xref: /linux/fs/f2fs/inode.c (revision bb5b94f5bbe75470912b70fb08880fc5273aa62d)
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 #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 		return -EFSCORRUPTED;
86 	return 0;
87 }
88 
89 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
90 {
91 	int extra_size = get_extra_isize(inode);
92 
93 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
94 		if (old_valid_dev(inode->i_rdev)) {
95 			ri->i_addr[extra_size] =
96 				cpu_to_le32(old_encode_dev(inode->i_rdev));
97 			ri->i_addr[extra_size + 1] = 0;
98 		} else {
99 			ri->i_addr[extra_size] = 0;
100 			ri->i_addr[extra_size + 1] =
101 				cpu_to_le32(new_encode_dev(inode->i_rdev));
102 			ri->i_addr[extra_size + 2] = 0;
103 		}
104 	}
105 }
106 
107 static void __recover_inline_status(struct inode *inode, struct page *ipage)
108 {
109 	void *inline_data = inline_data_addr(inode, ipage);
110 	__le32 *start = inline_data;
111 	__le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
112 
113 	while (start < end) {
114 		if (*start++) {
115 			f2fs_wait_on_page_writeback(ipage, NODE, true, true);
116 
117 			set_inode_flag(inode, FI_DATA_EXIST);
118 			set_raw_inline(inode, F2FS_INODE(ipage));
119 			set_page_dirty(ipage);
120 			return;
121 		}
122 	}
123 	return;
124 }
125 
126 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
127 {
128 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
129 
130 	if (!f2fs_sb_has_inode_chksum(sbi))
131 		return false;
132 
133 	if (!IS_INODE(page) || !(ri->i_inline & F2FS_EXTRA_ATTR))
134 		return false;
135 
136 	if (!F2FS_FITS_IN_INODE(ri, le16_to_cpu(ri->i_extra_isize),
137 				i_inode_checksum))
138 		return false;
139 
140 	return true;
141 }
142 
143 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
144 {
145 	struct f2fs_node *node = F2FS_NODE(page);
146 	struct f2fs_inode *ri = &node->i;
147 	__le32 ino = node->footer.ino;
148 	__le32 gen = ri->i_generation;
149 	__u32 chksum, chksum_seed;
150 	__u32 dummy_cs = 0;
151 	unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
152 	unsigned int cs_size = sizeof(dummy_cs);
153 
154 	chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
155 							sizeof(ino));
156 	chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
157 
158 	chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
159 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
160 	offset += cs_size;
161 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
162 						F2FS_BLKSIZE - offset);
163 	return chksum;
164 }
165 
166 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
167 {
168 	struct f2fs_inode *ri;
169 	__u32 provided, calculated;
170 
171 	if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
172 		return true;
173 
174 #ifdef CONFIG_F2FS_CHECK_FS
175 	if (!f2fs_enable_inode_chksum(sbi, page))
176 #else
177 	if (!f2fs_enable_inode_chksum(sbi, page) ||
178 			PageDirty(page) || PageWriteback(page))
179 #endif
180 		return true;
181 
182 	ri = &F2FS_NODE(page)->i;
183 	provided = le32_to_cpu(ri->i_inode_checksum);
184 	calculated = f2fs_inode_chksum(sbi, page);
185 
186 	if (provided != calculated)
187 		f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
188 			  page->index, ino_of_node(page), provided, calculated);
189 
190 	return provided == calculated;
191 }
192 
193 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
194 {
195 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
196 
197 	if (!f2fs_enable_inode_chksum(sbi, page))
198 		return;
199 
200 	ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
201 }
202 
203 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
