xref: /openbmc/linux/fs/f2fs/inode.c (revision 80d0624d)
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 #include <linux/lz4.h>
14 #include <linux/zstd.h>
15 
16 #include "f2fs.h"
17 #include "node.h"
18 #include "segment.h"
19 #include "xattr.h"
20 
21 #include <trace/events/f2fs.h>
22 
23 #ifdef CONFIG_F2FS_FS_COMPRESSION
24 extern const struct address_space_operations f2fs_compress_aops;
25 #endif
26 
27 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
28 {
29 	if (is_inode_flag_set(inode, FI_NEW_INODE))
30 		return;
31 
32 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
33 		return;
34 
35 	if (f2fs_inode_dirtied(inode, sync))
36 		return;
37 
38 	mark_inode_dirty_sync(inode);
39 }
40 
41 void f2fs_set_inode_flags(struct inode *inode)
42 {
43 	unsigned int flags = F2FS_I(inode)->i_flags;
44 	unsigned int new_fl = 0;
45 
46 	if (flags & F2FS_SYNC_FL)
47 		new_fl |= S_SYNC;
48 	if (flags & F2FS_APPEND_FL)
49 		new_fl |= S_APPEND;
50 	if (flags & F2FS_IMMUTABLE_FL)
51 		new_fl |= S_IMMUTABLE;
52 	if (flags & F2FS_NOATIME_FL)
53 		new_fl |= S_NOATIME;
54 	if (flags & F2FS_DIRSYNC_FL)
55 		new_fl |= S_DIRSYNC;
56 	if (file_is_encrypt(inode))
57 		new_fl |= S_ENCRYPTED;
58 	if (file_is_verity(inode))
59 		new_fl |= S_VERITY;
60 	if (flags & F2FS_CASEFOLD_FL)
61 		new_fl |= S_CASEFOLD;
62 	inode_set_flags(inode, new_fl,
63 			S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|
64 			S_ENCRYPTED|S_VERITY|S_CASEFOLD);
65 }
66 
67 static void __get_inode_rdev(struct inode *inode, struct page *node_page)
68 {
69 	__le32 *addr = get_dnode_addr(inode, node_page);
70 
71 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
72 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
73 		if (addr[0])
74 			inode->i_rdev = old_decode_dev(le32_to_cpu(addr[0]));
75 		else
76 			inode->i_rdev = new_decode_dev(le32_to_cpu(addr[1]));
77 	}
78 }
79 
80 static void __set_inode_rdev(struct inode *inode, struct page *node_page)
81 {
82 	__le32 *addr = get_dnode_addr(inode, node_page);
83 
84 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
85 		if (old_valid_dev(inode->i_rdev)) {
86 			addr[0] = cpu_to_le32(old_encode_dev(inode->i_rdev));
87 			addr[1] = 0;
88 		} else {
89 			addr[0] = 0;
90 			addr[1] = cpu_to_le32(new_encode_dev(inode->i_rdev));
91 			addr[2] = 0;
92 		}
93 	}
94 }
95 
96 static void __recover_inline_status(struct inode *inode, struct page *ipage)
97 {
98 	void *inline_data = inline_data_addr(inode, ipage);
99 	__le32 *start = inline_data;
100 	__le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
101 
102 	while (start < end) {
103 		if (*start++) {
104 			f2fs_wait_on_page_writeback(ipage, NODE, true, true);
105 
106 			set_inode_flag(inode, FI_DATA_EXIST);
107 			set_raw_inline(inode, F2FS_INODE(ipage));
108 			set_page_dirty(ipage);
109 			return;
110 		}
111 	}
112 	return;
113 }
114 
115 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
116 {
117 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
118 
119 	if (!f2fs_sb_has_inode_chksum(sbi))
120 		return false;
121 
122 	if (!IS_INODE(page) || !(ri->i_inline & F2FS_EXTRA_ATTR))
123 		return false;
124 
125 	if (!F2FS_FITS_IN_INODE(ri, le16_to_cpu(ri->i_extra_isize),
126 				i_inode_checksum))
127 		return false;
128 
129 	return true;
130 }
131 
132 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
133 {
134 	struct f2fs_node *node = F2FS_NODE(page);
135 	struct f2fs_inode *ri = &node->i;
136 	__le32 ino = node->footer.ino;
137 	__le32 gen = ri->i_generation;
