]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - fs/f2fs/segment.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: seperate read nat page from nat_tree_lock
[mirror_ubuntu-jammy-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);
971 wake_up(&SM_I(sbi)->dcc_info->discard_wait_queue);
972 return 0;
973}
974
975/* This should be covered by global mutex, &sit_i->sentry_lock */
976void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
977{
978 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
979 struct discard_cmd *dc;
980
981 mutex_lock(&dcc->cmd_lock);
982
983 dc = (struct discard_cmd *)__lookup_rb_tree(&dcc->root, NULL, blkaddr);
984 if (dc) {
985 if (dc->state != D_PREP)
986 wait_for_completion_io(&dc->wait);
987 __punch_discard_cmd(sbi, dc, blkaddr);
275b66b0 988 }
40465257 989
d431413f
CY		 990 mutex_unlock(&dcc->cmd_lock);
991}
992
993/* This comes from f2fs_put_super */
994void f2fs_wait_discard_bios(struct f2fs_sb_info *sbi)
995{
996 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
ba48a33e 997 struct list_head *pend_list;
46f84c2c 998 struct list_head *wait_list = &(dcc->wait_list);
d431413f
CY		 999 struct discard_cmd *dc, *tmp;
1000 struct blk_plug plug;
ba48a33e 1001 int i;
d431413f
CY		 1002
1003 mutex_lock(&dcc->cmd_lock);
1004
1005 blk_start_plug(&plug);
ba48a33e
CY		 1006 for (i = 0; i < MAX_PLIST_NUM; i++) {
1007 pend_list = &dcc->pend_list[i];
1008 list_for_each_entry_safe(dc, tmp, pend_list, list)
1009 __submit_discard_cmd(sbi, dc);
1010 }
d431413f
CY		 1011 blk_finish_plug(&plug);
1012
1013 list_for_each_entry_safe(dc, tmp, wait_list, list) {
1014 wait_for_completion_io(&dc->wait);
1015 __remove_discard_cmd(sbi, dc);
40465257 1016 }
d431413f 1017
15469963 1018 mutex_unlock(&dcc->cmd_lock);
275b66b0
CY		 1019}
1020
15469963
JK		 1021static int issue_discard_thread(void *data)
1022{
1023 struct f2fs_sb_info *sbi = data;
1024 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1025 wait_queue_head_t *q = &dcc->discard_wait_queue;
ba48a33e 1026 struct list_head *pend_list;
46f84c2c 1027 struct list_head *wait_list = &dcc->wait_list;
15469963
JK		 1028 struct discard_cmd *dc, *tmp;
1029 struct blk_plug plug;
ba48a33e 1030 int iter = 0, i;
15469963
JK		 1031repeat:
1032 if (kthread_should_stop())
1033 return 0;
1034
15469963 1035 mutex_lock(&dcc->cmd_lock);
a54455f5 1036 blk_start_plug(&plug);
ba48a33e
CY		 1037 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
1038 pend_list = &dcc->pend_list[i];
1039 list_for_each_entry_safe(dc, tmp, pend_list, list) {
1040 f2fs_bug_on(sbi, dc->state != D_PREP);
c81abe34 1041
ba48a33e
CY		 1042 if (is_idle(sbi))
1043 __submit_discard_cmd(sbi, dc);
c81abe34 1044
ba48a33e
CY		 1045 if (iter++ > DISCARD_ISSUE_RATE)
1046 goto next_step;
1047 }
22d375dd 1048 }
ba48a33e 1049next_step:
a54455f5 1050 blk_finish_plug(&plug);
22d375dd
CY		 1051
1052 list_for_each_entry_safe(dc, tmp, wait_list, list) {
fa64a003
CY		 1053 if (dc->state == D_DONE) {
1054 wait_for_completion_io(&dc->wait);
15469963 1055 __remove_discard_cmd(sbi, dc);
fa64a003 1056 }
15469963
JK		 1057 }
1058 mutex_unlock(&dcc->cmd_lock);
1059
15469963
JK		 1060 iter = 0;
1061 congestion_wait(BLK_RW_SYNC, HZ/50);
1062
22d375dd 1063 wait_event_interruptible(*q, kthread_should_stop() ||
ba48a33e 1064 atomic_read(&dcc->discard_cmd_cnt));
15469963
JK		 1065 goto repeat;
1066}
1067
f46e8809 1068#ifdef CONFIG_BLK_DEV_ZONED
3c62be17
JK		 1069static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
1070 struct block_device *bdev, block_t blkstart, block_t blklen)
f46e8809 1071{
92592285 1072 sector_t sector, nr_sects;
10a875f8 1073 block_t lblkstart = blkstart;
3c62be17
JK		 1074 int devi = 0;
1075
1076 if (sbi->s_ndevs) {
1077 devi = f2fs_target_device_index(sbi, blkstart);
1078 blkstart -= FDEV(devi).start_blk;
1079 }
f46e8809
DLM		 1080
1081 /*
1082 * We need to know the type of the zone: for conventional zones,
1083 * use regular discard if the drive supports it. For sequential
1084 * zones, reset the zone write pointer.
1085 */
3c62be17 1086 switch (get_blkz_type(sbi, bdev, blkstart)) {
f46e8809
DLM		 1087
1088 case BLK_ZONE_TYPE_CONVENTIONAL:
1089 if (!blk_queue_discard(bdev_get_queue(bdev)))
1090 return 0;
c81abe34 1091 return __queue_discard_cmd(sbi, bdev, lblkstart, blklen);
f46e8809
DLM		 1092 case BLK_ZONE_TYPE_SEQWRITE_REQ:
1093 case BLK_ZONE_TYPE_SEQWRITE_PREF:
92592285
JK		 1094 sector = SECTOR_FROM_BLOCK(blkstart);
1095 nr_sects = SECTOR_FROM_BLOCK(blklen);
1096
1097 if (sector & (bdev_zone_sectors(bdev) - 1) ||
1098 nr_sects != bdev_zone_sectors(bdev)) {
1099 f2fs_msg(sbi->sb, KERN_INFO,
1100 "(%d) %s: Unaligned discard attempted (block %x + %x)",
1101 devi, sbi->s_ndevs ? FDEV(devi).path: "",
1102 blkstart, blklen);
1103 return -EIO;
1104 }
d50aaeec 1105 trace_f2fs_issue_reset_zone(bdev, blkstart);
f46e8809
DLM		 1106 return blkdev_reset_zones(bdev, sector,
1107 nr_sects, GFP_NOFS);
1108 default:
1109 /* Unknown zone type: broken device ? */
1110 return -EIO;
1111 }
1112}
1113#endif
1114
3c62be17
JK		 1115static int __issue_discard_async(struct f2fs_sb_info *sbi,
1116 struct block_device *bdev, block_t blkstart, block_t blklen)
1117{
1118#ifdef CONFIG_BLK_DEV_ZONED
1119 if (f2fs_sb_mounted_blkzoned(sbi->sb) &&
1120 bdev_zoned_model(bdev) != BLK_ZONED_NONE)
1121 return __f2fs_issue_discard_zone(sbi, bdev, blkstart, blklen);
1122#endif
c81abe34 1123 return __queue_discard_cmd(sbi, bdev, blkstart, blklen);
3c62be17
JK		 1124}
1125
1e87a78d 1126static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
37208879
JK		 1127 block_t blkstart, block_t blklen)
1128{
3c62be17
JK		 1129 sector_t start = blkstart, len = 0;
1130 struct block_device *bdev;
a66cdd98
JK		 1131 struct seg_entry *se;
1132 unsigned int offset;
1133 block_t i;
3c62be17
JK		 1134 int err = 0;
1135
1136 bdev = f2fs_target_device(sbi, blkstart, NULL);
1137
1138 for (i = blkstart; i < blkstart + blklen; i++, len++) {
1139 if (i != start) {
1140 struct block_device *bdev2 =
1141 f2fs_target_device(sbi, i, NULL);
1142
1143 if (bdev2 != bdev) {
1144 err = __issue_discard_async(sbi, bdev,
1145 start, len);
1146 if (err)
1147 return err;
1148 bdev = bdev2;
1149 start = i;
1150 len = 0;
1151 }
1152 }
a66cdd98 1153
a66cdd98
JK		 1154 se = get_seg_entry(sbi, GET_SEGNO(sbi, i));
1155 offset = GET_BLKOFF_FROM_SEG0(sbi, i);
1156
1157 if (!f2fs_test_and_set_bit(offset, se->discard_map))
1158 sbi->discard_blks--;
1159 }
f46e8809 1160
3c62be17
JK		 1161 if (len)
1162 err = __issue_discard_async(sbi, bdev, start, len);
1163 return err;
1e87a78d
JK		 1164}
1165
25290fa5
JK		 1166static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc,
1167 bool check_only)
adf4983b 1168{
b2955550
JK		 1169 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
1170 int max_blocks = sbi->blocks_per_seg;
4b2fecc8 1171 struct seg_entry *se = get_seg_entry(sbi, cpc->trim_start);
b2955550
JK		 1172 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1173 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
a66cdd98 1174 unsigned long *discard_map = (unsigned long *)se->discard_map;
60a3b782 1175 unsigned long *dmap = SIT_I(sbi)->tmp_map;
b2955550 1176 unsigned int start = 0, end = -1;
4b2fecc8 1177 bool force = (cpc->reason == CP_DISCARD);
a7eeb823 1178 struct discard_entry *de = NULL;
46f84c2c 1179 struct list_head *head = &SM_I(sbi)->dcc_info->entry_list;
b2955550
JK		 1180 int i;
1181
3e025740 1182 if (se->valid_blocks == max_blocks || !f2fs_discard_en(sbi))
25290fa5 1183 return false;
b2955550 1184
a66cdd98
JK		 1185 if (!force) {
1186 if (!test_opt(sbi, DISCARD) || !se->valid_blocks ||
0b54fb84
JK		 1187 SM_I(sbi)->dcc_info->nr_discards >=
1188 SM_I(sbi)->dcc_info->max_discards)
25290fa5 1189 return false;
4b2fecc8
JK		 1190 }
1191
b2955550
JK		 1192 /* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
1193 for (i = 0; i < entries; i++)
a66cdd98 1194 dmap[i] = force ? ~ckpt_map[i] & ~discard_map[i] :
d7bc2484 1195 (cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
b2955550 1196
0b54fb84
JK		 1197 while (force || SM_I(sbi)->dcc_info->nr_discards <=
1198 SM_I(sbi)->dcc_info->max_discards) {
b2955550
JK		 1199 start = __find_rev_next_bit(dmap, max_blocks, end + 1);
1200 if (start >= max_blocks)
1201 break;
1202
1203 end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1);
c7b41e16
YH		 1204 if (force && start && end != max_blocks
1205 && (end - start) < cpc->trim_minlen)
1206 continue;
1207
25290fa5
JK		 1208 if (check_only)
1209 return true;
1210
a7eeb823
CY		 1211 if (!de) {
1212 de = f2fs_kmem_cache_alloc(discard_entry_slab,
1213 GFP_F2FS_ZERO);
1214 de->start_blkaddr = START_BLOCK(sbi, cpc->trim_start);
1215 list_add_tail(&de->list, head);
1216 }
1217
1218 for (i = start; i < end; i++)
1219 __set_bit_le(i, (void *)de->discard_map);
1220
1221 SM_I(sbi)->dcc_info->nr_discards += end - start;
b2955550 1222 }
25290fa5 1223 return false;
b2955550
JK		 1224}
1225
4b2fecc8
JK		 1226void release_discard_addrs(struct f2fs_sb_info *sbi)
1227{
46f84c2c 1228 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
4b2fecc8
JK		 1229 struct discard_entry *entry, *this;
1230
1231 /* drop caches */
1232 list_for_each_entry_safe(entry, this, head, list) {
1233 list_del(&entry->list);
1234 kmem_cache_free(discard_entry_slab, entry);
1235 }
1236}
1237
0a8165d7 1238/*
351df4b2
JK		 1239 * Should call clear_prefree_segments after checkpoint is done.
