]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - fs/f2fs/inode.c
f2fs: don't count inode block in in-memory inode.i_blocks
[mirror_ubuntu-jammy-kernel.git] / fs / f2fs / inode.c
1 /*
2 * fs/f2fs/inode.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11 #include <linux/fs.h>
12 #include <linux/f2fs_fs.h>
13 #include <linux/buffer_head.h>
14 #include <linux/backing-dev.h>
15 #include <linux/writeback.h>
16
17 #include "f2fs.h"
18 #include "node.h"
19 #include "segment.h"
20
21 #include <trace/events/f2fs.h>
22
23 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
24 {
25 if (f2fs_inode_dirtied(inode, sync))
26 return;
27
28 mark_inode_dirty_sync(inode);
29 }
30
31 void f2fs_set_inode_flags(struct inode *inode)
32 {
33 unsigned int flags = F2FS_I(inode)->i_flags;
34 unsigned int new_fl = 0;
35
36 if (flags & FS_SYNC_FL)
37 new_fl |= S_SYNC;
38 if (flags & FS_APPEND_FL)
39 new_fl |= S_APPEND;
40 if (flags & FS_IMMUTABLE_FL)
41 new_fl |= S_IMMUTABLE;
42 if (flags & FS_NOATIME_FL)
43 new_fl |= S_NOATIME;
44 if (flags & FS_DIRSYNC_FL)
45 new_fl |= S_DIRSYNC;
46 inode_set_flags(inode, new_fl,
47 S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
48 }
49
50 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
51 {
52 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
53 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
54 if (ri->i_addr[0])
55 inode->i_rdev =
56 old_decode_dev(le32_to_cpu(ri->i_addr[0]));
57 else
58 inode->i_rdev =
59 new_decode_dev(le32_to_cpu(ri->i_addr[1]));
60 }
61 }
62
63 static bool __written_first_block(struct f2fs_inode *ri)
64 {
65 block_t addr = le32_to_cpu(ri->i_addr[0]);
66
67 if (addr != NEW_ADDR && addr != NULL_ADDR)
68 return true;
69 return false;
70 }
71
72 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
73 {
74 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
75 if (old_valid_dev(inode->i_rdev)) {
76 ri->i_addr[0] =
77 cpu_to_le32(old_encode_dev(inode->i_rdev));
78 ri->i_addr[1] = 0;
79 } else {
80 ri->i_addr[0] = 0;
81 ri->i_addr[1] =
82 cpu_to_le32(new_encode_dev(inode->i_rdev));
83 ri->i_addr[2] = 0;
84 }
85 }
86 }
87
88 static void __recover_inline_status(struct inode *inode, struct page *ipage)
89 {
90 void *inline_data = inline_data_addr(ipage);
91 __le32 *start = inline_data;
92 __le32 *end = start + MAX_INLINE_DATA / sizeof(__le32);
93
94 while (start < end) {
95 if (*start++) {
96 f2fs_wait_on_page_writeback(ipage, NODE, true);
97
98 set_inode_flag(inode, FI_DATA_EXIST);
99 set_raw_inline(inode, F2FS_INODE(ipage));
100 set_page_dirty(ipage);
101 return;
102 }
103 }
104 return;
105 }
106
107 static int do_read_inode(struct inode *inode)
108 {
109 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
110 struct f2fs_inode_info *fi = F2FS_I(inode);
111 struct page *node_page;
112 struct f2fs_inode *ri;
113
114 /* Check if ino is within scope */
115 if (check_nid_range(sbi, inode->i_ino)) {
116 f2fs_msg(inode->i_sb, KERN_ERR, "bad inode number: %lu",
117 (unsigned long) inode->i_ino);
118 WARN_ON(1);
119 return -EINVAL;
120 }
121
122 node_page = get_node_page(sbi, inode->i_ino);
123 if (IS_ERR(node_page))
124 return PTR_ERR(node_page);
125
126 ri = F2FS_INODE(node_page);
127
128 inode->i_mode = le16_to_cpu(ri->i_mode);
129 i_uid_write(inode, le32_to_cpu(ri->i_uid));
130 i_gid_write(inode, le32_to_cpu(ri->i_gid));
