]> 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: remove buffer_head which has 32bits limit
[mirror_ubuntu-jammy-kernel.git] / fs / f2fs / segment.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
351df4b2
JK		 3 * fs/f2fs/segment.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
351df4b2
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/bio.h>
11#include <linux/blkdev.h>
690e4a3e 12#include <linux/prefetch.h>
6b4afdd7 13#include <linux/kthread.h>
74de593a 14#include <linux/swap.h>
60b99b48 15#include <linux/timer.h>
1d7be270 16#include <linux/freezer.h>
1eb1ef4a 17#include <linux/sched/signal.h>
351df4b2
JK		 18
19#include "f2fs.h"
20#include "segment.h"
21#include "node.h"
5f656541 22#include "gc.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
4d57b86d 169bool f2fs_need_SSR(struct f2fs_sb_info *sbi)
b3a97a2a
JK		 170{
171 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
172 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
173 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
174
b0332a0f 175 if (f2fs_lfs_mode(sbi))
b3a97a2a 176 return false;
0e5e8111 177 if (sbi->gc_mode == GC_URGENT_HIGH)
b3a97a2a 178 return true;
4354994f
DR		 179 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
180 return true;
b3a97a2a
JK		 181
182 return free_sections(sbi) <= (node_secs + 2 * dent_secs + imeta_secs +
a2a12b67 183 SM_I(sbi)->min_ssr_sections + reserved_sections(sbi));
b3a97a2a
JK		 184}
185
4d57b86d 186void f2fs_register_inmem_page(struct inode *inode, struct page *page)
88b88a66 187{
88b88a66 188 struct inmem_pages *new;
9be32d72 189
9e4ded3f 190 f2fs_trace_pid(page);
0722b101 191
e90027d2 192 f2fs_set_page_private(page, ATOMIC_WRITTEN_PAGE);
decd36b6 193
88b88a66
JK		 194 new = f2fs_kmem_cache_alloc(inmem_entry_slab, GFP_NOFS);
195
196 /* add atomic page indices to the list */
197 new->page = page;
198 INIT_LIST_HEAD(&new->list);
decd36b6 199
88b88a66 200 /* increase reference count with clean state */
88b88a66 201 get_page(page);
743b620c
JK		 202 mutex_lock(&F2FS_I(inode)->inmem_lock);
203 list_add_tail(&new->list, &F2FS_I(inode)->inmem_pages);
8dcf2ff7 204 inc_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
743b620c 205 mutex_unlock(&F2FS_I(inode)->inmem_lock);
8ce67cb0
JK		 206
207 trace_f2fs_register_inmem_page(page, INMEM);
88b88a66
JK		 208}
209
28bc106b 210static int __revoke_inmem_pages(struct inode *inode,
48432984
CY		 211 struct list_head *head, bool drop, bool recover,
212 bool trylock)
29b96b54 213{
28bc106b 214 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
29b96b54 215 struct inmem_pages *cur, *tmp;
28bc106b 216 int err = 0;
29b96b54
CY		 217
218 list_for_each_entry_safe(cur, tmp, head, list) {
28bc106b
CY		 219 struct page *page = cur->page;
220
221 if (drop)
222 trace_f2fs_commit_inmem_page(page, INMEM_DROP);
223
48432984
CY		 224 if (trylock) {
225 /*
226 * to avoid deadlock in between page lock and
227 * inmem_lock.
228 */
229 if (!trylock_page(page))
230 continue;
231 } else {
232 lock_page(page);
233 }
29b96b54 234
bae0ee7a 235 f2fs_wait_on_page_writeback(page, DATA, true, true);
e5e5732d 236
28bc106b
CY		 237 if (recover) {
238 struct dnode_of_data dn;
239 struct node_info ni;
240
241 trace_f2fs_commit_inmem_page(page, INMEM_REVOKE);
7f2b4e8e 242retry:
28bc106b 243 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d
CY		 244 err = f2fs_get_dnode_of_data(&dn, page->index,
245 LOOKUP_NODE);
7f2b4e8e
CY		 246 if (err) {
247 if (err == -ENOMEM) {
5df7731f
CY		 248 congestion_wait(BLK_RW_ASYNC,
249 DEFAULT_IO_TIMEOUT);
7f2b4e8e
CY		 250 cond_resched();
251 goto retry;
252 }
28bc106b
CY		 253 err = -EAGAIN;
254 goto next;
255 }
7735730d
CY		 256
257 err = f2fs_get_node_info(sbi, dn.nid, &ni);
258 if (err) {
259 f2fs_put_dnode(&dn);
260 return err;
261 }
262
f1d2564a 263 if (cur->old_addr == NEW_ADDR) {
4d57b86d 264 f2fs_invalidate_blocks(sbi, dn.data_blkaddr);
f1d2564a
DJ		 265 f2fs_update_data_blkaddr(&dn, NEW_ADDR);
266 } else
267 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
28bc106b
CY		 268 cur->old_addr, ni.version, true, true);
269 f2fs_put_dnode(&dn);
270 }
271next:
63c52d78 272 /* we don't need to invalidate this in the sccessful status */
2baf0781 273 if (drop || recover) {
63c52d78 274 ClearPageUptodate(page);
2baf0781
CY		 275 clear_cold_data(page);
276 }
240a5915 277 f2fs_clear_page_private(page);
28bc106b 278 f2fs_put_page(page, 1);
29b96b54
CY		 279
280 list_del(&cur->list);
281 kmem_cache_free(inmem_entry_slab, cur);
282 dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
283 }
28bc106b 284 return err;
29b96b54
CY		 285}
286
4d57b86d 287void f2fs_drop_inmem_pages_all(struct f2fs_sb_info *sbi, bool gc_failure)
57864ae5
JK		 288{
289 struct list_head *head = &sbi->inode_list[ATOMIC_FILE];
290 struct inode *inode;
291 struct f2fs_inode_info *fi;
677017d1
ST		 292 unsigned int count = sbi->atomic_files;
293 unsigned int looped = 0;
57864ae5
JK		 294next:
295 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
296 if (list_empty(head)) {
297 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
298 return;
299 }
300 fi = list_first_entry(head, struct f2fs_inode_info, inmem_ilist);
301 inode = igrab(&fi->vfs_inode);
677017d1
ST		 302 if (inode)
303 list_move_tail(&fi->inmem_ilist, head);
57864ae5
JK		 304 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
305
306 if (inode) {
2ef79ecb 307 if (gc_failure) {
677017d1
ST		 308 if (!fi->i_gc_failures[GC_FAILURE_ATOMIC])
309 goto skip;
2ef79ecb 310 }
2ef79ecb 311 set_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
4d57b86d 312 f2fs_drop_inmem_pages(inode);
677017d1 313skip:
57864ae5
JK		 314 iput(inode);
315 }
5df7731f 316 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
57864ae5 317 cond_resched();
677017d1
ST		 318 if (gc_failure) {
319 if (++looped >= count)
320 return;
321 }
57864ae5
JK		 322 goto next;
323}
324
4d57b86d 325void f2fs_drop_inmem_pages(struct inode *inode)
29b96b54 326{
57864ae5 327 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
29b96b54
CY		 328 struct f2fs_inode_info *fi = F2FS_I(inode);
329
48432984
CY		 330 while (!list_empty(&fi->inmem_pages)) {
331 mutex_lock(&fi->inmem_lock);
332 __revoke_inmem_pages(inode, &fi->inmem_pages,
333 true, false, true);
48432984
CY		 334 mutex_unlock(&fi->inmem_lock);
335 }
5fe45743 336
2ef79ecb 337 fi->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
743b620c
JK		 338
339 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
340 if (!list_empty(&fi->inmem_ilist))
341 list_del_init(&fi->inmem_ilist);
677017d1
ST		 342 if (f2fs_is_atomic_file(inode)) {
343 clear_inode_flag(inode, FI_ATOMIC_FILE);
344 sbi->atomic_files--;
345 }
743b620c 346 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
29b96b54
CY		 347}
348
4d57b86d 349void f2fs_drop_inmem_page(struct inode *inode, struct page *page)
8c242db9
JK		 350{
351 struct f2fs_inode_info *fi = F2FS_I(inode);
352 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
353 struct list_head *head = &fi->inmem_pages;
354 struct inmem_pages *cur = NULL;
355
356 f2fs_bug_on(sbi, !IS_ATOMIC_WRITTEN_PAGE(page));
357
358 mutex_lock(&fi->inmem_lock);
359 list_for_each_entry(cur, head, list) {
360 if (cur->page == page)
361 break;
362 }
363
d0891e84 364 f2fs_bug_on(sbi, list_empty(head) || cur->page != page);
8c242db9
JK		 365 list_del(&cur->list);
366 mutex_unlock(&fi->inmem_lock);
367
368 dec_page_count(sbi, F2FS_INMEM_PAGES);
369 kmem_cache_free(inmem_entry_slab, cur);
370
371 ClearPageUptodate(page);
240a5915 372 f2fs_clear_page_private(page);
8c242db9
JK		 373 f2fs_put_page(page, 0);
374
375 trace_f2fs_commit_inmem_page(page, INMEM_INVALIDATE);
376}
377
4d57b86d 378static int __f2fs_commit_inmem_pages(struct inode *inode)
88b88a66
JK		 379{
380 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
381 struct f2fs_inode_info *fi = F2FS_I(inode);
382 struct inmem_pages *cur, *tmp;
88b88a66 383 struct f2fs_io_info fio = {
05ca3632 384 .sbi = sbi,
39d787be 385 .ino = inode->i_ino,
88b88a66 386 .type = DATA,
04d328de 387 .op = REQ_OP_WRITE,
70fd7614 388 .op_flags = REQ_SYNC | REQ_PRIO,
b0af6d49 389 .io_type = FS_DATA_IO,
88b88a66 390 };
cf52b27a 391 struct list_head revoke_list;
bab475c5 392 bool submit_bio = false;
edb27dee 393 int err = 0;
88b88a66 394
cf52b27a
CY		 395 INIT_LIST_HEAD(&revoke_list);
396
88b88a66 397 list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
28bc106b
CY		 398 struct page *page = cur->page;
399
400 lock_page(page);
401 if (page->mapping == inode->i_mapping) {
402 trace_f2fs_commit_inmem_page(page, INMEM);
403
bae0ee7a 404 f2fs_wait_on_page_writeback(page, DATA, true, true);
8d64d365
CY		 405
406 set_page_dirty(page);
933439c8 407 if (clear_page_dirty_for_io(page)) {
29b96b54 408 inode_dec_dirty_pages(inode);
4d57b86d 409 f2fs_remove_dirty_inode(inode);
933439c8 410 }
640cc189 411retry:
28bc106b 412 fio.page = page;
e959c8f5 413 fio.old_blkaddr = NULL_ADDR;
4d978078 414 fio.encrypted_page = NULL;
cc15620b 415 fio.need_lock = LOCK_DONE;
4d57b86d 416 err = f2fs_do_write_data_page(&fio);
29b96b54 417 if (err) {
640cc189 418 if (err == -ENOMEM) {
5df7731f
CY		 419 congestion_wait(BLK_RW_ASYNC,
420 DEFAULT_IO_TIMEOUT);
640cc189
JK		 421 cond_resched();
422 goto retry;
423 }
28bc106b 424 unlock_page(page);
29b96b54 425 break;
70c640b1 426 }
28bc106b
CY		 427 /* record old blkaddr for revoking */
428 cur->old_addr = fio.old_blkaddr;
bab475c5 429 submit_bio = true;
28bc106b
CY		 430 }
431 unlock_page(page);
cf52b27a 432 list_move_tail(&cur->list, &revoke_list);
88b88a66 433 }
29b96b54 434
bab475c5
CY		 435 if (submit_bio)
436 f2fs_submit_merged_write_cond(sbi, inode, NULL, 0, DATA);
28bc106b 437
cf52b27a
CY		 438 if (err) {
439 /*
440 * try to revoke all committed pages, but still we could fail
441 * due to no memory or other reason, if that happened, EAGAIN
442 * will be returned, which means in such case, transaction is
443 * already not integrity, caller should use journal to do the
444 * recovery or rewrite & commit last transaction. For other
445 * error number, revoking was done by filesystem itself.
446 */
48432984
CY		 447 err = __revoke_inmem_pages(inode, &revoke_list,
448 false, true, false);
cf52b27a
CY		 449
450 /* drop all uncommitted pages */
48432984
CY		 451 __revoke_inmem_pages(inode, &fi->inmem_pages,
452 true, false, false);
cf52b27a 453 } else {
48432984
CY		 454 __revoke_inmem_pages(inode, &revoke_list,
455 false, false, false);
cf52b27a 456 }
28bc106b 457
29b96b54
CY		 458 return err;
459}
460
4d57b86d 461int f2fs_commit_inmem_pages(struct inode *inode)
29b96b54
CY		 462{
463 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
464 struct f2fs_inode_info *fi = F2FS_I(inode);
28bc106b 465 int err;
29b96b54
CY		 466
467 f2fs_balance_fs(sbi, true);
29b96b54 468
6f8d4455
JK		 469 down_write(&fi->i_gc_rwsem[WRITE]);
470
471 f2fs_lock_op(sbi);
5fe45743
CY		 472 set_inode_flag(inode, FI_ATOMIC_COMMIT);
473
29b96b54 474 mutex_lock(&fi->inmem_lock);
4d57b86d 475 err = __f2fs_commit_inmem_pages(inode);
88b88a66
JK		 476 mutex_unlock(&fi->inmem_lock);
477
5fe45743
CY		 478 clear_inode_flag(inode, FI_ATOMIC_COMMIT);
479
29b96b54 480 f2fs_unlock_op(sbi);
6f8d4455
JK		 481 up_write(&fi->i_gc_rwsem[WRITE]);
482
edb27dee 483 return err;
88b88a66
JK		 484}
485
0a8165d7 486/*
351df4b2
JK		 487 * This function balances dirty node and dentry pages.
488 * In addition, it controls garbage collection.
489 */
2c4db1a6 490void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
351df4b2 491{
55523519 492 if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
c45d6002 493 f2fs_show_injection_info(sbi, FAULT_CHECKPOINT);
0f348028 494 f2fs_stop_checkpoint(sbi, false);
55523519 495 }
0f348028 496
e589c2c4 497 /* balance_fs_bg is able to be pending */
a7881893 498 if (need && excess_cached_nats(sbi))
7bcd0cfa 499 f2fs_balance_fs_bg(sbi, false);
e589c2c4 500
00e09c0b 501 if (!f2fs_is_checkpoint_ready(sbi))
4354994f
DR		 502 return;
503
351df4b2 504 /*
029cd28c
JK		 505 * We should do GC or end up with checkpoint, if there are so many dirty
506 * dir/node pages without enough free segments.
351df4b2 507 */
7f3037a5 508 if (has_not_enough_free_secs(sbi, 0, 0)) {
fb24fea7 509 down_write(&sbi->gc_lock);
e066b83c 510 f2fs_gc(sbi, false, false, NULL_SEGNO);
351df4b2
JK		 511 }
512}
513
7bcd0cfa 514void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi, bool from_bg)
4660f9c0 515{
64c74a7a
CY		 516 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
517 return;
518
1dcc336b 519 /* try to shrink extent cache when there is no enough memory */
4d57b86d 520 if (!f2fs_available_free_memory(sbi, EXTENT_CACHE))
554df79e 521 f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER);
1dcc336b 522
1b38dc8e 523 /* check the # of cached NAT entries */
4d57b86d
CY		 524 if (!f2fs_available_free_memory(sbi, NAT_ENTRIES))
525 f2fs_try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK);
1b38dc8e 526
4d57b86d
CY		 527 if (!f2fs_available_free_memory(sbi, FREE_NIDS))
528 f2fs_try_to_free_nids(sbi, MAX_FREE_NIDS);
ad4edb83 529 else
4d57b86d 530 f2fs_build_free_nids(sbi, false, false);
31696580 531
493720a4
CY		 532 if (excess_dirty_nats(sbi) || excess_dirty_nodes(sbi) ||
533 excess_prefree_segs(sbi))
534 goto do_sync;
535
536 /* there is background inflight IO or foreground operation recently */
537 if (is_inflight_io(sbi, REQ_TIME) ||
538 (!f2fs_time_over(sbi, REQ_TIME) && rwsem_is_locked(&sbi->cp_rwsem)))
f455c8a5 539 return;
31696580 540
493720a4
CY		 541 /* exceed periodical checkpoint timeout threshold */
542 if (f2fs_time_over(sbi, CP_TIME))
543 goto do_sync;
544
1b38dc8e 545 /* checkpoint is the only way to shrink partial cached entries */
493720a4
CY		 546 if (f2fs_available_free_memory(sbi, NAT_ENTRIES) ||
547 f2fs_available_free_memory(sbi, INO_ENTRIES))
548 return;
549
550do_sync:
551 if (test_opt(sbi, DATA_FLUSH) && from_bg) {
552 struct blk_plug plug;
553
554 mutex_lock(&sbi->flush_lock);
555
556 blk_start_plug(&plug);
557 f2fs_sync_dirty_inodes(sbi, FILE_INODE);
558 blk_finish_plug(&plug);
559
560 mutex_unlock(&sbi->flush_lock);
36b35a0d 561 }
493720a4
CY		 562 f2fs_sync_fs(sbi->sb, true);
563 stat_inc_bg_cp_count(sbi->stat_info);
4660f9c0
JK		 564}
565
20fda56b
KM		 566static int __submit_flush_wait(struct f2fs_sb_info *sbi,
567 struct block_device *bdev)
3c62be17 568{
dc37910d 569 struct bio *bio;
3c62be17
JK		 570 int ret;
571
dc37910d
CY		 572 bio = f2fs_bio_alloc(sbi, 0, false);
573 if (!bio)
574 return -ENOMEM;
575
3adc5fcb 576 bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH;
74d46992 577 bio_set_dev(bio, bdev);
3c62be17
JK		 578 ret = submit_bio_wait(bio);
579 bio_put(bio);
20fda56b
KM		 580
581 trace_f2fs_issue_flush(bdev, test_opt(sbi, NOBARRIER),
582 test_opt(sbi, FLUSH_MERGE), ret);
3c62be17
JK		 583 return ret;
584}
585
39d787be 586static int submit_flush_wait(struct f2fs_sb_info *sbi, nid_t ino)
3c62be17 587{
39d787be 588 int ret = 0;
3c62be17
JK		 589 int i;
590
0916878d 591 if (!f2fs_is_multi_device(sbi))
39d787be 592 return __submit_flush_wait(sbi, sbi->sb->s_bdev);
20fda56b 593
39d787be 594 for (i = 0; i < sbi->s_ndevs; i++) {
4d57b86d 595 if (!f2fs_is_dirty_device(sbi, ino, i, FLUSH_INO))
39d787be 596 continue;
20fda56b
KM		 597 ret = __submit_flush_wait(sbi, FDEV(i).bdev);
598 if (ret)
599 break;
3c62be17
JK		 600 }
601 return ret;
602}
603
2163d198 604static int issue_flush_thread(void *data)
6b4afdd7
JK		 605{
606 struct f2fs_sb_info *sbi = data;
b01a9201 607 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
a688b9d9 608 wait_queue_head_t *q = &fcc->flush_wait_queue;
6b4afdd7
JK		 609repeat:
610 if (kthread_should_stop())
611 return 0;
612
dc6febb6
CY		 613 sb_start_intwrite(sbi->sb);
614
721bd4d5 615 if (!llist_empty(&fcc->issue_list)) {
6b4afdd7
JK		 616 struct flush_cmd *cmd, *next;
617 int ret;
618
721bd4d5
GZ		 619 fcc->dispatch_list = llist_del_all(&fcc->issue_list);
620 fcc->dispatch_list = llist_reverse_order(fcc->dispatch_list);
621
39d787be
CY		 622 cmd = llist_entry(fcc->dispatch_list, struct flush_cmd, llnode);
623
624 ret = submit_flush_wait(sbi, cmd->ino);
8b8dd65f
CY		 625 atomic_inc(&fcc->issued_flush);
626
721bd4d5
GZ		 627 llist_for_each_entry_safe(cmd, next,
628 fcc->dispatch_list, llnode) {
6b4afdd7 629 cmd->ret = ret;
6b4afdd7
JK		 630 complete(&cmd->wait);
631 }
a688b9d9 632 fcc->dispatch_list = NULL;
6b4afdd7
JK		 633 }
634
dc6febb6
CY		 635 sb_end_intwrite(sbi->sb);
636
a688b9d9 637 wait_event_interruptible(*q,
721bd4d5 638 kthread_should_stop() || !llist_empty(&fcc->issue_list));
6b4afdd7
JK		 639 goto repeat;
640}
641
39d787be 642int f2fs_issue_flush(struct f2fs_sb_info *sbi, nid_t ino)
6b4afdd7 643{
b01a9201 644 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
adf8d90b 645 struct flush_cmd cmd;
8b8dd65f 646 int ret;
6b4afdd7 647
0f7b2abd
JK		 648 if (test_opt(sbi, NOBARRIER))
649 return 0;
650
8b8dd65f 651 if (!test_opt(sbi, FLUSH_MERGE)) {
72691af6 652 atomic_inc(&fcc->queued_flush);
39d787be 653 ret = submit_flush_wait(sbi, ino);
72691af6 654 atomic_dec(&fcc->queued_flush);
8b8dd65f
CY		 655 atomic_inc(&fcc->issued_flush);
656 return ret;
657 }
740432f8 658
0916878d
DLM		 659 if (atomic_inc_return(&fcc->queued_flush) == 1 ||
660 f2fs_is_multi_device(sbi)) {
39d787be 661 ret = submit_flush_wait(sbi, ino);
72691af6 662 atomic_dec(&fcc->queued_flush);
8b8dd65f
CY		 663
664 atomic_inc(&fcc->issued_flush);
740432f8
JK		 665 return ret;
666 }
6b4afdd7 667
39d787be 668 cmd.ino = ino;
adf8d90b 669 init_completion(&cmd.wait);
6b4afdd7 670
721bd4d5 671 llist_add(&cmd.llnode, &fcc->issue_list);
6b4afdd7 672
6f890df0
CY		 673 /* update issue_list before we wake up issue_flush thread */
674 smp_mb();
675
676 if (waitqueue_active(&fcc->flush_wait_queue))
a688b9d9 677 wake_up(&fcc->flush_wait_queue);
6b4afdd7 678
5eba8c5d
JK		 679 if (fcc->f2fs_issue_flush) {
680 wait_for_completion(&cmd.wait);
72691af6 681 atomic_dec(&fcc->queued_flush);
5eba8c5d 682 } else {
d3238691
CY		 683 struct llist_node *list;
684
685 list = llist_del_all(&fcc->issue_list);
686 if (!list) {
687 wait_for_completion(&cmd.wait);
72691af6 688 atomic_dec(&fcc->queued_flush);
d3238691
CY		 689 } else {
690 struct flush_cmd *tmp, *next;
691
39d787be 692 ret = submit_flush_wait(sbi, ino);
d3238691
CY		 693
694 llist_for_each_entry_safe(tmp, next, list, llnode) {
695 if (tmp == &cmd) {
696 cmd.ret = ret;
72691af6 697 atomic_dec(&fcc->queued_flush);
d3238691
CY		 698 continue;
699 }
700 tmp->ret = ret;
701 complete(&tmp->wait);
702 }
703 }
5eba8c5d 704 }
adf8d90b
CY		 705
706 return cmd.ret;
6b4afdd7
JK		 707}
708
4d57b86d 709int f2fs_create_flush_cmd_control(struct f2fs_sb_info *sbi)
2163d198
GZ		 710{
711 dev_t dev = sbi->sb->s_bdev->bd_dev;
712 struct flush_cmd_control *fcc;
713 int err = 0;
714
b01a9201
JK		 715 if (SM_I(sbi)->fcc_info) {
716 fcc = SM_I(sbi)->fcc_info;
d871cd04
YS		 717 if (fcc->f2fs_issue_flush)
718 return err;
5eba8c5d
JK		 719 goto init_thread;
720 }
721
acbf054d 722 fcc = f2fs_kzalloc(sbi, sizeof(struct flush_cmd_control), GFP_KERNEL);
2163d198
GZ		 723 if (!fcc)
724 return -ENOMEM;
8b8dd65f 725 atomic_set(&fcc->issued_flush, 0);
72691af6 726 atomic_set(&fcc->queued_flush, 0);
2163d198 727 init_waitqueue_head(&fcc->flush_wait_queue);
721bd4d5 728 init_llist_head(&fcc->issue_list);
b01a9201 729 SM_I(sbi)->fcc_info = fcc;
d4fdf8ba
YH		 730 if (!test_opt(sbi, FLUSH_MERGE))
731 return err;
732
5eba8c5d 733init_thread:
2163d198
GZ		 734 fcc->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi,
735 "f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev));
736 if (IS_ERR(fcc->f2fs_issue_flush)) {
737 err = PTR_ERR(fcc->f2fs_issue_flush);
c8eb7024 738 kfree(fcc);
b01a9201 739 SM_I(sbi)->fcc_info = NULL;
2163d198
GZ		 740 return err;
741 }
2163d198
GZ		 742
743 return err;
744}
745
4d57b86d 746void f2fs_destroy_flush_cmd_control(struct f2fs_sb_info *sbi, bool free)
2163d198 747{
b01a9201 748 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
2163d198 749
5eba8c5d
JK		 750 if (fcc && fcc->f2fs_issue_flush) {
751 struct task_struct *flush_thread = fcc->f2fs_issue_flush;
752
753 fcc->f2fs_issue_flush = NULL;
754 kthread_stop(flush_thread);
755 }
756 if (free) {
c8eb7024 757 kfree(fcc);
b01a9201 758 SM_I(sbi)->fcc_info = NULL;
5eba8c5d 759 }
2163d198
GZ		 760}
761
1228b482
CY		 762int f2fs_flush_device_cache(struct f2fs_sb_info *sbi)
763{
764 int ret = 0, i;
765
0916878d 766 if (!f2fs_is_multi_device(sbi))
1228b482
CY		 767 return 0;
768
6ed29fe1
CY		 769 if (test_opt(sbi, NOBARRIER))
770 return 0;
771
1228b482
CY		 772 for (i = 1; i < sbi->s_ndevs; i++) {
773 if (!f2fs_test_bit(i, (char *)&sbi->dirty_device))
774 continue;
775 ret = __submit_flush_wait(sbi, FDEV(i).bdev);
776 if (ret)
777 break;
778
779 spin_lock(&sbi->dev_lock);
780 f2fs_clear_bit(i, (char *)&sbi->dirty_device);
781 spin_unlock(&sbi->dev_lock);
782 }
783
784 return ret;
785}
786
351df4b2
JK		 787static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
788 enum dirty_type dirty_type)
789{
790 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
791
792 /* need not be added */
793 if (IS_CURSEG(sbi, segno))
794 return;
795
796 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[dirty_type]))
797 dirty_i->nr_dirty[dirty_type]++;
798
799 if (dirty_type == DIRTY) {
800 struct seg_entry *sentry = get_seg_entry(sbi, segno);
4625d6aa 801 enum dirty_type t = sentry->type;
b2f2c390 802
ec325b52
JK		 803 if (unlikely(t >= DIRTY)) {
804 f2fs_bug_on(sbi, 1);
805 return;
806 }
4625d6aa
CL		 807 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[t]))
808 dirty_i->nr_dirty[t]++;
da52f8ad
JQ		 809
810 if (__is_large_section(sbi)) {
811 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
123aaf77 812 block_t valid_blocks =
da52f8ad
JQ		 813 get_valid_blocks(sbi, segno, true);
814
815 f2fs_bug_on(sbi, unlikely(!valid_blocks ||
816 valid_blocks == BLKS_PER_SEC(sbi)));
817
818 if (!IS_CURSEC(sbi, secno))
819 set_bit(secno, dirty_i->dirty_secmap);
820 }
351df4b2
JK		 821 }
822}
823
824static void __remove_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
825 enum dirty_type dirty_type)
826{
827 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
123aaf77 828 block_t valid_blocks;
351df4b2
JK		 829
830 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[dirty_type]))
831 dirty_i->nr_dirty[dirty_type]--;
832
833 if (dirty_type == DIRTY) {
4625d6aa
CL		 834 struct seg_entry *sentry = get_seg_entry(sbi, segno);
835 enum dirty_type t = sentry->type;
836
837 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
838 dirty_i->nr_dirty[t]--;
b2f2c390 839
da52f8ad
JQ		 840 valid_blocks = get_valid_blocks(sbi, segno, true);
841 if (valid_blocks == 0) {
4ddb1a4d 842 clear_bit(GET_SEC_FROM_SEG(sbi, segno),
5ec4e49f 843 dirty_i->victim_secmap);
bbf9f7d9
ST		 844#ifdef CONFIG_F2FS_CHECK_FS
845 clear_bit(segno, SIT_I(sbi)->invalid_segmap);
846#endif
847 }
da52f8ad
JQ		 848 if (__is_large_section(sbi)) {
849 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
850
851 if (!valid_blocks ||
852 valid_blocks == BLKS_PER_SEC(sbi)) {
853 clear_bit(secno, dirty_i->dirty_secmap);
854 return;
855 }
856
857 if (!IS_CURSEC(sbi, secno))
858 set_bit(secno, dirty_i->dirty_secmap);
859 }
351df4b2
JK		 860 }
861}
862
0a8165d7 863/*
351df4b2
JK		 864 * Should not occur error such as -ENOMEM.
