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->page));
112 folio_mark_dirty(ifolio);
113 return;
114 }
115 }
116 return;
117 }
118
f2fs_enable_inode_chksum(struct f2fs_sb_info * sbi,struct page * page)119 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
120 {
121 struct f2fs_inode *ri = &F2FS_NODE(page)->i;
122
123 if (!f2fs_sb_has_inode_chksum(sbi))
124 return false;
125
126 if (!IS_INODE(page) || !(ri->i_inline & F2FS_EXTRA_ATTR))
127 return false;
128
129 if (!F2FS_FITS_IN_INODE(ri, le16_to_cpu(ri->i_extra_isize),
130 i_inode_checksum))
131 return false;
132
133 return true;
134 }
135
f2fs_inode_chksum(struct f2fs_sb_info * sbi,struct page * page)136 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
137 {
138 struct f2fs_node *node = F2FS_NODE(page);
139 struct f2fs_inode *ri = &node->i;
140 __le32 ino = node->footer.ino;
141 __le32 gen = ri->i_generation;
142 __u32 chksum, chksum_seed;
143 __u32 dummy_cs = 0;
144 unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
145 unsigned int cs_size = sizeof(dummy_cs);
146
147 chksum = f2fs_chksum(sbi->s_chksum_seed, (__u8 *)&ino, sizeof(ino));
148 chksum_seed = f2fs_chksum(chksum, (__u8 *)&gen, sizeof(gen));
149
150 chksum = f2fs_chksum(chksum_seed, (__u8 *)ri, offset);
151 chksum = f2fs_chksum(chksum, (__u8 *)&dummy_cs, cs_size);
152 offset += cs_size;
153 chksum = f2fs_chksum(chksum, (__u8 *)ri + offset,
154 F2FS_BLKSIZE - offset);
155 return chksum;
156 }
157
f2fs_inode_chksum_verify(struct f2fs_sb_info * sbi,struct folio * folio)158 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct folio *folio)
159 {
160 struct f2fs_inode *ri;
161 __u32 provided, calculated;
162
163 if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
164 return true;
165
166 #ifdef CONFIG_F2FS_CHECK_FS
167 if (!f2fs_enable_inode_chksum(sbi, &folio->page))
168 #else
169 if (!f2fs_enable_inode_chksum(sbi, &folio->page) ||
170 folio_test_dirty(folio) ||
171 folio_test_writeback(folio))
172 #endif
173 return true;
174
175 ri = &F2FS_NODE(&folio->page)->i;
176 provided = le32_to_cpu(ri->i_inode_checksum);
177 calculated = f2fs_inode_chksum(sbi, &folio->page);
178
179 if (provided != calculated)
180 f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
181 folio->index, ino_of_node(&folio->page),
182 provided, calculated);
183
184 return provided == calculated;
185 }
186
f2fs_inode_chksum_set(struct f2fs_sb_info * sbi,struct page * page)187 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
188 {
189 struct f2fs_inode *ri = &F2FS_NODE(page)->i;
190
191 if (!f2fs_enable_inode_chksum(sbi, page))
192 return;
193
194 ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
195 }
196
sanity_check_compress_inode(struct inode * inode,struct f2fs_inode * ri)197 static bool sanity_check_compress_inode(struct inode *inode,
198 struct f2fs_inode *ri)
199 {
200 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
201 unsigned char clevel;
202
203 if (ri->i_compress_algorithm >= COMPRESS_MAX) {
204 f2fs_warn(sbi,
205 "%s: inode (ino=%lx) has unsupported compress algorithm: %u, run fsck to fix",
206 __func__, inode->i_ino, ri->i_compress_algorithm);
207 return false;
208 }
209 if (le64_to_cpu(ri->i_compr_blocks) >
210 SECTOR_TO_BLOCK(inode->i_blocks)) {
211 f2fs_warn(sbi,
212 "%s: inode (ino=%lx) has inconsistent i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
213 __func__, inode->i_ino, le64_to_cpu(ri->i_compr_blocks),
214 SECTOR_TO_BLOCK(inode->i_blocks));
215 return false;
216 }
217 if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
218 ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
219 f2fs_warn(sbi,
220 "%s: inode (ino=%lx) has unsupported log cluster size: %u, run fsck to fix",
221 __func__, inode->i_ino, ri->i_log_cluster_size);
222 return false;
223 }
224
225 clevel = le16_to_cpu(ri->i_compress_flag) >>
226 COMPRESS_LEVEL_OFFSET;
227 switch (ri->i_compress_algorithm) {
228 case COMPRESS_LZO:
