]> 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: add a new function get_ssr_cost
[mirror_ubuntu-jammy-kernel.git] / fs / f2fs / segment.c
CommitLineData
0a8165d7 1/*
351df4b2
JK		 2 * fs/f2fs/segment.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11#include <linux/fs.h>
12#include <linux/f2fs_fs.h>
13#include <linux/bio.h>
14#include <linux/blkdev.h>
690e4a3e 15#include <linux/prefetch.h>
6b4afdd7 16#include <linux/kthread.h>
74de593a 17#include <linux/swap.h>
60b99b48 18#include <linux/timer.h>
351df4b2
JK		 19
20#include "f2fs.h"
21#include "segment.h"
22#include "node.h"
9e4ded3f 23#include "trace.h"
6ec178da 24#include <trace/events/f2fs.h>
351df4b2 25
9a7f143a
CL		 26#define __reverse_ffz(x) __reverse_ffs(~(x))
27
7fd9e544 28static struct kmem_cache *discard_entry_slab;
b01a9201 29static struct kmem_cache *discard_cmd_slab;
184a5cd2 30static struct kmem_cache *sit_entry_set_slab;
88b88a66 31static struct kmem_cache *inmem_entry_slab;
7fd9e544 32
f96999c3
JK		 33static unsigned long __reverse_ulong(unsigned char *str)
34{
35 unsigned long tmp = 0;
36 int shift = 24, idx = 0;
37
38#if BITS_PER_LONG == 64
39 shift = 56;
40#endif
41 while (shift >= 0) {
42 tmp |= (unsigned long)str[idx++] << shift;
43 shift -= BITS_PER_BYTE;
44 }
45 return tmp;
46}
47
9a7f143a
CL		 48/*
49 * __reverse_ffs is copied from include/asm-generic/bitops/__ffs.h since
50 * MSB and LSB are reversed in a byte by f2fs_set_bit.
51 */
52static inline unsigned long __reverse_ffs(unsigned long word)
53{
54 int num = 0;
55
56#if BITS_PER_LONG == 64
f96999c3 57 if ((word & 0xffffffff00000000UL) == 0)
9a7f143a 58 num += 32;
f96999c3 59 else
9a7f143a 60 word >>= 32;
9a7f143a 61#endif
f96999c3 62 if ((word & 0xffff0000) == 0)
9a7f143a 63 num += 16;
f96999c3 64 else
9a7f143a 65 word >>= 16;
f96999c3
JK		 66
67 if ((word & 0xff00) == 0)
9a7f143a 68 num += 8;
f96999c3 69 else
9a7f143a 70 word >>= 8;
f96999c3 71
9a7f143a
CL		 72 if ((word & 0xf0) == 0)
73 num += 4;
74 else
75 word >>= 4;
f96999c3 76
9a7f143a
CL		 77 if ((word & 0xc) == 0)
78 num += 2;
79 else
80 word >>= 2;
f96999c3 81
9a7f143a
CL		 82 if ((word & 0x2) == 0)
83 num += 1;
84 return num;
85}
86
87/*
e1c42045 88 * __find_rev_next(_zero)_bit is copied from lib/find_next_bit.c because
9a7f143a 89 * f2fs_set_bit makes MSB and LSB reversed in a byte.
692223d1 90 * @size must be integral times of unsigned long.
9a7f143a 91 * Example:
f96999c3
JK		 92 * MSB <--> LSB
93 * f2fs_set_bit(0, bitmap) => 1000 0000
94 * f2fs_set_bit(7, bitmap) => 0000 0001
9a7f143a
CL		 95 */
96static unsigned long __find_rev_next_bit(const unsigned long *addr,
97 unsigned long size, unsigned long offset)
98{
99 const unsigned long *p = addr + BIT_WORD(offset);
692223d1 100 unsigned long result = size;
9a7f143a 101 unsigned long tmp;
9a7f143a
CL		 102
103 if (offset >= size)
104 return size;
105
692223d1 106 size -= (offset & ~(BITS_PER_LONG - 1));
9a7f143a 107 offset %= BITS_PER_LONG;
f96999c3 108
692223d1
FL		 109 while (1) {
110 if (*p == 0)
111 goto pass;
9a7f143a 112
f96999c3 113 tmp = __reverse_ulong((unsigned char *)p);
692223d1
FL		 114
115 tmp &= ~0UL >> offset;
116 if (size < BITS_PER_LONG)
117 tmp &= (~0UL << (BITS_PER_LONG - size));
9a7f143a 118 if (tmp)
692223d1
FL		 119 goto found;
120pass:
121 if (size <= BITS_PER_LONG)
122 break;
9a7f143a 123 size -= BITS_PER_LONG;
692223d1 124 offset = 0;
f96999c3 125 p++;
9a7f143a 126 }
692223d1
FL		 127 return result;
128found:
129 return result - size + __reverse_ffs(tmp);
9a7f143a
CL		 130}
131
132static unsigned long __find_rev_next_zero_bit(const unsigned long *addr,
133 unsigned long size, unsigned long offset)
134{
135 const unsigned long *p = addr + BIT_WORD(offset);
80609448 136 unsigned long result = size;
9a7f143a 137 unsigned long tmp;
9a7f143a
CL		 138
139 if (offset >= size)
140 return size;
141
80609448 142 size -= (offset & ~(BITS_PER_LONG - 1));
9a7f143a 143 offset %= BITS_PER_LONG;
80609448
JK		 144
145 while (1) {
146 if (*p == ~0UL)
147 goto pass;
148
f96999c3 149 tmp = __reverse_ulong((unsigned char *)p);
80609448
JK		 150
151 if (offset)
152 tmp |= ~0UL << (BITS_PER_LONG - offset);
153 if (size < BITS_PER_LONG)
154 tmp |= ~0UL >> size;
f96999c3 155 if (tmp != ~0UL)
80609448
JK		 156 goto found;
157pass:
158 if (size <= BITS_PER_LONG)
159 break;
9a7f143a 160 size -= BITS_PER_LONG;
80609448 161 offset = 0;
f96999c3 162 p++;
9a7f143a 163 }
80609448
JK		 164 return result;
165found:
166 return result - size + __reverse_ffz(tmp);
9a7f143a
CL		 167}
168
88b88a66
JK		 169void register_inmem_page(struct inode *inode, struct page *page)
170{
171 struct f2fs_inode_info *fi = F2FS_I(inode);
172 struct inmem_pages *new;
9be32d72 173
9e4ded3f 174 f2fs_trace_pid(page);
0722b101 175
decd36b6
CY		 176 set_page_private(page, (unsigned long)ATOMIC_WRITTEN_PAGE);
177 SetPagePrivate(page);
178
88b88a66
JK		 179 new = f2fs_kmem_cache_alloc(inmem_entry_slab, GFP_NOFS);
180
181 /* add atomic page indices to the list */
182 new->page = page;
183 INIT_LIST_HEAD(&new->list);
decd36b6 184
88b88a66
JK		 185 /* increase reference count with clean state */
186 mutex_lock(&fi->inmem_lock);
187 get_page(page);
188 list_add_tail(&new->list, &fi->inmem_pages);
8dcf2ff7 189 inc_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
88b88a66 190 mutex_unlock(&fi->inmem_lock);
8ce67cb0
JK		 191
192 trace_f2fs_register_inmem_page(page, INMEM);
88b88a66
JK		 193}
194
28bc106b
CY		 195static int __revoke_inmem_pages(struct inode *inode,
196 struct list_head *head, bool drop, bool recover)
29b96b54 197{
28bc106b 198 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
29b96b54 199 struct inmem_pages *cur, *tmp;
28bc106b 200 int err = 0;
29b96b54
CY		 201
202 list_for_each_entry_safe(cur, tmp, head, list) {
28bc106b
CY		 203 struct page *page = cur->page;
204
205 if (drop)
206 trace_f2fs_commit_inmem_page(page, INMEM_DROP);
207
208 lock_page(page);
29b96b54 209
28bc106b
CY		 210 if (recover) {
211 struct dnode_of_data dn;
212 struct node_info ni;
213
214 trace_f2fs_commit_inmem_page(page, INMEM_REVOKE);
215
216 set_new_dnode(&dn, inode, NULL, NULL, 0);
217 if (get_dnode_of_data(&dn, page->index, LOOKUP_NODE)) {
218 err = -EAGAIN;
219 goto next;
220 }
221 get_node_info(sbi, dn.nid, &ni);
222 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
223 cur->old_addr, ni.version, true, true);
224 f2fs_put_dnode(&dn);
225 }
226next:
63c52d78
JK		 227 /* we don't need to invalidate this in the sccessful status */
228 if (drop || recover)
229 ClearPageUptodate(page);
28bc106b 230 set_page_private(page, 0);
c81ced05 231 ClearPagePrivate(page);
28bc106b 232 f2fs_put_page(page, 1);
29b96b54
CY		 233
234 list_del(&cur->list);
235 kmem_cache_free(inmem_entry_slab, cur);
236 dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
237 }
28bc106b 238 return err;
29b96b54
CY		 239}
240
241void drop_inmem_pages(struct inode *inode)
242{
243 struct f2fs_inode_info *fi = F2FS_I(inode);
244
245 mutex_lock(&fi->inmem_lock);
28bc106b 246 __revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
29b96b54 247 mutex_unlock(&fi->inmem_lock);
5fe45743
CY		 248
249 clear_inode_flag(inode, FI_ATOMIC_FILE);
250 stat_dec_atomic_write(inode);
29b96b54
CY		 251}
252
8c242db9
JK		 253void drop_inmem_page(struct inode *inode, struct page *page)
254{
255 struct f2fs_inode_info *fi = F2FS_I(inode);
256 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
257 struct list_head *head = &fi->inmem_pages;
258 struct inmem_pages *cur = NULL;
259
260 f2fs_bug_on(sbi, !IS_ATOMIC_WRITTEN_PAGE(page));
261
262 mutex_lock(&fi->inmem_lock);
263 list_for_each_entry(cur, head, list) {
264 if (cur->page == page)
265 break;
266 }
267
268 f2fs_bug_on(sbi, !cur || cur->page != page);
269 list_del(&cur->list);
270 mutex_unlock(&fi->inmem_lock);
271
272 dec_page_count(sbi, F2FS_INMEM_PAGES);
273 kmem_cache_free(inmem_entry_slab, cur);
274
275 ClearPageUptodate(page);
276 set_page_private(page, 0);
277 ClearPagePrivate(page);
278 f2fs_put_page(page, 0);
279
280 trace_f2fs_commit_inmem_page(page, INMEM_INVALIDATE);
281}
282
28bc106b
CY		 283static int __commit_inmem_pages(struct inode *inode,
284 struct list_head *revoke_list)
88b88a66
JK		 285{
286 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
287 struct f2fs_inode_info *fi = F2FS_I(inode);
288 struct inmem_pages *cur, *tmp;
88b88a66 289 struct f2fs_io_info fio = {
05ca3632 290 .sbi = sbi,
88b88a66 291 .type = DATA,
04d328de 292 .op = REQ_OP_WRITE,
70fd7614 293 .op_flags = REQ_SYNC | REQ_PRIO,
88b88a66 294 };
942fd319 295 pgoff_t last_idx = ULONG_MAX;
edb27dee 296 int err = 0;
88b88a66 297
88b88a66 298 list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
28bc106b
CY		 299 struct page *page = cur->page;
300
301 lock_page(page);
302 if (page->mapping == inode->i_mapping) {
303 trace_f2fs_commit_inmem_page(page, INMEM);
304
305 set_page_dirty(page);
306 f2fs_wait_on_page_writeback(page, DATA, true);
933439c8 307 if (clear_page_dirty_for_io(page)) {
29b96b54 308 inode_dec_dirty_pages(inode);
933439c8
CY		 309 remove_dirty_inode(inode);
310 }
28bc106b
CY		 311
312 fio.page = page;
e959c8f5 313 fio.old_blkaddr = NULL_ADDR;
4d978078 314 fio.encrypted_page = NULL;
cc15620b 315 fio.need_lock = LOCK_DONE;
29b96b54
CY		 316 err = do_write_data_page(&fio);
317 if (err) {
28bc106b 318 unlock_page(page);
29b96b54 319 break;
70c640b1 320 }
29b96b54 321
28bc106b
CY		 322 /* record old blkaddr for revoking */
323 cur->old_addr = fio.old_blkaddr;
942fd319 324 last_idx = page->index;
28bc106b
CY		 325 }
326 unlock_page(page);
327 list_move_tail(&cur->list, revoke_list);
88b88a66 328 }
29b96b54 329
942fd319 330 if (last_idx != ULONG_MAX)
b9109b0e 331 f2fs_submit_merged_write_cond(sbi, inode, 0, last_idx, DATA);
28bc106b
CY		 332
333 if (!err)
334 __revoke_inmem_pages(inode, revoke_list, false, false);
335
29b96b54
CY		 336 return err;
337}
338
339int commit_inmem_pages(struct inode *inode)
340{
341 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
342 struct f2fs_inode_info *fi = F2FS_I(inode);
28bc106b
CY		 343 struct list_head revoke_list;
344 int err;
29b96b54 345
28bc106b 346 INIT_LIST_HEAD(&revoke_list);
29b96b54
CY		 347 f2fs_balance_fs(sbi, true);
348 f2fs_lock_op(sbi);
349
5fe45743
CY		 350 set_inode_flag(inode, FI_ATOMIC_COMMIT);
351
29b96b54 352 mutex_lock(&fi->inmem_lock);
28bc106b
CY		 353 err = __commit_inmem_pages(inode, &revoke_list);
354 if (err) {
355 int ret;
356 /*
357 * try to revoke all committed pages, but still we could fail
358 * due to no memory or other reason, if that happened, EAGAIN
359 * will be returned, which means in such case, transaction is
360 * already not integrity, caller should use journal to do the
361 * recovery or rewrite & commit last transaction. For other
362 * error number, revoking was done by filesystem itself.
363 */
364 ret = __revoke_inmem_pages(inode, &revoke_list, false, true);
365 if (ret)
366 err = ret;
367
368 /* drop all uncommitted pages */
369 __revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
370 }
88b88a66
JK		 371 mutex_unlock(&fi->inmem_lock);
372
5fe45743
CY		 373 clear_inode_flag(inode, FI_ATOMIC_COMMIT);
374
29b96b54 375 f2fs_unlock_op(sbi);
edb27dee 376 return err;
88b88a66
JK		 377}
378
0a8165d7 379/*
351df4b2
JK		 380 * This function balances dirty node and dentry pages.
381 * In addition, it controls garbage collection.
382 */
2c4db1a6 383void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
351df4b2 384{
0f348028 385#ifdef CONFIG_F2FS_FAULT_INJECTION
55523519
CY		 386 if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
387 f2fs_show_injection_info(FAULT_CHECKPOINT);
0f348028 388 f2fs_stop_checkpoint(sbi, false);
55523519 389 }
0f348028
CY		 390#endif
391
e589c2c4 392 /* balance_fs_bg is able to be pending */
a7881893 393 if (need && excess_cached_nats(sbi))
e589c2c4
JK		 394 f2fs_balance_fs_bg(sbi);
395
351df4b2 396 /*
029cd28c
JK		 397 * We should do GC or end up with checkpoint, if there are so many dirty
398 * dir/node pages without enough free segments.
