]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - fs/f2fs/segment.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: use printk_ratelimited for f2fs_msg
[mirror_ubuntu-jammy-kernel.git] / fs / f2fs / segment.c
CommitLineData
0a8165d7 1/*
351df4b2
JK		 2 * fs/f2fs/segment.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11#include <linux/fs.h>
12#include <linux/f2fs_fs.h>
13#include <linux/bio.h>
14#include <linux/blkdev.h>
690e4a3e 15#include <linux/prefetch.h>
6b4afdd7 16#include <linux/kthread.h>
74de593a 17#include <linux/swap.h>
60b99b48 18#include <linux/timer.h>
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 }
640cc189 312retry:
28bc106b 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) {
640cc189
JK		 319 if (err == -ENOMEM) {
320 congestion_wait(BLK_RW_ASYNC, HZ/50);
321 cond_resched();
322 goto retry;
323 }
28bc106b 324 unlock_page(page);
29b96b54 325 break;
70c640b1 326 }
28bc106b
CY		 327 /* record old blkaddr for revoking */
328 cur->old_addr = fio.old_blkaddr;
942fd319 329 last_idx = page->index;
28bc106b
CY		 330 }
331 unlock_page(page);
332 list_move_tail(&cur->list, revoke_list);
88b88a66 333 }
29b96b54 334
942fd319 335 if (last_idx != ULONG_MAX)
b9109b0e 336 f2fs_submit_merged_write_cond(sbi, inode, 0, last_idx, DATA);
28bc106b
CY		 337
338 if (!err)
339 __revoke_inmem_pages(inode, revoke_list, false, false);
340
29b96b54
CY		 341 return err;
342}
343
344int commit_inmem_pages(struct inode *inode)
345{
346 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
347 struct f2fs_inode_info *fi = F2FS_I(inode);
28bc106b
CY		 348 struct list_head revoke_list;
349 int err;
29b96b54 350
28bc106b 351 INIT_LIST_HEAD(&revoke_list);
29b96b54
CY		 352 f2fs_balance_fs(sbi, true);
353 f2fs_lock_op(sbi);
354
5fe45743
CY		 355 set_inode_flag(inode, FI_ATOMIC_COMMIT);
356
29b96b54 357 mutex_lock(&fi->inmem_lock);
28bc106b
CY		 358 err = __commit_inmem_pages(inode, &revoke_list);
359 if (err) {
360 int ret;
361 /*
362 * try to revoke all committed pages, but still we could fail
363 * due to no memory or other reason, if that happened, EAGAIN
364 * will be returned, which means in such case, transaction is
365 * already not integrity, caller should use journal to do the
366 * recovery or rewrite & commit last transaction. For other
367 * error number, revoking was done by filesystem itself.
368 */
369 ret = __revoke_inmem_pages(inode, &revoke_list, false, true);
370 if (ret)
371 err = ret;
372
373 /* drop all uncommitted pages */
374 __revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
375 }
88b88a66
JK		 376 mutex_unlock(&fi->inmem_lock);
377
5fe45743
CY		 378 clear_inode_flag(inode, FI_ATOMIC_COMMIT);
379
29b96b54 380 f2fs_unlock_op(sbi);
edb27dee 381 return err;
88b88a66
JK		 382}
383
0a8165d7 384/*
351df4b2
JK		 385 * This function balances dirty node and dentry pages.
386 * In addition, it controls garbage collection.
387 */
2c4db1a6 388void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
351df4b2 389{
0f348028 390#ifdef CONFIG_F2FS_FAULT_INJECTION
55523519
CY		 391 if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
392 f2fs_show_injection_info(FAULT_CHECKPOINT);
0f348028 393 f2fs_stop_checkpoint(sbi, false);
55523519 394 }
0f348028
CY		 395#endif
396
e589c2c4 397 /* balance_fs_bg is able to be pending */
a7881893 398 if (need && excess_cached_nats(sbi))
e589c2c4
JK		 399 f2fs_balance_fs_bg(sbi);
400
351df4b2 401 /*
029cd28c
JK		 402 * We should do GC or end up with checkpoint, if there are so many dirty
403 * dir/node pages without enough free segments.
351df4b2 404 */
7f3037a5 405 if (has_not_enough_free_secs(sbi, 0, 0)) {
351df4b2 406 mutex_lock(&sbi->gc_mutex);
e066b83c 407 f2fs_gc(sbi, false, false, NULL_SEGNO);
351df4b2
JK		 408 }
409}
410
4660f9c0
JK		 411void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi)
412{
1dcc336b 413 /* try to shrink extent cache when there is no enough memory */
554df79e
JK		 414 if (!available_free_memory(sbi, EXTENT_CACHE))
415 f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER);
1dcc336b 416
1b38dc8e
JK		 417 /* check the # of cached NAT entries */
418 if (!available_free_memory(sbi, NAT_ENTRIES))
419 try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK);
420
31696580 421 if (!available_free_memory(sbi, FREE_NIDS))
ad4edb83
JK		 422 try_to_free_nids(sbi, MAX_FREE_NIDS);
423 else
22ad0b6a 424 build_free_nids(sbi, false, false);
31696580 425
1c0f4bf5 426 if (!is_idle(sbi) && !excess_dirty_nats(sbi))
f455c8a5 427 return;
31696580 428
1b38dc8e
JK		 429 /* checkpoint is the only way to shrink partial cached entries */
430 if (!available_free_memory(sbi, NAT_ENTRIES) ||
60b99b48 431 !available_free_memory(sbi, INO_ENTRIES) ||
7d768d2c
CY		 432 excess_prefree_segs(sbi) ||
433 excess_dirty_nats(sbi) ||
f455c8a5 434 f2fs_time_over(sbi, CP_TIME)) {
e9f5b8b8
CY		 435 if (test_opt(sbi, DATA_FLUSH)) {
436 struct blk_plug plug;
437
438 blk_start_plug(&plug);
36b35a0d 439 sync_dirty_inodes(sbi, FILE_INODE);
e9f5b8b8
CY		 440 blk_finish_plug(&plug);
441 }
4660f9c0 442 f2fs_sync_fs(sbi->sb, true);
42190d2a 443 stat_inc_bg_cp_count(sbi->stat_info);
36b35a0d 444 }
4660f9c0
JK		 445}
446
20fda56b
KM		 447static int __submit_flush_wait(struct f2fs_sb_info *sbi,
448 struct block_device *bdev)
3c62be17
JK		 449{
450 struct bio *bio = f2fs_bio_alloc(0);
451 int ret;
452
3adc5fcb 453 bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH;
3c62be17
JK		 454 bio->bi_bdev = bdev;
455 ret = submit_bio_wait(bio);
456 bio_put(bio);
20fda56b
KM		 457
458 trace_f2fs_issue_flush(bdev, test_opt(sbi, NOBARRIER),
459 test_opt(sbi, FLUSH_MERGE), ret);
3c62be17
JK		 460 return ret;
461}
462
463static int submit_flush_wait(struct f2fs_sb_info *sbi)
464{
20fda56b 465 int ret = __submit_flush_wait(sbi, sbi->sb->s_bdev);
3c62be17
JK		 466 int i;
467
20fda56b
KM		 468 if (!sbi->s_ndevs || ret)
469 return ret;
470
471 for (i = 1; i < sbi->s_ndevs; i++) {
472 ret = __submit_flush_wait(sbi, FDEV(i).bdev);
473 if (ret)
474 break;
3c62be17
JK		 475 }
476 return ret;
477}
478
2163d198 479static int issue_flush_thread(void *data)
6b4afdd7
JK		 480{
481 struct f2fs_sb_info *sbi = data;
b01a9201 482 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
a688b9d9 483 wait_queue_head_t *q = &fcc->flush_wait_queue;
6b4afdd7
JK		 484repeat:
485 if (kthread_should_stop())
486 return 0;
487
dc6febb6
CY		 488 sb_start_intwrite(sbi->sb);
489
721bd4d5 490 if (!llist_empty(&fcc->issue_list)) {
6b4afdd7
JK		 491 struct flush_cmd *cmd, *next;
492 int ret;
493
721bd4d5
GZ		 494 fcc->dispatch_list = llist_del_all(&fcc->issue_list);
495 fcc->dispatch_list = llist_reverse_order(fcc->dispatch_list);
496
3c62be17 497 ret = submit_flush_wait(sbi);
8b8dd65f
CY		 498 atomic_inc(&fcc->issued_flush);
499
721bd4d5
GZ		 500 llist_for_each_entry_safe(cmd, next,
501 fcc->dispatch_list, llnode) {
6b4afdd7 502 cmd->ret = ret;
6b4afdd7
JK		 503 complete(&cmd->wait);
504 }
a688b9d9 505 fcc->dispatch_list = NULL;
6b4afdd7
JK		 506 }
507
dc6febb6
CY		 508 sb_end_intwrite(sbi->sb);
509
a688b9d9 510 wait_event_interruptible(*q,
721bd4d5 511 kthread_should_stop() || !llist_empty(&fcc->issue_list));
6b4afdd7
JK		 512 goto repeat;
513}
514
515int f2fs_issue_flush(struct f2fs_sb_info *sbi)
516{
b01a9201 517 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
adf8d90b 518 struct flush_cmd cmd;
8b8dd65f 519 int ret;
6b4afdd7 520
0f7b2abd
JK		 521 if (test_opt(sbi, NOBARRIER))
522 return 0;
523
8b8dd65f
CY		 524 if (!test_opt(sbi, FLUSH_MERGE)) {
525 ret = submit_flush_wait(sbi);
526 atomic_inc(&fcc->issued_flush);
527 return ret;
528 }
740432f8 529
8b8dd65f
CY		 530 if (!atomic_read(&fcc->issing_flush)) {
531 atomic_inc(&fcc->issing_flush);
3c62be17 532 ret = submit_flush_wait(sbi);
8b8dd65f
CY		 533 atomic_dec(&fcc->issing_flush);
534
535 atomic_inc(&fcc->issued_flush);
740432f8
JK		 536 return ret;
537 }
6b4afdd7 538
adf8d90b 539 init_completion(&cmd.wait);
6b4afdd7 540
8b8dd65f 541 atomic_inc(&fcc->issing_flush);
721bd4d5 542 llist_add(&cmd.llnode, &fcc->issue_list);
6b4afdd7 543
a688b9d9
GZ		 544 if (!fcc->dispatch_list)
545 wake_up(&fcc->flush_wait_queue);
6b4afdd7 546
5eba8c5d
JK		 547 if (fcc->f2fs_issue_flush) {
548 wait_for_completion(&cmd.wait);
8b8dd65f 549 atomic_dec(&fcc->issing_flush);
5eba8c5d
JK		 550 } else {
551 llist_del_all(&fcc->issue_list);
8b8dd65f 552 atomic_set(&fcc->issing_flush, 0);
5eba8c5d 553 }
adf8d90b
CY		 554
555 return cmd.ret;
6b4afdd7
JK		 556}
557
2163d198
GZ		 558int create_flush_cmd_control(struct f2fs_sb_info *sbi)
559{
560 dev_t dev = sbi->sb->s_bdev->bd_dev;
561 struct flush_cmd_control *fcc;
562 int err = 0;
563
b01a9201
JK		 564 if (SM_I(sbi)->fcc_info) {
565 fcc = SM_I(sbi)->fcc_info;
d871cd04
YS		 566 if (fcc->f2fs_issue_flush)
567 return err;
5eba8c5d
JK		 568 goto init_thread;
569 }
570
2163d198
GZ		 571 fcc = kzalloc(sizeof(struct flush_cmd_control), GFP_KERNEL);
572 if (!fcc)
573 return -ENOMEM;
8b8dd65f
CY		 574 atomic_set(&fcc->issued_flush, 0);
575 atomic_set(&fcc->issing_flush, 0);
2163d198 576 init_waitqueue_head(&fcc->flush_wait_queue);
721bd4d5 577 init_llist_head(&fcc->issue_list);
b01a9201 578 SM_I(sbi)->fcc_info = fcc;
d4fdf8ba
YH		 579 if (!test_opt(sbi, FLUSH_MERGE))
580 return err;
581
5eba8c5d 582init_thread:
2163d198
GZ		 583 fcc->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi,
584 "f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev));
585 if (IS_ERR(fcc->f2fs_issue_flush)) {
586 err = PTR_ERR(fcc->f2fs_issue_flush);
587 kfree(fcc);
b01a9201 588 SM_I(sbi)->fcc_info = NULL;
2163d198
GZ		 589 return err;
590 }
2163d198
GZ		 591
592 return err;
593}
594
5eba8c5d 595void destroy_flush_cmd_control(struct f2fs_sb_info *sbi, bool free)
2163d198 596{
b01a9201 597 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
2163d198 598
5eba8c5d
JK		 599 if (fcc && fcc->f2fs_issue_flush) {
600 struct task_struct *flush_thread = fcc->f2fs_issue_flush;
601
602 fcc->f2fs_issue_flush = NULL;
603 kthread_stop(flush_thread);
604 }
605 if (free) {
606 kfree(fcc);
b01a9201 607 SM_I(sbi)->fcc_info = NULL;
5eba8c5d 608 }
2163d198
GZ		 609}
610
351df4b2
JK		 611static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
612 enum dirty_type dirty_type)
613{
614 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
615
616 /* need not be added */
617 if (IS_CURSEG(sbi, segno))
618 return;
619
620 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[dirty_type]))
621 dirty_i->nr_dirty[dirty_type]++;
622
623 if (dirty_type == DIRTY) {
624 struct seg_entry *sentry = get_seg_entry(sbi, segno);
4625d6aa 625 enum dirty_type t = sentry->type;
b2f2c390 626
ec325b52
JK		 627 if (unlikely(t >= DIRTY)) {
628 f2fs_bug_on(sbi, 1);
629 return;
630 }
4625d6aa
CL		 631 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[t]))
632 dirty_i->nr_dirty[t]++;
351df4b2
JK		 633 }
634}
635
636static void __remove_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
637 enum dirty_type dirty_type)
638{
639 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
640
641 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[dirty_type]))
642 dirty_i->nr_dirty[dirty_type]--;
643
644 if (dirty_type == DIRTY) {
4625d6aa
CL		 645 struct seg_entry *sentry = get_seg_entry(sbi, segno);
646 enum dirty_type t = sentry->type;
647
648 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
649 dirty_i->nr_dirty[t]--;
b2f2c390 650
302bd348 651 if (get_valid_blocks(sbi, segno, true) == 0)
4ddb1a4d 652 clear_bit(GET_SEC_FROM_SEG(sbi, segno),
5ec4e49f 653 dirty_i->victim_secmap);
351df4b2
JK		 654 }
655}
656
0a8165d7 657/*
351df4b2
JK		 658 * Should not occur error such as -ENOMEM.
659 * Adding dirty entry into seglist is not critical operation.
660 * If a given segment is one of current working segments, it won't be added.