229 #ifdef CONFIG_F2FS_FS_LZO
230 if (clevel)
231 goto err_level;
232 #endif
233 break;
234 case COMPRESS_LZORLE:
235 #ifdef CONFIG_F2FS_FS_LZORLE
236 if (clevel)
237 goto err_level;
238 #endif
239 break;
240 case COMPRESS_LZ4:
241 #ifdef CONFIG_F2FS_FS_LZ4
242 #ifdef CONFIG_F2FS_FS_LZ4HC
243 if (clevel &&
244 (clevel < LZ4HC_MIN_CLEVEL || clevel > LZ4HC_MAX_CLEVEL))
245 goto err_level;
246 #else
247 if (clevel)
248 goto err_level;
249 #endif
250 #endif
251 break;
252 case COMPRESS_ZSTD:
253 #ifdef CONFIG_F2FS_FS_ZSTD
254 if (clevel < zstd_min_clevel() || clevel > zstd_max_clevel())
255 goto err_level;
256 #endif
257 break;
258 default:
259 goto err_level;
260 }
261
262 return true;
263 err_level:
264 f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported compress level: %u, run fsck to fix",
265 __func__, inode->i_ino, clevel);
266 return false;
267 }
268
sanity_check_inode(struct inode * inode,struct page * node_page)269 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
270 {
271 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
272 struct f2fs_inode_info *fi = F2FS_I(inode);
273 struct f2fs_inode *ri = F2FS_INODE(node_page);
274 unsigned long long iblocks;
275
276 iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
277 if (!iblocks) {
278 f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
279 __func__, inode->i_ino, iblocks);
280 return false;
281 }
282
283 if (ino_of_node(node_page) != nid_of_node(node_page)) {
284 f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
285 __func__, inode->i_ino,
286 ino_of_node(node_page), nid_of_node(node_page));
287 return false;
288 }
289
290 if (ino_of_node(node_page) == fi->i_xattr_nid) {
291 f2fs_warn(sbi, "%s: corrupted inode i_ino=%lx, xnid=%x, run fsck to fix.",
292 __func__, inode->i_ino, fi->i_xattr_nid);
293 return false;
294 }
295
296 if (f2fs_has_extra_attr(inode)) {
297 if (!f2fs_sb_has_extra_attr(sbi)) {
298 f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
299 __func__, inode->i_ino);
300 return false;
301 }
302 if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
303 fi->i_extra_isize < F2FS_MIN_EXTRA_ATTR_SIZE ||
304 fi->i_extra_isize % sizeof(__le32)) {
305 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
306 __func__, inode->i_ino, fi->i_extra_isize,
307 F2FS_TOTAL_EXTRA_ATTR_SIZE);
308 return false;
309 }
310 if (f2fs_sb_has_compression(sbi) &&
311 fi->i_flags & F2FS_COMPR_FL &&
312 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
313 i_compress_flag)) {
314 if (!sanity_check_compress_inode(inode, ri))
315 return false;
316 }
317 }
318
319 if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
320 f2fs_has_inline_xattr(inode) &&
321 (fi->i_inline_xattr_size < MIN_INLINE_XATTR_SIZE ||
322 fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
323 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, min: %zu, max: %lu",
324 __func__, inode->i_ino, fi->i_inline_xattr_size,
325 MIN_INLINE_XATTR_SIZE, MAX_INLINE_XATTR_SIZE);
326 return false;
327 }
328
329 if (!f2fs_sb_has_extra_attr(sbi)) {
330 if (f2fs_sb_has_project_quota(sbi)) {
331 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
332 __func__, inode->i_ino, F2FS_FEATURE_PRJQUOTA);
333 return false;
334 }
335 if (f2fs_sb_has_inode_chksum(sbi)) {
336 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
337 __func__, inode->i_ino, F2FS_FEATURE_INODE_CHKSUM);
338 return false;
339 }
340 if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
341 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
342 __func__, inode->i_ino, F2FS_FEATURE_FLEXIBLE_INLINE_XATTR);
343 return false;
344 }
345 if (f2fs_sb_has_inode_crtime(sbi)) {
346 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
347 __func__, inode->i_ino, F2FS_FEATURE_INODE_CRTIME);
348 return false;
349 }
350 if (f2fs_sb_has_compression(sbi)) {
351 f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
352 __func__, inode->i_ino, F2FS_FEATURE_COMPRESSION);
