1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/inode.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/buffer_head.h>
11 #include <linux/writeback.h>
12 #include <linux/sched/mm.h>
13 #include <linux/lz4.h>
14 #include <linux/zstd.h>
15
16 #include "f2fs.h"
17 #include "node.h"
18 #include "segment.h"
19 #include "xattr.h"
20
21 #include <trace/events/f2fs.h>
22
23 #ifdef CONFIG_F2FS_FS_COMPRESSION
24 extern const struct address_space_operations f2fs_compress_aops;
25 #endif
26
27 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
28 {
29 if (is_inode_flag_set(inode, FI_NEW_INODE))
30 return;
31
32 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
33 return;
34
35 if (f2fs_inode_dirtied(inode, sync))
36 return;
37
38 if (f2fs_is_atomic_file(inode)) {
39 set_inode_flag(inode, FI_ATOMIC_DIRTIED);
40 return;
41 }
42
43 mark_inode_dirty_sync(inode);
44 }
45
46 void f2fs_set_inode_flags(struct inode *inode)
47 {
48 unsigned int flags = F2FS_I(inode)->i_flags;
49 unsigned int new_fl = 0;
50
51 if (flags & F2FS_SYNC_FL)
52 new_fl |= S_SYNC;
53 if (flags & F2FS_APPEND_FL)
54 new_fl |= S_APPEND;
55 if (flags & F2FS_IMMUTABLE_FL)
56 new_fl |= S_IMMUTABLE;
57 if (flags & F2FS_NOATIME_FL)
58 new_fl |= S_NOATIME;
59 if (flags & F2FS_DIRSYNC_FL)
60 new_fl |= S_DIRSYNC;
61 if (file_is_encrypt(inode))
62 new_fl |= S_ENCRYPTED;
63 if (file_is_verity(inode))
64 new_fl |= S_VERITY;
65 if (flags & F2FS_CASEFOLD_FL)
66 new_fl |= S_CASEFOLD;
67 inode_set_flags(inode, new_fl,
68 S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|
69 S_ENCRYPTED|S_VERITY|S_CASEFOLD);
70 }
71
72 static void __get_inode_rdev(struct inode *inode, struct page *node_page)
73 {
74 __le32 *addr = get_dnode_addr(inode, node_page);
75
76 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
77 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
78 if (addr[0])
79 inode->i_rdev = old_decode_dev(le32_to_cpu(addr[0]));
80 else
81 inode->i_rdev = new_decode_dev(le32_to_cpu(addr[1]));
82 }
83 }
84
85 static void __set_inode_rdev(struct inode *inode, struct page *node_page)
86 {
87 __le32 *addr = get_dnode_addr(inode, node_page);
88
89 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
90 if (old_valid_dev(inode->i_rdev)) {
91 addr[0] = cpu_to_le32(old_encode_dev(inode->i_rdev));
92 addr[1] = 0;
93 } else {
94 addr[0] = 0;
95 addr[1] = cpu_to_le32(new_encode_dev(inode->i_rdev));
96 addr[2] = 0;
97 }
98 }
99 }
100
101 static void __recover_inline_status(struct inode *inode, struct page *ipage)
102 {
103 void *inline_data = inline_data_addr(inode, ipage);
104 __le32 *start = inline_data;
105 __le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
106
107 while (start < end) {
108 if (*start++) {
109 f2fs_wait_on_page_writeback(ipage, NODE, true, true);
110
111 set_inode_flag(inode, FI_DATA_EXIST);
112 set_raw_inline(inode, F2FS_INODE(ipage));
113 set_page_dirty(ipage);
114 return;
115 }
116 }
117 return;
118 }
119
120 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
121 {
122 struct f2fs_inode *ri = &F2FS_NODE(page)->i;
123
124 if (!f2fs_sb_has_inode_chksum(sbi))
125 return false;
126
127 if (!IS_INODE(page) || !(ri->i_inline & F2FS_EXTRA_ATTR))
128 return false;
129
130 if (!F2FS_FITS_IN_INODE(ri, le16_to_cpu(ri->i_extra_isize),
131 i_inode_checksum))
132 return false;
133
134 return true;
135 }
136
137 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
138 {
139 struct f2fs_node *node = F2FS_NODE(page);
140 struct f2fs_inode *ri = &node->i;
141 __le32 ino = node->footer.ino;
142 __le32 gen = ri->i_generation;
143 __u32 chksum, chksum_seed;
144 __u32 dummy_cs = 0;
145 unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
146 unsigned int cs_size = sizeof(dummy_cs);
147