661 */
8d8451af 662static void locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 663{
664 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
665 unsigned short valid_blocks;
666
667 if (segno == NULL_SEGNO || IS_CURSEG(sbi, segno))
668 return;
669
670 mutex_lock(&dirty_i->seglist_lock);
671
302bd348 672 valid_blocks = get_valid_blocks(sbi, segno, false);
351df4b2
JK		 673
674 if (valid_blocks == 0) {
675 __locate_dirty_segment(sbi, segno, PRE);
676 __remove_dirty_segment(sbi, segno, DIRTY);
677 } else if (valid_blocks < sbi->blocks_per_seg) {
678 __locate_dirty_segment(sbi, segno, DIRTY);
679 } else {
680 /* Recovery routine with SSR needs this */
681 __remove_dirty_segment(sbi, segno, DIRTY);
682 }
683
684 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 685}
686
004b6862 687static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi,
c81abe34
JK		 688 struct block_device *bdev, block_t lstart,
689 block_t start, block_t len)
275b66b0 690{
0b54fb84 691 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
ba48a33e 692 struct list_head *pend_list;
b01a9201 693 struct discard_cmd *dc;
275b66b0 694
ba48a33e
CY		 695 f2fs_bug_on(sbi, !len);
696
697 pend_list = &dcc->pend_list[plist_idx(len)];
698
b01a9201
JK		 699 dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS);
700 INIT_LIST_HEAD(&dc->list);
c81abe34 701 dc->bdev = bdev;
b01a9201 702 dc->lstart = lstart;
c81abe34 703 dc->start = start;
b01a9201 704 dc->len = len;
ec9895ad 705 dc->ref = 0;
15469963 706 dc->state = D_PREP;
c81abe34 707 dc->error = 0;
b01a9201 708 init_completion(&dc->wait);
22d375dd 709 list_add_tail(&dc->list, pend_list);
5f32366a 710 atomic_inc(&dcc->discard_cmd_cnt);
d84d1cbd 711 dcc->undiscard_blks += len;
004b6862
CY		 712
713 return dc;
714}
715
716static struct discard_cmd *__attach_discard_cmd(struct f2fs_sb_info *sbi,
717 struct block_device *bdev, block_t lstart,
718 block_t start, block_t len,
719 struct rb_node *parent, struct rb_node **p)
720{
721 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
722 struct discard_cmd *dc;
723
724 dc = __create_discard_cmd(sbi, bdev, lstart, start, len);
725
726 rb_link_node(&dc->rb_node, parent, p);
727 rb_insert_color(&dc->rb_node, &dcc->root);
728
729 return dc;
15469963
JK		 730}
731
004b6862
CY		 732static void __detach_discard_cmd(struct discard_cmd_control *dcc,
733 struct discard_cmd *dc)
15469963 734{
dcc9165d 735 if (dc->state == D_DONE)
004b6862
CY		 736 atomic_dec(&dcc->issing_discard);
737
738 list_del(&dc->list);
739 rb_erase(&dc->rb_node, &dcc->root);
d84d1cbd 740 dcc->undiscard_blks -= dc->len;
004b6862
CY		 741
742 kmem_cache_free(discard_cmd_slab, dc);
743
744 atomic_dec(&dcc->discard_cmd_cnt);
745}
746
747static void __remove_discard_cmd(struct f2fs_sb_info *sbi,
748 struct discard_cmd *dc)
749{
750 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
dcc9165d 751
d9703d90
CY		 752 f2fs_bug_on(sbi, dc->ref);
753
c81abe34
JK		 754 if (dc->error == -EOPNOTSUPP)
755 dc->error = 0;
15469963 756
c81abe34 757 if (dc->error)
15469963 758 f2fs_msg(sbi->sb, KERN_INFO,
04dfc230
CY		 759 "Issue discard(%u, %u, %u) failed, ret: %d",
760 dc->lstart, dc->start, dc->len, dc->error);
004b6862 761 __detach_discard_cmd(dcc, dc);
275b66b0
CY		 762}
763
c81abe34
JK		 764static void f2fs_submit_discard_endio(struct bio *bio)
765{
766 struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
767
4e4cbee9 768 dc->error = blk_status_to_errno(bio->bi_status);
c81abe34 769 dc->state = D_DONE;
e31b9821 770 complete_all(&dc->wait);
c81abe34
JK		 771 bio_put(bio);
772}
773
6915ea9d
CY		 774void __check_sit_bitmap(struct f2fs_sb_info *sbi,
775 block_t start, block_t end)
776{
777#ifdef CONFIG_F2FS_CHECK_FS
778 struct seg_entry *sentry;
779 unsigned int segno;
780 block_t blk = start;
781 unsigned long offset, size, max_blocks = sbi->blocks_per_seg;
782 unsigned long *map;
783
784 while (blk < end) {
785 segno = GET_SEGNO(sbi, blk);
786 sentry = get_seg_entry(sbi, segno);
787 offset = GET_BLKOFF_FROM_SEG0(sbi, blk);
788
789 size = min((unsigned long)(end - blk), max_blocks);
790 map = (unsigned long *)(sentry->cur_valid_map);
791 offset = __find_rev_next_bit(map, size, offset);
792 f2fs_bug_on(sbi, offset != size);
793 blk += size;
794 }
795#endif
796}
797
c81abe34
JK		 798/* this function is copied from blkdev_issue_discard from block/blk-lib.c */
799static void __submit_discard_cmd(struct f2fs_sb_info *sbi,
800 struct discard_cmd *dc)
801{
802 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
803 struct bio *bio = NULL;
804
805 if (dc->state != D_PREP)
806 return;
807
0243a5f9
CY		 808 trace_f2fs_issue_discard(dc->bdev, dc->start, dc->len);
809
c81abe34
JK		 810 dc->error = __blkdev_issue_discard(dc->bdev,
811 SECTOR_FROM_BLOCK(dc->start),
812 SECTOR_FROM_BLOCK(dc->len),
813 GFP_NOFS, 0, &bio);
814 if (!dc->error) {
815 /* should keep before submission to avoid D_DONE right away */
816 dc->state = D_SUBMIT;
8b8dd65f
CY		 817 atomic_inc(&dcc->issued_discard);
818 atomic_inc(&dcc->issing_discard);
c81abe34
JK		 819 if (bio) {
820 bio->bi_private = dc;
821 bio->bi_end_io = f2fs_submit_discard_endio;
822 bio->bi_opf |= REQ_SYNC;
823 submit_bio(bio);
46f84c2c 824 list_move_tail(&dc->list, &dcc->wait_list);
6915ea9d 825 __check_sit_bitmap(sbi, dc->start, dc->start + dc->len);
c81abe34
JK		 826 }
827 } else {
828 __remove_discard_cmd(sbi, dc);
829 }
830}
831
004b6862
CY		 832static struct discard_cmd *__insert_discard_tree(struct f2fs_sb_info *sbi,
833 struct block_device *bdev, block_t lstart,
834 block_t start, block_t len,
835 struct rb_node **insert_p,
836 struct rb_node *insert_parent)
c81abe34 837{
004b6862
CY		 838 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
839 struct rb_node **p = &dcc->root.rb_node;
840 struct rb_node *parent = NULL;
841 struct discard_cmd *dc = NULL;
842
843 if (insert_p && insert_parent) {
844 parent = insert_parent;
845 p = insert_p;
846 goto do_insert;
847 }
c81abe34 848
004b6862
CY		 849 p = __lookup_rb_tree_for_insert(sbi, &dcc->root, &parent, lstart);
850do_insert:
851 dc = __attach_discard_cmd(sbi, bdev, lstart, start, len, parent, p);
852 if (!dc)
853 return NULL;
c81abe34 854
004b6862 855 return dc;
c81abe34
JK		 856}
857
ba48a33e
CY		 858static void __relocate_discard_cmd(struct discard_cmd_control *dcc,
859 struct discard_cmd *dc)
860{
861 list_move_tail(&dc->list, &dcc->pend_list[plist_idx(dc->len)]);
862}
863
3d6a650f
YH		 864static void __punch_discard_cmd(struct f2fs_sb_info *sbi,
865 struct discard_cmd *dc, block_t blkaddr)
866{
ba48a33e 867 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 868 struct discard_info di = dc->di;
869 bool modified = false;
3d6a650f 870
004b6862 871 if (dc->state == D_DONE || dc->len == 1) {
3d6a650f
YH		 872 __remove_discard_cmd(sbi, dc);
873 return;
874 }
875
d84d1cbd
CY		 876 dcc->undiscard_blks -= di.len;
877
004b6862 878 if (blkaddr > di.lstart) {
3d6a650f 879 dc->len = blkaddr - dc->lstart;
d84d1cbd 880 dcc->undiscard_blks += dc->len;
ba48a33e 881 __relocate_discard_cmd(dcc, dc);
004b6862
CY		 882 modified = true;
883 }
884
885 if (blkaddr < di.lstart + di.len - 1) {
886 if (modified) {
887 __insert_discard_tree(sbi, dc->bdev, blkaddr + 1,
888 di.start + blkaddr + 1 - di.lstart,
889 di.lstart + di.len - 1 - blkaddr,
890 NULL, NULL);
891 } else {
892 dc->lstart++;
893 dc->len--;
894 dc->start++;
d84d1cbd 895 dcc->undiscard_blks += dc->len;
ba48a33e 896 __relocate_discard_cmd(dcc, dc);
004b6862 897 }
3d6a650f
YH		 898 }
899}
900
004b6862
CY		 901static void __update_discard_tree_range(struct f2fs_sb_info *sbi,
902 struct block_device *bdev, block_t lstart,
903 block_t start, block_t len)
275b66b0 904{
0b54fb84 905 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 906 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
907 struct discard_cmd *dc;
908 struct discard_info di = {0};
909 struct rb_node **insert_p = NULL, *insert_parent = NULL;
910 block_t end = lstart + len;
275b66b0 911
15469963 912 mutex_lock(&dcc->cmd_lock);
40465257 913
004b6862
CY		 914 dc = (struct discard_cmd *)__lookup_rb_tree_ret(&dcc->root,
915 NULL, lstart,
916 (struct rb_entry **)&prev_dc,
917 (struct rb_entry **)&next_dc,
918 &insert_p, &insert_parent, true);
919 if (dc)
920 prev_dc = dc;
921
922 if (!prev_dc) {
923 di.lstart = lstart;
924 di.len = next_dc ? next_dc->lstart - lstart : len;
925 di.len = min(di.len, len);
926 di.start = start;
22d375dd 927 }
15469963 928
004b6862
CY		 929 while (1) {
930 struct rb_node *node;
931 bool merged = false;
932 struct discard_cmd *tdc = NULL;
933
934 if (prev_dc) {
935 di.lstart = prev_dc->lstart + prev_dc->len;
936 if (di.lstart < lstart)
937 di.lstart = lstart;
938 if (di.lstart >= end)
939 break;
940
941 if (!next_dc || next_dc->lstart > end)
942 di.len = end - di.lstart;
943 else
944 di.len = next_dc->lstart - di.lstart;
945 di.start = start + di.lstart - lstart;
946 }
947
948 if (!di.len)
949 goto next;
950
951 if (prev_dc && prev_dc->state == D_PREP &&
952 prev_dc->bdev == bdev &&
953 __is_discard_back_mergeable(&di, &prev_dc->di)) {
954 prev_dc->di.len += di.len;
d84d1cbd 955 dcc->undiscard_blks += di.len;
ba48a33e 956 __relocate_discard_cmd(dcc, prev_dc);
004b6862
CY		 957 di = prev_dc->di;
958 tdc = prev_dc;
959 merged = true;
960 }
961
962 if (next_dc && next_dc->state == D_PREP &&
963 next_dc->bdev == bdev &&
964 __is_discard_front_mergeable(&di, &next_dc->di)) {
965 next_dc->di.lstart = di.lstart;
966 next_dc->di.len += di.len;
967 next_dc->di.start = di.start;
d84d1cbd 968 dcc->undiscard_blks += di.len;
ba48a33e 969 __relocate_discard_cmd(dcc, next_dc);
004b6862
CY		 970 if (tdc)
971 __remove_discard_cmd(sbi, tdc);
004b6862 972 merged = true;
4e6a8d9b 973 }
004b6862 974
df0f6b44 975 if (!merged) {
004b6862
CY		 976 __insert_discard_tree(sbi, bdev, di.lstart, di.start,
977 di.len, NULL, NULL);
df0f6b44 978 }
004b6862
CY		 979 next:
980 prev_dc = next_dc;
981 if (!prev_dc)
982 break;
983
984 node = rb_next(&prev_dc->rb_node);
985 next_dc = rb_entry_safe(node, struct discard_cmd, rb_node);
986 }
987
988 mutex_unlock(&dcc->cmd_lock);
989}
990
991static int __queue_discard_cmd(struct f2fs_sb_info *sbi,
992 struct block_device *bdev, block_t blkstart, block_t blklen)
993{
994 block_t lblkstart = blkstart;
995
0243a5f9 996 trace_f2fs_queue_discard(bdev, blkstart, blklen);
004b6862
CY		 997
998 if (sbi->s_ndevs) {
999 int devi = f2fs_target_device_index(sbi, blkstart);
1000
1001 blkstart -= FDEV(devi).start_blk;
1002 }
1003 __update_discard_tree_range(sbi, bdev, lblkstart, blkstart, blklen);
004b6862
CY		 1004 return 0;
1005}
1006
bd5b0738
CY		 1007static void __issue_discard_cmd(struct f2fs_sb_info *sbi, bool issue_cond)
1008{
1009 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1010 struct list_head *pend_list;
1011 struct discard_cmd *dc, *tmp;
1012 struct blk_plug plug;
1013 int i, iter = 0;
1014
1015 mutex_lock(&dcc->cmd_lock);
963932a9
YH		 1016 f2fs_bug_on(sbi,
1017 !__check_rb_tree_consistence(sbi, &dcc->root));
bd5b0738
CY		 1018 blk_start_plug(&plug);
1019 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
1020 pend_list = &dcc->pend_list[i];
1021 list_for_each_entry_safe(dc, tmp, pend_list, list) {
1022 f2fs_bug_on(sbi, dc->state != D_PREP);
1023
1024 if (!issue_cond || is_idle(sbi))
1025 __submit_discard_cmd(sbi, dc);
1026 if (issue_cond && iter++ > DISCARD_ISSUE_RATE)
1027 goto out;
1028 }
1029 }
1030out:
1031 blk_finish_plug(&plug);
1032 mutex_unlock(&dcc->cmd_lock);
1033}
1034
2a510c00
CY		 1035static void __wait_one_discard_bio(struct f2fs_sb_info *sbi,
1036 struct discard_cmd *dc)
1037{
1038 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1039
1040 wait_for_completion_io(&dc->wait);
1041 mutex_lock(&dcc->cmd_lock);
1042 f2fs_bug_on(sbi, dc->state != D_DONE);
1043 dc->ref--;
1044 if (!dc->ref)
1045 __remove_discard_cmd(sbi, dc);
1046 mutex_unlock(&dcc->cmd_lock);
1047}
1048
63a94fa1
CY		 1049static void __wait_discard_cmd(struct f2fs_sb_info *sbi, bool wait_cond)
1050{
1051 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1052 struct list_head *wait_list = &(dcc->wait_list);
1053 struct discard_cmd *dc, *tmp;
6afae633
CY		 1054 bool need_wait;
1055
1056next:
1057 need_wait = false;
63a94fa1
CY		 1058
1059 mutex_lock(&dcc->cmd_lock);
1060 list_for_each_entry_safe(dc, tmp, wait_list, list) {
6afae633 1061 if (!wait_cond || (dc->state == D_DONE && !dc->ref)) {
63a94fa1
CY		 1062 wait_for_completion_io(&dc->wait);
1063 __remove_discard_cmd(sbi, dc);
6afae633
CY		 1064 } else {
1065 dc->ref++;
1066 need_wait = true;
1067 break;
63a94fa1
CY		 1068 }
1069 }
1070 mutex_unlock(&dcc->cmd_lock);
6afae633
CY		 1071
1072 if (need_wait) {
2a510c00 1073 __wait_one_discard_bio(sbi, dc);
6afae633
CY		 1074 goto next;
1075 }
63a94fa1
CY		 1076}
1077
004b6862
CY		 1078/* This should be covered by global mutex, &sit_i->sentry_lock */
1079void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
1080{
1081 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1082 struct discard_cmd *dc;
ec9895ad 1083 bool need_wait = false;
004b6862
CY		 1084
1085 mutex_lock(&dcc->cmd_lock);
004b6862
CY		 1086 dc = (struct discard_cmd *)__lookup_rb_tree(&dcc->root, NULL, blkaddr);
1087 if (dc) {
ec9895ad
CY		 1088 if (dc->state == D_PREP) {
1089 __punch_discard_cmd(sbi, dc, blkaddr);
1090 } else {
1091 dc->ref++;
1092 need_wait = true;
1093 }
275b66b0 1094 }
d431413f 1095 mutex_unlock(&dcc->cmd_lock);
ec9895ad 1096
2a510c00
CY		 1097 if (need_wait)
1098 __wait_one_discard_bio(sbi, dc);
d431413f
CY		 1099}
1100
cce13252
CY		 1101void stop_discard_thread(struct f2fs_sb_info *sbi)
1102{
1103 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1104
1105 if (dcc && dcc->f2fs_issue_discard) {
1106 struct task_struct *discard_thread = dcc->f2fs_issue_discard;
1107
1108 dcc->f2fs_issue_discard = NULL;
1109 kthread_stop(discard_thread);
ec9895ad 1110 }
d431413f
CY		 1111}
1112
1113/* This comes from f2fs_put_super */
1114void f2fs_wait_discard_bios(struct f2fs_sb_info *sbi)
1115{
bd5b0738 1116 __issue_discard_cmd(sbi, false);
63a94fa1 1117 __wait_discard_cmd(sbi, false);
275b66b0
CY		 1118}
1119
15469963
JK		 1120static int issue_discard_thread(void *data)
1121{
1122 struct f2fs_sb_info *sbi = data;
1123 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1124 wait_queue_head_t *q = &dcc->discard_wait_queue;
15469963 1125
1d7be270 1126 set_freezable();
15469963 1127
1d7be270
JK		 1128 do {
1129 wait_event_interruptible(*q, kthread_should_stop() ||
1130 freezing(current) ||
1131 atomic_read(&dcc->discard_cmd_cnt));
1132 if (try_to_freeze())
1133 continue;
1134 if (kthread_should_stop())
1135 return 0;
15469963 1136
dc6febb6
CY		 1137 sb_start_intwrite(sbi->sb);
1138
1d7be270
JK		 1139 __issue_discard_cmd(sbi, true);
1140 __wait_discard_cmd(sbi, true);
1141
dc6febb6
CY		 1142 sb_end_intwrite(sbi->sb);
1143
1d7be270
JK		 1144 congestion_wait(BLK_RW_SYNC, HZ/50);
1145 } while (!kthread_should_stop());
1146 return 0;
15469963
JK		 1147}
1148
f46e8809 1149#ifdef CONFIG_BLK_DEV_ZONED
3c62be17
JK		 1150static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
1151 struct block_device *bdev, block_t blkstart, block_t blklen)
f46e8809 1152{
92592285 1153 sector_t sector, nr_sects;
10a875f8 1154 block_t lblkstart = blkstart;
3c62be17
JK		 1155 int devi = 0;
1156
1157 if (sbi->s_ndevs) {
1158 devi = f2fs_target_device_index(sbi, blkstart);
1159 blkstart -= FDEV(devi).start_blk;
1160 }
f46e8809
DLM		 1161
1162 /*
1163 * We need to know the type of the zone: for conventional zones,
1164 * use regular discard if the drive supports it. For sequential
1165 * zones, reset the zone write pointer.