353 return false;
354 }
355 }
356
357 if (f2fs_sanity_check_inline_data(inode, node_page)) {
358 f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
359 __func__, inode->i_ino, inode->i_mode);
360 return false;
361 }
362
363 if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
364 f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
365 __func__, inode->i_ino, inode->i_mode);
366 return false;
367 }
368
369 if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
370 f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
371 __func__, inode->i_ino);
372 return false;
373 }
374
375 if (fi->i_xattr_nid && f2fs_check_nid_range(sbi, fi->i_xattr_nid)) {
376 f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_xattr_nid: %u, run fsck to fix.",
377 __func__, inode->i_ino, fi->i_xattr_nid);
378 return false;
379 }
380
381 if (IS_DEVICE_ALIASING(inode)) {
382 if (!f2fs_sb_has_device_alias(sbi)) {
383 f2fs_warn(sbi, "%s: inode (ino=%lx) has device alias flag, but the feature is off",
384 __func__, inode->i_ino);
385 return false;
386 }
387 if (!f2fs_is_pinned_file(inode)) {
388 f2fs_warn(sbi, "%s: inode (ino=%lx) has device alias flag, but is not pinned",
389 __func__, inode->i_ino);
390 return false;
391 }
392 }
393
394 return true;
395 }
396
init_idisk_time(struct inode * inode)397 static void init_idisk_time(struct inode *inode)
398 {
399 struct f2fs_inode_info *fi = F2FS_I(inode);
400
401 fi->i_disk_time[0] = inode_get_atime(inode);
402 fi->i_disk_time[1] = inode_get_ctime(inode);
403 fi->i_disk_time[2] = inode_get_mtime(inode);
404 }
405
do_read_inode(struct inode * inode)406 static int do_read_inode(struct inode *inode)
407 {
408 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
409 struct f2fs_inode_info *fi = F2FS_I(inode);
410 struct folio *node_folio;
411 struct f2fs_inode *ri;
412 projid_t i_projid;
413
414 /* Check if ino is within scope */
415 if (f2fs_check_nid_range(sbi, inode->i_ino))
416 return -EINVAL;
417
418 node_folio = f2fs_get_inode_folio(sbi, inode->i_ino);
419 if (IS_ERR(node_folio))
420 return PTR_ERR(node_folio);
421
422 ri = F2FS_INODE(&node_folio->page);
423
424 inode->i_mode = le16_to_cpu(ri->i_mode);
425 i_uid_write(inode, le32_to_cpu(ri->i_uid));
426 i_gid_write(inode, le32_to_cpu(ri->i_gid));
427 set_nlink(inode, le32_to_cpu(ri->i_links));
428 inode->i_size = le64_to_cpu(ri->i_size);
429 inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
430
431 inode_set_atime(inode, le64_to_cpu(ri->i_atime),
432 le32_to_cpu(ri->i_atime_nsec));
433 inode_set_ctime(inode, le64_to_cpu(ri->i_ctime),
434 le32_to_cpu(ri->i_ctime_nsec));
435 inode_set_mtime(inode, le64_to_cpu(ri->i_mtime),
436 le32_to_cpu(ri->i_mtime_nsec));
437 inode->i_generation = le32_to_cpu(ri->i_generation);
438 if (S_ISDIR(inode->i_mode))
439 fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
440 else if (S_ISREG(inode->i_mode))
441 fi->i_gc_failures = le16_to_cpu(ri->i_gc_failures);
442 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
443 fi->i_flags = le32_to_cpu(ri->i_flags);
444 if (S_ISREG(inode->i_mode))
445 fi->i_flags &= ~F2FS_PROJINHERIT_FL;
446 bitmap_zero(fi->flags, FI_MAX);
447 fi->i_advise = ri->i_advise;
448 fi->i_pino = le32_to_cpu(ri->i_pino);
449 fi->i_dir_level = ri->i_dir_level;
450
451 get_inline_info(inode, ri);
452
453 fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
454 le16_to_cpu(ri->i_extra_isize) : 0;
455
456 if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
457 fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
458 } else if (f2fs_has_inline_xattr(inode) ||
459 f2fs_has_inline_dentry(inode)) {
460 fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
461 } else {
462
463 /*
464 * Previous inline data or directory always reserved 200 bytes
465 * in inode layout, even if inline_xattr is disabled. In order
466 * to keep inline_dentry's structure for backward compatibility,
467 * we get the space back only from inline_data.