865 * Adding dirty entry into seglist is not critical operation.
866 * If a given segment is one of current working segments, it won't be added.
867 */
8d8451af 868static void locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 869{
870 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
4354994f 871 unsigned short valid_blocks, ckpt_valid_blocks;
de881df9 872 unsigned int usable_blocks;
351df4b2
JK		 873
874 if (segno == NULL_SEGNO || IS_CURSEG(sbi, segno))
875 return;
876
de881df9 877 usable_blocks = f2fs_usable_blks_in_seg(sbi, segno);
351df4b2
JK		 878 mutex_lock(&dirty_i->seglist_lock);
879
302bd348 880 valid_blocks = get_valid_blocks(sbi, segno, false);
4354994f 881 ckpt_valid_blocks = get_ckpt_valid_blocks(sbi, segno);
351df4b2 882
4354994f 883 if (valid_blocks == 0 && (!is_sbi_flag_set(sbi, SBI_CP_DISABLED) ||
de881df9 884 ckpt_valid_blocks == usable_blocks)) {
351df4b2
JK		 885 __locate_dirty_segment(sbi, segno, PRE);
886 __remove_dirty_segment(sbi, segno, DIRTY);
de881df9 887 } else if (valid_blocks < usable_blocks) {
351df4b2
JK		 888 __locate_dirty_segment(sbi, segno, DIRTY);
889 } else {
890 /* Recovery routine with SSR needs this */
891 __remove_dirty_segment(sbi, segno, DIRTY);
892 }
893
894 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 895}
896
4354994f
DR		 897/* This moves currently empty dirty blocks to prefree. Must hold seglist_lock */
898void f2fs_dirty_to_prefree(struct f2fs_sb_info *sbi)
899{
900 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
901 unsigned int segno;
902
903 mutex_lock(&dirty_i->seglist_lock);
904 for_each_set_bit(segno, dirty_i->dirty_segmap[DIRTY], MAIN_SEGS(sbi)) {
905 if (get_valid_blocks(sbi, segno, false))
906 continue;
907 if (IS_CURSEG(sbi, segno))
908 continue;
909 __locate_dirty_segment(sbi, segno, PRE);
910 __remove_dirty_segment(sbi, segno, DIRTY);
911 }
912 mutex_unlock(&dirty_i->seglist_lock);
913}
914
4d3aed70 915block_t f2fs_get_unusable_blocks(struct f2fs_sb_info *sbi)
4354994f 916{
ae4ad7ea
DR		 917 int ovp_hole_segs =
918 (overprovision_segments(sbi) - reserved_segments(sbi));
919 block_t ovp_holes = ovp_hole_segs << sbi->log_blocks_per_seg;
4d3aed70 920 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
4354994f 921 block_t holes[2] = {0, 0}; /* DATA and NODE */
4d3aed70 922 block_t unusable;
4354994f
DR		 923 struct seg_entry *se;
924 unsigned int segno;
925
926 mutex_lock(&dirty_i->seglist_lock);
927 for_each_set_bit(segno, dirty_i->dirty_segmap[DIRTY], MAIN_SEGS(sbi)) {
928 se = get_seg_entry(sbi, segno);
929 if (IS_NODESEG(se->type))
de881df9
AR		 930 holes[NODE] += f2fs_usable_blks_in_seg(sbi, segno) -
931 se->valid_blocks;
4354994f 932 else
de881df9
AR		 933 holes[DATA] += f2fs_usable_blks_in_seg(sbi, segno) -
934 se->valid_blocks;
4354994f
DR		 935 }
936 mutex_unlock(&dirty_i->seglist_lock);
937
4d3aed70
DR		 938 unusable = holes[DATA] > holes[NODE] ? holes[DATA] : holes[NODE];
939 if (unusable > ovp_holes)
940 return unusable - ovp_holes;
941 return 0;
942}
943
944int f2fs_disable_cp_again(struct f2fs_sb_info *sbi, block_t unusable)
945{
946 int ovp_hole_segs =
947 (overprovision_segments(sbi) - reserved_segments(sbi));
948 if (unusable > F2FS_OPTION(sbi).unusable_cap)
4354994f 949 return -EAGAIN;
db610a64 950 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED_QUICK) &&
ae4ad7ea 951 dirty_segments(sbi) > ovp_hole_segs)
db610a64 952 return -EAGAIN;
4354994f
DR		 953 return 0;
954}
955
956/* This is only used by SBI_CP_DISABLED */
957static unsigned int get_free_segment(struct f2fs_sb_info *sbi)
958{
959 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
960 unsigned int segno = 0;
961
962 mutex_lock(&dirty_i->seglist_lock);
963 for_each_set_bit(segno, dirty_i->dirty_segmap[DIRTY], MAIN_SEGS(sbi)) {
964 if (get_valid_blocks(sbi, segno, false))
965 continue;
966 if (get_ckpt_valid_blocks(sbi, segno))
967 continue;
968 mutex_unlock(&dirty_i->seglist_lock);
969 return segno;
970 }
971 mutex_unlock(&dirty_i->seglist_lock);
972 return NULL_SEGNO;
973}
974
004b6862 975static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi,
c81abe34
JK		 976 struct block_device *bdev, block_t lstart,
977 block_t start, block_t len)
275b66b0 978{
0b54fb84 979 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
ba48a33e 980 struct list_head *pend_list;
b01a9201 981 struct discard_cmd *dc;
275b66b0 982
ba48a33e
CY		 983 f2fs_bug_on(sbi, !len);
984
985 pend_list = &dcc->pend_list[plist_idx(len)];
986
b01a9201
JK		 987 dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS);
988 INIT_LIST_HEAD(&dc->list);
c81abe34 989 dc->bdev = bdev;
b01a9201 990 dc->lstart = lstart;
c81abe34 991 dc->start = start;
b01a9201 992 dc->len = len;
ec9895ad 993 dc->ref = 0;
15469963 994 dc->state = D_PREP;
72691af6 995 dc->queued = 0;
c81abe34 996 dc->error = 0;
b01a9201 997 init_completion(&dc->wait);
22d375dd 998 list_add_tail(&dc->list, pend_list);
35ec7d57
CY		 999 spin_lock_init(&dc->lock);
1000 dc->bio_ref = 0;
5f32366a 1001 atomic_inc(&dcc->discard_cmd_cnt);
d84d1cbd 1002 dcc->undiscard_blks += len;
004b6862
CY		 1003
1004 return dc;
1005}
1006
1007static struct discard_cmd *__attach_discard_cmd(struct f2fs_sb_info *sbi,
1008 struct block_device *bdev, block_t lstart,
1009 block_t start, block_t len,
4dada3fd
CY		 1010 struct rb_node *parent, struct rb_node **p,
1011 bool leftmost)
004b6862
CY		 1012{
1013 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1014 struct discard_cmd *dc;
1015
1016 dc = __create_discard_cmd(sbi, bdev, lstart, start, len);
1017
1018 rb_link_node(&dc->rb_node, parent, p);
4dada3fd 1019 rb_insert_color_cached(&dc->rb_node, &dcc->root, leftmost);
004b6862
CY		 1020
1021 return dc;
15469963
JK		 1022}
1023
004b6862
CY		 1024static void __detach_discard_cmd(struct discard_cmd_control *dcc,
1025 struct discard_cmd *dc)
15469963 1026{
dcc9165d 1027 if (dc->state == D_DONE)
72691af6 1028 atomic_sub(dc->queued, &dcc->queued_discard);
004b6862
CY		 1029
1030 list_del(&dc->list);
4dada3fd 1031 rb_erase_cached(&dc->rb_node, &dcc->root);
d84d1cbd 1032 dcc->undiscard_blks -= dc->len;
004b6862
CY		 1033
1034 kmem_cache_free(discard_cmd_slab, dc);
1035
1036 atomic_dec(&dcc->discard_cmd_cnt);
1037}
1038
1039static void __remove_discard_cmd(struct f2fs_sb_info *sbi,
1040 struct discard_cmd *dc)
1041{
1042 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
35ec7d57 1043 unsigned long flags;
dcc9165d 1044
2ec6f2ef
CY		 1045 trace_f2fs_remove_discard(dc->bdev, dc->start, dc->len);
1046
35ec7d57
CY		 1047 spin_lock_irqsave(&dc->lock, flags);
1048 if (dc->bio_ref) {
1049 spin_unlock_irqrestore(&dc->lock, flags);
1050 return;
1051 }
1052 spin_unlock_irqrestore(&dc->lock, flags);
1053
d9703d90
CY		 1054 f2fs_bug_on(sbi, dc->ref);
1055
c81abe34
JK		 1056 if (dc->error == -EOPNOTSUPP)
1057 dc->error = 0;
15469963 1058
c81abe34 1059 if (dc->error)
22d7ea13 1060 printk_ratelimited(
c45d6002
CY		 1061 "%sF2FS-fs (%s): Issue discard(%u, %u, %u) failed, ret: %d",
1062 KERN_INFO, sbi->sb->s_id,
1063 dc->lstart, dc->start, dc->len, dc->error);
004b6862 1064 __detach_discard_cmd(dcc, dc);
275b66b0
CY		 1065}
1066
c81abe34
JK		 1067static void f2fs_submit_discard_endio(struct bio *bio)
1068{
1069 struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
35ec7d57 1070 unsigned long flags;
c81abe34 1071
35ec7d57 1072 spin_lock_irqsave(&dc->lock, flags);
3fa6a8c5
ST		 1073 if (!dc->error)
1074 dc->error = blk_status_to_errno(bio->bi_status);
35ec7d57
CY		 1075 dc->bio_ref--;
1076 if (!dc->bio_ref && dc->state == D_SUBMIT) {
1077 dc->state = D_DONE;
1078 complete_all(&dc->wait);
1079 }
1080 spin_unlock_irqrestore(&dc->lock, flags);
c81abe34
JK		 1081 bio_put(bio);
1082}
1083
94b1e10e 1084static void __check_sit_bitmap(struct f2fs_sb_info *sbi,
6915ea9d
CY		 1085 block_t start, block_t end)
1086{
1087#ifdef CONFIG_F2FS_CHECK_FS
1088 struct seg_entry *sentry;
1089 unsigned int segno;
1090 block_t blk = start;
1091 unsigned long offset, size, max_blocks = sbi->blocks_per_seg;
1092 unsigned long *map;
1093
1094 while (blk < end) {
1095 segno = GET_SEGNO(sbi, blk);
1096 sentry = get_seg_entry(sbi, segno);
1097 offset = GET_BLKOFF_FROM_SEG0(sbi, blk);
1098
008396e1
YS		 1099 if (end < START_BLOCK(sbi, segno + 1))
1100 size = GET_BLKOFF_FROM_SEG0(sbi, end);
1101 else
1102 size = max_blocks;
6915ea9d
CY		 1103 map = (unsigned long *)(sentry->cur_valid_map);
1104 offset = __find_rev_next_bit(map, size, offset);
1105 f2fs_bug_on(sbi, offset != size);
008396e1 1106 blk = START_BLOCK(sbi, segno + 1);
6915ea9d
CY		 1107 }
1108#endif
1109}
1110
8bb4f253
JK		 1111static void __init_discard_policy(struct f2fs_sb_info *sbi,
1112 struct discard_policy *dpolicy,
1113 int discard_type, unsigned int granularity)
1114{
1115 /* common policy */
1116 dpolicy->type = discard_type;
1117 dpolicy->sync = true;
20ee4382 1118 dpolicy->ordered = false;
8bb4f253
JK		 1119 dpolicy->granularity = granularity;
1120
1121 dpolicy->max_requests = DEF_MAX_DISCARD_REQUEST;
1122 dpolicy->io_aware_gran = MAX_PLIST_NUM;
6ce48b0c 1123 dpolicy->timeout = false;
8bb4f253
JK		 1124
1125 if (discard_type == DPOLICY_BG) {
1126 dpolicy->min_interval = DEF_MIN_DISCARD_ISSUE_TIME;
f9d1dced 1127 dpolicy->mid_interval = DEF_MID_DISCARD_ISSUE_TIME;
8bb4f253
JK		 1128 dpolicy->max_interval = DEF_MAX_DISCARD_ISSUE_TIME;
1129 dpolicy->io_aware = true;
cba60849 1130 dpolicy->sync = false;
20ee4382 1131 dpolicy->ordered = true;
8bb4f253
JK		 1132 if (utilization(sbi) > DEF_DISCARD_URGENT_UTIL) {
1133 dpolicy->granularity = 1;
1134 dpolicy->max_interval = DEF_MIN_DISCARD_ISSUE_TIME;
1135 }
1136 } else if (discard_type == DPOLICY_FORCE) {
1137 dpolicy->min_interval = DEF_MIN_DISCARD_ISSUE_TIME;
f9d1dced 1138 dpolicy->mid_interval = DEF_MID_DISCARD_ISSUE_TIME;
8bb4f253
JK		 1139 dpolicy->max_interval = DEF_MAX_DISCARD_ISSUE_TIME;
1140 dpolicy->io_aware = false;
1141 } else if (discard_type == DPOLICY_FSTRIM) {
1142 dpolicy->io_aware = false;
1143 } else if (discard_type == DPOLICY_UMOUNT) {
1144 dpolicy->io_aware = false;
b8623253
JK		 1145 /* we need to issue all to keep CP_TRIMMED_FLAG */
1146 dpolicy->granularity = 1;
6ce48b0c 1147 dpolicy->timeout = true;
8bb4f253
JK		 1148 }
1149}
1150
35ec7d57
CY		 1151static void __update_discard_tree_range(struct f2fs_sb_info *sbi,
1152 struct block_device *bdev, block_t lstart,
1153 block_t start, block_t len);
c81abe34 1154/* this function is copied from blkdev_issue_discard from block/blk-lib.c */
6b9cb124 1155static int __submit_discard_cmd(struct f2fs_sb_info *sbi,
78997b56 1156 struct discard_policy *dpolicy,
35ec7d57
CY		 1157 struct discard_cmd *dc,
1158 unsigned int *issued)
c81abe34 1159{
35ec7d57
CY		 1160 struct block_device *bdev = dc->bdev;
1161 struct request_queue *q = bdev_get_queue(bdev);
1162 unsigned int max_discard_blocks =
1163 SECTOR_TO_BLOCK(q->limits.max_discard_sectors);
c81abe34 1164 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
78997b56
CY		 1165 struct list_head *wait_list = (dpolicy->type == DPOLICY_FSTRIM) ?
1166 &(dcc->fstrim_list) : &(dcc->wait_list);
78997b56 1167 int flag = dpolicy->sync ? REQ_SYNC : 0;
35ec7d57
CY		 1168 block_t lstart, start, len, total_len;
1169 int err = 0;
c81abe34
JK		 1170
1171 if (dc->state != D_PREP)
6b9cb124 1172 return 0;
c81abe34 1173
d6184774 1174 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK))
6b9cb124 1175 return 0;
d6184774 1176
35ec7d57
CY		 1177 trace_f2fs_issue_discard(bdev, dc->start, dc->len);
1178
1179 lstart = dc->lstart;
1180 start = dc->start;
1181 len = dc->len;
1182 total_len = len;
1183
1184 dc->len = 0;
1185
1186 while (total_len && *issued < dpolicy->max_requests && !err) {
1187 struct bio *bio = NULL;
1188 unsigned long flags;
1189 bool last = true;
1190
1191 if (len > max_discard_blocks) {
1192 len = max_discard_blocks;
1193 last = false;
1194 }
1195
1196 (*issued)++;
1197 if (*issued == dpolicy->max_requests)
1198 last = true;
1199
1200 dc->len += len;
1201
b83dcfe6 1202 if (time_to_inject(sbi, FAULT_DISCARD)) {
c45d6002 1203 f2fs_show_injection_info(sbi, FAULT_DISCARD);
b83dcfe6
CY		 1204 err = -EIO;
1205 goto submit;
1206 }
35ec7d57
CY		 1207 err = __blkdev_issue_discard(bdev,
1208 SECTOR_FROM_BLOCK(start),
1209 SECTOR_FROM_BLOCK(len),
1210 GFP_NOFS, 0, &bio);
b83dcfe6 1211submit:
6b9cb124 1212 if (err) {
35ec7d57 1213 spin_lock_irqsave(&dc->lock, flags);
6b9cb124 1214 if (dc->state == D_PARTIAL)
35ec7d57 1215 dc->state = D_SUBMIT;
35ec7d57
CY		 1216 spin_unlock_irqrestore(&dc->lock, flags);
1217
6b9cb124
CY		 1218 break;
1219 }
35ec7d57 1220
6b9cb124 1221 f2fs_bug_on(sbi, !bio);
35ec7d57 1222
6b9cb124
CY		 1223 /*
1224 * should keep before submission to avoid D_DONE
1225 * right away
1226 */
1227 spin_lock_irqsave(&dc->lock, flags);
1228 if (last)
1229 dc->state = D_SUBMIT;
1230 else
1231 dc->state = D_PARTIAL;
1232 dc->bio_ref++;
1233 spin_unlock_irqrestore(&dc->lock, flags);
35ec7d57 1234
72691af6
JK		 1235 atomic_inc(&dcc->queued_discard);
1236 dc->queued++;
6b9cb124 1237 list_move_tail(&dc->list, wait_list);
b0af6d49 1238
6b9cb124 1239 /* sanity check on discard range */
9249dded 1240 __check_sit_bitmap(sbi, lstart, lstart + len);
35ec7d57 1241
6b9cb124
CY		 1242 bio->bi_private = dc;
1243 bio->bi_end_io = f2fs_submit_discard_endio;
1244 bio->bi_opf |= flag;
1245 submit_bio(bio);
1246
1247 atomic_inc(&dcc->issued_discard);
1248
1249 f2fs_update_iostat(sbi, FS_DISCARD, 1);
35ec7d57
CY		 1250
1251 lstart += len;
1252 start += len;
1253 total_len -= len;
1254 len = total_len;
c81abe34 1255 }
35ec7d57 1256
df423399
ST		 1257 if (!err && len) {
1258 dcc->undiscard_blks -= len;
35ec7d57 1259 __update_discard_tree_range(sbi, bdev, lstart, start, len);
df423399 1260 }
6b9cb124 1261 return err;
c81abe34
JK		 1262}
1263
47d0d7d7 1264static void __insert_discard_tree(struct f2fs_sb_info *sbi,
004b6862
CY		 1265 struct block_device *bdev, block_t lstart,
1266 block_t start, block_t len,
1267 struct rb_node **insert_p,
1268 struct rb_node *insert_parent)
c81abe34 1269{
004b6862 1270 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
dca6951f 1271 struct rb_node **p;
004b6862 1272 struct rb_node *parent = NULL;
4dada3fd 1273 bool leftmost = true;
004b6862
CY		 1274
1275 if (insert_p && insert_parent) {
1276 parent = insert_parent;
1277 p = insert_p;
1278 goto do_insert;
1279 }
c81abe34 1280
4dada3fd
CY		 1281 p = f2fs_lookup_rb_tree_for_insert(sbi, &dcc->root, &parent,
1282 lstart, &leftmost);
004b6862 1283do_insert:
47d0d7d7 1284 __attach_discard_cmd(sbi, bdev, lstart, start, len, parent,
4dada3fd 1285 p, leftmost);
c81abe34
JK		 1286}
1287
ba48a33e
CY		 1288static void __relocate_discard_cmd(struct discard_cmd_control *dcc,
1289 struct discard_cmd *dc)
1290{
1291 list_move_tail(&dc->list, &dcc->pend_list[plist_idx(dc->len)]);
1292}
1293
3d6a650f
YH		 1294static void __punch_discard_cmd(struct f2fs_sb_info *sbi,
1295 struct discard_cmd *dc, block_t blkaddr)
1296{
ba48a33e 1297 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 1298 struct discard_info di = dc->di;
1299 bool modified = false;
3d6a650f 1300
004b6862 1301 if (dc->state == D_DONE || dc->len == 1) {
3d6a650f
YH		 1302 __remove_discard_cmd(sbi, dc);
1303 return;
1304 }
1305
d84d1cbd
CY		 1306 dcc->undiscard_blks -= di.len;
1307
004b6862 1308 if (blkaddr > di.lstart) {
3d6a650f 1309 dc->len = blkaddr - dc->lstart;
d84d1cbd 1310 dcc->undiscard_blks += dc->len;
ba48a33e 1311 __relocate_discard_cmd(dcc, dc);
004b6862
CY		 1312 modified = true;
1313 }
1314
1315 if (blkaddr < di.lstart + di.len - 1) {
1316 if (modified) {
1317 __insert_discard_tree(sbi, dc->bdev, blkaddr + 1,
1318 di.start + blkaddr + 1 - di.lstart,
1319 di.lstart + di.len - 1 - blkaddr,
1320 NULL, NULL);
1321 } else {
1322 dc->lstart++;
1323 dc->len--;
1324 dc->start++;
d84d1cbd 1325 dcc->undiscard_blks += dc->len;
ba48a33e 1326 __relocate_discard_cmd(dcc, dc);
004b6862 1327 }
3d6a650f
YH		 1328 }
1329}
1330
004b6862
CY		 1331static void __update_discard_tree_range(struct f2fs_sb_info *sbi,
1332 struct block_device *bdev, block_t lstart,
1333 block_t start, block_t len)
275b66b0 1334{
0b54fb84 1335 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 1336 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
1337 struct discard_cmd *dc;
1338 struct discard_info di = {0};
1339 struct rb_node **insert_p = NULL, *insert_parent = NULL;
35ec7d57
CY		 1340 struct request_queue *q = bdev_get_queue(bdev);
1341 unsigned int max_discard_blocks =
1342 SECTOR_TO_BLOCK(q->limits.max_discard_sectors);
004b6862 1343 block_t end = lstart + len;
275b66b0 1344
4d57b86d 1345 dc = (struct discard_cmd *)f2fs_lookup_rb_tree_ret(&dcc->root,
004b6862
CY		 1346 NULL, lstart,
1347 (struct rb_entry **)&prev_dc,
1348 (struct rb_entry **)&next_dc,
4dada3fd 1349 &insert_p, &insert_parent, true, NULL);
004b6862
CY		 1350 if (dc)
1351 prev_dc = dc;
1352
1353 if (!prev_dc) {
1354 di.lstart = lstart;
1355 di.len = next_dc ? next_dc->lstart - lstart : len;
1356 di.len = min(di.len, len);
1357 di.start = start;
22d375dd 1358 }
15469963 1359
004b6862
CY		 1360 while (1) {
1361 struct rb_node *node;
1362 bool merged = false;
1363 struct discard_cmd *tdc = NULL;
1364
1365 if (prev_dc) {
1366 di.lstart = prev_dc->lstart + prev_dc->len;
1367 if (di.lstart < lstart)
1368 di.lstart = lstart;
1369 if (di.lstart >= end)
1370 break;
1371
1372 if (!next_dc || next_dc->lstart > end)
1373 di.len = end - di.lstart;
1374 else
1375 di.len = next_dc->lstart - di.lstart;
1376 di.start = start + di.lstart - lstart;
1377 }
1378
1379 if (!di.len)
1380 goto next;
1381
1382 if (prev_dc && prev_dc->state == D_PREP &&
1383 prev_dc->bdev == bdev &&
35ec7d57
CY		 1384 __is_discard_back_mergeable(&di, &prev_dc->di,
1385 max_discard_blocks)) {
004b6862 1386 prev_dc->di.len += di.len;
d84d1cbd 1387 dcc->undiscard_blks += di.len;
ba48a33e 1388 __relocate_discard_cmd(dcc, prev_dc);
004b6862
CY		 1389 di = prev_dc->di;
1390 tdc = prev_dc;
1391 merged = true;
1392 }
1393
1394 if (next_dc && next_dc->state == D_PREP &&
1395 next_dc->bdev == bdev &&
35ec7d57
CY		 1396 __is_discard_front_mergeable(&di, &next_dc->di,
1397 max_discard_blocks)) {
004b6862
CY		 1398 next_dc->di.lstart = di.lstart;
1399 next_dc->di.len += di.len;
1400 next_dc->di.start = di.start;
d84d1cbd 1401 dcc->undiscard_blks += di.len;
ba48a33e 1402 __relocate_discard_cmd(dcc, next_dc);
004b6862
CY		 1403 if (tdc)
1404 __remove_discard_cmd(sbi, tdc);
004b6862 1405 merged = true;
4e6a8d9b 1406 }
004b6862 1407
df0f6b44 1408 if (!merged) {
004b6862
CY		 1409 __insert_discard_tree(sbi, bdev, di.lstart, di.start,
1410 di.len, NULL, NULL);
df0f6b44 1411 }
004b6862
CY		 1412 next:
1413 prev_dc = next_dc;
1414 if (!prev_dc)
1415 break;
1416
1417 node = rb_next(&prev_dc->rb_node);
1418 next_dc = rb_entry_safe(node, struct discard_cmd, rb_node);
1419 }
004b6862
CY		 1420}
1421
1422static int __queue_discard_cmd(struct f2fs_sb_info *sbi,
1423 struct block_device *bdev, block_t blkstart, block_t blklen)
1424{
1425 block_t lblkstart = blkstart;
1426
7f3d7719
DLM		 1427 if (!f2fs_bdev_support_discard(bdev))
1428 return 0;
1429
0243a5f9 1430 trace_f2fs_queue_discard(bdev, blkstart, blklen);
004b6862 1431
0916878d 1432 if (f2fs_is_multi_device(sbi)) {
004b6862
CY		 1433 int devi = f2fs_target_device_index(sbi, blkstart);
1434
1435 blkstart -= FDEV(devi).start_blk;
1436 }
35ec7d57 1437 mutex_lock(&SM_I(sbi)->dcc_info->cmd_lock);
004b6862 1438 __update_discard_tree_range(sbi, bdev, lblkstart, blkstart, blklen);
35ec7d57 1439 mutex_unlock(&SM_I(sbi)->dcc_info->cmd_lock);
004b6862
CY		 1440 return 0;
1441}
1442
20ee4382
CY		 1443static unsigned int __issue_discard_cmd_orderly(struct f2fs_sb_info *sbi,
1444 struct discard_policy *dpolicy)
1445{
1446 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1447 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
1448 struct rb_node **insert_p = NULL, *insert_parent = NULL;
1449 struct discard_cmd *dc;
1450 struct blk_plug plug;
1451 unsigned int pos = dcc->next_pos;
1452 unsigned int issued = 0;
1453 bool io_interrupted = false;
1454
1455 mutex_lock(&dcc->cmd_lock);
1456 dc = (struct discard_cmd *)f2fs_lookup_rb_tree_ret(&dcc->root,
1457 NULL, pos,
1458 (struct rb_entry **)&prev_dc,
1459 (struct rb_entry **)&next_dc,
4dada3fd 1460 &insert_p, &insert_parent, true, NULL);
20ee4382
CY		 1461 if (!dc)
1462 dc = next_dc;
1463
1464 blk_start_plug(&plug);
1465
1466 while (dc) {
1467 struct rb_node *node;
6b9cb124 1468 int err = 0;
20ee4382
CY		 1469
1470 if (dc->state != D_PREP)
1471 goto next;
1472
a7d10cf3 1473 if (dpolicy->io_aware && !is_idle(sbi, DISCARD_TIME)) {
20ee4382
CY		 1474 io_interrupted = true;
1475 break;
1476 }
1477
1478 dcc->next_pos = dc->lstart + dc->len;
6b9cb124 1479 err = __submit_discard_cmd(sbi, dpolicy, dc, &issued);
20ee4382 1480
35ec7d57 1481 if (issued >= dpolicy->max_requests)
20ee4382
CY		 1482 break;
1483next:
1484 node = rb_next(&dc->rb_node);
6b9cb124
CY		 1485 if (err)
1486 __remove_discard_cmd(sbi, dc);
20ee4382
CY		 1487 dc = rb_entry_safe(node, struct discard_cmd, rb_node);
1488 }
1489
1490 blk_finish_plug(&plug);
1491
1492 if (!dc)
1493 dcc->next_pos = 0;
1494
1495 mutex_unlock(&dcc->cmd_lock);
1496
1497 if (!issued && io_interrupted)
1498 issued = -1;
1499
1500 return issued;
1501}
141af6ba
ST		 1502static unsigned int __wait_all_discard_cmd(struct f2fs_sb_info *sbi,
1503 struct discard_policy *dpolicy);
20ee4382 1504
78997b56
CY		 1505static int __issue_discard_cmd(struct f2fs_sb_info *sbi,
1506 struct discard_policy *dpolicy)
bd5b0738
CY		 1507{
1508 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1509 struct list_head *pend_list;
1510 struct discard_cmd *dc, *tmp;
1511 struct blk_plug plug;
141af6ba 1512 int i, issued;
e6c6de18 1513 bool io_interrupted = false;
bd5b0738 1514
6ce48b0c
CY		 1515 if (dpolicy->timeout)
1516 f2fs_update_time(sbi, UMOUNT_DISCARD_TIMEOUT);
03f2c02d 1517
141af6ba
ST		 1518retry:
1519 issued = 0;
78997b56 1520 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
6ce48b0c
CY		 1521 if (dpolicy->timeout &&
1522 f2fs_time_over(sbi, UMOUNT_DISCARD_TIMEOUT))
03f2c02d
JK		 1523 break;
1524
78997b56
CY		 1525 if (i + 1 < dpolicy->granularity)
1526 break;
20ee4382
CY		 1527
1528 if (i < DEFAULT_DISCARD_GRANULARITY && dpolicy->ordered)
1529 return __issue_discard_cmd_orderly(sbi, dpolicy);
1530
bd5b0738 1531 pend_list = &dcc->pend_list[i];
33da62cf
CY		 1532
1533 mutex_lock(&dcc->cmd_lock);
49c60c67
CY		 1534 if (list_empty(pend_list))
1535 goto next;
67fce70b
CY		 1536 if (unlikely(dcc->rbtree_check))
1537 f2fs_bug_on(sbi, !f2fs_check_rb_tree_consistence(sbi,
2e9b2bb2 1538 &dcc->root, false));
33da62cf 1539 blk_start_plug(&plug);
bd5b0738
CY		 1540 list_for_each_entry_safe(dc, tmp, pend_list, list) {
1541 f2fs_bug_on(sbi, dc->state != D_PREP);
1542
6ce48b0c
CY		 1543 if (dpolicy->timeout &&
1544 f2fs_time_over(sbi, UMOUNT_DISCARD_TIMEOUT))
6e0cd4a9
HX		 1545 break;
1546
ecc9aa00 1547 if (dpolicy->io_aware && i < dpolicy->io_aware_gran &&
a7d10cf3 1548 !is_idle(sbi, DISCARD_TIME)) {
e6c6de18 1549 io_interrupted = true;
522d1711 1550 break;
969d1b18 1551 }
e6c6de18 1552
35ec7d57 1553 __submit_discard_cmd(sbi, dpolicy, dc, &issued);
522d1711 1554
35ec7d57 1555 if (issued >= dpolicy->max_requests)
33da62cf 1556 break;
bd5b0738 1557 }
33da62cf 1558 blk_finish_plug(&plug);
49c60c67 1559next:
33da62cf
CY		 1560 mutex_unlock(&dcc->cmd_lock);
1561
522d1711 1562 if (issued >= dpolicy->max_requests || io_interrupted)
33da62cf 1563 break;
bd5b0738 1564 }
969d1b18 1565
141af6ba
ST		 1566 if (dpolicy->type == DPOLICY_UMOUNT && issued) {
1567 __wait_all_discard_cmd(sbi, dpolicy);
1568 goto retry;
1569 }
1570
e6c6de18
CY		 1571 if (!issued && io_interrupted)
1572 issued = -1;
1573
969d1b18
CY		 1574 return issued;
1575}
1576
cf5c759f 1577static bool __drop_discard_cmd(struct f2fs_sb_info *sbi)
969d1b18
CY		 1578{
1579 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1580 struct list_head *pend_list;
1581 struct discard_cmd *dc, *tmp;
1582 int i;
cf5c759f 1583 bool dropped = false;
969d1b18
CY		 1584
1585 mutex_lock(&dcc->cmd_lock);
1586 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
1587 pend_list = &dcc->pend_list[i];
1588 list_for_each_entry_safe(dc, tmp, pend_list, list) {
1589 f2fs_bug_on(sbi, dc->state != D_PREP);
1590 __remove_discard_cmd(sbi, dc);
cf5c759f 1591 dropped = true;
969d1b18
CY		 1592 }
1593 }
1594 mutex_unlock(&dcc->cmd_lock);
cf5c759f
CY		 1595
1596 return dropped;
bd5b0738
CY		 1597}
1598
4d57b86d 1599void f2fs_drop_discard_cmd(struct f2fs_sb_info *sbi)
7950e9ac
CY		 1600{
1601 __drop_discard_cmd(sbi);
1602}
1603
0ea80512 1604static unsigned int __wait_one_discard_bio(struct f2fs_sb_info *sbi,
2a510c00
CY		 1605 struct discard_cmd *dc)
1606{
1607 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
0ea80512 1608 unsigned int len = 0;
2a510c00
CY		 1609
1610 wait_for_completion_io(&dc->wait);
1611 mutex_lock(&dcc->cmd_lock);
1612 f2fs_bug_on(sbi, dc->state != D_DONE);
1613 dc->ref--;
0ea80512
CY		 1614 if (!dc->ref) {
1615 if (!dc->error)
1616 len = dc->len;
2a510c00 1617 __remove_discard_cmd(sbi, dc);
0ea80512 1618 }
2a510c00 1619 mutex_unlock(&dcc->cmd_lock);
0ea80512
CY		 1620
1621 return len;
2a510c00
CY		 1622}
1623
0ea80512 1624static unsigned int __wait_discard_cmd_range(struct f2fs_sb_info *sbi,
78997b56
CY		 1625 struct discard_policy *dpolicy,
1626 block_t start, block_t end)
63a94fa1
CY		 1627{
1628 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
78997b56
CY		 1629 struct list_head *wait_list = (dpolicy->type == DPOLICY_FSTRIM) ?
