xref: /openbmc/linux/fs/f2fs/inode.c (revision 11a163f2)
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/backing-dev.h>
12 #include <linux/writeback.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 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
22 {
23 	if (is_inode_flag_set(inode, FI_NEW_INODE))
24 		return;
25 
26 	if (f2fs_inode_dirtied(inode, sync))
27 		return;
28 
29 	mark_inode_dirty_sync(inode);
30 }
31 
32 void f2fs_set_inode_flags(struct inode *inode)
33 {
34 	unsigned int flags = F2FS_I(inode)->i_flags;
35 	unsigned int new_fl = 0;
36 
37 	if (flags & F2FS_SYNC_FL)
38 		new_fl |= S_SYNC;
39 	if (flags & F2FS_APPEND_FL)
40 		new_fl |= S_APPEND;
41 	if (flags & F2FS_IMMUTABLE_FL)
42 		new_fl |= S_IMMUTABLE;
43 	if (flags & F2FS_NOATIME_FL)
44 		new_fl |= S_NOATIME;
45 	if (flags & F2FS_DIRSYNC_FL)
46 		new_fl |= S_DIRSYNC;
47 	if (file_is_encrypt(inode))
48 		new_fl |= S_ENCRYPTED;
49 	if (file_is_verity(inode))
50 		new_fl |= S_VERITY;
51 	if (flags & F2FS_CASEFOLD_FL)
52 		new_fl |= S_CASEFOLD;
53 	inode_set_flags(inode, new_fl,
54 			S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|
55 			S_ENCRYPTED|S_VERITY|S_CASEFOLD);
56 }
57 
58 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
59 {
60 	int extra_size = get_extra_isize(inode);
61 
62 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
63 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
64 		if (ri->i_addr[extra_size])
65 			inode->i_rdev = old_decode_dev(
66 				le32_to_cpu(ri->i_addr[extra_size]));
67 		else
68 			inode->i_rdev = new_decode_dev(
69 				le32_to_cpu(ri->i_addr[extra_size + 1]));
70 	}
71 }
72 
73 static int __written_first_block(struct f2fs_sb_info *sbi,
74 					struct f2fs_inode *ri)
75 {
76 	block_t addr = le32_to_cpu(ri->i_addr[offset_in_addr(ri)]);
77 
78 	if (!__is_valid_data_blkaddr(addr))
79 		return 1;
80 	if (!f2fs_is_valid_blkaddr(sbi, addr, DATA_GENERIC_ENHANCE))
81 		return -EFSCORRUPTED;
82 	return 0;
83 }
84 
85 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
86 {
87 	int extra_size = get_extra_isize(inode);
88 
89 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
90 		if (old_valid_dev(inode->i_rdev)) {
91 			ri->i_addr[extra_size] =
92 				cpu_to_le32(old_encode_dev(inode->i_rdev));
93 			ri->i_addr[extra_size + 1] = 0;
94 		} else {
95 			ri->i_addr[extra_size] = 0;
96 			ri->i_addr[extra_size + 1] =
97 				cpu_to_le32(new_encode_dev(inode->i_rdev));
98 			ri->i_addr[extra_size + 2] = 0;
99 		}
100 	}
101 }
102 
103 static void __recover_inline_status(struct inode *inode, struct page *ipage)
104 {
105 	void *inline_data = inline_data_addr(inode, ipage);
106 	__le32 *start = inline_data;
107 	__le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
108 
109 	while (start < end) {
110 		if (*start++) {
111 			f2fs_wait_on_page_writeback(ipage, NODE, true, true);
112 
113 			set_inode_flag(inode, FI_DATA_EXIST);
114 			set_raw_inline(inode, F2FS_INODE(ipage));
115 			set_page_dirty(ipage);
116 			return;
117 		}
118 	}
119 	return;
120 }
121 
122 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
123 {
124 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
125 
126 	if (!f2fs_sb_has_inode_chksum(sbi))
127 		return false;
128 
129 	if (!IS_INODE(page) || !(ri->i_inline & F2FS_EXTRA_ATTR))
130 		return false;
131 
132 	if (!F2FS_FITS_IN_INODE(ri, le16_to_cpu(ri->i_extra_isize),
133 				i_inode_checksum))
134 		return false;
135 
136 	return true;
137 }
138 
139 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
140 {
141 	struct f2fs_node *node = F2FS_NODE(page);
142 	struct f2fs_inode *ri = &node->i;
143 	__le32 ino = node->footer.ino;
144 	__le32 gen = ri->i_generation;
145 	__u32 chksum, chksum_seed;
146 	__u32 dummy_cs = 0;