468 */
469 fi->i_inline_xattr_size = 0;
470 }
471
472 if (!sanity_check_inode(inode, &node_folio->page)) {
473 f2fs_folio_put(node_folio, true);
474 set_sbi_flag(sbi, SBI_NEED_FSCK);
475 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
476 return -EFSCORRUPTED;
477 }
478
479 /* check data exist */
480 if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
481 __recover_inline_status(inode, node_folio);
482
483 /* try to recover cold bit for non-dir inode */
484 if (!S_ISDIR(inode->i_mode) && !is_cold_node(&node_folio->page)) {
485 f2fs_folio_wait_writeback(node_folio, NODE, true, true);
486 set_cold_node(&node_folio->page, false);
487 folio_mark_dirty(node_folio);
488 }
489
490 /* get rdev by using inline_info */
491 __get_inode_rdev(inode, node_folio);
492
493 if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
494 fi->last_disk_size = inode->i_size;
495
496 if (fi->i_flags & F2FS_PROJINHERIT_FL)
497 set_inode_flag(inode, FI_PROJ_INHERIT);
498
499 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
500 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
501 i_projid = (projid_t)le32_to_cpu(ri->i_projid);
502 else
503 i_projid = F2FS_DEF_PROJID;
504 fi->i_projid = make_kprojid(&init_user_ns, i_projid);
505
506 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
507 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
508 fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
509 fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
510 }
511
512 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
513 (fi->i_flags & F2FS_COMPR_FL)) {
514 if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
515 i_compress_flag)) {
516 unsigned short compress_flag;
517
518 atomic_set(&fi->i_compr_blocks,
519 le64_to_cpu(ri->i_compr_blocks));
520 fi->i_compress_algorithm = ri->i_compress_algorithm;
521 fi->i_log_cluster_size = ri->i_log_cluster_size;
522 compress_flag = le16_to_cpu(ri->i_compress_flag);
523 fi->i_compress_level = compress_flag >>
524 COMPRESS_LEVEL_OFFSET;
525 fi->i_compress_flag = compress_flag &
526 GENMASK(COMPRESS_LEVEL_OFFSET - 1, 0);
527 fi->i_cluster_size = BIT(fi->i_log_cluster_size);
528 set_inode_flag(inode, FI_COMPRESSED_FILE);
529 }
530 }
531
532 init_idisk_time(inode);
533
534 if (!sanity_check_extent_cache(inode, &node_folio->page)) {
535 f2fs_folio_put(node_folio, true);
536 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
537 return -EFSCORRUPTED;
538 }
539
540 /* Need all the flag bits */
541 f2fs_init_read_extent_tree(inode, node_folio);
542 f2fs_init_age_extent_tree(inode);
543
544 f2fs_folio_put(node_folio, true);
545
546 stat_inc_inline_xattr(inode);
547 stat_inc_inline_inode(inode);
548 stat_inc_inline_dir(inode);
549 stat_inc_compr_inode(inode);
550 stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
551
552 return 0;
553 }
554
is_meta_ino(struct f2fs_sb_info * sbi,unsigned int ino)555 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
556 {
557 return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
558 ino == F2FS_COMPRESS_INO(sbi);
559 }
560
f2fs_iget(struct super_block * sb,unsigned long ino)561 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
562 {
563 struct f2fs_sb_info *sbi = F2FS_SB(sb);
564 struct inode *inode;
565 int ret = 0;
566
567 inode = iget_locked(sb, ino);
568 if (!inode)
569 return ERR_PTR(-ENOMEM);
570
571 if (!(inode->i_state & I_NEW)) {
572 if (is_meta_ino(sbi, ino)) {
573 f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
574 set_sbi_flag(sbi, SBI_NEED_FSCK);
575 ret = -EFSCORRUPTED;
576 trace_f2fs_iget_exit(inode, ret);
577 iput(inode);
578 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