148 chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
149 sizeof(ino));
150 chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
151
152 chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
153 chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
154 offset += cs_size;
155 chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
156 F2FS_BLKSIZE - offset);
157 return chksum;
158 }
159
160 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
161 {
162 struct f2fs_inode *ri;
163 __u32 provided, calculated;
164
165 if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
166 return true;
167
168 #ifdef CONFIG_F2FS_CHECK_FS
169 if (!f2fs_enable_inode_chksum(sbi, page))
170 #else
171 if (!f2fs_enable_inode_chksum(sbi, page) ||
172 PageDirty(page) ||
173 folio_test_writeback(page_folio(page)))
174 #endif
175 return true;
176
177 ri = &F2FS_NODE(page)->i;
178 provided = le32_to_cpu(ri->i_inode_checksum);
179 calculated = f2fs_inode_chksum(sbi, page);
180
181 if (provided != calculated)
182 f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
183 page->index, ino_of_node(page), provided, calculated);
184
185 return provided == calculated;
186 }
187
188 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
189 {
190 struct f2fs_inode *ri = &F2FS_NODE(page)->i;
191
192 if (!f2fs_enable_inode_chksum(sbi, page))
193 return;
194
195 ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
196 }
197
198 static bool sanity_check_compress_inode(struct inode *inode,
199 struct f2fs_inode *ri)
200 {
201 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
202 unsigned char clevel;
203
204 if (ri->i_compress_algorithm >= COMPRESS_MAX) {
205 f2fs_warn(sbi,
206 "%s: inode (ino=%lx) has unsupported compress algorithm: %u, run fsck to fix",
207 __func__, inode->i_ino, ri->i_compress_algorithm);
208 return false;
209 }
210 if (le64_to_cpu(ri->i_compr_blocks) >
211 SECTOR_TO_BLOCK(inode->i_blocks)) {
212 f2fs_warn(sbi,
213 "%s: inode (ino=%lx) has inconsistent i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
214 __func__, inode->i_ino, le64_to_cpu(ri->i_compr_blocks),
215 SECTOR_TO_BLOCK(inode->i_blocks));
216 return false;
217 }
218 if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
219 ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
220 f2fs_warn(sbi,
221 "%s: inode (ino=%lx) has unsupported log cluster size: %u, run fsck to fix",
222 __func__, inode->i_ino, ri->i_log_cluster_size);
223 return false;
224 }
225
226 clevel = le16_to_cpu(ri->i_compress_flag) >>
227 COMPRESS_LEVEL_OFFSET;
228 switch (ri->i_compress_algorithm) {
229 case COMPRESS_LZO:
230 #ifdef CONFIG_F2FS_FS_LZO
231 if (clevel)
232 goto err_level;
233 #endif
234 break;
235 case COMPRESS_LZORLE:
236 #ifdef CONFIG_F2FS_FS_LZORLE
237 if (clevel)
238 goto err_level;
239 #endif
240 break;
241 case COMPRESS_LZ4:
242 #ifdef CONFIG_F2FS_FS_LZ4
243 #ifdef CONFIG_F2FS_FS_LZ4HC
244 if (clevel &&
245 (clevel < LZ4HC_MIN_CLEVEL || clevel > LZ4HC_MAX_CLEVEL))
246 goto err_level;
247 #else
248 if (clevel)
249 goto err_level;
250 #endif
251 #endif
252 break;
253 case COMPRESS_ZSTD:
254 #ifdef CONFIG_F2FS_FS_ZSTD
255 if (clevel < zstd_min_clevel() || clevel > zstd_max_clevel())
256 goto err_level;
257 #endif
258 break;
259 default:
260 goto err_level;
261 }
262
263 return true;
264 err_level:
265 f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported compress level: %u, run fsck to fix",
266 __func__, inode->i_ino, clevel);
267 return false;
268 }
269
270 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
271 {
272 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
273 struct f2fs_inode_info *fi = F2FS_I(inode);
274 struct f2fs_inode *ri = F2FS_INODE(node_page);
275 unsigned long long iblocks;
276
277 iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
