]> 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
Merge tag 'microblaze-v5.13' of git://git.monstr.eu/linux-2.6-microblaze
[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"
6ec178da 23#include <trace/events/f2fs.h>
351df4b2 24
9a7f143a
CL		 25#define __reverse_ffz(x) __reverse_ffs(~(x))
26
7fd9e544 27static struct kmem_cache *discard_entry_slab;
b01a9201 28static struct kmem_cache *discard_cmd_slab;
184a5cd2 29static struct kmem_cache *sit_entry_set_slab;
88b88a66 30static struct kmem_cache *inmem_entry_slab;
7fd9e544 31
f96999c3
JK		 32static unsigned long __reverse_ulong(unsigned char *str)
33{
34 unsigned long tmp = 0;
35 int shift = 24, idx = 0;
36
37#if BITS_PER_LONG == 64
38 shift = 56;
39#endif
40 while (shift >= 0) {
41 tmp |= (unsigned long)str[idx++] << shift;
42 shift -= BITS_PER_BYTE;
43 }
44 return tmp;
45}
46
9a7f143a
CL		 47/*
48 * __reverse_ffs is copied from include/asm-generic/bitops/__ffs.h since
49 * MSB and LSB are reversed in a byte by f2fs_set_bit.
50 */
51static inline unsigned long __reverse_ffs(unsigned long word)
52{
53 int num = 0;
54
55#if BITS_PER_LONG == 64
f96999c3 56 if ((word & 0xffffffff00000000UL) == 0)
9a7f143a 57 num += 32;
f96999c3 58 else
9a7f143a 59 word >>= 32;
9a7f143a 60#endif
f96999c3 61 if ((word & 0xffff0000) == 0)
9a7f143a 62 num += 16;
f96999c3 63 else
9a7f143a 64 word >>= 16;
f96999c3
JK		 65
66 if ((word & 0xff00) == 0)
9a7f143a 67 num += 8;
f96999c3 68 else
9a7f143a 69 word >>= 8;
f96999c3 70
9a7f143a
CL		 71 if ((word & 0xf0) == 0)
72 num += 4;
73 else
74 word >>= 4;
f96999c3 75
9a7f143a
CL		 76 if ((word & 0xc) == 0)
77 num += 2;
78 else
79 word >>= 2;
f96999c3 80
9a7f143a
CL		 81 if ((word & 0x2) == 0)
82 num += 1;
83 return num;
84}
85
86/*
e1c42045 87 * __find_rev_next(_zero)_bit is copied from lib/find_next_bit.c because
9a7f143a 88 * f2fs_set_bit makes MSB and LSB reversed in a byte.
692223d1 89 * @size must be integral times of unsigned long.
9a7f143a 90 * Example:
f96999c3
JK		 91 * MSB <--> LSB
92 * f2fs_set_bit(0, bitmap) => 1000 0000
93 * f2fs_set_bit(7, bitmap) => 0000 0001
9a7f143a
CL		 94 */
95static unsigned long __find_rev_next_bit(const unsigned long *addr,
96 unsigned long size, unsigned long offset)
97{
98 const unsigned long *p = addr + BIT_WORD(offset);
692223d1 99 unsigned long result = size;
9a7f143a 100 unsigned long tmp;
9a7f143a
CL		 101
102 if (offset >= size)
103 return size;
104
692223d1 105 size -= (offset & ~(BITS_PER_LONG - 1));
9a7f143a 106 offset %= BITS_PER_LONG;
f96999c3 107
692223d1
FL		 108 while (1) {
109 if (*p == 0)
110 goto pass;
9a7f143a 111
f96999c3 112 tmp = __reverse_ulong((unsigned char *)p);
692223d1
FL		 113
114 tmp &= ~0UL >> offset;
115 if (size < BITS_PER_LONG)
116 tmp &= (~0UL << (BITS_PER_LONG - size));
9a7f143a 117 if (tmp)
692223d1
FL		 118 goto found;
119pass:
120 if (size <= BITS_PER_LONG)
121 break;
9a7f143a 122 size -= BITS_PER_LONG;
692223d1 123 offset = 0;
f96999c3 124 p++;
9a7f143a 125 }
692223d1
FL		 126 return result;
127found:
128 return result - size + __reverse_ffs(tmp);
9a7f143a
CL		 129}
130
131static unsigned long __find_rev_next_zero_bit(const unsigned long *addr,
132 unsigned long size, unsigned long offset)
133{
134 const unsigned long *p = addr + BIT_WORD(offset);
80609448 135 unsigned long result = size;
9a7f143a 136 unsigned long tmp;
9a7f143a
CL		 137
138 if (offset >= size)
139 return size;
140
80609448 141 size -= (offset & ~(BITS_PER_LONG - 1));
9a7f143a 142 offset %= BITS_PER_LONG;
80609448
JK		 143
144 while (1) {
145 if (*p == ~0UL)
146 goto pass;
147
f96999c3 148 tmp = __reverse_ulong((unsigned char *)p);
80609448
JK		 149
150 if (offset)
151 tmp |= ~0UL << (BITS_PER_LONG - offset);
152 if (size < BITS_PER_LONG)
153 tmp |= ~0UL >> size;
f96999c3 154 if (tmp != ~0UL)
80609448
JK		 155 goto found;
156pass:
157 if (size <= BITS_PER_LONG)
158 break;
9a7f143a 159 size -= BITS_PER_LONG;
80609448 160 offset = 0;
f96999c3 161 p++;
9a7f143a 162 }
80609448
JK		 163 return result;
164found:
165 return result - size + __reverse_ffz(tmp);
9a7f143a
CL		 166}
167
4d57b86d 168bool f2fs_need_SSR(struct f2fs_sb_info *sbi)
b3a97a2a
JK		 169{
170 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
171 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
172 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
173
b0332a0f 174 if (f2fs_lfs_mode(sbi))
b3a97a2a 175 return false;
0e5e8111 176 if (sbi->gc_mode == GC_URGENT_HIGH)
b3a97a2a 177 return true;
4354994f
DR		 178 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
179 return true;
b3a97a2a
JK		 180
181 return free_sections(sbi) <= (node_secs + 2 * dent_secs + imeta_secs +
a2a12b67 182 SM_I(sbi)->min_ssr_sections + reserved_sections(sbi));
b3a97a2a
JK		 183}
184
4d57b86d 185void f2fs_register_inmem_page(struct inode *inode, struct page *page)
88b88a66 186{
88b88a66 187 struct inmem_pages *new;
9be32d72 188
e90027d2 189 f2fs_set_page_private(page, ATOMIC_WRITTEN_PAGE);
decd36b6 190
88b88a66
JK		 191 new = f2fs_kmem_cache_alloc(inmem_entry_slab, GFP_NOFS);
192
193 /* add atomic page indices to the list */
194 new->page = page;
195 INIT_LIST_HEAD(&new->list);
decd36b6 196
88b88a66 197 /* increase reference count with clean state */
88b88a66 198 get_page(page);
743b620c
JK		 199 mutex_lock(&F2FS_I(inode)->inmem_lock);
200 list_add_tail(&new->list, &F2FS_I(inode)->inmem_pages);
8dcf2ff7 201 inc_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
743b620c 202 mutex_unlock(&F2FS_I(inode)->inmem_lock);
8ce67cb0
JK		 203
204 trace_f2fs_register_inmem_page(page, INMEM);
88b88a66
JK		 205}
206
28bc106b 207static int __revoke_inmem_pages(struct inode *inode,
48432984
CY		 208 struct list_head *head, bool drop, bool recover,
209 bool trylock)
29b96b54 210{
28bc106b 211 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
29b96b54 212 struct inmem_pages *cur, *tmp;
28bc106b 213 int err = 0;
29b96b54
CY		 214
215 list_for_each_entry_safe(cur, tmp, head, list) {
28bc106b
CY		 216 struct page *page = cur->page;
217
218 if (drop)
219 trace_f2fs_commit_inmem_page(page, INMEM_DROP);
220
48432984
CY		 221 if (trylock) {
222 /*
223 * to avoid deadlock in between page lock and
224 * inmem_lock.
225 */
226 if (!trylock_page(page))
227 continue;
228 } else {
229 lock_page(page);
230 }
29b96b54 231
bae0ee7a 232 f2fs_wait_on_page_writeback(page, DATA, true, true);
e5e5732d 233
28bc106b
CY		 234 if (recover) {
235 struct dnode_of_data dn;
236 struct node_info ni;
237
238 trace_f2fs_commit_inmem_page(page, INMEM_REVOKE);
7f2b4e8e 239retry:
28bc106b 240 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d
CY		 241 err = f2fs_get_dnode_of_data(&dn, page->index,
242 LOOKUP_NODE);
7f2b4e8e
CY		 243 if (err) {
244 if (err == -ENOMEM) {
5df7731f
CY		 245 congestion_wait(BLK_RW_ASYNC,
246 DEFAULT_IO_TIMEOUT);
7f2b4e8e
CY		 247 cond_resched();
248 goto retry;
249 }
28bc106b
CY		 250 err = -EAGAIN;
251 goto next;
252 }
7735730d
CY		 253
254 err = f2fs_get_node_info(sbi, dn.nid, &ni);
255 if (err) {
256 f2fs_put_dnode(&dn);
257 return err;
258 }
259
f1d2564a 260 if (cur->old_addr == NEW_ADDR) {
4d57b86d 261 f2fs_invalidate_blocks(sbi, dn.data_blkaddr);
f1d2564a
DJ		 262 f2fs_update_data_blkaddr(&dn, NEW_ADDR);
263 } else
264 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
28bc106b
CY		 265 cur->old_addr, ni.version, true, true);
266 f2fs_put_dnode(&dn);
267 }
268next:
63c52d78 269 /* we don't need to invalidate this in the sccessful status */
2baf0781 270 if (drop || recover) {
63c52d78 271 ClearPageUptodate(page);
2baf0781
CY		 272 clear_cold_data(page);
273 }
240a5915 274 f2fs_clear_page_private(page);
28bc106b 275 f2fs_put_page(page, 1);
29b96b54
CY		 276
277 list_del(&cur->list);
278 kmem_cache_free(inmem_entry_slab, cur);
279 dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
280 }
28bc106b 281 return err;
29b96b54
CY		 282}
283
4d57b86d 284void f2fs_drop_inmem_pages_all(struct f2fs_sb_info *sbi, bool gc_failure)
57864ae5
JK		 285{
286 struct list_head *head = &sbi->inode_list[ATOMIC_FILE];
287 struct inode *inode;
288 struct f2fs_inode_info *fi;
677017d1
ST		 289 unsigned int count = sbi->atomic_files;
290 unsigned int looped = 0;
57864ae5
JK		 291next:
292 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
293 if (list_empty(head)) {
294 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
295 return;
296 }
297 fi = list_first_entry(head, struct f2fs_inode_info, inmem_ilist);
298 inode = igrab(&fi->vfs_inode);
677017d1
ST		 299 if (inode)
300 list_move_tail(&fi->inmem_ilist, head);
57864ae5
JK		 301 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
302
303 if (inode) {
2ef79ecb 304 if (gc_failure) {
677017d1
ST		 305 if (!fi->i_gc_failures[GC_FAILURE_ATOMIC])
306 goto skip;
2ef79ecb 307 }
2ef79ecb 308 set_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
4d57b86d 309 f2fs_drop_inmem_pages(inode);
677017d1 310skip:
57864ae5
JK		 311 iput(inode);
312 }
5df7731f 313 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
57864ae5 314 cond_resched();
677017d1
ST		 315 if (gc_failure) {
316 if (++looped >= count)
317 return;
318 }
57864ae5
JK		 319 goto next;
320}
321
4d57b86d 322void f2fs_drop_inmem_pages(struct inode *inode)
29b96b54 323{
57864ae5 324 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
29b96b54
CY		 325 struct f2fs_inode_info *fi = F2FS_I(inode);
326
48432984
CY		 327 while (!list_empty(&fi->inmem_pages)) {
328 mutex_lock(&fi->inmem_lock);
329 __revoke_inmem_pages(inode, &fi->inmem_pages,
330 true, false, true);
48432984
CY		 331 mutex_unlock(&fi->inmem_lock);
332 }
5fe45743 333
2ef79ecb 334 fi->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
743b620c
JK		 335
336 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
337 if (!list_empty(&fi->inmem_ilist))
338 list_del_init(&fi->inmem_ilist);
677017d1
ST		 339 if (f2fs_is_atomic_file(inode)) {
340 clear_inode_flag(inode, FI_ATOMIC_FILE);
341 sbi->atomic_files--;
342 }
743b620c 343 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
29b96b54
CY		 344}
345
4d57b86d 346void f2fs_drop_inmem_page(struct inode *inode, struct page *page)
8c242db9
JK		 347{
348 struct f2fs_inode_info *fi = F2FS_I(inode);
349 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
350 struct list_head *head = &fi->inmem_pages;
351 struct inmem_pages *cur = NULL;
352
353 f2fs_bug_on(sbi, !IS_ATOMIC_WRITTEN_PAGE(page));
354
355 mutex_lock(&fi->inmem_lock);
356 list_for_each_entry(cur, head, list) {
357 if (cur->page == page)
358 break;
359 }
360
d0891e84 361 f2fs_bug_on(sbi, list_empty(head) || cur->page != page);
8c242db9
JK		 362 list_del(&cur->list);
363 mutex_unlock(&fi->inmem_lock);
364
365 dec_page_count(sbi, F2FS_INMEM_PAGES);
366 kmem_cache_free(inmem_entry_slab, cur);
367
368 ClearPageUptodate(page);
240a5915 369 f2fs_clear_page_private(page);
8c242db9
JK		 370 f2fs_put_page(page, 0);
371
372 trace_f2fs_commit_inmem_page(page, INMEM_INVALIDATE);
373}
374
4d57b86d 375static int __f2fs_commit_inmem_pages(struct inode *inode)
88b88a66
JK		 376{
377 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
378 struct f2fs_inode_info *fi = F2FS_I(inode);
379 struct inmem_pages *cur, *tmp;
88b88a66 380 struct f2fs_io_info fio = {
05ca3632 381 .sbi = sbi,
39d787be 382 .ino = inode->i_ino,
88b88a66 383 .type = DATA,
04d328de 384 .op = REQ_OP_WRITE,
70fd7614 385 .op_flags = REQ_SYNC | REQ_PRIO,
b0af6d49 386 .io_type = FS_DATA_IO,
88b88a66 387 };
cf52b27a 388 struct list_head revoke_list;
bab475c5 389 bool submit_bio = false;
edb27dee 390 int err = 0;
88b88a66 391
cf52b27a
CY		 392 INIT_LIST_HEAD(&revoke_list);
393
88b88a66 394 list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
28bc106b
CY		 395 struct page *page = cur->page;
396
397 lock_page(page);
398 if (page->mapping == inode->i_mapping) {
399 trace_f2fs_commit_inmem_page(page, INMEM);
400
bae0ee7a 401 f2fs_wait_on_page_writeback(page, DATA, true, true);
8d64d365
CY		 402
403 set_page_dirty(page);
933439c8 404 if (clear_page_dirty_for_io(page)) {
29b96b54 405 inode_dec_dirty_pages(inode);
4d57b86d 406 f2fs_remove_dirty_inode(inode);
933439c8 407 }
640cc189 408retry:
28bc106b 409 fio.page = page;
e959c8f5 410 fio.old_blkaddr = NULL_ADDR;
4d978078 411 fio.encrypted_page = NULL;
cc15620b 412 fio.need_lock = LOCK_DONE;
4d57b86d 413 err = f2fs_do_write_data_page(&fio);
29b96b54 414 if (err) {
640cc189 415 if (err == -ENOMEM) {
5df7731f
CY		 416 congestion_wait(BLK_RW_ASYNC,
417 DEFAULT_IO_TIMEOUT);
640cc189
JK		 418 cond_resched();
419 goto retry;
420 }
28bc106b 421 unlock_page(page);
29b96b54 422 break;
70c640b1 423 }
28bc106b
CY		 424 /* record old blkaddr for revoking */
425 cur->old_addr = fio.old_blkaddr;
bab475c5 426 submit_bio = true;
28bc106b
CY		 427 }
428 unlock_page(page);
cf52b27a 429 list_move_tail(&cur->list, &revoke_list);
88b88a66 430 }
29b96b54 431
bab475c5
CY		 432 if (submit_bio)
433 f2fs_submit_merged_write_cond(sbi, inode, NULL, 0, DATA);
28bc106b 434
cf52b27a
CY		 435 if (err) {
436 /*
437 * try to revoke all committed pages, but still we could fail
438 * due to no memory or other reason, if that happened, EAGAIN
439 * will be returned, which means in such case, transaction is
440 * already not integrity, caller should use journal to do the
441 * recovery or rewrite & commit last transaction. For other
442 * error number, revoking was done by filesystem itself.
443 */
48432984
CY		 444 err = __revoke_inmem_pages(inode, &revoke_list,
445 false, true, false);
cf52b27a
CY		 446
447 /* drop all uncommitted pages */
48432984
CY		 448 __revoke_inmem_pages(inode, &fi->inmem_pages,
449 true, false, false);
cf52b27a 450 } else {
48432984
CY		 451 __revoke_inmem_pages(inode, &revoke_list,
452 false, false, false);
cf52b27a 453 }
28bc106b 454
29b96b54
CY		 455 return err;
456}
457
4d57b86d 458int f2fs_commit_inmem_pages(struct inode *inode)
29b96b54
CY		 459{
460 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
461 struct f2fs_inode_info *fi = F2FS_I(inode);
28bc106b 462 int err;
29b96b54
CY		 463
464 f2fs_balance_fs(sbi, true);
29b96b54 465
6f8d4455
JK		 466 down_write(&fi->i_gc_rwsem[WRITE]);
467
468 f2fs_lock_op(sbi);
5fe45743
CY		 469 set_inode_flag(inode, FI_ATOMIC_COMMIT);
470
29b96b54 471 mutex_lock(&fi->inmem_lock);
4d57b86d 472 err = __f2fs_commit_inmem_pages(inode);
88b88a66
JK		 473 mutex_unlock(&fi->inmem_lock);
474
5fe45743
CY		 475 clear_inode_flag(inode, FI_ATOMIC_COMMIT);
476
29b96b54 477 f2fs_unlock_op(sbi);
6f8d4455
JK		 478 up_write(&fi->i_gc_rwsem[WRITE]);
479
edb27dee 480 return err;
88b88a66
JK		 481}
482
0a8165d7 483/*
351df4b2
JK		 484 * This function balances dirty node and dentry pages.
485 * In addition, it controls garbage collection.
486 */
2c4db1a6 487void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
351df4b2 488{
55523519 489 if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
c45d6002 490 f2fs_show_injection_info(sbi, FAULT_CHECKPOINT);
0f348028 491 f2fs_stop_checkpoint(sbi, false);
55523519 492 }
0f348028 493
e589c2c4 494 /* balance_fs_bg is able to be pending */
a7881893 495 if (need && excess_cached_nats(sbi))
7bcd0cfa 496 f2fs_balance_fs_bg(sbi, false);
e589c2c4 497
00e09c0b 498 if (!f2fs_is_checkpoint_ready(sbi))
4354994f
DR		 499 return;
500
351df4b2 501 /*
029cd28c
JK		 502 * We should do GC or end up with checkpoint, if there are so many dirty
503 * dir/node pages without enough free segments.
351df4b2 504 */
7f3037a5 505 if (has_not_enough_free_secs(sbi, 0, 0)) {
fb24fea7 506 down_write(&sbi->gc_lock);
e066b83c 507 f2fs_gc(sbi, false, false, NULL_SEGNO);
351df4b2
JK		 508 }
509}
510
7bcd0cfa 511void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi, bool from_bg)
4660f9c0 512{
64c74a7a
CY		 513 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
514 return;
515
1dcc336b 516 /* try to shrink extent cache when there is no enough memory */
4d57b86d 517 if (!f2fs_available_free_memory(sbi, EXTENT_CACHE))
554df79e 518 f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER);
1dcc336b 519
1b38dc8e 520 /* check the # of cached NAT entries */
4d57b86d
CY		 521 if (!f2fs_available_free_memory(sbi, NAT_ENTRIES))
522 f2fs_try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK);
1b38dc8e 523
4d57b86d
CY		 524 if (!f2fs_available_free_memory(sbi, FREE_NIDS))
525 f2fs_try_to_free_nids(sbi, MAX_FREE_NIDS);
ad4edb83 526 else
4d57b86d 527 f2fs_build_free_nids(sbi, false, false);
31696580 528
493720a4
CY		 529 if (excess_dirty_nats(sbi) || excess_dirty_nodes(sbi) ||
530 excess_prefree_segs(sbi))
531 goto do_sync;
532
533 /* there is background inflight IO or foreground operation recently */
534 if (is_inflight_io(sbi, REQ_TIME) ||
535 (!f2fs_time_over(sbi, REQ_TIME) && rwsem_is_locked(&sbi->cp_rwsem)))
f455c8a5 536 return;
31696580 537
493720a4
CY		 538 /* exceed periodical checkpoint timeout threshold */
539 if (f2fs_time_over(sbi, CP_TIME))
540 goto do_sync;
541
1b38dc8e 542 /* checkpoint is the only way to shrink partial cached entries */
493720a4
CY		 543 if (f2fs_available_free_memory(sbi, NAT_ENTRIES) ||
544 f2fs_available_free_memory(sbi, INO_ENTRIES))
545 return;
546
547do_sync:
548 if (test_opt(sbi, DATA_FLUSH) && from_bg) {
549 struct blk_plug plug;
550
551 mutex_lock(&sbi->flush_lock);
552
553 blk_start_plug(&plug);
554 f2fs_sync_dirty_inodes(sbi, FILE_INODE);
555 blk_finish_plug(&plug);
556
557 mutex_unlock(&sbi->flush_lock);
36b35a0d 558 }
493720a4
CY		 559 f2fs_sync_fs(sbi->sb, true);
560 stat_inc_bg_cp_count(sbi->stat_info);
4660f9c0
JK		 561}
562
20fda56b
KM		 563static int __submit_flush_wait(struct f2fs_sb_info *sbi,
564 struct block_device *bdev)
3c62be17 565{
25ac8426 566 int ret = blkdev_issue_flush(bdev);
20fda56b
KM		 567
568 trace_f2fs_issue_flush(bdev, test_opt(sbi, NOBARRIER),
569 test_opt(sbi, FLUSH_MERGE), ret);
3c62be17
JK		 570 return ret;
571}
572
39d787be 573static int submit_flush_wait(struct f2fs_sb_info *sbi, nid_t ino)
3c62be17 574{
39d787be 575 int ret = 0;
3c62be17
JK		 576 int i;
577
0916878d 578 if (!f2fs_is_multi_device(sbi))
39d787be 579 return __submit_flush_wait(sbi, sbi->sb->s_bdev);
20fda56b 580
39d787be 581 for (i = 0; i < sbi->s_ndevs; i++) {
4d57b86d 582 if (!f2fs_is_dirty_device(sbi, ino, i, FLUSH_INO))
39d787be 583 continue;
20fda56b
KM		 584 ret = __submit_flush_wait(sbi, FDEV(i).bdev);
585 if (ret)
586 break;
3c62be17
JK		 587 }
588 return ret;
589}
590
2163d198 591static int issue_flush_thread(void *data)
6b4afdd7
JK		 592{
593 struct f2fs_sb_info *sbi = data;
b01a9201 594 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
a688b9d9 595 wait_queue_head_t *q = &fcc->flush_wait_queue;
6b4afdd7
JK		 596repeat:
597 if (kthread_should_stop())
598 return 0;
599
721bd4d5 600 if (!llist_empty(&fcc->issue_list)) {
6b4afdd7
JK		 601 struct flush_cmd *cmd, *next;
602 int ret;
603
721bd4d5
GZ		 604 fcc->dispatch_list = llist_del_all(&fcc->issue_list);
605 fcc->dispatch_list = llist_reverse_order(fcc->dispatch_list);
606
39d787be
CY		 607 cmd = llist_entry(fcc->dispatch_list, struct flush_cmd, llnode);
608
609 ret = submit_flush_wait(sbi, cmd->ino);
8b8dd65f
CY		 610 atomic_inc(&fcc->issued_flush);
611
721bd4d5
GZ		 612 llist_for_each_entry_safe(cmd, next,
613 fcc->dispatch_list, llnode) {
6b4afdd7 614 cmd->ret = ret;
6b4afdd7
JK		 615 complete(&cmd->wait);
616 }
a688b9d9 617 fcc->dispatch_list = NULL;
6b4afdd7
JK		 618 }
619
a688b9d9 620 wait_event_interruptible(*q,
721bd4d5 621 kthread_should_stop() || !llist_empty(&fcc->issue_list));
6b4afdd7
JK		 622 goto repeat;
623}
624
39d787be 625int f2fs_issue_flush(struct f2fs_sb_info *sbi, nid_t ino)
6b4afdd7 626{
b01a9201 627 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
adf8d90b 628 struct flush_cmd cmd;
8b8dd65f 629 int ret;
6b4afdd7 630
0f7b2abd
JK		 631 if (test_opt(sbi, NOBARRIER))
632 return 0;
633
8b8dd65f 634 if (!test_opt(sbi, FLUSH_MERGE)) {
72691af6 635 atomic_inc(&fcc->queued_flush);
39d787be 636 ret = submit_flush_wait(sbi, ino);
72691af6 637 atomic_dec(&fcc->queued_flush);
8b8dd65f
CY		 638 atomic_inc(&fcc->issued_flush);
639 return ret;
640 }
740432f8 641
0916878d
DLM		 642 if (atomic_inc_return(&fcc->queued_flush) == 1 ||
643 f2fs_is_multi_device(sbi)) {
39d787be 644 ret = submit_flush_wait(sbi, ino);
72691af6 645 atomic_dec(&fcc->queued_flush);
8b8dd65f
CY		 646
647 atomic_inc(&fcc->issued_flush);
740432f8
JK		 648 return ret;
649 }
6b4afdd7 650
39d787be 651 cmd.ino = ino;
adf8d90b 652 init_completion(&cmd.wait);
6b4afdd7 653
721bd4d5 654 llist_add(&cmd.llnode, &fcc->issue_list);
6b4afdd7 655
6f890df0
CY		 656 /* update issue_list before we wake up issue_flush thread */
657 smp_mb();
658
659 if (waitqueue_active(&fcc->flush_wait_queue))
a688b9d9 660 wake_up(&fcc->flush_wait_queue);
6b4afdd7 661
5eba8c5d
JK		 662 if (fcc->f2fs_issue_flush) {
663 wait_for_completion(&cmd.wait);
72691af6 664 atomic_dec(&fcc->queued_flush);
5eba8c5d 665 } else {
d3238691
CY		 666 struct llist_node *list;
667
668 list = llist_del_all(&fcc->issue_list);
669 if (!list) {
670 wait_for_completion(&cmd.wait);
72691af6 671 atomic_dec(&fcc->queued_flush);
d3238691
CY		 672 } else {
673 struct flush_cmd *tmp, *next;
674
39d787be 675 ret = submit_flush_wait(sbi, ino);
d3238691
CY		 676
677 llist_for_each_entry_safe(tmp, next, list, llnode) {
678 if (tmp == &cmd) {
679 cmd.ret = ret;
72691af6 680 atomic_dec(&fcc->queued_flush);
d3238691
CY		 681 continue;
682 }
683 tmp->ret = ret;
684 complete(&tmp->wait);
685 }
686 }
5eba8c5d 687 }
adf8d90b
CY		 688
689 return cmd.ret;
6b4afdd7
JK		 690}
691
4d57b86d 692int f2fs_create_flush_cmd_control(struct f2fs_sb_info *sbi)
2163d198
GZ		 693{
694 dev_t dev = sbi->sb->s_bdev->bd_dev;
695 struct flush_cmd_control *fcc;
696 int err = 0;
697
b01a9201
JK		 698 if (SM_I(sbi)->fcc_info) {
699 fcc = SM_I(sbi)->fcc_info;
d871cd04
YS		 700 if (fcc->f2fs_issue_flush)
701 return err;
5eba8c5d
JK		 702 goto init_thread;
703 }
704
acbf054d 705 fcc = f2fs_kzalloc(sbi, sizeof(struct flush_cmd_control), GFP_KERNEL);
2163d198
GZ		 706 if (!fcc)
707 return -ENOMEM;
8b8dd65f 708 atomic_set(&fcc->issued_flush, 0);
72691af6 709 atomic_set(&fcc->queued_flush, 0);
2163d198 710 init_waitqueue_head(&fcc->flush_wait_queue);
721bd4d5 711 init_llist_head(&fcc->issue_list);
b01a9201 712 SM_I(sbi)->fcc_info = fcc;
d4fdf8ba
YH		 713 if (!test_opt(sbi, FLUSH_MERGE))
714 return err;
715
5eba8c5d 716init_thread:
2163d198
GZ		 717 fcc->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi,
718 "f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev));
719 if (IS_ERR(fcc->f2fs_issue_flush)) {
720 err = PTR_ERR(fcc->f2fs_issue_flush);
c8eb7024 721 kfree(fcc);
b01a9201 722 SM_I(sbi)->fcc_info = NULL;
2163d198
GZ		 723 return err;
724 }
2163d198
GZ		 725
726 return err;
727}
728
4d57b86d 729void f2fs_destroy_flush_cmd_control(struct f2fs_sb_info *sbi, bool free)
2163d198 730{
b01a9201 731 struct flush_cmd_control *fcc = SM_I(sbi)->fcc_info;
2163d198 732
5eba8c5d
JK		 733 if (fcc && fcc->f2fs_issue_flush) {
734 struct task_struct *flush_thread = fcc->f2fs_issue_flush;
735
736 fcc->f2fs_issue_flush = NULL;
737 kthread_stop(flush_thread);
738 }
739 if (free) {
c8eb7024 740 kfree(fcc);
b01a9201 741 SM_I(sbi)->fcc_info = NULL;
5eba8c5d 742 }
2163d198
GZ		 743}
744
1228b482
CY		 745int f2fs_flush_device_cache(struct f2fs_sb_info *sbi)
746{
747 int ret = 0, i;
748
0916878d 749 if (!f2fs_is_multi_device(sbi))
1228b482
CY		 750 return 0;
751
6ed29fe1
CY		 752 if (test_opt(sbi, NOBARRIER))
753 return 0;
754
1228b482
CY		 755 for (i = 1; i < sbi->s_ndevs; i++) {
756 if (!f2fs_test_bit(i, (char *)&sbi->dirty_device))
757 continue;
758 ret = __submit_flush_wait(sbi, FDEV(i).bdev);
759 if (ret)
760 break;
761
762 spin_lock(&sbi->dev_lock);
763 f2fs_clear_bit(i, (char *)&sbi->dirty_device);
764 spin_unlock(&sbi->dev_lock);
765 }
766
767 return ret;
768}
769
351df4b2
JK		 770static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
771 enum dirty_type dirty_type)
772{
773 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
774
775 /* need not be added */
776 if (IS_CURSEG(sbi, segno))
777 return;
778
779 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[dirty_type]))
780 dirty_i->nr_dirty[dirty_type]++;
781
782 if (dirty_type == DIRTY) {
783 struct seg_entry *sentry = get_seg_entry(sbi, segno);
4625d6aa 784 enum dirty_type t = sentry->type;
b2f2c390 785
ec325b52
JK		 786 if (unlikely(t >= DIRTY)) {
787 f2fs_bug_on(sbi, 1);
788 return;
789 }
4625d6aa
CL		 790 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[t]))
791 dirty_i->nr_dirty[t]++;
da52f8ad
JQ		 792
793 if (__is_large_section(sbi)) {
794 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
123aaf77 795 block_t valid_blocks =
da52f8ad
JQ		 796 get_valid_blocks(sbi, segno, true);
797
798 f2fs_bug_on(sbi, unlikely(!valid_blocks ||
799 valid_blocks == BLKS_PER_SEC(sbi)));
800
801 if (!IS_CURSEC(sbi, secno))
802 set_bit(secno, dirty_i->dirty_secmap);
803 }
351df4b2
JK		 804 }
805}
806
807static void __remove_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
808 enum dirty_type dirty_type)
809{
810 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
123aaf77 811 block_t valid_blocks;
351df4b2
JK		 812
813 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[dirty_type]))
814 dirty_i->nr_dirty[dirty_type]--;
815
816 if (dirty_type == DIRTY) {
4625d6aa
CL		 817 struct seg_entry *sentry = get_seg_entry(sbi, segno);
818 enum dirty_type t = sentry->type;
819
820 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
821 dirty_i->nr_dirty[t]--;
b2f2c390 822
da52f8ad
JQ		 823 valid_blocks = get_valid_blocks(sbi, segno, true);
824 if (valid_blocks == 0) {
4ddb1a4d 825 clear_bit(GET_SEC_FROM_SEG(sbi, segno),
5ec4e49f 826 dirty_i->victim_secmap);
bbf9f7d9
ST		 827#ifdef CONFIG_F2FS_CHECK_FS
828 clear_bit(segno, SIT_I(sbi)->invalid_segmap);
829#endif
830 }
da52f8ad
JQ		 831 if (__is_large_section(sbi)) {
832 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
833
834 if (!valid_blocks ||
835 valid_blocks == BLKS_PER_SEC(sbi)) {
836 clear_bit(secno, dirty_i->dirty_secmap);
837 return;
838 }
839
840 if (!IS_CURSEC(sbi, secno))
841 set_bit(secno, dirty_i->dirty_secmap);
842 }
351df4b2
JK		 843 }
844}
845
0a8165d7 846/*
351df4b2
JK		 847 * Should not occur error such as -ENOMEM.
