]> 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: fix a plint compile warning
[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
28bc106b
CY		 253static int __commit_inmem_pages(struct inode *inode,
254 struct list_head *revoke_list)
88b88a66
JK		 255{
256 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
257 struct f2fs_inode_info *fi = F2FS_I(inode);
258 struct inmem_pages *cur, *tmp;
88b88a66 259 struct f2fs_io_info fio = {
05ca3632 260 .sbi = sbi,
88b88a66 261 .type = DATA,
04d328de 262 .op = REQ_OP_WRITE,
70fd7614 263 .op_flags = REQ_SYNC | REQ_PRIO,
4375a336 264 .encrypted_page = NULL,
88b88a66 265 };
942fd319 266 pgoff_t last_idx = ULONG_MAX;
edb27dee 267 int err = 0;
88b88a66 268
88b88a66 269 list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
28bc106b
CY		 270 struct page *page = cur->page;
271
272 lock_page(page);
273 if (page->mapping == inode->i_mapping) {
274 trace_f2fs_commit_inmem_page(page, INMEM);
275
276 set_page_dirty(page);
277 f2fs_wait_on_page_writeback(page, DATA, true);
933439c8 278 if (clear_page_dirty_for_io(page)) {
29b96b54 279 inode_dec_dirty_pages(inode);
933439c8
CY		 280 remove_dirty_inode(inode);
281 }
28bc106b
CY		 282
283 fio.page = page;
29b96b54
CY		 284 err = do_write_data_page(&fio);
285 if (err) {
28bc106b 286 unlock_page(page);
29b96b54 287 break;
70c640b1 288 }
29b96b54 289
28bc106b
CY		 290 /* record old blkaddr for revoking */
291 cur->old_addr = fio.old_blkaddr;
942fd319 292 last_idx = page->index;
28bc106b
CY		 293 }
294 unlock_page(page);
295 list_move_tail(&cur->list, revoke_list);
88b88a66 296 }
29b96b54 297
942fd319
JK		 298 if (last_idx != ULONG_MAX)
299 f2fs_submit_merged_bio_cond(sbi, inode, 0, last_idx,
300 DATA, WRITE);
28bc106b
CY		 301
302 if (!err)
303 __revoke_inmem_pages(inode, revoke_list, false, false);
304
29b96b54
CY		 305 return err;
306}
307
308int commit_inmem_pages(struct inode *inode)
309{
310 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
311 struct f2fs_inode_info *fi = F2FS_I(inode);
28bc106b
CY		 312 struct list_head revoke_list;
313 int err;
29b96b54 314
28bc106b 315 INIT_LIST_HEAD(&revoke_list);
29b96b54
CY		 316 f2fs_balance_fs(sbi, true);
317 f2fs_lock_op(sbi);
318
5fe45743
CY		 319 set_inode_flag(inode, FI_ATOMIC_COMMIT);
320
29b96b54 321 mutex_lock(&fi->inmem_lock);
28bc106b
CY		 322 err = __commit_inmem_pages(inode, &revoke_list);
323 if (err) {
324 int ret;
325 /*
326 * try to revoke all committed pages, but still we could fail
327 * due to no memory or other reason, if that happened, EAGAIN
328 * will be returned, which means in such case, transaction is
329 * already not integrity, caller should use journal to do the
330 * recovery or rewrite & commit last transaction. For other
331 * error number, revoking was done by filesystem itself.
332 */
333 ret = __revoke_inmem_pages(inode, &revoke_list, false, true);
334 if (ret)
335 err = ret;
336
337 /* drop all uncommitted pages */
338 __revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
339 }
88b88a66
JK		 340 mutex_unlock(&fi->inmem_lock);
341
5fe45743
CY		 342 clear_inode_flag(inode, FI_ATOMIC_COMMIT);
343
29b96b54 344 f2fs_unlock_op(sbi);
edb27dee 345 return err;
88b88a66
JK		 346}
347
0a8165d7 348/*
351df4b2
JK		 349 * This function balances dirty node and dentry pages.
350 * In addition, it controls garbage collection.
351 */
2c4db1a6 352void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
351df4b2 353{
0f348028 354#ifdef CONFIG_F2FS_FAULT_INJECTION
55523519
CY		 355 if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
356 f2fs_show_injection_info(FAULT_CHECKPOINT);
0f348028 357 f2fs_stop_checkpoint(sbi, false);
55523519 358 }
0f348028
CY		 359#endif
360
2c4db1a6
JK		 361 if (!need)
362 return;
e589c2c4
JK		 363
364 /* balance_fs_bg is able to be pending */
365 if (excess_cached_nats(sbi))
366 f2fs_balance_fs_bg(sbi);
367
351df4b2 368 /*
029cd28c
JK		 369 * We should do GC or end up with checkpoint, if there are so many dirty
370 * dir/node pages without enough free segments.
351df4b2 371 */
7f3037a5 372 if (has_not_enough_free_secs(sbi, 0, 0)) {
351df4b2 373 mutex_lock(&sbi->gc_mutex);
7702bdbe 374 f2fs_gc(sbi, false, false);
351df4b2
JK		 375 }
376}
377
4660f9c0
JK		 378void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi)
379{
1dcc336b 380 /* try to shrink extent cache when there is no enough memory */
554df79e
JK		 381 if (!available_free_memory(sbi, EXTENT_CACHE))
382 f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER);
1dcc336b 383
1b38dc8e
JK		 384 /* check the # of cached NAT entries */
385 if (!available_free_memory(sbi, NAT_ENTRIES))
386 try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK);
387
31696580 388 if (!available_free_memory(sbi, FREE_NIDS))
ad4edb83
JK		 389 try_to_free_nids(sbi, MAX_FREE_NIDS);
390 else
22ad0b6a 391 build_free_nids(sbi, false, false);
31696580 392
f455c8a5
JK		 393 if (!is_idle(sbi))
394 return;
31696580 395
1b38dc8e
JK		 396 /* checkpoint is the only way to shrink partial cached entries */
397 if (!available_free_memory(sbi, NAT_ENTRIES) ||
60b99b48 398 !available_free_memory(sbi, INO_ENTRIES) ||
7d768d2c
CY		 399 excess_prefree_segs(sbi) ||
400 excess_dirty_nats(sbi) ||
f455c8a5 401 f2fs_time_over(sbi, CP_TIME)) {
e9f5b8b8
CY		 402 if (test_opt(sbi, DATA_FLUSH)) {
403 struct blk_plug plug;
404
405 blk_start_plug(&plug);
36b35a0d 406 sync_dirty_inodes(sbi, FILE_INODE);
e9f5b8b8
CY		 407 blk_finish_plug(&plug);
408 }
4660f9c0 409 f2fs_sync_fs(sbi->sb, true);
42190d2a 410 stat_inc_bg_cp_count(sbi->stat_info);
36b35a0d 411 }
4660f9c0
JK		 412}
413
3c62be17
JK		 414static int __submit_flush_wait(struct block_device *bdev)
415{
416 struct bio *bio = f2fs_bio_alloc(0);
417 int ret;
418
09cb6464 419 bio->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
3c62be17
JK		 420 bio->bi_bdev = bdev;
421 ret = submit_bio_wait(bio);
422 bio_put(bio);
423 return ret;
424}
425
426static int submit_flush_wait(struct f2fs_sb_info *sbi)
427{
428 int ret = __submit_flush_wait(sbi->sb->s_bdev);
429 int i;
430
431 if (sbi->s_ndevs && !ret) {
432 for (i = 1; i < sbi->s_ndevs; i++) {
d50aaeec
JK		 433 trace_f2fs_issue_flush(FDEV(i).bdev,
434 test_opt(sbi, NOBARRIER),
435 test_opt(sbi, FLUSH_MERGE));
3c62be17
JK		 436 ret = __submit_flush_wait(FDEV(i).bdev);
437 if (ret)
438 break;
439 }
440 }
441 return ret;
442}
443
2163d198 444static int issue_flush_thread(void *data)
6b4afdd7
JK		 445{
446 struct f2fs_sb_info *sbi = data;
b01a9201 447 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
a688b9d9 448 wait_queue_head_t *q = &fcc->flush_wait_queue;
6b4afdd7
JK		 449repeat:
450 if (kthread_should_stop())
451 return 0;
452
721bd4d5 453 if (!llist_empty(&fcc->issue_list)) {
6b4afdd7
JK		 454 struct flush_cmd *cmd, *next;
455 int ret;
456
721bd4d5
GZ		 457 fcc->dispatch_list = llist_del_all(&fcc->issue_list);
458 fcc->dispatch_list = llist_reverse_order(fcc->dispatch_list);
459
3c62be17 460 ret = submit_flush_wait(sbi);
721bd4d5
GZ		 461 llist_for_each_entry_safe(cmd, next,
462 fcc->dispatch_list, llnode) {
6b4afdd7 463 cmd->ret = ret;
6b4afdd7
JK		 464 complete(&cmd->wait);
465 }
a688b9d9 466 fcc->dispatch_list = NULL;
6b4afdd7
JK		 467 }
468
a688b9d9 469 wait_event_interruptible(*q,
721bd4d5 470 kthread_should_stop() || !llist_empty(&fcc->issue_list));
6b4afdd7
JK		 471 goto repeat;
472}
473
474int f2fs_issue_flush(struct f2fs_sb_info *sbi)
475{
b01a9201 476 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
adf8d90b 477 struct flush_cmd cmd;
6b4afdd7 478
0f7b2abd
JK		 479 if (test_opt(sbi, NOBARRIER))
480 return 0;
481
73545817
CY		 482 if (!test_opt(sbi, FLUSH_MERGE))
483 return submit_flush_wait(sbi);
484
485 if (!atomic_read(&fcc->submit_flush)) {
740432f8
JK		 486 int ret;
487
0a87f664 488 atomic_inc(&fcc->submit_flush);
3c62be17 489 ret = submit_flush_wait(sbi);
0a87f664 490 atomic_dec(&fcc->submit_flush);
740432f8
JK		 491 return ret;
492 }
6b4afdd7 493
adf8d90b 494 init_completion(&cmd.wait);
6b4afdd7 495
0a87f664 496 atomic_inc(&fcc->submit_flush);
721bd4d5 497 llist_add(&cmd.llnode, &fcc->issue_list);
6b4afdd7 498
a688b9d9
GZ		 499 if (!fcc->dispatch_list)
500 wake_up(&fcc->flush_wait_queue);
6b4afdd7 501
5eba8c5d
JK		 502 if (fcc->f2fs_issue_flush) {
503 wait_for_completion(&cmd.wait);
504 atomic_dec(&fcc->submit_flush);
505 } else {
506 llist_del_all(&fcc->issue_list);
507 atomic_set(&fcc->submit_flush, 0);
508 }
adf8d90b
CY		 509
510 return cmd.ret;
6b4afdd7
JK		 511}
512
2163d198
GZ		 513int create_flush_cmd_control(struct f2fs_sb_info *sbi)
514{
515 dev_t dev = sbi->sb->s_bdev->bd_dev;
516 struct flush_cmd_control *fcc;
517 int err = 0;
518
b01a9201
JK		 519 if (SM_I(sbi)->fcc_info) {
520 fcc = SM_I(sbi)->fcc_info;
5eba8c5d
JK		 521 goto init_thread;
522 }
523
2163d198
GZ		 524 fcc = kzalloc(sizeof(struct flush_cmd_control), GFP_KERNEL);
525 if (!fcc)
526 return -ENOMEM;
0a87f664 527 atomic_set(&fcc->submit_flush, 0);
2163d198 528 init_waitqueue_head(&fcc->flush_wait_queue);
721bd4d5 529 init_llist_head(&fcc->issue_list);
b01a9201 530 SM_I(sbi)->fcc_info = fcc;
5eba8c5d 531init_thread:
2163d198
GZ		 532 fcc->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi,
533 "f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev));
534 if (IS_ERR(fcc->f2fs_issue_flush)) {
535 err = PTR_ERR(fcc->f2fs_issue_flush);
536 kfree(fcc);
b01a9201 537 SM_I(sbi)->fcc_info = NULL;
2163d198
GZ		 538 return err;
539 }
2163d198
GZ		 540
541 return err;
542}
543
5eba8c5d 544void destroy_flush_cmd_control(struct f2fs_sb_info *sbi, bool free)
2163d198 545{
b01a9201 546 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
2163d198 547
5eba8c5d
JK		 548 if (fcc && fcc->f2fs_issue_flush) {
549 struct task_struct *flush_thread = fcc->f2fs_issue_flush;
550
551 fcc->f2fs_issue_flush = NULL;
552 kthread_stop(flush_thread);
553 }
554 if (free) {
555 kfree(fcc);
b01a9201 556 SM_I(sbi)->fcc_info = NULL;
5eba8c5d 557 }
2163d198
GZ		 558}
559
351df4b2
JK		 560static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
561 enum dirty_type dirty_type)
562{
563 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
564
565 /* need not be added */
566 if (IS_CURSEG(sbi, segno))
567 return;
568
569 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[dirty_type]))
570 dirty_i->nr_dirty[dirty_type]++;
571
572 if (dirty_type == DIRTY) {
573 struct seg_entry *sentry = get_seg_entry(sbi, segno);
4625d6aa 574 enum dirty_type t = sentry->type;
b2f2c390 575
ec325b52
JK		 576 if (unlikely(t >= DIRTY)) {
577 f2fs_bug_on(sbi, 1);
578 return;
579 }
4625d6aa
CL		 580 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[t]))
581 dirty_i->nr_dirty[t]++;
351df4b2
JK		 582 }
583}
584
585static void __remove_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
586 enum dirty_type dirty_type)
587{
588 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
589
590 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[dirty_type]))
591 dirty_i->nr_dirty[dirty_type]--;
592
593 if (dirty_type == DIRTY) {
4625d6aa
CL		 594 struct seg_entry *sentry = get_seg_entry(sbi, segno);
595 enum dirty_type t = sentry->type;
596
597 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
598 dirty_i->nr_dirty[t]--;
b2f2c390 599
5ec4e49f
JK		 600 if (get_valid_blocks(sbi, segno, sbi->segs_per_sec) == 0)
601 clear_bit(GET_SECNO(sbi, segno),
602 dirty_i->victim_secmap);
351df4b2
JK		 603 }
604}
605
0a8165d7 606/*
351df4b2
JK		 607 * Should not occur error such as -ENOMEM.
608 * Adding dirty entry into seglist is not critical operation.
609 * If a given segment is one of current working segments, it won't be added.