1240 */
1241static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
1242{
1243 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
b65ee148 1244 unsigned int segno;
351df4b2
JK		 1245
1246 mutex_lock(&dirty_i->seglist_lock);
7cd8558b 1247 for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], MAIN_SEGS(sbi))
351df4b2 1248 __set_test_and_free(sbi, segno);
351df4b2
JK		 1249 mutex_unlock(&dirty_i->seglist_lock);
1250}
1251
836b5a63 1252void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2 1253{
46f84c2c 1254 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
2d7b822a 1255 struct discard_entry *entry, *this;
351df4b2 1256 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
29e59c14 1257 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
29e59c14 1258 unsigned int start = 0, end = -1;
36abef4e 1259 unsigned int secno, start_segno;
c24a0fd6 1260 bool force = (cpc->reason == CP_DISCARD);
351df4b2
JK		 1261
1262 mutex_lock(&dirty_i->seglist_lock);
29e59c14 1263
351df4b2 1264 while (1) {
29e59c14 1265 int i;
7cd8558b
JK		 1266 start = find_next_bit(prefree_map, MAIN_SEGS(sbi), end + 1);
1267 if (start >= MAIN_SEGS(sbi))
351df4b2 1268 break;
7cd8558b
JK		 1269 end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
1270 start + 1);
29e59c14
CL		 1271
1272 for (i = start; i < end; i++)
1273 clear_bit(i, prefree_map);
1274
1275 dirty_i->nr_dirty[PRE] -= end - start;
1276
650d3c4e 1277 if (!test_opt(sbi, DISCARD))
29e59c14 1278 continue;
351df4b2 1279
650d3c4e
YH		 1280 if (force && start >= cpc->trim_start &&
1281 (end - 1) <= cpc->trim_end)
1282 continue;
1283
36abef4e
JK		 1284 if (!test_opt(sbi, LFS) || sbi->segs_per_sec == 1) {
1285 f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
37208879 1286 (end - start) << sbi->log_blocks_per_seg);
36abef4e
JK		 1287 continue;
1288 }
1289next:
4ddb1a4d
JK		 1290 secno = GET_SEC_FROM_SEG(sbi, start);
1291 start_segno = GET_SEG_FROM_SEC(sbi, secno);
36abef4e 1292 if (!IS_CURSEC(sbi, secno) &&
302bd348 1293 !get_valid_blocks(sbi, start, true))
36abef4e
JK		 1294 f2fs_issue_discard(sbi, START_BLOCK(sbi, start_segno),
1295 sbi->segs_per_sec << sbi->log_blocks_per_seg);
1296
1297 start = start_segno + sbi->segs_per_sec;
1298 if (start < end)
1299 goto next;
8b107f5b
JK		 1300 else
1301 end = start - 1;
351df4b2
JK		 1302 }
1303 mutex_unlock(&dirty_i->seglist_lock);
b2955550
JK		 1304
1305 /* send small discards */
2d7b822a 1306 list_for_each_entry_safe(entry, this, head, list) {
a7eeb823
CY		 1307 unsigned int cur_pos = 0, next_pos, len, total_len = 0;
1308 bool is_valid = test_bit_le(0, entry->discard_map);
1309
1310find_next:
1311 if (is_valid) {
1312 next_pos = find_next_zero_bit_le(entry->discard_map,
1313 sbi->blocks_per_seg, cur_pos);
1314 len = next_pos - cur_pos;
1315
1316 if (force && len < cpc->trim_minlen)
1317 goto skip;
1318
1319 f2fs_issue_discard(sbi, entry->start_blkaddr + cur_pos,
1320 len);
1321 cpc->trimmed += len;
1322 total_len += len;
1323 } else {
1324 next_pos = find_next_bit_le(entry->discard_map,
1325 sbi->blocks_per_seg, cur_pos);
1326 }
836b5a63 1327skip:
a7eeb823
CY		 1328 cur_pos = next_pos;
1329 is_valid = !is_valid;
1330
1331 if (cur_pos < sbi->blocks_per_seg)
1332 goto find_next;
1333
b2955550 1334 list_del(&entry->list);
a7eeb823 1335 SM_I(sbi)->dcc_info->nr_discards -= total_len;
b2955550
JK		 1336 kmem_cache_free(discard_entry_slab, entry);
1337 }
351df4b2
JK		 1338}
1339
8ed59745 1340static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84 1341{
15469963 1342 dev_t dev = sbi->sb->s_bdev->bd_dev;
0b54fb84 1343 struct discard_cmd_control *dcc;
ba48a33e 1344 int err = 0, i;
0b54fb84
JK		 1345
1346 if (SM_I(sbi)->dcc_info) {
1347 dcc = SM_I(sbi)->dcc_info;
1348 goto init_thread;
1349 }
1350
1351 dcc = kzalloc(sizeof(struct discard_cmd_control), GFP_KERNEL);
1352 if (!dcc)
1353 return -ENOMEM;
1354
46f84c2c 1355 INIT_LIST_HEAD(&dcc->entry_list);
ba48a33e
CY		 1356 for (i = 0; i < MAX_PLIST_NUM; i++)
1357 INIT_LIST_HEAD(&dcc->pend_list[i]);
46f84c2c 1358 INIT_LIST_HEAD(&dcc->wait_list);
15469963 1359 mutex_init(&dcc->cmd_lock);
8b8dd65f
CY		 1360 atomic_set(&dcc->issued_discard, 0);
1361 atomic_set(&dcc->issing_discard, 0);
5f32366a 1362 atomic_set(&dcc->discard_cmd_cnt, 0);
0b54fb84
JK		 1363 dcc->nr_discards = 0;
1364 dcc->max_discards = 0;
d84d1cbd 1365 dcc->undiscard_blks = 0;
004b6862 1366 dcc->root = RB_ROOT;
0b54fb84 1367
15469963 1368 init_waitqueue_head(&dcc->discard_wait_queue);
0b54fb84
JK		 1369 SM_I(sbi)->dcc_info = dcc;
1370init_thread:
15469963
JK		 1371 dcc->f2fs_issue_discard = kthread_run(issue_discard_thread, sbi,
1372 "f2fs_discard-%u:%u", MAJOR(dev), MINOR(dev));
1373 if (IS_ERR(dcc->f2fs_issue_discard)) {
1374 err = PTR_ERR(dcc->f2fs_issue_discard);
1375 kfree(dcc);
1376 SM_I(sbi)->dcc_info = NULL;
1377 return err;
1378 }
1379
0b54fb84
JK		 1380 return err;
1381}
1382
f099405f 1383static void destroy_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84
JK		 1384{
1385 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1386
f099405f
CY		 1387 if (!dcc)
1388 return;
1389
1390 if (dcc->f2fs_issue_discard) {
15469963
JK		 1391 struct task_struct *discard_thread = dcc->f2fs_issue_discard;
1392
1393 dcc->f2fs_issue_discard = NULL;
1394 kthread_stop(discard_thread);
1395 }
f099405f
CY		 1396
1397 kfree(dcc);
1398 SM_I(sbi)->dcc_info = NULL;
0b54fb84
JK		 1399}
1400
184a5cd2 1401static bool __mark_sit_entry_dirty(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 1402{
1403 struct sit_info *sit_i = SIT_I(sbi);
184a5cd2
CY		 1404
1405 if (!__test_and_set_bit(segno, sit_i->dirty_sentries_bitmap)) {
351df4b2 1406 sit_i->dirty_sentries++;
184a5cd2
CY		 1407 return false;
1408 }
1409
1410 return true;
351df4b2
JK		 1411}
1412
1413static void __set_sit_entry_type(struct f2fs_sb_info *sbi, int type,
1414 unsigned int segno, int modified)
1415{
1416 struct seg_entry *se = get_seg_entry(sbi, segno);
1417 se->type = type;
1418 if (modified)
1419 __mark_sit_entry_dirty(sbi, segno);
1420}
1421
1422static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
1423{
1424 struct seg_entry *se;
1425 unsigned int segno, offset;
1426 long int new_vblocks;
1427
1428 segno = GET_SEGNO(sbi, blkaddr);
1429
1430 se = get_seg_entry(sbi, segno);
1431 new_vblocks = se->valid_blocks + del;
491c0854 1432 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
351df4b2 1433
9850cf4a 1434 f2fs_bug_on(sbi, (new_vblocks >> (sizeof(unsigned short) << 3) ||
351df4b2
JK		 1435 (new_vblocks > sbi->blocks_per_seg)));
1436
1437 se->valid_blocks = new_vblocks;
1438 se->mtime = get_mtime(sbi);
1439 SIT_I(sbi)->max_mtime = se->mtime;
1440
1441 /* Update valid block bitmap */
1442 if (del > 0) {
355e7891
CY		 1443 if (f2fs_test_and_set_bit(offset, se->cur_valid_map)) {
1444#ifdef CONFIG_F2FS_CHECK_FS
1445 if (f2fs_test_and_set_bit(offset,
1446 se->cur_valid_map_mir))
1447 f2fs_bug_on(sbi, 1);
1448 else
1449 WARN_ON(1);
1450#else
05796763 1451 f2fs_bug_on(sbi, 1);
355e7891
CY		 1452#endif
1453 }
3e025740
JK		 1454 if (f2fs_discard_en(sbi) &&
1455 !f2fs_test_and_set_bit(offset, se->discard_map))
a66cdd98 1456 sbi->discard_blks--;
720037f9
JK		 1457
1458 /* don't overwrite by SSR to keep node chain */
1459 if (se->type == CURSEG_WARM_NODE) {
1460 if (!f2fs_test_and_set_bit(offset, se->ckpt_valid_map))
1461 se->ckpt_valid_blocks++;
1462 }
351df4b2 1463 } else {
355e7891
CY		 1464 if (!f2fs_test_and_clear_bit(offset, se->cur_valid_map)) {
1465#ifdef CONFIG_F2FS_CHECK_FS
1466 if (!f2fs_test_and_clear_bit(offset,
1467 se->cur_valid_map_mir))
1468 f2fs_bug_on(sbi, 1);
1469 else
1470 WARN_ON(1);
1471#else
05796763 1472 f2fs_bug_on(sbi, 1);
355e7891
CY		 1473#endif
1474 }
3e025740
JK		 1475 if (f2fs_discard_en(sbi) &&
1476 f2fs_test_and_clear_bit(offset, se->discard_map))
a66cdd98 1477 sbi->discard_blks++;
351df4b2
JK		 1478 }
1479 if (!f2fs_test_bit(offset, se->ckpt_valid_map))
1480 se->ckpt_valid_blocks += del;
1481
1482 __mark_sit_entry_dirty(sbi, segno);
1483
1484 /* update total number of valid blocks to be written in ckpt area */
1485 SIT_I(sbi)->written_valid_blocks += del;
1486
1487 if (sbi->segs_per_sec > 1)
1488 get_sec_entry(sbi, segno)->valid_blocks += del;
1489}
1490
5e443818 1491void refresh_sit_entry(struct f2fs_sb_info *sbi, block_t old, block_t new)
351df4b2 1492{
5e443818
JK		 1493 update_sit_entry(sbi, new, 1);
1494 if (GET_SEGNO(sbi, old) != NULL_SEGNO)
1495 update_sit_entry(sbi, old, -1);
1496
1497 locate_dirty_segment(sbi, GET_SEGNO(sbi, old));
1498 locate_dirty_segment(sbi, GET_SEGNO(sbi, new));
351df4b2
JK		 1499}
1500
1501void invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
1502{
1503 unsigned int segno = GET_SEGNO(sbi, addr);
1504 struct sit_info *sit_i = SIT_I(sbi);
1505
9850cf4a 1506 f2fs_bug_on(sbi, addr == NULL_ADDR);
351df4b2
JK		 1507 if (addr == NEW_ADDR)
1508 return;
1509