138 	__u32 chksum, chksum_seed;
139 	__u32 dummy_cs = 0;
140 	unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
141 	unsigned int cs_size = sizeof(dummy_cs);
142 
143 	chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
144 							sizeof(ino));
145 	chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
146 
147 	chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
148 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
149 	offset += cs_size;
150 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
151 						F2FS_BLKSIZE - offset);
152 	return chksum;
153 }
154 
155 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
156 {
157 	struct f2fs_inode *ri;
158 	__u32 provided, calculated;
159 
160 	if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
161 		return true;
162 
163 #ifdef CONFIG_F2FS_CHECK_FS
164 	if (!f2fs_enable_inode_chksum(sbi, page))
165 #else
166 	if (!f2fs_enable_inode_chksum(sbi, page) ||
167 			PageDirty(page) || PageWriteback(page))
168 #endif
169 		return true;
170 
171 	ri = &F2FS_NODE(page)->i;
172 	provided = le32_to_cpu(ri->i_inode_checksum);
173 	calculated = f2fs_inode_chksum(sbi, page);
174 
175 	if (provided != calculated)
176 		f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
177 			  page->index, ino_of_node(page), provided, calculated);
178 
179 	return provided == calculated;
180 }
181 
182 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
183 {
184 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
185 
186 	if (!f2fs_enable_inode_chksum(sbi, page))
187 		return;
188 
189 	ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
190 }
191 
192 static bool sanity_check_compress_inode(struct inode *inode,
193 			struct f2fs_inode *ri)
194 {
195 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
196 	unsigned char clevel;
197 
198 	if (ri->i_compress_algorithm >= COMPRESS_MAX) {
199 		f2fs_warn(sbi,
200 			"%s: inode (ino=%lx) has unsupported compress algorithm: %u, run fsck to fix",
201 			__func__, inode->i_ino, ri->i_compress_algorithm);
202 		return false;
203 	}
204 	if (le64_to_cpu(ri->i_compr_blocks) >
205 			SECTOR_TO_BLOCK(inode->i_blocks)) {
206 		f2fs_warn(sbi,
207 			"%s: inode (ino=%lx) has inconsistent i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
208 			__func__, inode->i_ino, le64_to_cpu(ri->i_compr_blocks),
209 			SECTOR_TO_BLOCK(inode->i_blocks));
210 		return false;
211 	}
212 	if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
213 		ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
214 		f2fs_warn(sbi,
215 			"%s: inode (ino=%lx) has unsupported log cluster size: %u, run fsck to fix",
216 			__func__, inode->i_ino, ri->i_log_cluster_size);
217 		return false;
218 	}
219 
220 	clevel = le16_to_cpu(ri->i_compress_flag) >>
221 				COMPRESS_LEVEL_OFFSET;
222 	switch (ri->i_compress_algorithm) {
223 	case COMPRESS_LZO:
224 #ifdef CONFIG_F2FS_FS_LZO
225 		if (clevel)
226 			goto err_level;
227 #endif
228 		break;
229 	case COMPRESS_LZORLE:
230 #ifdef CONFIG_F2FS_FS_LZORLE
231 		if (clevel)
232 			goto err_level;
233 #endif
234 		break;
235 	case COMPRESS_LZ4:
236 #ifdef CONFIG_F2FS_FS_LZ4
237 #ifdef CONFIG_F2FS_FS_LZ4HC
238 		if (clevel &&
239 		   (clevel < LZ4HC_MIN_CLEVEL || clevel > LZ4HC_MAX_CLEVEL))
240 			goto err_level;
241 #else
242 		if (clevel)
243 			goto err_level;
244 #endif
245 #endif
246 		break;
247 	case COMPRESS_ZSTD:
248 #ifdef CONFIG_F2FS_FS_ZSTD
249 		if (clevel < zstd_min_clevel() || clevel > zstd_max_clevel())
250 			goto err_level;
