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 (S_ISDIR(inode->i_mode) && unlikely(inode->i_nlink == 1)) {
298 f2fs_warn(sbi, "%s: directory inode (ino=%lx) has a single i_nlink",
299 __func__, inode->i_ino);
300 return false;
301 }
302
303 if (f2fs_has_extra_attr(inode)) {
304 if (!f2fs_sb_has_extra_attr(sbi)) {
305 f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
306 __func__, inode->i_ino);
307 return false;
308 }
309 if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
310 fi->i_extra_isize < F2FS_MIN_EXTRA_ATTR_SIZE ||
311 fi->i_extra_isize % sizeof(__le32)) {
312 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
313 __func__, inode->i_ino, fi->i_extra_isize,
314 F2FS_TOTAL_EXTRA_ATTR_SIZE);
315 return false;
316 }
317 if (f2fs_sb_has_compression(sbi) &&
318 fi->i_flags & F2FS_COMPR_FL &&
319 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
320 i_compress_flag)) {
321 if (!sanity_check_compress_inode(inode, ri))
322 return false;
323 }
324 }
325
326 if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
327 f2fs_has_inline_xattr(inode) &&
328 (fi->i_inline_xattr_size < MIN_INLINE_XATTR_SIZE ||
329 fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
330 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, min: %zu, max: %lu",
331 __func__, inode->i_ino, fi->i_inline_xattr_size,
332 MIN_INLINE_XATTR_SIZE, MAX_INLINE_XATTR_SIZE);
333 return false;
334 }
335
336 if (!f2fs_sb_has_extra_attr(sbi)) {
337 if (f2fs_sb_has_project_quota(sbi)) {
338 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
339 __func__, inode->i_ino, F2FS_FEATURE_PRJQUOTA);
340 return false;
341 }
342 if (f2fs_sb_has_inode_chksum(sbi)) {
343 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
344 __func__, inode->i_ino, F2FS_FEATURE_INODE_CHKSUM);
345 return false;
346 }
347 if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
348 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
349 __func__, inode->i_ino, F2FS_FEATURE_FLEXIBLE_INLINE_XATTR);
350 return false;
351 }
352 if (f2fs_sb_has_inode_crtime(sbi)) {
353 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
354 __func__, inode->i_ino, F2FS_FEATURE_INODE_CRTIME);
355 return false;
356 }
357 if (f2fs_sb_has_compression(sbi)) {
358 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
359 __func__, inode->i_ino, F2FS_FEATURE_COMPRESSION);
360 return false;
361 }
362 }
363
364 if (f2fs_sanity_check_inline_data(inode, node_folio)) {
365 f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
366 __func__, inode->i_ino, inode->i_mode);
367 return false;
368 }
369
370 if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
371 f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
372 __func__, inode->i_ino, inode->i_mode);
373 return false;
374 }
375
376 if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
377 f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
378 __func__, inode->i_ino);
379 return false;
380 }
381
382 if (fi->i_xattr_nid && f2fs_check_nid_range(sbi, fi->i_xattr_nid)) {
383 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_xattr_nid: %u, run fsck to fix.",
384 __func__, inode->i_ino, fi->i_xattr_nid);
