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/writeback.h>
11 #include <linux/sched/mm.h>
12 #include <linux/lz4.h>
13 #include <linux/zstd.h>
14
15 #include "f2fs.h"
16 #include "node.h"
17 #include "segment.h"
18 #include "xattr.h"
19
20 #include <trace/events/f2fs.h>
21
22 #ifdef CONFIG_F2FS_FS_COMPRESSION
23 extern const struct address_space_operations f2fs_compress_aops;
24 #endif
25
f2fs_mark_inode_dirty_sync(struct inode * inode,bool sync)26 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
27 {
28 if (is_inode_flag_set(inode, FI_NEW_INODE))
29 return;
30
31 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
32 return;
33
34 if (f2fs_inode_dirtied(inode, sync))
35 return;
36
37 /* only atomic file w/ FI_ATOMIC_COMMITTED can be set vfs dirty */
38 if (f2fs_is_atomic_file(inode) &&
39 !is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
40 return;
41
42 mark_inode_dirty_sync(inode);
43 }
44
f2fs_set_inode_flags(struct inode * inode)45 void f2fs_set_inode_flags(struct inode *inode)
46 {
47 unsigned int flags = F2FS_I(inode)->i_flags;
48 unsigned int new_fl = 0;
49
50 if (flags & F2FS_SYNC_FL)
51 new_fl |= S_SYNC;
52 if (flags & F2FS_APPEND_FL)
53 new_fl |= S_APPEND;
54 if (flags & F2FS_IMMUTABLE_FL)
55 new_fl |= S_IMMUTABLE;
56 if (flags & F2FS_NOATIME_FL)
57 new_fl |= S_NOATIME;
58 if (flags & F2FS_DIRSYNC_FL)
59 new_fl |= S_DIRSYNC;
60 if (file_is_encrypt(inode))
61 new_fl |= S_ENCRYPTED;
62 if (file_is_verity(inode))
63 new_fl |= S_VERITY;
64 if (flags & F2FS_CASEFOLD_FL)
65 new_fl |= S_CASEFOLD;
66 inode_set_flags(inode, new_fl,
67 S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|
68 S_ENCRYPTED|S_VERITY|S_CASEFOLD);
69 }
70
__get_inode_rdev(struct inode * inode,struct folio * node_folio)71 static void __get_inode_rdev(struct inode *inode, struct folio *node_folio)
72 {
73 __le32 *addr = get_dnode_addr(inode, node_folio);
74
75 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
76 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
77 if (addr[0])
78 inode->i_rdev = old_decode_dev(le32_to_cpu(addr[0]));
79 else
80 inode->i_rdev = new_decode_dev(le32_to_cpu(addr[1]));
81 }
82 }
83
__set_inode_rdev(struct inode * inode,struct folio * node_folio)84 static void __set_inode_rdev(struct inode *inode, struct folio *node_folio)
85 {
86 __le32 *addr = get_dnode_addr(inode, node_folio);
87
88 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
89 if (old_valid_dev(inode->i_rdev)) {
90 addr[0] = cpu_to_le32(old_encode_dev(inode->i_rdev));
91 addr[1] = 0;
92 } else {
93 addr[0] = 0;
94 addr[1] = cpu_to_le32(new_encode_dev(inode->i_rdev));
95 addr[2] = 0;
96 }
97 }
98 }
99
__recover_inline_status(struct inode * inode,struct folio * ifolio)100 static void __recover_inline_status(struct inode *inode, struct folio *ifolio)
101 {
102 void *inline_data = inline_data_addr(inode, ifolio);
103 __le32 *start = inline_data;
104 __le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
105
106 while (start < end) {
107 if (*start++) {
108 f2fs_folio_wait_writeback(ifolio, NODE, true, true);
109
110 set_inode_flag(inode, FI_DATA_EXIST);
111 set_raw_inline(inode, F2FS_INODE(ifolio));
112 folio_mark_dirty(ifolio);
113 return;
114 }
115 }
116 return;
117 }
118
119 static
f2fs_enable_inode_chksum(struct f2fs_sb_info * sbi,struct folio * folio)120 bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct folio *folio)
121 {
122 struct f2fs_inode *ri = &F2FS_NODE(folio)->i;
123
124 if (!f2fs_sb_has_inode_chksum(sbi))
125 return false;
126
127 if (!IS_INODE(folio) || !(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
f2fs_inode_chksum(struct f2fs_sb_info * sbi,struct folio * folio)137 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct folio *folio)
138 {
139 struct f2fs_node *node = F2FS_NODE(folio);
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->s_chksum_seed, (__u8 *)&ino, sizeof(ino));
149 chksum_seed = f2fs_chksum(chksum, (__u8 *)&gen, sizeof(gen));
150
151 chksum = f2fs_chksum(chksum_seed, (__u8 *)ri, offset);
152 chksum = f2fs_chksum(chksum, (__u8 *)&dummy_cs, cs_size);
153 offset += cs_size;