278 if (!iblocks) {
279 f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
280 __func__, inode->i_ino, iblocks);
281 return false;
282 }
283
284 if (ino_of_node(node_page) != nid_of_node(node_page)) {
285 f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
286 __func__, inode->i_ino,
287 ino_of_node(node_page), nid_of_node(node_page));
288 return false;
289 }
290
291 if (f2fs_has_extra_attr(inode)) {
292 if (!f2fs_sb_has_extra_attr(sbi)) {
293 f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
294 __func__, inode->i_ino);
295 return false;
296 }
297 if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
298 fi->i_extra_isize < F2FS_MIN_EXTRA_ATTR_SIZE ||
299 fi->i_extra_isize % sizeof(__le32)) {
300 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
301 __func__, inode->i_ino, fi->i_extra_isize,
302 F2FS_TOTAL_EXTRA_ATTR_SIZE);
303 return false;
304 }
305 if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
306 f2fs_has_inline_xattr(inode) &&
307 (!fi->i_inline_xattr_size ||
308 fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
309 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %lu",
310 __func__, inode->i_ino, fi->i_inline_xattr_size,
311 MAX_INLINE_XATTR_SIZE);
312 return false;
313 }
314 if (f2fs_sb_has_compression(sbi) &&
315 fi->i_flags & F2FS_COMPR_FL &&
316 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
317 i_compress_flag)) {
318 if (!sanity_check_compress_inode(inode, ri))
319 return false;
320 }
321 }
322
323 if (!f2fs_sb_has_extra_attr(sbi)) {
324 if (f2fs_sb_has_project_quota(sbi)) {
325 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
326 __func__, inode->i_ino, F2FS_FEATURE_PRJQUOTA);
327 return false;
328 }
329 if (f2fs_sb_has_inode_chksum(sbi)) {
330 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
331 __func__, inode->i_ino, F2FS_FEATURE_INODE_CHKSUM);
332 return false;
333 }
334 if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
335 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
336 __func__, inode->i_ino, F2FS_FEATURE_FLEXIBLE_INLINE_XATTR);
337 return false;
338 }
339 if (f2fs_sb_has_inode_crtime(sbi)) {
340 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
341 __func__, inode->i_ino, F2FS_FEATURE_INODE_CRTIME);
342 return false;
343 }
344 if (f2fs_sb_has_compression(sbi)) {
345 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
346 __func__, inode->i_ino, F2FS_FEATURE_COMPRESSION);
347 return false;
348 }
349 }
350
351 if (f2fs_sanity_check_inline_data(inode, node_page)) {
352 f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
353 __func__, inode->i_ino, inode->i_mode);
354 return false;
355 }
356
357 if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
358 f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
359 __func__, inode->i_ino, inode->i_mode);
360 return false;
361 }
362
363 if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
364 f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
365 __func__, inode->i_ino);
366 return false;
367 }
368
369 if (fi->i_xattr_nid && f2fs_check_nid_range(sbi, fi->i_xattr_nid)) {
370 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_xattr_nid: %u, run fsck to fix.",
371 __func__, inode->i_ino, fi->i_xattr_nid);
372 return false;
373 }
374
375 return true;
376 }
377
378 static void init_idisk_time(struct inode *inode)
379 {
380 struct f2fs_inode_info *fi = F2FS_I(inode);
381
382 fi->i_disk_time[0] = inode_get_atime(inode);
383 fi->i_disk_time[1] = inode_get_ctime(inode);
384 fi->i_disk_time[2] = inode_get_mtime(inode);
385 }
386
387 static int do_read_inode(struct inode *inode)
388 {
389 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
390 struct f2fs_inode_info *fi = F2FS_I(inode);
391 struct page *node_page;