251 #endif
252 		break;
253 	default:
254 		goto err_level;
255 	}
256 
257 	return true;
258 err_level:
259 	f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported compress level: %u, run fsck to fix",
260 		  __func__, inode->i_ino, clevel);
261 	return false;
262 }
263 
264 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
265 {
266 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
267 	struct f2fs_inode_info *fi = F2FS_I(inode);
268 	struct f2fs_inode *ri = F2FS_INODE(node_page);
269 	unsigned long long iblocks;
270 
271 	iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
272 	if (!iblocks) {
273 		f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
274 			  __func__, inode->i_ino, iblocks);
275 		return false;
276 	}
277 
278 	if (ino_of_node(node_page) != nid_of_node(node_page)) {
279 		f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
280 			  __func__, inode->i_ino,
281 			  ino_of_node(node_page), nid_of_node(node_page));
282 		return false;
283 	}
284 
285 	if (f2fs_has_extra_attr(inode)) {
286 		if (!f2fs_sb_has_extra_attr(sbi)) {
287 			f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
288 				  __func__, inode->i_ino);
289 			return false;
290 		}
291 		if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
292 			fi->i_extra_isize < F2FS_MIN_EXTRA_ATTR_SIZE ||
293 			fi->i_extra_isize % sizeof(__le32)) {
294 			f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
295 				  __func__, inode->i_ino, fi->i_extra_isize,
296 				  F2FS_TOTAL_EXTRA_ATTR_SIZE);
297 			return false;
298 		}
299 		if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
300 			f2fs_has_inline_xattr(inode) &&
301 			(!fi->i_inline_xattr_size ||
302 			fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
303 			f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %zu",
304 				  __func__, inode->i_ino, fi->i_inline_xattr_size,
305 				  MAX_INLINE_XATTR_SIZE);
306 			return false;
307 		}
308 		if (f2fs_sb_has_compression(sbi) &&
309 			fi->i_flags & F2FS_COMPR_FL &&
310 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
311 						i_compress_flag)) {
312 			if (!sanity_check_compress_inode(inode, ri))
313 				return false;
314 		}
315 	} else if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
316 		f2fs_warn(sbi, "%s: corrupted inode ino=%lx, run fsck to fix.",
317 			  __func__, inode->i_ino);
318 		return false;
319 	}
320 
321 	if (!f2fs_sb_has_extra_attr(sbi)) {
322 		if (f2fs_sb_has_project_quota(sbi)) {
323 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
324 				  __func__, inode->i_ino, F2FS_FEATURE_PRJQUOTA);
325 			return false;
326 		}
327 		if (f2fs_sb_has_inode_chksum(sbi)) {
328 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
329 				  __func__, inode->i_ino, F2FS_FEATURE_INODE_CHKSUM);
330 			return false;
331 		}
332 		if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
333 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
334 				  __func__, inode->i_ino, F2FS_FEATURE_FLEXIBLE_INLINE_XATTR);
335 			return false;
336 		}
337 		if (f2fs_sb_has_inode_crtime(sbi)) {
338 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
339 				  __func__, inode->i_ino, F2FS_FEATURE_INODE_CRTIME);
340 			return false;
341 		}
342 		if (f2fs_sb_has_compression(sbi)) {
343 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
344 				  __func__, inode->i_ino, F2FS_FEATURE_COMPRESSION);
345 			return false;
346 		}
347 	}
348 
349 	if (f2fs_sanity_check_inline_data(inode)) {
350 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