147 	unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
148 	unsigned int cs_size = sizeof(dummy_cs);
149 
150 	chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
151 							sizeof(ino));
152 	chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
153 
154 	chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
155 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
156 	offset += cs_size;
157 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
158 						F2FS_BLKSIZE - offset);
159 	return chksum;
160 }
161 
162 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
163 {
164 	struct f2fs_inode *ri;
165 	__u32 provided, calculated;
166 
167 	if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
168 		return true;
169 
170 #ifdef CONFIG_F2FS_CHECK_FS
171 	if (!f2fs_enable_inode_chksum(sbi, page))
172 #else
173 	if (!f2fs_enable_inode_chksum(sbi, page) ||
174 			PageDirty(page) || PageWriteback(page))
175 #endif
176 		return true;
177 
178 	ri = &F2FS_NODE(page)->i;
179 	provided = le32_to_cpu(ri->i_inode_checksum);
180 	calculated = f2fs_inode_chksum(sbi, page);
181 
182 	if (provided != calculated)
183 		f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
184 			  page->index, ino_of_node(page), provided, calculated);
185 
186 	return provided == calculated;
187 }
188 
189 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
190 {
191 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
192 
193 	if (!f2fs_enable_inode_chksum(sbi, page))
194 		return;
195 
196 	ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
197 }
198 
199 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
200 {
201 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
202 	struct f2fs_inode_info *fi = F2FS_I(inode);
203 	struct f2fs_inode *ri = F2FS_INODE(node_page);
204 	unsigned long long iblocks;
205 
206 	iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
207 	if (!iblocks) {
208 		set_sbi_flag(sbi, SBI_NEED_FSCK);
209 		f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
210 			  __func__, inode->i_ino, iblocks);
211 		return false;
212 	}
213 
214 	if (ino_of_node(node_page) != nid_of_node(node_page)) {
215 		set_sbi_flag(sbi, SBI_NEED_FSCK);
216 		f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
217 			  __func__, inode->i_ino,
218 			  ino_of_node(node_page), nid_of_node(node_page));
219 		return false;
220 	}
221 
222 	if (f2fs_sb_has_flexible_inline_xattr(sbi)
223 			&& !f2fs_has_extra_attr(inode)) {
224 		set_sbi_flag(sbi, SBI_NEED_FSCK);
225 		f2fs_warn(sbi, "%s: corrupted inode ino=%lx, run fsck to fix.",
226 			  __func__, inode->i_ino);
227 		return false;
228 	}
229 
230 	if (f2fs_has_extra_attr(inode) &&
231 			!f2fs_sb_has_extra_attr(sbi)) {
232 		set_sbi_flag(sbi, SBI_NEED_FSCK);
233 		f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
234 			  __func__, inode->i_ino);
235 		return false;
236 	}
237 
238 	if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
239 			fi->i_extra_isize % sizeof(__le32)) {
240 		set_sbi_flag(sbi, SBI_NEED_FSCK);
241 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
242 			  __func__, inode->i_ino, fi->i_extra_isize,
243 			  F2FS_TOTAL_EXTRA_ATTR_SIZE);
244 		return false;
245 	}
246 
247 	if (f2fs_has_extra_attr(inode) &&
248 		f2fs_sb_has_flexible_inline_xattr(sbi) &&
249 		f2fs_has_inline_xattr(inode) &&
250 		(!fi->i_inline_xattr_size ||
251 		fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
252 		set_sbi_flag(sbi, SBI_NEED_FSCK);
253 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %zu",
254 			  __func__, inode->i_ino, fi->i_inline_xattr_size,
255 			  MAX_INLINE_XATTR_SIZE);
256 		return false;
257 	}
258 
259 	if (F2FS_I(inode)->extent_tree) {
260 		struct extent_info *ei = &F2FS_I(inode)->extent_tree->largest;
261 
262 		if (ei->len &&