351df4b2 399 */
7f3037a5 400 if (has_not_enough_free_secs(sbi, 0, 0)) {
351df4b2 401 mutex_lock(&sbi->gc_mutex);
e066b83c 402 f2fs_gc(sbi, false, false, NULL_SEGNO);
351df4b2
JK		 403 }
404}
405
4660f9c0
JK		 406void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi)
407{
1dcc336b 408 /* try to shrink extent cache when there is no enough memory */
554df79e
JK		 409 if (!available_free_memory(sbi, EXTENT_CACHE))
410 f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER);
1dcc336b 411
1b38dc8e
JK		 412 /* check the # of cached NAT entries */
413 if (!available_free_memory(sbi, NAT_ENTRIES))
414 try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK);
415
31696580 416 if (!available_free_memory(sbi, FREE_NIDS))
ad4edb83
JK		 417 try_to_free_nids(sbi, MAX_FREE_NIDS);
418 else
22ad0b6a 419 build_free_nids(sbi, false, false);
31696580 420
1c0f4bf5 421 if (!is_idle(sbi) && !excess_dirty_nats(sbi))
f455c8a5 422 return;
31696580 423
1b38dc8e
JK		 424 /* checkpoint is the only way to shrink partial cached entries */
425 if (!available_free_memory(sbi, NAT_ENTRIES) ||
60b99b48 426 !available_free_memory(sbi, INO_ENTRIES) ||
7d768d2c
CY		 427 excess_prefree_segs(sbi) ||
428 excess_dirty_nats(sbi) ||
f455c8a5 429 f2fs_time_over(sbi, CP_TIME)) {
e9f5b8b8
CY		 430 if (test_opt(sbi, DATA_FLUSH)) {
431 struct blk_plug plug;
432
433 blk_start_plug(&plug);
36b35a0d 434 sync_dirty_inodes(sbi, FILE_INODE);
e9f5b8b8
CY		 435 blk_finish_plug(&plug);
436 }
4660f9c0 437 f2fs_sync_fs(sbi->sb, true);
42190d2a 438 stat_inc_bg_cp_count(sbi->stat_info);
36b35a0d 439 }
4660f9c0
JK		 440}
441
20fda56b
KM		 442static int __submit_flush_wait(struct f2fs_sb_info *sbi,
443 struct block_device *bdev)
3c62be17
JK		 444{
445 struct bio *bio = f2fs_bio_alloc(0);
446 int ret;
447
3adc5fcb 448 bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH;
3c62be17
JK		 449 bio->bi_bdev = bdev;
450 ret = submit_bio_wait(bio);
451 bio_put(bio);
20fda56b
KM		 452
453 trace_f2fs_issue_flush(bdev, test_opt(sbi, NOBARRIER),
454 test_opt(sbi, FLUSH_MERGE), ret);
3c62be17
JK		 455 return ret;
456}
457
458static int submit_flush_wait(struct f2fs_sb_info *sbi)
459{
20fda56b 460 int ret = __submit_flush_wait(sbi, sbi->sb->s_bdev);
3c62be17
JK		 461 int i;
462
20fda56b
KM		 463 if (!sbi->s_ndevs || ret)
464 return ret;
465
466 for (i = 1; i < sbi->s_ndevs; i++) {
467 ret = __submit_flush_wait(sbi, FDEV(i).bdev);
468 if (ret)
469 break;
3c62be17
JK		 470 }
471 return ret;
472}
473
2163d198 474static int issue_flush_thread(void *data)
6b4afdd7
JK		 475{
476 struct f2fs_sb_info *sbi = data;
b01a9201 477 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
a688b9d9 478 wait_queue_head_t *q = &fcc->flush_wait_queue;
6b4afdd7
JK		 479repeat:
480 if (kthread_should_stop())
481 return 0;
482
721bd4d5 483 if (!llist_empty(&fcc->issue_list)) {
6b4afdd7
JK		 484 struct flush_cmd *cmd, *next;
485 int ret;
486
721bd4d5
GZ		 487 fcc->dispatch_list = llist_del_all(&fcc->issue_list);
488 fcc->dispatch_list = llist_reverse_order(fcc->dispatch_list);
489
3c62be17 490 ret = submit_flush_wait(sbi);
8b8dd65f
CY		 491 atomic_inc(&fcc->issued_flush);
492
721bd4d5
GZ		 493 llist_for_each_entry_safe(cmd, next,
494 fcc->dispatch_list, llnode) {
6b4afdd7 495 cmd->ret = ret;
6b4afdd7
JK		 496 complete(&cmd->wait);
497 }
a688b9d9 498 fcc->dispatch_list = NULL;
6b4afdd7
JK		 499 }
500
a688b9d9 501 wait_event_interruptible(*q,
721bd4d5 502 kthread_should_stop() || !llist_empty(&fcc->issue_list));
6b4afdd7
JK		 503 goto repeat;
504}
505
506int f2fs_issue_flush(struct f2fs_sb_info *sbi)
507{
b01a9201 508 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
adf8d90b 509 struct flush_cmd cmd;
8b8dd65f 510 int ret;
6b4afdd7 511
0f7b2abd
JK		 512 if (test_opt(sbi, NOBARRIER))
513 return 0;
514
8b8dd65f
CY		 515 if (!test_opt(sbi, FLUSH_MERGE)) {
516 ret = submit_flush_wait(sbi);
517 atomic_inc(&fcc->issued_flush);
518 return ret;
519 }
740432f8 520
8b8dd65f
CY		 521 if (!atomic_read(&fcc->issing_flush)) {
522 atomic_inc(&fcc->issing_flush);
3c62be17 523 ret = submit_flush_wait(sbi);
8b8dd65f
CY		 524 atomic_dec(&fcc->issing_flush);
525
526 atomic_inc(&fcc->issued_flush);
740432f8
JK		 527 return ret;
528 }
6b4afdd7 529
adf8d90b 530 init_completion(&cmd.wait);
6b4afdd7 531
8b8dd65f 532 atomic_inc(&fcc->issing_flush);
721bd4d5 533 llist_add(&cmd.llnode, &fcc->issue_list);
6b4afdd7 534
a688b9d9
GZ		 535 if (!fcc->dispatch_list)
536 wake_up(&fcc->flush_wait_queue);
6b4afdd7 537
5eba8c5d
JK		 538 if (fcc->f2fs_issue_flush) {
539 wait_for_completion(&cmd.wait);
8b8dd65f 540 atomic_dec(&fcc->issing_flush);
5eba8c5d
JK		 541 } else {
542 llist_del_all(&fcc->issue_list);
8b8dd65f 543 atomic_set(&fcc->issing_flush, 0);
5eba8c5d 544 }
adf8d90b
CY		 545
546 return cmd.ret;
6b4afdd7
JK		 547}
548
2163d198
GZ		 549int create_flush_cmd_control(struct f2fs_sb_info *sbi)
550{
551 dev_t dev = sbi->sb->s_bdev->bd_dev;
552 struct flush_cmd_control *fcc;
553 int err = 0;
554
b01a9201
JK		 555 if (SM_I(sbi)->fcc_info) {
556 fcc = SM_I(sbi)->fcc_info;
5eba8c5d
JK		 557 goto init_thread;
558 }
559
2163d198
GZ		 560 fcc = kzalloc(sizeof(struct flush_cmd_control), GFP_KERNEL);
561 if (!fcc)
562 return -ENOMEM;
8b8dd65f
CY		 563 atomic_set(&fcc->issued_flush, 0);
564 atomic_set(&fcc->issing_flush, 0);
2163d198 565 init_waitqueue_head(&fcc->flush_wait_queue);
721bd4d5 566 init_llist_head(&fcc->issue_list);
b01a9201 567 SM_I(sbi)->fcc_info = fcc;
5eba8c5d 568init_thread:
2163d198
GZ		 569 fcc->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi,
570 "f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev));
571 if (IS_ERR(fcc->f2fs_issue_flush)) {
572 err = PTR_ERR(fcc->f2fs_issue_flush);
573 kfree(fcc);
b01a9201 574 SM_I(sbi)->fcc_info = NULL;
2163d198
GZ		 575 return err;
576 }
2163d198
GZ		 577
578 return err;
579}
580
5eba8c5d 581void destroy_flush_cmd_control(struct f2fs_sb_info *sbi, bool free)
2163d198 582{
b01a9201 583 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
2163d198 584
5eba8c5d
JK		 585 if (fcc && fcc->f2fs_issue_flush) {
586 struct task_struct *flush_thread = fcc->f2fs_issue_flush;
587
588 fcc->f2fs_issue_flush = NULL;
589 kthread_stop(flush_thread);
590 }
591 if (free) {
592 kfree(fcc);
b01a9201 593 SM_I(sbi)->fcc_info = NULL;
5eba8c5d 594 }
2163d198
GZ		 595}
596
351df4b2
JK		 597static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
598 enum dirty_type dirty_type)
599{
600 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
601
602 /* need not be added */
603 if (IS_CURSEG(sbi, segno))
604 return;
605
606 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[dirty_type]))
607 dirty_i->nr_dirty[dirty_type]++;
608
609 if (dirty_type == DIRTY) {
610 struct seg_entry *sentry = get_seg_entry(sbi, segno);
4625d6aa 611 enum dirty_type t = sentry->type;
b2f2c390 612
ec325b52
JK		 613 if (unlikely(t >= DIRTY)) {
614 f2fs_bug_on(sbi, 1);
615 return;
616 }
4625d6aa
CL		 617 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[t]))
618 dirty_i->nr_dirty[t]++;
351df4b2
JK		 619 }
620}
621
622static void __remove_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
623 enum dirty_type dirty_type)
624{
625 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
626
627 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[dirty_type]))
628 dirty_i->nr_dirty[dirty_type]--;
629
630 if (dirty_type == DIRTY) {
4625d6aa
CL		 631 struct seg_entry *sentry = get_seg_entry(sbi, segno);
632 enum dirty_type t = sentry->type;
633
634 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
635 dirty_i->nr_dirty[t]--;
b2f2c390 636
302bd348 637 if (get_valid_blocks(sbi, segno, true) == 0)
4ddb1a4d 638 clear_bit(GET_SEC_FROM_SEG(sbi, segno),
5ec4e49f 639 dirty_i->victim_secmap);
351df4b2
JK		 640 }
641}
642
0a8165d7 643/*
351df4b2
JK		 644 * Should not occur error such as -ENOMEM.
645 * Adding dirty entry into seglist is not critical operation.
646 * If a given segment is one of current working segments, it won't be added.
647 */
8d8451af 648static void locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 649{
650 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
651 unsigned short valid_blocks;
652
653 if (segno == NULL_SEGNO || IS_CURSEG(sbi, segno))
654 return;
655
656 mutex_lock(&dirty_i->seglist_lock);
657
302bd348 658 valid_blocks = get_valid_blocks(sbi, segno, false);
351df4b2
JK		 659
660 if (valid_blocks == 0) {
661 __locate_dirty_segment(sbi, segno, PRE);
662 __remove_dirty_segment(sbi, segno, DIRTY);
663 } else if (valid_blocks < sbi->blocks_per_seg) {
664 __locate_dirty_segment(sbi, segno, DIRTY);
665 } else {
666 /* Recovery routine with SSR needs this */
667 __remove_dirty_segment(sbi, segno, DIRTY);
668 }
669
670 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 671}
672
004b6862 673static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi,
c81abe34
JK		 674 struct block_device *bdev, block_t lstart,
675 block_t start, block_t len)
275b66b0 676{
0b54fb84 677 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
ba48a33e 678 struct list_head *pend_list;
b01a9201 679 struct discard_cmd *dc;
275b66b0 680
ba48a33e
CY		 681 f2fs_bug_on(sbi, !len);
682
683 pend_list = &dcc->pend_list[plist_idx(len)];
684
b01a9201
JK		 685 dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS);
686 INIT_LIST_HEAD(&dc->list);
c81abe34 687 dc->bdev = bdev;
b01a9201 688 dc->lstart = lstart;
c81abe34 689 dc->start = start;
b01a9201 690 dc->len = len;
ec9895ad 691 dc->ref = 0;
15469963 692 dc->state = D_PREP;
c81abe34 693 dc->error = 0;
b01a9201 694 init_completion(&dc->wait);
22d375dd 695 list_add_tail(&dc->list, pend_list);
5f32366a 696 atomic_inc(&dcc->discard_cmd_cnt);
d84d1cbd 697 dcc->undiscard_blks += len;
004b6862
CY		 698
699 return dc;
700}
701
702static struct discard_cmd *__attach_discard_cmd(struct f2fs_sb_info *sbi,
703 struct block_device *bdev, block_t lstart,
704 block_t start, block_t len,
705 struct rb_node *parent, struct rb_node **p)
706{
707 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
708 struct discard_cmd *dc;
709
710 dc = __create_discard_cmd(sbi, bdev, lstart, start, len);
711
712 rb_link_node(&dc->rb_node, parent, p);
713 rb_insert_color(&dc->rb_node, &dcc->root);
714
715 return dc;
15469963
JK		 716}
717
004b6862
CY		 718static void __detach_discard_cmd(struct discard_cmd_control *dcc,
719 struct discard_cmd *dc)
15469963 720{
dcc9165d 721 if (dc->state == D_DONE)
004b6862
CY		 722 atomic_dec(&dcc->issing_discard);
723
724 list_del(&dc->list);
725 rb_erase(&dc->rb_node, &dcc->root);
d84d1cbd 726 dcc->undiscard_blks -= dc->len;
004b6862
CY		 727
728 kmem_cache_free(discard_cmd_slab, dc);
729
730 atomic_dec(&dcc->discard_cmd_cnt);
731}
732
733static void __remove_discard_cmd(struct f2fs_sb_info *sbi,
734 struct discard_cmd *dc)
735{
736 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
dcc9165d 737
c81abe34
JK		 738 if (dc->error == -EOPNOTSUPP)
739 dc->error = 0;
15469963 740
c81abe34 741 if (dc->error)
15469963 742 f2fs_msg(sbi->sb, KERN_INFO,
c81abe34 743 "Issue discard failed, ret: %d", dc->error);
004b6862 744 __detach_discard_cmd(dcc, dc);
275b66b0
CY		 745}
746
c81abe34
JK		 747static void f2fs_submit_discard_endio(struct bio *bio)
748{
749 struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
750
c81abe34
JK		 751 dc->error = bio->bi_error;
752 dc->state = D_DONE;
fa64a003 753 complete(&dc->wait);
c81abe34
JK		 754 bio_put(bio);
755}
756
757/* this function is copied from blkdev_issue_discard from block/blk-lib.c */
758static void __submit_discard_cmd(struct f2fs_sb_info *sbi,
759 struct discard_cmd *dc)
760{
761 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
762 struct bio *bio = NULL;
763
764 if (dc->state != D_PREP)
765 return;
766
0243a5f9
CY		 767 trace_f2fs_issue_discard(dc->bdev, dc->start, dc->len);
768
c81abe34
JK		 769 dc->error = __blkdev_issue_discard(dc->bdev,
770 SECTOR_FROM_BLOCK(dc->start),
771 SECTOR_FROM_BLOCK(dc->len),
772 GFP_NOFS, 0, &bio);
773 if (!dc->error) {
774 /* should keep before submission to avoid D_DONE right away */
775 dc->state = D_SUBMIT;
8b8dd65f
CY		 776 atomic_inc(&dcc->issued_discard);
777 atomic_inc(&dcc->issing_discard);
c81abe34
JK		 778 if (bio) {
779 bio->bi_private = dc;
780 bio->bi_end_io = f2fs_submit_discard_endio;
781 bio->bi_opf |= REQ_SYNC;
782 submit_bio(bio);
46f84c2c 783 list_move_tail(&dc->list, &dcc->wait_list);
c81abe34
JK		 784 }
785 } else {
786 __remove_discard_cmd(sbi, dc);
787 }
788}
789
004b6862
CY		 790static struct discard_cmd *__insert_discard_tree(struct f2fs_sb_info *sbi,
791 struct block_device *bdev, block_t lstart,
792 block_t start, block_t len,
793 struct rb_node **insert_p,
794 struct rb_node *insert_parent)
c81abe34 795{
004b6862
CY		 796 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
797 struct rb_node **p = &dcc->root.rb_node;
798 struct rb_node *parent = NULL;
799 struct discard_cmd *dc = NULL;
800
801 if (insert_p && insert_parent) {
802 parent = insert_parent;
803 p = insert_p;
804 goto do_insert;
805 }
c81abe34 806
004b6862
CY		 807 p = __lookup_rb_tree_for_insert(sbi, &dcc->root, &parent, lstart);
808do_insert:
809 dc = __attach_discard_cmd(sbi, bdev, lstart, start, len, parent, p);
810 if (!dc)
811 return NULL;
c81abe34 812
004b6862 813 return dc;
c81abe34
JK		 814}
815
ba48a33e
CY		 816static void __relocate_discard_cmd(struct discard_cmd_control *dcc,
817 struct discard_cmd *dc)
818{
819 list_move_tail(&dc->list, &dcc->pend_list[plist_idx(dc->len)]);
820}
821
3d6a650f
YH		 822static void __punch_discard_cmd(struct f2fs_sb_info *sbi,
823 struct discard_cmd *dc, block_t blkaddr)
