblocks > sbi->blocks_per_seg)));
1430
1431 se->valid_blocks = new_vblocks;
1432 se->mtime = get_mtime(sbi);
1433 SIT_I(sbi)->max_mtime = se->mtime;
1434
1435 /* Update valid block bitmap */
1436 if (del > 0) {
355e7891
CY		 1437 if (f2fs_test_and_set_bit(offset, se->cur_valid_map)) {
1438#ifdef CONFIG_F2FS_CHECK_FS
1439 if (f2fs_test_and_set_bit(offset,
1440 se->cur_valid_map_mir))
1441 f2fs_bug_on(sbi, 1);
1442 else
1443 WARN_ON(1);
1444#else
05796763 1445 f2fs_bug_on(sbi, 1);
355e7891
CY		 1446#endif
1447 }
3e025740
JK		 1448 if (f2fs_discard_en(sbi) &&
1449 !f2fs_test_and_set_bit(offset, se->discard_map))
a66cdd98 1450 sbi->discard_blks--;
720037f9
JK		 1451
1452 /* don't overwrite by SSR to keep node chain */
1453 if (se->type == CURSEG_WARM_NODE) {
1454 if (!f2fs_test_and_set_bit(offset, se->ckpt_valid_map))
1455 se->ckpt_valid_blocks++;
1456 }
351df4b2 1457 } else {
355e7891
CY		 1458 if (!f2fs_test_and_clear_bit(offset, se->cur_valid_map)) {
1459#ifdef CONFIG_F2FS_CHECK_FS
1460 if (!f2fs_test_and_clear_bit(offset,
1461 se->cur_valid_map_mir))
1462 f2fs_bug_on(sbi, 1);
1463 else
1464 WARN_ON(1);
1465#else
05796763 1466 f2fs_bug_on(sbi, 1);
355e7891
CY		 1467#endif
1468 }
3e025740
JK		 1469 if (f2fs_discard_en(sbi) &&
1470 f2fs_test_and_clear_bit(offset, se->discard_map))
a66cdd98 1471 sbi->discard_blks++;
351df4b2
JK		 1472 }
1473 if (!f2fs_test_bit(offset, se->ckpt_valid_map))
1474 se->ckpt_valid_blocks += del;
1475
1476 __mark_sit_entry_dirty(sbi, segno);
1477
1478 /* update total number of valid blocks to be written in ckpt area */
1479 SIT_I(sbi)->written_valid_blocks += del;
1480
1481 if (sbi->segs_per_sec > 1)
1482 get_sec_entry(sbi, segno)->valid_blocks += del;
1483}
1484
5e443818 1485void refresh_sit_entry(struct f2fs_sb_info *sbi, block_t old, block_t new)
351df4b2 1486{
5e443818
JK		 1487 update_sit_entry(sbi, new, 1);
1488 if (GET_SEGNO(sbi, old) != NULL_SEGNO)
1489 update_sit_entry(sbi, old, -1);
1490
1491 locate_dirty_segment(sbi, GET_SEGNO(sbi, old));
1492 locate_dirty_segment(sbi, GET_SEGNO(sbi, new));
351df4b2
JK		 1493}
1494
1495void invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
1496{
1497 unsigned int segno = GET_SEGNO(sbi, addr);
1498 struct sit_info *sit_i = SIT_I(sbi);
1499
9850cf4a 1500 f2fs_bug_on(sbi, addr == NULL_ADDR);
351df4b2
JK		 1501 if (addr == NEW_ADDR)
1502 return;
1503
1504 /* add it into sit main buffer */
1505 mutex_lock(&sit_i->sentry_lock);
1506
1507 update_sit_entry(sbi, addr, -1);
1508
1509 /* add it into dirty seglist */
1510 locate_dirty_segment(sbi, segno);
1511
1512 mutex_unlock(&sit_i->sentry_lock);
1513}
1514
6e2c64ad
JK		 1515bool is_checkpointed_data(struct f2fs_sb_info *sbi, block_t blkaddr)
1516{
1517 struct sit_info *sit_i = SIT_I(sbi);
1518 unsigned int segno, offset;
1519 struct seg_entry *se;
1520 bool is_cp = false;
1521
1522 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
1523 return true;
1524
1525 mutex_lock(&sit_i->sentry_lock);
1526
1527 segno = GET_SEGNO(sbi, blkaddr);
1528 se = get_seg_entry(sbi, segno);
1529 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
1530
1531 if (f2fs_test_bit(offset, se->ckpt_valid_map))
1532 is_cp = true;
1533
1534 mutex_unlock(&sit_i->sentry_lock);
1535
1536 return is_cp;
1537}
1538
0a8165d7 1539/*
351df4b2
JK		 1540 * This function should be resided under the curseg_mutex lock
1541 */
1542static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
e79efe3b 1543 struct f2fs_summary *sum)
351df4b2
JK		 1544{
1545 struct curseg_info *curseg = CURSEG_I(sbi, type);
1546 void *addr = curseg->sum_blk;
e79efe3b 1547 addr += curseg->next_blkoff * sizeof(struct f2fs_summary);
351df4b2 1548 memcpy(addr, sum, sizeof(struct f2fs_summary));
351df4b2
JK		 1549}
1550
0a8165d7 1551/*
351df4b2
JK		 1552 * Calculate the number of current summary pages for writing
1553 */
3fa06d7b 1554int npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
351df4b2 1555{
351df4b2 1556 int valid_sum_count = 0;
9a47938b 1557 int i, sum_in_page;
351df4b2
JK		 1558
1559 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1560 if (sbi->ckpt->alloc_type[i] == SSR)
1561 valid_sum_count += sbi->blocks_per_seg;
3fa06d7b
CY		 1562 else {
1563 if (for_ra)
1564 valid_sum_count += le16_to_cpu(
1565 F2FS_CKPT(sbi)->cur_data_blkoff[i]);
1566 else
1567 valid_sum_count += curseg_blkoff(sbi, i);
1568 }
351df4b2
JK		 1569 }
1570
09cbfeaf 1571 sum_in_page = (PAGE_SIZE - 2 * SUM_JOURNAL_SIZE -
9a47938b
FL		 1572 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
1573 if (valid_sum_count <= sum_in_page)
351df4b2 1574 return 1;
9a47938b 1575 else if ((valid_sum_count - sum_in_page) <=
09cbfeaf 1576 (PAGE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
351df4b2
JK		 1577 return 2;
1578 return 3;
1579}
1580
0a8165d7 1581/*
351df4b2
JK		 1582 * Caller should put this summary page
1583 */
1584struct page *get_sum_page(struct f2fs_sb_info *sbi, unsigned int segno)
1585{
1586 return get_meta_page(sbi, GET_SUM_BLOCK(sbi, segno));
1587}
1588
381722d2 1589void update_meta_page(struct f2fs_sb_info *sbi, void *src, block_t blk_addr)
351df4b2
JK		 1590{
1591 struct page *page = grab_meta_page(sbi, blk_addr);
381722d2
CY		 1592 void *dst = page_address(page);
1593
1594 if (src)
09cbfeaf 1595 memcpy(dst, src, PAGE_SIZE);
381722d2 1596 else
09cbfeaf 1597 memset(dst, 0, PAGE_SIZE);
351df4b2
JK		 1598 set_page_dirty(page);
1599 f2fs_put_page(page, 1);
1600}
1601
381722d2
CY		 1602static void write_sum_page(struct f2fs_sb_info *sbi,
1603 struct f2fs_summary_block *sum_blk, block_t blk_addr)
1604{
1605 update_meta_page(sbi, (void *)sum_blk, blk_addr);
1606}
1607
b7ad7512
CY		 1608static void write_current_sum_page(struct f2fs_sb_info *sbi,
1609 int type, block_t blk_addr)
1610{
1611 struct curseg_info *curseg = CURSEG_I(sbi, type);
1612 struct page *page = grab_meta_page(sbi, blk_addr);
1613 struct f2fs_summary_block *src = curseg->sum_blk;
1614 struct f2fs_summary_block *dst;
1615
1616 dst = (struct f2fs_summary_block *)page_address(page);
1617
1618 mutex_lock(&curseg->curseg_mutex);
1619
1620 down_read(&curseg->journal_rwsem);
1621 memcpy(&dst->journal, curseg->journal, SUM_JOURNAL_SIZE);
1622 up_read(&curseg->journal_rwsem);
1623
1624 memcpy(dst->entries, src->entries, SUM_ENTRY_SIZE);
1625 memcpy(&dst->footer, &src->footer, SUM_FOOTER_SIZE);
1626
1627 mutex_unlock(&curseg->curseg_mutex);
1628
1629 set_page_dirty(page);
1630 f2fs_put_page(page, 1);
1631}
1632
a7881893
JK		 1633static int is_next_segment_free(struct f2fs_sb_info *sbi, int type)
1634{
1635 struct curseg_info *curseg = CURSEG_I(sbi, type);
1636 unsigned int segno = curseg->segno + 1;
1637 struct free_segmap_info *free_i = FREE_I(sbi);
1638
1639 if (segno < MAIN_SEGS(sbi) && segno % sbi->segs_per_sec)
1640 return !test_bit(segno, free_i->free_segmap);
1641 return 0;
1642}
1643
0a8165d7 1644/*
351df4b2
JK		 1645 * Find a new segment from the free segments bitmap to right order
1646 * This function should be returned with success, otherwise BUG
1647 */
1648static void get_new_segment(struct f2fs_sb_info *sbi,
1649 unsigned int *newseg, bool new_sec, int dir)
1650{
1651 struct free_segmap_info *free_i = FREE_I(sbi);
351df4b2 1652 unsigned int segno, secno, zoneno;
7cd8558b 1653 unsigned int total_zones = MAIN_SECS(sbi) / sbi->secs_per_zone;
4ddb1a4d
JK		 1654 unsigned int hint = GET_SEC_FROM_SEG(sbi, *newseg);
1655 unsigned int old_zoneno = GET_ZONE_FROM_SEG(sbi, *newseg);
351df4b2
JK		 1656 unsigned int left_start = hint;
1657 bool init = true;
1658 int go_left = 0;
1659 int i;
1660
1a118ccf 1661 spin_lock(&free_i->segmap_lock);
351df4b2
JK		 1662