154 chksum = f2fs_chksum(chksum, (__u8 *)ri + offset,
155 F2FS_BLKSIZE - offset);
156 return chksum;
157 }
158
f2fs_inode_chksum_verify(struct f2fs_sb_info * sbi,struct folio * folio)159 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct folio *folio)
160 {
161 struct f2fs_inode *ri;
162 __u32 provided, calculated;
163
164 if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
165 return true;
166
167 #ifdef CONFIG_F2FS_CHECK_FS
168 if (!f2fs_enable_inode_chksum(sbi, folio))
169 #else
170 if (!f2fs_enable_inode_chksum(sbi, folio) ||
171 folio_test_dirty(folio) ||
172 folio_test_writeback(folio))
173 #endif
174 return true;
175
176 ri = &F2FS_NODE(folio)->i;
177 provided = le32_to_cpu(ri->i_inode_checksum);
178 calculated = f2fs_inode_chksum(sbi, folio);
179
180 if (provided != calculated)
181 f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
182 folio->index, ino_of_node(folio),
183 provided, calculated);
184
185 return provided == calculated;
186 }
187
f2fs_inode_chksum_set(struct f2fs_sb_info * sbi,struct folio * folio)188 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct folio *folio)
189 {
190 struct f2fs_inode *ri = &F2FS_NODE(folio)->i;
191
192 if (!f2fs_enable_inode_chksum(sbi, folio))
193 return;
194
195 ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, folio));
196 }
197
sanity_check_compress_inode(struct inode * inode,struct f2fs_inode * ri)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
sanity_check_inode(struct inode * inode,struct folio * node_folio)270 static bool sanity_check_inode(struct inode *inode, struct folio *node_folio)
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_folio);
275 unsigned long long iblocks;
276
277 iblocks = le64_to_cpu(F2FS_INODE(node_folio)->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_folio) != nid_of_node(node_folio)) {
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_folio), nid_of_node(node_folio));
288 return false;
289 }
290
291 if (ino_of_node(node_folio) == fi->i_xattr_nid) {
292 f2fs_warn(sbi, "%s: corrupted inode i_ino=%lx, xnid=%x, run fsck to fix.",
293 __func__, inode->i_ino, fi->i_xattr_nid);
294 return false;
295 }
296
297 if (f2fs_has_extra_attr(inode)) {
298 if (!f2fs_sb_has_extra_attr(sbi)) {
299 f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
300 __func__, inode->i_ino);
301 return false;
302 }
303 if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
304 fi->i_extra_isize < F2FS_MIN_EXTRA_ATTR_SIZE ||
305 fi->i_extra_isize % sizeof(__le32)) {
306 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
307 __func__, inode->i_ino, fi->i_extra_isize,
308 F2FS_TOTAL_EXTRA_ATTR_SIZE);
309 return false;
310 }
311 if (f2fs_sb_has_compression(sbi) &&
312 fi->i_flags & F2FS_COMPR_FL &&
313 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
314 i_compress_flag)) {
315 if (!sanity_check_compress_inode(inode, ri))
316 return false;
317 }
318 }
319
320 if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
321 f2fs_has_inline_xattr(inode) &&
322 (fi->i_inline_xattr_size < MIN_INLINE_XATTR_SIZE ||
323 fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
324 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, min: %zu, max: %lu",
325 __func__, inode->i_ino, fi->i_inline_xattr_size,
326 MIN_INLINE_XATTR_SIZE, MAX_INLINE_XATTR_SIZE);
327 return false;
328 }
329
330 if (!f2fs_sb_has_extra_attr(sbi)) {
331 if (f2fs_sb_has_project_quota(sbi)) {
332 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
333 __func__, inode->i_ino, F2FS_FEATURE_PRJQUOTA);
334 return false;
335 }
336 if (f2fs_sb_has_inode_chksum(sbi)) {
337 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
338 __func__, inode->i_ino, F2FS_FEATURE_INODE_CHKSUM);
339 return false;
340 }
341 if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
342 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
343 __func__, inode->i_ino, F2FS_FEATURE_FLEXIBLE_INLINE_XATTR);
344 return false;
345 }
346 if (f2fs_sb_has_inode_crtime(sbi)) {
347 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
348 __func__, inode->i_ino, F2FS_FEATURE_INODE_CRTIME);
349 return false;
