xref: /linux/fs/f2fs/inode.c (revision c8b90d40d5bba8e6fba457b8a7c10d3c0d467e37)
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 
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 	if (f2fs_is_atomic_file(inode)) {
38 		set_inode_flag(inode, FI_ATOMIC_DIRTIED);
39 		return;
40 	}
41 
42 	mark_inode_dirty_sync(inode);
43 }
44 
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 
71 static void __get_inode_rdev(struct inode *inode, struct page *node_page)
72 {
73 	__le32 *addr = get_dnode_addr(inode, node_page);
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 
84 static void __set_inode_rdev(struct inode *inode, struct page *node_page)
85 {
86 	__le32 *addr = get_dnode_addr(inode, node_page);
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 
100 static void __recover_inline_status(struct inode *inode, struct page *ipage)
101 {
102 	void *inline_data = inline_data_addr(inode, ipage);
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_wait_on_page_writeback(ipage, NODE, true, true);
109 
110 			set_inode_flag(inode, FI_DATA_EXIST);
111 			set_raw_inline(inode, F2FS_INODE(ipage));
112 			set_page_dirty(ipage);
113 			return;
114 		}
115 	}
116 	return;
117 }
118 
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 
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, sbi->s_chksum_seed, (__u8 *)&ino,
148 							sizeof(ino));
149 	chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
150 
151 	chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
152 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
153 	offset += cs_size;
154 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
155 						F2FS_BLKSIZE - offset);
156 	return chksum;
157 }
158 
159 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
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, page))
169 #else
170 	if (!f2fs_enable_inode_chksum(sbi, page) ||
171 			PageDirty(page) ||
172 			folio_test_writeback(page_folio(page)))
173 #endif
174 		return true;
175 
176 	ri = &F2FS_NODE(page)->i;
177 	provided = le32_to_cpu(ri->i_inode_checksum);
178 	calculated = f2fs_inode_chksum(sbi, page);
179 
180 	if (provided != calculated)
181 		f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
182 			  page_folio(page)->index, ino_of_node(page),
183 			  provided, calculated);
184 
185 	return provided == calculated;
186 }
187 
188 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
189 {
190 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
191 
192 	if (!f2fs_enable_inode_chksum(sbi, page))
193 		return;
194 
195 	ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
196 }
197 
198 static bool sanity_check_compress_inode(struct inode *inode,
199 			struct f2fs_inode *ri)
200 {
201 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
202 	unsigned char clevel;
203 
204 	if (ri->i_compress_algorithm >= COMPRESS_MAX) {
205 		f2fs_warn(sbi,
206 			"%s: inode (ino=%lx) has unsupported compress algorithm: %u, run fsck to fix",
207 			__func__, inode->i_ino, ri->i_compress_algorithm);
208 		return false;
209 	}
210 	if (le64_to_cpu(ri->i_compr_blocks) >
211 			SECTOR_TO_BLOCK(inode->i_blocks)) {
212 		f2fs_warn(sbi,
213 			"%s: inode (ino=%lx) has inconsistent i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
214 			__func__, inode->i_ino, le64_to_cpu(ri->i_compr_blocks),
215 			SECTOR_TO_BLOCK(inode->i_blocks));
216 		return false;
217 	}
218 	if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
219 		ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
220 		f2fs_warn(sbi,
221 			"%s: inode (ino=%lx) has unsupported log cluster size: %u, run fsck to fix",
222 			__func__, inode->i_ino, ri->i_log_cluster_size);
223 		return false;
224 	}
225 
226 	clevel = le16_to_cpu(ri->i_compress_flag) >>
227 				COMPRESS_LEVEL_OFFSET;
228 	switch (ri->i_compress_algorithm) {
229 	case COMPRESS_LZO:
230 #ifdef CONFIG_F2FS_FS_LZO
231 		if (clevel)