848 * Adding dirty entry into seglist is not critical operation.
849 * If a given segment is one of current working segments, it won't be added.
850 */
8d8451af 851static void locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 852{
853 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
4354994f 854 unsigned short valid_blocks, ckpt_valid_blocks;
de881df9 855 unsigned int usable_blocks;
351df4b2
JK		 856
857 if (segno == NULL_SEGNO || IS_CURSEG(sbi, segno))
858 return;
859
de881df9 860 usable_blocks = f2fs_usable_blks_in_seg(sbi, segno);
351df4b2
JK		 861 mutex_lock(&dirty_i->seglist_lock);
862
302bd348 863 valid_blocks = get_valid_blocks(sbi, segno, false);
4354994f 864 ckpt_valid_blocks = get_ckpt_valid_blocks(sbi, segno);
351df4b2 865
4354994f 866 if (valid_blocks == 0 && (!is_sbi_flag_set(sbi, SBI_CP_DISABLED) ||
de881df9 867 ckpt_valid_blocks == usable_blocks)) {
351df4b2
JK		 868 __locate_dirty_segment(sbi, segno, PRE);
869 __remove_dirty_segment(sbi, segno, DIRTY);
de881df9 870 } else if (valid_blocks < usable_blocks) {
351df4b2
JK		 871 __locate_dirty_segment(sbi, segno, DIRTY);
872 } else {
873 /* Recovery routine with SSR needs this */
874 __remove_dirty_segment(sbi, segno, DIRTY);
875 }
876
877 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 878}
879
4354994f
DR		 880/* This moves currently empty dirty blocks to prefree. Must hold seglist_lock */
881void f2fs_dirty_to_prefree(struct f2fs_sb_info *sbi)
882{
883 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
884 unsigned int segno;
885
886 mutex_lock(&dirty_i->seglist_lock);
887 for_each_set_bit(segno, dirty_i->dirty_segmap[DIRTY], MAIN_SEGS(sbi)) {
888 if (get_valid_blocks(sbi, segno, false))
889 continue;
890 if (IS_CURSEG(sbi, segno))
891 continue;
892 __locate_dirty_segment(sbi, segno, PRE);
893 __remove_dirty_segment(sbi, segno, DIRTY);
894 }
895 mutex_unlock(&dirty_i->seglist_lock);
896}
897
4d3aed70 898block_t f2fs_get_unusable_blocks(struct f2fs_sb_info *sbi)
4354994f 899{
ae4ad7ea
DR		 900 int ovp_hole_segs =
901 (overprovision_segments(sbi) - reserved_segments(sbi));
902 block_t ovp_holes = ovp_hole_segs << sbi->log_blocks_per_seg;
4d3aed70 903 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
4354994f 904 block_t holes[2] = {0, 0}; /* DATA and NODE */
4d3aed70 905 block_t unusable;
4354994f
DR		 906 struct seg_entry *se;
907 unsigned int segno;
908
909 mutex_lock(&dirty_i->seglist_lock);
910 for_each_set_bit(segno, dirty_i->dirty_segmap[DIRTY], MAIN_SEGS(sbi)) {
911 se = get_seg_entry(sbi, segno);
912 if (IS_NODESEG(se->type))
de881df9
AR		 913 holes[NODE] += f2fs_usable_blks_in_seg(sbi, segno) -
914 se->valid_blocks;
4354994f 915 else
de881df9
AR		 916 holes[DATA] += f2fs_usable_blks_in_seg(sbi, segno) -
917 se->valid_blocks;
4354994f
DR		 918 }
919 mutex_unlock(&dirty_i->seglist_lock);
920
4d3aed70
DR		 921 unusable = holes[DATA] > holes[NODE] ? holes[DATA] : holes[NODE];
922 if (unusable > ovp_holes)
923 return unusable - ovp_holes;
924 return 0;
925}
926
927int f2fs_disable_cp_again(struct f2fs_sb_info *sbi, block_t unusable)
928{
929 int ovp_hole_segs =
930 (overprovision_segments(sbi) - reserved_segments(sbi));
931 if (unusable > F2FS_OPTION(sbi).unusable_cap)
4354994f 932 return -EAGAIN;
db610a64 933 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED_QUICK) &&
ae4ad7ea 934 dirty_segments(sbi) > ovp_hole_segs)
db610a64 935 return -EAGAIN;
4354994f
DR		 936 return 0;
937}
938
939/* This is only used by SBI_CP_DISABLED */
940static unsigned int get_free_segment(struct f2fs_sb_info *sbi)
941{
942 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
943 unsigned int segno = 0;
944
945 mutex_lock(&dirty_i->seglist_lock);
946 for_each_set_bit(segno, dirty_i->dirty_segmap[DIRTY], MAIN_SEGS(sbi)) {
947 if (get_valid_blocks(sbi, segno, false))
948 continue;
949 if (get_ckpt_valid_blocks(sbi, segno))
950 continue;
951 mutex_unlock(&dirty_i->seglist_lock);
952 return segno;
953 }
954 mutex_unlock(&dirty_i->seglist_lock);
955 return NULL_SEGNO;
956}
957
004b6862 958static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi,
c81abe34
JK		 959 struct block_device *bdev, block_t lstart,
960 block_t start, block_t len)
275b66b0 961{
0b54fb84 962 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
ba48a33e 963 struct list_head *pend_list;
b01a9201 964 struct discard_cmd *dc;
275b66b0 965
ba48a33e
CY		 966 f2fs_bug_on(sbi, !len);
967
968 pend_list = &dcc->pend_list[plist_idx(len)];
969
b01a9201
JK		 970 dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS);
971 INIT_LIST_HEAD(&dc->list);
c81abe34 972 dc->bdev = bdev;
b01a9201 973 dc->lstart = lstart;
c81abe34 974 dc->start = start;
b01a9201 975 dc->len = len;
ec9895ad 976 dc->ref = 0;
15469963 977 dc->state = D_PREP;
72691af6 978 dc->queued = 0;
c81abe34 979 dc->error = 0;
b01a9201 980 init_completion(&dc->wait);
22d375dd 981 list_add_tail(&dc->list, pend_list);
35ec7d57
CY		 982 spin_lock_init(&dc->lock);
983 dc->bio_ref = 0;
5f32366a 984 atomic_inc(&dcc->discard_cmd_cnt);
d84d1cbd 985 dcc->undiscard_blks += len;
004b6862
CY		 986
987 return dc;
988}
989
990static struct discard_cmd *__attach_discard_cmd(struct f2fs_sb_info *sbi,
991 struct block_device *bdev, block_t lstart,
992 block_t start, block_t len,
4dada3fd
CY		 993 struct rb_node *parent, struct rb_node **p,
994 bool leftmost)
004b6862
CY		 995{
996 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
997 struct discard_cmd *dc;
998
999 dc = __create_discard_cmd(sbi, bdev, lstart, start, len);
1000
1001 rb_link_node(&dc->rb_node, parent, p);
4dada3fd 1002 rb_insert_color_cached(&dc->rb_node, &dcc->root, leftmost);
004b6862
CY		 1003
1004 return dc;
15469963
JK		 1005}
1006
004b6862
CY		 1007static void __detach_discard_cmd(struct discard_cmd_control *dcc,
1008 struct discard_cmd *dc)
15469963 1009{
dcc9165d 1010 if (dc->state == D_DONE)
72691af6 1011 atomic_sub(dc->queued, &dcc->queued_discard);
004b6862
CY		 1012
1013 list_del(&dc->list);
4dada3fd 1014 rb_erase_cached(&dc->rb_node, &dcc->root);
d84d1cbd 1015 dcc->undiscard_blks -= dc->len;
004b6862
CY		 1016
1017 kmem_cache_free(discard_cmd_slab, dc);
1018
1019 atomic_dec(&dcc->discard_cmd_cnt);
1020}
1021
1022static void __remove_discard_cmd(struct f2fs_sb_info *sbi,
1023 struct discard_cmd *dc)
1024{
1025 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
35ec7d57 1026 unsigned long flags;
dcc9165d 1027
2ec6f2ef
CY		 1028 trace_f2fs_remove_discard(dc->bdev, dc->start, dc->len);
1029
35ec7d57
CY		 1030 spin_lock_irqsave(&dc->lock, flags);
1031 if (dc->bio_ref) {
1032 spin_unlock_irqrestore(&dc->lock, flags);
1033 return;
1034 }
1035 spin_unlock_irqrestore(&dc->lock, flags);
1036
d9703d90
CY		 1037 f2fs_bug_on(sbi, dc->ref);
1038
c81abe34
JK		 1039 if (dc->error == -EOPNOTSUPP)
1040 dc->error = 0;
15469963 1041
c81abe34 1042 if (dc->error)
22d7ea13 1043 printk_ratelimited(
c45d6002
CY		 1044 "%sF2FS-fs (%s): Issue discard(%u, %u, %u) failed, ret: %d",
1045 KERN_INFO, sbi->sb->s_id,
1046 dc->lstart, dc->start, dc->len, dc->error);
004b6862 1047 __detach_discard_cmd(dcc, dc);
275b66b0
CY		 1048}
1049
c81abe34
JK		 1050static void f2fs_submit_discard_endio(struct bio *bio)
1051{
1052 struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
35ec7d57 1053 unsigned long flags;
c81abe34 1054
35ec7d57 1055 spin_lock_irqsave(&dc->lock, flags);
3fa6a8c5
ST		 1056 if (!dc->error)
1057 dc->error = blk_status_to_errno(bio->bi_status);
35ec7d57
CY		 1058 dc->bio_ref--;
1059 if (!dc->bio_ref && dc->state == D_SUBMIT) {
1060 dc->state = D_DONE;
1061 complete_all(&dc->wait);
1062 }
1063 spin_unlock_irqrestore(&dc->lock, flags);
c81abe34
JK		 1064 bio_put(bio);
1065}
1066
94b1e10e 1067static void __check_sit_bitmap(struct f2fs_sb_info *sbi,
6915ea9d
CY		 1068 block_t start, block_t end)
1069{
1070#ifdef CONFIG_F2FS_CHECK_FS
1071 struct seg_entry *sentry;
1072 unsigned int segno;
1073 block_t blk = start;
1074 unsigned long offset, size, max_blocks = sbi->blocks_per_seg;
1075 unsigned long *map;
1076
1077 while (blk < end) {
1078 segno = GET_SEGNO(sbi, blk);
1079 sentry = get_seg_entry(sbi, segno);
1080 offset = GET_BLKOFF_FROM_SEG0(sbi, blk);
1081
008396e1
YS		 1082 if (end < START_BLOCK(sbi, segno + 1))
1083 size = GET_BLKOFF_FROM_SEG0(sbi, end);
1084 else
1085 size = max_blocks;
6915ea9d
CY		 1086 map = (unsigned long *)(sentry->cur_valid_map);
1087 offset = __find_rev_next_bit(map, size, offset);
1088 f2fs_bug_on(sbi, offset != size);
008396e1 1089 blk = START_BLOCK(sbi, segno + 1);
6915ea9d
CY		 1090 }
1091#endif
1092}
1093
8bb4f253
JK		 1094static void __init_discard_policy(struct f2fs_sb_info *sbi,
1095 struct discard_policy *dpolicy,
1096 int discard_type, unsigned int granularity)
1097{
1098 /* common policy */
1099 dpolicy->type = discard_type;
1100 dpolicy->sync = true;
20ee4382 1101 dpolicy->ordered = false;
8bb4f253
JK		 1102 dpolicy->granularity = granularity;
1103
1104 dpolicy->max_requests = DEF_MAX_DISCARD_REQUEST;
1105 dpolicy->io_aware_gran = MAX_PLIST_NUM;
6ce48b0c 1106 dpolicy->timeout = false;
8bb4f253
JK		 1107
1108 if (discard_type == DPOLICY_BG) {
1109 dpolicy->min_interval = DEF_MIN_DISCARD_ISSUE_TIME;
f9d1dced 1110 dpolicy->mid_interval = DEF_MID_DISCARD_ISSUE_TIME;
8bb4f253
JK		 1111 dpolicy->max_interval = DEF_MAX_DISCARD_ISSUE_TIME;
1112 dpolicy->io_aware = true;
cba60849 1113 dpolicy->sync = false;
20ee4382 1114 dpolicy->ordered = true;
8bb4f253
JK		 1115 if (utilization(sbi) > DEF_DISCARD_URGENT_UTIL) {
1116 dpolicy->granularity = 1;
1117 dpolicy->max_interval = DEF_MIN_DISCARD_ISSUE_TIME;
1118 }
1119 } else if (discard_type == DPOLICY_FORCE) {
1120 dpolicy->min_interval = DEF_MIN_DISCARD_ISSUE_TIME;
f9d1dced 1121 dpolicy->mid_interval = DEF_MID_DISCARD_ISSUE_TIME;
8bb4f253
JK		 1122 dpolicy->max_interval = DEF_MAX_DISCARD_ISSUE_TIME;
1123 dpolicy->io_aware = false;
1124 } else if (discard_type == DPOLICY_FSTRIM) {
1125 dpolicy->io_aware = false;
1126 } else if (discard_type == DPOLICY_UMOUNT) {
1127 dpolicy->io_aware = false;
b8623253
JK		 1128 /* we need to issue all to keep CP_TRIMMED_FLAG */
1129 dpolicy->granularity = 1;
6ce48b0c 1130 dpolicy->timeout = true;
8bb4f253
JK		 1131 }
1132}
1133
35ec7d57
CY		 1134static void __update_discard_tree_range(struct f2fs_sb_info *sbi,
1135 struct block_device *bdev, block_t lstart,
1136 block_t start, block_t len);
c81abe34 1137/* this function is copied from blkdev_issue_discard from block/blk-lib.c */
6b9cb124 1138static int __submit_discard_cmd(struct f2fs_sb_info *sbi,
78997b56 1139 struct discard_policy *dpolicy,
35ec7d57
CY		 1140 struct discard_cmd *dc,
1141 unsigned int *issued)
c81abe34 1142{
35ec7d57
CY		 1143 struct block_device *bdev = dc->bdev;
1144 struct request_queue *q = bdev_get_queue(bdev);
1145 unsigned int max_discard_blocks =
1146 SECTOR_TO_BLOCK(q->limits.max_discard_sectors);
c81abe34 1147 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
78997b56
CY		 1148 struct list_head *wait_list = (dpolicy->type == DPOLICY_FSTRIM) ?
1149 &(dcc->fstrim_list) : &(dcc->wait_list);
78997b56 1150 int flag = dpolicy->sync ? REQ_SYNC : 0;
35ec7d57
CY		 1151 block_t lstart, start, len, total_len;
1152 int err = 0;
c81abe34
JK		 1153
1154 if (dc->state != D_PREP)
6b9cb124 1155 return 0;
c81abe34 1156
d6184774 1157 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK))
6b9cb124 1158 return 0;
d6184774 1159
35ec7d57
CY		 1160 trace_f2fs_issue_discard(bdev, dc->start, dc->len);
1161
1162 lstart = dc->lstart;
1163 start = dc->start;
1164 len = dc->len;
1165 total_len = len;
1166
1167 dc->len = 0;
1168
1169 while (total_len && *issued < dpolicy->max_requests && !err) {
1170 struct bio *bio = NULL;
1171 unsigned long flags;
1172 bool last = true;
1173
1174 if (len > max_discard_blocks) {
1175 len = max_discard_blocks;
1176 last = false;
1177 }
1178
1179 (*issued)++;
1180 if (*issued == dpolicy->max_requests)
1181 last = true;
1182
1183 dc->len += len;
1184
b83dcfe6 1185 if (time_to_inject(sbi, FAULT_DISCARD)) {
c45d6002 1186 f2fs_show_injection_info(sbi, FAULT_DISCARD);
b83dcfe6
CY		 1187 err = -EIO;
1188 goto submit;
1189 }
35ec7d57
CY		 1190 err = __blkdev_issue_discard(bdev,
1191 SECTOR_FROM_BLOCK(start),
1192 SECTOR_FROM_BLOCK(len),
1193 GFP_NOFS, 0, &bio);
b83dcfe6 1194submit:
6b9cb124 1195 if (err) {
35ec7d57 1196 spin_lock_irqsave(&dc->lock, flags);
6b9cb124 1197 if (dc->state == D_PARTIAL)
35ec7d57 1198 dc->state = D_SUBMIT;
35ec7d57
CY		 1199 spin_unlock_irqrestore(&dc->lock, flags);
1200
6b9cb124
CY		 1201 break;
1202 }
35ec7d57 1203
6b9cb124 1204 f2fs_bug_on(sbi, !bio);
35ec7d57 1205
6b9cb124
CY		 1206 /*
1207 * should keep before submission to avoid D_DONE
1208 * right away
1209 */
1210 spin_lock_irqsave(&dc->lock, flags);
1211 if (last)
1212 dc->state = D_SUBMIT;
1213 else
1214 dc->state = D_PARTIAL;
1215 dc->bio_ref++;
1216 spin_unlock_irqrestore(&dc->lock, flags);
35ec7d57 1217
72691af6
JK		 1218 atomic_inc(&dcc->queued_discard);
1219 dc->queued++;
6b9cb124 1220 list_move_tail(&dc->list, wait_list);
b0af6d49 1221
6b9cb124 1222 /* sanity check on discard range */
9249dded 1223 __check_sit_bitmap(sbi, lstart, lstart + len);
35ec7d57 1224
6b9cb124
CY		 1225 bio->bi_private = dc;
1226 bio->bi_end_io = f2fs_submit_discard_endio;
1227 bio->bi_opf |= flag;
1228 submit_bio(bio);
1229
1230 atomic_inc(&dcc->issued_discard);
1231
1232 f2fs_update_iostat(sbi, FS_DISCARD, 1);
35ec7d57
CY		 1233
1234 lstart += len;
1235 start += len;
1236 total_len -= len;
1237 len = total_len;
c81abe34 1238 }
35ec7d57 1239
df423399
ST		 1240 if (!err && len) {
1241 dcc->undiscard_blks -= len;
35ec7d57 1242 __update_discard_tree_range(sbi, bdev, lstart, start, len);
df423399 1243 }
6b9cb124 1244 return err;
c81abe34
JK		 1245}
1246
47d0d7d7 1247static void __insert_discard_tree(struct f2fs_sb_info *sbi,
004b6862
CY		 1248 struct block_device *bdev, block_t lstart,
1249 block_t start, block_t len,
1250 struct rb_node **insert_p,
1251 struct rb_node *insert_parent)
c81abe34 1252{
004b6862 1253 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
dca6951f 1254 struct rb_node **p;
004b6862 1255 struct rb_node *parent = NULL;
4dada3fd 1256 bool leftmost = true;
004b6862
CY		 1257
1258 if (insert_p && insert_parent) {
1259 parent = insert_parent;
1260 p = insert_p;
1261 goto do_insert;
1262 }
c81abe34 1263
4dada3fd
CY		 1264 p = f2fs_lookup_rb_tree_for_insert(sbi, &dcc->root, &parent,
1265 lstart, &leftmost);
004b6862 1266do_insert:
47d0d7d7 1267 __attach_discard_cmd(sbi, bdev, lstart, start, len, parent,
4dada3fd 1268 p, leftmost);
c81abe34
JK		 1269}
1270
ba48a33e
CY		 1271static void __relocate_discard_cmd(struct discard_cmd_control *dcc,
1272 struct discard_cmd *dc)
1273{
1274 list_move_tail(&dc->list, &dcc->pend_list[plist_idx(dc->len)]);
1275}
1276
3d6a650f
YH		 1277static void __punch_discard_cmd(struct f2fs_sb_info *sbi,
1278 struct discard_cmd *dc, block_t blkaddr)
1279{
ba48a33e 1280 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 1281 struct discard_info di = dc->di;
1282 bool modified = false;
3d6a650f 1283
004b6862 1284 if (dc->state == D_DONE || dc->len == 1) {
3d6a650f
YH		 1285 __remove_discard_cmd(sbi, dc);
1286 return;
1287 }
1288
d84d1cbd
CY		 1289 dcc->undiscard_blks -= di.len;
1290
004b6862 1291 if (blkaddr > di.lstart) {
3d6a650f 1292 dc->len = blkaddr - dc->lstart;
d84d1cbd 1293 dcc->undiscard_blks += dc->len;
ba48a33e 1294 __relocate_discard_cmd(dcc, dc);
004b6862
CY		 1295 modified = true;
1296 }
1297
1298 if (blkaddr < di.lstart + di.len - 1) {
1299 if (modified) {
1300 __insert_discard_tree(sbi, dc->bdev, blkaddr + 1,
1301 di.start + blkaddr + 1 - di.lstart,
1302 di.lstart + di.len - 1 - blkaddr,
1303 NULL, NULL);
1304 } else {
1305 dc->lstart++;
1306 dc->len--;
1307 dc->start++;
d84d1cbd 1308 dcc->undiscard_blks += dc->len;
ba48a33e 1309 __relocate_discard_cmd(dcc, dc);
004b6862 1310 }
3d6a650f
YH		 1311 }
1312}
1313
004b6862
CY		 1314static void __update_discard_tree_range(struct f2fs_sb_info *sbi,
1315 struct block_device *bdev, block_t lstart,
1316 block_t start, block_t len)
275b66b0 1317{
0b54fb84 1318 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
004b6862
CY		 1319 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
1320 struct discard_cmd *dc;
1321 struct discard_info di = {0};
1322 struct rb_node **insert_p = NULL, *insert_parent = NULL;
35ec7d57
CY		 1323 struct request_queue *q = bdev_get_queue(bdev);
1324 unsigned int max_discard_blocks =
1325 SECTOR_TO_BLOCK(q->limits.max_discard_sectors);
004b6862 1326 block_t end = lstart + len;
275b66b0 1327
4d57b86d 1328 dc = (struct discard_cmd *)f2fs_lookup_rb_tree_ret(&dcc->root,
004b6862
CY		 1329 NULL, lstart,
1330 (struct rb_entry **)&prev_dc,
1331 (struct rb_entry **)&next_dc,
4dada3fd 1332 &insert_p, &insert_parent, true, NULL);
004b6862
CY		 1333 if (dc)
1334 prev_dc = dc;
1335
1336 if (!prev_dc) {
1337 di.lstart = lstart;
1338 di.len = next_dc ? next_dc->lstart - lstart : len;
1339 di.len = min(di.len, len);
1340 di.start = start;
22d375dd 1341 }
15469963 1342
004b6862
CY		 1343 while (1) {
1344 struct rb_node *node;
1345 bool merged = false;
1346 struct discard_cmd *tdc = NULL;
1347
1348 if (prev_dc) {
1349 di.lstart = prev_dc->lstart + prev_dc->len;
1350 if (di.lstart < lstart)
1351 di.lstart = lstart;
1352 if (di.lstart >= end)
1353 break;
1354
1355 if (!next_dc || next_dc->lstart > end)
1356 di.len = end - di.lstart;
1357 else
1358 di.len = next_dc->lstart - di.lstart;
1359 di.start = start + di.lstart - lstart;
1360 }
1361
1362 if (!di.len)
1363 goto next;
1364
1365 if (prev_dc && prev_dc->state == D_PREP &&
1366 prev_dc->bdev == bdev &&
35ec7d57
CY		 1367 __is_discard_back_mergeable(&di, &prev_dc->di,
1368 max_discard_blocks)) {
004b6862 1369 prev_dc->di.len += di.len;
d84d1cbd 1370 dcc->undiscard_blks += di.len;
ba48a33e 1371 __relocate_discard_cmd(dcc, prev_dc);
004b6862
CY		 1372 di = prev_dc->di;
1373 tdc = prev_dc;
1374 merged = true;
1375 }
1376
1377 if (next_dc && next_dc->state == D_PREP &&
1378 next_dc->bdev == bdev &&
35ec7d57
CY		 1379 __is_discard_front_mergeable(&di, &next_dc->di,
1380 max_discard_blocks)) {
004b6862
CY		 1381 next_dc->di.lstart = di.lstart;
1382 next_dc->di.len += di.len;
1383 next_dc->di.start = di.start;
d84d1cbd 1384 dcc->undiscard_blks += di.len;
ba48a33e 1385 __relocate_discard_cmd(dcc, next_dc);
004b6862
CY		 1386 if (tdc)
1387 __remove_discard_cmd(sbi, tdc);
004b6862 1388 merged = true;
4e6a8d9b 1389 }
004b6862 1390
df0f6b44 1391 if (!merged) {
004b6862
CY		 1392 __insert_discard_tree(sbi, bdev, di.lstart, di.start,
1393 di.len, NULL, NULL);
df0f6b44 1394 }
004b6862
CY		 1395 next:
1396 prev_dc = next_dc;
1397 if (!prev_dc)
1398 break;
1399
1400 node = rb_next(&prev_dc->rb_node);