824{
ba48a33e 825 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 826 struct discard_info di = dc->di;
827 bool modified = false;
3d6a650f 828
004b6862 829 if (dc->state == D_DONE || dc->len == 1) {
3d6a650f
YH		 830 __remove_discard_cmd(sbi, dc);
831 return;
832 }
833
d84d1cbd
CY		 834 dcc->undiscard_blks -= di.len;
835
004b6862 836 if (blkaddr > di.lstart) {
3d6a650f 837 dc->len = blkaddr - dc->lstart;
d84d1cbd 838 dcc->undiscard_blks += dc->len;
ba48a33e 839 __relocate_discard_cmd(dcc, dc);
df0f6b44 840 f2fs_bug_on(sbi, !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862
CY		 841 modified = true;
842 }
843
844 if (blkaddr < di.lstart + di.len - 1) {
845 if (modified) {
846 __insert_discard_tree(sbi, dc->bdev, blkaddr + 1,
847 di.start + blkaddr + 1 - di.lstart,
848 di.lstart + di.len - 1 - blkaddr,
849 NULL, NULL);
df0f6b44
CY		 850 f2fs_bug_on(sbi,
851 !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862
CY		 852 } else {
853 dc->lstart++;
854 dc->len--;
855 dc->start++;
d84d1cbd 856 dcc->undiscard_blks += dc->len;
ba48a33e 857 __relocate_discard_cmd(dcc, dc);
df0f6b44
CY		 858 f2fs_bug_on(sbi,
859 !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862 860 }
3d6a650f
YH		 861 }
862}
863
004b6862
CY		 864static void __update_discard_tree_range(struct f2fs_sb_info *sbi,
865 struct block_device *bdev, block_t lstart,
866 block_t start, block_t len)
275b66b0 867{
0b54fb84 868 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 869 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
870 struct discard_cmd *dc;
871 struct discard_info di = {0};
872 struct rb_node **insert_p = NULL, *insert_parent = NULL;
873 block_t end = lstart + len;
275b66b0 874
15469963 875 mutex_lock(&dcc->cmd_lock);
40465257 876
004b6862
CY		 877 dc = (struct discard_cmd *)__lookup_rb_tree_ret(&dcc->root,
878 NULL, lstart,
879 (struct rb_entry **)&prev_dc,
880 (struct rb_entry **)&next_dc,
881 &insert_p, &insert_parent, true);
882 if (dc)
883 prev_dc = dc;
884
885 if (!prev_dc) {
886 di.lstart = lstart;
887 di.len = next_dc ? next_dc->lstart - lstart : len;
888 di.len = min(di.len, len);
889 di.start = start;
22d375dd 890 }
15469963 891
004b6862
CY		 892 while (1) {
893 struct rb_node *node;
894 bool merged = false;
895 struct discard_cmd *tdc = NULL;
896
897 if (prev_dc) {
898 di.lstart = prev_dc->lstart + prev_dc->len;
899 if (di.lstart < lstart)
900 di.lstart = lstart;
901 if (di.lstart >= end)
902 break;
903
904 if (!next_dc || next_dc->lstart > end)
905 di.len = end - di.lstart;
906 else
907 di.len = next_dc->lstart - di.lstart;
908 di.start = start + di.lstart - lstart;
909 }
910
911 if (!di.len)
912 goto next;
913
914 if (prev_dc && prev_dc->state == D_PREP &&
915 prev_dc->bdev == bdev &&
916 __is_discard_back_mergeable(&di, &prev_dc->di)) {
917 prev_dc->di.len += di.len;
d84d1cbd 918 dcc->undiscard_blks += di.len;
ba48a33e 919 __relocate_discard_cmd(dcc, prev_dc);
df0f6b44
CY		 920 f2fs_bug_on(sbi,
921 !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862
CY		 922 di = prev_dc->di;
923 tdc = prev_dc;
924 merged = true;
925 }
926
927 if (next_dc && next_dc->state == D_PREP &&
928 next_dc->bdev == bdev &&
929 __is_discard_front_mergeable(&di, &next_dc->di)) {
930 next_dc->di.lstart = di.lstart;
931 next_dc->di.len += di.len;
932 next_dc->di.start = di.start;
d84d1cbd 933 dcc->undiscard_blks += di.len;
ba48a33e 934 __relocate_discard_cmd(dcc, next_dc);
004b6862
CY		 935 if (tdc)
936 __remove_discard_cmd(sbi, tdc);
df0f6b44
CY		 937 f2fs_bug_on(sbi,
938 !__check_rb_tree_consistence(sbi, &dcc->root));
004b6862 939 merged = true;
4e6a8d9b 940 }
004b6862 941
df0f6b44 942 if (!merged) {
004b6862
CY		 943 __insert_discard_tree(sbi, bdev, di.lstart, di.start,
944 di.len, NULL, NULL);
df0f6b44
CY		 945 f2fs_bug_on(sbi,
946 !__check_rb_tree_consistence(sbi, &dcc->root));
947 }
004b6862
CY		 948 next:
949 prev_dc = next_dc;
950 if (!prev_dc)
951 break;
952
953 node = rb_next(&prev_dc->rb_node);
954 next_dc = rb_entry_safe(node, struct discard_cmd, rb_node);
955 }
956
957 mutex_unlock(&dcc->cmd_lock);
958}
959
960static int __queue_discard_cmd(struct f2fs_sb_info *sbi,
961 struct block_device *bdev, block_t blkstart, block_t blklen)
962{
963 block_t lblkstart = blkstart;
964
0243a5f9 965 trace_f2fs_queue_discard(bdev, blkstart, blklen);
004b6862
CY		 966
967 if (sbi->s_ndevs) {
968 int devi = f2fs_target_device_index(sbi, blkstart);
969
970 blkstart -= FDEV(devi).start_blk;
971 }
972 __update_discard_tree_range(sbi, bdev, lblkstart, blkstart, blklen);
004b6862
CY		 973 return 0;
974}
975
bd5b0738
CY		 976static void __issue_discard_cmd(struct f2fs_sb_info *sbi, bool issue_cond)
977{
978 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
979 struct list_head *pend_list;
980 struct discard_cmd *dc, *tmp;
981 struct blk_plug plug;
982 int i, iter = 0;
983
984 mutex_lock(&dcc->cmd_lock);
985 blk_start_plug(&plug);
986 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
987 pend_list = &dcc->pend_list[i];
988 list_for_each_entry_safe(dc, tmp, pend_list, list) {
989 f2fs_bug_on(sbi, dc->state != D_PREP);
990
991 if (!issue_cond || is_idle(sbi))
992 __submit_discard_cmd(sbi, dc);
993 if (issue_cond && iter++ > DISCARD_ISSUE_RATE)
994 goto out;
995 }
996 }
997out:
998 blk_finish_plug(&plug);
999 mutex_unlock(&dcc->cmd_lock);
1000}
1001
63a94fa1
CY		 1002static void __wait_discard_cmd(struct f2fs_sb_info *sbi, bool wait_cond)
1003{
1004 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1005 struct list_head *wait_list = &(dcc->wait_list);
1006 struct discard_cmd *dc, *tmp;
1007
1008 mutex_lock(&dcc->cmd_lock);
1009 list_for_each_entry_safe(dc, tmp, wait_list, list) {
1010 if (!wait_cond || dc->state == D_DONE) {
ec9895ad
CY		 1011 if (dc->ref)
1012 continue;
63a94fa1
CY		 1013 wait_for_completion_io(&dc->wait);
1014 __remove_discard_cmd(sbi, dc);
1015 }
1016 }
1017 mutex_unlock(&dcc->cmd_lock);
1018}
1019
004b6862
CY		 1020/* This should be covered by global mutex, &sit_i->sentry_lock */
1021void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
1022{
1023 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1024 struct discard_cmd *dc;
ec9895ad 1025 bool need_wait = false;
004b6862
CY		 1026
1027 mutex_lock(&dcc->cmd_lock);
004b6862
CY		 1028 dc = (struct discard_cmd *)__lookup_rb_tree(&dcc->root, NULL, blkaddr);
1029 if (dc) {
ec9895ad
CY		 1030 if (dc->state == D_PREP) {
1031 __punch_discard_cmd(sbi, dc, blkaddr);
1032 } else {
1033 dc->ref++;
1034 need_wait = true;
1035 }
275b66b0 1036 }
d431413f 1037 mutex_unlock(&dcc->cmd_lock);
ec9895ad
CY		 1038
1039 if (need_wait) {
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 }
d431413f
CY		 1048}
1049
1050/* This comes from f2fs_put_super */
1051void f2fs_wait_discard_bios(struct f2fs_sb_info *sbi)
1052{
bd5b0738 1053 __issue_discard_cmd(sbi, false);
63a94fa1 1054 __wait_discard_cmd(sbi, false);
275b66b0
CY		 1055}
1056
15469963
JK		 1057static int issue_discard_thread(void *data)
1058{
1059 struct f2fs_sb_info *sbi = data;
1060 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1061 wait_queue_head_t *q = &dcc->discard_wait_queue;
15469963
JK		 1062repeat:
1063 if (kthread_should_stop())
1064 return 0;
1065
bd5b0738 1066 __issue_discard_cmd(sbi, true);
63a94fa1 1067 __wait_discard_cmd(sbi, true);
15469963 1068
15469963
JK		 1069 congestion_wait(BLK_RW_SYNC, HZ/50);
1070
22d375dd 1071 wait_event_interruptible(*q, kthread_should_stop() ||
ba48a33e 1072 atomic_read(&dcc->discard_cmd_cnt));
15469963
JK		 1073 goto repeat;
1074}
1075
f46e8809 1076#ifdef CONFIG_BLK_DEV_ZONED
3c62be17
JK		 1077static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
1078 struct block_device *bdev, block_t blkstart, block_t blklen)
f46e8809 1079{
92592285 1080 sector_t sector, nr_sects;
10a875f8 1081 block_t lblkstart = blkstart;
3c62be17
JK		 1082 int devi = 0;
1083
1084 if (sbi->s_ndevs) {
1085 devi = f2fs_target_device_index(sbi, blkstart);
1086 blkstart -= FDEV(devi).start_blk;
1087 }
f46e8809
DLM		 1088
1089 /*
1090 * We need to know the type of the zone: for conventional zones,
1091 * use regular discard if the drive supports it. For sequential
1092 * zones, reset the zone write pointer.
1093 */
3c62be17 1094 switch (get_blkz_type(sbi, bdev, blkstart)) {
f46e8809
DLM		 1095
1096 case BLK_ZONE_TYPE_CONVENTIONAL:
1097 if (!blk_queue_discard(bdev_get_queue(bdev)))
1098 return 0;
c81abe34 1099 return __queue_discard_cmd(sbi, bdev, lblkstart, blklen);
f46e8809
DLM		 1100 case BLK_ZONE_TYPE_SEQWRITE_REQ:
1101 case BLK_ZONE_TYPE_SEQWRITE_PREF:
92592285
JK		 1102 sector = SECTOR_FROM_BLOCK(blkstart);
1103 nr_sects = SECTOR_FROM_BLOCK(blklen);
1104
1105 if (sector & (bdev_zone_sectors(bdev) - 1) ||
1106 nr_sects != bdev_zone_sectors(bdev)) {
1107 f2fs_msg(sbi->sb, KERN_INFO,
1108 "(%d) %s: Unaligned discard attempted (block %x + %x)",
1109 devi, sbi->s_ndevs ? FDEV(devi).path: "",
1110 blkstart, blklen);
1111 return -EIO;
1112 }
d50aaeec 1113 trace_f2fs_issue_reset_zone(bdev, blkstart);
f46e8809
DLM		 1114 return blkdev_reset_zones(bdev, sector,
1115 nr_sects, GFP_NOFS);
1116 default:
1117 /* Unknown zone type: broken device ? */
1118 return -EIO;
1119 }
1120}
1121#endif
1122
3c62be17
JK		 1123static int __issue_discard_async(struct f2fs_sb_info *sbi,
1124 struct block_device *bdev, block_t blkstart, block_t blklen)
1125{
1126#ifdef CONFIG_BLK_DEV_ZONED
1127 if (f2fs_sb_mounted_blkzoned(sbi->sb) &&
1128 bdev_zoned_model(bdev) != BLK_ZONED_NONE)
1129 return __f2fs_issue_discard_zone(sbi, bdev, blkstart, blklen);
1130#endif
c81abe34 1131 return __queue_discard_cmd(sbi, bdev, blkstart, blklen);
3c62be17
JK		 1132}
1133
1e87a78d 1134static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
37208879
JK		 1135 block_t blkstart, block_t blklen)
1136{
3c62be17
JK		 1137 sector_t start = blkstart, len = 0;
1138 struct block_device *bdev;
a66cdd98
JK		 1139 struct seg_entry *se;
1140 unsigned int offset;
1141 block_t i;
3c62be17
JK		 1142 int err = 0;
1143
1144 bdev = f2fs_target_device(sbi, blkstart, NULL);
1145
1146 for (i = blkstart; i < blkstart + blklen; i++, len++) {
1147 if (i != start) {
1148 struct block_device *bdev2 =
1149 f2fs_target_device(sbi, i, NULL);
1150
1151 if (bdev2 != bdev) {
1152 err = __issue_discard_async(sbi, bdev,
1153 start, len);
1154 if (err)
1155 return err;
1156 bdev = bdev2;
1157 start = i;
1158 len = 0;
1159 }
1160 }
a66cdd98 1161
a66cdd98
JK		 1162 se = get_seg_entry(sbi, GET_SEGNO(sbi, i));
1163 offset = GET_BLKOFF_FROM_SEG0(sbi, i);
1164
1165 if (!f2fs_test_and_set_bit(offset, se->discard_map))
1166 sbi->discard_blks--;
1167 }
f46e8809 1168
3c62be17
JK		 1169 if (len)
1170 err = __issue_discard_async(sbi, bdev, start, len);
1171 return err;
1e87a78d
JK		 1172}
1173
25290fa5
JK		 1174static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc,
1175 bool check_only)
adf4983b 1176{
b2955550
JK		 1177 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
1178 int max_blocks = sbi->blocks_per_seg;
4b2fecc8 1179 struct seg_entry *se = get_seg_entry(sbi, cpc->trim_start);
b2955550
JK		 1180 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1181 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
a66cdd98 1182 unsigned long *discard_map = (unsigned long *)se->discard_map;
60a3b782 1183 unsigned long *dmap = SIT_I(sbi)->tmp_map;
b2955550 1184 unsigned int start = 0, end = -1;
c473f1a9 1185 bool force = (cpc->reason & CP_DISCARD);
a7eeb823 1186 struct discard_entry *de = NULL;
46f84c2c 1187 struct list_head *head = &SM_I(sbi)->dcc_info->entry_list;
b2955550
JK		 1188 int i;
1189
3e025740 1190 if (se->valid_blocks == max_blocks || !f2fs_discard_en(sbi))
25290fa5 1191 return false;
b2955550 1192
a66cdd98
JK		 1193 if (!force) {
1194 if (!test_opt(sbi, DISCARD) || !se->valid_blocks ||
0b54fb84
JK		 1195 SM_I(sbi)->dcc_info->nr_discards >=
1196 SM_I(sbi)->dcc_info->max_discards)
25290fa5 1197 return false;
4b2fecc8
JK		 1198 }
1199
b2955550
JK		 1200 /* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
1201 for (i = 0; i < entries; i++)
a66cdd98 1202 dmap[i] = force ? ~ckpt_map[i] & ~discard_map[i] :
d7bc2484 1203 (cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
b2955550 1204
0b54fb84
JK		 1205 while (force || SM_I(sbi)->dcc_info->nr_discards <=
1206 SM_I(sbi)->dcc_info->max_discards) {
b2955550
JK		 1207 start = __find_rev_next_bit(dmap, max_blocks, end + 1);
1208 if (start >= max_blocks)
1209 break;
1210
1211 end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1);
c7b41e16
YH		 1212 if (force && start && end != max_blocks
1213 && (end - start) < cpc->trim_minlen)
1214 continue;
1215
25290fa5
JK		 1216 if (check_only)
1217 return true;
1218
a7eeb823
CY		 1219 if (!de) {
1220 de = f2fs_kmem_cache_alloc(discard_entry_slab,
1221 GFP_F2FS_ZERO);
1222 de->start_blkaddr = START_BLOCK(sbi, cpc->trim_start);
1223 list_add_tail(&de->list, head);
1224 }
1225
1226 for (i = start; i < end; i++)
1227 __set_bit_le(i, (void *)de->discard_map);
1228
1229 SM_I(sbi)->dcc_info->nr_discards += end - start;
b2955550 1230 }
25290fa5 1231 return false;
b2955550
JK		 1232}
1233
4b2fecc8
JK		 1234void release_discard_addrs(struct f2fs_sb_info *sbi)
1235{
46f84c2c 1236 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
4b2fecc8
JK		 1237 struct discard_entry *entry, *this;
1238
1239 /* drop caches */
1240 list_for_each_entry_safe(entry, this, head, list) {
1241 list_del(&entry->list);
1242 kmem_cache_free(discard_entry_slab, entry);
1243 }
1244}
1245
0a8165d7 1246/*
351df4b2
JK		 1247 * Should call clear_prefree_segments after checkpoint is done.