350 }
351 if (f2fs_sb_has_compression(sbi)) {
352 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
353 __func__, inode->i_ino, F2FS_FEATURE_COMPRESSION);
354 return false;
355 }
356 }
357
358 if (f2fs_sanity_check_inline_data(inode, node_folio)) {
359 f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
360 __func__, inode->i_ino, inode->i_mode);
361 return false;
362 }
363
364 if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
365 f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
366 __func__, inode->i_ino, inode->i_mode);
367 return false;
368 }
369
370 if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
371 f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
372 __func__, inode->i_ino);
373 return false;
374 }
375
376 if (fi->i_xattr_nid && f2fs_check_nid_range(sbi, fi->i_xattr_nid)) {
377 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_xattr_nid: %u, run fsck to fix.",
378 __func__, inode->i_ino, fi->i_xattr_nid);
379 return false;
380 }
381
382 if (IS_DEVICE_ALIASING(inode)) {
383 if (!f2fs_sb_has_device_alias(sbi)) {
384 f2fs_warn(sbi, "%s: inode (ino=%lx) has device alias flag, but the feature is off",
385 __func__, inode->i_ino);
386 return false;
387 }
388 if (!f2fs_is_pinned_file(inode)) {
389 f2fs_warn(sbi, "%s: inode (ino=%lx) has device alias flag, but is not pinned",
390 __func__, inode->i_ino);
391 return false;
392 }
393 }
394
395 return true;
396 }
397
init_idisk_time(struct inode * inode)398 static void init_idisk_time(struct inode *inode)
399 {
400 struct f2fs_inode_info *fi = F2FS_I(inode);
401
402 fi->i_disk_time[0] = inode_get_atime(inode);
403 fi->i_disk_time[1] = inode_get_ctime(inode);
404 fi->i_disk_time[2] = inode_get_mtime(inode);
405 }
406
do_read_inode(struct inode * inode)407 static int do_read_inode(struct inode *inode)
408 {
409 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
410 struct f2fs_inode_info *fi = F2FS_I(inode);
411 struct folio *node_folio;
412 struct f2fs_inode *ri;
413 projid_t i_projid;
414
415 /* Check if ino is within scope */
416 if (f2fs_check_nid_range(sbi, inode->i_ino))
417 return -EINVAL;
418
419 node_folio = f2fs_get_inode_folio(sbi, inode->i_ino);
420 if (IS_ERR(node_folio))
421 return PTR_ERR(node_folio);
422
423 ri = F2FS_INODE(node_folio);
424
425 inode->i_mode = le16_to_cpu(ri->i_mode);
426 i_uid_write(inode, le32_to_cpu(ri->i_uid));
427 i_gid_write(inode, le32_to_cpu(ri->i_gid));
428 set_nlink(inode, le32_to_cpu(ri->i_links));
429 inode->i_size = le64_to_cpu(ri->i_size);
430 inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
431
432 inode_set_atime(inode, le64_to_cpu(ri->i_atime),
433 le32_to_cpu(ri->i_atime_nsec));
434 inode_set_ctime(inode, le64_to_cpu(ri->i_ctime),
435 le32_to_cpu(ri->i_ctime_nsec));
436 inode_set_mtime(inode, le64_to_cpu(ri->i_mtime),
437 le32_to_cpu(ri->i_mtime_nsec));
438 inode->i_generation = le32_to_cpu(ri->i_generation);
439 if (S_ISDIR(inode->i_mode))
440 fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
441 else if (S_ISREG(inode->i_mode))
442 fi->i_gc_failures = le16_to_cpu(ri->i_gc_failures);
443 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
444 fi->i_flags = le32_to_cpu(ri->i_flags);
445 if (S_ISREG(inode->i_mode))
446 fi->i_flags &= ~F2FS_PROJINHERIT_FL;
447 bitmap_zero(fi->flags, FI_MAX);
448 fi->i_advise = ri->i_advise;
449 fi->i_pino = le32_to_cpu(ri->i_pino);
450 fi->i_dir_level = ri->i_dir_level;
451
452 get_inline_info(inode, ri);
453
454 fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
455 le16_to_cpu(ri->i_extra_isize) : 0;
456
457 if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
458 fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
459 } else if (f2fs_has_inline_xattr(inode) ||
460 f2fs_has_inline_dentry(inode)) {
461 fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
462 } else {
463
464 /*
465 * Previous inline data or directory always reserved 200 bytes
466 * in inode layout, even if inline_xattr is disabled. In order
467 * to keep inline_dentry's structure for backward compatibility,
468 * we get the space back only from inline_data.