351 			  __func__, inode->i_ino, inode->i_mode);
352 		return false;
353 	}
354 
355 	if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
356 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
357 			  __func__, inode->i_ino, inode->i_mode);
358 		return false;
359 	}
360 
361 	if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
362 		f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
363 			  __func__, inode->i_ino);
364 		return false;
365 	}
366 
367 	if (fi->i_xattr_nid && f2fs_check_nid_range(sbi, fi->i_xattr_nid)) {
368 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_xattr_nid: %u, run fsck to fix.",
369 			  __func__, inode->i_ino, fi->i_xattr_nid);
370 		return false;
371 	}
372 
373 	return true;
374 }
375 
376 static void init_idisk_time(struct inode *inode)
377 {
378 	struct f2fs_inode_info *fi = F2FS_I(inode);
379 
380 	fi->i_disk_time[0] = inode->i_atime;
381 	fi->i_disk_time[1] = inode_get_ctime(inode);
382 	fi->i_disk_time[2] = inode->i_mtime;
383 }
384 
385 static int do_read_inode(struct inode *inode)
386 {
387 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
388 	struct f2fs_inode_info *fi = F2FS_I(inode);
389 	struct page *node_page;
390 	struct f2fs_inode *ri;
391 	projid_t i_projid;
392 
393 	/* Check if ino is within scope */
394 	if (f2fs_check_nid_range(sbi, inode->i_ino))
395 		return -EINVAL;
396 
397 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
398 	if (IS_ERR(node_page))
399 		return PTR_ERR(node_page);
400 
401 	ri = F2FS_INODE(node_page);
402 
403 	inode->i_mode = le16_to_cpu(ri->i_mode);
404 	i_uid_write(inode, le32_to_cpu(ri->i_uid));
405 	i_gid_write(inode, le32_to_cpu(ri->i_gid));
406 	set_nlink(inode, le32_to_cpu(ri->i_links));
407 	inode->i_size = le64_to_cpu(ri->i_size);
408 	inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
409 
410 	inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
411 	inode_set_ctime(inode, le64_to_cpu(ri->i_ctime),
412 			le32_to_cpu(ri->i_ctime_nsec));
413 	inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
414 	inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
415 	inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
416 	inode->i_generation = le32_to_cpu(ri->i_generation);
417 	if (S_ISDIR(inode->i_mode))
418 		fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
419 	else if (S_ISREG(inode->i_mode))
420 		fi->i_gc_failures[GC_FAILURE_PIN] =
421 					le16_to_cpu(ri->i_gc_failures);
422 	fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
423 	fi->i_flags = le32_to_cpu(ri->i_flags);
424 	if (S_ISREG(inode->i_mode))
425 		fi->i_flags &= ~F2FS_PROJINHERIT_FL;
426 	bitmap_zero(fi->flags, FI_MAX);
427 	fi->i_advise = ri->i_advise;
428 	fi->i_pino = le32_to_cpu(ri->i_pino);
429 	fi->i_dir_level = ri->i_dir_level;
430 
431 	get_inline_info(inode, ri);
432 
433 	fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
434 					le16_to_cpu(ri->i_extra_isize) : 0;
435 
436 	if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
437 		fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
438 	} else if (f2fs_has_inline_xattr(inode) ||
439 				f2fs_has_inline_dentry(inode)) {
440 		fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
441 	} else {
442 
443 		/*
444 		 * Previous inline data or directory always reserved 200 bytes
445 		 * in inode layout, even if inline_xattr is disabled. In order
446 		 * to keep inline_dentry's structure for backward compatibility,
447 		 * we get the space back only from inline_data.
