xref: /openbmc/linux/fs/f2fs/inode.c (revision af9b2ff010f593d81e2f5fb04155e9fc25b9dfd0)
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 
f2fs_mark_inode_dirty_sync(struct inode * inode,bool sync)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 	/* only atomic file w/ FI_ATOMIC_COMMITTED can be set vfs dirty */
39 	if (f2fs_is_atomic_file(inode) &&
40 			!is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
41 		return;
42 
43 	mark_inode_dirty_sync(inode);
44 }
45 
f2fs_set_inode_flags(struct inode * inode)46 void f2fs_set_inode_flags(struct inode *inode)
47 {
48 	unsigned int flags = F2FS_I(inode)->i_flags;
49 	unsigned int new_fl = 0;
50 
51 	if (flags & F2FS_SYNC_FL)
52 		new_fl |= S_SYNC;
53 	if (flags & F2FS_APPEND_FL)
54 		new_fl |= S_APPEND;
55 	if (flags & F2FS_IMMUTABLE_FL)
56 		new_fl |= S_IMMUTABLE;
57 	if (flags & F2FS_NOATIME_FL)
58 		new_fl |= S_NOATIME;
59 	if (flags & F2FS_DIRSYNC_FL)
60 		new_fl |= S_DIRSYNC;
61 	if (file_is_encrypt(inode))
62 		new_fl |= S_ENCRYPTED;
63 	if (file_is_verity(inode))
64 		new_fl |= S_VERITY;
65 	if (flags & F2FS_CASEFOLD_FL)
66 		new_fl |= S_CASEFOLD;
67 	inode_set_flags(inode, new_fl,
68 			S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|
69 			S_ENCRYPTED|S_VERITY|S_CASEFOLD);
70 }
71 
__get_inode_rdev(struct inode * inode,struct page * node_page)72 static void __get_inode_rdev(struct inode *inode, struct page *node_page)
73 {
74 	__le32 *addr = get_dnode_addr(inode, node_page);
75 
76 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
77 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
78 		if (addr[0])
79 			inode->i_rdev = old_decode_dev(le32_to_cpu(addr[0]));
80 		else
81 			inode->i_rdev = new_decode_dev(le32_to_cpu(addr[1]));
82 	}
83 }
84 
__set_inode_rdev(struct inode * inode,struct page * node_page)85 static void __set_inode_rdev(struct inode *inode, struct page *node_page)
86 {
87 	__le32 *addr = get_dnode_addr(inode, node_page);
88 
89 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
90 		if (old_valid_dev(inode->i_rdev)) {
91 			addr[0] = cpu_to_le32(old_encode_dev(inode->i_rdev));
92 			addr[1] = 0;
93 		} else {
94 			addr[0] = 0;
95 			addr[1] = cpu_to_le32(new_encode_dev(inode->i_rdev));
96 			addr[2] = 0;
97 		}
98 	}
99 }
100 
__recover_inline_status(struct inode * inode,struct page * ipage)101 static void __recover_inline_status(struct inode *inode, struct page *ipage)
102 {
103 	void *inline_data = inline_data_addr(inode, ipage);
104 	__le32 *start = inline_data;
105 	__le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
106 
107 	while (start < end) {
108 		if (*start++) {
109 			f2fs_wait_on_page_writeback(ipage, NODE, true, true);
110 
111 			set_inode_flag(inode, FI_DATA_EXIST);
112 			set_raw_inline(inode, F2FS_INODE(ipage));
113 			set_page_dirty(ipage);
114 			return;
115 		}
116 	}
117 	return;
118 }
119 
f2fs_enable_inode_chksum(struct f2fs_sb_info * sbi,struct page * page)120 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
121 {
122 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
123 
124 	if (!f2fs_sb_has_inode_chksum(sbi))
125 		return false;
126 
127 	if (!IS_INODE(page) || !(ri->i_inline & F2FS_EXTRA_ATTR))
128 		return false;
129 
130 	if (!F2FS_FITS_IN_INODE(ri, le16_to_cpu(ri->i_extra_isize),
131 				i_inode_checksum))
132 		return false;
133 
134 	return true;
135 }
136 
f2fs_inode_chksum(struct f2fs_sb_info * sbi,struct page * page)137 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
138 {
139 	struct f2fs_node *node = F2FS_NODE(page);
140 	struct f2fs_inode *ri = &node->i;
141 	__le32 ino = node->footer.ino;
142 	__le32 gen = ri->i_generation;
143 	__u32 chksum, chksum_seed;
144 	__u32 dummy_cs = 0;
145 	unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
146 	unsigned int cs_size = sizeof(dummy_cs);
147 
148 	chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
149 							sizeof(ino));
150 	chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
151 
152 	chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
153 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
154 	offset += cs_size;