232 			goto err_level;
233 #endif
234 		break;
235 	case COMPRESS_LZORLE:
236 #ifdef CONFIG_F2FS_FS_LZORLE
237 		if (clevel)
238 			goto err_level;
239 #endif
240 		break;
241 	case COMPRESS_LZ4:
242 #ifdef CONFIG_F2FS_FS_LZ4
243 #ifdef CONFIG_F2FS_FS_LZ4HC
244 		if (clevel &&
245 		   (clevel < LZ4HC_MIN_CLEVEL || clevel > LZ4HC_MAX_CLEVEL))
246 			goto err_level;
247 #else
248 		if (clevel)
249 			goto err_level;
250 #endif
251 #endif
252 		break;
253 	case COMPRESS_ZSTD:
254 #ifdef CONFIG_F2FS_FS_ZSTD
255 		if (clevel < zstd_min_clevel() || clevel > zstd_max_clevel())
256 			goto err_level;
257 #endif
258 		break;
259 	default:
260 		goto err_level;
261 	}
262 
263 	return true;
264 err_level:
265 	f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported compress level: %u, run fsck to fix",
266 		  __func__, inode->i_ino, clevel);
267 	return false;
268 }
269 
270 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
271 {
272 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
273 	struct f2fs_inode_info *fi = F2FS_I(inode);
274 	struct f2fs_inode *ri = F2FS_INODE(node_page);
275 	unsigned long long iblocks;
276 
277 	iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
278 	if (!iblocks) {
279 		f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
280 			  __func__, inode->i_ino, iblocks);
281 		return false;
282 	}
283 
284 	if (ino_of_node(node_page) != nid_of_node(node_page)) {
285 		f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
286 			  __func__, inode->i_ino,
287 			  ino_of_node(node_page), nid_of_node(node_page));
288 		return false;
289 	}
290 
291 	if (f2fs_has_extra_attr(inode)) {
292 		if (!f2fs_sb_has_extra_attr(sbi)) {
293 			f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
294 				  __func__, inode->i_ino);
295 			return false;
296 		}
297 		if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
298 			fi->i_extra_isize < F2FS_MIN_EXTRA_ATTR_SIZE ||
299 			fi->i_extra_isize % sizeof(__le32)) {
300 			f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
301 				  __func__, inode->i_ino, fi->i_extra_isize,
302 				  F2FS_TOTAL_EXTRA_ATTR_SIZE);
303 			return false;
304 		}
305 		if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
306 			f2fs_has_inline_xattr(inode) &&
307 			(!fi->i_inline_xattr_size ||
308 			fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
309 			f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %lu",
310 				  __func__, inode->i_ino, fi->i_inline_xattr_size,
311 				  MAX_INLINE_XATTR_SIZE);
312 			return false;
313 		}
314 		if (f2fs_sb_has_compression(sbi) &&
315 			fi->i_flags & F2FS_COMPR_FL &&
316 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
317 						i_compress_flag)) {
318 			if (!sanity_check_compress_inode(inode, ri))
319 				return false;
320 		}
321 	}
322 
323 	if (!f2fs_sb_has_extra_attr(sbi)) {
324 		if (f2fs_sb_has_project_quota(sbi)) {
325 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
326 				  __func__, inode->i_ino, F2FS_FEATURE_PRJQUOTA);
327 			return false;
328 		}
329 		if (f2fs_sb_has_inode_chksum(sbi)) {
330 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
331 				  __func__, inode->i_ino, F2FS_FEATURE_INODE_CHKSUM);
332 			return false;
333 		}
334 		if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
335 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
336 				  __func__, inode->i_ino, F2FS_FEATURE_FLEXIBLE_INLINE_XATTR);
337 			return false;
338 		}
339 		if (f2fs_sb_has_inode_crtime(sbi)) {
340 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
341 				  __func__, inode->i_ino, F2FS_FEATURE_INODE_CRTIME);
342 			return false;
343 		}
344 		if (f2fs_sb_has_compression(sbi)) {
345 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
346 				  __func__, inode->i_ino, F2FS_FEATURE_COMPRESSION);
347 			return false;
348 		}
349 	}
350 
351 	if (f2fs_sanity_check_inline_data(inode, node_page)) {