448 		 */
449 		fi->i_inline_xattr_size = 0;
450 	}
451 
452 	if (!sanity_check_inode(inode, node_page)) {
453 		f2fs_put_page(node_page, 1);
454 		set_sbi_flag(sbi, SBI_NEED_FSCK);
455 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
456 		return -EFSCORRUPTED;
457 	}
458 
459 	/* check data exist */
460 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
461 		__recover_inline_status(inode, node_page);
462 
463 	/* try to recover cold bit for non-dir inode */
464 	if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
465 		f2fs_wait_on_page_writeback(node_page, NODE, true, true);
466 		set_cold_node(node_page, false);
467 		set_page_dirty(node_page);
468 	}
469 
470 	/* get rdev by using inline_info */
471 	__get_inode_rdev(inode, node_page);
472 
473 	if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
474 		fi->last_disk_size = inode->i_size;
475 
476 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
477 		set_inode_flag(inode, FI_PROJ_INHERIT);
478 
479 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
480 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
481 		i_projid = (projid_t)le32_to_cpu(ri->i_projid);
482 	else
483 		i_projid = F2FS_DEF_PROJID;
484 	fi->i_projid = make_kprojid(&init_user_ns, i_projid);
485 
486 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
487 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
488 		fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
489 		fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
490 	}
491 
492 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
493 					(fi->i_flags & F2FS_COMPR_FL)) {
494 		if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
495 					i_compress_flag)) {
496 			unsigned short compress_flag;
497 
498 			atomic_set(&fi->i_compr_blocks,
499 					le64_to_cpu(ri->i_compr_blocks));
500 			fi->i_compress_algorithm = ri->i_compress_algorithm;
501 			fi->i_log_cluster_size = ri->i_log_cluster_size;
502 			compress_flag = le16_to_cpu(ri->i_compress_flag);
503 			fi->i_compress_level = compress_flag >>
504 						COMPRESS_LEVEL_OFFSET;
505 			fi->i_compress_flag = compress_flag &
506 					GENMASK(COMPRESS_LEVEL_OFFSET - 1, 0);
507 			fi->i_cluster_size = BIT(fi->i_log_cluster_size);
508 			set_inode_flag(inode, FI_COMPRESSED_FILE);
509 		}
510 	}
511 
512 	init_idisk_time(inode);
513 
514 	/* Need all the flag bits */
515 	f2fs_init_read_extent_tree(inode, node_page);
516 	f2fs_init_age_extent_tree(inode);
517 
518 	if (!sanity_check_extent_cache(inode)) {
519 		f2fs_put_page(node_page, 1);
520 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
521 		return -EFSCORRUPTED;
522 	}
523 
524 	f2fs_put_page(node_page, 1);
525 
526 	stat_inc_inline_xattr(inode);
527 	stat_inc_inline_inode(inode);
528 	stat_inc_inline_dir(inode);
529 	stat_inc_compr_inode(inode);
530 	stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
531 
532 	return 0;
533 }
534 
535 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
536 {
537 	return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
538 		ino == F2FS_COMPRESS_INO(sbi);
539 }
540 
541 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
542 {
543 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
544 	struct inode *inode;
545 	int ret = 0;
546 
547 	inode = iget_locked(sb, ino);
548 	if (!inode)
549 		return ERR_PTR(-ENOMEM);
550 
551 	if (!(inode->i_state & I_NEW)) {
552 		if (is_meta_ino(sbi, ino)) {
553 			f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
554 			set_sbi_flag(sbi, SBI_NEED_FSCK);
555 			ret = -EFSCORRUPTED;
556 			trace_f2fs_iget_exit(inode, ret);
557 			iput(inode);
558 			f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
559 			return ERR_PTR(ret);
560 		}
561 
562 		trace_f2fs_iget(inode);
563 		return inode;
564 	}
565 
566 	if (is_meta_ino(sbi, ino))
567 		goto make_now;
568 
569 	ret = do_read_inode(inode);
570 	if (ret)
571 		goto bad_inode;
572 make_now:
573 	if (ino == F2FS_NODE_INO(sbi)) {
574 		inode->i_mapping->a_ops = &f2fs_node_aops;
575 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
576 	} else if (ino == F2FS_META_INO(sbi)) {
577 		inode->i_mapping->a_ops = &f2fs_meta_aops;
578 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
579 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
580 #ifdef CONFIG_F2FS_FS_COMPRESSION
581 		inode->i_mapping->a_ops = &f2fs_compress_aops;
582 		/*
583 		 * generic_error_remove_page only truncates pages of regular
584 		 * inode
585 		 */
586 		inode->i_mode |= S_IFREG;
587 #endif
588 		mapping_set_gfp_mask(inode->i_mapping,