610 */
8d8451af 611static void locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 612{
613 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
614 unsigned short valid_blocks;
615
616 if (segno == NULL_SEGNO || IS_CURSEG(sbi, segno))
617 return;
618
619 mutex_lock(&dirty_i->seglist_lock);
620
621 valid_blocks = get_valid_blocks(sbi, segno, 0);
622
623 if (valid_blocks == 0) {
624 __locate_dirty_segment(sbi, segno, PRE);
625 __remove_dirty_segment(sbi, segno, DIRTY);
626 } else if (valid_blocks < sbi->blocks_per_seg) {
627 __locate_dirty_segment(sbi, segno, DIRTY);
628 } else {
629 /* Recovery routine with SSR needs this */
630 __remove_dirty_segment(sbi, segno, DIRTY);
631 }
632
633 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 634}
635
15469963 636static void __add_discard_cmd(struct f2fs_sb_info *sbi,
4e6a8d9b 637 struct bio *bio, block_t lstart, block_t len)
275b66b0 638{
0b54fb84
JK		 639 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
640 struct list_head *cmd_list = &(dcc->discard_cmd_list);
b01a9201 641 struct discard_cmd *dc;
275b66b0 642
b01a9201
JK		 643 dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS);
644 INIT_LIST_HEAD(&dc->list);
645 dc->bio = bio;
15469963 646 bio->bi_private = dc;
b01a9201
JK		 647 dc->lstart = lstart;
648 dc->len = len;
15469963 649 dc->state = D_PREP;
b01a9201 650 init_completion(&dc->wait);
15469963
JK		 651
652 mutex_lock(&dcc->cmd_lock);
0b54fb84 653 list_add_tail(&dc->list, cmd_list);
15469963
JK		 654 mutex_unlock(&dcc->cmd_lock);
655}
656
657static void __remove_discard_cmd(struct f2fs_sb_info *sbi, struct discard_cmd *dc)
658{
659 int err = dc->bio->bi_error;
275b66b0 660
dcc9165d
JK		 661 if (dc->state == D_DONE)
662 atomic_dec(&(SM_I(sbi)->dcc_info->submit_discard));
663
15469963
JK		 664 if (err == -EOPNOTSUPP)
665 err = 0;
666
667 if (err)
668 f2fs_msg(sbi->sb, KERN_INFO,
669 "Issue discard failed, ret: %d", err);
670 bio_put(dc->bio);
671 list_del(&dc->list);
672 kmem_cache_free(discard_cmd_slab, dc);
275b66b0
CY		 673}
674
4e6a8d9b
JK		 675/* This should be covered by global mutex, &sit_i->sentry_lock */
676void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
275b66b0 677{
0b54fb84
JK		 678 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
679 struct list_head *wait_list = &(dcc->discard_cmd_list);
b01a9201 680 struct discard_cmd *dc, *tmp;
40465257 681 struct blk_plug plug;
275b66b0 682
15469963 683 mutex_lock(&dcc->cmd_lock);
40465257
JK		 684
685 blk_start_plug(&plug);
686
b01a9201 687 list_for_each_entry_safe(dc, tmp, wait_list, list) {
275b66b0 688
15469963
JK		 689 if (blkaddr == NULL_ADDR) {
690 if (dc->state == D_PREP) {
691 dc->state = D_SUBMIT;
692 submit_bio(dc->bio);
dcc9165d 693 atomic_inc(&dcc->submit_discard);
15469963 694 }
15469963
JK		 695 continue;
696 }
697
698 if (dc->lstart <= blkaddr && blkaddr < dc->lstart + dc->len) {
699 if (dc->state == D_SUBMIT)
b01a9201 700 wait_for_completion_io(&dc->wait);
4e6a8d9b 701 else
15469963 702 __remove_discard_cmd(sbi, dc);
4e6a8d9b 703 }
275b66b0 704 }
40465257
JK		 705 blk_finish_plug(&plug);
706
707 /* this comes from f2fs_put_super */
708 if (blkaddr == NULL_ADDR) {
709 list_for_each_entry_safe(dc, tmp, wait_list, list) {
710 wait_for_completion_io(&dc->wait);
711 __remove_discard_cmd(sbi, dc);
712 }
713 }
15469963 714 mutex_unlock(&dcc->cmd_lock);
275b66b0
CY		 715}
716
b01a9201 717static void f2fs_submit_discard_endio(struct bio *bio)
275b66b0 718{
b01a9201 719 struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
275b66b0 720
b01a9201 721 complete(&dc->wait);
15469963 722 dc->state = D_DONE;
275b66b0
CY		 723}
724
15469963
JK		 725static int issue_discard_thread(void *data)
726{
727 struct f2fs_sb_info *sbi = data;
728 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
729 wait_queue_head_t *q = &dcc->discard_wait_queue;
730 struct list_head *cmd_list = &dcc->discard_cmd_list;
731 struct discard_cmd *dc, *tmp;
732 struct blk_plug plug;
733 int iter = 0;
734repeat:
735 if (kthread_should_stop())
736 return 0;
737
738 blk_start_plug(&plug);
739
740 mutex_lock(&dcc->cmd_lock);
741 list_for_each_entry_safe(dc, tmp, cmd_list, list) {
742 if (dc->state == D_PREP) {
743 dc->state = D_SUBMIT;
744 submit_bio(dc->bio);
dcc9165d 745 atomic_inc(&dcc->submit_discard);
15469963
JK		 746 if (iter++ > DISCARD_ISSUE_RATE)
747 break;
748 } else if (dc->state == D_DONE) {
749 __remove_discard_cmd(sbi, dc);
750 }
751 }
752 mutex_unlock(&dcc->cmd_lock);
753
754 blk_finish_plug(&plug);
755
756 iter = 0;
757 congestion_wait(BLK_RW_SYNC, HZ/50);
758
759 wait_event_interruptible(*q,
760 kthread_should_stop() || !list_empty(&dcc->discard_cmd_list));
761 goto repeat;
762}
763
764
275b66b0 765/* this function is copied from blkdev_issue_discard from block/blk-lib.c */
f46e8809 766static int __f2fs_issue_discard_async(struct f2fs_sb_info *sbi,
3c62be17 767 struct block_device *bdev, block_t blkstart, block_t blklen)
275b66b0 768{
275b66b0 769 struct bio *bio = NULL;
4e6a8d9b 770 block_t lblkstart = blkstart;
275b66b0
CY		 771 int err;
772
d50aaeec 773 trace_f2fs_issue_discard(bdev, blkstart, blklen);
f46e8809 774
3c62be17
JK		 775 if (sbi->s_ndevs) {
776 int devi = f2fs_target_device_index(sbi, blkstart);
777
778 blkstart -= FDEV(devi).start_blk;
779 }
f46e8809
DLM		 780 err = __blkdev_issue_discard(bdev,
781 SECTOR_FROM_BLOCK(blkstart),
782 SECTOR_FROM_BLOCK(blklen),
783 GFP_NOFS, 0, &bio);
275b66b0 784 if (!err && bio) {
b01a9201 785 bio->bi_end_io = f2fs_submit_discard_endio;
275b66b0 786 bio->bi_opf |= REQ_SYNC;
15469963
JK		 787
788 __add_discard_cmd(sbi, bio, lblkstart, blklen);
789 wake_up(&SM_I(sbi)->dcc_info->discard_wait_queue);
275b66b0 790 }
275b66b0
CY		 791 return err;
792}
793
f46e8809 794#ifdef CONFIG_BLK_DEV_ZONED
3c62be17
JK		 795static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
796 struct block_device *bdev, block_t blkstart, block_t blklen)
f46e8809 797{
92592285 798 sector_t sector, nr_sects;
3c62be17
JK		 799 int devi = 0;
800
801 if (sbi->s_ndevs) {
802 devi = f2fs_target_device_index(sbi, blkstart);
803 blkstart -= FDEV(devi).start_blk;
804 }
f46e8809
DLM		 805
806 /*
807 * We need to know the type of the zone: for conventional zones,
808 * use regular discard if the drive supports it. For sequential
809 * zones, reset the zone write pointer.
810 */
3c62be17 811 switch (get_blkz_type(sbi, bdev, blkstart)) {
f46e8809
DLM		 812
813 case BLK_ZONE_TYPE_CONVENTIONAL:
814 if (!blk_queue_discard(bdev_get_queue(bdev)))
815 return 0;
3c62be17 816 return __f2fs_issue_discard_async(sbi, bdev, blkstart, blklen);
f46e8809
DLM		 817 case BLK_ZONE_TYPE_SEQWRITE_REQ:
818 case BLK_ZONE_TYPE_SEQWRITE_PREF:
92592285
JK		 819 sector = SECTOR_FROM_BLOCK(blkstart);
820 nr_sects = SECTOR_FROM_BLOCK(blklen);
821
822 if (sector & (bdev_zone_sectors(bdev) - 1) ||
823 nr_sects != bdev_zone_sectors(bdev)) {
824 f2fs_msg(sbi->sb, KERN_INFO,
825 "(%d) %s: Unaligned discard attempted (block %x + %x)",
826 devi, sbi->s_ndevs ? FDEV(devi).path: "",
827 blkstart, blklen);
828 return -EIO;
829 }
d50aaeec 830 trace_f2fs_issue_reset_zone(bdev, blkstart);
f46e8809
DLM		 831 return blkdev_reset_zones(bdev, sector,
832 nr_sects, GFP_NOFS);
833 default:
834 /* Unknown zone type: broken device ? */
835 return -EIO;
836 }
837}
838#endif
839
3c62be17
JK		 840static int __issue_discard_async(struct f2fs_sb_info *sbi,
841 struct block_device *bdev, block_t blkstart, block_t blklen)
842{
843#ifdef CONFIG_BLK_DEV_ZONED
844 if (f2fs_sb_mounted_blkzoned(sbi->sb) &&
845 bdev_zoned_model(bdev) != BLK_ZONED_NONE)
846 return __f2fs_issue_discard_zone(sbi, bdev, blkstart, blklen);
847#endif
848 return __f2fs_issue_discard_async(sbi, bdev, blkstart, blklen);
849}
850
1e87a78d 851static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
37208879
JK		 852 block_t blkstart, block_t blklen)
853{
3c62be17
JK		 854 sector_t start = blkstart, len = 0;
855 struct block_device *bdev;
a66cdd98
JK		 856 struct seg_entry *se;
857 unsigned int offset;
858 block_t i;
3c62be17
JK		 859 int err = 0;
860
861 bdev = f2fs_target_device(sbi, blkstart, NULL);
862
863 for (i = blkstart; i < blkstart + blklen; i++, len++) {
864 if (i != start) {
865 struct block_device *bdev2 =
866 f2fs_target_device(sbi, i, NULL);
867
868 if (bdev2 != bdev) {
869 err = __issue_discard_async(sbi, bdev,
870 start, len);
871 if (err)
872 return err;
873 bdev = bdev2;
874 start = i;
875 len = 0;
876 }
877 }
a66cdd98 878
a66cdd98
JK		 879 se = get_seg_entry(sbi, GET_SEGNO(sbi, i));
880 offset = GET_BLKOFF_FROM_SEG0(sbi, i);
881
882 if (!f2fs_test_and_set_bit(offset, se->discard_map))
883 sbi->discard_blks--;
884 }
f46e8809 885
3c62be17
JK		 886 if (len)
887 err = __issue_discard_async(sbi, bdev, start, len);
888 return err;
1e87a78d
JK		 889}
890
adf4983b 891static void __add_discard_entry(struct f2fs_sb_info *sbi,
a66cdd98
JK		 892 struct cp_control *cpc, struct seg_entry *se,
893 unsigned int start, unsigned int end)
b2955550 894{
0b54fb84 895 struct list_head *head = &SM_I(sbi)->dcc_info->discard_entry_list;
adf4983b
JK		 896 struct discard_entry *new, *last;
897
898 if (!list_empty(head)) {
899 last = list_last_entry(head, struct discard_entry, list);
900 if (START_BLOCK(sbi, cpc->trim_start) + start ==
ad4d307f
JK		 901 last->blkaddr + last->len &&
902 last->len < MAX_DISCARD_BLOCKS(sbi)) {
adf4983b
JK		 903 last->len += end - start;
904 goto done;
905 }
906 }
907
908 new = f2fs_kmem_cache_alloc(discard_entry_slab, GFP_NOFS);
909 INIT_LIST_HEAD(&new->list);
910 new->blkaddr = START_BLOCK(sbi, cpc->trim_start) + start;
911 new->len = end - start;
912 list_add_tail(&new->list, head);
913done:
0b54fb84 914 SM_I(sbi)->dcc_info->nr_discards += end - start;
adf4983b
JK		 915}
916
25290fa5
JK		 917static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc,
918 bool check_only)
adf4983b 919{
b2955550
JK		 920 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
921 int max_blocks = sbi->blocks_per_seg;
4b2fecc8 922 struct seg_entry *se = get_seg_entry(sbi, cpc->trim_start);
b2955550
JK		 923 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
924 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
a66cdd98 925 unsigned long *discard_map = (unsigned long *)se->discard_map;
60a3b782 926 unsigned long *dmap = SIT_I(sbi)->tmp_map;
b2955550 927 unsigned int start = 0, end = -1;
4b2fecc8 928 bool force = (cpc->reason == CP_DISCARD);
b2955550
JK		 929 int i;
930
3e025740 931 if (se->valid_blocks == max_blocks || !f2fs_discard_en(sbi))
25290fa5 932 return false;
b2955550 933
a66cdd98
JK		 934 if (!force) {
935 if (!test_opt(sbi, DISCARD) || !se->valid_blocks ||
0b54fb84
JK		 936 SM_I(sbi)->dcc_info->nr_discards >=
937 SM_I(sbi)->dcc_info->max_discards)
25290fa5 938 return false;
4b2fecc8
JK		 939 }
940
b2955550
JK		 941 /* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
942 for (i = 0; i < entries; i++)
a66cdd98 943 dmap[i] = force ? ~ckpt_map[i] & ~discard_map[i] :
d7bc2484 944 (cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
b2955550 945
0b54fb84
JK		 946 while (force || SM_I(sbi)->dcc_info->nr_discards <=
947 SM_I(sbi)->dcc_info->max_discards) {
b2955550
JK		 948 start = __find_rev_next_bit(dmap, max_blocks, end + 1);
949 if (start >= max_blocks)
950 break;
951
952 end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1);
c7b41e16
YH		 953 if (force && start && end != max_blocks
954 && (end - start) < cpc->trim_minlen)
955 continue;
956
25290fa5
JK		 957 if (check_only)
958 return true;
959
a66cdd98 960 __add_discard_entry(sbi, cpc, se, start, end);
b2955550 961 }
25290fa5 962 return false;
b2955550
JK		 963}
964
4b2fecc8
JK		 965void release_discard_addrs(struct f2fs_sb_info *sbi)
966{
0b54fb84 967 struct list_head *head = &(SM_I(sbi)->dcc_info->discard_entry_list);
4b2fecc8
JK		 968 struct discard_entry *entry, *this;
969
970 /* drop caches */
971 list_for_each_entry_safe(entry, this, head, list) {
972 list_del(&entry->list);
973 kmem_cache_free(discard_entry_slab, entry);
974 }
975}
976
0a8165d7 977/*
351df4b2
JK		 978 * Should call clear_prefree_segments after checkpoint is done.