1248 */
1249static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
1250{
1251 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
b65ee148 1252 unsigned int segno;
351df4b2
JK		 1253
1254 mutex_lock(&dirty_i->seglist_lock);
7cd8558b 1255 for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], MAIN_SEGS(sbi))
351df4b2 1256 __set_test_and_free(sbi, segno);
351df4b2
JK		 1257 mutex_unlock(&dirty_i->seglist_lock);
1258}
1259
836b5a63 1260void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2 1261{
46f84c2c 1262 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
2d7b822a 1263 struct discard_entry *entry, *this;
351df4b2 1264 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
29e59c14 1265 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
29e59c14 1266 unsigned int start = 0, end = -1;
36abef4e 1267 unsigned int secno, start_segno;
c473f1a9 1268 bool force = (cpc->reason & CP_DISCARD);
351df4b2
JK		 1269
1270 mutex_lock(&dirty_i->seglist_lock);
29e59c14 1271
351df4b2 1272 while (1) {
29e59c14 1273 int i;
7cd8558b
JK		 1274 start = find_next_bit(prefree_map, MAIN_SEGS(sbi), end + 1);
1275 if (start >= MAIN_SEGS(sbi))
351df4b2 1276 break;
7cd8558b
JK		 1277 end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
1278 start + 1);
29e59c14
CL		 1279
1280 for (i = start; i < end; i++)
1281 clear_bit(i, prefree_map);
1282
1283 dirty_i->nr_dirty[PRE] -= end - start;
1284
650d3c4e 1285 if (!test_opt(sbi, DISCARD))
29e59c14 1286 continue;
351df4b2 1287
650d3c4e
YH		 1288 if (force && start >= cpc->trim_start &&
1289 (end - 1) <= cpc->trim_end)
1290 continue;
1291
36abef4e
JK		 1292 if (!test_opt(sbi, LFS) || sbi->segs_per_sec == 1) {
1293 f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
37208879 1294 (end - start) << sbi->log_blocks_per_seg);
36abef4e
JK		 1295 continue;
1296 }
1297next:
4ddb1a4d
JK		 1298 secno = GET_SEC_FROM_SEG(sbi, start);
1299 start_segno = GET_SEG_FROM_SEC(sbi, secno);
36abef4e 1300 if (!IS_CURSEC(sbi, secno) &&
302bd348 1301 !get_valid_blocks(sbi, start, true))
36abef4e
JK		 1302 f2fs_issue_discard(sbi, START_BLOCK(sbi, start_segno),
1303 sbi->segs_per_sec << sbi->log_blocks_per_seg);
1304
1305 start = start_segno + sbi->segs_per_sec;
1306 if (start < end)
1307 goto next;
8b107f5b
JK		 1308 else
1309 end = start - 1;
351df4b2
JK		 1310 }
1311 mutex_unlock(&dirty_i->seglist_lock);
b2955550
JK		 1312
1313 /* send small discards */
2d7b822a 1314 list_for_each_entry_safe(entry, this, head, list) {
a7eeb823
CY		 1315 unsigned int cur_pos = 0, next_pos, len, total_len = 0;
1316 bool is_valid = test_bit_le(0, entry->discard_map);
1317
1318find_next:
1319 if (is_valid) {
1320 next_pos = find_next_zero_bit_le(entry->discard_map,
1321 sbi->blocks_per_seg, cur_pos);
1322 len = next_pos - cur_pos;
1323
1324 if (force && len < cpc->trim_minlen)
1325 goto skip;
1326
1327 f2fs_issue_discard(sbi, entry->start_blkaddr + cur_pos,
1328 len);
1329 cpc->trimmed += len;
1330 total_len += len;
1331 } else {
1332 next_pos = find_next_bit_le(entry->discard_map,
1333 sbi->blocks_per_seg, cur_pos);
1334 }
836b5a63 1335skip:
a7eeb823
CY		 1336 cur_pos = next_pos;
1337 is_valid = !is_valid;
1338
1339 if (cur_pos < sbi->blocks_per_seg)
1340 goto find_next;
1341
b2955550 1342 list_del(&entry->list);
a7eeb823 1343 SM_I(sbi)->dcc_info->nr_discards -= total_len;
b2955550
JK		 1344 kmem_cache_free(discard_entry_slab, entry);
1345 }
34e159da
CY		 1346
1347 wake_up(&SM_I(sbi)->dcc_info->discard_wait_queue);
351df4b2
JK		 1348}
1349
8ed59745 1350static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84 1351{
15469963 1352 dev_t dev = sbi->sb->s_bdev->bd_dev;
0b54fb84 1353 struct discard_cmd_control *dcc;
ba48a33e 1354 int err = 0, i;
0b54fb84
JK		 1355
1356 if (SM_I(sbi)->dcc_info) {
1357 dcc = SM_I(sbi)->dcc_info;
1358 goto init_thread;
1359 }
1360
1361 dcc = kzalloc(sizeof(struct discard_cmd_control), GFP_KERNEL);
1362 if (!dcc)
1363 return -ENOMEM;
1364
46f84c2c 1365 INIT_LIST_HEAD(&dcc->entry_list);
ba48a33e
CY		 1366 for (i = 0; i < MAX_PLIST_NUM; i++)
1367 INIT_LIST_HEAD(&dcc->pend_list[i]);
46f84c2c 1368 INIT_LIST_HEAD(&dcc->wait_list);
15469963 1369 mutex_init(&dcc->cmd_lock);
8b8dd65f
CY		 1370 atomic_set(&dcc->issued_discard, 0);
1371 atomic_set(&dcc->issing_discard, 0);
5f32366a 1372 atomic_set(&dcc->discard_cmd_cnt, 0);
0b54fb84 1373 dcc->nr_discards = 0;
d618ebaf 1374 dcc->max_discards = MAIN_SEGS(sbi) << sbi->log_blocks_per_seg;
d84d1cbd 1375 dcc->undiscard_blks = 0;
004b6862 1376 dcc->root = RB_ROOT;
0b54fb84 1377
15469963 1378 init_waitqueue_head(&dcc->discard_wait_queue);
0b54fb84
JK		 1379 SM_I(sbi)->dcc_info = dcc;
1380init_thread:
15469963
JK		 1381 dcc->f2fs_issue_discard = kthread_run(issue_discard_thread, sbi,
1382 "f2fs_discard-%u:%u", MAJOR(dev), MINOR(dev));
1383 if (IS_ERR(dcc->f2fs_issue_discard)) {
1384 err = PTR_ERR(dcc->f2fs_issue_discard);
1385 kfree(dcc);
1386 SM_I(sbi)->dcc_info = NULL;
1387 return err;
1388 }
1389
0b54fb84
JK		 1390 return err;
1391}
1392
f099405f 1393static void destroy_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84
JK		 1394{
1395 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1396
f099405f
CY		 1397 if (!dcc)
1398 return;
1399
1400 if (dcc->f2fs_issue_discard) {
15469963
JK		 1401 struct task_struct *discard_thread = dcc->f2fs_issue_discard;
1402
1403 dcc->f2fs_issue_discard = NULL;
1404 kthread_stop(discard_thread);
1405 }
f099405f
CY		 1406
1407 kfree(dcc);
1408 SM_I(sbi)->dcc_info = NULL;
0b54fb84
JK		 1409}
1410
184a5cd2 1411static bool __mark_sit_entry_dirty(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 1412{
1413 struct sit_info *sit_i = SIT_I(sbi);
184a5cd2
CY		 1414
1415 if (!__test_and_set_bit(segno, sit_i->dirty_sentries_bitmap)) {
351df4b2 1416 sit_i->dirty_sentries++;
184a5cd2
CY		 1417 return false;
1418 }
1419
1420 return true;
351df4b2
JK		 1421}
1422
1423static void __set_sit_entry_type(struct f2fs_sb_info *sbi, int type,
1424 unsigned int segno, int modified)
1425{
1426 struct seg_entry *se = get_seg_entry(sbi, segno);
1427 se->type = type;
1428 if (modified)
1429 __mark_sit_entry_dirty(sbi, segno);
1430}
1431
1432static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
1433{
1434 struct seg_entry *se;
1435 unsigned int segno, offset;
1436 long int new_vblocks;
1437
1438 segno = GET_SEGNO(sbi, blkaddr);
1439
1440 se = get_seg_entry(sbi, segno);
1441 new_vblocks = se->valid_blocks + del;
491c0854 1442 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
351df4b2 1443
9850cf4a 1444 f2fs_bug_on(sbi, (new_vblocks >> (sizeof(unsigned short) << 3) ||
351df4b2
JK		 1445 (new_vblocks > sbi->blocks_per_seg)));
1446
1447 se->valid_blocks = new_vblocks;
1448 se->mtime = get_mtime(sbi);
1449 SIT_I(sbi)->max_mtime = se->mtime;
1450
1451 /* Update valid block bitmap */
1452 if (del > 0) {
355e7891
CY		 1453 if (f2fs_test_and_set_bit(offset, se->cur_valid_map)) {
1454#ifdef CONFIG_F2FS_CHECK_FS
1455 if (f2fs_test_and_set_bit(offset,
1456 se->cur_valid_map_mir))
1457 f2fs_bug_on(sbi, 1);
1458 else
1459 WARN_ON(1);
1460#else
05796763 1461 f2fs_bug_on(sbi, 1);
355e7891
CY		 1462#endif
1463 }
3e025740
JK		 1464 if (f2fs_discard_en(sbi) &&
1465 !f2fs_test_and_set_bit(offset, se->discard_map))
a66cdd98 1466 sbi->discard_blks--;
720037f9
JK		 1467
1468 /* don't overwrite by SSR to keep node chain */
1469 if (se->type == CURSEG_WARM_NODE) {
1470 if (!f2fs_test_and_set_bit(offset, se->ckpt_valid_map))
1471 se->ckpt_valid_blocks++;
1472 }
351df4b2 1473 } else {
355e7891
CY		 1474 if (!f2fs_test_and_clear_bit(offset, se->cur_valid_map)) {
1475#ifdef CONFIG_F2FS_CHECK_FS
1476 if (!f2fs_test_and_clear_bit(offset,
1477 se->cur_valid_map_mir))
1478 f2fs_bug_on(sbi, 1);
1479 else
1480 WARN_ON(1);
1481#else
05796763 1482 f2fs_bug_on(sbi, 1);
355e7891
CY		 1483#endif
1484 }
3e025740
JK		 1485 if (f2fs_discard_en(sbi) &&
1486 f2fs_test_and_clear_bit(offset, se->discard_map))
a66cdd98 1487 sbi->discard_blks++;
351df4b2
JK		 1488 }
1489 if (!f2fs_test_bit(offset, se->ckpt_valid_map))
1490 se->ckpt_valid_blocks += del;
1491
1492 __mark_sit_entry_dirty(sbi, segno);
1493
1494 /* update total number of valid blocks to be written in ckpt area */
1495 SIT_I(sbi)->written_valid_blocks += del;
1496
1497 if (sbi->segs_per_sec > 1)
1498 get_sec_entry(sbi, segno)->valid_blocks += del;
1499}
1500
5e443818 1501void refresh_sit_entry(struct f2fs_sb_info *sbi, block_t old, block_t new)
351df4b2 1502{
5e443818
JK		 1503 update_sit_entry(sbi, new, 1);
1504 if (GET_SEGNO(sbi, old) != NULL_SEGNO)
1505 update_sit_entry(sbi, old, -1);
1506
1507 locate_dirty_segment(sbi, GET_SEGNO(sbi, old));
1508 locate_dirty_segment(sbi, GET_SEGNO(sbi, new));
351df4b2
JK		 1509}
1510
1511void invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
1512{
1513 unsigned int segno = GET_SEGNO(sbi, addr);
1514 struct sit_info *sit_i = SIT_I(sbi);
1515
9850cf4a 1516 f2fs_bug_on(sbi, addr == NULL_ADDR);
351df4b2
JK		 1517 if (addr == NEW_ADDR)
1518 return;
1519
1520 /* add it into sit main buffer */
1521 mutex_lock(&sit_i->sentry_lock);
1522
1523 update_sit_entry(sbi, addr, -1);
1524
1525 /* add it into dirty seglist */
1526 locate_dirty_segment(sbi, segno);
1527
1528 mutex_unlock(&sit_i->sentry_lock);
1529}
1530
6e2c64ad
JK		 1531bool is_checkpointed_data(struct f2fs_sb_info *sbi, block_t blkaddr)
1532{
1533 struct sit_info *sit_i = SIT_I(sbi);
1534 unsigned int segno, offset;
1535 struct seg_entry *se;
1536 bool is_cp = false;
1537
1538 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
1539 return true;
1540
1541 mutex_lock(&sit_i->sentry_lock);
1542
1543 segno = GET_SEGNO(sbi, blkaddr);
1544 se = get_seg_entry(sbi, segno);
1545 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
1546
1547 if (f2fs_test_bit(offset, se->ckpt_valid_map))
1548 is_cp = true;
1549
1550 mutex_unlock(&sit_i->sentry_lock);
1551
1552 return is_cp;
1553}
1554
0a8165d7 1555/*
351df4b2
JK		 1556 * This function should be resided under the curseg_mutex lock
1557 */
1558static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
e79efe3b 1559 struct f2fs_summary *sum)
351df4b2
JK		 1560{
1561 struct curseg_info *curseg = CURSEG_I(sbi, type);
1562 void *addr = curseg->sum_blk;
e79efe3b 1563 addr += curseg->next_blkoff * sizeof(struct f2fs_summary);
351df4b2 1564 memcpy(addr, sum, sizeof(struct f2fs_summary));
351df4b2
JK		 1565}
1566
0a8165d7 1567/*
351df4b2
JK		 1568 * Calculate the number of current summary pages for writing
1569 */
3fa06d7b 1570int npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
351df4b2 1571{
351df4b2 1572 int valid_sum_count = 0;
9a47938b 1573 int i, sum_in_page;
351df4b2
JK		 1574
1575 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1576 if (sbi->ckpt->alloc_type[i] == SSR)
1577 valid_sum_count += sbi->blocks_per_seg;
3fa06d7b
CY		 1578 else {
1579 if (for_ra)
1580 valid_sum_count += le16_to_cpu(
1581 F2FS_CKPT(sbi)->cur_data_blkoff[i]);
1582 else
1583 valid_sum_count += curseg_blkoff(sbi, i);
1584 }
351df4b2
JK		 1585 }
1586
09cbfeaf 1587 sum_in_page = (PAGE_SIZE - 2 * SUM_JOURNAL_SIZE -
9a47938b
FL		 1588 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
1589 if (valid_sum_count <= sum_in_page)
351df4b2 1590 return 1;
9a47938b 1591 else if ((valid_sum_count - sum_in_page) <=
09cbfeaf 1592 (PAGE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
351df4b2
JK		 1593 return 2;
1594 return 3;
1595}
1596
0a8165d7 1597/*
351df4b2
JK		 1598 * Caller should put this summary page
1599 */
1600struct page *get_sum_page(struct f2fs_sb_info *sbi, unsigned int segno)
1601{
1602 return get_meta_page(sbi, GET_SUM_BLOCK(sbi, segno));
1603}
1604
381722d2 1605void update_meta_page(struct f2fs_sb_info *sbi, void *src, block_t blk_addr)
351df4b2
JK		 1606{
1607 struct page *page = grab_meta_page(sbi, blk_addr);
381722d2
CY		 1608 void *dst = page_address(page);
1609
1610 if (src)
09cbfeaf 1611 memcpy(dst, src, PAGE_SIZE);
381722d2 1612 else
09cbfeaf 1613 memset(dst, 0, PAGE_SIZE);
351df4b2
JK		 1614 set_page_dirty(page);
1615 f2fs_put_page(page, 1);
1616}
1617
381722d2
CY		 1618static void write_sum_page(struct f2fs_sb_info *sbi,
1619 struct f2fs_summary_block *sum_blk, block_t blk_addr)
1620{
1621 update_meta_page(sbi, (void *)sum_blk, blk_addr);
1622}
1623
b7ad7512
CY		 1624static void write_current_sum_page(struct f2fs_sb_info *sbi,
1625 int type, block_t blk_addr)
1626{
1627 struct curseg_info *curseg = CURSEG_I(sbi, type);
1628 struct page *page = grab_meta_page(sbi, blk_addr);
1629 struct f2fs_summary_block *src = curseg->sum_blk;
1630 struct f2fs_summary_block *dst;
1631
1632 dst = (struct f2fs_summary_block *)page_address(page);
1633
1634 mutex_lock(&curseg->curseg_mutex);
1635
1636 down_read(&curseg->journal_rwsem);
1637 memcpy(&dst->journal, curseg->journal, SUM_JOURNAL_SIZE);
1638 up_read(&curseg->journal_rwsem);
1639
1640 memcpy(dst->entries, src->entries, SUM_ENTRY_SIZE);
1641 memcpy(&dst->footer, &src->footer, SUM_FOOTER_SIZE);
1642
1643 mutex_unlock(&curseg->curseg_mutex);
1644
1645 set_page_dirty(page);
1646 f2fs_put_page(page, 1);
1647}
1648
a7881893
JK		 1649static int is_next_segment_free(struct f2fs_sb_info *sbi, int type)
1650{
1651 struct curseg_info *curseg = CURSEG_I(sbi, type);
1652 unsigned int segno = curseg->segno + 1;
1653 struct free_segmap_info *free_i = FREE_I(sbi);
1654
1655 if (segno < MAIN_SEGS(sbi) && segno % sbi->segs_per_sec)
1656 return !test_bit(segno, free_i->free_segmap);
1657 return 0;
1658}
1659
0a8165d7 1660/*
351df4b2
JK		 1661 * Find a new segment from the free segments bitmap to right order
1662 * This function should be returned with success, otherwise BUG
1663 */
1664static void get_new_segment(struct f2fs_sb_info *sbi,
1665 unsigned int *newseg, bool new_sec, int dir)
1666{
1667 struct free_segmap_info *free_i = FREE_I(sbi);
351df4b2 1668 unsigned int segno, secno, zoneno;
7cd8558b 1669 unsigned int total_zones = MAIN_SECS(sbi) / sbi->secs_per_zone;
4ddb1a4d
JK		 1670 unsigned int hint = GET_SEC_FROM_SEG(sbi, *newseg);
1671 unsigned int old_zoneno = GET_ZONE_FROM_SEG(sbi, *newseg);
351df4b2
JK		 1672 unsigned int left_start = hint;
1673 bool init = true;
1674 int go_left = 0;
1675 int i;
1676
1a118ccf 1677 spin_lock(&free_i->segmap_lock);
351df4b2
JK		 1678
1679 if (!new_sec && ((*newseg + 1) % sbi->segs_per_sec)) {
1680 segno = find_next_zero_bit(free_i->free_segmap,
4ddb1a4d
JK		 1681 GET_SEG_FROM_SEC(sbi, hint + 1), *newseg + 1);
1682 if (segno < GET_SEG_FROM_SEC(sbi, hint + 1))
351df4b2
JK		 1683 goto got_it;
1684 }
1685find_other_zone:
7cd8558b
JK		 1686 secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
1687 if (secno >= MAIN_SECS(sbi)) {
351df4b2
JK		 1688 if (dir == ALLOC_RIGHT) {
1689 secno = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1690 MAIN_SECS(sbi), 0);
1691 f2fs_bug_on(sbi, secno >= MAIN_SECS(sbi));
351df4b2
JK		 1692 } else {
1693 go_left = 1;
1694 left_start = hint - 1;
1695 }
1696 }
1697 if (go_left == 0)
1698 goto skip_left;
1699
1700 while (test_bit(left_start, free_i->free_secmap)) {
1701 if (left_start > 0) {
1702 left_start--;
1703 continue;
1704 }
1705 left_start = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1706 MAIN_SECS(sbi), 0);
1707 f2fs_bug_on(sbi, left_start >= MAIN_SECS(sbi));
351df4b2
JK		 1708 break;
1709 }
1710 secno = left_start;
1711skip_left:
1712 hint = secno;
4ddb1a4d
JK		 1713 segno = GET_SEG_FROM_SEC(sbi, secno);
1714 zoneno = GET_ZONE_FROM_SEC(sbi, secno);
351df4b2
JK		 1715
1716 /* give up on finding another zone */
1717 if (!init)
1718 goto got_it;
1719 if (sbi->secs_per_zone == 1)
1720 goto got_it;
1721 if (zoneno == old_zoneno)
1722 goto got_it;
1723 if (dir == ALLOC_LEFT) {
1724 if (!go_left && zoneno + 1 >= total_zones)
1725 goto got_it;
1726 if (go_left && zoneno == 0)
1727 goto got_it;
1728 }
1729 for (i = 0; i < NR_CURSEG_TYPE; i++)
1730 if (CURSEG_I(sbi, i)->zone == zoneno)
1731 break;
1732
1733 if (i < NR_CURSEG_TYPE) {
1734 /* zone is in user, try another */
1735 if (go_left)
1736 hint = zoneno * sbi->secs_per_zone - 1;
1737 else if (zoneno + 1 >= total_zones)
1738 hint = 0;
1739 else
1740 hint = (zoneno + 1) * sbi->secs_per_zone;
1741 init = false;
1742 goto find_other_zone;
1743 }
1744got_it:
1745 /* set it as dirty segment in free segmap */
9850cf4a 1746 f2fs_bug_on(sbi, test_bit(segno, free_i->free_segmap));
351df4b2
JK		 1747 __set_inuse(sbi, segno);
1748 *newseg = segno;
1a118ccf 1749 spin_unlock(&free_i->segmap_lock);
351df4b2
JK		 1750}
1751
1752static void reset_curseg(struct f2fs_sb_info *sbi, int type, int modified)
1753{
1754 struct curseg_info *curseg = CURSEG_I(sbi, type);
1755 struct summary_footer *sum_footer;
1756
1757 curseg->segno = curseg->next_segno;
4ddb1a4d 1758 curseg->zone = GET_ZONE_FROM_SEG(sbi, curseg->segno);
351df4b2
JK		 1759 curseg->next_blkoff = 0;
1760 curseg->next_segno = NULL_SEGNO;
1761
1762 sum_footer = &(curseg->sum_blk->footer);
1763 memset(sum_footer, 0, sizeof(struct summary_footer));
1764 if (IS_DATASEG(type))
1765 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
1766 if (IS_NODESEG(type))
1767 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
1768 __set_sit_entry_type(sbi, type, curseg->segno, modified);
1769}
1770
7a20b8a6
JK		 1771static unsigned int __get_next_segno(struct f2fs_sb_info *sbi, int type)
1772{
a7881893
JK		 1773 /* if segs_per_sec is large than 1, we need to keep original policy. */
1774 if (sbi->segs_per_sec != 1)
1775 return CURSEG_I(sbi, type)->segno;
1776
7a20b8a6
JK		 1777 if (type == CURSEG_HOT_DATA || IS_NODESEG(type))
1778 return 0;
1779
e066b83c
JK		 1780 if (SIT_I(sbi)->last_victim[ALLOC_NEXT])
1781 return SIT_I(sbi)->last_victim[ALLOC_NEXT];
7a20b8a6
JK		 1782 return CURSEG_I(sbi, type)->segno;
1783}
1784
0a8165d7 1785/*
351df4b2
JK		 1786 * Allocate a current working segment.