155 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
156 						F2FS_BLKSIZE - offset);
157 	return chksum;
158 }
159 
f2fs_inode_chksum_verify(struct f2fs_sb_info * sbi,struct page * page)160 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
161 {
162 	struct f2fs_inode *ri;
163 	__u32 provided, calculated;
164 
165 	if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
166 		return true;
167 
168 #ifdef CONFIG_F2FS_CHECK_FS
169 	if (!f2fs_enable_inode_chksum(sbi, page))
170 #else
171 	if (!f2fs_enable_inode_chksum(sbi, page) ||
172 			PageDirty(page) || PageWriteback(page))
173 #endif
174 		return true;
175 
176 	ri = &F2FS_NODE(page)->i;
177 	provided = le32_to_cpu(ri->i_inode_checksum);
178 	calculated = f2fs_inode_chksum(sbi, page);
179 
180 	if (provided != calculated)
181 		f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
182 			  page->index, ino_of_node(page), provided, calculated);
183 
184 	return provided == calculated;
185 }
186 
f2fs_inode_chksum_set(struct f2fs_sb_info * sbi,struct page * page)187 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
188 {
189 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
190 
191 	if (!f2fs_enable_inode_chksum(sbi, page))
192 		return;
193 
194 	ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
195 }
196 
sanity_check_compress_inode(struct inode * inode,struct f2fs_inode * ri)197 static bool sanity_check_compress_inode(struct inode *inode,
198 			struct f2fs_inode *ri)
199 {
200 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
201 	unsigned char clevel;
202 
203 	if (ri->i_compress_algorithm >= COMPRESS_MAX) {
204 		f2fs_warn(sbi,
205 			"%s: inode (ino=%lx) has unsupported compress algorithm: %u, run fsck to fix",
206 			__func__, inode->i_ino, ri->i_compress_algorithm);
207 		return false;
208 	}
209 	if (le64_to_cpu(ri->i_compr_blocks) >
210 			SECTOR_TO_BLOCK(inode->i_blocks)) {
211 		f2fs_warn(sbi,
212 			"%s: inode (ino=%lx) has inconsistent i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
213 			__func__, inode->i_ino, le64_to_cpu(ri->i_compr_blocks),
214 			SECTOR_TO_BLOCK(inode->i_blocks));
215 		return false;
216 	}
217 	if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
218 		ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
219 		f2fs_warn(sbi,
220 			"%s: inode (ino=%lx) has unsupported log cluster size: %u, run fsck to fix",
221 			__func__, inode->i_ino, ri->i_log_cluster_size);
222 		return false;
223 	}
224 
225 	clevel = le16_to_cpu(ri->i_compress_flag) >>
226 				COMPRESS_LEVEL_OFFSET;
227 	switch (ri->i_compress_algorithm) {
228 	case COMPRESS_LZO:
229 #ifdef CONFIG_F2FS_FS_LZO
230 		if (clevel)
231 			goto err_level;
232 #endif
233 		break;
234 	case COMPRESS_LZORLE:
235 #ifdef CONFIG_F2FS_FS_LZORLE
236 		if (clevel)
237 			goto err_level;
238 #endif
239 		break;
240 	case COMPRESS_LZ4:
241 #ifdef CONFIG_F2FS_FS_LZ4
242 #ifdef CONFIG_F2FS_FS_LZ4HC
243 		if (clevel &&
244 		   (clevel < LZ4HC_MIN_CLEVEL || clevel > LZ4HC_MAX_CLEVEL))
245 			goto err_level;
246 #else
247 		if (clevel)
248 			goto err_level;
249 #endif
250 #endif
251 		break;
252 	case COMPRESS_ZSTD:
253 #ifdef CONFIG_F2FS_FS_ZSTD
254 		if (clevel < zstd_min_clevel() || clevel > zstd_max_clevel())
255 			goto err_level;
256 #endif
257 		break;
258 	default:
259 		goto err_level;
260 	}
261 
262 	return true;
263 err_level:
264 	f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported compress level: %u, run fsck to fix",
265 		  __func__, inode->i_ino, clevel);
266 	return false;
267 }
268 
sanity_check_inode(struct inode * inode,struct page * node_page)269 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
270 {
271 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
272 	struct f2fs_inode_info *fi = F2FS_I(inode);
273 	struct f2fs_inode *ri = F2FS_INODE(node_page);
274 	unsigned long long iblocks;
275 
276 	iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
277 	if (!iblocks) {
278 		f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
279 			  __func__, inode->i_ino, iblocks);
280 		return false;
281 	}
282 
283 	if (ino_of_node(node_page) != nid_of_node(node_page)) {
