projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: stop gc/discard thread in prior during umount
[linux-2.6-block.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>
1d7be270 19#include <linux/freezer.h>
351df4b2
JK		 20
21#include "f2fs.h"
22#include "segment.h"
23#include "node.h"
9e4ded3f 24#include "trace.h"
6ec178da 25#include <trace/events/f2fs.h>
351df4b2 26
9a7f143a
CL		 27#define __reverse_ffz(x) __reverse_ffs(~(x))
28
7fd9e544 29static struct kmem_cache *discard_entry_slab;
b01a9201 30static struct kmem_cache *discard_cmd_slab;
184a5cd2 31static struct kmem_cache *sit_entry_set_slab;
88b88a66 32static struct kmem_cache *inmem_entry_slab;
7fd9e544 33
f96999c3
JK		 34static unsigned long __reverse_ulong(unsigned char *str)
35{
36 unsigned long tmp = 0;
37 int shift = 24, idx = 0;
38
39#if BITS_PER_LONG == 64
40 shift = 56;
41#endif
42 while (shift >= 0) {
43 tmp |= (unsigned long)str[idx++] << shift;
44 shift -= BITS_PER_BYTE;
45 }
46 return tmp;
47}
48
9a7f143a
CL		 49/*
50 * __reverse_ffs is copied from include/asm-generic/bitops/__ffs.h since
51 * MSB and LSB are reversed in a byte by f2fs_set_bit.
52 */
53static inline unsigned long __reverse_ffs(unsigned long word)
54{
55 int num = 0;
56
57#if BITS_PER_LONG == 64
f96999c3 58 if ((word & 0xffffffff00000000UL) == 0)
9a7f143a 59 num += 32;
f96999c3 60 else
9a7f143a 61 word >>= 32;
9a7f143a 62#endif
f96999c3 63 if ((word & 0xffff0000) == 0)
9a7f143a 64 num += 16;
f96999c3 65 else
9a7f143a 66 word >>= 16;
f96999c3
JK		 67
68 if ((word & 0xff00) == 0)
9a7f143a 69 num += 8;
f96999c3 70 else
9a7f143a 71 word >>= 8;
f96999c3 72
9a7f143a
CL		 73 if ((word & 0xf0) == 0)
74 num += 4;
75 else
76 word >>= 4;
f96999c3 77
9a7f143a
CL		 78 if ((word & 0xc) == 0)
79 num += 2;
80 else
81 word >>= 2;
f96999c3 82
9a7f143a
CL		 83 if ((word & 0x2) == 0)
84 num += 1;
85 return num;
86}
87
88/*
e1c42045 89 * __find_rev_next(_zero)_bit is copied from lib/find_next_bit.c because
9a7f143a 90 * f2fs_set_bit makes MSB and LSB reversed in a byte.
692223d1 91 * @size must be integral times of unsigned long.
9a7f143a 92 * Example:
f96999c3
JK		 93 * MSB <--> LSB
94 * f2fs_set_bit(0, bitmap) => 1000 0000
95 * f2fs_set_bit(7, bitmap) => 0000 0001
9a7f143a
CL		 96 */
97static unsigned long __find_rev_next_bit(const unsigned long *addr,
98 unsigned long size, unsigned long offset)
99{
100 const unsigned long *p = addr + BIT_WORD(offset);
692223d1 101 unsigned long result = size;
9a7f143a 102 unsigned long tmp;
9a7f143a
CL		 103
104 if (offset >= size)
105 return size;
106
692223d1 107 size -= (offset & ~(BITS_PER_LONG - 1));
9a7f143a 108 offset %= BITS_PER_LONG;
f96999c3 109
692223d1
FL		 110 while (1) {
111 if (*p == 0)
112 goto pass;
9a7f143a 113
f96999c3 114 tmp = __reverse_ulong((unsigned char *)p);
692223d1
FL		 115
116 tmp &= ~0UL >> offset;
117 if (size < BITS_PER_LONG)
118 tmp &= (~0UL << (BITS_PER_LONG - size));
9a7f143a 119 if (tmp)
692223d1
FL		 120 goto found;
121pass:
122 if (size <= BITS_PER_LONG)
123 break;
9a7f143a 124 size -= BITS_PER_LONG;
692223d1 125 offset = 0;
f96999c3 126 p++;
9a7f143a 127 }
692223d1
FL		 128 return result;
129found:
130 return result - size + __reverse_ffs(tmp);
9a7f143a
CL		 131}
132
133static unsigned long __find_rev_next_zero_bit(const unsigned long *addr,
134 unsigned long size, unsigned long offset)
135{
136 const unsigned long *p = addr + BIT_WORD(offset);
80609448 137 unsigned long result = size;
9a7f143a 138 unsigned long tmp;
9a7f143a
CL		 139
140 if (offset >= size)
141 return size;
142
80609448 143 size -= (offset & ~(BITS_PER_LONG - 1));
9a7f143a 144 offset %= BITS_PER_LONG;
80609448
JK		 145
146 while (1) {
147 if (*p == ~0UL)
148 goto pass;
149
f96999c3 150 tmp = __reverse_ulong((unsigned char *)p);
80609448
JK		 151
152 if (offset)
153 tmp |= ~0UL << (BITS_PER_LONG - offset);
154 if (size < BITS_PER_LONG)
155 tmp |= ~0UL >> size;
f96999c3 156 if (tmp != ~0UL)
80609448
JK		 157 goto found;
158pass:
159 if (size <= BITS_PER_LONG)
160 break;
9a7f143a 161 size -= BITS_PER_LONG;
80609448 162 offset = 0;
f96999c3 163 p++;
9a7f143a 164 }
80609448
JK		 165 return result;
166found:
167 return result - size + __reverse_ffz(tmp);
9a7f143a
CL		 168}
169
88b88a66
JK		 170void register_inmem_page(struct inode *inode, struct page *page)
171{
172 struct f2fs_inode_info *fi = F2FS_I(inode);
173 struct inmem_pages *new;
9be32d72 174
9e4ded3f 175 f2fs_trace_pid(page);
0722b101 176
decd36b6
CY		 177 set_page_private(page, (unsigned long)ATOMIC_WRITTEN_PAGE);
178 SetPagePrivate(page);
179
88b88a66
JK		 180 new = f2fs_kmem_cache_alloc(inmem_entry_slab, GFP_NOFS);
181
182 /* add atomic page indices to the list */
183 new->page = page;
184 INIT_LIST_HEAD(&new->list);
decd36b6 185
88b88a66
JK		 186 /* increase reference count with clean state */
187 mutex_lock(&fi->inmem_lock);
188 get_page(page);
189 list_add_tail(&new->list, &fi->inmem_pages);
8dcf2ff7 190 inc_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
88b88a66 191 mutex_unlock(&fi->inmem_lock);
8ce67cb0
JK		 192
193 trace_f2fs_register_inmem_page(page, INMEM);
88b88a66
JK		 194}
195
28bc106b
CY		 196static int __revoke_inmem_pages(struct inode *inode,
197 struct list_head *head, bool drop, bool recover)
29b96b54 198{
28bc106b 199 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
29b96b54 200 struct inmem_pages *cur, *tmp;
28bc106b 201 int err = 0;
29b96b54
CY		 202
203 list_for_each_entry_safe(cur, tmp, head, list) {
28bc106b
CY		 204 struct page *page = cur->page;
205
206 if (drop)
207 trace_f2fs_commit_inmem_page(page, INMEM_DROP);
208
209 lock_page(page);
29b96b54 210
28bc106b
CY		 211 if (recover) {
212 struct dnode_of_data dn;
213 struct node_info ni;
214
215 trace_f2fs_commit_inmem_page(page, INMEM_REVOKE);
216
217 set_new_dnode(&dn, inode, NULL, NULL, 0);
218 if (get_dnode_of_data(&dn, page->index, LOOKUP_NODE)) {
219 err = -EAGAIN;
220 goto next;
221 }
222 get_node_info(sbi, dn.nid, &ni);
223 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
224 cur->old_addr, ni.version, true, true);
225 f2fs_put_dnode(&dn);
226 }
227next:
63c52d78
JK		 228 /* we don't need to invalidate this in the sccessful status */
229 if (drop || recover)
230 ClearPageUptodate(page);
28bc106b 231 set_page_private(page, 0);
c81ced05 232 ClearPagePrivate(page);
28bc106b 233 f2fs_put_page(page, 1);
29b96b54
CY		 234
235 list_del(&cur->list);
236 kmem_cache_free(inmem_entry_slab, cur);
237 dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
238 }
28bc106b 239 return err;
29b96b54
CY		 240}
241
242void drop_inmem_pages(struct inode *inode)
243{
244 struct f2fs_inode_info *fi = F2FS_I(inode);
245
246 mutex_lock(&fi->inmem_lock);
28bc106b 247 __revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
29b96b54 248 mutex_unlock(&fi->inmem_lock);
5fe45743
CY		 249
250 clear_inode_flag(inode, FI_ATOMIC_FILE);
251 stat_dec_atomic_write(inode);
29b96b54
CY		 252}
253
8c242db9
JK		 254void drop_inmem_page(struct inode *inode, struct page *page)
255{
256 struct f2fs_inode_info *fi = F2FS_I(inode);
257 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
258 struct list_head *head = &fi->inmem_pages;
259 struct inmem_pages *cur = NULL;
260
261 f2fs_bug_on(sbi, !IS_ATOMIC_WRITTEN_PAGE(page));
262
263 mutex_lock(&fi->inmem_lock);
264 list_for_each_entry(cur, head, list) {
265 if (cur->page == page)
266 break;
267 }
268
269 f2fs_bug_on(sbi, !cur || cur->page != page);
270 list_del(&cur->list);
271 mutex_unlock(&fi->inmem_lock);
272
273 dec_page_count(sbi, F2FS_INMEM_PAGES);
274 kmem_cache_free(inmem_entry_slab, cur);
275
276 ClearPageUptodate(page);
277 set_page_private(page, 0);
278 ClearPagePrivate(page);
279 f2fs_put_page(page, 0);
280
281 trace_f2fs_commit_inmem_page(page, INMEM_INVALIDATE);
282}
283
28bc106b
CY		 284static int __commit_inmem_pages(struct inode *inode,
285 struct list_head *revoke_list)
88b88a66
JK		 286{
287 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
288 struct f2fs_inode_info *fi = F2FS_I(inode);
289 struct inmem_pages *cur, *tmp;
88b88a66 290 struct f2fs_io_info fio = {
05ca3632 291 .sbi = sbi,
88b88a66 292 .type = DATA,
04d328de 293 .op = REQ_OP_WRITE,
70fd7614 294 .op_flags = REQ_SYNC | REQ_PRIO,
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;
e959c8f5 314 fio.old_blkaddr = NULL_ADDR;
4d978078 315 fio.encrypted_page = NULL;
cc15620b 316 fio.need_lock = LOCK_DONE;
29b96b54
CY		 317 err = do_write_data_page(&fio);
318 if (err) {
28bc106b 319 unlock_page(page);
29b96b54 320 break;
70c640b1 321 }
29b96b54 322
28bc106b
CY		 323 /* record old blkaddr for revoking */
324 cur->old_addr = fio.old_blkaddr;
942fd319 325 last_idx = page->index;
28bc106b
CY		 326 }
327 unlock_page(page);
328 list_move_tail(&cur->list, revoke_list);
88b88a66 329 }
29b96b54 330
942fd319 331 if (last_idx != ULONG_MAX)
b9109b0e 332 f2fs_submit_merged_write_cond(sbi, inode, 0, last_idx, DATA);
28bc106b
CY		 333
334 if (!err)
335 __revoke_inmem_pages(inode, revoke_list, false, false);
336
29b96b54
CY		 337 return err;
338}
339
340int commit_inmem_pages(struct inode *inode)
341{
342 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
343 struct f2fs_inode_info *fi = F2FS_I(inode);
28bc106b
CY		 344 struct list_head revoke_list;
345 int err;
29b96b54 346
28bc106b 347 INIT_LIST_HEAD(&revoke_list);
29b96b54
CY		 348 f2fs_balance_fs(sbi, true);
349 f2fs_lock_op(sbi);
350
5fe45743
CY		 351 set_inode_flag(inode, FI_ATOMIC_COMMIT);
352
29b96b54 353 mutex_lock(&fi->inmem_lock);
28bc106b
CY		 354 err = __commit_inmem_pages(inode, &revoke_list);
355 if (err) {
356 int ret;
357 /*
358 * try to revoke all committed pages, but still we could fail
359 * due to no memory or other reason, if that happened, EAGAIN
360 * will be returned, which means in such case, transaction is
361 * already not integrity, caller should use journal to do the
362 * recovery or rewrite & commit last transaction. For other
363 * error number, revoking was done by filesystem itself.
364 */
365 ret = __revoke_inmem_pages(inode, &revoke_list, false, true);
366 if (ret)
367 err = ret;
368
369 /* drop all uncommitted pages */
370 __revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
371 }
88b88a66
JK		 372 mutex_unlock(&fi->inmem_lock);
373
5fe45743
CY		 374 clear_inode_flag(inode, FI_ATOMIC_COMMIT);
375
29b96b54 376 f2fs_unlock_op(sbi);
edb27dee 377 return err;
88b88a66
JK		 378}
379
0a8165d7 380/*
351df4b2
JK		 381 * This function balances dirty node and dentry pages.
382 * In addition, it controls garbage collection.
383 */
2c4db1a6 384void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
351df4b2 385{
0f348028 386#ifdef CONFIG_F2FS_FAULT_INJECTION
55523519
CY		 387 if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
388 f2fs_show_injection_info(FAULT_CHECKPOINT);
0f348028 389 f2fs_stop_checkpoint(sbi, false);
55523519 390 }
0f348028
CY		 391#endif
392
e589c2c4 393 /* balance_fs_bg is able to be pending */
a7881893 394 if (need && excess_cached_nats(sbi))
e589c2c4
JK		 395 f2fs_balance_fs_bg(sbi);
396
351df4b2 397 /*
029cd28c
JK		 398 * We should do GC or end up with checkpoint, if there are so many dirty
399 * dir/node pages without enough free segments.