469 */
470 fi->i_inline_xattr_size = 0;
471 }
472
473 if (!sanity_check_inode(inode, node_folio)) {
474 f2fs_folio_put(node_folio, true);
475 set_sbi_flag(sbi, SBI_NEED_FSCK);
476 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
477 return -EFSCORRUPTED;
478 }
479
480 /* check data exist */
481 if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
482 __recover_inline_status(inode, node_folio);
483
484 /* try to recover cold bit for non-dir inode */
485 if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_folio)) {
486 f2fs_folio_wait_writeback(node_folio, NODE, true, true);
487 set_cold_node(node_folio, false);
488 folio_mark_dirty(node_folio);
489 }
490
491 /* get rdev by using inline_info */
492 __get_inode_rdev(inode, node_folio);
493
494 if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
495 fi->last_disk_size = inode->i_size;
496
497 if (fi->i_flags & F2FS_PROJINHERIT_FL)
498 set_inode_flag(inode, FI_PROJ_INHERIT);
499
500 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
501 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
502 i_projid = (projid_t)le32_to_cpu(ri->i_projid);
503 else
504 i_projid = F2FS_DEF_PROJID;
505 fi->i_projid = make_kprojid(&init_user_ns, i_projid);
506
507 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
508 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
509 fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
510 fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
511 }
512
513 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
514 (fi->i_flags & F2FS_COMPR_FL)) {
515 if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
516 i_compress_flag)) {
517 unsigned short compress_flag;
518
519 atomic_set(&fi->i_compr_blocks,
520 le64_to_cpu(ri->i_compr_blocks));
521 fi->i_compress_algorithm = ri->i_compress_algorithm;
522 fi->i_log_cluster_size = ri->i_log_cluster_size;
523 compress_flag = le16_to_cpu(ri->i_compress_flag);
524 fi->i_compress_level = compress_flag >>
525 COMPRESS_LEVEL_OFFSET;
526 fi->i_compress_flag = compress_flag &
527 GENMASK(COMPRESS_LEVEL_OFFSET - 1, 0);
528 fi->i_cluster_size = BIT(fi->i_log_cluster_size);
529 set_inode_flag(inode, FI_COMPRESSED_FILE);
530 }
531 }
532
533 init_idisk_time(inode);
534
535 if (!sanity_check_extent_cache(inode, node_folio)) {
536 f2fs_folio_put(node_folio, true);
537 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
538 return -EFSCORRUPTED;
539 }
540
541 /* Need all the flag bits */
542 f2fs_init_read_extent_tree(inode, node_folio);
543 f2fs_init_age_extent_tree(inode);
544
545 f2fs_folio_put(node_folio, true);
546
547 stat_inc_inline_xattr(inode);
548 stat_inc_inline_inode(inode);
549 stat_inc_inline_dir(inode);
550 stat_inc_compr_inode(inode);
551 stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
552
553 return 0;
554 }
555
is_meta_ino(struct f2fs_sb_info * sbi,unsigned int ino)556 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
557 {
558 return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
559 ino == F2FS_COMPRESS_INO(sbi);
560 }
561
f2fs_iget(struct super_block * sb,unsigned long ino)562 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
563 {
564 struct f2fs_sb_info *sbi = F2FS_SB(sb);
565 struct inode *inode;
566 int ret = 0;
567
568 inode = iget_locked(sb, ino);
569 if (!inode)
570 return ERR_PTR(-ENOMEM);
571
572 if (!(inode->i_state & I_NEW)) {
573 if (is_meta_ino(sbi, ino)) {
574 f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
575 set_sbi_flag(sbi, SBI_NEED_FSCK);
576 ret = -EFSCORRUPTED;
577 trace_f2fs_iget_exit(inode, ret);
578 iput(inode);
579 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