589 			GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
590 	} else if (S_ISREG(inode->i_mode)) {
591 		inode->i_op = &f2fs_file_inode_operations;
592 		inode->i_fop = &f2fs_file_operations;
593 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
594 	} else if (S_ISDIR(inode->i_mode)) {
595 		inode->i_op = &f2fs_dir_inode_operations;
596 		inode->i_fop = &f2fs_dir_operations;
597 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
598 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
599 	} else if (S_ISLNK(inode->i_mode)) {
600 		if (file_is_encrypt(inode))
601 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
602 		else
603 			inode->i_op = &f2fs_symlink_inode_operations;
604 		inode_nohighmem(inode);
605 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
606 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
607 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
608 		inode->i_op = &f2fs_special_inode_operations;
609 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
610 	} else {
611 		ret = -EIO;
612 		goto bad_inode;
613 	}
614 	f2fs_set_inode_flags(inode);
615 
616 	unlock_new_inode(inode);
617 	trace_f2fs_iget(inode);
618 	return inode;
619 
620 bad_inode:
621 	f2fs_inode_synced(inode);
622 	iget_failed(inode);
623 	trace_f2fs_iget_exit(inode, ret);
624 	return ERR_PTR(ret);
625 }
626 
627 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
628 {
629 	struct inode *inode;
630 retry:
631 	inode = f2fs_iget(sb, ino);
632 	if (IS_ERR(inode)) {
633 		if (PTR_ERR(inode) == -ENOMEM) {
634 			memalloc_retry_wait(GFP_NOFS);
635 			goto retry;
636 		}
637 	}
638 	return inode;
639 }
640 
641 void f2fs_update_inode(struct inode *inode, struct page *node_page)
642 {
643 	struct f2fs_inode *ri;
644 	struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
645 
646 	f2fs_wait_on_page_writeback(node_page, NODE, true, true);
647 	set_page_dirty(node_page);
648 
649 	f2fs_inode_synced(inode);
650 
651 	ri = F2FS_INODE(node_page);
652 
653 	ri->i_mode = cpu_to_le16(inode->i_mode);
654 	ri->i_advise = F2FS_I(inode)->i_advise;
655 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
656 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
657 	ri->i_links = cpu_to_le32(inode->i_nlink);
658 	ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
659 
660 	if (!f2fs_is_atomic_file(inode) ||
661 			is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
662 		ri->i_size = cpu_to_le64(i_size_read(inode));
663 
664 	if (et) {
665 		read_lock(&et->lock);
666 		set_raw_read_extent(&et->largest, &ri->i_ext);
667 		read_unlock(&et->lock);
668 	} else {
669 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
670 	}
671 	set_raw_inline(inode, ri);
672 
673 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
674 	ri->i_ctime = cpu_to_le64(inode_get_ctime(inode).tv_sec);
675 	ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
676 	ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
677 	ri->i_ctime_nsec = cpu_to_le32(inode_get_ctime(inode).tv_nsec);
678 	ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
679 	if (S_ISDIR(inode->i_mode))
680 		ri->i_current_depth =
681 			cpu_to_le32(F2FS_I(inode)->i_current_depth);
682 	else if (S_ISREG(inode->i_mode))
683 		ri->i_gc_failures =
684 			cpu_to_le16(F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]);
685 	ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
686 	ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
687 	ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
688 	ri->i_generation = cpu_to_le32(inode->i_generation);
689 	ri->i_dir_level = F2FS_I(inode)->i_dir_level;
690 
691 	if (f2fs_has_extra_attr(inode)) {
692 		ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
693 
694 		if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
695 			ri->i_inline_xattr_size =
696 				cpu_to_le16(F2FS_I(inode)->i_inline_xattr_size);
697 
698 		if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
699 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
700 								i_projid)) {
701 			projid_t i_projid;
702 
703 			i_projid = from_kprojid(&init_user_ns,
704 						F2FS_I(inode)->i_projid);
705 			ri->i_projid = cpu_to_le32(i_projid);
706 		}
707 
708 		if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
709 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
710 								i_crtime)) {
711 			ri->i_crtime =
712 				cpu_to_le64(F2FS_I(inode)->i_crtime.tv_sec);
713 			ri->i_crtime_nsec =
714 				cpu_to_le32(F2FS_I(inode)->i_crtime.tv_nsec);
715 		}
716 
717 		if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