1663 if (!new_sec && ((*newseg + 1) % sbi->segs_per_sec)) {
1664 segno = find_next_zero_bit(free_i->free_segmap,
4ddb1a4d
JK		 1665 GET_SEG_FROM_SEC(sbi, hint + 1), *newseg + 1);
1666 if (segno < GET_SEG_FROM_SEC(sbi, hint + 1))
351df4b2
JK		 1667 goto got_it;
1668 }
1669find_other_zone:
7cd8558b
JK		 1670 secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
1671 if (secno >= MAIN_SECS(sbi)) {
351df4b2
JK		 1672 if (dir == ALLOC_RIGHT) {
1673 secno = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1674 MAIN_SECS(sbi), 0);
1675 f2fs_bug_on(sbi, secno >= MAIN_SECS(sbi));
351df4b2
JK		 1676 } else {
1677 go_left = 1;
1678 left_start = hint - 1;
1679 }
1680 }
1681 if (go_left == 0)
1682 goto skip_left;
1683
1684 while (test_bit(left_start, free_i->free_secmap)) {
1685 if (left_start > 0) {
1686 left_start--;
1687 continue;
1688 }
1689 left_start = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1690 MAIN_SECS(sbi), 0);
1691 f2fs_bug_on(sbi, left_start >= MAIN_SECS(sbi));
351df4b2
JK		 1692 break;
1693 }
1694 secno = left_start;
1695skip_left:
1696 hint = secno;
4ddb1a4d
JK		 1697 segno = GET_SEG_FROM_SEC(sbi, secno);
1698 zoneno = GET_ZONE_FROM_SEC(sbi, secno);
351df4b2
JK		 1699
1700 /* give up on finding another zone */
1701 if (!init)
1702 goto got_it;
1703 if (sbi->secs_per_zone == 1)
1704 goto got_it;
1705 if (zoneno == old_zoneno)
1706 goto got_it;
1707 if (dir == ALLOC_LEFT) {
1708 if (!go_left && zoneno + 1 >= total_zones)
1709 goto got_it;
1710 if (go_left && zoneno == 0)
1711 goto got_it;
1712 }
1713 for (i = 0; i < NR_CURSEG_TYPE; i++)
1714 if (CURSEG_I(sbi, i)->zone == zoneno)
1715 break;
1716
1717 if (i < NR_CURSEG_TYPE) {
1718 /* zone is in user, try another */
1719 if (go_left)
1720 hint = zoneno * sbi->secs_per_zone - 1;
1721 else if (zoneno + 1 >= total_zones)
1722 hint = 0;
1723 else
1724 hint = (zoneno + 1) * sbi->secs_per_zone;
1725 init = false;
1726 goto find_other_zone;
1727 }
1728got_it:
1729 /* set it as dirty segment in free segmap */
9850cf4a 1730 f2fs_bug_on(sbi, test_bit(segno, free_i->free_segmap));
351df4b2
JK		 1731 __set_inuse(sbi, segno);
1732 *newseg = segno;
1a118ccf 1733 spin_unlock(&free_i->segmap_lock);
351df4b2
JK		 1734}
1735
1736static void reset_curseg(struct f2fs_sb_info *sbi, int type, int modified)
1737{
1738 struct curseg_info *curseg = CURSEG_I(sbi, type);
1739 struct summary_footer *sum_footer;
1740
1741 curseg->segno = curseg->next_segno;
4ddb1a4d 1742 curseg->zone = GET_ZONE_FROM_SEG(sbi, curseg->segno);
351df4b2
JK		 1743 curseg->next_blkoff = 0;
1744 curseg->next_segno = NULL_SEGNO;
1745
1746 sum_footer = &(curseg->sum_blk->footer);
1747 memset(sum_footer, 0, sizeof(struct summary_footer));
1748 if (IS_DATASEG(type))
1749 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
1750 if (IS_NODESEG(type))
1751 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
1752 __set_sit_entry_type(sbi, type, curseg->segno, modified);
1753}
1754
7a20b8a6
JK		 1755static unsigned int __get_next_segno(struct f2fs_sb_info *sbi, int type)
1756{
a7881893
JK		 1757 /* if segs_per_sec is large than 1, we need to keep original policy. */
1758 if (sbi->segs_per_sec != 1)
1759 return CURSEG_I(sbi, type)->segno;
1760
7a20b8a6
JK		 1761 if (type == CURSEG_HOT_DATA || IS_NODESEG(type))
1762 return 0;
1763
e066b83c
JK		 1764 if (SIT_I(sbi)->last_victim[ALLOC_NEXT])
1765 return SIT_I(sbi)->last_victim[ALLOC_NEXT];
7a20b8a6
JK		 1766 return CURSEG_I(sbi, type)->segno;
1767}
1768
0a8165d7 1769/*
351df4b2
JK		 1770 * Allocate a current working segment.
1771 * This function always allocates a free segment in LFS manner.
1772 */
1773static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
1774{
1775 struct curseg_info *curseg = CURSEG_I(sbi, type);
1776 unsigned int segno = curseg->segno;
1777 int dir = ALLOC_LEFT;
1778
1779 write_sum_page(sbi, curseg->sum_blk,
81fb5e87 1780 GET_SUM_BLOCK(sbi, segno));
351df4b2
JK		 1781 if (type == CURSEG_WARM_DATA || type == CURSEG_COLD_DATA)
1782 dir = ALLOC_RIGHT;
1783
1784 if (test_opt(sbi, NOHEAP))
1785 dir = ALLOC_RIGHT;
1786
7a20b8a6 1787 segno = __get_next_segno(sbi, type);
351df4b2
JK		 1788 get_new_segment(sbi, &segno, new_sec, dir);
1789 curseg->next_segno = segno;
1790 reset_curseg(sbi, type, 1);
1791 curseg->alloc_type = LFS;
1792}
1793
1794static void __next_free_blkoff(struct f2fs_sb_info *sbi,
1795 struct curseg_info *seg, block_t start)
1796{
1797 struct seg_entry *se = get_seg_entry(sbi, seg->segno);
e81c93cf 1798 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
60a3b782 1799 unsigned long *target_map = SIT_I(sbi)->tmp_map;
e81c93cf
CL		 1800 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
1801 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1802 int i, pos;
1803
1804 for (i = 0; i < entries; i++)
1805 target_map[i] = ckpt_map[i] | cur_map[i];
1806
1807 pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, start);
1808
1809 seg->next_blkoff = pos;
351df4b2
JK		 1810}
1811
0a8165d7 1812/*
351df4b2
JK		 1813 * If a segment is written by LFS manner, next block offset is just obtained
1814 * by increasing the current block offset. However, if a segment is written by
1815 * SSR manner, next block offset obtained by calling __next_free_blkoff
1816 */
1817static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
1818 struct curseg_info *seg)
1819{
1820 if (seg->alloc_type == SSR)
1821 __next_free_blkoff(sbi, seg, seg->next_blkoff + 1);
1822 else
1823 seg->next_blkoff++;
1824}
1825
0a8165d7 1826/*
e1c42045 1827 * This function always allocates a used segment(from dirty seglist) by SSR
351df4b2
JK		 1828 * manner, so it should recover the existing segment information of valid blocks
1829 */
1830static void change_curseg(struct f2fs_sb_info *sbi, int type, bool reuse)
1831{
1832 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
1833 struct curseg_info *curseg = CURSEG_I(sbi, type);
1834 unsigned int new_segno = curseg->next_segno;
1835 struct f2fs_summary_block *sum_node;
1836 struct page *sum_page;
1837
1838 write_sum_page(sbi, curseg->sum_blk,
1839 GET_SUM_BLOCK(sbi, curseg->segno));
1840 __set_test_and_inuse(sbi, new_segno);
1841
1842 mutex_lock(&dirty_i->seglist_lock);
1843 __remove_dirty_segment(sbi, new_segno, PRE);
1844 __remove_dirty_segment(sbi, new_segno, DIRTY);
1845 mutex_unlock(&dirty_i->seglist_lock);
1846
1847 reset_curseg(sbi, type, 1);
1848 curseg->alloc_type = SSR;
1849 __next_free_blkoff(sbi, curseg, 0);
1850
1851 if (reuse) {
1852 sum_page = get_sum_page(sbi, new_segno);
1853 sum_node = (struct f2fs_summary_block *)page_address(sum_page);
1854 memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
1855 f2fs_put_page(sum_page, 1);
1856 }
1857}
1858
43727527
JK		 1859static int get_ssr_segment(struct f2fs_sb_info *sbi, int type)
1860{
1861 struct curseg_info *curseg = CURSEG_I(sbi, type);
1862 const struct victim_selection *v_ops = DIRTY_I(sbi)->v_ops;
e066b83c 1863 unsigned segno = NULL_SEGNO;
d27c3d89
CY		 1864 int i, cnt;
1865 bool reversed = false;
c192f7a4
JK		 1866
1867 /* need_SSR() already forces to do this */
e066b83c
JK		 1868 if (v_ops->get_victim(sbi, &segno, BG_GC, type, SSR)) {
1869 curseg->next_segno = segno;
c192f7a4 1870 return 1;
e066b83c 1871 }
43727527 1872
70d625cb
JK		 1873 /* For node segments, let's do SSR more intensively */
1874 if (IS_NODESEG(type)) {
d27c3d89
CY		 1875 if (type >= CURSEG_WARM_NODE) {
1876 reversed = true;
1877 i = CURSEG_COLD_NODE;
1878 } else {
1879 i = CURSEG_HOT_NODE;
1880 }
1881 cnt = NR_CURSEG_NODE_TYPE;
70d625cb 1882 } else {
d27c3d89
CY		 1883 if (type >= CURSEG_WARM_DATA) {
1884 reversed = true;