979 */
980static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
981{
982 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
b65ee148 983 unsigned int segno;
351df4b2
JK		 984
985 mutex_lock(&dirty_i->seglist_lock);
7cd8558b 986 for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], MAIN_SEGS(sbi))
351df4b2 987 __set_test_and_free(sbi, segno);
351df4b2
JK		 988 mutex_unlock(&dirty_i->seglist_lock);
989}
990
836b5a63 991void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2 992{
0b54fb84 993 struct list_head *head = &(SM_I(sbi)->dcc_info->discard_entry_list);
2d7b822a 994 struct discard_entry *entry, *this;
351df4b2 995 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
29e59c14 996 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
29e59c14 997 unsigned int start = 0, end = -1;
36abef4e 998 unsigned int secno, start_segno;
c24a0fd6 999 bool force = (cpc->reason == CP_DISCARD);
351df4b2
JK		 1000
1001 mutex_lock(&dirty_i->seglist_lock);
29e59c14 1002
351df4b2 1003 while (1) {
29e59c14 1004 int i;
7cd8558b
JK		 1005 start = find_next_bit(prefree_map, MAIN_SEGS(sbi), end + 1);
1006 if (start >= MAIN_SEGS(sbi))
351df4b2 1007 break;
7cd8558b
JK		 1008 end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
1009 start + 1);
29e59c14
CL		 1010
1011 for (i = start; i < end; i++)
1012 clear_bit(i, prefree_map);
1013
1014 dirty_i->nr_dirty[PRE] -= end - start;
1015
650d3c4e 1016 if (!test_opt(sbi, DISCARD))
29e59c14 1017 continue;
351df4b2 1018
650d3c4e
YH		 1019 if (force && start >= cpc->trim_start &&
1020 (end - 1) <= cpc->trim_end)
1021 continue;
1022
36abef4e
JK		 1023 if (!test_opt(sbi, LFS) || sbi->segs_per_sec == 1) {
1024 f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
37208879 1025 (end - start) << sbi->log_blocks_per_seg);
36abef4e
JK		 1026 continue;
1027 }
1028next:
1029 secno = GET_SECNO(sbi, start);
1030 start_segno = secno * sbi->segs_per_sec;
1031 if (!IS_CURSEC(sbi, secno) &&
1032 !get_valid_blocks(sbi, start, sbi->segs_per_sec))
1033 f2fs_issue_discard(sbi, START_BLOCK(sbi, start_segno),
1034 sbi->segs_per_sec << sbi->log_blocks_per_seg);
1035
1036 start = start_segno + sbi->segs_per_sec;
1037 if (start < end)
1038 goto next;
351df4b2
JK		 1039 }
1040 mutex_unlock(&dirty_i->seglist_lock);
b2955550
JK		 1041
1042 /* send small discards */
2d7b822a 1043 list_for_each_entry_safe(entry, this, head, list) {
c24a0fd6 1044 if (force && entry->len < cpc->trim_minlen)
836b5a63 1045 goto skip;
37208879 1046 f2fs_issue_discard(sbi, entry->blkaddr, entry->len);
f56aa1c5 1047 cpc->trimmed += entry->len;
836b5a63 1048skip:
b2955550 1049 list_del(&entry->list);
0b54fb84 1050 SM_I(sbi)->dcc_info->nr_discards -= entry->len;
b2955550
JK		 1051 kmem_cache_free(discard_entry_slab, entry);
1052 }
351df4b2
JK		 1053}
1054
8ed59745 1055static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84 1056{
15469963 1057 dev_t dev = sbi->sb->s_bdev->bd_dev;
0b54fb84
JK		 1058 struct discard_cmd_control *dcc;
1059 int err = 0;
1060
1061 if (SM_I(sbi)->dcc_info) {
1062 dcc = SM_I(sbi)->dcc_info;
1063 goto init_thread;
1064 }
1065
1066 dcc = kzalloc(sizeof(struct discard_cmd_control), GFP_KERNEL);
1067 if (!dcc)
1068 return -ENOMEM;
1069
1070 INIT_LIST_HEAD(&dcc->discard_entry_list);
1071 INIT_LIST_HEAD(&dcc->discard_cmd_list);
15469963 1072 mutex_init(&dcc->cmd_lock);
dcc9165d 1073 atomic_set(&dcc->submit_discard, 0);
0b54fb84
JK		 1074 dcc->nr_discards = 0;
1075 dcc->max_discards = 0;
1076
15469963 1077 init_waitqueue_head(&dcc->discard_wait_queue);
0b54fb84
JK		 1078 SM_I(sbi)->dcc_info = dcc;
1079init_thread:
15469963
JK		 1080 dcc->f2fs_issue_discard = kthread_run(issue_discard_thread, sbi,
1081 "f2fs_discard-%u:%u", MAJOR(dev), MINOR(dev));
1082 if (IS_ERR(dcc->f2fs_issue_discard)) {
1083 err = PTR_ERR(dcc->f2fs_issue_discard);
1084 kfree(dcc);
1085 SM_I(sbi)->dcc_info = NULL;
1086 return err;
1087 }
1088
0b54fb84
JK		 1089 return err;
1090}
1091
8ed59745 1092static void destroy_discard_cmd_control(struct f2fs_sb_info *sbi, bool free)
0b54fb84
JK		 1093{
1094 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1095
15469963
JK		 1096 if (dcc && dcc->f2fs_issue_discard) {
1097 struct task_struct *discard_thread = dcc->f2fs_issue_discard;
1098
1099 dcc->f2fs_issue_discard = NULL;
1100 kthread_stop(discard_thread);
1101 }
0b54fb84
JK		 1102 if (free) {
1103 kfree(dcc);
1104 SM_I(sbi)->dcc_info = NULL;
1105 }
1106}
1107
184a5cd2 1108static bool __mark_sit_entry_dirty(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 1109{
1110 struct sit_info *sit_i = SIT_I(sbi);
184a5cd2
CY		 1111
1112 if (!__test_and_set_bit(segno, sit_i->dirty_sentries_bitmap)) {
351df4b2 1113 sit_i->dirty_sentries++;
184a5cd2
CY		 1114 return false;
1115 }
1116
1117 return true;
351df4b2
JK		 1118}
1119
1120static void __set_sit_entry_type(struct f2fs_sb_info *sbi, int type,
1121 unsigned int segno, int modified)
1122{
1123 struct seg_entry *se = get_seg_entry(sbi, segno);
1124 se->type = type;
1125 if (modified)
1126 __mark_sit_entry_dirty(sbi, segno);
1127}
1128
1129static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
1130{
1131 struct seg_entry *se;
1132 unsigned int segno, offset;
1133 long int new_vblocks;
1134
1135 segno = GET_SEGNO(sbi, blkaddr);
1136
1137 se = get_seg_entry(sbi, segno);
1138 new_vblocks = se->valid_blocks + del;
491c0854 1139 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
351df4b2 1140
9850cf4a 1141 f2fs_bug_on(sbi, (new_vblocks >> (sizeof(unsigned short) << 3) ||
351df4b2
JK		 1142 (new_vblocks > sbi->blocks_per_seg)));
1143
1144 se->valid_blocks = new_vblocks;
1145 se->mtime = get_mtime(sbi);
1146 SIT_I(sbi)->max_mtime = se->mtime;
1147
1148 /* Update valid block bitmap */
1149 if (del > 0) {
355e7891
CY		 1150 if (f2fs_test_and_set_bit(offset, se->cur_valid_map)) {
1151#ifdef CONFIG_F2FS_CHECK_FS
1152 if (f2fs_test_and_set_bit(offset,
1153 se->cur_valid_map_mir))
1154 f2fs_bug_on(sbi, 1);
1155 else
1156 WARN_ON(1);
1157#else
05796763 1158 f2fs_bug_on(sbi, 1);
355e7891
CY		 1159#endif
1160 }
3e025740
JK		 1161 if (f2fs_discard_en(sbi) &&
1162 !f2fs_test_and_set_bit(offset, se->discard_map))
a66cdd98 1163 sbi->discard_blks--;
351df4b2 1164 } else {
355e7891
CY		 1165 if (!f2fs_test_and_clear_bit(offset, se->cur_valid_map)) {
1166#ifdef CONFIG_F2FS_CHECK_FS
1167 if (!f2fs_test_and_clear_bit(offset,
1168 se->cur_valid_map_mir))
1169 f2fs_bug_on(sbi, 1);
1170 else
1171 WARN_ON(1);
1172#else
05796763 1173 f2fs_bug_on(sbi, 1);
355e7891
CY		 1174#endif
1175 }
3e025740
JK		 1176 if (f2fs_discard_en(sbi) &&
1177 f2fs_test_and_clear_bit(offset, se->discard_map))
a66cdd98 1178 sbi->discard_blks++;
351df4b2
JK		 1179 }
1180 if (!f2fs_test_bit(offset, se->ckpt_valid_map))
1181 se->ckpt_valid_blocks += del;
1182
1183 __mark_sit_entry_dirty(sbi, segno);
1184
1185 /* update total number of valid blocks to be written in ckpt area */
1186 SIT_I(sbi)->written_valid_blocks += del;
1187
1188 if (sbi->segs_per_sec > 1)
1189 get_sec_entry(sbi, segno)->valid_blocks += del;
1190}
1191
5e443818 1192void refresh_sit_entry(struct f2fs_sb_info *sbi, block_t old, block_t new)
351df4b2 1193{
5e443818
JK		 1194 update_sit_entry(sbi, new, 1);
1195 if (GET_SEGNO(sbi, old) != NULL_SEGNO)
1196 update_sit_entry(sbi, old, -1);
1197
1198 locate_dirty_segment(sbi, GET_SEGNO(sbi, old));
1199 locate_dirty_segment(sbi, GET_SEGNO(sbi, new));
351df4b2
JK		 1200}
1201
1202void invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
1203{
1204 unsigned int segno = GET_SEGNO(sbi, addr);
1205 struct sit_info *sit_i = SIT_I(sbi);
1206
9850cf4a 1207 f2fs_bug_on(sbi, addr == NULL_ADDR);
351df4b2
JK		 1208 if (addr == NEW_ADDR)
1209 return;
1210
1211 /* add it into sit main buffer */
1212 mutex_lock(&sit_i->sentry_lock);
1213
1214 update_sit_entry(sbi, addr, -1);
1215
1216 /* add it into dirty seglist */
1217 locate_dirty_segment(sbi, segno);
1218
1219 mutex_unlock(&sit_i->sentry_lock);
1220}
1221
6e2c64ad
JK		 1222bool is_checkpointed_data(struct f2fs_sb_info *sbi, block_t blkaddr)
1223{
1224 struct sit_info *sit_i = SIT_I(sbi);
1225 unsigned int segno, offset;
1226 struct seg_entry *se;
1227 bool is_cp = false;
1228
1229 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
1230 return true;
1231
1232 mutex_lock(&sit_i->sentry_lock);
1233
1234 segno = GET_SEGNO(sbi, blkaddr);
1235 se = get_seg_entry(sbi, segno);
1236 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
1237
1238 if (f2fs_test_bit(offset, se->ckpt_valid_map))
1239 is_cp = true;
1240
1241 mutex_unlock(&sit_i->sentry_lock);
1242
1243 return is_cp;
1244}
1245
0a8165d7 1246/*
351df4b2
JK		 1247 * This function should be resided under the curseg_mutex lock
1248 */
1249static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
e79efe3b 1250 struct f2fs_summary *sum)
351df4b2
JK		 1251{
1252 struct curseg_info *curseg = CURSEG_I(sbi, type);
1253 void *addr = curseg->sum_blk;
e79efe3b 1254 addr += curseg->next_blkoff * sizeof(struct f2fs_summary);
351df4b2 1255 memcpy(addr, sum, sizeof(struct f2fs_summary));
351df4b2
JK		 1256}
1257
0a8165d7 1258/*
351df4b2
JK		 1259 * Calculate the number of current summary pages for writing
1260 */
3fa06d7b 1261int npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
351df4b2 1262{
351df4b2 1263 int valid_sum_count = 0;
9a47938b 1264 int i, sum_in_page;
351df4b2
JK		 1265
1266 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1267 if (sbi->ckpt->alloc_type[i] == SSR)
1268 valid_sum_count += sbi->blocks_per_seg;
3fa06d7b
CY		 1269 else {
1270 if (for_ra)
1271 valid_sum_count += le16_to_cpu(
1272 F2FS_CKPT(sbi)->cur_data_blkoff[i]);
1273 else
1274 valid_sum_count += curseg_blkoff(sbi, i);
1275 }
351df4b2
JK		 1276 }
1277
09cbfeaf 1278 sum_in_page = (PAGE_SIZE - 2 * SUM_JOURNAL_SIZE -
9a47938b
FL		 1279 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
1280 if (valid_sum_count <= sum_in_page)
351df4b2 1281 return 1;
9a47938b 1282 else if ((valid_sum_count - sum_in_page) <=
09cbfeaf 1283 (PAGE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
351df4b2
JK		 1284 return 2;
1285 return 3;
1286}
1287
0a8165d7 1288/*
351df4b2
JK		 1289 * Caller should put this summary page
1290 */
1291struct page *get_sum_page(struct f2fs_sb_info *sbi, unsigned int segno)
1292{
1293 return get_meta_page(sbi, GET_SUM_BLOCK(sbi, segno));
1294}
1295
381722d2 1296void update_meta_page(struct f2fs_sb_info *sbi, void *src, block_t blk_addr)
351df4b2
JK		 1297{
1298 struct page *page = grab_meta_page(sbi, blk_addr);
381722d2
CY		 1299 void *dst = page_address(page);
1300
1301 if (src)
09cbfeaf 1302 memcpy(dst, src, PAGE_SIZE);
381722d2 1303 else
09cbfeaf 1304 memset(dst, 0, PAGE_SIZE);
351df4b2
JK		 1305 set_page_dirty(page);
1306 f2fs_put_page(page, 1);
1307}
1308
381722d2
CY		 1309static void write_sum_page(struct f2fs_sb_info *sbi,
1310 struct f2fs_summary_block *sum_blk, block_t blk_addr)
1311{
1312 update_meta_page(sbi, (void *)sum_blk, blk_addr);
1313}
1314
b7ad7512
CY		 1315static void write_current_sum_page(struct f2fs_sb_info *sbi,
1316 int type, block_t blk_addr)
1317{
1318 struct curseg_info *curseg = CURSEG_I(sbi, type);
1319 struct page *page = grab_meta_page(sbi, blk_addr);
1320 struct f2fs_summary_block *src = curseg->sum_blk;
1321 struct f2fs_summary_block *dst;
1322
1323 dst = (struct f2fs_summary_block *)page_address(page);
1324
1325 mutex_lock(&curseg->curseg_mutex);
1326
1327 down_read(&curseg->journal_rwsem);