263 			(!f2fs_is_valid_blkaddr(sbi, ei->blk,
264 						DATA_GENERIC_ENHANCE) ||
265 			!f2fs_is_valid_blkaddr(sbi, ei->blk + ei->len - 1,
266 						DATA_GENERIC_ENHANCE))) {
267 			set_sbi_flag(sbi, SBI_NEED_FSCK);
268 			f2fs_warn(sbi, "%s: inode (ino=%lx) extent info [%u, %u, %u] is incorrect, run fsck to fix",
269 				  __func__, inode->i_ino,
270 				  ei->blk, ei->fofs, ei->len);
271 			return false;
272 		}
273 	}
274 
275 	if (f2fs_has_inline_data(inode) &&
276 			(!S_ISREG(inode->i_mode) && !S_ISLNK(inode->i_mode))) {
277 		set_sbi_flag(sbi, SBI_NEED_FSCK);
278 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
279 			  __func__, inode->i_ino, inode->i_mode);
280 		return false;
281 	}
282 
283 	if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
284 		set_sbi_flag(sbi, SBI_NEED_FSCK);
285 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
286 			  __func__, inode->i_ino, inode->i_mode);
287 		return false;
288 	}
289 
290 	if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
291 		set_sbi_flag(sbi, SBI_NEED_FSCK);
292 		f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
293 			  __func__, inode->i_ino);
294 		return false;
295 	}
296 
297 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
298 			fi->i_flags & F2FS_COMPR_FL &&
299 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
300 						i_log_cluster_size)) {
301 		if (ri->i_compress_algorithm >= COMPRESS_MAX) {
302 			set_sbi_flag(sbi, SBI_NEED_FSCK);
303 			f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported "
304 				"compress algorithm: %u, run fsck to fix",
305 				  __func__, inode->i_ino,
306 				  ri->i_compress_algorithm);
307 			return false;
308 		}
309 		if (le64_to_cpu(ri->i_compr_blocks) >
310 				SECTOR_TO_BLOCK(inode->i_blocks)) {
311 			set_sbi_flag(sbi, SBI_NEED_FSCK);
312 			f2fs_warn(sbi, "%s: inode (ino=%lx) has inconsistent "
313 				"i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
314 				  __func__, inode->i_ino,
315 				  le64_to_cpu(ri->i_compr_blocks),
316 				  SECTOR_TO_BLOCK(inode->i_blocks));
317 			return false;
318 		}
319 		if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
320 			ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
321 			set_sbi_flag(sbi, SBI_NEED_FSCK);
322 			f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported "
323 				"log cluster size: %u, run fsck to fix",
324 				  __func__, inode->i_ino,
325 				  ri->i_log_cluster_size);
326 			return false;
327 		}
328 	}
329 
330 	return true;
331 }
332 
333 static int do_read_inode(struct inode *inode)
334 {
335 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
336 	struct f2fs_inode_info *fi = F2FS_I(inode);
337 	struct page *node_page;
338 	struct f2fs_inode *ri;
339 	projid_t i_projid;
340 	int err;
341 
342 	/* Check if ino is within scope */
343 	if (f2fs_check_nid_range(sbi, inode->i_ino))
344 		return -EINVAL;
345 
346 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
347 	if (IS_ERR(node_page))
348 		return PTR_ERR(node_page);
349 
350 	ri = F2FS_INODE(node_page);
351 
352 	inode->i_mode = le16_to_cpu(ri->i_mode);
353 	i_uid_write(inode, le32_to_cpu(ri->i_uid));
354 	i_gid_write(inode, le32_to_cpu(ri->i_gid));
355 	set_nlink(inode, le32_to_cpu(ri->i_links));
356 	inode->i_size = le64_to_cpu(ri->i_size);
357 	inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
358 
359 	inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
360 	inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
361 	inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
362 	inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
363 	inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
364 	inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
365 	inode->i_generation = le32_to_cpu(ri->i_generation);