1787 * This function always allocates a free segment in LFS manner.
1788 */
1789static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
1790{
1791 struct curseg_info *curseg = CURSEG_I(sbi, type);
1792 unsigned int segno = curseg->segno;
1793 int dir = ALLOC_LEFT;
1794
1795 write_sum_page(sbi, curseg->sum_blk,
81fb5e87 1796 GET_SUM_BLOCK(sbi, segno));
351df4b2
JK		 1797 if (type == CURSEG_WARM_DATA || type == CURSEG_COLD_DATA)
1798 dir = ALLOC_RIGHT;
1799
1800 if (test_opt(sbi, NOHEAP))
1801 dir = ALLOC_RIGHT;
1802
7a20b8a6 1803 segno = __get_next_segno(sbi, type);
351df4b2
JK		 1804 get_new_segment(sbi, &segno, new_sec, dir);
1805 curseg->next_segno = segno;
1806 reset_curseg(sbi, type, 1);
1807 curseg->alloc_type = LFS;
1808}
1809
1810static void __next_free_blkoff(struct f2fs_sb_info *sbi,
1811 struct curseg_info *seg, block_t start)
1812{
1813 struct seg_entry *se = get_seg_entry(sbi, seg->segno);
e81c93cf 1814 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
60a3b782 1815 unsigned long *target_map = SIT_I(sbi)->tmp_map;
e81c93cf
CL		 1816 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
1817 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1818 int i, pos;
1819
1820 for (i = 0; i < entries; i++)
1821 target_map[i] = ckpt_map[i] | cur_map[i];
1822
1823 pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, start);
1824
1825 seg->next_blkoff = pos;
351df4b2
JK		 1826}
1827
0a8165d7 1828/*
351df4b2
JK		 1829 * If a segment is written by LFS manner, next block offset is just obtained
1830 * by increasing the current block offset. However, if a segment is written by
1831 * SSR manner, next block offset obtained by calling __next_free_blkoff
1832 */
1833static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
1834 struct curseg_info *seg)
1835{
1836 if (seg->alloc_type == SSR)
1837 __next_free_blkoff(sbi, seg, seg->next_blkoff + 1);
1838 else
1839 seg->next_blkoff++;
1840}
1841
0a8165d7 1842/*
e1c42045 1843 * This function always allocates a used segment(from dirty seglist) by SSR
351df4b2
JK		 1844 * manner, so it should recover the existing segment information of valid blocks
1845 */
1846static void change_curseg(struct f2fs_sb_info *sbi, int type, bool reuse)
1847{
1848 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
1849 struct curseg_info *curseg = CURSEG_I(sbi, type);
1850 unsigned int new_segno = curseg->next_segno;
1851 struct f2fs_summary_block *sum_node;
1852 struct page *sum_page;
1853
1854 write_sum_page(sbi, curseg->sum_blk,
1855 GET_SUM_BLOCK(sbi, curseg->segno));
1856 __set_test_and_inuse(sbi, new_segno);
1857
1858 mutex_lock(&dirty_i->seglist_lock);
1859 __remove_dirty_segment(sbi, new_segno, PRE);
1860 __remove_dirty_segment(sbi, new_segno, DIRTY);
1861 mutex_unlock(&dirty_i->seglist_lock);
1862
1863 reset_curseg(sbi, type, 1);
1864 curseg->alloc_type = SSR;
1865 __next_free_blkoff(sbi, curseg, 0);
1866
1867 if (reuse) {
1868 sum_page = get_sum_page(sbi, new_segno);
1869 sum_node = (struct f2fs_summary_block *)page_address(sum_page);
1870 memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
1871 f2fs_put_page(sum_page, 1);
1872 }
1873}
1874
43727527
JK		 1875static int get_ssr_segment(struct f2fs_sb_info *sbi, int type)
1876{
1877 struct curseg_info *curseg = CURSEG_I(sbi, type);
1878 const struct victim_selection *v_ops = DIRTY_I(sbi)->v_ops;
e066b83c 1879 unsigned segno = NULL_SEGNO;
d27c3d89
CY		 1880 int i, cnt;
1881 bool reversed = false;
c192f7a4
JK		 1882
1883 /* need_SSR() already forces to do this */
e066b83c
JK		 1884 if (v_ops->get_victim(sbi, &segno, BG_GC, type, SSR)) {
1885 curseg->next_segno = segno;
c192f7a4 1886 return 1;
e066b83c 1887 }
43727527 1888
70d625cb
JK		 1889 /* For node segments, let's do SSR more intensively */
1890 if (IS_NODESEG(type)) {
d27c3d89
CY		 1891 if (type >= CURSEG_WARM_NODE) {
1892 reversed = true;
1893 i = CURSEG_COLD_NODE;
1894 } else {
1895 i = CURSEG_HOT_NODE;
1896 }
1897 cnt = NR_CURSEG_NODE_TYPE;
70d625cb 1898 } else {
d27c3d89
CY		 1899 if (type >= CURSEG_WARM_DATA) {
1900 reversed = true;
1901 i = CURSEG_COLD_DATA;
1902 } else {
1903 i = CURSEG_HOT_DATA;
1904 }
1905 cnt = NR_CURSEG_DATA_TYPE;
70d625cb 1906 }
43727527 1907
d27c3d89 1908 for (; cnt-- > 0; reversed ? i-- : i++) {
c192f7a4
JK		 1909 if (i == type)
1910 continue;
e066b83c
JK		 1911 if (v_ops->get_victim(sbi, &segno, BG_GC, i, SSR)) {
1912 curseg->next_segno = segno;
43727527 1913 return 1;
e066b83c 1914 }
c192f7a4 1915 }
43727527
JK		 1916 return 0;
1917}
1918
351df4b2
JK		 1919/*
1920 * flush out current segment and replace it with new segment
1921 * This function should be returned with success, otherwise BUG
1922 */
1923static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
1924 int type, bool force)
1925{
a7881893
JK		 1926 struct curseg_info *curseg = CURSEG_I(sbi, type);
1927
7b405275 1928 if (force)
351df4b2 1929 new_curseg(sbi, type, true);
5b6c6be2
JK		 1930 else if (!is_set_ckpt_flags(sbi, CP_CRC_RECOVERY_FLAG) &&
1931 type == CURSEG_WARM_NODE)
351df4b2 1932 new_curseg(sbi, type, false);
a7881893
JK		 1933 else if (curseg->alloc_type == LFS && is_next_segment_free(sbi, type))
1934 new_curseg(sbi, type, false);
351df4b2
JK		 1935 else if (need_SSR(sbi) && get_ssr_segment(sbi, type))
1936 change_curseg(sbi, type, true);
1937 else
1938 new_curseg(sbi, type, false);
dcdfff65 1939
a7881893 1940 stat_inc_seg_type(sbi, curseg);
351df4b2
JK		 1941}
1942
1943void allocate_new_segments(struct f2fs_sb_info *sbi)
1944{
6ae1be13
JK		 1945 struct curseg_info *curseg;
1946 unsigned int old_segno;
351df4b2
JK		 1947 int i;
1948
6ae1be13
JK		 1949 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1950 curseg = CURSEG_I(sbi, i);
1951 old_segno = curseg->segno;
1952 SIT_I(sbi)->s_ops->allocate_segment(sbi, i, true);
1953 locate_dirty_segment(sbi, old_segno);
1954 }
351df4b2
JK		 1955}
1956
1957static const struct segment_allocation default_salloc_ops = {
1958 .allocate_segment = allocate_segment_by_default,
1959};
1960
25290fa5
JK		 1961bool exist_trim_candidates(struct f2fs_sb_info *sbi, struct cp_control *cpc)
1962{
1963 __u64 trim_start = cpc->trim_start;
1964 bool has_candidate = false;
1965
1966 mutex_lock(&SIT_I(sbi)->sentry_lock);
1967 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++) {
1968 if (add_discard_addrs(sbi, cpc, true)) {
1969 has_candidate = true;
1970 break;
1971 }
1972 }
1973 mutex_unlock(&SIT_I(sbi)->sentry_lock);
1974
1975 cpc->trim_start = trim_start;
1976 return has_candidate;
1977}
1978
4b2fecc8
JK		 1979int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
1980{
f7ef9b83
JK		 1981 __u64 start = F2FS_BYTES_TO_BLK(range->start);
1982 __u64 end = start + F2FS_BYTES_TO_BLK(range->len) - 1;
4b2fecc8
JK		 1983 unsigned int start_segno, end_segno;
1984 struct cp_control cpc;
c34f42e2 1985 int err = 0;
4b2fecc8 1986
836b5a63 1987 if (start >= MAX_BLKADDR(sbi) || range->len < sbi->blocksize)
4b2fecc8
JK		 1988 return -EINVAL;
1989
9bd27ae4 1990 cpc.trimmed = 0;
7cd8558b 1991 if (end <= MAIN_BLKADDR(sbi))
4b2fecc8
JK		 1992 goto out;
1993
ed214a11
YH		 1994 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
1995 f2fs_msg(sbi->sb, KERN_WARNING,
1996 "Found FS corruption, run fsck to fix.");
1997 goto out;
1998 }
1999
4b2fecc8 2000 /* start/end segment number in main_area */
7cd8558b
JK		 2001 start_segno = (start <= MAIN_BLKADDR(sbi)) ? 0 : GET_SEGNO(sbi, start);
2002 end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
2003 GET_SEGNO(sbi, end);
4b2fecc8 2004 cpc.reason = CP_DISCARD;
836b5a63 2005 cpc.trim_minlen = max_t(__u64, 1, F2FS_BYTES_TO_BLK(range->minlen));
4b2fecc8
JK		 2006
2007 /* do checkpoint to issue discard commands safely */
bba681cb
JK		 2008 for (; start_segno <= end_segno; start_segno = cpc.trim_end + 1) {
2009 cpc.trim_start = start_segno;
a66cdd98
JK		 2010
2011 if (sbi->discard_blks == 0)
2012 break;
2013 else if (sbi->discard_blks < BATCHED_TRIM_BLOCKS(sbi))
2014 cpc.trim_end = end_segno;
2015 else
2016 cpc.trim_end = min_t(unsigned int,
2017 rounddown(start_segno +
bba681cb
JK		 2018 BATCHED_TRIM_SEGMENTS(sbi),
2019 sbi->segs_per_sec) - 1, end_segno);
2020
2021 mutex_lock(&sbi->gc_mutex);
c34f42e2 2022 err = write_checkpoint(sbi, &cpc);
bba681cb 2023 mutex_unlock(&sbi->gc_mutex);
e9328353
CY		 2024 if (err)
2025 break;
74fa5f3d
CY		 2026
2027 schedule();
bba681cb 2028 }
4b2fecc8 2029out:
f7ef9b83 2030 range->len = F2FS_BLK_TO_BYTES(cpc.trimmed);
c34f42e2 2031 return err;
4b2fecc8
JK		 2032}
2033
351df4b2
JK		 2034static bool __has_curseg_space(struct f2fs_sb_info *sbi, int type)
2035{
2036 struct curseg_info *curseg = CURSEG_I(sbi, type);
2037 if (curseg->next_blkoff < sbi->blocks_per_seg)
2038 return true;
2039 return false;
2040}
2041
81377bd6 2042static int __get_segment_type_2(struct f2fs_io_info *fio)
351df4b2 2043{
81377bd6 2044 if (fio->type == DATA)
351df4b2
JK		 2045 return CURSEG_HOT_DATA;
2046 else
2047 return CURSEG_HOT_NODE;
2048}
2049
81377bd6 2050static int __get_segment_type_4(struct f2fs_io_info *fio)
351df4b2 2051{
81377bd6
JK		 2052 if (fio->type == DATA) {
2053 struct inode *inode = fio->page->mapping->host;
351df4b2
JK		 2054
2055 if (S_ISDIR(inode->i_mode))
2056 return CURSEG_HOT_DATA;
2057 else
2058 return CURSEG_COLD_DATA;
2059 } else {
81377bd6 2060 if (IS_DNODE(fio->page) && is_cold_node(fio->page))
a344b9fd 2061 return CURSEG_WARM_NODE;
351df4b2
JK		 2062 else
2063 return CURSEG_COLD_NODE;
2064 }
2065}
2066
81377bd6 2067static int __get_segment_type_6(struct f2fs_io_info *fio)
351df4b2 2068{
81377bd6
JK		 2069 if (fio->type == DATA) {
2070 struct inode *inode = fio->page->mapping->host;
351df4b2 2071
81377bd6 2072 if (is_cold_data(fio->page) || file_is_cold(inode))
351df4b2 2073 return CURSEG_COLD_DATA;
ef095d19
JK		 2074 if (is_inode_flag_set(inode, FI_HOT_DATA))
2075 return CURSEG_HOT_DATA;
2076 return CURSEG_WARM_DATA;
351df4b2 2077 } else {
81377bd6
JK		 2078 if (IS_DNODE(fio->page))
2079 return is_cold_node(fio->page) ? CURSEG_WARM_NODE :
351df4b2 2080 CURSEG_HOT_NODE;
ef095d19 2081 return CURSEG_COLD_NODE;
351df4b2
JK		 2082 }
2083}
2084
81377bd6 2085static int __get_segment_type(struct f2fs_io_info *fio)
351df4b2 2086{
a912b54d
JK		 2087 int type = 0;
2088
81377bd6 2089 switch (fio->sbi->active_logs) {
351df4b2 2090 case 2:
a912b54d
JK		 2091 type = __get_segment_type_2(fio);
2092 break;
351df4b2 2093 case 4:
a912b54d
JK		 2094 type = __get_segment_type_4(fio);
2095 break;
2096 case 6:
2097 type = __get_segment_type_6(fio);
2098 break;
2099 default:
2100 f2fs_bug_on(fio->sbi, true);
351df4b2 2101 }
81377bd6 2102
a912b54d
JK		 2103 if (IS_HOT(type))
2104 fio->temp = HOT;
2105 else if (IS_WARM(type))
2106 fio->temp = WARM;
2107 else
2108 fio->temp = COLD;
2109 return type;
351df4b2
JK		 2110}
2111
bfad7c2d
JK		 2112void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
2113 block_t old_blkaddr, block_t *new_blkaddr,
2114 struct f2fs_summary *sum, int type)
351df4b2
JK		 2115{
2116 struct sit_info *sit_i = SIT_I(sbi);
6ae1be13 2117 struct curseg_info *curseg = CURSEG_I(sbi, type);
351df4b2
JK		 2118
2119 mutex_lock(&curseg->curseg_mutex);
21cb1d99 2120 mutex_lock(&sit_i->sentry_lock);
351df4b2
JK		 2121
2122 *new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
351df4b2 2123
4e6a8d9b
JK		 2124 f2fs_wait_discard_bio(sbi, *new_blkaddr);
2125
351df4b2
JK		 2126 /*
2127 * __add_sum_entry should be resided under the curseg_mutex
2128 * because, this function updates a summary entry in the
2129 * current summary block.
2130 */
e79efe3b 2131 __add_sum_entry(sbi, type, sum);
351df4b2 2132
351df4b2 2133 __refresh_next_blkoff(sbi, curseg);
dcdfff65
JK		 2134
2135 stat_inc_block_count(sbi, curseg);
351df4b2 2136
c6f82fe9
JK		 2137 if (!__has_curseg_space(sbi, type))
2138 sit_i->s_ops->allocate_segment(sbi, type, false);
351df4b2 2139 /*
c6f82fe9
JK		 2140 * SIT information should be updated after segment allocation,
2141 * since we need to keep dirty segments precisely under SSR.
