xref: /openbmc/linux/fs/f2fs/inode.c (revision b868a02e)
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) {
266 		struct extent_info *ei = &fi->extent_tree->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 	f2fs_init_extent_tree(inode, node_page);
396 
397 	get_inline_info(inode, ri);
398 
399 	fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
400 					le16_to_cpu(ri->i_extra_isize) : 0;
401 
402 	if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
403 		fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
404 	} else if (f2fs_has_inline_xattr(inode) ||
405 				f2fs_has_inline_dentry(inode)) {
406 		fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
407 	} else {
408 
409 		/*
410 		 * Previous inline data or directory always reserved 200 bytes
411 		 * in inode layout, even if inline_xattr is disabled. In order
412 		 * to keep inline_dentry's structure for backward compatibility,
413 		 * we get the space back only from inline_data.
414 		 */
415 		fi->i_inline_xattr_size = 0;
416 	}
417 
418 	if (!sanity_check_inode(inode, node_page)) {
419 		f2fs_put_page(node_page, 1);
420 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
421 		return -EFSCORRUPTED;
422 	}
423 
424 	/* check data exist */
425 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
426 		__recover_inline_status(inode, node_page);
427 
428 	/* try to recover cold bit for non-dir inode */
429 	if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
430 		f2fs_wait_on_page_writeback(node_page, NODE, true, true);
431 		set_cold_node(node_page, false);
432 		set_page_dirty(node_page);
433 	}
434 
435 	/* get rdev by using inline_info */
436 	__get_inode_rdev(inode, ri);
437 
438 	if (S_ISREG(inode->i_mode)) {
439 		err = __written_first_block(sbi, ri);
440 		if (err < 0) {
441 			f2fs_put_page(node_page, 1);
442 			return err;
443 		}
444 		if (!err)
445 			set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
446 	}
447 
448 	if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
449 		fi->last_disk_size = inode->i_size;
450 
451 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
452 		set_inode_flag(inode, FI_PROJ_INHERIT);
453 
454 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
455 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
456 		i_projid = (projid_t)le32_to_cpu(ri->i_projid);
457 	else
458 		i_projid = F2FS_DEF_PROJID;
459 	fi->i_projid = make_kprojid(&init_user_ns, i_projid);
460 
461 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
462 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
463 		fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
464 		fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
465 	}
466 
467 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
468 					(fi->i_flags & F2FS_COMPR_FL)) {
469 		if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
470 					i_log_cluster_size)) {
471 			atomic_set(&fi->i_compr_blocks,
472 					le64_to_cpu(ri->i_compr_blocks));
473 			fi->i_compress_algorithm = ri->i_compress_algorithm;
474 			fi->i_log_cluster_size = ri->i_log_cluster_size;
475 			fi->i_compress_flag = le16_to_cpu(ri->i_compress_flag);
476 			fi->i_cluster_size = 1 << fi->i_log_cluster_size;
477 			set_inode_flag(inode, FI_COMPRESSED_FILE);
478 		}
479 	}
480 
481 	init_idisk_time(inode);
482 	f2fs_put_page(node_page, 1);
483 
484 	stat_inc_inline_xattr(inode);
485 	stat_inc_inline_inode(inode);
486 	stat_inc_inline_dir(inode);
487 	stat_inc_compr_inode(inode);
488 	stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
489 
490 	return 0;
491 }
492 
493 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
494 {
495 	return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
496 		ino == F2FS_COMPRESS_INO(sbi);
497 }
498 
499 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
500 {
501 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
502 	struct inode *inode;
503 	int ret = 0;
504 
505 	inode = iget_locked(sb, ino);
506 	if (!inode)
507 		return ERR_PTR(-ENOMEM);
508 
509 	if (!(inode->i_state & I_NEW)) {
510 		if (is_meta_ino(sbi, ino)) {
511 			f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
512 			set_sbi_flag(sbi, SBI_NEED_FSCK);
513 			ret = -EFSCORRUPTED;
514 			trace_f2fs_iget_exit(inode, ret);
515 			iput(inode);
516 			f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
517 			return ERR_PTR(ret);
518 		}
519 
520 		trace_f2fs_iget(inode);
521 		return inode;
522 	}
523 
524 	if (is_meta_ino(sbi, ino))
525 		goto make_now;
526 
527 	ret = do_read_inode(inode);
528 	if (ret)
529 		goto bad_inode;
530 make_now:
531 	if (ino == F2FS_NODE_INO(sbi)) {
532 		inode->i_mapping->a_ops = &f2fs_node_aops;