352 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
353 			  __func__, inode->i_ino, inode->i_mode);
354 		return false;
355 	}
356 
357 	if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
358 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
359 			  __func__, inode->i_ino, inode->i_mode);
360 		return false;
361 	}
362 
363 	if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
364 		f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
365 			  __func__, inode->i_ino);
366 		return false;
367 	}
368 
369 	if (fi->i_xattr_nid && f2fs_check_nid_range(sbi, fi->i_xattr_nid)) {
370 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_xattr_nid: %u, run fsck to fix.",
371 			  __func__, inode->i_ino, fi->i_xattr_nid);
372 		return false;
373 	}
374 
375 	if (IS_DEVICE_ALIASING(inode)) {
376 		if (!f2fs_sb_has_device_alias(sbi)) {
377 			f2fs_warn(sbi, "%s: inode (ino=%lx) has device alias flag, but the feature is off",
378 				  __func__, inode->i_ino);
379 			return false;
380 		}
381 		if (!f2fs_is_pinned_file(inode)) {
382 			f2fs_warn(sbi, "%s: inode (ino=%lx) has device alias flag, but is not pinned",
383 				  __func__, inode->i_ino);
384 			return false;
385 		}
386 	}
387 
388 	return true;
389 }
390 
391 static void init_idisk_time(struct inode *inode)
392 {
393 	struct f2fs_inode_info *fi = F2FS_I(inode);
394 
395 	fi->i_disk_time[0] = inode_get_atime(inode);
396 	fi->i_disk_time[1] = inode_get_ctime(inode);
397 	fi->i_disk_time[2] = inode_get_mtime(inode);
398 }
399 
400 static int do_read_inode(struct inode *inode)
401 {
402 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
403 	struct f2fs_inode_info *fi = F2FS_I(inode);
404 	struct page *node_page;
405 	struct f2fs_inode *ri;
406 	projid_t i_projid;
407 
408 	/* Check if ino is within scope */
409 	if (f2fs_check_nid_range(sbi, inode->i_ino))
410 		return -EINVAL;
411 
412 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
413 	if (IS_ERR(node_page))
414 		return PTR_ERR(node_page);
415 
416 	ri = F2FS_INODE(node_page);
417 
418 	inode->i_mode = le16_to_cpu(ri->i_mode);
419 	i_uid_write(inode, le32_to_cpu(ri->i_uid));
420 	i_gid_write(inode, le32_to_cpu(ri->i_gid));
421 	set_nlink(inode, le32_to_cpu(ri->i_links));
422 	inode->i_size = le64_to_cpu(ri->i_size);
423 	inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
424 
425 	inode_set_atime(inode, le64_to_cpu(ri->i_atime),
426 			le32_to_cpu(ri->i_atime_nsec));
427 	inode_set_ctime(inode, le64_to_cpu(ri->i_ctime),
428 			le32_to_cpu(ri->i_ctime_nsec));
429 	inode_set_mtime(inode, le64_to_cpu(ri->i_mtime),
430 			le32_to_cpu(ri->i_mtime_nsec));
431 	inode->i_generation = le32_to_cpu(ri->i_generation);
432 	if (S_ISDIR(inode->i_mode))
433 		fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
434 	else if (S_ISREG(inode->i_mode))
435 		fi->i_gc_failures = le16_to_cpu(ri->i_gc_failures);
436 	fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
437 	fi->i_flags = le32_to_cpu(ri->i_flags);
438 	if (S_ISREG(inode->i_mode))
439 		fi->i_flags &= ~F2FS_PROJINHERIT_FL;
440 	bitmap_zero(fi->flags, FI_MAX);
441 	fi->i_advise = ri->i_advise;
442 	fi->i_pino = le32_to_cpu(ri->i_pino);
443 	fi->i_dir_level = ri->i_dir_level;
444 
445 	get_inline_info(inode, ri);
446 
447 	fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
448 					le16_to_cpu(ri->i_extra_isize) : 0;
449 
450 	if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
451 		fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
452 	} else if (f2fs_has_inline_xattr(inode) ||
453 				f2fs_has_inline_dentry(inode)) {
454 		fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
455 	} else {
456 
457 		/*
458 		 * Previous inline data or directory always reserved 200 bytes
459 		 * in inode layout, even if inline_xattr is disabled. In order
460 		 * to keep inline_dentry's structure for backward compatibility,
461 		 * we get the space back only from inline_data.