204 {
205 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
206 	struct f2fs_inode_info *fi = F2FS_I(inode);
207 	struct f2fs_inode *ri = F2FS_INODE(node_page);
208 	unsigned long long iblocks;
209 
210 	iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
211 	if (!iblocks) {
212 		set_sbi_flag(sbi, SBI_NEED_FSCK);
213 		f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
214 			  __func__, inode->i_ino, iblocks);
215 		return false;
216 	}
217 
218 	if (ino_of_node(node_page) != nid_of_node(node_page)) {
219 		set_sbi_flag(sbi, SBI_NEED_FSCK);
220 		f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
221 			  __func__, inode->i_ino,
222 			  ino_of_node(node_page), nid_of_node(node_page));
223 		return false;
224 	}
225 
226 	if (f2fs_sb_has_flexible_inline_xattr(sbi)
227 			&& !f2fs_has_extra_attr(inode)) {
228 		set_sbi_flag(sbi, SBI_NEED_FSCK);
229 		f2fs_warn(sbi, "%s: corrupted inode ino=%lx, run fsck to fix.",
230 			  __func__, inode->i_ino);
231 		return false;
232 	}
233 
234 	if (f2fs_has_extra_attr(inode) &&
235 			!f2fs_sb_has_extra_attr(sbi)) {
236 		set_sbi_flag(sbi, SBI_NEED_FSCK);
237 		f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
238 			  __func__, inode->i_ino);
239 		return false;
240 	}
241 
242 	if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
243 			fi->i_extra_isize % sizeof(__le32)) {
244 		set_sbi_flag(sbi, SBI_NEED_FSCK);
245 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
246 			  __func__, inode->i_ino, fi->i_extra_isize,
247 			  F2FS_TOTAL_EXTRA_ATTR_SIZE);
248 		return false;
249 	}
250 
251 	if (f2fs_has_extra_attr(inode) &&
252 		f2fs_sb_has_flexible_inline_xattr(sbi) &&
253 		f2fs_has_inline_xattr(inode) &&
254 		(!fi->i_inline_xattr_size ||
255 		fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
256 		set_sbi_flag(sbi, SBI_NEED_FSCK);
257 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %zu",
258 			  __func__, inode->i_ino, fi->i_inline_xattr_size,
259 			  MAX_INLINE_XATTR_SIZE);
260 		return false;
261 	}
262 
263 	if (F2FS_I(inode)->extent_tree) {
264 		struct extent_info *ei = &F2FS_I(inode)->extent_tree->largest;
265 
266 		if (ei->len &&
267 			(!f2fs_is_valid_blkaddr(sbi, ei->blk,
268 						DATA_GENERIC_ENHANCE) ||
269 			!f2fs_is_valid_blkaddr(sbi, ei->blk + ei->len - 1,
270 						DATA_GENERIC_ENHANCE))) {
271 			set_sbi_flag(sbi, SBI_NEED_FSCK);
272 			f2fs_warn(sbi, "%s: inode (ino=%lx) extent info [%u, %u, %u] is incorrect, run fsck to fix",
273 				  __func__, inode->i_ino,
274 				  ei->blk, ei->fofs, ei->len);
275 			return false;
276 		}
277 	}
278 
279 	if (f2fs_has_inline_data(inode) &&
280 			(!S_ISREG(inode->i_mode) && !S_ISLNK(inode->i_mode))) {
281 		set_sbi_flag(sbi, SBI_NEED_FSCK);
282 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
283 			  __func__, inode->i_ino, inode->i_mode);
284 		return false;
285 	}
286 
287 	if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
288 		set_sbi_flag(sbi, SBI_NEED_FSCK);
289 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
290 			  __func__, inode->i_ino, inode->i_mode);
291 		return false;
292 	}
293 
294 	if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
295 		set_sbi_flag(sbi, SBI_NEED_FSCK);
296 		f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
297 			  __func__, inode->i_ino);
298 		return false;
299 	}
300 
301 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
302 			fi->i_flags & F2FS_COMPR_FL &&
303 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
304 						i_log_cluster_size)) {
305 		if (ri->i_compress_algorithm >= COMPRESS_MAX) {
306 			set_sbi_flag(sbi, SBI_NEED_FSCK);