1630 &(dcc->fstrim_list) : &(dcc->wait_list);
63a94fa1 1631 struct discard_cmd *dc, *tmp;
6afae633 1632 bool need_wait;
0ea80512 1633 unsigned int trimmed = 0;
6afae633
CY		 1634
1635next:
1636 need_wait = false;
63a94fa1
CY		 1637
1638 mutex_lock(&dcc->cmd_lock);
1639 list_for_each_entry_safe(dc, tmp, wait_list, list) {
8412663d
CY		 1640 if (dc->lstart + dc->len <= start || end <= dc->lstart)
1641 continue;
78997b56 1642 if (dc->len < dpolicy->granularity)
8412663d 1643 continue;
78997b56 1644 if (dc->state == D_DONE && !dc->ref) {
63a94fa1 1645 wait_for_completion_io(&dc->wait);
0ea80512
CY		 1646 if (!dc->error)
1647 trimmed += dc->len;
63a94fa1 1648 __remove_discard_cmd(sbi, dc);
6afae633
CY		 1649 } else {
1650 dc->ref++;
1651 need_wait = true;
1652 break;
63a94fa1
CY		 1653 }
1654 }
1655 mutex_unlock(&dcc->cmd_lock);
6afae633
CY		 1656
1657 if (need_wait) {
0ea80512 1658 trimmed += __wait_one_discard_bio(sbi, dc);
6afae633
CY		 1659 goto next;
1660 }
0ea80512
CY		 1661
1662 return trimmed;
63a94fa1
CY		 1663}
1664
01f9cf6d 1665static unsigned int __wait_all_discard_cmd(struct f2fs_sb_info *sbi,
78997b56 1666 struct discard_policy *dpolicy)
8412663d 1667{
9a997188 1668 struct discard_policy dp;
01f9cf6d 1669 unsigned int discard_blks;
9a997188 1670
01f9cf6d
CY		 1671 if (dpolicy)
1672 return __wait_discard_cmd_range(sbi, dpolicy, 0, UINT_MAX);
9a997188
JK		 1673
1674 /* wait all */
8bb4f253 1675 __init_discard_policy(sbi, &dp, DPOLICY_FSTRIM, 1);
01f9cf6d 1676 discard_blks = __wait_discard_cmd_range(sbi, &dp, 0, UINT_MAX);
8bb4f253 1677 __init_discard_policy(sbi, &dp, DPOLICY_UMOUNT, 1);
01f9cf6d
CY		 1678 discard_blks += __wait_discard_cmd_range(sbi, &dp, 0, UINT_MAX);
1679
1680 return discard_blks;
8412663d
CY		 1681}
1682
004b6862 1683/* This should be covered by global mutex, &sit_i->sentry_lock */
94b1e10e 1684static void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
004b6862
CY		 1685{
1686 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1687 struct discard_cmd *dc;
ec9895ad 1688 bool need_wait = false;
004b6862
CY		 1689
1690 mutex_lock(&dcc->cmd_lock);
4d57b86d
CY		 1691 dc = (struct discard_cmd *)f2fs_lookup_rb_tree(&dcc->root,
1692 NULL, blkaddr);
004b6862 1693 if (dc) {
ec9895ad
CY		 1694 if (dc->state == D_PREP) {
1695 __punch_discard_cmd(sbi, dc, blkaddr);
1696 } else {
1697 dc->ref++;
1698 need_wait = true;
1699 }
275b66b0 1700 }
d431413f 1701 mutex_unlock(&dcc->cmd_lock);
ec9895ad 1702
2a510c00
CY		 1703 if (need_wait)
1704 __wait_one_discard_bio(sbi, dc);
d431413f
CY		 1705}
1706
4d57b86d 1707void f2fs_stop_discard_thread(struct f2fs_sb_info *sbi)
cce13252
CY		 1708{
1709 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1710
1711 if (dcc && dcc->f2fs_issue_discard) {
1712 struct task_struct *discard_thread = dcc->f2fs_issue_discard;
1713
1714 dcc->f2fs_issue_discard = NULL;
1715 kthread_stop(discard_thread);
ec9895ad 1716 }
d431413f
CY		 1717}
1718
8412663d 1719/* This comes from f2fs_put_super */
03f2c02d 1720bool f2fs_issue_discard_timeout(struct f2fs_sb_info *sbi)
969d1b18
CY		 1721{
1722 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
78997b56 1723 struct discard_policy dpolicy;
cf5c759f 1724 bool dropped;
969d1b18 1725
8bb4f253
JK		 1726 __init_discard_policy(sbi, &dpolicy, DPOLICY_UMOUNT,
1727 dcc->discard_granularity);
78997b56 1728 __issue_discard_cmd(sbi, &dpolicy);
cf5c759f 1729 dropped = __drop_discard_cmd(sbi);
cf5c759f 1730
9a997188
JK		 1731 /* just to make sure there is no pending discard commands */
1732 __wait_all_discard_cmd(sbi, NULL);
2482c432
CY		 1733
1734 f2fs_bug_on(sbi, atomic_read(&dcc->discard_cmd_cnt));
cf5c759f 1735 return dropped;
969d1b18
CY		 1736}
1737
15469963
JK		 1738static int issue_discard_thread(void *data)
1739{
1740 struct f2fs_sb_info *sbi = data;
1741 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1742 wait_queue_head_t *q = &dcc->discard_wait_queue;
78997b56 1743 struct discard_policy dpolicy;
969d1b18
CY		 1744 unsigned int wait_ms = DEF_MIN_DISCARD_ISSUE_TIME;
1745 int issued;
15469963 1746
1d7be270 1747 set_freezable();
15469963 1748
1d7be270 1749 do {
8bb4f253 1750 __init_discard_policy(sbi, &dpolicy, DPOLICY_BG,
78997b56
CY		 1751 dcc->discard_granularity);
1752
969d1b18
CY		 1753 wait_event_interruptible_timeout(*q,
1754 kthread_should_stop() || freezing(current) ||
1755 dcc->discard_wake,
1756 msecs_to_jiffies(wait_ms));
35a9a766
SY		 1757
1758 if (dcc->discard_wake)
1759 dcc->discard_wake = 0;
1760
76c7bfb3
JK		 1761 /* clean up pending candidates before going to sleep */
1762 if (atomic_read(&dcc->queued_discard))
1763 __wait_all_discard_cmd(sbi, NULL);
1764
1d7be270
JK		 1765 if (try_to_freeze())
1766 continue;
3b60d802
CY		 1767 if (f2fs_readonly(sbi->sb))
1768 continue;
1d7be270
JK		 1769 if (kthread_should_stop())
1770 return 0;
d6184774
YH		 1771 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
1772 wait_ms = dpolicy.max_interval;
1773 continue;
1774 }
15469963 1775
0e5e8111 1776 if (sbi->gc_mode == GC_URGENT_HIGH)
8bb4f253 1777 __init_discard_policy(sbi, &dpolicy, DPOLICY_FORCE, 1);
969d1b18 1778
dc6febb6
CY		 1779 sb_start_intwrite(sbi->sb);
1780
78997b56 1781 issued = __issue_discard_cmd(sbi, &dpolicy);
f9d1dced 1782 if (issued > 0) {
78997b56
CY		 1783 __wait_all_discard_cmd(sbi, &dpolicy);
1784 wait_ms = dpolicy.min_interval;
f9d1dced 1785 } else if (issued == -1){
a7d10cf3
ST		 1786 wait_ms = f2fs_time_to_wait(sbi, DISCARD_TIME);
1787 if (!wait_ms)
abde73c7 1788 wait_ms = dpolicy.mid_interval;
969d1b18 1789 } else {
78997b56 1790 wait_ms = dpolicy.max_interval;
969d1b18 1791 }
1d7be270 1792
dc6febb6 1793 sb_end_intwrite(sbi->sb);
1d7be270 1794
1d7be270
JK		 1795 } while (!kthread_should_stop());
1796 return 0;
15469963
JK		 1797}
1798
f46e8809 1799#ifdef CONFIG_BLK_DEV_ZONED
3c62be17
JK		 1800static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
1801 struct block_device *bdev, block_t blkstart, block_t blklen)
f46e8809 1802{
92592285 1803 sector_t sector, nr_sects;
10a875f8 1804 block_t lblkstart = blkstart;
3c62be17
JK		 1805 int devi = 0;
1806
0916878d 1807 if (f2fs_is_multi_device(sbi)) {
3c62be17 1808 devi = f2fs_target_device_index(sbi, blkstart);
95175daf
DLM		 1809 if (blkstart < FDEV(devi).start_blk ||
1810 blkstart > FDEV(devi).end_blk) {
dcbb4c10 1811 f2fs_err(sbi, "Invalid block %x", blkstart);
95175daf
DLM		 1812 return -EIO;
1813 }
3c62be17
JK		 1814 blkstart -= FDEV(devi).start_blk;
1815 }
f46e8809 1816
95175daf
DLM		 1817 /* For sequential zones, reset the zone write pointer */
1818 if (f2fs_blkz_is_seq(sbi, devi, blkstart)) {
92592285
JK		 1819 sector = SECTOR_FROM_BLOCK(blkstart);
1820 nr_sects = SECTOR_FROM_BLOCK(blklen);
1821
1822 if (sector & (bdev_zone_sectors(bdev) - 1) ||
1823 nr_sects != bdev_zone_sectors(bdev)) {
dcbb4c10
JP		 1824 f2fs_err(sbi, "(%d) %s: Unaligned zone reset attempted (block %x + %x)",
1825 devi, sbi->s_ndevs ? FDEV(devi).path : "",
1826 blkstart, blklen);
92592285
JK		 1827 return -EIO;
1828 }
d50aaeec 1829 trace_f2fs_issue_reset_zone(bdev, blkstart);
6c1b1da5
AJ		 1830 return blkdev_zone_mgmt(bdev, REQ_OP_ZONE_RESET,
1831 sector, nr_sects, GFP_NOFS);
f46e8809 1832 }
95175daf
DLM		 1833
1834 /* For conventional zones, use regular discard if supported */
95175daf 1835 return __queue_discard_cmd(sbi, bdev, lblkstart, blklen);
f46e8809
DLM		 1836}
1837#endif
1838
3c62be17
JK		 1839static int __issue_discard_async(struct f2fs_sb_info *sbi,
1840 struct block_device *bdev, block_t blkstart, block_t blklen)
1841{
1842#ifdef CONFIG_BLK_DEV_ZONED
7f3d7719 1843 if (f2fs_sb_has_blkzoned(sbi) && bdev_is_zoned(bdev))
3c62be17
JK		 1844 return __f2fs_issue_discard_zone(sbi, bdev, blkstart, blklen);
1845#endif
c81abe34 1846 return __queue_discard_cmd(sbi, bdev, blkstart, blklen);
3c62be17
JK		 1847}
1848
1e87a78d 1849static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
37208879
JK		 1850 block_t blkstart, block_t blklen)
1851{
3c62be17
JK		 1852 sector_t start = blkstart, len = 0;
1853 struct block_device *bdev;
a66cdd98
JK		 1854 struct seg_entry *se;
1855 unsigned int offset;
1856 block_t i;
3c62be17
JK		 1857 int err = 0;
1858
1859 bdev = f2fs_target_device(sbi, blkstart, NULL);
1860
1861 for (i = blkstart; i < blkstart + blklen; i++, len++) {
1862 if (i != start) {
1863 struct block_device *bdev2 =
1864 f2fs_target_device(sbi, i, NULL);
1865
1866 if (bdev2 != bdev) {
1867 err = __issue_discard_async(sbi, bdev,
1868 start, len);
1869 if (err)
1870 return err;
1871 bdev = bdev2;
1872 start = i;
1873 len = 0;
1874 }
1875 }
a66cdd98 1876
a66cdd98
JK		 1877 se = get_seg_entry(sbi, GET_SEGNO(sbi, i));
1878 offset = GET_BLKOFF_FROM_SEG0(sbi, i);
1879
1880 if (!f2fs_test_and_set_bit(offset, se->discard_map))
1881 sbi->discard_blks--;
1882 }
f46e8809 1883
3c62be17
JK		 1884 if (len)
1885 err = __issue_discard_async(sbi, bdev, start, len);
1886 return err;
1e87a78d
JK		 1887}
1888
25290fa5
JK		 1889static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc,
1890 bool check_only)
adf4983b 1891{
b2955550
JK		 1892 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
1893 int max_blocks = sbi->blocks_per_seg;
4b2fecc8 1894 struct seg_entry *se = get_seg_entry(sbi, cpc->trim_start);
b2955550
JK		 1895 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1896 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
a66cdd98 1897 unsigned long *discard_map = (unsigned long *)se->discard_map;
60a3b782 1898 unsigned long *dmap = SIT_I(sbi)->tmp_map;
b2955550 1899 unsigned int start = 0, end = -1;
c473f1a9 1900 bool force = (cpc->reason & CP_DISCARD);
a7eeb823 1901 struct discard_entry *de = NULL;
46f84c2c 1902 struct list_head *head = &SM_I(sbi)->dcc_info->entry_list;
b2955550
JK		 1903 int i;
1904
7d20c8ab 1905 if (se->valid_blocks == max_blocks || !f2fs_hw_support_discard(sbi))
25290fa5 1906 return false;
b2955550 1907
a66cdd98 1908 if (!force) {
7d20c8ab 1909 if (!f2fs_realtime_discard_enable(sbi) || !se->valid_blocks ||
0b54fb84
JK		 1910 SM_I(sbi)->dcc_info->nr_discards >=
1911 SM_I(sbi)->dcc_info->max_discards)
25290fa5 1912 return false;
4b2fecc8
JK		 1913 }
1914
b2955550
JK		 1915 /* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
1916 for (i = 0; i < entries; i++)
a66cdd98 1917 dmap[i] = force ? ~ckpt_map[i] & ~discard_map[i] :
d7bc2484 1918 (cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
b2955550 1919
0b54fb84
JK		 1920 while (force || SM_I(sbi)->dcc_info->nr_discards <=
1921 SM_I(sbi)->dcc_info->max_discards) {
b2955550
JK		 1922 start = __find_rev_next_bit(dmap, max_blocks, end + 1);
1923 if (start >= max_blocks)
1924 break;
1925
1926 end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1);
c7b41e16
YH		 1927 if (force && start && end != max_blocks
1928 && (end - start) < cpc->trim_minlen)
1929 continue;
1930
25290fa5
JK		 1931 if (check_only)
1932 return true;
1933
a7eeb823
CY		 1934 if (!de) {
1935 de = f2fs_kmem_cache_alloc(discard_entry_slab,
1936 GFP_F2FS_ZERO);
1937 de->start_blkaddr = START_BLOCK(sbi, cpc->trim_start);
1938 list_add_tail(&de->list, head);
1939 }
1940
1941 for (i = start; i < end; i++)
1942 __set_bit_le(i, (void *)de->discard_map);
1943
1944 SM_I(sbi)->dcc_info->nr_discards += end - start;
b2955550 1945 }
25290fa5 1946 return false;
b2955550
JK		 1947}
1948
af8ff65b
CY		 1949static void release_discard_addr(struct discard_entry *entry)
1950{
1951 list_del(&entry->list);
1952 kmem_cache_free(discard_entry_slab, entry);
1953}
1954
4d57b86d 1955void f2fs_release_discard_addrs(struct f2fs_sb_info *sbi)
4b2fecc8 1956{
46f84c2c 1957 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
4b2fecc8
JK		 1958 struct discard_entry *entry, *this;
1959
1960 /* drop caches */
af8ff65b
CY		 1961 list_for_each_entry_safe(entry, this, head, list)
1962 release_discard_addr(entry);
4b2fecc8
JK		 1963}
1964
0a8165d7 1965/*
4d57b86d 1966 * Should call f2fs_clear_prefree_segments after checkpoint is done.
351df4b2
JK		 1967 */
1968static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
1969{
1970 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
b65ee148 1971 unsigned int segno;
351df4b2
JK		 1972
1973 mutex_lock(&dirty_i->seglist_lock);
7cd8558b 1974 for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], MAIN_SEGS(sbi))
d0b9e42a 1975 __set_test_and_free(sbi, segno, false);
351df4b2
JK		 1976 mutex_unlock(&dirty_i->seglist_lock);
1977}
1978
4d57b86d
CY		 1979void f2fs_clear_prefree_segments(struct f2fs_sb_info *sbi,
1980 struct cp_control *cpc)
351df4b2 1981{
969d1b18
CY		 1982 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1983 struct list_head *head = &dcc->entry_list;
2d7b822a 1984 struct discard_entry *entry, *this;
351df4b2 1985 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
29e59c14 1986 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
29e59c14 1987 unsigned int start = 0, end = -1;
36abef4e 1988 unsigned int secno, start_segno;
c473f1a9 1989 bool force = (cpc->reason & CP_DISCARD);
b0332a0f 1990 bool need_align = f2fs_lfs_mode(sbi) && __is_large_section(sbi);
351df4b2
JK		 1991
1992 mutex_lock(&dirty_i->seglist_lock);
29e59c14 1993
351df4b2 1994 while (1) {
29e59c14 1995 int i;
ad6672bb
YS		 1996
1997 if (need_align && end != -1)
1998 end--;
7cd8558b
JK		 1999 start = find_next_bit(prefree_map, MAIN_SEGS(sbi), end + 1);
2000 if (start >= MAIN_SEGS(sbi))
351df4b2 2001 break;
7cd8558b
JK		 2002 end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
2003 start + 1);
29e59c14 2004
ad6672bb
YS		 2005 if (need_align) {
2006 start = rounddown(start, sbi->segs_per_sec);
2007 end = roundup(end, sbi->segs_per_sec);
2008 }
29e59c14 2009
ad6672bb
YS		 2010 for (i = start; i < end; i++) {
2011 if (test_and_clear_bit(i, prefree_map))
2012 dirty_i->nr_dirty[PRE]--;
2013 }
29e59c14 2014
7d20c8ab 2015 if (!f2fs_realtime_discard_enable(sbi))
29e59c14 2016 continue;
351df4b2 2017
650d3c4e
YH		 2018 if (force && start >= cpc->trim_start &&
2019 (end - 1) <= cpc->trim_end)
2020 continue;
2021
b0332a0f 2022 if (!f2fs_lfs_mode(sbi) || !__is_large_section(sbi)) {
36abef4e 2023 f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
37208879 2024 (end - start) << sbi->log_blocks_per_seg);
36abef4e
JK		 2025 continue;
2026 }
2027next:
4ddb1a4d
JK		 2028 secno = GET_SEC_FROM_SEG(sbi, start);
2029 start_segno = GET_SEG_FROM_SEC(sbi, secno);
36abef4e 2030 if (!IS_CURSEC(sbi, secno) &&
302bd348 2031 !get_valid_blocks(sbi, start, true))
36abef4e
JK		 2032 f2fs_issue_discard(sbi, START_BLOCK(sbi, start_segno),
2033 sbi->segs_per_sec << sbi->log_blocks_per_seg);
2034
2035 start = start_segno + sbi->segs_per_sec;
2036 if (start < end)
2037 goto next;
8b107f5b
JK		 2038 else
2039 end = start - 1;
351df4b2
JK		 2040 }
2041 mutex_unlock(&dirty_i->seglist_lock);
b2955550
JK		 2042
2043 /* send small discards */
2d7b822a 2044 list_for_each_entry_safe(entry, this, head, list) {
a7eeb823
CY		 2045 unsigned int cur_pos = 0, next_pos, len, total_len = 0;
2046 bool is_valid = test_bit_le(0, entry->discard_map);
2047
2048find_next:
2049 if (is_valid) {
2050 next_pos = find_next_zero_bit_le(entry->discard_map,
2051 sbi->blocks_per_seg, cur_pos);
2052 len = next_pos - cur_pos;
2053
7beb01f7 2054 if (f2fs_sb_has_blkzoned(sbi) ||
acfd2810 2055 (force && len < cpc->trim_minlen))
a7eeb823
CY		 2056 goto skip;
2057
2058 f2fs_issue_discard(sbi, entry->start_blkaddr + cur_pos,
2059 len);
a7eeb823
CY		 2060 total_len += len;
2061 } else {
2062 next_pos = find_next_bit_le(entry->discard_map,
2063 sbi->blocks_per_seg, cur_pos);
2064 }
836b5a63 2065skip:
a7eeb823
CY		 2066 cur_pos = next_pos;
2067 is_valid = !is_valid;
2068
2069 if (cur_pos < sbi->blocks_per_seg)
2070 goto find_next;
2071
af8ff65b 2072 release_discard_addr(entry);
969d1b18 2073 dcc->nr_discards -= total_len;
b2955550 2074 }
34e159da 2075
01983c71 2076 wake_up_discard_thread(sbi, false);
351df4b2
JK		 2077}
2078
8ed59745 2079static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84 2080{
15469963 2081 dev_t dev = sbi->sb->s_bdev->bd_dev;
0b54fb84 2082 struct discard_cmd_control *dcc;
ba48a33e 2083 int err = 0, i;
0b54fb84
JK		 2084
2085 if (SM_I(sbi)->dcc_info) {
2086 dcc = SM_I(sbi)->dcc_info;
2087 goto init_thread;
2088 }
2089
acbf054d 2090 dcc = f2fs_kzalloc(sbi, sizeof(struct discard_cmd_control), GFP_KERNEL);
0b54fb84
JK		 2091 if (!dcc)
2092 return -ENOMEM;
2093
969d1b18 2094 dcc->discard_granularity = DEFAULT_DISCARD_GRANULARITY;
46f84c2c 2095 INIT_LIST_HEAD(&dcc->entry_list);
78997b56 2096 for (i = 0; i < MAX_PLIST_NUM; i++)
ba48a33e 2097 INIT_LIST_HEAD(&dcc->pend_list[i]);
46f84c2c 2098 INIT_LIST_HEAD(&dcc->wait_list);
8412663d 2099 INIT_LIST_HEAD(&dcc->fstrim_list);
15469963 2100 mutex_init(&dcc->cmd_lock);
8b8dd65f 2101 atomic_set(&dcc->issued_discard, 0);
72691af6 2102 atomic_set(&dcc->queued_discard, 0);
5f32366a 2103 atomic_set(&dcc->discard_cmd_cnt, 0);
0b54fb84 2104 dcc->nr_discards = 0;
d618ebaf 2105 dcc->max_discards = MAIN_SEGS(sbi) << sbi->log_blocks_per_seg;
d84d1cbd 2106 dcc->undiscard_blks = 0;
20ee4382 2107 dcc->next_pos = 0;
4dada3fd 2108 dcc->root = RB_ROOT_CACHED;
67fce70b 2109 dcc->rbtree_check = false;
0b54fb84 2110
15469963 2111 init_waitqueue_head(&dcc->discard_wait_queue);
0b54fb84
JK		 2112 SM_I(sbi)->dcc_info = dcc;
2113init_thread:
15469963
JK		 2114 dcc->f2fs_issue_discard = kthread_run(issue_discard_thread, sbi,
2115 "f2fs_discard-%u:%u", MAJOR(dev), MINOR(dev));
2116 if (IS_ERR(dcc->f2fs_issue_discard)) {
2117 err = PTR_ERR(dcc->f2fs_issue_discard);
c8eb7024 2118 kfree(dcc);
15469963
JK		 2119 SM_I(sbi)->dcc_info = NULL;
2120 return err;
2121 }
2122
0b54fb84
JK		 2123 return err;
2124}
2125
f099405f 2126static void destroy_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84
JK		 2127{
2128 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
2129
f099405f
CY		 2130 if (!dcc)
2131 return;
2132
4d57b86d 2133 f2fs_stop_discard_thread(sbi);
f099405f 2134
04f9287a
CY		 2135 /*
2136 * Recovery can cache discard commands, so in error path of
2137 * fill_super(), it needs to give a chance to handle them.
