xref: /openbmc/linux/fs/f2fs/inode.c (revision 54618888)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/inode.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/buffer_head.h>
11 #include <linux/writeback.h>
12 #include <linux/sched/mm.h>
13 
14 #include "f2fs.h"
15 #include "node.h"
16 #include "segment.h"
17 #include "xattr.h"
18 
19 #include <trace/events/f2fs.h>
20 
21 #ifdef CONFIG_F2FS_FS_COMPRESSION
22 extern const struct address_space_operations f2fs_compress_aops;
23 #endif
24 
25 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
26 {
27 	if (is_inode_flag_set(inode, FI_NEW_INODE))
28 		return;
29 
30 	if (f2fs_inode_dirtied(inode, sync))
31 		return;
32 
33 	mark_inode_dirty_sync(inode);
34 }
35 
36 void f2fs_set_inode_flags(struct inode *inode)
37 {
38 	unsigned int flags = F2FS_I(inode)->i_flags;
39 	unsigned int new_fl = 0;
40 
41 	if (flags & F2FS_SYNC_FL)
42 		new_fl |= S_SYNC;
43 	if (flags & F2FS_APPEND_FL)
44 		new_fl |= S_APPEND;
45 	if (flags & F2FS_IMMUTABLE_FL)
46 		new_fl |= S_IMMUTABLE;
47 	if (flags & F2FS_NOATIME_FL)
48 		new_fl |= S_NOATIME;
49 	if (flags & F2FS_DIRSYNC_FL)
50 		new_fl |= S_DIRSYNC;
51 	if (file_is_encrypt(inode))
52 		new_fl |= S_ENCRYPTED;
53 	if (file_is_verity(inode))
54 		new_fl |= S_VERITY;
55 	if (flags & F2FS_CASEFOLD_FL)
56 		new_fl |= S_CASEFOLD;
57 	inode_set_flags(inode, new_fl,
58 			S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|
59 			S_ENCRYPTED|S_VERITY|S_CASEFOLD);
60 }
61 
62 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
63 {
64 	int extra_size = get_extra_isize(inode);
65 
66 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
67 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
68 		if (ri->i_addr[extra_size])
69 			inode->i_rdev = old_decode_dev(
70 				le32_to_cpu(ri->i_addr[extra_size]));
71 		else
72 			inode->i_rdev = new_decode_dev(
73 				le32_to_cpu(ri->i_addr[extra_size + 1]));
74 	}
75 }
76 
77 static int __written_first_block(struct f2fs_sb_info *sbi,
78 					struct f2fs_inode *ri)
79 {
80 	block_t addr = le32_to_cpu(ri->i_addr[offset_in_addr(ri)]);
81 
82 	if (!__is_valid_data_blkaddr(addr))
83 		return 1;
84 	if (!f2fs_is_valid_blkaddr(sbi, addr, DATA_GENERIC_ENHANCE)) {
85 		f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
86 		return -EFSCORRUPTED;
87 	}
88 	return 0;
89 }
90 
91 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
92 {
93 	int extra_size = get_extra_isize(inode);
94 
95 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
96 		if (old_valid_dev(inode->i_rdev)) {
97 			ri->i_addr[extra_size] =
98 				cpu_to_le32(old_encode_dev(inode->i_rdev));
99 			ri->i_addr[extra_size + 1] = 0;
100 		} else {
101 			ri->i_addr[extra_size] = 0;
102 			ri->i_addr[extra_size + 1] =
103 				cpu_to_le32(new_encode_dev(inode->i_rdev));
104 			ri->i_addr[extra_size + 2] = 0;
105 		}
106 	}
107 }
108 
109 static void __recover_inline_status(struct inode *inode, struct page *ipage)
110 {
111 	void *inline_data = inline_data_addr(inode, ipage);
112 	__le32 *start = inline_data;
113 	__le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
114 
115 	while (start < end) {
116 		if (*start++) {
117 			f2fs_wait_on_page_writeback(ipage, NODE, true, true);
118 
119 			set_inode_flag(inode, FI_DATA_EXIST);
120 			set_raw_inline(inode, F2FS_INODE(ipage));
121 			set_page_dirty(ipage);
122 			return;
123 		}
124 	}
125 	return;
126 }
127 
128 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
129 {
130 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
131 
132 	if (!f2fs_sb_has_inode_chksum(sbi))
133 		return false;
134 
135 	if (!IS_INODE(page) || !(ri->i_inline & F2FS_EXTRA_ATTR))
136 		return false;
137 
138 	if (!F2FS_FITS_IN_INODE(ri, le16_to_cpu(ri->i_extra_isize),
139 				i_inode_checksum))