284 		f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
285 			  __func__, inode->i_ino,
286 			  ino_of_node(node_page), nid_of_node(node_page));
287 		return false;
288 	}
289 
290 	if (ino_of_node(node_page) == fi->i_xattr_nid) {
291 		f2fs_warn(sbi, "%s: corrupted inode i_ino=%lx, xnid=%x, run fsck to fix.",
292 			  __func__, inode->i_ino, fi->i_xattr_nid);
293 		return false;
294 	}
295 
296 	if (f2fs_has_extra_attr(inode)) {
297 		if (!f2fs_sb_has_extra_attr(sbi)) {
298 			f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
299 				  __func__, inode->i_ino);
300 			return false;
301 		}
302 		if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
303 			fi->i_extra_isize < F2FS_MIN_EXTRA_ATTR_SIZE ||
304 			fi->i_extra_isize % sizeof(__le32)) {
305 			f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
306 				  __func__, inode->i_ino, fi->i_extra_isize,
307 				  F2FS_TOTAL_EXTRA_ATTR_SIZE);
308 			return false;
309 		}
310 		if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
311 			f2fs_has_inline_xattr(inode) &&
312 			(!fi->i_inline_xattr_size ||
313 			fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
314 			f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %zu",
315 				  __func__, inode->i_ino, fi->i_inline_xattr_size,
316 				  MAX_INLINE_XATTR_SIZE);
317 			return false;
318 		}
319 		if (f2fs_sb_has_compression(sbi) &&
320 			fi->i_flags & F2FS_COMPR_FL &&
321 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
322 						i_compress_flag)) {
323 			if (!sanity_check_compress_inode(inode, ri))
324 				return false;
325 		}
326 	} else if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
327 		f2fs_warn(sbi, "%s: corrupted inode ino=%lx, run fsck to fix.",
328 			  __func__, inode->i_ino);
329 		return false;
330 	}
331 
332 	if (!f2fs_sb_has_extra_attr(sbi)) {
333 		if (f2fs_sb_has_project_quota(sbi)) {
334 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
335 				  __func__, inode->i_ino, F2FS_FEATURE_PRJQUOTA);
336 			return false;
337 		}
338 		if (f2fs_sb_has_inode_chksum(sbi)) {
339 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
340 				  __func__, inode->i_ino, F2FS_FEATURE_INODE_CHKSUM);
341 			return false;
342 		}
343 		if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
344 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
345 				  __func__, inode->i_ino, F2FS_FEATURE_FLEXIBLE_INLINE_XATTR);
346 			return false;
347 		}
348 		if (f2fs_sb_has_inode_crtime(sbi)) {
349 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
350 				  __func__, inode->i_ino, F2FS_FEATURE_INODE_CRTIME);
351 			return false;
352 		}
353 		if (f2fs_sb_has_compression(sbi)) {
354 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
355 				  __func__, inode->i_ino, F2FS_FEATURE_COMPRESSION);
356 			return false;
357 		}
358 	}
359 
360 	if (f2fs_sanity_check_inline_data(inode, node_page)) {
361 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
362 			  __func__, inode->i_ino, inode->i_mode);
363 		return false;
364 	}
365 
366 	if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
367 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
368 			  __func__, inode->i_ino, inode->i_mode);
369 		return false;
370 	}
371 
372 	if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
373 		f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
374 			  __func__, inode->i_ino);
375 		return false;
376 	}
377 
378 	if (fi->i_xattr_nid && f2fs_check_nid_range(sbi, fi->i_xattr_nid)) {
379 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_xattr_nid: %u, run fsck to fix.",
380 			  __func__, inode->i_ino, fi->i_xattr_nid);
381 		return false;
382 	}
383 
384 	return true;
385 }
386 
init_idisk_time(struct inode * inode)387 static void init_idisk_time(struct inode *inode)
388 {
389 	struct f2fs_inode_info *fi = F2FS_I(inode);
390 
391 	fi->i_disk_time[0] = inode->i_atime;
392 	fi->i_disk_time[1] = inode_get_ctime(inode);
393 	fi->i_disk_time[2] = inode->i_mtime;
394 }
395 
do_read_inode(struct inode * inode)396 static int do_read_inode(struct inode *inode)
397 {
398 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
399 	struct f2fs_inode_info *fi = F2FS_I(inode);