1401 next_dc = rb_entry_safe(node, struct discard_cmd, rb_node);
1402 }
004b6862
CY		 1403}
1404
1405static int __queue_discard_cmd(struct f2fs_sb_info *sbi,
1406 struct block_device *bdev, block_t blkstart, block_t blklen)
1407{
1408 block_t lblkstart = blkstart;
1409
7f3d7719
DLM		 1410 if (!f2fs_bdev_support_discard(bdev))
1411 return 0;
1412
0243a5f9 1413 trace_f2fs_queue_discard(bdev, blkstart, blklen);
004b6862 1414
0916878d 1415 if (f2fs_is_multi_device(sbi)) {
004b6862
CY		 1416 int devi = f2fs_target_device_index(sbi, blkstart);
1417
1418 blkstart -= FDEV(devi).start_blk;
1419 }
35ec7d57 1420 mutex_lock(&SM_I(sbi)->dcc_info->cmd_lock);
004b6862 1421 __update_discard_tree_range(sbi, bdev, lblkstart, blkstart, blklen);
35ec7d57 1422 mutex_unlock(&SM_I(sbi)->dcc_info->cmd_lock);
004b6862
CY		 1423 return 0;
1424}
1425
20ee4382
CY		 1426static unsigned int __issue_discard_cmd_orderly(struct f2fs_sb_info *sbi,
1427 struct discard_policy *dpolicy)
1428{
1429 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1430 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
1431 struct rb_node **insert_p = NULL, *insert_parent = NULL;
1432 struct discard_cmd *dc;
1433 struct blk_plug plug;
1434 unsigned int pos = dcc->next_pos;
1435 unsigned int issued = 0;
1436 bool io_interrupted = false;
1437
1438 mutex_lock(&dcc->cmd_lock);
1439 dc = (struct discard_cmd *)f2fs_lookup_rb_tree_ret(&dcc->root,
1440 NULL, pos,
1441 (struct rb_entry **)&prev_dc,
1442 (struct rb_entry **)&next_dc,
4dada3fd 1443 &insert_p, &insert_parent, true, NULL);
20ee4382
CY		 1444 if (!dc)
1445 dc = next_dc;
1446
1447 blk_start_plug(&plug);
1448
1449 while (dc) {
1450 struct rb_node *node;
6b9cb124 1451 int err = 0;
20ee4382
CY		 1452
1453 if (dc->state != D_PREP)
1454 goto next;
1455
a7d10cf3 1456 if (dpolicy->io_aware && !is_idle(sbi, DISCARD_TIME)) {
20ee4382
CY		 1457 io_interrupted = true;
1458 break;
1459 }
1460
1461 dcc->next_pos = dc->lstart + dc->len;
6b9cb124 1462 err = __submit_discard_cmd(sbi, dpolicy, dc, &issued);
20ee4382 1463
35ec7d57 1464 if (issued >= dpolicy->max_requests)
20ee4382
CY		 1465 break;
1466next:
1467 node = rb_next(&dc->rb_node);
6b9cb124
CY		 1468 if (err)
1469 __remove_discard_cmd(sbi, dc);
20ee4382
CY		 1470 dc = rb_entry_safe(node, struct discard_cmd, rb_node);
1471 }
1472
1473 blk_finish_plug(&plug);
1474
1475 if (!dc)
1476 dcc->next_pos = 0;
1477
1478 mutex_unlock(&dcc->cmd_lock);
1479
1480 if (!issued && io_interrupted)
1481 issued = -1;
1482
1483 return issued;
1484}
141af6ba
ST		 1485static unsigned int __wait_all_discard_cmd(struct f2fs_sb_info *sbi,
1486 struct discard_policy *dpolicy);
20ee4382 1487
78997b56
CY		 1488static int __issue_discard_cmd(struct f2fs_sb_info *sbi,
1489 struct discard_policy *dpolicy)
bd5b0738
CY		 1490{
1491 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1492 struct list_head *pend_list;
1493 struct discard_cmd *dc, *tmp;
1494 struct blk_plug plug;
141af6ba 1495 int i, issued;
e6c6de18 1496 bool io_interrupted = false;
bd5b0738 1497
6ce48b0c
CY		 1498 if (dpolicy->timeout)
1499 f2fs_update_time(sbi, UMOUNT_DISCARD_TIMEOUT);
03f2c02d 1500
141af6ba
ST		 1501retry:
1502 issued = 0;
78997b56 1503 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
6ce48b0c
CY		 1504 if (dpolicy->timeout &&
1505 f2fs_time_over(sbi, UMOUNT_DISCARD_TIMEOUT))
03f2c02d
JK		 1506 break;
1507
78997b56
CY		 1508 if (i + 1 < dpolicy->granularity)
1509 break;
20ee4382
CY		 1510
1511 if (i < DEFAULT_DISCARD_GRANULARITY && dpolicy->ordered)
1512 return __issue_discard_cmd_orderly(sbi, dpolicy);
1513
bd5b0738 1514 pend_list = &dcc->pend_list[i];
33da62cf
CY		 1515
1516 mutex_lock(&dcc->cmd_lock);
49c60c67
CY		 1517 if (list_empty(pend_list))
1518 goto next;
67fce70b
CY		 1519 if (unlikely(dcc->rbtree_check))
1520 f2fs_bug_on(sbi, !f2fs_check_rb_tree_consistence(sbi,
2e9b2bb2 1521 &dcc->root, false));
33da62cf 1522 blk_start_plug(&plug);
bd5b0738
CY		 1523 list_for_each_entry_safe(dc, tmp, pend_list, list) {
1524 f2fs_bug_on(sbi, dc->state != D_PREP);
1525
6ce48b0c
CY		 1526 if (dpolicy->timeout &&
1527 f2fs_time_over(sbi, UMOUNT_DISCARD_TIMEOUT))
6e0cd4a9
HX		 1528 break;
1529
ecc9aa00 1530 if (dpolicy->io_aware && i < dpolicy->io_aware_gran &&
a7d10cf3 1531 !is_idle(sbi, DISCARD_TIME)) {
e6c6de18 1532 io_interrupted = true;
522d1711 1533 break;
969d1b18 1534 }
e6c6de18 1535
35ec7d57 1536 __submit_discard_cmd(sbi, dpolicy, dc, &issued);
522d1711 1537
35ec7d57 1538 if (issued >= dpolicy->max_requests)
33da62cf 1539 break;
bd5b0738 1540 }
33da62cf 1541 blk_finish_plug(&plug);
49c60c67 1542next:
33da62cf
CY		 1543 mutex_unlock(&dcc->cmd_lock);
1544
522d1711 1545 if (issued >= dpolicy->max_requests || io_interrupted)
33da62cf 1546 break;
bd5b0738 1547 }
969d1b18 1548
141af6ba
ST		 1549 if (dpolicy->type == DPOLICY_UMOUNT && issued) {
1550 __wait_all_discard_cmd(sbi, dpolicy);
1551 goto retry;
1552 }
1553
e6c6de18
CY		 1554 if (!issued && io_interrupted)
1555 issued = -1;
1556
969d1b18
CY		 1557 return issued;
1558}
1559
cf5c759f 1560static bool __drop_discard_cmd(struct f2fs_sb_info *sbi)
969d1b18
CY		 1561{
1562 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1563 struct list_head *pend_list;
1564 struct discard_cmd *dc, *tmp;
1565 int i;
cf5c759f 1566 bool dropped = false;
969d1b18
CY		 1567
1568 mutex_lock(&dcc->cmd_lock);
1569 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
1570 pend_list = &dcc->pend_list[i];
1571 list_for_each_entry_safe(dc, tmp, pend_list, list) {
1572 f2fs_bug_on(sbi, dc->state != D_PREP);
1573 __remove_discard_cmd(sbi, dc);
cf5c759f 1574 dropped = true;
969d1b18
CY		 1575 }
1576 }
1577 mutex_unlock(&dcc->cmd_lock);
cf5c759f
CY		 1578
1579 return dropped;
bd5b0738
CY		 1580}
1581
4d57b86d 1582void f2fs_drop_discard_cmd(struct f2fs_sb_info *sbi)
7950e9ac
CY		 1583{
1584 __drop_discard_cmd(sbi);
1585}
1586
0ea80512 1587static unsigned int __wait_one_discard_bio(struct f2fs_sb_info *sbi,
2a510c00
CY		 1588 struct discard_cmd *dc)
1589{
1590 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
0ea80512 1591 unsigned int len = 0;
2a510c00
CY		 1592
1593 wait_for_completion_io(&dc->wait);
1594 mutex_lock(&dcc->cmd_lock);
1595 f2fs_bug_on(sbi, dc->state != D_DONE);
1596 dc->ref--;
0ea80512
CY		 1597 if (!dc->ref) {
1598 if (!dc->error)
1599 len = dc->len;
2a510c00 1600 __remove_discard_cmd(sbi, dc);
0ea80512 1601 }
2a510c00 1602 mutex_unlock(&dcc->cmd_lock);
0ea80512
CY		 1603
1604 return len;
2a510c00
CY		 1605}
1606
0ea80512 1607static unsigned int __wait_discard_cmd_range(struct f2fs_sb_info *sbi,
78997b56
CY		 1608 struct discard_policy *dpolicy,
1609 block_t start, block_t end)
63a94fa1
CY		 1610{
1611 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
78997b56
CY		 1612 struct list_head *wait_list = (dpolicy->type == DPOLICY_FSTRIM) ?
1613 &(dcc->fstrim_list) : &(dcc->wait_list);
63a94fa1 1614 struct discard_cmd *dc, *tmp;
6afae633 1615 bool need_wait;
0ea80512 1616 unsigned int trimmed = 0;
6afae633
CY		 1617
1618next:
1619 need_wait = false;
63a94fa1
CY		 1620
1621 mutex_lock(&dcc->cmd_lock);
1622 list_for_each_entry_safe(dc, tmp, wait_list, list) {
8412663d
CY		 1623 if (dc->lstart + dc->len <= start || end <= dc->lstart)
1624 continue;
78997b56 1625 if (dc->len < dpolicy->granularity)
8412663d 1626 continue;
78997b56 1627 if (dc->state == D_DONE && !dc->ref) {
63a94fa1 1628 wait_for_completion_io(&dc->wait);
0ea80512
CY		 1629 if (!dc->error)
1630 trimmed += dc->len;
63a94fa1 1631 __remove_discard_cmd(sbi, dc);
6afae633
CY		 1632 } else {
1633 dc->ref++;
1634 need_wait = true;
1635 break;
63a94fa1
CY		 1636 }
1637 }
1638 mutex_unlock(&dcc->cmd_lock);
6afae633
CY		 1639
1640 if (need_wait) {
0ea80512 1641 trimmed += __wait_one_discard_bio(sbi, dc);
6afae633
CY		 1642 goto next;
1643 }
0ea80512
CY		 1644
1645 return trimmed;
63a94fa1
CY		 1646}
1647
01f9cf6d 1648static unsigned int __wait_all_discard_cmd(struct f2fs_sb_info *sbi,
78997b56 1649 struct discard_policy *dpolicy)
8412663d 1650{
9a997188 1651 struct discard_policy dp;
01f9cf6d 1652 unsigned int discard_blks;
9a997188 1653
01f9cf6d
CY		 1654 if (dpolicy)
1655 return __wait_discard_cmd_range(sbi, dpolicy, 0, UINT_MAX);
9a997188
JK		 1656
1657 /* wait all */
8bb4f253 1658 __init_discard_policy(sbi, &dp, DPOLICY_FSTRIM, 1);
01f9cf6d 1659 discard_blks = __wait_discard_cmd_range(sbi, &dp, 0, UINT_MAX);
8bb4f253 1660 __init_discard_policy(sbi, &dp, DPOLICY_UMOUNT, 1);
01f9cf6d
CY		 1661 discard_blks += __wait_discard_cmd_range(sbi, &dp, 0, UINT_MAX);
1662
1663 return discard_blks;
8412663d
CY		 1664}
1665
004b6862 1666/* This should be covered by global mutex, &sit_i->sentry_lock */
94b1e10e 1667static void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
004b6862
CY		 1668{
1669 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1670 struct discard_cmd *dc;
ec9895ad 1671 bool need_wait = false;
004b6862
CY		 1672
1673 mutex_lock(&dcc->cmd_lock);
4d57b86d
CY		 1674 dc = (struct discard_cmd *)f2fs_lookup_rb_tree(&dcc->root,
1675 NULL, blkaddr);
004b6862 1676 if (dc) {
ec9895ad
CY		 1677 if (dc->state == D_PREP) {
1678 __punch_discard_cmd(sbi, dc, blkaddr);
1679 } else {
1680 dc->ref++;
1681 need_wait = true;
1682 }
275b66b0 1683 }
d431413f 1684 mutex_unlock(&dcc->cmd_lock);
ec9895ad 1685
2a510c00
CY		 1686 if (need_wait)
1687 __wait_one_discard_bio(sbi, dc);
d431413f
CY		 1688}
1689
4d57b86d 1690void f2fs_stop_discard_thread(struct f2fs_sb_info *sbi)
cce13252
CY		 1691{
1692 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1693
1694 if (dcc && dcc->f2fs_issue_discard) {
1695 struct task_struct *discard_thread = dcc->f2fs_issue_discard;
1696
1697 dcc->f2fs_issue_discard = NULL;
1698 kthread_stop(discard_thread);
ec9895ad 1699 }
d431413f
CY		 1700}
1701
8412663d 1702/* This comes from f2fs_put_super */
03f2c02d 1703bool f2fs_issue_discard_timeout(struct f2fs_sb_info *sbi)
969d1b18
CY		 1704{
1705 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
78997b56 1706 struct discard_policy dpolicy;
cf5c759f 1707 bool dropped;
969d1b18 1708
8bb4f253
JK		 1709 __init_discard_policy(sbi, &dpolicy, DPOLICY_UMOUNT,
1710 dcc->discard_granularity);
78997b56 1711 __issue_discard_cmd(sbi, &dpolicy);
cf5c759f 1712 dropped = __drop_discard_cmd(sbi);
cf5c759f 1713
9a997188
JK		 1714 /* just to make sure there is no pending discard commands */
1715 __wait_all_discard_cmd(sbi, NULL);
2482c432
CY		 1716
1717 f2fs_bug_on(sbi, atomic_read(&dcc->discard_cmd_cnt));
cf5c759f 1718 return dropped;
969d1b18
CY		 1719}
1720
15469963
JK		 1721static int issue_discard_thread(void *data)
1722{
1723 struct f2fs_sb_info *sbi = data;
1724 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1725 wait_queue_head_t *q = &dcc->discard_wait_queue;
78997b56 1726 struct discard_policy dpolicy;
969d1b18
CY		 1727 unsigned int wait_ms = DEF_MIN_DISCARD_ISSUE_TIME;
1728 int issued;
15469963 1729
1d7be270 1730 set_freezable();
15469963 1731
1d7be270 1732 do {
8bb4f253 1733 __init_discard_policy(sbi, &dpolicy, DPOLICY_BG,
78997b56
CY		 1734 dcc->discard_granularity);
1735
969d1b18
CY		 1736 wait_event_interruptible_timeout(*q,
1737 kthread_should_stop() || freezing(current) ||
1738 dcc->discard_wake,
1739 msecs_to_jiffies(wait_ms));
35a9a766
SY		 1740
1741 if (dcc->discard_wake)
1742 dcc->discard_wake = 0;
1743
76c7bfb3
JK		 1744 /* clean up pending candidates before going to sleep */
1745 if (atomic_read(&dcc->queued_discard))
1746 __wait_all_discard_cmd(sbi, NULL);
1747
1d7be270
JK		 1748 if (try_to_freeze())
1749 continue;
3b60d802
CY		 1750 if (f2fs_readonly(sbi->sb))
1751 continue;
1d7be270
JK		 1752 if (kthread_should_stop())
1753 return 0;
d6184774
YH		 1754 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
1755 wait_ms = dpolicy.max_interval;
1756 continue;
1757 }
15469963 1758
0e5e8111 1759 if (sbi->gc_mode == GC_URGENT_HIGH)
8bb4f253 1760 __init_discard_policy(sbi, &dpolicy, DPOLICY_FORCE, 1);
969d1b18 1761
dc6febb6
CY		 1762 sb_start_intwrite(sbi->sb);
1763
78997b56 1764 issued = __issue_discard_cmd(sbi, &dpolicy);
f9d1dced 1765 if (issued > 0) {
78997b56
CY		 1766 __wait_all_discard_cmd(sbi, &dpolicy);
1767 wait_ms = dpolicy.min_interval;
f9d1dced 1768 } else if (issued == -1){
a7d10cf3
ST		 1769 wait_ms = f2fs_time_to_wait(sbi, DISCARD_TIME);
1770 if (!wait_ms)
abde73c7 1771 wait_ms = dpolicy.mid_interval;
969d1b18 1772 } else {
78997b56 1773 wait_ms = dpolicy.max_interval;
969d1b18 1774 }
1d7be270 1775
dc6febb6 1776 sb_end_intwrite(sbi->sb);
1d7be270 1777
1d7be270
JK		 1778 } while (!kthread_should_stop());
1779 return 0;
15469963
JK		 1780}
1781
f46e8809 1782#ifdef CONFIG_BLK_DEV_ZONED
3c62be17
JK		 1783static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
1784 struct block_device *bdev, block_t blkstart, block_t blklen)
f46e8809 1785{
92592285 1786 sector_t sector, nr_sects;
10a875f8 1787 block_t lblkstart = blkstart;
3c62be17
JK		 1788 int devi = 0;
1789
0916878d 1790 if (f2fs_is_multi_device(sbi)) {
3c62be17 1791 devi = f2fs_target_device_index(sbi, blkstart);
95175daf
DLM		 1792 if (blkstart < FDEV(devi).start_blk ||
1793 blkstart > FDEV(devi).end_blk) {
dcbb4c10 1794 f2fs_err(sbi, "Invalid block %x", blkstart);
95175daf
DLM		 1795 return -EIO;
1796 }
3c62be17
JK		 1797 blkstart -= FDEV(devi).start_blk;
1798 }
f46e8809 1799
95175daf
DLM		 1800 /* For sequential zones, reset the zone write pointer */
1801 if (f2fs_blkz_is_seq(sbi, devi, blkstart)) {
92592285
JK		 1802 sector = SECTOR_FROM_BLOCK(blkstart);
1803 nr_sects = SECTOR_FROM_BLOCK(blklen);
1804
1805 if (sector & (bdev_zone_sectors(bdev) - 1) ||
1806 nr_sects != bdev_zone_sectors(bdev)) {
dcbb4c10
JP		 1807 f2fs_err(sbi, "(%d) %s: Unaligned zone reset attempted (block %x + %x)",
1808 devi, sbi->s_ndevs ? FDEV(devi).path : "",
1809 blkstart, blklen);
92592285
JK		 1810 return -EIO;
1811 }
d50aaeec 1812 trace_f2fs_issue_reset_zone(bdev, blkstart);
6c1b1da5
AJ		 1813 return blkdev_zone_mgmt(bdev, REQ_OP_ZONE_RESET,
1814 sector, nr_sects, GFP_NOFS);
f46e8809 1815 }
95175daf
DLM		 1816
1817 /* For conventional zones, use regular discard if supported */
95175daf 1818 return __queue_discard_cmd(sbi, bdev, lblkstart, blklen);
f46e8809
DLM		 1819}
1820#endif
1821
3c62be17
JK		 1822static int __issue_discard_async(struct f2fs_sb_info *sbi,
1823 struct block_device *bdev, block_t blkstart, block_t blklen)
1824{
1825#ifdef CONFIG_BLK_DEV_ZONED
7f3d7719 1826 if (f2fs_sb_has_blkzoned(sbi) && bdev_is_zoned(bdev))
3c62be17
JK		 1827 return __f2fs_issue_discard_zone(sbi, bdev, blkstart, blklen);
1828#endif
c81abe34 1829 return __queue_discard_cmd(sbi, bdev, blkstart, blklen);
3c62be17
JK		 1830}
1831
1e87a78d 1832static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
37208879
JK		 1833 block_t blkstart, block_t blklen)
1834{
3c62be17
JK		 1835 sector_t start = blkstart, len = 0;
1836 struct block_device *bdev;
a66cdd98
JK		 1837 struct seg_entry *se;
1838 unsigned int offset;
1839 block_t i;
3c62be17
JK		 1840 int err = 0;
1841
1842 bdev = f2fs_target_device(sbi, blkstart, NULL);
1843
1844 for (i = blkstart; i < blkstart + blklen; i++, len++) {
1845 if (i != start) {
1846 struct block_device *bdev2 =
1847 f2fs_target_device(sbi, i, NULL);
1848
1849 if (bdev2 != bdev) {
1850 err = __issue_discard_async(sbi, bdev,
1851 start, len);
1852 if (err)
1853 return err;
1854 bdev = bdev2;
1855 start = i;
1856 len = 0;
1857 }
1858 }
a66cdd98 1859
a66cdd98
JK		 1860 se = get_seg_entry(sbi, GET_SEGNO(sbi, i));
1861 offset = GET_BLKOFF_FROM_SEG0(sbi, i);
1862
1863 if (!f2fs_test_and_set_bit(offset, se->discard_map))
1864 sbi->discard_blks--;
1865 }
f46e8809 1866
3c62be17
JK		 1867 if (len)
1868 err = __issue_discard_async(sbi, bdev, start, len);
1869 return err;
1e87a78d
JK		 1870}
1871
25290fa5
JK		 1872static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc,
1873 bool check_only)
adf4983b 1874{
b2955550
JK		 1875 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
1876 int max_blocks = sbi->blocks_per_seg;
4b2fecc8 1877 struct seg_entry *se = get_seg_entry(sbi, cpc->trim_start);
b2955550
JK		 1878 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
1879 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
a66cdd98 1880 unsigned long *discard_map = (unsigned long *)se->discard_map;
60a3b782 1881 unsigned long *dmap = SIT_I(sbi)->tmp_map;
b2955550 1882 unsigned int start = 0, end = -1;
c473f1a9 1883 bool force = (cpc->reason & CP_DISCARD);
a7eeb823 1884 struct discard_entry *de = NULL;
46f84c2c 1885 struct list_head *head = &SM_I(sbi)->dcc_info->entry_list;
b2955550
JK		 1886 int i;
1887
7d20c8ab 1888 if (se->valid_blocks == max_blocks || !f2fs_hw_support_discard(sbi))
25290fa5 1889 return false;
b2955550 1890
a66cdd98 1891 if (!force) {
7d20c8ab 1892 if (!f2fs_realtime_discard_enable(sbi) || !se->valid_blocks ||
0b54fb84
JK		 1893 SM_I(sbi)->dcc_info->nr_discards >=
1894 SM_I(sbi)->dcc_info->max_discards)
25290fa5 1895 return false;
4b2fecc8
JK		 1896 }
1897
b2955550
JK		 1898 /* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
1899 for (i = 0; i < entries; i++)
a66cdd98 1900 dmap[i] = force ? ~ckpt_map[i] & ~discard_map[i] :
d7bc2484 1901 (cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
b2955550 1902
0b54fb84
JK		 1903 while (force || SM_I(sbi)->dcc_info->nr_discards <=
1904 SM_I(sbi)->dcc_info->max_discards) {
b2955550
JK		 1905 start = __find_rev_next_bit(dmap, max_blocks, end + 1);
1906 if (start >= max_blocks)
1907 break;
1908
1909 end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1);
c7b41e16
YH		 1910 if (force && start && end != max_blocks
1911 && (end - start) < cpc->trim_minlen)
1912 continue;
1913
25290fa5
JK		 1914 if (check_only)
1915 return true;
1916
a7eeb823
CY		 1917 if (!de) {
1918 de = f2fs_kmem_cache_alloc(discard_entry_slab,
1919 GFP_F2FS_ZERO);
1920 de->start_blkaddr = START_BLOCK(sbi, cpc->trim_start);
1921 list_add_tail(&de->list, head);
1922 }
1923
1924 for (i = start; i < end; i++)
1925 __set_bit_le(i, (void *)de->discard_map);
1926
1927 SM_I(sbi)->dcc_info->nr_discards += end - start;
b2955550 1928 }
25290fa5 1929 return false;
b2955550
JK		 1930}
1931
af8ff65b
CY		 1932static void release_discard_addr(struct discard_entry *entry)
1933{
1934 list_del(&entry->list);
1935 kmem_cache_free(discard_entry_slab, entry);
1936}
1937
4d57b86d 1938void f2fs_release_discard_addrs(struct f2fs_sb_info *sbi)
4b2fecc8 1939{
46f84c2c 1940 struct list_head *head = &(SM_I(sbi)->dcc_info->entry_list);
4b2fecc8
JK		 1941 struct discard_entry *entry, *this;
1942
1943 /* drop caches */
af8ff65b
CY		 1944 list_for_each_entry_safe(entry, this, head, list)
1945 release_discard_addr(entry);
4b2fecc8
JK		 1946}
1947
0a8165d7 1948/*
4d57b86d 1949 * Should call f2fs_clear_prefree_segments after checkpoint is done.