1510 /* add it into sit main buffer */
1511 mutex_lock(&sit_i->sentry_lock);
1512
1513 update_sit_entry(sbi, addr, -1);
1514
1515 /* add it into dirty seglist */
1516 locate_dirty_segment(sbi, segno);
1517
1518 mutex_unlock(&sit_i->sentry_lock);
1519}
1520
6e2c64ad
JK		 1521bool is_checkpointed_data(struct f2fs_sb_info *sbi, block_t blkaddr)
1522{
1523 struct sit_info *sit_i = SIT_I(sbi);
1524 unsigned int segno, offset;
1525 struct seg_entry *se;
1526 bool is_cp = false;
1527
1528 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
1529 return true;
1530
1531 mutex_lock(&sit_i->sentry_lock);
1532
1533 segno = GET_SEGNO(sbi, blkaddr);
1534 se = get_seg_entry(sbi, segno);
1535 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
1536
1537 if (f2fs_test_bit(offset, se->ckpt_valid_map))
1538 is_cp = true;
1539
1540 mutex_unlock(&sit_i->sentry_lock);
1541
1542 return is_cp;
1543}
1544
0a8165d7 1545/*
351df4b2
JK		 1546 * This function should be resided under the curseg_mutex lock
1547 */
1548static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
e79efe3b 1549 struct f2fs_summary *sum)
351df4b2
JK		 1550{
1551 struct curseg_info *curseg = CURSEG_I(sbi, type);
1552 void *addr = curseg->sum_blk;
e79efe3b 1553 addr += curseg->next_blkoff * sizeof(struct f2fs_summary);
351df4b2 1554 memcpy(addr, sum, sizeof(struct f2fs_summary));
351df4b2
JK		 1555}
1556
0a8165d7 1557/*
351df4b2
JK		 1558 * Calculate the number of current summary pages for writing
1559 */
3fa06d7b 1560int npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
351df4b2 1561{
351df4b2 1562 int valid_sum_count = 0;
9a47938b 1563 int i, sum_in_page;
351df4b2
JK		 1564
1565 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1566 if (sbi->ckpt->alloc_type[i] == SSR)
1567 valid_sum_count += sbi->blocks_per_seg;
3fa06d7b
CY		 1568 else {
1569 if (for_ra)
1570 valid_sum_count += le16_to_cpu(
1571 F2FS_CKPT(sbi)->cur_data_blkoff[i]);
1572 else
1573 valid_sum_count += curseg_blkoff(sbi, i);
1574 }
351df4b2
JK		 1575 }
1576
09cbfeaf 1577 sum_in_page = (PAGE_SIZE - 2 * SUM_JOURNAL_SIZE -
9a47938b
FL		 1578 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
1579 if (valid_sum_count <= sum_in_page)
351df4b2 1580 return 1;
9a47938b 1581 else if ((valid_sum_count - sum_in_page) <=
09cbfeaf 1582 (PAGE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
351df4b2
JK		 1583 return 2;
1584 return 3;
1585}
1586
0a8165d7 1587/*
351df4b2
JK		 1588 * Caller should put this summary page
1589 */
1590struct page *get_sum_page(struct f2fs_sb_info *sbi, unsigned int segno)
1591{
1592 return get_meta_page(sbi, GET_SUM_BLOCK(sbi, segno));
1593}
1594
381722d2 1595void update_meta_page(struct f2fs_sb_info *sbi, void *src, block_t blk_addr)
351df4b2
JK		 1596{
1597 struct page *page = grab_meta_page(sbi, blk_addr);
381722d2
CY		 1598 void *dst = page_address(page);
1599
1600 if (src)
09cbfeaf 1601 memcpy(dst, src, PAGE_SIZE);
381722d2 1602 else
09cbfeaf 1603 memset(dst, 0, PAGE_SIZE);
351df4b2
JK		 1604 set_page_dirty(page);
1605 f2fs_put_page(page, 1);
1606}
1607
381722d2
CY		 1608static void write_sum_page(struct f2fs_sb_info *sbi,
1609 struct f2fs_summary_block *sum_blk, block_t blk_addr)
1610{
1611 update_meta_page(sbi, (void *)sum_blk, blk_addr);
1612}
1613
b7ad7512
CY		 1614static void write_current_sum_page(struct f2fs_sb_info *sbi,
1615 int type, block_t blk_addr)
1616{
1617 struct curseg_info *curseg = CURSEG_I(sbi, type);
1618 struct page *page = grab_meta_page(sbi, blk_addr);
1619 struct f2fs_summary_block *src = curseg->sum_blk;
1620 struct f2fs_summary_block *dst;
1621
1622 dst = (struct f2fs_summary_block *)page_address(page);
1623
1624 mutex_lock(&curseg->curseg_mutex);
1625
1626 down_read(&curseg->journal_rwsem);
1627 memcpy(&dst->journal, curseg->journal, SUM_JOURNAL_SIZE);
1628 up_read(&curseg->journal_rwsem);
1629
1630 memcpy(dst->entries, src->entries, SUM_ENTRY_SIZE);
1631 memcpy(&dst->footer, &src->footer, SUM_FOOTER_SIZE);
1632
1633 mutex_unlock(&curseg->curseg_mutex);
1634
1635 set_page_dirty(page);
1636 f2fs_put_page(page, 1);
1637}
1638
a7881893
JK		 1639static int is_next_segment_free(struct f2fs_sb_info *sbi, int type)
1640{
1641 struct curseg_info *curseg = CURSEG_I(sbi, type);
1642 unsigned int segno = curseg->segno + 1;
1643 struct free_segmap_info *free_i = FREE_I(sbi);
1644
1645 if (segno < MAIN_SEGS(sbi) && segno % sbi->segs_per_sec)
1646 return !test_bit(segno, free_i->free_segmap);
1647 return 0;
1648}
1649
0a8165d7 1650/*
351df4b2
JK		 1651 * Find a new segment from the free segments bitmap to right order
1652 * This function should be returned with success, otherwise BUG
1653 */
1654static void get_new_segment(struct f2fs_sb_info *sbi,
1655 unsigned int *newseg, bool new_sec, int dir)
1656{
1657 struct free_segmap_info *free_i = FREE_I(sbi);
351df4b2 1658 unsigned int segno, secno, zoneno;
7cd8558b 1659 unsigned int total_zones = MAIN_SECS(sbi) / sbi->secs_per_zone;
4ddb1a4d
JK		 1660 unsigned int hint = GET_SEC_FROM_SEG(sbi, *newseg);
1661 unsigned int old_zoneno = GET_ZONE_FROM_SEG(sbi, *newseg);
351df4b2
JK		 1662 unsigned int left_start = hint;
1663 bool init = true;
1664 int go_left = 0;
1665 int i;
1666
1a118ccf 1667 spin_lock(&free_i->segmap_lock);
351df4b2
JK		 1668
1669 if (!new_sec && ((*newseg + 1) % sbi->segs_per_sec)) {
1670 segno = find_next_zero_bit(free_i->free_segmap,
4ddb1a4d
JK		 1671 GET_SEG_FROM_SEC(sbi, hint + 1), *newseg + 1);
1672 if (segno < GET_SEG_FROM_SEC(sbi, hint + 1))
351df4b2
JK		 1673 goto got_it;
1674 }
1675find_other_zone:
7cd8558b
JK		 1676 secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
1677 if (secno >= MAIN_SECS(sbi)) {
351df4b2
JK		 1678 if (dir == ALLOC_RIGHT) {
1679 secno = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1680 MAIN_SECS(sbi), 0);
1681 f2fs_bug_on(sbi, secno >= MAIN_SECS(sbi));
351df4b2
JK		 1682 } else {
1683 go_left = 1;
1684 left_start = hint - 1;
1685 }
1686 }
1687 if (go_left == 0)
1688 goto skip_left;
1689
1690 while (test_bit(left_start, free_i->free_secmap)) {
1691 if (left_start > 0) {
1692 left_start--;
1693 continue;
1694 }
1695 left_start = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1696 MAIN_SECS(sbi), 0);
1697 f2fs_bug_on(sbi, left_start >= MAIN_SECS(sbi));
351df4b2
JK		 1698 break;
1699 }
1700 secno = left_start;
1701skip_left:
1702 hint = secno;
4ddb1a4d
JK		 1703 segno = GET_SEG_FROM_SEC(sbi, secno);
1704 zoneno = GET_ZONE_FROM_SEC(sbi, secno);
351df4b2
JK		 1705
1706 /* give up on finding another zone */
1707 if (!init)
1708 goto got_it;
1709 if (sbi->secs_per_zone == 1)
1710 goto got_it;
1711 if (zoneno == old_zoneno)
1712 goto got_it;
1713 if (dir == ALLOC_LEFT) {
1714 if (!go_left && zoneno + 1 >= total_zones)
1715 goto got_it;
1716 if (go_left && zoneno == 0)
1717 goto got_it;
1718 }
1719 for (i = 0; i < NR_CURSEG_TYPE; i++)
1720 if (CURSEG_I(sbi, i)->zone == zoneno)
1721 break;
1722
1723 if (i < NR_CURSEG_TYPE) {
1724 /* zone is in user, try another */
1725 if (go_left)
1726 hint = zoneno * sbi->secs_per_zone - 1;
1727 else if (zoneno + 1 >= total_zones)
1728 hint = 0;
1729 else
1730 hint = (zoneno + 1) * sbi->secs_per_zone;
1731 init = false;
1732 goto find_other_zone;
1733 }
1734got_it:
1735 /* set it as dirty segment in free segmap */
9850cf4a 1736 f2fs_bug_on(sbi, test_bit(segno, free_i->free_segmap));
351df4b2
JK		 1737 __set_inuse(sbi, segno);
1738 *newseg = segno;
1a118ccf 1739 spin_unlock(&free_i->segmap_lock);
351df4b2
JK		 1740}
1741
1742static void reset_curseg(struct f2fs_sb_info *sbi, int type, int modified)
1743{
1744 struct curseg_info *curseg = CURSEG_I(sbi, type);
1745 struct summary_footer *sum_footer;
1746
1747 curseg->segno = curseg->next_segno;
4ddb1a4d 1748 curseg->zone = GET_ZONE_FROM_SEG(sbi, curseg->segno);
351df4b2
JK		 1749 curseg->next_blkoff = 0;
1750 curseg->next_segno = NULL_SEGNO;
1751
1752 sum_footer = &(curseg->sum_blk->footer);
1753 memset(sum_footer, 0, sizeof(struct summary_footer));
1754 if (IS_DATASEG(type))
1755 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
1756 if (IS_NODESEG(type))
1757 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
1758 __set_sit_entry_type(sbi, type, curseg->segno, modified);
1759}
1760
7a20b8a6
JK		 1761static unsigned int __get_next_segno(struct f2fs_sb_info *sbi, int type)
1762{
a7881893
JK		 1763 /* if segs_per_sec is large than 1, we need to keep original policy. */
1764 if (sbi->segs_per_sec != 1)
1765 return CURSEG_I(sbi, type)->segno;
1766
7a20b8a6
JK		 1767 if (type == CURSEG_HOT_DATA || IS_NODESEG(type))
1768 return 0;
1769
e066b83c
JK		 1770 if (SIT_I(sbi)->last_victim[ALLOC_NEXT])
1771 return SIT_I(sbi)->last_victim[ALLOC_NEXT];
7a20b8a6
JK		 1772 return CURSEG_I(sbi, type)->segno;
1773}
1774
0a8165d7 1775/*
351df4b2
JK		 1776 * Allocate a current working segment.
1777 * This function always allocates a free segment in LFS manner.