140 		return false;
141 
142 	return true;
143 }
144 
145 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
146 {
147 	struct f2fs_node *node = F2FS_NODE(page);
148 	struct f2fs_inode *ri = &node->i;
149 	__le32 ino = node->footer.ino;
150 	__le32 gen = ri->i_generation;
151 	__u32 chksum, chksum_seed;
152 	__u32 dummy_cs = 0;
153 	unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
154 	unsigned int cs_size = sizeof(dummy_cs);
155 
156 	chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
157 							sizeof(ino));
158 	chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
159 
160 	chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
161 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
162 	offset += cs_size;
163 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
164 						F2FS_BLKSIZE - offset);
165 	return chksum;
166 }
167 
168 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
169 {
170 	struct f2fs_inode *ri;
171 	__u32 provided, calculated;
172 
173 	if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
174 		return true;
175 
176 #ifdef CONFIG_F2FS_CHECK_FS
177 	if (!f2fs_enable_inode_chksum(sbi, page))
178 #else
179 	if (!f2fs_enable_inode_chksum(sbi, page) ||
180 			PageDirty(page) || PageWriteback(page))
181 #endif
182 		return true;
183 
184 	ri = &F2FS_NODE(page)->i;
185 	provided = le32_to_cpu(ri->i_inode_checksum);
186 	calculated = f2fs_inode_chksum(sbi, page);
187 
188 	if (provided != calculated)
189 		f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
190 			  page->index, ino_of_node(page), provided, calculated);
191 
192 	return provided == calculated;
193 }
194 
195 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
196 {
197 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
198 
199 	if (!f2fs_enable_inode_chksum(sbi, page))
200 		return;
201 
202 	ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
203 }
204 
205 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
206 {
207 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
208 	struct f2fs_inode_info *fi = F2FS_I(inode);
209 	struct f2fs_inode *ri = F2FS_INODE(node_page);
210 	unsigned long long iblocks;
211 
212 	iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
213 	if (!iblocks) {
214 		set_sbi_flag(sbi, SBI_NEED_FSCK);
215 		f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
216 			  __func__, inode->i_ino, iblocks);
217 		return false;
218 	}
219 
220 	if (ino_of_node(node_page) != nid_of_node(node_page)) {
221 		set_sbi_flag(sbi, SBI_NEED_FSCK);
222 		f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
223 			  __func__, inode->i_ino,
224 			  ino_of_node(node_page), nid_of_node(node_page));
225 		return false;
226 	}
227 
228 	if (f2fs_sb_has_flexible_inline_xattr(sbi)
229 			&& !f2fs_has_extra_attr(inode)) {
230 		set_sbi_flag(sbi, SBI_NEED_FSCK);
231 		f2fs_warn(sbi, "%s: corrupted inode ino=%lx, run fsck to fix.",
232 			  __func__, inode->i_ino);
233 		return false;
234 	}
235 
236 	if (f2fs_has_extra_attr(inode) &&
237 			!f2fs_sb_has_extra_attr(sbi)) {
238 		set_sbi_flag(sbi, SBI_NEED_FSCK);
239 		f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
240 			  __func__, inode->i_ino);
241 		return false;
242 	}
243 
244 	if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
245 			fi->i_extra_isize % sizeof(__le32)) {
246 		set_sbi_flag(sbi, SBI_NEED_FSCK);
247 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
248 			  __func__, inode->i_ino, fi->i_extra_isize,
249 			  F2FS_TOTAL_EXTRA_ATTR_SIZE);
250 		return false;
251 	}
252 
253 	if (f2fs_has_extra_attr(inode) &&
254 		f2fs_sb_has_flexible_inline_xattr(sbi) &&
255 		f2fs_has_inline_xattr(inode) &&
256 		(!fi->i_inline_xattr_size ||
257 		fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
258 		set_sbi_flag(sbi, SBI_NEED_FSCK);