351df4b2
JK		 1950 */
1951static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
1952{
1953 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
b65ee148 1954 unsigned int segno;
351df4b2
JK		 1955
1956 mutex_lock(&dirty_i->seglist_lock);
7cd8558b 1957 for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], MAIN_SEGS(sbi))
d0b9e42a 1958 __set_test_and_free(sbi, segno, false);
351df4b2
JK		 1959 mutex_unlock(&dirty_i->seglist_lock);
1960}
1961
4d57b86d
CY		 1962void f2fs_clear_prefree_segments(struct f2fs_sb_info *sbi,
1963 struct cp_control *cpc)
351df4b2 1964{
969d1b18
CY		 1965 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1966 struct list_head *head = &dcc->entry_list;
2d7b822a 1967 struct discard_entry *entry, *this;
351df4b2 1968 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
29e59c14 1969 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
29e59c14 1970 unsigned int start = 0, end = -1;
36abef4e 1971 unsigned int secno, start_segno;
c473f1a9 1972 bool force = (cpc->reason & CP_DISCARD);
b0332a0f 1973 bool need_align = f2fs_lfs_mode(sbi) && __is_large_section(sbi);
351df4b2
JK		 1974
1975 mutex_lock(&dirty_i->seglist_lock);
29e59c14 1976
351df4b2 1977 while (1) {
29e59c14 1978 int i;
ad6672bb
YS		 1979
1980 if (need_align && end != -1)
1981 end--;
7cd8558b
JK		 1982 start = find_next_bit(prefree_map, MAIN_SEGS(sbi), end + 1);
1983 if (start >= MAIN_SEGS(sbi))
351df4b2 1984 break;
7cd8558b
JK		 1985 end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
1986 start + 1);
29e59c14 1987
ad6672bb
YS		 1988 if (need_align) {
1989 start = rounddown(start, sbi->segs_per_sec);
1990 end = roundup(end, sbi->segs_per_sec);
1991 }
29e59c14 1992
ad6672bb
YS		 1993 for (i = start; i < end; i++) {
1994 if (test_and_clear_bit(i, prefree_map))
1995 dirty_i->nr_dirty[PRE]--;
1996 }
29e59c14 1997
7d20c8ab 1998 if (!f2fs_realtime_discard_enable(sbi))
29e59c14 1999 continue;
351df4b2 2000
650d3c4e
YH		 2001 if (force && start >= cpc->trim_start &&
2002 (end - 1) <= cpc->trim_end)
2003 continue;
2004
b0332a0f 2005 if (!f2fs_lfs_mode(sbi) || !__is_large_section(sbi)) {
36abef4e 2006 f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
37208879 2007 (end - start) << sbi->log_blocks_per_seg);
36abef4e
JK		 2008 continue;
2009 }
2010next:
4ddb1a4d
JK		 2011 secno = GET_SEC_FROM_SEG(sbi, start);
2012 start_segno = GET_SEG_FROM_SEC(sbi, secno);
36abef4e 2013 if (!IS_CURSEC(sbi, secno) &&
302bd348 2014 !get_valid_blocks(sbi, start, true))
36abef4e
JK		 2015 f2fs_issue_discard(sbi, START_BLOCK(sbi, start_segno),
2016 sbi->segs_per_sec << sbi->log_blocks_per_seg);
2017
2018 start = start_segno + sbi->segs_per_sec;
2019 if (start < end)
2020 goto next;
8b107f5b
JK		 2021 else
2022 end = start - 1;
351df4b2
JK		 2023 }
2024 mutex_unlock(&dirty_i->seglist_lock);
b2955550
JK		 2025
2026 /* send small discards */
2d7b822a 2027 list_for_each_entry_safe(entry, this, head, list) {
a7eeb823
CY		 2028 unsigned int cur_pos = 0, next_pos, len, total_len = 0;
2029 bool is_valid = test_bit_le(0, entry->discard_map);
2030
2031find_next:
2032 if (is_valid) {
2033 next_pos = find_next_zero_bit_le(entry->discard_map,
2034 sbi->blocks_per_seg, cur_pos);
2035 len = next_pos - cur_pos;
2036
7beb01f7 2037 if (f2fs_sb_has_blkzoned(sbi) ||
acfd2810 2038 (force && len < cpc->trim_minlen))
a7eeb823
CY		 2039 goto skip;
2040
2041 f2fs_issue_discard(sbi, entry->start_blkaddr + cur_pos,
2042 len);
a7eeb823
CY		 2043 total_len += len;
2044 } else {
2045 next_pos = find_next_bit_le(entry->discard_map,
2046 sbi->blocks_per_seg, cur_pos);
2047 }
836b5a63 2048skip:
a7eeb823
CY		 2049 cur_pos = next_pos;
2050 is_valid = !is_valid;
2051
2052 if (cur_pos < sbi->blocks_per_seg)
2053 goto find_next;
2054
af8ff65b 2055 release_discard_addr(entry);
969d1b18 2056 dcc->nr_discards -= total_len;
b2955550 2057 }
34e159da 2058
01983c71 2059 wake_up_discard_thread(sbi, false);
351df4b2
JK		 2060}
2061
8ed59745 2062static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84 2063{
15469963 2064 dev_t dev = sbi->sb->s_bdev->bd_dev;
0b54fb84 2065 struct discard_cmd_control *dcc;
ba48a33e 2066 int err = 0, i;
0b54fb84
JK		 2067
2068 if (SM_I(sbi)->dcc_info) {
2069 dcc = SM_I(sbi)->dcc_info;
2070 goto init_thread;
2071 }
2072
acbf054d 2073 dcc = f2fs_kzalloc(sbi, sizeof(struct discard_cmd_control), GFP_KERNEL);
0b54fb84
JK		 2074 if (!dcc)
2075 return -ENOMEM;
2076
969d1b18 2077 dcc->discard_granularity = DEFAULT_DISCARD_GRANULARITY;
46f84c2c 2078 INIT_LIST_HEAD(&dcc->entry_list);
78997b56 2079 for (i = 0; i < MAX_PLIST_NUM; i++)
ba48a33e 2080 INIT_LIST_HEAD(&dcc->pend_list[i]);
46f84c2c 2081 INIT_LIST_HEAD(&dcc->wait_list);
8412663d 2082 INIT_LIST_HEAD(&dcc->fstrim_list);
15469963 2083 mutex_init(&dcc->cmd_lock);
8b8dd65f 2084 atomic_set(&dcc->issued_discard, 0);
72691af6 2085 atomic_set(&dcc->queued_discard, 0);
5f32366a 2086 atomic_set(&dcc->discard_cmd_cnt, 0);
0b54fb84 2087 dcc->nr_discards = 0;
d618ebaf 2088 dcc->max_discards = MAIN_SEGS(sbi) << sbi->log_blocks_per_seg;
d84d1cbd 2089 dcc->undiscard_blks = 0;
20ee4382 2090 dcc->next_pos = 0;
4dada3fd 2091 dcc->root = RB_ROOT_CACHED;
67fce70b 2092 dcc->rbtree_check = false;
0b54fb84 2093
15469963 2094 init_waitqueue_head(&dcc->discard_wait_queue);
0b54fb84
JK		 2095 SM_I(sbi)->dcc_info = dcc;
2096init_thread:
15469963
JK		 2097 dcc->f2fs_issue_discard = kthread_run(issue_discard_thread, sbi,
2098 "f2fs_discard-%u:%u", MAJOR(dev), MINOR(dev));
2099 if (IS_ERR(dcc->f2fs_issue_discard)) {
2100 err = PTR_ERR(dcc->f2fs_issue_discard);
c8eb7024 2101 kfree(dcc);
15469963
JK		 2102 SM_I(sbi)->dcc_info = NULL;
2103 return err;
2104 }
2105
0b54fb84
JK		 2106 return err;
2107}
2108
f099405f 2109static void destroy_discard_cmd_control(struct f2fs_sb_info *sbi)
0b54fb84
JK		 2110{
2111 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
2112
f099405f
CY		 2113 if (!dcc)
2114 return;
2115
4d57b86d 2116 f2fs_stop_discard_thread(sbi);
f099405f 2117
04f9287a
CY		 2118 /*
2119 * Recovery can cache discard commands, so in error path of
2120 * fill_super(), it needs to give a chance to handle them.
2121 */
2122 if (unlikely(atomic_read(&dcc->discard_cmd_cnt)))
2123 f2fs_issue_discard_timeout(sbi);
2124
c8eb7024 2125 kfree(dcc);
f099405f 2126 SM_I(sbi)->dcc_info = NULL;
0b54fb84
JK		 2127}
2128
184a5cd2 2129static bool __mark_sit_entry_dirty(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2
JK		 2130{
2131 struct sit_info *sit_i = SIT_I(sbi);
184a5cd2
CY		 2132
2133 if (!__test_and_set_bit(segno, sit_i->dirty_sentries_bitmap)) {
351df4b2 2134 sit_i->dirty_sentries++;
184a5cd2
CY		 2135 return false;
2136 }
2137
2138 return true;
351df4b2
JK		 2139}
2140
2141static void __set_sit_entry_type(struct f2fs_sb_info *sbi, int type,
2142 unsigned int segno, int modified)
2143{
2144 struct seg_entry *se = get_seg_entry(sbi, segno);
2145 se->type = type;
2146 if (modified)
2147 __mark_sit_entry_dirty(sbi, segno);
2148}
2149
c5d02785
CY		 2150static inline unsigned long long get_segment_mtime(struct f2fs_sb_info *sbi,
2151 block_t blkaddr)
6f3a01ae
CY		 2152{
2153 unsigned int segno = GET_SEGNO(sbi, blkaddr);
c5d02785
CY		 2154
2155 if (segno == NULL_SEGNO)
2156 return 0;
2157 return get_seg_entry(sbi, segno)->mtime;
2158}
2159
2160static void update_segment_mtime(struct f2fs_sb_info *sbi, block_t blkaddr,
2161 unsigned long long old_mtime)
2162{
2163 struct seg_entry *se;
2164 unsigned int segno = GET_SEGNO(sbi, blkaddr);
2165 unsigned long long ctime = get_mtime(sbi, false);
2166 unsigned long long mtime = old_mtime ? old_mtime : ctime;
2167
2168 if (segno == NULL_SEGNO)
2169 return;
2170
2171 se = get_seg_entry(sbi, segno);
6f3a01ae
CY		 2172
2173 if (!se->mtime)
2174 se->mtime = mtime;
2175 else
2176 se->mtime = div_u64(se->mtime * se->valid_blocks + mtime,
2177 se->valid_blocks + 1);
2178
c5d02785
CY		 2179 if (ctime > SIT_I(sbi)->max_mtime)
2180 SIT_I(sbi)->max_mtime = ctime;
6f3a01ae
CY		 2181}
2182
351df4b2
JK		 2183static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
2184{
2185 struct seg_entry *se;
2186 unsigned int segno, offset;
2187 long int new_vblocks;
6415fedc
YS		 2188 bool exist;
2189#ifdef CONFIG_F2FS_CHECK_FS
2190 bool mir_exist;
2191#endif
351df4b2
JK		 2192
2193 segno = GET_SEGNO(sbi, blkaddr);
2194
2195 se = get_seg_entry(sbi, segno);
2196 new_vblocks = se->valid_blocks + del;
491c0854 2197 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
351df4b2 2198
9feffe14 2199 f2fs_bug_on(sbi, (new_vblocks < 0 ||
de881df9 2200 (new_vblocks > f2fs_usable_blks_in_seg(sbi, segno))));
351df4b2
JK		 2201
2202 se->valid_blocks = new_vblocks;
351df4b2
JK		 2203
2204 /* Update valid block bitmap */
2205 if (del > 0) {
6415fedc 2206 exist = f2fs_test_and_set_bit(offset, se->cur_valid_map);
355e7891 2207#ifdef CONFIG_F2FS_CHECK_FS
6415fedc
YS		 2208 mir_exist = f2fs_test_and_set_bit(offset,
2209 se->cur_valid_map_mir);
2210 if (unlikely(exist != mir_exist)) {
dcbb4c10
JP		 2211 f2fs_err(sbi, "Inconsistent error when setting bitmap, blk:%u, old bit:%d",
2212 blkaddr, exist);
05796763 2213 f2fs_bug_on(sbi, 1);
6415fedc 2214 }
355e7891 2215#endif
6415fedc 2216 if (unlikely(exist)) {
dcbb4c10
JP		 2217 f2fs_err(sbi, "Bitmap was wrongly set, blk:%u",
2218 blkaddr);
6415fedc 2219 f2fs_bug_on(sbi, 1);
35ee82ca
YS		 2220 se->valid_blocks--;
2221 del = 0;
355e7891 2222 }
6415fedc 2223
7d20c8ab 2224 if (!f2fs_test_and_set_bit(offset, se->discard_map))
a66cdd98 2225 sbi->discard_blks--;
720037f9 2226
899fee36
CY		 2227 /*
2228 * SSR should never reuse block which is checkpointed
2229 * or newly invalidated.
2230 */
2231 if (!is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
720037f9
JK		 2232 if (!f2fs_test_and_set_bit(offset, se->ckpt_valid_map))
2233 se->ckpt_valid_blocks++;
2234 }
351df4b2 2235 } else {
6415fedc 2236 exist = f2fs_test_and_clear_bit(offset, se->cur_valid_map);
355e7891 2237#ifdef CONFIG_F2FS_CHECK_FS
6415fedc
YS		 2238 mir_exist = f2fs_test_and_clear_bit(offset,
2239 se->cur_valid_map_mir);
2240 if (unlikely(exist != mir_exist)) {
dcbb4c10
JP		 2241 f2fs_err(sbi, "Inconsistent error when clearing bitmap, blk:%u, old bit:%d",
2242 blkaddr, exist);
05796763 2243 f2fs_bug_on(sbi, 1);
6415fedc 2244 }
355e7891 2245#endif
6415fedc 2246 if (unlikely(!exist)) {
dcbb4c10
JP		 2247 f2fs_err(sbi, "Bitmap was wrongly cleared, blk:%u",
2248 blkaddr);
6415fedc 2249 f2fs_bug_on(sbi, 1);
35ee82ca
YS		 2250 se->valid_blocks++;
2251 del = 0;
4354994f
DR		 2252 } else if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2253 /*
2254 * If checkpoints are off, we must not reuse data that
2255 * was used in the previous checkpoint. If it was used
2256 * before, we must track that to know how much space we
2257 * really have.
2258 */
c9c8ed50
CY		 2259 if (f2fs_test_bit(offset, se->ckpt_valid_map)) {
2260 spin_lock(&sbi->stat_lock);
4354994f 2261 sbi->unusable_block_count++;
c9c8ed50
CY		 2262 spin_unlock(&sbi->stat_lock);
2263 }
355e7891 2264 }
6415fedc 2265
7d20c8ab 2266 if (f2fs_test_and_clear_bit(offset, se->discard_map))
a66cdd98 2267 sbi->discard_blks++;
351df4b2
JK		 2268 }
2269 if (!f2fs_test_bit(offset, se->ckpt_valid_map))
2270 se->ckpt_valid_blocks += del;
2271
2272 __mark_sit_entry_dirty(sbi, segno);
2273
2274 /* update total number of valid blocks to be written in ckpt area */
2275 SIT_I(sbi)->written_valid_blocks += del;
2276
2c70c5e3 2277 if (__is_large_section(sbi))
351df4b2
JK		 2278 get_sec_entry(sbi, segno)->valid_blocks += del;
2279}
2280
4d57b86d 2281void f2fs_invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
351df4b2
JK		 2282{
2283 unsigned int segno = GET_SEGNO(sbi, addr);
2284 struct sit_info *sit_i = SIT_I(sbi);
2285
9850cf4a 2286 f2fs_bug_on(sbi, addr == NULL_ADDR);
4c8ff709 2287 if (addr == NEW_ADDR || addr == COMPRESS_ADDR)
351df4b2
JK		 2288 return;
2289
6aa58d8a
CY		 2290 invalidate_mapping_pages(META_MAPPING(sbi), addr, addr);
2291
351df4b2 2292 /* add it into sit main buffer */
3d26fa6b 2293 down_write(&sit_i->sentry_lock);
351df4b2 2294
c5d02785 2295 update_segment_mtime(sbi, addr, 0);
351df4b2
JK		 2296 update_sit_entry(sbi, addr, -1);
2297
2298 /* add it into dirty seglist */
2299 locate_dirty_segment(sbi, segno);
2300
3d26fa6b 2301 up_write(&sit_i->sentry_lock);
351df4b2
JK		 2302}
2303
4d57b86d 2304bool f2fs_is_checkpointed_data(struct f2fs_sb_info *sbi, block_t blkaddr)
6e2c64ad
JK		 2305{
2306 struct sit_info *sit_i = SIT_I(sbi);
2307 unsigned int segno, offset;
2308 struct seg_entry *se;
2309 bool is_cp = false;
2310
93770ab7 2311 if (!__is_valid_data_blkaddr(blkaddr))
6e2c64ad
JK		 2312 return true;
2313
3d26fa6b 2314 down_read(&sit_i->sentry_lock);
6e2c64ad
JK		 2315
2316 segno = GET_SEGNO(sbi, blkaddr);
2317 se = get_seg_entry(sbi, segno);
2318 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
2319
2320 if (f2fs_test_bit(offset, se->ckpt_valid_map))
2321 is_cp = true;
2322
3d26fa6b 2323 up_read(&sit_i->sentry_lock);
6e2c64ad
JK		 2324
2325 return is_cp;
2326}
2327
0a8165d7 2328/*
351df4b2
JK		 2329 * This function should be resided under the curseg_mutex lock
2330 */
2331static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
e79efe3b 2332 struct f2fs_summary *sum)
351df4b2
JK		 2333{
2334 struct curseg_info *curseg = CURSEG_I(sbi, type);
2335 void *addr = curseg->sum_blk;
e79efe3b 2336 addr += curseg->next_blkoff * sizeof(struct f2fs_summary);
351df4b2 2337 memcpy(addr, sum, sizeof(struct f2fs_summary));
351df4b2
JK		 2338}
2339
0a8165d7 2340/*
351df4b2
JK		 2341 * Calculate the number of current summary pages for writing
2342 */
4d57b86d 2343int f2fs_npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
351df4b2 2344{
351df4b2 2345 int valid_sum_count = 0;
9a47938b 2346 int i, sum_in_page;
351df4b2
JK		 2347
2348 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
2349 if (sbi->ckpt->alloc_type[i] == SSR)
2350 valid_sum_count += sbi->blocks_per_seg;
3fa06d7b
CY		 2351 else {
2352 if (for_ra)
2353 valid_sum_count += le16_to_cpu(
2354 F2FS_CKPT(sbi)->cur_data_blkoff[i]);
2355 else
2356 valid_sum_count += curseg_blkoff(sbi, i);
2357 }
351df4b2
JK		 2358 }
2359
09cbfeaf 2360 sum_in_page = (PAGE_SIZE - 2 * SUM_JOURNAL_SIZE -
9a47938b
FL		 2361 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
2362 if (valid_sum_count <= sum_in_page)
351df4b2 2363 return 1;
9a47938b 2364 else if ((valid_sum_count - sum_in_page) <=
09cbfeaf 2365 (PAGE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
351df4b2
JK		 2366 return 2;
2367 return 3;
2368}
2369
0a8165d7 2370/*
351df4b2
JK		 2371 * Caller should put this summary page
2372 */
4d57b86d 2373struct page *f2fs_get_sum_page(struct f2fs_sb_info *sbi, unsigned int segno)
351df4b2 2374{
86f33603
JK		 2375 if (unlikely(f2fs_cp_error(sbi)))
2376 return ERR_PTR(-EIO);
2377 return f2fs_get_meta_page_retry(sbi, GET_SUM_BLOCK(sbi, segno));
351df4b2
JK		 2378}
2379
4d57b86d
CY		 2380void f2fs_update_meta_page(struct f2fs_sb_info *sbi,
2381 void *src, block_t blk_addr)
351df4b2 2382{
4d57b86d 2383 struct page *page = f2fs_grab_meta_page(sbi, blk_addr);
381722d2 2384
0537b811 2385 memcpy(page_address(page), src, PAGE_SIZE);
351df4b2
JK		 2386 set_page_dirty(page);
2387 f2fs_put_page(page, 1);
2388}
2389
381722d2
CY		 2390static void write_sum_page(struct f2fs_sb_info *sbi,
2391 struct f2fs_summary_block *sum_blk, block_t blk_addr)
2392{
4d57b86d 2393 f2fs_update_meta_page(sbi, (void *)sum_blk, blk_addr);
381722d2
CY		 2394}
2395
b7ad7512
CY		 2396static void write_current_sum_page(struct f2fs_sb_info *sbi,
2397 int type, block_t blk_addr)
2398{
2399 struct curseg_info *curseg = CURSEG_I(sbi, type);
4d57b86d 2400 struct page *page = f2fs_grab_meta_page(sbi, blk_addr);
b7ad7512
CY		 2401 struct f2fs_summary_block *src = curseg->sum_blk;
2402 struct f2fs_summary_block *dst;
2403
2404 dst = (struct f2fs_summary_block *)page_address(page);
81114baa 2405 memset(dst, 0, PAGE_SIZE);
b7ad7512
CY		 2406
2407 mutex_lock(&curseg->curseg_mutex);
2408
2409 down_read(&curseg->journal_rwsem);
2410 memcpy(&dst->journal, curseg->journal, SUM_JOURNAL_SIZE);
2411 up_read(&curseg->journal_rwsem);
2412
2413 memcpy(dst->entries, src->entries, SUM_ENTRY_SIZE);
2414 memcpy(&dst->footer, &src->footer, SUM_FOOTER_SIZE);
2415
2416 mutex_unlock(&curseg->curseg_mutex);
2417
2418 set_page_dirty(page);
2419 f2fs_put_page(page, 1);
2420}
2421
093749e2
CY		 2422static int is_next_segment_free(struct f2fs_sb_info *sbi,
2423 struct curseg_info *curseg, int type)
a7881893 2424{
a7881893
JK		 2425 unsigned int segno = curseg->segno + 1;
2426 struct free_segmap_info *free_i = FREE_I(sbi);
2427
2428 if (segno < MAIN_SEGS(sbi) && segno % sbi->segs_per_sec)
2429 return !test_bit(segno, free_i->free_segmap);
2430 return 0;
2431}
2432
0a8165d7 2433/*
351df4b2
JK		 2434 * Find a new segment from the free segments bitmap to right order
2435 * This function should be returned with success, otherwise BUG
2436 */
2437static void get_new_segment(struct f2fs_sb_info *sbi,
2438 unsigned int *newseg, bool new_sec, int dir)
2439{
2440 struct free_segmap_info *free_i = FREE_I(sbi);
351df4b2 2441 unsigned int segno, secno, zoneno;
7cd8558b 2442 unsigned int total_zones = MAIN_SECS(sbi) / sbi->secs_per_zone;
4ddb1a4d
JK		 2443 unsigned int hint = GET_SEC_FROM_SEG(sbi, *newseg);
2444 unsigned int old_zoneno = GET_ZONE_FROM_SEG(sbi, *newseg);
351df4b2
JK		 2445 unsigned int left_start = hint;
2446 bool init = true;
2447 int go_left = 0;
2448 int i;
2449
1a118ccf 2450 spin_lock(&free_i->segmap_lock);
351df4b2
JK		 2451
2452 if (!new_sec && ((*newseg + 1) % sbi->segs_per_sec)) {
2453 segno = find_next_zero_bit(free_i->free_segmap,
4ddb1a4d
JK		 2454 GET_SEG_FROM_SEC(sbi, hint + 1), *newseg + 1);
2455 if (segno < GET_SEG_FROM_SEC(sbi, hint + 1))
351df4b2
JK		 2456 goto got_it;
2457 }
2458find_other_zone:
7cd8558b
JK		 2459 secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
2460 if (secno >= MAIN_SECS(sbi)) {
351df4b2
JK		 2461 if (dir == ALLOC_RIGHT) {
2462 secno = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 2463 MAIN_SECS(sbi), 0);
2464 f2fs_bug_on(sbi, secno >= MAIN_SECS(sbi));
351df4b2
JK		 2465 } else {
2466 go_left = 1;
2467 left_start = hint - 1;
2468 }
2469 }
2470 if (go_left == 0)
2471 goto skip_left;
2472
2473 while (test_bit(left_start, free_i->free_secmap)) {
2474 if (left_start > 0) {
2475 left_start--;
2476 continue;
2477 }
2478 left_start = find_next_zero_bit(free_i->free_secmap,
7cd8558b
JK		 2479 MAIN_SECS(sbi), 0);
2480 f2fs_bug_on(sbi, left_start >= MAIN_SECS(sbi));
351df4b2
JK		 2481 break;
2482 }
2483 secno = left_start;
2484skip_left:
4ddb1a4d
JK		 2485 segno = GET_SEG_FROM_SEC(sbi, secno);
2486 zoneno = GET_ZONE_FROM_SEC(sbi, secno);
351df4b2
JK		 2487
2488 /* give up on finding another zone */
2489 if (!init)
2490 goto got_it;
2491 if (sbi->secs_per_zone == 1)
2492 goto got_it;
2493 if (zoneno == old_zoneno)
2494 goto got_it;
2495 if (dir == ALLOC_LEFT) {
2496 if (!go_left && zoneno + 1 >= total_zones)
2497 goto got_it;
2498 if (go_left && zoneno == 0)
2499 goto got_it;
2500 }
2501 for (i = 0; i < NR_CURSEG_TYPE; i++)
2502 if (CURSEG_I(sbi, i)->zone == zoneno)
2503 break;
2504
2505 if (i < NR_CURSEG_TYPE) {
2506 /* zone is in user, try another */
2507 if (go_left)
2508 hint = zoneno * sbi->secs_per_zone - 1;
2509 else if (zoneno + 1 >= total_zones)
2510 hint = 0;
2511 else
2512 hint = (zoneno + 1) * sbi->secs_per_zone;
2513 init = false;
2514 goto find_other_zone;
2515 }
2516got_it:
2517 /* set it as dirty segment in free segmap */
9850cf4a 2518 f2fs_bug_on(sbi, test_bit(segno, free_i->free_segmap));
351df4b2
JK		 2519 __set_inuse(sbi, segno);
2520 *newseg = segno;
1a118ccf 2521 spin_unlock(&free_i->segmap_lock);
351df4b2
JK		 2522}
2523
2524static void reset_curseg(struct f2fs_sb_info *sbi, int type, int modified)
2525{
2526 struct curseg_info *curseg = CURSEG_I(sbi, type);
2527 struct summary_footer *sum_footer;
093749e2 2528 unsigned short seg_type = curseg->seg_type;
351df4b2 2529
d0b9e42a 2530 curseg->inited = true;
351df4b2 2531 curseg->segno = curseg->next_segno;
4ddb1a4d 2532 curseg->zone = GET_ZONE_FROM_SEG(sbi, curseg->segno);
351df4b2
JK		 2533 curseg->next_blkoff = 0;
2534 curseg->next_segno = NULL_SEGNO;
2535
2536 sum_footer = &(curseg->sum_blk->footer);
2537 memset(sum_footer, 0, sizeof(struct summary_footer));
093749e2
CY		 2538
2539 sanity_check_seg_type(sbi, seg_type);
2540
2541 if (IS_DATASEG(seg_type))
351df4b2 2542 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
093749e2 2543 if (IS_NODESEG(seg_type))
351df4b2 2544 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
093749e2 2545 __set_sit_entry_type(sbi, seg_type, curseg->segno, modified);
351df4b2
JK		 2546}
2547
7a20b8a6
JK		 2548static unsigned int __get_next_segno(struct f2fs_sb_info *sbi, int type)
2549{
d0b9e42a 2550 struct curseg_info *curseg = CURSEG_I(sbi, type);
093749e2
CY		 2551 unsigned short seg_type = curseg->seg_type;
2552
2553 sanity_check_seg_type(sbi, seg_type);
d0b9e42a 2554
a7881893 2555 /* if segs_per_sec is large than 1, we need to keep original policy. */
2c70c5e3 2556 if (__is_large_section(sbi))
d0b9e42a
CY		 2557 return curseg->segno;
2558
2559 /* inmem log may not locate on any segment after mount */
2560 if (!curseg->inited)
2561 return 0;
a7881893 2562
4354994f
DR		 2563 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2564 return 0;
2565
b94929d9 2566 if (test_opt(sbi, NOHEAP) &&
093749e2 2567 (seg_type == CURSEG_HOT_DATA || IS_NODESEG(seg_type)))
7a20b8a6
JK		 2568 return 0;
2569
e066b83c
JK		 2570 if (SIT_I(sbi)->last_victim[ALLOC_NEXT])
2571 return SIT_I(sbi)->last_victim[ALLOC_NEXT];
07939627
JK		 2572
2573 /* find segments from 0 to reuse freed segments */
63189b78 2574 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
07939627
JK		 2575 return 0;
2576
d0b9e42a 2577 return curseg->segno;
7a20b8a6
JK		 2578}
2579
0a8165d7 2580/*
351df4b2
JK		 2581 * Allocate a current working segment.
2582 * This function always allocates a free segment in LFS manner.