385 return false;
386 }
387
388 if (IS_DEVICE_ALIASING(inode)) {
389 if (!f2fs_sb_has_device_alias(sbi)) {
390 f2fs_warn(sbi, "%s: inode (ino=%lx) has device alias flag, but the feature is off",
391 __func__, inode->i_ino);
392 return false;
393 }
394 if (!f2fs_is_pinned_file(inode)) {
395 f2fs_warn(sbi, "%s: inode (ino=%lx) has device alias flag, but is not pinned",
396 __func__, inode->i_ino);
397 return false;
398 }
399 }
400
401 return true;
402 }
403
init_idisk_time(struct inode * inode)404 static void init_idisk_time(struct inode *inode)
405 {
406 struct f2fs_inode_info *fi = F2FS_I(inode);
407
408 fi->i_disk_time[0] = inode_get_atime(inode);
409 fi->i_disk_time[1] = inode_get_ctime(inode);
410 fi->i_disk_time[2] = inode_get_mtime(inode);
411 }
412
do_read_inode(struct inode * inode)413 static int do_read_inode(struct inode *inode)
414 {
415 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
416 struct f2fs_inode_info *fi = F2FS_I(inode);
417 struct folio *node_folio;
418 struct f2fs_inode *ri;
419 projid_t i_projid;
420
421 /* Check if ino is within scope */
422 if (f2fs_check_nid_range(sbi, inode->i_ino))
423 return -EINVAL;
424
425 node_folio = f2fs_get_inode_folio(sbi, inode->i_ino);
426 if (IS_ERR(node_folio))
427 return PTR_ERR(node_folio);
428
429 ri = F2FS_INODE(node_folio);
430
431 inode->i_mode = le16_to_cpu(ri->i_mode);
432 i_uid_write(inode, le32_to_cpu(ri->i_uid));
433 i_gid_write(inode, le32_to_cpu(ri->i_gid));
434 set_nlink(inode, le32_to_cpu(ri->i_links));
435 inode->i_size = le64_to_cpu(ri->i_size);
436 inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
437
438 inode_set_atime(inode, le64_to_cpu(ri->i_atime),
439 le32_to_cpu(ri->i_atime_nsec));
440 inode_set_ctime(inode, le64_to_cpu(ri->i_ctime),
441 le32_to_cpu(ri->i_ctime_nsec));
442 inode_set_mtime(inode, le64_to_cpu(ri->i_mtime),
443 le32_to_cpu(ri->i_mtime_nsec));
444 inode->i_generation = le32_to_cpu(ri->i_generation);
445 if (S_ISDIR(inode->i_mode))
446 fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
447 else if (S_ISREG(inode->i_mode))
448 fi->i_gc_failures = le16_to_cpu(ri->i_gc_failures);
449 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
450 fi->i_flags = le32_to_cpu(ri->i_flags);
451 if (S_ISREG(inode->i_mode))
452 fi->i_flags &= ~F2FS_PROJINHERIT_FL;
453 bitmap_zero(fi->flags, FI_MAX);
454 fi->i_advise = ri->i_advise;
455 fi->i_pino = le32_to_cpu(ri->i_pino);
456 fi->i_dir_level = ri->i_dir_level;
457
458 get_inline_info(inode, ri);
459
460 fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
461 le16_to_cpu(ri->i_extra_isize) : 0;
462
463 if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
464 fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
465 } else if (f2fs_has_inline_xattr(inode) ||
466 f2fs_has_inline_dentry(inode)) {
467 fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
468 } else {
469
470 /*
471 * Previous inline data or directory always reserved 200 bytes
472 * in inode layout, even if inline_xattr is disabled. In order
473 * to keep inline_dentry's structure for backward compatibility,
474 * we get the space back only from inline_data.