366 	if (S_ISDIR(inode->i_mode))
367 		fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
368 	else if (S_ISREG(inode->i_mode))
369 		fi->i_gc_failures[GC_FAILURE_PIN] =
370 					le16_to_cpu(ri->i_gc_failures);
371 	fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
372 	fi->i_flags = le32_to_cpu(ri->i_flags);
373 	if (S_ISREG(inode->i_mode))
374 		fi->i_flags &= ~F2FS_PROJINHERIT_FL;
375 	bitmap_zero(fi->flags, FI_MAX);
376 	fi->i_advise = ri->i_advise;
377 	fi->i_pino = le32_to_cpu(ri->i_pino);
378 	fi->i_dir_level = ri->i_dir_level;
379 
380 	f2fs_init_extent_tree(inode, node_page);
381 
382 	get_inline_info(inode, ri);
383 
384 	fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
385 					le16_to_cpu(ri->i_extra_isize) : 0;
386 
387 	if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
388 		fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
389 	} else if (f2fs_has_inline_xattr(inode) ||
390 				f2fs_has_inline_dentry(inode)) {
391 		fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
392 	} else {
393 
394 		/*
395 		 * Previous inline data or directory always reserved 200 bytes
396 		 * in inode layout, even if inline_xattr is disabled. In order
397 		 * to keep inline_dentry's structure for backward compatibility,
398 		 * we get the space back only from inline_data.
399 		 */
400 		fi->i_inline_xattr_size = 0;
401 	}
402 
403 	if (!sanity_check_inode(inode, node_page)) {
404 		f2fs_put_page(node_page, 1);
405 		return -EFSCORRUPTED;
406 	}
407 
408 	/* check data exist */
409 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
410 		__recover_inline_status(inode, node_page);
411 
412 	/* try to recover cold bit for non-dir inode */
413 	if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
414 		f2fs_wait_on_page_writeback(node_page, NODE, true, true);
415 		set_cold_node(node_page, false);
416 		set_page_dirty(node_page);
417 	}
418 
419 	/* get rdev by using inline_info */
420 	__get_inode_rdev(inode, ri);
421 
422 	if (S_ISREG(inode->i_mode)) {
423 		err = __written_first_block(sbi, ri);
424 		if (err < 0) {
425 			f2fs_put_page(node_page, 1);
426 			return err;
427 		}
428 		if (!err)
429 			set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
430 	}
431 
432 	if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
433 		fi->last_disk_size = inode->i_size;
434 
435 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
436 		set_inode_flag(inode, FI_PROJ_INHERIT);
437 
438 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
439 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
440 		i_projid = (projid_t)le32_to_cpu(ri->i_projid);
441 	else
442 		i_projid = F2FS_DEF_PROJID;
443 	fi->i_projid = make_kprojid(&init_user_ns, i_projid);
444 
445 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
446 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
447 		fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
448 		fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
449 	}
450 
451 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
452 					(fi->i_flags & F2FS_COMPR_FL)) {
453 		if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
454 					i_log_cluster_size)) {
455 			atomic_set(&fi->i_compr_blocks,
456 					le64_to_cpu(ri->i_compr_blocks));
457 			fi->i_compress_algorithm = ri->i_compress_algorithm;
458 			fi->i_log_cluster_size = ri->i_log_cluster_size;
459 			fi->i_cluster_size = 1 << fi->i_log_cluster_size;
460 			set_inode_flag(inode, FI_COMPRESSED_FILE);
461 		}
462 	}
463 
464 	F2FS_I(inode)->i_disk_time[0] = inode->i_atime;
465 	F2FS_I(inode)->i_disk_time[1] = inode->i_ctime;
466 	F2FS_I(inode)->i_disk_time[2] = inode->i_mtime;
467 	F2FS_I(inode)->i_disk_time[3] = F2FS_I(inode)->i_crtime;
468 	f2fs_put_page(node_page, 1);