351df4b2 400 */
7f3037a5 401 if (has_not_enough_free_secs(sbi, 0, 0)) {
351df4b2 402 mutex_lock(&sbi->gc_mutex);
e066b83c 403 f2fs_gc(sbi, false, false, NULL_SEGNO);
351df4b2
JK		 404 }
405}
406
4660f9c0
JK		 407void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi)
408{
1dcc336b 409 /* try to shrink extent cache when there is no enough memory */
554df79e
JK		 410 if (!available_free_memory(sbi, EXTENT_CACHE))
411 f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER);
1dcc336b 412
1b38dc8e
JK		 413 /* check the # of cached NAT entries */
414 if (!available_free_memory(sbi, NAT_ENTRIES))
415 try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK);
416
31696580 417 if (!available_free_memory(sbi, FREE_NIDS))
ad4edb83
JK		 418 try_to_free_nids(sbi, MAX_FREE_NIDS);
419 else
22ad0b6a 420 build_free_nids(sbi, false, false);
31696580 421
1c0f4bf5 422 if (!is_idle(sbi) && !excess_dirty_nats(sbi))
f455c8a5 423 return;
31696580 424
1b38dc8e
JK		 425 /* checkpoint is the only way to shrink partial cached entries */
426 if (!available_free_memory(sbi, NAT_ENTRIES) ||
60b99b48 427 !available_free_memory(sbi, INO_ENTRIES) ||
7d768d2c
CY		 428 excess_prefree_segs(sbi) ||
429 excess_dirty_nats(sbi) ||
f455c8a5 430 f2fs_time_over(sbi, CP_TIME)) {
e9f5b8b8
CY		 431 if (test_opt(sbi, DATA_FLUSH)) {
432 struct blk_plug plug;
433
434 blk_start_plug(&plug);
36b35a0d 435 sync_dirty_inodes(sbi, FILE_INODE);
e9f5b8b8
CY		 436 blk_finish_plug(&plug);
437 }
4660f9c0 438 f2fs_sync_fs(sbi->sb, true);
42190d2a 439 stat_inc_bg_cp_count(sbi->stat_info);
36b35a0d 440 }
4660f9c0
JK		 441}
442
20fda56b
KM		 443static int __submit_flush_wait(struct f2fs_sb_info *sbi,
444 struct block_device *bdev)
3c62be17
JK		 445{
446 struct bio *bio = f2fs_bio_alloc(0);
447 int ret;
448
3adc5fcb 449 bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH;
3c62be17
JK		 450 bio->bi_bdev = bdev;
451 ret = submit_bio_wait(bio);
452 bio_put(bio);
20fda56b
KM		 453
454 trace_f2fs_issue_flush(bdev, test_opt(sbi, NOBARRIER),
455 test_opt(sbi, FLUSH_MERGE), ret);
3c62be17
JK		 456 return ret;
457}
458
459static int submit_flush_wait(struct f2fs_sb_info *sbi)
460{
20fda56b 461 int ret = __submit_flush_wait(sbi, sbi->sb->s_bdev);
3c62be17
JK		 462 int i;
463
20fda56b
KM		 464 if (!sbi->s_ndevs || ret)
465 return ret;
466
467 for (i = 1; i < sbi->s_ndevs; i++) {
468 ret = __submit_flush_wait(sbi, FDEV(i).bdev);
469 if (ret)
470 break;
3c62be17
JK		 471 }
472 return ret;
473}
474
2163d198 475static int issue_flush_thread(void *data)
6b4afdd7
JK		 476{
477 struct f2fs_sb_info *sbi = data;
b01a9201 478 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
a688b9d9 479 wait_queue_head_t *q = &fcc->flush_wait_queue;
6b4afdd7
JK		 480repeat:
481 if (kthread_should_stop())
482 return 0;
483
721bd4d5 484 if (!llist_empty(&fcc->issue_list)) {
6b4afdd7
JK		 485 struct flush_cmd *cmd, *next;
486 int ret;
487
721bd4d5
GZ		 488 fcc->dispatch_list = llist_del_all(&fcc->issue_list);
489 fcc->dispatch_list = llist_reverse_order(fcc->dispatch_list);
490
3c62be17 491 ret = submit_flush_wait(sbi);
8b8dd65f
CY		 492 atomic_inc(&fcc->issued_flush);
493
721bd4d5
GZ		 494 llist_for_each_entry_safe(cmd, next,
495 fcc->dispatch_list, llnode) {
6b4afdd7 496 cmd->ret = ret;
6b4afdd7
JK		 497 complete(&cmd->wait);
498 }
a688b9d9 499 fcc->dispatch_list = NULL;
6b4afdd7
JK		 500 }
501
a688b9d9 502 wait_event_interruptible(*q,
721bd4d5 503 kthread_should_stop() || !llist_empty(&fcc->issue_list));
6b4afdd7
JK		 504 goto repeat;
505}
506
507int f2fs_issue_flush(struct f2fs_sb_info *sbi)
508{
b01a9201 509 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
adf8d90b 510 struct flush_cmd cmd;
8b8dd65f 511 int ret;
6b4afdd7 512
0f7b2abd
JK		 513 if (test_opt(sbi, NOBARRIER))
514 return 0;
515
8b8dd65f
CY		 516 if (!test_opt(sbi, FLUSH_MERGE)) {
517 ret = submit_flush_wait(sbi);
518 atomic_inc(&fcc->issued_flush);
519 return ret;
520 }
740432f8 521
8b8dd65f
CY		 522 if (!atomic_read(&fcc->issing_flush)) {
523 atomic_inc(&fcc->issing_flush);
3c62be17 524 ret = submit_flush_wait(sbi);
8b8dd65f
CY		 525 atomic_dec(&fcc->issing_flush);
526
527 atomic_inc(&fcc->issued_flush);
740432f8
JK		 528 return ret;
529 }
6b4afdd7 530
adf8d90b 531 init_completion(&cmd.wait);
6b4afdd7 532
8b8dd65f 533 atomic_inc(&fcc->issing_flush);
721bd4d5 534 llist_add(&cmd.llnode, &fcc->issue_list);
6b4afdd7 535
a688b9d9
GZ		 536 if (!fcc->dispatch_list)
537 wake_up(&fcc->flush_wait_queue);
6b4afdd7 538
5eba8c5d
JK		 539 if (fcc->f2fs_issue_flush) {
540 wait_for_completion(&cmd.wait);
8b8dd65f 541 atomic_dec(&fcc->issing_flush);
5eba8c5d
JK		 542 } else {
543 llist_del_all(&fcc->issue_list);
8b8dd65f 544 atomic_set(&fcc->issing_flush, 0);
5eba8c5d 545 }
adf8d90b
CY		 546
547 return cmd.ret;
6b4afdd7
JK		 548}
549
2163d198
GZ		 550int create_flush_cmd_control(struct f2fs_sb_info *sbi)
551{
552 dev_t dev = sbi->sb->s_bdev->bd_dev;
553 struct flush_cmd_control *fcc;
554 int err = 0;
555
b01a9201
JK		 556 if (SM_I(sbi)->fcc_info) {
557 fcc = SM_I(sbi)->fcc_info;
d871cd04
YS		 558 if (fcc->f2fs_issue_flush)
559 return err;
5eba8c5d
JK		 560 goto init_thread;
561 }
562
2163d198
GZ		 563 fcc = kzalloc(sizeof(struct flush_cmd_control), GFP_KERNEL);
564 if (!fcc)
565 return -ENOMEM;
8b8dd65f
CY		 566 atomic_set(&fcc->issued_flush, 0);
567 atomic_set(&fcc->issing_flush, 0);
2163d198 568 init_waitqueue_head(&fcc->flush_wait_queue);
721bd4d5 569 init_llist_head(&fcc->issue_list);
b01a9201 570 SM_I(sbi)->fcc_info = fcc;
d4fdf8ba
YH		 571 if (!test_opt(sbi, FLUSH_MERGE))
572 return err;
573
5eba8c5d 574init_thread:
2163d198
GZ		 575 fcc->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi,
576 "f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev));
577 if (IS_ERR(fcc->f2fs_issue_flush)) {
578 err = PTR_ERR(fcc->f2fs_issue_flush);
579 kfree(fcc);
b01a9201 580 SM_I(sbi)->fcc_info = NULL;
2163d198
GZ		 581 return err;
582 }
2163d198
GZ		 583
584 return err;
585}
586
5eba8c5d 587void destroy_flush_cmd_control(struct f2fs_sb_info *sbi, bool free)
2163d198 588{
b01a9201 589 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
2163d198 590
5eba8c5d
JK		 591 if (fcc && fcc->f2fs_issue_flush) {
592 struct task_struct *flush_thread = fcc->f2fs_issue_flush;
593
594 fcc->f2fs_issue_flush = NULL;
595 kthread_stop(flush_thread);
596 }
597 if (free) {
598 kfree(fcc);
b01a9201 599 SM_I(sbi)->fcc_info = NULL;
5eba8c5d 600 }
2163d198
GZ		 601}
602
351df4b2
JK		 603static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
604 enum dirty_type dirty_type)
605{
606 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
607
608 /* need not be added */
609 if (IS_CURSEG(sbi, segno))
610 return;
611
612 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[dirty_type]))
613 dirty_i->nr_dirty[dirty_type]++;
614
615 if (dirty_type == DIRTY) {
616 struct seg_entry *sentry = get_seg_entry(sbi, segno);
4625d6aa 617 enum dirty_type t = sentry->type;
b2f2c390 618
ec325b52
JK		 619 if (unlikely(t >= DIRTY)) {
620 f2fs_bug_on(sbi, 1);
621 return;
622 }
4625d6aa
CL		 623 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[t]))
624 dirty_i->nr_dirty[t]++;
351df4b2
JK		 625 }
626}
627
628static void __remove_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
629 enum dirty_type dirty_type)
630{
631 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
632
633 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[dirty_type]))
634 dirty_i->nr_dirty[dirty_type]--;
635
636 if (dirty_type == DIRTY) {
4625d6aa
CL		 637 struct seg_entry *sentry = get_seg_entry(sbi, segno);
638 enum dirty_type t = sentry->type;
639
640 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
641 dirty_i->nr_dirty[t]--;
b2f2c390 642
302bd348 643 if (get_valid_blocks(sbi, segno, true) == 0)
4ddb1a4d 644 clear_bit(GET_SEC_FROM_SEG(sbi, segno),
5ec4e49f 645 dirty_i->victim_secmap);
351df4b2
JK		 646 }
647}
648
0a8165d7 649/*
351df4b2
JK		 650 * Should not occur error such as -ENOMEM.
651 * Adding dirty entry into seglist is not critical operation.
652 * If a given segment is one of current working segments, it won't be added.
653 */
8d8451af 654static void locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 655{
656 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
657 unsigned short valid_blocks;
658
659 if (segno == NULL_SEGNO || IS_CURSEG(sbi, segno))
660 return;
661
662 mutex_lock(&dirty_i->seglist_lock);
663
302bd348 664 valid_blocks = get_valid_blocks(sbi, segno, false);
351df4b2
JK		 665
666 if (valid_blocks == 0) {
667 __locate_dirty_segment(sbi, segno, PRE);
668 __remove_dirty_segment(sbi, segno, DIRTY);
669 } else if (valid_blocks < sbi->blocks_per_seg) {
670 __locate_dirty_segment(sbi, segno, DIRTY);
671 } else {
672 /* Recovery routine with SSR needs this */
673 __remove_dirty_segment(sbi, segno, DIRTY);
674 }
675
676 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 677}
678
004b6862 679static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi,
c81abe34
JK		 680 struct block_device *bdev, block_t lstart,
681 block_t start, block_t len)
275b66b0 682{
0b54fb84 683 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
ba48a33e 684 struct list_head *pend_list;
b01a9201 685 struct discard_cmd *dc;
275b66b0 686
ba48a33e
CY		 687 f2fs_bug_on(sbi, !len);
688
689 pend_list = &dcc->pend_list[plist_idx(len)];
690
b01a9201
JK		 691 dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS);
692 INIT_LIST_HEAD(&dc->list);
c81abe34 693 dc->bdev = bdev;
b01a9201 694 dc->lstart = lstart;
c81abe34 695 dc->start = start;
b01a9201 696 dc->len = len;
ec9895ad 697 dc->ref = 0;
15469963 698 dc->state = D_PREP;
c81abe34 699 dc->error = 0;
b01a9201 700 init_completion(&dc->wait);
22d375dd 701 list_add_tail(&dc->list, pend_list);
5f32366a 702 atomic_inc(&dcc->discard_cmd_cnt);
d84d1cbd 703 dcc->undiscard_blks += len;
004b6862
CY		 704
705 return dc;
706}
707
708static struct discard_cmd *__attach_discard_cmd(struct f2fs_sb_info *sbi,
709 struct block_device *bdev, block_t lstart,
710 block_t start, block_t len,
711 struct rb_node *parent, struct rb_node **p)
712{
713 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
714 struct discard_cmd *dc;
715
716 dc = __create_discard_cmd(sbi, bdev, lstart, start, len);
717
718 rb_link_node(&dc->rb_node, parent, p);
719 rb_insert_color(&dc->rb_node, &dcc->root);
720
721 return dc;
15469963
JK		 722}
723
004b6862
CY		 724static void __detach_discard_cmd(struct discard_cmd_control *dcc,
725 struct discard_cmd *dc)
15469963 726{
dcc9165d 727 if (dc->state == D_DONE)
004b6862
CY		 728 atomic_dec(&dcc->issing_discard);
729
730 list_del(&dc->list);
731 rb_erase(&dc->rb_node, &dcc->root);
d84d1cbd 732 dcc->undiscard_blks -= dc->len;
004b6862
CY		 733
734 kmem_cache_free(discard_cmd_slab, dc);
735
736 atomic_dec(&dcc->discard_cmd_cnt);
737}
738
739static void __remove_discard_cmd(struct f2fs_sb_info *sbi,
740 struct discard_cmd *dc)
741{
742 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
dcc9165d 743
d9703d90
CY		 744 f2fs_bug_on(sbi, dc->ref);
745
c81abe34
JK		 746 if (dc->error == -EOPNOTSUPP)
747 dc->error = 0;
15469963 748
c81abe34 749 if (dc->error)
15469963 750 f2fs_msg(sbi->sb, KERN_INFO,
04dfc230
CY		 751 "Issue discard(%u, %u, %u) failed, ret: %d",
752 dc->lstart, dc->start, dc->len, dc->error);
004b6862 753 __detach_discard_cmd(dcc, dc);
275b66b0
CY		 754}
755
c81abe34
JK		 756static void f2fs_submit_discard_endio(struct bio *bio)
757{
758 struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
759
c81abe34
JK		 760 dc->error = bio->bi_error;
761 dc->state = D_DONE;
e31b9821 762 complete_all(&dc->wait);
c81abe34
JK		 763 bio_put(bio);
764}
765
766/* this function is copied from blkdev_issue_discard from block/blk-lib.c */
767static void __submit_discard_cmd(struct f2fs_sb_info *sbi,
768 struct discard_cmd *dc)
769{
770 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
771 struct bio *bio = NULL;
772
773 if (dc->state != D_PREP)
774 return;
775
0243a5f9
CY		 776 trace_f2fs_issue_discard(dc->bdev, dc->start, dc->len);
777
c81abe34
JK		 778 dc->error = __blkdev_issue_discard(dc->bdev,
779 SECTOR_FROM_BLOCK(dc->start),
780 SECTOR_FROM_BLOCK(dc->len),
781 GFP_NOFS, 0, &bio);
782 if (!dc->error) {
783 /* should keep before submission to avoid D_DONE right away */
784 dc->state = D_SUBMIT;
8b8dd65f
CY		 785 atomic_inc(&dcc->issued_discard);
786 atomic_inc(&dcc->issing_discard);
c81abe34
JK		 787 if (bio) {
788 bio->bi_private = dc;
789 bio->bi_end_io = f2fs_submit_discard_endio;
790 bio->bi_opf |= REQ_SYNC;
791 submit_bio(bio);
46f84c2c 792 list_move_tail(&dc->list, &dcc->wait_list);
c81abe34
JK		 793 }
794 } else {
795 __remove_discard_cmd(sbi, dc);
796 }
797}
798
004b6862
CY		 799static struct discard_cmd *__insert_discard_tree(struct f2fs_sb_info *sbi,
800 struct block_device *bdev, block_t lstart,
801 block_t start, block_t len,
802 struct rb_node **insert_p,
803 struct rb_node *insert_parent)
c81abe34 804{
004b6862
CY		 805 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
806 struct rb_node **p = &dcc->root.rb_node;
807 struct rb_node *parent = NULL;
808 struct discard_cmd *dc = NULL;
809
810 if (insert_p && insert_parent) {
811 parent = insert_parent;
812 p = insert_p;
813 goto do_insert;
814 }
c81abe34 815
004b6862
CY		 816 p = __lookup_rb_tree_for_insert(sbi, &dcc->root, &parent, lstart);
817do_insert:
818 dc = __attach_discard_cmd(sbi, bdev, lstart, start, len, parent, p);
819 if (!dc)
820 return NULL;
c81abe34 821
004b6862 822 return dc;
c81abe34
JK		 823}
824
ba48a33e
CY		 825static void __relocate_discard_cmd(struct discard_cmd_control *dcc,
826 struct discard_cmd *dc)
827{
828 list_move_tail(&dc->list, &dcc->pend_list[plist_idx(dc->len)]);
829}
830
3d6a650f
YH		 831static void __punch_discard_cmd(struct f2fs_sb_info *sbi,
832 struct discard_cmd *dc, block_t blkaddr)
833{
ba48a33e 834 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 835 struct discard_info di = dc->di;
836 bool modified = false;
3d6a650f 837
004b6862 838 if (dc->state == D_DONE || dc->len == 1) {
3d6a650f
YH		 839 __remove_discard_cmd(sbi, dc);