400 	struct page *node_page;
401 	struct f2fs_inode *ri;
402 	projid_t i_projid;
403 
404 	/* Check if ino is within scope */
405 	if (f2fs_check_nid_range(sbi, inode->i_ino))
406 		return -EINVAL;
407 
408 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
409 	if (IS_ERR(node_page))
410 		return PTR_ERR(node_page);
411 
412 	ri = F2FS_INODE(node_page);
413 
414 	inode->i_mode = le16_to_cpu(ri->i_mode);
415 	i_uid_write(inode, le32_to_cpu(ri->i_uid));
416 	i_gid_write(inode, le32_to_cpu(ri->i_gid));
417 	set_nlink(inode, le32_to_cpu(ri->i_links));
418 	inode->i_size = le64_to_cpu(ri->i_size);
419 	inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
420 
421 	inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
422 	inode_set_ctime(inode, le64_to_cpu(ri->i_ctime),
423 			le32_to_cpu(ri->i_ctime_nsec));
424 	inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
425 	inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
426 	inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
427 	inode->i_generation = le32_to_cpu(ri->i_generation);
428 	if (S_ISDIR(inode->i_mode))
429 		fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
430 	else if (S_ISREG(inode->i_mode))
431 		fi->i_gc_failures[GC_FAILURE_PIN] =
432 					le16_to_cpu(ri->i_gc_failures);
433 	fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
434 	fi->i_flags = le32_to_cpu(ri->i_flags);
435 	if (S_ISREG(inode->i_mode))
436 		fi->i_flags &= ~F2FS_PROJINHERIT_FL;
437 	bitmap_zero(fi->flags, FI_MAX);
438 	fi->i_advise = ri->i_advise;
439 	fi->i_pino = le32_to_cpu(ri->i_pino);
440 	fi->i_dir_level = ri->i_dir_level;
441 
442 	get_inline_info(inode, ri);
443 
444 	fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
445 					le16_to_cpu(ri->i_extra_isize) : 0;
446 
447 	if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
448 		fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
449 	} else if (f2fs_has_inline_xattr(inode) ||
450 				f2fs_has_inline_dentry(inode)) {
451 		fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
452 	} else {
453 
454 		/*
455 		 * Previous inline data or directory always reserved 200 bytes
456 		 * in inode layout, even if inline_xattr is disabled. In order
457 		 * to keep inline_dentry's structure for backward compatibility,
458 		 * we get the space back only from inline_data.
459 		 */
460 		fi->i_inline_xattr_size = 0;
461 	}
462 
463 	if (!sanity_check_inode(inode, node_page)) {
464 		f2fs_put_page(node_page, 1);
465 		set_sbi_flag(sbi, SBI_NEED_FSCK);
466 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
467 		return -EFSCORRUPTED;
468 	}
469 
470 	/* check data exist */
471 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
472 		__recover_inline_status(inode, node_page);
473 
474 	/* try to recover cold bit for non-dir inode */
475 	if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
476 		f2fs_wait_on_page_writeback(node_page, NODE, true, true);
477 		set_cold_node(node_page, false);
478 		set_page_dirty(node_page);
479 	}
480 
481 	/* get rdev by using inline_info */
482 	__get_inode_rdev(inode, node_page);
483 
484 	if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
485 		fi->last_disk_size = inode->i_size;
486 
487 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
488 		set_inode_flag(inode, FI_PROJ_INHERIT);
489 
490 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
491 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
492 		i_projid = (projid_t)le32_to_cpu(ri->i_projid);
493 	else
494 		i_projid = F2FS_DEF_PROJID;
495 	fi->i_projid = make_kprojid(&init_user_ns, i_projid);
496 
497 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
498 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
499 		fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
500 		fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
501 	}
502 
503 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
504 					(fi->i_flags & F2FS_COMPR_FL)) {
505 		if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
506 					i_compress_flag)) {
507 			unsigned short compress_flag;
508 
509 			atomic_set(&fi->i_compr_blocks,