307 			f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported "
308 				"compress algorithm: %u, run fsck to fix",
309 				  __func__, inode->i_ino,
310 				  ri->i_compress_algorithm);
311 			return false;
312 		}
313 		if (le64_to_cpu(ri->i_compr_blocks) >
314 				SECTOR_TO_BLOCK(inode->i_blocks)) {
315 			set_sbi_flag(sbi, SBI_NEED_FSCK);
316 			f2fs_warn(sbi, "%s: inode (ino=%lx) has inconsistent "
317 				"i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
318 				  __func__, inode->i_ino,
319 				  le64_to_cpu(ri->i_compr_blocks),
320 				  SECTOR_TO_BLOCK(inode->i_blocks));
321 			return false;
322 		}
323 		if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
324 			ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
325 			set_sbi_flag(sbi, SBI_NEED_FSCK);
326 			f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported "
327 				"log cluster size: %u, run fsck to fix",
328 				  __func__, inode->i_ino,
329 				  ri->i_log_cluster_size);
330 			return false;
331 		}
332 	}
333 
334 	return true;
335 }
336 
337 static int do_read_inode(struct inode *inode)
338 {
339 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
340 	struct f2fs_inode_info *fi = F2FS_I(inode);
341 	struct page *node_page;
342 	struct f2fs_inode *ri;
343 	projid_t i_projid;
344 	int err;
345 
346 	/* Check if ino is within scope */
347 	if (f2fs_check_nid_range(sbi, inode->i_ino))
348 		return -EINVAL;
349 
350 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
351 	if (IS_ERR(node_page))
352 		return PTR_ERR(node_page);
353 
354 	ri = F2FS_INODE(node_page);
355 
356 	inode->i_mode = le16_to_cpu(ri->i_mode);
357 	i_uid_write(inode, le32_to_cpu(ri->i_uid));
358 	i_gid_write(inode, le32_to_cpu(ri->i_gid));
359 	set_nlink(inode, le32_to_cpu(ri->i_links));
360 	inode->i_size = le64_to_cpu(ri->i_size);
361 	inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
362 
363 	inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
364 	inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
365 	inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
366 	inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
367 	inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
368 	inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
369 	inode->i_generation = le32_to_cpu(ri->i_generation);
370 	if (S_ISDIR(inode->i_mode))
371 		fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
372 	else if (S_ISREG(inode->i_mode))
373 		fi->i_gc_failures[GC_FAILURE_PIN] =
374 					le16_to_cpu(ri->i_gc_failures);
375 	fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
376 	fi->i_flags = le32_to_cpu(ri->i_flags);
377 	if (S_ISREG(inode->i_mode))
378 		fi->i_flags &= ~F2FS_PROJINHERIT_FL;
379 	bitmap_zero(fi->flags, FI_MAX);
380 	fi->i_advise = ri->i_advise;
381 	fi->i_pino = le32_to_cpu(ri->i_pino);
382 	fi->i_dir_level = ri->i_dir_level;
383 
384 	f2fs_init_extent_tree(inode, node_page);
385 
386 	get_inline_info(inode, ri);
387 
388 	fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
389 					le16_to_cpu(ri->i_extra_isize) : 0;
390 
391 	if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
392 		fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
393 	} else if (f2fs_has_inline_xattr(inode) ||
394 				f2fs_has_inline_dentry(inode)) {
395 		fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
396 	} else {
397 
398 		/*
399 		 * Previous inline data or directory always reserved 200 bytes
400 		 * in inode layout, even if inline_xattr is disabled. In order
401 		 * to keep inline_dentry's structure for backward compatibility,
402 		 * we get the space back only from inline_data.