1166 */
3c62be17 1167 switch (get_blkz_type(sbi, bdev, blkstart)) {
f46e8809
DLM		 1168
1169 case BLK_ZONE_TYPE_CONVENTIONAL:
1170 if (!blk_queue_discard(bdev_get_queue(bdev)))
1171 return 0;
c81abe34 1172 return __queue_discard_cmd(sbi, bdev, lblkstart, blklen);
f46e8809
DLM		 1173 case BLK_ZONE_TYPE_SEQWRITE_REQ:
1174 case BLK_ZONE_TYPE_SEQWRITE_PREF:
92592285
JK		 1175 sector = SECTOR_FROM_BLOCK(blkstart);
1176 nr_sects = SECTOR_FROM_BLOCK(blklen);
1177
1178 if (sector & (bdev_zone_sectors(bdev) - 1) ||
1179 nr_sects != bdev_zone_sectors(bdev)) {
1180 f2fs_msg(sbi->sb, KERN_INFO,
1181 "(%d) %s: Unaligned discard attempted (block %x + %x)",
1182 devi, sbi->s_ndevs ? FDEV(devi).path: "",
1183 blkstart, blklen);
1184 return -EIO;
1185 }
d50aaeec 1186 trace_f2fs_issue_reset_zone(bdev, blkstart);
f46e8809
DLM		 1187 return blkdev_reset_zones(bdev, sector,
1188 nr_sects, GFP_NOFS);
1189 default:
1190 /* Unknown zone type: broken device ? */
1191 return -EIO;
1192 }
1193}
1194#endif
1195
3c62be17
JK		 1196static int __issue_discard_async(struct f2fs_sb_info *sbi,
1197 struct block_device *bdev, block_t blkstart, block_t blklen)
1198{
1199#ifdef CONFIG_BLK_DEV_ZONED
1200 if (f2fs_sb_mounted_blkzoned(sbi->sb) &&
1201 bdev_zoned_model(bdev) != BLK_ZONED_NONE)
1202 return __f2fs_issue_discard_zone(sbi, bdev, blkstart, blklen);
1203#endif
c81abe34 1204 return __queue_discard_cmd(sbi, bdev, blkstart, blklen);
3c62be17
JK		 1205}
1206
1e87a78d 1207static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
37208879
JK		 1208 block_t blkstart, block_t blklen)
1209{
3c62be17
JK		 1210 sector_t start = blkstart, len = 0;
1211 struct block_device *bdev;
a66cdd98
JK		 1212 struct seg_entry *se;
1213 unsigned int offset;
1214 block_t i;
3c62be17
JK		 1215 int err = 0;
1216
1217 bdev = f2fs_target_device(sbi, blkstart, NULL);
1218
1219 for (i = blkstart; i < blkstart + blklen; i++, len++) {
1220 if (i != start) {
1221 struct block_device *bdev2 =
1222 f2fs_target_device(sbi, i, NULL);
1223
1224 if (bdev2 != bdev) {
1225 err = __issue_discard_async(sbi, bdev,
1226 start, len);
1227 if (err)
1228 return err;
1229 bdev = bdev2;
1230 start = i;
1231 len = 0;
1232 }
1233 }
a66cdd98 1234
a66cdd98
JK		 1235 se = get_seg_entry(sbi, GET_SEGNO(sbi, i));
1236 offset = GET_BLKOFF_FROM_SEG0(sbi, i);
1237
1238 if (!f2fs_test_and_set_bit(offset, se->discard_map))
1239 sbi->discard_blks--;
1240 }
f46e8809 1241
3c62be17
JK		 1242 if (len)
1243 err = __issue_discard_async(sbi, bdev, start, len);
1244 return err;
1e87a78d
JK		 1245}
1246
25290fa5
JK		 1247static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc,
1248 bool check_only)
adf4983b 1249{
b2955550
JK		 1250 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
1251 int max_blocks = sbi->blocks_per_seg;
4b2fecc8 1252 struct seg_entry *se = get_seg_entry(sbi, cpc->trim_start);
b2955550
JK		 1253 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1254 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
a66cdd98 1255 unsigned long *discard_map = (unsigned long *)se->discard_map;
60a3b782 1256 unsigned long *dmap = SIT_I(sbi)->tmp_map;
b2955550 1257 unsigned int start = 0, end = -1;
c473f1a9 1258 bool force = (cpc->reason & CP_DISCARD);
a7eeb823 1259 struct discard_entry *de = NULL;
46f84c2c 1260 struct list_head *head = &SM_I(sbi)->dcc_info->entry_list;
b2955550
JK		 1261 int i;
1262
3e025740 1263 if (se->valid_blocks == max_blocks || !f2fs_discard_en(sbi))
25290fa5 1264 return false;
b2955550 1265
a66cdd98
JK		 1266 if (!force) {
1267 if (!test_opt(sbi, DISCARD) || !se->valid_blocks ||
0b54fb84
JK		 1268 SM_I(sbi)->dcc_info->nr_discards >=
1269 SM_I(sbi)->dcc_info->max_discards)
25290fa5 1270 return false;
4b2fecc8
JK		 1271 }
1272
b2955550
JK		 1273 /* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
1274 for (i = 0; i < entries; i++)
a66cdd98 1275 dmap[i] = force ? ~ckpt_map[i] & ~discard_map[i] :
d7bc2484 1276 (cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
b2955550 1277
0b54fb84
JK		 1278 while (force || SM_I(sbi)->dcc_info->nr_discards <=
1279 SM_I(sbi)->dcc_info->max_discards) {
b2955550
JK		 1280 start = __find_rev_next_bit(dmap, max_blocks, end + 1);
1281 if (start >= max_blocks)
1282 break;
1283
1284 end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1);
c7b41e16
YH		 1285 if (force && start && end != max_blocks
1286 && (end - start) < cpc->trim_minlen)
1287 continue;
1288
25290fa5
JK		 1289 if (check_only)
1290 return true;
1291
a7eeb823
CY		 1292 if (!de) {
1293 de = f2fs_kmem_cache_alloc(discard_entry_slab,
1294 GFP_F2FS_ZERO);
1295 de->start_blkaddr = START_BLOCK(sbi, cpc->trim_start);
1296 list_add_tail(&de->list, head);
1297 }
1298
1299 for (i = start; i < end; i++)
1300 __set_bit_le(i, (void *)de->discard_map);
1301
1302 SM_I(sbi)->dcc_info->nr_discards += end - start;
b2955550 1303 }
25290fa5 1304 return false;
b2955550
JK		 1305}
1306
4b2fecc8
JK		 1307void release_discard_addrs(struct f2fs_sb_info *sbi)
1308{
46f84c2c 1309 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
4b2fecc8
JK		 1310 struct discard_entry *entry, *this;
1311
1312 /* drop caches */
1313 list_for_each_entry_safe(entry, this, head, list) {
1314 list_del(&entry->list);
1315 kmem_cache_free(discard_entry_slab, entry);
1316 }
1317}
1318
0a8165d7 1319/*
351df4b2
JK		 1320 * Should call clear_prefree_segments after checkpoint is done.
1321 */
1322static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
1323{
1324 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
b65ee148 1325 unsigned int segno;
351df4b2
JK		 1326
1327 mutex_lock(&dirty_i->seglist_lock);
7cd8558b 1328 for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], MAIN_SEGS(sbi))
351df4b2 1329 __set_test_and_free(sbi, segno);
351df4b2
JK		 1330 mutex_unlock(&dirty_i->seglist_lock);
1331}
1332
836b5a63 1333void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2 1334{
46f84c2c 1335 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
2d7b822a 1336 struct discard_entry *entry, *this;
351df4b2 1337 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
29e59c14 1338 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
29e59c14 1339 unsigned int start = 0, end = -1;
36abef4e 1340 unsigned int secno, start_segno;
c473f1a9 1341 bool force = (cpc->reason & CP_DISCARD);
351df4b2
JK		 1342
1343 mutex_lock(&dirty_i->seglist_lock);
29e59c14 1344
351df4b2 1345 while (1) {
29e59c14 1346 int i;
7cd8558b
JK		 1347 start = find_next_bit(prefree_map, MAIN_SEGS(sbi), end + 1);
1348 if (start >= MAIN_SEGS(sbi))
351df4b2 1349 break;
7cd8558b
JK		 1350 end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
1351 start + 1);
29e59c14
CL		 1352
1353 for (i = start; i < end; i++)
1354 clear_bit(i, prefree_map);
1355
1356 dirty_i->nr_dirty[PRE] -= end - start;
1357
650d3c4e 1358 if (!test_opt(sbi, DISCARD))
29e59c14 1359 continue;
351df4b2 1360
650d3c4e
YH		 1361 if (force && start >= cpc->trim_start &&
1362 (end - 1) <= cpc->trim_end)
1363 continue;
1364
36abef4e
JK		 1365 if (!test_opt(sbi, LFS) || sbi->segs_per_sec == 1) {
1366 f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
37208879 1367 (end - start) << sbi->log_blocks_per_seg);
36abef4e
JK		 1368 continue;
1369 }
1370next:
4ddb1a4d
JK		 1371 secno = GET_SEC_FROM_SEG(sbi, start);
1372 start_segno = GET_SEG_FROM_SEC(sbi, secno);
36abef4e 1373 if (!IS_CURSEC(sbi, secno) &&
302bd348 1374 !get_valid_blocks(sbi, start, true))
36abef4e
JK		 1375 f2fs_issue_discard(sbi, START_BLOCK(sbi, start_segno),
1376 sbi->segs_per_sec << sbi->log_blocks_per_seg);
1377
1378 start = start_segno + sbi->segs_per_sec;
1379 if (start < end)
1380 goto next;
8b107f5b
JK		 1381 else
1382 end = start - 1;
351df4b2
JK		 1383 }
1384 mutex_unlock(&dirty_i->seglist_lock);
b2955550
JK		 1385
1386 /* send small discards */
2d7b822a 1387 list_for_each_entry_safe(entry, this, head, list) {
a7eeb823
CY		 1388 unsigned int cur_pos = 0, next_pos, len, total_len = 0;
1389 bool is_valid = test_bit_le(0, entry->discard_map);
1390
1391find_next:
1392 if (is_valid) {
1393 next_pos = find_next_zero_bit_le(entry->discard_map,
1394 sbi->blocks_per_seg, cur_pos);
1395 len = next_pos - cur_pos;
1396
acfd2810
DLM		 1397 if (f2fs_sb_mounted_blkzoned(sbi->sb) ||
1398 (force && len < cpc->trim_minlen))
a7eeb823
CY		 1399 goto skip;
1400
1401 f2fs_issue_discard(sbi, entry->start_blkaddr + cur_pos,
1402 len);
1403 cpc->trimmed += len;
1404 total_len += len;
1405 } else {
1406 next_pos = find_next_bit_le(entry->discard_map,
1407 sbi->blocks_per_seg, cur_pos);
1408 }
836b5a63 1409skip:
a7eeb823
CY		 1410 cur_pos = next_pos;
1411 is_valid = !is_valid;
1412
1413 if (cur_pos < sbi->blocks_per_seg)
1414 goto find_next;
1415
b2955550 1416 list_del(&entry->list);
a7eeb823 1417 SM_I(sbi)->dcc_info->nr_discards -= total_len;
b2955550
JK		 1418 kmem_cache_free(discard_entry_slab, entry);
1419 }
34e159da
CY		 1420
1421 wake_up(&SM_I(sbi)->dcc_info->discard_wait_queue);
351df4b2
JK		 1422}
1423
8ed59745 1424static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84 1425{
15469963 1426 dev_t dev = sbi->sb->s_bdev->bd_dev;
0b54fb84 1427 struct discard_cmd_control *dcc;
ba48a33e 1428 int err = 0, i;
0b54fb84
JK		 1429
1430 if (SM_I(sbi)->dcc_info) {
1431 dcc = SM_I(sbi)->dcc_info;
1432 goto init_thread;
1433 }
1434
1435 dcc = kzalloc(sizeof(struct discard_cmd_control), GFP_KERNEL);
1436 if (!dcc)
1437 return -ENOMEM;
1438
46f84c2c 1439 INIT_LIST_HEAD(&dcc->entry_list);
ba48a33e
CY		 1440 for (i = 0; i < MAX_PLIST_NUM; i++)
1441 INIT_LIST_HEAD(&dcc->pend_list[i]);
46f84c2c 1442 INIT_LIST_HEAD(&dcc->wait_list);
15469963 1443 mutex_init(&dcc->cmd_lock);
8b8dd65f
CY		 1444 atomic_set(&dcc->issued_discard, 0);
1445 atomic_set(&dcc->issing_discard, 0);
5f32366a 1446 atomic_set(&dcc->discard_cmd_cnt, 0);
0b54fb84 1447 dcc->nr_discards = 0;
d618ebaf 1448 dcc->max_discards = MAIN_SEGS(sbi) << sbi->log_blocks_per_seg;
d84d1cbd 1449 dcc->undiscard_blks = 0;
004b6862 1450 dcc->root = RB_ROOT;
0b54fb84 1451
15469963 1452 init_waitqueue_head(&dcc->discard_wait_queue);
0b54fb84
JK		 1453 SM_I(sbi)->dcc_info = dcc;
1454init_thread:
15469963
JK		 1455 dcc->f2fs_issue_discard = kthread_run(issue_discard_thread, sbi,
1456 "f2fs_discard-%u:%u", MAJOR(dev), MINOR(dev));
1457 if (IS_ERR(dcc->f2fs_issue_discard)) {
1458 err = PTR_ERR(dcc->f2fs_issue_discard);
1459 kfree(dcc);
1460 SM_I(sbi)->dcc_info = NULL;
1461 return err;
1462 }
1463
0b54fb84
JK		 1464 return err;
1465}
1466
f099405f 1467static void destroy_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84
JK		 1468{
1469 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1470
f099405f
CY		 1471 if (!dcc)
1472 return;
1473
cce13252 1474 stop_discard_thread(sbi);
f099405f
CY		 1475
1476 kfree(dcc);
1477 SM_I(sbi)->dcc_info = NULL;
0b54fb84
JK		 1478}
1479
184a5cd2 1480static bool __mark_sit_entry_dirty(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 1481{
1482 struct sit_info *sit_i = SIT_I(sbi);
184a5cd2
CY		 1483
1484 if (!__test_and_set_bit(segno, sit_i->dirty_sentries_bitmap)) {
351df4b2 1485 sit_i->dirty_sentries++;
184a5cd2
CY		 1486 return false;
1487 }
1488
1489 return true;
351df4b2
JK		 1490}
1491
1492static void __set_sit_entry_type(struct f2fs_sb_info *sbi, int type,
1493 unsigned int segno, int modified)
1494{
1495 struct seg_entry *se = get_seg_entry(sbi, segno);
1496 se->type = type;
1497 if (modified)
1498 __mark_sit_entry_dirty(sbi, segno);
1499}
1500
1501static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
1502{
1503 struct seg_entry *se;
1504 unsigned int segno, offset;
1505 long int new_vblocks;