718 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
719 							i_compress_flag)) {
720 			unsigned short compress_flag;
721 
722 			ri->i_compr_blocks =
723 				cpu_to_le64(atomic_read(
724 					&F2FS_I(inode)->i_compr_blocks));
725 			ri->i_compress_algorithm =
726 				F2FS_I(inode)->i_compress_algorithm;
727 			compress_flag = F2FS_I(inode)->i_compress_flag |
728 				F2FS_I(inode)->i_compress_level <<
729 						COMPRESS_LEVEL_OFFSET;
730 			ri->i_compress_flag = cpu_to_le16(compress_flag);
731 			ri->i_log_cluster_size =
732 				F2FS_I(inode)->i_log_cluster_size;
733 		}
734 	}
735 
736 	__set_inode_rdev(inode, node_page);
737 
738 	/* deleted inode */
739 	if (inode->i_nlink == 0)
740 		clear_page_private_inline(node_page);
741 
742 	init_idisk_time(inode);
743 #ifdef CONFIG_F2FS_CHECK_FS
744 	f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
745 #endif
746 }
747 
748 void f2fs_update_inode_page(struct inode *inode)
749 {
750 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
751 	struct page *node_page;
752 	int count = 0;
753 retry:
754 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
755 	if (IS_ERR(node_page)) {
756 		int err = PTR_ERR(node_page);
757 
758 		/* The node block was truncated. */
759 		if (err == -ENOENT)
760 			return;
761 
762 		if (err == -ENOMEM || ++count <= DEFAULT_RETRY_IO_COUNT)
763 			goto retry;
764 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_UPDATE_INODE);
765 		return;
766 	}
767 	f2fs_update_inode(inode, node_page);
768 	f2fs_put_page(node_page, 1);
769 }
770 
771 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
772 {
773 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
774 
775 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
776 			inode->i_ino == F2FS_META_INO(sbi))
777 		return 0;
778 
779 	/*
780 	 * atime could be updated without dirtying f2fs inode in lazytime mode
781 	 */
782 	if (f2fs_is_time_consistent(inode) &&
783 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
784 		return 0;
785 
786 	if (!f2fs_is_checkpoint_ready(sbi))
787 		return -ENOSPC;
788 
789 	/*
790 	 * We need to balance fs here to prevent from producing dirty node pages
791 	 * during the urgent cleaning time when running out of free sections.
792 	 */
793 	f2fs_update_inode_page(inode);
794 	if (wbc && wbc->nr_to_write)
795 		f2fs_balance_fs(sbi, true);
796 	return 0;
797 }
798 
799 /*
800  * Called at the last iput() if i_nlink is zero
801  */
802 void f2fs_evict_inode(struct inode *inode)
803 {
804 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
805 	struct f2fs_inode_info *fi = F2FS_I(inode);
806 	nid_t xnid = fi->i_xattr_nid;
807 	int err = 0;
808 
809 	f2fs_abort_atomic_write(inode, true);
810 
811 	if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
812 		clear_inode_flag(fi->cow_inode, FI_COW_FILE);
813 		F2FS_I(fi->cow_inode)->atomic_inode = NULL;
814 		iput(fi->cow_inode);
815 		fi->cow_inode = NULL;
816 	}
817 
818 	trace_f2fs_evict_inode(inode);
819 	truncate_inode_pages_final(&inode->i_data);
820 
821 	if ((inode->i_nlink || is_bad_inode(inode)) &&
822 		test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
823 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
824 
825 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
826 			inode->i_ino == F2FS_META_INO(sbi) ||
827 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
828 		goto out_clear;
829 
830 	f2fs_bug_on(sbi, get_dirty_pages(inode));
831 	f2fs_remove_dirty_inode(inode);
832 
833 	f2fs_destroy_extent_tree(inode);
834 
835 	if (inode->i_nlink || is_bad_inode(inode))
836 		goto no_delete;
837 
838 	err = f2fs_dquot_initialize(inode);
839 	if (err) {
840 		err = 0;
841 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
842 	}
843 
844 	f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
845 	f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
846 	f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
847 
848 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
849 		sb_start_intwrite(inode->i_sb);
850 	set_inode_flag(inode, FI_NO_ALLOC);
851 	i_size_write(inode, 0);
852 retry:
853 	if (F2FS_HAS_BLOCKS(inode))
854 		err = f2fs_truncate(inode);
855 
856 	if (time_to_inject(sbi, FAULT_EVICT_INODE))
857 		err = -EIO;
858 
859 	if (!err) {
860 		f2fs_lock_op(sbi);
861 		err = f2fs_remove_inode_page(inode);
862 		f2fs_unlock_op(sbi);
863 		if (err == -ENOENT) {
864 			err = 0;
865 
866 			/*
867 			 * in fuzzed image, another node may has the same
868 			 * block address as inode's, if it was truncated
869 			 * previously, truncation of inode node will fail.