840 return;
841 }
842
d84d1cbd
CY		 843 dcc->undiscard_blks -= di.len;
844
004b6862 845 if (blkaddr > di.lstart) {
3d6a650f 846 dc->len = blkaddr - dc->lstart;
d84d1cbd 847 dcc->undiscard_blks += dc->len;
ba48a33e 848 __relocate_discard_cmd(dcc, dc);
004b6862
CY		 849 modified = true;
850 }
851
852 if (blkaddr < di.lstart + di.len - 1) {
853 if (modified) {
854 __insert_discard_tree(sbi, dc->bdev, blkaddr + 1,
855 di.start + blkaddr + 1 - di.lstart,
856 di.lstart + di.len - 1 - blkaddr,
857 NULL, NULL);
858 } else {
859 dc->lstart++;
860 dc->len--;
861 dc->start++;
d84d1cbd 862 dcc->undiscard_blks += dc->len;
ba48a33e 863 __relocate_discard_cmd(dcc, dc);
004b6862 864 }
3d6a650f
YH		 865 }
866}
867
004b6862
CY		 868static void __update_discard_tree_range(struct f2fs_sb_info *sbi,
869 struct block_device *bdev, block_t lstart,
870 block_t start, block_t len)
275b66b0 871{
0b54fb84 872 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 873 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
874 struct discard_cmd *dc;
875 struct discard_info di = {0};
876 struct rb_node **insert_p = NULL, *insert_parent = NULL;
877 block_t end = lstart + len;
275b66b0 878
15469963 879 mutex_lock(&dcc->cmd_lock);
40465257 880
004b6862
CY		 881 dc = (struct discard_cmd *)__lookup_rb_tree_ret(&dcc->root,
882 NULL, lstart,
883 (struct rb_entry **)&prev_dc,
884 (struct rb_entry **)&next_dc,
885 &insert_p, &insert_parent, true);
886 if (dc)
887 prev_dc = dc;
888
889 if (!prev_dc) {
890 di.lstart = lstart;
891 di.len = next_dc ? next_dc->lstart - lstart : len;
892 di.len = min(di.len, len);
893 di.start = start;
22d375dd 894 }
15469963 895
004b6862
CY		 896 while (1) {
897 struct rb_node *node;
898 bool merged = false;
899 struct discard_cmd *tdc = NULL;
900
901 if (prev_dc) {
902 di.lstart = prev_dc->lstart + prev_dc->len;
903 if (di.lstart < lstart)
904 di.lstart = lstart;
905 if (di.lstart >= end)
906 break;
907
908 if (!next_dc || next_dc->lstart > end)
909 di.len = end - di.lstart;
910 else
911 di.len = next_dc->lstart - di.lstart;
912 di.start = start + di.lstart - lstart;
913 }
914
915 if (!di.len)
916 goto next;
917
918 if (prev_dc && prev_dc->state == D_PREP &&
919 prev_dc->bdev == bdev &&
920 __is_discard_back_mergeable(&di, &prev_dc->di)) {
921 prev_dc->di.len += di.len;
d84d1cbd 922 dcc->undiscard_blks += di.len;
ba48a33e 923 __relocate_discard_cmd(dcc, prev_dc);
004b6862
CY		 924 di = prev_dc->di;
925 tdc = prev_dc;
926 merged = true;
927 }
928
929 if (next_dc && next_dc->state == D_PREP &&
930 next_dc->bdev == bdev &&
931 __is_discard_front_mergeable(&di, &next_dc->di)) {
932 next_dc->di.lstart = di.lstart;
933 next_dc->di.len += di.len;
934 next_dc->di.start = di.start;
d84d1cbd 935 dcc->undiscard_blks += di.len;
ba48a33e 936 __relocate_discard_cmd(dcc, next_dc);
004b6862
CY		 937 if (tdc)
938 __remove_discard_cmd(sbi, tdc);
004b6862 939 merged = true;
4e6a8d9b 940 }
004b6862 941
df0f6b44 942 if (!merged) {
004b6862
CY		 943 __insert_discard_tree(sbi, bdev, di.lstart, di.start,
944 di.len, NULL, NULL);
df0f6b44 945 }
004b6862
CY		 946 next:
947 prev_dc = next_dc;
948 if (!prev_dc)
949 break;
950
951 node = rb_next(&prev_dc->rb_node);
952 next_dc = rb_entry_safe(node, struct discard_cmd, rb_node);
953 }
954
955 mutex_unlock(&dcc->cmd_lock);
956}
957
958static int __queue_discard_cmd(struct f2fs_sb_info *sbi,
959 struct block_device *bdev, block_t blkstart, block_t blklen)
960{
961 block_t lblkstart = blkstart;
962
0243a5f9 963 trace_f2fs_queue_discard(bdev, blkstart, blklen);
004b6862
CY		 964
965 if (sbi->s_ndevs) {
966 int devi = f2fs_target_device_index(sbi, blkstart);
967
968 blkstart -= FDEV(devi).start_blk;
969 }
970 __update_discard_tree_range(sbi, bdev, lblkstart, blkstart, blklen);
004b6862
CY		 971 return 0;
972}
973
bd5b0738
CY		 974static void __issue_discard_cmd(struct f2fs_sb_info *sbi, bool issue_cond)
975{
976 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
977 struct list_head *pend_list;
978 struct discard_cmd *dc, *tmp;
979 struct blk_plug plug;
980 int i, iter = 0;
981
982 mutex_lock(&dcc->cmd_lock);
963932a9
YH		 983 f2fs_bug_on(sbi,
984 !__check_rb_tree_consistence(sbi, &dcc->root));
bd5b0738
CY		 985 blk_start_plug(&plug);
986 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
987 pend_list = &dcc->pend_list[i];
988 list_for_each_entry_safe(dc, tmp, pend_list, list) {
989 f2fs_bug_on(sbi, dc->state != D_PREP);
990
991 if (!issue_cond || is_idle(sbi))
992 __submit_discard_cmd(sbi, dc);
993 if (issue_cond && iter++ > DISCARD_ISSUE_RATE)
994 goto out;
995 }
996 }
997out:
998 blk_finish_plug(&plug);
999 mutex_unlock(&dcc->cmd_lock);
1000}
1001
2a510c00
CY		 1002static void __wait_one_discard_bio(struct f2fs_sb_info *sbi,
1003 struct discard_cmd *dc)
1004{
1005 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1006
1007 wait_for_completion_io(&dc->wait);
1008 mutex_lock(&dcc->cmd_lock);
1009 f2fs_bug_on(sbi, dc->state != D_DONE);
1010 dc->ref--;
1011 if (!dc->ref)
1012 __remove_discard_cmd(sbi, dc);
1013 mutex_unlock(&dcc->cmd_lock);
1014}
1015
63a94fa1
CY		 1016static void __wait_discard_cmd(struct f2fs_sb_info *sbi, bool wait_cond)
1017{
1018 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1019 struct list_head *wait_list = &(dcc->wait_list);
1020 struct discard_cmd *dc, *tmp;
6afae633
CY		 1021 bool need_wait;
1022
1023next:
1024 need_wait = false;
63a94fa1
CY		 1025
1026 mutex_lock(&dcc->cmd_lock);
1027 list_for_each_entry_safe(dc, tmp, wait_list, list) {
6afae633 1028 if (!wait_cond || (dc->state == D_DONE && !dc->ref)) {
63a94fa1
CY		 1029 wait_for_completion_io(&dc->wait);
1030 __remove_discard_cmd(sbi, dc);
6afae633
CY		 1031 } else {
1032 dc->ref++;
1033 need_wait = true;
1034 break;
63a94fa1
CY		 1035 }
1036 }
1037 mutex_unlock(&dcc->cmd_lock);
6afae633
CY		 1038
1039 if (need_wait) {
2a510c00 1040 __wait_one_discard_bio(sbi, dc);
6afae633
CY		 1041 goto next;
1042 }
63a94fa1
CY		 1043}
1044
004b6862
CY		 1045/* This should be covered by global mutex, &sit_i->sentry_lock */
1046void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
1047{
1048 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1049 struct discard_cmd *dc;
ec9895ad 1050 bool need_wait = false;
004b6862
CY		 1051
1052 mutex_lock(&dcc->cmd_lock);
004b6862
CY		 1053 dc = (struct discard_cmd *)__lookup_rb_tree(&dcc->root, NULL, blkaddr);
1054 if (dc) {
ec9895ad
CY		 1055 if (dc->state == D_PREP) {
1056 __punch_discard_cmd(sbi, dc, blkaddr);
1057 } else {
1058 dc->ref++;
1059 need_wait = true;
1060 }
275b66b0 1061 }
d431413f 1062 mutex_unlock(&dcc->cmd_lock);
ec9895ad 1063
2a510c00
CY		 1064 if (need_wait)
1065 __wait_one_discard_bio(sbi, dc);
d431413f
CY		 1066}
1067
cce13252
CY		 1068void stop_discard_thread(struct f2fs_sb_info *sbi)
1069{
1070 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1071
1072 if (dcc && dcc->f2fs_issue_discard) {
1073 struct task_struct *discard_thread = dcc->f2fs_issue_discard;
1074
1075 dcc->f2fs_issue_discard = NULL;
1076 kthread_stop(discard_thread);
1077 }
1078}
1079
d431413f
CY		 1080/* This comes from f2fs_put_super */
1081void f2fs_wait_discard_bios(struct f2fs_sb_info *sbi)
1082{
bd5b0738 1083 __issue_discard_cmd(sbi, false);
63a94fa1 1084 __wait_discard_cmd(sbi, false);
275b66b0
CY		 1085}
1086
15469963
JK		 1087static int issue_discard_thread(void *data)
1088{
1089 struct f2fs_sb_info *sbi = data;
1090 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1091 wait_queue_head_t *q = &dcc->discard_wait_queue;
15469963 1092
1d7be270 1093 set_freezable();
15469963 1094
1d7be270
JK		 1095 do {
1096 wait_event_interruptible(*q, kthread_should_stop() ||
1097 freezing(current) ||
1098 atomic_read(&dcc->discard_cmd_cnt));
1099 if (try_to_freeze())
1100 continue;
1101 if (kthread_should_stop())
1102 return 0;
15469963 1103
1d7be270
JK		 1104 __issue_discard_cmd(sbi, true);
1105 __wait_discard_cmd(sbi, true);
1106
1107 congestion_wait(BLK_RW_SYNC, HZ/50);
1108 } while (!kthread_should_stop());
1109 return 0;
15469963
JK		 1110}
1111
f46e8809 1112#ifdef CONFIG_BLK_DEV_ZONED
3c62be17
JK		 1113static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
1114 struct block_device *bdev, block_t blkstart, block_t blklen)
f46e8809 1115{
92592285 1116 sector_t sector, nr_sects;
10a875f8 1117 block_t lblkstart = blkstart;
3c62be17
JK		 1118 int devi = 0;
1119
1120 if (sbi->s_ndevs) {
1121 devi = f2fs_target_device_index(sbi, blkstart);
1122 blkstart -= FDEV(devi).start_blk;
1123 }
f46e8809
DLM		 1124
1125 /*
1126 * We need to know the type of the zone: for conventional zones,
1127 * use regular discard if the drive supports it. For sequential
1128 * zones, reset the zone write pointer.
1129 */
3c62be17 1130 switch (get_blkz_type(sbi, bdev, blkstart)) {
f46e8809
DLM		 1131
1132 case BLK_ZONE_TYPE_CONVENTIONAL:
1133 if (!blk_queue_discard(bdev_get_queue(bdev)))
1134 return 0;
c81abe34 1135 return __queue_discard_cmd(sbi, bdev, lblkstart, blklen);
f46e8809
DLM		 1136 case BLK_ZONE_TYPE_SEQWRITE_REQ:
1137 case BLK_ZONE_TYPE_SEQWRITE_PREF:
92592285
JK		 1138 sector = SECTOR_FROM_BLOCK(blkstart);
1139 nr_sects = SECTOR_FROM_BLOCK(blklen);
1140
1141 if (sector & (bdev_zone_sectors(bdev) - 1) ||
1142 nr_sects != bdev_zone_sectors(bdev)) {
1143 f2fs_msg(sbi->sb, KERN_INFO,
1144 "(%d) %s: Unaligned discard attempted (block %x + %x)",
1145 devi, sbi->s_ndevs ? FDEV(devi).path: "",
1146 blkstart, blklen);
1147 return -EIO;
1148 }
d50aaeec 1149 trace_f2fs_issue_reset_zone(bdev, blkstart);
f46e8809
DLM		 1150 return blkdev_reset_zones(bdev, sector,
1151 nr_sects, GFP_NOFS);
1152 default:
1153 /* Unknown zone type: broken device ? */
1154 return -EIO;
1155 }
1156}
1157#endif
1158
3c62be17
JK		 1159static int __issue_discard_async(struct f2fs_sb_info *sbi,
1160 struct block_device *bdev, block_t blkstart, block_t blklen)
1161{
1162#ifdef CONFIG_BLK_DEV_ZONED
1163 if (f2fs_sb_mounted_blkzoned(sbi->sb) &&
1164 bdev_zoned_model(bdev) != BLK_ZONED_NONE)
1165 return __f2fs_issue_discard_zone(sbi, bdev, blkstart, blklen);
1166#endif
c81abe34 1167 return __queue_discard_cmd(sbi, bdev, blkstart, blklen);
3c62be17
JK		 1168}
1169
1e87a78d 1170static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
37208879
JK		 1171 block_t blkstart, block_t blklen)
1172{
3c62be17
JK		 1173 sector_t start = blkstart, len = 0;
1174 struct block_device *bdev;
a66cdd98
JK		 1175 struct seg_entry *se;
1176 unsigned int offset;
1177 block_t i;
3c62be17
JK		 1178 int err = 0;
1179
1180 bdev = f2fs_target_device(sbi, blkstart, NULL);
1181
1182 for (i = blkstart; i < blkstart + blklen; i++, len++) {
1183 if (i != start) {
1184 struct block_device *bdev2 =
1185 f2fs_target_device(sbi, i, NULL);
1186
1187 if (bdev2 != bdev) {
1188 err = __issue_discard_async(sbi, bdev,
1189 start, len);
1190 if (err)
1191 return err;
1192 bdev = bdev2;
1193 start = i;
1194 len = 0;
1195 }
1196 }
a66cdd98 1197
a66cdd98
JK		 1198 se = get_seg_entry(sbi, GET_SEGNO(sbi, i));
1199 offset = GET_BLKOFF_FROM_SEG0(sbi, i);
1200
1201 if (!f2fs_test_and_set_bit(offset, se->discard_map))
1202 sbi->discard_blks--;
1203 }
f46e8809 1204
3c62be17
JK		 1205 if (len)
1206 err = __issue_discard_async(sbi, bdev, start, len);
1207 return err;
1e87a78d
JK		 1208}
1209
25290fa5
JK		 1210static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc,
1211 bool check_only)
adf4983b 1212{
b2955550
JK		 1213 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
1214 int max_blocks = sbi->blocks_per_seg;
4b2fecc8 1215 struct seg_entry *se = get_seg_entry(sbi, cpc->trim_start);
b2955550
JK		 1216 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1217 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
a66cdd98 1218 unsigned long *discard_map = (unsigned long *)se->discard_map;
60a3b782 1219 unsigned long *dmap = SIT_I(sbi)->tmp_map;
b2955550 1220 unsigned int start = 0, end = -1;
c473f1a9 1221 bool force = (cpc->reason & CP_DISCARD);
a7eeb823 1222 struct discard_entry *de = NULL;
46f84c2c 1223 struct list_head *head = &SM_I(sbi)->dcc_info->entry_list;
b2955550
JK		 1224 int i;
1225
3e025740 1226 if (se->valid_blocks == max_blocks || !f2fs_discard_en(sbi))
25290fa5 1227 return false;
b2955550 1228
a66cdd98
JK		 1229 if (!force) {
1230 if (!test_opt(sbi, DISCARD) || !se->valid_blocks ||
0b54fb84
JK		 1231 SM_I(sbi)->dcc_info->nr_discards >=
1232 SM_I(sbi)->dcc_info->max_discards)
25290fa5 1233 return false;
4b2fecc8
JK		 1234 }
1235
b2955550
JK		 1236 /* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
1237 for (i = 0; i < entries; i++)
a66cdd98 1238 dmap[i] = force ? ~ckpt_map[i] & ~discard_map[i] :
d7bc2484 1239 (cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
b2955550 1240
0b54fb84
JK		 1241 while (force || SM_I(sbi)->dcc_info->nr_discards <=
1242 SM_I(sbi)->dcc_info->max_discards) {
b2955550
JK		 1243 start = __find_rev_next_bit(dmap, max_blocks, end + 1);
1244 if (start >= max_blocks)
1245 break;
1246
1247 end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1);
c7b41e16
YH		 1248 if (force && start && end != max_blocks
1249 && (end - start) < cpc->trim_minlen)
1250 continue;
1251
25290fa5
JK		 1252 if (check_only)
1253 return true;
1254
a7eeb823
CY		 1255 if (!de) {
1256 de = f2fs_kmem_cache_alloc(discard_entry_slab,
1257 GFP_F2FS_ZERO);
1258 de->start_blkaddr = START_BLOCK(sbi, cpc->trim_start);
1259 list_add_tail(&de->list, head);
1260 }
1261
1262 for (i = start; i < end; i++)
1263 __set_bit_le(i, (void *)de->discard_map);
1264
1265 SM_I(sbi)->dcc_info->nr_discards += end - start;
b2955550 1266 }
25290fa5 1267 return false;
b2955550
JK		 1268}
1269
4b2fecc8
JK		 1270void release_discard_addrs(struct f2fs_sb_info *sbi)
1271{
46f84c2c 1272 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
4b2fecc8
JK		 1273 struct discard_entry *entry, *this;
1274
1275 /* drop caches */
1276 list_for_each_entry_safe(entry, this, head, list) {
1277 list_del(&entry->list);
1278 kmem_cache_free(discard_entry_slab, entry);
1279 }
1280}
1281
0a8165d7 1282/*
351df4b2
JK		 1283 * Should call clear_prefree_segments after checkpoint is done.