1885 i = CURSEG_COLD_DATA;
1886 } else {
1887 i = CURSEG_HOT_DATA;
1888 }
1889 cnt = NR_CURSEG_DATA_TYPE;
70d625cb 1890 }
43727527 1891
d27c3d89 1892 for (; cnt-- > 0; reversed ? i-- : i++) {
c192f7a4
JK		 1893 if (i == type)
1894 continue;
e066b83c
JK		 1895 if (v_ops->get_victim(sbi, &segno, BG_GC, i, SSR)) {
1896 curseg->next_segno = segno;
43727527 1897 return 1;
e066b83c 1898 }
c192f7a4 1899 }
43727527
JK		 1900 return 0;
1901}
1902
351df4b2
JK		 1903/*
1904 * flush out current segment and replace it with new segment
1905 * This function should be returned with success, otherwise BUG
1906 */
1907static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
1908 int type, bool force)
1909{
a7881893
JK		 1910 struct curseg_info *curseg = CURSEG_I(sbi, type);
1911
7b405275 1912 if (force)
351df4b2 1913 new_curseg(sbi, type, true);
5b6c6be2
JK		 1914 else if (!is_set_ckpt_flags(sbi, CP_CRC_RECOVERY_FLAG) &&
1915 type == CURSEG_WARM_NODE)
351df4b2 1916 new_curseg(sbi, type, false);
a7881893
JK		 1917 else if (curseg->alloc_type == LFS && is_next_segment_free(sbi, type))
1918 new_curseg(sbi, type, false);
351df4b2
JK		 1919 else if (need_SSR(sbi) && get_ssr_segment(sbi, type))
1920 change_curseg(sbi, type, true);
1921 else
1922 new_curseg(sbi, type, false);
dcdfff65 1923
a7881893 1924 stat_inc_seg_type(sbi, curseg);
351df4b2
JK		 1925}
1926
1927void allocate_new_segments(struct f2fs_sb_info *sbi)
1928{
6ae1be13
JK		 1929 struct curseg_info *curseg;
1930 unsigned int old_segno;
351df4b2
JK		 1931 int i;
1932
6ae1be13
JK		 1933 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1934 curseg = CURSEG_I(sbi, i);
1935 old_segno = curseg->segno;
1936 SIT_I(sbi)->s_ops->allocate_segment(sbi, i, true);
1937 locate_dirty_segment(sbi, old_segno);
1938 }
351df4b2
JK		 1939}
1940
1941static const struct segment_allocation default_salloc_ops = {
1942 .allocate_segment = allocate_segment_by_default,
1943};
1944
25290fa5
JK		 1945bool exist_trim_candidates(struct f2fs_sb_info *sbi, struct cp_control *cpc)
1946{
1947 __u64 trim_start = cpc->trim_start;
1948 bool has_candidate = false;
1949
1950 mutex_lock(&SIT_I(sbi)->sentry_lock);
1951 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++) {
1952 if (add_discard_addrs(sbi, cpc, true)) {
1953 has_candidate = true;
1954 break;
1955 }
1956 }
1957 mutex_unlock(&SIT_I(sbi)->sentry_lock);
1958
1959 cpc->trim_start = trim_start;
1960 return has_candidate;
1961}
1962
4b2fecc8
JK		 1963int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
1964{
f7ef9b83
JK		 1965 __u64 start = F2FS_BYTES_TO_BLK(range->start);
1966 __u64 end = start + F2FS_BYTES_TO_BLK(range->len) - 1;
4b2fecc8
JK		 1967 unsigned int start_segno, end_segno;
1968 struct cp_control cpc;
c34f42e2 1969 int err = 0;
4b2fecc8 1970
836b5a63 1971 if (start >= MAX_BLKADDR(sbi) || range->len < sbi->blocksize)
4b2fecc8
JK		 1972 return -EINVAL;
1973
9bd27ae4 1974 cpc.trimmed = 0;
7cd8558b 1975 if (end <= MAIN_BLKADDR(sbi))
4b2fecc8
JK		 1976 goto out;
1977
ed214a11
YH		 1978 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
1979 f2fs_msg(sbi->sb, KERN_WARNING,
1980 "Found FS corruption, run fsck to fix.");
1981 goto out;
1982 }
1983
4b2fecc8 1984 /* start/end segment number in main_area */
7cd8558b
JK		 1985 start_segno = (start <= MAIN_BLKADDR(sbi)) ? 0 : GET_SEGNO(sbi, start);
1986 end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
1987 GET_SEGNO(sbi, end);
4b2fecc8 1988 cpc.reason = CP_DISCARD;
836b5a63 1989 cpc.trim_minlen = max_t(__u64, 1, F2FS_BYTES_TO_BLK(range->minlen));
4b2fecc8
JK		 1990
1991 /* do checkpoint to issue discard commands safely */
bba681cb
JK		 1992 for (; start_segno <= end_segno; start_segno = cpc.trim_end + 1) {
1993 cpc.trim_start = start_segno;
a66cdd98
JK		 1994
1995 if (sbi->discard_blks == 0)
1996 break;
1997 else if (sbi->discard_blks < BATCHED_TRIM_BLOCKS(sbi))
1998 cpc.trim_end = end_segno;
1999 else
2000 cpc.trim_end = min_t(unsigned int,
2001 rounddown(start_segno +
bba681cb
JK		 2002 BATCHED_TRIM_SEGMENTS(sbi),
2003 sbi->segs_per_sec) - 1, end_segno);
2004
2005 mutex_lock(&sbi->gc_mutex);
c34f42e2 2006 err = write_checkpoint(sbi, &cpc);
bba681cb 2007 mutex_unlock(&sbi->gc_mutex);
e9328353
CY		 2008 if (err)
2009 break;
74fa5f3d
CY		 2010
2011 schedule();
bba681cb 2012 }
4b2fecc8 2013out:
f7ef9b83 2014 range->len = F2FS_BLK_TO_BYTES(cpc.trimmed);
c34f42e2 2015 return err;
4b2fecc8
JK		 2016}
2017
351df4b2
JK		 2018static bool __has_curseg_space(struct f2fs_sb_info *sbi, int type)
2019{
2020 struct curseg_info *curseg = CURSEG_I(sbi, type);
2021 if (curseg->next_blkoff < sbi->blocks_per_seg)
2022 return true;
2023 return false;
2024}
2025
2026static int __get_segment_type_2(struct page *page, enum page_type p_type)
2027{
2028 if (p_type == DATA)
2029 return CURSEG_HOT_DATA;
2030 else
2031 return CURSEG_HOT_NODE;
2032}
2033
2034static int __get_segment_type_4(struct page *page, enum page_type p_type)
2035{
2036 if (p_type == DATA) {
2037 struct inode *inode = page->mapping->host;
2038
2039 if (S_ISDIR(inode->i_mode))
2040 return CURSEG_HOT_DATA;
2041 else
2042 return CURSEG_COLD_DATA;
2043 } else {
a344b9fd
JK		 2044 if (IS_DNODE(page) && is_cold_node(page))
2045 return CURSEG_WARM_NODE;
351df4b2
JK		 2046 else
2047 return CURSEG_COLD_NODE;
2048 }
2049}
2050
2051static int __get_segment_type_6(struct page *page, enum page_type p_type)
2052{
2053 if (p_type == DATA) {
2054 struct inode *inode = page->mapping->host;
2055
ef095d19 2056 if (is_cold_data(page) || file_is_cold(inode))
351df4b2 2057 return CURSEG_COLD_DATA;
ef095d19
JK		 2058 if (is_inode_flag_set(inode, FI_HOT_DATA))
2059 return CURSEG_HOT_DATA;
2060 return CURSEG_WARM_DATA;
351df4b2
JK		 2061 } else {
2062 if (IS_DNODE(page))
2063 return is_cold_node(page) ? CURSEG_WARM_NODE :
2064 CURSEG_HOT_NODE;
ef095d19 2065 return CURSEG_COLD_NODE;
351df4b2
JK		 2066 }
2067}
2068
2069static int __get_segment_type(struct page *page, enum page_type p_type)
2070{
4081363f 2071 switch (F2FS_P_SB(page)->active_logs) {
351df4b2
JK		 2072 case 2:
2073 return __get_segment_type_2(page, p_type);
2074 case 4:
2075 return __get_segment_type_4(page, p_type);
351df4b2 2076 }
12a67146 2077 /* NR_CURSEG_TYPE(6) logs by default */
9850cf4a
JK		 2078 f2fs_bug_on(F2FS_P_SB(page),
2079 F2FS_P_SB(page)->active_logs != NR_CURSEG_TYPE);
12a67146 2080 return __get_segment_type_6(page, p_type);
351df4b2
JK		 2081}
2082
bfad7c2d
JK		 2083void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
2084 block_t old_blkaddr, block_t *new_blkaddr,
2085 struct f2fs_summary *sum, int type)
351df4b2
JK		 2086{
2087 struct sit_info *sit_i = SIT_I(sbi);
6ae1be13 2088 struct curseg_info *curseg = CURSEG_I(sbi, type);
351df4b2
JK		 2089
2090 mutex_lock(&curseg->curseg_mutex);
21cb1d99 2091 mutex_lock(&sit_i->sentry_lock);
351df4b2
JK		 2092
2093 *new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
351df4b2 2094
4e6a8d9b
JK		 2095 f2fs_wait_discard_bio(sbi, *new_blkaddr);
2096
351df4b2
JK		 2097 /*
2098 * __add_sum_entry should be resided under the curseg_mutex
2099 * because, this function updates a summary entry in the
2100 * current summary block.
2101 */
e79efe3b 2102 __add_sum_entry(sbi, type, sum);
351df4b2 2103
351df4b2 2104 __refresh_next_blkoff(sbi, curseg);
dcdfff65
JK		 2105
2106 stat_inc_block_count(sbi, curseg);
351df4b2 2107
c6f82fe9
JK		 2108 if (!__has_curseg_space(sbi, type))
2109 sit_i->s_ops->allocate_segment(sbi, type, false);
351df4b2 2110 /*
c6f82fe9
JK		 2111 * SIT information should be updated after segment allocation,
2112 * since we need to keep dirty segments precisely under SSR.