1778 */
1779static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
1780{
1781 struct curseg_info *curseg = CURSEG_I(sbi, type);
1782 unsigned int segno = curseg->segno;
1783 int dir = ALLOC_LEFT;
1784
1785 write_sum_page(sbi, curseg->sum_blk,
81fb5e87 1786 GET_SUM_BLOCK(sbi, segno));
351df4b2
JK		 1787 if (type == CURSEG_WARM_DATA || type == CURSEG_COLD_DATA)
1788 dir = ALLOC_RIGHT;
1789
1790 if (test_opt(sbi, NOHEAP))
1791 dir = ALLOC_RIGHT;
1792
7a20b8a6 1793 segno = __get_next_segno(sbi, type);
351df4b2
JK		 1794 get_new_segment(sbi, &segno, new_sec, dir);
1795 curseg->next_segno = segno;
1796 reset_curseg(sbi, type, 1);
1797 curseg->alloc_type = LFS;
1798}
1799
1800static void __next_free_blkoff(struct f2fs_sb_info *sbi,
1801 struct curseg_info *seg, block_t start)
1802{
1803 struct seg_entry *se = get_seg_entry(sbi, seg->segno);
e81c93cf 1804 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
60a3b782 1805 unsigned long *target_map = SIT_I(sbi)->tmp_map;
e81c93cf
CL		 1806 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
1807 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1808 int i, pos;
1809
1810 for (i = 0; i < entries; i++)
1811 target_map[i] = ckpt_map[i] | cur_map[i];
1812
1813 pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, start);
1814
1815 seg->next_blkoff = pos;
351df4b2
JK		 1816}
1817
0a8165d7 1818/*
351df4b2
JK		 1819 * If a segment is written by LFS manner, next block offset is just obtained
1820 * by increasing the current block offset. However, if a segment is written by
1821 * SSR manner, next block offset obtained by calling __next_free_blkoff
1822 */
1823static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
1824 struct curseg_info *seg)
1825{
1826 if (seg->alloc_type == SSR)
1827 __next_free_blkoff(sbi, seg, seg->next_blkoff + 1);
1828 else
1829 seg->next_blkoff++;
1830}
1831
0a8165d7 1832/*
e1c42045 1833 * This function always allocates a used segment(from dirty seglist) by SSR
351df4b2
JK		 1834 * manner, so it should recover the existing segment information of valid blocks
1835 */
1836static void change_curseg(struct f2fs_sb_info *sbi, int type, bool reuse)
1837{
1838 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
1839 struct curseg_info *curseg = CURSEG_I(sbi, type);
1840 unsigned int new_segno = curseg->next_segno;
1841 struct f2fs_summary_block *sum_node;
1842 struct page *sum_page;
1843
1844 write_sum_page(sbi, curseg->sum_blk,
1845 GET_SUM_BLOCK(sbi, curseg->segno));
1846 __set_test_and_inuse(sbi, new_segno);
1847
1848 mutex_lock(&dirty_i->seglist_lock);
1849 __remove_dirty_segment(sbi, new_segno, PRE);
1850 __remove_dirty_segment(sbi, new_segno, DIRTY);
1851 mutex_unlock(&dirty_i->seglist_lock);
1852
1853 reset_curseg(sbi, type, 1);
1854 curseg->alloc_type = SSR;
1855 __next_free_blkoff(sbi, curseg, 0);
1856
1857 if (reuse) {
1858 sum_page = get_sum_page(sbi, new_segno);
1859 sum_node = (struct f2fs_summary_block *)page_address(sum_page);
1860 memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
1861 f2fs_put_page(sum_page, 1);
1862 }
1863}
1864
43727527
JK		 1865static int get_ssr_segment(struct f2fs_sb_info *sbi, int type)
1866{
1867 struct curseg_info *curseg = CURSEG_I(sbi, type);
1868 const struct victim_selection *v_ops = DIRTY_I(sbi)->v_ops;
e066b83c 1869 unsigned segno = NULL_SEGNO;
d27c3d89
CY		 1870 int i, cnt;
1871 bool reversed = false;
c192f7a4
JK		 1872
1873 /* need_SSR() already forces to do this */
e066b83c
JK		 1874 if (v_ops->get_victim(sbi, &segno, BG_GC, type, SSR)) {
1875 curseg->next_segno = segno;
c192f7a4 1876 return 1;
e066b83c 1877 }
43727527 1878
70d625cb
JK		 1879 /* For node segments, let's do SSR more intensively */
1880 if (IS_NODESEG(type)) {
d27c3d89
CY		 1881 if (type >= CURSEG_WARM_NODE) {
1882 reversed = true;
1883 i = CURSEG_COLD_NODE;
1884 } else {
1885 i = CURSEG_HOT_NODE;
1886 }
1887 cnt = NR_CURSEG_NODE_TYPE;
70d625cb 1888 } else {
d27c3d89
CY		 1889 if (type >= CURSEG_WARM_DATA) {
1890 reversed = true;
1891 i = CURSEG_COLD_DATA;
1892 } else {
1893 i = CURSEG_HOT_DATA;
1894 }
1895 cnt = NR_CURSEG_DATA_TYPE;
70d625cb 1896 }
43727527 1897
d27c3d89 1898 for (; cnt-- > 0; reversed ? i-- : i++) {
c192f7a4
JK		 1899 if (i == type)
1900 continue;
e066b83c
JK		 1901 if (v_ops->get_victim(sbi, &segno, BG_GC, i, SSR)) {
1902 curseg->next_segno = segno;
43727527 1903 return 1;
e066b83c 1904 }
c192f7a4 1905 }
43727527
JK		 1906 return 0;
1907}
1908
351df4b2
JK		 1909/*
1910 * flush out current segment and replace it with new segment
1911 * This function should be returned with success, otherwise BUG
1912 */
1913static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
1914 int type, bool force)
1915{
a7881893
JK		 1916 struct curseg_info *curseg = CURSEG_I(sbi, type);
1917
7b405275 1918 if (force)
351df4b2 1919 new_curseg(sbi, type, true);
5b6c6be2
JK		 1920 else if (!is_set_ckpt_flags(sbi, CP_CRC_RECOVERY_FLAG) &&
1921 type == CURSEG_WARM_NODE)
351df4b2 1922 new_curseg(sbi, type, false);
a7881893
JK		 1923 else if (curseg->alloc_type == LFS && is_next_segment_free(sbi, type))
1924 new_curseg(sbi, type, false);
351df4b2
JK		 1925 else if (need_SSR(sbi) && get_ssr_segment(sbi, type))
1926 change_curseg(sbi, type, true);
1927 else
1928 new_curseg(sbi, type, false);
dcdfff65 1929
a7881893 1930 stat_inc_seg_type(sbi, curseg);
351df4b2
JK		 1931}
1932
1933void allocate_new_segments(struct f2fs_sb_info *sbi)
1934{
6ae1be13
JK		 1935 struct curseg_info *curseg;
1936 unsigned int old_segno;
351df4b2
JK		 1937 int i;
1938
6ae1be13
JK		 1939 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1940 curseg = CURSEG_I(sbi, i);
1941 old_segno = curseg->segno;
1942 SIT_I(sbi)->s_ops->allocate_segment(sbi, i, true);
1943 locate_dirty_segment(sbi, old_segno);
1944 }
351df4b2
JK		 1945}
1946
1947static const struct segment_allocation default_salloc_ops = {
1948 .allocate_segment = allocate_segment_by_default,
1949};
1950
25290fa5
JK		 1951bool exist_trim_candidates(struct f2fs_sb_info *sbi, struct cp_control *cpc)
1952{
1953 __u64 trim_start = cpc->trim_start;
1954 bool has_candidate = false;
1955
1956 mutex_lock(&SIT_I(sbi)->sentry_lock);
1957 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++) {
1958 if (add_discard_addrs(sbi, cpc, true)) {
1959 has_candidate = true;
1960 break;
1961 }
1962 }
1963 mutex_unlock(&SIT_I(sbi)->sentry_lock);
1964
1965 cpc->trim_start = trim_start;
1966 return has_candidate;
1967}
1968
4b2fecc8
JK		 1969int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
1970{
f7ef9b83
JK		 1971 __u64 start = F2FS_BYTES_TO_BLK(range->start);
1972 __u64 end = start + F2FS_BYTES_TO_BLK(range->len) - 1;
4b2fecc8
JK		 1973 unsigned int start_segno, end_segno;
1974 struct cp_control cpc;
c34f42e2 1975 int err = 0;
4b2fecc8 1976
836b5a63 1977 if (start >= MAX_BLKADDR(sbi) || range->len < sbi->blocksize)
4b2fecc8
JK		 1978 return -EINVAL;
1979
9bd27ae4 1980 cpc.trimmed = 0;
7cd8558b 1981 if (end <= MAIN_BLKADDR(sbi))
4b2fecc8
JK		 1982 goto out;
1983
ed214a11
YH		 1984 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
1985 f2fs_msg(sbi->sb, KERN_WARNING,
1986 "Found FS corruption, run fsck to fix.");
1987 goto out;
1988 }
1989
4b2fecc8 1990 /* start/end segment number in main_area */
7cd8558b
JK		 1991 start_segno = (start <= MAIN_BLKADDR(sbi)) ? 0 : GET_SEGNO(sbi, start);
1992 end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
1993 GET_SEGNO(sbi, end);
4b2fecc8 1994 cpc.reason = CP_DISCARD;
836b5a63 1995 cpc.trim_minlen = max_t(__u64, 1, F2FS_BYTES_TO_BLK(range->minlen));
4b2fecc8
JK		 1996
1997 /* do checkpoint to issue discard commands safely */
bba681cb
JK		 1998 for (; start_segno <= end_segno; start_segno = cpc.trim_end + 1) {
1999 cpc.trim_start = start_segno;
a66cdd98
JK		 2000
2001 if (sbi->discard_blks == 0)
2002 break;
2003 else if (sbi->discard_blks < BATCHED_TRIM_BLOCKS(sbi))
2004 cpc.trim_end = end_segno;
2005 else
2006 cpc.trim_end = min_t(unsigned int,
2007 rounddown(start_segno +
bba681cb
JK		 2008 BATCHED_TRIM_SEGMENTS(sbi),
2009 sbi->segs_per_sec) - 1, end_segno);
2010
2011 mutex_lock(&sbi->gc_mutex);
c34f42e2 2012 err = write_checkpoint(sbi, &cpc);
bba681cb 2013 mutex_unlock(&sbi->gc_mutex);
e9328353
CY		 2014 if (err)
2015 break;
74fa5f3d
CY		 2016
2017 schedule();
bba681cb 2018 }
4b2fecc8 2019out:
f7ef9b83 2020 range->len = F2FS_BLK_TO_BYTES(cpc.trimmed);
c34f42e2 2021 return err;
4b2fecc8
JK		 2022}
2023
351df4b2
JK		 2024static bool __has_curseg_space(struct f2fs_sb_info *sbi, int type)
2025{
2026 struct curseg_info *curseg = CURSEG_I(sbi, type);
2027 if (curseg->next_blkoff < sbi->blocks_per_seg)
2028 return true;
2029 return false;
2030}
2031
2032static int __get_segment_type_2(struct page *page, enum page_type p_type)
2033{
2034 if (p_type == DATA)
2035 return CURSEG_HOT_DATA;
2036 else
2037 return CURSEG_HOT_NODE;
2038}
2039
2040static int __get_segment_type_4(struct page *page, enum page_type p_type)
2041{
2042 if (p_type == DATA) {
2043 struct inode *inode = page->mapping->host;
2044
2045 if (S_ISDIR(inode->i_mode))
2046 return CURSEG_HOT_DATA;
2047 else
2048 return CURSEG_COLD_DATA;
2049 } else {
a344b9fd
JK		 2050 if (IS_DNODE(page) && is_cold_node(page))
2051 return CURSEG_WARM_NODE;
351df4b2
JK		 2052 else
2053 return CURSEG_COLD_NODE;
2054 }
2055}
2056
2057static int __get_segment_type_6(struct page *page, enum page_type p_type)
2058{
2059 if (p_type == DATA) {
2060 struct inode *inode = page->mapping->host;
2061
ef095d19 2062 if (is_cold_data(page) || file_is_cold(inode))
351df4b2 2063 return CURSEG_COLD_DATA;
ef095d19
JK		 2064 if (is_inode_flag_set(inode, FI_HOT_DATA))
2065 return CURSEG_HOT_DATA;
2066 return CURSEG_WARM_DATA;
351df4b2
JK		 2067 } else {
2068 if (IS_DNODE(page))
2069 return is_cold_node(page) ? CURSEG_WARM_NODE :
2070 CURSEG_HOT_NODE;
ef095d19 2071 return CURSEG_COLD_NODE;
351df4b2
JK		 2072 }
2073}
2074
2075static int __get_segment_type(struct page *page, enum page_type p_type)
2076{
4081363f 2077 switch (F2FS_P_SB(page)->active_logs) {
351df4b2
JK		 2078 case 2:
2079 return __get_segment_type_2(page, p_type);
2080 case 4:
2081 return __get_segment_type_4(page, p_type);
351df4b2 2082 }
12a67146 2083 /* NR_CURSEG_TYPE(6) logs by default */
9850cf4a
JK		 2084 f2fs_bug_on(F2FS_P_SB(page),
2085 F2FS_P_SB(page)->active_logs != NR_CURSEG_TYPE);
12a67146 2086 return __get_segment_type_6(page, p_type);
351df4b2
JK		 2087}
2088
bfad7c2d
JK		 2089void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
2090 block_t old_blkaddr, block_t *new_blkaddr,
2091 struct f2fs_summary *sum, int type)
351df4b2
JK		 2092{
2093 struct sit_info *sit_i = SIT_I(sbi);
6ae1be13 2094 struct curseg_info *curseg = CURSEG_I(sbi, type);
351df4b2
JK		 2095
2096 mutex_lock(&curseg->curseg_mutex);
21cb1d99 2097 mutex_lock(&sit_i->sentry_lock);
351df4b2
JK		 2098
2099 *new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
351df4b2 2100
4e6a8d9b
JK		 2101 f2fs_wait_discard_bio(sbi, *new_blkaddr);
2102
351df4b2
JK		 2103 /*
2104 * __add_sum_entry should be resided under the curseg_mutex
2105 * because, this function updates a summary entry in the
2106 * current summary block.
2107 */
e79efe3b 2108 __add_sum_entry(sbi, type, sum);
351df4b2 2109
351df4b2 2110 __refresh_next_blkoff(sbi, curseg);
dcdfff65
JK		 2111
2112 stat_inc_block_count(sbi, curseg);
351df4b2 2113
c6f82fe9
JK		 2114 if (!__has_curseg_space(sbi, type))
2115 sit_i->s_ops->allocate_segment(sbi, type, false);
351df4b2 2116 /*
c6f82fe9
JK		 2117 * SIT information should be updated after segment allocation,
2118 * since we need to keep dirty segments precisely under SSR.