1284 */
1285static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
1286{
1287 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
b65ee148 1288 unsigned int segno;
351df4b2
JK		 1289
1290 mutex_lock(&dirty_i->seglist_lock);
7cd8558b 1291 for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], MAIN_SEGS(sbi))
351df4b2 1292 __set_test_and_free(sbi, segno);
351df4b2
JK		 1293 mutex_unlock(&dirty_i->seglist_lock);
1294}
1295
836b5a63 1296void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2 1297{
46f84c2c 1298 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
2d7b822a 1299 struct discard_entry *entry, *this;
351df4b2 1300 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
29e59c14 1301 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
29e59c14 1302 unsigned int start = 0, end = -1;
36abef4e 1303 unsigned int secno, start_segno;
c473f1a9 1304 bool force = (cpc->reason & CP_DISCARD);
351df4b2
JK		 1305
1306 mutex_lock(&dirty_i->seglist_lock);
29e59c14 1307
351df4b2 1308 while (1) {
29e59c14 1309 int i;
7cd8558b
JK		 1310 start = find_next_bit(prefree_map, MAIN_SEGS(sbi), end + 1);
1311 if (start >= MAIN_SEGS(sbi))
351df4b2 1312 break;
7cd8558b
JK		 1313 end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
1314 start + 1);
29e59c14
CL		 1315
1316 for (i = start; i < end; i++)
1317 clear_bit(i, prefree_map);
1318
1319 dirty_i->nr_dirty[PRE] -= end - start;
1320
650d3c4e 1321 if (!test_opt(sbi, DISCARD))
29e59c14 1322 continue;
351df4b2 1323
650d3c4e
YH		 1324 if (force && start >= cpc->trim_start &&
1325 (end - 1) <= cpc->trim_end)
1326 continue;
1327
36abef4e
JK		 1328 if (!test_opt(sbi, LFS) || sbi->segs_per_sec == 1) {
1329 f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
37208879 1330 (end - start) << sbi->log_blocks_per_seg);
36abef4e
JK		 1331 continue;
1332 }
1333next:
4ddb1a4d
JK		 1334 secno = GET_SEC_FROM_SEG(sbi, start);
1335 start_segno = GET_SEG_FROM_SEC(sbi, secno);
36abef4e 1336 if (!IS_CURSEC(sbi, secno) &&
302bd348 1337 !get_valid_blocks(sbi, start, true))
36abef4e
JK		 1338 f2fs_issue_discard(sbi, START_BLOCK(sbi, start_segno),
1339 sbi->segs_per_sec << sbi->log_blocks_per_seg);
1340
1341 start = start_segno + sbi->segs_per_sec;
1342 if (start < end)
1343 goto next;
8b107f5b
JK		 1344 else
1345 end = start - 1;
351df4b2
JK		 1346 }
1347 mutex_unlock(&dirty_i->seglist_lock);
b2955550
JK		 1348
1349 /* send small discards */
2d7b822a 1350 list_for_each_entry_safe(entry, this, head, list) {
a7eeb823
CY		 1351 unsigned int cur_pos = 0, next_pos, len, total_len = 0;
1352 bool is_valid = test_bit_le(0, entry->discard_map);
1353
1354find_next:
1355 if (is_valid) {
1356 next_pos = find_next_zero_bit_le(entry->discard_map,
1357 sbi->blocks_per_seg, cur_pos);
1358 len = next_pos - cur_pos;
1359
acfd2810
DLM		 1360 if (f2fs_sb_mounted_blkzoned(sbi->sb) ||
1361 (force && len < cpc->trim_minlen))
a7eeb823
CY		 1362 goto skip;
1363
1364 f2fs_issue_discard(sbi, entry->start_blkaddr + cur_pos,
1365 len);
1366 cpc->trimmed += len;
1367 total_len += len;
1368 } else {
1369 next_pos = find_next_bit_le(entry->discard_map,
1370 sbi->blocks_per_seg, cur_pos);
1371 }
836b5a63 1372skip:
a7eeb823
CY		 1373 cur_pos = next_pos;
1374 is_valid = !is_valid;
1375
1376 if (cur_pos < sbi->blocks_per_seg)
1377 goto find_next;
1378
b2955550 1379 list_del(&entry->list);
a7eeb823 1380 SM_I(sbi)->dcc_info->nr_discards -= total_len;
b2955550
JK		 1381 kmem_cache_free(discard_entry_slab, entry);
1382 }
34e159da
CY		 1383
1384 wake_up(&SM_I(sbi)->dcc_info->discard_wait_queue);
351df4b2
JK		 1385}
1386
8ed59745 1387static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84 1388{
15469963 1389 dev_t dev = sbi->sb->s_bdev->bd_dev;
0b54fb84 1390 struct discard_cmd_control *dcc;
ba48a33e 1391 int err = 0, i;
0b54fb84
JK		 1392
1393 if (SM_I(sbi)->dcc_info) {
1394 dcc = SM_I(sbi)->dcc_info;
1395 goto init_thread;
1396 }
1397
1398 dcc = kzalloc(sizeof(struct discard_cmd_control), GFP_KERNEL);
1399 if (!dcc)
1400 return -ENOMEM;
1401
46f84c2c 1402 INIT_LIST_HEAD(&dcc->entry_list);
ba48a33e
CY		 1403 for (i = 0; i < MAX_PLIST_NUM; i++)
1404 INIT_LIST_HEAD(&dcc->pend_list[i]);
46f84c2c 1405 INIT_LIST_HEAD(&dcc->wait_list);
15469963 1406 mutex_init(&dcc->cmd_lock);
8b8dd65f
CY		 1407 atomic_set(&dcc->issued_discard, 0);
1408 atomic_set(&dcc->issing_discard, 0);
5f32366a 1409 atomic_set(&dcc->discard_cmd_cnt, 0);
0b54fb84 1410 dcc->nr_discards = 0;
d618ebaf 1411 dcc->max_discards = MAIN_SEGS(sbi) << sbi->log_blocks_per_seg;
d84d1cbd 1412 dcc->undiscard_blks = 0;
004b6862 1413 dcc->root = RB_ROOT;
0b54fb84 1414
15469963 1415 init_waitqueue_head(&dcc->discard_wait_queue);
0b54fb84
JK		 1416 SM_I(sbi)->dcc_info = dcc;
1417init_thread:
15469963
JK		 1418 dcc->f2fs_issue_discard = kthread_run(issue_discard_thread, sbi,
1419 "f2fs_discard-%u:%u", MAJOR(dev), MINOR(dev));
1420 if (IS_ERR(dcc->f2fs_issue_discard)) {
1421 err = PTR_ERR(dcc->f2fs_issue_discard);
1422 kfree(dcc);
1423 SM_I(sbi)->dcc_info = NULL;
1424 return err;
1425 }
1426
0b54fb84
JK		 1427 return err;
1428}
1429
f099405f 1430static void destroy_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84
JK		 1431{
1432 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1433
f099405f
CY		 1434 if (!dcc)
1435 return;
1436
cce13252 1437 stop_discard_thread(sbi);
f099405f
CY		 1438
1439 kfree(dcc);
1440 SM_I(sbi)->dcc_info = NULL;
0b54fb84
JK		 1441}
1442
184a5cd2 1443static bool __mark_sit_entry_dirty(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 1444{
1445 struct sit_info *sit_i = SIT_I(sbi);
184a5cd2
CY		 1446
1447 if (!__test_and_set_bit(segno, sit_i->dirty_sentries_bitmap)) {
351df4b2 1448 sit_i->dirty_sentries++;
184a5cd2
CY		 1449 return false;
1450 }
1451
1452 return true;
351df4b2
JK		 1453}
1454
1455static void __set_sit_entry_type(struct f2fs_sb_info *sbi, int type,
1456 unsigned int segno, int modified)
1457{
1458 struct seg_entry *se = get_seg_entry(sbi, segno);
1459 se->type = type;
1460 if (modified)
1461 __mark_sit_entry_dirty(sbi, segno);
1462}
1463
1464static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
1465{
1466 struct seg_entry *se;
1467 unsigned int segno, offset;
1468 long int new_vblocks;
1469
1470 segno = GET_SEGNO(sbi, blkaddr);
1471
1472 se = get_seg_entry(sbi, segno);
1473 new_vblocks = se->valid_blocks + del;
491c0854 1474 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
351df4b2 1475
9850cf4a 1476 f2fs_bug_on(sbi, (new_vblocks >> (sizeof(unsigned short) << 3) ||
351df4b2
JK		 1477 (new_vblocks > sbi->blocks_per_seg)));
1478
1479 se->valid_blocks = new_vblocks;
1480 se->mtime = get_mtime(sbi);
1481 SIT_I(sbi)->max_mtime = se->mtime;
1482
1483 /* Update valid block bitmap */
1484 if (del > 0) {
355e7891
CY		 1485 if (f2fs_test_and_set_bit(offset, se->cur_valid_map)) {
1486#ifdef CONFIG_F2FS_CHECK_FS
1487 if (f2fs_test_and_set_bit(offset,
1488 se->cur_valid_map_mir))
1489 f2fs_bug_on(sbi, 1);
1490 else
1491 WARN_ON(1);
1492#else
05796763 1493 f2fs_bug_on(sbi, 1);
355e7891
CY		 1494#endif
1495 }
3e025740
JK		 1496 if (f2fs_discard_en(sbi) &&
1497 !f2fs_test_and_set_bit(offset, se->discard_map))
a66cdd98 1498 sbi->discard_blks--;
720037f9
JK		 1499
1500 /* don't overwrite by SSR to keep node chain */
1501 if (se->type == CURSEG_WARM_NODE) {
1502 if (!f2fs_test_and_set_bit(offset, se->ckpt_valid_map))
1503 se->ckpt_valid_blocks++;
1504 }
351df4b2 1505 } else {
355e7891
CY		 1506 if (!f2fs_test_and_clear_bit(offset, se->cur_valid_map)) {
1507#ifdef CONFIG_F2FS_CHECK_FS
1508 if (!f2fs_test_and_clear_bit(offset,
1509 se->cur_valid_map_mir))
1510 f2fs_bug_on(sbi, 1);
1511 else
1512 WARN_ON(1);
1513#else
05796763 1514 f2fs_bug_on(sbi, 1);
355e7891
CY		 1515#endif
1516 }
3e025740
JK		 1517 if (f2fs_discard_en(sbi) &&
1518 f2fs_test_and_clear_bit(offset, se->discard_map))
a66cdd98 1519 sbi->discard_blks++;
351df4b2
JK		 1520 }
1521 if (!f2fs_test_bit(offset, se->ckpt_valid_map))
1522 se->ckpt_valid_blocks += del;
1523
1524 __mark_sit_entry_dirty(sbi, segno);
1525
1526 /* update total number of valid blocks to be written in ckpt area */
1527 SIT_I(sbi)->written_valid_blocks += del;
1528
1529 if (sbi->segs_per_sec > 1)
1530 get_sec_entry(sbi, segno)->valid_blocks += del;
1531}
1532
5e443818 1533void refresh_sit_entry(struct f2fs_sb_info *sbi, block_t old, block_t new)
351df4b2 1534{
5e443818
JK		 1535 update_sit_entry(sbi, new, 1);
1536 if (GET_SEGNO(sbi, old) != NULL_SEGNO)
1537 update_sit_entry(sbi, old, -1);
1538
1539 locate_dirty_segment(sbi, GET_SEGNO(sbi, old));
1540 locate_dirty_segment(sbi, GET_SEGNO(sbi, new));
351df4b2
JK		 1541}
1542
1543void invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
1544{
1545 unsigned int segno = GET_SEGNO(sbi, addr);
1546 struct sit_info *sit_i = SIT_I(sbi);
1547
9850cf4a 1548 f2fs_bug_on(sbi, addr == NULL_ADDR);
351df4b2
JK		 1549 if (addr == NEW_ADDR)
1550 return;
1551
1552 /* add it into sit main buffer */
1553 mutex_lock(&sit_i->sentry_lock);
1554
1555 update_sit_entry(sbi, addr, -1);
1556
1557 /* add it into dirty seglist */
1558 locate_dirty_segment(sbi, segno);
1559
1560 mutex_unlock(&sit_i->sentry_lock);
1561}
1562
6e2c64ad
JK		 1563bool is_checkpointed_data(struct f2fs_sb_info *sbi, block_t blkaddr)
1564{
1565 struct sit_info *sit_i = SIT_I(sbi);
1566 unsigned int segno, offset;
1567 struct seg_entry *se;
1568 bool is_cp = false;
1569
1570 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
1571 return true;
1572
1573 mutex_lock(&sit_i->sentry_lock);
1574
1575 segno = GET_SEGNO(sbi, blkaddr);
1576 se = get_seg_entry(sbi, segno);
1577 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
1578
1579 if (f2fs_test_bit(offset, se->ckpt_valid_map))
1580 is_cp = true;
1581
1582 mutex_unlock(&sit_i->sentry_lock);
1583
1584 return is_cp;
1585}
1586
0a8165d7 1587/*
351df4b2
JK		 1588 * This function should be resided under the curseg_mutex lock
1589 */
1590static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
e79efe3b 1591 struct f2fs_summary *sum)
351df4b2
JK		 1592{
1593 struct curseg_info *curseg = CURSEG_I(sbi, type);
1594 void *addr = curseg->sum_blk;
e79efe3b 1595 addr += curseg->next_blkoff * sizeof(struct f2fs_summary);
351df4b2 1596 memcpy(addr, sum, sizeof(struct f2fs_summary));
351df4b2
JK		 1597}
1598
0a8165d7 1599/*
351df4b2
JK		 1600 * Calculate the number of current summary pages for writing
1601 */
3fa06d7b 1602int npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
351df4b2 1603{
351df4b2 1604 int valid_sum_count = 0;
9a47938b 1605 int i, sum_in_page;
351df4b2
JK		 1606
1607 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1608 if (sbi->ckpt->alloc_type[i] == SSR)
1609 valid_sum_count += sbi->blocks_per_seg;
3fa06d7b
CY		 1610 else {
1611 if (for_ra)
1612 valid_sum_count += le16_to_cpu(
1613 F2FS_CKPT(sbi)->cur_data_blkoff[i]);
1614 else
1615 valid_sum_count += curseg_blkoff(sbi, i);
1616 }
351df4b2
JK		 1617 }
1618
09cbfeaf 1619 sum_in_page = (PAGE_SIZE - 2 * SUM_JOURNAL_SIZE -
9a47938b
FL		 1620 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
1621 if (valid_sum_count <= sum_in_page)
351df4b2 1622 return 1;
9a47938b 1623 else if ((valid_sum_count - sum_in_page) <=
09cbfeaf 1624 (PAGE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
351df4b2
JK		 1625 return 2;
1626 return 3;
1627}
1628
0a8165d7 1629/*
351df4b2
JK		 1630 * Caller should put this summary page
1631 */
1632struct page *get_sum_page(struct f2fs_sb_info *sbi, unsigned int segno)
1633{
1634 return get_meta_page(sbi, GET_SUM_BLOCK(sbi, segno));
1635}
1636
381722d2 1637void update_meta_page(struct f2fs_sb_info *sbi, void *src, block_t blk_addr)
351df4b2
JK		 1638{
1639 struct page *page = grab_meta_page(sbi, blk_addr);
381722d2
CY		 1640 void *dst = page_address(page);
1641
1642 if (src)
09cbfeaf 1643 memcpy(dst, src, PAGE_SIZE);
381722d2 1644 else
09cbfeaf 1645 memset(dst, 0, PAGE_SIZE);
351df4b2
JK		 1646 set_page_dirty(page);
1647 f2fs_put_page(page, 1);
1648}
1649
381722d2
CY		 1650static void write_sum_page(struct f2fs_sb_info *sbi,
1651 struct f2fs_summary_block *sum_blk, block_t blk_addr)
1652{
1653 update_meta_page(sbi, (void *)sum_blk, blk_addr);
1654}
1655
b7ad7512
CY		 1656static void write_current_sum_page(struct f2fs_sb_info *sbi,
1657 int type, block_t blk_addr)
1658{
1659 struct curseg_info *curseg = CURSEG_I(sbi, type);
1660 struct page *page = grab_meta_page(sbi, blk_addr);
1661 struct f2fs_summary_block *src = curseg->sum_blk;
1662 struct f2fs_summary_block *dst;
1663
1664 dst = (struct f2fs_summary_block *)page_address(page);
1665
1666 mutex_lock(&curseg->curseg_mutex);
1667
1668 down_read(&curseg->journal_rwsem);
1669 memcpy(&dst->journal, curseg->journal, SUM_JOURNAL_SIZE);
1670 up_read(&curseg->journal_rwsem);
1671
1672 memcpy(dst->entries, src->entries, SUM_ENTRY_SIZE);
1673 memcpy(&dst->footer, &src->footer, SUM_FOOTER_SIZE);
1674
1675 mutex_unlock(&curseg->curseg_mutex);
1676
1677 set_page_dirty(page);
1678 f2fs_put_page(page, 1);
1679}
1680
a7881893
JK		 1681static int is_next_segment_free(struct f2fs_sb_info *sbi, int type)
1682{
1683 struct curseg_info *curseg = CURSEG_I(sbi, type);
1684 unsigned int segno = curseg->segno + 1;
1685 struct free_segmap_info *free_i = FREE_I(sbi);
1686
1687 if (segno < MAIN_SEGS(sbi) && segno % sbi->segs_per_sec)
1688 return !test_bit(segno, free_i->free_segmap);
1689 return 0;
1690}
1691
0a8165d7 1692/*
351df4b2
JK		 1693 * Find a new segment from the free segments bitmap to right order
1694 * This function should be returned with success, otherwise BUG
1695 */
1696static void get_new_segment(struct f2fs_sb_info *sbi,
1697 unsigned int *newseg, bool new_sec, int dir)
1698{
1699 struct free_segmap_info *free_i = FREE_I(sbi);
351df4b2 1700 unsigned int segno, secno, zoneno;
7cd8558b 1701 unsigned int total_zones = MAIN_SECS(sbi) / sbi->secs_per_zone;
4ddb1a4d
JK		 1702 unsigned int hint = GET_SEC_FROM_SEG(sbi, *newseg);
1703 unsigned int old_zoneno = GET_ZONE_FROM_SEG(sbi, *newseg);
351df4b2
JK		 1704 unsigned int left_start = hint;
1705 bool init = true;
1706 int go_left = 0;
1707 int i;
1708
1a118ccf 1709 spin_lock(&free_i->segmap_lock);
351df4b2
JK		 1710
1711 if (!new_sec && ((*newseg + 1) % sbi->segs_per_sec)) {
1712 segno = find_next_zero_bit(free_i->free_segmap,
4ddb1a4d
JK		 1713 GET_SEG_FROM_SEC(sbi, hint + 1), *newseg + 1);
1714 if (segno < GET_SEG_FROM_SEC(sbi, hint + 1))
351df4b2
JK		 1715 goto got_it;
1716 }
1717find_other_zone:
7cd8558b
JK		 1718 secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
1719 if (secno >= MAIN_SECS(sbi)) {
351df4b2
JK		 1720 if (dir == ALLOC_RIGHT) {
1721 secno = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1722 MAIN_SECS(sbi), 0);
1723 f2fs_bug_on(sbi, secno >= MAIN_SECS(sbi));
351df4b2
JK		 1724 } else {
1725 go_left = 1;
1726 left_start = hint - 1;
1727 }
1728 }
1729 if (go_left == 0)
1730 goto skip_left;
1731
1732 while (test_bit(left_start, free_i->free_secmap)) {
1733 if (left_start > 0) {
1734 left_start--;
1735 continue;
1736 }
1737 left_start = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1738 MAIN_SECS(sbi), 0);
1739 f2fs_bug_on(sbi, left_start >= MAIN_SECS(sbi));
351df4b2
JK		 1740 break;
1741 }
1742 secno = left_start;
1743skip_left:
1744 hint = secno;
4ddb1a4d
JK		 1745 segno = GET_SEG_FROM_SEC(sbi, secno);
1746 zoneno = GET_ZONE_FROM_SEC(sbi, secno);
351df4b2
JK		 1747
1748 /* give up on finding another zone */
1749 if (!init)
1750 goto got_it;
1751 if (sbi->secs_per_zone == 1)
1752 goto got_it;
1753 if (zoneno == old_zoneno)
1754 goto got_it;
1755 if (dir == ALLOC_LEFT) {
1756 if (!go_left && zoneno + 1 >= total_zones)
1757 goto got_it;
1758 if (go_left && zoneno == 0)
1759 goto got_it;
1760 }
1761 for (i = 0; i < NR_CURSEG_TYPE; i++)
1762 if (CURSEG_I(sbi, i)->zone == zoneno)
1763 break;
1764
1765 if (i < NR_CURSEG_TYPE) {
1766 /* zone is in user, try another */
1767 if (go_left)
1768 hint = zoneno * sbi->secs_per_zone - 1;
1769 else if (zoneno + 1 >= total_zones)
1770 hint = 0;
1771 else
1772 hint = (zoneno + 1) * sbi->secs_per_zone;
1773 init = false;
1774 goto find_other_zone;
1775 }
1776got_it:
1777 /* set it as dirty segment in free segmap */
9850cf4a 1778 f2fs_bug_on(sbi, test_bit(segno, free_i->free_segmap));
351df4b2
JK		 1779 __set_inuse(sbi, segno);
1780 *newseg = segno;
1a118ccf 1781 spin_unlock(&free_i->segmap_lock);
351df4b2
JK		 1782}
1783
1784static void reset_curseg(struct f2fs_sb_info *sbi, int type, int modified)
1785{
1786 struct curseg_info *curseg = CURSEG_I(sbi, type);
1787 struct summary_footer *sum_footer;
1788
1789 curseg->segno = curseg->next_segno;
4ddb1a4d 1790 curseg->zone = GET_ZONE_FROM_SEG(sbi, curseg->segno);
351df4b2
JK		 1791 curseg->next_blkoff = 0;
1792 curseg->next_segno = NULL_SEGNO;
1793
1794 sum_footer = &(curseg->sum_blk->footer);
1795 memset(sum_footer, 0, sizeof(struct summary_footer));
1796 if (IS_DATASEG(type))
1797 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
1798 if (IS_NODESEG(type))
1799 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
1800 __set_sit_entry_type(sbi, type, curseg->segno, modified);
1801}
1802
7a20b8a6
JK		 1803static unsigned int __get_next_segno(struct f2fs_sb_info *sbi, int type)
1804{
a7881893
JK		 1805 /* if segs_per_sec is large than 1, we need to keep original policy. */
1806 if (sbi->segs_per_sec != 1)
1807 return CURSEG_I(sbi, type)->segno;
1808
7a20b8a6
JK		 1809 if (type == CURSEG_HOT_DATA || IS_NODESEG(type))
1810 return 0;
1811
e066b83c
JK		 1812 if (SIT_I(sbi)->last_victim[ALLOC_NEXT])
1813 return SIT_I(sbi)->last_victim[ALLOC_NEXT];
7a20b8a6
JK		 1814 return CURSEG_I(sbi, type)->segno;
1815}
1816
0a8165d7 1817/*
351df4b2
JK		 1818 * Allocate a current working segment.