259 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %zu",
260 			  __func__, inode->i_ino, fi->i_inline_xattr_size,
261 			  MAX_INLINE_XATTR_SIZE);
262 		return false;
263 	}
264 
265 	if (fi->extent_tree[EX_READ]) {
266 		struct extent_info *ei = &fi->extent_tree[EX_READ]->largest;
267 
268 		if (ei->len &&
269 			(!f2fs_is_valid_blkaddr(sbi, ei->blk,
270 						DATA_GENERIC_ENHANCE) ||
271 			!f2fs_is_valid_blkaddr(sbi, ei->blk + ei->len - 1,
272 						DATA_GENERIC_ENHANCE))) {
273 			set_sbi_flag(sbi, SBI_NEED_FSCK);
274 			f2fs_warn(sbi, "%s: inode (ino=%lx) extent info [%u, %u, %u] is incorrect, run fsck to fix",
275 				  __func__, inode->i_ino,
276 				  ei->blk, ei->fofs, ei->len);
277 			return false;
278 		}
279 	}
280 
281 	if (f2fs_sanity_check_inline_data(inode)) {
282 		set_sbi_flag(sbi, SBI_NEED_FSCK);
283 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
284 			  __func__, inode->i_ino, inode->i_mode);
285 		return false;
286 	}
287 
288 	if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
289 		set_sbi_flag(sbi, SBI_NEED_FSCK);
290 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
291 			  __func__, inode->i_ino, inode->i_mode);
292 		return false;
293 	}
294 
295 	if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
296 		set_sbi_flag(sbi, SBI_NEED_FSCK);
297 		f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
298 			  __func__, inode->i_ino);
299 		return false;
300 	}
301 
302 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
303 			fi->i_flags & F2FS_COMPR_FL &&
304 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
305 						i_log_cluster_size)) {
306 		if (ri->i_compress_algorithm >= COMPRESS_MAX) {
307 			set_sbi_flag(sbi, SBI_NEED_FSCK);
308 			f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported "
309 				"compress algorithm: %u, run fsck to fix",
310 				  __func__, inode->i_ino,
311 				  ri->i_compress_algorithm);
312 			return false;
313 		}
314 		if (le64_to_cpu(ri->i_compr_blocks) >
315 				SECTOR_TO_BLOCK(inode->i_blocks)) {
316 			set_sbi_flag(sbi, SBI_NEED_FSCK);
317 			f2fs_warn(sbi, "%s: inode (ino=%lx) has inconsistent "
318 				"i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
319 				  __func__, inode->i_ino,
320 				  le64_to_cpu(ri->i_compr_blocks),
321 				  SECTOR_TO_BLOCK(inode->i_blocks));
322 			return false;
323 		}
324 		if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
325 			ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
326 			set_sbi_flag(sbi, SBI_NEED_FSCK);
327 			f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported "
328 				"log cluster size: %u, run fsck to fix",
329 				  __func__, inode->i_ino,
330 				  ri->i_log_cluster_size);
331 			return false;
332 		}
333 	}
334 
335 	return true;
336 }
337 
338 static void init_idisk_time(struct inode *inode)
339 {
340 	struct f2fs_inode_info *fi = F2FS_I(inode);
341 
342 	fi->i_disk_time[0] = inode->i_atime;
343 	fi->i_disk_time[1] = inode->i_ctime;
344 	fi->i_disk_time[2] = inode->i_mtime;
345 	fi->i_disk_time[3] = fi->i_crtime;
346 }
347 
348 static int do_read_inode(struct inode *inode)
349 {
350 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
351 	struct f2fs_inode_info *fi = F2FS_I(inode);
352 	struct page *node_page;
353 	struct f2fs_inode *ri;
354 	projid_t i_projid;
355 	int err;
356 
357 	/* Check if ino is within scope */
358 	if (f2fs_check_nid_range(sbi, inode->i_ino))
359 		return -EINVAL;
360 
361 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
362 	if (IS_ERR(node_page))
363 		return PTR_ERR(node_page);
364 
365 	ri = F2FS_INODE(node_page);
366 
367 	inode->i_mode = le16_to_cpu(ri->i_mode);
368 	i_uid_write(inode, le32_to_cpu(ri->i_uid));
369 	i_gid_write(inode, le32_to_cpu(ri->i_gid));
370 	set_nlink(inode, le32_to_cpu(ri->i_links));