2583 */
2584static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
2585{
2586 struct curseg_info *curseg = CURSEG_I(sbi, type);
d0b9e42a 2587 unsigned short seg_type = curseg->seg_type;
351df4b2
JK		 2588 unsigned int segno = curseg->segno;
2589 int dir = ALLOC_LEFT;
2590
d0b9e42a
CY		 2591 if (curseg->inited)
2592 write_sum_page(sbi, curseg->sum_blk,
81fb5e87 2593 GET_SUM_BLOCK(sbi, segno));
d0b9e42a 2594 if (seg_type == CURSEG_WARM_DATA || seg_type == CURSEG_COLD_DATA)
351df4b2
JK		 2595 dir = ALLOC_RIGHT;
2596
2597 if (test_opt(sbi, NOHEAP))
2598 dir = ALLOC_RIGHT;
2599
7a20b8a6 2600 segno = __get_next_segno(sbi, type);
351df4b2
JK		 2601 get_new_segment(sbi, &segno, new_sec, dir);
2602 curseg->next_segno = segno;
2603 reset_curseg(sbi, type, 1);
2604 curseg->alloc_type = LFS;
2605}
2606
2607static void __next_free_blkoff(struct f2fs_sb_info *sbi,
2608 struct curseg_info *seg, block_t start)
2609{
2610 struct seg_entry *se = get_seg_entry(sbi, seg->segno);
e81c93cf 2611 int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
60a3b782 2612 unsigned long *target_map = SIT_I(sbi)->tmp_map;
e81c93cf
CL		 2613 unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
2614 unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
2615 int i, pos;
2616
2617 for (i = 0; i < entries; i++)
2618 target_map[i] = ckpt_map[i] | cur_map[i];
2619
2620 pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, start);
2621
2622 seg->next_blkoff = pos;
351df4b2
JK		 2623}
2624
0a8165d7 2625/*
351df4b2
JK		 2626 * If a segment is written by LFS manner, next block offset is just obtained
2627 * by increasing the current block offset. However, if a segment is written by
2628 * SSR manner, next block offset obtained by calling __next_free_blkoff
2629 */
2630static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
2631 struct curseg_info *seg)
2632{
2633 if (seg->alloc_type == SSR)
2634 __next_free_blkoff(sbi, seg, seg->next_blkoff + 1);
2635 else
2636 seg->next_blkoff++;
2637}
2638
0a8165d7 2639/*
e1c42045 2640 * This function always allocates a used segment(from dirty seglist) by SSR
351df4b2
JK		 2641 * manner, so it should recover the existing segment information of valid blocks
2642 */
093749e2 2643static void change_curseg(struct f2fs_sb_info *sbi, int type, bool flush)
351df4b2
JK		 2644{
2645 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
2646 struct curseg_info *curseg = CURSEG_I(sbi, type);
2647 unsigned int new_segno = curseg->next_segno;
2648 struct f2fs_summary_block *sum_node;
2649 struct page *sum_page;
2650
093749e2
CY		 2651 if (flush)
2652 write_sum_page(sbi, curseg->sum_blk,
2653 GET_SUM_BLOCK(sbi, curseg->segno));
2654
351df4b2
JK		 2655 __set_test_and_inuse(sbi, new_segno);
2656
2657 mutex_lock(&dirty_i->seglist_lock);
2658 __remove_dirty_segment(sbi, new_segno, PRE);
2659 __remove_dirty_segment(sbi, new_segno, DIRTY);
2660 mutex_unlock(&dirty_i->seglist_lock);
2661
2662 reset_curseg(sbi, type, 1);
2663 curseg->alloc_type = SSR;
2664 __next_free_blkoff(sbi, curseg, 0);
2665
4d57b86d 2666 sum_page = f2fs_get_sum_page(sbi, new_segno);
86f33603
JK		 2667 if (IS_ERR(sum_page)) {
2668 /* GC won't be able to use stale summary pages by cp_error */
2669 memset(curseg->sum_blk, 0, SUM_ENTRY_SIZE);
2670 return;
2671 }
025d63a4
CY		 2672 sum_node = (struct f2fs_summary_block *)page_address(sum_page);
2673 memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
2674 f2fs_put_page(sum_page, 1);
351df4b2
JK		 2675}
2676
093749e2
CY		 2677static int get_ssr_segment(struct f2fs_sb_info *sbi, int type,
2678 int alloc_mode, unsigned long long age);
2679
2680static void get_atssr_segment(struct f2fs_sb_info *sbi, int type,
2681 int target_type, int alloc_mode,
2682 unsigned long long age)
2683{
2684 struct curseg_info *curseg = CURSEG_I(sbi, type);
2685
2686 curseg->seg_type = target_type;
2687
2688 if (get_ssr_segment(sbi, type, alloc_mode, age)) {
2689 struct seg_entry *se = get_seg_entry(sbi, curseg->next_segno);
2690
2691 curseg->seg_type = se->type;
2692 change_curseg(sbi, type, true);
2693 } else {
2694 /* allocate cold segment by default */
2695 curseg->seg_type = CURSEG_COLD_DATA;
2696 new_curseg(sbi, type, true);
2697 }
2698 stat_inc_seg_type(sbi, curseg);
2699}
2700
2701static void __f2fs_init_atgc_curseg(struct f2fs_sb_info *sbi)
2702{
2703 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC);
2704
2705 if (!sbi->am.atgc_enabled)
2706 return;
2707
2708 down_read(&SM_I(sbi)->curseg_lock);
2709
2710 mutex_lock(&curseg->curseg_mutex);
2711 down_write(&SIT_I(sbi)->sentry_lock);
2712
2713 get_atssr_segment(sbi, CURSEG_ALL_DATA_ATGC, CURSEG_COLD_DATA, SSR, 0);
2714
2715 up_write(&SIT_I(sbi)->sentry_lock);
2716 mutex_unlock(&curseg->curseg_mutex);
2717
2718 up_read(&SM_I(sbi)->curseg_lock);
2719
2720}
2721void f2fs_init_inmem_curseg(struct f2fs_sb_info *sbi)
2722{
2723 __f2fs_init_atgc_curseg(sbi);
2724}
2725
2726static void __f2fs_save_inmem_curseg(struct f2fs_sb_info *sbi, int type)
d0b9e42a
CY		 2727{
2728 struct curseg_info *curseg = CURSEG_I(sbi, type);
2729
2730 mutex_lock(&curseg->curseg_mutex);
2731 if (!curseg->inited)
2732 goto out;
2733
2734 if (get_valid_blocks(sbi, curseg->segno, false)) {
2735 write_sum_page(sbi, curseg->sum_blk,
2736 GET_SUM_BLOCK(sbi, curseg->segno));
2737 } else {
2738 mutex_lock(&DIRTY_I(sbi)->seglist_lock);
2739 __set_test_and_free(sbi, curseg->segno, true);
2740 mutex_unlock(&DIRTY_I(sbi)->seglist_lock);
2741 }
2742out:
2743 mutex_unlock(&curseg->curseg_mutex);
2744}
2745
093749e2
CY		 2746void f2fs_save_inmem_curseg(struct f2fs_sb_info *sbi)
2747{
2748 __f2fs_save_inmem_curseg(sbi, CURSEG_COLD_DATA_PINNED);
2749
2750 if (sbi->am.atgc_enabled)
2751 __f2fs_save_inmem_curseg(sbi, CURSEG_ALL_DATA_ATGC);
2752}
2753
2754static void __f2fs_restore_inmem_curseg(struct f2fs_sb_info *sbi, int type)
d0b9e42a
CY		 2755{
2756 struct curseg_info *curseg = CURSEG_I(sbi, type);
2757
2758 mutex_lock(&curseg->curseg_mutex);
2759 if (!curseg->inited)
2760 goto out;
2761 if (get_valid_blocks(sbi, curseg->segno, false))
2762 goto out;
2763
2764 mutex_lock(&DIRTY_I(sbi)->seglist_lock);
2765 __set_test_and_inuse(sbi, curseg->segno);
2766 mutex_unlock(&DIRTY_I(sbi)->seglist_lock);
2767out:
2768 mutex_unlock(&curseg->curseg_mutex);
2769}
2770
093749e2
CY		 2771void f2fs_restore_inmem_curseg(struct f2fs_sb_info *sbi)
2772{
2773 __f2fs_restore_inmem_curseg(sbi, CURSEG_COLD_DATA_PINNED);
2774
2775 if (sbi->am.atgc_enabled)
2776 __f2fs_restore_inmem_curseg(sbi, CURSEG_ALL_DATA_ATGC);
2777}
2778
2779static int get_ssr_segment(struct f2fs_sb_info *sbi, int type,
2780 int alloc_mode, unsigned long long age)
43727527
JK		 2781{
2782 struct curseg_info *curseg = CURSEG_I(sbi, type);
2783 const struct victim_selection *v_ops = DIRTY_I(sbi)->v_ops;
e066b83c 2784 unsigned segno = NULL_SEGNO;
093749e2 2785 unsigned short seg_type = curseg->seg_type;
d27c3d89
CY		 2786 int i, cnt;
2787 bool reversed = false;
c192f7a4 2788
093749e2
CY		 2789 sanity_check_seg_type(sbi, seg_type);
2790
4d57b86d 2791 /* f2fs_need_SSR() already forces to do this */
093749e2 2792 if (!v_ops->get_victim(sbi, &segno, BG_GC, seg_type, alloc_mode, age)) {
e066b83c 2793 curseg->next_segno = segno;
c192f7a4 2794 return 1;
e066b83c 2795 }
43727527 2796
70d625cb 2797 /* For node segments, let's do SSR more intensively */
093749e2
CY		 2798 if (IS_NODESEG(seg_type)) {
2799 if (seg_type >= CURSEG_WARM_NODE) {
d27c3d89
CY		 2800 reversed = true;
2801 i = CURSEG_COLD_NODE;
2802 } else {
2803 i = CURSEG_HOT_NODE;
2804 }
2805 cnt = NR_CURSEG_NODE_TYPE;
70d625cb 2806 } else {
093749e2 2807 if (seg_type >= CURSEG_WARM_DATA) {
d27c3d89
CY		 2808 reversed = true;
2809 i = CURSEG_COLD_DATA;
2810 } else {
2811 i = CURSEG_HOT_DATA;
2812 }
2813 cnt = NR_CURSEG_DATA_TYPE;
70d625cb 2814 }
43727527 2815
d27c3d89 2816 for (; cnt-- > 0; reversed ? i-- : i++) {
093749e2 2817 if (i == seg_type)
c192f7a4 2818 continue;
093749e2 2819 if (!v_ops->get_victim(sbi, &segno, BG_GC, i, alloc_mode, age)) {
e066b83c 2820 curseg->next_segno = segno;
43727527 2821 return 1;
e066b83c 2822 }
c192f7a4 2823 }
4354994f
DR		 2824
2825 /* find valid_blocks=0 in dirty list */
2826 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2827 segno = get_free_segment(sbi);
2828 if (segno != NULL_SEGNO) {
2829 curseg->next_segno = segno;
2830 return 1;
2831 }
2832 }
43727527
JK		 2833 return 0;
2834}
2835
351df4b2
JK		 2836/*
2837 * flush out current segment and replace it with new segment
2838 * This function should be returned with success, otherwise BUG
2839 */
2840static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
2841 int type, bool force)
2842{
a7881893
JK		 2843 struct curseg_info *curseg = CURSEG_I(sbi, type);
2844
7b405275 2845 if (force)
351df4b2 2846 new_curseg(sbi, type, true);
5b6c6be2 2847 else if (!is_set_ckpt_flags(sbi, CP_CRC_RECOVERY_FLAG) &&
093749e2 2848 curseg->seg_type == CURSEG_WARM_NODE)
351df4b2 2849 new_curseg(sbi, type, false);
093749e2
CY		 2850 else if (curseg->alloc_type == LFS &&
2851 is_next_segment_free(sbi, curseg, type) &&
4354994f 2852 likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
a7881893 2853 new_curseg(sbi, type, false);
093749e2
CY		 2854 else if (f2fs_need_SSR(sbi) &&
2855 get_ssr_segment(sbi, type, SSR, 0))
2856 change_curseg(sbi, type, true);
351df4b2
JK		 2857 else
2858 new_curseg(sbi, type, false);
dcdfff65 2859
a7881893 2860 stat_inc_seg_type(sbi, curseg);
351df4b2
JK		 2861}
2862
0ef81833 2863void f2fs_allocate_segment_for_resize(struct f2fs_sb_info *sbi, int type,
04f0b2ea
QS		 2864 unsigned int start, unsigned int end)
2865{
2866 struct curseg_info *curseg = CURSEG_I(sbi, type);
2867 unsigned int segno;
2868
2869 down_read(&SM_I(sbi)->curseg_lock);
2870 mutex_lock(&curseg->curseg_mutex);
2871 down_write(&SIT_I(sbi)->sentry_lock);
2872
2873 segno = CURSEG_I(sbi, type)->segno;
2874 if (segno < start || segno > end)
2875 goto unlock;
2876
093749e2
CY		 2877 if (f2fs_need_SSR(sbi) && get_ssr_segment(sbi, type, SSR, 0))
2878 change_curseg(sbi, type, true);
04f0b2ea
QS		 2879 else
2880 new_curseg(sbi, type, true);
2881
2882 stat_inc_seg_type(sbi, curseg);
2883
2884 locate_dirty_segment(sbi, segno);
2885unlock:
2886 up_write(&SIT_I(sbi)->sentry_lock);
2887
2888 if (segno != curseg->segno)
dcbb4c10
JP		 2889 f2fs_notice(sbi, "For resize: curseg of type %d: %u ==> %u",
2890 type, segno, curseg->segno);
04f0b2ea
QS		 2891
2892 mutex_unlock(&curseg->curseg_mutex);
2893 up_read(&SM_I(sbi)->curseg_lock);
2894}
2895
901d745f 2896static void __allocate_new_segment(struct f2fs_sb_info *sbi, int type)
351df4b2 2897{
901d745f 2898 struct curseg_info *curseg = CURSEG_I(sbi, type);
6ae1be13 2899 unsigned int old_segno;
351df4b2 2900
d0b9e42a
CY		 2901 if (!curseg->inited)
2902 goto alloc;
2903
901d745f
CY		 2904 if (!curseg->next_blkoff &&
2905 !get_valid_blocks(sbi, curseg->segno, false) &&
2906 !get_ckpt_valid_blocks(sbi, curseg->segno))
2907 return;
3d26fa6b 2908
d0b9e42a 2909alloc:
901d745f
CY		 2910 old_segno = curseg->segno;
2911 SIT_I(sbi)->s_ops->allocate_segment(sbi, type, true);
2912 locate_dirty_segment(sbi, old_segno);
2913}
f5a53edc 2914
901d745f
CY		 2915void f2fs_allocate_new_segment(struct f2fs_sb_info *sbi, int type)
2916{
2917 down_write(&SIT_I(sbi)->sentry_lock);
2918 __allocate_new_segment(sbi, type);
2919 up_write(&SIT_I(sbi)->sentry_lock);
2920}
3d26fa6b 2921
901d745f
CY		 2922void f2fs_allocate_new_segments(struct f2fs_sb_info *sbi)
2923{
2924 int i;
2925
2926 down_write(&SIT_I(sbi)->sentry_lock);
2927 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++)
2928 __allocate_new_segment(sbi, i);
3d26fa6b 2929 up_write(&SIT_I(sbi)->sentry_lock);
351df4b2
JK		 2930}
2931
2932static const struct segment_allocation default_salloc_ops = {
2933 .allocate_segment = allocate_segment_by_default,
2934};
2935
4d57b86d
CY		 2936bool f2fs_exist_trim_candidates(struct f2fs_sb_info *sbi,
2937 struct cp_control *cpc)
25290fa5
JK		 2938{
2939 __u64 trim_start = cpc->trim_start;
2940 bool has_candidate = false;
2941
3d26fa6b 2942 down_write(&SIT_I(sbi)->sentry_lock);
25290fa5
JK		 2943 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++) {
2944 if (add_discard_addrs(sbi, cpc, true)) {
2945 has_candidate = true;
2946 break;
2947 }
2948 }
3d26fa6b 2949 up_write(&SIT_I(sbi)->sentry_lock);
25290fa5
JK		 2950
2951 cpc->trim_start = trim_start;
2952 return has_candidate;
2953}
2954
01f9cf6d 2955static unsigned int __issue_discard_cmd_range(struct f2fs_sb_info *sbi,
9a997188
JK		 2956 struct discard_policy *dpolicy,
2957 unsigned int start, unsigned int end)
2958{
2959 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
2960 struct discard_cmd *prev_dc = NULL, *next_dc = NULL;
2961 struct rb_node **insert_p = NULL, *insert_parent = NULL;
2962 struct discard_cmd *dc;
2963 struct blk_plug plug;
2964 int issued;
01f9cf6d 2965 unsigned int trimmed = 0;
9a997188
JK		 2966
2967next:
2968 issued = 0;
2969
2970 mutex_lock(&dcc->cmd_lock);
67fce70b
CY		 2971 if (unlikely(dcc->rbtree_check))
2972 f2fs_bug_on(sbi, !f2fs_check_rb_tree_consistence(sbi,
2e9b2bb2 2973 &dcc->root, false));
9a997188 2974
4d57b86d 2975 dc = (struct discard_cmd *)f2fs_lookup_rb_tree_ret(&dcc->root,
9a997188
JK		 2976 NULL, start,
2977 (struct rb_entry **)&prev_dc,
2978 (struct rb_entry **)&next_dc,
4dada3fd 2979 &insert_p, &insert_parent, true, NULL);
9a997188
JK		 2980 if (!dc)
2981 dc = next_dc;
2982
2983 blk_start_plug(&plug);
2984
2985 while (dc && dc->lstart <= end) {
2986 struct rb_node *node;
6b9cb124 2987 int err = 0;
9a997188
JK		 2988
2989 if (dc->len < dpolicy->granularity)
2990 goto skip;
2991
2992 if (dc->state != D_PREP) {
2993 list_move_tail(&dc->list, &dcc->fstrim_list);
2994 goto skip;
2995 }
2996
6b9cb124 2997 err = __submit_discard_cmd(sbi, dpolicy, dc, &issued);
9a997188 2998
35ec7d57 2999 if (issued >= dpolicy->max_requests) {
9a997188
JK		 3000 start = dc->lstart + dc->len;
3001
6b9cb124
CY		 3002 if (err)
3003 __remove_discard_cmd(sbi, dc);
3004
9a997188
JK		 3005 blk_finish_plug(&plug);
3006 mutex_unlock(&dcc->cmd_lock);
01f9cf6d 3007 trimmed += __wait_all_discard_cmd(sbi, NULL);
5df7731f 3008 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
9a997188
JK		 3009 goto next;
3010 }
3011skip:
3012 node = rb_next(&dc->rb_node);
6b9cb124
CY		 3013 if (err)
3014 __remove_discard_cmd(sbi, dc);
9a997188
JK		 3015 dc = rb_entry_safe(node, struct discard_cmd, rb_node);
3016
3017 if (fatal_signal_pending(current))
3018 break;
3019 }
3020
3021 blk_finish_plug(&plug);
3022 mutex_unlock(&dcc->cmd_lock);
01f9cf6d
CY		 3023
3024 return trimmed;
9a997188
JK		 3025}
3026
4b2fecc8
JK		 3027int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
3028{
f7ef9b83
JK		 3029 __u64 start = F2FS_BYTES_TO_BLK(range->start);
3030 __u64 end = start + F2FS_BYTES_TO_BLK(range->len) - 1;
377224c4 3031 unsigned int start_segno, end_segno;
8412663d 3032 block_t start_block, end_block;
4b2fecc8 3033 struct cp_control cpc;
78997b56 3034 struct discard_policy dpolicy;
0ea80512 3035 unsigned long long trimmed = 0;
c34f42e2 3036 int err = 0;
b0332a0f 3037 bool need_align = f2fs_lfs_mode(sbi) && __is_large_section(sbi);
4b2fecc8 3038
836b5a63 3039 if (start >= MAX_BLKADDR(sbi) || range->len < sbi->blocksize)
4b2fecc8
JK		 3040 return -EINVAL;
3041
3f16ecd9
CY		 3042 if (end < MAIN_BLKADDR(sbi))
3043 goto out;
4b2fecc8 3044
ed214a11 3045 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK)) {
dcbb4c10 3046 f2fs_warn(sbi, "Found FS corruption, run fsck to fix.");
10f966bb 3047 return -EFSCORRUPTED;
ed214a11
YH		 3048 }
3049
4b2fecc8 3050 /* start/end segment number in main_area */
7cd8558b
JK		 3051 start_segno = (start <= MAIN_BLKADDR(sbi)) ? 0 : GET_SEGNO(sbi, start);
3052 end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
3053 GET_SEGNO(sbi, end);
ad6672bb
YS		 3054 if (need_align) {
3055 start_segno = rounddown(start_segno, sbi->segs_per_sec);
3056 end_segno = roundup(end_segno + 1, sbi->segs_per_sec) - 1;
3057 }
8412663d 3058
4b2fecc8 3059 cpc.reason = CP_DISCARD;
836b5a63 3060 cpc.trim_minlen = max_t(__u64, 1, F2FS_BYTES_TO_BLK(range->minlen));
377224c4
CY		 3061 cpc.trim_start = start_segno;
3062 cpc.trim_end = end_segno;
4b2fecc8 3063
377224c4
CY		 3064 if (sbi->discard_blks == 0)
3065 goto out;
74fa5f3d 3066
fb24fea7 3067 down_write(&sbi->gc_lock);
4d57b86d 3068 err = f2fs_write_checkpoint(sbi, &cpc);
fb24fea7 3069 up_write(&sbi->gc_lock);
377224c4
CY		 3070 if (err)
3071 goto out;
8412663d 3072
e555da9f
JK		 3073 /*
3074 * We filed discard candidates, but actually we don't need to wait for
3075 * all of them, since they'll be issued in idle time along with runtime
3076 * discard option. User configuration looks like using runtime discard
3077 * or periodic fstrim instead of it.
3078 */
7d20c8ab 3079 if (f2fs_realtime_discard_enable(sbi))
5a615492
JK		 3080 goto out;
3081
3082 start_block = START_BLOCK(sbi, start_segno);
3083 end_block = START_BLOCK(sbi, end_segno + 1);
3084
3085 __init_discard_policy(sbi, &dpolicy, DPOLICY_FSTRIM, cpc.trim_minlen);
01f9cf6d
CY		 3086 trimmed = __issue_discard_cmd_range(sbi, &dpolicy,
3087 start_block, end_block);
5a615492 3088
01f9cf6d 3089 trimmed += __wait_discard_cmd_range(sbi, &dpolicy,
0ea80512 3090 start_block, end_block);
377224c4 3091out:
6eae2694
CY		 3092 if (!err)
3093 range->len = F2FS_BLK_TO_BYTES(trimmed);
c34f42e2 3094 return err;
4b2fecc8
JK		 3095}
3096
093749e2
CY		 3097static bool __has_curseg_space(struct f2fs_sb_info *sbi,
3098 struct curseg_info *curseg)
351df4b2 3099{
de881df9
AR		 3100 return curseg->next_blkoff < f2fs_usable_blks_in_seg(sbi,
3101 curseg->segno);
351df4b2
JK		 3102}
3103
4d57b86d 3104int f2fs_rw_hint_to_seg_type(enum rw_hint hint)
4f0a03d3
HL		 3105{
3106 switch (hint) {
3107 case WRITE_LIFE_SHORT:
3108 return CURSEG_HOT_DATA;
3109 case WRITE_LIFE_EXTREME:
3110 return CURSEG_COLD_DATA;
3111 default:
3112 return CURSEG_WARM_DATA;
3113 }
3114}
3115
0cdd3195
HL		 3116/* This returns write hints for each segment type. This hints will be
3117 * passed down to block layer. There are mapping tables which depend on
3118 * the mount option 'whint_mode'.
3119 *
3120 * 1) whint_mode=off. F2FS only passes down WRITE_LIFE_NOT_SET.
3121 *
3122 * 2) whint_mode=user-based. F2FS tries to pass down hints given by users.
3123 *
3124 * User F2FS Block
3125 * ---- ---- -----
3126 * META WRITE_LIFE_NOT_SET
3127 * HOT_NODE "
3128 * WARM_NODE "
3129 * COLD_NODE "
3130 * ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
3131 * extension list " "
3132 *
3133 * -- buffered io
3134 * WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
3135 * WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
3136 * WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
3137 * WRITE_LIFE_NONE " "
3138 * WRITE_LIFE_MEDIUM " "
3139 * WRITE_LIFE_LONG " "
3140 *
3141 * -- direct io
3142 * WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
3143 * WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
3144 * WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
3145 * WRITE_LIFE_NONE " WRITE_LIFE_NONE
3146 * WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
3147 * WRITE_LIFE_LONG " WRITE_LIFE_LONG
3148 *
f2e703f9
HL		 3149 * 3) whint_mode=fs-based. F2FS passes down hints with its policy.
3150 *
3151 * User F2FS Block
3152 * ---- ---- -----
3153 * META WRITE_LIFE_MEDIUM;
3154 * HOT_NODE WRITE_LIFE_NOT_SET
3155 * WARM_NODE "
3156 * COLD_NODE WRITE_LIFE_NONE
3157 * ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
3158 * extension list " "
3159 *
3160 * -- buffered io
3161 * WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
3162 * WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
3163 * WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_LONG
3164 * WRITE_LIFE_NONE " "
3165 * WRITE_LIFE_MEDIUM " "
3166 * WRITE_LIFE_LONG " "
3167 *
3168 * -- direct io
3169 * WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
3170 * WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
3171 * WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
3172 * WRITE_LIFE_NONE " WRITE_LIFE_NONE
3173 * WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
3174 * WRITE_LIFE_LONG " WRITE_LIFE_LONG
0cdd3195
HL		 3175 */
3176
4d57b86d 3177enum rw_hint f2fs_io_type_to_rw_hint(struct f2fs_sb_info *sbi,
0cdd3195
HL		 3178 enum page_type type, enum temp_type temp)
3179{
63189b78 3180 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) {
0cdd3195 3181 if (type == DATA) {
f2e703f9 3182 if (temp == WARM)
0cdd3195 3183 return WRITE_LIFE_NOT_SET;
f2e703f9
HL		 3184 else if (temp == HOT)
3185 return WRITE_LIFE_SHORT;
3186 else if (temp == COLD)
3187 return WRITE_LIFE_EXTREME;
0cdd3195
HL		 3188 } else {
3189 return WRITE_LIFE_NOT_SET;
3190 }
63189b78 3191 } else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) {
f2e703f9
HL		 3192 if (type == DATA) {
3193 if (temp == WARM)
3194 return WRITE_LIFE_LONG;
3195 else if (temp == HOT)
3196 return WRITE_LIFE_SHORT;
3197 else if (temp == COLD)
3198 return WRITE_LIFE_EXTREME;
3199 } else if (type == NODE) {
3200 if (temp == WARM || temp == HOT)
3201 return WRITE_LIFE_NOT_SET;
3202 else if (temp == COLD)
3203 return WRITE_LIFE_NONE;
3204 } else if (type == META) {
3205 return WRITE_LIFE_MEDIUM;
3206 }
0cdd3195 3207 }
f2e703f9 3208 return WRITE_LIFE_NOT_SET;
0cdd3195
HL		 3209}
3210
81377bd6 3211static int __get_segment_type_2(struct f2fs_io_info *fio)
351df4b2 3212{
81377bd6 3213 if (fio->type == DATA)
351df4b2
JK		 3214 return CURSEG_HOT_DATA;
3215 else
3216 return CURSEG_HOT_NODE;
3217}
3218
81377bd6 3219static int __get_segment_type_4(struct f2fs_io_info *fio)
351df4b2 3220{
81377bd6
JK		 3221 if (fio->type == DATA) {
3222 struct inode *inode = fio->page->mapping->host;
351df4b2
JK		 3223
3224 if (S_ISDIR(inode->i_mode))
3225 return CURSEG_HOT_DATA;
3226 else
3227 return CURSEG_COLD_DATA;
3228 } else {
81377bd6 3229 if (IS_DNODE(fio->page) && is_cold_node(fio->page))
a344b9fd 3230 return CURSEG_WARM_NODE;
351df4b2
JK		 3231 else
3232 return CURSEG_COLD_NODE;
3233 }
3234}
3235
81377bd6 3236static int __get_segment_type_6(struct f2fs_io_info *fio)
351df4b2 3237{
81377bd6
JK		 3238 if (fio->type == DATA) {
3239 struct inode *inode = fio->page->mapping->host;
351df4b2 3240
093749e2
CY		 3241 if (is_cold_data(fio->page)) {
3242 if (fio->sbi->am.atgc_enabled)
3243 return CURSEG_ALL_DATA_ATGC;
3244 else
3245 return CURSEG_COLD_DATA;
3246 }
602a16d5 3247 if (file_is_cold(inode) || f2fs_need_compress_data(inode))
351df4b2 3248 return CURSEG_COLD_DATA;
b6a06cbb 3249 if (file_is_hot(inode) ||
b4c3ca8b 3250 is_inode_flag_set(inode, FI_HOT_DATA) ||
2079f115
CY		 3251 f2fs_is_atomic_file(inode) ||
3252 f2fs_is_volatile_file(inode))
ef095d19 3253 return CURSEG_HOT_DATA;
4d57b86d 3254 return f2fs_rw_hint_to_seg_type(inode->i_write_hint);
351df4b2 3255 } else {
81377bd6
JK		 3256 if (IS_DNODE(fio->page))
3257 return is_cold_node(fio->page) ? CURSEG_WARM_NODE :
351df4b2 3258 CURSEG_HOT_NODE;
ef095d19 3259 return CURSEG_COLD_NODE;
351df4b2
JK		 3260 }
3261}
3262
81377bd6 3263static int __get_segment_type(struct f2fs_io_info *fio)
351df4b2 3264{
a912b54d
JK		 3265 int type = 0;
3266
63189b78 3267 switch (F2FS_OPTION(fio->sbi).active_logs) {
351df4b2 3268 case 2:
a912b54d
JK		 3269 type = __get_segment_type_2(fio);
3270 break;
351df4b2 3271 case 4:
a912b54d
JK		 3272 type = __get_segment_type_4(fio);
3273 break;
3274 case 6:
3275 type = __get_segment_type_6(fio);
3276 break;
3277 default:
3278 f2fs_bug_on(fio->sbi, true);
351df4b2 3279 }
81377bd6 3280
a912b54d
JK		 3281 if (IS_HOT(type))
3282 fio->temp = HOT;
3283 else if (IS_WARM(type))
3284 fio->temp = WARM;
3285 else
3286 fio->temp = COLD;
3287 return type;
351df4b2
JK		 3288}
3289
4d57b86d 3290void f2fs_allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
bfad7c2d 3291 block_t old_blkaddr, block_t *new_blkaddr,
fb830fc5 3292 struct f2fs_summary *sum, int type,
093749e2 3293 struct f2fs_io_info *fio)
351df4b2
JK		 3294{
3295 struct sit_info *sit_i = SIT_I(sbi);
6ae1be13 3296 struct curseg_info *curseg = CURSEG_I(sbi, type);
c5d02785 3297 unsigned long long old_mtime;
093749e2
CY		 3298 bool from_gc = (type == CURSEG_ALL_DATA_ATGC);
3299 struct seg_entry *se = NULL;
351df4b2 3300
2b60311d
CY		 3301 down_read(&SM_I(sbi)->curseg_lock);
3302
351df4b2 3303 mutex_lock(&curseg->curseg_mutex);
3d26fa6b 3304 down_write(&sit_i->sentry_lock);
351df4b2 3305
093749e2
CY		 3306 if (from_gc) {
3307 f2fs_bug_on(sbi, GET_SEGNO(sbi, old_blkaddr) == NULL_SEGNO);
3308 se = get_seg_entry(sbi, GET_SEGNO(sbi, old_blkaddr));
3309 sanity_check_seg_type(sbi, se->type);
3310 f2fs_bug_on(sbi, IS_NODESEG(se->type));
3311 }
351df4b2 3312 *new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
351df4b2 3313
093749e2
CY		 3314 f2fs_bug_on(sbi, curseg->next_blkoff >= sbi->blocks_per_seg);
3315
4e6a8d9b
JK		 3316 f2fs_wait_discard_bio(sbi, *new_blkaddr);
3317
351df4b2
JK		 3318 /*
3319 * __add_sum_entry should be resided under the curseg_mutex
3320 * because, this function updates a summary entry in the
3321 * current summary block.
3322 */
e79efe3b 3323 __add_sum_entry(sbi, type, sum);
351df4b2 3324
351df4b2 3325 __refresh_next_blkoff(sbi, curseg);
dcdfff65
JK		 3326
3327 stat_inc_block_count(sbi, curseg);
351df4b2 3328
c5d02785
CY		 3329 if (from_gc) {
3330 old_mtime = get_segment_mtime(sbi, old_blkaddr);
3331 } else {
3332 update_segment_mtime(sbi, old_blkaddr, 0);
3333 old_mtime = 0;
3334 }
3335 update_segment_mtime(sbi, *new_blkaddr, old_mtime);
3336
65f1b80b
YS		 3337 /*
3338 * SIT information should be updated before segment allocation,
3339 * since SSR needs latest valid block information.