403 		 */
404 		fi->i_inline_xattr_size = 0;
405 	}
406 
407 	if (!sanity_check_inode(inode, node_page)) {
408 		f2fs_put_page(node_page, 1);
409 		return -EFSCORRUPTED;
410 	}
411 
412 	/* check data exist */
413 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
414 		__recover_inline_status(inode, node_page);
415 
416 	/* try to recover cold bit for non-dir inode */
417 	if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
418 		f2fs_wait_on_page_writeback(node_page, NODE, true, true);
419 		set_cold_node(node_page, false);
420 		set_page_dirty(node_page);
421 	}
422 
423 	/* get rdev by using inline_info */
424 	__get_inode_rdev(inode, ri);
425 
426 	if (S_ISREG(inode->i_mode)) {
427 		err = __written_first_block(sbi, ri);
428 		if (err < 0) {
429 			f2fs_put_page(node_page, 1);
430 			return err;
431 		}
432 		if (!err)
433 			set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
434 	}
435 
436 	if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
437 		fi->last_disk_size = inode->i_size;
438 
439 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
440 		set_inode_flag(inode, FI_PROJ_INHERIT);
441 
442 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
443 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
444 		i_projid = (projid_t)le32_to_cpu(ri->i_projid);
445 	else
446 		i_projid = F2FS_DEF_PROJID;
447 	fi->i_projid = make_kprojid(&init_user_ns, i_projid);
448 
449 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
450 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
451 		fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
452 		fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
453 	}
454 
455 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
456 					(fi->i_flags & F2FS_COMPR_FL)) {
457 		if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
458 					i_log_cluster_size)) {
459 			atomic_set(&fi->i_compr_blocks,
460 					le64_to_cpu(ri->i_compr_blocks));
461 			fi->i_compress_algorithm = ri->i_compress_algorithm;
462 			fi->i_log_cluster_size = ri->i_log_cluster_size;
463 			fi->i_compress_flag = le16_to_cpu(ri->i_compress_flag);
464 			fi->i_cluster_size = 1 << fi->i_log_cluster_size;
465 			set_inode_flag(inode, FI_COMPRESSED_FILE);
466 		}
467 	}
468 
469 	F2FS_I(inode)->i_disk_time[0] = inode->i_atime;
470 	F2FS_I(inode)->i_disk_time[1] = inode->i_ctime;
471 	F2FS_I(inode)->i_disk_time[2] = inode->i_mtime;
472 	F2FS_I(inode)->i_disk_time[3] = F2FS_I(inode)->i_crtime;
473 	f2fs_put_page(node_page, 1);
474 
475 	stat_inc_inline_xattr(inode);
476 	stat_inc_inline_inode(inode);
477 	stat_inc_inline_dir(inode);
478 	stat_inc_compr_inode(inode);
479 	stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
480 
481 	return 0;
482 }
483 
484 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
485 {
486 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
487 	struct inode *inode;
488 	int ret = 0;
489 
490 	inode = iget_locked(sb, ino);
491 	if (!inode)
492 		return ERR_PTR(-ENOMEM);
493 
494 	if (!(inode->i_state & I_NEW)) {
495 		trace_f2fs_iget(inode);
496 		return inode;
497 	}
498 	if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
499 		goto make_now;
500 
501 #ifdef CONFIG_F2FS_FS_COMPRESSION
502 	if (ino == F2FS_COMPRESS_INO(sbi))
503 		goto make_now;
504 #endif
505 
506 	ret = do_read_inode(inode);
507 	if (ret)
508 		goto bad_inode;
509 make_now:
510 	if (ino == F2FS_NODE_INO(sbi)) {
511 		inode->i_mapping->a_ops = &f2fs_node_aops;
512 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
513 	} else if (ino == F2FS_META_INO(sbi)) {
514 		inode->i_mapping->a_ops = &f2fs_meta_aops;
515 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
516 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
517 #ifdef CONFIG_F2FS_FS_COMPRESSION
518 		inode->i_mapping->a_ops = &f2fs_compress_aops;
519 #endif
520 		mapping_set_gfp_mask(inode->i_mapping,
521 			GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
522 	} else if (S_ISREG(inode->i_mode)) {