351df4b2
JK		 2119 */
2120 refresh_sit_entry(sbi, old_blkaddr, *new_blkaddr);
5e443818 2121
351df4b2
JK		 2122 mutex_unlock(&sit_i->sentry_lock);
2123
bfad7c2d 2124 if (page && IS_NODESEG(type))
351df4b2
JK		 2125 fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
2126
bfad7c2d
JK		 2127 mutex_unlock(&curseg->curseg_mutex);
2128}
2129
05ca3632 2130static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
bfad7c2d 2131{
05ca3632 2132 int type = __get_segment_type(fio->page, fio->type);
0a595eba 2133 int err;
bfad7c2d 2134
7dfeaa32
JK		 2135 if (fio->type == NODE || fio->type == DATA)
2136 mutex_lock(&fio->sbi->wio_mutex[fio->type]);
0a595eba 2137reallocate:
7a9d7548
CY		 2138 allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
2139 &fio->new_blkaddr, sum, type);
bfad7c2d 2140
351df4b2 2141 /* writeout dirty page into bdev */
0a595eba
JK		 2142 err = f2fs_submit_page_mbio(fio);
2143 if (err == -EAGAIN) {
2144 fio->old_blkaddr = fio->new_blkaddr;
2145 goto reallocate;
2146 }
7dfeaa32
JK		 2147
2148 if (fio->type == NODE || fio->type == DATA)
2149 mutex_unlock(&fio->sbi->wio_mutex[fio->type]);
351df4b2
JK		 2150}
2151
577e3495 2152void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
351df4b2 2153{
458e6197 2154 struct f2fs_io_info fio = {
05ca3632 2155 .sbi = sbi,
458e6197 2156 .type = META,
04d328de 2157 .op = REQ_OP_WRITE,
70fd7614 2158 .op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
7a9d7548
CY		 2159 .old_blkaddr = page->index,
2160 .new_blkaddr = page->index,
05ca3632 2161 .page = page,
4375a336 2162 .encrypted_page = NULL,
458e6197
JK		 2163 };
2164
2b947003 2165 if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
04d328de 2166 fio.op_flags &= ~REQ_META;
2b947003 2167
351df4b2 2168 set_page_writeback(page);
05ca3632 2169 f2fs_submit_page_mbio(&fio);
351df4b2
JK		 2170}
2171
05ca3632 2172void write_node_page(unsigned int nid, struct f2fs_io_info *fio)
351df4b2
JK		 2173{
2174 struct f2fs_summary sum;
05ca3632 2175
351df4b2 2176 set_summary(&sum, nid, 0, 0);
05ca3632 2177 do_write_page(&sum, fio);
351df4b2
JK		 2178}
2179
05ca3632 2180void write_data_page(struct dnode_of_data *dn, struct f2fs_io_info *fio)
351df4b2 2181{
05ca3632 2182 struct f2fs_sb_info *sbi = fio->sbi;
351df4b2
JK		 2183 struct f2fs_summary sum;
2184 struct node_info ni;
2185
9850cf4a 2186 f2fs_bug_on(sbi, dn->data_blkaddr == NULL_ADDR);
351df4b2
JK		 2187 get_node_info(sbi, dn->nid, &ni);
2188 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
05ca3632 2189 do_write_page(&sum, fio);
f28b3434 2190 f2fs_update_data_blkaddr(dn, fio->new_blkaddr);
351df4b2
JK		 2191}
2192
d1b3e72d 2193int rewrite_data_page(struct f2fs_io_info *fio)
351df4b2 2194{
7a9d7548 2195 fio->new_blkaddr = fio->old_blkaddr;
05ca3632 2196 stat_inc_inplace_blocks(fio->sbi);
d1b3e72d 2197 return f2fs_submit_page_bio(fio);
351df4b2
JK		 2198}
2199
4356e48e 2200void __f2fs_replace_block(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
19f106bc 2201 block_t old_blkaddr, block_t new_blkaddr,
28bc106b 2202 bool recover_curseg, bool recover_newaddr)
351df4b2
JK		 2203{
2204 struct sit_info *sit_i = SIT_I(sbi);
2205 struct curseg_info *curseg;
2206 unsigned int segno, old_cursegno;
2207 struct seg_entry *se;
2208 int type;
19f106bc 2209 unsigned short old_blkoff;
351df4b2
JK		 2210
2211 segno = GET_SEGNO(sbi, new_blkaddr);
2212 se = get_seg_entry(sbi, segno);
2213 type = se->type;
2214
19f106bc
CY		 2215 if (!recover_curseg) {
2216 /* for recovery flow */
2217 if (se->valid_blocks == 0 && !IS_CURSEG(sbi, segno)) {
2218 if (old_blkaddr == NULL_ADDR)
2219 type = CURSEG_COLD_DATA;
2220 else
2221 type = CURSEG_WARM_DATA;
2222 }
2223 } else {
2224 if (!IS_CURSEG(sbi, segno))
351df4b2
JK		 2225 type = CURSEG_WARM_DATA;
2226 }
19f106bc 2227
351df4b2
JK		 2228 curseg = CURSEG_I(sbi, type);
2229
2230 mutex_lock(&curseg->curseg_mutex);
2231 mutex_lock(&sit_i->sentry_lock);
2232
2233 old_cursegno = curseg->segno;
19f106bc 2234 old_blkoff = curseg->next_blkoff;
351df4b2
JK		 2235
2236 /* change the current segment */
2237 if (segno != curseg->segno) {
2238 curseg->next_segno = segno;
2239 change_curseg(sbi, type, true);
2240 }
2241
491c0854 2242 curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
e79efe3b 2243 __add_sum_entry(sbi, type, sum);
351df4b2 2244
28bc106b 2245 if (!recover_curseg || recover_newaddr)
6e2c64ad
JK		 2246 update_sit_entry(sbi, new_blkaddr, 1);
2247 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
2248 update_sit_entry(sbi, old_blkaddr, -1);
2249
2250 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
2251 locate_dirty_segment(sbi, GET_SEGNO(sbi, new_blkaddr));
2252
351df4b2 2253 locate_dirty_segment(sbi, old_cursegno);
351df4b2 2254
19f106bc
CY		 2255 if (recover_curseg) {
2256 if (old_cursegno != curseg->segno) {
2257 curseg->next_segno = old_cursegno;
2258 change_curseg(sbi, type, true);
2259 }
2260 curseg->next_blkoff = old_blkoff;
2261 }
2262
351df4b2
JK		 2263 mutex_unlock(&sit_i->sentry_lock);
2264 mutex_unlock(&curseg->curseg_mutex);
2265}
2266
528e3459
CY		 2267void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
2268 block_t old_addr, block_t new_addr,
28bc106b
CY		 2269 unsigned char version, bool recover_curseg,
2270 bool recover_newaddr)
528e3459
CY		 2271{
2272 struct f2fs_summary sum;
2273
2274 set_summary(&sum, dn->nid, dn->ofs_in_node, version);
2275
28bc106b
CY		 2276 __f2fs_replace_block(sbi, &sum, old_addr, new_addr,
2277 recover_curseg, recover_newaddr);
528e3459 2278
f28b3434 2279 f2fs_update_data_blkaddr(dn, new_addr);
528e3459
CY		 2280}
2281
93dfe2ac 2282void f2fs_wait_on_page_writeback(struct page *page,
fec1d657 2283 enum page_type type, bool ordered)
93dfe2ac 2284{
93dfe2ac 2285 if (PageWriteback(page)) {
4081363f
JK		 2286 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
2287
942fd319
JK		 2288 f2fs_submit_merged_bio_cond(sbi, page->mapping->host,
2289 0, page->index, type, WRITE);
fec1d657
JK		 2290 if (ordered)
2291 wait_on_page_writeback(page);
2292 else
2293 wait_for_stable_page(page);
93dfe2ac
JK		 2294 }
2295}
2296
08b39fbd
CY		 2297void f2fs_wait_on_encrypted_page_writeback(struct f2fs_sb_info *sbi,
2298 block_t blkaddr)
2299{
2300 struct page *cpage;
2301
5d4c0af4 2302 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
08b39fbd
CY		 2303 return;
2304
08b39fbd
CY		 2305 cpage = find_lock_page(META_MAPPING(sbi), blkaddr);
2306 if (cpage) {
fec1d657 2307 f2fs_wait_on_page_writeback(cpage, DATA, true);
08b39fbd
CY		 2308 f2fs_put_page(cpage, 1);
2309 }
2310}
2311
351df4b2
JK		 2312static int read_compacted_summaries(struct f2fs_sb_info *sbi)
2313{
2314 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2315 struct curseg_info *seg_i;
2316 unsigned char *kaddr;
2317 struct page *page;
2318 block_t start;
2319 int i, j, offset;
2320
2321 start = start_sum_block(sbi);
2322
2323 page = get_meta_page(sbi, start++);
2324 kaddr = (unsigned char *)page_address(page);
2325
2326 /* Step 1: restore nat cache */
2327 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2328 memcpy(seg_i->journal, kaddr, SUM_JOURNAL_SIZE);
351df4b2
JK		 2329
2330 /* Step 2: restore sit cache */
2331 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2332 memcpy(seg_i->journal, kaddr + SUM_JOURNAL_SIZE, SUM_JOURNAL_SIZE);
351df4b2
JK		 2333 offset = 2 * SUM_JOURNAL_SIZE;
2334
2335 /* Step 3: restore summary entries */
2336 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2337 unsigned short blk_off;
2338 unsigned int segno;
2339
2340 seg_i = CURSEG_I(sbi, i);
2341 segno = le32_to_cpu(ckpt->cur_data_segno[i]);
2342 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[i]);
2343 seg_i->next_segno = segno;
2344 reset_curseg(sbi, i, 0);
2345 seg_i->alloc_type = ckpt->alloc_type[i];
2346 seg_i->next_blkoff = blk_off;
2347
2348 if (seg_i->alloc_type == SSR)
2349 blk_off = sbi->blocks_per_seg;
2350
2351 for (j = 0; j < blk_off; j++) {
2352 struct f2fs_summary *s;
2353 s = (struct f2fs_summary *)(kaddr + offset);
2354 seg_i->sum_blk->entries[j] = *s;
2355 offset += SUMMARY_SIZE;
09cbfeaf 2356 if (offset + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2357 SUM_FOOTER_SIZE)
2358 continue;
2359
2360 f2fs_put_page(page, 1);
2361 page = NULL;
2362
2363 page = get_meta_page(sbi, start++);
2364 kaddr = (unsigned char *)page_address(page);
2365 offset = 0;
2366 }
2367 }
2368 f2fs_put_page(page, 1);
2369 return 0;
2370}
2371
2372static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
2373{
2374 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2375 struct f2fs_summary_block *sum;
2376 struct curseg_info *curseg;
2377 struct page *new;
2378 unsigned short blk_off;
2379 unsigned int segno = 0;
2380 block_t blk_addr = 0;
2381
2382 /* get segment number and block addr */
2383 if (IS_DATASEG(type)) {
2384 segno = le32_to_cpu(ckpt->cur_data_segno[type]);
2385 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[type -
2386 CURSEG_HOT_DATA]);
119ee914 2387 if (__exist_node_summaries(sbi))
351df4b2
JK		 2388 blk_addr = sum_blk_addr(sbi, NR_CURSEG_TYPE, type);
2389 else
2390 blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
2391 } else {
2392 segno = le32_to_cpu(ckpt->cur_node_segno[type -
2393 CURSEG_HOT_NODE]);
2394 blk_off = le16_to_cpu(ckpt->cur_node_blkoff[type -
2395 CURSEG_HOT_NODE]);
119ee914 2396 if (__exist_node_summaries(sbi))
351df4b2
JK		 2397 blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
2398 type - CURSEG_HOT_NODE);
2399 else
2400 blk_addr = GET_SUM_BLOCK(sbi, segno);
2401 }
2402
2403 new = get_meta_page(sbi, blk_addr);
2404 sum = (struct f2fs_summary_block *)page_address(new);
2405
2406 if (IS_NODESEG(type)) {
119ee914 2407 if (__exist_node_summaries(sbi)) {
351df4b2
JK		 2408 struct f2fs_summary *ns = &sum->entries[0];
2409 int i;
2410 for (i = 0; i < sbi->blocks_per_seg; i++, ns++) {
2411 ns->version = 0;
2412 ns->ofs_in_node = 0;
2413 }
2414 } else {
d653788a
GZ		 2415 int err;
2416
2417 err = restore_node_summary(sbi, segno, sum);
2418 if (err) {
351df4b2 2419 f2fs_put_page(new, 1);
d653788a 2420 return err;
351df4b2
JK		 2421 }
2422 }
2423 }
2424
2425 /* set uncompleted segment to curseg */