462 		 */
463 		fi->i_inline_xattr_size = 0;
464 	}
465 
466 	if (!sanity_check_inode(inode, node_page)) {
467 		f2fs_put_page(node_page, 1);
468 		set_sbi_flag(sbi, SBI_NEED_FSCK);
469 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
470 		return -EFSCORRUPTED;
471 	}
472 
473 	/* check data exist */
474 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
475 		__recover_inline_status(inode, node_page);
476 
477 	/* try to recover cold bit for non-dir inode */
478 	if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
479 		f2fs_wait_on_page_writeback(node_page, NODE, true, true);
480 		set_cold_node(node_page, false);
481 		set_page_dirty(node_page);
482 	}
483 
484 	/* get rdev by using inline_info */
485 	__get_inode_rdev(inode, node_page);
486 
487 	if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
488 		fi->last_disk_size = inode->i_size;
489 
490 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
491 		set_inode_flag(inode, FI_PROJ_INHERIT);
492 
493 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
494 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
495 		i_projid = (projid_t)le32_to_cpu(ri->i_projid);
496 	else
497 		i_projid = F2FS_DEF_PROJID;
498 	fi->i_projid = make_kprojid(&init_user_ns, i_projid);
499 
500 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
501 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
502 		fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
503 		fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
504 	}
505 
506 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
507 					(fi->i_flags & F2FS_COMPR_FL)) {
508 		if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
509 					i_compress_flag)) {
510 			unsigned short compress_flag;
511 
512 			atomic_set(&fi->i_compr_blocks,
513 					le64_to_cpu(ri->i_compr_blocks));
514 			fi->i_compress_algorithm = ri->i_compress_algorithm;
515 			fi->i_log_cluster_size = ri->i_log_cluster_size;
516 			compress_flag = le16_to_cpu(ri->i_compress_flag);
517 			fi->i_compress_level = compress_flag >>
518 						COMPRESS_LEVEL_OFFSET;
519 			fi->i_compress_flag = compress_flag &
520 					GENMASK(COMPRESS_LEVEL_OFFSET - 1, 0);
521 			fi->i_cluster_size = BIT(fi->i_log_cluster_size);
522 			set_inode_flag(inode, FI_COMPRESSED_FILE);
523 		}
524 	}
525 
526 	init_idisk_time(inode);
527 
528 	if (!sanity_check_extent_cache(inode, node_page)) {
529 		f2fs_put_page(node_page, 1);
530 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
531 		return -EFSCORRUPTED;
532 	}
533 
534 	/* Need all the flag bits */
535 	f2fs_init_read_extent_tree(inode, node_page);
536 	f2fs_init_age_extent_tree(inode);
537 
538 	f2fs_put_page(node_page, 1);
539 
540 	stat_inc_inline_xattr(inode);
541 	stat_inc_inline_inode(inode);
542 	stat_inc_inline_dir(inode);
543 	stat_inc_compr_inode(inode);
544 	stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
545 
546 	return 0;
547 }
548 
549 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
550 {
551 	return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
552 		ino == F2FS_COMPRESS_INO(sbi);
553 }
554 
555 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
556 {
557 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
558 	struct inode *inode;
559 	int ret = 0;
560 
561 	inode = iget_locked(sb, ino);
562 	if (!inode)
563 		return ERR_PTR(-ENOMEM);
564 
565 	if (!(inode->i_state & I_NEW)) {
566 		if (is_meta_ino(sbi, ino)) {
567 			f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
568 			set_sbi_flag(sbi, SBI_NEED_FSCK);
569 			ret = -EFSCORRUPTED;
570 			trace_f2fs_iget_exit(inode, ret);
571 			iput(inode);
572 			f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
573 			return ERR_PTR(ret);
574 		}
575 