2138 */
2139 if (unlikely(atomic_read(&dcc->discard_cmd_cnt)))
2140 f2fs_issue_discard_timeout(sbi);
2141
c8eb7024 2142 kfree(dcc);
f099405f 2143 SM_I(sbi)->dcc_info = NULL;
0b54fb84
JK		 2144}
2145
184a5cd2 2146static bool __mark_sit_entry_dirty(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 2147{
2148 struct sit_info *sit_i = SIT_I(sbi);
184a5cd2
CY		 2149
2150 if (!__test_and_set_bit(segno, sit_i->dirty_sentries_bitmap)) {
351df4b2 2151 sit_i->dirty_sentries++;
184a5cd2
CY		 2152 return false;
2153 }
2154
2155 return true;
351df4b2
JK		 2156}
2157
2158static void __set_sit_entry_type(struct f2fs_sb_info *sbi, int type,
2159 unsigned int segno, int modified)
2160{
2161 struct seg_entry *se = get_seg_entry(sbi, segno);
2162 se->type = type;
2163 if (modified)
2164 __mark_sit_entry_dirty(sbi, segno);
2165}
2166
c5d02785
CY		 2167static inline unsigned long long get_segment_mtime(struct f2fs_sb_info *sbi,
2168 block_t blkaddr)
6f3a01ae
CY		 2169{
2170 unsigned int segno = GET_SEGNO(sbi, blkaddr);
c5d02785
CY		 2171
2172 if (segno == NULL_SEGNO)
2173 return 0;
2174 return get_seg_entry(sbi, segno)->mtime;
2175}
2176
2177static void update_segment_mtime(struct f2fs_sb_info *sbi, block_t blkaddr,
2178 unsigned long long old_mtime)
2179{
2180 struct seg_entry *se;
2181 unsigned int segno = GET_SEGNO(sbi, blkaddr);
2182 unsigned long long ctime = get_mtime(sbi, false);
2183 unsigned long long mtime = old_mtime ? old_mtime : ctime;
2184
2185 if (segno == NULL_SEGNO)
2186 return;
2187
2188 se = get_seg_entry(sbi, segno);
6f3a01ae
CY		 2189
2190 if (!se->mtime)
2191 se->mtime = mtime;
2192 else
2193 se->mtime = div_u64(se->mtime * se->valid_blocks + mtime,
2194 se->valid_blocks + 1);
2195
c5d02785
CY		 2196 if (ctime > SIT_I(sbi)->max_mtime)
2197 SIT_I(sbi)->max_mtime = ctime;
6f3a01ae
CY		 2198}
2199
351df4b2
JK		 2200static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
2201{
2202 struct seg_entry *se;
2203 unsigned int segno, offset;
2204 long int new_vblocks;
6415fedc
YS		 2205 bool exist;
2206#ifdef CONFIG_F2FS_CHECK_FS
2207 bool mir_exist;
2208#endif
351df4b2
JK		 2209
2210 segno = GET_SEGNO(sbi, blkaddr);
2211
2212 se = get_seg_entry(sbi, segno);
2213 new_vblocks = se->valid_blocks + del;
491c0854 2214 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
351df4b2 2215
9feffe14 2216 f2fs_bug_on(sbi, (new_vblocks < 0 ||
de881df9 2217 (new_vblocks > f2fs_usable_blks_in_seg(sbi, segno))));
351df4b2
JK		 2218
2219 se->valid_blocks = new_vblocks;
351df4b2
JK		 2220
2221 /* Update valid block bitmap */
2222 if (del > 0) {
6415fedc 2223 exist = f2fs_test_and_set_bit(offset, se->cur_valid_map);
355e7891 2224#ifdef CONFIG_F2FS_CHECK_FS
6415fedc
YS		 2225 mir_exist = f2fs_test_and_set_bit(offset,
2226 se->cur_valid_map_mir);
2227 if (unlikely(exist != mir_exist)) {
dcbb4c10
JP		 2228 f2fs_err(sbi, "Inconsistent error when setting bitmap, blk:%u, old bit:%d",
2229 blkaddr, exist);
05796763 2230 f2fs_bug_on(sbi, 1);
6415fedc 2231 }
355e7891 2232#endif
6415fedc 2233 if (unlikely(exist)) {
dcbb4c10
JP		 2234 f2fs_err(sbi, "Bitmap was wrongly set, blk:%u",
2235 blkaddr);
6415fedc 2236 f2fs_bug_on(sbi, 1);
35ee82ca
YS		 2237 se->valid_blocks--;
2238 del = 0;
355e7891 2239 }
6415fedc 2240
7d20c8ab 2241 if (!f2fs_test_and_set_bit(offset, se->discard_map))
a66cdd98 2242 sbi->discard_blks--;
720037f9 2243
899fee36
CY		 2244 /*
2245 * SSR should never reuse block which is checkpointed
2246 * or newly invalidated.
2247 */
2248 if (!is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
720037f9
JK		 2249 if (!f2fs_test_and_set_bit(offset, se->ckpt_valid_map))
2250 se->ckpt_valid_blocks++;
2251 }
351df4b2 2252 } else {
6415fedc 2253 exist = f2fs_test_and_clear_bit(offset, se->cur_valid_map);
355e7891 2254#ifdef CONFIG_F2FS_CHECK_FS
6415fedc
YS		 2255 mir_exist = f2fs_test_and_clear_bit(offset,
2256 se->cur_valid_map_mir);
2257 if (unlikely(exist != mir_exist)) {
dcbb4c10
JP		 2258 f2fs_err(sbi, "Inconsistent error when clearing bitmap, blk:%u, old bit:%d",
2259 blkaddr, exist);
05796763 2260 f2fs_bug_on(sbi, 1);
6415fedc 2261 }
355e7891 2262#endif
6415fedc 2263 if (unlikely(!exist)) {
dcbb4c10
JP		 2264 f2fs_err(sbi, "Bitmap was wrongly cleared, blk:%u",
2265 blkaddr);
6415fedc 2266 f2fs_bug_on(sbi, 1);
35ee82ca
YS		 2267 se->valid_blocks++;
2268 del = 0;
4354994f
DR		 2269 } else if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2270 /*
2271 * If checkpoints are off, we must not reuse data that
2272 * was used in the previous checkpoint. If it was used
2273 * before, we must track that to know how much space we
2274 * really have.
2275 */
c9c8ed50
CY		 2276 if (f2fs_test_bit(offset, se->ckpt_valid_map)) {
2277 spin_lock(&sbi->stat_lock);
4354994f 2278 sbi->unusable_block_count++;
c9c8ed50
CY		 2279 spin_unlock(&sbi->stat_lock);
2280 }
355e7891 2281 }
6415fedc 2282
7d20c8ab 2283 if (f2fs_test_and_clear_bit(offset, se->discard_map))
a66cdd98 2284 sbi->discard_blks++;
351df4b2
JK		 2285 }
2286 if (!f2fs_test_bit(offset, se->ckpt_valid_map))
2287 se->ckpt_valid_blocks += del;
2288
2289 __mark_sit_entry_dirty(sbi, segno);
2290
2291 /* update total number of valid blocks to be written in ckpt area */
2292 SIT_I(sbi)->written_valid_blocks += del;
2293
2c70c5e3 2294 if (__is_large_section(sbi))
351df4b2
JK		 2295 get_sec_entry(sbi, segno)->valid_blocks += del;
2296}
2297
4d57b86d 2298void f2fs_invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
351df4b2
JK		 2299{
2300 unsigned int segno = GET_SEGNO(sbi, addr);
2301 struct sit_info *sit_i = SIT_I(sbi);
2302
9850cf4a 2303 f2fs_bug_on(sbi, addr == NULL_ADDR);
4c8ff709 2304 if (addr == NEW_ADDR || addr == COMPRESS_ADDR)
351df4b2
JK		 2305 return;
2306
6aa58d8a
CY		 2307 invalidate_mapping_pages(META_MAPPING(sbi), addr, addr);
2308
351df4b2 2309 /* add it into sit main buffer */
3d26fa6b 2310 down_write(&sit_i->sentry_lock);
351df4b2 2311
c5d02785 2312 update_segment_mtime(sbi, addr, 0);
351df4b2
JK		 2313 update_sit_entry(sbi, addr, -1);
2314
2315 /* add it into dirty seglist */
2316 locate_dirty_segment(sbi, segno);
2317
3d26fa6b 2318 up_write(&sit_i->sentry_lock);
351df4b2
JK		 2319}
2320
4d57b86d 2321bool f2fs_is_checkpointed_data(struct f2fs_sb_info *sbi, block_t blkaddr)
6e2c64ad
JK		 2322{
2323 struct sit_info *sit_i = SIT_I(sbi);
2324 unsigned int segno, offset;
2325 struct seg_entry *se;
2326 bool is_cp = false;
2327
93770ab7 2328 if (!__is_valid_data_blkaddr(blkaddr))
6e2c64ad
JK		 2329 return true;
2330
3d26fa6b 2331 down_read(&sit_i->sentry_lock);
6e2c64ad
JK		 2332
2333 segno = GET_SEGNO(sbi, blkaddr);
2334 se = get_seg_entry(sbi, segno);
2335 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
2336
2337 if (f2fs_test_bit(offset, se->ckpt_valid_map))
2338 is_cp = true;
2339
3d26fa6b 2340 up_read(&sit_i->sentry_lock);
6e2c64ad
JK		 2341
2342 return is_cp;
2343}
2344
0a8165d7 2345/*
351df4b2
JK		 2346 * This function should be resided under the curseg_mutex lock
2347 */
2348static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
e79efe3b 2349 struct f2fs_summary *sum)
351df4b2
JK		 2350{
2351 struct curseg_info *curseg = CURSEG_I(sbi, type);
2352 void *addr = curseg->sum_blk;
e79efe3b 2353 addr += curseg->next_blkoff * sizeof(struct f2fs_summary);
351df4b2 2354 memcpy(addr, sum, sizeof(struct f2fs_summary));
351df4b2
JK		 2355}
2356
0a8165d7 2357/*
351df4b2
JK		 2358 * Calculate the number of current summary pages for writing
2359 */
4d57b86d 2360int f2fs_npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
351df4b2 2361{
351df4b2 2362 int valid_sum_count = 0;
9a47938b 2363 int i, sum_in_page;
351df4b2
JK		 2364
2365 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2366 if (sbi->ckpt->alloc_type[i] == SSR)
2367 valid_sum_count += sbi->blocks_per_seg;
3fa06d7b
CY		 2368 else {
2369 if (for_ra)
2370 valid_sum_count += le16_to_cpu(
2371 F2FS_CKPT(sbi)->cur_data_blkoff[i]);
2372 else
2373 valid_sum_count += curseg_blkoff(sbi, i);
2374 }
351df4b2
JK		 2375 }
2376
09cbfeaf 2377 sum_in_page = (PAGE_SIZE - 2 * SUM_JOURNAL_SIZE -
9a47938b
FL		 2378 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
2379 if (valid_sum_count <= sum_in_page)
351df4b2 2380 return 1;
9a47938b 2381 else if ((valid_sum_count - sum_in_page) <=
09cbfeaf 2382 (PAGE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
351df4b2
JK		 2383 return 2;
2384 return 3;
2385}
2386
0a8165d7 2387/*
351df4b2
JK		 2388 * Caller should put this summary page
2389 */
4d57b86d 2390struct page *f2fs_get_sum_page(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2 2391{
86f33603
JK		 2392 if (unlikely(f2fs_cp_error(sbi)))
2393 return ERR_PTR(-EIO);
2394 return f2fs_get_meta_page_retry(sbi, GET_SUM_BLOCK(sbi, segno));
351df4b2
JK		 2395}
2396
4d57b86d
CY		 2397void f2fs_update_meta_page(struct f2fs_sb_info *sbi,
2398 void *src, block_t blk_addr)
351df4b2 2399{
4d57b86d 2400 struct page *page = f2fs_grab_meta_page(sbi, blk_addr);
381722d2 2401
0537b811 2402 memcpy(page_address(page), src, PAGE_SIZE);
351df4b2
JK		 2403 set_page_dirty(page);
2404 f2fs_put_page(page, 1);
2405}
2406
381722d2
CY		 2407static void write_sum_page(struct f2fs_sb_info *sbi,
2408 struct f2fs_summary_block *sum_blk, block_t blk_addr)
2409{
4d57b86d 2410 f2fs_update_meta_page(sbi, (void *)sum_blk, blk_addr);
381722d2
CY		 2411}
2412
b7ad7512
CY		 2413static void write_current_sum_page(struct f2fs_sb_info *sbi,
2414 int type, block_t blk_addr)
2415{
2416 struct curseg_info *curseg = CURSEG_I(sbi, type);
4d57b86d 2417 struct page *page = f2fs_grab_meta_page(sbi, blk_addr);
b7ad7512
CY		 2418 struct f2fs_summary_block *src = curseg->sum_blk;
2419 struct f2fs_summary_block *dst;
2420
2421 dst = (struct f2fs_summary_block *)page_address(page);
81114baa 2422 memset(dst, 0, PAGE_SIZE);
b7ad7512
CY		 2423
2424 mutex_lock(&curseg->curseg_mutex);
2425
2426 down_read(&curseg->journal_rwsem);
2427 memcpy(&dst->journal, curseg->journal, SUM_JOURNAL_SIZE);
2428 up_read(&curseg->journal_rwsem);
2429
2430 memcpy(dst->entries, src->entries, SUM_ENTRY_SIZE);
2431 memcpy(&dst->footer, &src->footer, SUM_FOOTER_SIZE);
2432
2433 mutex_unlock(&curseg->curseg_mutex);
2434
2435 set_page_dirty(page);
2436 f2fs_put_page(page, 1);
2437}
2438
093749e2
CY		 2439static int is_next_segment_free(struct f2fs_sb_info *sbi,
2440 struct curseg_info *curseg, int type)
a7881893 2441{
a7881893
JK		 2442 unsigned int segno = curseg->segno + 1;
2443 struct free_segmap_info *free_i = FREE_I(sbi);
2444
2445 if (segno < MAIN_SEGS(sbi) && segno % sbi->segs_per_sec)
2446 return !test_bit(segno, free_i->free_segmap);
2447 return 0;
2448}
2449
0a8165d7 2450/*
351df4b2
JK		 2451 * Find a new segment from the free segments bitmap to right order
2452 * This function should be returned with success, otherwise BUG
2453 */
2454static void get_new_segment(struct f2fs_sb_info *sbi,
2455 unsigned int *newseg, bool new_sec, int dir)
2456{
2457 struct free_segmap_info *free_i = FREE_I(sbi);
351df4b2 2458 unsigned int segno, secno, zoneno;
7cd8558b 2459 unsigned int total_zones = MAIN_SECS(sbi) / sbi->secs_per_zone;
4ddb1a4d
JK		 2460 unsigned int hint = GET_SEC_FROM_SEG(sbi, *newseg);
2461 unsigned int old_zoneno = GET_ZONE_FROM_SEG(sbi, *newseg);
351df4b2
JK		 2462 unsigned int left_start = hint;
2463 bool init = true;
2464 int go_left = 0;
2465 int i;
2466
1a118ccf 2467 spin_lock(&free_i->segmap_lock);
351df4b2
JK		 2468
2469 if (!new_sec && ((*newseg + 1) % sbi->segs_per_sec)) {
2470 segno = find_next_zero_bit(free_i->free_segmap,
4ddb1a4d
JK		 2471 GET_SEG_FROM_SEC(sbi, hint + 1), *newseg + 1);
2472 if (segno < GET_SEG_FROM_SEC(sbi, hint + 1))
351df4b2
JK		 2473 goto got_it;
2474 }
2475find_other_zone:
7cd8558b
JK		 2476 secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
2477 if (secno >= MAIN_SECS(sbi)) {
351df4b2
JK		 2478 if (dir == ALLOC_RIGHT) {
2479 secno = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 2480 MAIN_SECS(sbi), 0);
2481 f2fs_bug_on(sbi, secno >= MAIN_SECS(sbi));
351df4b2
JK		 2482 } else {
2483 go_left = 1;
2484 left_start = hint - 1;
2485 }
2486 }
2487 if (go_left == 0)
2488 goto skip_left;
2489
2490 while (test_bit(left_start, free_i->free_secmap)) {
2491 if (left_start > 0) {
2492 left_start--;
2493 continue;
2494 }
2495 left_start = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 2496 MAIN_SECS(sbi), 0);
2497 f2fs_bug_on(sbi, left_start >= MAIN_SECS(sbi));
351df4b2
JK		 2498 break;
2499 }
2500 secno = left_start;
2501skip_left:
4ddb1a4d
JK		 2502 segno = GET_SEG_FROM_SEC(sbi, secno);
2503 zoneno = GET_ZONE_FROM_SEC(sbi, secno);
351df4b2
JK		 2504
2505 /* give up on finding another zone */
2506 if (!init)
2507 goto got_it;
2508 if (sbi->secs_per_zone == 1)
2509 goto got_it;
2510 if (zoneno == old_zoneno)
2511 goto got_it;
2512 if (dir == ALLOC_LEFT) {
2513 if (!go_left && zoneno + 1 >= total_zones)
2514 goto got_it;
2515 if (go_left && zoneno == 0)
2516 goto got_it;
2517 }
2518 for (i = 0; i < NR_CURSEG_TYPE; i++)
2519 if (CURSEG_I(sbi, i)->zone == zoneno)
2520 break;
2521
2522 if (i < NR_CURSEG_TYPE) {
2523 /* zone is in user, try another */
2524 if (go_left)
2525 hint = zoneno * sbi->secs_per_zone - 1;
2526 else if (zoneno + 1 >= total_zones)
2527 hint = 0;
2528 else
2529 hint = (zoneno + 1) * sbi->secs_per_zone;
2530 init = false;
2531 goto find_other_zone;
2532 }
2533got_it:
2534 /* set it as dirty segment in free segmap */
9850cf4a 2535 f2fs_bug_on(sbi, test_bit(segno, free_i->free_segmap));
351df4b2
JK		 2536 __set_inuse(sbi, segno);
2537 *newseg = segno;
1a118ccf 2538 spin_unlock(&free_i->segmap_lock);
351df4b2
JK		 2539}
2540
2541static void reset_curseg(struct f2fs_sb_info *sbi, int type, int modified)
2542{
2543 struct curseg_info *curseg = CURSEG_I(sbi, type);
2544 struct summary_footer *sum_footer;
093749e2 2545 unsigned short seg_type = curseg->seg_type;
351df4b2 2546
d0b9e42a 2547 curseg->inited = true;
351df4b2 2548 curseg->segno = curseg->next_segno;
4ddb1a4d 2549 curseg->zone = GET_ZONE_FROM_SEG(sbi, curseg->segno);
351df4b2
JK		 2550 curseg->next_blkoff = 0;
2551 curseg->next_segno = NULL_SEGNO;
2552
2553 sum_footer = &(curseg->sum_blk->footer);
2554 memset(sum_footer, 0, sizeof(struct summary_footer));
093749e2
CY		 2555
2556 sanity_check_seg_type(sbi, seg_type);
2557
2558 if (IS_DATASEG(seg_type))
351df4b2 2559 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
093749e2 2560 if (IS_NODESEG(seg_type))
351df4b2 2561 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
093749e2 2562 __set_sit_entry_type(sbi, seg_type, curseg->segno, modified);
351df4b2
JK		 2563}
2564
7a20b8a6
JK		 2565static unsigned int __get_next_segno(struct f2fs_sb_info *sbi, int type)
2566{
d0b9e42a 2567 struct curseg_info *curseg = CURSEG_I(sbi, type);
093749e2
CY		 2568 unsigned short seg_type = curseg->seg_type;
2569
2570 sanity_check_seg_type(sbi, seg_type);
d0b9e42a 2571
a7881893 2572 /* if segs_per_sec is large than 1, we need to keep original policy. */
2c70c5e3 2573 if (__is_large_section(sbi))
d0b9e42a
CY		 2574 return curseg->segno;
2575
2576 /* inmem log may not locate on any segment after mount */
2577 if (!curseg->inited)
2578 return 0;
a7881893 2579
4354994f
DR		 2580 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2581 return 0;
2582
b94929d9 2583 if (test_opt(sbi, NOHEAP) &&
093749e2 2584 (seg_type == CURSEG_HOT_DATA || IS_NODESEG(seg_type)))
7a20b8a6
JK		 2585 return 0;
2586
e066b83c
JK		 2587 if (SIT_I(sbi)->last_victim[ALLOC_NEXT])
2588 return SIT_I(sbi)->last_victim[ALLOC_NEXT];
07939627
JK		 2589
2590 /* find segments from 0 to reuse freed segments */
63189b78 2591 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
07939627
JK		 2592 return 0;
2593
d0b9e42a 2594 return curseg->segno;
7a20b8a6
JK		 2595}
2596
0a8165d7 2597/*
351df4b2
JK		 2598 * Allocate a current working segment.
2599 * This function always allocates a free segment in LFS manner.