1506
1507 segno = GET_SEGNO(sbi, blkaddr);
1508
1509 se = get_seg_entry(sbi, segno);
1510 new_vblocks = se->valid_blocks + del;
491c0854 1511 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
351df4b2 1512
9850cf4a 1513 f2fs_bug_on(sbi, (new_vblocks >> (sizeof(unsigned short) << 3) ||
351df4b2
JK		 1514 (new_vblocks > sbi->blocks_per_seg)));
1515
1516 se->valid_blocks = new_vblocks;
1517 se->mtime = get_mtime(sbi);
1518 SIT_I(sbi)->max_mtime = se->mtime;
1519
1520 /* Update valid block bitmap */
1521 if (del > 0) {
355e7891
CY		 1522 if (f2fs_test_and_set_bit(offset, se->cur_valid_map)) {
1523#ifdef CONFIG_F2FS_CHECK_FS
1524 if (f2fs_test_and_set_bit(offset,
1525 se->cur_valid_map_mir))
1526 f2fs_bug_on(sbi, 1);
1527 else
1528 WARN_ON(1);
1529#else
05796763 1530 f2fs_bug_on(sbi, 1);
355e7891
CY		 1531#endif
1532 }
3e025740
JK		 1533 if (f2fs_discard_en(sbi) &&
1534 !f2fs_test_and_set_bit(offset, se->discard_map))
a66cdd98 1535 sbi->discard_blks--;
720037f9
JK		 1536
1537 /* don't overwrite by SSR to keep node chain */
1538 if (se->type == CURSEG_WARM_NODE) {
1539 if (!f2fs_test_and_set_bit(offset, se->ckpt_valid_map))
1540 se->ckpt_valid_blocks++;
1541 }
351df4b2 1542 } else {
355e7891
CY		 1543 if (!f2fs_test_and_clear_bit(offset, se->cur_valid_map)) {
1544#ifdef CONFIG_F2FS_CHECK_FS
1545 if (!f2fs_test_and_clear_bit(offset,
1546 se->cur_valid_map_mir))
1547 f2fs_bug_on(sbi, 1);
1548 else
1549 WARN_ON(1);
1550#else
05796763 1551 f2fs_bug_on(sbi, 1);
355e7891
CY		 1552#endif
1553 }
3e025740
JK		 1554 if (f2fs_discard_en(sbi) &&
1555 f2fs_test_and_clear_bit(offset, se->discard_map))
a66cdd98 1556 sbi->discard_blks++;
351df4b2
JK		 1557 }
1558 if (!f2fs_test_bit(offset, se->ckpt_valid_map))
1559 se->ckpt_valid_blocks += del;
1560
1561 __mark_sit_entry_dirty(sbi, segno);
1562
1563 /* update total number of valid blocks to be written in ckpt area */
1564 SIT_I(sbi)->written_valid_blocks += del;
1565
1566 if (sbi->segs_per_sec > 1)
1567 get_sec_entry(sbi, segno)->valid_blocks += del;
1568}
1569
5e443818 1570void refresh_sit_entry(struct f2fs_sb_info *sbi, block_t old, block_t new)
351df4b2 1571{
5e443818
JK		 1572 update_sit_entry(sbi, new, 1);
1573 if (GET_SEGNO(sbi, old) != NULL_SEGNO)
1574 update_sit_entry(sbi, old, -1);
1575
1576 locate_dirty_segment(sbi, GET_SEGNO(sbi, old));
1577 locate_dirty_segment(sbi, GET_SEGNO(sbi, new));
351df4b2
JK		 1578}
1579
1580void invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
1581{
1582 unsigned int segno = GET_SEGNO(sbi, addr);
1583 struct sit_info *sit_i = SIT_I(sbi);
1584
9850cf4a 1585 f2fs_bug_on(sbi, addr == NULL_ADDR);
351df4b2
JK		 1586 if (addr == NEW_ADDR)
1587 return;
1588
1589 /* add it into sit main buffer */
1590 mutex_lock(&sit_i->sentry_lock);
1591
1592 update_sit_entry(sbi, addr, -1);
1593
1594 /* add it into dirty seglist */
1595 locate_dirty_segment(sbi, segno);
1596
1597 mutex_unlock(&sit_i->sentry_lock);
1598}
1599
6e2c64ad
JK		 1600bool is_checkpointed_data(struct f2fs_sb_info *sbi, block_t blkaddr)
1601{
1602 struct sit_info *sit_i = SIT_I(sbi);
1603 unsigned int segno, offset;
1604 struct seg_entry *se;
1605 bool is_cp = false;
1606
1607 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
1608 return true;
1609
1610 mutex_lock(&sit_i->sentry_lock);
1611
1612 segno = GET_SEGNO(sbi, blkaddr);
1613 se = get_seg_entry(sbi, segno);
1614 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
1615
1616 if (f2fs_test_bit(offset, se->ckpt_valid_map))
1617 is_cp = true;
1618
1619 mutex_unlock(&sit_i->sentry_lock);
1620
1621 return is_cp;
1622}
1623
0a8165d7 1624/*
351df4b2
JK		 1625 * This function should be resided under the curseg_mutex lock
1626 */
1627static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
e79efe3b 1628 struct f2fs_summary *sum)
351df4b2
JK		 1629{
1630 struct curseg_info *curseg = CURSEG_I(sbi, type);
1631 void *addr = curseg->sum_blk;
e79efe3b 1632 addr += curseg->next_blkoff * sizeof(struct f2fs_summary);
351df4b2 1633 memcpy(addr, sum, sizeof(struct f2fs_summary));
351df4b2
JK		 1634}
1635
0a8165d7 1636/*
351df4b2
JK		 1637 * Calculate the number of current summary pages for writing
1638 */
3fa06d7b 1639int npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
351df4b2 1640{
351df4b2 1641 int valid_sum_count = 0;
9a47938b 1642 int i, sum_in_page;
351df4b2
JK		 1643
1644 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1645 if (sbi->ckpt->alloc_type[i] == SSR)
1646 valid_sum_count += sbi->blocks_per_seg;
3fa06d7b
CY		 1647 else {
1648 if (for_ra)
1649 valid_sum_count += le16_to_cpu(
1650 F2FS_CKPT(sbi)->cur_data_blkoff[i]);
1651 else
1652 valid_sum_count += curseg_blkoff(sbi, i);
1653 }
351df4b2
JK		 1654 }
1655
09cbfeaf 1656 sum_in_page = (PAGE_SIZE - 2 * SUM_JOURNAL_SIZE -
9a47938b
FL		 1657 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
1658 if (valid_sum_count <= sum_in_page)
351df4b2 1659 return 1;
9a47938b 1660 else if ((valid_sum_count - sum_in_page) <=
09cbfeaf 1661 (PAGE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
351df4b2
JK		 1662 return 2;
1663 return 3;
1664}
1665
0a8165d7 1666/*
351df4b2
JK		 1667 * Caller should put this summary page
1668 */
1669struct page *get_sum_page(struct f2fs_sb_info *sbi, unsigned int segno)
1670{
1671 return get_meta_page(sbi, GET_SUM_BLOCK(sbi, segno));
1672}
1673
381722d2 1674void update_meta_page(struct f2fs_sb_info *sbi, void *src, block_t blk_addr)
351df4b2
JK		 1675{
1676 struct page *page = grab_meta_page(sbi, blk_addr);
381722d2
CY		 1677 void *dst = page_address(page);
1678
1679 if (src)
09cbfeaf 1680 memcpy(dst, src, PAGE_SIZE);
381722d2 1681 else
09cbfeaf 1682 memset(dst, 0, PAGE_SIZE);
351df4b2
JK		 1683 set_page_dirty(page);
1684 f2fs_put_page(page, 1);
1685}
1686
381722d2
CY		 1687static void write_sum_page(struct f2fs_sb_info *sbi,
1688 struct f2fs_summary_block *sum_blk, block_t blk_addr)
1689{
1690 update_meta_page(sbi, (void *)sum_blk, blk_addr);
1691}
1692
b7ad7512
CY		 1693static void write_current_sum_page(struct f2fs_sb_info *sbi,
1694 int type, block_t blk_addr)
1695{
1696 struct curseg_info *curseg = CURSEG_I(sbi, type);
1697 struct page *page = grab_meta_page(sbi, blk_addr);
1698 struct f2fs_summary_block *src = curseg->sum_blk;
1699 struct f2fs_summary_block *dst;
1700
1701 dst = (struct f2fs_summary_block *)page_address(page);
1702
1703 mutex_lock(&curseg->curseg_mutex);
1704
1705 down_read(&curseg->journal_rwsem);
1706 memcpy(&dst->journal, curseg->journal, SUM_JOURNAL_SIZE);
1707 up_read(&curseg->journal_rwsem);
1708
1709 memcpy(dst->entries, src->entries, SUM_ENTRY_SIZE);
1710 memcpy(&dst->footer, &src->footer, SUM_FOOTER_SIZE);
1711
1712 mutex_unlock(&curseg->curseg_mutex);
1713
1714 set_page_dirty(page);
1715 f2fs_put_page(page, 1);
1716}
1717
a7881893
JK		 1718static int is_next_segment_free(struct f2fs_sb_info *sbi, int type)
1719{
1720 struct curseg_info *curseg = CURSEG_I(sbi, type);
1721 unsigned int segno = curseg->segno + 1;
1722 struct free_segmap_info *free_i = FREE_I(sbi);
1723
1724 if (segno < MAIN_SEGS(sbi) && segno % sbi->segs_per_sec)
1725 return !test_bit(segno, free_i->free_segmap);
1726 return 0;
1727}
1728
0a8165d7 1729/*
351df4b2
JK		 1730 * Find a new segment from the free segments bitmap to right order
1731 * This function should be returned with success, otherwise BUG
1732 */
1733static void get_new_segment(struct f2fs_sb_info *sbi,
1734 unsigned int *newseg, bool new_sec, int dir)
1735{
1736 struct free_segmap_info *free_i = FREE_I(sbi);
351df4b2 1737 unsigned int segno, secno, zoneno;
7cd8558b 1738 unsigned int total_zones = MAIN_SECS(sbi) / sbi->secs_per_zone;
4ddb1a4d
JK		 1739 unsigned int hint = GET_SEC_FROM_SEG(sbi, *newseg);
1740 unsigned int old_zoneno = GET_ZONE_FROM_SEG(sbi, *newseg);
351df4b2
JK		 1741 unsigned int left_start = hint;
1742 bool init = true;
1743 int go_left = 0;
1744 int i;
1745
1a118ccf 1746 spin_lock(&free_i->segmap_lock);
351df4b2
JK		 1747
1748 if (!new_sec && ((*newseg + 1) % sbi->segs_per_sec)) {
1749 segno = find_next_zero_bit(free_i->free_segmap,
4ddb1a4d
JK		 1750 GET_SEG_FROM_SEC(sbi, hint + 1), *newseg + 1);
1751 if (segno < GET_SEG_FROM_SEC(sbi, hint + 1))
351df4b2
JK		 1752 goto got_it;
1753 }
1754find_other_zone:
7cd8558b
JK		 1755 secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
1756 if (secno >= MAIN_SECS(sbi)) {
351df4b2
JK		 1757 if (dir == ALLOC_RIGHT) {
1758 secno = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1759 MAIN_SECS(sbi), 0);
1760 f2fs_bug_on(sbi, secno >= MAIN_SECS(sbi));
351df4b2
JK		 1761 } else {
1762 go_left = 1;
1763 left_start = hint - 1;
1764 }
1765 }
1766 if (go_left == 0)
1767 goto skip_left;
1768
1769 while (test_bit(left_start, free_i->free_secmap)) {
1770 if (left_start > 0) {
1771 left_start--;
1772 continue;
1773 }
1774 left_start = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1775 MAIN_SECS(sbi), 0);
1776 f2fs_bug_on(sbi, left_start >= MAIN_SECS(sbi));
351df4b2
JK		 1777 break;
1778 }
1779 secno = left_start;
1780skip_left:
1781 hint = secno;
4ddb1a4d
JK		 1782 segno = GET_SEG_FROM_SEC(sbi, secno);
1783 zoneno = GET_ZONE_FROM_SEC(sbi, secno);
351df4b2
JK		 1784
1785 /* give up on finding another zone */
1786 if (!init)
1787 goto got_it;
1788 if (sbi->secs_per_zone == 1)
1789 goto got_it;
1790 if (zoneno == old_zoneno)
1791 goto got_it;
1792 if (dir == ALLOC_LEFT) {
1793 if (!go_left && zoneno + 1 >= total_zones)
1794 goto got_it;
1795 if (go_left && zoneno == 0)
1796 goto got_it;
1797 }
1798 for (i = 0; i < NR_CURSEG_TYPE; i++)
1799 if (CURSEG_I(sbi, i)->zone == zoneno)
1800 break;
1801
1802 if (i < NR_CURSEG_TYPE) {
1803 /* zone is in user, try another */
1804 if (go_left)
1805 hint = zoneno * sbi->secs_per_zone - 1;
1806 else if (zoneno + 1 >= total_zones)
1807 hint = 0;
1808 else
1809 hint = (zoneno + 1) * sbi->secs_per_zone;
1810 init = false;
1811 goto find_other_zone;
1812 }
1813got_it:
1814 /* set it as dirty segment in free segmap */
9850cf4a 1815 f2fs_bug_on(sbi, test_bit(segno, free_i->free_segmap));
351df4b2
JK		 1816 __set_inuse(sbi, segno);
1817 *newseg = segno;
1a118ccf 1818 spin_unlock(&free_i->segmap_lock);
351df4b2
JK		 1819}
1820
1821static void reset_curseg(struct f2fs_sb_info *sbi, int type, int modified)
1822{
1823 struct curseg_info *curseg = CURSEG_I(sbi, type);
1824 struct summary_footer *sum_footer;
1825
1826 curseg->segno = curseg->next_segno;
4ddb1a4d 1827 curseg->zone = GET_ZONE_FROM_SEG(sbi, curseg->segno);
351df4b2
JK		 1828 curseg->next_blkoff = 0;
1829 curseg->next_segno = NULL_SEGNO;
1830
1831 sum_footer = &(curseg->sum_blk->footer);
1832 memset(sum_footer, 0, sizeof(struct summary_footer));
1833 if (IS_DATASEG(type))
1834 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
1835 if (IS_NODESEG(type))
1836 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
1837 __set_sit_entry_type(sbi, type, curseg->segno, modified);
1838}
1839
7a20b8a6
JK		 1840static unsigned int __get_next_segno(struct f2fs_sb_info *sbi, int type)
1841{
a7881893
JK		 1842 /* if segs_per_sec is large than 1, we need to keep original policy. */
1843 if (sbi->segs_per_sec != 1)
1844 return CURSEG_I(sbi, type)->segno;
1845
7a20b8a6
JK		 1846 if (type == CURSEG_HOT_DATA || IS_NODESEG(type))
1847 return 0;
1848
e066b83c
JK		 1849 if (SIT_I(sbi)->last_victim[ALLOC_NEXT])
1850 return SIT_I(sbi)->last_victim[ALLOC_NEXT];
7a20b8a6
JK		 1851 return CURSEG_I(sbi, type)->segno;
1852}
1853
0a8165d7 1854/*
351df4b2
JK		 1855 * Allocate a current working segment.
1856 * This function always allocates a free segment in LFS manner.