371 	inode->i_size = le64_to_cpu(ri->i_size);
372 	inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
373 
374 	inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
375 	inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
376 	inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
377 	inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
378 	inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
379 	inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
380 	inode->i_generation = le32_to_cpu(ri->i_generation);
381 	if (S_ISDIR(inode->i_mode))
382 		fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
383 	else if (S_ISREG(inode->i_mode))
384 		fi->i_gc_failures[GC_FAILURE_PIN] =
385 					le16_to_cpu(ri->i_gc_failures);
386 	fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
387 	fi->i_flags = le32_to_cpu(ri->i_flags);
388 	if (S_ISREG(inode->i_mode))
389 		fi->i_flags &= ~F2FS_PROJINHERIT_FL;
390 	bitmap_zero(fi->flags, FI_MAX);
391 	fi->i_advise = ri->i_advise;
392 	fi->i_pino = le32_to_cpu(ri->i_pino);
393 	fi->i_dir_level = ri->i_dir_level;
394 
395 	get_inline_info(inode, ri);
396 
397 	fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
398 					le16_to_cpu(ri->i_extra_isize) : 0;
399 
400 	if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
401 		fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
402 	} else if (f2fs_has_inline_xattr(inode) ||
403 				f2fs_has_inline_dentry(inode)) {
404 		fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
405 	} else {
406 
407 		/*
408 		 * Previous inline data or directory always reserved 200 bytes
409 		 * in inode layout, even if inline_xattr is disabled. In order
410 		 * to keep inline_dentry's structure for backward compatibility,
411 		 * we get the space back only from inline_data.
412 		 */
413 		fi->i_inline_xattr_size = 0;
414 	}
415 
416 	if (!sanity_check_inode(inode, node_page)) {
417 		f2fs_put_page(node_page, 1);
418 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
419 		return -EFSCORRUPTED;
420 	}
421 
422 	/* check data exist */
423 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
424 		__recover_inline_status(inode, node_page);
425 
426 	/* try to recover cold bit for non-dir inode */
427 	if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
428 		f2fs_wait_on_page_writeback(node_page, NODE, true, true);
429 		set_cold_node(node_page, false);
430 		set_page_dirty(node_page);
431 	}
432 
433 	/* get rdev by using inline_info */
434 	__get_inode_rdev(inode, ri);
435 
436 	if (S_ISREG(inode->i_mode)) {
437 		err = __written_first_block(sbi, ri);
438 		if (err < 0) {
439 			f2fs_put_page(node_page, 1);
440 			return err;
441 		}
442 		if (!err)
443 			set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
444 	}
445 
446 	if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
447 		fi->last_disk_size = inode->i_size;
448 
449 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
450 		set_inode_flag(inode, FI_PROJ_INHERIT);
451 
452 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
453 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
454 		i_projid = (projid_t)le32_to_cpu(ri->i_projid);
455 	else
456 		i_projid = F2FS_DEF_PROJID;
457 	fi->i_projid = make_kprojid(&init_user_ns, i_projid);
458 
459 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
460 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
461 		fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
462 		fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
463 	}
464 
465 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
466 					(fi->i_flags & F2FS_COMPR_FL)) {
467 		if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
468 					i_log_cluster_size)) {
469 			atomic_set(&fi->i_compr_blocks,