1328 memcpy(&dst->journal, curseg->journal, SUM_JOURNAL_SIZE);
1329 up_read(&curseg->journal_rwsem);
1330
1331 memcpy(dst->entries, src->entries, SUM_ENTRY_SIZE);
1332 memcpy(&dst->footer, &src->footer, SUM_FOOTER_SIZE);
1333
1334 mutex_unlock(&curseg->curseg_mutex);
1335
1336 set_page_dirty(page);
1337 f2fs_put_page(page, 1);
1338}
1339
0a8165d7 1340/*
351df4b2
JK		 1341 * Find a new segment from the free segments bitmap to right order
1342 * This function should be returned with success, otherwise BUG
1343 */
1344static void get_new_segment(struct f2fs_sb_info *sbi,
1345 unsigned int *newseg, bool new_sec, int dir)
1346{
1347 struct free_segmap_info *free_i = FREE_I(sbi);
351df4b2 1348 unsigned int segno, secno, zoneno;
7cd8558b 1349 unsigned int total_zones = MAIN_SECS(sbi) / sbi->secs_per_zone;
351df4b2
JK		 1350 unsigned int hint = *newseg / sbi->segs_per_sec;
1351 unsigned int old_zoneno = GET_ZONENO_FROM_SEGNO(sbi, *newseg);
1352 unsigned int left_start = hint;
1353 bool init = true;
1354 int go_left = 0;
1355 int i;
1356
1a118ccf 1357 spin_lock(&free_i->segmap_lock);
351df4b2
JK		 1358
1359 if (!new_sec && ((*newseg + 1) % sbi->segs_per_sec)) {
1360 segno = find_next_zero_bit(free_i->free_segmap,
0ab14356
CY		 1361 (hint + 1) * sbi->segs_per_sec, *newseg + 1);
1362 if (segno < (hint + 1) * sbi->segs_per_sec)
351df4b2
JK		 1363 goto got_it;
1364 }
1365find_other_zone:
7cd8558b
JK		 1366 secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
1367 if (secno >= MAIN_SECS(sbi)) {
351df4b2
JK		 1368 if (dir == ALLOC_RIGHT) {
1369 secno = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1370 MAIN_SECS(sbi), 0);
1371 f2fs_bug_on(sbi, secno >= MAIN_SECS(sbi));
351df4b2
JK		 1372 } else {
1373 go_left = 1;
1374 left_start = hint - 1;
1375 }
1376 }
1377 if (go_left == 0)
1378 goto skip_left;
1379
1380 while (test_bit(left_start, free_i->free_secmap)) {
1381 if (left_start > 0) {
1382 left_start--;
1383 continue;
1384 }
1385 left_start = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 1386 MAIN_SECS(sbi), 0);
1387 f2fs_bug_on(sbi, left_start >= MAIN_SECS(sbi));
351df4b2
JK		 1388 break;
1389 }
1390 secno = left_start;
1391skip_left:
1392 hint = secno;
1393 segno = secno * sbi->segs_per_sec;
1394 zoneno = secno / sbi->secs_per_zone;
1395
1396 /* give up on finding another zone */
1397 if (!init)
1398 goto got_it;
1399 if (sbi->secs_per_zone == 1)
1400 goto got_it;
1401 if (zoneno == old_zoneno)
1402 goto got_it;
1403 if (dir == ALLOC_LEFT) {
1404 if (!go_left && zoneno + 1 >= total_zones)
1405 goto got_it;
1406 if (go_left && zoneno == 0)
1407 goto got_it;
1408 }
1409 for (i = 0; i < NR_CURSEG_TYPE; i++)
1410 if (CURSEG_I(sbi, i)->zone == zoneno)
1411 break;
1412
1413 if (i < NR_CURSEG_TYPE) {
1414 /* zone is in user, try another */
1415 if (go_left)
1416 hint = zoneno * sbi->secs_per_zone - 1;
1417 else if (zoneno + 1 >= total_zones)
1418 hint = 0;
1419 else
1420 hint = (zoneno + 1) * sbi->secs_per_zone;
1421 init = false;
1422 goto find_other_zone;
1423 }
1424got_it:
1425 /* set it as dirty segment in free segmap */
9850cf4a 1426 f2fs_bug_on(sbi, test_bit(segno, free_i->free_segmap));
351df4b2
JK		 1427 __set_inuse(sbi, segno);
1428 *newseg = segno;
1a118ccf 1429 spin_unlock(&free_i->segmap_lock);
351df4b2
JK		 1430}
1431
1432static void reset_curseg(struct f2fs_sb_info *sbi, int type, int modified)
1433{
1434 struct curseg_info *curseg = CURSEG_I(sbi, type);
1435 struct summary_footer *sum_footer;
1436
1437 curseg->segno = curseg->next_segno;
1438 curseg->zone = GET_ZONENO_FROM_SEGNO(sbi, curseg->segno);
1439 curseg->next_blkoff = 0;
1440 curseg->next_segno = NULL_SEGNO;
1441
1442 sum_footer = &(curseg->sum_blk->footer);
1443 memset(sum_footer, 0, sizeof(struct summary_footer));
1444 if (IS_DATASEG(type))
1445 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
1446 if (IS_NODESEG(type))
1447 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
1448 __set_sit_entry_type(sbi, type, curseg->segno, modified);
1449}
1450
0a8165d7 1451/*
351df4b2
JK		 1452 * Allocate a current working segment.
1453 * This function always allocates a free segment in LFS manner.
1454 */
1455static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
1456{
1457 struct curseg_info *curseg = CURSEG_I(sbi, type);
1458 unsigned int segno = curseg->segno;
1459 int dir = ALLOC_LEFT;
1460
1461 write_sum_page(sbi, curseg->sum_blk,
81fb5e87 1462 GET_SUM_BLOCK(sbi, segno));
351df4b2
JK		 1463 if (type == CURSEG_WARM_DATA || type == CURSEG_COLD_DATA)
1464 dir = ALLOC_RIGHT;
1465
1466 if (test_opt(sbi, NOHEAP))
1467 dir = ALLOC_RIGHT;
1468
1469 get_new_segment(sbi, &segno, new_sec, dir);
1470 curseg->next_segno = segno;
1471 reset_curseg(sbi, type, 1);
1472 curseg->alloc_type = LFS;
1473}
1474
1475static void __next_free_blkoff(struct f2fs_sb_info *sbi,
1476 struct curseg_info *seg, block_t start)
1477{
1478 struct seg_entry *se = get_seg_entry(sbi, seg->segno);
e81c93cf 1479 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
60a3b782 1480 unsigned long *target_map = SIT_I(sbi)->tmp_map;
e81c93cf
CL		 1481 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
1482 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1483 int i, pos;
1484
1485 for (i = 0; i < entries; i++)
1486 target_map[i] = ckpt_map[i] | cur_map[i];
1487
1488 pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, start);
1489
1490 seg->next_blkoff = pos;
351df4b2
JK		 1491}
1492
0a8165d7 1493/*
351df4b2
JK		 1494 * If a segment is written by LFS manner, next block offset is just obtained
1495 * by increasing the current block offset. However, if a segment is written by
1496 * SSR manner, next block offset obtained by calling __next_free_blkoff
1497 */
1498static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
1499 struct curseg_info *seg)
1500{
1501 if (seg->alloc_type == SSR)
1502 __next_free_blkoff(sbi, seg, seg->next_blkoff + 1);
1503 else
1504 seg->next_blkoff++;
1505}
1506
0a8165d7 1507/*
e1c42045 1508 * This function always allocates a used segment(from dirty seglist) by SSR
351df4b2
JK		 1509 * manner, so it should recover the existing segment information of valid blocks
1510 */
1511static void change_curseg(struct f2fs_sb_info *sbi, int type, bool reuse)
1512{
1513 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
1514 struct curseg_info *curseg = CURSEG_I(sbi, type);
1515 unsigned int new_segno = curseg->next_segno;
1516 struct f2fs_summary_block *sum_node;
1517 struct page *sum_page;
1518
1519 write_sum_page(sbi, curseg->sum_blk,
1520 GET_SUM_BLOCK(sbi, curseg->segno));
1521 __set_test_and_inuse(sbi, new_segno);
1522
1523 mutex_lock(&dirty_i->seglist_lock);
1524 __remove_dirty_segment(sbi, new_segno, PRE);
1525 __remove_dirty_segment(sbi, new_segno, DIRTY);
1526 mutex_unlock(&dirty_i->seglist_lock);
1527
1528 reset_curseg(sbi, type, 1);
1529 curseg->alloc_type = SSR;
1530 __next_free_blkoff(sbi, curseg, 0);
1531
1532 if (reuse) {
1533 sum_page = get_sum_page(sbi, new_segno);
1534 sum_node = (struct f2fs_summary_block *)page_address(sum_page);
1535 memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
1536 f2fs_put_page(sum_page, 1);
1537 }
1538}
1539
43727527
JK		 1540static int get_ssr_segment(struct f2fs_sb_info *sbi, int type)
1541{
1542 struct curseg_info *curseg = CURSEG_I(sbi, type);
1543 const struct victim_selection *v_ops = DIRTY_I(sbi)->v_ops;
d27c3d89
CY		 1544 int i, cnt;
1545 bool reversed = false;
c192f7a4
JK		 1546
1547 /* need_SSR() already forces to do this */
1548 if (v_ops->get_victim(sbi, &(curseg)->next_segno, BG_GC, type, SSR))
1549 return 1;
43727527 1550
70d625cb
JK		 1551 /* For node segments, let's do SSR more intensively */
1552 if (IS_NODESEG(type)) {
d27c3d89
CY		 1553 if (type >= CURSEG_WARM_NODE) {
1554 reversed = true;
1555 i = CURSEG_COLD_NODE;
1556 } else {
1557 i = CURSEG_HOT_NODE;
1558 }
1559 cnt = NR_CURSEG_NODE_TYPE;
70d625cb 1560 } else {
d27c3d89
CY		 1561 if (type >= CURSEG_WARM_DATA) {
1562 reversed = true;
1563 i = CURSEG_COLD_DATA;
1564 } else {
1565 i = CURSEG_HOT_DATA;
1566 }
1567 cnt = NR_CURSEG_DATA_TYPE;
70d625cb 1568 }
43727527 1569
d27c3d89 1570 for (; cnt-- > 0; reversed ? i-- : i++) {
c192f7a4
JK		 1571 if (i == type)
1572 continue;
43727527 1573 if (v_ops->get_victim(sbi, &(curseg)->next_segno,
c192f7a4 1574 BG_GC, i, SSR))
43727527 1575 return 1;
c192f7a4 1576 }
43727527
JK		 1577 return 0;
1578}
1579
351df4b2
JK		 1580/*
1581 * flush out current segment and replace it with new segment
1582 * This function should be returned with success, otherwise BUG
1583 */
1584static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
1585 int type, bool force)
1586{
7b405275 1587 if (force)
351df4b2 1588 new_curseg(sbi, type, true);
5b6c6be2
JK		 1589 else if (!is_set_ckpt_flags(sbi, CP_CRC_RECOVERY_FLAG) &&
1590 type == CURSEG_WARM_NODE)
351df4b2
JK		 1591 new_curseg(sbi, type, false);
1592 else if (need_SSR(sbi) && get_ssr_segment(sbi, type))
1593 change_curseg(sbi, type, true);
1594 else
1595 new_curseg(sbi, type, false);
dcdfff65 1596
d0db7703 1597 stat_inc_seg_type(sbi, CURSEG_I(sbi, type));
351df4b2
JK		 1598}
1599
1600void allocate_new_segments(struct f2fs_sb_info *sbi)
1601{
6ae1be13
JK		 1602 struct curseg_info *curseg;
1603 unsigned int old_segno;
351df4b2
JK		 1604 int i;
1605
6ae1be13
JK		 1606 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1607 curseg = CURSEG_I(sbi, i);
1608 old_segno = curseg->segno;
1609 SIT_I(sbi)->s_ops->allocate_segment(sbi, i, true);
1610 locate_dirty_segment(sbi, old_segno);
1611 }
351df4b2
JK		 1612}
1613
1614static const struct segment_allocation default_salloc_ops = {
1615 .allocate_segment = allocate_segment_by_default,
1616};
1617
25290fa5
JK		 1618bool exist_trim_candidates(struct f2fs_sb_info *sbi, struct cp_control *cpc)
1619{
1620 __u64 trim_start = cpc->trim_start;
1621 bool has_candidate = false;
1622
1623 mutex_lock(&SIT_I(sbi)->sentry_lock);
1624 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++) {
1625 if (add_discard_addrs(sbi, cpc, true)) {
1626 has_candidate = true;
1627 break;
1628 }
1629 }
1630 mutex_unlock(&SIT_I(sbi)->sentry_lock);
1631
1632 cpc->trim_start = trim_start;
1633 return has_candidate;
1634}
1635
4b2fecc8
JK		 1636int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
1637{
f7ef9b83
JK		 1638 __u64 start = F2FS_BYTES_TO_BLK(range->start);
1639 __u64 end = start + F2FS_BYTES_TO_BLK(range->len) - 1;
4b2fecc8
JK		 1640 unsigned int start_segno, end_segno;
1641 struct cp_control cpc;
c34f42e2 1642 int err = 0;
4b2fecc8 1643
836b5a63 1644 if (start >= MAX_BLKADDR(sbi) || range->len < sbi->blocksize)
4b2fecc8
JK		 1645 return -EINVAL;
1646
9bd27ae4 1647 cpc.trimmed = 0;
7cd8558b 1648 if (end <= MAIN_BLKADDR(sbi))
4b2fecc8
JK		 1649 goto out;
1650
ed214a11
YH		 1651 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
1652 f2fs_msg(sbi->sb, KERN_WARNING,
1653 "Found FS corruption, run fsck to fix.");
1654 goto out;
1655 }
1656
4b2fecc8 1657 /* start/end segment number in main_area */
7cd8558b
JK		 1658 start_segno = (start <= MAIN_BLKADDR(sbi)) ? 0 : GET_SEGNO(sbi, start);
1659 end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
1660 GET_SEGNO(sbi, end);
4b2fecc8 1661 cpc.reason = CP_DISCARD;
836b5a63 1662 cpc.trim_minlen = max_t(__u64, 1, F2FS_BYTES_TO_BLK(range->minlen));
4b2fecc8
JK		 1663
1664 /* do checkpoint to issue discard commands safely */
bba681cb