579 return ERR_PTR(ret);
580 }
581
582 trace_f2fs_iget(inode);
583 return inode;
584 }
585
586 if (is_meta_ino(sbi, ino))
587 goto make_now;
588
589 ret = do_read_inode(inode);
590 if (ret)
591 goto bad_inode;
592 make_now:
593 if (ino == F2FS_NODE_INO(sbi)) {
594 inode->i_mapping->a_ops = &f2fs_node_aops;
595 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
596 } else if (ino == F2FS_META_INO(sbi)) {
597 inode->i_mapping->a_ops = &f2fs_meta_aops;
598 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
599 } else if (ino == F2FS_COMPRESS_INO(sbi)) {
600 #ifdef CONFIG_F2FS_FS_COMPRESSION
601 inode->i_mapping->a_ops = &f2fs_compress_aops;
602 /*
603 * generic_error_remove_folio only truncates pages of regular
604 * inode
605 */
606 inode->i_mode |= S_IFREG;
607 #endif
608 mapping_set_gfp_mask(inode->i_mapping,
609 GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
610 } else if (S_ISREG(inode->i_mode)) {
611 inode->i_op = &f2fs_file_inode_operations;
612 inode->i_fop = &f2fs_file_operations;
613 inode->i_mapping->a_ops = &f2fs_dblock_aops;
614 } else if (S_ISDIR(inode->i_mode)) {
615 inode->i_op = &f2fs_dir_inode_operations;
616 inode->i_fop = &f2fs_dir_operations;
617 inode->i_mapping->a_ops = &f2fs_dblock_aops;
618 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
619 } else if (S_ISLNK(inode->i_mode)) {
620 if (file_is_encrypt(inode))
621 inode->i_op = &f2fs_encrypted_symlink_inode_operations;
622 else
623 inode->i_op = &f2fs_symlink_inode_operations;
624 inode_nohighmem(inode);
625 inode->i_mapping->a_ops = &f2fs_dblock_aops;
626 } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
627 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
628 inode->i_op = &f2fs_special_inode_operations;
629 init_special_inode(inode, inode->i_mode, inode->i_rdev);
630 } else {
631 ret = -EIO;
632 goto bad_inode;
633 }
634 f2fs_set_inode_flags(inode);
635
636 unlock_new_inode(inode);
637 trace_f2fs_iget(inode);
638 return inode;
639
640 bad_inode:
641 f2fs_inode_synced(inode);
642 iget_failed(inode);
643 trace_f2fs_iget_exit(inode, ret);
644 return ERR_PTR(ret);
645 }
646
f2fs_iget_retry(struct super_block * sb,unsigned long ino)647 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
648 {
649 struct inode *inode;
650 retry:
651 inode = f2fs_iget(sb, ino);
652 if (IS_ERR(inode)) {
653 if (PTR_ERR(inode) == -ENOMEM) {
654 memalloc_retry_wait(GFP_NOFS);
655 goto retry;
656 }
657 }
658 return inode;
659 }
660
f2fs_update_inode(struct inode * inode,struct folio * node_folio)661 void f2fs_update_inode(struct inode *inode, struct folio *node_folio)
662 {
663 struct f2fs_inode_info *fi = F2FS_I(inode);
664 struct f2fs_inode *ri;
665 struct extent_tree *et = fi->extent_tree[EX_READ];
666
667 f2fs_folio_wait_writeback(node_folio, NODE, true, true);
668 folio_mark_dirty(node_folio);
669
670 f2fs_inode_synced(inode);
671
672 ri = F2FS_INODE(&node_folio->page);
673
674 ri->i_mode = cpu_to_le16(inode->i_mode);
675 ri->i_advise = fi->i_advise;
676 ri->i_uid = cpu_to_le32(i_uid_read(inode));
677 ri->i_gid = cpu_to_le32(i_gid_read(inode));
678 ri->i_links = cpu_to_le32(inode->i_nlink);
679 ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
680
681 if (!f2fs_is_atomic_file(inode) ||
682 is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
683 ri->i_size = cpu_to_le64(i_size_read(inode));
684
685 if (et) {
686 read_lock(&et->lock);
687 set_raw_read_extent(&et->largest, &ri->i_ext);
688 read_unlock(&et->lock);
689 } else {
690 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
691 }
692 set_raw_inline(inode, ri);
693
694 ri->i_atime = cpu_to_le64(inode_get_atime_sec(inode));