870 			 */
871 			if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
872 				f2fs_warn(F2FS_I_SB(inode),
873 					"f2fs_evict_inode: inconsistent node id, ino:%lu",
874 					inode->i_ino);
875 				f2fs_inode_synced(inode);
876 				set_sbi_flag(sbi, SBI_NEED_FSCK);
877 			}
878 		}
879 	}
880 
881 	/* give more chances, if ENOMEM case */
882 	if (err == -ENOMEM) {
883 		err = 0;
884 		goto retry;
885 	}
886 
887 	if (err) {
888 		f2fs_update_inode_page(inode);
889 		if (dquot_initialize_needed(inode))
890 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
891 	}
892 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
893 		sb_end_intwrite(inode->i_sb);
894 no_delete:
895 	dquot_drop(inode);
896 
897 	stat_dec_inline_xattr(inode);
898 	stat_dec_inline_dir(inode);
899 	stat_dec_inline_inode(inode);
900 	stat_dec_compr_inode(inode);
901 	stat_sub_compr_blocks(inode,
902 			atomic_read(&fi->i_compr_blocks));
903 
904 	if (likely(!f2fs_cp_error(sbi) &&
905 				!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
906 		f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
907 	else
908 		f2fs_inode_synced(inode);
909 
910 	/* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
911 	if (inode->i_ino)
912 		invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
913 							inode->i_ino);
914 	if (xnid)
915 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
916 	if (inode->i_nlink) {
917 		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
918 			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
919 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
920 			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
921 	}
922 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
923 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
924 		clear_inode_flag(inode, FI_FREE_NID);
925 	} else {
926 		/*
927 		 * If xattr nid is corrupted, we can reach out error condition,
928 		 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
929 		 * In that case, f2fs_check_nid_range() is enough to give a clue.
930 		 */
931 	}
932 out_clear:
933 	fscrypt_put_encryption_info(inode);
934 	fsverity_cleanup_inode(inode);
935 	clear_inode(inode);
936 }
937 
938 /* caller should call f2fs_lock_op() */
939 void f2fs_handle_failed_inode(struct inode *inode)
940 {
941 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
942 	struct node_info ni;
943 	int err;
944 
945 	/*
946 	 * clear nlink of inode in order to release resource of inode
947 	 * immediately.
948 	 */
949 	clear_nlink(inode);
950 
951 	/*
952 	 * we must call this to avoid inode being remained as dirty, resulting
953 	 * in a panic when flushing dirty inodes in gdirty_list.
954 	 */
955 	f2fs_update_inode_page(inode);
956 	f2fs_inode_synced(inode);
957 
958 	/* don't make bad inode, since it becomes a regular file. */
959 	unlock_new_inode(inode);
960 
961 	/*
962 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
963 	 * so we can prevent losing this orphan when encoutering checkpoint
964 	 * and following suddenly power-off.
965 	 */
966 	err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
967 	if (err) {
968 		set_sbi_flag(sbi, SBI_NEED_FSCK);
969 		set_inode_flag(inode, FI_FREE_NID);
970 		f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
971 		goto out;
972 	}
973 
974 	if (ni.blk_addr != NULL_ADDR) {
975 		err = f2fs_acquire_orphan_inode(sbi);
976 		if (err) {
977 			set_sbi_flag(sbi, SBI_NEED_FSCK);
978 			f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
979 		} else {
980 			f2fs_add_orphan_inode(inode);
981 		}
982 		f2fs_alloc_nid_done(sbi, inode->i_ino);
983 	} else {
984 		set_inode_flag(inode, FI_FREE_NID);
985 	}
986 
987 out:
988 	f2fs_unlock_op(sbi);
989 
990 	/* iput will drop the inode object */
991 	iput(inode);
992 }
993