533 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
534 	} else if (ino == F2FS_META_INO(sbi)) {
535 		inode->i_mapping->a_ops = &f2fs_meta_aops;
536 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
537 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
538 #ifdef CONFIG_F2FS_FS_COMPRESSION
539 		inode->i_mapping->a_ops = &f2fs_compress_aops;
540 		/*
541 		 * generic_error_remove_page only truncates pages of regular
542 		 * inode
543 		 */
544 		inode->i_mode |= S_IFREG;
545 #endif
546 		mapping_set_gfp_mask(inode->i_mapping,
547 			GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
548 	} else if (S_ISREG(inode->i_mode)) {
549 		inode->i_op = &f2fs_file_inode_operations;
550 		inode->i_fop = &f2fs_file_operations;
551 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
552 	} else if (S_ISDIR(inode->i_mode)) {
553 		inode->i_op = &f2fs_dir_inode_operations;
554 		inode->i_fop = &f2fs_dir_operations;
555 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
556 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
557 	} else if (S_ISLNK(inode->i_mode)) {
558 		if (file_is_encrypt(inode))
559 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
560 		else
561 			inode->i_op = &f2fs_symlink_inode_operations;
562 		inode_nohighmem(inode);
563 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
564 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
565 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
566 		inode->i_op = &f2fs_special_inode_operations;
567 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
568 	} else {
569 		ret = -EIO;
570 		goto bad_inode;
571 	}
572 	f2fs_set_inode_flags(inode);
573 
574 	if (file_should_truncate(inode) &&
575 			!is_sbi_flag_set(sbi, SBI_POR_DOING)) {
576 		ret = f2fs_truncate(inode);
577 		if (ret)
578 			goto bad_inode;
579 		file_dont_truncate(inode);
580 	}
581 
582 	unlock_new_inode(inode);
583 	trace_f2fs_iget(inode);
584 	return inode;
585 
586 bad_inode:
587 	f2fs_inode_synced(inode);
588 	iget_failed(inode);
589 	trace_f2fs_iget_exit(inode, ret);
590 	return ERR_PTR(ret);
591 }
592 
593 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
594 {
595 	struct inode *inode;
596 retry:
597 	inode = f2fs_iget(sb, ino);
598 	if (IS_ERR(inode)) {
599 		if (PTR_ERR(inode) == -ENOMEM) {
600 			memalloc_retry_wait(GFP_NOFS);
601 			goto retry;
602 		}
603 	}
604 	return inode;
605 }
606 
607 void f2fs_update_inode(struct inode *inode, struct page *node_page)
608 {
609 	struct f2fs_inode *ri;
610 	struct extent_tree *et = F2FS_I(inode)->extent_tree;
611 
612 	f2fs_wait_on_page_writeback(node_page, NODE, true, true);
613 	set_page_dirty(node_page);
614 
615 	f2fs_inode_synced(inode);
616 
617 	ri = F2FS_INODE(node_page);
618 
619 	ri->i_mode = cpu_to_le16(inode->i_mode);
620 	ri->i_advise = F2FS_I(inode)->i_advise;
621 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
622 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
623 	ri->i_links = cpu_to_le32(inode->i_nlink);
624 	ri->i_size = cpu_to_le64(i_size_read(inode));
625 	ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
626 
627 	if (et) {
628 		read_lock(&et->lock);
629 		set_raw_extent(&et->largest, &ri->i_ext);
630 		read_unlock(&et->lock);
631 	} else {
632 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
633 	}
634 	set_raw_inline(inode, ri);
635 
636 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
637 	ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
638 	ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
639 	ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
640 	ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
641 	ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
642 	if (S_ISDIR(inode->i_mode))
643 		ri->i_current_depth =
644 			cpu_to_le32(F2FS_I(inode)->i_current_depth);
645 	else if (S_ISREG(inode->i_mode))
646 		ri->i_gc_failures =
647 			cpu_to_le16(F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]);
648 	ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
649 	ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
650 	ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
651 	ri->i_generation = cpu_to_le32(inode->i_generation);
652 	ri->i_dir_level = F2FS_I(inode)->i_dir_level;
653 
654 	if (f2fs_has_extra_attr(inode)) {
655 		ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
656 
657 		if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
658 			ri->i_inline_xattr_size =
659 				cpu_to_le16(F2FS_I(inode)->i_inline_xattr_size);
660 
661 		if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
662 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
663 								i_projid)) {
664 			projid_t i_projid;
665 