131 set_nlink(inode, le32_to_cpu(ri->i_links));
132 inode->i_size = le64_to_cpu(ri->i_size);
133 inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
134
135 inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
136 inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
137 inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
138 inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
139 inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
140 inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
141 inode->i_generation = le32_to_cpu(ri->i_generation);
142
143 fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
144 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
145 fi->i_flags = le32_to_cpu(ri->i_flags);
146 fi->flags = 0;
147 fi->i_advise = ri->i_advise;
148 fi->i_pino = le32_to_cpu(ri->i_pino);
149 fi->i_dir_level = ri->i_dir_level;
150
151 if (f2fs_init_extent_tree(inode, &ri->i_ext))
152 set_page_dirty(node_page);
153
154 get_inline_info(inode, ri);
155
156 /* check data exist */
157 if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
158 __recover_inline_status(inode, node_page);
159
160 /* get rdev by using inline_info */
161 __get_inode_rdev(inode, ri);
162
163 if (__written_first_block(ri))
164 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
165
166 if (!need_inode_block_update(sbi, inode->i_ino))
167 fi->last_disk_size = inode->i_size;
168
169 f2fs_put_page(node_page, 1);
170
171 stat_inc_inline_xattr(inode);
172 stat_inc_inline_inode(inode);
173 stat_inc_inline_dir(inode);
174
175 return 0;
176 }
177
178 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
179 {
180 struct f2fs_sb_info *sbi = F2FS_SB(sb);
181 struct inode *inode;
182 int ret = 0;
183
184 inode = iget_locked(sb, ino);
185 if (!inode)
186 return ERR_PTR(-ENOMEM);
187
188 if (!(inode->i_state & I_NEW)) {
189 trace_f2fs_iget(inode);
190 return inode;
191 }
192 if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
193 goto make_now;
194
195 ret = do_read_inode(inode);
196 if (ret)
197 goto bad_inode;
198 make_now:
199 if (ino == F2FS_NODE_INO(sbi)) {
200 inode->i_mapping->a_ops = &f2fs_node_aops;
201 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
202 } else if (ino == F2FS_META_INO(sbi)) {
203 inode->i_mapping->a_ops = &f2fs_meta_aops;
204 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
205 } else if (S_ISREG(inode->i_mode)) {
206 inode->i_op = &f2fs_file_inode_operations;
207 inode->i_fop = &f2fs_file_operations;
208 inode->i_mapping->a_ops = &f2fs_dblock_aops;
209 } else if (S_ISDIR(inode->i_mode)) {
210 inode->i_op = &f2fs_dir_inode_operations;
211 inode->i_fop = &f2fs_dir_operations;
212 inode->i_mapping->a_ops = &f2fs_dblock_aops;
213 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_HIGH_ZERO);
214 } else if (S_ISLNK(inode->i_mode)) {
215 if (f2fs_encrypted_inode(inode))
216 inode->i_op = &f2fs_encrypted_symlink_inode_operations;
217 else
218 inode->i_op = &f2fs_symlink_inode_operations;
219 inode_nohighmem(inode);
220 inode->i_mapping->a_ops = &f2fs_dblock_aops;
221 } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
222 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
223 inode->i_op = &f2fs_special_inode_operations;
224 init_special_inode(inode, inode->i_mode, inode->i_rdev);
225 } else {
226 ret = -EIO;