695 ri->i_ctime = cpu_to_le64(inode_get_ctime_sec(inode));
696 ri->i_mtime = cpu_to_le64(inode_get_mtime_sec(inode));
697 ri->i_atime_nsec = cpu_to_le32(inode_get_atime_nsec(inode));
698 ri->i_ctime_nsec = cpu_to_le32(inode_get_ctime_nsec(inode));
699 ri->i_mtime_nsec = cpu_to_le32(inode_get_mtime_nsec(inode));
700 if (S_ISDIR(inode->i_mode))
701 ri->i_current_depth = cpu_to_le32(fi->i_current_depth);
702 else if (S_ISREG(inode->i_mode))
703 ri->i_gc_failures = cpu_to_le16(fi->i_gc_failures);
704 ri->i_xattr_nid = cpu_to_le32(fi->i_xattr_nid);
705 ri->i_flags = cpu_to_le32(fi->i_flags);
706 ri->i_pino = cpu_to_le32(fi->i_pino);
707 ri->i_generation = cpu_to_le32(inode->i_generation);
708 ri->i_dir_level = fi->i_dir_level;
709
710 if (f2fs_has_extra_attr(inode)) {
711 ri->i_extra_isize = cpu_to_le16(fi->i_extra_isize);
712
713 if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
714 ri->i_inline_xattr_size =
715 cpu_to_le16(fi->i_inline_xattr_size);
716
717 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
718 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) {
719 projid_t i_projid;
720
721 i_projid = from_kprojid(&init_user_ns, fi->i_projid);
722 ri->i_projid = cpu_to_le32(i_projid);
723 }
724
725 if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
726 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
727 ri->i_crtime = cpu_to_le64(fi->i_crtime.tv_sec);
728 ri->i_crtime_nsec = cpu_to_le32(fi->i_crtime.tv_nsec);
729 }
730
731 if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
732 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
733 i_compress_flag)) {
734 unsigned short compress_flag;
735
736 ri->i_compr_blocks = cpu_to_le64(
737 atomic_read(&fi->i_compr_blocks));
738 ri->i_compress_algorithm = fi->i_compress_algorithm;
739 compress_flag = fi->i_compress_flag |
740 fi->i_compress_level <<
741 COMPRESS_LEVEL_OFFSET;
742 ri->i_compress_flag = cpu_to_le16(compress_flag);
743 ri->i_log_cluster_size = fi->i_log_cluster_size;
744 }
745 }
746
747 __set_inode_rdev(inode, node_folio);
748
749 /* deleted inode */
750 if (inode->i_nlink == 0)
751 clear_page_private_inline(&node_folio->page);
752
753 init_idisk_time(inode);
754 #ifdef CONFIG_F2FS_CHECK_FS
755 f2fs_inode_chksum_set(F2FS_I_SB(inode), &node_folio->page);
756 #endif
757 }
758
f2fs_update_inode_page(struct inode * inode)759 void f2fs_update_inode_page(struct inode *inode)
760 {
761 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
762 struct folio *node_folio;
763 int count = 0;
764 retry:
765 node_folio = f2fs_get_inode_folio(sbi, inode->i_ino);
766 if (IS_ERR(node_folio)) {
767 int err = PTR_ERR(node_folio);
768
769 /* The node block was truncated. */
770 if (err == -ENOENT)
771 return;
772
773 if (err == -EFSCORRUPTED)
774 goto stop_checkpoint;
775
776 if (err == -ENOMEM || ++count <= DEFAULT_RETRY_IO_COUNT)
777 goto retry;
778 stop_checkpoint:
779 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_UPDATE_INODE);
780 return;
781 }
782 f2fs_update_inode(inode, node_folio);
783 f2fs_folio_put(node_folio, true);
784 }
785
f2fs_write_inode(struct inode * inode,struct writeback_control * wbc)786 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
787 {
788 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
789
790 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
791 inode->i_ino == F2FS_META_INO(sbi))
792 return 0;
793
794 /*
795 * atime could be updated without dirtying f2fs inode in lazytime mode
796 */
797 if (f2fs_is_time_consistent(inode) &&
798 !is_inode_flag_set(inode, FI_DIRTY_INODE))
799 return 0;
800
801 /*
802 * no need to update inode page, ultimately f2fs_evict_inode() will
803 * clear dirty status of inode.