2600 */
2601static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
2602{
2603 struct curseg_info *curseg = CURSEG_I(sbi, type);
d0b9e42a 2604 unsigned short seg_type = curseg->seg_type;
351df4b2
JK		 2605 unsigned int segno = curseg->segno;
2606 int dir = ALLOC_LEFT;
2607
d0b9e42a
CY		 2608 if (curseg->inited)
2609 write_sum_page(sbi, curseg->sum_blk,
81fb5e87 2610 GET_SUM_BLOCK(sbi, segno));
d0b9e42a 2611 if (seg_type == CURSEG_WARM_DATA || seg_type == CURSEG_COLD_DATA)
351df4b2
JK		 2612 dir = ALLOC_RIGHT;
2613
2614 if (test_opt(sbi, NOHEAP))
2615 dir = ALLOC_RIGHT;
2616
7a20b8a6 2617 segno = __get_next_segno(sbi, type);
351df4b2
JK		 2618 get_new_segment(sbi, &segno, new_sec, dir);
2619 curseg->next_segno = segno;
2620 reset_curseg(sbi, type, 1);
2621 curseg->alloc_type = LFS;
2622}
2623
2624static void __next_free_blkoff(struct f2fs_sb_info *sbi,
2625 struct curseg_info *seg, block_t start)
2626{
2627 struct seg_entry *se = get_seg_entry(sbi, seg->segno);
e81c93cf 2628 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
60a3b782 2629 unsigned long *target_map = SIT_I(sbi)->tmp_map;
e81c93cf
CL		 2630 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
2631 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
2632 int i, pos;
2633
2634 for (i = 0; i < entries; i++)
2635 target_map[i] = ckpt_map[i] | cur_map[i];
2636
2637 pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, start);
2638
2639 seg->next_blkoff = pos;
351df4b2
JK		 2640}
2641
0a8165d7 2642/*
351df4b2
JK		 2643 * If a segment is written by LFS manner, next block offset is just obtained
2644 * by increasing the current block offset. However, if a segment is written by
2645 * SSR manner, next block offset obtained by calling __next_free_blkoff
2646 */
2647static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
2648 struct curseg_info *seg)
2649{
2650 if (seg->alloc_type == SSR)
2651 __next_free_blkoff(sbi, seg, seg->next_blkoff + 1);
2652 else
2653 seg->next_blkoff++;
2654}
2655
0a8165d7 2656/*
e1c42045 2657 * This function always allocates a used segment(from dirty seglist) by SSR
351df4b2
JK		 2658 * manner, so it should recover the existing segment information of valid blocks
2659 */
093749e2 2660static void change_curseg(struct f2fs_sb_info *sbi, int type, bool flush)
351df4b2
JK		 2661{
2662 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
2663 struct curseg_info *curseg = CURSEG_I(sbi, type);
2664 unsigned int new_segno = curseg->next_segno;
2665 struct f2fs_summary_block *sum_node;
2666 struct page *sum_page;
2667
093749e2
CY		 2668 if (flush)
2669 write_sum_page(sbi, curseg->sum_blk,
2670 GET_SUM_BLOCK(sbi, curseg->segno));
2671
351df4b2
JK		 2672 __set_test_and_inuse(sbi, new_segno);
2673
2674 mutex_lock(&dirty_i->seglist_lock);
2675 __remove_dirty_segment(sbi, new_segno, PRE);
2676 __remove_dirty_segment(sbi, new_segno, DIRTY);
2677 mutex_unlock(&dirty_i->seglist_lock);
2678
2679 reset_curseg(sbi, type, 1);
2680 curseg->alloc_type = SSR;
2681 __next_free_blkoff(sbi, curseg, 0);
2682
4d57b86d 2683 sum_page = f2fs_get_sum_page(sbi, new_segno);
86f33603
JK		 2684 if (IS_ERR(sum_page)) {
2685 /* GC won't be able to use stale summary pages by cp_error */
2686 memset(curseg->sum_blk, 0, SUM_ENTRY_SIZE);
2687 return;
2688 }
025d63a4
CY		 2689 sum_node = (struct f2fs_summary_block *)page_address(sum_page);
2690 memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
2691 f2fs_put_page(sum_page, 1);
351df4b2
JK		 2692}
2693
093749e2
CY		 2694static int get_ssr_segment(struct f2fs_sb_info *sbi, int type,
2695 int alloc_mode, unsigned long long age);
2696
2697static void get_atssr_segment(struct f2fs_sb_info *sbi, int type,
2698 int target_type, int alloc_mode,
2699 unsigned long long age)
2700{
2701 struct curseg_info *curseg = CURSEG_I(sbi, type);
2702
2703 curseg->seg_type = target_type;
2704
2705 if (get_ssr_segment(sbi, type, alloc_mode, age)) {
2706 struct seg_entry *se = get_seg_entry(sbi, curseg->next_segno);
2707
2708 curseg->seg_type = se->type;
2709 change_curseg(sbi, type, true);
2710 } else {
2711 /* allocate cold segment by default */
2712 curseg->seg_type = CURSEG_COLD_DATA;
2713 new_curseg(sbi, type, true);
2714 }
2715 stat_inc_seg_type(sbi, curseg);
2716}
2717
2718static void __f2fs_init_atgc_curseg(struct f2fs_sb_info *sbi)
2719{
2720 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC);
2721
2722 if (!sbi->am.atgc_enabled)
2723 return;
2724
2725 down_read(&SM_I(sbi)->curseg_lock);
2726
2727 mutex_lock(&curseg->curseg_mutex);
2728 down_write(&SIT_I(sbi)->sentry_lock);
2729
2730 get_atssr_segment(sbi, CURSEG_ALL_DATA_ATGC, CURSEG_COLD_DATA, SSR, 0);
2731
2732 up_write(&SIT_I(sbi)->sentry_lock);
2733 mutex_unlock(&curseg->curseg_mutex);
2734
2735 up_read(&SM_I(sbi)->curseg_lock);
2736
2737}
2738void f2fs_init_inmem_curseg(struct f2fs_sb_info *sbi)
2739{
2740 __f2fs_init_atgc_curseg(sbi);
2741}
2742
2743static void __f2fs_save_inmem_curseg(struct f2fs_sb_info *sbi, int type)
d0b9e42a
CY		 2744{
2745 struct curseg_info *curseg = CURSEG_I(sbi, type);
2746
2747 mutex_lock(&curseg->curseg_mutex);
2748 if (!curseg->inited)
2749 goto out;
2750
2751 if (get_valid_blocks(sbi, curseg->segno, false)) {
2752 write_sum_page(sbi, curseg->sum_blk,
2753 GET_SUM_BLOCK(sbi, curseg->segno));
2754 } else {
2755 mutex_lock(&DIRTY_I(sbi)->seglist_lock);
2756 __set_test_and_free(sbi, curseg->segno, true);
2757 mutex_unlock(&DIRTY_I(sbi)->seglist_lock);
2758 }
2759out:
2760 mutex_unlock(&curseg->curseg_mutex);
2761}
2762
093749e2
CY		 2763void f2fs_save_inmem_curseg(struct f2fs_sb_info *sbi)
2764{
2765 __f2fs_save_inmem_curseg(sbi, CURSEG_COLD_DATA_PINNED);
2766
2767 if (sbi->am.atgc_enabled)
2768 __f2fs_save_inmem_curseg(sbi, CURSEG_ALL_DATA_ATGC);
2769}
2770
2771static void __f2fs_restore_inmem_curseg(struct f2fs_sb_info *sbi, int type)
d0b9e42a
CY		 2772{
2773 struct curseg_info *curseg = CURSEG_I(sbi, type);
2774
2775 mutex_lock(&curseg->curseg_mutex);
2776 if (!curseg->inited)
2777 goto out;
2778 if (get_valid_blocks(sbi, curseg->segno, false))
2779 goto out;
2780
2781 mutex_lock(&DIRTY_I(sbi)->seglist_lock);
2782 __set_test_and_inuse(sbi, curseg->segno);
2783 mutex_unlock(&DIRTY_I(sbi)->seglist_lock);
2784out:
2785 mutex_unlock(&curseg->curseg_mutex);
2786}
2787
093749e2
CY		 2788void f2fs_restore_inmem_curseg(struct f2fs_sb_info *sbi)
2789{
2790 __f2fs_restore_inmem_curseg(sbi, CURSEG_COLD_DATA_PINNED);
2791
2792 if (sbi->am.atgc_enabled)
2793 __f2fs_restore_inmem_curseg(sbi, CURSEG_ALL_DATA_ATGC);
2794}
2795
2796static int get_ssr_segment(struct f2fs_sb_info *sbi, int type,
2797 int alloc_mode, unsigned long long age)
43727527
JK		 2798{
2799 struct curseg_info *curseg = CURSEG_I(sbi, type);
2800 const struct victim_selection *v_ops = DIRTY_I(sbi)->v_ops;
e066b83c 2801 unsigned segno = NULL_SEGNO;
093749e2 2802 unsigned short seg_type = curseg->seg_type;
d27c3d89
CY		 2803 int i, cnt;
2804 bool reversed = false;
c192f7a4 2805
093749e2
CY		 2806 sanity_check_seg_type(sbi, seg_type);
2807
4d57b86d 2808 /* f2fs_need_SSR() already forces to do this */
093749e2 2809 if (!v_ops->get_victim(sbi, &segno, BG_GC, seg_type, alloc_mode, age)) {
e066b83c 2810 curseg->next_segno = segno;
c192f7a4 2811 return 1;
e066b83c 2812 }
43727527 2813
70d625cb 2814 /* For node segments, let's do SSR more intensively */
093749e2
CY		 2815 if (IS_NODESEG(seg_type)) {
2816 if (seg_type >= CURSEG_WARM_NODE) {
d27c3d89
CY		 2817 reversed = true;
2818 i = CURSEG_COLD_NODE;
2819 } else {
2820 i = CURSEG_HOT_NODE;
2821 }
2822 cnt = NR_CURSEG_NODE_TYPE;
70d625cb 2823 } else {
093749e2 2824 if (seg_type >= CURSEG_WARM_DATA) {
d27c3d89
CY		 2825 reversed = true;
2826 i = CURSEG_COLD_DATA;
2827 } else {
2828 i = CURSEG_HOT_DATA;
2829 }
2830 cnt = NR_CURSEG_DATA_TYPE;
70d625cb 2831 }
43727527 2832
d27c3d89 2833 for (; cnt-- > 0; reversed ? i-- : i++) {
093749e2 2834 if (i == seg_type)
c192f7a4 2835 continue;
093749e2 2836 if (!v_ops->get_victim(sbi, &segno, BG_GC, i, alloc_mode, age)) {
e066b83c 2837 curseg->next_segno = segno;
43727527 2838 return 1;
e066b83c 2839 }
c192f7a4 2840 }
4354994f
DR		 2841
2842 /* find valid_blocks=0 in dirty list */
2843 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2844 segno = get_free_segment(sbi);
2845 if (segno != NULL_SEGNO) {
2846 curseg->next_segno = segno;
2847 return 1;
2848 }
2849 }
43727527
JK		 2850 return 0;
2851}
2852
351df4b2
JK		 2853/*
2854 * flush out current segment and replace it with new segment
2855 * This function should be returned with success, otherwise BUG
2856 */
2857static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
2858 int type, bool force)
2859{
a7881893
JK		 2860 struct curseg_info *curseg = CURSEG_I(sbi, type);
2861
7b405275 2862 if (force)
351df4b2 2863 new_curseg(sbi, type, true);
5b6c6be2 2864 else if (!is_set_ckpt_flags(sbi, CP_CRC_RECOVERY_FLAG) &&
093749e2 2865 curseg->seg_type == CURSEG_WARM_NODE)
351df4b2 2866 new_curseg(sbi, type, false);
093749e2
CY		 2867 else if (curseg->alloc_type == LFS &&
2868 is_next_segment_free(sbi, curseg, type) &&
4354994f 2869 likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
a7881893 2870 new_curseg(sbi, type, false);
093749e2
CY		 2871 else if (f2fs_need_SSR(sbi) &&
2872 get_ssr_segment(sbi, type, SSR, 0))
2873 change_curseg(sbi, type, true);
351df4b2
JK		 2874 else
2875 new_curseg(sbi, type, false);
dcdfff65 2876
a7881893 2877 stat_inc_seg_type(sbi, curseg);
351df4b2
JK		 2878}
2879
0ef81833 2880void f2fs_allocate_segment_for_resize(struct f2fs_sb_info *sbi, int type,
04f0b2ea
QS		 2881 unsigned int start, unsigned int end)
2882{
2883 struct curseg_info *curseg = CURSEG_I(sbi, type);
2884 unsigned int segno;
2885
2886 down_read(&SM_I(sbi)->curseg_lock);
2887 mutex_lock(&curseg->curseg_mutex);
2888 down_write(&SIT_I(sbi)->sentry_lock);
2889
2890 segno = CURSEG_I(sbi, type)->segno;
2891 if (segno < start || segno > end)
2892 goto unlock;
2893
093749e2
CY		 2894 if (f2fs_need_SSR(sbi) && get_ssr_segment(sbi, type, SSR, 0))
2895 change_curseg(sbi, type, true);
04f0b2ea
QS		 2896 else
2897 new_curseg(sbi, type, true);
2898
2899 stat_inc_seg_type(sbi, curseg);
2900
2901 locate_dirty_segment(sbi, segno);
2902unlock:
2903 up_write(&SIT_I(sbi)->sentry_lock);
2904
2905 if (segno != curseg->segno)
dcbb4c10
JP		 2906 f2fs_notice(sbi, "For resize: curseg of type %d: %u ==> %u",
2907 type, segno, curseg->segno);
04f0b2ea
QS		 2908
2909 mutex_unlock(&curseg->curseg_mutex);
2910 up_read(&SM_I(sbi)->curseg_lock);
2911}
2912
901d745f 2913static void __allocate_new_segment(struct f2fs_sb_info *sbi, int type)
351df4b2 2914{
901d745f 2915 struct curseg_info *curseg = CURSEG_I(sbi, type);
6ae1be13 2916 unsigned int old_segno;
351df4b2 2917
d0b9e42a
CY		 2918 if (!curseg->inited)
2919 goto alloc;
2920
901d745f
CY		 2921 if (!curseg->next_blkoff &&
2922 !get_valid_blocks(sbi, curseg->segno, false) &&
2923 !get_ckpt_valid_blocks(sbi, curseg->segno))
2924 return;
3d26fa6b 2925
d0b9e42a 2926alloc:
901d745f
CY		 2927 old_segno = curseg->segno;
2928 SIT_I(sbi)->s_ops->allocate_segment(sbi, type, true);
2929 locate_dirty_segment(sbi, old_segno);
2930}
f5a53edc 2931
901d745f
CY		 2932void f2fs_allocate_new_segment(struct f2fs_sb_info *sbi, int type)
2933{
2934 down_write(&SIT_I(sbi)->sentry_lock);
2935 __allocate_new_segment(sbi, type);
2936 up_write(&SIT_I(sbi)->sentry_lock);
2937}
3d26fa6b 2938
901d745f
CY		 2939void f2fs_allocate_new_segments(struct f2fs_sb_info *sbi)
2940{
2941 int i;
2942
2943 down_write(&SIT_I(sbi)->sentry_lock);
2944 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++)
2945 __allocate_new_segment(sbi, i);
3d26fa6b 2946 up_write(&SIT_I(sbi)->sentry_lock);
351df4b2
JK		 2947}
2948
2949static const struct segment_allocation default_salloc_ops = {
2950 .allocate_segment = allocate_segment_by_default,
2951};
2952
4d57b86d
CY		 2953bool f2fs_exist_trim_candidates(struct f2fs_sb_info *sbi,
2954 struct cp_control *cpc)
25290fa5
JK		 2955{
2956 __u64 trim_start = cpc->trim_start;
2957 bool has_candidate = false;
2958
3d26fa6b 2959 down_write(&SIT_I(sbi)->sentry_lock);
25290fa5
JK		 2960 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++) {
2961 if (add_discard_addrs(sbi, cpc, true)) {
2962 has_candidate = true;
2963 break;
2964 }
2965 }
3d26fa6b 2966 up_write(&SIT_I(sbi)->sentry_lock);
25290fa5
JK		 2967
2968 cpc->trim_start = trim_start;
2969 return has_candidate;
2970}
2971
01f9cf6d 2972static unsigned int __issue_discard_cmd_range(struct f2fs_sb_info *sbi,
9a997188
JK		 2973 struct discard_policy *dpolicy,
2974 unsigned int start, unsigned int end)
2975{
2976 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
2977 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
2978 struct rb_node **insert_p = NULL, *insert_parent = NULL;
2979 struct discard_cmd *dc;
2980 struct blk_plug plug;
2981 int issued;
01f9cf6d 2982 unsigned int trimmed = 0;
9a997188
JK		 2983
2984next:
2985 issued = 0;
2986
2987 mutex_lock(&dcc->cmd_lock);
67fce70b
CY		 2988 if (unlikely(dcc->rbtree_check))
2989 f2fs_bug_on(sbi, !f2fs_check_rb_tree_consistence(sbi,
2e9b2bb2 2990 &dcc->root, false));
9a997188 2991
4d57b86d 2992 dc = (struct discard_cmd *)f2fs_lookup_rb_tree_ret(&dcc->root,
9a997188
JK		 2993 NULL, start,
2994 (struct rb_entry **)&prev_dc,
2995 (struct rb_entry **)&next_dc,
4dada3fd 2996 &insert_p, &insert_parent, true, NULL);
9a997188
JK		 2997 if (!dc)
2998 dc = next_dc;
2999
3000 blk_start_plug(&plug);
3001
3002 while (dc && dc->lstart <= end) {
3003 struct rb_node *node;
6b9cb124 3004 int err = 0;
9a997188
JK		 3005
3006 if (dc->len < dpolicy->granularity)
3007 goto skip;
3008
3009 if (dc->state != D_PREP) {
3010 list_move_tail(&dc->list, &dcc->fstrim_list);
3011 goto skip;
3012 }
3013
6b9cb124 3014 err = __submit_discard_cmd(sbi, dpolicy, dc, &issued);
9a997188 3015
35ec7d57 3016 if (issued >= dpolicy->max_requests) {
9a997188
JK		 3017 start = dc->lstart + dc->len;
3018
6b9cb124
CY		 3019 if (err)
3020 __remove_discard_cmd(sbi, dc);
3021
9a997188
JK		 3022 blk_finish_plug(&plug);
3023 mutex_unlock(&dcc->cmd_lock);
01f9cf6d 3024 trimmed += __wait_all_discard_cmd(sbi, NULL);
5df7731f 3025 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
9a997188
JK		 3026 goto next;
3027 }
3028skip:
3029 node = rb_next(&dc->rb_node);
6b9cb124
CY		 3030 if (err)
3031 __remove_discard_cmd(sbi, dc);
9a997188
JK		 3032 dc = rb_entry_safe(node, struct discard_cmd, rb_node);
3033
3034 if (fatal_signal_pending(current))
3035 break;
3036 }
3037
3038 blk_finish_plug(&plug);
3039 mutex_unlock(&dcc->cmd_lock);
01f9cf6d
CY		 3040
3041 return trimmed;
9a997188
JK		 3042}
3043
4b2fecc8
JK		 3044int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
3045{
f7ef9b83
JK		 3046 __u64 start = F2FS_BYTES_TO_BLK(range->start);
3047 __u64 end = start + F2FS_BYTES_TO_BLK(range->len) - 1;
377224c4 3048 unsigned int start_segno, end_segno;
8412663d 3049 block_t start_block, end_block;
4b2fecc8 3050 struct cp_control cpc;
78997b56 3051 struct discard_policy dpolicy;
0ea80512 3052 unsigned long long trimmed = 0;
c34f42e2 3053 int err = 0;
b0332a0f 3054 bool need_align = f2fs_lfs_mode(sbi) && __is_large_section(sbi);
4b2fecc8 3055
836b5a63 3056 if (start >= MAX_BLKADDR(sbi) || range->len < sbi->blocksize)
4b2fecc8
JK		 3057 return -EINVAL;
3058
3f16ecd9
CY		 3059 if (end < MAIN_BLKADDR(sbi))
3060 goto out;
4b2fecc8 3061
ed214a11 3062 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
dcbb4c10 3063 f2fs_warn(sbi, "Found FS corruption, run fsck to fix.");
10f966bb 3064 return -EFSCORRUPTED;
ed214a11
YH		 3065 }
3066
4b2fecc8 3067 /* start/end segment number in main_area */
7cd8558b
JK		 3068 start_segno = (start <= MAIN_BLKADDR(sbi)) ? 0 : GET_SEGNO(sbi, start);
3069 end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
3070 GET_SEGNO(sbi, end);
ad6672bb
YS		 3071 if (need_align) {
3072 start_segno = rounddown(start_segno, sbi->segs_per_sec);
3073 end_segno = roundup(end_segno + 1, sbi->segs_per_sec) - 1;
3074 }
8412663d 3075
4b2fecc8 3076 cpc.reason = CP_DISCARD;
836b5a63 3077 cpc.trim_minlen = max_t(__u64, 1, F2FS_BYTES_TO_BLK(range->minlen));
377224c4
CY		 3078 cpc.trim_start = start_segno;
3079 cpc.trim_end = end_segno;
4b2fecc8 3080
377224c4
CY		 3081 if (sbi->discard_blks == 0)
3082 goto out;
74fa5f3d 3083
fb24fea7 3084 down_write(&sbi->gc_lock);
4d57b86d 3085 err = f2fs_write_checkpoint(sbi, &cpc);
fb24fea7 3086 up_write(&sbi->gc_lock);
377224c4
CY		 3087 if (err)
3088 goto out;
8412663d 3089
e555da9f
JK		 3090 /*
3091 * We filed discard candidates, but actually we don't need to wait for
3092 * all of them, since they'll be issued in idle time along with runtime
3093 * discard option. User configuration looks like using runtime discard
3094 * or periodic fstrim instead of it.
3095 */
7d20c8ab 3096 if (f2fs_realtime_discard_enable(sbi))
5a615492
JK		 3097 goto out;
3098
3099 start_block = START_BLOCK(sbi, start_segno);
3100 end_block = START_BLOCK(sbi, end_segno + 1);
3101
3102 __init_discard_policy(sbi, &dpolicy, DPOLICY_FSTRIM, cpc.trim_minlen);
01f9cf6d
CY		 3103 trimmed = __issue_discard_cmd_range(sbi, &dpolicy,
3104 start_block, end_block);
5a615492 3105
01f9cf6d 3106 trimmed += __wait_discard_cmd_range(sbi, &dpolicy,
0ea80512 3107 start_block, end_block);
377224c4 3108out:
6eae2694
CY		 3109 if (!err)
3110 range->len = F2FS_BLK_TO_BYTES(trimmed);
c34f42e2 3111 return err;
4b2fecc8
JK		 3112}
3113
093749e2
CY		 3114static bool __has_curseg_space(struct f2fs_sb_info *sbi,
3115 struct curseg_info *curseg)
351df4b2 3116{
de881df9
AR		 3117 return curseg->next_blkoff < f2fs_usable_blks_in_seg(sbi,
3118 curseg->segno);
351df4b2
JK		 3119}
3120
4d57b86d 3121int f2fs_rw_hint_to_seg_type(enum rw_hint hint)
4f0a03d3
HL		 3122{
3123 switch (hint) {
3124 case WRITE_LIFE_SHORT:
3125 return CURSEG_HOT_DATA;
3126 case WRITE_LIFE_EXTREME:
3127 return CURSEG_COLD_DATA;
3128 default:
3129 return CURSEG_WARM_DATA;
3130 }
3131}
3132
0cdd3195
HL		 3133/* This returns write hints for each segment type. This hints will be
3134 * passed down to block layer. There are mapping tables which depend on
3135 * the mount option 'whint_mode'.
3136 *
3137 * 1) whint_mode=off. F2FS only passes down WRITE_LIFE_NOT_SET.
3138 *
3139 * 2) whint_mode=user-based. F2FS tries to pass down hints given by users.
3140 *
3141 * User F2FS Block
3142 * ---- ---- -----
3143 * META WRITE_LIFE_NOT_SET
3144 * HOT_NODE "
3145 * WARM_NODE "
3146 * COLD_NODE "
3147 * ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
3148 * extension list " "
3149 *
3150 * -- buffered io
3151 * WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
3152 * WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
3153 * WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
3154 * WRITE_LIFE_NONE " "
3155 * WRITE_LIFE_MEDIUM " "
3156 * WRITE_LIFE_LONG " "
3157 *
3158 * -- direct io
3159 * WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
3160 * WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
3161 * WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
3162 * WRITE_LIFE_NONE " WRITE_LIFE_NONE
3163 * WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
3164 * WRITE_LIFE_LONG " WRITE_LIFE_LONG
3165 *
f2e703f9
HL		 3166 * 3) whint_mode=fs-based. F2FS passes down hints with its policy.
3167 *
3168 * User F2FS Block
3169 * ---- ---- -----
3170 * META WRITE_LIFE_MEDIUM;
3171 * HOT_NODE WRITE_LIFE_NOT_SET
3172 * WARM_NODE "
3173 * COLD_NODE WRITE_LIFE_NONE
3174 * ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
3175 * extension list " "
3176 *
3177 * -- buffered io
3178 * WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
3179 * WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
3180 * WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_LONG
3181 * WRITE_LIFE_NONE " "
3182 * WRITE_LIFE_MEDIUM " "
3183 * WRITE_LIFE_LONG " "
3184 *
3185 * -- direct io
3186 * WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
3187 * WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
3188 * WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
3189 * WRITE_LIFE_NONE " WRITE_LIFE_NONE
3190 * WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
3191 * WRITE_LIFE_LONG " WRITE_LIFE_LONG
0cdd3195
HL		 3192 */
3193
4d57b86d 3194enum rw_hint f2fs_io_type_to_rw_hint(struct f2fs_sb_info *sbi,
0cdd3195
HL		 3195 enum page_type type, enum temp_type temp)
3196{
63189b78 3197 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) {
0cdd3195 3198 if (type == DATA) {
f2e703f9 3199 if (temp == WARM)
0cdd3195 3200 return WRITE_LIFE_NOT_SET;
f2e703f9
HL		 3201 else if (temp == HOT)
3202 return WRITE_LIFE_SHORT;
3203 else if (temp == COLD)
3204 return WRITE_LIFE_EXTREME;
0cdd3195
HL		 3205 } else {
3206 return WRITE_LIFE_NOT_SET;
3207 }
63189b78 3208 } else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) {
f2e703f9
HL		 3209 if (type == DATA) {
3210 if (temp == WARM)
3211 return WRITE_LIFE_LONG;
3212 else if (temp == HOT)
3213 return WRITE_LIFE_SHORT;
3214 else if (temp == COLD)
3215 return WRITE_LIFE_EXTREME;
3216 } else if (type == NODE) {
3217 if (temp == WARM || temp == HOT)
3218 return WRITE_LIFE_NOT_SET;
3219 else if (temp == COLD)
3220 return WRITE_LIFE_NONE;
3221 } else if (type == META) {
3222 return WRITE_LIFE_MEDIUM;
3223 }
0cdd3195 3224 }
f2e703f9 3225 return WRITE_LIFE_NOT_SET;
0cdd3195
HL		 3226}
3227
81377bd6 3228static int __get_segment_type_2(struct f2fs_io_info *fio)
351df4b2 3229{
81377bd6 3230 if (fio->type == DATA)
351df4b2
JK		 3231 return CURSEG_HOT_DATA;
3232 else
3233 return CURSEG_HOT_NODE;
3234}
3235
81377bd6 3236static int __get_segment_type_4(struct f2fs_io_info *fio)
351df4b2 3237{
81377bd6
JK		 3238 if (fio->type == DATA) {
3239 struct inode *inode = fio->page->mapping->host;
351df4b2
JK		 3240
3241 if (S_ISDIR(inode->i_mode))
3242 return CURSEG_HOT_DATA;
3243 else
3244 return CURSEG_COLD_DATA;
3245 } else {
81377bd6 3246 if (IS_DNODE(fio->page) && is_cold_node(fio->page))
a344b9fd 3247 return CURSEG_WARM_NODE;
351df4b2
JK		 3248 else
3249 return CURSEG_COLD_NODE;
3250 }
3251}
3252
81377bd6 3253static int __get_segment_type_6(struct f2fs_io_info *fio)
351df4b2 3254{
81377bd6
JK		 3255 if (fio->type == DATA) {
3256 struct inode *inode = fio->page->mapping->host;
351df4b2 3257
093749e2
CY		 3258 if (is_cold_data(fio->page)) {
3259 if (fio->sbi->am.atgc_enabled)
3260 return CURSEG_ALL_DATA_ATGC;
3261 else
3262 return CURSEG_COLD_DATA;
3263 }
3264 if (file_is_cold(inode) || f2fs_compressed_file(inode))
351df4b2 3265 return CURSEG_COLD_DATA;
b6a06cbb 3266 if (file_is_hot(inode) ||
b4c3ca8b 3267 is_inode_flag_set(inode, FI_HOT_DATA) ||
2079f115
CY		 3268 f2fs_is_atomic_file(inode) ||
3269 f2fs_is_volatile_file(inode))
ef095d19 3270 return CURSEG_HOT_DATA;
4d57b86d 3271 return f2fs_rw_hint_to_seg_type(inode->i_write_hint);
351df4b2 3272 } else {
81377bd6
JK		 3273 if (IS_DNODE(fio->page))
3274 return is_cold_node(fio->page) ? CURSEG_WARM_NODE :
351df4b2 3275 CURSEG_HOT_NODE;
ef095d19 3276 return CURSEG_COLD_NODE;
351df4b2
JK		 3277 }
3278}
3279
81377bd6 3280static int __get_segment_type(struct f2fs_io_info *fio)
351df4b2 3281{
a912b54d
JK		 3282 int type = 0;
3283
63189b78 3284 switch (F2FS_OPTION(fio->sbi).active_logs) {
351df4b2 3285 case 2:
a912b54d
JK		 3286 type = __get_segment_type_2(fio);
3287 break;
351df4b2 3288 case 4:
a912b54d
JK		 3289 type = __get_segment_type_4(fio);
3290 break;
3291 case 6:
3292 type = __get_segment_type_6(fio);
3293 break;
3294 default:
3295 f2fs_bug_on(fio->sbi, true);
351df4b2 3296 }
81377bd6 3297
a912b54d
JK		 3298 if (IS_HOT(type))
3299 fio->temp = HOT;
3300 else if (IS_WARM(type))
3301 fio->temp = WARM;
3302 else
3303 fio->temp = COLD;
3304 return type;
351df4b2
JK		 3305}
3306
4d57b86d 3307void f2fs_allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
bfad7c2d 3308 block_t old_blkaddr, block_t *new_blkaddr,
fb830fc5 3309 struct f2fs_summary *sum, int type,
093749e2 3310 struct f2fs_io_info *fio)
351df4b2
JK		 3311{
3312 struct sit_info *sit_i = SIT_I(sbi);
6ae1be13 3313 struct curseg_info *curseg = CURSEG_I(sbi, type);
c5d02785 3314 unsigned long long old_mtime;
093749e2
CY		 3315 bool from_gc = (type == CURSEG_ALL_DATA_ATGC);
3316 struct seg_entry *se = NULL;
351df4b2 3317
2b60311d
CY		 3318 down_read(&SM_I(sbi)->curseg_lock);
3319
351df4b2 3320 mutex_lock(&curseg->curseg_mutex);
3d26fa6b 3321 down_write(&sit_i->sentry_lock);
351df4b2 3322
093749e2
CY		 3323 if (from_gc) {
3324 f2fs_bug_on(sbi, GET_SEGNO(sbi, old_blkaddr) == NULL_SEGNO);
3325 se = get_seg_entry(sbi, GET_SEGNO(sbi, old_blkaddr));
3326 sanity_check_seg_type(sbi, se->type);
3327 f2fs_bug_on(sbi, IS_NODESEG(se->type));
3328 }
351df4b2 3329 *new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
351df4b2 3330
093749e2
CY		 3331 f2fs_bug_on(sbi, curseg->next_blkoff >= sbi->blocks_per_seg);
3332
4e6a8d9b
JK		 3333 f2fs_wait_discard_bio(sbi, *new_blkaddr);
3334
351df4b2
JK		 3335 /*
3336 * __add_sum_entry should be resided under the curseg_mutex
3337 * because, this function updates a summary entry in the
3338 * current summary block.
3339 */
e79efe3b 3340 __add_sum_entry(sbi, type, sum);
351df4b2 3341
351df4b2 3342 __refresh_next_blkoff(sbi, curseg);
dcdfff65
JK		 3343
3344 stat_inc_block_count(sbi, curseg);
351df4b2 3345
c5d02785
CY		 3346 if (from_gc) {
3347 old_mtime = get_segment_mtime(sbi, old_blkaddr);
3348 } else {
3349 update_segment_mtime(sbi, old_blkaddr, 0);
3350 old_mtime = 0;
3351 }
3352 update_segment_mtime(sbi, *new_blkaddr, old_mtime);
3353
65f1b80b
YS		 3354 /*
3355 * SIT information should be updated before segment allocation,
3356 * since SSR needs latest valid block information.
3357 */
3358 update_sit_entry(sbi, *new_blkaddr, 1);
3359 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
3360 update_sit_entry(sbi, old_blkaddr, -1);
3361
093749e2
CY		 3362 if (!__has_curseg_space(sbi, curseg)) {
3363 if (from_gc)
3364 get_atssr_segment(sbi, type, se->type,
3365 AT_SSR, se->mtime);
3366 else
3367 sit_i->s_ops->allocate_segment(sbi, type, false);
3368 }
351df4b2 3369 /*
65f1b80b
YS		 3370 * segment dirty status should be updated after segment allocation,
3371 * so we just need to update status only one time after previous
3372 * segment being closed.