1857 */
1858static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
1859{
1860 struct curseg_info *curseg = CURSEG_I(sbi, type);
1861 unsigned int segno = curseg->segno;
1862 int dir = ALLOC_LEFT;
1863
1864 write_sum_page(sbi, curseg->sum_blk,
81fb5e87 1865 GET_SUM_BLOCK(sbi, segno));
351df4b2
JK		 1866 if (type == CURSEG_WARM_DATA || type == CURSEG_COLD_DATA)
1867 dir = ALLOC_RIGHT;
1868
1869 if (test_opt(sbi, NOHEAP))
1870 dir = ALLOC_RIGHT;
1871
7a20b8a6 1872 segno = __get_next_segno(sbi, type);
351df4b2
JK		 1873 get_new_segment(sbi, &segno, new_sec, dir);
1874 curseg->next_segno = segno;
1875 reset_curseg(sbi, type, 1);
1876 curseg->alloc_type = LFS;
1877}
1878
1879static void __next_free_blkoff(struct f2fs_sb_info *sbi,
1880 struct curseg_info *seg, block_t start)
1881{
1882 struct seg_entry *se = get_seg_entry(sbi, seg->segno);
e81c93cf 1883 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
60a3b782 1884 unsigned long *target_map = SIT_I(sbi)->tmp_map;
e81c93cf
CL		 1885 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
1886 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1887 int i, pos;
1888
1889 for (i = 0; i < entries; i++)
1890 target_map[i] = ckpt_map[i] | cur_map[i];
1891
1892 pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, start);
1893
1894 seg->next_blkoff = pos;
351df4b2
JK		 1895}
1896
0a8165d7 1897/*
351df4b2
JK		 1898 * If a segment is written by LFS manner, next block offset is just obtained
1899 * by increasing the current block offset. However, if a segment is written by
1900 * SSR manner, next block offset obtained by calling __next_free_blkoff
1901 */
1902static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
1903 struct curseg_info *seg)
1904{
1905 if (seg->alloc_type == SSR)
1906 __next_free_blkoff(sbi, seg, seg->next_blkoff + 1);
1907 else
1908 seg->next_blkoff++;
1909}
1910
0a8165d7 1911/*
e1c42045 1912 * This function always allocates a used segment(from dirty seglist) by SSR
351df4b2
JK		 1913 * manner, so it should recover the existing segment information of valid blocks
1914 */
1915static void change_curseg(struct f2fs_sb_info *sbi, int type, bool reuse)
1916{
1917 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
1918 struct curseg_info *curseg = CURSEG_I(sbi, type);
1919 unsigned int new_segno = curseg->next_segno;
1920 struct f2fs_summary_block *sum_node;
1921 struct page *sum_page;
1922
1923 write_sum_page(sbi, curseg->sum_blk,
1924 GET_SUM_BLOCK(sbi, curseg->segno));
1925 __set_test_and_inuse(sbi, new_segno);
1926
1927 mutex_lock(&dirty_i->seglist_lock);
1928 __remove_dirty_segment(sbi, new_segno, PRE);
1929 __remove_dirty_segment(sbi, new_segno, DIRTY);
1930 mutex_unlock(&dirty_i->seglist_lock);
1931
1932 reset_curseg(sbi, type, 1);
1933 curseg->alloc_type = SSR;
1934 __next_free_blkoff(sbi, curseg, 0);
1935
1936 if (reuse) {
1937 sum_page = get_sum_page(sbi, new_segno);
1938 sum_node = (struct f2fs_summary_block *)page_address(sum_page);
1939 memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
1940 f2fs_put_page(sum_page, 1);
1941 }
1942}
1943
43727527
JK		 1944static int get_ssr_segment(struct f2fs_sb_info *sbi, int type)
1945{
1946 struct curseg_info *curseg = CURSEG_I(sbi, type);
1947 const struct victim_selection *v_ops = DIRTY_I(sbi)->v_ops;
e066b83c 1948 unsigned segno = NULL_SEGNO;
d27c3d89
CY		 1949 int i, cnt;
1950 bool reversed = false;
c192f7a4
JK		 1951
1952 /* need_SSR() already forces to do this */
e066b83c
JK		 1953 if (v_ops->get_victim(sbi, &segno, BG_GC, type, SSR)) {
1954 curseg->next_segno = segno;
c192f7a4 1955 return 1;
e066b83c 1956 }
43727527 1957
70d625cb
JK		 1958 /* For node segments, let's do SSR more intensively */
1959 if (IS_NODESEG(type)) {
d27c3d89
CY		 1960 if (type >= CURSEG_WARM_NODE) {
1961 reversed = true;
1962 i = CURSEG_COLD_NODE;
1963 } else {
1964 i = CURSEG_HOT_NODE;
1965 }
1966 cnt = NR_CURSEG_NODE_TYPE;
70d625cb 1967 } else {
d27c3d89
CY		 1968 if (type >= CURSEG_WARM_DATA) {
1969 reversed = true;
1970 i = CURSEG_COLD_DATA;
1971 } else {
1972 i = CURSEG_HOT_DATA;
1973 }
1974 cnt = NR_CURSEG_DATA_TYPE;
70d625cb 1975 }
43727527 1976
d27c3d89 1977 for (; cnt-- > 0; reversed ? i-- : i++) {
c192f7a4
JK		 1978 if (i == type)
1979 continue;
e066b83c
JK		 1980 if (v_ops->get_victim(sbi, &segno, BG_GC, i, SSR)) {
1981 curseg->next_segno = segno;
43727527 1982 return 1;
e066b83c 1983 }
c192f7a4 1984 }
43727527
JK		 1985 return 0;
1986}
1987
351df4b2
JK		 1988/*
1989 * flush out current segment and replace it with new segment
1990 * This function should be returned with success, otherwise BUG
1991 */
1992static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
1993 int type, bool force)
1994{
a7881893
JK		 1995 struct curseg_info *curseg = CURSEG_I(sbi, type);
1996
7b405275 1997 if (force)
351df4b2 1998 new_curseg(sbi, type, true);
5b6c6be2
JK		 1999 else if (!is_set_ckpt_flags(sbi, CP_CRC_RECOVERY_FLAG) &&
2000 type == CURSEG_WARM_NODE)
351df4b2 2001 new_curseg(sbi, type, false);
a7881893
JK		 2002 else if (curseg->alloc_type == LFS && is_next_segment_free(sbi, type))
2003 new_curseg(sbi, type, false);
351df4b2
JK		 2004 else if (need_SSR(sbi) && get_ssr_segment(sbi, type))
2005 change_curseg(sbi, type, true);
2006 else
2007 new_curseg(sbi, type, false);
dcdfff65 2008
a7881893 2009 stat_inc_seg_type(sbi, curseg);
351df4b2
JK		 2010}
2011
2012void allocate_new_segments(struct f2fs_sb_info *sbi)
2013{
6ae1be13
JK		 2014 struct curseg_info *curseg;
2015 unsigned int old_segno;
351df4b2
JK		 2016 int i;
2017
6ae1be13
JK		 2018 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2019 curseg = CURSEG_I(sbi, i);
2020 old_segno = curseg->segno;
2021 SIT_I(sbi)->s_ops->allocate_segment(sbi, i, true);
2022 locate_dirty_segment(sbi, old_segno);
2023 }
351df4b2
JK		 2024}
2025
2026static const struct segment_allocation default_salloc_ops = {
2027 .allocate_segment = allocate_segment_by_default,
2028};
2029
25290fa5
JK		 2030bool exist_trim_candidates(struct f2fs_sb_info *sbi, struct cp_control *cpc)
2031{
2032 __u64 trim_start = cpc->trim_start;
2033 bool has_candidate = false;
2034
2035 mutex_lock(&SIT_I(sbi)->sentry_lock);
2036 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++) {
2037 if (add_discard_addrs(sbi, cpc, true)) {
2038 has_candidate = true;
2039 break;
2040 }
2041 }
2042 mutex_unlock(&SIT_I(sbi)->sentry_lock);
2043
2044 cpc->trim_start = trim_start;
2045 return has_candidate;
2046}
2047
4b2fecc8
JK		 2048int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
2049{
f7ef9b83
JK		 2050 __u64 start = F2FS_BYTES_TO_BLK(range->start);
2051 __u64 end = start + F2FS_BYTES_TO_BLK(range->len) - 1;
4b2fecc8
JK		 2052 unsigned int start_segno, end_segno;
2053 struct cp_control cpc;
c34f42e2 2054 int err = 0;
4b2fecc8 2055
836b5a63 2056 if (start >= MAX_BLKADDR(sbi) || range->len < sbi->blocksize)
4b2fecc8
JK		 2057 return -EINVAL;
2058
9bd27ae4 2059 cpc.trimmed = 0;
7cd8558b 2060 if (end <= MAIN_BLKADDR(sbi))
4b2fecc8
JK		 2061 goto out;
2062
ed214a11
YH		 2063 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
2064 f2fs_msg(sbi->sb, KERN_WARNING,
2065 "Found FS corruption, run fsck to fix.");
2066 goto out;
2067 }
2068
4b2fecc8 2069 /* start/end segment number in main_area */
7cd8558b
JK		 2070 start_segno = (start <= MAIN_BLKADDR(sbi)) ? 0 : GET_SEGNO(sbi, start);
2071 end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
2072 GET_SEGNO(sbi, end);
4b2fecc8 2073 cpc.reason = CP_DISCARD;
836b5a63 2074 cpc.trim_minlen = max_t(__u64, 1, F2FS_BYTES_TO_BLK(range->minlen));
4b2fecc8
JK		 2075
2076 /* do checkpoint to issue discard commands safely */
bba681cb
JK		 2077 for (; start_segno <= end_segno; start_segno = cpc.trim_end + 1) {
2078 cpc.trim_start = start_segno;
a66cdd98
JK		 2079
2080 if (sbi->discard_blks == 0)
2081 break;
2082 else if (sbi->discard_blks < BATCHED_TRIM_BLOCKS(sbi))
2083 cpc.trim_end = end_segno;
2084 else
2085 cpc.trim_end = min_t(unsigned int,
2086 rounddown(start_segno +
bba681cb
JK		 2087 BATCHED_TRIM_SEGMENTS(sbi),
2088 sbi->segs_per_sec) - 1, end_segno);
2089
2090 mutex_lock(&sbi->gc_mutex);
c34f42e2 2091 err = write_checkpoint(sbi, &cpc);
bba681cb 2092 mutex_unlock(&sbi->gc_mutex);
e9328353
CY		 2093 if (err)
2094 break;
74fa5f3d
CY		 2095
2096 schedule();
bba681cb 2097 }
4b2fecc8 2098out:
f7ef9b83 2099 range->len = F2FS_BLK_TO_BYTES(cpc.trimmed);
c34f42e2 2100 return err;
4b2fecc8
JK		 2101}
2102
351df4b2
JK		 2103static bool __has_curseg_space(struct f2fs_sb_info *sbi, int type)
2104{
2105 struct curseg_info *curseg = CURSEG_I(sbi, type);
2106 if (curseg->next_blkoff < sbi->blocks_per_seg)
2107 return true;
2108 return false;
2109}
2110
81377bd6 2111static int __get_segment_type_2(struct f2fs_io_info *fio)
351df4b2 2112{
81377bd6 2113 if (fio->type == DATA)
351df4b2
JK		 2114 return CURSEG_HOT_DATA;
2115 else
2116 return CURSEG_HOT_NODE;
2117}
2118
81377bd6 2119static int __get_segment_type_4(struct f2fs_io_info *fio)
351df4b2 2120{
81377bd6
JK		 2121 if (fio->type == DATA) {
2122 struct inode *inode = fio->page->mapping->host;
351df4b2
JK		 2123
2124 if (S_ISDIR(inode->i_mode))
2125 return CURSEG_HOT_DATA;
2126 else
2127 return CURSEG_COLD_DATA;
2128 } else {
81377bd6 2129 if (IS_DNODE(fio->page) && is_cold_node(fio->page))
a344b9fd 2130 return CURSEG_WARM_NODE;
351df4b2
JK		 2131 else
2132 return CURSEG_COLD_NODE;
2133 }
2134}
2135
81377bd6 2136static int __get_segment_type_6(struct f2fs_io_info *fio)
351df4b2 2137{
81377bd6
JK		 2138 if (fio->type == DATA) {
2139 struct inode *inode = fio->page->mapping->host;
351df4b2 2140
81377bd6 2141 if (is_cold_data(fio->page) || file_is_cold(inode))
351df4b2 2142 return CURSEG_COLD_DATA;
ef095d19
JK		 2143 if (is_inode_flag_set(inode, FI_HOT_DATA))
2144 return CURSEG_HOT_DATA;
2145 return CURSEG_WARM_DATA;
351df4b2 2146 } else {
81377bd6
JK		 2147 if (IS_DNODE(fio->page))
2148 return is_cold_node(fio->page) ? CURSEG_WARM_NODE :
351df4b2 2149 CURSEG_HOT_NODE;
ef095d19 2150 return CURSEG_COLD_NODE;
351df4b2
JK		 2151 }
2152}
2153
81377bd6 2154static int __get_segment_type(struct f2fs_io_info *fio)
351df4b2 2155{
a912b54d
JK		 2156 int type = 0;
2157
81377bd6 2158 switch (fio->sbi->active_logs) {
351df4b2 2159 case 2:
a912b54d
JK		 2160 type = __get_segment_type_2(fio);
2161 break;
351df4b2 2162 case 4:
a912b54d
JK		 2163 type = __get_segment_type_4(fio);
2164 break;
2165 case 6:
2166 type = __get_segment_type_6(fio);
2167 break;
2168 default:
2169 f2fs_bug_on(fio->sbi, true);
351df4b2 2170 }
81377bd6 2171
a912b54d
JK		 2172 if (IS_HOT(type))
2173 fio->temp = HOT;
2174 else if (IS_WARM(type))
2175 fio->temp = WARM;
2176 else
2177 fio->temp = COLD;
2178 return type;
351df4b2
JK		 2179}
2180
bfad7c2d
JK		 2181void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
2182 block_t old_blkaddr, block_t *new_blkaddr,
fb830fc5
CY		 2183 struct f2fs_summary *sum, int type,
2184 struct f2fs_io_info *fio, bool add_list)
351df4b2
JK		 2185{
2186 struct sit_info *sit_i = SIT_I(sbi);
6ae1be13 2187 struct curseg_info *curseg = CURSEG_I(sbi, type);
351df4b2
JK		 2188
2189 mutex_lock(&curseg->curseg_mutex);
21cb1d99 2190 mutex_lock(&sit_i->sentry_lock);
351df4b2
JK		 2191
2192 *new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
351df4b2 2193
4e6a8d9b
JK		 2194 f2fs_wait_discard_bio(sbi, *new_blkaddr);
2195
351df4b2
JK		 2196 /*
2197 * __add_sum_entry should be resided under the curseg_mutex
2198 * because, this function updates a summary entry in the
2199 * current summary block.
2200 */
e79efe3b 2201 __add_sum_entry(sbi, type, sum);
351df4b2 2202
351df4b2 2203 __refresh_next_blkoff(sbi, curseg);
dcdfff65
JK		 2204
2205 stat_inc_block_count(sbi, curseg);
351df4b2 2206
c6f82fe9
JK		 2207 if (!__has_curseg_space(sbi, type))
2208 sit_i->s_ops->allocate_segment(sbi, type, false);
351df4b2 2209 /*
c6f82fe9
JK		 2210 * SIT information should be updated after segment allocation,
2211 * since we need to keep dirty segments precisely under SSR.