475 */
476 fi->i_inline_xattr_size = 0;
477 }
478
479 if (!sanity_check_inode(inode, node_folio)) {
480 f2fs_folio_put(node_folio, true);
481 set_sbi_flag(sbi, SBI_NEED_FSCK);
482 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
483 return -EFSCORRUPTED;
484 }
485
486 /* check data exist */
487 if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
488 __recover_inline_status(inode, node_folio);
489
490 /* try to recover cold bit for non-dir inode */
491 if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_folio)) {
492 f2fs_folio_wait_writeback(node_folio, NODE, true, true);
493 set_cold_node(node_folio, false);
494 folio_mark_dirty(node_folio);
495 }
496
497 /* get rdev by using inline_info */
498 __get_inode_rdev(inode, node_folio);
499
500 if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
501 fi->last_disk_size = inode->i_size;
502
503 if (fi->i_flags & F2FS_PROJINHERIT_FL)
504 set_inode_flag(inode, FI_PROJ_INHERIT);
505
506 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
507 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
508 i_projid = (projid_t)le32_to_cpu(ri->i_projid);
509 else
510 i_projid = F2FS_DEF_PROJID;
511 fi->i_projid = make_kprojid(&init_user_ns, i_projid);
512
513 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
514 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
515 fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
516 fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
517 }
518
519 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
520 (fi->i_flags & F2FS_COMPR_FL)) {
521 if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
522 i_compress_flag)) {
523 unsigned short compress_flag;
524
525 atomic_set(&fi->i_compr_blocks,
526 le64_to_cpu(ri->i_compr_blocks));
527 fi->i_compress_algorithm = ri->i_compress_algorithm;
528 fi->i_log_cluster_size = ri->i_log_cluster_size;
529 compress_flag = le16_to_cpu(ri->i_compress_flag);
530 fi->i_compress_level = compress_flag >>
531 COMPRESS_LEVEL_OFFSET;
532 fi->i_compress_flag = compress_flag &
533 GENMASK(COMPRESS_LEVEL_OFFSET - 1, 0);
534 fi->i_cluster_size = BIT(fi->i_log_cluster_size);
535 set_inode_flag(inode, FI_COMPRESSED_FILE);
536 }
537 }
538
539 init_idisk_time(inode);
540
541 if (!sanity_check_extent_cache(inode, node_folio)) {
542 f2fs_folio_put(node_folio, true);
543 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
544 return -EFSCORRUPTED;
545 }
546
547 /* Need all the flag bits */
548 f2fs_init_read_extent_tree(inode, node_folio);
549 f2fs_init_age_extent_tree(inode);
550
551 f2fs_folio_put(node_folio, true);
552
553 stat_inc_inline_xattr(inode);
554 stat_inc_inline_inode(inode);
555 stat_inc_inline_dir(inode);
556 stat_inc_compr_inode(inode);
557 stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
558
559 return 0;
560 }
561
is_meta_ino(struct f2fs_sb_info * sbi,unsigned int ino)562 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
563 {
564 return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
565 ino == F2FS_COMPRESS_INO(sbi);
566 }
567
f2fs_iget(struct super_block * sb,unsigned long ino)568 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
569 {
570 struct f2fs_sb_info *sbi = F2FS_SB(sb);
571 struct inode *inode;
572 int ret = 0;
573
574 inode = iget_locked(sb, ino);
575 if (!inode)
576 return ERR_PTR(-ENOMEM);
577
578 if (!(inode_state_read_once(inode) & I_NEW)) {
579 if (is_meta_ino(sbi, ino)) {
580 f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
581 set_sbi_flag(sbi, SBI_NEED_FSCK);
582 ret = -EFSCORRUPTED;
583 trace_f2fs_iget_exit(inode, ret);
584 iput(inode);
585 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
586 return ERR_PTR(ret);
587 }
588
589 trace_f2fs_iget(inode);
590 return inode;
591 }
592
593 if (is_meta_ino(sbi, ino))