JK		 1665 for (; start_segno <= end_segno; start_segno = cpc.trim_end + 1) {
1666 cpc.trim_start = start_segno;
a66cdd98
JK		 1667
1668 if (sbi->discard_blks == 0)
1669 break;
1670 else if (sbi->discard_blks < BATCHED_TRIM_BLOCKS(sbi))
1671 cpc.trim_end = end_segno;
1672 else
1673 cpc.trim_end = min_t(unsigned int,
1674 rounddown(start_segno +
bba681cb
JK		 1675 BATCHED_TRIM_SEGMENTS(sbi),
1676 sbi->segs_per_sec) - 1, end_segno);
1677
1678 mutex_lock(&sbi->gc_mutex);
c34f42e2 1679 err = write_checkpoint(sbi, &cpc);
bba681cb 1680 mutex_unlock(&sbi->gc_mutex);
e9328353
CY		 1681 if (err)
1682 break;
74fa5f3d
CY		 1683
1684 schedule();
bba681cb 1685 }
4b2fecc8 1686out:
f7ef9b83 1687 range->len = F2FS_BLK_TO_BYTES(cpc.trimmed);
c34f42e2 1688 return err;
4b2fecc8
JK		 1689}
1690
351df4b2
JK		 1691static bool __has_curseg_space(struct f2fs_sb_info *sbi, int type)
1692{
1693 struct curseg_info *curseg = CURSEG_I(sbi, type);
1694 if (curseg->next_blkoff < sbi->blocks_per_seg)
1695 return true;
1696 return false;
1697}
1698
1699static int __get_segment_type_2(struct page *page, enum page_type p_type)
1700{
1701 if (p_type == DATA)
1702 return CURSEG_HOT_DATA;
1703 else
1704 return CURSEG_HOT_NODE;
1705}
1706
1707static int __get_segment_type_4(struct page *page, enum page_type p_type)
1708{
1709 if (p_type == DATA) {
1710 struct inode *inode = page->mapping->host;
1711
1712 if (S_ISDIR(inode->i_mode))
1713 return CURSEG_HOT_DATA;
1714 else
1715 return CURSEG_COLD_DATA;
1716 } else {
a344b9fd
JK		 1717 if (IS_DNODE(page) && is_cold_node(page))
1718 return CURSEG_WARM_NODE;
351df4b2
JK		 1719 else
1720 return CURSEG_COLD_NODE;
1721 }
1722}
1723
1724static int __get_segment_type_6(struct page *page, enum page_type p_type)
1725{
1726 if (p_type == DATA) {
1727 struct inode *inode = page->mapping->host;
1728
1729 if (S_ISDIR(inode->i_mode))
1730 return CURSEG_HOT_DATA;
354a3399 1731 else if (is_cold_data(page) || file_is_cold(inode))
351df4b2
JK		 1732 return CURSEG_COLD_DATA;
1733 else
1734 return CURSEG_WARM_DATA;
1735 } else {
1736 if (IS_DNODE(page))
1737 return is_cold_node(page) ? CURSEG_WARM_NODE :
1738 CURSEG_HOT_NODE;
1739 else
1740 return CURSEG_COLD_NODE;
1741 }
1742}
1743
1744static int __get_segment_type(struct page *page, enum page_type p_type)
1745{
4081363f 1746 switch (F2FS_P_SB(page)->active_logs) {
351df4b2
JK		 1747 case 2:
1748 return __get_segment_type_2(page, p_type);
1749 case 4:
1750 return __get_segment_type_4(page, p_type);
351df4b2 1751 }
12a67146 1752 /* NR_CURSEG_TYPE(6) logs by default */
9850cf4a
JK		 1753 f2fs_bug_on(F2FS_P_SB(page),
1754 F2FS_P_SB(page)->active_logs != NR_CURSEG_TYPE);
12a67146 1755 return __get_segment_type_6(page, p_type);
351df4b2
JK		 1756}
1757
bfad7c2d
JK		 1758void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
1759 block_t old_blkaddr, block_t *new_blkaddr,
1760 struct f2fs_summary *sum, int type)
351df4b2
JK		 1761{
1762 struct sit_info *sit_i = SIT_I(sbi);
6ae1be13 1763 struct curseg_info *curseg = CURSEG_I(sbi, type);
351df4b2
JK		 1764
1765 mutex_lock(&curseg->curseg_mutex);
21cb1d99 1766 mutex_lock(&sit_i->sentry_lock);
351df4b2
JK		 1767
1768 *new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
351df4b2 1769
4e6a8d9b
JK		 1770 f2fs_wait_discard_bio(sbi, *new_blkaddr);
1771
351df4b2
JK		 1772 /*
1773 * __add_sum_entry should be resided under the curseg_mutex
1774 * because, this function updates a summary entry in the
1775 * current summary block.
1776 */
e79efe3b 1777 __add_sum_entry(sbi, type, sum);
351df4b2 1778
351df4b2 1779 __refresh_next_blkoff(sbi, curseg);
dcdfff65
JK		 1780
1781 stat_inc_block_count(sbi, curseg);
351df4b2
JK		 1782
1783 /*
1784 * SIT information should be updated before segment allocation,
1785 * since SSR needs latest valid block information.
1786 */
1787 refresh_sit_entry(sbi, old_blkaddr, *new_blkaddr);
5e443818 1788
3436c4bd
YS		 1789 if (!__has_curseg_space(sbi, type))
1790 sit_i->s_ops->allocate_segment(sbi, type, false);
1791
351df4b2
JK		 1792 mutex_unlock(&sit_i->sentry_lock);
1793
bfad7c2d 1794 if (page && IS_NODESEG(type))
351df4b2
JK		 1795 fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
1796
bfad7c2d
JK		 1797 mutex_unlock(&curseg->curseg_mutex);
1798}
1799
05ca3632 1800static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
bfad7c2d 1801{
05ca3632 1802 int type = __get_segment_type(fio->page, fio->type);
0a595eba 1803 int err;
bfad7c2d 1804
7dfeaa32
JK		 1805 if (fio->type == NODE || fio->type == DATA)
1806 mutex_lock(&fio->sbi->wio_mutex[fio->type]);
0a595eba 1807reallocate:
7a9d7548
CY		 1808 allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
1809 &fio->new_blkaddr, sum, type);
bfad7c2d 1810
351df4b2 1811 /* writeout dirty page into bdev */
0a595eba
JK		 1812 err = f2fs_submit_page_mbio(fio);
1813 if (err == -EAGAIN) {
1814 fio->old_blkaddr = fio->new_blkaddr;
1815 goto reallocate;
1816 }
7dfeaa32
JK		 1817
1818 if (fio->type == NODE || fio->type == DATA)
1819 mutex_unlock(&fio->sbi->wio_mutex[fio->type]);
351df4b2
JK		 1820}
1821
577e3495 1822void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
351df4b2 1823{
458e6197 1824 struct f2fs_io_info fio = {
05ca3632 1825 .sbi = sbi,
458e6197 1826 .type = META,
04d328de 1827 .op = REQ_OP_WRITE,
70fd7614 1828 .op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
7a9d7548
CY		 1829 .old_blkaddr = page->index,
1830 .new_blkaddr = page->index,
05ca3632 1831 .page = page,
4375a336 1832 .encrypted_page = NULL,
458e6197
JK		 1833 };
1834
2b947003 1835 if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
04d328de 1836 fio.op_flags &= ~REQ_META;
2b947003 1837
351df4b2 1838 set_page_writeback(page);
05ca3632 1839 f2fs_submit_page_mbio(&fio);
351df4b2
JK		 1840}
1841
05ca3632 1842void write_node_page(unsigned int nid, struct f2fs_io_info *fio)
351df4b2
JK		 1843{
1844 struct f2fs_summary sum;
05ca3632 1845
351df4b2 1846 set_summary(&sum, nid, 0, 0);
05ca3632 1847 do_write_page(&sum, fio);
351df4b2
JK		 1848}
1849
05ca3632 1850void write_data_page(struct dnode_of_data *dn, struct f2fs_io_info *fio)
351df4b2 1851{
05ca3632 1852 struct f2fs_sb_info *sbi = fio->sbi;
351df4b2
JK		 1853 struct f2fs_summary sum;
1854 struct node_info ni;
1855
9850cf4a 1856 f2fs_bug_on(sbi, dn->data_blkaddr == NULL_ADDR);
351df4b2
JK		 1857 get_node_info(sbi, dn->nid, &ni);
1858 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
05ca3632 1859 do_write_page(&sum, fio);
f28b3434 1860 f2fs_update_data_blkaddr(dn, fio->new_blkaddr);
351df4b2
JK		 1861}
1862
05ca3632 1863void rewrite_data_page(struct f2fs_io_info *fio)
351df4b2 1864{
7a9d7548 1865 fio->new_blkaddr = fio->old_blkaddr;
05ca3632
JK		 1866 stat_inc_inplace_blocks(fio->sbi);
1867 f2fs_submit_page_mbio(fio);
351df4b2
JK		 1868}
1869
4356e48e 1870void __f2fs_replace_block(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
19f106bc 1871 block_t old_blkaddr, block_t new_blkaddr,
28bc106b 1872 bool recover_curseg, bool recover_newaddr)
351df4b2
JK		 1873{
1874 struct sit_info *sit_i = SIT_I(sbi);
1875 struct curseg_info *curseg;
1876 unsigned int segno, old_cursegno;
1877 struct seg_entry *se;
1878 int type;
19f106bc 1879 unsigned short old_blkoff;
351df4b2
JK		 1880
1881 segno = GET_SEGNO(sbi, new_blkaddr);
1882 se = get_seg_entry(sbi, segno);
1883 type = se->type;
1884
19f106bc
CY		 1885 if (!recover_curseg) {
1886 /* for recovery flow */
1887 if (se->valid_blocks == 0 && !IS_CURSEG(sbi, segno)) {
1888 if (old_blkaddr == NULL_ADDR)
1889 type = CURSEG_COLD_DATA;
1890 else
1891 type = CURSEG_WARM_DATA;
1892 }
1893 } else {
1894 if (!IS_CURSEG(sbi, segno))
351df4b2
JK		 1895 type = CURSEG_WARM_DATA;
1896 }
19f106bc 1897
351df4b2
JK		 1898 curseg = CURSEG_I(sbi, type);
1899
1900 mutex_lock(&curseg->curseg_mutex);
1901 mutex_lock(&sit_i->sentry_lock);
1902
1903 old_cursegno = curseg->segno;
19f106bc 1904 old_blkoff = curseg->next_blkoff;
351df4b2
JK		 1905
1906 /* change the current segment */
1907 if (segno != curseg->segno) {
1908 curseg->next_segno = segno;
1909 change_curseg(sbi, type, true);
1910 }
1911
491c0854 1912 curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
e79efe3b 1913 __add_sum_entry(sbi, type, sum);
351df4b2 1914
28bc106b 1915 if (!recover_curseg || recover_newaddr)
6e2c64ad
JK		 1916 update_sit_entry(sbi, new_blkaddr, 1);
1917 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
1918 update_sit_entry(sbi, old_blkaddr, -1);
1919
1920 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
1921 locate_dirty_segment(sbi, GET_SEGNO(sbi, new_blkaddr));
1922
351df4b2 1923 locate_dirty_segment(sbi, old_cursegno);
351df4b2 1924
19f106bc
CY		 1925 if (recover_curseg) {
1926 if (old_cursegno != curseg->segno) {
1927 curseg->next_segno = old_cursegno;
1928 change_curseg(sbi, type, true);
1929 }
1930 curseg->next_blkoff = old_blkoff;
1931 }
1932
351df4b2
JK		 1933 mutex_unlock(&sit_i->sentry_lock);
1934 mutex_unlock(&curseg->curseg_mutex);
1935}
1936
528e3459
CY		 1937void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
1938 block_t old_addr, block_t new_addr,
28bc106b
CY		 1939 unsigned char version, bool recover_curseg,
1940 bool recover_newaddr)
528e3459
CY		 1941{
1942 struct f2fs_summary sum;
1943
1944 set_summary(&sum, dn->nid, dn->ofs_in_node, version);
1945
28bc106b
CY		 1946 __f2fs_replace_block(sbi, &sum, old_addr, new_addr,
1947 recover_curseg, recover_newaddr);
528e3459 1948
f28b3434 1949 f2fs_update_data_blkaddr(dn, new_addr);
528e3459
CY		 1950}
1951
93dfe2ac 1952void f2fs_wait_on_page_writeback(struct page *page,
fec1d657 1953 enum page_type type, bool ordered)
93dfe2ac 1954{
93dfe2ac 1955 if (PageWriteback(page)) {
4081363f
JK		 1956 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1957
942fd319
JK		 1958 f2fs_submit_merged_bio_cond(sbi, page->mapping->host,
1959 0, page->index, type, WRITE);
fec1d657
JK		 1960 if (ordered)
1961 wait_on_page_writeback(page);
1962 else
1963 wait_for_stable_page(page);
93dfe2ac
JK		 1964 }
1965}
1966
08b39fbd
CY		 1967void f2fs_wait_on_encrypted_page_writeback(struct f2fs_sb_info *sbi,
1968 block_t blkaddr)
1969{
1970 struct page *cpage;
1971
5d4c0af4 1972 if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR)
08b39fbd
CY		 1973 return;
1974
08b39fbd
CY		 1975 cpage = find_lock_page(META_MAPPING(sbi), blkaddr);
1976 if (cpage) {
fec1d657 1977 f2fs_wait_on_page_writeback(cpage, DATA, true);
08b39fbd
CY		 1978 f2fs_put_page(cpage, 1);
1979 }
1980}
1981
351df4b2
JK		 1982static int read_compacted_summaries(struct f2fs_sb_info *sbi)
1983{
1984 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1985 struct curseg_info *seg_i;
1986 unsigned char *kaddr;
1987 struct page *page;
1988 block_t start;
1989 int i, j, offset;
1990
1991 start = start_sum_block(sbi);
1992
1993 page = get_meta_page(sbi, start++);
1994 kaddr = (unsigned char *)page_address(page);
1995
1996 /* Step 1: restore nat cache */
1997 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 1998 memcpy(seg_i->journal, kaddr, SUM_JOURNAL_SIZE);
351df4b2
JK		 1999
2000 /* Step 2: restore sit cache */
2001 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2002 memcpy(seg_i->journal, kaddr + SUM_JOURNAL_SIZE, SUM_JOURNAL_SIZE);
351df4b2
JK		 2003 offset = 2 * SUM_JOURNAL_SIZE;
2004
2005 /* Step 3: restore summary entries */
2006 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2007 unsigned short blk_off;
2008 unsigned int segno;
2009