351df4b2 3373 */
65f1b80b
YS		 3374 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
3375 locate_dirty_segment(sbi, GET_SEGNO(sbi, *new_blkaddr));
5e443818 3376
3d26fa6b 3377 up_write(&sit_i->sentry_lock);
351df4b2 3378
704956ec 3379 if (page && IS_NODESEG(type)) {
351df4b2
JK		 3380 fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
3381
704956ec
CY		 3382 f2fs_inode_chksum_set(sbi, page);
3383 }
3384
8223ecc4
CY		 3385 if (F2FS_IO_ALIGNED(sbi))
3386 fio->retry = false;
3387
f608c38c 3388 if (fio) {
fb830fc5
CY		 3389 struct f2fs_bio_info *io;
3390
3391 INIT_LIST_HEAD(&fio->list);
3392 fio->in_list = true;
3393 io = sbi->write_io[fio->type] + fio->temp;
3394 spin_lock(&io->io_lock);
3395 list_add_tail(&fio->list, &io->io_list);
3396 spin_unlock(&io->io_lock);
3397 }
3398
bfad7c2d 3399 mutex_unlock(&curseg->curseg_mutex);
2b60311d
CY		 3400
3401 up_read(&SM_I(sbi)->curseg_lock);
bfad7c2d
JK		 3402}
3403
39d787be
CY		 3404static void update_device_state(struct f2fs_io_info *fio)
3405{
3406 struct f2fs_sb_info *sbi = fio->sbi;
3407 unsigned int devidx;
3408
0916878d 3409 if (!f2fs_is_multi_device(sbi))
39d787be
CY		 3410 return;
3411
3412 devidx = f2fs_target_device_index(sbi, fio->new_blkaddr);
3413
3414 /* update device state for fsync */
4d57b86d 3415 f2fs_set_dirty_device(sbi, fio->ino, devidx, FLUSH_INO);
1228b482
CY		 3416
3417 /* update device state for checkpoint */
3418 if (!f2fs_test_bit(devidx, (char *)&sbi->dirty_device)) {
3419 spin_lock(&sbi->dev_lock);
3420 f2fs_set_bit(devidx, (char *)&sbi->dirty_device);
3421 spin_unlock(&sbi->dev_lock);
3422 }
39d787be
CY		 3423}
3424
05ca3632 3425static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
bfad7c2d 3426{
81377bd6 3427 int type = __get_segment_type(fio);
b0332a0f 3428 bool keep_order = (f2fs_lfs_mode(fio->sbi) && type == CURSEG_COLD_DATA);
bfad7c2d 3429
107a805d
CY		 3430 if (keep_order)
3431 down_read(&fio->sbi->io_order_lock);
0a595eba 3432reallocate:
4d57b86d 3433 f2fs_allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
093749e2 3434 &fio->new_blkaddr, sum, type, fio);
6aa58d8a
CY		 3435 if (GET_SEGNO(fio->sbi, fio->old_blkaddr) != NULL_SEGNO)
3436 invalidate_mapping_pages(META_MAPPING(fio->sbi),
3437 fio->old_blkaddr, fio->old_blkaddr);
bfad7c2d 3438
351df4b2 3439 /* writeout dirty page into bdev */
fe16efe6
CY		 3440 f2fs_submit_page_write(fio);
3441 if (fio->retry) {
0a595eba
JK		 3442 fio->old_blkaddr = fio->new_blkaddr;
3443 goto reallocate;
3444 }
fe16efe6
CY		 3445
3446 update_device_state(fio);
3447
107a805d
CY		 3448 if (keep_order)
3449 up_read(&fio->sbi->io_order_lock);
351df4b2
JK		 3450}
3451
4d57b86d 3452void f2fs_do_write_meta_page(struct f2fs_sb_info *sbi, struct page *page,
b0af6d49 3453 enum iostat_type io_type)
351df4b2 3454{
458e6197 3455 struct f2fs_io_info fio = {
05ca3632 3456 .sbi = sbi,
458e6197 3457 .type = META,
0cdd3195 3458 .temp = HOT,
04d328de 3459 .op = REQ_OP_WRITE,
70fd7614 3460 .op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
7a9d7548
CY		 3461 .old_blkaddr = page->index,
3462 .new_blkaddr = page->index,
05ca3632 3463 .page = page,
4375a336 3464 .encrypted_page = NULL,
fb830fc5 3465 .in_list = false,
458e6197
JK		 3466 };
3467
2b947003 3468 if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
04d328de 3469 fio.op_flags &= ~REQ_META;
2b947003 3470
351df4b2 3471 set_page_writeback(page);
17c50035 3472 ClearPageError(page);
b9109b0e 3473 f2fs_submit_page_write(&fio);
b0af6d49 3474
b63e7be5 3475 stat_inc_meta_count(sbi, page->index);
b0af6d49 3476 f2fs_update_iostat(sbi, io_type, F2FS_BLKSIZE);
351df4b2
JK		 3477}
3478
4d57b86d 3479void f2fs_do_write_node_page(unsigned int nid, struct f2fs_io_info *fio)
351df4b2
JK		 3480{
3481 struct f2fs_summary sum;
05ca3632 3482
351df4b2 3483 set_summary(&sum, nid, 0, 0);
05ca3632 3484 do_write_page(&sum, fio);
b0af6d49
CY		 3485
3486 f2fs_update_iostat(fio->sbi, fio->io_type, F2FS_BLKSIZE);
351df4b2
JK		 3487}
3488
4d57b86d
CY		 3489void f2fs_outplace_write_data(struct dnode_of_data *dn,
3490 struct f2fs_io_info *fio)
351df4b2 3491{
05ca3632 3492 struct f2fs_sb_info *sbi = fio->sbi;
351df4b2 3493 struct f2fs_summary sum;
351df4b2 3494
9850cf4a 3495 f2fs_bug_on(sbi, dn->data_blkaddr == NULL_ADDR);
7735730d 3496 set_summary(&sum, dn->nid, dn->ofs_in_node, fio->version);
05ca3632 3497 do_write_page(&sum, fio);
f28b3434 3498 f2fs_update_data_blkaddr(dn, fio->new_blkaddr);
b0af6d49
CY		 3499
3500 f2fs_update_iostat(sbi, fio->io_type, F2FS_BLKSIZE);
351df4b2
JK		 3501}
3502
4d57b86d 3503int f2fs_inplace_write_data(struct f2fs_io_info *fio)
351df4b2 3504{
b0af6d49 3505 int err;
d21b0f23 3506 struct f2fs_sb_info *sbi = fio->sbi;
05573d6c 3507 unsigned int segno;
b0af6d49 3508
7a9d7548 3509 fio->new_blkaddr = fio->old_blkaddr;
0cdd3195
HL		 3510 /* i/o temperature is needed for passing down write hints */
3511 __get_segment_type(fio);
d21b0f23 3512
05573d6c
CY		 3513 segno = GET_SEGNO(sbi, fio->new_blkaddr);
3514
3515 if (!IS_DATASEG(get_seg_entry(sbi, segno)->type)) {
3516 set_sbi_flag(sbi, SBI_NEED_FSCK);
2d821c12
CY		 3517 f2fs_warn(sbi, "%s: incorrect segment(%u) type, run fsck to fix.",
3518 __func__, segno);
10f966bb 3519 return -EFSCORRUPTED;
05573d6c 3520 }
d21b0f23 3521
05ca3632 3522 stat_inc_inplace_blocks(fio->sbi);
b0af6d49 3523
0e7f4197 3524 if (fio->bio && !(SM_I(sbi)->ipu_policy & (1 << F2FS_IPU_NOCACHE)))
8648de2c
CY		 3525 err = f2fs_merge_page_bio(fio);
3526 else
3527 err = f2fs_submit_page_bio(fio);
e46f6bd8 3528 if (!err) {
39d787be 3529 update_device_state(fio);
e46f6bd8
CY		 3530 f2fs_update_iostat(fio->sbi, fio->io_type, F2FS_BLKSIZE);
3531 }
b0af6d49
CY		 3532
3533 return err;
351df4b2
JK		 3534}
3535
2b60311d
CY		 3536static inline int __f2fs_get_curseg(struct f2fs_sb_info *sbi,
3537 unsigned int segno)
3538{
3539 int i;
3540
3541 for (i = CURSEG_HOT_DATA; i < NO_CHECK_TYPE; i++) {
3542 if (CURSEG_I(sbi, i)->segno == segno)
3543 break;
3544 }
3545 return i;
3546}
3547
4d57b86d 3548void f2fs_do_replace_block(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
19f106bc 3549 block_t old_blkaddr, block_t new_blkaddr,
c5d02785
CY		 3550 bool recover_curseg, bool recover_newaddr,
3551 bool from_gc)
351df4b2
JK		 3552{
3553 struct sit_info *sit_i = SIT_I(sbi);
3554 struct curseg_info *curseg;
3555 unsigned int segno, old_cursegno;
3556 struct seg_entry *se;
3557 int type;
19f106bc 3558 unsigned short old_blkoff;
351df4b2
JK		 3559
3560 segno = GET_SEGNO(sbi, new_blkaddr);
3561 se = get_seg_entry(sbi, segno);
3562 type = se->type;
3563
2b60311d
CY		 3564 down_write(&SM_I(sbi)->curseg_lock);
3565
19f106bc
CY		 3566 if (!recover_curseg) {
3567 /* for recovery flow */
3568 if (se->valid_blocks == 0 && !IS_CURSEG(sbi, segno)) {
3569 if (old_blkaddr == NULL_ADDR)
3570 type = CURSEG_COLD_DATA;
3571 else
3572 type = CURSEG_WARM_DATA;
3573 }
3574 } else {
2b60311d
CY		 3575 if (IS_CURSEG(sbi, segno)) {
3576 /* se->type is volatile as SSR allocation */
3577 type = __f2fs_get_curseg(sbi, segno);
3578 f2fs_bug_on(sbi, type == NO_CHECK_TYPE);
3579 } else {
351df4b2 3580 type = CURSEG_WARM_DATA;
2b60311d 3581 }
351df4b2 3582 }
19f106bc 3583
2c190504 3584 f2fs_bug_on(sbi, !IS_DATASEG(type));
351df4b2
JK		 3585 curseg = CURSEG_I(sbi, type);
3586
3587 mutex_lock(&curseg->curseg_mutex);
3d26fa6b 3588 down_write(&sit_i->sentry_lock);
351df4b2
JK		 3589
3590 old_cursegno = curseg->segno;
19f106bc 3591 old_blkoff = curseg->next_blkoff;
351df4b2
JK		 3592
3593 /* change the current segment */
3594 if (segno != curseg->segno) {
3595 curseg->next_segno = segno;
093749e2 3596 change_curseg(sbi, type, true);
351df4b2
JK		 3597 }
3598
491c0854 3599 curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
e79efe3b 3600 __add_sum_entry(sbi, type, sum);
351df4b2 3601
c5d02785
CY		 3602 if (!recover_curseg || recover_newaddr) {
3603 if (!from_gc)
3604 update_segment_mtime(sbi, new_blkaddr, 0);
6e2c64ad 3605 update_sit_entry(sbi, new_blkaddr, 1);
c5d02785 3606 }
6aa58d8a
CY		 3607 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO) {
3608 invalidate_mapping_pages(META_MAPPING(sbi),
3609 old_blkaddr, old_blkaddr);
c5d02785
CY		 3610 if (!from_gc)
3611 update_segment_mtime(sbi, old_blkaddr, 0);
6e2c64ad 3612 update_sit_entry(sbi, old_blkaddr, -1);
6aa58d8a 3613 }
6e2c64ad
JK		 3614
3615 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
3616 locate_dirty_segment(sbi, GET_SEGNO(sbi, new_blkaddr));
3617
351df4b2 3618 locate_dirty_segment(sbi, old_cursegno);
351df4b2 3619
19f106bc
CY		 3620 if (recover_curseg) {
3621 if (old_cursegno != curseg->segno) {
3622 curseg->next_segno = old_cursegno;
093749e2 3623 change_curseg(sbi, type, true);
19f106bc
CY		 3624 }
3625 curseg->next_blkoff = old_blkoff;
3626 }
3627
3d26fa6b 3628 up_write(&sit_i->sentry_lock);
351df4b2 3629 mutex_unlock(&curseg->curseg_mutex);
2b60311d 3630 up_write(&SM_I(sbi)->curseg_lock);
351df4b2
JK		 3631}
3632
528e3459
CY		 3633void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
3634 block_t old_addr, block_t new_addr,
28bc106b
CY		 3635 unsigned char version, bool recover_curseg,
3636 bool recover_newaddr)
528e3459
CY		 3637{
3638 struct f2fs_summary sum;
3639
3640 set_summary(&sum, dn->nid, dn->ofs_in_node, version);
3641
4d57b86d 3642 f2fs_do_replace_block(sbi, &sum, old_addr, new_addr,
c5d02785 3643 recover_curseg, recover_newaddr, false);
528e3459 3644
f28b3434 3645 f2fs_update_data_blkaddr(dn, new_addr);
528e3459
CY		 3646}
3647
93dfe2ac 3648void f2fs_wait_on_page_writeback(struct page *page,
bae0ee7a 3649 enum page_type type, bool ordered, bool locked)
93dfe2ac 3650{
93dfe2ac 3651 if (PageWriteback(page)) {
4081363f
JK		 3652 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
3653
0b20fcec 3654 /* submit cached LFS IO */
bab475c5 3655 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, type);
0b20fcec
CY		 3656 /* sbumit cached IPU IO */
3657 f2fs_submit_merged_ipu_write(sbi, NULL, page);
bae0ee7a 3658 if (ordered) {
fec1d657 3659 wait_on_page_writeback(page);
bae0ee7a
CY		 3660 f2fs_bug_on(sbi, locked && PageWriteback(page));
3661 } else {
fec1d657 3662 wait_for_stable_page(page);
bae0ee7a 3663 }
93dfe2ac
JK		 3664 }
3665}
3666
0ded69f6 3667void f2fs_wait_on_block_writeback(struct inode *inode, block_t blkaddr)
08b39fbd 3668{
0ded69f6 3669 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
08b39fbd
CY		 3670 struct page *cpage;
3671
0ded69f6
JK		 3672 if (!f2fs_post_read_required(inode))
3673 return;
3674
93770ab7 3675 if (!__is_valid_data_blkaddr(blkaddr))
08b39fbd
CY		 3676 return;
3677
08b39fbd
CY		 3678 cpage = find_lock_page(META_MAPPING(sbi), blkaddr);
3679 if (cpage) {
bae0ee7a 3680 f2fs_wait_on_page_writeback(cpage, DATA, true, true);
08b39fbd
CY		 3681 f2fs_put_page(cpage, 1);
3682 }
3683}
3684
1e78e8bd
ST		 3685void f2fs_wait_on_block_writeback_range(struct inode *inode, block_t blkaddr,
3686 block_t len)
3687{
3688 block_t i;
3689
3690 for (i = 0; i < len; i++)
3691 f2fs_wait_on_block_writeback(inode, blkaddr + i);
3692}
3693
7735730d 3694static int read_compacted_summaries(struct f2fs_sb_info *sbi)
351df4b2
JK		 3695{
3696 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3697 struct curseg_info *seg_i;
3698 unsigned char *kaddr;
3699 struct page *page;
3700 block_t start;
3701 int i, j, offset;
3702
3703 start = start_sum_block(sbi);
3704
4d57b86d 3705 page = f2fs_get_meta_page(sbi, start++);
7735730d
CY		 3706 if (IS_ERR(page))
3707 return PTR_ERR(page);
351df4b2
JK		 3708 kaddr = (unsigned char *)page_address(page);
3709
3710 /* Step 1: restore nat cache */
3711 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 3712 memcpy(seg_i->journal, kaddr, SUM_JOURNAL_SIZE);
351df4b2
JK		 3713
3714 /* Step 2: restore sit cache */
3715 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 3716 memcpy(seg_i->journal, kaddr + SUM_JOURNAL_SIZE, SUM_JOURNAL_SIZE);
351df4b2
JK		 3717 offset = 2 * SUM_JOURNAL_SIZE;
3718
3719 /* Step 3: restore summary entries */
3720 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
3721 unsigned short blk_off;
3722 unsigned int segno;
3723
3724 seg_i = CURSEG_I(sbi, i);
3725 segno = le32_to_cpu(ckpt->cur_data_segno[i]);
3726 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[i]);
3727 seg_i->next_segno = segno;
3728 reset_curseg(sbi, i, 0);
3729 seg_i->alloc_type = ckpt->alloc_type[i];
3730 seg_i->next_blkoff = blk_off;
3731
3732 if (seg_i->alloc_type == SSR)
3733 blk_off = sbi->blocks_per_seg;
3734
3735 for (j = 0; j < blk_off; j++) {
3736 struct f2fs_summary *s;
3737 s = (struct f2fs_summary *)(kaddr + offset);
3738 seg_i->sum_blk->entries[j] = *s;
3739 offset += SUMMARY_SIZE;
09cbfeaf 3740 if (offset + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 3741 SUM_FOOTER_SIZE)
3742 continue;
3743
3744 f2fs_put_page(page, 1);
3745 page = NULL;
3746
4d57b86d 3747 page = f2fs_get_meta_page(sbi, start++);
7735730d
CY		 3748 if (IS_ERR(page))
3749 return PTR_ERR(page);
351df4b2
JK		 3750 kaddr = (unsigned char *)page_address(page);
3751 offset = 0;
3752 }
3753 }
3754 f2fs_put_page(page, 1);
7735730d 3755 return 0;
351df4b2
JK		 3756}
3757
3758static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
3759{
3760 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3761 struct f2fs_summary_block *sum;
3762 struct curseg_info *curseg;
3763 struct page *new;
3764 unsigned short blk_off;
3765 unsigned int segno = 0;
3766 block_t blk_addr = 0;
7735730d 3767 int err = 0;
351df4b2
JK		 3768
3769 /* get segment number and block addr */
3770 if (IS_DATASEG(type)) {
3771 segno = le32_to_cpu(ckpt->cur_data_segno[type]);
3772 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[type -
3773 CURSEG_HOT_DATA]);
119ee914 3774 if (__exist_node_summaries(sbi))
d0b9e42a 3775 blk_addr = sum_blk_addr(sbi, NR_CURSEG_PERSIST_TYPE, type);
351df4b2
JK		 3776 else
3777 blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
3778 } else {
3779 segno = le32_to_cpu(ckpt->cur_node_segno[type -
3780 CURSEG_HOT_NODE]);
3781 blk_off = le16_to_cpu(ckpt->cur_node_blkoff[type -
3782 CURSEG_HOT_NODE]);
119ee914 3783 if (__exist_node_summaries(sbi))
351df4b2
JK		 3784 blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
3785 type - CURSEG_HOT_NODE);
3786 else
3787 blk_addr = GET_SUM_BLOCK(sbi, segno);
3788 }
3789
4d57b86d 3790 new = f2fs_get_meta_page(sbi, blk_addr);
7735730d
CY		 3791 if (IS_ERR(new))
3792 return PTR_ERR(new);
351df4b2
JK		 3793 sum = (struct f2fs_summary_block *)page_address(new);
3794
3795 if (IS_NODESEG(type)) {
119ee914 3796 if (__exist_node_summaries(sbi)) {
351df4b2
JK		 3797 struct f2fs_summary *ns = &sum->entries[0];
3798 int i;
3799 for (i = 0; i < sbi->blocks_per_seg; i++, ns++) {
3800 ns->version = 0;
3801 ns->ofs_in_node = 0;
3802 }
3803 } else {
7735730d
CY		 3804 err = f2fs_restore_node_summary(sbi, segno, sum);
3805 if (err)
3806 goto out;
351df4b2
JK		 3807 }
3808 }
3809
3810 /* set uncompleted segment to curseg */
3811 curseg = CURSEG_I(sbi, type);
3812 mutex_lock(&curseg->curseg_mutex);
b7ad7512
CY		 3813
3814 /* update journal info */
3815 down_write(&curseg->journal_rwsem);
3816 memcpy(curseg->journal, &sum->journal, SUM_JOURNAL_SIZE);
3817 up_write(&curseg->journal_rwsem);
3818
3819 memcpy(curseg->sum_blk->entries, sum->entries, SUM_ENTRY_SIZE);
3820 memcpy(&curseg->sum_blk->footer, &sum->footer, SUM_FOOTER_SIZE);
351df4b2
JK		 3821 curseg->next_segno = segno;
3822 reset_curseg(sbi, type, 0);
3823 curseg->alloc_type = ckpt->alloc_type[type];
3824 curseg->next_blkoff = blk_off;
3825 mutex_unlock(&curseg->curseg_mutex);
7735730d 3826out:
351df4b2 3827 f2fs_put_page(new, 1);
7735730d 3828 return err;
351df4b2
JK		 3829}
3830
3831static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
3832{
21d3f8e1
JQ		 3833 struct f2fs_journal *sit_j = CURSEG_I(sbi, CURSEG_COLD_DATA)->journal;
3834 struct f2fs_journal *nat_j = CURSEG_I(sbi, CURSEG_HOT_DATA)->journal;
351df4b2 3835 int type = CURSEG_HOT_DATA;
e4fc5fbf 3836 int err;
351df4b2 3837
aaec2b1d 3838 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG)) {
4d57b86d 3839 int npages = f2fs_npages_for_summary_flush(sbi, true);
3fa06d7b
CY		 3840
3841 if (npages >= 2)
4d57b86d 3842 f2fs_ra_meta_pages(sbi, start_sum_block(sbi), npages,
26879fb1 3843 META_CP, true);
3fa06d7b 3844
351df4b2 3845 /* restore for compacted data summary */
7735730d
CY		 3846 err = read_compacted_summaries(sbi);
3847 if (err)
3848 return err;
351df4b2
JK		 3849 type = CURSEG_HOT_NODE;
3850 }
3851
119ee914 3852 if (__exist_node_summaries(sbi))
d0b9e42a
CY		 3853 f2fs_ra_meta_pages(sbi,
3854 sum_blk_addr(sbi, NR_CURSEG_PERSIST_TYPE, type),
3855 NR_CURSEG_PERSIST_TYPE - type, META_CP, true);
3fa06d7b 3856
e4fc5fbf
CY		 3857 for (; type <= CURSEG_COLD_NODE; type++) {
3858 err = read_normal_summaries(sbi, type);
3859 if (err)
3860 return err;
3861 }
3862
21d3f8e1
JQ		 3863 /* sanity check for summary blocks */
3864 if (nats_in_cursum(nat_j) > NAT_JOURNAL_ENTRIES ||
9227d522 3865 sits_in_cursum(sit_j) > SIT_JOURNAL_ENTRIES) {
dcbb4c10
JP		 3866 f2fs_err(sbi, "invalid journal entries nats %u sits %u\n",
3867 nats_in_cursum(nat_j), sits_in_cursum(sit_j));
21d3f8e1 3868 return -EINVAL;
9227d522 3869 }
21d3f8e1 3870
351df4b2
JK		 3871 return 0;
3872}
3873
3874static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
3875{
3876 struct page *page;
3877 unsigned char *kaddr;
3878 struct f2fs_summary *summary;
3879 struct curseg_info *seg_i;
3880 int written_size = 0;
3881 int i, j;
3882
4d57b86d 3883 page = f2fs_grab_meta_page(sbi, blkaddr++);
351df4b2 3884 kaddr = (unsigned char *)page_address(page);
81114baa 3885 memset(kaddr, 0, PAGE_SIZE);
351df4b2
JK		 3886
3887 /* Step 1: write nat cache */
3888 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 3889 memcpy(kaddr, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 3890 written_size += SUM_JOURNAL_SIZE;
3891
3892 /* Step 2: write sit cache */
3893 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 3894 memcpy(kaddr + written_size, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 3895 written_size += SUM_JOURNAL_SIZE;
3896
351df4b2
JK		 3897 /* Step 3: write summary entries */
3898 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
3899 unsigned short blkoff;
3900 seg_i = CURSEG_I(sbi, i);
3901 if (sbi->ckpt->alloc_type[i] == SSR)
3902 blkoff = sbi->blocks_per_seg;
3903 else
3904 blkoff = curseg_blkoff(sbi, i);
3905
3906 for (j = 0; j < blkoff; j++) {
3907 if (!page) {
4d57b86d 3908 page = f2fs_grab_meta_page(sbi, blkaddr++);
351df4b2 3909 kaddr = (unsigned char *)page_address(page);
81114baa 3910 memset(kaddr, 0, PAGE_SIZE);
351df4b2
JK		 3911 written_size = 0;
3912 }
3913 summary = (struct f2fs_summary *)(kaddr + written_size);
3914 *summary = seg_i->sum_blk->entries[j];
3915 written_size += SUMMARY_SIZE;
351df4b2 3916
09cbfeaf 3917 if (written_size + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 3918 SUM_FOOTER_SIZE)
3919 continue;
3920
e8d61a74 3921 set_page_dirty(page);
351df4b2
JK		 3922 f2fs_put_page(page, 1);
3923 page = NULL;
3924 }
3925 }
e8d61a74
CY		 3926 if (page) {
3927 set_page_dirty(page);
351df4b2 3928 f2fs_put_page(page, 1);
e8d61a74 3929 }
351df4b2
JK		 3930}
3931
3932static void write_normal_summaries(struct f2fs_sb_info *sbi,
3933 block_t blkaddr, int type)
3934{
3935 int i, end;
3936 if (IS_DATASEG(type))
3937 end = type + NR_CURSEG_DATA_TYPE;
3938 else
3939 end = type + NR_CURSEG_NODE_TYPE;
3940
b7ad7512
CY		 3941 for (i = type; i < end; i++)
3942 write_current_sum_page(sbi, i, blkaddr + (i - type));
351df4b2
JK		 3943}
3944
4d57b86d 3945void f2fs_write_data_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
351df4b2 3946{
aaec2b1d 3947 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG))
351df4b2
JK		 3948 write_compacted_summaries(sbi, start_blk);
3949 else
3950 write_normal_summaries(sbi, start_blk, CURSEG_HOT_DATA);
3951}
3952
4d57b86d 3953void f2fs_write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
351df4b2 3954{
119ee914 3955 write_normal_summaries(sbi, start_blk, CURSEG_HOT_NODE);
351df4b2
JK		 3956}
3957
4d57b86d 3958int f2fs_lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
351df4b2
JK		 3959 unsigned int val, int alloc)
3960{
3961 int i;
3962
3963 if (type == NAT_JOURNAL) {
dfc08a12
CY		 3964 for (i = 0; i < nats_in_cursum(journal); i++) {
3965 if (le32_to_cpu(nid_in_journal(journal, i)) == val)
351df4b2
JK		 3966 return i;
3967 }
dfc08a12
CY		 3968 if (alloc && __has_cursum_space(journal, 1, NAT_JOURNAL))
3969 return update_nats_in_cursum(journal, 1);
351df4b2 3970 } else if (type == SIT_JOURNAL) {
dfc08a12
CY		 3971 for (i = 0; i < sits_in_cursum(journal); i++)
3972 if (le32_to_cpu(segno_in_journal(journal, i)) == val)
351df4b2 3973 return i;
dfc08a12
CY		 3974 if (alloc && __has_cursum_space(journal, 1, SIT_JOURNAL))
3975 return update_sits_in_cursum(journal, 1);
351df4b2
JK		 3976 }
3977 return -1;
3978}
3979
3980static struct page *get_current_sit_page(struct f2fs_sb_info *sbi,
3981 unsigned int segno)
3982{
86f33603 3983 return f2fs_get_meta_page(sbi, current_sit_addr(sbi, segno));
351df4b2
JK		 3984}
3985
3986static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
3987 unsigned int start)
3988{
3989 struct sit_info *sit_i = SIT_I(sbi);
068c3cd8 3990 struct page *page;
351df4b2 3991 pgoff_t src_off, dst_off;
351df4b2
JK		 3992
3993 src_off = current_sit_addr(sbi, start);
3994 dst_off = next_sit_addr(sbi, src_off);
3995
4d57b86d 3996 page = f2fs_grab_meta_page(sbi, dst_off);
068c3cd8 3997 seg_info_to_sit_page(sbi, page, start);
351df4b2 3998
068c3cd8 3999 set_page_dirty(page);
351df4b2
JK		 4000 set_to_next_sit(sit_i, start);
4001
068c3cd8 4002 return page;
351df4b2
JK		 4003}
4004
184a5cd2
CY		 4005static struct sit_entry_set *grab_sit_entry_set(void)
4006{
4007 struct sit_entry_set *ses =
80c54505 4008 f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS);
184a5cd2
CY		 4009
4010 ses->entry_cnt = 0;
4011 INIT_LIST_HEAD(&ses->set_list);
4012 return ses;
4013}
4014
4015static void release_sit_entry_set(struct sit_entry_set *ses)
4016{
4017 list_del(&ses->set_list);
4018 kmem_cache_free(sit_entry_set_slab, ses);
4019}
4020
4021static void adjust_sit_entry_set(struct sit_entry_set *ses,
4022 struct list_head *head)
4023{
4024 struct sit_entry_set *next = ses;
4025
4026 if (list_is_last(&ses->set_list, head))
4027 return;
4028
4029 list_for_each_entry_continue(next, head, set_list)
4030 if (ses->entry_cnt <= next->entry_cnt)
4031 break;
4032
4033 list_move_tail(&ses->set_list, &next->set_list);
4034}
4035
4036static void add_sit_entry(unsigned int segno, struct list_head *head)
4037{
4038 struct sit_entry_set *ses;
4039 unsigned int start_segno = START_SEGNO(segno);
4040
4041 list_for_each_entry(ses, head, set_list) {
4042 if (ses->start_segno == start_segno) {
4043 ses->entry_cnt++;
4044 adjust_sit_entry_set(ses, head);
4045 return;
4046 }
4047 }
4048
4049 ses = grab_sit_entry_set();
4050
4051 ses->start_segno = start_segno;
4052 ses->entry_cnt++;
4053 list_add(&ses->set_list, head);
4054}
4055
4056static void add_sits_in_set(struct f2fs_sb_info *sbi)
4057{
4058 struct f2fs_sm_info *sm_info = SM_I(sbi);
4059 struct list_head *set_list = &sm_info->sit_entry_set;
4060 unsigned long *bitmap = SIT_I(sbi)->dirty_sentries_bitmap;
184a5cd2
CY		 4061 unsigned int segno;
4062
7cd8558b 4063 for_each_set_bit(segno, bitmap, MAIN_SEGS(sbi))
184a5cd2
CY		 4064 add_sit_entry(segno, set_list);
4065}
4066
4067static void remove_sits_in_journal(struct f2fs_sb_info *sbi)
351df4b2
JK		 4068{
4069 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 4070 struct f2fs_journal *journal = curseg->journal;
351df4b2
JK		 4071 int i;
4072
b7ad7512 4073 down_write(&curseg->journal_rwsem);
dfc08a12 4074 for (i = 0; i < sits_in_cursum(journal); i++) {
184a5cd2
CY		 4075 unsigned int segno;
4076 bool dirtied;
4077
dfc08a12 4078 segno = le32_to_cpu(segno_in_journal(journal, i));
184a5cd2
CY		 4079 dirtied = __mark_sit_entry_dirty(sbi, segno);
4080
4081 if (!dirtied)
4082 add_sit_entry(segno, &SM_I(sbi)->sit_entry_set);
351df4b2 4083 }
dfc08a12 4084 update_sits_in_cursum(journal, -i);
b7ad7512 4085 up_write(&curseg->journal_rwsem);
351df4b2
JK		 4086}
4087
0a8165d7 4088/*
351df4b2
JK		 4089 * CP calls this function, which flushes SIT entries including sit_journal,
4090 * and moves prefree segs to free segs.