594 goto make_now;
595
596 ret = do_read_inode(inode);
597 if (ret)
598 goto bad_inode;
599 make_now:
600 if (ino == F2FS_NODE_INO(sbi)) {
601 inode->i_mapping->a_ops = &f2fs_node_aops;
602 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
603 } else if (ino == F2FS_META_INO(sbi)) {
604 inode->i_mapping->a_ops = &f2fs_meta_aops;
605 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
606 } else if (ino == F2FS_COMPRESS_INO(sbi)) {
607 #ifdef CONFIG_F2FS_FS_COMPRESSION
608 inode->i_mapping->a_ops = &f2fs_compress_aops;
609 /*
610 * generic_error_remove_folio only truncates pages of regular
611 * inode
612 */
613 inode->i_mode |= S_IFREG;
614 #endif
615 mapping_set_gfp_mask(inode->i_mapping,
616 GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
617 } else if (S_ISREG(inode->i_mode)) {
618 inode->i_op = &f2fs_file_inode_operations;
619 inode->i_fop = &f2fs_file_operations;
620 inode->i_mapping->a_ops = &f2fs_dblock_aops;
621 } else if (S_ISDIR(inode->i_mode)) {
622 inode->i_op = &f2fs_dir_inode_operations;
623 inode->i_fop = &f2fs_dir_operations;
624 inode->i_mapping->a_ops = &f2fs_dblock_aops;
625 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
626 } else if (S_ISLNK(inode->i_mode)) {
627 if (file_is_encrypt(inode))
628 inode->i_op = &f2fs_encrypted_symlink_inode_operations;
629 else
630 inode->i_op = &f2fs_symlink_inode_operations;
631 inode_nohighmem(inode);
632 inode->i_mapping->a_ops = &f2fs_dblock_aops;
633 } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
634 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
635 inode->i_op = &f2fs_special_inode_operations;
636 init_special_inode(inode, inode->i_mode, inode->i_rdev);
637 } else {
638 ret = -EIO;
639 goto bad_inode;
640 }
641 f2fs_set_inode_flags(inode);
642
643 unlock_new_inode(inode);
644 trace_f2fs_iget(inode);
645 return inode;
646
647 bad_inode:
648 f2fs_inode_synced(inode);
649 iget_failed(inode);
650 trace_f2fs_iget_exit(inode, ret);
651 return ERR_PTR(ret);
652 }
653
f2fs_iget_retry(struct super_block * sb,unsigned long ino)654 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
655 {
656 struct inode *inode;
657 retry:
658 inode = f2fs_iget(sb, ino);
659 if (IS_ERR(inode)) {
660 if (PTR_ERR(inode) == -ENOMEM) {
661 memalloc_retry_wait(GFP_NOFS);
662 goto retry;
663 }
664 }
665 return inode;
666 }
667
f2fs_update_inode(struct inode * inode,struct folio * node_folio)668 void f2fs_update_inode(struct inode *inode, struct folio *node_folio)
669 {
670 struct f2fs_inode_info *fi = F2FS_I(inode);
671 struct f2fs_inode *ri;
672 struct extent_tree *et = fi->extent_tree[EX_READ];
673
674 f2fs_folio_wait_writeback(node_folio, NODE, true, true);
675 folio_mark_dirty(node_folio);
676
677 f2fs_inode_synced(inode);
678
679 ri = F2FS_INODE(node_folio);
680
681 ri->i_mode = cpu_to_le16(inode->i_mode);
682 ri->i_advise = fi->i_advise;
683 ri->i_uid = cpu_to_le32(i_uid_read(inode));
684 ri->i_gid = cpu_to_le32(i_gid_read(inode));
685 ri->i_links = cpu_to_le32(inode->i_nlink);
686 ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
687
688 if (!f2fs_is_atomic_file(inode) ||
689 is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
690 ri->i_size = cpu_to_le64(i_size_read(inode));
691
692 if (et) {
693 read_lock(&et->lock);
694 set_raw_read_extent(&et->largest, &ri->i_ext);
695 read_unlock(&et->lock);
696 } else {
697 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
698 }
699 set_raw_inline(inode, ri);
700
701 ri->i_atime = cpu_to_le64(inode_get_atime_sec(inode));
702 ri->i_ctime = cpu_to_le64(inode_get_ctime_sec(inode));
703 ri->i_mtime = cpu_to_le64(inode_get_mtime_sec(inode));
704 ri->i_atime_nsec = cpu_to_le32(inode_get_atime_nsec(inode));
705 ri->i_ctime_nsec = cpu_to_le32(inode_get_ctime_nsec(inode));