576 		trace_f2fs_iget(inode);
577 		return inode;
578 	}
579 
580 	if (is_meta_ino(sbi, ino))
581 		goto make_now;
582 
583 	ret = do_read_inode(inode);
584 	if (ret)
585 		goto bad_inode;
586 make_now:
587 	if (ino == F2FS_NODE_INO(sbi)) {
588 		inode->i_mapping->a_ops = &f2fs_node_aops;
589 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
590 	} else if (ino == F2FS_META_INO(sbi)) {
591 		inode->i_mapping->a_ops = &f2fs_meta_aops;
592 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
593 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
594 #ifdef CONFIG_F2FS_FS_COMPRESSION
595 		inode->i_mapping->a_ops = &f2fs_compress_aops;
596 		/*
597 		 * generic_error_remove_folio only truncates pages of regular
598 		 * inode
599 		 */
600 		inode->i_mode |= S_IFREG;
601 #endif
602 		mapping_set_gfp_mask(inode->i_mapping,
603 			GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
604 	} else if (S_ISREG(inode->i_mode)) {
605 		inode->i_op = &f2fs_file_inode_operations;
606 		inode->i_fop = &f2fs_file_operations;
607 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
608 	} else if (S_ISDIR(inode->i_mode)) {
609 		inode->i_op = &f2fs_dir_inode_operations;
610 		inode->i_fop = &f2fs_dir_operations;
611 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
612 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
613 	} else if (S_ISLNK(inode->i_mode)) {
614 		if (file_is_encrypt(inode))
615 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
616 		else
617 			inode->i_op = &f2fs_symlink_inode_operations;
618 		inode_nohighmem(inode);
619 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
620 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
621 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
622 		inode->i_op = &f2fs_special_inode_operations;
623 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
624 	} else {
625 		ret = -EIO;
626 		goto bad_inode;
627 	}
628 	f2fs_set_inode_flags(inode);
629 
630 	unlock_new_inode(inode);
631 	trace_f2fs_iget(inode);
632 	return inode;
633 
634 bad_inode:
635 	f2fs_inode_synced(inode);
636 	iget_failed(inode);
637 	trace_f2fs_iget_exit(inode, ret);
638 	return ERR_PTR(ret);
639 }
640 
641 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
642 {
643 	struct inode *inode;
644 retry:
645 	inode = f2fs_iget(sb, ino);
646 	if (IS_ERR(inode)) {
647 		if (PTR_ERR(inode) == -ENOMEM) {
648 			memalloc_retry_wait(GFP_NOFS);
649 			goto retry;
650 		}
651 	}
652 	return inode;
653 }
654 
655 void f2fs_update_inode(struct inode *inode, struct page *node_page)
656 {
657 	struct f2fs_inode_info *fi = F2FS_I(inode);
658 	struct f2fs_inode *ri;
659 	struct extent_tree *et = fi->extent_tree[EX_READ];
660 
661 	f2fs_wait_on_page_writeback(node_page, NODE, true, true);
662 	set_page_dirty(node_page);
663 
664 	f2fs_inode_synced(inode);
665 
666 	ri = F2FS_INODE(node_page);
667 
668 	ri->i_mode = cpu_to_le16(inode->i_mode);
669 	ri->i_advise = fi->i_advise;
670 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
671 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
672 	ri->i_links = cpu_to_le32(inode->i_nlink);
673 	ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
674 
675 	if (!f2fs_is_atomic_file(inode) ||
676 			is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
677 		ri->i_size = cpu_to_le64(i_size_read(inode));
678 
679 	if (et) {
680 		read_lock(&et->lock);
681 		set_raw_read_extent(&et->largest, &ri->i_ext);
682 		read_unlock(&et->lock);
683 	} else {
684 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
685 	}
686 	set_raw_inline(inode, ri);
687 
688 	ri->i_atime = cpu_to_le64(inode_get_atime_sec(inode));
689 	ri->i_ctime = cpu_to_le64(inode_get_ctime_sec(inode));