1819 * This function always allocates a free segment in LFS manner.
1820 */
1821static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
1822{
1823 struct curseg_info *curseg = CURSEG_I(sbi, type);
1824 unsigned int segno = curseg->segno;
1825 int dir = ALLOC_LEFT;
1826
1827 write_sum_page(sbi, curseg->sum_blk,
81fb5e87 1828 GET_SUM_BLOCK(sbi, segno));
351df4b2
JK		 1829 if (type == CURSEG_WARM_DATA || type == CURSEG_COLD_DATA)
1830 dir = ALLOC_RIGHT;
1831
1832 if (test_opt(sbi, NOHEAP))
1833 dir = ALLOC_RIGHT;
1834
7a20b8a6 1835 segno = __get_next_segno(sbi, type);
351df4b2
JK		 1836 get_new_segment(sbi, &segno, new_sec, dir);
1837 curseg->next_segno = segno;
1838 reset_curseg(sbi, type, 1);
1839 curseg->alloc_type = LFS;
1840}
1841
1842static void __next_free_blkoff(struct f2fs_sb_info *sbi,
1843 struct curseg_info *seg, block_t start)
1844{
1845 struct seg_entry *se = get_seg_entry(sbi, seg->segno);
e81c93cf 1846 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
60a3b782 1847 unsigned long *target_map = SIT_I(sbi)->tmp_map;
e81c93cf
CL		 1848 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
1849 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1850 int i, pos;
1851
1852 for (i = 0; i < entries; i++)
1853 target_map[i] = ckpt_map[i] | cur_map[i];
1854
1855 pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, start);
1856
1857 seg->next_blkoff = pos;
351df4b2
JK		 1858}
1859
0a8165d7 1860/*
351df4b2
JK		 1861 * If a segment is written by LFS manner, next block offset is just obtained
1862 * by increasing the current block offset. However, if a segment is written by
1863 * SSR manner, next block offset obtained by calling __next_free_blkoff
1864 */
1865static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
1866 struct curseg_info *seg)
1867{
1868 if (seg->alloc_type == SSR)
1869 __next_free_blkoff(sbi, seg, seg->next_blkoff + 1);
1870 else
1871 seg->next_blkoff++;
1872}
1873
0a8165d7 1874/*
e1c42045 1875 * This function always allocates a used segment(from dirty seglist) by SSR
351df4b2
JK		 1876 * manner, so it should recover the existing segment information of valid blocks
1877 */
1878static void change_curseg(struct f2fs_sb_info *sbi, int type, bool reuse)
1879{
1880 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
1881 struct curseg_info *curseg = CURSEG_I(sbi, type);
1882 unsigned int new_segno = curseg->next_segno;
1883 struct f2fs_summary_block *sum_node;
1884 struct page *sum_page;
1885
1886 write_sum_page(sbi, curseg->sum_blk,
1887 GET_SUM_BLOCK(sbi, curseg->segno));
1888 __set_test_and_inuse(sbi, new_segno);
1889
1890 mutex_lock(&dirty_i->seglist_lock);
1891 __remove_dirty_segment(sbi, new_segno, PRE);
1892 __remove_dirty_segment(sbi, new_segno, DIRTY);
1893 mutex_unlock(&dirty_i->seglist_lock);
1894
1895 reset_curseg(sbi, type, 1);
1896 curseg->alloc_type = SSR;
1897 __next_free_blkoff(sbi, curseg, 0);
1898
1899 if (reuse) {
1900 sum_page = get_sum_page(sbi, new_segno);
1901 sum_node = (struct f2fs_summary_block *)page_address(sum_page);
1902 memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
1903 f2fs_put_page(sum_page, 1);
1904 }
1905}
1906
43727527
JK		 1907static int get_ssr_segment(struct f2fs_sb_info *sbi, int type)
1908{
1909 struct curseg_info *curseg = CURSEG_I(sbi, type);
1910 const struct victim_selection *v_ops = DIRTY_I(sbi)->v_ops;
e066b83c 1911 unsigned segno = NULL_SEGNO;
d27c3d89
CY		 1912 int i, cnt;
1913 bool reversed = false;
c192f7a4
JK		 1914
1915 /* need_SSR() already forces to do this */
e066b83c
JK		 1916 if (v_ops->get_victim(sbi, &segno, BG_GC, type, SSR)) {
1917 curseg->next_segno = segno;
c192f7a4 1918 return 1;
e066b83c 1919 }
43727527 1920
70d625cb
JK		 1921 /* For node segments, let's do SSR more intensively */
1922 if (IS_NODESEG(type)) {
d27c3d89
CY		 1923 if (type >= CURSEG_WARM_NODE) {
1924 reversed = true;
1925 i = CURSEG_COLD_NODE;
1926 } else {
1927 i = CURSEG_HOT_NODE;
1928 }
1929 cnt = NR_CURSEG_NODE_TYPE;
70d625cb 1930 } else {
d27c3d89
CY		 1931 if (type >= CURSEG_WARM_DATA) {
1932 reversed = true;
1933 i = CURSEG_COLD_DATA;
1934 } else {
1935 i = CURSEG_HOT_DATA;
1936 }
1937 cnt = NR_CURSEG_DATA_TYPE;
70d625cb 1938 }
43727527 1939
d27c3d89 1940 for (; cnt-- > 0; reversed ? i-- : i++) {
c192f7a4
JK		 1941 if (i == type)
1942 continue;
e066b83c
JK		 1943 if (v_ops->get_victim(sbi, &segno, BG_GC, i, SSR)) {
1944 curseg->next_segno = segno;
43727527 1945 return 1;
e066b83c 1946 }
c192f7a4 1947 }
43727527
JK		 1948 return 0;
1949}
1950
351df4b2
JK		 1951/*
1952 * flush out current segment and replace it with new segment
1953 * This function should be returned with success, otherwise BUG
1954 */
1955static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
1956 int type, bool force)
1957{
a7881893
JK		 1958 struct curseg_info *curseg = CURSEG_I(sbi, type);
1959
7b405275 1960 if (force)
351df4b2 1961 new_curseg(sbi, type, true);
5b6c6be2
JK		 1962 else if (!is_set_ckpt_flags(sbi, CP_CRC_RECOVERY_FLAG) &&
1963 type == CURSEG_WARM_NODE)
351df4b2 1964 new_curseg(sbi, type, false);
a7881893
JK		 1965 else if (curseg->alloc_type == LFS && is_next_segment_free(sbi, type))
1966 new_curseg(sbi, type, false);
351df4b2
JK		 1967 else if (need_SSR(sbi) && get_ssr_segment(sbi, type))
1968 change_curseg(sbi, type, true);
1969 else
1970 new_curseg(sbi, type, false);
dcdfff65 1971
a7881893 1972 stat_inc_seg_type(sbi, curseg);
351df4b2
JK		 1973}
1974
1975void allocate_new_segments(struct f2fs_sb_info *sbi)
1976{
6ae1be13
JK		 1977 struct curseg_info *curseg;
1978 unsigned int old_segno;
351df4b2
JK		 1979 int i;
1980
6ae1be13
JK		 1981 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1982 curseg = CURSEG_I(sbi, i);
1983 old_segno = curseg->segno;
1984 SIT_I(sbi)->s_ops->allocate_segment(sbi, i, true);
1985 locate_dirty_segment(sbi, old_segno);
1986 }
351df4b2
JK		 1987}
1988
1989static const struct segment_allocation default_salloc_ops = {
1990 .allocate_segment = allocate_segment_by_default,
1991};
1992
25290fa5
JK		 1993bool exist_trim_candidates(struct f2fs_sb_info *sbi, struct cp_control *cpc)
1994{
1995 __u64 trim_start = cpc->trim_start;
1996 bool has_candidate = false;
1997
1998 mutex_lock(&SIT_I(sbi)->sentry_lock);
1999 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++) {
2000 if (add_discard_addrs(sbi, cpc, true)) {
2001 has_candidate = true;
2002 break;
2003 }
2004 }
2005 mutex_unlock(&SIT_I(sbi)->sentry_lock);
2006
2007 cpc->trim_start = trim_start;
2008 return has_candidate;
2009}
2010
4b2fecc8
JK		 2011int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
2012{
f7ef9b83
JK		 2013 __u64 start = F2FS_BYTES_TO_BLK(range->start);
2014 __u64 end = start + F2FS_BYTES_TO_BLK(range->len) - 1;
4b2fecc8
JK		 2015 unsigned int start_segno, end_segno;
2016 struct cp_control cpc;
c34f42e2 2017 int err = 0;
4b2fecc8 2018
836b5a63 2019 if (start >= MAX_BLKADDR(sbi) || range->len < sbi->blocksize)
4b2fecc8
JK		 2020 return -EINVAL;
2021
9bd27ae4 2022 cpc.trimmed = 0;
7cd8558b 2023 if (end <= MAIN_BLKADDR(sbi))
4b2fecc8
JK		 2024 goto out;
2025
ed214a11
YH		 2026 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
2027 f2fs_msg(sbi->sb, KERN_WARNING,
2028 "Found FS corruption, run fsck to fix.");
2029 goto out;
2030 }
2031
4b2fecc8 2032 /* start/end segment number in main_area */
7cd8558b
JK		 2033 start_segno = (start <= MAIN_BLKADDR(sbi)) ? 0 : GET_SEGNO(sbi, start);
2034 end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
2035 GET_SEGNO(sbi, end);
4b2fecc8 2036 cpc.reason = CP_DISCARD;
836b5a63 2037 cpc.trim_minlen = max_t(__u64, 1, F2FS_BYTES_TO_BLK(range->minlen));
4b2fecc8
JK		 2038
2039 /* do checkpoint to issue discard commands safely */
bba681cb
JK		 2040 for (; start_segno <= end_segno; start_segno = cpc.trim_end + 1) {
2041 cpc.trim_start = start_segno;
a66cdd98
JK		 2042
2043 if (sbi->discard_blks == 0)
2044 break;
2045 else if (sbi->discard_blks < BATCHED_TRIM_BLOCKS(sbi))
2046 cpc.trim_end = end_segno;
2047 else
2048 cpc.trim_end = min_t(unsigned int,
2049 rounddown(start_segno +
bba681cb
JK		 2050 BATCHED_TRIM_SEGMENTS(sbi),
2051 sbi->segs_per_sec) - 1, end_segno);
2052
2053 mutex_lock(&sbi->gc_mutex);
c34f42e2 2054 err = write_checkpoint(sbi, &cpc);
bba681cb 2055 mutex_unlock(&sbi->gc_mutex);
e9328353
CY		 2056 if (err)
2057 break;
74fa5f3d
CY		 2058
2059 schedule();
bba681cb 2060 }
4b2fecc8 2061out:
f7ef9b83 2062 range->len = F2FS_BLK_TO_BYTES(cpc.trimmed);
c34f42e2 2063 return err;
4b2fecc8
JK		 2064}
2065
351df4b2
JK		 2066static bool __has_curseg_space(struct f2fs_sb_info *sbi, int type)
2067{
2068 struct curseg_info *curseg = CURSEG_I(sbi, type);
2069 if (curseg->next_blkoff < sbi->blocks_per_seg)
2070 return true;
2071 return false;
2072}
2073
81377bd6 2074static int __get_segment_type_2(struct f2fs_io_info *fio)
351df4b2 2075{
81377bd6 2076 if (fio->type == DATA)
351df4b2
JK		 2077 return CURSEG_HOT_DATA;
2078 else
2079 return CURSEG_HOT_NODE;
2080}
2081
81377bd6 2082static int __get_segment_type_4(struct f2fs_io_info *fio)
351df4b2 2083{
81377bd6
JK		 2084 if (fio->type == DATA) {
2085 struct inode *inode = fio->page->mapping->host;
351df4b2
JK		 2086
2087 if (S_ISDIR(inode->i_mode))
2088 return CURSEG_HOT_DATA;
2089 else
2090 return CURSEG_COLD_DATA;
2091 } else {
81377bd6 2092 if (IS_DNODE(fio->page) && is_cold_node(fio->page))
a344b9fd 2093 return CURSEG_WARM_NODE;
351df4b2
JK		 2094 else
2095 return CURSEG_COLD_NODE;
2096 }
2097}
2098
81377bd6 2099static int __get_segment_type_6(struct f2fs_io_info *fio)
351df4b2 2100{
81377bd6
JK		 2101 if (fio->type == DATA) {
2102 struct inode *inode = fio->page->mapping->host;
351df4b2 2103
81377bd6 2104 if (is_cold_data(fio->page) || file_is_cold(inode))
351df4b2 2105 return CURSEG_COLD_DATA;
ef095d19
JK		 2106 if (is_inode_flag_set(inode, FI_HOT_DATA))
2107 return CURSEG_HOT_DATA;
2108 return CURSEG_WARM_DATA;
351df4b2 2109 } else {
81377bd6
JK		 2110 if (IS_DNODE(fio->page))
2111 return is_cold_node(fio->page) ? CURSEG_WARM_NODE :
351df4b2 2112 CURSEG_HOT_NODE;
ef095d19 2113 return CURSEG_COLD_NODE;
351df4b2
JK		 2114 }
2115}
2116
81377bd6 2117static int __get_segment_type(struct f2fs_io_info *fio)
351df4b2 2118{
a912b54d
JK		 2119 int type = 0;
2120
81377bd6 2121 switch (fio->sbi->active_logs) {
351df4b2 2122 case 2:
a912b54d
JK		 2123 type = __get_segment_type_2(fio);
2124 break;
351df4b2 2125 case 4:
a912b54d
JK		 2126 type = __get_segment_type_4(fio);
2127 break;
2128 case 6:
2129 type = __get_segment_type_6(fio);
2130 break;
2131 default:
2132 f2fs_bug_on(fio->sbi, true);
351df4b2 2133 }
81377bd6 2134
a912b54d
JK		 2135 if (IS_HOT(type))
2136 fio->temp = HOT;
2137 else if (IS_WARM(type))
2138 fio->temp = WARM;
2139 else
2140 fio->temp = COLD;
2141 return type;
351df4b2
JK		 2142}
2143
bfad7c2d
JK		 2144void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
2145 block_t old_blkaddr, block_t *new_blkaddr,
fb830fc5
CY		 2146 struct f2fs_summary *sum, int type,
2147 struct f2fs_io_info *fio, bool add_list)
351df4b2
JK		 2148{
2149 struct sit_info *sit_i = SIT_I(sbi);
6ae1be13 2150 struct curseg_info *curseg = CURSEG_I(sbi, type);
351df4b2
JK		 2151
2152 mutex_lock(&curseg->curseg_mutex);
21cb1d99 2153 mutex_lock(&sit_i->sentry_lock);
351df4b2
JK		 2154
2155 *new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
351df4b2 2156
4e6a8d9b
JK		 2157 f2fs_wait_discard_bio(sbi, *new_blkaddr);
2158
351df4b2
JK		 2159 /*
2160 * __add_sum_entry should be resided under the curseg_mutex
2161 * because, this function updates a summary entry in the
2162 * current summary block.
2163 */
e79efe3b 2164 __add_sum_entry(sbi, type, sum);
351df4b2 2165
351df4b2 2166 __refresh_next_blkoff(sbi, curseg);
dcdfff65
JK		 2167
2168 stat_inc_block_count(sbi, curseg);
351df4b2 2169
c6f82fe9
JK		 2170 if (!__has_curseg_space(sbi, type))
2171 sit_i->s_ops->allocate_segment(sbi, type, false);
351df4b2 2172 /*
c6f82fe9
JK		 2173 * SIT information should be updated after segment allocation,
2174 * since we need to keep dirty segments precisely under SSR.