523 		inode->i_op = &f2fs_file_inode_operations;
524 		inode->i_fop = &f2fs_file_operations;
525 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
526 	} else if (S_ISDIR(inode->i_mode)) {
527 		inode->i_op = &f2fs_dir_inode_operations;
528 		inode->i_fop = &f2fs_dir_operations;
529 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
530 		inode_nohighmem(inode);
531 	} else if (S_ISLNK(inode->i_mode)) {
532 		if (file_is_encrypt(inode))
533 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
534 		else
535 			inode->i_op = &f2fs_symlink_inode_operations;
536 		inode_nohighmem(inode);
537 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
538 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
539 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
540 		inode->i_op = &f2fs_special_inode_operations;
541 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
542 	} else {
543 		ret = -EIO;
544 		goto bad_inode;
545 	}
546 	f2fs_set_inode_flags(inode);
547 	unlock_new_inode(inode);
548 	trace_f2fs_iget(inode);
549 	return inode;
550 
551 bad_inode:
552 	f2fs_inode_synced(inode);
553 	iget_failed(inode);
554 	trace_f2fs_iget_exit(inode, ret);
555 	return ERR_PTR(ret);
556 }
557 
558 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
559 {
560 	struct inode *inode;
561 retry:
562 	inode = f2fs_iget(sb, ino);
563 	if (IS_ERR(inode)) {
564 		if (PTR_ERR(inode) == -ENOMEM) {
565 			congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
566 			goto retry;
567 		}
568 	}
569 	return inode;
570 }
571 
572 void f2fs_update_inode(struct inode *inode, struct page *node_page)
573 {
574 	struct f2fs_inode *ri;
575 	struct extent_tree *et = F2FS_I(inode)->extent_tree;
576 
577 	f2fs_wait_on_page_writeback(node_page, NODE, true, true);
578 	set_page_dirty(node_page);
579 
580 	f2fs_inode_synced(inode);
581 
582 	ri = F2FS_INODE(node_page);
583 
584 	ri->i_mode = cpu_to_le16(inode->i_mode);
585 	ri->i_advise = F2FS_I(inode)->i_advise;
586 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
587 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
588 	ri->i_links = cpu_to_le32(inode->i_nlink);
589 	ri->i_size = cpu_to_le64(i_size_read(inode));
590 	ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
591 
592 	if (et) {
593 		read_lock(&et->lock);
594 		set_raw_extent(&et->largest, &ri->i_ext);
595 		read_unlock(&et->lock);
596 	} else {
597 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
598 	}
599 	set_raw_inline(inode, ri);
600 
601 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
602 	ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
603 	ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
604 	ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
605 	ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
606 	ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
607 	if (S_ISDIR(inode->i_mode))
608 		ri->i_current_depth =
609 			cpu_to_le32(F2FS_I(inode)->i_current_depth);
610 	else if (S_ISREG(inode->i_mode))
611 		ri->i_gc_failures =
612 			cpu_to_le16(F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]);
613 	ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
614 	ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
615 	ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
616 	ri->i_generation = cpu_to_le32(inode->i_generation);
617 	ri->i_dir_level = F2FS_I(inode)->i_dir_level;
618 
619 	if (f2fs_has_extra_attr(inode)) {
620 		ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
621 
622 		if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
623 			ri->i_inline_xattr_size =
624 				cpu_to_le16(F2FS_I(inode)->i_inline_xattr_size);
625 
626 		if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
627 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
628 								i_projid)) {
629 			projid_t i_projid;
630 
631 			i_projid = from_kprojid(&init_user_ns,
632 						F2FS_I(inode)->i_projid);
633 			ri->i_projid = cpu_to_le32(i_projid);