351df4b2
JK		 2142 */
2143 refresh_sit_entry(sbi, old_blkaddr, *new_blkaddr);
5e443818 2144
351df4b2
JK		 2145 mutex_unlock(&sit_i->sentry_lock);
2146
bfad7c2d 2147 if (page && IS_NODESEG(type))
351df4b2
JK		 2148 fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
2149
bfad7c2d
JK		 2150 mutex_unlock(&curseg->curseg_mutex);
2151}
2152
05ca3632 2153static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
bfad7c2d 2154{
81377bd6 2155 int type = __get_segment_type(fio);
0a595eba 2156 int err;
bfad7c2d 2157
7dfeaa32
JK		 2158 if (fio->type == NODE || fio->type == DATA)
2159 mutex_lock(&fio->sbi->wio_mutex[fio->type]);
0a595eba 2160reallocate:
7a9d7548
CY		 2161 allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
2162 &fio->new_blkaddr, sum, type);
bfad7c2d 2163
351df4b2 2164 /* writeout dirty page into bdev */
b9109b0e 2165 err = f2fs_submit_page_write(fio);
0a595eba
JK		 2166 if (err == -EAGAIN) {
2167 fio->old_blkaddr = fio->new_blkaddr;
2168 goto reallocate;
2169 }
7dfeaa32
JK		 2170
2171 if (fio->type == NODE || fio->type == DATA)
2172 mutex_unlock(&fio->sbi->wio_mutex[fio->type]);
351df4b2
JK		 2173}
2174
577e3495 2175void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
351df4b2 2176{
458e6197 2177 struct f2fs_io_info fio = {
05ca3632 2178 .sbi = sbi,
458e6197 2179 .type = META,
04d328de 2180 .op = REQ_OP_WRITE,
70fd7614 2181 .op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
7a9d7548
CY		 2182 .old_blkaddr = page->index,
2183 .new_blkaddr = page->index,
05ca3632 2184 .page = page,
4375a336 2185 .encrypted_page = NULL,
458e6197
JK		 2186 };
2187
2b947003 2188 if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
04d328de 2189 fio.op_flags &= ~REQ_META;
2b947003 2190
351df4b2 2191 set_page_writeback(page);
b9109b0e 2192 f2fs_submit_page_write(&fio);
351df4b2
JK		 2193}
2194
05ca3632 2195void write_node_page(unsigned int nid, struct f2fs_io_info *fio)
351df4b2
JK		 2196{
2197 struct f2fs_summary sum;
05ca3632 2198
351df4b2 2199 set_summary(&sum, nid, 0, 0);
05ca3632 2200 do_write_page(&sum, fio);
351df4b2
JK		 2201}
2202
05ca3632 2203void write_data_page(struct dnode_of_data *dn, struct f2fs_io_info *fio)
351df4b2 2204{
05ca3632 2205 struct f2fs_sb_info *sbi = fio->sbi;
351df4b2
JK		 2206 struct f2fs_summary sum;
2207 struct node_info ni;
2208
9850cf4a 2209 f2fs_bug_on(sbi, dn->data_blkaddr == NULL_ADDR);
351df4b2
JK		 2210 get_node_info(sbi, dn->nid, &ni);
2211 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
05ca3632 2212 do_write_page(&sum, fio);
f28b3434 2213 f2fs_update_data_blkaddr(dn, fio->new_blkaddr);
351df4b2
JK		 2214}
2215
d1b3e72d 2216int rewrite_data_page(struct f2fs_io_info *fio)
351df4b2 2217{
7a9d7548 2218 fio->new_blkaddr = fio->old_blkaddr;
05ca3632 2219 stat_inc_inplace_blocks(fio->sbi);
d1b3e72d 2220 return f2fs_submit_page_bio(fio);
351df4b2
JK		 2221}
2222
4356e48e 2223void __f2fs_replace_block(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
19f106bc 2224 block_t old_blkaddr, block_t new_blkaddr,
28bc106b 2225 bool recover_curseg, bool recover_newaddr)
351df4b2
JK		 2226{
2227 struct sit_info *sit_i = SIT_I(sbi);
2228 struct curseg_info *curseg;
2229 unsigned int segno, old_cursegno;
2230 struct seg_entry *se;
2231 int type;
19f106bc 2232 unsigned short old_blkoff;
351df4b2
JK		 2233
2234 segno = GET_SEGNO(sbi, new_blkaddr);
2235 se = get_seg_entry(sbi, segno);
2236 type = se->type;
2237
19f106bc
CY		 2238 if (!recover_curseg) {
2239 /* for recovery flow */
2240 if (se->valid_blocks == 0 && !IS_CURSEG(sbi, segno)) {
2241 if (old_blkaddr == NULL_ADDR)
2242 type = CURSEG_COLD_DATA;
2243 else
2244 type = CURSEG_WARM_DATA;
2245 }
2246 } else {
2247 if (!IS_CURSEG(sbi, segno))
351df4b2
JK		 2248 type = CURSEG_WARM_DATA;
2249 }
19f106bc 2250
351df4b2
JK		 2251 curseg = CURSEG_I(sbi, type);
2252
2253 mutex_lock(&curseg->curseg_mutex);
2254 mutex_lock(&sit_i->sentry_lock);
2255
2256 old_cursegno = curseg->segno;
19f106bc 2257 old_blkoff = curseg->next_blkoff;
351df4b2
JK		 2258
2259 /* change the current segment */
2260 if (segno != curseg->segno) {
2261 curseg->next_segno = segno;
2262 change_curseg(sbi, type, true);
2263 }
2264
491c0854 2265 curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
e79efe3b 2266 __add_sum_entry(sbi, type, sum);
351df4b2 2267
28bc106b 2268 if (!recover_curseg || recover_newaddr)
6e2c64ad
JK		 2269 update_sit_entry(sbi, new_blkaddr, 1);
2270 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
2271 update_sit_entry(sbi, old_blkaddr, -1);
2272
2273 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
2274 locate_dirty_segment(sbi, GET_SEGNO(sbi, new_blkaddr));
2275
351df4b2 2276 locate_dirty_segment(sbi, old_cursegno);
351df4b2 2277
19f106bc
CY		 2278 if (recover_curseg) {
2279 if (old_cursegno != curseg->segno) {
2280 curseg->next_segno = old_cursegno;
2281 change_curseg(sbi, type, true);
2282 }
2283 curseg->next_blkoff = old_blkoff;
2284 }
2285
351df4b2
JK		 2286 mutex_unlock(&sit_i->sentry_lock);
2287 mutex_unlock(&curseg->curseg_mutex);
2288}
2289
528e3459
CY		 2290void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
2291 block_t old_addr, block_t new_addr,
28bc106b
CY		 2292 unsigned char version, bool recover_curseg,
2293 bool recover_newaddr)
528e3459
CY		 2294{
2295 struct f2fs_summary sum;
2296
2297 set_summary(&sum, dn->nid, dn->ofs_in_node, version);
2298
28bc106b
CY		 2299 __f2fs_replace_block(sbi, &sum, old_addr, new_addr,
2300 recover_curseg, recover_newaddr);
528e3459 2301
f28b3434 2302 f2fs_update_data_blkaddr(dn, new_addr);
528e3459
CY		 2303}
2304
93dfe2ac 2305void f2fs_wait_on_page_writeback(struct page *page,
fec1d657 2306 enum page_type type, bool ordered)
93dfe2ac 2307{
93dfe2ac 2308 if (PageWriteback(page)) {
4081363f
JK		 2309 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
2310
b9109b0e
JK		 2311 f2fs_submit_merged_write_cond(sbi, page->mapping->host,
2312 0, page->index, type);
fec1d657
JK		 2313 if (ordered)
2314 wait_on_page_writeback(page);
2315 else
2316 wait_for_stable_page(page);
93dfe2ac
JK		 2317 }
2318}
2319
08b39fbd
CY		 2320void f2fs_wait_on_encrypted_page_writeback(struct f2fs_sb_info *sbi,
2321 block_t blkaddr)
2322{
2323 struct page *cpage;
2324
5d4c0af4 2325 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
08b39fbd
CY		 2326 return;
2327
08b39fbd
CY		 2328 cpage = find_lock_page(META_MAPPING(sbi), blkaddr);
2329 if (cpage) {
fec1d657 2330 f2fs_wait_on_page_writeback(cpage, DATA, true);
08b39fbd
CY		 2331 f2fs_put_page(cpage, 1);
2332 }
2333}
2334
351df4b2
JK		 2335static int read_compacted_summaries(struct f2fs_sb_info *sbi)
2336{
2337 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2338 struct curseg_info *seg_i;
2339 unsigned char *kaddr;
2340 struct page *page;
2341 block_t start;
2342 int i, j, offset;
2343
2344 start = start_sum_block(sbi);
2345
2346 page = get_meta_page(sbi, start++);
2347 kaddr = (unsigned char *)page_address(page);
2348
2349 /* Step 1: restore nat cache */
2350 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2351 memcpy(seg_i->journal, kaddr, SUM_JOURNAL_SIZE);
351df4b2
JK		 2352
2353 /* Step 2: restore sit cache */
2354 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2355 memcpy(seg_i->journal, kaddr + SUM_JOURNAL_SIZE, SUM_JOURNAL_SIZE);
351df4b2
JK		 2356 offset = 2 * SUM_JOURNAL_SIZE;
2357
2358 /* Step 3: restore summary entries */
2359 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2360 unsigned short blk_off;
2361 unsigned int segno;
2362
2363 seg_i = CURSEG_I(sbi, i);
2364 segno = le32_to_cpu(ckpt->cur_data_segno[i]);
2365 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[i]);
2366 seg_i->next_segno = segno;
2367 reset_curseg(sbi, i, 0);
2368 seg_i->alloc_type = ckpt->alloc_type[i];
2369 seg_i->next_blkoff = blk_off;
2370
2371 if (seg_i->alloc_type == SSR)
2372 blk_off = sbi->blocks_per_seg;
2373
2374 for (j = 0; j < blk_off; j++) {
2375 struct f2fs_summary *s;
2376 s = (struct f2fs_summary *)(kaddr + offset);
2377 seg_i->sum_blk->entries[j] = *s;
2378 offset += SUMMARY_SIZE;
09cbfeaf 2379 if (offset + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2380 SUM_FOOTER_SIZE)
2381 continue;
2382
2383 f2fs_put_page(page, 1);
2384 page = NULL;
2385
2386 page = get_meta_page(sbi, start++);
2387 kaddr = (unsigned char *)page_address(page);
2388 offset = 0;
2389 }
2390 }
2391 f2fs_put_page(page, 1);
2392 return 0;
2393}
2394
2395static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
2396{
2397 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2398 struct f2fs_summary_block *sum;
2399 struct curseg_info *curseg;
2400 struct page *new;
2401 unsigned short blk_off;
2402 unsigned int segno = 0;
2403 block_t blk_addr = 0;
2404
2405 /* get segment number and block addr */
2406 if (IS_DATASEG(type)) {
2407 segno = le32_to_cpu(ckpt->cur_data_segno[type]);
2408 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[type -
2409 CURSEG_HOT_DATA]);
119ee914 2410 if (__exist_node_summaries(sbi))
351df4b2
JK		 2411 blk_addr = sum_blk_addr(sbi, NR_CURSEG_TYPE, type);
2412 else
2413 blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
2414 } else {
2415 segno = le32_to_cpu(ckpt->cur_node_segno[type -
2416 CURSEG_HOT_NODE]);
2417 blk_off = le16_to_cpu(ckpt->cur_node_blkoff[type -
2418 CURSEG_HOT_NODE]);
119ee914 2419 if (__exist_node_summaries(sbi))
351df4b2
JK		 2420 blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
2421 type - CURSEG_HOT_NODE);
2422 else
2423 blk_addr = GET_SUM_BLOCK(sbi, segno);
2424 }
2425
2426 new = get_meta_page(sbi, blk_addr);
2427 sum = (struct f2fs_summary_block *)page_address(new);
2428
2429 if (IS_NODESEG(type)) {
119ee914 2430 if (__exist_node_summaries(sbi)) {
351df4b2
JK		 2431 struct f2fs_summary *ns = &sum->entries[0];
2432 int i;
2433 for (i = 0; i < sbi->blocks_per_seg; i++, ns++) {
2434 ns->version = 0;
2435 ns->ofs_in_node = 0;
2436 }
2437 } else {
d653788a
GZ		 2438 int err;
2439
2440 err = restore_node_summary(sbi, segno, sum);
2441 if (err) {
351df4b2 2442 f2fs_put_page(new, 1);
d653788a 2443 return err;
351df4b2
JK		 2444 }
2445 }
2446 }
2447
2448 /* set uncompleted segment to curseg */
2449 curseg = CURSEG_I(sbi, type);
2450 mutex_lock(&curseg->curseg_mutex);
b7ad7512
CY		 2451
2452 /* update journal info */
2453 down_write(&curseg->journal_rwsem);
2454 memcpy(curseg->journal, &sum->journal, SUM_JOURNAL_SIZE);
2455 up_write(&curseg->journal_rwsem);
2456
2457 memcpy(curseg->sum_blk->entries, sum->entries, SUM_ENTRY_SIZE);
2458 memcpy(&curseg->sum_blk->footer, &sum->footer, SUM_FOOTER_SIZE);
351df4b2
JK		 2459 curseg->next_segno = segno;
2460 reset_curseg(sbi, type, 0);
2461 curseg->alloc_type = ckpt->alloc_type[type];
2462 curseg->next_blkoff = blk_off;
2463 mutex_unlock(&curseg->curseg_mutex);
2464 f2fs_put_page(new, 1);
2465 return 0;
2466}
2467
2468static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
2469{
2470 int type = CURSEG_HOT_DATA;
e4fc5fbf 2471 int err;
351df4b2 2472
aaec2b1d 2473 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG)) {
3fa06d7b
CY		 2474 int npages = npages_for_summary_flush(sbi, true);
2475
2476 if (npages >= 2)
2477 ra_meta_pages(sbi, start_sum_block(sbi), npages,
26879fb1 2478 META_CP, true);
3fa06d7b 2479
351df4b2
JK		 2480 /* restore for compacted data summary */
2481 if (read_compacted_summaries(sbi))
2482 return -EINVAL;
2483 type = CURSEG_HOT_NODE;
2484 }
2485
119ee914 2486 if (__exist_node_summaries(sbi))
3fa06d7b 2487 ra_meta_pages(sbi, sum_blk_addr(sbi, NR_CURSEG_TYPE, type),
26879fb1 2488 NR_CURSEG_TYPE - type, META_CP, true);
3fa06d7b 2489
e4fc5fbf
CY		 2490 for (; type <= CURSEG_COLD_NODE; type++) {
2491 err = read_normal_summaries(sbi, type);
2492 if (err)
2493 return err;
2494 }
2495
351df4b2
JK		 2496 return 0;
2497}
2498
2499static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
2500{
2501 struct page *page;
2502 unsigned char *kaddr;
2503 struct f2fs_summary *summary;
2504 struct curseg_info *seg_i;
2505 int written_size = 0;
2506 int i, j;
2507
2508 page = grab_meta_page(sbi, blkaddr++);
2509 kaddr = (unsigned char *)page_address(page);
2510
2511 /* Step 1: write nat cache */
2512 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2513 memcpy(kaddr, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2514 written_size += SUM_JOURNAL_SIZE;
2515
2516 /* Step 2: write sit cache */
2517 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2518 memcpy(kaddr + written_size, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2519 written_size += SUM_JOURNAL_SIZE;
2520
351df4b2
JK		 2521 /* Step 3: write summary entries */
2522 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2523 unsigned short blkoff;
2524 seg_i = CURSEG_I(sbi, i);
2525 if (sbi->ckpt->alloc_type[i] == SSR)
2526 blkoff = sbi->blocks_per_seg;
2527 else
2528 blkoff = curseg_blkoff(sbi, i);
2529
2530 for (j = 0; j < blkoff; j++) {
2531 if (!page) {
2532 page = grab_meta_page(sbi, blkaddr++);
2533 kaddr = (unsigned char *)page_address(page);
2534 written_size = 0;
2535 }
2536 summary = (struct f2fs_summary *)(kaddr + written_size);
2537 *summary = seg_i->sum_blk->entries[j];
2538 written_size += SUMMARY_SIZE;
351df4b2 2539
09cbfeaf 2540 if (written_size + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2541 SUM_FOOTER_SIZE)
2542 continue;
2543
e8d61a74 2544 set_page_dirty(page);
351df4b2
JK		 2545 f2fs_put_page(page, 1);
2546 page = NULL;
2547 }
2548 }
e8d61a74
CY		 2549 if (page) {
2550 set_page_dirty(page);
351df4b2 2551 f2fs_put_page(page, 1);
e8d61a74 2552 }
351df4b2
JK		 2553}
2554
2555static void write_normal_summaries(struct f2fs_sb_info *sbi,
2556 block_t blkaddr, int type)
2557{
2558 int i, end;
2559 if (IS_DATASEG(type))
2560 end = type + NR_CURSEG_DATA_TYPE;
2561 else
2562 end = type + NR_CURSEG_NODE_TYPE;
2563
b7ad7512
CY		 2564 for (i = type; i < end; i++)
2565 write_current_sum_page(sbi, i, blkaddr + (i - type));
351df4b2
JK		 2566}
2567