2426 curseg = CURSEG_I(sbi, type);
2427 mutex_lock(&curseg->curseg_mutex);
b7ad7512
CY		 2428
2429 /* update journal info */
2430 down_write(&curseg->journal_rwsem);
2431 memcpy(curseg->journal, &sum->journal, SUM_JOURNAL_SIZE);
2432 up_write(&curseg->journal_rwsem);
2433
2434 memcpy(curseg->sum_blk->entries, sum->entries, SUM_ENTRY_SIZE);
2435 memcpy(&curseg->sum_blk->footer, &sum->footer, SUM_FOOTER_SIZE);
351df4b2
JK		 2436 curseg->next_segno = segno;
2437 reset_curseg(sbi, type, 0);
2438 curseg->alloc_type = ckpt->alloc_type[type];
2439 curseg->next_blkoff = blk_off;
2440 mutex_unlock(&curseg->curseg_mutex);
2441 f2fs_put_page(new, 1);
2442 return 0;
2443}
2444
2445static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
2446{
2447 int type = CURSEG_HOT_DATA;
e4fc5fbf 2448 int err;
351df4b2 2449
aaec2b1d 2450 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG)) {
3fa06d7b
CY		 2451 int npages = npages_for_summary_flush(sbi, true);
2452
2453 if (npages >= 2)
2454 ra_meta_pages(sbi, start_sum_block(sbi), npages,
26879fb1 2455 META_CP, true);
3fa06d7b 2456
351df4b2
JK		 2457 /* restore for compacted data summary */
2458 if (read_compacted_summaries(sbi))
2459 return -EINVAL;
2460 type = CURSEG_HOT_NODE;
2461 }
2462
119ee914 2463 if (__exist_node_summaries(sbi))
3fa06d7b 2464 ra_meta_pages(sbi, sum_blk_addr(sbi, NR_CURSEG_TYPE, type),
26879fb1 2465 NR_CURSEG_TYPE - type, META_CP, true);
3fa06d7b 2466
e4fc5fbf
CY		 2467 for (; type <= CURSEG_COLD_NODE; type++) {
2468 err = read_normal_summaries(sbi, type);
2469 if (err)
2470 return err;
2471 }
2472
351df4b2
JK		 2473 return 0;
2474}
2475
2476static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
2477{
2478 struct page *page;
2479 unsigned char *kaddr;
2480 struct f2fs_summary *summary;
2481 struct curseg_info *seg_i;
2482 int written_size = 0;
2483 int i, j;
2484
2485 page = grab_meta_page(sbi, blkaddr++);
2486 kaddr = (unsigned char *)page_address(page);
2487
2488 /* Step 1: write nat cache */
2489 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2490 memcpy(kaddr, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2491 written_size += SUM_JOURNAL_SIZE;
2492
2493 /* Step 2: write sit cache */
2494 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2495 memcpy(kaddr + written_size, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2496 written_size += SUM_JOURNAL_SIZE;
2497
351df4b2
JK		 2498 /* Step 3: write summary entries */
2499 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2500 unsigned short blkoff;
2501 seg_i = CURSEG_I(sbi, i);
2502 if (sbi->ckpt->alloc_type[i] == SSR)
2503 blkoff = sbi->blocks_per_seg;
2504 else
2505 blkoff = curseg_blkoff(sbi, i);
2506
2507 for (j = 0; j < blkoff; j++) {
2508 if (!page) {
2509 page = grab_meta_page(sbi, blkaddr++);
2510 kaddr = (unsigned char *)page_address(page);
2511 written_size = 0;
2512 }
2513 summary = (struct f2fs_summary *)(kaddr + written_size);
2514 *summary = seg_i->sum_blk->entries[j];
2515 written_size += SUMMARY_SIZE;
351df4b2 2516
09cbfeaf 2517 if (written_size + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2518 SUM_FOOTER_SIZE)
2519 continue;
2520
e8d61a74 2521 set_page_dirty(page);
351df4b2
JK		 2522 f2fs_put_page(page, 1);
2523 page = NULL;
2524 }
2525 }
e8d61a74
CY		 2526 if (page) {
2527 set_page_dirty(page);
351df4b2 2528 f2fs_put_page(page, 1);
e8d61a74 2529 }
351df4b2
JK		 2530}
2531
2532static void write_normal_summaries(struct f2fs_sb_info *sbi,
2533 block_t blkaddr, int type)
2534{
2535 int i, end;
2536 if (IS_DATASEG(type))
2537 end = type + NR_CURSEG_DATA_TYPE;
2538 else
2539 end = type + NR_CURSEG_NODE_TYPE;
2540
b7ad7512
CY		 2541 for (i = type; i < end; i++)
2542 write_current_sum_page(sbi, i, blkaddr + (i - type));
351df4b2
JK		 2543}
2544
2545void write_data_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2546{
aaec2b1d 2547 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG))
351df4b2
JK		 2548 write_compacted_summaries(sbi, start_blk);
2549 else
2550 write_normal_summaries(sbi, start_blk, CURSEG_HOT_DATA);
2551}
2552
2553void write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2554{
119ee914 2555 write_normal_summaries(sbi, start_blk, CURSEG_HOT_NODE);
351df4b2
JK		 2556}
2557
dfc08a12 2558int lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
351df4b2
JK		 2559 unsigned int val, int alloc)
2560{
2561 int i;
2562
2563 if (type == NAT_JOURNAL) {
dfc08a12
CY		 2564 for (i = 0; i < nats_in_cursum(journal); i++) {
2565 if (le32_to_cpu(nid_in_journal(journal, i)) == val)
351df4b2
JK		 2566 return i;
2567 }
dfc08a12
CY		 2568 if (alloc && __has_cursum_space(journal, 1, NAT_JOURNAL))
2569 return update_nats_in_cursum(journal, 1);
351df4b2 2570 } else if (type == SIT_JOURNAL) {
dfc08a12
CY		 2571 for (i = 0; i < sits_in_cursum(journal); i++)
2572 if (le32_to_cpu(segno_in_journal(journal, i)) == val)
351df4b2 2573 return i;
dfc08a12
CY		 2574 if (alloc && __has_cursum_space(journal, 1, SIT_JOURNAL))
2575 return update_sits_in_cursum(journal, 1);
351df4b2
JK		 2576 }
2577 return -1;
2578}
2579
2580static struct page *get_current_sit_page(struct f2fs_sb_info *sbi,
2581 unsigned int segno)
2582{
2cc22186 2583 return get_meta_page(sbi, current_sit_addr(sbi, segno));
351df4b2
JK		 2584}
2585
2586static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
2587 unsigned int start)
2588{
2589 struct sit_info *sit_i = SIT_I(sbi);
2590 struct page *src_page, *dst_page;
2591 pgoff_t src_off, dst_off;
2592 void *src_addr, *dst_addr;
2593
2594 src_off = current_sit_addr(sbi, start);
2595 dst_off = next_sit_addr(sbi, src_off);
2596
2597 /* get current sit block page without lock */
2598 src_page = get_meta_page(sbi, src_off);
2599 dst_page = grab_meta_page(sbi, dst_off);
9850cf4a 2600 f2fs_bug_on(sbi, PageDirty(src_page));
351df4b2
JK		 2601
2602 src_addr = page_address(src_page);
2603 dst_addr = page_address(dst_page);
09cbfeaf 2604 memcpy(dst_addr, src_addr, PAGE_SIZE);
351df4b2
JK		 2605
2606 set_page_dirty(dst_page);
2607 f2fs_put_page(src_page, 1);
2608
2609 set_to_next_sit(sit_i, start);
2610
2611 return dst_page;
2612}
2613
184a5cd2
CY		 2614static struct sit_entry_set *grab_sit_entry_set(void)
2615{
2616 struct sit_entry_set *ses =
80c54505 2617 f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS);
184a5cd2
CY		 2618
2619 ses->entry_cnt = 0;
2620 INIT_LIST_HEAD(&ses->set_list);
2621 return ses;
2622}
2623
2624static void release_sit_entry_set(struct sit_entry_set *ses)
2625{
2626 list_del(&ses->set_list);
2627 kmem_cache_free(sit_entry_set_slab, ses);
2628}
2629
2630static void adjust_sit_entry_set(struct sit_entry_set *ses,
2631 struct list_head *head)
2632{
2633 struct sit_entry_set *next = ses;
2634
2635 if (list_is_last(&ses->set_list, head))
2636 return;
2637
2638 list_for_each_entry_continue(next, head, set_list)
2639 if (ses->entry_cnt <= next->entry_cnt)
2640 break;
2641
2642 list_move_tail(&ses->set_list, &next->set_list);
2643}
2644
2645static void add_sit_entry(unsigned int segno, struct list_head *head)
2646{
2647 struct sit_entry_set *ses;
2648 unsigned int start_segno = START_SEGNO(segno);
2649
2650 list_for_each_entry(ses, head, set_list) {
2651 if (ses->start_segno == start_segno) {
2652 ses->entry_cnt++;
2653 adjust_sit_entry_set(ses, head);
2654 return;
2655 }
2656 }
2657
2658 ses = grab_sit_entry_set();
2659
2660 ses->start_segno = start_segno;
2661 ses->entry_cnt++;
2662 list_add(&ses->set_list, head);
2663}
2664
2665static void add_sits_in_set(struct f2fs_sb_info *sbi)
2666{
2667 struct f2fs_sm_info *sm_info = SM_I(sbi);
2668 struct list_head *set_list = &sm_info->sit_entry_set;
2669 unsigned long *bitmap = SIT_I(sbi)->dirty_sentries_bitmap;
184a5cd2
CY		 2670 unsigned int segno;
2671
7cd8558b 2672 for_each_set_bit(segno, bitmap, MAIN_SEGS(sbi))
184a5cd2
CY		 2673 add_sit_entry(segno, set_list);
2674}
2675
2676static void remove_sits_in_journal(struct f2fs_sb_info *sbi)
351df4b2
JK		 2677{
2678 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2679 struct f2fs_journal *journal = curseg->journal;
351df4b2
JK		 2680 int i;
2681
b7ad7512 2682 down_write(&curseg->journal_rwsem);
dfc08a12 2683 for (i = 0; i < sits_in_cursum(journal); i++) {
184a5cd2
CY		 2684 unsigned int segno;
2685 bool dirtied;
2686
dfc08a12 2687 segno = le32_to_cpu(segno_in_journal(journal, i));
184a5cd2
CY		 2688 dirtied = __mark_sit_entry_dirty(sbi, segno);
2689
2690 if (!dirtied)
2691 add_sit_entry(segno, &SM_I(sbi)->sit_entry_set);
351df4b2 2692 }
dfc08a12 2693 update_sits_in_cursum(journal, -i);
b7ad7512 2694 up_write(&curseg->journal_rwsem);
351df4b2
JK		 2695}
2696
0a8165d7 2697/*
351df4b2
JK		 2698 * CP calls this function, which flushes SIT entries including sit_journal,
2699 * and moves prefree segs to free segs.
2700 */
4b2fecc8 2701void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2
JK		 2702{
2703 struct sit_info *sit_i = SIT_I(sbi);
2704 unsigned long *bitmap = sit_i->dirty_sentries_bitmap;
2705 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2706 struct f2fs_journal *journal = curseg->journal;
184a5cd2
CY		 2707 struct sit_entry_set *ses, *tmp;
2708 struct list_head *head = &SM_I(sbi)->sit_entry_set;
184a5cd2 2709 bool to_journal = true;
4b2fecc8 2710 struct seg_entry *se;
351df4b2 2711
351df4b2
JK		 2712 mutex_lock(&sit_i->sentry_lock);
2713
2b11a74b
WL		 2714 if (!sit_i->dirty_sentries)
2715 goto out;
2716
351df4b2 2717 /*
184a5cd2
CY		 2718 * add and account sit entries of dirty bitmap in sit entry
2719 * set temporarily
351df4b2 2720 */
184a5cd2 2721 add_sits_in_set(sbi);
351df4b2 2722
184a5cd2
CY		 2723 /*
2724 * if there are no enough space in journal to store dirty sit
2725 * entries, remove all entries from journal and add and account
2726 * them in sit entry set.
2727 */
dfc08a12 2728 if (!__has_cursum_space(journal, sit_i->dirty_sentries, SIT_JOURNAL))
184a5cd2 2729 remove_sits_in_journal(sbi);
b2955550 2730
184a5cd2
CY		 2731 /*
2732 * there are two steps to flush sit entries:
2733 * #1, flush sit entries to journal in current cold data summary block.
2734 * #2, flush sit entries to sit page.