634 		}
635 
636 		if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
637 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
638 								i_crtime)) {
639 			ri->i_crtime =
640 				cpu_to_le64(F2FS_I(inode)->i_crtime.tv_sec);
641 			ri->i_crtime_nsec =
642 				cpu_to_le32(F2FS_I(inode)->i_crtime.tv_nsec);
643 		}
644 
645 		if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
646 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
647 							i_log_cluster_size)) {
648 			ri->i_compr_blocks =
649 				cpu_to_le64(atomic_read(
650 					&F2FS_I(inode)->i_compr_blocks));
651 			ri->i_compress_algorithm =
652 				F2FS_I(inode)->i_compress_algorithm;
653 			ri->i_compress_flag =
654 				cpu_to_le16(F2FS_I(inode)->i_compress_flag);
655 			ri->i_log_cluster_size =
656 				F2FS_I(inode)->i_log_cluster_size;
657 		}
658 	}
659 
660 	__set_inode_rdev(inode, ri);
661 
662 	/* deleted inode */
663 	if (inode->i_nlink == 0)
664 		clear_page_private_inline(node_page);
665 
666 	F2FS_I(inode)->i_disk_time[0] = inode->i_atime;
667 	F2FS_I(inode)->i_disk_time[1] = inode->i_ctime;
668 	F2FS_I(inode)->i_disk_time[2] = inode->i_mtime;
669 	F2FS_I(inode)->i_disk_time[3] = F2FS_I(inode)->i_crtime;
670 
671 #ifdef CONFIG_F2FS_CHECK_FS
672 	f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
673 #endif
674 }
675 
676 void f2fs_update_inode_page(struct inode *inode)
677 {
678 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
679 	struct page *node_page;
680 retry:
681 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
682 	if (IS_ERR(node_page)) {
683 		int err = PTR_ERR(node_page);
684 
685 		if (err == -ENOMEM) {
686 			cond_resched();
687 			goto retry;
688 		} else if (err != -ENOENT) {
689 			f2fs_stop_checkpoint(sbi, false);
690 		}
691 		return;
692 	}
693 	f2fs_update_inode(inode, node_page);
694 	f2fs_put_page(node_page, 1);
695 }
696 
697 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
698 {
699 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
700 
701 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
702 			inode->i_ino == F2FS_META_INO(sbi))
703 		return 0;
704 
705 	/*
706 	 * atime could be updated without dirtying f2fs inode in lazytime mode
707 	 */
708 	if (f2fs_is_time_consistent(inode) &&
709 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
710 		return 0;
711 
712 	if (!f2fs_is_checkpoint_ready(sbi))
713 		return -ENOSPC;
714 
715 	/*
716 	 * We need to balance fs here to prevent from producing dirty node pages
717 	 * during the urgent cleaning time when running out of free sections.
718 	 */
719 	f2fs_update_inode_page(inode);
720 	if (wbc && wbc->nr_to_write)
721 		f2fs_balance_fs(sbi, true);
722 	return 0;
723 }
724 
725 /*
726  * Called at the last iput() if i_nlink is zero
727  */
728 void f2fs_evict_inode(struct inode *inode)
729 {
730 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
731 	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
732 	int err = 0;
733 
734 	/* some remained atomic pages should discarded */
735 	if (f2fs_is_atomic_file(inode))
736 		f2fs_drop_inmem_pages(inode);
737 
738 	trace_f2fs_evict_inode(inode);
739 	truncate_inode_pages_final(&inode->i_data);
740 
741 	if (test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
742 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
743 
744 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
745 			inode->i_ino == F2FS_META_INO(sbi) ||
746 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
747 		goto out_clear;
748 
749 	f2fs_bug_on(sbi, get_dirty_pages(inode));
750 	f2fs_remove_dirty_inode(inode);
751 
752 	f2fs_destroy_extent_tree(inode);
753 
754 	if (inode->i_nlink || is_bad_inode(inode))
755 		goto no_delete;
756 
757 	err = dquot_initialize(inode);
758 	if (err) {
759 		err = 0;
760 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
761 	}
762 
763 	f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
764 	f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