470 					le64_to_cpu(ri->i_compr_blocks));
471 			fi->i_compress_algorithm = ri->i_compress_algorithm;
472 			fi->i_log_cluster_size = ri->i_log_cluster_size;
473 			fi->i_compress_flag = le16_to_cpu(ri->i_compress_flag);
474 			fi->i_cluster_size = 1 << fi->i_log_cluster_size;
475 			set_inode_flag(inode, FI_COMPRESSED_FILE);
476 		}
477 	}
478 
479 	init_idisk_time(inode);
480 
481 	/* Need all the flag bits */
482 	f2fs_init_read_extent_tree(inode, node_page);
483 	f2fs_init_age_extent_tree(inode);
484 
485 	f2fs_put_page(node_page, 1);
486 
487 	stat_inc_inline_xattr(inode);
488 	stat_inc_inline_inode(inode);
489 	stat_inc_inline_dir(inode);
490 	stat_inc_compr_inode(inode);
491 	stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
492 
493 	return 0;
494 }
495 
496 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
497 {
498 	return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
499 		ino == F2FS_COMPRESS_INO(sbi);
500 }
501 
502 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
503 {
504 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
505 	struct inode *inode;
506 	int ret = 0;
507 
508 	inode = iget_locked(sb, ino);
509 	if (!inode)
510 		return ERR_PTR(-ENOMEM);
511 
512 	if (!(inode->i_state & I_NEW)) {
513 		if (is_meta_ino(sbi, ino)) {
514 			f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
515 			set_sbi_flag(sbi, SBI_NEED_FSCK);
516 			ret = -EFSCORRUPTED;
517 			trace_f2fs_iget_exit(inode, ret);
518 			iput(inode);
519 			f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
520 			return ERR_PTR(ret);
521 		}
522 
523 		trace_f2fs_iget(inode);
524 		return inode;
525 	}
526 
527 	if (is_meta_ino(sbi, ino))
528 		goto make_now;
529 
530 	ret = do_read_inode(inode);
531 	if (ret)
532 		goto bad_inode;
533 make_now:
534 	if (ino == F2FS_NODE_INO(sbi)) {
535 		inode->i_mapping->a_ops = &f2fs_node_aops;
536 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
537 	} else if (ino == F2FS_META_INO(sbi)) {
538 		inode->i_mapping->a_ops = &f2fs_meta_aops;
539 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
540 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
541 #ifdef CONFIG_F2FS_FS_COMPRESSION
542 		inode->i_mapping->a_ops = &f2fs_compress_aops;
543 		/*
544 		 * generic_error_remove_page only truncates pages of regular
545 		 * inode
546 		 */
547 		inode->i_mode |= S_IFREG;
548 #endif
549 		mapping_set_gfp_mask(inode->i_mapping,
550 			GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
551 	} else if (S_ISREG(inode->i_mode)) {
552 		inode->i_op = &f2fs_file_inode_operations;
553 		inode->i_fop = &f2fs_file_operations;
554 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
555 	} else if (S_ISDIR(inode->i_mode)) {
556 		inode->i_op = &f2fs_dir_inode_operations;
557 		inode->i_fop = &f2fs_dir_operations;
558 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
559 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
560 	} else if (S_ISLNK(inode->i_mode)) {
561 		if (file_is_encrypt(inode))
562 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
563 		else
564 			inode->i_op = &f2fs_symlink_inode_operations;
565 		inode_nohighmem(inode);
566 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
567 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
568 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
569 		inode->i_op = &f2fs_special_inode_operations;
570 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
571 	} else {
572 		ret = -EIO;
573 		goto bad_inode;
574 	}
575 	f2fs_set_inode_flags(inode);
576 
577 	if (file_should_truncate(inode) &&
578 			!is_sbi_flag_set(sbi, SBI_POR_DOING)) {
579 		ret = f2fs_truncate(inode);
580 		if (ret)
581 			goto bad_inode;
582 		file_dont_truncate(inode);
583 	}
584 
585 	unlock_new_inode(inode);
586 	trace_f2fs_iget(inode);