2568void write_data_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2569{
aaec2b1d 2570 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG))
351df4b2
JK		 2571 write_compacted_summaries(sbi, start_blk);
2572 else
2573 write_normal_summaries(sbi, start_blk, CURSEG_HOT_DATA);
2574}
2575
2576void write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2577{
119ee914 2578 write_normal_summaries(sbi, start_blk, CURSEG_HOT_NODE);
351df4b2
JK		 2579}
2580
dfc08a12 2581int lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
351df4b2
JK		 2582 unsigned int val, int alloc)
2583{
2584 int i;
2585
2586 if (type == NAT_JOURNAL) {
dfc08a12
CY		 2587 for (i = 0; i < nats_in_cursum(journal); i++) {
2588 if (le32_to_cpu(nid_in_journal(journal, i)) == val)
351df4b2
JK		 2589 return i;
2590 }
dfc08a12
CY		 2591 if (alloc && __has_cursum_space(journal, 1, NAT_JOURNAL))
2592 return update_nats_in_cursum(journal, 1);
351df4b2 2593 } else if (type == SIT_JOURNAL) {
dfc08a12
CY		 2594 for (i = 0; i < sits_in_cursum(journal); i++)
2595 if (le32_to_cpu(segno_in_journal(journal, i)) == val)
351df4b2 2596 return i;
dfc08a12
CY		 2597 if (alloc && __has_cursum_space(journal, 1, SIT_JOURNAL))
2598 return update_sits_in_cursum(journal, 1);
351df4b2
JK		 2599 }
2600 return -1;
2601}
2602
2603static struct page *get_current_sit_page(struct f2fs_sb_info *sbi,
2604 unsigned int segno)
2605{
2cc22186 2606 return get_meta_page(sbi, current_sit_addr(sbi, segno));
351df4b2
JK		 2607}
2608
2609static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
2610 unsigned int start)
2611{
2612 struct sit_info *sit_i = SIT_I(sbi);
2613 struct page *src_page, *dst_page;
2614 pgoff_t src_off, dst_off;
2615 void *src_addr, *dst_addr;
2616
2617 src_off = current_sit_addr(sbi, start);
2618 dst_off = next_sit_addr(sbi, src_off);
2619
2620 /* get current sit block page without lock */
2621 src_page = get_meta_page(sbi, src_off);
2622 dst_page = grab_meta_page(sbi, dst_off);
9850cf4a 2623 f2fs_bug_on(sbi, PageDirty(src_page));
351df4b2
JK		 2624
2625 src_addr = page_address(src_page);
2626 dst_addr = page_address(dst_page);
09cbfeaf 2627 memcpy(dst_addr, src_addr, PAGE_SIZE);
351df4b2
JK		 2628
2629 set_page_dirty(dst_page);
2630 f2fs_put_page(src_page, 1);
2631
2632 set_to_next_sit(sit_i, start);
2633
2634 return dst_page;
2635}
2636
184a5cd2
CY		 2637static struct sit_entry_set *grab_sit_entry_set(void)
2638{
2639 struct sit_entry_set *ses =
80c54505 2640 f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS);
184a5cd2
CY		 2641
2642 ses->entry_cnt = 0;
2643 INIT_LIST_HEAD(&ses->set_list);
2644 return ses;
2645}
2646
2647static void release_sit_entry_set(struct sit_entry_set *ses)
2648{
2649 list_del(&ses->set_list);
2650 kmem_cache_free(sit_entry_set_slab, ses);
2651}
2652
2653static void adjust_sit_entry_set(struct sit_entry_set *ses,
2654 struct list_head *head)
2655{
2656 struct sit_entry_set *next = ses;
2657
2658 if (list_is_last(&ses->set_list, head))
2659 return;
2660
2661 list_for_each_entry_continue(next, head, set_list)
2662 if (ses->entry_cnt <= next->entry_cnt)
2663 break;
2664
2665 list_move_tail(&ses->set_list, &next->set_list);
2666}
2667
2668static void add_sit_entry(unsigned int segno, struct list_head *head)
2669{
2670 struct sit_entry_set *ses;
2671 unsigned int start_segno = START_SEGNO(segno);
2672
2673 list_for_each_entry(ses, head, set_list) {
2674 if (ses->start_segno == start_segno) {
2675 ses->entry_cnt++;
2676 adjust_sit_entry_set(ses, head);
2677 return;
2678 }
2679 }
2680
2681 ses = grab_sit_entry_set();
2682
2683 ses->start_segno = start_segno;
2684 ses->entry_cnt++;
2685 list_add(&ses->set_list, head);
2686}
2687
2688static void add_sits_in_set(struct f2fs_sb_info *sbi)
2689{
2690 struct f2fs_sm_info *sm_info = SM_I(sbi);
2691 struct list_head *set_list = &sm_info->sit_entry_set;
2692 unsigned long *bitmap = SIT_I(sbi)->dirty_sentries_bitmap;
184a5cd2
CY		 2693 unsigned int segno;
2694
7cd8558b 2695 for_each_set_bit(segno, bitmap, MAIN_SEGS(sbi))
184a5cd2
CY		 2696 add_sit_entry(segno, set_list);
2697}
2698
2699static void remove_sits_in_journal(struct f2fs_sb_info *sbi)
351df4b2
JK		 2700{
2701 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2702 struct f2fs_journal *journal = curseg->journal;
351df4b2
JK		 2703 int i;
2704
b7ad7512 2705 down_write(&curseg->journal_rwsem);
dfc08a12 2706 for (i = 0; i < sits_in_cursum(journal); i++) {
184a5cd2
CY		 2707 unsigned int segno;
2708 bool dirtied;
2709
dfc08a12 2710 segno = le32_to_cpu(segno_in_journal(journal, i));
184a5cd2
CY		 2711 dirtied = __mark_sit_entry_dirty(sbi, segno);
2712
2713 if (!dirtied)
2714 add_sit_entry(segno, &SM_I(sbi)->sit_entry_set);
351df4b2 2715 }
dfc08a12 2716 update_sits_in_cursum(journal, -i);
b7ad7512 2717 up_write(&curseg->journal_rwsem);
351df4b2
JK		 2718}
2719
0a8165d7 2720/*
351df4b2
JK		 2721 * CP calls this function, which flushes SIT entries including sit_journal,
2722 * and moves prefree segs to free segs.
2723 */
4b2fecc8 2724void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2
JK		 2725{
2726 struct sit_info *sit_i = SIT_I(sbi);
2727 unsigned long *bitmap = sit_i->dirty_sentries_bitmap;
2728 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2729 struct f2fs_journal *journal = curseg->journal;
184a5cd2
CY		 2730 struct sit_entry_set *ses, *tmp;
2731 struct list_head *head = &SM_I(sbi)->sit_entry_set;
184a5cd2 2732 bool to_journal = true;
4b2fecc8 2733 struct seg_entry *se;
351df4b2 2734
351df4b2
JK		 2735 mutex_lock(&sit_i->sentry_lock);
2736
2b11a74b
WL		 2737 if (!sit_i->dirty_sentries)
2738 goto out;
2739
351df4b2 2740 /*
184a5cd2
CY		 2741 * add and account sit entries of dirty bitmap in sit entry
2742 * set temporarily
351df4b2 2743 */
184a5cd2 2744 add_sits_in_set(sbi);
351df4b2 2745
184a5cd2
CY		 2746 /*
2747 * if there are no enough space in journal to store dirty sit
2748 * entries, remove all entries from journal and add and account
2749 * them in sit entry set.
2750 */
dfc08a12 2751 if (!__has_cursum_space(journal, sit_i->dirty_sentries, SIT_JOURNAL))
184a5cd2 2752 remove_sits_in_journal(sbi);
b2955550 2753
184a5cd2
CY		 2754 /*
2755 * there are two steps to flush sit entries:
2756 * #1, flush sit entries to journal in current cold data summary block.
2757 * #2, flush sit entries to sit page.
2758 */
2759 list_for_each_entry_safe(ses, tmp, head, set_list) {
4a257ed6 2760 struct page *page = NULL;
184a5cd2
CY		 2761 struct f2fs_sit_block *raw_sit = NULL;
2762 unsigned int start_segno = ses->start_segno;
2763 unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
7cd8558b 2764 (unsigned long)MAIN_SEGS(sbi));
184a5cd2
CY		 2765 unsigned int segno = start_segno;
2766
2767 if (to_journal &&
dfc08a12 2768 !__has_cursum_space(journal, ses->entry_cnt, SIT_JOURNAL))
184a5cd2
CY		 2769 to_journal = false;
2770
b7ad7512
CY		 2771 if (to_journal) {
2772 down_write(&curseg->journal_rwsem);
2773 } else {
184a5cd2
CY		 2774 page = get_next_sit_page(sbi, start_segno);
2775 raw_sit = page_address(page);
351df4b2 2776 }
351df4b2 2777
184a5cd2
CY		 2778 /* flush dirty sit entries in region of current sit set */
2779 for_each_set_bit_from(segno, bitmap, end) {
2780 int offset, sit_offset;
4b2fecc8
JK		 2781
2782 se = get_seg_entry(sbi, segno);
184a5cd2
CY		 2783
2784 /* add discard candidates */
c473f1a9 2785 if (!(cpc->reason & CP_DISCARD)) {
4b2fecc8 2786 cpc->trim_start = segno;
25290fa5 2787 add_discard_addrs(sbi, cpc, false);
4b2fecc8 2788 }
184a5cd2
CY		 2789
2790 if (to_journal) {
dfc08a12 2791 offset = lookup_journal_in_cursum(journal,
184a5cd2
CY		 2792 SIT_JOURNAL, segno, 1);
2793 f2fs_bug_on(sbi, offset < 0);
dfc08a12 2794 segno_in_journal(journal, offset) =
184a5cd2
CY		 2795 cpu_to_le32(segno);
2796 seg_info_to_raw_sit(se,
dfc08a12 2797 &sit_in_journal(journal, offset));
184a5cd2
CY		 2798 } else {
2799 sit_offset = SIT_ENTRY_OFFSET(sit_i, segno);
2800 seg_info_to_raw_sit(se,
2801 &raw_sit->entries[sit_offset]);
2802 }
351df4b2 2803
184a5cd2
CY		 2804 __clear_bit(segno, bitmap);
2805 sit_i->dirty_sentries--;
2806 ses->entry_cnt--;
351df4b2
JK		 2807 }
2808
b7ad7512
CY		 2809 if (to_journal)
2810 up_write(&curseg->journal_rwsem);
2811 else
184a5cd2
CY		 2812 f2fs_put_page(page, 1);
2813
2814 f2fs_bug_on(sbi, ses->entry_cnt);
2815 release_sit_entry_set(ses);
351df4b2 2816 }
184a5cd2
CY		 2817
2818 f2fs_bug_on(sbi, !list_empty(head));
2819 f2fs_bug_on(sbi, sit_i->dirty_sentries);
184a5cd2 2820out:
c473f1a9 2821 if (cpc->reason & CP_DISCARD) {
650d3c4e
YH		 2822 __u64 trim_start = cpc->trim_start;
2823
4b2fecc8 2824 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++)
25290fa5 2825 add_discard_addrs(sbi, cpc, false);
650d3c4e
YH		 2826
2827 cpc->trim_start = trim_start;
4b2fecc8 2828 }
351df4b2 2829 mutex_unlock(&sit_i->sentry_lock);
351df4b2 2830
351df4b2
JK		 2831 set_prefree_as_free_segments(sbi);
2832}
2833
2834static int build_sit_info(struct f2fs_sb_info *sbi)
2835{
2836 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
351df4b2
JK		 2837 struct sit_info *sit_i;
2838 unsigned int sit_segs, start;
ae27d62e 2839 char *src_bitmap;
351df4b2
JK		 2840 unsigned int bitmap_size;
2841
2842 /* allocate memory for SIT information */
2843 sit_i = kzalloc(sizeof(struct sit_info), GFP_KERNEL);
2844 if (!sit_i)
2845 return -ENOMEM;
2846
2847 SM_I(sbi)->sit_info = sit_i;
2848
39307a8e
JK		 2849 sit_i->sentries = f2fs_kvzalloc(MAIN_SEGS(sbi) *
2850 sizeof(struct seg_entry), GFP_KERNEL);
351df4b2
JK		 2851 if (!sit_i->sentries)
2852 return -ENOMEM;
2853
7cd8558b 2854 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2855 sit_i->dirty_sentries_bitmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2856 if (!sit_i->dirty_sentries_bitmap)
2857 return -ENOMEM;
2858
7cd8558b 2859 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 2860 sit_i->sentries[start].cur_valid_map
2861 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2862 sit_i->sentries[start].ckpt_valid_map
2863 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
a66cdd98 2864 if (!sit_i->sentries[start].cur_valid_map ||
3e025740 2865 !sit_i->sentries[start].ckpt_valid_map)
351df4b2 2866 return -ENOMEM;
3e025740 2867
355e7891
CY		 2868#ifdef CONFIG_F2FS_CHECK_FS
2869 sit_i->sentries[start].cur_valid_map_mir
2870 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2871 if (!sit_i->sentries[start].cur_valid_map_mir)
2872 return -ENOMEM;
2873#endif
2874
3e025740
JK		 2875 if (f2fs_discard_en(sbi)) {
2876 sit_i->sentries[start].discard_map
2877 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2878 if (!sit_i->sentries[start].discard_map)
2879 return -ENOMEM;
2880 }
351df4b2
JK		 2881 }
2882
60a3b782
JK		 2883 sit_i->tmp_map = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2884 if (!sit_i->tmp_map)
2885 return -ENOMEM;
2886
351df4b2 2887 if (sbi->segs_per_sec > 1) {
39307a8e
JK		 2888 sit_i->sec_entries = f2fs_kvzalloc(MAIN_SECS(sbi) *
2889 sizeof(struct sec_entry), GFP_KERNEL);
351df4b2
JK		 2890 if (!sit_i->sec_entries)
2891 return -ENOMEM;
2892 }
2893
2894 /* get information related with SIT */
2895 sit_segs = le32_to_cpu(raw_super->segment_count_sit) >> 1;
2896
2897 /* setup SIT bitmap from ckeckpoint pack */
2898 bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
2899 src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
2900
ae27d62e
CY		 2901 sit_i->sit_bitmap = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2902 if (!sit_i->sit_bitmap)
351df4b2 2903 return -ENOMEM;
351df4b2 2904
ae27d62e
CY		 2905#ifdef CONFIG_F2FS_CHECK_FS
2906 sit_i->sit_bitmap_mir = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2907 if (!sit_i->sit_bitmap_mir)
2908 return -ENOMEM;
2909#endif
2910
351df4b2
JK		 2911 /* init SIT information */
2912 sit_i->s_ops = &default_salloc_ops;
2913
2914 sit_i->sit_base_addr = le32_to_cpu(raw_super->sit_blkaddr);
2915 sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
c79b7ff1 2916 sit_i->written_valid_blocks = 0;
351df4b2
JK		 2917 sit_i->bitmap_size = bitmap_size;
2918 sit_i->dirty_sentries = 0;
2919 sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
2920 sit_i->elapsed_time = le64_to_cpu(sbi->ckpt->elapsed_time);
2921 sit_i->mounted_time = CURRENT_TIME_SEC.tv_sec;
2922 mutex_init(&sit_i->sentry_lock);
2923 return 0;
2924}
2925
2926static int build_free_segmap(struct f2fs_sb_info *sbi)
2927{
351df4b2
JK		 2928 struct free_segmap_info *free_i;
2929 unsigned int bitmap_size, sec_bitmap_size;
2930
2931 /* allocate memory for free segmap information */
2932 free_i = kzalloc(sizeof(struct free_segmap_info), GFP_KERNEL);
2933 if (!free_i)
2934 return -ENOMEM;
2935
2936 SM_I(sbi)->free_info = free_i;
2937
7cd8558b 2938 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2939 free_i->free_segmap = f2fs_kvmalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2940 if (!free_i->free_segmap)
2941 return -ENOMEM;
2942
7cd8558b 2943 sec_bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
39307a8e 2944 free_i->free_secmap = f2fs_kvmalloc(sec_bitmap_size, GFP_KERNEL);
351df4b2
JK		 2945 if (!free_i->free_secmap)
2946 return -ENOMEM;
2947
2948 /* set all segments as dirty temporarily */
2949 memset(free_i->free_segmap, 0xff, bitmap_size);
2950 memset(free_i->free_secmap, 0xff, sec_bitmap_size);
2951
2952 /* init free segmap information */
7cd8558b 2953 free_i->start_segno = GET_SEGNO_FROM_SEG0(sbi, MAIN_BLKADDR(sbi));
351df4b2
JK		 2954 free_i->free_segments = 0;
2955 free_i->free_sections = 0;
1a118ccf 2956 spin_lock_init(&free_i->segmap_lock);
351df4b2
JK		 2957 return 0;
2958}
2959
2960static int build_curseg(struct f2fs_sb_info *sbi)
2961{
1042d60f 2962 struct curseg_info *array;
351df4b2
JK		 2963 int i;
2964
b434babf 2965 array = kcalloc(NR_CURSEG_TYPE, sizeof(*array), GFP_KERNEL);
351df4b2
JK		 2966 if (!array)
2967 return -ENOMEM;
2968
2969 SM_I(sbi)->curseg_array = array;
2970
2971 for (i = 0; i < NR_CURSEG_TYPE; i++) {
2972 mutex_init(&array[i].curseg_mutex);