804 */
805 if (f2fs_cp_error(sbi))
806 return -EIO;
807
808 if (!f2fs_is_checkpoint_ready(sbi)) {
809 f2fs_mark_inode_dirty_sync(inode, true);
810 return -ENOSPC;
811 }
812
813 /*
814 * We need to balance fs here to prevent from producing dirty node pages
815 * during the urgent cleaning time when running out of free sections.
816 */
817 f2fs_update_inode_page(inode);
818 if (wbc && wbc->nr_to_write)
819 f2fs_balance_fs(sbi, true);
820 return 0;
821 }
822
f2fs_remove_donate_inode(struct inode * inode)823 static void f2fs_remove_donate_inode(struct inode *inode)
824 {
825 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
826
827 if (list_empty(&F2FS_I(inode)->gdonate_list))
828 return;
829
830 spin_lock(&sbi->inode_lock[DONATE_INODE]);
831 list_del_init(&F2FS_I(inode)->gdonate_list);
832 sbi->donate_files--;
833 spin_unlock(&sbi->inode_lock[DONATE_INODE]);
834 }
835
836 /*
837 * Called at the last iput() if i_nlink is zero
838 */
f2fs_evict_inode(struct inode * inode)839 void f2fs_evict_inode(struct inode *inode)
840 {
841 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
842 struct f2fs_inode_info *fi = F2FS_I(inode);
843 nid_t xnid = fi->i_xattr_nid;
844 int err = 0;
845 bool freeze_protected = false;
846
847 f2fs_abort_atomic_write(inode, true);
848
849 if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
850 clear_inode_flag(fi->cow_inode, FI_COW_FILE);
851 F2FS_I(fi->cow_inode)->atomic_inode = NULL;
852 iput(fi->cow_inode);
853 fi->cow_inode = NULL;
854 }
855
856 trace_f2fs_evict_inode(inode);
857 truncate_inode_pages_final(&inode->i_data);
858
859 if ((inode->i_nlink || is_bad_inode(inode)) &&
860 test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
861 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
862
863 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
864 inode->i_ino == F2FS_META_INO(sbi) ||
865 inode->i_ino == F2FS_COMPRESS_INO(sbi))
866 goto out_clear;
867
868 f2fs_bug_on(sbi, get_dirty_pages(inode));
869 f2fs_remove_dirty_inode(inode);
870 f2fs_remove_donate_inode(inode);
871
872 if (!IS_DEVICE_ALIASING(inode))
873 f2fs_destroy_extent_tree(inode);
874
875 if (inode->i_nlink || is_bad_inode(inode))
876 goto no_delete;
877
878 err = f2fs_dquot_initialize(inode);
879 if (err) {
880 err = 0;
881 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
882 }
883
884 f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
885 f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
886 f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
887
888 if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING)) {
889 sb_start_intwrite(inode->i_sb);
890 freeze_protected = true;
891 }
892 set_inode_flag(inode, FI_NO_ALLOC);
893 i_size_write(inode, 0);
894 retry:
895 if (F2FS_HAS_BLOCKS(inode))
896 err = f2fs_truncate(inode);
897
898 if (time_to_inject(sbi, FAULT_EVICT_INODE))
899 err = -EIO;
900
901 if (!err) {
902 f2fs_lock_op(sbi);
903 err = f2fs_remove_inode_page(inode);
904 f2fs_unlock_op(sbi);
905 if (err == -ENOENT) {
906 err = 0;
907
908 /*
909 * in fuzzed image, another node may has the same
910 * block address as inode's, if it was truncated
911 * previously, truncation of inode node will fail.