587 	return inode;
588 
589 bad_inode:
590 	f2fs_inode_synced(inode);
591 	iget_failed(inode);
592 	trace_f2fs_iget_exit(inode, ret);
593 	return ERR_PTR(ret);
594 }
595 
596 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
597 {
598 	struct inode *inode;
599 retry:
600 	inode = f2fs_iget(sb, ino);
601 	if (IS_ERR(inode)) {
602 		if (PTR_ERR(inode) == -ENOMEM) {
603 			memalloc_retry_wait(GFP_NOFS);
604 			goto retry;
605 		}
606 	}
607 	return inode;
608 }
609 
610 void f2fs_update_inode(struct inode *inode, struct page *node_page)
611 {
612 	struct f2fs_inode *ri;
613 	struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
614 
615 	f2fs_wait_on_page_writeback(node_page, NODE, true, true);
616 	set_page_dirty(node_page);
617 
618 	f2fs_inode_synced(inode);
619 
620 	ri = F2FS_INODE(node_page);
621 
622 	ri->i_mode = cpu_to_le16(inode->i_mode);
623 	ri->i_advise = F2FS_I(inode)->i_advise;
624 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
625 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
626 	ri->i_links = cpu_to_le32(inode->i_nlink);
627 	ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
628 
629 	if (!f2fs_is_atomic_file(inode) ||
630 			is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
631 		ri->i_size = cpu_to_le64(i_size_read(inode));
632 
633 	if (et) {
634 		read_lock(&et->lock);
635 		set_raw_read_extent(&et->largest, &ri->i_ext);
636 		read_unlock(&et->lock);
637 	} else {
638 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
639 	}
640 	set_raw_inline(inode, ri);
641 
642 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
643 	ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
644 	ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
645 	ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
646 	ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
647 	ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
648 	if (S_ISDIR(inode->i_mode))
649 		ri->i_current_depth =
650 			cpu_to_le32(F2FS_I(inode)->i_current_depth);
651 	else if (S_ISREG(inode->i_mode))
652 		ri->i_gc_failures =
653 			cpu_to_le16(F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]);
654 	ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
655 	ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
656 	ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
657 	ri->i_generation = cpu_to_le32(inode->i_generation);
658 	ri->i_dir_level = F2FS_I(inode)->i_dir_level;
659 
660 	if (f2fs_has_extra_attr(inode)) {
661 		ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
662 
663 		if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
664 			ri->i_inline_xattr_size =
665 				cpu_to_le16(F2FS_I(inode)->i_inline_xattr_size);
666 
667 		if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
668 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
669 								i_projid)) {
670 			projid_t i_projid;
671 
672 			i_projid = from_kprojid(&init_user_ns,
673 						F2FS_I(inode)->i_projid);
674 			ri->i_projid = cpu_to_le32(i_projid);
675 		}
676 
677 		if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
678 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
679 								i_crtime)) {
680 			ri->i_crtime =
681 				cpu_to_le64(F2FS_I(inode)->i_crtime.tv_sec);
682 			ri->i_crtime_nsec =
683 				cpu_to_le32(F2FS_I(inode)->i_crtime.tv_nsec);
684 		}
685 
686 		if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
687 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
688 							i_log_cluster_size)) {
689 			ri->i_compr_blocks =
690 				cpu_to_le64(atomic_read(
691 					&F2FS_I(inode)->i_compr_blocks));
692 			ri->i_compress_algorithm =
693 				F2FS_I(inode)->i_compress_algorithm;
694 			ri->i_compress_flag =
695 				cpu_to_le16(F2FS_I(inode)->i_compress_flag);
696 			ri->i_log_cluster_size =