392 struct f2fs_inode *ri;
393 projid_t i_projid;
394
395 /* Check if ino is within scope */
396 if (f2fs_check_nid_range(sbi, inode->i_ino))
397 return -EINVAL;
398
399 node_page = f2fs_get_node_page(sbi, inode->i_ino);
400 if (IS_ERR(node_page))
401 return PTR_ERR(node_page);
402
403 ri = F2FS_INODE(node_page);
404
405 inode->i_mode = le16_to_cpu(ri->i_mode);
406 i_uid_write(inode, le32_to_cpu(ri->i_uid));
407 i_gid_write(inode, le32_to_cpu(ri->i_gid));
408 set_nlink(inode, le32_to_cpu(ri->i_links));
409 inode->i_size = le64_to_cpu(ri->i_size);
410 inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
411
412 inode_set_atime(inode, le64_to_cpu(ri->i_atime),
413 le32_to_cpu(ri->i_atime_nsec));
414 inode_set_ctime(inode, le64_to_cpu(ri->i_ctime),
415 le32_to_cpu(ri->i_ctime_nsec));
416 inode_set_mtime(inode, le64_to_cpu(ri->i_mtime),
417 le32_to_cpu(ri->i_mtime_nsec));
418 inode->i_generation = le32_to_cpu(ri->i_generation);
419 if (S_ISDIR(inode->i_mode))
420 fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
421 else if (S_ISREG(inode->i_mode))
422 fi->i_gc_failures = le16_to_cpu(ri->i_gc_failures);
423 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
424 fi->i_flags = le32_to_cpu(ri->i_flags);
425 if (S_ISREG(inode->i_mode))
426 fi->i_flags &= ~F2FS_PROJINHERIT_FL;
427 bitmap_zero(fi->flags, FI_MAX);
428 fi->i_advise = ri->i_advise;
429 fi->i_pino = le32_to_cpu(ri->i_pino);
430 fi->i_dir_level = ri->i_dir_level;
431
432 get_inline_info(inode, ri);
433
434 fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
435 le16_to_cpu(ri->i_extra_isize) : 0;
436
437 if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
438 fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
439 } else if (f2fs_has_inline_xattr(inode) ||
440 f2fs_has_inline_dentry(inode)) {
441 fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
442 } else {
443
444 /*
445 * Previous inline data or directory always reserved 200 bytes
446 * in inode layout, even if inline_xattr is disabled. In order
447 * to keep inline_dentry's structure for backward compatibility,
448 * we get the space back only from inline_data.
449 */
450 fi->i_inline_xattr_size = 0;
451 }
452
453 if (!sanity_check_inode(inode, node_page)) {
454 f2fs_put_page(node_page, 1);
455 set_sbi_flag(sbi, SBI_NEED_FSCK);
456 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
457 return -EFSCORRUPTED;
458 }
459
460 /* check data exist */
461 if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
462 __recover_inline_status(inode, node_page);
463
464 /* try to recover cold bit for non-dir inode */
465 if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
466 f2fs_wait_on_page_writeback(node_page, NODE, true, true);
467 set_cold_node(node_page, false);
468 set_page_dirty(node_page);
469 }
470
471 /* get rdev by using inline_info */
472 __get_inode_rdev(inode, node_page);
473
474 if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
475 fi->last_disk_size = inode->i_size;
476
477 if (fi->i_flags & F2FS_PROJINHERIT_FL)
478 set_inode_flag(inode, FI_PROJ_INHERIT);
479
480 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
481 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
482 i_projid = (projid_t)le32_to_cpu(ri->i_projid);
483 else
484 i_projid = F2FS_DEF_PROJID;
485 fi->i_projid = make_kprojid(&init_user_ns, i_projid);
486
487 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
488 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
489 fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
490 fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
491 }
492
493 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
494 (fi->i_flags & F2FS_COMPR_FL)) {
495 if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
496 i_compress_flag)) {