351df4b2
JK		 2113 */
2114 refresh_sit_entry(sbi, old_blkaddr, *new_blkaddr);
5e443818 2115
351df4b2
JK		 2116 mutex_unlock(&sit_i->sentry_lock);
2117
bfad7c2d 2118 if (page && IS_NODESEG(type))
351df4b2
JK		 2119 fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
2120
bfad7c2d
JK		 2121 mutex_unlock(&curseg->curseg_mutex);
2122}
2123
05ca3632 2124static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
bfad7c2d 2125{
05ca3632 2126 int type = __get_segment_type(fio->page, fio->type);
0a595eba 2127 int err;
bfad7c2d 2128
7dfeaa32
JK		 2129 if (fio->type == NODE || fio->type == DATA)
2130 mutex_lock(&fio->sbi->wio_mutex[fio->type]);
0a595eba 2131reallocate:
7a9d7548
CY		 2132 allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
2133 &fio->new_blkaddr, sum, type);
bfad7c2d 2134
351df4b2 2135 /* writeout dirty page into bdev */
0a595eba
JK		 2136 err = f2fs_submit_page_mbio(fio);
2137 if (err == -EAGAIN) {
2138 fio->old_blkaddr = fio->new_blkaddr;
2139 goto reallocate;
2140 }
7dfeaa32
JK		 2141
2142 if (fio->type == NODE || fio->type == DATA)
2143 mutex_unlock(&fio->sbi->wio_mutex[fio->type]);
351df4b2
JK		 2144}
2145
577e3495 2146void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
351df4b2 2147{
458e6197 2148 struct f2fs_io_info fio = {
05ca3632 2149 .sbi = sbi,
458e6197 2150 .type = META,
04d328de 2151 .op = REQ_OP_WRITE,
70fd7614 2152 .op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
7a9d7548
CY		 2153 .old_blkaddr = page->index,
2154 .new_blkaddr = page->index,
05ca3632 2155 .page = page,
4375a336 2156 .encrypted_page = NULL,
458e6197
JK		 2157 };
2158
2b947003 2159 if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
04d328de 2160 fio.op_flags &= ~REQ_META;
2b947003 2161
351df4b2 2162 set_page_writeback(page);
05ca3632 2163 f2fs_submit_page_mbio(&fio);
351df4b2
JK		 2164}
2165
05ca3632 2166void write_node_page(unsigned int nid, struct f2fs_io_info *fio)
351df4b2
JK		 2167{
2168 struct f2fs_summary sum;
05ca3632 2169
351df4b2 2170 set_summary(&sum, nid, 0, 0);
05ca3632 2171 do_write_page(&sum, fio);
351df4b2
JK		 2172}
2173
05ca3632 2174void write_data_page(struct dnode_of_data *dn, struct f2fs_io_info *fio)
351df4b2 2175{
05ca3632 2176 struct f2fs_sb_info *sbi = fio->sbi;
351df4b2
JK		 2177 struct f2fs_summary sum;
2178 struct node_info ni;
2179
9850cf4a 2180 f2fs_bug_on(sbi, dn->data_blkaddr == NULL_ADDR);
351df4b2
JK		 2181 get_node_info(sbi, dn->nid, &ni);
2182 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
05ca3632 2183 do_write_page(&sum, fio);
f28b3434 2184 f2fs_update_data_blkaddr(dn, fio->new_blkaddr);
351df4b2
JK		 2185}
2186
d1b3e72d 2187int rewrite_data_page(struct f2fs_io_info *fio)
351df4b2 2188{
7a9d7548 2189 fio->new_blkaddr = fio->old_blkaddr;
05ca3632 2190 stat_inc_inplace_blocks(fio->sbi);
d1b3e72d 2191 return f2fs_submit_page_bio(fio);
351df4b2
JK		 2192}
2193
4356e48e 2194void __f2fs_replace_block(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
19f106bc 2195 block_t old_blkaddr, block_t new_blkaddr,
28bc106b 2196 bool recover_curseg, bool recover_newaddr)
351df4b2
JK		 2197{
2198 struct sit_info *sit_i = SIT_I(sbi);
2199 struct curseg_info *curseg;
2200 unsigned int segno, old_cursegno;
2201 struct seg_entry *se;
2202 int type;
19f106bc 2203 unsigned short old_blkoff;
351df4b2
JK		 2204
2205 segno = GET_SEGNO(sbi, new_blkaddr);
2206 se = get_seg_entry(sbi, segno);
2207 type = se->type;
2208
19f106bc
CY		 2209 if (!recover_curseg) {
2210 /* for recovery flow */
2211 if (se->valid_blocks == 0 && !IS_CURSEG(sbi, segno)) {
2212 if (old_blkaddr == NULL_ADDR)
2213 type = CURSEG_COLD_DATA;
2214 else
2215 type = CURSEG_WARM_DATA;
2216 }
2217 } else {
2218 if (!IS_CURSEG(sbi, segno))
351df4b2
JK		 2219 type = CURSEG_WARM_DATA;
2220 }
19f106bc 2221
351df4b2
JK		 2222 curseg = CURSEG_I(sbi, type);
2223
2224 mutex_lock(&curseg->curseg_mutex);
2225 mutex_lock(&sit_i->sentry_lock);
2226
2227 old_cursegno = curseg->segno;
19f106bc 2228 old_blkoff = curseg->next_blkoff;
351df4b2
JK		 2229
2230 /* change the current segment */
2231 if (segno != curseg->segno) {
2232 curseg->next_segno = segno;
2233 change_curseg(sbi, type, true);
2234 }
2235
491c0854 2236 curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
e79efe3b 2237 __add_sum_entry(sbi, type, sum);
351df4b2 2238
28bc106b 2239 if (!recover_curseg || recover_newaddr)
6e2c64ad
JK		 2240 update_sit_entry(sbi, new_blkaddr, 1);
2241 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
2242 update_sit_entry(sbi, old_blkaddr, -1);
2243
2244 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
2245 locate_dirty_segment(sbi, GET_SEGNO(sbi, new_blkaddr));
2246
351df4b2 2247 locate_dirty_segment(sbi, old_cursegno);
351df4b2 2248
19f106bc
CY		 2249 if (recover_curseg) {
2250 if (old_cursegno != curseg->segno) {
2251 curseg->next_segno = old_cursegno;
2252 change_curseg(sbi, type, true);
2253 }
2254 curseg->next_blkoff = old_blkoff;
2255 }
2256
351df4b2
JK		 2257 mutex_unlock(&sit_i->sentry_lock);
2258 mutex_unlock(&curseg->curseg_mutex);
2259}
2260
528e3459
CY		 2261void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
2262 block_t old_addr, block_t new_addr,
28bc106b
CY		 2263 unsigned char version, bool recover_curseg,
2264 bool recover_newaddr)
528e3459
CY		 2265{
2266 struct f2fs_summary sum;
2267
2268 set_summary(&sum, dn->nid, dn->ofs_in_node, version);
2269
28bc106b
CY		 2270 __f2fs_replace_block(sbi, &sum, old_addr, new_addr,
2271 recover_curseg, recover_newaddr);
528e3459 2272
f28b3434 2273 f2fs_update_data_blkaddr(dn, new_addr);
528e3459
CY		 2274}
2275
93dfe2ac 2276void f2fs_wait_on_page_writeback(struct page *page,
fec1d657 2277 enum page_type type, bool ordered)
93dfe2ac 2278{
93dfe2ac 2279 if (PageWriteback(page)) {
4081363f
JK		 2280 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
2281
942fd319
JK		 2282 f2fs_submit_merged_bio_cond(sbi, page->mapping->host,
2283 0, page->index, type, WRITE);
fec1d657
JK		 2284 if (ordered)
2285 wait_on_page_writeback(page);
2286 else
2287 wait_for_stable_page(page);
93dfe2ac
JK		 2288 }
2289}
2290
08b39fbd
CY		 2291void f2fs_wait_on_encrypted_page_writeback(struct f2fs_sb_info *sbi,
2292 block_t blkaddr)
2293{
2294 struct page *cpage;
2295
5d4c0af4 2296 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
08b39fbd
CY		 2297 return;
2298
08b39fbd
CY		 2299 cpage = find_lock_page(META_MAPPING(sbi), blkaddr);
2300 if (cpage) {
fec1d657 2301 f2fs_wait_on_page_writeback(cpage, DATA, true);
08b39fbd
CY		 2302 f2fs_put_page(cpage, 1);
2303 }
2304}
2305
351df4b2
JK		 2306static int read_compacted_summaries(struct f2fs_sb_info *sbi)
2307{
2308 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2309 struct curseg_info *seg_i;
2310 unsigned char *kaddr;
2311 struct page *page;
2312 block_t start;
2313 int i, j, offset;
2314
2315 start = start_sum_block(sbi);
2316
2317 page = get_meta_page(sbi, start++);
2318 kaddr = (unsigned char *)page_address(page);
2319
2320 /* Step 1: restore nat cache */
2321 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2322 memcpy(seg_i->journal, kaddr, SUM_JOURNAL_SIZE);
351df4b2
JK		 2323
2324 /* Step 2: restore sit cache */
2325 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2326 memcpy(seg_i->journal, kaddr + SUM_JOURNAL_SIZE, SUM_JOURNAL_SIZE);
351df4b2
JK		 2327 offset = 2 * SUM_JOURNAL_SIZE;
2328
2329 /* Step 3: restore summary entries */
2330 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2331 unsigned short blk_off;
2332 unsigned int segno;
2333
2334 seg_i = CURSEG_I(sbi, i);
2335 segno = le32_to_cpu(ckpt->cur_data_segno[i]);
2336 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[i]);
2337 seg_i->next_segno = segno;
2338 reset_curseg(sbi, i, 0);
2339 seg_i->alloc_type = ckpt->alloc_type[i];
2340 seg_i->next_blkoff = blk_off;
2341
2342 if (seg_i->alloc_type == SSR)
2343 blk_off = sbi->blocks_per_seg;
2344
2345 for (j = 0; j < blk_off; j++) {
2346 struct f2fs_summary *s;
2347 s = (struct f2fs_summary *)(kaddr + offset);
2348 seg_i->sum_blk->entries[j] = *s;
2349 offset += SUMMARY_SIZE;
09cbfeaf 2350 if (offset + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2351 SUM_FOOTER_SIZE)
2352 continue;
2353
2354 f2fs_put_page(page, 1);
2355 page = NULL;
2356
2357 page = get_meta_page(sbi, start++);
2358 kaddr = (unsigned char *)page_address(page);
2359 offset = 0;
2360 }
2361 }
2362 f2fs_put_page(page, 1);
2363 return 0;
2364}
2365
2366static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
2367{
2368 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2369 struct f2fs_summary_block *sum;
2370 struct curseg_info *curseg;
2371 struct page *new;
2372 unsigned short blk_off;
2373 unsigned int segno = 0;
2374 block_t blk_addr = 0;
2375
2376 /* get segment number and block addr */
2377 if (IS_DATASEG(type)) {
2378 segno = le32_to_cpu(ckpt->cur_data_segno[type]);
2379 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[type -
2380 CURSEG_HOT_DATA]);
119ee914 2381 if (__exist_node_summaries(sbi))
351df4b2
JK		 2382 blk_addr = sum_blk_addr(sbi, NR_CURSEG_TYPE, type);
2383 else
2384 blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
2385 } else {
2386 segno = le32_to_cpu(ckpt->cur_node_segno[type -
2387 CURSEG_HOT_NODE]);
2388 blk_off = le16_to_cpu(ckpt->cur_node_blkoff[type -
2389 CURSEG_HOT_NODE]);
119ee914 2390 if (__exist_node_summaries(sbi))
351df4b2
JK		 2391 blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
2392 type - CURSEG_HOT_NODE);
2393 else