351df4b2
JK		 2175 */
2176 refresh_sit_entry(sbi, old_blkaddr, *new_blkaddr);
5e443818 2177
351df4b2
JK		 2178 mutex_unlock(&sit_i->sentry_lock);
2179
bfad7c2d 2180 if (page && IS_NODESEG(type))
351df4b2
JK		 2181 fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
2182
fb830fc5
CY		 2183 if (add_list) {
2184 struct f2fs_bio_info *io;
2185
2186 INIT_LIST_HEAD(&fio->list);
2187 fio->in_list = true;
2188 io = sbi->write_io[fio->type] + fio->temp;
2189 spin_lock(&io->io_lock);
2190 list_add_tail(&fio->list, &io->io_list);
2191 spin_unlock(&io->io_lock);
2192 }
2193
bfad7c2d
JK		 2194 mutex_unlock(&curseg->curseg_mutex);
2195}
2196
05ca3632 2197static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
bfad7c2d 2198{
81377bd6 2199 int type = __get_segment_type(fio);
0a595eba 2200 int err;
bfad7c2d 2201
0a595eba 2202reallocate:
7a9d7548 2203 allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
fb830fc5 2204 &fio->new_blkaddr, sum, type, fio, true);
bfad7c2d 2205
351df4b2 2206 /* writeout dirty page into bdev */
b9109b0e 2207 err = f2fs_submit_page_write(fio);
0a595eba
JK		 2208 if (err == -EAGAIN) {
2209 fio->old_blkaddr = fio->new_blkaddr;
2210 goto reallocate;
2211 }
351df4b2
JK		 2212}
2213
577e3495 2214void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
351df4b2 2215{
458e6197 2216 struct f2fs_io_info fio = {
05ca3632 2217 .sbi = sbi,
458e6197 2218 .type = META,
04d328de 2219 .op = REQ_OP_WRITE,
70fd7614 2220 .op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
7a9d7548
CY		 2221 .old_blkaddr = page->index,
2222 .new_blkaddr = page->index,
05ca3632 2223 .page = page,
4375a336 2224 .encrypted_page = NULL,
fb830fc5 2225 .in_list = false,
458e6197
JK		 2226 };
2227
2b947003 2228 if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
04d328de 2229 fio.op_flags &= ~REQ_META;
2b947003 2230
351df4b2 2231 set_page_writeback(page);
b9109b0e 2232 f2fs_submit_page_write(&fio);
351df4b2
JK		 2233}
2234
05ca3632 2235void write_node_page(unsigned int nid, struct f2fs_io_info *fio)
351df4b2
JK		 2236{
2237 struct f2fs_summary sum;
05ca3632 2238
351df4b2 2239 set_summary(&sum, nid, 0, 0);
05ca3632 2240 do_write_page(&sum, fio);
351df4b2
JK		 2241}
2242
05ca3632 2243void write_data_page(struct dnode_of_data *dn, struct f2fs_io_info *fio)
351df4b2 2244{
05ca3632 2245 struct f2fs_sb_info *sbi = fio->sbi;
351df4b2
JK		 2246 struct f2fs_summary sum;
2247 struct node_info ni;
2248
9850cf4a 2249 f2fs_bug_on(sbi, dn->data_blkaddr == NULL_ADDR);
351df4b2
JK		 2250 get_node_info(sbi, dn->nid, &ni);
2251 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
05ca3632 2252 do_write_page(&sum, fio);
f28b3434 2253 f2fs_update_data_blkaddr(dn, fio->new_blkaddr);
351df4b2
JK		 2254}
2255
d1b3e72d 2256int rewrite_data_page(struct f2fs_io_info *fio)
351df4b2 2257{
7a9d7548 2258 fio->new_blkaddr = fio->old_blkaddr;
05ca3632 2259 stat_inc_inplace_blocks(fio->sbi);
d1b3e72d 2260 return f2fs_submit_page_bio(fio);
351df4b2
JK		 2261}
2262
4356e48e 2263void __f2fs_replace_block(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
19f106bc 2264 block_t old_blkaddr, block_t new_blkaddr,
28bc106b 2265 bool recover_curseg, bool recover_newaddr)
351df4b2
JK		 2266{
2267 struct sit_info *sit_i = SIT_I(sbi);
2268 struct curseg_info *curseg;
2269 unsigned int segno, old_cursegno;
2270 struct seg_entry *se;
2271 int type;
19f106bc 2272 unsigned short old_blkoff;
351df4b2
JK		 2273
2274 segno = GET_SEGNO(sbi, new_blkaddr);
2275 se = get_seg_entry(sbi, segno);
2276 type = se->type;
2277
19f106bc
CY		 2278 if (!recover_curseg) {
2279 /* for recovery flow */
2280 if (se->valid_blocks == 0 && !IS_CURSEG(sbi, segno)) {
2281 if (old_blkaddr == NULL_ADDR)
2282 type = CURSEG_COLD_DATA;
2283 else
2284 type = CURSEG_WARM_DATA;
2285 }
2286 } else {
2287 if (!IS_CURSEG(sbi, segno))
351df4b2
JK		 2288 type = CURSEG_WARM_DATA;
2289 }
19f106bc 2290
351df4b2
JK		 2291 curseg = CURSEG_I(sbi, type);
2292
2293 mutex_lock(&curseg->curseg_mutex);
2294 mutex_lock(&sit_i->sentry_lock);
2295
2296 old_cursegno = curseg->segno;
19f106bc 2297 old_blkoff = curseg->next_blkoff;
351df4b2
JK		 2298
2299 /* change the current segment */
2300 if (segno != curseg->segno) {
2301 curseg->next_segno = segno;
2302 change_curseg(sbi, type, true);
2303 }
2304
491c0854 2305 curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
e79efe3b 2306 __add_sum_entry(sbi, type, sum);
351df4b2 2307
28bc106b 2308 if (!recover_curseg || recover_newaddr)
6e2c64ad
JK		 2309 update_sit_entry(sbi, new_blkaddr, 1);
2310 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
2311 update_sit_entry(sbi, old_blkaddr, -1);
2312
2313 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
2314 locate_dirty_segment(sbi, GET_SEGNO(sbi, new_blkaddr));
2315
351df4b2 2316 locate_dirty_segment(sbi, old_cursegno);
351df4b2 2317
19f106bc
CY		 2318 if (recover_curseg) {
2319 if (old_cursegno != curseg->segno) {
2320 curseg->next_segno = old_cursegno;
2321 change_curseg(sbi, type, true);
2322 }
2323 curseg->next_blkoff = old_blkoff;
2324 }
2325
351df4b2
JK		 2326 mutex_unlock(&sit_i->sentry_lock);
2327 mutex_unlock(&curseg->curseg_mutex);
2328}
2329
528e3459
CY		 2330void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
2331 block_t old_addr, block_t new_addr,
28bc106b
CY		 2332 unsigned char version, bool recover_curseg,
2333 bool recover_newaddr)
528e3459
CY		 2334{
2335 struct f2fs_summary sum;
2336
2337 set_summary(&sum, dn->nid, dn->ofs_in_node, version);
2338
28bc106b
CY		 2339 __f2fs_replace_block(sbi, &sum, old_addr, new_addr,
2340 recover_curseg, recover_newaddr);
528e3459 2341
f28b3434 2342 f2fs_update_data_blkaddr(dn, new_addr);
528e3459
CY		 2343}
2344
93dfe2ac 2345void f2fs_wait_on_page_writeback(struct page *page,
fec1d657 2346 enum page_type type, bool ordered)
93dfe2ac 2347{
93dfe2ac 2348 if (PageWriteback(page)) {
4081363f
JK		 2349 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
2350
b9109b0e
JK		 2351 f2fs_submit_merged_write_cond(sbi, page->mapping->host,
2352 0, page->index, type);
fec1d657
JK		 2353 if (ordered)
2354 wait_on_page_writeback(page);
2355 else
2356 wait_for_stable_page(page);
93dfe2ac
JK		 2357 }
2358}
2359
08b39fbd
CY		 2360void f2fs_wait_on_encrypted_page_writeback(struct f2fs_sb_info *sbi,
2361 block_t blkaddr)
2362{
2363 struct page *cpage;
2364
5d4c0af4 2365 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
08b39fbd
CY		 2366 return;
2367
08b39fbd
CY		 2368 cpage = find_lock_page(META_MAPPING(sbi), blkaddr);
2369 if (cpage) {
fec1d657 2370 f2fs_wait_on_page_writeback(cpage, DATA, true);
08b39fbd
CY		 2371 f2fs_put_page(cpage, 1);
2372 }
2373}
2374
351df4b2
JK		 2375static int read_compacted_summaries(struct f2fs_sb_info *sbi)
2376{
2377 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2378 struct curseg_info *seg_i;
2379 unsigned char *kaddr;
2380 struct page *page;
2381 block_t start;
2382 int i, j, offset;
2383
2384 start = start_sum_block(sbi);
2385
2386 page = get_meta_page(sbi, start++);
2387 kaddr = (unsigned char *)page_address(page);
2388
2389 /* Step 1: restore nat cache */
2390 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2391 memcpy(seg_i->journal, kaddr, SUM_JOURNAL_SIZE);
351df4b2
JK		 2392
2393 /* Step 2: restore sit cache */
2394 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2395 memcpy(seg_i->journal, kaddr + SUM_JOURNAL_SIZE, SUM_JOURNAL_SIZE);
351df4b2
JK		 2396 offset = 2 * SUM_JOURNAL_SIZE;
2397
2398 /* Step 3: restore summary entries */
2399 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2400 unsigned short blk_off;
2401 unsigned int segno;
2402
2403 seg_i = CURSEG_I(sbi, i);
2404 segno = le32_to_cpu(ckpt->cur_data_segno[i]);
2405 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[i]);
2406 seg_i->next_segno = segno;
2407 reset_curseg(sbi, i, 0);
2408 seg_i->alloc_type = ckpt->alloc_type[i];
2409 seg_i->next_blkoff = blk_off;
2410
2411 if (seg_i->alloc_type == SSR)
2412 blk_off = sbi->blocks_per_seg;
2413
2414 for (j = 0; j < blk_off; j++) {
2415 struct f2fs_summary *s;
2416 s = (struct f2fs_summary *)(kaddr + offset);
2417 seg_i->sum_blk->entries[j] = *s;
2418 offset += SUMMARY_SIZE;
09cbfeaf 2419 if (offset + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2420 SUM_FOOTER_SIZE)
2421 continue;
2422
2423 f2fs_put_page(page, 1);
2424 page = NULL;
2425
2426 page = get_meta_page(sbi, start++);
2427 kaddr = (unsigned char *)page_address(page);
2428 offset = 0;
2429 }
2430 }
2431 f2fs_put_page(page, 1);
2432 return 0;
2433}
2434
2435static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
2436{
2437 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2438 struct f2fs_summary_block *sum;
2439 struct curseg_info *curseg;
2440 struct page *new;
2441 unsigned short blk_off;
2442 unsigned int segno = 0;
2443 block_t blk_addr = 0;
2444
2445 /* get segment number and block addr */
2446 if (IS_DATASEG(type)) {
2447 segno = le32_to_cpu(ckpt->cur_data_segno[type]);
2448 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[type -
2449 CURSEG_HOT_DATA]);
119ee914 2450 if (__exist_node_summaries(sbi))
351df4b2
JK		 2451 blk_addr = sum_blk_addr(sbi, NR_CURSEG_TYPE, type);
2452 else
2453 blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
2454 } else {
2455 segno = le32_to_cpu(ckpt->cur_node_segno[type -
2456 CURSEG_HOT_NODE]);
2457 blk_off = le16_to_cpu(ckpt->cur_node_blkoff[type -
2458 CURSEG_HOT_NODE]);
119ee914 2459 if (__exist_node_summaries(sbi))
351df4b2
JK		 2460 blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
2461 type - CURSEG_HOT_NODE);
2462 else
2463 blk_addr = GET_SUM_BLOCK(sbi, segno);
2464 }
2465
2466 new = get_meta_page(sbi, blk_addr);
2467 sum = (struct f2fs_summary_block *)page_address(new);
2468
2469 if (IS_NODESEG(type)) {
119ee914 2470 if (__exist_node_summaries(sbi)) {
351df4b2
JK		 2471 struct f2fs_summary *ns = &sum->entries[0];
2472 int i;
2473 for (i = 0; i < sbi->blocks_per_seg; i++, ns++) {
2474 ns->version = 0;
2475 ns->ofs_in_node = 0;
2476 }
2477 } else {
d653788a
GZ		 2478 int err;
2479
2480 err = restore_node_summary(sbi, segno, sum);
2481 if (err) {
351df4b2 2482 f2fs_put_page(new, 1);
d653788a 2483 return err;
351df4b2
JK		 2484 }
2485 }
2486 }
2487
2488 /* set uncompleted segment to curseg */
2489 curseg = CURSEG_I(sbi, type);
2490 mutex_lock(&curseg->curseg_mutex);
b7ad7512
CY		 2491
2492 /* update journal info */
2493 down_write(&curseg->journal_rwsem);
2494 memcpy(curseg->journal, &sum->journal, SUM_JOURNAL_SIZE);
2495 up_write(&curseg->journal_rwsem);
2496
2497 memcpy(curseg->sum_blk->entries, sum->entries, SUM_ENTRY_SIZE);
2498 memcpy(&curseg->sum_blk->footer, &sum->footer, SUM_FOOTER_SIZE);
351df4b2
JK		 2499 curseg->next_segno = segno;
2500 reset_curseg(sbi, type, 0);
2501 curseg->alloc_type = ckpt->alloc_type[type];
2502 curseg->next_blkoff = blk_off;
2503 mutex_unlock(&curseg->curseg_mutex);
2504 f2fs_put_page(new, 1);
2505 return 0;
2506}
2507
2508static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
2509{
21d3f8e1
JQ		 2510 struct f2fs_journal *sit_j = CURSEG_I(sbi, CURSEG_COLD_DATA)->journal;
2511 struct f2fs_journal *nat_j = CURSEG_I(sbi, CURSEG_HOT_DATA)->journal;
351df4b2 2512 int type = CURSEG_HOT_DATA;
e4fc5fbf 2513 int err;
351df4b2 2514
aaec2b1d 2515 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG)) {
3fa06d7b
CY		 2516 int npages = npages_for_summary_flush(sbi, true);
2517
2518 if (npages >= 2)
2519 ra_meta_pages(sbi, start_sum_block(sbi), npages,
26879fb1 2520 META_CP, true);
3fa06d7b 2521
351df4b2
JK		 2522 /* restore for compacted data summary */
2523 if (read_compacted_summaries(sbi))
2524 return -EINVAL;
2525 type = CURSEG_HOT_NODE;
2526 }
2527
119ee914 2528 if (__exist_node_summaries(sbi))
3fa06d7b 2529 ra_meta_pages(sbi, sum_blk_addr(sbi, NR_CURSEG_TYPE, type),
26879fb1 2530 NR_CURSEG_TYPE - type, META_CP, true);
3fa06d7b 2531
e4fc5fbf
CY		 2532 for (; type <= CURSEG_COLD_NODE; type++) {
2533 err = read_normal_summaries(sbi, type);
2534 if (err)
2535 return err;
2536 }
2537
21d3f8e1
JQ		 2538 /* sanity check for summary blocks */
2539 if (nats_in_cursum(nat_j) > NAT_JOURNAL_ENTRIES ||
2540 sits_in_cursum(sit_j) > SIT_JOURNAL_ENTRIES)
2541 return -EINVAL;
2542
351df4b2
JK		 2543 return 0;
2544}
2545
2546static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
2547{
2548 struct page *page;
2549 unsigned char *kaddr;
2550 struct f2fs_summary *summary;
2551 struct curseg_info *seg_i;
2552 int written_size = 0;
2553 int i, j;
2554
2555 page = grab_meta_page(sbi, blkaddr++);
2556 kaddr = (unsigned char *)page_address(page);
2557
2558 /* Step 1: write nat cache */
2559 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2560 memcpy(kaddr, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2561 written_size += SUM_JOURNAL_SIZE;
2562
2563 /* Step 2: write sit cache */
2564 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2565 memcpy(kaddr + written_size, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2566 written_size += SUM_JOURNAL_SIZE;
2567
351df4b2
JK		 2568 /* Step 3: write summary entries */
2569 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2570 unsigned short blkoff;
2571 seg_i = CURSEG_I(sbi, i);
2572 if (sbi->ckpt->alloc_type[i] == SSR)
2573 blkoff = sbi->blocks_per_seg;
2574 else
2575 blkoff = curseg_blkoff(sbi, i);
2576
2577 for (j = 0; j < blkoff; j++) {
2578 if (!page) {
2579 page = grab_meta_page(sbi, blkaddr++);
2580 kaddr = (unsigned char *)page_address(page);
2581 written_size = 0;
2582 }
2583 summary = (struct f2fs_summary *)(kaddr + written_size);
2584 *summary = seg_i->sum_blk->entries[j];
2585 written_size += SUMMARY_SIZE;
351df4b2 2586
09cbfeaf 2587 if (written_size + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2588 SUM_FOOTER_SIZE)
2589 continue;
2590
e8d61a74 2591 set_page_dirty(page);
351df4b2
JK		 2592 f2fs_put_page(page, 1);
2593 page = NULL;
2594 }
2595 }
e8d61a74
CY		 2596 if (page) {
2597 set_page_dirty(page);
351df4b2 2598 f2fs_put_page(page, 1);
e8d61a74 2599 }
351df4b2
JK		 2600}
2601
2602static void write_normal_summaries(struct f2fs_sb_info *sbi,