580 return ERR_PTR(ret);
581 }
582
583 trace_f2fs_iget(inode);
584 return inode;
585 }
586
587 if (is_meta_ino(sbi, ino))
588 goto make_now;
589
590 ret = do_read_inode(inode);
591 if (ret)
592 goto bad_inode;
593 make_now:
594 if (ino == F2FS_NODE_INO(sbi)) {
595 inode->i_mapping->a_ops = &f2fs_node_aops;
596 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
597 } else if (ino == F2FS_META_INO(sbi)) {
598 inode->i_mapping->a_ops = &f2fs_meta_aops;
599 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
600 } else if (ino == F2FS_COMPRESS_INO(sbi)) {
601 #ifdef CONFIG_F2FS_FS_COMPRESSION
602 inode->i_mapping->a_ops = &f2fs_compress_aops;
603 /*
604 * generic_error_remove_folio only truncates pages of regular
605 * inode
606 */
607 inode->i_mode |= S_IFREG;
608 #endif
609 mapping_set_gfp_mask(inode->i_mapping,
610 GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
611 } else if (S_ISREG(inode->i_mode)) {
612 inode->i_op = &f2fs_file_inode_operations;
613 inode->i_fop = &f2fs_file_operations;
614 inode->i_mapping->a_ops = &f2fs_dblock_aops;
615 } else if (S_ISDIR(inode->i_mode)) {
616 inode->i_op = &f2fs_dir_inode_operations;
617 inode->i_fop = &f2fs_dir_operations;
618 inode->i_mapping->a_ops = &f2fs_dblock_aops;
619 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
620 } else if (S_ISLNK(inode->i_mode)) {
621 if (file_is_encrypt(inode))
622 inode->i_op = &f2fs_encrypted_symlink_inode_operations;
623 else
624 inode->i_op = &f2fs_symlink_inode_operations;
625 inode_nohighmem(inode);
626 inode->i_mapping->a_ops = &f2fs_dblock_aops;
627 } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
628 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
629 inode->i_op = &f2fs_special_inode_operations;
630 init_special_inode(inode, inode->i_mode, inode->i_rdev);
631 } else {
632 ret = -EIO;
633 goto bad_inode;
634 }
635 f2fs_set_inode_flags(inode);
636
637 unlock_new_inode(inode);
638 trace_f2fs_iget(inode);
639 return inode;
640
641 bad_inode:
642 f2fs_inode_synced(inode);
643 iget_failed(inode);
644 trace_f2fs_iget_exit(inode, ret);
645 return ERR_PTR(ret);
646 }
647
f2fs_iget_retry(struct super_block * sb,unsigned long ino)648 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
649 {
650 struct inode *inode;
651 retry:
652 inode = f2fs_iget(sb, ino);
653 if (IS_ERR(inode)) {
654 if (PTR_ERR(inode) == -ENOMEM) {
655 memalloc_retry_wait(GFP_NOFS);
656 goto retry;
657 }
658 }
659 return inode;
660 }
661
f2fs_update_inode(struct inode * inode,struct folio * node_folio)662 void f2fs_update_inode(struct inode *inode, struct folio *node_folio)
663 {
664 struct f2fs_inode_info *fi = F2FS_I(inode);
665 struct f2fs_inode *ri;
666 struct extent_tree *et = fi->extent_tree[EX_READ];
667
668 f2fs_folio_wait_writeback(node_folio, NODE, true, true);
669 folio_mark_dirty(node_folio);
670
671 f2fs_inode_synced(inode);
672
673 ri = F2FS_INODE(node_folio);
674
675 ri->i_mode = cpu_to_le16(inode->i_mode);
676 ri->i_advise = fi->i_advise;
677 ri->i_uid = cpu_to_le32(i_uid_read(inode));
678 ri->i_gid = cpu_to_le32(i_gid_read(inode));
679 ri->i_links = cpu_to_le32(inode->i_nlink);
680 ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
681
682 if (!f2fs_is_atomic_file(inode) ||
683 is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
684 ri->i_size = cpu_to_le64(i_size_read(inode));
685
686 if (et) {
687 read_lock(&et->lock);
688 set_raw_read_extent(&et->largest, &ri->i_ext);
689 read_unlock(&et->lock);
690 } else {
691 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
692 }
693 set_raw_inline(inode, ri);
694
695 ri->i_atime = cpu_to_le64(inode_get_atime_sec(inode));