351df4b2
JK		 2212 */
2213 refresh_sit_entry(sbi, old_blkaddr, *new_blkaddr);
5e443818 2214
351df4b2
JK		 2215 mutex_unlock(&sit_i->sentry_lock);
2216
704956ec 2217 if (page && IS_NODESEG(type)) {
351df4b2
JK		 2218 fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
2219
704956ec
CY		 2220 f2fs_inode_chksum_set(sbi, page);
2221 }
2222
fb830fc5
CY		 2223 if (add_list) {
2224 struct f2fs_bio_info *io;
2225
2226 INIT_LIST_HEAD(&fio->list);
2227 fio->in_list = true;
2228 io = sbi->write_io[fio->type] + fio->temp;
2229 spin_lock(&io->io_lock);
2230 list_add_tail(&fio->list, &io->io_list);
2231 spin_unlock(&io->io_lock);
2232 }
2233
bfad7c2d
JK		 2234 mutex_unlock(&curseg->curseg_mutex);
2235}
2236
05ca3632 2237static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
bfad7c2d 2238{
81377bd6 2239 int type = __get_segment_type(fio);
0a595eba 2240 int err;
bfad7c2d 2241
0a595eba 2242reallocate:
7a9d7548 2243 allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
fb830fc5 2244 &fio->new_blkaddr, sum, type, fio, true);
bfad7c2d 2245
351df4b2 2246 /* writeout dirty page into bdev */
b9109b0e 2247 err = f2fs_submit_page_write(fio);
0a595eba
JK		 2248 if (err == -EAGAIN) {
2249 fio->old_blkaddr = fio->new_blkaddr;
2250 goto reallocate;
2251 }
351df4b2
JK		 2252}
2253
577e3495 2254void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
351df4b2 2255{
458e6197 2256 struct f2fs_io_info fio = {
05ca3632 2257 .sbi = sbi,
458e6197 2258 .type = META,
04d328de 2259 .op = REQ_OP_WRITE,
70fd7614 2260 .op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
7a9d7548
CY		 2261 .old_blkaddr = page->index,
2262 .new_blkaddr = page->index,
05ca3632 2263 .page = page,
4375a336 2264 .encrypted_page = NULL,
fb830fc5 2265 .in_list = false,
458e6197
JK		 2266 };
2267
2b947003 2268 if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
04d328de 2269 fio.op_flags &= ~REQ_META;
2b947003 2270
351df4b2 2271 set_page_writeback(page);
b9109b0e 2272 f2fs_submit_page_write(&fio);
351df4b2
JK		 2273}
2274
05ca3632 2275void write_node_page(unsigned int nid, struct f2fs_io_info *fio)
351df4b2
JK		 2276{
2277 struct f2fs_summary sum;
05ca3632 2278
351df4b2 2279 set_summary(&sum, nid, 0, 0);
05ca3632 2280 do_write_page(&sum, fio);
351df4b2
JK		 2281}
2282
05ca3632 2283void write_data_page(struct dnode_of_data *dn, struct f2fs_io_info *fio)
351df4b2 2284{
05ca3632 2285 struct f2fs_sb_info *sbi = fio->sbi;
351df4b2
JK		 2286 struct f2fs_summary sum;
2287 struct node_info ni;
2288
9850cf4a 2289 f2fs_bug_on(sbi, dn->data_blkaddr == NULL_ADDR);
351df4b2
JK		 2290 get_node_info(sbi, dn->nid, &ni);
2291 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
05ca3632 2292 do_write_page(&sum, fio);
f28b3434 2293 f2fs_update_data_blkaddr(dn, fio->new_blkaddr);
351df4b2
JK		 2294}
2295
d1b3e72d 2296int rewrite_data_page(struct f2fs_io_info *fio)
351df4b2 2297{
7a9d7548 2298 fio->new_blkaddr = fio->old_blkaddr;
05ca3632 2299 stat_inc_inplace_blocks(fio->sbi);
d1b3e72d 2300 return f2fs_submit_page_bio(fio);
351df4b2
JK		 2301}
2302
4356e48e 2303void __f2fs_replace_block(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
19f106bc 2304 block_t old_blkaddr, block_t new_blkaddr,
28bc106b 2305 bool recover_curseg, bool recover_newaddr)
351df4b2
JK		 2306{
2307 struct sit_info *sit_i = SIT_I(sbi);
2308 struct curseg_info *curseg;
2309 unsigned int segno, old_cursegno;
2310 struct seg_entry *se;
2311 int type;
19f106bc 2312 unsigned short old_blkoff;
351df4b2
JK		 2313
2314 segno = GET_SEGNO(sbi, new_blkaddr);
2315 se = get_seg_entry(sbi, segno);
2316 type = se->type;
2317
19f106bc
CY		 2318 if (!recover_curseg) {
2319 /* for recovery flow */
2320 if (se->valid_blocks == 0 && !IS_CURSEG(sbi, segno)) {
2321 if (old_blkaddr == NULL_ADDR)
2322 type = CURSEG_COLD_DATA;
2323 else
2324 type = CURSEG_WARM_DATA;
2325 }
2326 } else {
2327 if (!IS_CURSEG(sbi, segno))
351df4b2
JK		 2328 type = CURSEG_WARM_DATA;
2329 }
19f106bc 2330
351df4b2
JK		 2331 curseg = CURSEG_I(sbi, type);
2332
2333 mutex_lock(&curseg->curseg_mutex);
2334 mutex_lock(&sit_i->sentry_lock);
2335
2336 old_cursegno = curseg->segno;
19f106bc 2337 old_blkoff = curseg->next_blkoff;
351df4b2
JK		 2338
2339 /* change the current segment */
2340 if (segno != curseg->segno) {
2341 curseg->next_segno = segno;
2342 change_curseg(sbi, type, true);
2343 }
2344
491c0854 2345 curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
e79efe3b 2346 __add_sum_entry(sbi, type, sum);
351df4b2 2347
28bc106b 2348 if (!recover_curseg || recover_newaddr)
6e2c64ad
JK		 2349 update_sit_entry(sbi, new_blkaddr, 1);
2350 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
2351 update_sit_entry(sbi, old_blkaddr, -1);
2352
2353 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
2354 locate_dirty_segment(sbi, GET_SEGNO(sbi, new_blkaddr));
2355
351df4b2 2356 locate_dirty_segment(sbi, old_cursegno);
351df4b2 2357
19f106bc
CY		 2358 if (recover_curseg) {
2359 if (old_cursegno != curseg->segno) {
2360 curseg->next_segno = old_cursegno;
2361 change_curseg(sbi, type, true);
2362 }
2363 curseg->next_blkoff = old_blkoff;
2364 }
2365
351df4b2
JK		 2366 mutex_unlock(&sit_i->sentry_lock);
2367 mutex_unlock(&curseg->curseg_mutex);
2368}
2369
528e3459
CY		 2370void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
2371 block_t old_addr, block_t new_addr,
28bc106b
CY		 2372 unsigned char version, bool recover_curseg,
2373 bool recover_newaddr)
528e3459
CY		 2374{
2375 struct f2fs_summary sum;
2376
2377 set_summary(&sum, dn->nid, dn->ofs_in_node, version);
2378
28bc106b
CY		 2379 __f2fs_replace_block(sbi, &sum, old_addr, new_addr,
2380 recover_curseg, recover_newaddr);
528e3459 2381
f28b3434 2382 f2fs_update_data_blkaddr(dn, new_addr);
528e3459
CY		 2383}
2384
93dfe2ac 2385void f2fs_wait_on_page_writeback(struct page *page,
fec1d657 2386 enum page_type type, bool ordered)
93dfe2ac 2387{
93dfe2ac 2388 if (PageWriteback(page)) {
4081363f
JK		 2389 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
2390
b9109b0e
JK		 2391 f2fs_submit_merged_write_cond(sbi, page->mapping->host,
2392 0, page->index, type);
fec1d657
JK		 2393 if (ordered)
2394 wait_on_page_writeback(page);
2395 else
2396 wait_for_stable_page(page);
93dfe2ac
JK		 2397 }
2398}
2399
08b39fbd
CY		 2400void f2fs_wait_on_encrypted_page_writeback(struct f2fs_sb_info *sbi,
2401 block_t blkaddr)
2402{
2403 struct page *cpage;
2404
5d4c0af4 2405 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
08b39fbd
CY		 2406 return;
2407
08b39fbd
CY		 2408 cpage = find_lock_page(META_MAPPING(sbi), blkaddr);
2409 if (cpage) {
fec1d657 2410 f2fs_wait_on_page_writeback(cpage, DATA, true);
08b39fbd
CY		 2411 f2fs_put_page(cpage, 1);
2412 }
2413}
2414
351df4b2
JK		 2415static int read_compacted_summaries(struct f2fs_sb_info *sbi)
2416{
2417 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2418 struct curseg_info *seg_i;
2419 unsigned char *kaddr;
2420 struct page *page;
2421 block_t start;
2422 int i, j, offset;
2423
2424 start = start_sum_block(sbi);
2425
2426 page = get_meta_page(sbi, start++);
2427 kaddr = (unsigned char *)page_address(page);
2428
2429 /* Step 1: restore nat cache */
2430 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2431 memcpy(seg_i->journal, kaddr, SUM_JOURNAL_SIZE);
351df4b2
JK		 2432
2433 /* Step 2: restore sit cache */
2434 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2435 memcpy(seg_i->journal, kaddr + SUM_JOURNAL_SIZE, SUM_JOURNAL_SIZE);
351df4b2
JK		 2436 offset = 2 * SUM_JOURNAL_SIZE;
2437
2438 /* Step 3: restore summary entries */
2439 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2440 unsigned short blk_off;
2441 unsigned int segno;
2442
2443 seg_i = CURSEG_I(sbi, i);
2444 segno = le32_to_cpu(ckpt->cur_data_segno[i]);
2445 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[i]);
2446 seg_i->next_segno = segno;
2447 reset_curseg(sbi, i, 0);
2448 seg_i->alloc_type = ckpt->alloc_type[i];
2449 seg_i->next_blkoff = blk_off;
2450
2451 if (seg_i->alloc_type == SSR)
2452 blk_off = sbi->blocks_per_seg;
2453
2454 for (j = 0; j < blk_off; j++) {
2455 struct f2fs_summary *s;
2456 s = (struct f2fs_summary *)(kaddr + offset);
2457 seg_i->sum_blk->entries[j] = *s;
2458 offset += SUMMARY_SIZE;
09cbfeaf 2459 if (offset + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2460 SUM_FOOTER_SIZE)
2461 continue;
2462
2463 f2fs_put_page(page, 1);
2464 page = NULL;
2465
2466 page = get_meta_page(sbi, start++);
2467 kaddr = (unsigned char *)page_address(page);
2468 offset = 0;
2469 }
2470 }
2471 f2fs_put_page(page, 1);
2472 return 0;
2473}
2474
2475static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
2476{
2477 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2478 struct f2fs_summary_block *sum;
2479 struct curseg_info *curseg;
2480 struct page *new;
2481 unsigned short blk_off;
2482 unsigned int segno = 0;
2483 block_t blk_addr = 0;
2484
2485 /* get segment number and block addr */
2486 if (IS_DATASEG(type)) {
2487 segno = le32_to_cpu(ckpt->cur_data_segno[type]);
2488 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[type -
2489 CURSEG_HOT_DATA]);
119ee914 2490 if (__exist_node_summaries(sbi))
351df4b2
JK		 2491 blk_addr = sum_blk_addr(sbi, NR_CURSEG_TYPE, type);
2492 else
2493 blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
2494 } else {
2495 segno = le32_to_cpu(ckpt->cur_node_segno[type -
2496 CURSEG_HOT_NODE]);
2497 blk_off = le16_to_cpu(ckpt->cur_node_blkoff[type -
2498 CURSEG_HOT_NODE]);
119ee914 2499 if (__exist_node_summaries(sbi))
351df4b2
JK		 2500 blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
2501 type - CURSEG_HOT_NODE);
2502 else
2503 blk_addr = GET_SUM_BLOCK(sbi, segno);
2504 }
2505
2506 new = get_meta_page(sbi, blk_addr);
2507 sum = (struct f2fs_summary_block *)page_address(new);
2508
2509 if (IS_NODESEG(type)) {
119ee914 2510 if (__exist_node_summaries(sbi)) {
351df4b2
JK		 2511 struct f2fs_summary *ns = &sum->entries[0];
2512 int i;
2513 for (i = 0; i < sbi->blocks_per_seg; i++, ns++) {
2514 ns->version = 0;
2515 ns->ofs_in_node = 0;
2516 }
2517 } else {
d653788a
GZ		 2518 int err;
2519
2520 err = restore_node_summary(sbi, segno, sum);
2521 if (err) {
351df4b2 2522 f2fs_put_page(new, 1);
d653788a 2523 return err;
351df4b2
JK		 2524 }
2525 }
2526 }
2527
2528 /* set uncompleted segment to curseg */
2529 curseg = CURSEG_I(sbi, type);
2530 mutex_lock(&curseg->curseg_mutex);
b7ad7512
CY		 2531
2532 /* update journal info */
2533 down_write(&curseg->journal_rwsem);
2534 memcpy(curseg->journal, &sum->journal, SUM_JOURNAL_SIZE);
2535 up_write(&curseg->journal_rwsem);
2536
2537 memcpy(curseg->sum_blk->entries, sum->entries, SUM_ENTRY_SIZE);
2538 memcpy(&curseg->sum_blk->footer, &sum->footer, SUM_FOOTER_SIZE);
351df4b2
JK		 2539 curseg->next_segno = segno;
2540 reset_curseg(sbi, type, 0);
2541 curseg->alloc_type = ckpt->alloc_type[type];
2542 curseg->next_blkoff = blk_off;
2543 mutex_unlock(&curseg->curseg_mutex);
2544 f2fs_put_page(new, 1);
2545 return 0;
2546}
2547
2548static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
2549{
21d3f8e1
JQ		 2550 struct f2fs_journal *sit_j = CURSEG_I(sbi, CURSEG_COLD_DATA)->journal;
2551 struct f2fs_journal *nat_j = CURSEG_I(sbi, CURSEG_HOT_DATA)->journal;
351df4b2 2552 int type = CURSEG_HOT_DATA;
e4fc5fbf 2553 int err;
351df4b2 2554
aaec2b1d 2555 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG)) {
3fa06d7b
CY		 2556 int npages = npages_for_summary_flush(sbi, true);
2557
2558 if (npages >= 2)
2559 ra_meta_pages(sbi, start_sum_block(sbi), npages,
26879fb1 2560 META_CP, true);
3fa06d7b 2561
351df4b2
JK		 2562 /* restore for compacted data summary */
2563 if (read_compacted_summaries(sbi))
2564 return -EINVAL;
2565 type = CURSEG_HOT_NODE;
2566 }
2567
119ee914 2568 if (__exist_node_summaries(sbi))
3fa06d7b 2569 ra_meta_pages(sbi, sum_blk_addr(sbi, NR_CURSEG_TYPE, type),
26879fb1 2570 NR_CURSEG_TYPE - type, META_CP, true);
3fa06d7b 2571
e4fc5fbf
CY		 2572 for (; type <= CURSEG_COLD_NODE; type++) {
2573 err = read_normal_summaries(sbi, type);
2574 if (err)
2575 return err;
2576 }
2577
21d3f8e1
JQ		 2578 /* sanity check for summary blocks */
2579 if (nats_in_cursum(nat_j) > NAT_JOURNAL_ENTRIES ||
2580 sits_in_cursum(sit_j) > SIT_JOURNAL_ENTRIES)
2581 return -EINVAL;
2582
351df4b2
JK		 2583 return 0;
2584}
2585
2586static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
2587{
2588 struct page *page;
2589 unsigned char *kaddr;
2590 struct f2fs_summary *summary;
2591 struct curseg_info *seg_i;
2592 int written_size = 0;
2593 int i, j;
2594
2595 page = grab_meta_page(sbi, blkaddr++);
2596 kaddr = (unsigned char *)page_address(page);
2597
2598 /* Step 1: write nat cache */
2599 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2600 memcpy(kaddr, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2601 written_size += SUM_JOURNAL_SIZE;
2602
2603 /* Step 2: write sit cache */
2604 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2605 memcpy(kaddr + written_size, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2606 written_size += SUM_JOURNAL_SIZE;
2607
351df4b2
JK		 2608 /* Step 3: write summary entries */
2609 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2610 unsigned short blkoff;
2611 seg_i = CURSEG_I(sbi, i);
2612 if (sbi->ckpt->alloc_type[i] == SSR)
2613 blkoff = sbi->blocks_per_seg;
2614 else
2615 blkoff = curseg_blkoff(sbi, i);
2616
2617 for (j = 0; j < blkoff; j++) {
2618 if (!page) {
2619 page = grab_meta_page(sbi, blkaddr++);
2620 kaddr = (unsigned char *)page_address(page);
2621 written_size = 0;
2622 }
2623 summary = (struct f2fs_summary *)(kaddr + written_size);
2624 *summary = seg_i->sum_blk->entries[j];
2625 written_size += SUMMARY_SIZE;
351df4b2 2626
09cbfeaf 2627 if (written_size + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2628 SUM_FOOTER_SIZE)
2629 continue;
2630
e8d61a74 2631 set_page_dirty(page);
351df4b2
JK		 2632 f2fs_put_page(page, 1);
2633 page = NULL;
2634 }
2635 }
e8d61a74
CY		 2636 if (page) {
2637 set_page_dirty(page);
351df4b2 2638 f2fs_put_page(page, 1);
e8d61a74 2639 }
351df4b2
JK		 2640}
2641
2642static void write_normal_summaries(struct f2fs_sb_info *sbi,
2643 block_t blkaddr, int type)
2644{
2645 int i, end;
2646 if (IS_DATASEG(type))
2647 end = type + NR_CURSEG_DATA_TYPE;
2648 else
2649 end = type + NR_CURSEG_NODE_TYPE;
2650
b7ad7512
CY		 2651 for (i = type; i < end; i++)
2652 write_current_sum_page(sbi, i, blkaddr + (i - type));
351df4b2
JK		 2653}
2654
2655void write_data_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2656{
aaec2b1d 2657 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG))
351df4b2
JK		 2658 write_compacted_summaries(sbi, start_blk);
2659 else
2660 write_normal_summaries(sbi, start_blk, CURSEG_HOT_DATA);
2661}
2662
2663void write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2664{
119ee914 2665 write_normal_summaries(sbi, start_blk, CURSEG_HOT_NODE);
351df4b2
JK		 2666}
2667
dfc08a12 2668int lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
351df4b2
JK		 2669 unsigned int val, int alloc)
2670{
2671 int i;
2672
2673 if (type == NAT_JOURNAL) {
dfc08a12
CY		 2674 for (i = 0; i < nats_in_cursum(journal); i++) {
2675 if (le32_to_cpu(nid_in_journal(journal, i)) == val)
351df4b2
JK		 2676 return i;
2677 }
dfc08a12
CY		 2678 if (alloc && __has_cursum_space(journal, 1, NAT_JOURNAL))
2679 return update_nats_in_cursum(journal, 1);
351df4b2 2680 } else if (type == SIT_JOURNAL) {
dfc08a12
CY		 2681 for (i = 0; i < sits_in_cursum(journal); i++)
2682 if (le32_to_cpu(segno_in_journal(journal, i)) == val)
351df4b2 2683 return i;
dfc08a12
CY		 2684 if (alloc && __has_cursum_space(journal, 1, SIT_JOURNAL))
2685 return update_sits_in_cursum(journal, 1);
351df4b2
JK		 2686 }
2687 return -1;
2688}
2689
2690static struct page *get_current_sit_page(struct f2fs_sb_info *sbi,
2691 unsigned int segno)
2692{
2cc22186 2693 return get_meta_page(sbi, current_sit_addr(sbi, segno));
351df4b2
JK		 2694}
2695
2696static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
2697 unsigned int start)
2698{
2699 struct sit_info *sit_i = SIT_I(sbi);
2700 struct page *src_page, *dst_page;
2701 pgoff_t src_off, dst_off;
2702 void *src_addr, *dst_addr;
2703
2704 src_off = current_sit_addr(sbi, start);
2705 dst_off = next_sit_addr(sbi, src_off);
2706
2707 /* get current sit block page without lock */
2708 src_page = get_meta_page(sbi, src_off);
2709 dst_page = grab_meta_page(sbi, dst_off);
9850cf4a 2710 f2fs_bug_on(sbi, PageDirty(src_page));
351df4b2
JK		 2711
2712 src_addr = page_address(src_page);
2713 dst_addr = page_address(dst_page);
09cbfeaf 2714 memcpy(dst_addr, src_addr, PAGE_SIZE);
351df4b2
JK		 2715
2716 set_page_dirty(dst_page);
2717 f2fs_put_page(src_page, 1);
2718
2719 set_to_next_sit(sit_i, start);
2720
2721 return dst_page;
2722}
2723
184a5cd2
CY		 2724static struct sit_entry_set *grab_sit_entry_set(void)
2725{
2726 struct sit_entry_set *ses =
80c54505 2727 f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS);
184a5cd2
CY		 2728
2729 ses->entry_cnt = 0;
2730 INIT_LIST_HEAD(&ses->set_list);
2731 return ses;
2732}
2733
2734static void release_sit_entry_set(struct sit_entry_set *ses)
2735{
2736 list_del(&ses->set_list);
2737 kmem_cache_free(sit_entry_set_slab, ses);
2738}
2739
2740static void adjust_sit_entry_set(struct sit_entry_set *ses,
2741 struct list_head *head)
2742{
2743 struct sit_entry_set *next = ses;
2744
2745 if (list_is_last(&ses->set_list, head))
2746 return;
2747
2748 list_for_each_entry_continue(next, head, set_list)
2749 if (ses->entry_cnt <= next->entry_cnt)
2750 break;
2751
2752 list_move_tail(&ses->set_list, &next->set_list);
2753}
2754
2755static void add_sit_entry(unsigned int segno, struct list_head *head)
2756{
2757 struct sit_entry_set *ses;
2758 unsigned int start_segno = START_SEGNO(segno);
2759
2760 list_for_each_entry(ses, head, set_list) {
2761 if (ses->start_segno == start_segno) {
2762 ses->entry_cnt++;
2763 adjust_sit_entry_set(ses, head);
2764 return;
2765 }
2766 }
2767
2768 ses = grab_sit_entry_set();
2769
2770 ses->start_segno = start_segno;
2771 ses->entry_cnt++;
2772 list_add(&ses->set_list, head);
2773}
2774
2775static void add_sits_in_set(struct f2fs_sb_info *sbi)
2776{
2777 struct f2fs_sm_info *sm_info = SM_I(sbi);
2778 struct list_head *set_list = &sm_info->sit_entry_set;
2779 unsigned long *bitmap = SIT_I(sbi)->dirty_sentries_bitmap;
184a5cd2
CY		 2780 unsigned int segno;
2781
7cd8558b 2782 for_each_set_bit(segno, bitmap, MAIN_SEGS(sbi))
184a5cd2
CY		 2783 add_sit_entry(segno, set_list);
2784}
2785
2786static void remove_sits_in_journal(struct f2fs_sb_info *sbi)
351df4b2
JK		 2787{
2788 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2789 struct f2fs_journal *journal = curseg->journal;
351df4b2
JK		 2790 int i;
2791
b7ad7512 2792 down_write(&curseg->journal_rwsem);
dfc08a12 2793 for (i = 0; i < sits_in_cursum(journal); i++) {
184a5cd2
CY		 2794 unsigned int segno;
2795 bool dirtied;
2796
dfc08a12 2797 segno = le32_to_cpu(segno_in_journal(journal, i));
184a5cd2
CY		 2798 dirtied = __mark_sit_entry_dirty(sbi, segno);
2799
2800 if (!dirtied)
2801 add_sit_entry(segno, &SM_I(sbi)->sit_entry_set);
351df4b2 2802 }
dfc08a12 2803 update_sits_in_cursum(journal, -i);
b7ad7512 2804 up_write(&curseg->journal_rwsem);
351df4b2
JK		 2805}
2806
0a8165d7 2807/*
351df4b2
JK		 2808 * CP calls this function, which flushes SIT entries including sit_journal,
2809 * and moves prefree segs to free segs.