2603 block_t blkaddr, int type)
2604{
2605 int i, end;
2606 if (IS_DATASEG(type))
2607 end = type + NR_CURSEG_DATA_TYPE;
2608 else
2609 end = type + NR_CURSEG_NODE_TYPE;
2610
b7ad7512
CY		 2611 for (i = type; i < end; i++)
2612 write_current_sum_page(sbi, i, blkaddr + (i - type));
351df4b2
JK		 2613}
2614
2615void write_data_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2616{
aaec2b1d 2617 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG))
351df4b2
JK		 2618 write_compacted_summaries(sbi, start_blk);
2619 else
2620 write_normal_summaries(sbi, start_blk, CURSEG_HOT_DATA);
2621}
2622
2623void write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2624{
119ee914 2625 write_normal_summaries(sbi, start_blk, CURSEG_HOT_NODE);
351df4b2
JK		 2626}
2627
dfc08a12 2628int lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
351df4b2
JK		 2629 unsigned int val, int alloc)
2630{
2631 int i;
2632
2633 if (type == NAT_JOURNAL) {
dfc08a12
CY		 2634 for (i = 0; i < nats_in_cursum(journal); i++) {
2635 if (le32_to_cpu(nid_in_journal(journal, i)) == val)
351df4b2
JK		 2636 return i;
2637 }
dfc08a12
CY		 2638 if (alloc && __has_cursum_space(journal, 1, NAT_JOURNAL))
2639 return update_nats_in_cursum(journal, 1);
351df4b2 2640 } else if (type == SIT_JOURNAL) {
dfc08a12
CY		 2641 for (i = 0; i < sits_in_cursum(journal); i++)
2642 if (le32_to_cpu(segno_in_journal(journal, i)) == val)
351df4b2 2643 return i;
dfc08a12
CY		 2644 if (alloc && __has_cursum_space(journal, 1, SIT_JOURNAL))
2645 return update_sits_in_cursum(journal, 1);
351df4b2
JK		 2646 }
2647 return -1;
2648}
2649
2650static struct page *get_current_sit_page(struct f2fs_sb_info *sbi,
2651 unsigned int segno)
2652{
2cc22186 2653 return get_meta_page(sbi, current_sit_addr(sbi, segno));
351df4b2
JK		 2654}
2655
2656static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
2657 unsigned int start)
2658{
2659 struct sit_info *sit_i = SIT_I(sbi);
2660 struct page *src_page, *dst_page;
2661 pgoff_t src_off, dst_off;
2662 void *src_addr, *dst_addr;
2663
2664 src_off = current_sit_addr(sbi, start);
2665 dst_off = next_sit_addr(sbi, src_off);
2666
2667 /* get current sit block page without lock */
2668 src_page = get_meta_page(sbi, src_off);
2669 dst_page = grab_meta_page(sbi, dst_off);
9850cf4a 2670 f2fs_bug_on(sbi, PageDirty(src_page));
351df4b2
JK		 2671
2672 src_addr = page_address(src_page);
2673 dst_addr = page_address(dst_page);
09cbfeaf 2674 memcpy(dst_addr, src_addr, PAGE_SIZE);
351df4b2
JK		 2675
2676 set_page_dirty(dst_page);
2677 f2fs_put_page(src_page, 1);
2678
2679 set_to_next_sit(sit_i, start);
2680
2681 return dst_page;
2682}
2683
184a5cd2
CY		 2684static struct sit_entry_set *grab_sit_entry_set(void)
2685{
2686 struct sit_entry_set *ses =
80c54505 2687 f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS);
184a5cd2
CY		 2688
2689 ses->entry_cnt = 0;
2690 INIT_LIST_HEAD(&ses->set_list);
2691 return ses;
2692}
2693
2694static void release_sit_entry_set(struct sit_entry_set *ses)
2695{
2696 list_del(&ses->set_list);
2697 kmem_cache_free(sit_entry_set_slab, ses);
2698}
2699
2700static void adjust_sit_entry_set(struct sit_entry_set *ses,
2701 struct list_head *head)
2702{
2703 struct sit_entry_set *next = ses;
2704
2705 if (list_is_last(&ses->set_list, head))
2706 return;
2707
2708 list_for_each_entry_continue(next, head, set_list)
2709 if (ses->entry_cnt <= next->entry_cnt)
2710 break;
2711
2712 list_move_tail(&ses->set_list, &next->set_list);
2713}
2714
2715static void add_sit_entry(unsigned int segno, struct list_head *head)
2716{
2717 struct sit_entry_set *ses;
2718 unsigned int start_segno = START_SEGNO(segno);
2719
2720 list_for_each_entry(ses, head, set_list) {
2721 if (ses->start_segno == start_segno) {
2722 ses->entry_cnt++;
2723 adjust_sit_entry_set(ses, head);
2724 return;
2725 }
2726 }
2727
2728 ses = grab_sit_entry_set();
2729
2730 ses->start_segno = start_segno;
2731 ses->entry_cnt++;
2732 list_add(&ses->set_list, head);
2733}
2734
2735static void add_sits_in_set(struct f2fs_sb_info *sbi)
2736{
2737 struct f2fs_sm_info *sm_info = SM_I(sbi);
2738 struct list_head *set_list = &sm_info->sit_entry_set;
2739 unsigned long *bitmap = SIT_I(sbi)->dirty_sentries_bitmap;
184a5cd2
CY		 2740 unsigned int segno;
2741
7cd8558b 2742 for_each_set_bit(segno, bitmap, MAIN_SEGS(sbi))
184a5cd2
CY		 2743 add_sit_entry(segno, set_list);
2744}
2745
2746static void remove_sits_in_journal(struct f2fs_sb_info *sbi)
351df4b2
JK		 2747{
2748 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2749 struct f2fs_journal *journal = curseg->journal;
351df4b2
JK		 2750 int i;
2751
b7ad7512 2752 down_write(&curseg->journal_rwsem);
dfc08a12 2753 for (i = 0; i < sits_in_cursum(journal); i++) {
184a5cd2
CY		 2754 unsigned int segno;
2755 bool dirtied;
2756
dfc08a12 2757 segno = le32_to_cpu(segno_in_journal(journal, i));
184a5cd2
CY		 2758 dirtied = __mark_sit_entry_dirty(sbi, segno);
2759
2760 if (!dirtied)
2761 add_sit_entry(segno, &SM_I(sbi)->sit_entry_set);
351df4b2 2762 }
dfc08a12 2763 update_sits_in_cursum(journal, -i);
b7ad7512 2764 up_write(&curseg->journal_rwsem);
351df4b2
JK		 2765}
2766
0a8165d7 2767/*
351df4b2
JK		 2768 * CP calls this function, which flushes SIT entries including sit_journal,
2769 * and moves prefree segs to free segs.
2770 */
4b2fecc8 2771void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2
JK		 2772{
2773 struct sit_info *sit_i = SIT_I(sbi);
2774 unsigned long *bitmap = sit_i->dirty_sentries_bitmap;
2775 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2776 struct f2fs_journal *journal = curseg->journal;
184a5cd2
CY		 2777 struct sit_entry_set *ses, *tmp;
2778 struct list_head *head = &SM_I(sbi)->sit_entry_set;
184a5cd2 2779 bool to_journal = true;
4b2fecc8 2780 struct seg_entry *se;
351df4b2 2781
351df4b2
JK		 2782 mutex_lock(&sit_i->sentry_lock);
2783
2b11a74b
WL		 2784 if (!sit_i->dirty_sentries)
2785 goto out;
2786
351df4b2 2787 /*
184a5cd2
CY		 2788 * add and account sit entries of dirty bitmap in sit entry
2789 * set temporarily
351df4b2 2790 */
184a5cd2 2791 add_sits_in_set(sbi);
351df4b2 2792
184a5cd2
CY		 2793 /*
2794 * if there are no enough space in journal to store dirty sit
2795 * entries, remove all entries from journal and add and account
2796 * them in sit entry set.
2797 */
dfc08a12 2798 if (!__has_cursum_space(journal, sit_i->dirty_sentries, SIT_JOURNAL))
184a5cd2 2799 remove_sits_in_journal(sbi);
b2955550 2800
184a5cd2
CY		 2801 /*
2802 * there are two steps to flush sit entries:
2803 * #1, flush sit entries to journal in current cold data summary block.
2804 * #2, flush sit entries to sit page.
2805 */
2806 list_for_each_entry_safe(ses, tmp, head, set_list) {
4a257ed6 2807 struct page *page = NULL;
184a5cd2
CY		 2808 struct f2fs_sit_block *raw_sit = NULL;
2809 unsigned int start_segno = ses->start_segno;
2810 unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
7cd8558b 2811 (unsigned long)MAIN_SEGS(sbi));
184a5cd2
CY		 2812 unsigned int segno = start_segno;
2813
2814 if (to_journal &&
dfc08a12 2815 !__has_cursum_space(journal, ses->entry_cnt, SIT_JOURNAL))
184a5cd2
CY		 2816 to_journal = false;
2817
b7ad7512
CY		 2818 if (to_journal) {
2819 down_write(&curseg->journal_rwsem);
2820 } else {
184a5cd2
CY		 2821 page = get_next_sit_page(sbi, start_segno);
2822 raw_sit = page_address(page);
351df4b2 2823 }
351df4b2 2824
184a5cd2
CY		 2825 /* flush dirty sit entries in region of current sit set */
2826 for_each_set_bit_from(segno, bitmap, end) {
2827 int offset, sit_offset;
4b2fecc8
JK		 2828
2829 se = get_seg_entry(sbi, segno);
184a5cd2
CY		 2830
2831 /* add discard candidates */
c473f1a9 2832 if (!(cpc->reason & CP_DISCARD)) {
4b2fecc8 2833 cpc->trim_start = segno;
25290fa5 2834 add_discard_addrs(sbi, cpc, false);
4b2fecc8 2835 }
184a5cd2
CY		 2836
2837 if (to_journal) {
dfc08a12 2838 offset = lookup_journal_in_cursum(journal,
184a5cd2
CY		 2839 SIT_JOURNAL, segno, 1);
2840 f2fs_bug_on(sbi, offset < 0);
dfc08a12 2841 segno_in_journal(journal, offset) =
184a5cd2
CY		 2842 cpu_to_le32(segno);
2843 seg_info_to_raw_sit(se,
dfc08a12 2844 &sit_in_journal(journal, offset));
184a5cd2
CY		 2845 } else {
2846 sit_offset = SIT_ENTRY_OFFSET(sit_i, segno);
2847 seg_info_to_raw_sit(se,
2848 &raw_sit->entries[sit_offset]);
2849 }
351df4b2 2850
184a5cd2
CY		 2851 __clear_bit(segno, bitmap);
2852 sit_i->dirty_sentries--;
2853 ses->entry_cnt--;
351df4b2
JK		 2854 }
2855
b7ad7512
CY		 2856 if (to_journal)
2857 up_write(&curseg->journal_rwsem);
2858 else
184a5cd2
CY		 2859 f2fs_put_page(page, 1);
2860
2861 f2fs_bug_on(sbi, ses->entry_cnt);
2862 release_sit_entry_set(ses);
351df4b2 2863 }
184a5cd2
CY		 2864
2865 f2fs_bug_on(sbi, !list_empty(head));
2866 f2fs_bug_on(sbi, sit_i->dirty_sentries);
184a5cd2 2867out:
c473f1a9 2868 if (cpc->reason & CP_DISCARD) {
650d3c4e
YH		 2869 __u64 trim_start = cpc->trim_start;
2870
4b2fecc8 2871 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++)
25290fa5 2872 add_discard_addrs(sbi, cpc, false);
650d3c4e
YH		 2873
2874 cpc->trim_start = trim_start;
4b2fecc8 2875 }
351df4b2 2876 mutex_unlock(&sit_i->sentry_lock);
351df4b2 2877
351df4b2
JK		 2878 set_prefree_as_free_segments(sbi);
2879}
2880
2881static int build_sit_info(struct f2fs_sb_info *sbi)
2882{
2883 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
351df4b2
JK		 2884 struct sit_info *sit_i;
2885 unsigned int sit_segs, start;
ae27d62e 2886 char *src_bitmap;
351df4b2
JK		 2887 unsigned int bitmap_size;
2888
2889 /* allocate memory for SIT information */
2890 sit_i = kzalloc(sizeof(struct sit_info), GFP_KERNEL);
2891 if (!sit_i)
2892 return -ENOMEM;
2893
2894 SM_I(sbi)->sit_info = sit_i;
2895
39307a8e
JK		 2896 sit_i->sentries = f2fs_kvzalloc(MAIN_SEGS(sbi) *
2897 sizeof(struct seg_entry), GFP_KERNEL);
351df4b2
JK		 2898 if (!sit_i->sentries)
2899 return -ENOMEM;
2900
7cd8558b 2901 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2902 sit_i->dirty_sentries_bitmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2903 if (!sit_i->dirty_sentries_bitmap)
2904 return -ENOMEM;
2905
7cd8558b 2906 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 2907 sit_i->sentries[start].cur_valid_map
2908 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2909 sit_i->sentries[start].ckpt_valid_map
2910 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
a66cdd98 2911 if (!sit_i->sentries[start].cur_valid_map ||
3e025740 2912 !sit_i->sentries[start].ckpt_valid_map)
351df4b2 2913 return -ENOMEM;
3e025740 2914
355e7891
CY		 2915#ifdef CONFIG_F2FS_CHECK_FS
2916 sit_i->sentries[start].cur_valid_map_mir
2917 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2918 if (!sit_i->sentries[start].cur_valid_map_mir)
2919 return -ENOMEM;
2920#endif
2921
3e025740
JK		 2922 if (f2fs_discard_en(sbi)) {
2923 sit_i->sentries[start].discard_map
2924 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2925 if (!sit_i->sentries[start].discard_map)
2926 return -ENOMEM;
2927 }
351df4b2
JK		 2928 }
2929
60a3b782
JK		 2930 sit_i->tmp_map = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2931 if (!sit_i->tmp_map)
2932 return -ENOMEM;
2933
351df4b2 2934 if (sbi->segs_per_sec > 1) {
39307a8e
JK		 2935 sit_i->sec_entries = f2fs_kvzalloc(MAIN_SECS(sbi) *
2936 sizeof(struct sec_entry), GFP_KERNEL);
351df4b2
JK		 2937 if (!sit_i->sec_entries)
2938 return -ENOMEM;
2939 }
2940
2941 /* get information related with SIT */
2942 sit_segs = le32_to_cpu(raw_super->segment_count_sit) >> 1;
2943
2944 /* setup SIT bitmap from ckeckpoint pack */
2945 bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
2946 src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
2947
ae27d62e
CY		 2948 sit_i->sit_bitmap = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2949 if (!sit_i->sit_bitmap)
351df4b2 2950 return -ENOMEM;
351df4b2 2951
ae27d62e
CY		 2952#ifdef CONFIG_F2FS_CHECK_FS
2953 sit_i->sit_bitmap_mir = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2954 if (!sit_i->sit_bitmap_mir)
2955 return -ENOMEM;
2956#endif
2957
351df4b2
JK		 2958 /* init SIT information */
2959 sit_i->s_ops = &default_salloc_ops;
2960
2961 sit_i->sit_base_addr = le32_to_cpu(raw_super->sit_blkaddr);
2962 sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
c79b7ff1 2963 sit_i->written_valid_blocks = 0;
351df4b2
JK		 2964 sit_i->bitmap_size = bitmap_size;
2965 sit_i->dirty_sentries = 0;
2966 sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
2967 sit_i->elapsed_time = le64_to_cpu(sbi->ckpt->elapsed_time);
2968 sit_i->mounted_time = CURRENT_TIME_SEC.tv_sec;
2969 mutex_init(&sit_i->sentry_lock);
2970 return 0;
2971}
2972
2973static int build_free_segmap(struct f2fs_sb_info *sbi)
2974{
351df4b2
JK		 2975 struct free_segmap_info *free_i;
2976 unsigned int bitmap_size, sec_bitmap_size;
2977
2978 /* allocate memory for free segmap information */
2979 free_i = kzalloc(sizeof(struct free_segmap_info), GFP_KERNEL);
2980 if (!free_i)
2981 return -ENOMEM;
2982
2983 SM_I(sbi)->free_info = free_i;
2984
7cd8558b 2985 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2986 free_i->free_segmap = f2fs_kvmalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2987 if (!free_i->free_segmap)
2988 return -ENOMEM;
2989
7cd8558b 2990 sec_bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
39307a8e 2991 free_i->free_secmap = f2fs_kvmalloc(sec_bitmap_size, GFP_KERNEL);
351df4b2
JK		 2992 if (!free_i->free_secmap)
2993 return -ENOMEM;
2994
2995 /* set all segments as dirty temporarily */
2996 memset(free_i->free_segmap, 0xff, bitmap_size);
2997 memset(free_i->free_secmap, 0xff, sec_bitmap_size);
2998
2999 /* init free segmap information */
7cd8558b 3000 free_i->start_segno = GET_SEGNO_FROM_SEG0(sbi, MAIN_BLKADDR(sbi));
351df4b2
JK		 3001 free_i->free_segments = 0;
3002 free_i->free_sections = 0;
1a118ccf 3003 spin_lock_init(&free_i->segmap_lock);
351df4b2
JK		 3004 return 0;
3005}
3006
3007static int build_curseg(struct f2fs_sb_info *sbi)
3008{
1042d60f 3009 struct curseg_info *array;
351df4b2
JK		 3010 int i;
3011
b434babf 3012 array = kcalloc(NR_CURSEG_TYPE, sizeof(*array), GFP_KERNEL);
351df4b2