696 ri->i_ctime = cpu_to_le64(inode_get_ctime_sec(inode));
697 ri->i_mtime = cpu_to_le64(inode_get_mtime_sec(inode));
698 ri->i_atime_nsec = cpu_to_le32(inode_get_atime_nsec(inode));
699 ri->i_ctime_nsec = cpu_to_le32(inode_get_ctime_nsec(inode));
700 ri->i_mtime_nsec = cpu_to_le32(inode_get_mtime_nsec(inode));
701 if (S_ISDIR(inode->i_mode))
702 ri->i_current_depth = cpu_to_le32(fi->i_current_depth);
703 else if (S_ISREG(inode->i_mode))
704 ri->i_gc_failures = cpu_to_le16(fi->i_gc_failures);
705 ri->i_xattr_nid = cpu_to_le32(fi->i_xattr_nid);
706 ri->i_flags = cpu_to_le32(fi->i_flags);
707 ri->i_pino = cpu_to_le32(fi->i_pino);
708 ri->i_generation = cpu_to_le32(inode->i_generation);
709 ri->i_dir_level = fi->i_dir_level;
710
711 if (f2fs_has_extra_attr(inode)) {
712 ri->i_extra_isize = cpu_to_le16(fi->i_extra_isize);
713
714 if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
715 ri->i_inline_xattr_size =
716 cpu_to_le16(fi->i_inline_xattr_size);
717
718 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
719 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) {
720 projid_t i_projid;
721
722 i_projid = from_kprojid(&init_user_ns, fi->i_projid);
723 ri->i_projid = cpu_to_le32(i_projid);
724 }
725
726 if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
727 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
728 ri->i_crtime = cpu_to_le64(fi->i_crtime.tv_sec);
729 ri->i_crtime_nsec = cpu_to_le32(fi->i_crtime.tv_nsec);
730 }
731
732 if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
733 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
734 i_compress_flag)) {
735 unsigned short compress_flag;
736
737 ri->i_compr_blocks = cpu_to_le64(
738 atomic_read(&fi->i_compr_blocks));
739 ri->i_compress_algorithm = fi->i_compress_algorithm;
740 compress_flag = fi->i_compress_flag |
741 fi->i_compress_level <<
742 COMPRESS_LEVEL_OFFSET;
743 ri->i_compress_flag = cpu_to_le16(compress_flag);
744 ri->i_log_cluster_size = fi->i_log_cluster_size;
745 }
746 }
747
748 __set_inode_rdev(inode, node_folio);
749
750 /* deleted inode */
751 if (inode->i_nlink == 0)
752 folio_clear_f2fs_inline(node_folio);
753
754 init_idisk_time(inode);
755 #ifdef CONFIG_F2FS_CHECK_FS
756 f2fs_inode_chksum_set(F2FS_I_SB(inode), node_folio);
757 #endif
758 }
759
f2fs_update_inode_page(struct inode * inode)760 void f2fs_update_inode_page(struct inode *inode)
761 {
762 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
763 struct folio *node_folio;
764 int count = 0;
765 retry:
766 node_folio = f2fs_get_inode_folio(sbi, inode->i_ino);
767 if (IS_ERR(node_folio)) {
768 int err = PTR_ERR(node_folio);
769
770 /* The node block was truncated. */
771 if (err == -ENOENT)
772 return;
773
774 if (err == -EFSCORRUPTED)
775 goto stop_checkpoint;
776
777 if (err == -ENOMEM || ++count <= DEFAULT_RETRY_IO_COUNT)
778 goto retry;
779 stop_checkpoint:
780 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_UPDATE_INODE);
781 return;
782 }
783 f2fs_update_inode(inode, node_folio);
784 f2fs_folio_put(node_folio, true);
785 }
786
f2fs_write_inode(struct inode * inode,struct writeback_control * wbc)787 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
788 {
789 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
790
791 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
792 inode->i_ino == F2FS_META_INO(sbi))
793 return 0;
794
795 /*
796 * atime could be updated without dirtying f2fs inode in lazytime mode
797 */
798 if (f2fs_is_time_consistent(inode) &&
799 !is_inode_flag_set(inode, FI_DIRTY_INODE))
800 return 0;
801
802 /*
803 * no need to update inode page, ultimately f2fs_evict_inode() will
804 * clear dirty status of inode.