227 goto bad_inode;
228 }
229 f2fs_set_inode_flags(inode);
230 unlock_new_inode(inode);
231 trace_f2fs_iget(inode);
232 return inode;
233
234 bad_inode:
235 iget_failed(inode);
236 trace_f2fs_iget_exit(inode, ret);
237 return ERR_PTR(ret);
238 }
239
240 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
241 {
242 struct inode *inode;
243 retry:
244 inode = f2fs_iget(sb, ino);
245 if (IS_ERR(inode)) {
246 if (PTR_ERR(inode) == -ENOMEM) {
247 congestion_wait(BLK_RW_ASYNC, HZ/50);
248 goto retry;
249 }
250 }
251 return inode;
252 }
253
254 int update_inode(struct inode *inode, struct page *node_page)
255 {
256 struct f2fs_inode *ri;
257 struct extent_tree *et = F2FS_I(inode)->extent_tree;
258
259 f2fs_inode_synced(inode);
260
261 f2fs_wait_on_page_writeback(node_page, NODE, true);
262
263 ri = F2FS_INODE(node_page);
264
265 ri->i_mode = cpu_to_le16(inode->i_mode);
266 ri->i_advise = F2FS_I(inode)->i_advise;
267 ri->i_uid = cpu_to_le32(i_uid_read(inode));
268 ri->i_gid = cpu_to_le32(i_gid_read(inode));
269 ri->i_links = cpu_to_le32(inode->i_nlink);
270 ri->i_size = cpu_to_le64(i_size_read(inode));
271 ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
272
273 if (et) {
274 read_lock(&et->lock);
275 set_raw_extent(&et->largest, &ri->i_ext);
276 read_unlock(&et->lock);
277 } else {
278 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
279 }
280 set_raw_inline(inode, ri);
281
282 ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
283 ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
284 ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
285 ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
286 ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
287 ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
288 ri->i_current_depth = cpu_to_le32(F2FS_I(inode)->i_current_depth);
289 ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
290 ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
291 ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
292 ri->i_generation = cpu_to_le32(inode->i_generation);
293 ri->i_dir_level = F2FS_I(inode)->i_dir_level;
294
295 __set_inode_rdev(inode, ri);
296 set_cold_node(inode, node_page);
297
298 /* deleted inode */
299 if (inode->i_nlink == 0)
300 clear_inline_node(node_page);
301
302 return set_page_dirty(node_page);
303 }
304
305 int update_inode_page(struct inode *inode)
306 {
307 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
308 struct page *node_page;
309 int ret = 0;
310 retry:
311 node_page = get_node_page(sbi, inode->i_ino);
312 if (IS_ERR(node_page)) {
313 int err = PTR_ERR(node_page);
314 if (err == -ENOMEM) {
315 cond_resched();
316 goto retry;
317 } else if (err != -ENOENT) {
318 f2fs_stop_checkpoint(sbi, false);
319 }
320 return 0;
321 }
322 ret = update_inode(inode, node_page);
323 f2fs_put_page(node_page, 1);
324 return ret;
325 }
326
327 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
328 {
329 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
330
331 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
332 inode->i_ino == F2FS_META_INO(sbi))
333 return 0;
334
335 if (!is_inode_flag_set(inode, FI_DIRTY_INODE))
336 return 0;
337
338 /*
339 * We need to balance fs here to prevent from producing dirty node pages
340 * during the urgent cleaning time when runing out of free sections.