09cbfeaf 2973 array[i].sum_blk = kzalloc(PAGE_SIZE, GFP_KERNEL);
351df4b2
JK		 2974 if (!array[i].sum_blk)
2975 return -ENOMEM;
b7ad7512
CY		 2976 init_rwsem(&array[i].journal_rwsem);
2977 array[i].journal = kzalloc(sizeof(struct f2fs_journal),
2978 GFP_KERNEL);
2979 if (!array[i].journal)
2980 return -ENOMEM;
351df4b2
JK		 2981 array[i].segno = NULL_SEGNO;
2982 array[i].next_blkoff = 0;
2983 }
2984 return restore_curseg_summaries(sbi);
2985}
2986
2987static void build_sit_entries(struct f2fs_sb_info *sbi)
2988{
2989 struct sit_info *sit_i = SIT_I(sbi);
2990 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2991 struct f2fs_journal *journal = curseg->journal;
9c094040
YH		 2992 struct seg_entry *se;
2993 struct f2fs_sit_entry sit;
74de593a
CY		 2994 int sit_blk_cnt = SIT_BLK_CNT(sbi);
2995 unsigned int i, start, end;
2996 unsigned int readed, start_blk = 0;
351df4b2 2997
74de593a 2998 do {
664ba972
JK		 2999 readed = ra_meta_pages(sbi, start_blk, BIO_MAX_PAGES,
3000 META_SIT, true);
74de593a
CY		 3001
3002 start = start_blk * sit_i->sents_per_block;
3003 end = (start_blk + readed) * sit_i->sents_per_block;
3004
7cd8558b 3005 for (; start < end && start < MAIN_SEGS(sbi); start++) {
74de593a 3006 struct f2fs_sit_block *sit_blk;
74de593a
CY		 3007 struct page *page;
3008
9c094040 3009 se = &sit_i->sentries[start];
74de593a
CY		 3010 page = get_current_sit_page(sbi, start);
3011 sit_blk = (struct f2fs_sit_block *)page_address(page);
3012 sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
3013 f2fs_put_page(page, 1);
d600af23 3014
74de593a
CY		 3015 check_block_count(sbi, start, &sit);
3016 seg_info_from_raw_sit(se, &sit);
a66cdd98
JK		 3017
3018 /* build discard map only one time */
3e025740 3019 if (f2fs_discard_en(sbi)) {
1f43e2ad
CY		 3020 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
3021 memset(se->discard_map, 0xff,
3022 SIT_VBLOCK_MAP_SIZE);
3023 } else {
3024 memcpy(se->discard_map,
3025 se->cur_valid_map,
3026 SIT_VBLOCK_MAP_SIZE);
3027 sbi->discard_blks +=
3028 sbi->blocks_per_seg -
3029 se->valid_blocks;
3030 }
3e025740 3031 }
a66cdd98 3032
d600af23
CY		 3033 if (sbi->segs_per_sec > 1)
3034 get_sec_entry(sbi, start)->valid_blocks +=
3035 se->valid_blocks;
351df4b2 3036 }
74de593a
CY		 3037 start_blk += readed;
3038 } while (start_blk < sit_blk_cnt);
d600af23
CY		 3039
3040 down_read(&curseg->journal_rwsem);
3041 for (i = 0; i < sits_in_cursum(journal); i++) {
d600af23
CY		 3042 unsigned int old_valid_blocks;
3043
3044 start = le32_to_cpu(segno_in_journal(journal, i));
3045 se = &sit_i->sentries[start];
3046 sit = sit_in_journal(journal, i);
3047
3048 old_valid_blocks = se->valid_blocks;
3049
3050 check_block_count(sbi, start, &sit);
3051 seg_info_from_raw_sit(se, &sit);
3052
3053 if (f2fs_discard_en(sbi)) {
1f43e2ad
CY		 3054 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
3055 memset(se->discard_map, 0xff,
3056 SIT_VBLOCK_MAP_SIZE);
3057 } else {
3058 memcpy(se->discard_map, se->cur_valid_map,
3059 SIT_VBLOCK_MAP_SIZE);
3060 sbi->discard_blks += old_valid_blocks -
3061 se->valid_blocks;
3062 }
d600af23
CY		 3063 }
3064
3065 if (sbi->segs_per_sec > 1)
3066 get_sec_entry(sbi, start)->valid_blocks +=
3067 se->valid_blocks - old_valid_blocks;
3068 }
3069 up_read(&curseg->journal_rwsem);
351df4b2
JK		 3070}
3071
3072static void init_free_segmap(struct f2fs_sb_info *sbi)
3073{
3074 unsigned int start;
3075 int type;
3076
7cd8558b 3077 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 3078 struct seg_entry *sentry = get_seg_entry(sbi, start);
3079 if (!sentry->valid_blocks)
3080 __set_free(sbi, start);
c79b7ff1
JK		 3081 else
3082 SIT_I(sbi)->written_valid_blocks +=
3083 sentry->valid_blocks;
351df4b2
JK		 3084 }
3085
3086 /* set use the current segments */
3087 for (type = CURSEG_HOT_DATA; type <= CURSEG_COLD_NODE; type++) {
3088 struct curseg_info *curseg_t = CURSEG_I(sbi, type);
3089 __set_test_and_inuse(sbi, curseg_t->segno);
3090 }
3091}
3092
3093static void init_dirty_segmap(struct f2fs_sb_info *sbi)
3094{
3095 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3096 struct free_segmap_info *free_i = FREE_I(sbi);
7cd8558b 3097 unsigned int segno = 0, offset = 0;
351df4b2
JK		 3098 unsigned short valid_blocks;
3099
8736fbf0 3100 while (1) {
351df4b2 3101 /* find dirty segment based on free segmap */
7cd8558b
JK		 3102 segno = find_next_inuse(free_i, MAIN_SEGS(sbi), offset);
3103 if (segno >= MAIN_SEGS(sbi))
351df4b2
JK		 3104 break;
3105 offset = segno + 1;
302bd348 3106 valid_blocks = get_valid_blocks(sbi, segno, false);
ec325b52 3107 if (valid_blocks == sbi->blocks_per_seg || !valid_blocks)
351df4b2 3108 continue;
ec325b52
JK		 3109 if (valid_blocks > sbi->blocks_per_seg) {
3110 f2fs_bug_on(sbi, 1);
3111 continue;
3112 }
351df4b2
JK		 3113 mutex_lock(&dirty_i->seglist_lock);
3114 __locate_dirty_segment(sbi, segno, DIRTY);
3115 mutex_unlock(&dirty_i->seglist_lock);
3116 }
3117}
3118
5ec4e49f 3119static int init_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3120{
3121 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
7cd8558b 3122 unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
351df4b2 3123
39307a8e 3124 dirty_i->victim_secmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
5ec4e49f 3125 if (!dirty_i->victim_secmap)
351df4b2
JK		 3126 return -ENOMEM;
3127 return 0;
3128}
3129
3130static int build_dirty_segmap(struct f2fs_sb_info *sbi)
3131{
3132 struct dirty_seglist_info *dirty_i;
3133 unsigned int bitmap_size, i;
3134
3135 /* allocate memory for dirty segments list information */
3136 dirty_i = kzalloc(sizeof(struct dirty_seglist_info), GFP_KERNEL);
3137 if (!dirty_i)
3138 return -ENOMEM;
3139
3140 SM_I(sbi)->dirty_info = dirty_i;
3141 mutex_init(&dirty_i->seglist_lock);
3142
7cd8558b 3143 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
351df4b2
JK		 3144
3145 for (i = 0; i < NR_DIRTY_TYPE; i++) {
39307a8e 3146 dirty_i->dirty_segmap[i] = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 3147 if (!dirty_i->dirty_segmap[i])
3148 return -ENOMEM;
3149 }
3150
3151 init_dirty_segmap(sbi);
5ec4e49f 3152 return init_victim_secmap(sbi);
351df4b2
JK		 3153}
3154
0a8165d7 3155/*
351df4b2
JK		 3156 * Update min, max modified time for cost-benefit GC algorithm
3157 */
3158static void init_min_max_mtime(struct f2fs_sb_info *sbi)
3159{
3160 struct sit_info *sit_i = SIT_I(sbi);
3161 unsigned int segno;
3162
3163 mutex_lock(&sit_i->sentry_lock);
3164
3165 sit_i->min_mtime = LLONG_MAX;
3166
7cd8558b 3167 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
351df4b2
JK		 3168 unsigned int i;
3169 unsigned long long mtime = 0;
3170
3171 for (i = 0; i < sbi->segs_per_sec; i++)
3172 mtime += get_seg_entry(sbi, segno + i)->mtime;
3173
3174 mtime = div_u64(mtime, sbi->segs_per_sec);
3175
3176 if (sit_i->min_mtime > mtime)
3177 sit_i->min_mtime = mtime;
3178 }
3179 sit_i->max_mtime = get_mtime(sbi);
3180 mutex_unlock(&sit_i->sentry_lock);
3181}
3182
3183int build_segment_manager(struct f2fs_sb_info *sbi)
3184{
3185 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3186 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1042d60f 3187 struct f2fs_sm_info *sm_info;
351df4b2
JK		 3188 int err;
3189
3190 sm_info = kzalloc(sizeof(struct f2fs_sm_info), GFP_KERNEL);
3191 if (!sm_info)
3192 return -ENOMEM;
3193
3194 /* init sm info */
3195 sbi->sm_info = sm_info;
351df4b2
JK		 3196 sm_info->seg0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3197 sm_info->main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3198 sm_info->segment_count = le32_to_cpu(raw_super->segment_count);
3199 sm_info->reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3200 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3201 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
3202 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
58c41035
JK		 3203 sm_info->rec_prefree_segments = sm_info->main_segments *
3204 DEF_RECLAIM_PREFREE_SEGMENTS / 100;
44a83499
JK		 3205 if (sm_info->rec_prefree_segments > DEF_MAX_RECLAIM_PREFREE_SEGMENTS)
3206 sm_info->rec_prefree_segments = DEF_MAX_RECLAIM_PREFREE_SEGMENTS;
3207
52763a4b
JK		 3208 if (!test_opt(sbi, LFS))
3209 sm_info->ipu_policy = 1 << F2FS_IPU_FSYNC;
216fbd64 3210 sm_info->min_ipu_util = DEF_MIN_IPU_UTIL;
c1ce1b02 3211 sm_info->min_fsync_blocks = DEF_MIN_FSYNC_BLOCKS;
ef095d19 3212 sm_info->min_hot_blocks = DEF_MIN_HOT_BLOCKS;
351df4b2 3213
bba681cb
JK		 3214 sm_info->trim_sections = DEF_BATCHED_TRIM_SECTIONS;
3215
184a5cd2
CY		 3216 INIT_LIST_HEAD(&sm_info->sit_entry_set);
3217
b270ad6f 3218 if (test_opt(sbi, FLUSH_MERGE) && !f2fs_readonly(sbi->sb)) {
2163d198
GZ		 3219 err = create_flush_cmd_control(sbi);
3220 if (err)
a688b9d9 3221 return err;
6b4afdd7
JK		 3222 }
3223
0b54fb84
JK		 3224 err = create_discard_cmd_control(sbi);
3225 if (err)
3226 return err;
3227
351df4b2
JK		 3228 err = build_sit_info(sbi);
3229 if (err)
3230 return err;
3231 err = build_free_segmap(sbi);
3232 if (err)
3233 return err;
3234 err = build_curseg(sbi);
3235 if (err)
3236 return err;
3237
3238 /* reinit free segmap based on SIT */
3239 build_sit_entries(sbi);
3240
3241 init_free_segmap(sbi);
3242 err = build_dirty_segmap(sbi);
3243 if (err)
3244 return err;
3245
3246 init_min_max_mtime(sbi);
3247 return 0;
3248}
3249
3250static void discard_dirty_segmap(struct f2fs_sb_info *sbi,
3251 enum dirty_type dirty_type)
3252{
3253 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3254
3255 mutex_lock(&dirty_i->seglist_lock);
39307a8e 3256 kvfree(dirty_i->dirty_segmap[dirty_type]);
351df4b2
JK		 3257 dirty_i->nr_dirty[dirty_type] = 0;
3258 mutex_unlock(&dirty_i->seglist_lock);
3259}
3260
5ec4e49f 3261static void destroy_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 3262{
3263 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
39307a8e 3264 kvfree(dirty_i->victim_secmap);
351df4b2
JK		 3265}
3266
3267static void destroy_dirty_segmap(struct f2fs_sb_info *sbi)
3268{
3269 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
3270 int i;
3271
3272 if (!dirty_i)
3273 return;
3274
3275 /* discard pre-free/dirty segments list */
3276 for (i = 0; i < NR_DIRTY_TYPE; i++)
3277 discard_dirty_segmap(sbi, i);
3278
5ec4e49f 3279 destroy_victim_secmap(sbi);
351df4b2
JK		 3280 SM_I(sbi)->dirty_info = NULL;
3281 kfree(dirty_i);
3282}
3283
3284static void destroy_curseg(struct f2fs_sb_info *sbi)
3285{
3286 struct curseg_info *array = SM_I(sbi)->curseg_array;
3287 int i;
3288
3289 if (!array)
3290 return;
3291 SM_I(sbi)->curseg_array = NULL;
b7ad7512 3292 for (i = 0; i < NR_CURSEG_TYPE; i++) {
351df4b2 3293 kfree(array[i].sum_blk);
b7ad7512
CY		 3294 kfree(array[i].journal);
3295 }
351df4b2
JK		 3296 kfree(array);
3297}
3298
3299static void destroy_free_segmap(struct f2fs_sb_info *sbi)
3300{
3301 struct free_segmap_info *free_i = SM_I(sbi)->free_info;
3302 if (!free_i)
3303 return;
3304 SM_I(sbi)->free_info = NULL;
39307a8e
JK		 3305 kvfree(free_i->free_segmap);
3306 kvfree(free_i->free_secmap);
351df4b2
JK		 3307 kfree(free_i);
3308}
3309
3310static void destroy_sit_info(struct f2fs_sb_info *sbi)
3311{
3312 struct sit_info *sit_i = SIT_I(sbi);
3313 unsigned int start;
3314
3315 if (!sit_i)
3316 return;
3317
3318 if (sit_i->sentries) {
7cd8558b 3319 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2 3320 kfree(sit_i->sentries[start].cur_valid_map);
355e7891
CY		 3321#ifdef CONFIG_F2FS_CHECK_FS
3322 kfree(sit_i->sentries[start].cur_valid_map_mir);
3323#endif
351df4b2 3324 kfree(sit_i->sentries[start].ckpt_valid_map);
a66cdd98 3325 kfree(sit_i->sentries[start].discard_map);
351df4b2
JK		 3326 }
3327 }
60a3b782
JK		 3328 kfree(sit_i->tmp_map);
3329
39307a8e
JK		 3330 kvfree(sit_i->sentries);
3331 kvfree(sit_i->sec_entries);
3332 kvfree(sit_i->dirty_sentries_bitmap);
351df4b2
JK		 3333
3334 SM_I(sbi)->sit_info = NULL;
3335 kfree(sit_i->sit_bitmap);
ae27d62e
CY		 3336#ifdef CONFIG_F2FS_CHECK_FS
3337 kfree(sit_i->sit_bitmap_mir);
3338#endif
351df4b2
JK		 3339 kfree(sit_i);
3340}
3341
3342void destroy_segment_manager(struct f2fs_sb_info *sbi)
3343{
3344 struct f2fs_sm_info *sm_info = SM_I(sbi);
a688b9d9 3345
3b03f724
CY		 3346 if (!sm_info)
3347 return;
5eba8c5d 3348 destroy_flush_cmd_control(sbi, true);
f099405f 3349 destroy_discard_cmd_control(sbi);
351df4b2
JK		 3350 destroy_dirty_segmap(sbi);
3351 destroy_curseg(sbi);
3352 destroy_free_segmap(sbi);
3353 destroy_sit_info(sbi);
3354 sbi->sm_info = NULL;
3355 kfree(sm_info);
3356}
7fd9e544
JK		 3357
3358int __init create_segment_manager_caches(void)
3359{
3360 discard_entry_slab = f2fs_kmem_cache_create("discard_entry",
e8512d2e 3361 sizeof(struct discard_entry));
7fd9e544 3362 if (!discard_entry_slab)
184a5cd2
CY		 3363 goto fail;
3364
b01a9201
JK		 3365 discard_cmd_slab = f2fs_kmem_cache_create("discard_cmd",
3366 sizeof(struct discard_cmd));
3367 if (!discard_cmd_slab)
6ab2a308 3368 goto destroy_discard_entry;
275b66b0 3369
184a5cd2 3370 sit_entry_set_slab = f2fs_kmem_cache_create("sit_entry_set",
c9ee0085 3371 sizeof(struct sit_entry_set));
184a5cd2 3372 if (!sit_entry_set_slab)
b01a9201 3373 goto destroy_discard_cmd;
88b88a66
JK		 3374
3375 inmem_entry_slab = f2fs_kmem_cache_create("inmem_page_entry",
3376 sizeof(struct inmem_pages));
3377 if (!inmem_entry_slab)
3378 goto destroy_sit_entry_set;
7fd9e544 3379 return 0;
184a5cd2 3380
88b88a66
JK		 3381destroy_sit_entry_set:
3382 kmem_cache_destroy(sit_entry_set_slab);
b01a9201
JK		 3383destroy_discard_cmd:
3384 kmem_cache_destroy(discard_cmd_slab);
6ab2a308 3385destroy_discard_entry:
184a5cd2
CY		 3386 kmem_cache_destroy(discard_entry_slab);
3387fail:
3388 return -ENOMEM;
7fd9e544
JK		 3389}
3390
3391void destroy_segment_manager_caches(void)
3392{
184a5cd2 3393 kmem_cache_destroy(sit_entry_set_slab);
b01a9201 3394 kmem_cache_destroy(discard_cmd_slab);
7fd9e544 3395 kmem_cache_destroy(discard_entry_slab);
88b88a66 3396 kmem_cache_destroy(inmem_entry_slab);
7fd9e544 3397}
Ubuntu Jammy Linux kernel mirror