2394 blk_addr = GET_SUM_BLOCK(sbi, segno);
2395 }
2396
2397 new = get_meta_page(sbi, blk_addr);
2398 sum = (struct f2fs_summary_block *)page_address(new);
2399
2400 if (IS_NODESEG(type)) {
119ee914 2401 if (__exist_node_summaries(sbi)) {
351df4b2
JK		 2402 struct f2fs_summary *ns = &sum->entries[0];
2403 int i;
2404 for (i = 0; i < sbi->blocks_per_seg; i++, ns++) {
2405 ns->version = 0;
2406 ns->ofs_in_node = 0;
2407 }
2408 } else {
d653788a
GZ		 2409 int err;
2410
2411 err = restore_node_summary(sbi, segno, sum);
2412 if (err) {
351df4b2 2413 f2fs_put_page(new, 1);
d653788a 2414 return err;
351df4b2
JK		 2415 }
2416 }
2417 }
2418
2419 /* set uncompleted segment to curseg */
2420 curseg = CURSEG_I(sbi, type);
2421 mutex_lock(&curseg->curseg_mutex);
b7ad7512
CY		 2422
2423 /* update journal info */
2424 down_write(&curseg->journal_rwsem);
2425 memcpy(curseg->journal, &sum->journal, SUM_JOURNAL_SIZE);
2426 up_write(&curseg->journal_rwsem);
2427
2428 memcpy(curseg->sum_blk->entries, sum->entries, SUM_ENTRY_SIZE);
2429 memcpy(&curseg->sum_blk->footer, &sum->footer, SUM_FOOTER_SIZE);
351df4b2
JK		 2430 curseg->next_segno = segno;
2431 reset_curseg(sbi, type, 0);
2432 curseg->alloc_type = ckpt->alloc_type[type];
2433 curseg->next_blkoff = blk_off;
2434 mutex_unlock(&curseg->curseg_mutex);
2435 f2fs_put_page(new, 1);
2436 return 0;
2437}
2438
2439static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
2440{
2441 int type = CURSEG_HOT_DATA;
e4fc5fbf 2442 int err;
351df4b2 2443
aaec2b1d 2444 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG)) {
3fa06d7b
CY		 2445 int npages = npages_for_summary_flush(sbi, true);
2446
2447 if (npages >= 2)
2448 ra_meta_pages(sbi, start_sum_block(sbi), npages,
26879fb1 2449 META_CP, true);
3fa06d7b 2450
351df4b2
JK		 2451 /* restore for compacted data summary */
2452 if (read_compacted_summaries(sbi))
2453 return -EINVAL;
2454 type = CURSEG_HOT_NODE;
2455 }
2456
119ee914 2457 if (__exist_node_summaries(sbi))
3fa06d7b 2458 ra_meta_pages(sbi, sum_blk_addr(sbi, NR_CURSEG_TYPE, type),
26879fb1 2459 NR_CURSEG_TYPE - type, META_CP, true);
3fa06d7b 2460
e4fc5fbf
CY		 2461 for (; type <= CURSEG_COLD_NODE; type++) {
2462 err = read_normal_summaries(sbi, type);
2463 if (err)
2464 return err;
2465 }
2466
351df4b2
JK		 2467 return 0;
2468}
2469
2470static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
2471{
2472 struct page *page;
2473 unsigned char *kaddr;
2474 struct f2fs_summary *summary;
2475 struct curseg_info *seg_i;
2476 int written_size = 0;
2477 int i, j;
2478
2479 page = grab_meta_page(sbi, blkaddr++);
2480 kaddr = (unsigned char *)page_address(page);
2481
2482 /* Step 1: write nat cache */
2483 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2484 memcpy(kaddr, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2485 written_size += SUM_JOURNAL_SIZE;
2486
2487 /* Step 2: write sit cache */
2488 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2489 memcpy(kaddr + written_size, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2490 written_size += SUM_JOURNAL_SIZE;
2491
351df4b2
JK		 2492 /* Step 3: write summary entries */
2493 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2494 unsigned short blkoff;
2495 seg_i = CURSEG_I(sbi, i);
2496 if (sbi->ckpt->alloc_type[i] == SSR)
2497 blkoff = sbi->blocks_per_seg;
2498 else
2499 blkoff = curseg_blkoff(sbi, i);
2500
2501 for (j = 0; j < blkoff; j++) {
2502 if (!page) {
2503 page = grab_meta_page(sbi, blkaddr++);
2504 kaddr = (unsigned char *)page_address(page);
2505 written_size = 0;
2506 }
2507 summary = (struct f2fs_summary *)(kaddr + written_size);
2508 *summary = seg_i->sum_blk->entries[j];
2509 written_size += SUMMARY_SIZE;
351df4b2 2510
09cbfeaf 2511 if (written_size + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2512 SUM_FOOTER_SIZE)
2513 continue;
2514
e8d61a74 2515 set_page_dirty(page);
351df4b2
JK		 2516 f2fs_put_page(page, 1);
2517 page = NULL;
2518 }
2519 }
e8d61a74
CY		 2520 if (page) {
2521 set_page_dirty(page);
351df4b2 2522 f2fs_put_page(page, 1);
e8d61a74 2523 }
351df4b2
JK		 2524}
2525
2526static void write_normal_summaries(struct f2fs_sb_info *sbi,
2527 block_t blkaddr, int type)
2528{
2529 int i, end;
2530 if (IS_DATASEG(type))
2531 end = type + NR_CURSEG_DATA_TYPE;
2532 else
2533 end = type + NR_CURSEG_NODE_TYPE;
2534
b7ad7512
CY		 2535 for (i = type; i < end; i++)
2536 write_current_sum_page(sbi, i, blkaddr + (i - type));
351df4b2
JK		 2537}
2538
2539void write_data_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2540{
aaec2b1d 2541 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG))
351df4b2
JK		 2542 write_compacted_summaries(sbi, start_blk);
2543 else
2544 write_normal_summaries(sbi, start_blk, CURSEG_HOT_DATA);
2545}
2546
2547void write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2548{
119ee914 2549 write_normal_summaries(sbi, start_blk, CURSEG_HOT_NODE);
351df4b2
JK		 2550}
2551
dfc08a12 2552int lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
351df4b2
JK		 2553 unsigned int val, int alloc)
2554{
2555 int i;
2556
2557 if (type == NAT_JOURNAL) {
dfc08a12
CY		 2558 for (i = 0; i < nats_in_cursum(journal); i++) {
2559 if (le32_to_cpu(nid_in_journal(journal, i)) == val)
351df4b2
JK		 2560 return i;
2561 }
dfc08a12
CY		 2562 if (alloc && __has_cursum_space(journal, 1, NAT_JOURNAL))
2563 return update_nats_in_cursum(journal, 1);
351df4b2 2564 } else if (type == SIT_JOURNAL) {
dfc08a12
CY		 2565 for (i = 0; i < sits_in_cursum(journal); i++)
2566 if (le32_to_cpu(segno_in_journal(journal, i)) == val)
351df4b2 2567 return i;
dfc08a12
CY		 2568 if (alloc && __has_cursum_space(journal, 1, SIT_JOURNAL))
2569 return update_sits_in_cursum(journal, 1);
351df4b2
JK		 2570 }
2571 return -1;
2572}
2573
2574static struct page *get_current_sit_page(struct f2fs_sb_info *sbi,
2575 unsigned int segno)
2576{
2cc22186 2577 return get_meta_page(sbi, current_sit_addr(sbi, segno));
351df4b2
JK		 2578}
2579
2580static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
2581 unsigned int start)
2582{
2583 struct sit_info *sit_i = SIT_I(sbi);
2584 struct page *src_page, *dst_page;
2585 pgoff_t src_off, dst_off;
2586 void *src_addr, *dst_addr;
2587
2588 src_off = current_sit_addr(sbi, start);
2589 dst_off = next_sit_addr(sbi, src_off);
2590
2591 /* get current sit block page without lock */
2592 src_page = get_meta_page(sbi, src_off);
2593 dst_page = grab_meta_page(sbi, dst_off);
9850cf4a 2594 f2fs_bug_on(sbi, PageDirty(src_page));
351df4b2
JK		 2595
2596 src_addr = page_address(src_page);
2597 dst_addr = page_address(dst_page);
09cbfeaf 2598 memcpy(dst_addr, src_addr, PAGE_SIZE);
351df4b2
JK		 2599
2600 set_page_dirty(dst_page);
2601 f2fs_put_page(src_page, 1);
2602
2603 set_to_next_sit(sit_i, start);
2604
2605 return dst_page;
2606}
2607
184a5cd2
CY		 2608static struct sit_entry_set *grab_sit_entry_set(void)
2609{
2610 struct sit_entry_set *ses =
80c54505 2611 f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS);
184a5cd2
CY		 2612
2613 ses->entry_cnt = 0;
2614 INIT_LIST_HEAD(&ses->set_list);
2615 return ses;
2616}
2617
2618static void release_sit_entry_set(struct sit_entry_set *ses)
2619{
2620 list_del(&ses->set_list);
2621 kmem_cache_free(sit_entry_set_slab, ses);
2622}
2623
2624static void adjust_sit_entry_set(struct sit_entry_set *ses,
2625 struct list_head *head)
2626{
2627 struct sit_entry_set *next = ses;
2628
2629 if (list_is_last(&ses->set_list, head))
2630 return;
2631
2632 list_for_each_entry_continue(next, head, set_list)
2633 if (ses->entry_cnt <= next->entry_cnt)
2634 break;
2635
2636 list_move_tail(&ses->set_list, &next->set_list);
2637}
2638
2639static void add_sit_entry(unsigned int segno, struct list_head *head)
2640{
2641 struct sit_entry_set *ses;
2642 unsigned int start_segno = START_SEGNO(segno);
2643
2644 list_for_each_entry(ses, head, set_list) {
2645 if (ses->start_segno == start_segno) {
2646 ses->entry_cnt++;
2647 adjust_sit_entry_set(ses, head);
2648 return;
2649 }
2650 }
2651
2652 ses = grab_sit_entry_set();
2653
2654 ses->start_segno = start_segno;
2655 ses->entry_cnt++;
2656 list_add(&ses->set_list, head);
2657}
2658
2659static void add_sits_in_set(struct f2fs_sb_info *sbi)
2660{
2661 struct f2fs_sm_info *sm_info = SM_I(sbi);
2662 struct list_head *set_list = &sm_info->sit_entry_set;
2663 unsigned long *bitmap = SIT_I(sbi)->dirty_sentries_bitmap;
184a5cd2
CY		 2664 unsigned int segno;
2665
7cd8558b 2666 for_each_set_bit(segno, bitmap, MAIN_SEGS(sbi))
184a5cd2
CY		 2667 add_sit_entry(segno, set_list);
2668}
2669
2670static void remove_sits_in_journal(struct f2fs_sb_info *sbi)
351df4b2
JK		 2671{
2672 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2673 struct f2fs_journal *journal = curseg->journal;
351df4b2
JK		 2674 int i;
2675
b7ad7512 2676 down_write(&curseg->journal_rwsem);
dfc08a12 2677 for (i = 0; i < sits_in_cursum(journal); i++) {
184a5cd2
CY		 2678 unsigned int segno;
2679 bool dirtied;
2680
dfc08a12 2681 segno = le32_to_cpu(segno_in_journal(journal, i));
184a5cd2
CY		 2682 dirtied = __mark_sit_entry_dirty(sbi, segno);
2683
2684 if (!dirtied)
2685 add_sit_entry(segno, &SM_I(sbi)->sit_entry_set);
351df4b2 2686 }
dfc08a12 2687 update_sits_in_cursum(journal, -i);
b7ad7512 2688 up_write(&curseg->journal_rwsem);
351df4b2
JK		 2689}
2690
0a8165d7 2691/*
351df4b2
JK		 2692 * CP calls this function, which flushes SIT entries including sit_journal,
2693 * and moves prefree segs to free segs.