497 unsigned short compress_flag;
498
499 atomic_set(&fi->i_compr_blocks,
500 le64_to_cpu(ri->i_compr_blocks));
501 fi->i_compress_algorithm = ri->i_compress_algorithm;
502 fi->i_log_cluster_size = ri->i_log_cluster_size;
503 compress_flag = le16_to_cpu(ri->i_compress_flag);
504 fi->i_compress_level = compress_flag >>
505 COMPRESS_LEVEL_OFFSET;
506 fi->i_compress_flag = compress_flag &
507 GENMASK(COMPRESS_LEVEL_OFFSET - 1, 0);
508 fi->i_cluster_size = BIT(fi->i_log_cluster_size);
509 set_inode_flag(inode, FI_COMPRESSED_FILE);
510 }
511 }
512
513 init_idisk_time(inode);
514
515 if (!sanity_check_extent_cache(inode, node_page)) {
516 f2fs_put_page(node_page, 1);
517 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
518 return -EFSCORRUPTED;
519 }
520
521 /* Need all the flag bits */
522 f2fs_init_read_extent_tree(inode, node_page);
523 f2fs_init_age_extent_tree(inode);
524
525 f2fs_put_page(node_page, 1);
526
527 stat_inc_inline_xattr(inode);
528 stat_inc_inline_inode(inode);
529 stat_inc_inline_dir(inode);
530 stat_inc_compr_inode(inode);
531 stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
532
533 return 0;
534 }
535
536 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
537 {
538 return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
539 ino == F2FS_COMPRESS_INO(sbi);
540 }
541
542 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
543 {
544 struct f2fs_sb_info *sbi = F2FS_SB(sb);
545 struct inode *inode;
546 int ret = 0;
547
548 inode = iget_locked(sb, ino);
549 if (!inode)
550 return ERR_PTR(-ENOMEM);
551
552 if (!(inode->i_state & I_NEW)) {
553 if (is_meta_ino(sbi, ino)) {
554 f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
555 set_sbi_flag(sbi, SBI_NEED_FSCK);
556 ret = -EFSCORRUPTED;
557 trace_f2fs_iget_exit(inode, ret);
558 iput(inode);
559 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
560 return ERR_PTR(ret);
561 }
562
563 trace_f2fs_iget(inode);
564 return inode;
565 }
566
567 if (is_meta_ino(sbi, ino))
568 goto make_now;
569
570 ret = do_read_inode(inode);
571 if (ret)
572 goto bad_inode;
573 make_now:
574 if (ino == F2FS_NODE_INO(sbi)) {
575 inode->i_mapping->a_ops = &f2fs_node_aops;
576 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
577 } else if (ino == F2FS_META_INO(sbi)) {
578 inode->i_mapping->a_ops = &f2fs_meta_aops;
579 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
580 } else if (ino == F2FS_COMPRESS_INO(sbi)) {
581 #ifdef CONFIG_F2FS_FS_COMPRESSION
582 inode->i_mapping->a_ops = &f2fs_compress_aops;
583 /*
584 * generic_error_remove_folio only truncates pages of regular
585 * inode
586 */
587 inode->i_mode |= S_IFREG;
588 #endif
589 mapping_set_gfp_mask(inode->i_mapping,
590 GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
591 } else if (S_ISREG(inode->i_mode)) {
592 inode->i_op = &f2fs_file_inode_operations;
593 inode->i_fop = &f2fs_file_operations;
594 inode->i_mapping->a_ops = &f2fs_dblock_aops;
595 } else if (S_ISDIR(inode->i_mode)) {
596 inode->i_op = &f2fs_dir_inode_operations;
597 inode->i_fop = &f2fs_dir_operations;
598 inode->i_mapping->a_ops = &f2fs_dblock_aops;
599 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
600 } else if (S_ISLNK(inode->i_mode)) {
601 if (file_is_encrypt(inode))
602 inode->i_op = &f2fs_encrypted_symlink_inode_operations;
603 else
604 inode->i_op = &f2fs_symlink_inode_operations;
605 inode_nohighmem(inode);
606 inode->i_mapping->a_ops = &f2fs_dblock_aops;
607 } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
608 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
609 inode->i_op = &f2fs_special_inode_operations;
610 init_special_inode(inode, inode->i_mode, inode->i_rdev);
611 } else {
612 ret = -EIO;
613 goto bad_inode;
614 }
615 f2fs_set_inode_flags(inode);
616