JK		 3013 if (!array)
3014 return -ENOMEM;
3015
3016 SM_I(sbi)->curseg_array = array;
3017
3018 for (i = 0; i < NR_CURSEG_TYPE; i++) {
3019 mutex_init(&array[i].curseg_mutex);
09cbfeaf 3020 array[i].sum_blk = kzalloc(PAGE_SIZE, GFP_KERNEL);
351df4b2
JK		 3021 if (!array[i].sum_blk)
3022 return -ENOMEM;
b7ad7512
CY		 3023 init_rwsem(&array[i].journal_rwsem);
3024 array[i].journal = kzalloc(sizeof(struct f2fs_journal),
3025 GFP_KERNEL);
3026 if (!array[i].journal)
3027 return -ENOMEM;
351df4b2
JK		 3028 array[i].segno = NULL_SEGNO;
3029 array[i].next_blkoff = 0;
3030 }
3031 return restore_curseg_summaries(sbi);
3032}
3033
3034static void build_sit_entries(struct f2fs_sb_info *sbi)
3035{
3036 struct sit_info *sit_i = SIT_I(sbi);
3037 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 3038 struct f2fs_journal *journal = curseg->journal;
9c094040
YH		 3039 struct seg_entry *se;
3040 struct f2fs_sit_entry sit;
74de593a
CY		 3041 int sit_blk_cnt = SIT_BLK_CNT(sbi);
3042 unsigned int i, start, end;
3043 unsigned int readed, start_blk = 0;
351df4b2 3044
74de593a 3045 do {
664ba972
JK		 3046 readed = ra_meta_pages(sbi, start_blk, BIO_MAX_PAGES,
3047 META_SIT, true);
74de593a
CY		 3048
3049 start = start_blk * sit_i->sents_per_block;
3050 end = (start_blk + readed) * sit_i->sents_per_block;
3051
7cd8558b 3052 for (; start < end && start < MAIN_SEGS(sbi); start++) {
74de593a 3053 struct f2fs_sit_block *sit_blk;
74de593a
CY		 3054 struct page *page;
3055
9c094040 3056 se = &sit_i->sentries[start];
74de593a
CY		 3057 page = get_current_sit_page(sbi, start);
3058 sit_blk = (struct f2fs_sit_block *)page_address(page);
3059 sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
3060 f2fs_put_page(page, 1);
d600af23 3061
74de593a
CY		 3062 check_block_count(sbi, start, &sit);
3063 seg_info_from_raw_sit(se, &sit);
a66cdd98
JK		 3064
3065 /* build discard map only one time */
3e025740 3066 if (f2fs_discard_en(sbi)) {
1f43e2ad
CY		 3067 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
3068 memset(se->discard_map, 0xff,
3069 SIT_VBLOCK_MAP_SIZE);
3070 } else {
3071 memcpy(se->discard_map,
3072 se->cur_valid_map,
3073 SIT_VBLOCK_MAP_SIZE);
3074 sbi->discard_blks +=
3075 sbi->blocks_per_seg -
3076 se->valid_blocks;
3077 }
3e025740 3078 }
a66cdd98 3079
d600af23
CY		 3080 if (sbi->segs_per_sec > 1)
3081 get_sec_entry(sbi, start)->valid_blocks +=
3082 se->valid_blocks;
351df4b2 3083 }
74de593a
CY		 3084 start_blk += readed;
3085 } while (start_blk < sit_blk_cnt);
d600af23
CY		 3086
3087 down_read(&curseg->journal_rwsem);
3088 for (i = 0; i < sits_in_cursum(journal); i++) {
d600af23
CY		 3089 unsigned int old_valid_blocks;
3090
3091 start = le32_to_cpu(segno_in_journal(journal, i));
3092 se = &sit_i->sentries[start];
3093 sit = sit_in_journal(journal, i);
3094
3095 old_valid_blocks = se->valid_blocks;
3096
3097 check_block_count(sbi, start, &sit);
3098 seg_info_from_raw_sit(se, &sit);
3099
3100 if (f2fs_discard_en(sbi)) {
1f43e2ad
CY		 3101 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
3102 memset(se->discard_map, 0xff,
3103 SIT_VBLOCK_MAP_SIZE);
3104 } else {
3105 memcpy(se->discard_map, se->cur_valid_map,
3106 SIT_VBLOCK_MAP_SIZE);
3107 sbi->discard_blks += old_valid_blocks -
3108 se->valid_blocks;
3109 }
d600af23
CY		 3110 }
3111
3112 if (sbi->segs_per_sec > 1)
3113 get_sec_entry(sbi, start)->valid_blocks +=
3114 se->valid_blocks - old_valid_blocks;
3115 }
3116 up_read(&curseg->journal_rwsem);
351df4b2
JK		 3117}
3118
3119static void init_free_segmap(struct f2fs_sb_info *sbi)
3120{
3121 unsigned int start;
3122 int type;
3123
7cd8558b 3124 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 3125 struct seg_entry *sentry = get_seg_entry(sbi, start);
3126 if (!sentry->valid_blocks)
3127 __set_free(sbi, start);
c79b7ff1
JK		 3128 else
3129 SIT_I(sbi)->written_valid_blocks +=
3130 sentry->valid_blocks;
351df4b2
JK		 3131 }
3132
3133 /* set use the current segments */
3134 for (type = CURSEG_HOT_DATA; type <= CURSEG_COLD_NODE; type++) {
3135 struct curseg_info *curseg_t = CURSEG_I(sbi, type);
3136 __set_test_and_inuse(sbi, curseg_t->segno);
3137 }
3138}
3139
3140static void init_dirty_segmap(struct f2fs_sb_info *sbi)
3141{
3142 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3143 struct free_segmap_info *free_i = FREE_I(sbi);
7cd8558b 3144 unsigned int segno = 0, offset = 0;
351df4b2
JK		 3145 unsigned short valid_blocks;
3146
8736fbf0 3147 while (1) {
351df4b2 3148 /* find dirty segment based on free segmap */
7cd8558b
JK		 3149 segno = find_next_inuse(free_i, MAIN_SEGS(sbi), offset);
3150 if (segno >= MAIN_SEGS(sbi))
351df4b2
JK		 3151 break;
3152 offset = segno + 1;
302bd348 3153 valid_blocks = get_valid_blocks(sbi, segno, false);
ec325b52 3154 if (valid_blocks == sbi->blocks_per_seg || !valid_blocks)
351df4b2 3155 continue;
ec325b52
JK		 3156 if (valid_blocks > sbi->blocks_per_seg) {
3157 f2fs_bug_on(sbi, 1);
3158 continue;
3159 }
351df4b2
JK		 3160 mutex_lock(&dirty_i->seglist_lock);
3161 __locate_dirty_segment(sbi, segno, DIRTY);
3162 mutex_unlock(&dirty_i->seglist_lock);
3163 }
3164}
3165
5ec4e49f 3166static int init_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3167{
3168 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
7cd8558b 3169 unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
351df4b2 3170
39307a8e 3171 dirty_i->victim_secmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
5ec4e49f 3172 if (!dirty_i->victim_secmap)
351df4b2
JK		 3173 return -ENOMEM;
3174 return 0;
3175}
3176
3177static int build_dirty_segmap(struct f2fs_sb_info *sbi)
3178{
3179 struct dirty_seglist_info *dirty_i;
3180 unsigned int bitmap_size, i;
3181
3182 /* allocate memory for dirty segments list information */
3183 dirty_i = kzalloc(sizeof(struct dirty_seglist_info), GFP_KERNEL);
3184 if (!dirty_i)
3185 return -ENOMEM;
3186
3187 SM_I(sbi)->dirty_info = dirty_i;
3188 mutex_init(&dirty_i->seglist_lock);
3189
7cd8558b 3190 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
351df4b2
JK		 3191
3192 for (i = 0; i < NR_DIRTY_TYPE; i++) {
39307a8e 3193 dirty_i->dirty_segmap[i] = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 3194 if (!dirty_i->dirty_segmap[i])
3195 return -ENOMEM;
3196 }
3197
3198 init_dirty_segmap(sbi);
5ec4e49f 3199 return init_victim_secmap(sbi);
351df4b2
JK		 3200}
3201
0a8165d7 3202/*
351df4b2
JK		 3203 * Update min, max modified time for cost-benefit GC algorithm
3204 */
3205static void init_min_max_mtime(struct f2fs_sb_info *sbi)
3206{
3207 struct sit_info *sit_i = SIT_I(sbi);
3208 unsigned int segno;
3209
3210 mutex_lock(&sit_i->sentry_lock);
3211
3212 sit_i->min_mtime = LLONG_MAX;
3213
7cd8558b 3214 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
351df4b2
JK		 3215 unsigned int i;
3216 unsigned long long mtime = 0;
3217
3218 for (i = 0; i < sbi->segs_per_sec; i++)
3219 mtime += get_seg_entry(sbi, segno + i)->mtime;
3220
3221 mtime = div_u64(mtime, sbi->segs_per_sec);
3222
3223 if (sit_i->min_mtime > mtime)
3224 sit_i->min_mtime = mtime;
3225 }
3226 sit_i->max_mtime = get_mtime(sbi);
3227 mutex_unlock(&sit_i->sentry_lock);
3228}
3229
3230int build_segment_manager(struct f2fs_sb_info *sbi)
3231{
3232 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3233 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1042d60f 3234 struct f2fs_sm_info *sm_info;
351df4b2
JK		 3235 int err;
3236
3237 sm_info = kzalloc(sizeof(struct f2fs_sm_info), GFP_KERNEL);
3238 if (!sm_info)
3239 return -ENOMEM;
3240
3241 /* init sm info */
3242 sbi->sm_info = sm_info;
351df4b2
JK		 3243 sm_info->seg0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3244 sm_info->main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3245 sm_info->segment_count = le32_to_cpu(raw_super->segment_count);
3246 sm_info->reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3247 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3248 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
3249 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
58c41035
JK		 3250 sm_info->rec_prefree_segments = sm_info->main_segments *
3251 DEF_RECLAIM_PREFREE_SEGMENTS / 100;
44a83499
JK		 3252 if (sm_info->rec_prefree_segments > DEF_MAX_RECLAIM_PREFREE_SEGMENTS)
3253 sm_info->rec_prefree_segments = DEF_MAX_RECLAIM_PREFREE_SEGMENTS;
3254
52763a4b
JK		 3255 if (!test_opt(sbi, LFS))
3256 sm_info->ipu_policy = 1 << F2FS_IPU_FSYNC;
216fbd64 3257 sm_info->min_ipu_util = DEF_MIN_IPU_UTIL;
c1ce1b02 3258 sm_info->min_fsync_blocks = DEF_MIN_FSYNC_BLOCKS;
ef095d19 3259 sm_info->min_hot_blocks = DEF_MIN_HOT_BLOCKS;
351df4b2 3260
bba681cb
JK		 3261 sm_info->trim_sections = DEF_BATCHED_TRIM_SECTIONS;
3262
184a5cd2
CY		 3263 INIT_LIST_HEAD(&sm_info->sit_entry_set);
3264
d4fdf8ba 3265 if (!f2fs_readonly(sbi->sb)) {
2163d198
GZ		 3266 err = create_flush_cmd_control(sbi);
3267 if (err)
a688b9d9 3268 return err;
6b4afdd7
JK		 3269 }
3270
0b54fb84
JK		 3271 err = create_discard_cmd_control(sbi);
3272 if (err)
3273 return err;
3274
351df4b2
JK		 3275 err = build_sit_info(sbi);
3276 if (err)
3277 return err;
3278 err = build_free_segmap(sbi);
3279 if (err)
3280 return err;
3281 err = build_curseg(sbi);
3282 if (err)
3283 return err;
3284
3285 /* reinit free segmap based on SIT */
3286 build_sit_entries(sbi);
3287
3288 init_free_segmap(sbi);
3289 err = build_dirty_segmap(sbi);
3290 if (err)
3291 return err;
3292
3293 init_min_max_mtime(sbi);
3294 return 0;
3295}
3296
3297static void discard_dirty_segmap(struct f2fs_sb_info *sbi,
3298 enum dirty_type dirty_type)
3299{
3300 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3301
3302 mutex_lock(&dirty_i->seglist_lock);
39307a8e 3303 kvfree(dirty_i->dirty_segmap[dirty_type]);
351df4b2
JK		 3304 dirty_i->nr_dirty[dirty_type] = 0;
3305 mutex_unlock(&dirty_i->seglist_lock);
3306}
3307
5ec4e49f 3308static void destroy_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3309{
3310 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
39307a8e 3311 kvfree(dirty_i->victim_secmap);
351df4b2
JK		 3312}
3313
3314static void destroy_dirty_segmap(struct f2fs_sb_info *sbi)
3315{
3316 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3317 int i;
3318
3319 if (!dirty_i)
3320 return;
3321
3322 /* discard pre-free/dirty segments list */
3323 for (i = 0; i < NR_DIRTY_TYPE; i++)
3324 discard_dirty_segmap(sbi, i);
3325
5ec4e49f 3326 destroy_victim_secmap(sbi);
351df4b2
JK		 3327 SM_I(sbi)->dirty_info = NULL;
3328 kfree(dirty_i);
3329}
3330
3331static void destroy_curseg(struct f2fs_sb_info *sbi)
3332{
3333 struct curseg_info *array = SM_I(sbi)->curseg_array;
3334 int i;
3335
3336 if (!array)
3337 return;
3338 SM_I(sbi)->curseg_array = NULL;
b7ad7512 3339 for (i = 0; i < NR_CURSEG_TYPE; i++) {
351df4b2 3340 kfree(array[i].sum_blk);
b7ad7512
CY		 3341 kfree(array[i].journal);
3342 }
351df4b2
JK		 3343 kfree(array);
3344}
3345
3346static void destroy_free_segmap(struct f2fs_sb_info *sbi)
3347{
3348 struct free_segmap_info *free_i = SM_I(sbi)->free_info;
3349 if (!free_i)
3350 return;
3351 SM_I(sbi)->free_info = NULL;
39307a8e
JK		 3352 kvfree(free_i->free_segmap);
3353 kvfree(free_i->free_secmap);
351df4b2
JK		 3354 kfree(free_i);
3355}
3356
3357static void destroy_sit_info(struct f2fs_sb_info *sbi)
3358{
3359 struct sit_info *sit_i = SIT_I(sbi);
3360 unsigned int start;
3361
3362 if (!sit_i)
3363 return;
3364
3365 if (sit_i->sentries) {
7cd8558b 3366 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2 3367 kfree(sit_i->sentries[start].cur_valid_map);
355e7891
CY		 3368#ifdef CONFIG_F2FS_CHECK_FS
3369 kfree(sit_i->sentries[start].cur_valid_map_mir);
3370#endif
351df4b2 3371 kfree(sit_i->sentries[start].ckpt_valid_map);
a66cdd98 3372 kfree(sit_i->sentries[start].discard_map);
351df4b2
JK		 3373 }
3374 }
60a3b782
JK		 3375 kfree(sit_i->tmp_map);
3376
39307a8e
JK		 3377 kvfree(sit_i->sentries);
3378 kvfree(sit_i->sec_entries);
3379 kvfree(sit_i->dirty_sentries_bitmap);
351df4b2
JK		 3380
3381 SM_I(sbi)->sit_info = NULL;
3382 kfree(sit_i->sit_bitmap);
ae27d62e
CY		 3383#ifdef CONFIG_F2FS_CHECK_FS
3384 kfree(sit_i->sit_bitmap_mir);
3385#endif
351df4b2
JK		 3386 kfree(sit_i);
3387}
3388
3389void destroy_segment_manager(struct f2fs_sb_info *sbi)
3390{
3391 struct f2fs_sm_info *sm_info = SM_I(sbi);
a688b9d9 3392
3b03f724
CY		 3393 if (!sm_info)
3394 return;
5eba8c5d 3395 destroy_flush_cmd_control(sbi, true);
f099405f 3396 destroy_discard_cmd_control(sbi);
351df4b2
JK		 3397 destroy_dirty_segmap(sbi);
3398 destroy_curseg(sbi);
3399 destroy_free_segmap(sbi);
3400 destroy_sit_info(sbi);
3401 sbi->sm_info = NULL;
3402 kfree(sm_info);
3403}
7fd9e544
JK		 3404
3405int __init create_segment_manager_caches(void)
3406{
3407 discard_entry_slab = f2fs_kmem_cache_create("discard_entry",
e8512d2e 3408 sizeof(struct discard_entry));
7fd9e544 3409 if (!discard_entry_slab)
184a5cd2
CY		 3410 goto fail;
3411
b01a9201
JK		 3412 discard_cmd_slab = f2fs_kmem_cache_create("discard_cmd",
3413 sizeof(struct discard_cmd));
3414 if (!discard_cmd_slab)
6ab2a308 3415 goto destroy_discard_entry;
275b66b0 3416
184a5cd2 3417 sit_entry_set_slab = f2fs_kmem_cache_create("sit_entry_set",
c9ee0085 3418 sizeof(struct sit_entry_set));
184a5cd2 3419 if (!sit_entry_set_slab)
b01a9201 3420 goto destroy_discard_cmd;
88b88a66
JK		 3421
3422 inmem_entry_slab = f2fs_kmem_cache_create("inmem_page_entry",
3423 sizeof(struct inmem_pages));
3424 if (!inmem_entry_slab)
3425 goto destroy_sit_entry_set;
7fd9e544 3426 return 0;
184a5cd2 3427
88b88a66
JK		 3428destroy_sit_entry_set:
3429 kmem_cache_destroy(sit_entry_set_slab);
b01a9201
JK		 3430destroy_discard_cmd:
3431 kmem_cache_destroy(discard_cmd_slab);
6ab2a308 3432destroy_discard_entry:
184a5cd2
CY		 3433 kmem_cache_destroy(discard_entry_slab);
3434fail:
3435 return -ENOMEM;
7fd9e544
JK		 3436}
3437
3438void destroy_segment_manager_caches(void)
3439{
184a5cd2 3440 kmem_cache_destroy(sit_entry_set_slab);
b01a9201 3441 kmem_cache_destroy(discard_cmd_slab);
7fd9e544 3442 kmem_cache_destroy(discard_entry_slab);
88b88a66 3443 kmem_cache_destroy(inmem_entry_slab);
7fd9e544 3444}
Linux 6.x block layer and io_uring tree(s)