2694 */
4b2fecc8 2695void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2
JK		 2696{
2697 struct sit_info *sit_i = SIT_I(sbi);
2698 unsigned long *bitmap = sit_i->dirty_sentries_bitmap;
2699 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2700 struct f2fs_journal *journal = curseg->journal;
184a5cd2
CY		 2701 struct sit_entry_set *ses, *tmp;
2702 struct list_head *head = &SM_I(sbi)->sit_entry_set;
184a5cd2 2703 bool to_journal = true;
4b2fecc8 2704 struct seg_entry *se;
351df4b2 2705
351df4b2
JK		 2706 mutex_lock(&sit_i->sentry_lock);
2707
2b11a74b
WL		 2708 if (!sit_i->dirty_sentries)
2709 goto out;
2710
351df4b2 2711 /*
184a5cd2
CY		 2712 * add and account sit entries of dirty bitmap in sit entry
2713 * set temporarily
351df4b2 2714 */
184a5cd2 2715 add_sits_in_set(sbi);
351df4b2 2716
184a5cd2
CY		 2717 /*
2718 * if there are no enough space in journal to store dirty sit
2719 * entries, remove all entries from journal and add and account
2720 * them in sit entry set.
2721 */
dfc08a12 2722 if (!__has_cursum_space(journal, sit_i->dirty_sentries, SIT_JOURNAL))
184a5cd2 2723 remove_sits_in_journal(sbi);
b2955550 2724
184a5cd2
CY		 2725 /*
2726 * there are two steps to flush sit entries:
2727 * #1, flush sit entries to journal in current cold data summary block.
2728 * #2, flush sit entries to sit page.
2729 */
2730 list_for_each_entry_safe(ses, tmp, head, set_list) {
4a257ed6 2731 struct page *page = NULL;
184a5cd2
CY		 2732 struct f2fs_sit_block *raw_sit = NULL;
2733 unsigned int start_segno = ses->start_segno;
2734 unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
7cd8558b 2735 (unsigned long)MAIN_SEGS(sbi));
184a5cd2
CY		 2736 unsigned int segno = start_segno;
2737
2738 if (to_journal &&
dfc08a12 2739 !__has_cursum_space(journal, ses->entry_cnt, SIT_JOURNAL))
184a5cd2
CY		 2740 to_journal = false;
2741
b7ad7512
CY		 2742 if (to_journal) {
2743 down_write(&curseg->journal_rwsem);
2744 } else {
184a5cd2
CY		 2745 page = get_next_sit_page(sbi, start_segno);
2746 raw_sit = page_address(page);
351df4b2 2747 }
351df4b2 2748
184a5cd2
CY		 2749 /* flush dirty sit entries in region of current sit set */
2750 for_each_set_bit_from(segno, bitmap, end) {
2751 int offset, sit_offset;
4b2fecc8
JK		 2752
2753 se = get_seg_entry(sbi, segno);
184a5cd2
CY		 2754
2755 /* add discard candidates */
d7bc2484 2756 if (cpc->reason != CP_DISCARD) {
4b2fecc8 2757 cpc->trim_start = segno;
25290fa5 2758 add_discard_addrs(sbi, cpc, false);
4b2fecc8 2759 }
184a5cd2
CY		 2760
2761 if (to_journal) {
dfc08a12 2762 offset = lookup_journal_in_cursum(journal,
184a5cd2
CY		 2763 SIT_JOURNAL, segno, 1);
2764 f2fs_bug_on(sbi, offset < 0);
dfc08a12 2765 segno_in_journal(journal, offset) =
184a5cd2
CY		 2766 cpu_to_le32(segno);
2767 seg_info_to_raw_sit(se,
dfc08a12 2768 &sit_in_journal(journal, offset));
184a5cd2
CY		 2769 } else {
2770 sit_offset = SIT_ENTRY_OFFSET(sit_i, segno);
2771 seg_info_to_raw_sit(se,
2772 &raw_sit->entries[sit_offset]);
2773 }
351df4b2 2774
184a5cd2
CY		 2775 __clear_bit(segno, bitmap);
2776 sit_i->dirty_sentries--;
2777 ses->entry_cnt--;
351df4b2
JK		 2778 }
2779
b7ad7512
CY		 2780 if (to_journal)
2781 up_write(&curseg->journal_rwsem);
2782 else
184a5cd2
CY		 2783 f2fs_put_page(page, 1);
2784
2785 f2fs_bug_on(sbi, ses->entry_cnt);
2786 release_sit_entry_set(ses);
351df4b2 2787 }
184a5cd2
CY		 2788
2789 f2fs_bug_on(sbi, !list_empty(head));
2790 f2fs_bug_on(sbi, sit_i->dirty_sentries);
184a5cd2 2791out:
4b2fecc8 2792 if (cpc->reason == CP_DISCARD) {
650d3c4e
YH		 2793 __u64 trim_start = cpc->trim_start;
2794
4b2fecc8 2795 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++)
25290fa5 2796 add_discard_addrs(sbi, cpc, false);
650d3c4e
YH		 2797
2798 cpc->trim_start = trim_start;
4b2fecc8 2799 }
351df4b2 2800 mutex_unlock(&sit_i->sentry_lock);
351df4b2 2801
351df4b2
JK		 2802 set_prefree_as_free_segments(sbi);
2803}
2804
2805static int build_sit_info(struct f2fs_sb_info *sbi)
2806{
2807 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
351df4b2
JK		 2808 struct sit_info *sit_i;
2809 unsigned int sit_segs, start;
ae27d62e 2810 char *src_bitmap;
351df4b2
JK		 2811 unsigned int bitmap_size;
2812
2813 /* allocate memory for SIT information */
2814 sit_i = kzalloc(sizeof(struct sit_info), GFP_KERNEL);
2815 if (!sit_i)
2816 return -ENOMEM;
2817
2818 SM_I(sbi)->sit_info = sit_i;
2819
39307a8e
JK		 2820 sit_i->sentries = f2fs_kvzalloc(MAIN_SEGS(sbi) *
2821 sizeof(struct seg_entry), GFP_KERNEL);
351df4b2
JK		 2822 if (!sit_i->sentries)
2823 return -ENOMEM;
2824
7cd8558b 2825 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2826 sit_i->dirty_sentries_bitmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2827 if (!sit_i->dirty_sentries_bitmap)
2828 return -ENOMEM;
2829
7cd8558b 2830 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 2831 sit_i->sentries[start].cur_valid_map
2832 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2833 sit_i->sentries[start].ckpt_valid_map
2834 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
a66cdd98 2835 if (!sit_i->sentries[start].cur_valid_map ||
3e025740 2836 !sit_i->sentries[start].ckpt_valid_map)
351df4b2 2837 return -ENOMEM;
3e025740 2838
355e7891
CY		 2839#ifdef CONFIG_F2FS_CHECK_FS
2840 sit_i->sentries[start].cur_valid_map_mir
2841 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2842 if (!sit_i->sentries[start].cur_valid_map_mir)
2843 return -ENOMEM;
2844#endif
2845
3e025740
JK		 2846 if (f2fs_discard_en(sbi)) {
2847 sit_i->sentries[start].discard_map
2848 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2849 if (!sit_i->sentries[start].discard_map)
2850 return -ENOMEM;
2851 }
351df4b2
JK		 2852 }
2853
60a3b782
JK		 2854 sit_i->tmp_map = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2855 if (!sit_i->tmp_map)
2856 return -ENOMEM;
2857
351df4b2 2858 if (sbi->segs_per_sec > 1) {
39307a8e
JK		 2859 sit_i->sec_entries = f2fs_kvzalloc(MAIN_SECS(sbi) *
2860 sizeof(struct sec_entry), GFP_KERNEL);
351df4b2
JK		 2861 if (!sit_i->sec_entries)
2862 return -ENOMEM;
2863 }
2864
2865 /* get information related with SIT */
2866 sit_segs = le32_to_cpu(raw_super->segment_count_sit) >> 1;
2867
2868 /* setup SIT bitmap from ckeckpoint pack */
2869 bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
2870 src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
2871
ae27d62e
CY		 2872 sit_i->sit_bitmap = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2873 if (!sit_i->sit_bitmap)
351df4b2 2874 return -ENOMEM;
351df4b2 2875
ae27d62e
CY		 2876#ifdef CONFIG_F2FS_CHECK_FS
2877 sit_i->sit_bitmap_mir = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2878 if (!sit_i->sit_bitmap_mir)
2879 return -ENOMEM;
2880#endif
2881
351df4b2
JK		 2882 /* init SIT information */
2883 sit_i->s_ops = &default_salloc_ops;
2884
2885 sit_i->sit_base_addr = le32_to_cpu(raw_super->sit_blkaddr);
2886 sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
c79b7ff1 2887 sit_i->written_valid_blocks = 0;
351df4b2
JK		 2888 sit_i->bitmap_size = bitmap_size;
2889 sit_i->dirty_sentries = 0;
2890 sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
2891 sit_i->elapsed_time = le64_to_cpu(sbi->ckpt->elapsed_time);
2892 sit_i->mounted_time = CURRENT_TIME_SEC.tv_sec;
2893 mutex_init(&sit_i->sentry_lock);
2894 return 0;
2895}
2896
2897static int build_free_segmap(struct f2fs_sb_info *sbi)
2898{
351df4b2
JK		 2899 struct free_segmap_info *free_i;
2900 unsigned int bitmap_size, sec_bitmap_size;
2901
2902 /* allocate memory for free segmap information */