469 
470 	stat_inc_inline_xattr(inode);
471 	stat_inc_inline_inode(inode);
472 	stat_inc_inline_dir(inode);
473 	stat_inc_compr_inode(inode);
474 	stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
475 
476 	return 0;
477 }
478 
479 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
480 {
481 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
482 	struct inode *inode;
483 	int ret = 0;
484 
485 	inode = iget_locked(sb, ino);
486 	if (!inode)
487 		return ERR_PTR(-ENOMEM);
488 
489 	if (!(inode->i_state & I_NEW)) {
490 		trace_f2fs_iget(inode);
491 		return inode;
492 	}
493 	if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
494 		goto make_now;
495 
496 	ret = do_read_inode(inode);
497 	if (ret)
498 		goto bad_inode;
499 make_now:
500 	if (ino == F2FS_NODE_INO(sbi)) {
501 		inode->i_mapping->a_ops = &f2fs_node_aops;
502 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
503 	} else if (ino == F2FS_META_INO(sbi)) {
504 		inode->i_mapping->a_ops = &f2fs_meta_aops;
505 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
506 	} else if (S_ISREG(inode->i_mode)) {
507 		inode->i_op = &f2fs_file_inode_operations;
508 		inode->i_fop = &f2fs_file_operations;
509 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
510 	} else if (S_ISDIR(inode->i_mode)) {
511 		inode->i_op = &f2fs_dir_inode_operations;
512 		inode->i_fop = &f2fs_dir_operations;
513 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
514 		inode_nohighmem(inode);
515 	} else if (S_ISLNK(inode->i_mode)) {
516 		if (file_is_encrypt(inode))
517 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
518 		else
519 			inode->i_op = &f2fs_symlink_inode_operations;
520 		inode_nohighmem(inode);
521 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
522 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
523 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
524 		inode->i_op = &f2fs_special_inode_operations;
525 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
526 	} else {
527 		ret = -EIO;
528 		goto bad_inode;
529 	}
530 	f2fs_set_inode_flags(inode);
531 	unlock_new_inode(inode);
532 	trace_f2fs_iget(inode);
533 	return inode;
534 
535 bad_inode:
536 	f2fs_inode_synced(inode);
537 	iget_failed(inode);
538 	trace_f2fs_iget_exit(inode, ret);
539 	return ERR_PTR(ret);
540 }
541 
542 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
543 {
544 	struct inode *inode;
545 retry:
546 	inode = f2fs_iget(sb, ino);
547 	if (IS_ERR(inode)) {
548 		if (PTR_ERR(inode) == -ENOMEM) {
549 			congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
550 			goto retry;
551 		}
552 	}
553 	return inode;
554 }
555 
556 void f2fs_update_inode(struct inode *inode, struct page *node_page)
557 {
558 	struct f2fs_inode *ri;
559 	struct extent_tree *et = F2FS_I(inode)->extent_tree;
560 
561 	f2fs_wait_on_page_writeback(node_page, NODE, true, true);
562 	set_page_dirty(node_page);
563 
564 	f2fs_inode_synced(inode);
565 
566 	ri = F2FS_INODE(node_page);
567 
568 	ri->i_mode = cpu_to_le16(inode->i_mode);
569 	ri->i_advise = F2FS_I(inode)->i_advise;
570 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
571 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
572 	ri->i_links = cpu_to_le32(inode->i_nlink);
573 	ri->i_size = cpu_to_le64(i_size_read(inode));
574 	ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
575 
576 	if (et) {
577 		read_lock(&et->lock);
578 		set_raw_extent(&et->largest, &ri->i_ext);
579 		read_unlock(&et->lock);
580 	} else {
581 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
582 	}
583 	set_raw_inline(inode, ri);
584 
585 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
586 	ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
587 	ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