912 */
913 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
914 f2fs_warn(F2FS_I_SB(inode),
915 "f2fs_evict_inode: inconsistent node id, ino:%lu",
916 inode->i_ino);
917 f2fs_inode_synced(inode);
918 set_sbi_flag(sbi, SBI_NEED_FSCK);
919 }
920 }
921 }
922
923 /* give more chances, if ENOMEM case */
924 if (err == -ENOMEM) {
925 err = 0;
926 goto retry;
927 }
928
929 if (IS_DEVICE_ALIASING(inode))
930 f2fs_destroy_extent_tree(inode);
931
932 if (err) {
933 f2fs_update_inode_page(inode);
934 if (dquot_initialize_needed(inode))
935 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
936 }
937 if (freeze_protected)
938 sb_end_intwrite(inode->i_sb);
939 no_delete:
940 dquot_drop(inode);
941
942 stat_dec_inline_xattr(inode);
943 stat_dec_inline_dir(inode);
944 stat_dec_inline_inode(inode);
945 stat_dec_compr_inode(inode);
946 stat_sub_compr_blocks(inode,
947 atomic_read(&fi->i_compr_blocks));
948
949 if (likely(!f2fs_cp_error(sbi) &&
950 !is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
951 f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
952 else
953 f2fs_inode_synced(inode);
954
955 /* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
956 if (inode->i_ino)
957 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
958 inode->i_ino);
959 if (xnid)
960 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
961 if (inode->i_nlink) {
962 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
963 f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
964 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
965 f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
966 }
967 if (is_inode_flag_set(inode, FI_FREE_NID)) {
968 f2fs_alloc_nid_failed(sbi, inode->i_ino);
969 clear_inode_flag(inode, FI_FREE_NID);
970 } else {
971 /*
972 * If xattr nid is corrupted, we can reach out error condition,
973 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
974 * In that case, f2fs_check_nid_range() is enough to give a clue.
975 */
976 }
977 out_clear:
978 fscrypt_put_encryption_info(inode);
979 fsverity_cleanup_inode(inode);
980 clear_inode(inode);
981 }
982
983 /* caller should call f2fs_lock_op() */
f2fs_handle_failed_inode(struct inode * inode)984 void f2fs_handle_failed_inode(struct inode *inode)
985 {
986 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
987 struct node_info ni;
988 int err;
989
990 /*
991 * clear nlink of inode in order to release resource of inode
992 * immediately.
993 */
994 clear_nlink(inode);
995
996 /*
997 * we must call this to avoid inode being remained as dirty, resulting
998 * in a panic when flushing dirty inodes in gdirty_list.
999 */
1000 f2fs_update_inode_page(inode);
1001 f2fs_inode_synced(inode);
1002
1003 /* don't make bad inode, since it becomes a regular file. */
1004 unlock_new_inode(inode);
1005
1006 /*
1007 * Note: we should add inode to orphan list before f2fs_unlock_op()
1008 * so we can prevent losing this orphan when encoutering checkpoint
1009 * and following suddenly power-off.
1010 */
1011 err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
1012 if (err) {
1013 set_sbi_flag(sbi, SBI_NEED_FSCK);
1014 set_inode_flag(inode, FI_FREE_NID);
1015 f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
1016 goto out;
1017 }
1018
1019 if (ni.blk_addr != NULL_ADDR) {
1020 err = f2fs_acquire_orphan_inode(sbi);
1021 if (err) {
1022 set_sbi_flag(sbi, SBI_NEED_FSCK);
1023 f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
1024 } else {
1025 f2fs_add_orphan_inode(inode);
1026 }
1027 f2fs_alloc_nid_done(sbi, inode->i_ino);
1028 } else {
1029 set_inode_flag(inode, FI_FREE_NID);
1030 }
1031
1032 out:
1033 f2fs_unlock_op(sbi);
1034
1035 /* iput will drop the inode object */
1036 iput(inode);
1037 }
1038