2903 free_i = kzalloc(sizeof(struct free_segmap_info), GFP_KERNEL);
2904 if (!free_i)
2905 return -ENOMEM;
2906
2907 SM_I(sbi)->free_info = free_i;
2908
7cd8558b 2909 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2910 free_i->free_segmap = f2fs_kvmalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2911 if (!free_i->free_segmap)
2912 return -ENOMEM;
2913
7cd8558b 2914 sec_bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
39307a8e 2915 free_i->free_secmap = f2fs_kvmalloc(sec_bitmap_size, GFP_KERNEL);
351df4b2
JK		 2916 if (!free_i->free_secmap)
2917 return -ENOMEM;
2918
2919 /* set all segments as dirty temporarily */
2920 memset(free_i->free_segmap, 0xff, bitmap_size);
2921 memset(free_i->free_secmap, 0xff, sec_bitmap_size);
2922
2923 /* init free segmap information */
7cd8558b 2924 free_i->start_segno = GET_SEGNO_FROM_SEG0(sbi, MAIN_BLKADDR(sbi));
351df4b2
JK		 2925 free_i->free_segments = 0;
2926 free_i->free_sections = 0;
1a118ccf 2927 spin_lock_init(&free_i->segmap_lock);
351df4b2
JK		 2928 return 0;
2929}
2930
2931static int build_curseg(struct f2fs_sb_info *sbi)
2932{
1042d60f 2933 struct curseg_info *array;
351df4b2
JK		 2934 int i;
2935
b434babf 2936 array = kcalloc(NR_CURSEG_TYPE, sizeof(*array), GFP_KERNEL);
351df4b2
JK		 2937 if (!array)
2938 return -ENOMEM;
2939
2940 SM_I(sbi)->curseg_array = array;
2941
2942 for (i = 0; i < NR_CURSEG_TYPE; i++) {
2943 mutex_init(&array[i].curseg_mutex);
09cbfeaf 2944 array[i].sum_blk = kzalloc(PAGE_SIZE, GFP_KERNEL);
351df4b2
JK		 2945 if (!array[i].sum_blk)
2946 return -ENOMEM;
b7ad7512
CY		 2947 init_rwsem(&array[i].journal_rwsem);
2948 array[i].journal = kzalloc(sizeof(struct f2fs_journal),
2949 GFP_KERNEL);
2950 if (!array[i].journal)
2951 return -ENOMEM;
351df4b2
JK		 2952 array[i].segno = NULL_SEGNO;
2953 array[i].next_blkoff = 0;
2954 }
2955 return restore_curseg_summaries(sbi);
2956}
2957
2958static void build_sit_entries(struct f2fs_sb_info *sbi)
2959{
2960 struct sit_info *sit_i = SIT_I(sbi);
2961 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2962 struct f2fs_journal *journal = curseg->journal;
9c094040
YH		 2963 struct seg_entry *se;
2964 struct f2fs_sit_entry sit;
74de593a
CY		 2965 int sit_blk_cnt = SIT_BLK_CNT(sbi);
2966 unsigned int i, start, end;
2967 unsigned int readed, start_blk = 0;
351df4b2 2968
74de593a 2969 do {
664ba972
JK		 2970 readed = ra_meta_pages(sbi, start_blk, BIO_MAX_PAGES,
2971 META_SIT, true);
74de593a
CY		 2972
2973 start = start_blk * sit_i->sents_per_block;
2974 end = (start_blk + readed) * sit_i->sents_per_block;
2975
7cd8558b 2976 for (; start < end && start < MAIN_SEGS(sbi); start++) {
74de593a 2977 struct f2fs_sit_block *sit_blk;
74de593a
CY		 2978 struct page *page;
2979
9c094040 2980 se = &sit_i->sentries[start];
74de593a
CY		 2981 page = get_current_sit_page(sbi, start);
2982 sit_blk = (struct f2fs_sit_block *)page_address(page);
2983 sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
2984 f2fs_put_page(page, 1);
d600af23 2985
74de593a
CY		 2986 check_block_count(sbi, start, &sit);
2987 seg_info_from_raw_sit(se, &sit);
a66cdd98
JK		 2988
2989 /* build discard map only one time */
3e025740
JK		 2990 if (f2fs_discard_en(sbi)) {
2991 memcpy(se->discard_map, se->cur_valid_map,
2992 SIT_VBLOCK_MAP_SIZE);
2993 sbi->discard_blks += sbi->blocks_per_seg -
2994 se->valid_blocks;
2995 }
a66cdd98 2996
d600af23
CY		 2997 if (sbi->segs_per_sec > 1)
2998 get_sec_entry(sbi, start)->valid_blocks +=
2999 se->valid_blocks;
351df4b2 3000 }
74de593a
CY		 3001 start_blk += readed;
3002 } while (start_blk < sit_blk_cnt);
d600af23
CY		 3003
3004 down_read(&curseg->journal_rwsem);
3005 for (i = 0; i < sits_in_cursum(journal); i++) {
d600af23
CY		 3006 unsigned int old_valid_blocks;
3007
3008 start = le32_to_cpu(segno_in_journal(journal, i));
3009 se = &sit_i->sentries[start];
3010 sit = sit_in_journal(journal, i);
3011
3012 old_valid_blocks = se->valid_blocks;
3013
3014 check_block_count(sbi, start, &sit);
3015 seg_info_from_raw_sit(se, &sit);
3016
3017 if (f2fs_discard_en(sbi)) {
3018 memcpy(se->discard_map, se->cur_valid_map,
3019 SIT_VBLOCK_MAP_SIZE);
3020 sbi->discard_blks += old_valid_blocks -
3021 se->valid_blocks;
3022 }
3023
3024 if (sbi->segs_per_sec > 1)
3025 get_sec_entry(sbi, start)->valid_blocks +=
3026 se->valid_blocks - old_valid_blocks;
3027 }
3028 up_read(&curseg->journal_rwsem);
351df4b2
JK		 3029}
3030
3031static void init_free_segmap(struct f2fs_sb_info *sbi)
3032{
3033 unsigned int start;
3034 int type;
3035
7cd8558b 3036 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 3037 struct seg_entry *sentry = get_seg_entry(sbi, start);
3038 if (!sentry->valid_blocks)
3039 __set_free(sbi, start);
c79b7ff1
JK		 3040 else
3041 SIT_I(sbi)->written_valid_blocks +=
3042 sentry->valid_blocks;
351df4b2
JK		 3043 }
3044
3045 /* set use the current segments */
3046 for (type = CURSEG_HOT_DATA; type <= CURSEG_COLD_NODE; type++) {
3047 struct curseg_info *curseg_t = CURSEG_I(sbi, type);
3048 __set_test_and_inuse(sbi, curseg_t->segno);
3049 }
3050}
3051
3052static void init_dirty_segmap(struct f2fs_sb_info *sbi)
3053{
3054 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3055 struct free_segmap_info *free_i = FREE_I(sbi);
7cd8558b 3056 unsigned int segno = 0, offset = 0;
351df4b2
JK		 3057 unsigned short valid_blocks;
3058
8736fbf0 3059 while (1) {
351df4b2 3060 /* find dirty segment based on free segmap */
7cd8558b
JK		 3061 segno = find_next_inuse(free_i, MAIN_SEGS(sbi), offset);
3062 if (segno >= MAIN_SEGS(sbi))
351df4b2
JK		 3063 break;
3064 offset = segno + 1;
302bd348 3065 valid_blocks = get_valid_blocks(sbi, segno, false);
ec325b52 3066 if (valid_blocks == sbi->blocks_per_seg || !valid_blocks)
351df4b2 3067 continue;
ec325b52
JK		 3068 if (valid_blocks > sbi->blocks_per_seg) {
3069 f2fs_bug_on(sbi, 1);
3070 continue;
3071 }
351df4b2
JK		 3072 mutex_lock(&dirty_i->seglist_lock);
3073 __locate_dirty_segment(sbi, segno, DIRTY);
3074 mutex_unlock(&dirty_i->seglist_lock);
3075 }
3076}
3077
5ec4e49f 3078static int init_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3079{
3080 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
7cd8558b 3081 unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
351df4b2 3082
39307a8e 3083 dirty_i->victim_secmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
5ec4e49f 3084 if (!dirty_i->victim_secmap)
351df4b2
JK		 3085 return -ENOMEM;
3086 return 0;
3087}
3088
3089static int build_dirty_segmap(struct f2fs_sb_info *sbi)
3090{
3091 struct dirty_seglist_info *dirty_i;
3092 unsigned int bitmap_size, i;
3093
3094 /* allocate memory for dirty segments list information */
3095 dirty_i = kzalloc(sizeof(struct dirty_seglist_info), GFP_KERNEL);
3096 if (!dirty_i)
3097 return -ENOMEM;
3098
3099 SM_I(sbi)->dirty_info = dirty_i;
3100 mutex_init(&dirty_i->seglist_lock);
3101
7cd8558b 3102 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
351df4b2
JK		 3103
3104 for (i = 0; i < NR_DIRTY_TYPE; i++) {
39307a8e 3105 dirty_i->dirty_segmap[i] = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 3106 if (!dirty_i->dirty_segmap[i])
3107 return -ENOMEM;
3108 }
3109
3110 init_dirty_segmap(sbi);
5ec4e49f 3111 return init_victim_secmap(sbi);
351df4b2
JK		 3112}
3113
0a8165d7 3114/*
351df4b2
JK		 3115 * Update min, max modified time for cost-benefit GC algorithm
3116 */
3117static void init_min_max_mtime(struct f2fs_sb_info *sbi)
3118{
3119 struct sit_info *sit_i = SIT_I(sbi);
3120 unsigned int segno;
3121
3122 mutex_lock(&sit_i->sentry_lock);