706 ri->i_mtime_nsec = cpu_to_le32(inode_get_mtime_nsec(inode));
707 if (S_ISDIR(inode->i_mode))
708 ri->i_current_depth = cpu_to_le32(fi->i_current_depth);
709 else if (S_ISREG(inode->i_mode))
710 ri->i_gc_failures = cpu_to_le16(fi->i_gc_failures);
711 ri->i_xattr_nid = cpu_to_le32(fi->i_xattr_nid);
712 ri->i_flags = cpu_to_le32(fi->i_flags);
713 ri->i_pino = cpu_to_le32(fi->i_pino);
714 ri->i_generation = cpu_to_le32(inode->i_generation);
715 ri->i_dir_level = fi->i_dir_level;
716
717 if (f2fs_has_extra_attr(inode)) {
718 ri->i_extra_isize = cpu_to_le16(fi->i_extra_isize);
719
720 if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
721 ri->i_inline_xattr_size =
722 cpu_to_le16(fi->i_inline_xattr_size);
723
724 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
725 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) {
726 projid_t i_projid;
727
728 i_projid = from_kprojid(&init_user_ns, fi->i_projid);
729 ri->i_projid = cpu_to_le32(i_projid);
730 }
731
732 if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
733 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
734 ri->i_crtime = cpu_to_le64(fi->i_crtime.tv_sec);
735 ri->i_crtime_nsec = cpu_to_le32(fi->i_crtime.tv_nsec);
736 }
737
738 if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
739 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
740 i_compress_flag)) {
741 unsigned short compress_flag;
742
743 ri->i_compr_blocks = cpu_to_le64(
744 atomic_read(&fi->i_compr_blocks));
745 ri->i_compress_algorithm = fi->i_compress_algorithm;
746 compress_flag = fi->i_compress_flag |
747 fi->i_compress_level <<
748 COMPRESS_LEVEL_OFFSET;
749 ri->i_compress_flag = cpu_to_le16(compress_flag);
750 ri->i_log_cluster_size = fi->i_log_cluster_size;
751 }
752 }
753
754 __set_inode_rdev(inode, node_folio);
755
756 /* deleted inode */
757 if (inode->i_nlink == 0)
758 folio_clear_f2fs_inline(node_folio);
759
760 init_idisk_time(inode);
761 #ifdef CONFIG_F2FS_CHECK_FS
762 f2fs_inode_chksum_set(F2FS_I_SB(inode), node_folio);
763 #endif
764 }
765
f2fs_update_inode_page(struct inode * inode)766 void f2fs_update_inode_page(struct inode *inode)
767 {
768 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
769 struct folio *node_folio;
770 int count = 0;
771 retry:
772 node_folio = f2fs_get_inode_folio(sbi, inode->i_ino);
773 if (IS_ERR(node_folio)) {
774 int err = PTR_ERR(node_folio);
775
776 /* The node block was truncated. */
777 if (err == -ENOENT)
778 return;
779
780 if (err == -EFSCORRUPTED)
781 goto stop_checkpoint;
782
783 if (err == -ENOMEM || ++count <= DEFAULT_RETRY_IO_COUNT)
784 goto retry;
785 stop_checkpoint:
786 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_UPDATE_INODE);
787 return;
788 }
789 f2fs_update_inode(inode, node_folio);
790 f2fs_folio_put(node_folio, true);
791 }
792
f2fs_write_inode(struct inode * inode,struct writeback_control * wbc)793 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
794 {
795 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
796
797 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
798 inode->i_ino == F2FS_META_INO(sbi))
799 return 0;
800
801 /*
802 * atime could be updated without dirtying f2fs inode in lazytime mode
803 */
804 if (f2fs_is_time_consistent(inode) &&
805 !is_inode_flag_set(inode, FI_DIRTY_INODE))
806 return 0;
807
808 /*
809 * no need to update inode page, ultimately f2fs_evict_inode() will
810 * clear dirty status of inode.
811 */
812 if (f2fs_cp_error(sbi))
813 return -EIO;
814
815 if (!f2fs_is_checkpoint_ready(sbi)) {
816 f2fs_mark_inode_dirty_sync(inode, true);
817 return -ENOSPC;
818 }
819
820 /*
821 * We need to balance fs here to prevent from producing dirty node pages
822 * during the urgent cleaning time when running out of free sections.