690 	ri->i_mtime = cpu_to_le64(inode_get_mtime_sec(inode));
691 	ri->i_atime_nsec = cpu_to_le32(inode_get_atime_nsec(inode));
692 	ri->i_ctime_nsec = cpu_to_le32(inode_get_ctime_nsec(inode));
693 	ri->i_mtime_nsec = cpu_to_le32(inode_get_mtime_nsec(inode));
694 	if (S_ISDIR(inode->i_mode))
695 		ri->i_current_depth = cpu_to_le32(fi->i_current_depth);
696 	else if (S_ISREG(inode->i_mode))
697 		ri->i_gc_failures = cpu_to_le16(fi->i_gc_failures);
698 	ri->i_xattr_nid = cpu_to_le32(fi->i_xattr_nid);
699 	ri->i_flags = cpu_to_le32(fi->i_flags);
700 	ri->i_pino = cpu_to_le32(fi->i_pino);
701 	ri->i_generation = cpu_to_le32(inode->i_generation);
702 	ri->i_dir_level = fi->i_dir_level;
703 
704 	if (f2fs_has_extra_attr(inode)) {
705 		ri->i_extra_isize = cpu_to_le16(fi->i_extra_isize);
706 
707 		if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
708 			ri->i_inline_xattr_size =
709 				cpu_to_le16(fi->i_inline_xattr_size);
710 
711 		if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
712 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) {
713 			projid_t i_projid;
714 
715 			i_projid = from_kprojid(&init_user_ns, fi->i_projid);
716 			ri->i_projid = cpu_to_le32(i_projid);
717 		}
718 
719 		if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
720 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
721 			ri->i_crtime = cpu_to_le64(fi->i_crtime.tv_sec);
722 			ri->i_crtime_nsec = cpu_to_le32(fi->i_crtime.tv_nsec);
723 		}
724 
725 		if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
726 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
727 							i_compress_flag)) {
728 			unsigned short compress_flag;
729 
730 			ri->i_compr_blocks = cpu_to_le64(
731 					atomic_read(&fi->i_compr_blocks));
732 			ri->i_compress_algorithm = fi->i_compress_algorithm;
733 			compress_flag = fi->i_compress_flag |
734 						fi->i_compress_level <<
735 						COMPRESS_LEVEL_OFFSET;
736 			ri->i_compress_flag = cpu_to_le16(compress_flag);
737 			ri->i_log_cluster_size = fi->i_log_cluster_size;
738 		}
739 	}
740 
741 	__set_inode_rdev(inode, node_page);
742 
743 	/* deleted inode */
744 	if (inode->i_nlink == 0)
745 		clear_page_private_inline(node_page);
746 
747 	init_idisk_time(inode);
748 #ifdef CONFIG_F2FS_CHECK_FS
749 	f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
750 #endif
751 }
752 
753 void f2fs_update_inode_page(struct inode *inode)
754 {
755 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
756 	struct page *node_page;
757 	int count = 0;
758 retry:
759 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
760 	if (IS_ERR(node_page)) {
761 		int err = PTR_ERR(node_page);
762 
763 		/* The node block was truncated. */
764 		if (err == -ENOENT)
765 			return;
766 
767 		if (err == -ENOMEM || ++count <= DEFAULT_RETRY_IO_COUNT)
768 			goto retry;
769 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_UPDATE_INODE);
770 		return;
771 	}
772 	f2fs_update_inode(inode, node_page);
773 	f2fs_put_page(node_page, 1);
774 }
775 
776 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
777 {
778 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
779 
780 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
781 			inode->i_ino == F2FS_META_INO(sbi))
782 		return 0;
783 
784 	/*
785 	 * atime could be updated without dirtying f2fs inode in lazytime mode
786 	 */
787 	if (f2fs_is_time_consistent(inode) &&
788 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
789 		return 0;
790 
791 	if (!f2fs_is_checkpoint_ready(sbi)) {
792 		f2fs_mark_inode_dirty_sync(inode, true);
793 		return -ENOSPC;
794 	}
795 
796 	/*
797 	 * We need to balance fs here to prevent from producing dirty node pages
798 	 * during the urgent cleaning time when running out of free sections.