4091 */
4d57b86d 4092void f2fs_flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2
JK		 4093{
4094 struct sit_info *sit_i = SIT_I(sbi);
4095 unsigned long *bitmap = sit_i->dirty_sentries_bitmap;
4096 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 4097 struct f2fs_journal *journal = curseg->journal;
184a5cd2
CY		 4098 struct sit_entry_set *ses, *tmp;
4099 struct list_head *head = &SM_I(sbi)->sit_entry_set;
04f0b2ea 4100 bool to_journal = !is_sbi_flag_set(sbi, SBI_IS_RESIZEFS);
4b2fecc8 4101 struct seg_entry *se;
351df4b2 4102
3d26fa6b 4103 down_write(&sit_i->sentry_lock);
351df4b2 4104
2b11a74b
WL		 4105 if (!sit_i->dirty_sentries)
4106 goto out;
4107
351df4b2 4108 /*
184a5cd2
CY		 4109 * add and account sit entries of dirty bitmap in sit entry
4110 * set temporarily
351df4b2 4111 */
184a5cd2 4112 add_sits_in_set(sbi);
351df4b2 4113
184a5cd2
CY		 4114 /*
4115 * if there are no enough space in journal to store dirty sit
4116 * entries, remove all entries from journal and add and account
4117 * them in sit entry set.
4118 */
04f0b2ea
QS		 4119 if (!__has_cursum_space(journal, sit_i->dirty_sentries, SIT_JOURNAL) ||
4120 !to_journal)
184a5cd2 4121 remove_sits_in_journal(sbi);
b2955550 4122
184a5cd2
CY		 4123 /*
4124 * there are two steps to flush sit entries:
4125 * #1, flush sit entries to journal in current cold data summary block.
4126 * #2, flush sit entries to sit page.
4127 */
4128 list_for_each_entry_safe(ses, tmp, head, set_list) {
4a257ed6 4129 struct page *page = NULL;
184a5cd2
CY		 4130 struct f2fs_sit_block *raw_sit = NULL;
4131 unsigned int start_segno = ses->start_segno;
4132 unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
7cd8558b 4133 (unsigned long)MAIN_SEGS(sbi));
184a5cd2
CY		 4134 unsigned int segno = start_segno;
4135
4136 if (to_journal &&
dfc08a12 4137 !__has_cursum_space(journal, ses->entry_cnt, SIT_JOURNAL))
184a5cd2
CY		 4138 to_journal = false;
4139
b7ad7512
CY		 4140 if (to_journal) {
4141 down_write(&curseg->journal_rwsem);
4142 } else {
184a5cd2
CY		 4143 page = get_next_sit_page(sbi, start_segno);
4144 raw_sit = page_address(page);
351df4b2 4145 }
351df4b2 4146
184a5cd2
CY		 4147 /* flush dirty sit entries in region of current sit set */
4148 for_each_set_bit_from(segno, bitmap, end) {
4149 int offset, sit_offset;
4b2fecc8
JK		 4150
4151 se = get_seg_entry(sbi, segno);
56b07e7e
ZZ		 4152#ifdef CONFIG_F2FS_CHECK_FS
4153 if (memcmp(se->cur_valid_map, se->cur_valid_map_mir,
4154 SIT_VBLOCK_MAP_SIZE))
4155 f2fs_bug_on(sbi, 1);
4156#endif
184a5cd2
CY		 4157
4158 /* add discard candidates */
c473f1a9 4159 if (!(cpc->reason & CP_DISCARD)) {
4b2fecc8 4160 cpc->trim_start = segno;
25290fa5 4161 add_discard_addrs(sbi, cpc, false);
4b2fecc8 4162 }
184a5cd2
CY		 4163
4164 if (to_journal) {
4d57b86d 4165 offset = f2fs_lookup_journal_in_cursum(journal,
184a5cd2
CY		 4166 SIT_JOURNAL, segno, 1);
4167 f2fs_bug_on(sbi, offset < 0);
dfc08a12 4168 segno_in_journal(journal, offset) =
184a5cd2
CY		 4169 cpu_to_le32(segno);
4170 seg_info_to_raw_sit(se,
dfc08a12 4171 &sit_in_journal(journal, offset));
56b07e7e
ZZ		 4172 check_block_count(sbi, segno,
4173 &sit_in_journal(journal, offset));
184a5cd2
CY		 4174 } else {
4175 sit_offset = SIT_ENTRY_OFFSET(sit_i, segno);
4176 seg_info_to_raw_sit(se,
4177 &raw_sit->entries[sit_offset]);
56b07e7e
ZZ		 4178 check_block_count(sbi, segno,
4179 &raw_sit->entries[sit_offset]);
184a5cd2 4180 }
351df4b2 4181
184a5cd2
CY		 4182 __clear_bit(segno, bitmap);
4183 sit_i->dirty_sentries--;
4184 ses->entry_cnt--;
351df4b2
JK		 4185 }
4186
b7ad7512
CY		 4187 if (to_journal)
4188 up_write(&curseg->journal_rwsem);
4189 else
184a5cd2
CY		 4190 f2fs_put_page(page, 1);
4191
4192 f2fs_bug_on(sbi, ses->entry_cnt);
4193 release_sit_entry_set(ses);
351df4b2 4194 }
184a5cd2
CY		 4195
4196 f2fs_bug_on(sbi, !list_empty(head));
4197 f2fs_bug_on(sbi, sit_i->dirty_sentries);
184a5cd2 4198out:
c473f1a9 4199 if (cpc->reason & CP_DISCARD) {
650d3c4e
YH		 4200 __u64 trim_start = cpc->trim_start;
4201
4b2fecc8 4202 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++)
25290fa5 4203 add_discard_addrs(sbi, cpc, false);
650d3c4e
YH		 4204
4205 cpc->trim_start = trim_start;
4b2fecc8 4206 }
3d26fa6b 4207 up_write(&sit_i->sentry_lock);
351df4b2 4208
351df4b2
JK		 4209 set_prefree_as_free_segments(sbi);
4210}
4211
4212static int build_sit_info(struct f2fs_sb_info *sbi)
4213{
4214 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
351df4b2
JK		 4215 struct sit_info *sit_i;
4216 unsigned int sit_segs, start;
2fde3dd1 4217 char *src_bitmap, *bitmap;
bbf9f7d9 4218 unsigned int bitmap_size, main_bitmap_size, sit_bitmap_size;
351df4b2
JK		 4219
4220 /* allocate memory for SIT information */
acbf054d 4221 sit_i = f2fs_kzalloc(sbi, sizeof(struct sit_info), GFP_KERNEL);
351df4b2
JK		 4222 if (!sit_i)
4223 return -ENOMEM;
4224
4225 SM_I(sbi)->sit_info = sit_i;
4226
9d2a789c
KC		 4227 sit_i->sentries =
4228 f2fs_kvzalloc(sbi, array_size(sizeof(struct seg_entry),
4229 MAIN_SEGS(sbi)),
4230 GFP_KERNEL);
351df4b2
JK		 4231 if (!sit_i->sentries)
4232 return -ENOMEM;
4233
bbf9f7d9
ST		 4234 main_bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
4235 sit_i->dirty_sentries_bitmap = f2fs_kvzalloc(sbi, main_bitmap_size,
628b3d14 4236 GFP_KERNEL);
351df4b2
JK		 4237 if (!sit_i->dirty_sentries_bitmap)
4238 return -ENOMEM;
4239
2fde3dd1
CY		 4240#ifdef CONFIG_F2FS_CHECK_FS
4241 bitmap_size = MAIN_SEGS(sbi) * SIT_VBLOCK_MAP_SIZE * 4;
4242#else
4243 bitmap_size = MAIN_SEGS(sbi) * SIT_VBLOCK_MAP_SIZE * 3;
4244#endif
4245 sit_i->bitmap = f2fs_kvzalloc(sbi, bitmap_size, GFP_KERNEL);
4246 if (!sit_i->bitmap)
4247 return -ENOMEM;
4248
4249 bitmap = sit_i->bitmap;
4250
7cd8558b 4251 for (start = 0; start < MAIN_SEGS(sbi); start++) {
2fde3dd1
CY		 4252 sit_i->sentries[start].cur_valid_map = bitmap;
4253 bitmap += SIT_VBLOCK_MAP_SIZE;
4254
4255 sit_i->sentries[start].ckpt_valid_map = bitmap;
4256 bitmap += SIT_VBLOCK_MAP_SIZE;
3e025740 4257
355e7891 4258#ifdef CONFIG_F2FS_CHECK_FS
2fde3dd1
CY		 4259 sit_i->sentries[start].cur_valid_map_mir = bitmap;
4260 bitmap += SIT_VBLOCK_MAP_SIZE;
355e7891
CY		 4261#endif
4262
2fde3dd1
CY		 4263 sit_i->sentries[start].discard_map = bitmap;
4264 bitmap += SIT_VBLOCK_MAP_SIZE;
351df4b2
JK		 4265 }
4266
acbf054d 4267 sit_i->tmp_map = f2fs_kzalloc(sbi, SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
60a3b782
JK		 4268 if (!sit_i->tmp_map)
4269 return -ENOMEM;
4270
2c70c5e3 4271 if (__is_large_section(sbi)) {
9d2a789c
KC		 4272 sit_i->sec_entries =
4273 f2fs_kvzalloc(sbi, array_size(sizeof(struct sec_entry),
4274 MAIN_SECS(sbi)),
4275 GFP_KERNEL);
351df4b2
JK		 4276 if (!sit_i->sec_entries)
4277 return -ENOMEM;
4278 }
4279
4280 /* get information related with SIT */
4281 sit_segs = le32_to_cpu(raw_super->segment_count_sit) >> 1;
4282
4283 /* setup SIT bitmap from ckeckpoint pack */
bbf9f7d9 4284 sit_bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
351df4b2
JK		 4285 src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
4286
bbf9f7d9 4287 sit_i->sit_bitmap = kmemdup(src_bitmap, sit_bitmap_size, GFP_KERNEL);
ae27d62e 4288 if (!sit_i->sit_bitmap)
351df4b2 4289 return -ENOMEM;
351df4b2 4290
ae27d62e 4291#ifdef CONFIG_F2FS_CHECK_FS
bbf9f7d9
ST		 4292 sit_i->sit_bitmap_mir = kmemdup(src_bitmap,
4293 sit_bitmap_size, GFP_KERNEL);
ae27d62e
CY		 4294 if (!sit_i->sit_bitmap_mir)
4295 return -ENOMEM;
bbf9f7d9
ST		 4296
4297 sit_i->invalid_segmap = f2fs_kvzalloc(sbi,
4298 main_bitmap_size, GFP_KERNEL);
4299 if (!sit_i->invalid_segmap)
4300 return -ENOMEM;
ae27d62e
CY		 4301#endif
4302
351df4b2
JK		 4303 /* init SIT information */
4304 sit_i->s_ops = &default_salloc_ops;
4305
4306 sit_i->sit_base_addr = le32_to_cpu(raw_super->sit_blkaddr);
4307 sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
c79b7ff1 4308 sit_i->written_valid_blocks = 0;
bbf9f7d9 4309 sit_i->bitmap_size = sit_bitmap_size;
351df4b2
JK		 4310 sit_i->dirty_sentries = 0;
4311 sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
4312 sit_i->elapsed_time = le64_to_cpu(sbi->ckpt->elapsed_time);
a7e679b5 4313 sit_i->mounted_time = ktime_get_boottime_seconds();
3d26fa6b 4314 init_rwsem(&sit_i->sentry_lock);
351df4b2
JK		 4315 return 0;
4316}
4317
4318static int build_free_segmap(struct f2fs_sb_info *sbi)
4319{
351df4b2
JK		 4320 struct free_segmap_info *free_i;
4321 unsigned int bitmap_size, sec_bitmap_size;
4322
4323 /* allocate memory for free segmap information */
acbf054d 4324 free_i = f2fs_kzalloc(sbi, sizeof(struct free_segmap_info), GFP_KERNEL);
351df4b2
JK		 4325 if (!free_i)
4326 return -ENOMEM;
4327
4328 SM_I(sbi)->free_info = free_i;
4329
7cd8558b 4330 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
628b3d14 4331 free_i->free_segmap = f2fs_kvmalloc(sbi, bitmap_size, GFP_KERNEL);
351df4b2
JK		 4332 if (!free_i->free_segmap)
4333 return -ENOMEM;
4334
7cd8558b 4335 sec_bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
628b3d14 4336 free_i->free_secmap = f2fs_kvmalloc(sbi, sec_bitmap_size, GFP_KERNEL);
351df4b2
JK		 4337 if (!free_i->free_secmap)
4338 return -ENOMEM;
4339
4340 /* set all segments as dirty temporarily */
4341 memset(free_i->free_segmap, 0xff, bitmap_size);
4342 memset(free_i->free_secmap, 0xff, sec_bitmap_size);
4343
4344 /* init free segmap information */
7cd8558b 4345 free_i->start_segno = GET_SEGNO_FROM_SEG0(sbi, MAIN_BLKADDR(sbi));
351df4b2
JK		 4346 free_i->free_segments = 0;
4347 free_i->free_sections = 0;
1a118ccf 4348 spin_lock_init(&free_i->segmap_lock);
351df4b2
JK		 4349 return 0;
4350}
4351
4352static int build_curseg(struct f2fs_sb_info *sbi)
4353{
1042d60f 4354 struct curseg_info *array;
351df4b2
JK		 4355 int i;
4356
d0b9e42a
CY		 4357 array = f2fs_kzalloc(sbi, array_size(NR_CURSEG_TYPE,
4358 sizeof(*array)), GFP_KERNEL);
351df4b2
JK		 4359 if (!array)
4360 return -ENOMEM;
4361
4362 SM_I(sbi)->curseg_array = array;
4363
d0b9e42a 4364 for (i = 0; i < NO_CHECK_TYPE; i++) {
351df4b2 4365 mutex_init(&array[i].curseg_mutex);
acbf054d 4366 array[i].sum_blk = f2fs_kzalloc(sbi, PAGE_SIZE, GFP_KERNEL);
351df4b2
JK		 4367 if (!array[i].sum_blk)
4368 return -ENOMEM;
b7ad7512 4369 init_rwsem(&array[i].journal_rwsem);
acbf054d
CY		 4370 array[i].journal = f2fs_kzalloc(sbi,
4371 sizeof(struct f2fs_journal), GFP_KERNEL);
b7ad7512
CY		 4372 if (!array[i].journal)
4373 return -ENOMEM;
d0b9e42a
CY		 4374 if (i < NR_PERSISTENT_LOG)
4375 array[i].seg_type = CURSEG_HOT_DATA + i;
4376 else if (i == CURSEG_COLD_DATA_PINNED)
4377 array[i].seg_type = CURSEG_COLD_DATA;
093749e2
CY		 4378 else if (i == CURSEG_ALL_DATA_ATGC)
4379 array[i].seg_type = CURSEG_COLD_DATA;
351df4b2
JK		 4380 array[i].segno = NULL_SEGNO;
4381 array[i].next_blkoff = 0;
d0b9e42a 4382 array[i].inited = false;
351df4b2
JK		 4383 }
4384 return restore_curseg_summaries(sbi);
4385}
4386
c39a1b34 4387static int build_sit_entries(struct f2fs_sb_info *sbi)
351df4b2
JK		 4388{
4389 struct sit_info *sit_i = SIT_I(sbi);
4390 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 4391 struct f2fs_journal *journal = curseg->journal;
9c094040
YH		 4392 struct seg_entry *se;
4393 struct f2fs_sit_entry sit;
74de593a
CY		 4394 int sit_blk_cnt = SIT_BLK_CNT(sbi);
4395 unsigned int i, start, end;
4396 unsigned int readed, start_blk = 0;
c39a1b34 4397 int err = 0;
8a29c126 4398 block_t total_node_blocks = 0;
351df4b2 4399
74de593a 4400 do {
4d57b86d 4401 readed = f2fs_ra_meta_pages(sbi, start_blk, BIO_MAX_PAGES,
664ba972 4402 META_SIT, true);
74de593a
CY		 4403
4404 start = start_blk * sit_i->sents_per_block;
4405 end = (start_blk + readed) * sit_i->sents_per_block;
4406
7cd8558b 4407 for (; start < end && start < MAIN_SEGS(sbi); start++) {
74de593a 4408 struct f2fs_sit_block *sit_blk;
74de593a
CY		 4409 struct page *page;
4410
9c094040 4411 se = &sit_i->sentries[start];
74de593a 4412 page = get_current_sit_page(sbi, start);
edc55aaf
JK		 4413 if (IS_ERR(page))
4414 return PTR_ERR(page);
74de593a
CY		 4415 sit_blk = (struct f2fs_sit_block *)page_address(page);
4416 sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
4417 f2fs_put_page(page, 1);
d600af23 4418
c39a1b34
JK		 4419 err = check_block_count(sbi, start, &sit);
4420 if (err)
4421 return err;
74de593a 4422 seg_info_from_raw_sit(se, &sit);
8a29c126
JK		 4423 if (IS_NODESEG(se->type))
4424 total_node_blocks += se->valid_blocks;
a66cdd98
JK		 4425
4426 /* build discard map only one time */
7d20c8ab
CY		 4427 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
4428 memset(se->discard_map, 0xff,
4429 SIT_VBLOCK_MAP_SIZE);
4430 } else {
4431 memcpy(se->discard_map,
4432 se->cur_valid_map,
4433 SIT_VBLOCK_MAP_SIZE);
4434 sbi->discard_blks +=
4435 sbi->blocks_per_seg -
4436 se->valid_blocks;
3e025740 4437 }
a66cdd98 4438
2c70c5e3 4439 if (__is_large_section(sbi))
d600af23
CY		 4440 get_sec_entry(sbi, start)->valid_blocks +=
4441 se->valid_blocks;
351df4b2 4442 }
74de593a
CY		 4443 start_blk += readed;
4444 } while (start_blk < sit_blk_cnt);
d600af23
CY		 4445
4446 down_read(&curseg->journal_rwsem);
4447 for (i = 0; i < sits_in_cursum(journal); i++) {
d600af23
CY		 4448 unsigned int old_valid_blocks;
4449
4450 start = le32_to_cpu(segno_in_journal(journal, i));
b2ca374f 4451 if (start >= MAIN_SEGS(sbi)) {
dcbb4c10
JP		 4452 f2fs_err(sbi, "Wrong journal entry on segno %u",
4453 start);
10f966bb 4454 err = -EFSCORRUPTED;
b2ca374f
JK		 4455 break;
4456 }
4457
d600af23
CY		 4458 se = &sit_i->sentries[start];
4459 sit = sit_in_journal(journal, i);
4460
4461 old_valid_blocks = se->valid_blocks;
8a29c126
JK		 4462 if (IS_NODESEG(se->type))
4463 total_node_blocks -= old_valid_blocks;
d600af23 4464
c39a1b34
JK		 4465 err = check_block_count(sbi, start, &sit);
4466 if (err)
4467 break;
d600af23 4468 seg_info_from_raw_sit(se, &sit);
8a29c126
JK		 4469 if (IS_NODESEG(se->type))
4470 total_node_blocks += se->valid_blocks;
d600af23 4471
7d20c8ab
CY		 4472 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
4473 memset(se->discard_map, 0xff, SIT_VBLOCK_MAP_SIZE);
4474 } else {
4475 memcpy(se->discard_map, se->cur_valid_map,
4476 SIT_VBLOCK_MAP_SIZE);
4477 sbi->discard_blks += old_valid_blocks;
4478 sbi->discard_blks -= se->valid_blocks;
d600af23
CY		 4479 }
4480
2c70c5e3 4481 if (__is_large_section(sbi)) {
d600af23 4482 get_sec_entry(sbi, start)->valid_blocks +=
a9af3fdc
CY		 4483 se->valid_blocks;
4484 get_sec_entry(sbi, start)->valid_blocks -=
4485 old_valid_blocks;
4486 }
d600af23
CY		 4487 }
4488 up_read(&curseg->journal_rwsem);
8a29c126
JK		 4489
4490 if (!err && total_node_blocks != valid_node_count(sbi)) {
dcbb4c10
JP		 4491 f2fs_err(sbi, "SIT is corrupted node# %u vs %u",
4492 total_node_blocks, valid_node_count(sbi));
10f966bb 4493 err = -EFSCORRUPTED;
8a29c126
JK		 4494 }
4495
c39a1b34 4496 return err;
351df4b2
JK		 4497}
4498
4499static void init_free_segmap(struct f2fs_sb_info *sbi)
4500{
4501 unsigned int start;
4502 int type;
de881df9 4503 struct seg_entry *sentry;
351df4b2 4504
7cd8558b 4505 for (start = 0; start < MAIN_SEGS(sbi); start++) {
de881df9
AR		 4506 if (f2fs_usable_blks_in_seg(sbi, start) == 0)
4507 continue;
4508 sentry = get_seg_entry(sbi, start);
351df4b2
JK		 4509 if (!sentry->valid_blocks)
4510 __set_free(sbi, start);
c79b7ff1
JK		 4511 else
4512 SIT_I(sbi)->written_valid_blocks +=
4513 sentry->valid_blocks;
351df4b2
JK		 4514 }
4515
4516 /* set use the current segments */
4517 for (type = CURSEG_HOT_DATA; type <= CURSEG_COLD_NODE; type++) {
4518 struct curseg_info *curseg_t = CURSEG_I(sbi, type);
4519 __set_test_and_inuse(sbi, curseg_t->segno);
4520 }
4521}
4522
4523static void init_dirty_segmap(struct f2fs_sb_info *sbi)
4524{
4525 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
4526 struct free_segmap_info *free_i = FREE_I(sbi);
da52f8ad 4527 unsigned int segno = 0, offset = 0, secno;
de881df9 4528 block_t valid_blocks, usable_blks_in_seg;
123aaf77 4529 block_t blks_per_sec = BLKS_PER_SEC(sbi);
351df4b2 4530
8736fbf0 4531 while (1) {
351df4b2 4532 /* find dirty segment based on free segmap */
7cd8558b
JK		 4533 segno = find_next_inuse(free_i, MAIN_SEGS(sbi), offset);
4534 if (segno >= MAIN_SEGS(sbi))
351df4b2
JK		 4535 break;
4536 offset = segno + 1;
302bd348 4537 valid_blocks = get_valid_blocks(sbi, segno, false);
de881df9
AR		 4538 usable_blks_in_seg = f2fs_usable_blks_in_seg(sbi, segno);
4539 if (valid_blocks == usable_blks_in_seg || !valid_blocks)
351df4b2 4540 continue;
de881df9 4541 if (valid_blocks > usable_blks_in_seg) {
ec325b52
JK		 4542 f2fs_bug_on(sbi, 1);
4543 continue;
4544 }
351df4b2
JK		 4545 mutex_lock(&dirty_i->seglist_lock);
4546 __locate_dirty_segment(sbi, segno, DIRTY);
4547 mutex_unlock(&dirty_i->seglist_lock);
4548 }
da52f8ad
JQ		 4549
4550 if (!__is_large_section(sbi))
4551 return;
4552
4553 mutex_lock(&dirty_i->seglist_lock);
4554 for (segno = 0; segno < MAIN_SECS(sbi); segno += blks_per_sec) {
4555 valid_blocks = get_valid_blocks(sbi, segno, true);
4556 secno = GET_SEC_FROM_SEG(sbi, segno);
4557
4558 if (!valid_blocks || valid_blocks == blks_per_sec)
4559 continue;
4560 if (IS_CURSEC(sbi, secno))
4561 continue;
4562 set_bit(secno, dirty_i->dirty_secmap);
4563 }
4564 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 4565}
4566
5ec4e49f 4567static int init_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 4568{
4569 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
7cd8558b 4570 unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
351df4b2 4571
628b3d14 4572 dirty_i->victim_secmap = f2fs_kvzalloc(sbi, bitmap_size, GFP_KERNEL);
5ec4e49f 4573 if (!dirty_i->victim_secmap)
351df4b2
JK		 4574 return -ENOMEM;
4575 return 0;
4576}
4577
4578static int build_dirty_segmap(struct f2fs_sb_info *sbi)
4579{
4580 struct dirty_seglist_info *dirty_i;
4581 unsigned int bitmap_size, i;
4582
4583 /* allocate memory for dirty segments list information */
acbf054d
CY		 4584 dirty_i = f2fs_kzalloc(sbi, sizeof(struct dirty_seglist_info),
4585 GFP_KERNEL);
351df4b2
JK		 4586 if (!dirty_i)
4587 return -ENOMEM;
4588
4589 SM_I(sbi)->dirty_info = dirty_i;
4590 mutex_init(&dirty_i->seglist_lock);
4591
7cd8558b 4592 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
351df4b2
JK		 4593
4594 for (i = 0; i < NR_DIRTY_TYPE; i++) {
628b3d14
CY		 4595 dirty_i->dirty_segmap[i] = f2fs_kvzalloc(sbi, bitmap_size,
4596 GFP_KERNEL);
351df4b2
JK		 4597 if (!dirty_i->dirty_segmap[i])
4598 return -ENOMEM;
4599 }
4600
da52f8ad
JQ		 4601 if (__is_large_section(sbi)) {
4602 bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
4603 dirty_i->dirty_secmap = f2fs_kvzalloc(sbi,
4604 bitmap_size, GFP_KERNEL);
4605 if (!dirty_i->dirty_secmap)
4606 return -ENOMEM;
4607 }
4608
351df4b2 4609 init_dirty_segmap(sbi);
5ec4e49f 4610 return init_victim_secmap(sbi);
351df4b2
JK		 4611}
4612
c854f4d6
CY		 4613static int sanity_check_curseg(struct f2fs_sb_info *sbi)
4614{
4615 int i;
4616
4617 /*
4618 * In LFS/SSR curseg, .next_blkoff should point to an unused blkaddr;
4619 * In LFS curseg, all blkaddr after .next_blkoff should be unused.