588 	ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
589 	ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
590 	ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
591 	if (S_ISDIR(inode->i_mode))
592 		ri->i_current_depth =
593 			cpu_to_le32(F2FS_I(inode)->i_current_depth);
594 	else if (S_ISREG(inode->i_mode))
595 		ri->i_gc_failures =
596 			cpu_to_le16(F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]);
597 	ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
598 	ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
599 	ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
600 	ri->i_generation = cpu_to_le32(inode->i_generation);
601 	ri->i_dir_level = F2FS_I(inode)->i_dir_level;
602 
603 	if (f2fs_has_extra_attr(inode)) {
604 		ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
605 
606 		if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
607 			ri->i_inline_xattr_size =
608 				cpu_to_le16(F2FS_I(inode)->i_inline_xattr_size);
609 
610 		if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
611 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
612 								i_projid)) {
613 			projid_t i_projid;
614 
615 			i_projid = from_kprojid(&init_user_ns,
616 						F2FS_I(inode)->i_projid);
617 			ri->i_projid = cpu_to_le32(i_projid);
618 		}
619 
620 		if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
621 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
622 								i_crtime)) {
623 			ri->i_crtime =
624 				cpu_to_le64(F2FS_I(inode)->i_crtime.tv_sec);
625 			ri->i_crtime_nsec =
626 				cpu_to_le32(F2FS_I(inode)->i_crtime.tv_nsec);
627 		}
628 
629 		if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
630 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
631 							i_log_cluster_size)) {
632 			ri->i_compr_blocks =
633 				cpu_to_le64(atomic_read(
634 					&F2FS_I(inode)->i_compr_blocks));
635 			ri->i_compress_algorithm =
636 				F2FS_I(inode)->i_compress_algorithm;
637 			ri->i_log_cluster_size =
638 				F2FS_I(inode)->i_log_cluster_size;
639 		}
640 	}
641 
642 	__set_inode_rdev(inode, ri);
643 
644 	/* deleted inode */
645 	if (inode->i_nlink == 0)
646 		clear_inline_node(node_page);
647 
648 	F2FS_I(inode)->i_disk_time[0] = inode->i_atime;
649 	F2FS_I(inode)->i_disk_time[1] = inode->i_ctime;
650 	F2FS_I(inode)->i_disk_time[2] = inode->i_mtime;
651 	F2FS_I(inode)->i_disk_time[3] = F2FS_I(inode)->i_crtime;
652 
653 #ifdef CONFIG_F2FS_CHECK_FS
654 	f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
655 #endif
656 }
657 
658 void f2fs_update_inode_page(struct inode *inode)
659 {
660 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
661 	struct page *node_page;
662 retry:
663 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
664 	if (IS_ERR(node_page)) {
665 		int err = PTR_ERR(node_page);
666 		if (err == -ENOMEM) {
667 			cond_resched();
668 			goto retry;
669 		} else if (err != -ENOENT) {
670 			f2fs_stop_checkpoint(sbi, false);
671 		}
672 		return;
673 	}
674 	f2fs_update_inode(inode, node_page);
675 	f2fs_put_page(node_page, 1);
676 }
677 
678 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
679 {
680 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
681 
682 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
683 			inode->i_ino == F2FS_META_INO(sbi))
684 		return 0;
685 
686 	/*
687 	 * atime could be updated without dirtying f2fs inode in lazytime mode
688 	 */
689 	if (f2fs_is_time_consistent(inode) &&
690 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
691 		return 0;
692 
693 	if (!f2fs_is_checkpoint_ready(sbi))
694 		return -ENOSPC;
695 
696 	/*
697 	 * We need to balance fs here to prevent from producing dirty node pages
698 	 * during the urgent cleaning time when runing out of free sections.