3340 */
3341 update_sit_entry(sbi, *new_blkaddr, 1);
3342 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
3343 update_sit_entry(sbi, old_blkaddr, -1);
3344
093749e2
CY		 3345 if (!__has_curseg_space(sbi, curseg)) {
3346 if (from_gc)
3347 get_atssr_segment(sbi, type, se->type,
3348 AT_SSR, se->mtime);
3349 else
3350 sit_i->s_ops->allocate_segment(sbi, type, false);
3351 }
351df4b2 3352 /*
65f1b80b
YS		 3353 * segment dirty status should be updated after segment allocation,
3354 * so we just need to update status only one time after previous
3355 * segment being closed.
351df4b2 3356 */
65f1b80b
YS		 3357 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
3358 locate_dirty_segment(sbi, GET_SEGNO(sbi, *new_blkaddr));
5e443818 3359
3d26fa6b 3360 up_write(&sit_i->sentry_lock);
351df4b2 3361
704956ec 3362 if (page && IS_NODESEG(type)) {
351df4b2
JK		 3363 fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
3364
704956ec
CY		 3365 f2fs_inode_chksum_set(sbi, page);
3366 }
3367
8223ecc4
CY		 3368 if (F2FS_IO_ALIGNED(sbi))
3369 fio->retry = false;
3370
f608c38c 3371 if (fio) {
fb830fc5
CY		 3372 struct f2fs_bio_info *io;
3373
3374 INIT_LIST_HEAD(&fio->list);
3375 fio->in_list = true;
3376 io = sbi->write_io[fio->type] + fio->temp;
3377 spin_lock(&io->io_lock);
3378 list_add_tail(&fio->list, &io->io_list);
3379 spin_unlock(&io->io_lock);
3380 }
3381
bfad7c2d 3382 mutex_unlock(&curseg->curseg_mutex);
2b60311d
CY		 3383
3384 up_read(&SM_I(sbi)->curseg_lock);
bfad7c2d
JK		 3385}
3386
39d787be
CY		 3387static void update_device_state(struct f2fs_io_info *fio)
3388{
3389 struct f2fs_sb_info *sbi = fio->sbi;
3390 unsigned int devidx;
3391
0916878d 3392 if (!f2fs_is_multi_device(sbi))
39d787be
CY		 3393 return;
3394
3395 devidx = f2fs_target_device_index(sbi, fio->new_blkaddr);
3396
3397 /* update device state for fsync */
4d57b86d 3398 f2fs_set_dirty_device(sbi, fio->ino, devidx, FLUSH_INO);
1228b482
CY		 3399
3400 /* update device state for checkpoint */
3401 if (!f2fs_test_bit(devidx, (char *)&sbi->dirty_device)) {
3402 spin_lock(&sbi->dev_lock);
3403 f2fs_set_bit(devidx, (char *)&sbi->dirty_device);
3404 spin_unlock(&sbi->dev_lock);
3405 }
39d787be
CY		 3406}
3407
05ca3632 3408static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
bfad7c2d 3409{
81377bd6 3410 int type = __get_segment_type(fio);
b0332a0f 3411 bool keep_order = (f2fs_lfs_mode(fio->sbi) && type == CURSEG_COLD_DATA);
bfad7c2d 3412
107a805d
CY		 3413 if (keep_order)
3414 down_read(&fio->sbi->io_order_lock);
0a595eba 3415reallocate:
4d57b86d 3416 f2fs_allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
093749e2 3417 &fio->new_blkaddr, sum, type, fio);
6aa58d8a
CY		 3418 if (GET_SEGNO(fio->sbi, fio->old_blkaddr) != NULL_SEGNO)
3419 invalidate_mapping_pages(META_MAPPING(fio->sbi),
3420 fio->old_blkaddr, fio->old_blkaddr);
bfad7c2d 3421
351df4b2 3422 /* writeout dirty page into bdev */
fe16efe6
CY		 3423 f2fs_submit_page_write(fio);
3424 if (fio->retry) {
0a595eba
JK		 3425 fio->old_blkaddr = fio->new_blkaddr;
3426 goto reallocate;
3427 }
fe16efe6
CY		 3428
3429 update_device_state(fio);
3430
107a805d
CY		 3431 if (keep_order)
3432 up_read(&fio->sbi->io_order_lock);
351df4b2
JK		 3433}
3434
4d57b86d 3435void f2fs_do_write_meta_page(struct f2fs_sb_info *sbi, struct page *page,
b0af6d49 3436 enum iostat_type io_type)
351df4b2 3437{
458e6197 3438 struct f2fs_io_info fio = {
05ca3632 3439 .sbi = sbi,
458e6197 3440 .type = META,
0cdd3195 3441 .temp = HOT,
04d328de 3442 .op = REQ_OP_WRITE,
70fd7614 3443 .op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
7a9d7548
CY		 3444 .old_blkaddr = page->index,
3445 .new_blkaddr = page->index,
05ca3632 3446 .page = page,
4375a336 3447 .encrypted_page = NULL,
fb830fc5 3448 .in_list = false,
458e6197
JK		 3449 };
3450
2b947003 3451 if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
04d328de 3452 fio.op_flags &= ~REQ_META;
2b947003 3453
351df4b2 3454 set_page_writeback(page);
17c50035 3455 ClearPageError(page);
b9109b0e 3456 f2fs_submit_page_write(&fio);
b0af6d49 3457
b63e7be5 3458 stat_inc_meta_count(sbi, page->index);
b0af6d49 3459 f2fs_update_iostat(sbi, io_type, F2FS_BLKSIZE);
351df4b2
JK		 3460}
3461
4d57b86d 3462void f2fs_do_write_node_page(unsigned int nid, struct f2fs_io_info *fio)
351df4b2
JK		 3463{
3464 struct f2fs_summary sum;
05ca3632 3465
351df4b2 3466 set_summary(&sum, nid, 0, 0);
05ca3632 3467 do_write_page(&sum, fio);
b0af6d49
CY		 3468
3469 f2fs_update_iostat(fio->sbi, fio->io_type, F2FS_BLKSIZE);
351df4b2
JK		 3470}
3471
4d57b86d
CY		 3472void f2fs_outplace_write_data(struct dnode_of_data *dn,
3473 struct f2fs_io_info *fio)
351df4b2 3474{
05ca3632 3475 struct f2fs_sb_info *sbi = fio->sbi;
351df4b2 3476 struct f2fs_summary sum;
351df4b2 3477
9850cf4a 3478 f2fs_bug_on(sbi, dn->data_blkaddr == NULL_ADDR);
7735730d 3479 set_summary(&sum, dn->nid, dn->ofs_in_node, fio->version);
05ca3632 3480 do_write_page(&sum, fio);
f28b3434 3481 f2fs_update_data_blkaddr(dn, fio->new_blkaddr);
b0af6d49
CY		 3482
3483 f2fs_update_iostat(sbi, fio->io_type, F2FS_BLKSIZE);
351df4b2
JK		 3484}
3485
4d57b86d 3486int f2fs_inplace_write_data(struct f2fs_io_info *fio)
351df4b2 3487{
b0af6d49 3488 int err;
d21b0f23 3489 struct f2fs_sb_info *sbi = fio->sbi;
05573d6c 3490 unsigned int segno;
b0af6d49 3491
7a9d7548 3492 fio->new_blkaddr = fio->old_blkaddr;
0cdd3195
HL		 3493 /* i/o temperature is needed for passing down write hints */
3494 __get_segment_type(fio);
d21b0f23 3495
05573d6c
CY		 3496 segno = GET_SEGNO(sbi, fio->new_blkaddr);
3497
3498 if (!IS_DATASEG(get_seg_entry(sbi, segno)->type)) {
3499 set_sbi_flag(sbi, SBI_NEED_FSCK);
2d821c12
CY		 3500 f2fs_warn(sbi, "%s: incorrect segment(%u) type, run fsck to fix.",
3501 __func__, segno);
10f966bb 3502 return -EFSCORRUPTED;
05573d6c 3503 }
d21b0f23 3504
05ca3632 3505 stat_inc_inplace_blocks(fio->sbi);
b0af6d49 3506
0e7f4197 3507 if (fio->bio && !(SM_I(sbi)->ipu_policy & (1 << F2FS_IPU_NOCACHE)))
8648de2c
CY		 3508 err = f2fs_merge_page_bio(fio);
3509 else
3510 err = f2fs_submit_page_bio(fio);
e46f6bd8 3511 if (!err) {
39d787be 3512 update_device_state(fio);
e46f6bd8
CY		 3513 f2fs_update_iostat(fio->sbi, fio->io_type, F2FS_BLKSIZE);
3514 }
b0af6d49
CY		 3515
3516 return err;
351df4b2
JK		 3517}
3518
2b60311d
CY		 3519static inline int __f2fs_get_curseg(struct f2fs_sb_info *sbi,
3520 unsigned int segno)
3521{
3522 int i;
3523
3524 for (i = CURSEG_HOT_DATA; i < NO_CHECK_TYPE; i++) {
3525 if (CURSEG_I(sbi, i)->segno == segno)
3526 break;
3527 }
3528 return i;
3529}
3530
4d57b86d 3531void f2fs_do_replace_block(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
19f106bc 3532 block_t old_blkaddr, block_t new_blkaddr,
c5d02785
CY		 3533 bool recover_curseg, bool recover_newaddr,
3534 bool from_gc)
351df4b2
JK		 3535{
3536 struct sit_info *sit_i = SIT_I(sbi);
3537 struct curseg_info *curseg;
3538 unsigned int segno, old_cursegno;
3539 struct seg_entry *se;
3540 int type;
19f106bc 3541 unsigned short old_blkoff;
351df4b2
JK		 3542
3543 segno = GET_SEGNO(sbi, new_blkaddr);
3544 se = get_seg_entry(sbi, segno);
3545 type = se->type;
3546
2b60311d
CY		 3547 down_write(&SM_I(sbi)->curseg_lock);
3548
19f106bc
CY		 3549 if (!recover_curseg) {
3550 /* for recovery flow */
3551 if (se->valid_blocks == 0 && !IS_CURSEG(sbi, segno)) {
3552 if (old_blkaddr == NULL_ADDR)
3553 type = CURSEG_COLD_DATA;
3554 else
3555 type = CURSEG_WARM_DATA;
3556 }
3557 } else {
2b60311d
CY		 3558 if (IS_CURSEG(sbi, segno)) {
3559 /* se->type is volatile as SSR allocation */
3560 type = __f2fs_get_curseg(sbi, segno);
3561 f2fs_bug_on(sbi, type == NO_CHECK_TYPE);
3562 } else {
351df4b2 3563 type = CURSEG_WARM_DATA;
2b60311d 3564 }
351df4b2 3565 }
19f106bc 3566
2c190504 3567 f2fs_bug_on(sbi, !IS_DATASEG(type));
351df4b2
JK		 3568 curseg = CURSEG_I(sbi, type);
3569
3570 mutex_lock(&curseg->curseg_mutex);
3d26fa6b 3571 down_write(&sit_i->sentry_lock);
351df4b2
JK		 3572
3573 old_cursegno = curseg->segno;
19f106bc 3574 old_blkoff = curseg->next_blkoff;
351df4b2
JK		 3575
3576 /* change the current segment */
3577 if (segno != curseg->segno) {
3578 curseg->next_segno = segno;
093749e2 3579 change_curseg(sbi, type, true);
351df4b2
JK		 3580 }
3581
491c0854 3582 curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
e79efe3b 3583 __add_sum_entry(sbi, type, sum);
351df4b2 3584
c5d02785
CY		 3585 if (!recover_curseg || recover_newaddr) {
3586 if (!from_gc)
3587 update_segment_mtime(sbi, new_blkaddr, 0);
6e2c64ad 3588 update_sit_entry(sbi, new_blkaddr, 1);
c5d02785 3589 }
6aa58d8a
CY		 3590 if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO) {
3591 invalidate_mapping_pages(META_MAPPING(sbi),
3592 old_blkaddr, old_blkaddr);
c5d02785
CY		 3593 if (!from_gc)
3594 update_segment_mtime(sbi, old_blkaddr, 0);
6e2c64ad 3595 update_sit_entry(sbi, old_blkaddr, -1);
6aa58d8a 3596 }
6e2c64ad
JK		 3597
3598 locate_dirty_segment(sbi, GET_SEGNO(sbi, old_blkaddr));
3599 locate_dirty_segment(sbi, GET_SEGNO(sbi, new_blkaddr));
3600
351df4b2 3601 locate_dirty_segment(sbi, old_cursegno);
351df4b2 3602
19f106bc
CY		 3603 if (recover_curseg) {
3604 if (old_cursegno != curseg->segno) {
3605 curseg->next_segno = old_cursegno;
093749e2 3606 change_curseg(sbi, type, true);
19f106bc
CY		 3607 }
3608 curseg->next_blkoff = old_blkoff;
3609 }
3610
3d26fa6b 3611 up_write(&sit_i->sentry_lock);
351df4b2 3612 mutex_unlock(&curseg->curseg_mutex);
2b60311d 3613 up_write(&SM_I(sbi)->curseg_lock);
351df4b2
JK		 3614}
3615
528e3459
CY		 3616void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
3617 block_t old_addr, block_t new_addr,
28bc106b
CY		 3618 unsigned char version, bool recover_curseg,
3619 bool recover_newaddr)
528e3459
CY		 3620{
3621 struct f2fs_summary sum;
3622
3623 set_summary(&sum, dn->nid, dn->ofs_in_node, version);
3624
4d57b86d 3625 f2fs_do_replace_block(sbi, &sum, old_addr, new_addr,
c5d02785 3626 recover_curseg, recover_newaddr, false);
528e3459 3627
f28b3434 3628 f2fs_update_data_blkaddr(dn, new_addr);
528e3459
CY		 3629}
3630
93dfe2ac 3631void f2fs_wait_on_page_writeback(struct page *page,
bae0ee7a 3632 enum page_type type, bool ordered, bool locked)
93dfe2ac 3633{
93dfe2ac 3634 if (PageWriteback(page)) {
4081363f
JK		 3635 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
3636
0b20fcec 3637 /* submit cached LFS IO */
bab475c5 3638 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, type);
0b20fcec
CY		 3639 /* sbumit cached IPU IO */
3640 f2fs_submit_merged_ipu_write(sbi, NULL, page);
bae0ee7a 3641 if (ordered) {
fec1d657 3642 wait_on_page_writeback(page);
bae0ee7a
CY		 3643 f2fs_bug_on(sbi, locked && PageWriteback(page));
3644 } else {
fec1d657 3645 wait_for_stable_page(page);
bae0ee7a 3646 }
93dfe2ac
JK		 3647 }
3648}
3649
0ded69f6 3650void f2fs_wait_on_block_writeback(struct inode *inode, block_t blkaddr)
08b39fbd 3651{
0ded69f6 3652 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
08b39fbd
CY		 3653 struct page *cpage;
3654
0ded69f6
JK		 3655 if (!f2fs_post_read_required(inode))
3656 return;
3657
93770ab7 3658 if (!__is_valid_data_blkaddr(blkaddr))
08b39fbd
CY		 3659 return;
3660
08b39fbd
CY		 3661 cpage = find_lock_page(META_MAPPING(sbi), blkaddr);
3662 if (cpage) {
bae0ee7a 3663 f2fs_wait_on_page_writeback(cpage, DATA, true, true);
08b39fbd
CY		 3664 f2fs_put_page(cpage, 1);
3665 }
3666}
3667
1e78e8bd
ST		 3668void f2fs_wait_on_block_writeback_range(struct inode *inode, block_t blkaddr,
3669 block_t len)
3670{
3671 block_t i;
3672
3673 for (i = 0; i < len; i++)
3674 f2fs_wait_on_block_writeback(inode, blkaddr + i);
3675}
3676
7735730d 3677static int read_compacted_summaries(struct f2fs_sb_info *sbi)
351df4b2
JK		 3678{
3679 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3680 struct curseg_info *seg_i;
3681 unsigned char *kaddr;
3682 struct page *page;
3683 block_t start;
3684 int i, j, offset;
3685
3686 start = start_sum_block(sbi);
3687
4d57b86d 3688 page = f2fs_get_meta_page(sbi, start++);
7735730d
CY		 3689 if (IS_ERR(page))
3690 return PTR_ERR(page);
351df4b2
JK		 3691 kaddr = (unsigned char *)page_address(page);
3692
3693 /* Step 1: restore nat cache */
3694 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 3695 memcpy(seg_i->journal, kaddr, SUM_JOURNAL_SIZE);
351df4b2
JK		 3696
3697 /* Step 2: restore sit cache */
3698 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 3699 memcpy(seg_i->journal, kaddr + SUM_JOURNAL_SIZE, SUM_JOURNAL_SIZE);
351df4b2
JK		 3700 offset = 2 * SUM_JOURNAL_SIZE;
3701
3702 /* Step 3: restore summary entries */
3703 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
3704 unsigned short blk_off;
3705 unsigned int segno;
3706
3707 seg_i = CURSEG_I(sbi, i);
3708 segno = le32_to_cpu(ckpt->cur_data_segno[i]);
3709 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[i]);
3710 seg_i->next_segno = segno;
3711 reset_curseg(sbi, i, 0);
3712 seg_i->alloc_type = ckpt->alloc_type[i];
3713 seg_i->next_blkoff = blk_off;
3714
3715 if (seg_i->alloc_type == SSR)
3716 blk_off = sbi->blocks_per_seg;
3717
3718 for (j = 0; j < blk_off; j++) {
3719 struct f2fs_summary *s;
3720 s = (struct f2fs_summary *)(kaddr + offset);
3721 seg_i->sum_blk->entries[j] = *s;
3722 offset += SUMMARY_SIZE;
09cbfeaf 3723 if (offset + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 3724 SUM_FOOTER_SIZE)
3725 continue;
3726
3727 f2fs_put_page(page, 1);
3728 page = NULL;
3729
4d57b86d 3730 page = f2fs_get_meta_page(sbi, start++);
7735730d
CY		 3731 if (IS_ERR(page))
3732 return PTR_ERR(page);
351df4b2
JK		 3733 kaddr = (unsigned char *)page_address(page);
3734 offset = 0;
3735 }
3736 }
3737 f2fs_put_page(page, 1);
7735730d 3738 return 0;
351df4b2
JK		 3739}
3740
3741static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
3742{
3743 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3744 struct f2fs_summary_block *sum;
3745 struct curseg_info *curseg;
3746 struct page *new;
3747 unsigned short blk_off;
3748 unsigned int segno = 0;
3749 block_t blk_addr = 0;
7735730d 3750 int err = 0;
351df4b2
JK		 3751
3752 /* get segment number and block addr */
3753 if (IS_DATASEG(type)) {
3754 segno = le32_to_cpu(ckpt->cur_data_segno[type]);
3755 blk_off = le16_to_cpu(ckpt->cur_data_blkoff[type -
3756 CURSEG_HOT_DATA]);
119ee914 3757 if (__exist_node_summaries(sbi))
d0b9e42a 3758 blk_addr = sum_blk_addr(sbi, NR_CURSEG_PERSIST_TYPE, type);
351df4b2
JK		 3759 else
3760 blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
3761 } else {
3762 segno = le32_to_cpu(ckpt->cur_node_segno[type -
3763 CURSEG_HOT_NODE]);
3764 blk_off = le16_to_cpu(ckpt->cur_node_blkoff[type -
3765 CURSEG_HOT_NODE]);
119ee914 3766 if (__exist_node_summaries(sbi))
351df4b2
JK		 3767 blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
3768 type - CURSEG_HOT_NODE);
3769 else
3770 blk_addr = GET_SUM_BLOCK(sbi, segno);
3771 }
3772
4d57b86d 3773 new = f2fs_get_meta_page(sbi, blk_addr);
7735730d
CY		 3774 if (IS_ERR(new))
3775 return PTR_ERR(new);
351df4b2
JK		 3776 sum = (struct f2fs_summary_block *)page_address(new);
3777
3778 if (IS_NODESEG(type)) {
119ee914 3779 if (__exist_node_summaries(sbi)) {
351df4b2
JK		 3780 struct f2fs_summary *ns = &sum->entries[0];
3781 int i;
3782 for (i = 0; i < sbi->blocks_per_seg; i++, ns++) {
3783 ns->version = 0;
3784 ns->ofs_in_node = 0;
3785 }
3786 } else {
7735730d
CY		 3787 err = f2fs_restore_node_summary(sbi, segno, sum);
3788 if (err)
3789 goto out;
351df4b2
JK		 3790 }
3791 }
3792
3793 /* set uncompleted segment to curseg */
3794 curseg = CURSEG_I(sbi, type);
3795 mutex_lock(&curseg->curseg_mutex);
b7ad7512
CY		 3796
3797 /* update journal info */
3798 down_write(&curseg->journal_rwsem);
3799 memcpy(curseg->journal, &sum->journal, SUM_JOURNAL_SIZE);
3800 up_write(&curseg->journal_rwsem);
3801
3802 memcpy(curseg->sum_blk->entries, sum->entries, SUM_ENTRY_SIZE);
3803 memcpy(&curseg->sum_blk->footer, &sum->footer, SUM_FOOTER_SIZE);
351df4b2
JK		 3804 curseg->next_segno = segno;
3805 reset_curseg(sbi, type, 0);
3806 curseg->alloc_type = ckpt->alloc_type[type];
3807 curseg->next_blkoff = blk_off;
3808 mutex_unlock(&curseg->curseg_mutex);
7735730d 3809out:
351df4b2 3810 f2fs_put_page(new, 1);
7735730d 3811 return err;
351df4b2
JK		 3812}
3813
3814static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
3815{
21d3f8e1
JQ		 3816 struct f2fs_journal *sit_j = CURSEG_I(sbi, CURSEG_COLD_DATA)->journal;
3817 struct f2fs_journal *nat_j = CURSEG_I(sbi, CURSEG_HOT_DATA)->journal;
351df4b2 3818 int type = CURSEG_HOT_DATA;
e4fc5fbf 3819 int err;
351df4b2 3820
aaec2b1d 3821 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG)) {
4d57b86d 3822 int npages = f2fs_npages_for_summary_flush(sbi, true);
3fa06d7b
CY		 3823
3824 if (npages >= 2)
4d57b86d 3825 f2fs_ra_meta_pages(sbi, start_sum_block(sbi), npages,
26879fb1 3826 META_CP, true);
3fa06d7b 3827
351df4b2 3828 /* restore for compacted data summary */
7735730d
CY		 3829 err = read_compacted_summaries(sbi);
3830 if (err)
3831 return err;
351df4b2
JK		 3832 type = CURSEG_HOT_NODE;
3833 }
3834
119ee914 3835 if (__exist_node_summaries(sbi))
d0b9e42a
CY		 3836 f2fs_ra_meta_pages(sbi,
3837 sum_blk_addr(sbi, NR_CURSEG_PERSIST_TYPE, type),
3838 NR_CURSEG_PERSIST_TYPE - type, META_CP, true);
3fa06d7b 3839
e4fc5fbf
CY		 3840 for (; type <= CURSEG_COLD_NODE; type++) {
3841 err = read_normal_summaries(sbi, type);
3842 if (err)
3843 return err;
3844 }
3845
21d3f8e1
JQ		 3846 /* sanity check for summary blocks */
3847 if (nats_in_cursum(nat_j) > NAT_JOURNAL_ENTRIES ||
9227d522 3848 sits_in_cursum(sit_j) > SIT_JOURNAL_ENTRIES) {
dcbb4c10
JP		 3849 f2fs_err(sbi, "invalid journal entries nats %u sits %u\n",
3850 nats_in_cursum(nat_j), sits_in_cursum(sit_j));
21d3f8e1 3851 return -EINVAL;
9227d522 3852 }
21d3f8e1 3853
351df4b2
JK		 3854 return 0;
3855}
3856
3857static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
3858{
3859 struct page *page;
3860 unsigned char *kaddr;
3861 struct f2fs_summary *summary;
3862 struct curseg_info *seg_i;
3863 int written_size = 0;
3864 int i, j;
3865
4d57b86d 3866 page = f2fs_grab_meta_page(sbi, blkaddr++);
351df4b2 3867 kaddr = (unsigned char *)page_address(page);
81114baa 3868 memset(kaddr, 0, PAGE_SIZE);
351df4b2
JK		 3869
3870 /* Step 1: write nat cache */
3871 seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 3872 memcpy(kaddr, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 3873 written_size += SUM_JOURNAL_SIZE;
3874
3875 /* Step 2: write sit cache */
3876 seg_i = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 3877 memcpy(kaddr + written_size, seg_i->journal, SUM_JOURNAL_SIZE);
351df4b2
JK		 3878 written_size += SUM_JOURNAL_SIZE;
3879
351df4b2
JK		 3880 /* Step 3: write summary entries */
3881 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
3882 unsigned short blkoff;
3883 seg_i = CURSEG_I(sbi, i);
3884 if (sbi->ckpt->alloc_type[i] == SSR)
3885 blkoff = sbi->blocks_per_seg;
3886 else
3887 blkoff = curseg_blkoff(sbi, i);
3888
3889 for (j = 0; j < blkoff; j++) {
3890 if (!page) {
4d57b86d 3891 page = f2fs_grab_meta_page(sbi, blkaddr++);
351df4b2 3892 kaddr = (unsigned char *)page_address(page);
81114baa 3893 memset(kaddr, 0, PAGE_SIZE);
351df4b2
JK		 3894 written_size = 0;
3895 }
3896 summary = (struct f2fs_summary *)(kaddr + written_size);
3897 *summary = seg_i->sum_blk->entries[j];
3898 written_size += SUMMARY_SIZE;
351df4b2 3899
09cbfeaf 3900 if (written_size + SUMMARY_SIZE <= PAGE_SIZE -
351df4b2
JK		 3901 SUM_FOOTER_SIZE)
3902 continue;
3903
e8d61a74 3904 set_page_dirty(page);
351df4b2
JK		 3905 f2fs_put_page(page, 1);
3906 page = NULL;
3907 }
3908 }
e8d61a74
CY		 3909 if (page) {
3910 set_page_dirty(page);
351df4b2 3911 f2fs_put_page(page, 1);
e8d61a74 3912 }
351df4b2
JK		 3913}
3914
3915static void write_normal_summaries(struct f2fs_sb_info *sbi,
3916 block_t blkaddr, int type)
3917{
3918 int i, end;
3919 if (IS_DATASEG(type))
3920 end = type + NR_CURSEG_DATA_TYPE;
3921 else
3922 end = type + NR_CURSEG_NODE_TYPE;
3923
b7ad7512
CY		 3924 for (i = type; i < end; i++)
3925 write_current_sum_page(sbi, i, blkaddr + (i - type));
351df4b2
JK		 3926}
3927
4d57b86d 3928void f2fs_write_data_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
351df4b2 3929{
aaec2b1d 3930 if (is_set_ckpt_flags(sbi, CP_COMPACT_SUM_FLAG))
351df4b2
JK		 3931 write_compacted_summaries(sbi, start_blk);
3932 else
3933 write_normal_summaries(sbi, start_blk, CURSEG_HOT_DATA);
3934}
3935
4d57b86d 3936void f2fs_write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk)
351df4b2 3937{
119ee914 3938 write_normal_summaries(sbi, start_blk, CURSEG_HOT_NODE);
351df4b2
JK		 3939}
3940
4d57b86d 3941int f2fs_lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
351df4b2
JK		 3942 unsigned int val, int alloc)
3943{
3944 int i;
3945
3946 if (type == NAT_JOURNAL) {
dfc08a12
CY		 3947 for (i = 0; i < nats_in_cursum(journal); i++) {
3948 if (le32_to_cpu(nid_in_journal(journal, i)) == val)
351df4b2
JK		 3949 return i;
3950 }
dfc08a12
CY		 3951 if (alloc && __has_cursum_space(journal, 1, NAT_JOURNAL))
3952 return update_nats_in_cursum(journal, 1);
351df4b2 3953 } else if (type == SIT_JOURNAL) {
dfc08a12
CY		 3954 for (i = 0; i < sits_in_cursum(journal); i++)
3955 if (le32_to_cpu(segno_in_journal(journal, i)) == val)
351df4b2 3956 return i;
dfc08a12
CY		 3957 if (alloc && __has_cursum_space(journal, 1, SIT_JOURNAL))
3958 return update_sits_in_cursum(journal, 1);
351df4b2
JK		 3959 }
3960 return -1;
3961}
3962
3963static struct page *get_current_sit_page(struct f2fs_sb_info *sbi,
3964 unsigned int segno)
3965{
86f33603 3966 return f2fs_get_meta_page(sbi, current_sit_addr(sbi, segno));
351df4b2
JK		 3967}
3968
3969static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
3970 unsigned int start)
3971{
3972 struct sit_info *sit_i = SIT_I(sbi);
068c3cd8 3973 struct page *page;
351df4b2 3974 pgoff_t src_off, dst_off;
351df4b2
JK		 3975
3976 src_off = current_sit_addr(sbi, start);
3977 dst_off = next_sit_addr(sbi, src_off);
3978
4d57b86d 3979 page = f2fs_grab_meta_page(sbi, dst_off);
068c3cd8 3980 seg_info_to_sit_page(sbi, page, start);
351df4b2 3981
068c3cd8 3982 set_page_dirty(page);
351df4b2
JK		 3983 set_to_next_sit(sit_i, start);
3984
068c3cd8 3985 return page;
351df4b2
JK		 3986}
3987
184a5cd2
CY		 3988static struct sit_entry_set *grab_sit_entry_set(void)
3989{
3990 struct sit_entry_set *ses =
80c54505 3991 f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS);
184a5cd2
CY		 3992
3993 ses->entry_cnt = 0;
3994 INIT_LIST_HEAD(&ses->set_list);
3995 return ses;
3996}
3997
3998static void release_sit_entry_set(struct sit_entry_set *ses)
3999{
4000 list_del(&ses->set_list);
4001 kmem_cache_free(sit_entry_set_slab, ses);
4002}
4003
4004static void adjust_sit_entry_set(struct sit_entry_set *ses,
4005 struct list_head *head)
4006{
4007 struct sit_entry_set *next = ses;
4008
4009 if (list_is_last(&ses->set_list, head))
4010 return;
4011
4012 list_for_each_entry_continue(next, head, set_list)
4013 if (ses->entry_cnt <= next->entry_cnt)
4014 break;
4015
4016 list_move_tail(&ses->set_list, &next->set_list);
4017}
4018
4019static void add_sit_entry(unsigned int segno, struct list_head *head)
4020{
4021 struct sit_entry_set *ses;
4022 unsigned int start_segno = START_SEGNO(segno);
4023
4024 list_for_each_entry(ses, head, set_list) {
4025 if (ses->start_segno == start_segno) {
4026 ses->entry_cnt++;
4027 adjust_sit_entry_set(ses, head);
4028 return;
4029 }
4030 }
4031
4032 ses = grab_sit_entry_set();
4033
4034 ses->start_segno = start_segno;
4035 ses->entry_cnt++;
4036 list_add(&ses->set_list, head);
4037}
4038
4039static void add_sits_in_set(struct f2fs_sb_info *sbi)
4040{
4041 struct f2fs_sm_info *sm_info = SM_I(sbi);
4042 struct list_head *set_list = &sm_info->sit_entry_set;
4043 unsigned long *bitmap = SIT_I(sbi)->dirty_sentries_bitmap;
184a5cd2
CY		 4044 unsigned int segno;
4045
7cd8558b 4046 for_each_set_bit(segno, bitmap, MAIN_SEGS(sbi))
184a5cd2
CY		 4047 add_sit_entry(segno, set_list);
4048}
4049
4050static void remove_sits_in_journal(struct f2fs_sb_info *sbi)
351df4b2
JK		 4051{
4052 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 4053 struct f2fs_journal *journal = curseg->journal;
351df4b2
JK		 4054 int i;
4055
b7ad7512 4056 down_write(&curseg->journal_rwsem);
dfc08a12 4057 for (i = 0; i < sits_in_cursum(journal); i++) {
184a5cd2
CY		 4058 unsigned int segno;
4059 bool dirtied;
4060
dfc08a12 4061 segno = le32_to_cpu(segno_in_journal(journal, i));
184a5cd2
CY		 4062 dirtied = __mark_sit_entry_dirty(sbi, segno);
4063
4064 if (!dirtied)
4065 add_sit_entry(segno, &SM_I(sbi)->sit_entry_set);
351df4b2 4066 }
dfc08a12 4067 update_sits_in_cursum(journal, -i);
b7ad7512 4068 up_write(&curseg->journal_rwsem);
351df4b2
JK		 4069}
4070
0a8165d7 4071/*
351df4b2
JK		 4072 * CP calls this function, which flushes SIT entries including sit_journal,
4073 * and moves prefree segs to free segs.