2010 seg_i = CURSEG_I(sbi, i);
2011 segno = le32_to_cpu(ckpt->cur_data_segno[i]);
2012 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[i]);
2013 seg_i->next_segno = segno;
2014 reset_curseg(sbi, i, 0);
2015 seg_i->alloc_type = ckpt->alloc_type[i];
2016 seg_i->next_blkoff = blk_off;
2017
2018 if (seg_i->alloc_type == SSR)
2019 blk_off = sbi->blocks_per_seg;
2020
2021 for (j = 0; j < blk_off; j++) {
2022 struct f2fs_summary *s;
2023 s = (struct f2fs_summary *)(kaddr + offset);
2024 seg_i->sum_blk->entries[j] = *s;
2025 offset += SUMMARY_SIZE;
09cbfeaf 2026 if (offset + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2027 SUM_FOOTER_SIZE)
2028 continue;
2029
2030 f2fs_put_page(page, 1);
2031 page = NULL;
2032
2033 page = get_meta_page(sbi, start++);
2034 kaddr = (unsigned char *)page_address(page);
2035 offset = 0;
2036 }
2037 }
2038 f2fs_put_page(page, 1);
2039 return 0;
2040}
2041
2042static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
2043{
2044 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2045 struct f2fs_summary_block *sum;
2046 struct curseg_info *curseg;
2047 struct page *new;
2048 unsigned short blk_off;
2049 unsigned int segno = 0;
2050 block_t blk_addr = 0;
2051
2052 /* get segment number and block addr */
2053 if (IS_DATASEG(type)) {
2054 segno = le32_to_cpu(ckpt->cur_data_segno[type]);
2055 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[type -
2056 CURSEG_HOT_DATA]);
119ee914 2057 if (__exist_node_summaries(sbi))
351df4b2
JK		 2058 blk_addr = sum_blk_addr(sbi, NR_CURSEG_TYPE, type);
2059 else
2060 blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
2061 } else {
2062 segno = le32_to_cpu(ckpt->cur_node_segno[type -
2063 CURSEG_HOT_NODE]);
2064 blk_off = le16_to_cpu(ckpt->cur_node_blkoff[type -
2065 CURSEG_HOT_NODE]);
119ee914 2066 if (__exist_node_summaries(sbi))
351df4b2
JK		 2067 blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
2068 type - CURSEG_HOT_NODE);
2069 else
2070 blk_addr = GET_SUM_BLOCK(sbi, segno);
2071 }
2072
2073 new = get_meta_page(sbi, blk_addr);
2074 sum = (struct f2fs_summary_block *)page_address(new);
2075
2076 if (IS_NODESEG(type)) {
119ee914 2077 if (__exist_node_summaries(sbi)) {
351df4b2
JK		 2078 struct f2fs_summary *ns = &sum->entries[0];
2079 int i;
2080 for (i = 0; i < sbi->blocks_per_seg; i++, ns++) {
2081 ns->version = 0;
2082 ns->ofs_in_node = 0;
2083 }
2084 } else {
d653788a
GZ		 2085 int err;
2086
2087 err = restore_node_summary(sbi, segno, sum);
2088 if (err) {
351df4b2 2089 f2fs_put_page(new, 1);
d653788a 2090 return err;
351df4b2
JK		 2091 }
2092 }
2093 }
2094
2095 /* set uncompleted segment to curseg */
2096 curseg = CURSEG_I(sbi, type);
2097 mutex_lock(&curseg->curseg_mutex);
b7ad7512
CY		 2098
2099 /* update journal info */
2100 down_write(&curseg->journal_rwsem);
2101 memcpy(curseg->journal, &sum->journal, SUM_JOURNAL_SIZE);
2102 up_write(&curseg->journal_rwsem);
2103
2104 memcpy(curseg->sum_blk->entries, sum->entries, SUM_ENTRY_SIZE);
2105 memcpy(&curseg->sum_blk->footer, &sum->footer, SUM_FOOTER_SIZE);
351df4b2
JK		 2106 curseg->next_segno = segno;
2107 reset_curseg(sbi, type, 0);
2108 curseg->alloc_type = ckpt->alloc_type[type];
2109 curseg->next_blkoff = blk_off;
2110 mutex_unlock(&curseg->curseg_mutex);
2111 f2fs_put_page(new, 1);
2112 return 0;
2113}
2114
2115static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
2116{
2117 int type = CURSEG_HOT_DATA;
e4fc5fbf 2118 int err;
351df4b2 2119
aaec2b1d 2120 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG)) {
3fa06d7b
CY		 2121 int npages = npages_for_summary_flush(sbi, true);
2122
2123 if (npages >= 2)
2124 ra_meta_pages(sbi, start_sum_block(sbi), npages,
26879fb1 2125 META_CP, true);
3fa06d7b 2126
351df4b2
JK		 2127 /* restore for compacted data summary */
2128 if (read_compacted_summaries(sbi))
2129 return -EINVAL;
2130 type = CURSEG_HOT_NODE;
2131 }
2132
119ee914 2133 if (__exist_node_summaries(sbi))
3fa06d7b 2134 ra_meta_pages(sbi, sum_blk_addr(sbi, NR_CURSEG_TYPE, type),
26879fb1 2135 NR_CURSEG_TYPE - type, META_CP, true);
3fa06d7b 2136
e4fc5fbf
CY		 2137 for (; type <= CURSEG_COLD_NODE; type++) {
2138 err = read_normal_summaries(sbi, type);
2139 if (err)
2140 return err;
2141 }
2142
351df4b2
JK		 2143 return 0;
2144}
2145
2146static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
2147{
2148 struct page *page;
2149 unsigned char *kaddr;
2150 struct f2fs_summary *summary;
2151 struct curseg_info *seg_i;
2152 int written_size = 0;
2153 int i, j;
2154
2155 page = grab_meta_page(sbi, blkaddr++);
2156 kaddr = (unsigned char *)page_address(page);
2157
2158 /* Step 1: write nat cache */
2159 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2160 memcpy(kaddr, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2161 written_size += SUM_JOURNAL_SIZE;
2162
2163 /* Step 2: write sit cache */
2164 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2165 memcpy(kaddr + written_size, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 2166 written_size += SUM_JOURNAL_SIZE;
2167
351df4b2
JK		 2168 /* Step 3: write summary entries */
2169 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2170 unsigned short blkoff;
2171 seg_i = CURSEG_I(sbi, i);
2172 if (sbi->ckpt->alloc_type[i] == SSR)
2173 blkoff = sbi->blocks_per_seg;
2174 else
2175 blkoff = curseg_blkoff(sbi, i);
2176
2177 for (j = 0; j < blkoff; j++) {
2178 if (!page) {
2179 page = grab_meta_page(sbi, blkaddr++);
2180 kaddr = (unsigned char *)page_address(page);
2181 written_size = 0;
2182 }
2183 summary = (struct f2fs_summary *)(kaddr + written_size);
2184 *summary = seg_i->sum_blk->entries[j];
2185 written_size += SUMMARY_SIZE;
351df4b2 2186
09cbfeaf 2187 if (written_size + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 2188 SUM_FOOTER_SIZE)
2189 continue;
2190
e8d61a74 2191 set_page_dirty(page);
351df4b2
JK		 2192 f2fs_put_page(page, 1);
2193 page = NULL;
2194 }
2195 }
e8d61a74
CY		 2196 if (page) {
2197 set_page_dirty(page);
351df4b2 2198 f2fs_put_page(page, 1);
e8d61a74 2199 }
351df4b2
JK		 2200}
2201
2202static void write_normal_summaries(struct f2fs_sb_info *sbi,
2203 block_t blkaddr, int type)
2204{
2205 int i, end;
2206 if (IS_DATASEG(type))
2207 end = type + NR_CURSEG_DATA_TYPE;
2208 else
2209 end = type + NR_CURSEG_NODE_TYPE;
2210
b7ad7512
CY		 2211 for (i = type; i < end; i++)
2212 write_current_sum_page(sbi, i, blkaddr + (i - type));
351df4b2
JK		 2213}
2214
2215void write_data_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2216{
aaec2b1d 2217 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG))
351df4b2
JK		 2218 write_compacted_summaries(sbi, start_blk);
2219 else
2220 write_normal_summaries(sbi, start_blk, CURSEG_HOT_DATA);
2221}
2222
2223void write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
2224{
119ee914 2225 write_normal_summaries(sbi, start_blk, CURSEG_HOT_NODE);
351df4b2
JK		 2226}
2227
dfc08a12 2228int lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
351df4b2
JK		 2229 unsigned int val, int alloc)
2230{
2231 int i;
2232
2233 if (type == NAT_JOURNAL) {
dfc08a12
CY		 2234 for (i = 0; i < nats_in_cursum(journal); i++) {
2235 if (le32_to_cpu(nid_in_journal(journal, i)) == val)
351df4b2
JK		 2236 return i;
2237 }
dfc08a12
CY		 2238 if (alloc && __has_cursum_space(journal, 1, NAT_JOURNAL))
2239 return update_nats_in_cursum(journal, 1);
351df4b2 2240 } else if (type == SIT_JOURNAL) {
dfc08a12
CY		 2241 for (i = 0; i < sits_in_cursum(journal); i++)
2242 if (le32_to_cpu(segno_in_journal(journal, i)) == val)
351df4b2 2243 return i;
dfc08a12
CY		 2244 if (alloc && __has_cursum_space(journal, 1, SIT_JOURNAL))
2245 return update_sits_in_cursum(journal, 1);
351df4b2
JK		 2246 }
2247 return -1;
2248}
2249
2250static struct page *get_current_sit_page(struct f2fs_sb_info *sbi,
2251 unsigned int segno)
2252{
2cc22186 2253 return get_meta_page(sbi, current_sit_addr(sbi, segno));
351df4b2
JK		 2254}
2255
2256static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
2257 unsigned int start)
2258{
2259 struct sit_info *sit_i = SIT_I(sbi);
2260 struct page *src_page, *dst_page;
2261 pgoff_t src_off, dst_off;
2262 void *src_addr, *dst_addr;
2263
2264 src_off = current_sit_addr(sbi, start);
2265 dst_off = next_sit_addr(sbi, src_off);
2266
2267 /* get current sit block page without lock */
2268 src_page = get_meta_page(sbi, src_off);
2269 dst_page = grab_meta_page(sbi, dst_off);
9850cf4a 2270 f2fs_bug_on(sbi, PageDirty(src_page));
351df4b2
JK		 2271
2272 src_addr = page_address(src_page);
2273 dst_addr = page_address(dst_page);
09cbfeaf 2274 memcpy(dst_addr, src_addr, PAGE_SIZE);
351df4b2
JK		 2275
2276 set_page_dirty(dst_page);
2277 f2fs_put_page(src_page, 1);
2278
2279 set_to_next_sit(sit_i, start);
2280
2281 return dst_page;
2282}
2283
184a5cd2
CY		 2284static struct sit_entry_set *grab_sit_entry_set(void)
2285{
2286 struct sit_entry_set *ses =
80c54505 2287 f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS);
184a5cd2
CY		 2288
2289 ses->entry_cnt = 0;
2290 INIT_LIST_HEAD(&ses->set_list);
2291 return ses;
2292}
2293
2294static void release_sit_entry_set(struct sit_entry_set *ses)
2295{
2296 list_del(&ses->set_list);
2297 kmem_cache_free(sit_entry_set_slab, ses);
2298}
2299
2300static void adjust_sit_entry_set(struct sit_entry_set *ses,
2301 struct list_head *head)
2302{
2303 struct sit_entry_set *next = ses;
2304
2305 if (list_is_last(&ses->set_list, head))
2306 return;
2307
2308 list_for_each_entry_continue(next, head, set_list)
2309 if (ses->entry_cnt <= next->entry_cnt)
2310 break;
2311
2312 list_move_tail(&ses->set_list, &next->set_list);
2313}
2314
2315static void add_sit_entry(unsigned int segno, struct list_head *head)
2316{
2317 struct sit_entry_set *ses;
2318 unsigned int start_segno = START_SEGNO(segno);
2319
2320 list_for_each_entry(ses, head, set_list) {
2321 if (ses->start_segno == start_segno) {
2322 ses->entry_cnt++;
2323 adjust_sit_entry_set(ses, head);
2324 return;
2325 }
2326 }
2327
2328 ses = grab_sit_entry_set();
2329
2330 ses->start_segno = start_segno;
2331 ses->entry_cnt++;
2332 list_add(&ses->set_list, head);
2333}
2334
2335static void add_sits_in_set(struct f2fs_sb_info *sbi)
2336{
2337 struct f2fs_sm_info *sm_info = SM_I(sbi);
2338 struct list_head *set_list = &sm_info->sit_entry_set;
2339 unsigned long *bitmap = SIT_I(sbi)->dirty_sentries_bitmap;
184a5cd2
CY		 2340 unsigned int segno;
2341
7cd8558b 2342 for_each_set_bit(segno, bitmap, MAIN_SEGS(sbi))
184a5cd2
CY		 2343 add_sit_entry(segno, set_list);
2344}
2345
2346static void remove_sits_in_journal(struct f2fs_sb_info *sbi)
351df4b2
JK		 2347{
2348 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2349 struct f2fs_journal *journal = curseg->journal;
351df4b2
JK		 2350 int i;
2351
b7ad7512 2352 down_write(&curseg->journal_rwsem);
dfc08a12 2353 for (i = 0; i < sits_in_cursum(journal); i++) {
184a5cd2
CY		 2354 unsigned int segno;
2355 bool dirtied;
2356
dfc08a12 2357 segno = le32_to_cpu(segno_in_journal(journal, i));
184a5cd2
CY		 2358 dirtied = __mark_sit_entry_dirty(sbi, segno);
2359
2360 if (!dirtied)
2361 add_sit_entry(segno, &SM_I(sbi)->sit_entry_set);
351df4b2 2362 }
dfc08a12 2363 update_sits_in_cursum(journal, -i);
b7ad7512 2364 up_write(&curseg->journal_rwsem);
351df4b2
JK		 2365}
2366
0a8165d7 2367/*
351df4b2
JK		 2368 * CP calls this function, which flushes SIT entries including sit_journal,
2369 * and moves prefree segs to free segs.