666 			i_projid = from_kprojid(&init_user_ns,
667 						F2FS_I(inode)->i_projid);
668 			ri->i_projid = cpu_to_le32(i_projid);
669 		}
670 
671 		if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
672 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
673 								i_crtime)) {
674 			ri->i_crtime =
675 				cpu_to_le64(F2FS_I(inode)->i_crtime.tv_sec);
676 			ri->i_crtime_nsec =
677 				cpu_to_le32(F2FS_I(inode)->i_crtime.tv_nsec);
678 		}
679 
680 		if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
681 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
682 							i_log_cluster_size)) {
683 			ri->i_compr_blocks =
684 				cpu_to_le64(atomic_read(
685 					&F2FS_I(inode)->i_compr_blocks));
686 			ri->i_compress_algorithm =
687 				F2FS_I(inode)->i_compress_algorithm;
688 			ri->i_compress_flag =
689 				cpu_to_le16(F2FS_I(inode)->i_compress_flag);
690 			ri->i_log_cluster_size =
691 				F2FS_I(inode)->i_log_cluster_size;
692 		}
693 	}
694 
695 	__set_inode_rdev(inode, ri);
696 
697 	/* deleted inode */
698 	if (inode->i_nlink == 0)
699 		clear_page_private_inline(node_page);
700 
701 	init_idisk_time(inode);
702 #ifdef CONFIG_F2FS_CHECK_FS
703 	f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
704 #endif
705 }
706 
707 void f2fs_update_inode_page(struct inode *inode)
708 {
709 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
710 	struct page *node_page;
711 retry:
712 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
713 	if (IS_ERR(node_page)) {
714 		int err = PTR_ERR(node_page);
715 
716 		if (err == -ENOMEM) {
717 			cond_resched();
718 			goto retry;
719 		} else if (err != -ENOENT) {
720 			f2fs_stop_checkpoint(sbi, false,
721 					STOP_CP_REASON_UPDATE_INODE);
722 		}
723 		return;
724 	}
725 	f2fs_update_inode(inode, node_page);
726 	f2fs_put_page(node_page, 1);
727 }
728 
729 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
730 {
731 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
732 
733 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
734 			inode->i_ino == F2FS_META_INO(sbi))
735 		return 0;
736 
737 	/*
738 	 * atime could be updated without dirtying f2fs inode in lazytime mode
739 	 */
740 	if (f2fs_is_time_consistent(inode) &&
741 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
742 		return 0;
743 
744 	if (!f2fs_is_checkpoint_ready(sbi))
745 		return -ENOSPC;
746 
747 	/*
748 	 * We need to balance fs here to prevent from producing dirty node pages
749 	 * during the urgent cleaning time when running out of free sections.
750 	 */
751 	f2fs_update_inode_page(inode);
752 	if (wbc && wbc->nr_to_write)
753 		f2fs_balance_fs(sbi, true);
754 	return 0;
755 }
756 
757 /*
758  * Called at the last iput() if i_nlink is zero
759  */
760 void f2fs_evict_inode(struct inode *inode)
761 {
762 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
763 	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
764 	int err = 0;
765 
766 	f2fs_abort_atomic_write(inode, true);
767 
768 	trace_f2fs_evict_inode(inode);
769 	truncate_inode_pages_final(&inode->i_data);
770 
771 	if ((inode->i_nlink || is_bad_inode(inode)) &&
772 		test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
773 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
774 
775 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
776 			inode->i_ino == F2FS_META_INO(sbi) ||
777 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
778 		goto out_clear;
779 
780 	f2fs_bug_on(sbi, get_dirty_pages(inode));
781 	f2fs_remove_dirty_inode(inode);
782 
783 	f2fs_destroy_extent_tree(inode);
784 
785 	if (inode->i_nlink || is_bad_inode(inode))
786 		goto no_delete;
787 
788 	err = f2fs_dquot_initialize(inode);
789 	if (err) {
790 		err = 0;
791 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
792 	}
793 
794 	f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
795 	f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
796 	f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
797 
798 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
799 		sb_start_intwrite(inode->i_sb);
800 	set_inode_flag(inode, FI_NO_ALLOC);
801 	i_size_write(inode, 0);
802 retry:
803 	if (F2FS_HAS_BLOCKS(inode))
804 		err = f2fs_truncate(inode);
805 
806 	if (time_to_inject(sbi, FAULT_EVICT_INODE)) {
807 		f2fs_show_injection_info(sbi, FAULT_EVICT_INODE);
808 		err = -EIO;
809 	}
810 
811 	if (!err) {
812 		f2fs_lock_op(sbi);
813 		err = f2fs_remove_inode_page(inode);
814 		f2fs_unlock_op(sbi);
815 		if (err == -ENOENT) {
816 			err = 0;
817 
818 			/*
819 			 * in fuzzed image, another node may has the same
820 			 * block address as inode's, if it was truncated
821 			 * previously, truncation of inode node will fail.