823 */
824 f2fs_update_inode_page(inode);
825 if (wbc && wbc->nr_to_write)
826 f2fs_balance_fs(sbi, true);
827 return 0;
828 }
829
f2fs_remove_donate_inode(struct inode * inode)830 void f2fs_remove_donate_inode(struct inode *inode)
831 {
832 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
833
834 if (list_empty(&F2FS_I(inode)->gdonate_list))
835 return;
836
837 spin_lock(&sbi->inode_lock[DONATE_INODE]);
838 list_del_init(&F2FS_I(inode)->gdonate_list);
839 sbi->donate_files--;
840 spin_unlock(&sbi->inode_lock[DONATE_INODE]);
841 }
842
843 /*
844 * Called at the last iput() if i_nlink is zero
845 */
f2fs_evict_inode(struct inode * inode)846 void f2fs_evict_inode(struct inode *inode)
847 {
848 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
849 struct f2fs_inode_info *fi = F2FS_I(inode);
850 nid_t xnid = fi->i_xattr_nid;
851 int err = 0;
852 bool freeze_protected = false;
853
854 f2fs_abort_atomic_write(inode, true);
855
856 if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
857 clear_inode_flag(fi->cow_inode, FI_COW_FILE);
858 F2FS_I(fi->cow_inode)->atomic_inode = NULL;
859 iput(fi->cow_inode);
860 fi->cow_inode = NULL;
861 }
862
863 trace_f2fs_evict_inode(inode);
864 truncate_inode_pages_final(&inode->i_data);
865
866 if ((inode->i_nlink || is_bad_inode(inode)) &&
867 test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
868 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
869
870 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
871 inode->i_ino == F2FS_META_INO(sbi) ||
872 inode->i_ino == F2FS_COMPRESS_INO(sbi))
873 goto out_clear;
874
875 f2fs_bug_on(sbi, get_dirty_pages(inode));
876 f2fs_remove_dirty_inode(inode);
877 f2fs_remove_donate_inode(inode);
878
879 if (!IS_DEVICE_ALIASING(inode))
880 f2fs_destroy_extent_tree(inode);
881
882 if (inode->i_nlink || is_bad_inode(inode))
883 goto no_delete;
884
885 err = f2fs_dquot_initialize(inode);
886 if (err) {
887 err = 0;
888 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
889 }
890
891 f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
892 f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
893 f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
894
895 if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING)) {
896 sb_start_intwrite(inode->i_sb);
897 freeze_protected = true;
898 }
899 set_inode_flag(inode, FI_NO_ALLOC);
900 i_size_write(inode, 0);
901 retry:
902 if (F2FS_HAS_BLOCKS(inode))
903 err = f2fs_truncate(inode);
904
905 if (time_to_inject(sbi, FAULT_EVICT_INODE))
906 err = -EIO;
907
908 if (!err) {
909 f2fs_lock_op(sbi);
910 err = f2fs_remove_inode_page(inode);
911 f2fs_unlock_op(sbi);
912 if (err == -ENOENT) {
913 err = 0;
914
915 /*
916 * in fuzzed image, another node may has the same
917 * block address as inode's, if it was truncated
918 * previously, truncation of inode node will fail.
919 */
920 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
921 f2fs_warn(F2FS_I_SB(inode),
922 "f2fs_evict_inode: inconsistent node id, ino:%lu",
923 inode->i_ino);
924 f2fs_inode_synced(inode);
925 set_sbi_flag(sbi, SBI_NEED_FSCK);
926 }
927 }
928 }
929
930 /* give more chances, if ENOMEM case */
931 if (err == -ENOMEM) {
932 err = 0;
933 goto retry;
934 }
935
936 if (IS_DEVICE_ALIASING(inode))
937 f2fs_destroy_extent_tree(inode);
938
939 if (err) {
940 f2fs_update_inode_page(inode);
941 if (dquot_initialize_needed(inode))
942 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
943
944 /*
945 * If both f2fs_truncate() and f2fs_update_inode_page() failed
946 * due to fuzzed corrupted inode, call f2fs_inode_synced() to
947 * avoid triggering later f2fs_bug_on().