2735 */
2736 list_for_each_entry_safe(ses, tmp, head, set_list) {
4a257ed6 2737 struct page *page = NULL;
184a5cd2
CY		 2738 struct f2fs_sit_block *raw_sit = NULL;
2739 unsigned int start_segno = ses->start_segno;
2740 unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
7cd8558b 2741 (unsigned long)MAIN_SEGS(sbi));
184a5cd2
CY		 2742 unsigned int segno = start_segno;
2743
2744 if (to_journal &&
dfc08a12 2745 !__has_cursum_space(journal, ses->entry_cnt, SIT_JOURNAL))
184a5cd2
CY		 2746 to_journal = false;
2747
b7ad7512
CY		 2748 if (to_journal) {
2749 down_write(&curseg->journal_rwsem);
2750 } else {
184a5cd2
CY		 2751 page = get_next_sit_page(sbi, start_segno);
2752 raw_sit = page_address(page);
351df4b2 2753 }
351df4b2 2754
184a5cd2
CY		 2755 /* flush dirty sit entries in region of current sit set */
2756 for_each_set_bit_from(segno, bitmap, end) {
2757 int offset, sit_offset;
4b2fecc8
JK		 2758
2759 se = get_seg_entry(sbi, segno);
184a5cd2
CY		 2760
2761 /* add discard candidates */
d7bc2484 2762 if (cpc->reason != CP_DISCARD) {
4b2fecc8 2763 cpc->trim_start = segno;
25290fa5 2764 add_discard_addrs(sbi, cpc, false);
4b2fecc8 2765 }
184a5cd2
CY		 2766
2767 if (to_journal) {
dfc08a12 2768 offset = lookup_journal_in_cursum(journal,
184a5cd2
CY		 2769 SIT_JOURNAL, segno, 1);
2770 f2fs_bug_on(sbi, offset < 0);
dfc08a12 2771 segno_in_journal(journal, offset) =
184a5cd2
CY		 2772 cpu_to_le32(segno);
2773 seg_info_to_raw_sit(se,
dfc08a12 2774 &sit_in_journal(journal, offset));
184a5cd2
CY		 2775 } else {
2776 sit_offset = SIT_ENTRY_OFFSET(sit_i, segno);
2777 seg_info_to_raw_sit(se,
2778 &raw_sit->entries[sit_offset]);
2779 }
351df4b2 2780
184a5cd2
CY		 2781 __clear_bit(segno, bitmap);
2782 sit_i->dirty_sentries--;
2783 ses->entry_cnt--;
351df4b2
JK		 2784 }
2785
b7ad7512
CY		 2786 if (to_journal)
2787 up_write(&curseg->journal_rwsem);
2788 else
184a5cd2
CY		 2789 f2fs_put_page(page, 1);
2790
2791 f2fs_bug_on(sbi, ses->entry_cnt);
2792 release_sit_entry_set(ses);
351df4b2 2793 }
184a5cd2
CY		 2794
2795 f2fs_bug_on(sbi, !list_empty(head));
2796 f2fs_bug_on(sbi, sit_i->dirty_sentries);
184a5cd2 2797out:
4b2fecc8 2798 if (cpc->reason == CP_DISCARD) {
650d3c4e
YH		 2799 __u64 trim_start = cpc->trim_start;
2800
4b2fecc8 2801 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++)
25290fa5 2802 add_discard_addrs(sbi, cpc, false);
650d3c4e
YH		 2803
2804 cpc->trim_start = trim_start;
4b2fecc8 2805 }
351df4b2 2806 mutex_unlock(&sit_i->sentry_lock);
351df4b2 2807
351df4b2
JK		 2808 set_prefree_as_free_segments(sbi);
2809}
2810
2811static int build_sit_info(struct f2fs_sb_info *sbi)
2812{
2813 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
351df4b2
JK		 2814 struct sit_info *sit_i;
2815 unsigned int sit_segs, start;
ae27d62e 2816 char *src_bitmap;
351df4b2
JK		 2817 unsigned int bitmap_size;
2818
2819 /* allocate memory for SIT information */
2820 sit_i = kzalloc(sizeof(struct sit_info), GFP_KERNEL);
2821 if (!sit_i)
2822 return -ENOMEM;
2823
2824 SM_I(sbi)->sit_info = sit_i;
2825
39307a8e
JK		 2826 sit_i->sentries = f2fs_kvzalloc(MAIN_SEGS(sbi) *
2827 sizeof(struct seg_entry), GFP_KERNEL);
351df4b2
JK		 2828 if (!sit_i->sentries)
2829 return -ENOMEM;
2830
7cd8558b 2831 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2832 sit_i->dirty_sentries_bitmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2833 if (!sit_i->dirty_sentries_bitmap)
2834 return -ENOMEM;
2835
7cd8558b 2836 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 2837 sit_i->sentries[start].cur_valid_map
2838 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2839 sit_i->sentries[start].ckpt_valid_map
2840 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
a66cdd98 2841 if (!sit_i->sentries[start].cur_valid_map ||
3e025740 2842 !sit_i->sentries[start].ckpt_valid_map)
351df4b2 2843 return -ENOMEM;
3e025740 2844
355e7891
CY		 2845#ifdef CONFIG_F2FS_CHECK_FS
2846 sit_i->sentries[start].cur_valid_map_mir
2847 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2848 if (!sit_i->sentries[start].cur_valid_map_mir)
2849 return -ENOMEM;
2850#endif
2851
3e025740
JK		 2852 if (f2fs_discard_en(sbi)) {
2853 sit_i->sentries[start].discard_map
2854 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2855 if (!sit_i->sentries[start].discard_map)
2856 return -ENOMEM;
2857 }
351df4b2
JK		 2858 }
2859
60a3b782
JK		 2860 sit_i->tmp_map = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2861 if (!sit_i->tmp_map)
2862 return -ENOMEM;
2863
351df4b2 2864 if (sbi->segs_per_sec > 1) {
39307a8e
JK		 2865 sit_i->sec_entries = f2fs_kvzalloc(MAIN_SECS(sbi) *
2866 sizeof(struct sec_entry), GFP_KERNEL);
351df4b2
JK		 2867 if (!sit_i->sec_entries)
2868 return -ENOMEM;
2869 }
2870
2871 /* get information related with SIT */
2872 sit_segs = le32_to_cpu(raw_super->segment_count_sit) >> 1;
2873
2874 /* setup SIT bitmap from ckeckpoint pack */
2875 bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
2876 src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
2877
ae27d62e
CY		 2878 sit_i->sit_bitmap = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2879 if (!sit_i->sit_bitmap)
351df4b2 2880 return -ENOMEM;
351df4b2 2881
ae27d62e
CY		 2882#ifdef CONFIG_F2FS_CHECK_FS
2883 sit_i->sit_bitmap_mir = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2884 if (!sit_i->sit_bitmap_mir)
2885 return -ENOMEM;
2886#endif
2887
351df4b2
JK		 2888 /* init SIT information */
2889 sit_i->s_ops = &default_salloc_ops;
2890
2891 sit_i->sit_base_addr = le32_to_cpu(raw_super->sit_blkaddr);
2892 sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
c79b7ff1 2893 sit_i->written_valid_blocks = 0;
351df4b2
JK		 2894 sit_i->bitmap_size = bitmap_size;
2895 sit_i->dirty_sentries = 0;
2896 sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
2897 sit_i->elapsed_time = le64_to_cpu(sbi->ckpt->elapsed_time);
2898 sit_i->mounted_time = CURRENT_TIME_SEC.tv_sec;
2899 mutex_init(&sit_i->sentry_lock);
2900 return 0;
2901}
2902
2903static int build_free_segmap(struct f2fs_sb_info *sbi)
2904{
351df4b2
JK		 2905 struct free_segmap_info *free_i;
2906 unsigned int bitmap_size, sec_bitmap_size;
2907
2908 /* allocate memory for free segmap information */
2909 free_i = kzalloc(sizeof(struct free_segmap_info), GFP_KERNEL);
2910 if (!free_i)
2911 return -ENOMEM;
2912
2913 SM_I(sbi)->free_info = free_i;
2914
7cd8558b 2915 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2916 free_i->free_segmap = f2fs_kvmalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2917 if (!free_i->free_segmap)
2918 return -ENOMEM;
2919
7cd8558b 2920 sec_bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
39307a8e 2921 free_i->free_secmap = f2fs_kvmalloc(sec_bitmap_size, GFP_KERNEL);
351df4b2
JK		 2922 if (!free_i->free_secmap)
2923 return -ENOMEM;
2924
2925 /* set all segments as dirty temporarily */
2926 memset(free_i->free_segmap, 0xff, bitmap_size);
2927 memset(free_i->free_secmap, 0xff, sec_bitmap_size);
2928
2929 /* init free segmap information */
7cd8558b 2930 free_i->start_segno = GET_SEGNO_FROM_SEG0(sbi, MAIN_BLKADDR(sbi));
351df4b2
JK		 2931 free_i->free_segments = 0;
2932 free_i->free_sections = 0;
1a118ccf 2933 spin_lock_init(&free_i->segmap_lock);
351df4b2
JK		 2934 return 0;
2935}
2936
2937static int build_curseg(struct f2fs_sb_info *sbi)
2938{
1042d60f 2939 struct curseg_info *array;
351df4b2
JK		 2940 int i;
2941
b434babf 2942 array = kcalloc(NR_CURSEG_TYPE, sizeof(*array), GFP_KERNEL);
351df4b2
JK		 2943 if (!array)
2944 return -ENOMEM;
2945
2946 SM_I(sbi)->curseg_array = array;
2947
2948 for (i = 0; i < NR_CURSEG_TYPE; i++) {
2949 mutex_init(&array[i].curseg_mutex);
09cbfeaf 2950 array[i].sum_blk = kzalloc(PAGE_SIZE, GFP_KERNEL);
351df4b2
JK		 2951 if (!array[i].sum_blk)
2952 return -ENOMEM;
b7ad7512
CY		 2953 init_rwsem(&array[i].journal_rwsem);
2954 array[i].journal = kzalloc(sizeof(struct f2fs_journal),
2955 GFP_KERNEL);
2956 if (!array[i].journal)
2957 return -ENOMEM;
351df4b2
JK		 2958 array[i].segno = NULL_SEGNO;
2959 array[i].next_blkoff = 0;
2960 }
2961 return restore_curseg_summaries(sbi);
2962}
2963
2964static void build_sit_entries(struct f2fs_sb_info *sbi)
2965{
2966 struct sit_info *sit_i = SIT_I(sbi);
2967 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2968 struct f2fs_journal *journal = curseg->journal;
9c094040
YH		 2969 struct seg_entry *se;
2970 struct f2fs_sit_entry sit;
74de593a
CY		 2971 int sit_blk_cnt = SIT_BLK_CNT(sbi);
2972 unsigned int i, start, end;
2973 unsigned int readed, start_blk = 0;
351df4b2 2974
74de593a 2975 do {
664ba972
JK		 2976 readed = ra_meta_pages(sbi, start_blk, BIO_MAX_PAGES,
2977 META_SIT, true);
74de593a
CY		 2978
2979 start = start_blk * sit_i->sents_per_block;
2980 end = (start_blk + readed) * sit_i->sents_per_block;
2981
7cd8558b 2982 for (; start < end && start < MAIN_SEGS(sbi); start++) {
74de593a 2983 struct f2fs_sit_block *sit_blk;
74de593a
CY		 2984 struct page *page;
2985
9c094040 2986 se = &sit_i->sentries[start];
74de593a
CY		 2987 page = get_current_sit_page(sbi, start);
2988 sit_blk = (struct f2fs_sit_block *)page_address(page);
2989 sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
2990 f2fs_put_page(page, 1);
d600af23 2991
74de593a
CY		 2992 check_block_count(sbi, start, &sit);
2993 seg_info_from_raw_sit(se, &sit);
a66cdd98
JK		 2994
2995 /* build discard map only one time */
3e025740
JK		 2996 if (f2fs_discard_en(sbi)) {
2997 memcpy(se->discard_map, se->cur_valid_map,
2998 SIT_VBLOCK_MAP_SIZE);
2999 sbi->discard_blks += sbi->blocks_per_seg -
3000 se->valid_blocks;
3001 }
a66cdd98 3002
d600af23
CY		 3003 if (sbi->segs_per_sec > 1)
3004 get_sec_entry(sbi, start)->valid_blocks +=
3005 se->valid_blocks;
351df4b2 3006 }
74de593a
CY		 3007 start_blk += readed;
3008 } while (start_blk < sit_blk_cnt);
d600af23
CY		 3009
3010 down_read(&curseg->journal_rwsem);
3011 for (i = 0; i < sits_in_cursum(journal); i++) {
d600af23
CY		 3012 unsigned int old_valid_blocks;
3013
3014 start = le32_to_cpu(segno_in_journal(journal, i));
3015 se = &sit_i->sentries[start];
3016 sit = sit_in_journal(journal, i);
3017
3018 old_valid_blocks = se->valid_blocks;
3019
3020 check_block_count(sbi, start, &sit);
3021 seg_info_from_raw_sit(se, &sit);
3022
3023 if (f2fs_discard_en(sbi)) {
3024 memcpy(se->discard_map, se->cur_valid_map,
3025 SIT_VBLOCK_MAP_SIZE);
3026 sbi->discard_blks += old_valid_blocks -
3027 se->valid_blocks;
3028 }
3029
3030 if (sbi->segs_per_sec > 1)
3031 get_sec_entry(sbi, start)->valid_blocks +=
3032 se->valid_blocks - old_valid_blocks;
3033 }
3034 up_read(&curseg->journal_rwsem);
351df4b2
JK		 3035}
3036