617 unlock_new_inode(inode);
618 trace_f2fs_iget(inode);
619 return inode;
620
621 bad_inode:
622 f2fs_inode_synced(inode);
623 iget_failed(inode);
624 trace_f2fs_iget_exit(inode, ret);
625 return ERR_PTR(ret);
626 }
627
628 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
629 {
630 struct inode *inode;
631 retry:
632 inode = f2fs_iget(sb, ino);
633 if (IS_ERR(inode)) {
634 if (PTR_ERR(inode) == -ENOMEM) {
635 memalloc_retry_wait(GFP_NOFS);
636 goto retry;
637 }
638 }
639 return inode;
640 }
641
642 void f2fs_update_inode(struct inode *inode, struct page *node_page)
643 {
644 struct f2fs_inode_info *fi = F2FS_I(inode);
645 struct f2fs_inode *ri;
646 struct extent_tree *et = fi->extent_tree[EX_READ];
647
648 f2fs_wait_on_page_writeback(node_page, NODE, true, true);
649 set_page_dirty(node_page);
650
651 f2fs_inode_synced(inode);
652
653 ri = F2FS_INODE(node_page);
654
655 ri->i_mode = cpu_to_le16(inode->i_mode);
656 ri->i_advise = fi->i_advise;
657 ri->i_uid = cpu_to_le32(i_uid_read(inode));
658 ri->i_gid = cpu_to_le32(i_gid_read(inode));
659 ri->i_links = cpu_to_le32(inode->i_nlink);
660 ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
661
662 if (!f2fs_is_atomic_file(inode) ||
663 is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
664 ri->i_size = cpu_to_le64(i_size_read(inode));
665
666 if (et) {
667 read_lock(&et->lock);
668 set_raw_read_extent(&et->largest, &ri->i_ext);
669 read_unlock(&et->lock);
670 } else {
671 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
672 }
673 set_raw_inline(inode, ri);
674
675 ri->i_atime = cpu_to_le64(inode_get_atime_sec(inode));
676 ri->i_ctime = cpu_to_le64(inode_get_ctime_sec(inode));
677 ri->i_mtime = cpu_to_le64(inode_get_mtime_sec(inode));
678 ri->i_atime_nsec = cpu_to_le32(inode_get_atime_nsec(inode));
679 ri->i_ctime_nsec = cpu_to_le32(inode_get_ctime_nsec(inode));
680 ri->i_mtime_nsec = cpu_to_le32(inode_get_mtime_nsec(inode));
681 if (S_ISDIR(inode->i_mode))
682 ri->i_current_depth = cpu_to_le32(fi->i_current_depth);
683 else if (S_ISREG(inode->i_mode))
684 ri->i_gc_failures = cpu_to_le16(fi->i_gc_failures);
685 ri->i_xattr_nid = cpu_to_le32(fi->i_xattr_nid);
686 ri->i_flags = cpu_to_le32(fi->i_flags);
687 ri->i_pino = cpu_to_le32(fi->i_pino);
688 ri->i_generation = cpu_to_le32(inode->i_generation);
689 ri->i_dir_level = fi->i_dir_level;
690
691 if (f2fs_has_extra_attr(inode)) {
692 ri->i_extra_isize = cpu_to_le16(fi->i_extra_isize);
693
694 if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
695 ri->i_inline_xattr_size =
696 cpu_to_le16(fi->i_inline_xattr_size);
697
698 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
699 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) {
700 projid_t i_projid;
701
702 i_projid = from_kprojid(&init_user_ns, fi->i_projid);
703 ri->i_projid = cpu_to_le32(i_projid);
704 }
705
706 if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
707 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
708 ri->i_crtime = cpu_to_le64(fi->i_crtime.tv_sec);
709 ri->i_crtime_nsec = cpu_to_le32(fi->i_crtime.tv_nsec);
710 }
711
712 if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
713 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
714 i_compress_flag)) {
715 unsigned short compress_flag;
716
717 ri->i_compr_blocks = cpu_to_le64(
718 atomic_read(&fi->i_compr_blocks));
719 ri->i_compress_algorithm = fi->i_compress_algorithm;
720 compress_flag = fi->i_compress_flag |
721 fi->i_compress_level <<
722 COMPRESS_LEVEL_OFFSET;
723 ri->i_compress_flag = cpu_to_le16(compress_flag);
724 ri->i_log_cluster_size = fi->i_log_cluster_size;
725 }
726 }
727
728 __set_inode_rdev(inode, node_page);
729
730 /* deleted inode */
731 if (inode->i_nlink == 0)
732 clear_page_private_inline(node_page);
733