697 				F2FS_I(inode)->i_log_cluster_size;
698 		}
699 	}
700 
701 	__set_inode_rdev(inode, ri);
702 
703 	/* deleted inode */
704 	if (inode->i_nlink == 0)
705 		clear_page_private_inline(node_page);
706 
707 	init_idisk_time(inode);
708 #ifdef CONFIG_F2FS_CHECK_FS
709 	f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
710 #endif
711 }
712 
713 void f2fs_update_inode_page(struct inode *inode)
714 {
715 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
716 	struct page *node_page;
717 retry:
718 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
719 	if (IS_ERR(node_page)) {
720 		int err = PTR_ERR(node_page);
721 
722 		if (err == -ENOMEM) {
723 			cond_resched();
724 			goto retry;
725 		} else if (err != -ENOENT) {
726 			f2fs_stop_checkpoint(sbi, false,
727 					STOP_CP_REASON_UPDATE_INODE);
728 		}
729 		return;
730 	}
731 	f2fs_update_inode(inode, node_page);
732 	f2fs_put_page(node_page, 1);
733 }
734 
735 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
736 {
737 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
738 
739 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
740 			inode->i_ino == F2FS_META_INO(sbi))
741 		return 0;
742 
743 	/*
744 	 * atime could be updated without dirtying f2fs inode in lazytime mode
745 	 */
746 	if (f2fs_is_time_consistent(inode) &&
747 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
748 		return 0;
749 
750 	if (!f2fs_is_checkpoint_ready(sbi))
751 		return -ENOSPC;
752 
753 	/*
754 	 * We need to balance fs here to prevent from producing dirty node pages
755 	 * during the urgent cleaning time when running out of free sections.
756 	 */
757 	f2fs_update_inode_page(inode);
758 	if (wbc && wbc->nr_to_write)
759 		f2fs_balance_fs(sbi, true);
760 	return 0;
761 }
762 
763 /*
764  * Called at the last iput() if i_nlink is zero
765  */
766 void f2fs_evict_inode(struct inode *inode)
767 {
768 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
769 	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
770 	int err = 0;
771 
772 	f2fs_abort_atomic_write(inode, true);
773 
774 	trace_f2fs_evict_inode(inode);
775 	truncate_inode_pages_final(&inode->i_data);
776 
777 	if ((inode->i_nlink || is_bad_inode(inode)) &&
778 		test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
779 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
780 
781 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
782 			inode->i_ino == F2FS_META_INO(sbi) ||
783 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
784 		goto out_clear;
785 
786 	f2fs_bug_on(sbi, get_dirty_pages(inode));
787 	f2fs_remove_dirty_inode(inode);
788 
789 	f2fs_destroy_extent_tree(inode);
790 
791 	if (inode->i_nlink || is_bad_inode(inode))
792 		goto no_delete;
793 
794 	err = f2fs_dquot_initialize(inode);
795 	if (err) {
796 		err = 0;
797 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
798 	}
799 
800 	f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
801 	f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
802 	f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
803 
804 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
805 		sb_start_intwrite(inode->i_sb);
806 	set_inode_flag(inode, FI_NO_ALLOC);
807 	i_size_write(inode, 0);
808 retry:
809 	if (F2FS_HAS_BLOCKS(inode))
810 		err = f2fs_truncate(inode);
811 
812 	if (time_to_inject(sbi, FAULT_EVICT_INODE)) {
813 		f2fs_show_injection_info(sbi, FAULT_EVICT_INODE);
814 		err = -EIO;
815 	}
816 
817 	if (!err) {
818 		f2fs_lock_op(sbi);
819 		err = f2fs_remove_inode_page(inode);
820 		f2fs_unlock_op(sbi);
821 		if (err == -ENOENT) {
822 			err = 0;
823 
824 			/*
825 			 * in fuzzed image, another node may has the same
826 			 * block address as inode's, if it was truncated
827 			 * previously, truncation of inode node will fail.