3123
3124 sit_i->min_mtime = LLONG_MAX;
3125
7cd8558b 3126 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
351df4b2
JK		 3127 unsigned int i;
3128 unsigned long long mtime = 0;
3129
3130 for (i = 0; i < sbi->segs_per_sec; i++)
3131 mtime += get_seg_entry(sbi, segno + i)->mtime;
3132
3133 mtime = div_u64(mtime, sbi->segs_per_sec);
3134
3135 if (sit_i->min_mtime > mtime)
3136 sit_i->min_mtime = mtime;
3137 }
3138 sit_i->max_mtime = get_mtime(sbi);
3139 mutex_unlock(&sit_i->sentry_lock);
3140}
3141
3142int build_segment_manager(struct f2fs_sb_info *sbi)
3143{
3144 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3145 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1042d60f 3146 struct f2fs_sm_info *sm_info;
351df4b2
JK		 3147 int err;
3148
3149 sm_info = kzalloc(sizeof(struct f2fs_sm_info), GFP_KERNEL);
3150 if (!sm_info)
3151 return -ENOMEM;
3152
3153 /* init sm info */
3154 sbi->sm_info = sm_info;
351df4b2
JK		 3155 sm_info->seg0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3156 sm_info->main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3157 sm_info->segment_count = le32_to_cpu(raw_super->segment_count);
3158 sm_info->reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3159 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3160 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
3161 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
58c41035
JK		 3162 sm_info->rec_prefree_segments = sm_info->main_segments *
3163 DEF_RECLAIM_PREFREE_SEGMENTS / 100;
44a83499
JK		 3164 if (sm_info->rec_prefree_segments > DEF_MAX_RECLAIM_PREFREE_SEGMENTS)
3165 sm_info->rec_prefree_segments = DEF_MAX_RECLAIM_PREFREE_SEGMENTS;
3166
52763a4b
JK		 3167 if (!test_opt(sbi, LFS))
3168 sm_info->ipu_policy = 1 << F2FS_IPU_FSYNC;
216fbd64 3169 sm_info->min_ipu_util = DEF_MIN_IPU_UTIL;
c1ce1b02 3170 sm_info->min_fsync_blocks = DEF_MIN_FSYNC_BLOCKS;
ef095d19 3171 sm_info->min_hot_blocks = DEF_MIN_HOT_BLOCKS;
351df4b2 3172
bba681cb
JK		 3173 sm_info->trim_sections = DEF_BATCHED_TRIM_SECTIONS;
3174
184a5cd2
CY		 3175 INIT_LIST_HEAD(&sm_info->sit_entry_set);
3176
b270ad6f 3177 if (test_opt(sbi, FLUSH_MERGE) && !f2fs_readonly(sbi->sb)) {
2163d198
GZ		 3178 err = create_flush_cmd_control(sbi);
3179 if (err)
a688b9d9 3180 return err;
6b4afdd7
JK		 3181 }
3182
0b54fb84
JK		 3183 err = create_discard_cmd_control(sbi);
3184 if (err)
3185 return err;
3186
351df4b2
JK		 3187 err = build_sit_info(sbi);
3188 if (err)
3189 return err;
3190 err = build_free_segmap(sbi);
3191 if (err)
3192 return err;
3193 err = build_curseg(sbi);
3194 if (err)
3195 return err;
3196
3197 /* reinit free segmap based on SIT */
3198 build_sit_entries(sbi);
3199
3200 init_free_segmap(sbi);
3201 err = build_dirty_segmap(sbi);
3202 if (err)
3203 return err;
3204
3205 init_min_max_mtime(sbi);
3206 return 0;
3207}
3208
3209static void discard_dirty_segmap(struct f2fs_sb_info *sbi,
3210 enum dirty_type dirty_type)
3211{
3212 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3213
3214 mutex_lock(&dirty_i->seglist_lock);
39307a8e 3215 kvfree(dirty_i->dirty_segmap[dirty_type]);
351df4b2
JK		 3216 dirty_i->nr_dirty[dirty_type] = 0;
3217 mutex_unlock(&dirty_i->seglist_lock);
3218}
3219
5ec4e49f 3220static void destroy_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3221{
3222 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
39307a8e 3223 kvfree(dirty_i->victim_secmap);
351df4b2
JK		 3224}
3225
3226static void destroy_dirty_segmap(struct f2fs_sb_info *sbi)
3227{
3228 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3229 int i;
3230
3231 if (!dirty_i)
3232 return;
3233
3234 /* discard pre-free/dirty segments list */
3235 for (i = 0; i < NR_DIRTY_TYPE; i++)
3236 discard_dirty_segmap(sbi, i);
3237
5ec4e49f 3238 destroy_victim_secmap(sbi);
351df4b2
JK		 3239 SM_I(sbi)->dirty_info = NULL;
3240 kfree(dirty_i);
3241}
3242
3243static void destroy_curseg(struct f2fs_sb_info *sbi)
3244{
3245 struct curseg_info *array = SM_I(sbi)->curseg_array;
3246 int i;
3247
3248 if (!array)
3249 return;
3250 SM_I(sbi)->curseg_array = NULL;
b7ad7512 3251 for (i = 0; i < NR_CURSEG_TYPE; i++) {
351df4b2 3252 kfree(array[i].sum_blk);
b7ad7512
CY		 3253 kfree(array[i].journal);
3254 }
351df4b2
JK		 3255 kfree(array);
3256}
3257
3258static void destroy_free_segmap(struct f2fs_sb_info *sbi)
3259{
3260 struct free_segmap_info *free_i = SM_I(sbi)->free_info;
3261 if (!free_i)
3262 return;
3263 SM_I(sbi)->free_info = NULL;
39307a8e
JK		 3264 kvfree(free_i->free_segmap);
3265 kvfree(free_i->free_secmap);
351df4b2
JK		 3266 kfree(free_i);
3267}
3268
3269static void destroy_sit_info(struct f2fs_sb_info *sbi)
3270{
3271 struct sit_info *sit_i = SIT_I(sbi);
3272 unsigned int start;
3273
3274 if (!sit_i)
3275 return;
3276
3277 if (sit_i->sentries) {
7cd8558b 3278 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2 3279 kfree(sit_i->sentries[start].cur_valid_map);
355e7891
CY		 3280#ifdef CONFIG_F2FS_CHECK_FS
3281 kfree(sit_i->sentries[start].cur_valid_map_mir);
3282#endif
351df4b2 3283 kfree(sit_i->sentries[start].ckpt_valid_map);
a66cdd98 3284 kfree(sit_i->sentries[start].discard_map);
351df4b2
JK		 3285 }
3286 }
60a3b782
JK		 3287 kfree(sit_i->tmp_map);
3288
39307a8e
JK		 3289 kvfree(sit_i->sentries);
3290 kvfree(sit_i->sec_entries);
3291 kvfree(sit_i->dirty_sentries_bitmap);
351df4b2
JK		 3292
3293 SM_I(sbi)->sit_info = NULL;
3294 kfree(sit_i->sit_bitmap);
ae27d62e
CY		 3295#ifdef CONFIG_F2FS_CHECK_FS
3296 kfree(sit_i->sit_bitmap_mir);
3297#endif
351df4b2
JK		 3298 kfree(sit_i);
3299}
3300
3301void destroy_segment_manager(struct f2fs_sb_info *sbi)
3302{
3303 struct f2fs_sm_info *sm_info = SM_I(sbi);
a688b9d9 3304
3b03f724
CY		 3305 if (!sm_info)
3306 return;
5eba8c5d 3307 destroy_flush_cmd_control(sbi, true);
f099405f 3308 destroy_discard_cmd_control(sbi);
351df4b2
JK		 3309 destroy_dirty_segmap(sbi);
3310 destroy_curseg(sbi);
3311 destroy_free_segmap(sbi);
3312 destroy_sit_info(sbi);
3313 sbi->sm_info = NULL;
3314 kfree(sm_info);
3315}
7fd9e544
JK		 3316
3317int __init create_segment_manager_caches(void)
3318{
3319 discard_entry_slab = f2fs_kmem_cache_create("discard_entry",
e8512d2e 3320 sizeof(struct discard_entry));
7fd9e544 3321 if (!discard_entry_slab)
184a5cd2
CY		 3322 goto fail;
3323
b01a9201
JK		 3324 discard_cmd_slab = f2fs_kmem_cache_create("discard_cmd",
3325 sizeof(struct discard_cmd));
3326 if (!discard_cmd_slab)
6ab2a308 3327 goto destroy_discard_entry;
275b66b0 3328
184a5cd2 3329 sit_entry_set_slab = f2fs_kmem_cache_create("sit_entry_set",
c9ee0085 3330 sizeof(struct sit_entry_set));
184a5cd2 3331 if (!sit_entry_set_slab)
b01a9201 3332 goto destroy_discard_cmd;
88b88a66
JK		 3333
3334 inmem_entry_slab = f2fs_kmem_cache_create("inmem_page_entry",
3335 sizeof(struct inmem_pages));
3336 if (!inmem_entry_slab)
3337 goto destroy_sit_entry_set;
7fd9e544 3338 return 0;
184a5cd2 3339
88b88a66
JK		 3340destroy_sit_entry_set:
3341 kmem_cache_destroy(sit_entry_set_slab);
b01a9201
JK		 3342destroy_discard_cmd:
3343 kmem_cache_destroy(discard_cmd_slab);
6ab2a308 3344destroy_discard_entry:
184a5cd2
CY		 3345 kmem_cache_destroy(discard_entry_slab);
3346fail:
3347 return -ENOMEM;
7fd9e544
JK		 3348}
3349
3350void destroy_segment_manager_caches(void)
3351{
184a5cd2 3352 kmem_cache_destroy(sit_entry_set_slab);
b01a9201 3353 kmem_cache_destroy(discard_cmd_slab);
7fd9e544 3354 kmem_cache_destroy(discard_entry_slab);
88b88a66 3355 kmem_cache_destroy(inmem_entry_slab);
7fd9e544 3356}
Ubuntu Artful kernel mirror