734 init_idisk_time(inode);
735 #ifdef CONFIG_F2FS_CHECK_FS
736 f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
737 #endif
738 }
739
740 void f2fs_update_inode_page(struct inode *inode)
741 {
742 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
743 struct page *node_page;
744 int count = 0;
745 retry:
746 node_page = f2fs_get_node_page(sbi, inode->i_ino);
747 if (IS_ERR(node_page)) {
748 int err = PTR_ERR(node_page);
749
750 /* The node block was truncated. */
751 if (err == -ENOENT)
752 return;
753
754 if (err == -ENOMEM || ++count <= DEFAULT_RETRY_IO_COUNT)
755 goto retry;
756 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_UPDATE_INODE);
757 return;
758 }
759 f2fs_update_inode(inode, node_page);
760 f2fs_put_page(node_page, 1);
761 }
762
763 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
764 {
765 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
766
767 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
768 inode->i_ino == F2FS_META_INO(sbi))
769 return 0;
770
771 /*
772 * atime could be updated without dirtying f2fs inode in lazytime mode
773 */
774 if (f2fs_is_time_consistent(inode) &&
775 !is_inode_flag_set(inode, FI_DIRTY_INODE))
776 return 0;
777
778 if (!f2fs_is_checkpoint_ready(sbi))
779 return -ENOSPC;
780
781 /*
782 * We need to balance fs here to prevent from producing dirty node pages
783 * during the urgent cleaning time when running out of free sections.
784 */
785 f2fs_update_inode_page(inode);
786 if (wbc && wbc->nr_to_write)
787 f2fs_balance_fs(sbi, true);
788 return 0;
789 }
790
791 /*
792 * Called at the last iput() if i_nlink is zero
793 */
794 void f2fs_evict_inode(struct inode *inode)
795 {
796 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
797 struct f2fs_inode_info *fi = F2FS_I(inode);
798 nid_t xnid = fi->i_xattr_nid;
799 int err = 0;
800 bool freeze_protected = false;
801
802 f2fs_abort_atomic_write(inode, true);
803
804 if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
805 clear_inode_flag(fi->cow_inode, FI_COW_FILE);
806 F2FS_I(fi->cow_inode)->atomic_inode = NULL;
807 iput(fi->cow_inode);
808 fi->cow_inode = NULL;
809 }
810
811 trace_f2fs_evict_inode(inode);
812 truncate_inode_pages_final(&inode->i_data);
813
814 if ((inode->i_nlink || is_bad_inode(inode)) &&
815 test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
816 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
817
818 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
819 inode->i_ino == F2FS_META_INO(sbi) ||
820 inode->i_ino == F2FS_COMPRESS_INO(sbi))
821 goto out_clear;
822
823 f2fs_bug_on(sbi, get_dirty_pages(inode));
824 f2fs_remove_dirty_inode(inode);
825
826 f2fs_destroy_extent_tree(inode);
827
828 if (inode->i_nlink || is_bad_inode(inode))
829 goto no_delete;
830
831 err = f2fs_dquot_initialize(inode);
832 if (err) {
833 err = 0;
834 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
835 }
836
837 f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
838 f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
839 f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
840
841 if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING)) {
842 sb_start_intwrite(inode->i_sb);
843 freeze_protected = true;
844 }
845 set_inode_flag(inode, FI_NO_ALLOC);
846 i_size_write(inode, 0);
847 retry:
848 if (F2FS_HAS_BLOCKS(inode))
849 err = f2fs_truncate(inode);
850
851 if (time_to_inject(sbi, FAULT_EVICT_INODE))
852 err = -EIO;
853
854 if (!err) {
855 f2fs_lock_op(sbi);
856 err = f2fs_remove_inode_page(inode);
857 f2fs_unlock_op(sbi);
858 if (err == -ENOENT) {
859 err = 0;
860
861 /*
862 * in fuzzed image, another node may has the same
863 * block address as inode's, if it was truncated
864 * previously, truncation of inode node will fail.