828 			 */
829 			if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
830 				f2fs_warn(F2FS_I_SB(inode),
831 					"f2fs_evict_inode: inconsistent node id, ino:%lu",
832 					inode->i_ino);
833 				f2fs_inode_synced(inode);
834 				set_sbi_flag(sbi, SBI_NEED_FSCK);
835 			}
836 		}
837 	}
838 
839 	/* give more chances, if ENOMEM case */
840 	if (err == -ENOMEM) {
841 		err = 0;
842 		goto retry;
843 	}
844 
845 	if (err) {
846 		f2fs_update_inode_page(inode);
847 		if (dquot_initialize_needed(inode))
848 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
849 	}
850 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
851 		sb_end_intwrite(inode->i_sb);
852 no_delete:
853 	dquot_drop(inode);
854 
855 	stat_dec_inline_xattr(inode);
856 	stat_dec_inline_dir(inode);
857 	stat_dec_inline_inode(inode);
858 	stat_dec_compr_inode(inode);
859 	stat_sub_compr_blocks(inode,
860 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
861 
862 	if (likely(!f2fs_cp_error(sbi) &&
863 				!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
864 		f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
865 	else
866 		f2fs_inode_synced(inode);
867 
868 	/* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
869 	if (inode->i_ino)
870 		invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
871 							inode->i_ino);
872 	if (xnid)
873 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
874 	if (inode->i_nlink) {
875 		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
876 			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
877 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
878 			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
879 	}
880 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
881 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
882 		clear_inode_flag(inode, FI_FREE_NID);
883 	} else {
884 		/*
885 		 * If xattr nid is corrupted, we can reach out error condition,
886 		 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
887 		 * In that case, f2fs_check_nid_range() is enough to give a clue.
888 		 */
889 	}
890 out_clear:
891 	fscrypt_put_encryption_info(inode);
892 	fsverity_cleanup_inode(inode);
893 	clear_inode(inode);
894 }
895 
896 /* caller should call f2fs_lock_op() */
897 void f2fs_handle_failed_inode(struct inode *inode)
898 {
899 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
900 	struct node_info ni;
901 	int err;
902 
903 	/*
904 	 * clear nlink of inode in order to release resource of inode
905 	 * immediately.
906 	 */
907 	clear_nlink(inode);
908 
909 	/*
910 	 * we must call this to avoid inode being remained as dirty, resulting
911 	 * in a panic when flushing dirty inodes in gdirty_list.
912 	 */
913 	f2fs_update_inode_page(inode);
914 	f2fs_inode_synced(inode);
915 
916 	/* don't make bad inode, since it becomes a regular file. */
917 	unlock_new_inode(inode);
918 
919 	/*
920 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
921 	 * so we can prevent losing this orphan when encoutering checkpoint
922 	 * and following suddenly power-off.
923 	 */
924 	err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
925 	if (err) {
926 		set_sbi_flag(sbi, SBI_NEED_FSCK);
927 		set_inode_flag(inode, FI_FREE_NID);
928 		f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
929 		goto out;
930 	}
931 
932 	if (ni.blk_addr != NULL_ADDR) {
933 		err = f2fs_acquire_orphan_inode(sbi);
934 		if (err) {
935 			set_sbi_flag(sbi, SBI_NEED_FSCK);
936 			f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
937 		} else {
938 			f2fs_add_orphan_inode(inode);
939 		}
940 		f2fs_alloc_nid_done(sbi, inode->i_ino);
941 	} else {
942 		set_inode_flag(inode, FI_FREE_NID);
943 	}
944 
945 out:
946 	f2fs_unlock_op(sbi);
947 
948 	/* iput will drop the inode object */
949 	iput(inode);
950 }
951