805 */
806 if (f2fs_cp_error(sbi))
807 return -EIO;
808
809 if (!f2fs_is_checkpoint_ready(sbi)) {
810 f2fs_mark_inode_dirty_sync(inode, true);
811 return -ENOSPC;
812 }
813
814 /*
815 * We need to balance fs here to prevent from producing dirty node pages
816 * during the urgent cleaning time when running out of free sections.
817 */
818 f2fs_update_inode_page(inode);
819 if (wbc && wbc->nr_to_write)
820 f2fs_balance_fs(sbi, true);
821 return 0;
822 }
823
f2fs_remove_donate_inode(struct inode * inode)824 void f2fs_remove_donate_inode(struct inode *inode)
825 {
826 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
827
828 if (list_empty(&F2FS_I(inode)->gdonate_list))
829 return;
830
831 spin_lock(&sbi->inode_lock[DONATE_INODE]);
832 list_del_init(&F2FS_I(inode)->gdonate_list);
833 sbi->donate_files--;
834 spin_unlock(&sbi->inode_lock[DONATE_INODE]);
835 }
836
837 /*
838 * Called at the last iput() if i_nlink is zero
839 */
f2fs_evict_inode(struct inode * inode)840 void f2fs_evict_inode(struct inode *inode)
841 {
842 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
843 struct f2fs_inode_info *fi = F2FS_I(inode);
844 nid_t xnid = fi->i_xattr_nid;
845 int err = 0;
846 bool freeze_protected = false;
847
848 f2fs_abort_atomic_write(inode, true);
849
850 if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
851 clear_inode_flag(fi->cow_inode, FI_COW_FILE);
852 F2FS_I(fi->cow_inode)->atomic_inode = NULL;
853 iput(fi->cow_inode);
854 fi->cow_inode = NULL;
855 }
856
857 trace_f2fs_evict_inode(inode);
858 truncate_inode_pages_final(&inode->i_data);
859
860 if ((inode->i_nlink || is_bad_inode(inode)) &&
861 test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
862 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
863
864 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
865 inode->i_ino == F2FS_META_INO(sbi) ||
866 inode->i_ino == F2FS_COMPRESS_INO(sbi))
867 goto out_clear;
868
869 f2fs_bug_on(sbi, get_dirty_pages(inode));
870 f2fs_remove_dirty_inode(inode);
871 f2fs_remove_donate_inode(inode);
872
873 if (!IS_DEVICE_ALIASING(inode))
874 f2fs_destroy_extent_tree(inode);
875
876 if (inode->i_nlink || is_bad_inode(inode))
877 goto no_delete;
878
879 err = f2fs_dquot_initialize(inode);
880 if (err) {
881 err = 0;
882 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
883 }
884
885 f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
886 f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
887 f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
888
889 if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING)) {
890 sb_start_intwrite(inode->i_sb);
891 freeze_protected = true;
892 }
893 set_inode_flag(inode, FI_NO_ALLOC);
894 i_size_write(inode, 0);
895 retry:
896 if (F2FS_HAS_BLOCKS(inode))
897 err = f2fs_truncate(inode);
898
899 if (time_to_inject(sbi, FAULT_EVICT_INODE))
900 err = -EIO;
901
902 if (!err) {
903 f2fs_lock_op(sbi);
904 err = f2fs_remove_inode_page(inode);
905 f2fs_unlock_op(sbi);
906 if (err == -ENOENT) {
907 err = 0;
908
909 /*
910 * in fuzzed image, another node may has the same
911 * block address as inode's, if it was truncated
912 * previously, truncation of inode node will fail.