2370 */
4b2fecc8 2371void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2
JK		 2372{
2373 struct sit_info *sit_i = SIT_I(sbi);
2374 unsigned long *bitmap = sit_i->dirty_sentries_bitmap;
2375 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2376 struct f2fs_journal *journal = curseg->journal;
184a5cd2
CY		 2377 struct sit_entry_set *ses, *tmp;
2378 struct list_head *head = &SM_I(sbi)->sit_entry_set;
184a5cd2 2379 bool to_journal = true;
4b2fecc8 2380 struct seg_entry *se;
351df4b2 2381
351df4b2
JK		 2382 mutex_lock(&sit_i->sentry_lock);
2383
2b11a74b
WL		 2384 if (!sit_i->dirty_sentries)
2385 goto out;
2386
351df4b2 2387 /*
184a5cd2
CY		 2388 * add and account sit entries of dirty bitmap in sit entry
2389 * set temporarily
351df4b2 2390 */
184a5cd2 2391 add_sits_in_set(sbi);
351df4b2 2392
184a5cd2
CY		 2393 /*
2394 * if there are no enough space in journal to store dirty sit
2395 * entries, remove all entries from journal and add and account
2396 * them in sit entry set.
2397 */
dfc08a12 2398 if (!__has_cursum_space(journal, sit_i->dirty_sentries, SIT_JOURNAL))
184a5cd2 2399 remove_sits_in_journal(sbi);
b2955550 2400
184a5cd2
CY		 2401 /*
2402 * there are two steps to flush sit entries:
2403 * #1, flush sit entries to journal in current cold data summary block.
2404 * #2, flush sit entries to sit page.
2405 */
2406 list_for_each_entry_safe(ses, tmp, head, set_list) {
4a257ed6 2407 struct page *page = NULL;
184a5cd2
CY		 2408 struct f2fs_sit_block *raw_sit = NULL;
2409 unsigned int start_segno = ses->start_segno;
2410 unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
7cd8558b 2411 (unsigned long)MAIN_SEGS(sbi));
184a5cd2
CY		 2412 unsigned int segno = start_segno;
2413
2414 if (to_journal &&
dfc08a12 2415 !__has_cursum_space(journal, ses->entry_cnt, SIT_JOURNAL))
184a5cd2
CY		 2416 to_journal = false;
2417
b7ad7512
CY		 2418 if (to_journal) {
2419 down_write(&curseg->journal_rwsem);
2420 } else {
184a5cd2
CY		 2421 page = get_next_sit_page(sbi, start_segno);
2422 raw_sit = page_address(page);
351df4b2 2423 }
351df4b2 2424
184a5cd2
CY		 2425 /* flush dirty sit entries in region of current sit set */
2426 for_each_set_bit_from(segno, bitmap, end) {
2427 int offset, sit_offset;
4b2fecc8
JK		 2428
2429 se = get_seg_entry(sbi, segno);
184a5cd2
CY		 2430
2431 /* add discard candidates */
d7bc2484 2432 if (cpc->reason != CP_DISCARD) {
4b2fecc8 2433 cpc->trim_start = segno;
25290fa5 2434 add_discard_addrs(sbi, cpc, false);
4b2fecc8 2435 }
184a5cd2
CY		 2436
2437 if (to_journal) {
dfc08a12 2438 offset = lookup_journal_in_cursum(journal,
184a5cd2
CY		 2439 SIT_JOURNAL, segno, 1);
2440 f2fs_bug_on(sbi, offset < 0);
dfc08a12 2441 segno_in_journal(journal, offset) =
184a5cd2
CY		 2442 cpu_to_le32(segno);
2443 seg_info_to_raw_sit(se,
dfc08a12 2444 &sit_in_journal(journal, offset));
184a5cd2
CY		 2445 } else {
2446 sit_offset = SIT_ENTRY_OFFSET(sit_i, segno);
2447 seg_info_to_raw_sit(se,
2448 &raw_sit->entries[sit_offset]);
2449 }
351df4b2 2450
184a5cd2
CY		 2451 __clear_bit(segno, bitmap);
2452 sit_i->dirty_sentries--;
2453 ses->entry_cnt--;
351df4b2
JK		 2454 }
2455
b7ad7512
CY		 2456 if (to_journal)
2457 up_write(&curseg->journal_rwsem);
2458 else
184a5cd2
CY		 2459 f2fs_put_page(page, 1);
2460
2461 f2fs_bug_on(sbi, ses->entry_cnt);
2462 release_sit_entry_set(ses);
351df4b2 2463 }
184a5cd2
CY		 2464
2465 f2fs_bug_on(sbi, !list_empty(head));
2466 f2fs_bug_on(sbi, sit_i->dirty_sentries);
184a5cd2 2467out:
4b2fecc8 2468 if (cpc->reason == CP_DISCARD) {
650d3c4e
YH		 2469 __u64 trim_start = cpc->trim_start;
2470
4b2fecc8 2471 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++)
25290fa5 2472 add_discard_addrs(sbi, cpc, false);
650d3c4e
YH		 2473
2474 cpc->trim_start = trim_start;
4b2fecc8 2475 }
351df4b2 2476 mutex_unlock(&sit_i->sentry_lock);
351df4b2 2477
351df4b2
JK		 2478 set_prefree_as_free_segments(sbi);
2479}
2480
2481static int build_sit_info(struct f2fs_sb_info *sbi)
2482{
2483 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
351df4b2
JK		 2484 struct sit_info *sit_i;
2485 unsigned int sit_segs, start;
ae27d62e 2486 char *src_bitmap;
351df4b2
JK		 2487 unsigned int bitmap_size;
2488
2489 /* allocate memory for SIT information */
2490 sit_i = kzalloc(sizeof(struct sit_info), GFP_KERNEL);
2491 if (!sit_i)
2492 return -ENOMEM;
2493
2494 SM_I(sbi)->sit_info = sit_i;
2495
39307a8e
JK		 2496 sit_i->sentries = f2fs_kvzalloc(MAIN_SEGS(sbi) *
2497 sizeof(struct seg_entry), GFP_KERNEL);
351df4b2
JK		 2498 if (!sit_i->sentries)
2499 return -ENOMEM;
2500
7cd8558b 2501 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2502 sit_i->dirty_sentries_bitmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2503 if (!sit_i->dirty_sentries_bitmap)
2504 return -ENOMEM;
2505
7cd8558b 2506 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 2507 sit_i->sentries[start].cur_valid_map
2508 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2509 sit_i->sentries[start].ckpt_valid_map
2510 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
a66cdd98 2511 if (!sit_i->sentries[start].cur_valid_map ||
3e025740 2512 !sit_i->sentries[start].ckpt_valid_map)
351df4b2 2513 return -ENOMEM;
3e025740 2514
355e7891
CY		 2515#ifdef CONFIG_F2FS_CHECK_FS
2516 sit_i->sentries[start].cur_valid_map_mir
2517 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2518 if (!sit_i->sentries[start].cur_valid_map_mir)
2519 return -ENOMEM;
2520#endif
2521
3e025740
JK		 2522 if (f2fs_discard_en(sbi)) {
2523 sit_i->sentries[start].discard_map
2524 = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2525 if (!sit_i->sentries[start].discard_map)
2526 return -ENOMEM;
2527 }
351df4b2
JK		 2528 }
2529
60a3b782
JK		 2530 sit_i->tmp_map = kzalloc(SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
2531 if (!sit_i->tmp_map)
2532 return -ENOMEM;
2533
351df4b2 2534 if (sbi->segs_per_sec > 1) {
39307a8e
JK		 2535 sit_i->sec_entries = f2fs_kvzalloc(MAIN_SECS(sbi) *
2536 sizeof(struct sec_entry), GFP_KERNEL);
351df4b2
JK		 2537 if (!sit_i->sec_entries)
2538 return -ENOMEM;
2539 }
2540
2541 /* get information related with SIT */
2542 sit_segs = le32_to_cpu(raw_super->segment_count_sit) >> 1;
2543
2544 /* setup SIT bitmap from ckeckpoint pack */
2545 bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
2546 src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
2547
ae27d62e
CY		 2548 sit_i->sit_bitmap = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2549 if (!sit_i->sit_bitmap)
351df4b2 2550 return -ENOMEM;
351df4b2 2551
ae27d62e
CY		 2552#ifdef CONFIG_F2FS_CHECK_FS
2553 sit_i->sit_bitmap_mir = kmemdup(src_bitmap, bitmap_size, GFP_KERNEL);
2554 if (!sit_i->sit_bitmap_mir)
2555 return -ENOMEM;
2556#endif
2557
351df4b2
JK		 2558 /* init SIT information */
2559 sit_i->s_ops = &default_salloc_ops;
2560
2561 sit_i->sit_base_addr = le32_to_cpu(raw_super->sit_blkaddr);
2562 sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
c79b7ff1 2563 sit_i->written_valid_blocks = 0;
351df4b2
JK		 2564 sit_i->bitmap_size = bitmap_size;
2565 sit_i->dirty_sentries = 0;
2566 sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
2567 sit_i->elapsed_time = le64_to_cpu(sbi->ckpt->elapsed_time);
2568 sit_i->mounted_time = CURRENT_TIME_SEC.tv_sec;
2569 mutex_init(&sit_i->sentry_lock);
2570 return 0;
2571}
2572
2573static int build_free_segmap(struct f2fs_sb_info *sbi)
2574{
351df4b2
JK		 2575 struct free_segmap_info *free_i;
2576 unsigned int bitmap_size, sec_bitmap_size;
2577
2578 /* allocate memory for free segmap information */
2579 free_i = kzalloc(sizeof(struct free_segmap_info), GFP_KERNEL);
2580 if (!free_i)
2581 return -ENOMEM;
2582
2583 SM_I(sbi)->free_info = free_i;
2584
7cd8558b 2585 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
39307a8e 2586 free_i->free_segmap = f2fs_kvmalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2587 if (!free_i->free_segmap)
2588 return -ENOMEM;
2589
7cd8558b 2590 sec_bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
39307a8e 2591 free_i->free_secmap = f2fs_kvmalloc(sec_bitmap_size, GFP_KERNEL);
351df4b2
JK		 2592 if (!free_i->free_secmap)
2593 return -ENOMEM;
2594
2595 /* set all segments as dirty temporarily */
2596 memset(free_i->free_segmap, 0xff, bitmap_size);
2597 memset(free_i->free_secmap, 0xff, sec_bitmap_size);
2598
2599 /* init free segmap information */
7cd8558b 2600 free_i->start_segno = GET_SEGNO_FROM_SEG0(sbi, MAIN_BLKADDR(sbi));
351df4b2
JK		 2601 free_i->free_segments = 0;
2602 free_i->free_sections = 0;
1a118ccf 2603 spin_lock_init(&free_i->segmap_lock);
351df4b2
JK		 2604 return 0;
2605}
2606
2607static int build_curseg(struct f2fs_sb_info *sbi)
2608{
1042d60f 2609 struct curseg_info *array;
351df4b2
JK		 2610 int i;
2611
b434babf 2612 array = kcalloc(NR_CURSEG_TYPE, sizeof(*array), GFP_KERNEL);
351df4b2
JK		 2613 if (!array)
2614 return -ENOMEM;
2615
2616 SM_I(sbi)->curseg_array = array;
2617
2618 for (i = 0; i < NR_CURSEG_TYPE; i++) {
2619 mutex_init(&array[i].curseg_mutex);
09cbfeaf 2620 array[i].sum_blk = kzalloc(PAGE_SIZE, GFP_KERNEL);
351df4b2
JK		 2621 if (!array[i].sum_blk)
2622 return -ENOMEM;
b7ad7512
CY		 2623 init_rwsem(&array[i].journal_rwsem);
2624 array[i].journal = kzalloc(sizeof(struct f2fs_journal),
2625 GFP_KERNEL);
2626 if (!array[i].journal)
2627 return -ENOMEM;
351df4b2
JK		 2628 array[i].segno = NULL_SEGNO;
2629 array[i].next_blkoff = 0;
2630 }
2631 return restore_curseg_summaries(sbi);
2632}
2633
2634static void build_sit_entries(struct f2fs_sb_info *sbi)
2635{
2636 struct sit_info *sit_i = SIT_I(sbi);
2637 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 2638 struct f2fs_journal *journal = curseg->journal;
9c094040
YH		 2639 struct seg_entry *se;
2640 struct f2fs_sit_entry sit;
74de593a
CY		 2641 int sit_blk_cnt = SIT_BLK_CNT(sbi);
2642 unsigned int i, start, end;
2643 unsigned int readed, start_blk = 0;
351df4b2 2644
74de593a 2645 do {
664ba972
JK		 2646 readed = ra_meta_pages(sbi, start_blk, BIO_MAX_PAGES,
2647 META_SIT, true);
74de593a
CY		 2648
2649 start = start_blk * sit_i->sents_per_block;
2650 end = (start_blk + readed) * sit_i->sents_per_block;
2651
7cd8558b 2652 for (; start < end && start < MAIN_SEGS(sbi); start++) {
74de593a 2653 struct f2fs_sit_block *sit_blk;
74de593a
CY		 2654 struct page *page;
2655
9c094040 2656 se = &sit_i->sentries[start];
74de593a
CY		 2657 page = get_current_sit_page(sbi, start);
2658 sit_blk = (struct f2fs_sit_block *)page_address(page);
2659 sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
2660 f2fs_put_page(page, 1);
d600af23 2661
74de593a
CY		 2662 check_block_count(sbi, start, &sit);
2663 seg_info_from_raw_sit(se, &sit);
a66cdd98
JK		 2664
2665 /* build discard map only one time */
3e025740
JK		 2666 if (f2fs_discard_en(sbi)) {
2667 memcpy(se->discard_map, se->cur_valid_map,
2668 SIT_VBLOCK_MAP_SIZE);
2669 sbi->discard_blks += sbi->blocks_per_seg -
2670 se->valid_blocks;
2671 }
a66cdd98 2672
d600af23
CY		 2673 if (sbi->segs_per_sec > 1)
2674 get_sec_entry(sbi, start)->valid_blocks +=
2675 se->valid_blocks;
351df4b2 2676 }
74de593a
CY		 2677 start_blk += readed;
2678 } while (start_blk < sit_blk_cnt);
d600af23
CY		 2679
2680 down_read(&curseg->journal_rwsem);
2681 for (i = 0; i < sits_in_cursum(journal); i++) {
d600af23