699 	 */
700 	f2fs_update_inode_page(inode);
701 	if (wbc && wbc->nr_to_write)
702 		f2fs_balance_fs(sbi, true);
703 	return 0;
704 }
705 
706 /*
707  * Called at the last iput() if i_nlink is zero
708  */
709 void f2fs_evict_inode(struct inode *inode)
710 {
711 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
712 	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
713 	int err = 0;
714 
715 	/* some remained atomic pages should discarded */
716 	if (f2fs_is_atomic_file(inode))
717 		f2fs_drop_inmem_pages(inode);
718 
719 	trace_f2fs_evict_inode(inode);
720 	truncate_inode_pages_final(&inode->i_data);
721 
722 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
723 			inode->i_ino == F2FS_META_INO(sbi))
724 		goto out_clear;
725 
726 	f2fs_bug_on(sbi, get_dirty_pages(inode));
727 	f2fs_remove_dirty_inode(inode);
728 
729 	f2fs_destroy_extent_tree(inode);
730 
731 	if (inode->i_nlink || is_bad_inode(inode))
732 		goto no_delete;
733 
734 	err = dquot_initialize(inode);
735 	if (err) {
736 		err = 0;
737 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
738 	}
739 
740 	f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
741 	f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
742 	f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
743 
744 	sb_start_intwrite(inode->i_sb);
745 	set_inode_flag(inode, FI_NO_ALLOC);
746 	i_size_write(inode, 0);
747 retry:
748 	if (F2FS_HAS_BLOCKS(inode))
749 		err = f2fs_truncate(inode);
750 
751 	if (time_to_inject(sbi, FAULT_EVICT_INODE)) {
752 		f2fs_show_injection_info(sbi, FAULT_EVICT_INODE);
753 		err = -EIO;
754 	}
755 
756 	if (!err) {
757 		f2fs_lock_op(sbi);
758 		err = f2fs_remove_inode_page(inode);
759 		f2fs_unlock_op(sbi);
760 		if (err == -ENOENT)
761 			err = 0;
762 	}
763 
764 	/* give more chances, if ENOMEM case */
765 	if (err == -ENOMEM) {
766 		err = 0;
767 		goto retry;
768 	}
769 
770 	if (err) {
771 		f2fs_update_inode_page(inode);
772 		if (dquot_initialize_needed(inode))
773 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
774 	}
775 	sb_end_intwrite(inode->i_sb);
776 no_delete:
777 	dquot_drop(inode);
778 
779 	stat_dec_inline_xattr(inode);
780 	stat_dec_inline_dir(inode);
781 	stat_dec_inline_inode(inode);
782 	stat_dec_compr_inode(inode);
783 	stat_sub_compr_blocks(inode,
784 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
785 
786 	if (likely(!f2fs_cp_error(sbi) &&
787 				!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
788 		f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
789 	else
790 		f2fs_inode_synced(inode);
791 
792 	/* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
793 	if (inode->i_ino)
794 		invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
795 							inode->i_ino);
796 	if (xnid)
797 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
798 	if (inode->i_nlink) {
799 		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
800 			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
801 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
802 			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
803 	}
804 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
805 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
806 		clear_inode_flag(inode, FI_FREE_NID);
807 	} else {
808 		/*
809 		 * If xattr nid is corrupted, we can reach out error condition,
810 		 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
811 		 * In that case, f2fs_check_nid_range() is enough to give a clue.
812 		 */
813 	}
814 out_clear:
815 	fscrypt_put_encryption_info(inode);
816 	fsverity_cleanup_inode(inode);
817 	clear_inode(inode);
818 }
819 
820 /* caller should call f2fs_lock_op() */
821 void f2fs_handle_failed_inode(struct inode *inode)
822 {
823 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
824 	struct node_info ni;
825 	int err;
826 
827 	/*
828 	 * clear nlink of inode in order to release resource of inode
829 	 * immediately.
830 	 */
831 	clear_nlink(inode);
832 
833 	/*
834 	 * we must call this to avoid inode being remained as dirty, resulting
835 	 * in a panic when flushing dirty inodes in gdirty_list.
836 	 */
837 	f2fs_update_inode_page(inode);
838 	f2fs_inode_synced(inode);
839 
840 	/* don't make bad inode, since it becomes a regular file. */
841 	unlock_new_inode(inode);
842 
843 	/*
844 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
845 	 * so we can prevent losing this orphan when encoutering checkpoint
846 	 * and following suddenly power-off.
847 	 */
848 	err = f2fs_get_node_info(sbi, inode->i_ino, &ni);
849 	if (err) {
850 		set_sbi_flag(sbi, SBI_NEED_FSCK);
851 		f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
852 		goto out;
853 	}
854 
855 	if (ni.blk_addr != NULL_ADDR) {
856 		err = f2fs_acquire_orphan_inode(sbi);
857 		if (err) {
858 			set_sbi_flag(sbi, SBI_NEED_FSCK);
859 			f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
860 		} else {
861 			f2fs_add_orphan_inode(inode);
862 		}
863 		f2fs_alloc_nid_done(sbi, inode->i_ino);
864 	} else {
865 		set_inode_flag(inode, FI_FREE_NID);
866 	}
867 
868 out:
869 	f2fs_unlock_op(sbi);
870 
871 	/* iput will drop the inode object */
872 	iput(inode);
873 }
874