4074 */
4d57b86d 4075void f2fs_flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
351df4b2
JK		 4076{
4077 struct sit_info *sit_i = SIT_I(sbi);
4078 unsigned long *bitmap = sit_i->dirty_sentries_bitmap;
4079 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 4080 struct f2fs_journal *journal = curseg->journal;
184a5cd2
CY		 4081 struct sit_entry_set *ses, *tmp;
4082 struct list_head *head = &SM_I(sbi)->sit_entry_set;
04f0b2ea 4083 bool to_journal = !is_sbi_flag_set(sbi, SBI_IS_RESIZEFS);
4b2fecc8 4084 struct seg_entry *se;
351df4b2 4085
3d26fa6b 4086 down_write(&sit_i->sentry_lock);
351df4b2 4087
2b11a74b
WL		 4088 if (!sit_i->dirty_sentries)
4089 goto out;
4090
351df4b2 4091 /*
184a5cd2
CY		 4092 * add and account sit entries of dirty bitmap in sit entry
4093 * set temporarily
351df4b2 4094 */
184a5cd2 4095 add_sits_in_set(sbi);
351df4b2 4096
184a5cd2
CY		 4097 /*
4098 * if there are no enough space in journal to store dirty sit
4099 * entries, remove all entries from journal and add and account
4100 * them in sit entry set.
4101 */
04f0b2ea
QS		 4102 if (!__has_cursum_space(journal, sit_i->dirty_sentries, SIT_JOURNAL) ||
4103 !to_journal)
184a5cd2 4104 remove_sits_in_journal(sbi);
b2955550 4105
184a5cd2
CY		 4106 /*
4107 * there are two steps to flush sit entries:
4108 * #1, flush sit entries to journal in current cold data summary block.
4109 * #2, flush sit entries to sit page.
4110 */
4111 list_for_each_entry_safe(ses, tmp, head, set_list) {
4a257ed6 4112 struct page *page = NULL;
184a5cd2
CY		 4113 struct f2fs_sit_block *raw_sit = NULL;
4114 unsigned int start_segno = ses->start_segno;
4115 unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
7cd8558b 4116 (unsigned long)MAIN_SEGS(sbi));
184a5cd2
CY		 4117 unsigned int segno = start_segno;
4118
4119 if (to_journal &&
dfc08a12 4120 !__has_cursum_space(journal, ses->entry_cnt, SIT_JOURNAL))
184a5cd2
CY		 4121 to_journal = false;
4122
b7ad7512
CY		 4123 if (to_journal) {
4124 down_write(&curseg->journal_rwsem);
4125 } else {
184a5cd2
CY		 4126 page = get_next_sit_page(sbi, start_segno);
4127 raw_sit = page_address(page);
351df4b2 4128 }
351df4b2 4129
184a5cd2
CY		 4130 /* flush dirty sit entries in region of current sit set */
4131 for_each_set_bit_from(segno, bitmap, end) {
4132 int offset, sit_offset;
4b2fecc8
JK		 4133
4134 se = get_seg_entry(sbi, segno);
56b07e7e
ZZ		 4135#ifdef CONFIG_F2FS_CHECK_FS
4136 if (memcmp(se->cur_valid_map, se->cur_valid_map_mir,
4137 SIT_VBLOCK_MAP_SIZE))
4138 f2fs_bug_on(sbi, 1);
4139#endif
184a5cd2
CY		 4140
4141 /* add discard candidates */
c473f1a9 4142 if (!(cpc->reason & CP_DISCARD)) {
4b2fecc8 4143 cpc->trim_start = segno;
25290fa5 4144 add_discard_addrs(sbi, cpc, false);
4b2fecc8 4145 }
184a5cd2
CY		 4146
4147 if (to_journal) {
4d57b86d 4148 offset = f2fs_lookup_journal_in_cursum(journal,
184a5cd2
CY		 4149 SIT_JOURNAL, segno, 1);
4150 f2fs_bug_on(sbi, offset < 0);
dfc08a12 4151 segno_in_journal(journal, offset) =
184a5cd2
CY		 4152 cpu_to_le32(segno);
4153 seg_info_to_raw_sit(se,
dfc08a12 4154 &sit_in_journal(journal, offset));
56b07e7e
ZZ		 4155 check_block_count(sbi, segno,
4156 &sit_in_journal(journal, offset));
184a5cd2
CY		 4157 } else {
4158 sit_offset = SIT_ENTRY_OFFSET(sit_i, segno);
4159 seg_info_to_raw_sit(se,
4160 &raw_sit->entries[sit_offset]);
56b07e7e
ZZ		 4161 check_block_count(sbi, segno,
4162 &raw_sit->entries[sit_offset]);
184a5cd2 4163 }
351df4b2 4164
184a5cd2
CY		 4165 __clear_bit(segno, bitmap);
4166 sit_i->dirty_sentries--;
4167 ses->entry_cnt--;
351df4b2
JK		 4168 }
4169
b7ad7512
CY		 4170 if (to_journal)
4171 up_write(&curseg->journal_rwsem);
4172 else
184a5cd2
CY		 4173 f2fs_put_page(page, 1);
4174
4175 f2fs_bug_on(sbi, ses->entry_cnt);
4176 release_sit_entry_set(ses);
351df4b2 4177 }
184a5cd2
CY		 4178
4179 f2fs_bug_on(sbi, !list_empty(head));
4180 f2fs_bug_on(sbi, sit_i->dirty_sentries);
184a5cd2 4181out:
c473f1a9 4182 if (cpc->reason & CP_DISCARD) {
650d3c4e
YH		 4183 __u64 trim_start = cpc->trim_start;
4184
4b2fecc8 4185 for (; cpc->trim_start <= cpc->trim_end; cpc->trim_start++)
25290fa5 4186 add_discard_addrs(sbi, cpc, false);
650d3c4e
YH		 4187
4188 cpc->trim_start = trim_start;
4b2fecc8 4189 }
3d26fa6b 4190 up_write(&sit_i->sentry_lock);
351df4b2 4191
351df4b2
JK		 4192 set_prefree_as_free_segments(sbi);
4193}
4194
4195static int build_sit_info(struct f2fs_sb_info *sbi)
4196{
4197 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
351df4b2
JK		 4198 struct sit_info *sit_i;
4199 unsigned int sit_segs, start;
2fde3dd1 4200 char *src_bitmap, *bitmap;
bbf9f7d9 4201 unsigned int bitmap_size, main_bitmap_size, sit_bitmap_size;
351df4b2
JK		 4202
4203 /* allocate memory for SIT information */
acbf054d 4204 sit_i = f2fs_kzalloc(sbi, sizeof(struct sit_info), GFP_KERNEL);
351df4b2
JK		 4205 if (!sit_i)
4206 return -ENOMEM;
4207
4208 SM_I(sbi)->sit_info = sit_i;
4209
9d2a789c
KC		 4210 sit_i->sentries =
4211 f2fs_kvzalloc(sbi, array_size(sizeof(struct seg_entry),
4212 MAIN_SEGS(sbi)),
4213 GFP_KERNEL);
351df4b2
JK		 4214 if (!sit_i->sentries)
4215 return -ENOMEM;
4216
bbf9f7d9
ST		 4217 main_bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
4218 sit_i->dirty_sentries_bitmap = f2fs_kvzalloc(sbi, main_bitmap_size,
628b3d14 4219 GFP_KERNEL);
351df4b2
JK		 4220 if (!sit_i->dirty_sentries_bitmap)
4221 return -ENOMEM;
4222
2fde3dd1
CY		 4223#ifdef CONFIG_F2FS_CHECK_FS
4224 bitmap_size = MAIN_SEGS(sbi) * SIT_VBLOCK_MAP_SIZE * 4;
4225#else
4226 bitmap_size = MAIN_SEGS(sbi) * SIT_VBLOCK_MAP_SIZE * 3;
4227#endif
4228 sit_i->bitmap = f2fs_kvzalloc(sbi, bitmap_size, GFP_KERNEL);
4229 if (!sit_i->bitmap)
4230 return -ENOMEM;
4231
4232 bitmap = sit_i->bitmap;
4233
7cd8558b 4234 for (start = 0; start < MAIN_SEGS(sbi); start++) {
2fde3dd1
CY		 4235 sit_i->sentries[start].cur_valid_map = bitmap;
4236 bitmap += SIT_VBLOCK_MAP_SIZE;
4237
4238 sit_i->sentries[start].ckpt_valid_map = bitmap;
4239 bitmap += SIT_VBLOCK_MAP_SIZE;
3e025740 4240
355e7891 4241#ifdef CONFIG_F2FS_CHECK_FS
2fde3dd1
CY		 4242 sit_i->sentries[start].cur_valid_map_mir = bitmap;
4243 bitmap += SIT_VBLOCK_MAP_SIZE;
355e7891
CY		 4244#endif
4245
2fde3dd1
CY		 4246 sit_i->sentries[start].discard_map = bitmap;
4247 bitmap += SIT_VBLOCK_MAP_SIZE;
351df4b2
JK		 4248 }
4249
acbf054d 4250 sit_i->tmp_map = f2fs_kzalloc(sbi, SIT_VBLOCK_MAP_SIZE, GFP_KERNEL);
60a3b782
JK		 4251 if (!sit_i->tmp_map)
4252 return -ENOMEM;
4253
2c70c5e3 4254 if (__is_large_section(sbi)) {
9d2a789c
KC		 4255 sit_i->sec_entries =
4256 f2fs_kvzalloc(sbi, array_size(sizeof(struct sec_entry),
4257 MAIN_SECS(sbi)),
4258 GFP_KERNEL);
351df4b2
JK		 4259 if (!sit_i->sec_entries)
4260 return -ENOMEM;
4261 }
4262
4263 /* get information related with SIT */
4264 sit_segs = le32_to_cpu(raw_super->segment_count_sit) >> 1;
4265
4266 /* setup SIT bitmap from ckeckpoint pack */
bbf9f7d9 4267 sit_bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
351df4b2
JK		 4268 src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
4269
bbf9f7d9 4270 sit_i->sit_bitmap = kmemdup(src_bitmap, sit_bitmap_size, GFP_KERNEL);
ae27d62e 4271 if (!sit_i->sit_bitmap)
351df4b2 4272 return -ENOMEM;
351df4b2 4273
ae27d62e 4274#ifdef CONFIG_F2FS_CHECK_FS
bbf9f7d9
ST		 4275 sit_i->sit_bitmap_mir = kmemdup(src_bitmap,
4276 sit_bitmap_size, GFP_KERNEL);
ae27d62e
CY		 4277 if (!sit_i->sit_bitmap_mir)
4278 return -ENOMEM;
bbf9f7d9
ST		 4279
4280 sit_i->invalid_segmap = f2fs_kvzalloc(sbi,
4281 main_bitmap_size, GFP_KERNEL);
4282 if (!sit_i->invalid_segmap)
4283 return -ENOMEM;
ae27d62e
CY		 4284#endif
4285
351df4b2
JK		 4286 /* init SIT information */
4287 sit_i->s_ops = &default_salloc_ops;
4288
4289 sit_i->sit_base_addr = le32_to_cpu(raw_super->sit_blkaddr);
4290 sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
c79b7ff1 4291 sit_i->written_valid_blocks = 0;
bbf9f7d9 4292 sit_i->bitmap_size = sit_bitmap_size;
351df4b2
JK		 4293 sit_i->dirty_sentries = 0;
4294 sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
4295 sit_i->elapsed_time = le64_to_cpu(sbi->ckpt->elapsed_time);
a7e679b5 4296 sit_i->mounted_time = ktime_get_boottime_seconds();
3d26fa6b 4297 init_rwsem(&sit_i->sentry_lock);
351df4b2
JK		 4298 return 0;
4299}
4300
4301static int build_free_segmap(struct f2fs_sb_info *sbi)
4302{
351df4b2
JK		 4303 struct free_segmap_info *free_i;
4304 unsigned int bitmap_size, sec_bitmap_size;
4305
4306 /* allocate memory for free segmap information */
acbf054d 4307 free_i = f2fs_kzalloc(sbi, sizeof(struct free_segmap_info), GFP_KERNEL);
351df4b2
JK		 4308 if (!free_i)
4309 return -ENOMEM;
4310
4311 SM_I(sbi)->free_info = free_i;
4312
7cd8558b 4313 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
628b3d14 4314 free_i->free_segmap = f2fs_kvmalloc(sbi, bitmap_size, GFP_KERNEL);
351df4b2
JK		 4315 if (!free_i->free_segmap)
4316 return -ENOMEM;
4317
7cd8558b 4318 sec_bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
628b3d14 4319 free_i->free_secmap = f2fs_kvmalloc(sbi, sec_bitmap_size, GFP_KERNEL);
351df4b2
JK		 4320 if (!free_i->free_secmap)
4321 return -ENOMEM;
4322
4323 /* set all segments as dirty temporarily */
4324 memset(free_i->free_segmap, 0xff, bitmap_size);
4325 memset(free_i->free_secmap, 0xff, sec_bitmap_size);
4326
4327 /* init free segmap information */
7cd8558b 4328 free_i->start_segno = GET_SEGNO_FROM_SEG0(sbi, MAIN_BLKADDR(sbi));
351df4b2
JK		 4329 free_i->free_segments = 0;
4330 free_i->free_sections = 0;
1a118ccf 4331 spin_lock_init(&free_i->segmap_lock);
351df4b2
JK		 4332 return 0;
4333}
4334
4335static int build_curseg(struct f2fs_sb_info *sbi)
4336{
1042d60f 4337 struct curseg_info *array;
351df4b2
JK		 4338 int i;
4339
d0b9e42a
CY		 4340 array = f2fs_kzalloc(sbi, array_size(NR_CURSEG_TYPE,
4341 sizeof(*array)), GFP_KERNEL);
351df4b2
JK		 4342 if (!array)
4343 return -ENOMEM;
4344
4345 SM_I(sbi)->curseg_array = array;
4346
d0b9e42a 4347 for (i = 0; i < NO_CHECK_TYPE; i++) {
351df4b2 4348 mutex_init(&array[i].curseg_mutex);
acbf054d 4349 array[i].sum_blk = f2fs_kzalloc(sbi, PAGE_SIZE, GFP_KERNEL);
351df4b2
JK		 4350 if (!array[i].sum_blk)
4351 return -ENOMEM;
b7ad7512 4352 init_rwsem(&array[i].journal_rwsem);
acbf054d
CY		 4353 array[i].journal = f2fs_kzalloc(sbi,
4354 sizeof(struct f2fs_journal), GFP_KERNEL);
b7ad7512
CY		 4355 if (!array[i].journal)
4356 return -ENOMEM;
d0b9e42a
CY		 4357 if (i < NR_PERSISTENT_LOG)
4358 array[i].seg_type = CURSEG_HOT_DATA + i;
4359 else if (i == CURSEG_COLD_DATA_PINNED)
4360 array[i].seg_type = CURSEG_COLD_DATA;
093749e2
CY		 4361 else if (i == CURSEG_ALL_DATA_ATGC)
4362 array[i].seg_type = CURSEG_COLD_DATA;
351df4b2
JK		 4363 array[i].segno = NULL_SEGNO;
4364 array[i].next_blkoff = 0;
d0b9e42a 4365 array[i].inited = false;
351df4b2
JK		 4366 }
4367 return restore_curseg_summaries(sbi);
4368}
4369
c39a1b34 4370static int build_sit_entries(struct f2fs_sb_info *sbi)
351df4b2
JK		 4371{
4372 struct sit_info *sit_i = SIT_I(sbi);
4373 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
b7ad7512 4374 struct f2fs_journal *journal = curseg->journal;
9c094040
YH		 4375 struct seg_entry *se;
4376 struct f2fs_sit_entry sit;
74de593a
CY		 4377 int sit_blk_cnt = SIT_BLK_CNT(sbi);
4378 unsigned int i, start, end;
4379 unsigned int readed, start_blk = 0;
c39a1b34 4380 int err = 0;
8a29c126 4381 block_t total_node_blocks = 0;
351df4b2 4382
74de593a 4383 do {
a8affc03 4384 readed = f2fs_ra_meta_pages(sbi, start_blk, BIO_MAX_VECS,
664ba972 4385 META_SIT, true);
74de593a
CY		 4386
4387 start = start_blk * sit_i->sents_per_block;
4388 end = (start_blk + readed) * sit_i->sents_per_block;
4389
7cd8558b 4390 for (; start < end && start < MAIN_SEGS(sbi); start++) {
74de593a 4391 struct f2fs_sit_block *sit_blk;
74de593a
CY		 4392 struct page *page;
4393
9c094040 4394 se = &sit_i->sentries[start];
74de593a 4395 page = get_current_sit_page(sbi, start);
edc55aaf
JK		 4396 if (IS_ERR(page))
4397 return PTR_ERR(page);
74de593a
CY		 4398 sit_blk = (struct f2fs_sit_block *)page_address(page);
4399 sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
4400 f2fs_put_page(page, 1);
d600af23 4401
c39a1b34
JK		 4402 err = check_block_count(sbi, start, &sit);
4403 if (err)
4404 return err;
74de593a 4405 seg_info_from_raw_sit(se, &sit);
8a29c126
JK		 4406 if (IS_NODESEG(se->type))
4407 total_node_blocks += se->valid_blocks;
a66cdd98
JK		 4408
4409 /* build discard map only one time */
7d20c8ab
CY		 4410 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
4411 memset(se->discard_map, 0xff,
4412 SIT_VBLOCK_MAP_SIZE);
4413 } else {
4414 memcpy(se->discard_map,
4415 se->cur_valid_map,
4416 SIT_VBLOCK_MAP_SIZE);
4417 sbi->discard_blks +=
4418 sbi->blocks_per_seg -
4419 se->valid_blocks;
3e025740 4420 }
a66cdd98 4421
2c70c5e3 4422 if (__is_large_section(sbi))
d600af23
CY		 4423 get_sec_entry(sbi, start)->valid_blocks +=
4424 se->valid_blocks;
351df4b2 4425 }
74de593a
CY		 4426 start_blk += readed;
4427 } while (start_blk < sit_blk_cnt);
d600af23
CY		 4428
4429 down_read(&curseg->journal_rwsem);
4430 for (i = 0; i < sits_in_cursum(journal); i++) {
d600af23
CY		 4431 unsigned int old_valid_blocks;
4432
4433 start = le32_to_cpu(segno_in_journal(journal, i));
b2ca374f 4434 if (start >= MAIN_SEGS(sbi)) {
dcbb4c10
JP		 4435 f2fs_err(sbi, "Wrong journal entry on segno %u",
4436 start);
10f966bb 4437 err = -EFSCORRUPTED;
b2ca374f
JK		 4438 break;
4439 }
4440
d600af23
CY		 4441 se = &sit_i->sentries[start];
4442 sit = sit_in_journal(journal, i);
4443
4444 old_valid_blocks = se->valid_blocks;
8a29c126
JK		 4445 if (IS_NODESEG(se->type))
4446 total_node_blocks -= old_valid_blocks;
d600af23 4447
c39a1b34
JK		 4448 err = check_block_count(sbi, start, &sit);
4449 if (err)
4450 break;
d600af23 4451 seg_info_from_raw_sit(se, &sit);
8a29c126
JK		 4452 if (IS_NODESEG(se->type))
4453 total_node_blocks += se->valid_blocks;
d600af23 4454
7d20c8ab
CY		 4455 if (is_set_ckpt_flags(sbi, CP_TRIMMED_FLAG)) {
4456 memset(se->discard_map, 0xff, SIT_VBLOCK_MAP_SIZE);
4457 } else {
4458 memcpy(se->discard_map, se->cur_valid_map,
4459 SIT_VBLOCK_MAP_SIZE);
4460 sbi->discard_blks += old_valid_blocks;
4461 sbi->discard_blks -= se->valid_blocks;
d600af23
CY		 4462 }
4463
2c70c5e3 4464 if (__is_large_section(sbi)) {
d600af23 4465 get_sec_entry(sbi, start)->valid_blocks +=
a9af3fdc
CY		 4466 se->valid_blocks;
4467 get_sec_entry(sbi, start)->valid_blocks -=
4468 old_valid_blocks;
4469 }
d600af23
CY		 4470 }
4471 up_read(&curseg->journal_rwsem);
8a29c126
JK		 4472
4473 if (!err && total_node_blocks != valid_node_count(sbi)) {
dcbb4c10
JP		 4474 f2fs_err(sbi, "SIT is corrupted node# %u vs %u",
4475 total_node_blocks, valid_node_count(sbi));
10f966bb 4476 err = -EFSCORRUPTED;
8a29c126
JK		 4477 }
4478
c39a1b34 4479 return err;
351df4b2
JK		 4480}
4481
4482static void init_free_segmap(struct f2fs_sb_info *sbi)
4483{
4484 unsigned int start;
4485 int type;
de881df9 4486 struct seg_entry *sentry;
351df4b2 4487
7cd8558b 4488 for (start = 0; start < MAIN_SEGS(sbi); start++) {
de881df9
AR		 4489 if (f2fs_usable_blks_in_seg(sbi, start) == 0)
4490 continue;
4491 sentry = get_seg_entry(sbi, start);
351df4b2
JK		 4492 if (!sentry->valid_blocks)
4493 __set_free(sbi, start);
c79b7ff1
JK		 4494 else
4495 SIT_I(sbi)->written_valid_blocks +=
4496 sentry->valid_blocks;
351df4b2
JK		 4497 }
4498
4499 /* set use the current segments */
4500 for (type = CURSEG_HOT_DATA; type <= CURSEG_COLD_NODE; type++) {
4501 struct curseg_info *curseg_t = CURSEG_I(sbi, type);
4502 __set_test_and_inuse(sbi, curseg_t->segno);
4503 }
4504}
4505
4506static void init_dirty_segmap(struct f2fs_sb_info *sbi)
4507{
4508 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
4509 struct free_segmap_info *free_i = FREE_I(sbi);
da52f8ad 4510 unsigned int segno = 0, offset = 0, secno;
de881df9 4511 block_t valid_blocks, usable_blks_in_seg;
123aaf77 4512 block_t blks_per_sec = BLKS_PER_SEC(sbi);
351df4b2 4513
8736fbf0 4514 while (1) {
351df4b2 4515 /* find dirty segment based on free segmap */
7cd8558b
JK		 4516 segno = find_next_inuse(free_i, MAIN_SEGS(sbi), offset);
4517 if (segno >= MAIN_SEGS(sbi))
351df4b2
JK		 4518 break;
4519 offset = segno + 1;
302bd348 4520 valid_blocks = get_valid_blocks(sbi, segno, false);
de881df9
AR		 4521 usable_blks_in_seg = f2fs_usable_blks_in_seg(sbi, segno);
4522 if (valid_blocks == usable_blks_in_seg || !valid_blocks)
351df4b2 4523 continue;
de881df9 4524 if (valid_blocks > usable_blks_in_seg) {
ec325b52
JK		 4525 f2fs_bug_on(sbi, 1);
4526 continue;
4527 }
351df4b2
JK		 4528 mutex_lock(&dirty_i->seglist_lock);
4529 __locate_dirty_segment(sbi, segno, DIRTY);
4530 mutex_unlock(&dirty_i->seglist_lock);
4531 }
da52f8ad
JQ		 4532
4533 if (!__is_large_section(sbi))
4534 return;
4535
4536 mutex_lock(&dirty_i->seglist_lock);
5335bfc6 4537 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
da52f8ad
JQ		 4538 valid_blocks = get_valid_blocks(sbi, segno, true);
4539 secno = GET_SEC_FROM_SEG(sbi, segno);
4540
4541 if (!valid_blocks || valid_blocks == blks_per_sec)
4542 continue;
4543 if (IS_CURSEC(sbi, secno))
4544 continue;
4545 set_bit(secno, dirty_i->dirty_secmap);
4546 }
4547 mutex_unlock(&dirty_i->seglist_lock);
351df4b2
JK		 4548}
4549
5ec4e49f 4550static int init_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 4551{
4552 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
7cd8558b 4553 unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
351df4b2 4554
628b3d14 4555 dirty_i->victim_secmap = f2fs_kvzalloc(sbi, bitmap_size, GFP_KERNEL);
5ec4e49f 4556 if (!dirty_i->victim_secmap)
351df4b2
JK		 4557 return -ENOMEM;
4558 return 0;
4559}
4560
4561static int build_dirty_segmap(struct f2fs_sb_info *sbi)
4562{
4563 struct dirty_seglist_info *dirty_i;
4564 unsigned int bitmap_size, i;
4565
4566 /* allocate memory for dirty segments list information */
acbf054d
CY		 4567 dirty_i = f2fs_kzalloc(sbi, sizeof(struct dirty_seglist_info),
4568 GFP_KERNEL);
351df4b2
JK		 4569 if (!dirty_i)
4570 return -ENOMEM;
4571
4572 SM_I(sbi)->dirty_info = dirty_i;
4573 mutex_init(&dirty_i->seglist_lock);
4574
7cd8558b 4575 bitmap_size = f2fs_bitmap_size(MAIN_SEGS(sbi));
351df4b2
JK		 4576
4577 for (i = 0; i < NR_DIRTY_TYPE; i++) {
628b3d14
CY		 4578 dirty_i->dirty_segmap[i] = f2fs_kvzalloc(sbi, bitmap_size,
4579 GFP_KERNEL);
351df4b2
JK		 4580 if (!dirty_i->dirty_segmap[i])
4581 return -ENOMEM;
4582 }
4583
da52f8ad
JQ		 4584 if (__is_large_section(sbi)) {
4585 bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi));
4586 dirty_i->dirty_secmap = f2fs_kvzalloc(sbi,
4587 bitmap_size, GFP_KERNEL);
4588 if (!dirty_i->dirty_secmap)
4589 return -ENOMEM;
4590 }
4591
351df4b2 4592 init_dirty_segmap(sbi);
5ec4e49f 4593 return init_victim_secmap(sbi);
351df4b2
JK		 4594}
4595
c854f4d6
CY		 4596static int sanity_check_curseg(struct f2fs_sb_info *sbi)
4597{
4598 int i;
4599
4600 /*
4601 * In LFS/SSR curseg, .next_blkoff should point to an unused blkaddr;
4602 * In LFS curseg, all blkaddr after .next_blkoff should be unused.