913 */
914 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
915 f2fs_warn(F2FS_I_SB(inode),
916 "f2fs_evict_inode: inconsistent node id, ino:%lu",
917 inode->i_ino);
918 f2fs_inode_synced(inode);
919 set_sbi_flag(sbi, SBI_NEED_FSCK);
920 }
921 }
922 }
923
924 /* give more chances, if ENOMEM case */
925 if (err == -ENOMEM) {
926 err = 0;
927 goto retry;
928 }
929
930 if (IS_DEVICE_ALIASING(inode))
931 f2fs_destroy_extent_tree(inode);
932
933 if (err) {
934 f2fs_update_inode_page(inode);
935 if (dquot_initialize_needed(inode))
936 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
937
938 /*
939 * If both f2fs_truncate() and f2fs_update_inode_page() failed
940 * due to fuzzed corrupted inode, call f2fs_inode_synced() to
941 * avoid triggering later f2fs_bug_on().
942 */
943 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
944 f2fs_warn(sbi,
945 "f2fs_evict_inode: inode is dirty, ino:%lu",
946 inode->i_ino);
947 f2fs_inode_synced(inode);
948 set_sbi_flag(sbi, SBI_NEED_FSCK);
949 }
950 }
951 if (freeze_protected)
952 sb_end_intwrite(inode->i_sb);
953 no_delete:
954 dquot_drop(inode);
955
956 stat_dec_inline_xattr(inode);
957 stat_dec_inline_dir(inode);
958 stat_dec_inline_inode(inode);
959 stat_dec_compr_inode(inode);
960 stat_sub_compr_blocks(inode,
961 atomic_read(&fi->i_compr_blocks));
962
963 if (likely(!f2fs_cp_error(sbi) &&
964 !is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
965 f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
966
967 /*
968 * anyway, it needs to remove the inode from sbi->inode_list[DIRTY_META]
969 * list to avoid UAF in f2fs_sync_inode_meta() during checkpoint.
970 */
971 f2fs_inode_synced(inode);
972
973 /* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
974 if (inode->i_ino)
975 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
976 inode->i_ino);
977 if (xnid)
978 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
979 if (inode->i_nlink) {
980 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
981 f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
982 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
983 f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
984 }
985 if (is_inode_flag_set(inode, FI_FREE_NID)) {
986 f2fs_alloc_nid_failed(sbi, inode->i_ino);
987 clear_inode_flag(inode, FI_FREE_NID);
988 } else {
989 /*
990 * If xattr nid is corrupted, we can reach out error condition,
991 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
992 * In that case, f2fs_check_nid_range() is enough to give a clue.
993 */
994 }
995 out_clear:
996 fscrypt_put_encryption_info(inode);
997 fsverity_cleanup_inode(inode);
998 clear_inode(inode);
999 }
1000
1001 /* caller should call f2fs_lock_op() */
f2fs_handle_failed_inode(struct inode * inode)1002 void f2fs_handle_failed_inode(struct inode *inode)
1003 {
1004 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1005 struct node_info ni;
1006 int err;
1007
1008 /*
1009 * clear nlink of inode in order to release resource of inode
1010 * immediately.
1011 */
1012 clear_nlink(inode);
1013
1014 /*
1015 * we must call this to avoid inode being remained as dirty, resulting
1016 * in a panic when flushing dirty inodes in gdirty_list.
1017 */
1018 f2fs_update_inode_page(inode);
1019 f2fs_inode_synced(inode);
1020
1021 /* don't make bad inode, since it becomes a regular file. */
1022 unlock_new_inode(inode);
1023
1024 /*
1025 * Note: we should add inode to orphan list before f2fs_unlock_op()
1026 * so we can prevent losing this orphan when encoutering checkpoint
1027 * and following suddenly power-off.
1028 */
1029 err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
1030 if (err) {
1031 set_sbi_flag(sbi, SBI_NEED_FSCK);
1032 set_inode_flag(inode, FI_FREE_NID);
1033 f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
1034 goto out;
1035 }
1036
1037 if (ni.blk_addr != NULL_ADDR) {
1038 err = f2fs_acquire_orphan_inode(sbi);
1039 if (err) {
1040 set_sbi_flag(sbi, SBI_NEED_FSCK);
1041 f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
1042 } else {
1043 f2fs_add_orphan_inode(inode);
1044 }
1045 f2fs_alloc_nid_done(sbi, inode->i_ino);
1046 } else {
1047 set_inode_flag(inode, FI_FREE_NID);
1048 }
1049
1050 out:
1051 f2fs_unlock_op(sbi);
1052
1053 /* iput will drop the inode object */
1054 iput(inode);
1055 }
1056