948 */
949 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
950 f2fs_warn(sbi,
951 "f2fs_evict_inode: inode is dirty, ino:%lu",
952 inode->i_ino);
953 f2fs_inode_synced(inode);
954 set_sbi_flag(sbi, SBI_NEED_FSCK);
955 }
956 }
957 if (freeze_protected)
958 sb_end_intwrite(inode->i_sb);
959 no_delete:
960 dquot_drop(inode);
961
962 stat_dec_inline_xattr(inode);
963 stat_dec_inline_dir(inode);
964 stat_dec_inline_inode(inode);
965 stat_dec_compr_inode(inode);
966 stat_sub_compr_blocks(inode,
967 atomic_read(&fi->i_compr_blocks));
968
969 if (likely(!f2fs_cp_error(sbi) &&
970 !is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
971 f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
972
973 /*
974 * anyway, it needs to remove the inode from sbi->inode_list[DIRTY_META]
975 * list to avoid UAF in f2fs_sync_inode_meta() during checkpoint.
976 */
977 f2fs_inode_synced(inode);
978
979 /* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
980 if (inode->i_ino)
981 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
982 inode->i_ino);
983 if (xnid)
984 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
985 if (inode->i_nlink) {
986 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
987 f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
988 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
989 f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
990 }
991 if (is_inode_flag_set(inode, FI_FREE_NID)) {
992 f2fs_alloc_nid_failed(sbi, inode->i_ino);
993 clear_inode_flag(inode, FI_FREE_NID);
994 } else {
995 /*
996 * If xattr nid is corrupted, we can reach out error condition,
997 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
998 * In that case, f2fs_check_nid_range() is enough to give a clue.
999 */
1000 }
1001 out_clear:
1002 fscrypt_put_encryption_info(inode);
1003 fsverity_cleanup_inode(inode);
1004 clear_inode(inode);
1005 }
1006
1007 /* caller should call f2fs_lock_op() */
f2fs_handle_failed_inode(struct inode * inode)1008 void f2fs_handle_failed_inode(struct inode *inode)
1009 {
1010 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1011 struct node_info ni;
1012 int err;
1013
1014 /*
1015 * clear nlink of inode in order to release resource of inode
1016 * immediately.
1017 */
1018 clear_nlink(inode);
1019
1020 /*
1021 * we must call this to avoid inode being remained as dirty, resulting
1022 * in a panic when flushing dirty inodes in gdirty_list.
1023 */
1024 f2fs_update_inode_page(inode);
1025 f2fs_inode_synced(inode);
1026
1027 /* don't make bad inode, since it becomes a regular file. */
1028 unlock_new_inode(inode);
1029
1030 /*
1031 * Note: we should add inode to orphan list before f2fs_unlock_op()
1032 * so we can prevent losing this orphan when encoutering checkpoint
1033 * and following suddenly power-off.
1034 */
1035 err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
1036 if (err) {
1037 set_sbi_flag(sbi, SBI_NEED_FSCK);
1038 set_inode_flag(inode, FI_FREE_NID);
1039 f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
1040 goto out;
1041 }
1042
1043 if (ni.blk_addr != NULL_ADDR) {
1044 err = f2fs_acquire_orphan_inode(sbi);
1045 if (err) {
1046 set_sbi_flag(sbi, SBI_NEED_FSCK);
1047 f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
1048 } else {
1049 f2fs_add_orphan_inode(inode);
1050 }
1051 f2fs_alloc_nid_done(sbi, inode->i_ino);
1052 } else {
1053 set_inode_flag(inode, FI_FREE_NID);
1054 }
1055
1056 out:
1057 f2fs_unlock_op(sbi);
1058
1059 /* iput will drop the inode object */
1060 iput(inode);
1061 }
1062