3037static void init_free_segmap(struct f2fs_sb_info *sbi)
3038{
3039 unsigned int start;
3040 int type;
3041
7cd8558b 3042 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 3043 struct seg_entry *sentry = get_seg_entry(sbi, start);
3044 if (!sentry->valid_blocks)
3045 __set_free(sbi, start);
c79b7ff1
JK		 3046 else
3047 SIT_I(sbi)->written_valid_blocks +=
3048 sentry->valid_blocks;
351df4b2
JK		 3049 }
3050
3051 /* set use the current segments */
3052 for (type = CURSEG_HOT_DATA; type <= CURSEG_COLD_NODE; type++) {
3053 struct curseg_info *curseg_t = CURSEG_I(sbi, type);
3054 __set_test_and_inuse(sbi, curseg_t->segno);
3055 }
3056}
3057
3058static void init_dirty_segmap(struct f2fs_sb_info *sbi)
3059{
3060 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3061 struct free_segmap_info *free_i = FREE_I(sbi);
7cd8558b 3062 unsigned int segno = 0, offset = 0;
351df4b2
JK		 3063 unsigned short valid_blocks;
3064
8736fbf0 3065 while (1) {
351df4b2 3066 /* find dirty segment based on free segmap */
7cd8558b
JK		 3067 segno = find_next_inuse(free_i, MAIN_SEGS(sbi), offset);
3068 if (segno >= MAIN_SEGS(sbi))
351df4b2
JK		 3069 break;
3070 offset = segno + 1;
302bd348 3071 valid_blocks = get_valid_blocks(sbi, segno, false);
ec325b52 3072 if (valid_blocks == sbi->blocks_per_seg || !valid_blocks)
351df4b2 3073 continue;
ec325b52
JK		 3074 if (valid_blocks > sbi->blocks_per_seg) {
3075 f2fs_bug_on(sbi, 1);
3076 continue;
3077 }
351df4b2
JK		 3078 mutex_lock(&dirty_i->seglist_lock);
3079 __locate_dirty_segment(sbi, segno, DIRTY);
3080 mutex_unlock(&dirty_i->seglist_lock);
3081 }
3082}
3083
5ec4e49f 3084static int init_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3085{
3086 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
7cd8558b 3087 unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
351df4b2 3088
39307a8e 3089 dirty_i->victim_secmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
5ec4e49f 3090 if (!dirty_i->victim_secmap)
351df4b2
JK		 3091 return -ENOMEM;
3092 return 0;
3093}
3094
3095static int build_dirty_segmap(struct f2fs_sb_info *sbi)
3096{
3097 struct dirty_seglist_info *dirty_i;
3098 unsigned int bitmap_size, i;
3099
3100 /* allocate memory for dirty segments list information */
3101 dirty_i = kzalloc(sizeof(struct dirty_seglist_info), GFP_KERNEL);
3102 if (!dirty_i)
3103 return -ENOMEM;
3104
3105 SM_I(sbi)->dirty_info = dirty_i;
3106 mutex_init(&dirty_i->seglist_lock);
3107
7cd8558b 3108 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
351df4b2
JK		 3109
3110 for (i = 0; i < NR_DIRTY_TYPE; i++) {
39307a8e 3111 dirty_i->dirty_segmap[i] = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 3112 if (!dirty_i->dirty_segmap[i])
3113 return -ENOMEM;
3114 }
3115
3116 init_dirty_segmap(sbi);
5ec4e49f 3117 return init_victim_secmap(sbi);
351df4b2
JK		 3118}
3119
0a8165d7 3120/*
351df4b2
JK		 3121 * Update min, max modified time for cost-benefit GC algorithm
3122 */
3123static void init_min_max_mtime(struct f2fs_sb_info *sbi)
3124{
3125 struct sit_info *sit_i = SIT_I(sbi);
3126 unsigned int segno;
3127
3128 mutex_lock(&sit_i->sentry_lock);
3129
3130 sit_i->min_mtime = LLONG_MAX;
3131
7cd8558b 3132 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
351df4b2
JK		 3133 unsigned int i;
3134 unsigned long long mtime = 0;
3135
3136 for (i = 0; i < sbi->segs_per_sec; i++)
3137 mtime += get_seg_entry(sbi, segno + i)->mtime;
3138
3139 mtime = div_u64(mtime, sbi->segs_per_sec);
3140
3141 if (sit_i->min_mtime > mtime)
3142 sit_i->min_mtime = mtime;
3143 }
3144 sit_i->max_mtime = get_mtime(sbi);
3145 mutex_unlock(&sit_i->sentry_lock);
3146}
3147
3148int build_segment_manager(struct f2fs_sb_info *sbi)
3149{
3150 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3151 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1042d60f 3152 struct f2fs_sm_info *sm_info;
351df4b2
JK		 3153 int err;
3154
3155 sm_info = kzalloc(sizeof(struct f2fs_sm_info), GFP_KERNEL);
3156 if (!sm_info)
3157 return -ENOMEM;
3158
3159 /* init sm info */
3160 sbi->sm_info = sm_info;
351df4b2
JK		 3161 sm_info->seg0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3162 sm_info->main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3163 sm_info->segment_count = le32_to_cpu(raw_super->segment_count);
3164 sm_info->reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3165 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3166 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
3167 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
58c41035
JK		 3168 sm_info->rec_prefree_segments = sm_info->main_segments *
3169 DEF_RECLAIM_PREFREE_SEGMENTS / 100;
44a83499
JK		 3170 if (sm_info->rec_prefree_segments > DEF_MAX_RECLAIM_PREFREE_SEGMENTS)
3171 sm_info->rec_prefree_segments = DEF_MAX_RECLAIM_PREFREE_SEGMENTS;
3172
52763a4b
JK		 3173 if (!test_opt(sbi, LFS))
3174 sm_info->ipu_policy = 1 << F2FS_IPU_FSYNC;
216fbd64 3175 sm_info->min_ipu_util = DEF_MIN_IPU_UTIL;
c1ce1b02 3176 sm_info->min_fsync_blocks = DEF_MIN_FSYNC_BLOCKS;
ef095d19 3177 sm_info->min_hot_blocks = DEF_MIN_HOT_BLOCKS;
351df4b2 3178
bba681cb
JK		 3179 sm_info->trim_sections = DEF_BATCHED_TRIM_SECTIONS;
3180
184a5cd2
CY		 3181 INIT_LIST_HEAD(&sm_info->sit_entry_set);
3182
b270ad6f 3183 if (test_opt(sbi, FLUSH_MERGE) && !f2fs_readonly(sbi->sb)) {
2163d198
GZ		 3184 err = create_flush_cmd_control(sbi);
3185 if (err)
a688b9d9 3186 return err;
6b4afdd7
JK		 3187 }
3188
0b54fb84
JK		 3189 err = create_discard_cmd_control(sbi);
3190 if (err)
3191 return err;
3192
351df4b2
JK		 3193 err = build_sit_info(sbi);
3194 if (err)
3195 return err;
3196 err = build_free_segmap(sbi);
3197 if (err)
3198 return err;
3199 err = build_curseg(sbi);
3200 if (err)
3201 return err;
3202
3203 /* reinit free segmap based on SIT */
3204 build_sit_entries(sbi);
3205
3206 init_free_segmap(sbi);
3207 err = build_dirty_segmap(sbi);
3208 if (err)
3209 return err;
3210
3211 init_min_max_mtime(sbi);
3212 return 0;
3213}
3214
3215static void discard_dirty_segmap(struct f2fs_sb_info *sbi,
3216 enum dirty_type dirty_type)
3217{
3218 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3219
3220 mutex_lock(&dirty_i->seglist_lock);
39307a8e 3221 kvfree(dirty_i->dirty_segmap[dirty_type]);
351df4b2
JK		 3222 dirty_i->nr_dirty[dirty_type] = 0;
3223 mutex_unlock(&dirty_i->seglist_lock);
3224}
3225
5ec4e49f 3226static void destroy_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3227{
3228 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
39307a8e 3229 kvfree(dirty_i->victim_secmap);
351df4b2
JK		 3230}
3231
3232static void destroy_dirty_segmap(struct f2fs_sb_info *sbi)
3233{
3234 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3235 int i;
3236
3237 if (!dirty_i)
3238 return;
3239
3240 /* discard pre-free/dirty segments list */
3241 for (i = 0; i < NR_DIRTY_TYPE; i++)
3242 discard_dirty_segmap(sbi, i);
3243
5ec4e49f 3244 destroy_victim_secmap(sbi);
351df4b2
JK		 3245 SM_I(sbi)->dirty_info = NULL;
3246 kfree(dirty_i);
3247}
3248
3249static void destroy_curseg(struct f2fs_sb_info *sbi)
3250{
3251 struct curseg_info *array = SM_I(sbi)->curseg_array;
3252 int i;
3253
3254 if (!array)
3255 return;
3256 SM_I(sbi)->curseg_array = NULL;
b7ad7512 3257 for (i = 0; i < NR_CURSEG_TYPE; i++) {
351df4b2 3258 kfree(array[i].sum_blk);
b7ad7512
CY		 3259 kfree(array[i].journal);
3260 }
351df4b2
JK		 3261 kfree(array);
3262}
3263
3264static void destroy_free_segmap(struct f2fs_sb_info *sbi)
3265{
3266 struct free_segmap_info *free_i = SM_I(sbi)->free_info;
3267 if (!free_i)
3268 return;
3269 SM_I(sbi)->free_info = NULL;
39307a8e
JK		 3270 kvfree(free_i->free_segmap);
3271 kvfree(free_i->free_secmap);
351df4b2
JK		 3272 kfree(free_i);
3273}
3274
3275static void destroy_sit_info(struct f2fs_sb_info *sbi)
3276{
3277 struct sit_info *sit_i = SIT_I(sbi);
3278 unsigned int start;
3279
3280 if (!sit_i)
3281 return;
3282
3283 if (sit_i->sentries) {
7cd8558b 3284 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2 3285 kfree(sit_i->sentries[start].cur_valid_map);
355e7891
CY		 3286#ifdef CONFIG_F2FS_CHECK_FS
3287 kfree(sit_i->sentries[start].cur_valid_map_mir);
3288#endif
351df4b2 3289 kfree(sit_i->sentries[start].ckpt_valid_map);
a66cdd98 3290 kfree(sit_i->sentries[start].discard_map);
351df4b2
JK		 3291 }
3292 }
60a3b782
JK		 3293 kfree(sit_i->tmp_map);
3294
39307a8e
JK		 3295 kvfree(sit_i->sentries);
3296 kvfree(sit_i->sec_entries);
3297 kvfree(sit_i->dirty_sentries_bitmap);
351df4b2
JK		 3298
3299 SM_I(sbi)->sit_info = NULL;
3300 kfree(sit_i->sit_bitmap);
ae27d62e
CY		 3301#ifdef CONFIG_F2FS_CHECK_FS
3302 kfree(sit_i->sit_bitmap_mir);
3303#endif
351df4b2
JK		 3304 kfree(sit_i);
3305}
3306
3307void destroy_segment_manager(struct f2fs_sb_info *sbi)
3308{
3309 struct f2fs_sm_info *sm_info = SM_I(sbi);
a688b9d9 3310
3b03f724
CY		 3311 if (!sm_info)
3312 return;
5eba8c5d 3313 destroy_flush_cmd_control(sbi, true);
f099405f 3314 destroy_discard_cmd_control(sbi);
351df4b2
JK		 3315 destroy_dirty_segmap(sbi);
3316 destroy_curseg(sbi);
3317 destroy_free_segmap(sbi);
3318 destroy_sit_info(sbi);
3319 sbi->sm_info = NULL;
3320 kfree(sm_info);
3321}
7fd9e544
JK		 3322
3323int __init create_segment_manager_caches(void)
3324{
3325 discard_entry_slab = f2fs_kmem_cache_create("discard_entry",
e8512d2e 3326 sizeof(struct discard_entry));
7fd9e544 3327 if (!discard_entry_slab)
184a5cd2
CY		 3328 goto fail;
3329
b01a9201
JK		 3330 discard_cmd_slab = f2fs_kmem_cache_create("discard_cmd",
3331 sizeof(struct discard_cmd));
3332 if (!discard_cmd_slab)
6ab2a308 3333 goto destroy_discard_entry;
275b66b0 3334
184a5cd2 3335 sit_entry_set_slab = f2fs_kmem_cache_create("sit_entry_set",
c9ee0085 3336 sizeof(struct sit_entry_set));
184a5cd2 3337 if (!sit_entry_set_slab)
b01a9201 3338 goto destroy_discard_cmd;
88b88a66
JK		 3339
3340 inmem_entry_slab = f2fs_kmem_cache_create("inmem_page_entry",
3341 sizeof(struct inmem_pages));
3342 if (!inmem_entry_slab)
3343 goto destroy_sit_entry_set;
7fd9e544 3344 return 0;
184a5cd2 3345
88b88a66
JK		 3346destroy_sit_entry_set:
3347 kmem_cache_destroy(sit_entry_set_slab);
b01a9201
JK		 3348destroy_discard_cmd:
3349 kmem_cache_destroy(discard_cmd_slab);
6ab2a308 3350destroy_discard_entry:
184a5cd2
CY		 3351 kmem_cache_destroy(discard_entry_slab);
3352fail:
3353 return -ENOMEM;
7fd9e544
JK		 3354}
3355
3356void destroy_segment_manager_caches(void)
3357{
184a5cd2 3358 kmem_cache_destroy(sit_entry_set_slab);
b01a9201 3359 kmem_cache_destroy(discard_cmd_slab);
7fd9e544 3360 kmem_cache_destroy(discard_entry_slab);
88b88a66 3361 kmem_cache_destroy(inmem_entry_slab);
7fd9e544 3362}
Ubuntu Jammy Linux kernel mirror