4603 */
d0b9e42a 4604 for (i = 0; i < NR_PERSISTENT_LOG; i++) {
c854f4d6
CY		 4605 struct curseg_info *curseg = CURSEG_I(sbi, i);
4606 struct seg_entry *se = get_seg_entry(sbi, curseg->segno);
4607 unsigned int blkofs = curseg->next_blkoff;
4608
093749e2
CY		 4609 sanity_check_seg_type(sbi, curseg->seg_type);
4610
c854f4d6
CY		 4611 if (f2fs_test_bit(blkofs, se->cur_valid_map))
4612 goto out;
4613
4614 if (curseg->alloc_type == SSR)
4615 continue;
4616
4617 for (blkofs += 1; blkofs < sbi->blocks_per_seg; blkofs++) {
4618 if (!f2fs_test_bit(blkofs, se->cur_valid_map))
4619 continue;
4620out:
dcbb4c10
JP		 4621 f2fs_err(sbi,
4622 "Current segment's next free block offset is inconsistent with bitmap, logtype:%u, segno:%u, type:%u, next_blkoff:%u, blkofs:%u",
4623 i, curseg->segno, curseg->alloc_type,
4624 curseg->next_blkoff, blkofs);
10f966bb 4625 return -EFSCORRUPTED;
c854f4d6
CY		 4626 }
4627 }
4628 return 0;
4629}
4630
c426d991
SK		 4631#ifdef CONFIG_BLK_DEV_ZONED
4632
d508c94e
SK		 4633static int check_zone_write_pointer(struct f2fs_sb_info *sbi,
4634 struct f2fs_dev_info *fdev,
4635 struct blk_zone *zone)
4636{
4637 unsigned int wp_segno, wp_blkoff, zone_secno, zone_segno, segno;
4638 block_t zone_block, wp_block, last_valid_block;
4639 unsigned int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
4640 int i, s, b, ret;
4641 struct seg_entry *se;
4642
4643 if (zone->type != BLK_ZONE_TYPE_SEQWRITE_REQ)
4644 return 0;
4645
4646 wp_block = fdev->start_blk + (zone->wp >> log_sectors_per_block);
4647 wp_segno = GET_SEGNO(sbi, wp_block);
4648 wp_blkoff = wp_block - START_BLOCK(sbi, wp_segno);
4649 zone_block = fdev->start_blk + (zone->start >> log_sectors_per_block);
4650 zone_segno = GET_SEGNO(sbi, zone_block);
4651 zone_secno = GET_SEC_FROM_SEG(sbi, zone_segno);
4652
4653 if (zone_segno >= MAIN_SEGS(sbi))
4654 return 0;
4655
4656 /*
4657 * Skip check of zones cursegs point to, since
4658 * fix_curseg_write_pointer() checks them.
4659 */
4660 for (i = 0; i < NO_CHECK_TYPE; i++)
4661 if (zone_secno == GET_SEC_FROM_SEG(sbi,
4662 CURSEG_I(sbi, i)->segno))
4663 return 0;
4664
4665 /*
4666 * Get last valid block of the zone.
4667 */
4668 last_valid_block = zone_block - 1;
4669 for (s = sbi->segs_per_sec - 1; s >= 0; s--) {
4670 segno = zone_segno + s;
4671 se = get_seg_entry(sbi, segno);
4672 for (b = sbi->blocks_per_seg - 1; b >= 0; b--)
4673 if (f2fs_test_bit(b, se->cur_valid_map)) {
4674 last_valid_block = START_BLOCK(sbi, segno) + b;
4675 break;
4676 }
4677 if (last_valid_block >= zone_block)
4678 break;
4679 }
4680
4681 /*
4682 * If last valid block is beyond the write pointer, report the
4683 * inconsistency. This inconsistency does not cause write error
4684 * because the zone will not be selected for write operation until
4685 * it get discarded. Just report it.
4686 */
4687 if (last_valid_block >= wp_block) {
4688 f2fs_notice(sbi, "Valid block beyond write pointer: "
4689 "valid block[0x%x,0x%x] wp[0x%x,0x%x]",
4690 GET_SEGNO(sbi, last_valid_block),
4691 GET_BLKOFF_FROM_SEG0(sbi, last_valid_block),
4692 wp_segno, wp_blkoff);
4693 return 0;
4694 }
4695
4696 /*
4697 * If there is no valid block in the zone and if write pointer is
4698 * not at zone start, reset the write pointer.
4699 */
4700 if (last_valid_block + 1 == zone_block && zone->wp != zone->start) {
4701 f2fs_notice(sbi,
4702 "Zone without valid block has non-zero write "
4703 "pointer. Reset the write pointer: wp[0x%x,0x%x]",
4704 wp_segno, wp_blkoff);
4705 ret = __f2fs_issue_discard_zone(sbi, fdev->bdev, zone_block,
4706 zone->len >> log_sectors_per_block);
4707 if (ret) {
4708 f2fs_err(sbi, "Discard zone failed: %s (errno=%d)",
4709 fdev->path, ret);
4710 return ret;
4711 }
4712 }
4713
4714 return 0;
4715}
4716
c426d991
SK		 4717static struct f2fs_dev_info *get_target_zoned_dev(struct f2fs_sb_info *sbi,
4718 block_t zone_blkaddr)
4719{
4720 int i;
4721
4722 for (i = 0; i < sbi->s_ndevs; i++) {
4723 if (!bdev_is_zoned(FDEV(i).bdev))
4724 continue;
4725 if (sbi->s_ndevs == 1 || (FDEV(i).start_blk <= zone_blkaddr &&
4726 zone_blkaddr <= FDEV(i).end_blk))
4727 return &FDEV(i);
4728 }
4729
4730 return NULL;
4731}
4732
4733static int report_one_zone_cb(struct blk_zone *zone, unsigned int idx,
4734 void *data) {
4735 memcpy(data, zone, sizeof(struct blk_zone));
4736 return 0;
4737}
4738
4739static int fix_curseg_write_pointer(struct f2fs_sb_info *sbi, int type)
4740{
4741 struct curseg_info *cs = CURSEG_I(sbi, type);
4742 struct f2fs_dev_info *zbd;
4743 struct blk_zone zone;
4744 unsigned int cs_section, wp_segno, wp_blkoff, wp_sector_off;
4745 block_t cs_zone_block, wp_block;
4746 unsigned int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
4747 sector_t zone_sector;
4748 int err;
4749
4750 cs_section = GET_SEC_FROM_SEG(sbi, cs->segno);
4751 cs_zone_block = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, cs_section));
4752
4753 zbd = get_target_zoned_dev(sbi, cs_zone_block);
4754 if (!zbd)
4755 return 0;
4756
4757 /* report zone for the sector the curseg points to */
4758 zone_sector = (sector_t)(cs_zone_block - zbd->start_blk)
4759 << log_sectors_per_block;
4760 err = blkdev_report_zones(zbd->bdev, zone_sector, 1,
4761 report_one_zone_cb, &zone);
4762 if (err != 1) {
4763 f2fs_err(sbi, "Report zone failed: %s errno=(%d)",
4764 zbd->path, err);
4765 return err;
4766 }
4767
4768 if (zone.type != BLK_ZONE_TYPE_SEQWRITE_REQ)
4769 return 0;
4770
4771 wp_block = zbd->start_blk + (zone.wp >> log_sectors_per_block);
4772 wp_segno = GET_SEGNO(sbi, wp_block);
4773 wp_blkoff = wp_block - START_BLOCK(sbi, wp_segno);
4774 wp_sector_off = zone.wp & GENMASK(log_sectors_per_block - 1, 0);
4775
4776 if (cs->segno == wp_segno && cs->next_blkoff == wp_blkoff &&
4777 wp_sector_off == 0)
4778 return 0;
4779
4780 f2fs_notice(sbi, "Unaligned curseg[%d] with write pointer: "
4781 "curseg[0x%x,0x%x] wp[0x%x,0x%x]",
4782 type, cs->segno, cs->next_blkoff, wp_segno, wp_blkoff);
4783
4784 f2fs_notice(sbi, "Assign new section to curseg[%d]: "
4785 "curseg[0x%x,0x%x]", type, cs->segno, cs->next_blkoff);
4786 allocate_segment_by_default(sbi, type, true);
4787
d508c94e
SK		 4788 /* check consistency of the zone curseg pointed to */
4789 if (check_zone_write_pointer(sbi, zbd, &zone))
4790 return -EIO;
4791
c426d991
SK		 4792 /* check newly assigned zone */
4793 cs_section = GET_SEC_FROM_SEG(sbi, cs->segno);
4794 cs_zone_block = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, cs_section));
4795
4796 zbd = get_target_zoned_dev(sbi, cs_zone_block);
4797 if (!zbd)
4798 return 0;
4799
4800 zone_sector = (sector_t)(cs_zone_block - zbd->start_blk)
4801 << log_sectors_per_block;
4802 err = blkdev_report_zones(zbd->bdev, zone_sector, 1,
4803 report_one_zone_cb, &zone);
4804 if (err != 1) {
4805 f2fs_err(sbi, "Report zone failed: %s errno=(%d)",
4806 zbd->path, err);
4807 return err;
4808 }
4809
4810 if (zone.type != BLK_ZONE_TYPE_SEQWRITE_REQ)
4811 return 0;
4812
4813 if (zone.wp != zone.start) {
4814 f2fs_notice(sbi,
4815 "New zone for curseg[%d] is not yet discarded. "
4816 "Reset the zone: curseg[0x%x,0x%x]",
4817 type, cs->segno, cs->next_blkoff);
4818 err = __f2fs_issue_discard_zone(sbi, zbd->bdev,
4819 zone_sector >> log_sectors_per_block,
4820 zone.len >> log_sectors_per_block);
4821 if (err) {
4822 f2fs_err(sbi, "Discard zone failed: %s (errno=%d)",
4823 zbd->path, err);
4824 return err;
4825 }
4826 }
4827
4828 return 0;
4829}
4830
4831int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi)
4832{
4833 int i, ret;
4834
d0b9e42a 4835 for (i = 0; i < NR_PERSISTENT_LOG; i++) {
c426d991
SK		 4836 ret = fix_curseg_write_pointer(sbi, i);
4837 if (ret)
4838 return ret;
4839 }
4840
4841 return 0;
4842}
d508c94e
SK		 4843
4844struct check_zone_write_pointer_args {
4845 struct f2fs_sb_info *sbi;
4846 struct f2fs_dev_info *fdev;
4847};
4848
4849static int check_zone_write_pointer_cb(struct blk_zone *zone, unsigned int idx,
4850 void *data) {
4851 struct check_zone_write_pointer_args *args;
4852 args = (struct check_zone_write_pointer_args *)data;
4853
4854 return check_zone_write_pointer(args->sbi, args->fdev, zone);
4855}
4856
4857int f2fs_check_write_pointer(struct f2fs_sb_info *sbi)
4858{
4859 int i, ret;
4860 struct check_zone_write_pointer_args args;
4861
4862 for (i = 0; i < sbi->s_ndevs; i++) {
4863 if (!bdev_is_zoned(FDEV(i).bdev))
4864 continue;
4865
4866 args.sbi = sbi;
4867 args.fdev = &FDEV(i);
4868 ret = blkdev_report_zones(FDEV(i).bdev, 0, BLK_ALL_ZONES,
4869 check_zone_write_pointer_cb, &args);
4870 if (ret < 0)
4871 return ret;
4872 }
4873
4874 return 0;
4875}
de881df9
AR		 4876
4877static bool is_conv_zone(struct f2fs_sb_info *sbi, unsigned int zone_idx,
4878 unsigned int dev_idx)
4879{
4880 if (!bdev_is_zoned(FDEV(dev_idx).bdev))
4881 return true;
4882 return !test_bit(zone_idx, FDEV(dev_idx).blkz_seq);
4883}
4884
4885/* Return the zone index in the given device */
4886static unsigned int get_zone_idx(struct f2fs_sb_info *sbi, unsigned int secno,
4887 int dev_idx)
4888{
4889 block_t sec_start_blkaddr = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, secno));
4890
4891 return (sec_start_blkaddr - FDEV(dev_idx).start_blk) >>
4892 sbi->log_blocks_per_blkz;
4893}
4894
4895/*
4896 * Return the usable segments in a section based on the zone's
4897 * corresponding zone capacity. Zone is equal to a section.
4898 */
4899static inline unsigned int f2fs_usable_zone_segs_in_sec(
4900 struct f2fs_sb_info *sbi, unsigned int segno)
4901{
4902 unsigned int dev_idx, zone_idx, unusable_segs_in_sec;
4903
4904 dev_idx = f2fs_target_device_index(sbi, START_BLOCK(sbi, segno));
4905 zone_idx = get_zone_idx(sbi, GET_SEC_FROM_SEG(sbi, segno), dev_idx);
4906
4907 /* Conventional zone's capacity is always equal to zone size */
4908 if (is_conv_zone(sbi, zone_idx, dev_idx))
4909 return sbi->segs_per_sec;
4910
4911 /*
4912 * If the zone_capacity_blocks array is NULL, then zone capacity
4913 * is equal to the zone size for all zones
4914 */
4915 if (!FDEV(dev_idx).zone_capacity_blocks)
4916 return sbi->segs_per_sec;
4917
4918 /* Get the segment count beyond zone capacity block */
4919 unusable_segs_in_sec = (sbi->blocks_per_blkz -
4920 FDEV(dev_idx).zone_capacity_blocks[zone_idx]) >>
4921 sbi->log_blocks_per_seg;
4922 return sbi->segs_per_sec - unusable_segs_in_sec;
4923}
4924
4925/*
4926 * Return the number of usable blocks in a segment. The number of blocks
4927 * returned is always equal to the number of blocks in a segment for
4928 * segments fully contained within a sequential zone capacity or a
4929 * conventional zone. For segments partially contained in a sequential
4930 * zone capacity, the number of usable blocks up to the zone capacity
4931 * is returned. 0 is returned in all other cases.
4932 */
4933static inline unsigned int f2fs_usable_zone_blks_in_seg(
4934 struct f2fs_sb_info *sbi, unsigned int segno)
4935{
4936 block_t seg_start, sec_start_blkaddr, sec_cap_blkaddr;
4937 unsigned int zone_idx, dev_idx, secno;
4938
4939 secno = GET_SEC_FROM_SEG(sbi, segno);
4940 seg_start = START_BLOCK(sbi, segno);
4941 dev_idx = f2fs_target_device_index(sbi, seg_start);
4942 zone_idx = get_zone_idx(sbi, secno, dev_idx);
4943
4944 /*
4945 * Conventional zone's capacity is always equal to zone size,
4946 * so, blocks per segment is unchanged.
4947 */
4948 if (is_conv_zone(sbi, zone_idx, dev_idx))
4949 return sbi->blocks_per_seg;
4950
4951 if (!FDEV(dev_idx).zone_capacity_blocks)
4952 return sbi->blocks_per_seg;
4953
4954 sec_start_blkaddr = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, secno));
4955 sec_cap_blkaddr = sec_start_blkaddr +
4956 FDEV(dev_idx).zone_capacity_blocks[zone_idx];
4957
4958 /*
4959 * If segment starts before zone capacity and spans beyond
4960 * zone capacity, then usable blocks are from seg start to
4961 * zone capacity. If the segment starts after the zone capacity,
4962 * then there are no usable blocks.
4963 */
4964 if (seg_start >= sec_cap_blkaddr)
4965 return 0;
4966 if (seg_start + sbi->blocks_per_seg > sec_cap_blkaddr)
4967 return sec_cap_blkaddr - seg_start;
4968
4969 return sbi->blocks_per_seg;
4970}
c426d991
SK		 4971#else
4972int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi)
4973{
4974 return 0;
4975}
d508c94e
SK		 4976
4977int f2fs_check_write_pointer(struct f2fs_sb_info *sbi)
4978{
4979 return 0;
4980}
de881df9
AR		 4981
4982static inline unsigned int f2fs_usable_zone_blks_in_seg(struct f2fs_sb_info *sbi,
4983 unsigned int segno)
4984{
4985 return 0;
4986}
4987
4988static inline unsigned int f2fs_usable_zone_segs_in_sec(struct f2fs_sb_info *sbi,
4989 unsigned int segno)
4990{
4991 return 0;
4992}
c426d991 4993#endif
de881df9
AR		 4994unsigned int f2fs_usable_blks_in_seg(struct f2fs_sb_info *sbi,
4995 unsigned int segno)
4996{
4997 if (f2fs_sb_has_blkzoned(sbi))
4998 return f2fs_usable_zone_blks_in_seg(sbi, segno);
4999
5000 return sbi->blocks_per_seg;
5001}
5002
5003unsigned int f2fs_usable_segs_in_sec(struct f2fs_sb_info *sbi,
5004 unsigned int segno)
5005{
5006 if (f2fs_sb_has_blkzoned(sbi))
5007 return f2fs_usable_zone_segs_in_sec(sbi, segno);
5008
5009 return sbi->segs_per_sec;
5010}
c426d991 5011
0a8165d7 5012/*
351df4b2
JK		 5013 * Update min, max modified time for cost-benefit GC algorithm
5014 */
5015static void init_min_max_mtime(struct f2fs_sb_info *sbi)
5016{
5017 struct sit_info *sit_i = SIT_I(sbi);
5018 unsigned int segno;
5019
3d26fa6b 5020 down_write(&sit_i->sentry_lock);
351df4b2 5021
5ad25442 5022 sit_i->min_mtime = ULLONG_MAX;
351df4b2 5023
7cd8558b 5024 for (segno = 0; segno < MAIN_SEGS(sbi); segno += sbi->segs_per_sec) {
351df4b2
JK		 5025 unsigned int i;
5026 unsigned long long mtime = 0;
5027
5028 for (i = 0; i < sbi->segs_per_sec; i++)
5029 mtime += get_seg_entry(sbi, segno + i)->mtime;
5030
5031 mtime = div_u64(mtime, sbi->segs_per_sec);
5032
5033 if (sit_i->min_mtime > mtime)
5034 sit_i->min_mtime = mtime;
5035 }
a1f72ac2 5036 sit_i->max_mtime = get_mtime(sbi, false);
093749e2 5037 sit_i->dirty_max_mtime = 0;
3d26fa6b 5038 up_write(&sit_i->sentry_lock);
351df4b2
JK		 5039}
5040
4d57b86d 5041int f2fs_build_segment_manager(struct f2fs_sb_info *sbi)
351df4b2
JK		 5042{
5043 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
5044 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1042d60f 5045 struct f2fs_sm_info *sm_info;
351df4b2
JK		 5046 int err;
5047
acbf054d 5048 sm_info = f2fs_kzalloc(sbi, sizeof(struct f2fs_sm_info), GFP_KERNEL);
351df4b2
JK		 5049 if (!sm_info)
5050 return -ENOMEM;
5051
5052 /* init sm info */
5053 sbi->sm_info = sm_info;
351df4b2
JK		 5054 sm_info->seg0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
5055 sm_info->main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
5056 sm_info->segment_count = le32_to_cpu(raw_super->segment_count);
5057 sm_info->reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
5058 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
5059 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
5060 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
58c41035
JK		 5061 sm_info->rec_prefree_segments = sm_info->main_segments *
5062 DEF_RECLAIM_PREFREE_SEGMENTS / 100;
44a83499
JK		 5063 if (sm_info->rec_prefree_segments > DEF_MAX_RECLAIM_PREFREE_SEGMENTS)
5064 sm_info->rec_prefree_segments = DEF_MAX_RECLAIM_PREFREE_SEGMENTS;
5065
b0332a0f 5066 if (!f2fs_lfs_mode(sbi))
52763a4b 5067 sm_info->ipu_policy = 1 << F2FS_IPU_FSYNC;
216fbd64 5068 sm_info->min_ipu_util = DEF_MIN_IPU_UTIL;
c1ce1b02 5069 sm_info->min_fsync_blocks = DEF_MIN_FSYNC_BLOCKS;
853137ce 5070 sm_info->min_seq_blocks = sbi->blocks_per_seg * sbi->segs_per_sec;
ef095d19 5071 sm_info->min_hot_blocks = DEF_MIN_HOT_BLOCKS;
a2a12b67 5072 sm_info->min_ssr_sections = reserved_sections(sbi);
351df4b2 5073
184a5cd2
CY		 5074 INIT_LIST_HEAD(&sm_info->sit_entry_set);
5075
2b60311d
CY		 5076 init_rwsem(&sm_info->curseg_lock);
5077
d4fdf8ba 5078 if (!f2fs_readonly(sbi->sb)) {
4d57b86d 5079 err = f2fs_create_flush_cmd_control(sbi);
2163d198 5080 if (err)
a688b9d9 5081 return err;
6b4afdd7
JK		 5082 }
5083
0b54fb84
JK		 5084 err = create_discard_cmd_control(sbi);
5085 if (err)
5086 return err;
5087
351df4b2
JK		 5088 err = build_sit_info(sbi);
5089 if (err)
5090 return err;
5091 err = build_free_segmap(sbi);
5092 if (err)
5093 return err;
5094 err = build_curseg(sbi);
5095 if (err)
5096 return err;
5097
5098 /* reinit free segmap based on SIT */
c39a1b34
JK		 5099 err = build_sit_entries(sbi);
5100 if (err)
5101 return err;
351df4b2
JK		 5102
5103 init_free_segmap(sbi);
5104 err = build_dirty_segmap(sbi);
5105 if (err)
5106 return err;
5107
c854f4d6
CY		 5108 err = sanity_check_curseg(sbi);
5109 if (err)
5110 return err;
5111
351df4b2
JK		 5112 init_min_max_mtime(sbi);
5113 return 0;
5114}
5115
5116static void discard_dirty_segmap(struct f2fs_sb_info *sbi,
5117 enum dirty_type dirty_type)
5118{
5119 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
5120
5121 mutex_lock(&dirty_i->seglist_lock);
39307a8e 5122 kvfree(dirty_i->dirty_segmap[dirty_type]);
351df4b2
JK		 5123 dirty_i->nr_dirty[dirty_type] = 0;
5124 mutex_unlock(&dirty_i->seglist_lock);
5125}
5126
5ec4e49f 5127static void destroy_victim_secmap(struct f2fs_sb_info *sbi)
351df4b2
JK		 5128{
5129 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
39307a8e 5130 kvfree(dirty_i->victim_secmap);
351df4b2
JK		 5131}
5132
5133static void destroy_dirty_segmap(struct f2fs_sb_info *sbi)
5134{
5135 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
5136 int i;
5137
5138 if (!dirty_i)
5139 return;
5140
5141 /* discard pre-free/dirty segments list */
5142 for (i = 0; i < NR_DIRTY_TYPE; i++)
5143 discard_dirty_segmap(sbi, i);
5144
da52f8ad
JQ		 5145 if (__is_large_section(sbi)) {
5146 mutex_lock(&dirty_i->seglist_lock);
5147 kvfree(dirty_i->dirty_secmap);
5148 mutex_unlock(&dirty_i->seglist_lock);
5149 }
5150
5ec4e49f 5151 destroy_victim_secmap(sbi);
351df4b2 5152 SM_I(sbi)->dirty_info = NULL;
c8eb7024 5153 kfree(dirty_i);
351df4b2
JK		 5154}
5155
5156static void destroy_curseg(struct f2fs_sb_info *sbi)
5157{
5158 struct curseg_info *array = SM_I(sbi)->curseg_array;
5159 int i;
5160
5161 if (!array)
5162 return;
5163 SM_I(sbi)->curseg_array = NULL;
b7ad7512 5164 for (i = 0; i < NR_CURSEG_TYPE; i++) {
c8eb7024
CY		 5165 kfree(array[i].sum_blk);
5166 kfree(array[i].journal);
b7ad7512 5167 }
c8eb7024 5168 kfree(array);
351df4b2
JK		 5169}
5170
5171static void destroy_free_segmap(struct f2fs_sb_info *sbi)
5172{
5173 struct free_segmap_info *free_i = SM_I(sbi)->free_info;
5174 if (!free_i)
5175 return;
5176 SM_I(sbi)->free_info = NULL;
39307a8e
JK		 5177 kvfree(free_i->free_segmap);
5178 kvfree(free_i->free_secmap);
c8eb7024 5179 kfree(free_i);
351df4b2
JK		 5180}
5181
5182static void destroy_sit_info(struct f2fs_sb_info *sbi)
5183{
5184 struct sit_info *sit_i = SIT_I(sbi);
351df4b2
JK		 5185
5186 if (!sit_i)
5187 return;
5188
2fde3dd1
CY		 5189 if (sit_i->sentries)
5190 kvfree(sit_i->bitmap);
c8eb7024 5191 kfree(sit_i->tmp_map);
60a3b782 5192
39307a8e
JK		 5193 kvfree(sit_i->sentries);
5194 kvfree(sit_i->sec_entries);
5195 kvfree(sit_i->dirty_sentries_bitmap);
351df4b2
JK		 5196
5197 SM_I(sbi)->sit_info = NULL;
5222595d 5198 kvfree(sit_i->sit_bitmap);
ae27d62e 5199#ifdef CONFIG_F2FS_CHECK_FS
5222595d 5200 kvfree(sit_i->sit_bitmap_mir);
bbf9f7d9 5201 kvfree(sit_i->invalid_segmap);
ae27d62e 5202#endif
c8eb7024 5203 kfree(sit_i);
351df4b2
JK		 5204}
5205
4d57b86d 5206void f2fs_destroy_segment_manager(struct f2fs_sb_info *sbi)
351df4b2
JK		 5207{
5208 struct f2fs_sm_info *sm_info = SM_I(sbi);
a688b9d9 5209
3b03f724
CY		 5210 if (!sm_info)
5211 return;
4d57b86d 5212 f2fs_destroy_flush_cmd_control(sbi, true);
f099405f 5213 destroy_discard_cmd_control(sbi);
351df4b2
JK		 5214 destroy_dirty_segmap(sbi);
5215 destroy_curseg(sbi);
5216 destroy_free_segmap(sbi);
5217 destroy_sit_info(sbi);
5218 sbi->sm_info = NULL;
c8eb7024 5219 kfree(sm_info);
351df4b2 5220}
7fd9e544 5221
4d57b86d 5222int __init f2fs_create_segment_manager_caches(void)
7fd9e544 5223{
98510003 5224 discard_entry_slab = f2fs_kmem_cache_create("f2fs_discard_entry",
e8512d2e 5225 sizeof(struct discard_entry));
7fd9e544 5226 if (!discard_entry_slab)
184a5cd2
CY		 5227 goto fail;
5228
98510003 5229 discard_cmd_slab = f2fs_kmem_cache_create("f2fs_discard_cmd",
b01a9201
JK		 5230 sizeof(struct discard_cmd));
5231 if (!discard_cmd_slab)
6ab2a308 5232 goto destroy_discard_entry;
275b66b0 5233
98510003 5234 sit_entry_set_slab = f2fs_kmem_cache_create("f2fs_sit_entry_set",
c9ee0085 5235 sizeof(struct sit_entry_set));
184a5cd2 5236 if (!sit_entry_set_slab)
b01a9201 5237 goto destroy_discard_cmd;
88b88a66 5238
98510003 5239 inmem_entry_slab = f2fs_kmem_cache_create("f2fs_inmem_page_entry",
88b88a66
JK		 5240 sizeof(struct inmem_pages));
5241 if (!inmem_entry_slab)
5242 goto destroy_sit_entry_set;
7fd9e544 5243 return 0;
184a5cd2 5244
88b88a66
JK		 5245destroy_sit_entry_set:
5246 kmem_cache_destroy(sit_entry_set_slab);
b01a9201
JK		 5247destroy_discard_cmd:
5248 kmem_cache_destroy(discard_cmd_slab);
6ab2a308 5249destroy_discard_entry:
184a5cd2
CY		 5250 kmem_cache_destroy(discard_entry_slab);
5251fail:
5252 return -ENOMEM;
7fd9e544
JK		 5253}
5254
4d57b86d 5255void f2fs_destroy_segment_manager_caches(void)
7fd9e544 5256{
184a5cd2 5257 kmem_cache_destroy(sit_entry_set_slab);
b01a9201 5258 kmem_cache_destroy(discard_cmd_slab);
7fd9e544 5259 kmem_cache_destroy(discard_entry_slab);
88b88a66 5260 kmem_cache_destroy(inmem_entry_slab);
7fd9e544 5261}
Ubuntu Jammy Linux kernel mirror