822 			 */
823 			if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
824 				f2fs_warn(F2FS_I_SB(inode),
825 					"f2fs_evict_inode: inconsistent node id, ino:%lu",
826 					inode->i_ino);
827 				f2fs_inode_synced(inode);
828 				set_sbi_flag(sbi, SBI_NEED_FSCK);
829 			}
830 		}
831 	}
832 
833 	/* give more chances, if ENOMEM case */
834 	if (err == -ENOMEM) {
835 		err = 0;
836 		goto retry;
837 	}
838 
839 	if (err) {
840 		f2fs_update_inode_page(inode);
841 		if (dquot_initialize_needed(inode))
842 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
843 	}
844 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
845 		sb_end_intwrite(inode->i_sb);
846 no_delete:
847 	dquot_drop(inode);
848 
849 	stat_dec_inline_xattr(inode);
850 	stat_dec_inline_dir(inode);
851 	stat_dec_inline_inode(inode);
852 	stat_dec_compr_inode(inode);
853 	stat_sub_compr_blocks(inode,
854 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
855 
856 	if (likely(!f2fs_cp_error(sbi) &&
857 				!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
858 		f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
859 	else
860 		f2fs_inode_synced(inode);
861 
862 	/* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
863 	if (inode->i_ino)
864 		invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
865 							inode->i_ino);
866 	if (xnid)
867 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
868 	if (inode->i_nlink) {
869 		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
870 			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
871 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
872 			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
873 	}
874 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
875 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
876 		clear_inode_flag(inode, FI_FREE_NID);
877 	} else {
878 		/*
879 		 * If xattr nid is corrupted, we can reach out error condition,
880 		 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
881 		 * In that case, f2fs_check_nid_range() is enough to give a clue.
882 		 */
883 	}
884 out_clear:
885 	fscrypt_put_encryption_info(inode);
886 	fsverity_cleanup_inode(inode);
887 	clear_inode(inode);
888 }
889 
890 /* caller should call f2fs_lock_op() */
891 void f2fs_handle_failed_inode(struct inode *inode)
892 {
893 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
894 	struct node_info ni;
895 	int err;
896 
897 	/*
898 	 * clear nlink of inode in order to release resource of inode
899 	 * immediately.
900 	 */
901 	clear_nlink(inode);
902 
903 	/*
904 	 * we must call this to avoid inode being remained as dirty, resulting
905 	 * in a panic when flushing dirty inodes in gdirty_list.
906 	 */
907 	f2fs_update_inode_page(inode);
908 	f2fs_inode_synced(inode);
909 
910 	/* don't make bad inode, since it becomes a regular file. */
911 	unlock_new_inode(inode);
912 
913 	/*
914 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
915 	 * so we can prevent losing this orphan when encoutering checkpoint
916 	 * and following suddenly power-off.
917 	 */
918 	err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
919 	if (err) {
920 		set_sbi_flag(sbi, SBI_NEED_FSCK);
921 		set_inode_flag(inode, FI_FREE_NID);
922 		f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
923 		goto out;
924 	}
925 
926 	if (ni.blk_addr != NULL_ADDR) {
927 		err = f2fs_acquire_orphan_inode(sbi);
928 		if (err) {
929 			set_sbi_flag(sbi, SBI_NEED_FSCK);
930 			f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
931 		} else {
932 			f2fs_add_orphan_inode(inode);
933 		}
934 		f2fs_alloc_nid_done(sbi, inode->i_ino);
935 	} else {
936 		set_inode_flag(inode, FI_FREE_NID);
937 	}
938 
939 out:
940 	f2fs_unlock_op(sbi);
941 
942 	/* iput will drop the inode object */
943 	iput(inode);
944 }
945