799 	 */
800 	f2fs_update_inode_page(inode);
801 	if (wbc && wbc->nr_to_write)
802 		f2fs_balance_fs(sbi, true);
803 	return 0;
804 }
805 
806 /*
807  * Called at the last iput() if i_nlink is zero
808  */
809 void f2fs_evict_inode(struct inode *inode)
810 {
811 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
812 	struct f2fs_inode_info *fi = F2FS_I(inode);
813 	nid_t xnid = fi->i_xattr_nid;
814 	int err = 0;
815 	bool freeze_protected = false;
816 
817 	f2fs_abort_atomic_write(inode, true);
818 
819 	if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
820 		clear_inode_flag(fi->cow_inode, FI_COW_FILE);
821 		F2FS_I(fi->cow_inode)->atomic_inode = NULL;
822 		iput(fi->cow_inode);
823 		fi->cow_inode = NULL;
824 	}
825 
826 	trace_f2fs_evict_inode(inode);
827 	truncate_inode_pages_final(&inode->i_data);
828 
829 	if ((inode->i_nlink || is_bad_inode(inode)) &&
830 		test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
831 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
832 
833 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
834 			inode->i_ino == F2FS_META_INO(sbi) ||
835 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
836 		goto out_clear;
837 
838 	f2fs_bug_on(sbi, get_dirty_pages(inode));
839 	f2fs_remove_dirty_inode(inode);
840 
841 	if (!IS_DEVICE_ALIASING(inode))
842 		f2fs_destroy_extent_tree(inode);
843 
844 	if (inode->i_nlink || is_bad_inode(inode))
845 		goto no_delete;
846 
847 	err = f2fs_dquot_initialize(inode);
848 	if (err) {
849 		err = 0;
850 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
851 	}
852 
853 	f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
854 	f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
855 	f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
856 
857 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING)) {
858 		sb_start_intwrite(inode->i_sb);
859 		freeze_protected = true;
860 	}
861 	set_inode_flag(inode, FI_NO_ALLOC);
862 	i_size_write(inode, 0);
863 retry:
864 	if (F2FS_HAS_BLOCKS(inode))
865 		err = f2fs_truncate(inode);
866 
867 	if (time_to_inject(sbi, FAULT_EVICT_INODE))
868 		err = -EIO;
869 
870 	if (!err) {
871 		f2fs_lock_op(sbi);
872 		err = f2fs_remove_inode_page(inode);
873 		f2fs_unlock_op(sbi);
874 		if (err == -ENOENT) {
875 			err = 0;
876 
877 			/*
878 			 * in fuzzed image, another node may has the same
879 			 * block address as inode's, if it was truncated
880 			 * previously, truncation of inode node will fail.