CY		 2682 unsigned int old_valid_blocks;
2683
2684 start = le32_to_cpu(segno_in_journal(journal, i));
2685 se = &sit_i->sentries[start];
2686 sit = sit_in_journal(journal, i);
2687
2688 old_valid_blocks = se->valid_blocks;
2689
2690 check_block_count(sbi, start, &sit);
2691 seg_info_from_raw_sit(se, &sit);
2692
2693 if (f2fs_discard_en(sbi)) {
2694 memcpy(se->discard_map, se->cur_valid_map,
2695 SIT_VBLOCK_MAP_SIZE);
2696 sbi->discard_blks += old_valid_blocks -
2697 se->valid_blocks;
2698 }
2699
2700 if (sbi->segs_per_sec > 1)
2701 get_sec_entry(sbi, start)->valid_blocks +=
2702 se->valid_blocks - old_valid_blocks;
2703 }
2704 up_read(&curseg->journal_rwsem);
351df4b2
JK		 2705}
2706
2707static void init_free_segmap(struct f2fs_sb_info *sbi)
2708{
2709 unsigned int start;
2710 int type;
2711
7cd8558b 2712 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2
JK		 2713 struct seg_entry *sentry = get_seg_entry(sbi, start);
2714 if (!sentry->valid_blocks)
2715 __set_free(sbi, start);
c79b7ff1
JK		 2716 else
2717 SIT_I(sbi)->written_valid_blocks +=
2718 sentry->valid_blocks;
351df4b2
JK		 2719 }
2720
2721 /* set use the current segments */
2722 for (type = CURSEG_HOT_DATA; type <= CURSEG_COLD_NODE; type++) {
2723 struct curseg_info *curseg_t = CURSEG_I(sbi, type);
2724 __set_test_and_inuse(sbi, curseg_t->segno);
2725 }
2726}
2727
2728static void init_dirty_segmap(struct f2fs_sb_info *sbi)
2729{
2730 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
2731 struct free_segmap_info *free_i = FREE_I(sbi);
7cd8558b 2732 unsigned int segno = 0, offset = 0;
351df4b2
JK		 2733 unsigned short valid_blocks;
2734
8736fbf0 2735 while (1) {
351df4b2 2736 /* find dirty segment based on free segmap */
7cd8558b
JK		 2737 segno = find_next_inuse(free_i, MAIN_SEGS(sbi), offset);
2738 if (segno >= MAIN_SEGS(sbi))
351df4b2
JK		 2739 break;
2740 offset = segno + 1;
2741 valid_blocks = get_valid_blocks(sbi, segno, 0);
ec325b52 2742 if (valid_blocks == sbi->blocks_per_seg || !valid_blocks)
351df4b2 2743 continue;
ec325b52
JK		 2744 if (valid_blocks > sbi->blocks_per_seg) {
2745 f2fs_bug_on(sbi, 1);
2746 continue;
2747 }
351df4b2
JK		 2748 mutex_lock(&dirty_i->seglist_lock);
2749 __locate_dirty_segment(sbi, segno, DIRTY);
2750 mutex_unlock(&dirty_i->seglist_lock);
2751 }
2752}
2753
5ec4e49f 2754static int init_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 2755{
2756 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
7cd8558b 2757 unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
351df4b2 2758
39307a8e 2759 dirty_i->victim_secmap = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
5ec4e49f 2760 if (!dirty_i->victim_secmap)
351df4b2
JK		 2761 return -ENOMEM;
2762 return 0;
2763}
2764
2765static int build_dirty_segmap(struct f2fs_sb_info *sbi)
2766{
2767 struct dirty_seglist_info *dirty_i;
2768 unsigned int bitmap_size, i;
2769
2770 /* allocate memory for dirty segments list information */
2771 dirty_i = kzalloc(sizeof(struct dirty_seglist_info), GFP_KERNEL);
2772 if (!dirty_i)
2773 return -ENOMEM;
2774
2775 SM_I(sbi)->dirty_info = dirty_i;
2776 mutex_init(&dirty_i->seglist_lock);
2777
7cd8558b 2778 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
351df4b2
JK		 2779
2780 for (i = 0; i < NR_DIRTY_TYPE; i++) {
39307a8e 2781 dirty_i->dirty_segmap[i] = f2fs_kvzalloc(bitmap_size, GFP_KERNEL);
351df4b2
JK		 2782 if (!dirty_i->dirty_segmap[i])
2783 return -ENOMEM;
2784 }
2785
2786 init_dirty_segmap(sbi);
5ec4e49f 2787 return init_victim_secmap(sbi);
351df4b2
JK		 2788}
2789
0a8165d7 2790/*
351df4b2
JK		 2791 * Update min, max modified time for cost-benefit GC algorithm
2792 */
2793static void init_min_max_mtime(struct f2fs_sb_info *sbi)
2794{
2795 struct sit_info *sit_i = SIT_I(sbi);
2796 unsigned int segno;
2797
2798 mutex_lock(&sit_i->sentry_lock);
2799
2800 sit_i->min_mtime = LLONG_MAX;
2801
7cd8558b 2802 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
351df4b2
JK		 2803 unsigned int i;
2804 unsigned long long mtime = 0;
2805
2806 for (i = 0; i < sbi->segs_per_sec; i++)
2807 mtime += get_seg_entry(sbi, segno + i)->mtime;
2808
2809 mtime = div_u64(mtime, sbi->segs_per_sec);
2810
2811 if (sit_i->min_mtime > mtime)
2812 sit_i->min_mtime = mtime;
2813 }
2814 sit_i->max_mtime = get_mtime(sbi);
2815 mutex_unlock(&sit_i->sentry_lock);
2816}
2817
2818int build_segment_manager(struct f2fs_sb_info *sbi)
2819{
2820 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2821 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1042d60f 2822 struct f2fs_sm_info *sm_info;
351df4b2
JK		 2823 int err;
2824
2825 sm_info = kzalloc(sizeof(struct f2fs_sm_info), GFP_KERNEL);
2826 if (!sm_info)
2827 return -ENOMEM;
2828
2829 /* init sm info */
2830 sbi->sm_info = sm_info;
351df4b2
JK		 2831 sm_info->seg0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2832 sm_info->main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2833 sm_info->segment_count = le32_to_cpu(raw_super->segment_count);
2834 sm_info->reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2835 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2836 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
2837 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
58c41035
JK		 2838 sm_info->rec_prefree_segments = sm_info->main_segments *
2839 DEF_RECLAIM_PREFREE_SEGMENTS / 100;
44a83499
JK		 2840 if (sm_info->rec_prefree_segments > DEF_MAX_RECLAIM_PREFREE_SEGMENTS)
2841 sm_info->rec_prefree_segments = DEF_MAX_RECLAIM_PREFREE_SEGMENTS;
2842
52763a4b
JK		 2843 if (!test_opt(sbi, LFS))
2844 sm_info->ipu_policy = 1 << F2FS_IPU_FSYNC;
216fbd64 2845 sm_info->min_ipu_util = DEF_MIN_IPU_UTIL;
c1ce1b02 2846 sm_info->min_fsync_blocks = DEF_MIN_FSYNC_BLOCKS;
351df4b2 2847
bba681cb
JK		 2848 sm_info->trim_sections = DEF_BATCHED_TRIM_SECTIONS;
2849
184a5cd2
CY		 2850 INIT_LIST_HEAD(&sm_info->sit_entry_set);
2851
b270ad6f 2852 if (test_opt(sbi, FLUSH_MERGE) && !f2fs_readonly(sbi->sb)) {
2163d198
GZ		 2853 err = create_flush_cmd_control(sbi);
2854 if (err)
a688b9d9 2855 return err;
6b4afdd7
JK		 2856 }
2857
0b54fb84
JK		 2858 err = create_discard_cmd_control(sbi);
2859 if (err)
2860 return err;
2861
351df4b2
JK		 2862 err = build_sit_info(sbi);
2863 if (err)
2864 return err;
2865 err = build_free_segmap(sbi);
2866 if (err)
2867 return err;
2868 err = build_curseg(sbi);
2869 if (err)
2870 return err;
2871
2872 /* reinit free segmap based on SIT */
2873 build_sit_entries(sbi);
2874
2875 init_free_segmap(sbi);
2876 err = build_dirty_segmap(sbi);
2877 if (err)
2878 return err;
2879
2880 init_min_max_mtime(sbi);
2881 return 0;
2882}
2883
2884static void discard_dirty_segmap(struct f2fs_sb_info *sbi,
2885 enum dirty_type dirty_type)
2886{
2887 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
2888
2889 mutex_lock(&dirty_i->seglist_lock);
39307a8e 2890 kvfree(dirty_i->dirty_segmap[dirty_type]);
351df4b2
JK		 2891 dirty_i->nr_dirty[dirty_type] = 0;
2892 mutex_unlock(&dirty_i->seglist_lock);
2893}
2894
5ec4e49f 2895static void destroy_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 2896{
2897 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
39307a8e 2898 kvfree(dirty_i->victim_secmap);
351df4b2
JK		 2899}
2900
2901static void destroy_dirty_segmap(struct f2fs_sb_info *sbi)
2902{
2903 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
2904 int i;
2905
2906 if (!dirty_i)
2907 return;
2908
2909 /* discard pre-free/dirty segments list */
2910 for (i = 0; i < NR_DIRTY_TYPE; i++)
2911 discard_dirty_segmap(sbi, i);
2912
5ec4e49f 2913 destroy_victim_secmap(sbi);
351df4b2
JK		 2914 SM_I(sbi)->dirty_info = NULL;
2915 kfree(dirty_i);
2916}
2917
2918static void destroy_curseg(struct f2fs_sb_info *sbi)
2919{
2920 struct curseg_info *array = SM_I(sbi)->curseg_array;
2921 int i;
2922
2923 if (!array)
2924 return;
2925 SM_I(sbi)->curseg_array = NULL;
b7ad7512 2926 for (i = 0; i < NR_CURSEG_TYPE; i++) {
351df4b2 2927 kfree(array[i].sum_blk);
b7ad7512
CY		 2928 kfree(array[i].journal);
2929 }
351df4b2
JK		 2930 kfree(array);
2931}
2932
2933static void destroy_free_segmap(struct f2fs_sb_info *sbi)
2934{
2935 struct free_segmap_info *free_i = SM_I(sbi)->free_info;
2936 if (!free_i)
2937 return;
2938 SM_I(sbi)->free_info = NULL;
39307a8e
JK		 2939 kvfree(free_i->free_segmap);
2940 kvfree(free_i->free_secmap);
351df4b2
JK		 2941 kfree(free_i);
2942}
2943
2944static void destroy_sit_info(struct f2fs_sb_info *sbi)
2945{
2946 struct sit_info *sit_i = SIT_I(sbi);
2947 unsigned int start;
2948
2949 if (!sit_i)
2950 return;
2951
2952 if (sit_i->sentries) {
7cd8558b 2953 for (start = 0; start < MAIN_SEGS(sbi); start++) {
351df4b2 2954 kfree(sit_i->sentries[start].cur_valid_map);
355e7891
CY		 2955#ifdef CONFIG_F2FS_CHECK_FS
2956 kfree(sit_i->sentries[start].cur_valid_map_mir);
2957#endif
351df4b2 2958 kfree(sit_i->sentries[start].ckpt_valid_map);
a66cdd98 2959 kfree(sit_i->sentries[start].discard_map);
351df4b2
JK		 2960 }
2961 }
60a3b782
JK		 2962 kfree(sit_i->tmp_map);
2963
39307a8e
JK		 2964 kvfree(sit_i->sentries);
2965 kvfree(sit_i->sec_entries);
2966 kvfree(sit_i->dirty_sentries_bitmap);
351df4b2
JK		 2967
2968 SM_I(sbi)->sit_info = NULL;
2969 kfree(sit_i->sit_bitmap);
ae27d62e
CY		 2970#ifdef CONFIG_F2FS_CHECK_FS
2971 kfree(sit_i->sit_bitmap_mir);
2972#endif
351df4b2
JK		 2973 kfree(sit_i);
2974}
2975
2976void destroy_segment_manager(struct f2fs_sb_info *sbi)
2977{
2978 struct f2fs_sm_info *sm_info = SM_I(sbi);
a688b9d9 2979
3b03f724
CY		 2980 if (!sm_info)
2981 return;
5eba8c5d 2982 destroy_flush_cmd_control(sbi, true);
0b54fb84 2983 destroy_discard_cmd_control(sbi, true);
351df4b2
JK		 2984 destroy_dirty_segmap(sbi);
2985 destroy_curseg(sbi);
2986 destroy_free_segmap(sbi);
2987 destroy_sit_info(sbi);
2988 sbi->sm_info = NULL;
2989 kfree(sm_info);
2990}
7fd9e544
JK		 2991
2992int __init create_segment_manager_caches(void)
2993{
2994 discard_entry_slab = f2fs_kmem_cache_create("discard_entry",
e8512d2e 2995 sizeof(struct discard_entry));
7fd9e544 2996 if (!discard_entry_slab)
184a5cd2
CY		 2997 goto fail;
2998
b01a9201
JK		 2999 discard_cmd_slab = f2fs_kmem_cache_create("discard_cmd",
3000 sizeof(struct discard_cmd));
3001 if (!discard_cmd_slab)
6ab2a308 3002 goto destroy_discard_entry;
275b66b0 3003
184a5cd2 3004 sit_entry_set_slab = f2fs_kmem_cache_create("sit_entry_set",
c9ee0085 3005 sizeof(struct sit_entry_set));
184a5cd2 3006 if (!sit_entry_set_slab)
b01a9201 3007 goto destroy_discard_cmd;
88b88a66
JK		 3008
3009 inmem_entry_slab = f2fs_kmem_cache_create("inmem_page_entry",
3010 sizeof(struct inmem_pages));
3011 if (!inmem_entry_slab)
3012 goto destroy_sit_entry_set;
7fd9e544 3013 return 0;
184a5cd2 3014
88b88a66
JK		 3015destroy_sit_entry_set:
3016 kmem_cache_destroy(sit_entry_set_slab);
b01a9201
JK		 3017destroy_discard_cmd:
3018 kmem_cache_destroy(discard_cmd_slab);
6ab2a308 3019destroy_discard_entry:
184a5cd2
CY		 3020 kmem_cache_destroy(discard_entry_slab);
3021fail:
3022 return -ENOMEM;
7fd9e544
JK		 3023}
3024
3025void destroy_segment_manager_caches(void)
3026{
184a5cd2 3027 kmem_cache_destroy(sit_entry_set_slab);
b01a9201 3028 kmem_cache_destroy(discard_cmd_slab);
7fd9e544 3029 kmem_cache_destroy(discard_entry_slab);
88b88a66 3030 kmem_cache_destroy(inmem_entry_slab);
7fd9e544 3031}
Ubuntu Jammy Linux kernel mirror