765 	f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
766 
767 	sb_start_intwrite(inode->i_sb);
768 	set_inode_flag(inode, FI_NO_ALLOC);
769 	i_size_write(inode, 0);
770 retry:
771 	if (F2FS_HAS_BLOCKS(inode))
772 		err = f2fs_truncate(inode);
773 
774 	if (time_to_inject(sbi, FAULT_EVICT_INODE)) {
775 		f2fs_show_injection_info(sbi, FAULT_EVICT_INODE);
776 		err = -EIO;
777 	}
778 
779 	if (!err) {
780 		f2fs_lock_op(sbi);
781 		err = f2fs_remove_inode_page(inode);
782 		f2fs_unlock_op(sbi);
783 		if (err == -ENOENT)
784 			err = 0;
785 	}
786 
787 	/* give more chances, if ENOMEM case */
788 	if (err == -ENOMEM) {
789 		err = 0;
790 		goto retry;
791 	}
792 
793 	if (err) {
794 		f2fs_update_inode_page(inode);
795 		if (dquot_initialize_needed(inode))
796 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
797 	}
798 	sb_end_intwrite(inode->i_sb);
799 no_delete:
800 	dquot_drop(inode);
801 
802 	stat_dec_inline_xattr(inode);
803 	stat_dec_inline_dir(inode);
804 	stat_dec_inline_inode(inode);
805 	stat_dec_compr_inode(inode);
806 	stat_sub_compr_blocks(inode,
807 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
808 
809 	if (likely(!f2fs_cp_error(sbi) &&
810 				!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
811 		f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
812 	else
813 		f2fs_inode_synced(inode);
814 
815 	/* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
816 	if (inode->i_ino)
817 		invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
818 							inode->i_ino);
819 	if (xnid)
820 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
821 	if (inode->i_nlink) {
822 		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
823 			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
824 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
825 			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
826 	}
827 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
828 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
829 		clear_inode_flag(inode, FI_FREE_NID);
830 	} else {
831 		/*
832 		 * If xattr nid is corrupted, we can reach out error condition,
833 		 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
834 		 * In that case, f2fs_check_nid_range() is enough to give a clue.
835 		 */
836 	}
837 out_clear:
838 	fscrypt_put_encryption_info(inode);
839 	fsverity_cleanup_inode(inode);
840 	clear_inode(inode);
841 }
842 
843 /* caller should call f2fs_lock_op() */
844 void f2fs_handle_failed_inode(struct inode *inode)
845 {
846 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
847 	struct node_info ni;
848 	int err;
849 
850 	/*
851 	 * clear nlink of inode in order to release resource of inode
852 	 * immediately.
853 	 */
854 	clear_nlink(inode);
855 
856 	/*
857 	 * we must call this to avoid inode being remained as dirty, resulting
858 	 * in a panic when flushing dirty inodes in gdirty_list.
859 	 */
860 	f2fs_update_inode_page(inode);
861 	f2fs_inode_synced(inode);
862 
863 	/* don't make bad inode, since it becomes a regular file. */
864 	unlock_new_inode(inode);
865 
866 	/*
867 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
868 	 * so we can prevent losing this orphan when encoutering checkpoint
869 	 * and following suddenly power-off.
870 	 */
871 	err = f2fs_get_node_info(sbi, inode->i_ino, &ni);
872 	if (err) {
873 		set_sbi_flag(sbi, SBI_NEED_FSCK);
874 		f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
875 		goto out;
876 	}
877 
878 	if (ni.blk_addr != NULL_ADDR) {
879 		err = f2fs_acquire_orphan_inode(sbi);
880 		if (err) {
881 			set_sbi_flag(sbi, SBI_NEED_FSCK);
882 			f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
883 		} else {
884 			f2fs_add_orphan_inode(inode);
885 		}
886 		f2fs_alloc_nid_done(sbi, inode->i_ino);
887 	} else {
888 		set_inode_flag(inode, FI_FREE_NID);
889 	}
890 
891 out:
892 	f2fs_unlock_op(sbi);
893 
894 	/* iput will drop the inode object */
895 	iput(inode);
896 }
897