4620 */
d0b9e42a 4621 for (i = 0; i < NR_PERSISTENT_LOG; i++) {
c854f4d6
CY		 4622 struct curseg_info *curseg = CURSEG_I(sbi, i);
4623 struct seg_entry *se = get_seg_entry(sbi, curseg->segno);
4624 unsigned int blkofs = curseg->next_blkoff;
4625
093749e2
CY		 4626 sanity_check_seg_type(sbi, curseg->seg_type);
4627
c854f4d6
CY		 4628 if (f2fs_test_bit(blkofs, se->cur_valid_map))
4629 goto out;
4630
4631 if (curseg->alloc_type == SSR)
4632 continue;
4633
4634 for (blkofs += 1; blkofs < sbi->blocks_per_seg; blkofs++) {
4635 if (!f2fs_test_bit(blkofs, se->cur_valid_map))
4636 continue;
4637out:
dcbb4c10
JP		 4638 f2fs_err(sbi,
4639 "Current segment's next free block offset is inconsistent with bitmap, logtype:%u, segno:%u, type:%u, next_blkoff:%u, blkofs:%u",
4640 i, curseg->segno, curseg->alloc_type,
4641 curseg->next_blkoff, blkofs);
10f966bb 4642 return -EFSCORRUPTED;
c854f4d6
CY		 4643 }
4644 }
4645 return 0;
4646}
4647
c426d991
SK		 4648#ifdef CONFIG_BLK_DEV_ZONED
4649
d508c94e
SK		 4650static int check_zone_write_pointer(struct f2fs_sb_info *sbi,
4651 struct f2fs_dev_info *fdev,
4652 struct blk_zone *zone)
4653{
4654 unsigned int wp_segno, wp_blkoff, zone_secno, zone_segno, segno;
4655 block_t zone_block, wp_block, last_valid_block;
4656 unsigned int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
4657 int i, s, b, ret;
4658 struct seg_entry *se;
4659
4660 if (zone->type != BLK_ZONE_TYPE_SEQWRITE_REQ)
4661 return 0;
4662
4663 wp_block = fdev->start_blk + (zone->wp >> log_sectors_per_block);
4664 wp_segno = GET_SEGNO(sbi, wp_block);
4665 wp_blkoff = wp_block - START_BLOCK(sbi, wp_segno);
4666 zone_block = fdev->start_blk + (zone->start >> log_sectors_per_block);
4667 zone_segno = GET_SEGNO(sbi, zone_block);
4668 zone_secno = GET_SEC_FROM_SEG(sbi, zone_segno);
4669
4670 if (zone_segno >= MAIN_SEGS(sbi))
4671 return 0;
4672
4673 /*
4674 * Skip check of zones cursegs point to, since
4675 * fix_curseg_write_pointer() checks them.
4676 */
4677 for (i = 0; i < NO_CHECK_TYPE; i++)
4678 if (zone_secno == GET_SEC_FROM_SEG(sbi,
4679 CURSEG_I(sbi, i)->segno))
4680 return 0;
4681
4682 /*
4683 * Get last valid block of the zone.
4684 */
4685 last_valid_block = zone_block - 1;
4686 for (s = sbi->segs_per_sec - 1; s >= 0; s--) {
4687 segno = zone_segno + s;
4688 se = get_seg_entry(sbi, segno);
4689 for (b = sbi->blocks_per_seg - 1; b >= 0; b--)
4690 if (f2fs_test_bit(b, se->cur_valid_map)) {
4691 last_valid_block = START_BLOCK(sbi, segno) + b;
4692 break;
4693 }
4694 if (last_valid_block >= zone_block)
4695 break;
4696 }
4697
4698 /*
4699 * If last valid block is beyond the write pointer, report the
4700 * inconsistency. This inconsistency does not cause write error
4701 * because the zone will not be selected for write operation until
4702 * it get discarded. Just report it.
4703 */
4704 if (last_valid_block >= wp_block) {
4705 f2fs_notice(sbi, "Valid block beyond write pointer: "
4706 "valid block[0x%x,0x%x] wp[0x%x,0x%x]",
4707 GET_SEGNO(sbi, last_valid_block),
4708 GET_BLKOFF_FROM_SEG0(sbi, last_valid_block),
4709 wp_segno, wp_blkoff);
4710 return 0;
4711 }
4712
4713 /*
4714 * If there is no valid block in the zone and if write pointer is
4715 * not at zone start, reset the write pointer.
4716 */
4717 if (last_valid_block + 1 == zone_block && zone->wp != zone->start) {
4718 f2fs_notice(sbi,
4719 "Zone without valid block has non-zero write "
4720 "pointer. Reset the write pointer: wp[0x%x,0x%x]",
4721 wp_segno, wp_blkoff);
4722 ret = __f2fs_issue_discard_zone(sbi, fdev->bdev, zone_block,
4723 zone->len >> log_sectors_per_block);
4724 if (ret) {
4725 f2fs_err(sbi, "Discard zone failed: %s (errno=%d)",
4726 fdev->path, ret);
4727 return ret;
4728 }
4729 }
4730
4731 return 0;
4732}
4733
c426d991
SK		 4734static struct f2fs_dev_info *get_target_zoned_dev(struct f2fs_sb_info *sbi,
4735 block_t zone_blkaddr)
4736{
4737 int i;
4738
4739 for (i = 0; i < sbi->s_ndevs; i++) {
4740 if (!bdev_is_zoned(FDEV(i).bdev))
4741 continue;
4742 if (sbi->s_ndevs == 1 || (FDEV(i).start_blk <= zone_blkaddr &&
4743 zone_blkaddr <= FDEV(i).end_blk))
4744 return &FDEV(i);
4745 }
4746
4747 return NULL;
4748}
4749
4750static int report_one_zone_cb(struct blk_zone *zone, unsigned int idx,
4751 void *data) {
4752 memcpy(data, zone, sizeof(struct blk_zone));
4753 return 0;
4754}
4755
4756static int fix_curseg_write_pointer(struct f2fs_sb_info *sbi, int type)
4757{
4758 struct curseg_info *cs = CURSEG_I(sbi, type);
4759 struct f2fs_dev_info *zbd;
4760 struct blk_zone zone;
4761 unsigned int cs_section, wp_segno, wp_blkoff, wp_sector_off;
4762 block_t cs_zone_block, wp_block;
4763 unsigned int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
4764 sector_t zone_sector;
4765 int err;
4766
4767 cs_section = GET_SEC_FROM_SEG(sbi, cs->segno);
4768 cs_zone_block = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, cs_section));
4769
4770 zbd = get_target_zoned_dev(sbi, cs_zone_block);
4771 if (!zbd)
4772 return 0;
4773
4774 /* report zone for the sector the curseg points to */
4775 zone_sector = (sector_t)(cs_zone_block - zbd->start_blk)
4776 << log_sectors_per_block;
4777 err = blkdev_report_zones(zbd->bdev, zone_sector, 1,
4778 report_one_zone_cb, &zone);
4779 if (err != 1) {
4780 f2fs_err(sbi, "Report zone failed: %s errno=(%d)",
4781 zbd->path, err);
4782 return err;
4783 }
4784
4785 if (zone.type != BLK_ZONE_TYPE_SEQWRITE_REQ)
4786 return 0;
4787
4788 wp_block = zbd->start_blk + (zone.wp >> log_sectors_per_block);
4789 wp_segno = GET_SEGNO(sbi, wp_block);
4790 wp_blkoff = wp_block - START_BLOCK(sbi, wp_segno);
4791 wp_sector_off = zone.wp & GENMASK(log_sectors_per_block - 1, 0);
4792
4793 if (cs->segno == wp_segno && cs->next_blkoff == wp_blkoff &&
4794 wp_sector_off == 0)
4795 return 0;
4796
4797 f2fs_notice(sbi, "Unaligned curseg[%d] with write pointer: "
4798 "curseg[0x%x,0x%x] wp[0x%x,0x%x]",
4799 type, cs->segno, cs->next_blkoff, wp_segno, wp_blkoff);
4800
4801 f2fs_notice(sbi, "Assign new section to curseg[%d]: "
4802 "curseg[0x%x,0x%x]", type, cs->segno, cs->next_blkoff);
4803 allocate_segment_by_default(sbi, type, true);
4804
d508c94e
SK		 4805 /* check consistency of the zone curseg pointed to */
4806 if (check_zone_write_pointer(sbi, zbd, &zone))
4807 return -EIO;
4808
c426d991
SK		 4809 /* check newly assigned zone */
4810 cs_section = GET_SEC_FROM_SEG(sbi, cs->segno);
4811 cs_zone_block = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, cs_section));
4812
4813 zbd = get_target_zoned_dev(sbi, cs_zone_block);
4814 if (!zbd)
4815 return 0;
4816
4817 zone_sector = (sector_t)(cs_zone_block - zbd->start_blk)
4818 << log_sectors_per_block;
4819 err = blkdev_report_zones(zbd->bdev, zone_sector, 1,
4820 report_one_zone_cb, &zone);
4821 if (err != 1) {
4822 f2fs_err(sbi, "Report zone failed: %s errno=(%d)",
4823 zbd->path, err);
4824 return err;
4825 }
4826
4827 if (zone.type != BLK_ZONE_TYPE_SEQWRITE_REQ)
4828 return 0;
4829
4830 if (zone.wp != zone.start) {
4831 f2fs_notice(sbi,
4832 "New zone for curseg[%d] is not yet discarded. "
4833 "Reset the zone: curseg[0x%x,0x%x]",
4834 type, cs->segno, cs->next_blkoff);
4835 err = __f2fs_issue_discard_zone(sbi, zbd->bdev,
4836 zone_sector >> log_sectors_per_block,
4837 zone.len >> log_sectors_per_block);
4838 if (err) {
4839 f2fs_err(sbi, "Discard zone failed: %s (errno=%d)",
4840 zbd->path, err);
4841 return err;
4842 }
4843 }
4844
4845 return 0;
4846}
4847
4848int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi)
4849{
4850 int i, ret;
4851
d0b9e42a 4852 for (i = 0; i < NR_PERSISTENT_LOG; i++) {
c426d991
SK		 4853 ret = fix_curseg_write_pointer(sbi, i);
4854 if (ret)
4855 return ret;
4856 }
4857
4858 return 0;
4859}
d508c94e
SK		 4860
4861struct check_zone_write_pointer_args {
4862 struct f2fs_sb_info *sbi;
4863 struct f2fs_dev_info *fdev;
4864};
4865
4866static int check_zone_write_pointer_cb(struct blk_zone *zone, unsigned int idx,
4867 void *data) {
4868 struct check_zone_write_pointer_args *args;
4869 args = (struct check_zone_write_pointer_args *)data;
4870
4871 return check_zone_write_pointer(args->sbi, args->fdev, zone);
4872}
4873
4874int f2fs_check_write_pointer(struct f2fs_sb_info *sbi)
4875{
4876 int i, ret;
4877 struct check_zone_write_pointer_args args;
4878
4879 for (i = 0; i < sbi->s_ndevs; i++) {
4880 if (!bdev_is_zoned(FDEV(i).bdev))
4881 continue;
4882
4883 args.sbi = sbi;
4884 args.fdev = &FDEV(i);
4885 ret = blkdev_report_zones(FDEV(i).bdev, 0, BLK_ALL_ZONES,
4886 check_zone_write_pointer_cb, &args);
4887 if (ret < 0)
4888 return ret;
4889 }
4890
4891 return 0;
4892}
de881df9
AR		 4893
4894static bool is_conv_zone(struct f2fs_sb_info *sbi, unsigned int zone_idx,
4895 unsigned int dev_idx)
4896{
4897 if (!bdev_is_zoned(FDEV(dev_idx).bdev))
4898 return true;
4899 return !test_bit(zone_idx, FDEV(dev_idx).blkz_seq);
4900}
4901
4902/* Return the zone index in the given device */
4903static unsigned int get_zone_idx(struct f2fs_sb_info *sbi, unsigned int secno,
4904 int dev_idx)
4905{
4906 block_t sec_start_blkaddr = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, secno));
4907
4908 return (sec_start_blkaddr - FDEV(dev_idx).start_blk) >>
4909 sbi->log_blocks_per_blkz;
4910}
4911
4912/*
4913 * Return the usable segments in a section based on the zone's
4914 * corresponding zone capacity. Zone is equal to a section.
4915 */
4916static inline unsigned int f2fs_usable_zone_segs_in_sec(
4917 struct f2fs_sb_info *sbi, unsigned int segno)
4918{
4919 unsigned int dev_idx, zone_idx, unusable_segs_in_sec;
4920
4921 dev_idx = f2fs_target_device_index(sbi, START_BLOCK(sbi, segno));
4922 zone_idx = get_zone_idx(sbi, GET_SEC_FROM_SEG(sbi, segno), dev_idx);
4923
4924 /* Conventional zone's capacity is always equal to zone size */
4925 if (is_conv_zone(sbi, zone_idx, dev_idx))
4926 return sbi->segs_per_sec;
4927
4928 /*
4929 * If the zone_capacity_blocks array is NULL, then zone capacity
4930 * is equal to the zone size for all zones
4931 */
4932 if (!FDEV(dev_idx).zone_capacity_blocks)
4933 return sbi->segs_per_sec;
4934
4935 /* Get the segment count beyond zone capacity block */
4936 unusable_segs_in_sec = (sbi->blocks_per_blkz -
4937 FDEV(dev_idx).zone_capacity_blocks[zone_idx]) >>
4938 sbi->log_blocks_per_seg;
4939 return sbi->segs_per_sec - unusable_segs_in_sec;
4940}
4941
4942/*
4943 * Return the number of usable blocks in a segment. The number of blocks
4944 * returned is always equal to the number of blocks in a segment for
4945 * segments fully contained within a sequential zone capacity or a
4946 * conventional zone. For segments partially contained in a sequential
4947 * zone capacity, the number of usable blocks up to the zone capacity
4948 * is returned. 0 is returned in all other cases.
4949 */
4950static inline unsigned int f2fs_usable_zone_blks_in_seg(
4951 struct f2fs_sb_info *sbi, unsigned int segno)
4952{
4953 block_t seg_start, sec_start_blkaddr, sec_cap_blkaddr;
4954 unsigned int zone_idx, dev_idx, secno;
4955
4956 secno = GET_SEC_FROM_SEG(sbi, segno);
4957 seg_start = START_BLOCK(sbi, segno);
4958 dev_idx = f2fs_target_device_index(sbi, seg_start);
4959 zone_idx = get_zone_idx(sbi, secno, dev_idx);
4960
4961 /*
4962 * Conventional zone's capacity is always equal to zone size,
4963 * so, blocks per segment is unchanged.
4964 */
4965 if (is_conv_zone(sbi, zone_idx, dev_idx))
4966 return sbi->blocks_per_seg;
4967
4968 if (!FDEV(dev_idx).zone_capacity_blocks)
4969 return sbi->blocks_per_seg;
4970
4971 sec_start_blkaddr = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, secno));
4972 sec_cap_blkaddr = sec_start_blkaddr +
4973 FDEV(dev_idx).zone_capacity_blocks[zone_idx];
4974
4975 /*
4976 * If segment starts before zone capacity and spans beyond
4977 * zone capacity, then usable blocks are from seg start to
4978 * zone capacity. If the segment starts after the zone capacity,
4979 * then there are no usable blocks.
4980 */
4981 if (seg_start >= sec_cap_blkaddr)
4982 return 0;
4983 if (seg_start + sbi->blocks_per_seg > sec_cap_blkaddr)
4984 return sec_cap_blkaddr - seg_start;
4985
4986 return sbi->blocks_per_seg;
4987}
c426d991
SK		 4988#else
4989int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi)
4990{
4991 return 0;
4992}
d508c94e
SK		 4993
4994int f2fs_check_write_pointer(struct f2fs_sb_info *sbi)
4995{
4996 return 0;
4997}
de881df9
AR		 4998
4999static inline unsigned int f2fs_usable_zone_blks_in_seg(struct f2fs_sb_info *sbi,
5000 unsigned int segno)
5001{
5002 return 0;
5003}
5004
5005static inline unsigned int f2fs_usable_zone_segs_in_sec(struct f2fs_sb_info *sbi,
5006 unsigned int segno)
5007{
5008 return 0;
5009}
c426d991 5010#endif
de881df9
AR		 5011unsigned int f2fs_usable_blks_in_seg(struct f2fs_sb_info *sbi,
5012 unsigned int segno)
5013{
5014 if (f2fs_sb_has_blkzoned(sbi))
5015 return f2fs_usable_zone_blks_in_seg(sbi, segno);
5016
5017 return sbi->blocks_per_seg;
5018}
5019
5020unsigned int f2fs_usable_segs_in_sec(struct f2fs_sb_info *sbi,
5021 unsigned int segno)
5022{
5023 if (f2fs_sb_has_blkzoned(sbi))
5024 return f2fs_usable_zone_segs_in_sec(sbi, segno);
5025
5026 return sbi->segs_per_sec;
5027}
c426d991 5028
0a8165d7 5029/*
351df4b2
JK		 5030 * Update min, max modified time for cost-benefit GC algorithm
5031 */
5032static void init_min_max_mtime(struct f2fs_sb_info *sbi)
5033{
5034 struct sit_info *sit_i = SIT_I(sbi);
5035 unsigned int segno;
5036
3d26fa6b 5037 down_write(&sit_i->sentry_lock);
351df4b2 5038
5ad25442 5039 sit_i->min_mtime = ULLONG_MAX;
351df4b2 5040
7cd8558b 5041 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
351df4b2
JK		 5042 unsigned int i;
5043 unsigned long long mtime = 0;
5044
5045 for (i = 0; i < sbi->segs_per_sec; i++)
5046 mtime += get_seg_entry(sbi, segno + i)->mtime;
5047
5048 mtime = div_u64(mtime, sbi->segs_per_sec);
5049
5050 if (sit_i->min_mtime > mtime)
5051 sit_i->min_mtime = mtime;
5052 }
a1f72ac2 5053 sit_i->max_mtime = get_mtime(sbi, false);
093749e2 5054 sit_i->dirty_max_mtime = 0;
3d26fa6b 5055 up_write(&sit_i->sentry_lock);
351df4b2
JK		 5056}
5057
4d57b86d 5058int f2fs_build_segment_manager(struct f2fs_sb_info *sbi)
351df4b2
JK		 5059{
5060 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
5061 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1042d60f 5062 struct f2fs_sm_info *sm_info;
351df4b2
JK		 5063 int err;
5064
acbf054d 5065 sm_info = f2fs_kzalloc(sbi, sizeof(struct f2fs_sm_info), GFP_KERNEL);
351df4b2
JK		 5066 if (!sm_info)
5067 return -ENOMEM;
5068
5069 /* init sm info */
5070 sbi->sm_info = sm_info;
351df4b2
JK		 5071 sm_info->seg0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
5072 sm_info->main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
5073 sm_info->segment_count = le32_to_cpu(raw_super->segment_count);
5074 sm_info->reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
5075 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
5076 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
5077 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
58c41035
JK		 5078 sm_info->rec_prefree_segments = sm_info->main_segments *
5079 DEF_RECLAIM_PREFREE_SEGMENTS / 100;
44a83499
JK		 5080 if (sm_info->rec_prefree_segments > DEF_MAX_RECLAIM_PREFREE_SEGMENTS)
5081 sm_info->rec_prefree_segments = DEF_MAX_RECLAIM_PREFREE_SEGMENTS;
5082
b0332a0f 5083 if (!f2fs_lfs_mode(sbi))
52763a4b 5084 sm_info->ipu_policy = 1 << F2FS_IPU_FSYNC;
216fbd64 5085 sm_info->min_ipu_util = DEF_MIN_IPU_UTIL;
c1ce1b02 5086 sm_info->min_fsync_blocks = DEF_MIN_FSYNC_BLOCKS;
853137ce 5087 sm_info->min_seq_blocks = sbi->blocks_per_seg * sbi->segs_per_sec;
ef095d19 5088 sm_info->min_hot_blocks = DEF_MIN_HOT_BLOCKS;
a2a12b67 5089 sm_info->min_ssr_sections = reserved_sections(sbi);
351df4b2 5090
184a5cd2
CY		 5091 INIT_LIST_HEAD(&sm_info->sit_entry_set);
5092
2b60311d
CY		 5093 init_rwsem(&sm_info->curseg_lock);
5094
d4fdf8ba 5095 if (!f2fs_readonly(sbi->sb)) {
4d57b86d 5096 err = f2fs_create_flush_cmd_control(sbi);
2163d198 5097 if (err)
a688b9d9 5098 return err;
6b4afdd7
JK		 5099 }
5100
0b54fb84
JK		 5101 err = create_discard_cmd_control(sbi);
5102 if (err)
5103 return err;
5104
351df4b2
JK		 5105 err = build_sit_info(sbi);
5106 if (err)
5107 return err;
5108 err = build_free_segmap(sbi);
5109 if (err)
5110 return err;
5111 err = build_curseg(sbi);
5112 if (err)
5113 return err;
5114
5115 /* reinit free segmap based on SIT */
c39a1b34
JK		 5116 err = build_sit_entries(sbi);
5117 if (err)
5118 return err;
351df4b2
JK		 5119
5120 init_free_segmap(sbi);
5121 err = build_dirty_segmap(sbi);
5122 if (err)
5123 return err;
5124
c854f4d6
CY		 5125 err = sanity_check_curseg(sbi);
5126 if (err)
5127 return err;
5128
351df4b2
JK		 5129 init_min_max_mtime(sbi);
5130 return 0;
5131}
5132
5133static void discard_dirty_segmap(struct f2fs_sb_info *sbi,
5134 enum dirty_type dirty_type)
5135{
5136 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
5137
5138 mutex_lock(&dirty_i->seglist_lock);
39307a8e 5139 kvfree(dirty_i->dirty_segmap[dirty_type]);
351df4b2
JK		 5140 dirty_i->nr_dirty[dirty_type] = 0;
5141 mutex_unlock(&dirty_i->seglist_lock);
5142}
5143
5ec4e49f 5144static void destroy_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 5145{
5146 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
39307a8e 5147 kvfree(dirty_i->victim_secmap);
351df4b2
JK		 5148}
5149
5150static void destroy_dirty_segmap(struct f2fs_sb_info *sbi)
5151{
5152 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
5153 int i;
5154
5155 if (!dirty_i)
5156 return;
5157
5158 /* discard pre-free/dirty segments list */
5159 for (i = 0; i < NR_DIRTY_TYPE; i++)
5160 discard_dirty_segmap(sbi, i);
5161
da52f8ad
JQ		 5162 if (__is_large_section(sbi)) {
5163 mutex_lock(&dirty_i->seglist_lock);
5164 kvfree(dirty_i->dirty_secmap);
5165 mutex_unlock(&dirty_i->seglist_lock);
5166 }
5167
5ec4e49f 5168 destroy_victim_secmap(sbi);
351df4b2 5169 SM_I(sbi)->dirty_info = NULL;
c8eb7024 5170 kfree(dirty_i);
351df4b2
JK		 5171}
5172
5173static void destroy_curseg(struct f2fs_sb_info *sbi)
5174{
5175 struct curseg_info *array = SM_I(sbi)->curseg_array;
5176 int i;
5177
5178 if (!array)
5179 return;
5180 SM_I(sbi)->curseg_array = NULL;
b7ad7512 5181 for (i = 0; i < NR_CURSEG_TYPE; i++) {
c8eb7024
CY		 5182 kfree(array[i].sum_blk);
5183 kfree(array[i].journal);
b7ad7512 5184 }
c8eb7024 5185 kfree(array);
351df4b2
JK		 5186}
5187
5188static void destroy_free_segmap(struct f2fs_sb_info *sbi)
5189{
5190 struct free_segmap_info *free_i = SM_I(sbi)->free_info;
5191 if (!free_i)
5192 return;
5193 SM_I(sbi)->free_info = NULL;
39307a8e
JK		 5194 kvfree(free_i->free_segmap);
5195 kvfree(free_i->free_secmap);
c8eb7024 5196 kfree(free_i);
351df4b2
JK		 5197}
5198
5199static void destroy_sit_info(struct f2fs_sb_info *sbi)
5200{
5201 struct sit_info *sit_i = SIT_I(sbi);
351df4b2
JK		 5202
5203 if (!sit_i)
5204 return;
5205
2fde3dd1
CY		 5206 if (sit_i->sentries)
5207 kvfree(sit_i->bitmap);
c8eb7024 5208 kfree(sit_i->tmp_map);
60a3b782 5209
39307a8e
JK		 5210 kvfree(sit_i->sentries);
5211 kvfree(sit_i->sec_entries);
5212 kvfree(sit_i->dirty_sentries_bitmap);
351df4b2
JK		 5213
5214 SM_I(sbi)->sit_info = NULL;
5222595d 5215 kvfree(sit_i->sit_bitmap);
ae27d62e 5216#ifdef CONFIG_F2FS_CHECK_FS
5222595d 5217 kvfree(sit_i->sit_bitmap_mir);
bbf9f7d9 5218 kvfree(sit_i->invalid_segmap);
ae27d62e 5219#endif
c8eb7024 5220 kfree(sit_i);
351df4b2
JK		 5221}
5222
4d57b86d 5223void f2fs_destroy_segment_manager(struct f2fs_sb_info *sbi)
351df4b2
JK		 5224{
5225 struct f2fs_sm_info *sm_info = SM_I(sbi);
a688b9d9 5226
3b03f724
CY		 5227 if (!sm_info)
5228 return;
4d57b86d 5229 f2fs_destroy_flush_cmd_control(sbi, true);
f099405f 5230 destroy_discard_cmd_control(sbi);
351df4b2
JK		 5231 destroy_dirty_segmap(sbi);
5232 destroy_curseg(sbi);
5233 destroy_free_segmap(sbi);
5234 destroy_sit_info(sbi);
5235 sbi->sm_info = NULL;
c8eb7024 5236 kfree(sm_info);
351df4b2 5237}
7fd9e544 5238
4d57b86d 5239int __init f2fs_create_segment_manager_caches(void)
7fd9e544 5240{
98510003 5241 discard_entry_slab = f2fs_kmem_cache_create("f2fs_discard_entry",
e8512d2e 5242 sizeof(struct discard_entry));
7fd9e544 5243 if (!discard_entry_slab)
184a5cd2
CY		 5244 goto fail;
5245
98510003 5246 discard_cmd_slab = f2fs_kmem_cache_create("f2fs_discard_cmd",
b01a9201
JK		 5247 sizeof(struct discard_cmd));
5248 if (!discard_cmd_slab)
6ab2a308 5249 goto destroy_discard_entry;
275b66b0 5250
98510003 5251 sit_entry_set_slab = f2fs_kmem_cache_create("f2fs_sit_entry_set",
c9ee0085 5252 sizeof(struct sit_entry_set));
184a5cd2 5253 if (!sit_entry_set_slab)
b01a9201 5254 goto destroy_discard_cmd;
88b88a66 5255
98510003 5256 inmem_entry_slab = f2fs_kmem_cache_create("f2fs_inmem_page_entry",
88b88a66
JK		 5257 sizeof(struct inmem_pages));
5258 if (!inmem_entry_slab)
5259 goto destroy_sit_entry_set;
7fd9e544 5260 return 0;
184a5cd2 5261
88b88a66
JK		 5262destroy_sit_entry_set:
5263 kmem_cache_destroy(sit_entry_set_slab);
b01a9201
JK		 5264destroy_discard_cmd:
5265 kmem_cache_destroy(discard_cmd_slab);
6ab2a308 5266destroy_discard_entry:
184a5cd2
CY		 5267 kmem_cache_destroy(discard_entry_slab);
5268fail:
5269 return -ENOMEM;
7fd9e544
JK		 5270}
5271
4d57b86d 5272void f2fs_destroy_segment_manager_caches(void)
7fd9e544 5273{
184a5cd2 5274 kmem_cache_destroy(sit_entry_set_slab);
b01a9201 5275 kmem_cache_destroy(discard_cmd_slab);
7fd9e544 5276 kmem_cache_destroy(discard_entry_slab);
88b88a66 5277 kmem_cache_destroy(inmem_entry_slab);
7fd9e544 5278}
Ubuntu Jammy Linux kernel mirror