2810 */
4b2fecc8 2811void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2
JK		 2812{
2813 struct sit_info *sit_i = SIT_I(sbi);
2814 unsigned long *bitmap = sit_i->dirty_sentries_bitmap;
2815 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2816 struct f2fs_journal *journal = curseg->journal;
184a5cd2
CY		 2817 struct sit_entry_set *ses, *tmp;
2818 struct list_head *head = &SM_I(sbi)->sit_entry_set;
184a5cd2 2819 bool to_journal = true;
4b2fecc8 2820 struct seg_entry *se;
351df4b2 2821
351df4b2
JK		 2822 mutex_lock(&sit_i->sentry_lock);
2823
2b11a74b
WL		 2824 if (!sit_i->dirty_sentries)
2825 goto out;
2826
351df4b2 2827 /*
184a5cd2
CY		 2828 * add and account sit entries of dirty bitmap in sit entry
2829 * set temporarily
351df4b2 2830 */
184a5cd2 2831 add_sits_in_set(sbi);
351df4b2 2832
184a5cd2
CY		 2833 /*
2834 * if there are no enough space in journal to store dirty sit
2835 * entries, remove all entries from journal and add and account
2836 * them in sit entry set.
2837 */
dfc08a12 2838 if (!__has_cursum_space(journal, sit_i->dirty_sentries, SIT_JOURNAL))
184a5cd2 2839 remove_sits_in_journal(sbi);
b2955550 2840
184a5cd2
CY		 2841 /*
2842 * there are two steps to flush sit entries:
2843 * #1, flush sit entries to journal in current cold data summary block.
2844 * #2, flush sit entries to sit page.
2845 */
2846 list_for_each_entry_safe(ses, tmp, head, set_list) {
4a257ed6 2847 struct page *page = NULL;
184a5cd2
CY		 2848 struct f2fs_sit_block *raw_sit = NULL;
2849 unsigned int start_segno = ses->start_segno;
2850 unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
7cd8558b 2851 (unsigned long)MAIN_SEGS(sbi));
184a5cd2
CY		 2852 unsigned int segno = start_segno;
2853
2854 if (to_journal &&
dfc08a12 2855 !__has_cursum_space(journal, ses->entry_cnt, SIT_JOURNAL))
184a5cd2
CY		 2856 to_journal = false;
2857
b7ad7512
CY		 2858 if (to_journal) {
2859 down_write(&curseg->journal_rwsem);
2860 } else {
184a5cd2
CY		 2861 page = get_next_sit_page(sbi, start_segno);
2862 raw_sit = page_address(page);
351df4b2 2863 }
351df4b2 2864
184a5cd2
CY		 2865 /* flush dirty sit entries in region of current sit set */
2866 for_each_set_bit_from(segno, bitmap, end) {
2867 int offset, sit_offset;
4b2fecc8
JK		 2868
2869 se = get_seg_entry(sbi, segno);
184a5cd2
CY		 2870
2871 /* add discard candidates */
c473f1a9 2872 if (!(cpc->reason & CP_DISCARD)) {
4b2fecc8 2873 cpc->trim_start = segno;
25290fa5 2874 add_discard_addrs(sbi, cpc, false);
4b2fecc8 2875 }
184a5cd2
CY		 2876
2877 if (to_journal) {
dfc08a12 2878 offset = lookup_journal_in_cursum(journal,
184a5cd2
CY		 2879 SIT_JOURNAL, segno, 1);
2880 f2fs_bug_on(sbi, offset < 0);
dfc08a12 2881 segno_in_journal(journal, offset) =
184a5cd2
CY		 2882 cpu_to_le32(segno);
2883 seg_info_to_raw_sit(se,
dfc08a12 2884 &sit_in_journal(journal, offset));
184a5cd2
CY		 2885 } else {
2886 sit_offset = SIT_ENTRY_OFFSET(sit_i, segno);
2887 seg_info_to_raw_sit(se,
2888 &raw_sit->entries[sit_offset]);
2889 }
351df4b2 2890
184a5cd2
CY		 2891 __clear_bit(segno, bitmap);
2892 sit_i->dirty_sentries--;
2893 ses->entry_cnt--;
351df4b2
JK		 2894 }
2895
b7ad7512
CY		 2896 if (to_journal)
2897 up_write(&curseg->journal_rwsem);
2898 else
184a5cd2
CY		 2899 f2fs_put_page(page, 1);
2900
2901 f2fs_bug_on(sbi, ses->entry_cnt);
2902 release_sit_entry_set(ses);
351df4b2 2903 }
184a5cd2
CY		 2904
2905 f2fs_bug_on(sbi, !list_empty(head));
2906 f2fs_bug_on(sbi, sit_i->dirty_sentries);
184a5cd2 2907out:
c473f1a9 2908 if (cpc->reason & CP_DISCARD) {
650d3c4e
YH		 2909 __u64 trim_start = cpc->trim_start;
2910
4b2fecc8 2911 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++)
25290fa5 2912 add_discard_addrs(sbi, cpc, false);
650d3c4e
YH		 2913
2914 cpc->trim_start = trim_start;
4b2fecc8 2915 }
351df4b2 2916 mutex_unlock(&sit_i->sentry_lock);
351df4b2 2917
351df4b2
JK		 2918 set_prefree_as_free_segments(sbi);
2919}
2920
2921static int build_sit_info(struct f2fs_sb_info *sbi)
2922{
2923 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
351df4b2
JK		 2924 struct sit_info *sit_i;
2925 unsigned int sit_segs, start;
ae27d62e 2926 char *src_bitmap;
351df4b2
JK		 2927 unsigned int bitmap_size;
2928
2929 /* allocate memory for SIT information */
2930 sit_i = kzalloc(sizeof(struct sit_info), GFP_KERNEL);
2931 if (!sit_i)
2932 return -ENOMEM;
2933
2934 SM_I(sbi)->sit_info = sit_i;
2935
a7c3e901 2936 sit_i->sentries = kvzalloc(MAIN_SEGS(sbi) *
39307a8e 2937 sizeof(struct seg_entry), GFP_KERNEL);
351df4b2
JK		 2938 if (!sit_i->sentries)
2939 return -ENOMEM;
2940
7cd8558b 2941 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
a7c3e901 2942 sit_i->dirty_sentries_bitmap = kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2943 if (!sit_i->dirty_sentries_bitmap)
2944 return -ENOMEM;
2945
7cd8558b 2946 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 2947 sit_i->sentries[start].cur_valid_map
2948 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2949 sit_i->sentries[start].ckpt_valid_map
2950 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
a66cdd98 2951 if (!sit_i->sentries[start].cur_valid_map ||
3e025740 2952 !sit_i->sentries[start].ckpt_valid_map)
351df4b2 2953 return -ENOMEM;
3e025740 2954
355e7891
CY		 2955#ifdef CONFIG_F2FS_CHECK_FS
2956 sit_i->sentries[start].cur_valid_map_mir
2957 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2958 if (!sit_i->sentries[start].cur_valid_map_mir)
2959 return -ENOMEM;
2960#endif
2961
3e025740
JK		 2962 if (f2fs_discard_en(sbi)) {
2963 sit_i->sentries[start].discard_map
2964 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2965 if (!sit_i->sentries[start].discard_map)
2966 return -ENOMEM;
2967 }
351df4b2
JK		 2968 }
2969
60a3b782
JK		 2970 sit_i->tmp_map = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2971 if (!sit_i->tmp_map)
2972 return -ENOMEM;
2973
351df4b2 2974 if (sbi->segs_per_sec > 1) {
a7c3e901 2975 sit_i->sec_entries = kvzalloc(MAIN_SECS(sbi) *
39307a8e 2976 sizeof(struct sec_entry), GFP_KERNEL);
351df4b2
JK		 2977 if (!sit_i->sec_entries)
2978 return -ENOMEM;
2979 }
2980
2981 /* get information related with SIT */
2982 sit_segs = le32_to_cpu(raw_super->segment_count_sit) >> 1;
2983
2984 /* setup SIT bitmap from ckeckpoint pack */
2985 bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
2986 src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
2987
ae27d62e
CY		 2988 sit_i->sit_bitmap = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2989 if (!sit_i->sit_bitmap)
351df4b2 2990 return -ENOMEM;
351df4b2 2991
ae27d62e
CY		 2992#ifdef CONFIG_F2FS_CHECK_FS
2993 sit_i->sit_bitmap_mir = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2994 if (!sit_i->sit_bitmap_mir)
2995 return -ENOMEM;
2996#endif
2997
351df4b2
JK		 2998 /* init SIT information */
2999 sit_i->s_ops = &default_salloc_ops;
3000
3001 sit_i->sit_base_addr = le32_to_cpu(raw_super->sit_blkaddr);
3002 sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
c79b7ff1 3003 sit_i->written_valid_blocks = 0;
351df4b2
JK		 3004 sit_i->bitmap_size = bitmap_size;
3005 sit_i->dirty_sentries = 0;
3006 sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
3007 sit_i->elapsed_time = le64_to_cpu(sbi->ckpt->elapsed_time);
48fbfe50 3008 sit_i->mounted_time = ktime_get_real_seconds();
351df4b2
JK		 3009 mutex_init(&sit_i->sentry_lock);
3010 return 0;
3011}
3012
3013static int build_free_segmap(struct f2fs_sb_info *sbi)
3014{
351df4b2
JK		 3015 struct free_segmap_info *free_i;
3016 unsigned int bitmap_size, sec_bitmap_size;
3017
3018 /* allocate memory for free segmap information */
3019 free_i = kzalloc(sizeof(struct free_segmap_info), GFP_KERNEL);
3020 if (!free_i)
3021 return -ENOMEM;
3022
3023 SM_I(sbi)->free_info = free_i;
3024
7cd8558b 3025 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
a7c3e901 3026 free_i->free_segmap = kvmalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 3027 if (!free_i->free_segmap)
3028 return -ENOMEM;
3029
7cd8558b 3030 sec_bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
a7c3e901 3031 free_i->free_secmap = kvmalloc(sec_bitmap_size, GFP_KERNEL);
351df4b2
JK		 3032 if (!free_i->free_secmap)
3033 return -ENOMEM;
3034
3035 /* set all segments as dirty temporarily */
3036 memset(free_i->free_segmap, 0xff, bitmap_size);
3037 memset(free_i->free_secmap, 0xff, sec_bitmap_size);
3038
3039 /* init free segmap information */
7cd8558b 3040 free_i->start_segno = GET_SEGNO_FROM_SEG0(sbi, MAIN_BLKADDR(sbi));
351df4b2
JK		 3041 free_i->free_segments = 0;
3042 free_i->free_sections = 0;
1a118ccf 3043 spin_lock_init(&free_i->segmap_lock);
351df4b2
JK		 3044 return 0;
3045}
3046
3047static int build_curseg(struct f2fs_sb_info *sbi)
3048{
1042d60f 3049 struct curseg_info *array;
351df4b2
JK		 3050 int i;
3051
b434babf 3052 array = kcalloc(NR_CURSEG_TYPE, sizeof(*array), GFP_KERNEL);
351df4b2
JK		 3053 if (!array)
3054 return -ENOMEM;
3055
3056 SM_I(sbi)->curseg_array = array;
3057
3058 for (i = 0; i < NR_CURSEG_TYPE; i++) {
3059 mutex_init(&array[i].curseg_mutex);
09cbfeaf 3060 array[i].sum_blk = kzalloc(PAGE_SIZE, GFP_KERNEL);
351df4b2
JK		 3061 if (!array[i].sum_blk)
3062 return -ENOMEM;
b7ad7512
CY		 3063 init_rwsem(&array[i].journal_rwsem);
3064 array[i].journal = kzalloc(sizeof(struct f2fs_journal),
3065 GFP_KERNEL);
3066 if (!array[i].journal)
3067 return -ENOMEM;
351df4b2
JK		 3068 array[i].segno = NULL_SEGNO;
3069 array[i].next_blkoff = 0;
3070 }
3071 return restore_curseg_summaries(sbi);
3072}
3073
3074static void build_sit_entries(struct f2fs_sb_info *sbi)
3075{
3076 struct sit_info *sit_i = SIT_I(sbi);
3077 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 3078 struct f2fs_journal *journal = curseg->journal;
9c094040
YH		 3079 struct seg_entry *se;
3080 struct f2fs_sit_entry sit;
74de593a
CY		 3081 int sit_blk_cnt = SIT_BLK_CNT(sbi);
3082 unsigned int i, start, end;
3083 unsigned int readed, start_blk = 0;
351df4b2 3084
74de593a 3085 do {
664ba972
JK		 3086 readed = ra_meta_pages(sbi, start_blk, BIO_MAX_PAGES,
3087 META_SIT, true);
74de593a
CY		 3088
3089 start = start_blk * sit_i->sents_per_block;
3090 end = (start_blk + readed) * sit_i->sents_per_block;
3091
7cd8558b 3092 for (; start < end && start < MAIN_SEGS(sbi); start++) {
74de593a 3093 struct f2fs_sit_block *sit_blk;
74de593a
CY		 3094 struct page *page;
3095
9c094040 3096 se = &sit_i->sentries[start];
74de593a
CY		 3097 page = get_current_sit_page(sbi, start);
3098 sit_blk = (struct f2fs_sit_block *)page_address(page);
3099 sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
3100 f2fs_put_page(page, 1);
d600af23 3101
74de593a
CY		 3102 check_block_count(sbi, start, &sit);
3103 seg_info_from_raw_sit(se, &sit);
a66cdd98
JK		 3104
3105 /* build discard map only one time */
3e025740 3106 if (f2fs_discard_en(sbi)) {
1f43e2ad
CY		 3107 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
3108 memset(se->discard_map, 0xff,
3109 SIT_VBLOCK_MAP_SIZE);
3110 } else {
3111 memcpy(se->discard_map,
3112 se->cur_valid_map,
3113 SIT_VBLOCK_MAP_SIZE);
3114 sbi->discard_blks +=
3115 sbi->blocks_per_seg -
3116 se->valid_blocks;
3117 }
3e025740 3118 }
a66cdd98 3119
d600af23
CY		 3120 if (sbi->segs_per_sec > 1)
3121 get_sec_entry(sbi, start)->valid_blocks +=
3122 se->valid_blocks;
351df4b2 3123 }
74de593a
CY		 3124 start_blk += readed;
3125 } while (start_blk < sit_blk_cnt);
d600af23
CY		 3126
3127 down_read(&curseg->journal_rwsem);
3128 for (i = 0; i < sits_in_cursum(journal); i++) {
d600af23