341 */
342 update_inode_page(inode);
343 if (wbc && wbc->nr_to_write)
344 f2fs_balance_fs(sbi, true);
345 return 0;
346 }
347
348 /*
349 * Called at the last iput() if i_nlink is zero
350 */
351 void f2fs_evict_inode(struct inode *inode)
352 {
353 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
354 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
355 int err = 0;
356
357 /* some remained atomic pages should discarded */
358 if (f2fs_is_atomic_file(inode))
359 drop_inmem_pages(inode);
360
361 trace_f2fs_evict_inode(inode);
362 truncate_inode_pages_final(&inode->i_data);
363
364 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
365 inode->i_ino == F2FS_META_INO(sbi))
366 goto out_clear;
367
368 f2fs_bug_on(sbi, get_dirty_pages(inode));
369 remove_dirty_inode(inode);
370
371 f2fs_destroy_extent_tree(inode);
372
373 if (inode->i_nlink || is_bad_inode(inode))
374 goto no_delete;
375
376 remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
377 remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
378
379 sb_start_intwrite(inode->i_sb);
380 set_inode_flag(inode, FI_NO_ALLOC);
381 i_size_write(inode, 0);
382 retry:
383 if (F2FS_HAS_BLOCKS(inode))
384 err = f2fs_truncate(inode);
385
386 #ifdef CONFIG_F2FS_FAULT_INJECTION
387 if (time_to_inject(sbi, FAULT_EVICT_INODE)) {
388 f2fs_show_injection_info(FAULT_EVICT_INODE);
389 err = -EIO;
390 }
391 #endif
392 if (!err) {
393 f2fs_lock_op(sbi);
394 err = remove_inode_page(inode);
395 f2fs_unlock_op(sbi);
396 if (err == -ENOENT)
397 err = 0;
398 }
399
400 /* give more chances, if ENOMEM case */
401 if (err == -ENOMEM) {
402 err = 0;
403 goto retry;
404 }
405
406 if (err)
407 update_inode_page(inode);
408 sb_end_intwrite(inode->i_sb);
409 no_delete:
410 stat_dec_inline_xattr(inode);
411 stat_dec_inline_dir(inode);
412 stat_dec_inline_inode(inode);
413
414 /* ino == 0, if f2fs_new_inode() was failed t*/
415 if (inode->i_ino)
416 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
417 inode->i_ino);
418 if (xnid)
419 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
420 if (inode->i_nlink) {
421 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
422 add_ino_entry(sbi, inode->i_ino, APPEND_INO);
423 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
424 add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
425 }
426 if (is_inode_flag_set(inode, FI_FREE_NID)) {
427 alloc_nid_failed(sbi, inode->i_ino);
428 clear_inode_flag(inode, FI_FREE_NID);
429 } else {
430 f2fs_bug_on(sbi, err &&
431 !exist_written_data(sbi, inode->i_ino, ORPHAN_INO));
432 }
433 out_clear:
434 fscrypt_put_encryption_info(inode, NULL);
435 clear_inode(inode);
436 }
437
438 /* caller should call f2fs_lock_op() */
439 void handle_failed_inode(struct inode *inode)
440 {
441 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
442 struct node_info ni;
443
444 /*
445 * clear nlink of inode in order to release resource of inode
446 * immediately.
447 */
448 clear_nlink(inode);
449
450 /*
451 * we must call this to avoid inode being remained as dirty, resulting
452 * in a panic when flushing dirty inodes in gdirty_list.
453 */
454 update_inode_page(inode);
455 f2fs_inode_synced(inode);
456
457 /* don't make bad inode, since it becomes a regular file. */
458 unlock_new_inode(inode);
459
460 /*
461 * Note: we should add inode to orphan list before f2fs_unlock_op()
462 * so we can prevent losing this orphan when encoutering checkpoint
463 * and following suddenly power-off.
464 */
465 get_node_info(sbi, inode->i_ino, &ni);
466
467 if (ni.blk_addr != NULL_ADDR) {
468 int err = acquire_orphan_inode(sbi);
469 if (err) {
470 set_sbi_flag(sbi, SBI_NEED_FSCK);
471 f2fs_msg(sbi->sb, KERN_WARNING,
472 "Too many orphan inodes, run fsck to fix.");
473 } else {
474 add_orphan_inode(inode);
475 }
476 alloc_nid_done(sbi, inode->i_ino);
477 } else {
478 set_inode_flag(inode, FI_FREE_NID);
479 }
480
481 f2fs_unlock_op(sbi);
482
483 /* iput will drop the inode object */
484 iput(inode);
485 }