881 			 */
882 			if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
883 				f2fs_warn(F2FS_I_SB(inode),
884 					"f2fs_evict_inode: inconsistent node id, ino:%lu",
885 					inode->i_ino);
886 				f2fs_inode_synced(inode);
887 				set_sbi_flag(sbi, SBI_NEED_FSCK);
888 			}
889 		}
890 	}
891 
892 	/* give more chances, if ENOMEM case */
893 	if (err == -ENOMEM) {
894 		err = 0;
895 		goto retry;
896 	}
897 
898 	if (IS_DEVICE_ALIASING(inode))
899 		f2fs_destroy_extent_tree(inode);
900 
901 	if (err) {
902 		f2fs_update_inode_page(inode);
903 		if (dquot_initialize_needed(inode))
904 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
905 	}
906 	if (freeze_protected)
907 		sb_end_intwrite(inode->i_sb);
908 no_delete:
909 	dquot_drop(inode);
910 
911 	stat_dec_inline_xattr(inode);
912 	stat_dec_inline_dir(inode);
913 	stat_dec_inline_inode(inode);
914 	stat_dec_compr_inode(inode);
915 	stat_sub_compr_blocks(inode,
916 			atomic_read(&fi->i_compr_blocks));
917 
918 	if (likely(!f2fs_cp_error(sbi) &&
919 				!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
920 		f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
921 	else
922 		f2fs_inode_synced(inode);
923 
924 	/* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
925 	if (inode->i_ino)
926 		invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
927 							inode->i_ino);
928 	if (xnid)
929 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
930 	if (inode->i_nlink) {
931 		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
932 			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
933 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
934 			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
935 	}
936 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
937 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
938 		clear_inode_flag(inode, FI_FREE_NID);
939 	} else {
940 		/*
941 		 * If xattr nid is corrupted, we can reach out error condition,
942 		 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
943 		 * In that case, f2fs_check_nid_range() is enough to give a clue.
944 		 */
945 	}
946 out_clear:
947 	fscrypt_put_encryption_info(inode);
948 	fsverity_cleanup_inode(inode);
949 	clear_inode(inode);
950 }
951 
952 /* caller should call f2fs_lock_op() */
953 void f2fs_handle_failed_inode(struct inode *inode)
954 {
955 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
956 	struct node_info ni;
957 	int err;
958 
959 	/*
960 	 * clear nlink of inode in order to release resource of inode
961 	 * immediately.
962 	 */
963 	clear_nlink(inode);
964 
965 	/*
966 	 * we must call this to avoid inode being remained as dirty, resulting
967 	 * in a panic when flushing dirty inodes in gdirty_list.
968 	 */
969 	f2fs_update_inode_page(inode);
970 	f2fs_inode_synced(inode);
971 
972 	/* don't make bad inode, since it becomes a regular file. */
973 	unlock_new_inode(inode);
974 
975 	/*
976 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
977 	 * so we can prevent losing this orphan when encoutering checkpoint
978 	 * and following suddenly power-off.
979 	 */
980 	err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
981 	if (err) {
982 		set_sbi_flag(sbi, SBI_NEED_FSCK);
983 		set_inode_flag(inode, FI_FREE_NID);
984 		f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
985 		goto out;
986 	}
987 
988 	if (ni.blk_addr != NULL_ADDR) {
989 		err = f2fs_acquire_orphan_inode(sbi);
990 		if (err) {
991 			set_sbi_flag(sbi, SBI_NEED_FSCK);
992 			f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
993 		} else {
994 			f2fs_add_orphan_inode(inode);
995 		}
996 		f2fs_alloc_nid_done(sbi, inode->i_ino);
997 	} else {
998 		set_inode_flag(inode, FI_FREE_NID);
999 	}
1000 
1001 out:
1002 	f2fs_unlock_op(sbi);
1003 
1004 	/* iput will drop the inode object */
1005 	iput(inode);
1006 }
1007