865 */
866 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
867 f2fs_warn(F2FS_I_SB(inode),
868 "f2fs_evict_inode: inconsistent node id, ino:%lu",
869 inode->i_ino);
870 f2fs_inode_synced(inode);
871 set_sbi_flag(sbi, SBI_NEED_FSCK);
872 }
873 }
874 }
875
876 /* give more chances, if ENOMEM case */
877 if (err == -ENOMEM) {
878 err = 0;
879 goto retry;
880 }
881
882 if (err) {
883 f2fs_update_inode_page(inode);
884 if (dquot_initialize_needed(inode))
885 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
886 }
887 if (freeze_protected)
888 sb_end_intwrite(inode->i_sb);
889 no_delete:
890 dquot_drop(inode);
891
892 stat_dec_inline_xattr(inode);
893 stat_dec_inline_dir(inode);
894 stat_dec_inline_inode(inode);
895 stat_dec_compr_inode(inode);
896 stat_sub_compr_blocks(inode,
897 atomic_read(&fi->i_compr_blocks));
898
899 if (likely(!f2fs_cp_error(sbi) &&
900 !is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
901 f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
902 else
903 f2fs_inode_synced(inode);
904
905 /* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
906 if (inode->i_ino)
907 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
908 inode->i_ino);
909 if (xnid)
910 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
911 if (inode->i_nlink) {
912 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
913 f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
914 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
915 f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
916 }
917 if (is_inode_flag_set(inode, FI_FREE_NID)) {
918 f2fs_alloc_nid_failed(sbi, inode->i_ino);
919 clear_inode_flag(inode, FI_FREE_NID);
920 } else {
921 /*
922 * If xattr nid is corrupted, we can reach out error condition,
923 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
924 * In that case, f2fs_check_nid_range() is enough to give a clue.
925 */
926 }
927 out_clear:
928 fscrypt_put_encryption_info(inode);
929 fsverity_cleanup_inode(inode);
930 clear_inode(inode);
931 }
932
933 /* caller should call f2fs_lock_op() */
934 void f2fs_handle_failed_inode(struct inode *inode)
935 {
936 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
937 struct node_info ni;
938 int err;
939
940 /*
941 * clear nlink of inode in order to release resource of inode
942 * immediately.
943 */
944 clear_nlink(inode);
945
946 /*
947 * we must call this to avoid inode being remained as dirty, resulting
948 * in a panic when flushing dirty inodes in gdirty_list.
949 */
950 f2fs_update_inode_page(inode);
951 f2fs_inode_synced(inode);
952
953 /* don't make bad inode, since it becomes a regular file. */
954 unlock_new_inode(inode);
955
956 /*
957 * Note: we should add inode to orphan list before f2fs_unlock_op()
958 * so we can prevent losing this orphan when encoutering checkpoint
959 * and following suddenly power-off.
960 */
961 err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
962 if (err) {
963 set_sbi_flag(sbi, SBI_NEED_FSCK);
964 set_inode_flag(inode, FI_FREE_NID);
965 f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
966 goto out;
967 }
968
969 if (ni.blk_addr != NULL_ADDR) {
970 err = f2fs_acquire_orphan_inode(sbi);
971 if (err) {
972 set_sbi_flag(sbi, SBI_NEED_FSCK);
973 f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
974 } else {
975 f2fs_add_orphan_inode(inode);
976 }
977 f2fs_alloc_nid_done(sbi, inode->i_ino);
978 } else {
979 set_inode_flag(inode, FI_FREE_NID);
980 }
981
982 out:
983 f2fs_unlock_op(sbi);
984
985 /* iput will drop the inode object */
986 iput(inode);
987 }
988
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.