CY		 3129 unsigned int old_valid_blocks;
3130
3131 start = le32_to_cpu(segno_in_journal(journal, i));
3132 se = &sit_i->sentries[start];
3133 sit = sit_in_journal(journal, i);
3134
3135 old_valid_blocks = se->valid_blocks;
3136
3137 check_block_count(sbi, start, &sit);
3138 seg_info_from_raw_sit(se, &sit);
3139
3140 if (f2fs_discard_en(sbi)) {
1f43e2ad
CY		 3141 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
3142 memset(se->discard_map, 0xff,
3143 SIT_VBLOCK_MAP_SIZE);
3144 } else {
3145 memcpy(se->discard_map, se->cur_valid_map,
3146 SIT_VBLOCK_MAP_SIZE);
3147 sbi->discard_blks += old_valid_blocks -
3148 se->valid_blocks;
3149 }
d600af23
CY		 3150 }
3151
3152 if (sbi->segs_per_sec > 1)
3153 get_sec_entry(sbi, start)->valid_blocks +=
3154 se->valid_blocks - old_valid_blocks;
3155 }
3156 up_read(&curseg->journal_rwsem);
351df4b2
JK		 3157}
3158
3159static void init_free_segmap(struct f2fs_sb_info *sbi)
3160{
3161 unsigned int start;
3162 int type;
3163
7cd8558b 3164 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 3165 struct seg_entry *sentry = get_seg_entry(sbi, start);
3166 if (!sentry->valid_blocks)
3167 __set_free(sbi, start);
c79b7ff1
JK		 3168 else
3169 SIT_I(sbi)->written_valid_blocks +=
3170 sentry->valid_blocks;
351df4b2
JK		 3171 }
3172
3173 /* set use the current segments */
3174 for (type = CURSEG_HOT_DATA; type <= CURSEG_COLD_NODE; type++) {
3175 struct curseg_info *curseg_t = CURSEG_I(sbi, type);
3176 __set_test_and_inuse(sbi, curseg_t->segno);
3177 }
3178}
3179
3180static void init_dirty_segmap(struct f2fs_sb_info *sbi)
3181{
3182 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3183 struct free_segmap_info *free_i = FREE_I(sbi);
7cd8558b 3184 unsigned int segno = 0, offset = 0;
351df4b2
JK		 3185 unsigned short valid_blocks;
3186
8736fbf0 3187 while (1) {
351df4b2 3188 /* find dirty segment based on free segmap */
7cd8558b
JK		 3189 segno = find_next_inuse(free_i, MAIN_SEGS(sbi), offset);
3190 if (segno >= MAIN_SEGS(sbi))
351df4b2
JK		 3191 break;
3192 offset = segno + 1;
302bd348 3193 valid_blocks = get_valid_blocks(sbi, segno, false);
ec325b52 3194 if (valid_blocks == sbi->blocks_per_seg || !valid_blocks)
351df4b2 3195 continue;
ec325b52
JK		 3196 if (valid_blocks > sbi->blocks_per_seg) {
3197 f2fs_bug_on(sbi, 1);
3198 continue;
3199 }
351df4b2
JK		 3200 mutex_lock(&dirty_i->seglist_lock);
3201 __locate_dirty_segment(sbi, segno, DIRTY);
3202 mutex_unlock(&dirty_i->seglist_lock);
3203 }
3204}
3205
5ec4e49f 3206static int init_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3207{
3208 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
7cd8558b 3209 unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
351df4b2 3210
a7c3e901 3211 dirty_i->victim_secmap = kvzalloc(bitmap_size, GFP_KERNEL);
5ec4e49f 3212 if (!dirty_i->victim_secmap)
351df4b2
JK		 3213 return -ENOMEM;
3214 return 0;
3215}
3216
3217static int build_dirty_segmap(struct f2fs_sb_info *sbi)
3218{
3219 struct dirty_seglist_info *dirty_i;
3220 unsigned int bitmap_size, i;
3221
3222 /* allocate memory for dirty segments list information */
3223 dirty_i = kzalloc(sizeof(struct dirty_seglist_info), GFP_KERNEL);
3224 if (!dirty_i)
3225 return -ENOMEM;
3226
3227 SM_I(sbi)->dirty_info = dirty_i;
3228 mutex_init(&dirty_i->seglist_lock);
3229
7cd8558b 3230 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
351df4b2
JK		 3231
3232 for (i = 0; i < NR_DIRTY_TYPE; i++) {
a7c3e901 3233 dirty_i->dirty_segmap[i] = kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 3234 if (!dirty_i->dirty_segmap[i])
3235 return -ENOMEM;
3236 }
3237
3238 init_dirty_segmap(sbi);
5ec4e49f 3239 return init_victim_secmap(sbi);
351df4b2
JK		 3240}
3241
0a8165d7 3242/*
351df4b2
JK		 3243 * Update min, max modified time for cost-benefit GC algorithm
3244 */
3245static void init_min_max_mtime(struct f2fs_sb_info *sbi)
3246{
3247 struct sit_info *sit_i = SIT_I(sbi);
3248 unsigned int segno;
3249
3250 mutex_lock(&sit_i->sentry_lock);
3251
3252 sit_i->min_mtime = LLONG_MAX;
3253
7cd8558b 3254 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
351df4b2
JK		 3255 unsigned int i;
3256 unsigned long long mtime = 0;
3257
3258 for (i = 0; i < sbi->segs_per_sec; i++)
3259 mtime += get_seg_entry(sbi, segno + i)->mtime;
3260
3261 mtime = div_u64(mtime, sbi->segs_per_sec);
3262
3263 if (sit_i->min_mtime > mtime)
3264 sit_i->min_mtime = mtime;
3265 }
3266 sit_i->max_mtime = get_mtime(sbi);
3267 mutex_unlock(&sit_i->sentry_lock);
3268}
3269
3270int build_segment_manager(struct f2fs_sb_info *sbi)
3271{
3272 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3273 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1042d60f 3274 struct f2fs_sm_info *sm_info;
351df4b2
JK		 3275 int err;
3276
3277 sm_info = kzalloc(sizeof(struct f2fs_sm_info), GFP_KERNEL);
3278 if (!sm_info)
3279 return -ENOMEM;
3280
3281 /* init sm info */
3282 sbi->sm_info = sm_info;
351df4b2
JK		 3283 sm_info->seg0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3284 sm_info->main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3285 sm_info->segment_count = le32_to_cpu(raw_super->segment_count);
3286 sm_info->reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3287 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3288 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
3289 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
58c41035
JK		 3290 sm_info->rec_prefree_segments = sm_info->main_segments *
3291 DEF_RECLAIM_PREFREE_SEGMENTS / 100;
44a83499
JK		 3292 if (sm_info->rec_prefree_segments > DEF_MAX_RECLAIM_PREFREE_SEGMENTS)
3293 sm_info->rec_prefree_segments = DEF_MAX_RECLAIM_PREFREE_SEGMENTS;
3294
52763a4b
JK		 3295 if (!test_opt(sbi, LFS))
3296 sm_info->ipu_policy = 1 << F2FS_IPU_FSYNC;
216fbd64 3297 sm_info->min_ipu_util = DEF_MIN_IPU_UTIL;
c1ce1b02 3298 sm_info->min_fsync_blocks = DEF_MIN_FSYNC_BLOCKS;
ef095d19 3299 sm_info->min_hot_blocks = DEF_MIN_HOT_BLOCKS;
351df4b2 3300
bba681cb
JK		 3301 sm_info->trim_sections = DEF_BATCHED_TRIM_SECTIONS;
3302
184a5cd2
CY		 3303 INIT_LIST_HEAD(&sm_info->sit_entry_set);
3304
d4fdf8ba 3305 if (!f2fs_readonly(sbi->sb)) {
2163d198
GZ		 3306 err = create_flush_cmd_control(sbi);
3307 if (err)
a688b9d9 3308 return err;
6b4afdd7
JK		 3309 }
3310
0b54fb84
JK		 3311 err = create_discard_cmd_control(sbi);
3312 if (err)
3313 return err;
3314
351df4b2
JK		 3315 err = build_sit_info(sbi);
3316 if (err)
3317 return err;
3318 err = build_free_segmap(sbi);
3319 if (err)
3320 return err;
3321 err = build_curseg(sbi);
3322 if (err)
3323 return err;
3324
3325 /* reinit free segmap based on SIT */
3326 build_sit_entries(sbi);
3327
3328 init_free_segmap(sbi);
3329 err = build_dirty_segmap(sbi);
3330 if (err)
3331 return err;
3332
3333 init_min_max_mtime(sbi);
3334 return 0;
3335}
3336
3337static void discard_dirty_segmap(struct f2fs_sb_info *sbi,
3338 enum dirty_type dirty_type)
3339{
3340 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3341
3342 mutex_lock(&dirty_i->seglist_lock);
39307a8e 3343 kvfree(dirty_i->dirty_segmap[dirty_type]);
351df4b2
JK		 3344 dirty_i->nr_dirty[dirty_type] = 0;
3345 mutex_unlock(&dirty_i->seglist_lock);
3346}
3347
5ec4e49f 3348static void destroy_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3349{
3350 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
39307a8e 3351 kvfree(dirty_i->victim_secmap);
351df4b2
JK		 3352}
3353
3354static void destroy_dirty_segmap(struct f2fs_sb_info *sbi)
3355{
3356 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3357 int i;
3358
3359 if (!dirty_i)
3360 return;
3361
3362 /* discard pre-free/dirty segments list */
3363 for (i = 0; i < NR_DIRTY_TYPE; i++)
3364 discard_dirty_segmap(sbi, i);
3365
5ec4e49f 3366 destroy_victim_secmap(sbi);
351df4b2
JK		 3367 SM_I(sbi)->dirty_info = NULL;
3368 kfree(dirty_i);
3369}
3370
3371static void destroy_curseg(struct f2fs_sb_info *sbi)
3372{
3373 struct curseg_info *array = SM_I(sbi)->curseg_array;
3374 int i;
3375
3376 if (!array)
3377 return;
3378 SM_I(sbi)->curseg_array = NULL;
b7ad7512 3379 for (i = 0; i < NR_CURSEG_TYPE; i++) {
351df4b2 3380 kfree(array[i].sum_blk);
b7ad7512
CY		 3381 kfree(array[i].journal);
3382 }
351df4b2
JK		 3383 kfree(array);
3384}
3385
3386static void destroy_free_segmap(struct f2fs_sb_info *sbi)
3387{
3388 struct free_segmap_info *free_i = SM_I(sbi)->free_info;
3389 if (!free_i)
3390 return;
3391 SM_I(sbi)->free_info = NULL;
39307a8e
JK		 3392 kvfree(free_i->free_segmap);
3393 kvfree(free_i->free_secmap);
351df4b2
JK		 3394 kfree(free_i);
3395}
3396
3397static void destroy_sit_info(struct f2fs_sb_info *sbi)
3398{
3399 struct sit_info *sit_i = SIT_I(sbi);
3400 unsigned int start;
3401
3402 if (!sit_i)
3403 return;
3404
3405 if (sit_i->sentries) {
7cd8558b 3406 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2 3407 kfree(sit_i->sentries[start].cur_valid_map);
355e7891
CY		 3408#ifdef CONFIG_F2FS_CHECK_FS
3409 kfree(sit_i->sentries[start].cur_valid_map_mir);
3410#endif
351df4b2 3411 kfree(sit_i->sentries[start].ckpt_valid_map);
a66cdd98 3412 kfree(sit_i->sentries[start].discard_map);
351df4b2
JK		 3413 }
3414 }
60a3b782
JK		 3415 kfree(sit_i->tmp_map);
3416
39307a8e
JK		 3417 kvfree(sit_i->sentries);
3418 kvfree(sit_i->sec_entries);
3419 kvfree(sit_i->dirty_sentries_bitmap);
351df4b2
JK		 3420
3421 SM_I(sbi)->sit_info = NULL;
3422 kfree(sit_i->sit_bitmap);
ae27d62e
CY		 3423#ifdef CONFIG_F2FS_CHECK_FS
3424 kfree(sit_i->sit_bitmap_mir);
3425#endif
351df4b2
JK		 3426 kfree(sit_i);
3427}
3428
3429void destroy_segment_manager(struct f2fs_sb_info *sbi)
3430{
3431 struct f2fs_sm_info *sm_info = SM_I(sbi);
a688b9d9 3432
3b03f724
CY		 3433 if (!sm_info)
3434 return;
5eba8c5d 3435 destroy_flush_cmd_control(sbi, true);
f099405f 3436 destroy_discard_cmd_control(sbi);
351df4b2
JK		 3437 destroy_dirty_segmap(sbi);
3438 destroy_curseg(sbi);
3439 destroy_free_segmap(sbi);
3440 destroy_sit_info(sbi);
3441 sbi->sm_info = NULL;
3442 kfree(sm_info);
3443}
7fd9e544
JK		 3444
3445int __init create_segment_manager_caches(void)
3446{
3447 discard_entry_slab = f2fs_kmem_cache_create("discard_entry",
e8512d2e 3448 sizeof(struct discard_entry));
7fd9e544 3449 if (!discard_entry_slab)
184a5cd2
CY		 3450 goto fail;
3451
b01a9201
JK		 3452 discard_cmd_slab = f2fs_kmem_cache_create("discard_cmd",
3453 sizeof(struct discard_cmd));
3454 if (!discard_cmd_slab)
6ab2a308 3455 goto destroy_discard_entry;
275b66b0 3456
184a5cd2 3457 sit_entry_set_slab = f2fs_kmem_cache_create("sit_entry_set",
c9ee0085 3458 sizeof(struct sit_entry_set));
184a5cd2 3459 if (!sit_entry_set_slab)
b01a9201 3460 goto destroy_discard_cmd;
88b88a66
JK		 3461
3462 inmem_entry_slab = f2fs_kmem_cache_create("inmem_page_entry",
3463 sizeof(struct inmem_pages));
3464 if (!inmem_entry_slab)
3465 goto destroy_sit_entry_set;
7fd9e544 3466 return 0;
184a5cd2 3467
88b88a66
JK		 3468destroy_sit_entry_set:
3469 kmem_cache_destroy(sit_entry_set_slab);
b01a9201
JK		 3470destroy_discard_cmd:
3471 kmem_cache_destroy(discard_cmd_slab);
6ab2a308 3472destroy_discard_entry:
184a5cd2
CY		 3473 kmem_cache_destroy(discard_entry_slab);
3474fail:
3475 return -ENOMEM;
7fd9e544
JK		 3476}
3477
3478void destroy_segment_manager_caches(void)
3479{
184a5cd2 3480 kmem_cache_destroy(sit_entry_set_slab);
b01a9201 3481 kmem_cache_destroy(discard_cmd_slab);
7fd9e544 3482 kmem_cache_destroy(discard_entry_slab);
88b88a66 3483 kmem_cache_destroy(inmem_entry_slab);
7fd9e544 3484}
Ubuntu Jammy Linux kernel mirror