510 					le64_to_cpu(ri->i_compr_blocks));
511 			fi->i_compress_algorithm = ri->i_compress_algorithm;
512 			fi->i_log_cluster_size = ri->i_log_cluster_size;
513 			compress_flag = le16_to_cpu(ri->i_compress_flag);
514 			fi->i_compress_level = compress_flag >>
515 						COMPRESS_LEVEL_OFFSET;
516 			fi->i_compress_flag = compress_flag &
517 					GENMASK(COMPRESS_LEVEL_OFFSET - 1, 0);
518 			fi->i_cluster_size = BIT(fi->i_log_cluster_size);
519 			set_inode_flag(inode, FI_COMPRESSED_FILE);
520 		}
521 	}
522 
523 	init_idisk_time(inode);
524 
525 	if (!sanity_check_extent_cache(inode, node_page)) {
526 		f2fs_put_page(node_page, 1);
527 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
528 		return -EFSCORRUPTED;
529 	}
530 
531 	/* Need all the flag bits */
532 	f2fs_init_read_extent_tree(inode, node_page);
533 	f2fs_init_age_extent_tree(inode);
534 
535 	f2fs_put_page(node_page, 1);
536 
537 	stat_inc_inline_xattr(inode);
538 	stat_inc_inline_inode(inode);
539 	stat_inc_inline_dir(inode);
540 	stat_inc_compr_inode(inode);
541 	stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
542 
543 	return 0;
544 }
545 
is_meta_ino(struct f2fs_sb_info * sbi,unsigned int ino)546 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
547 {
548 	return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
549 		ino == F2FS_COMPRESS_INO(sbi);
550 }
551 
f2fs_iget(struct super_block * sb,unsigned long ino)552 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
553 {
554 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
555 	struct inode *inode;
556 	int ret = 0;
557 
558 	inode = iget_locked(sb, ino);
559 	if (!inode)
560 		return ERR_PTR(-ENOMEM);
561 
562 	if (!(inode->i_state & I_NEW)) {
563 		if (is_meta_ino(sbi, ino)) {
564 			f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
565 			set_sbi_flag(sbi, SBI_NEED_FSCK);
566 			ret = -EFSCORRUPTED;
567 			trace_f2fs_iget_exit(inode, ret);
568 			iput(inode);
569 			f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
570 			return ERR_PTR(ret);
571 		}
572 
573 		trace_f2fs_iget(inode);
574 		return inode;
575 	}
576 
577 	if (is_meta_ino(sbi, ino))
578 		goto make_now;
579 
580 	ret = do_read_inode(inode);
581 	if (ret)
582 		goto bad_inode;
583 make_now:
584 	if (ino == F2FS_NODE_INO(sbi)) {
585 		inode->i_mapping->a_ops = &f2fs_node_aops;
586 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
587 	} else if (ino == F2FS_META_INO(sbi)) {
588 		inode->i_mapping->a_ops = &f2fs_meta_aops;
589 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
590 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
591 #ifdef CONFIG_F2FS_FS_COMPRESSION
592 		inode->i_mapping->a_ops = &f2fs_compress_aops;
593 		/*
594 		 * generic_error_remove_page only truncates pages of regular
595 		 * inode
596 		 */
597 		inode->i_mode |= S_IFREG;
598 #endif
599 		mapping_set_gfp_mask(inode->i_mapping,
600 			GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
601 	} else if (S_ISREG(inode->i_mode)) {
602 		inode->i_op = &f2fs_file_inode_operations;
603 		inode->i_fop = &f2fs_file_operations;
604 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
605 	} else if (S_ISDIR(inode->i_mode)) {
606 		inode->i_op = &f2fs_dir_inode_operations;
607 		inode->i_fop = &f2fs_dir_operations;
608 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
609 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
610 	} else if (S_ISLNK(inode->i_mode)) {
611 		if (file_is_encrypt(inode))
612 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
613 		else
614 			inode->i_op = &f2fs_symlink_inode_operations;
615 		inode_nohighmem(inode);
616 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
617 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
618 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
619 		inode->i_op = &f2fs_special_inode_operations;
620 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
621 	} else {
622 		ret = -EIO;
623 		goto bad_inode;
624 	}
625 	f2fs_set_inode_flags(inode);
626 
627 	unlock_new_inode(inode);
628 	trace_f2fs_iget(inode);
629 	return inode;
630 
631 bad_inode:
632 	f2fs_inode_synced(inode);
633 	iget_failed(inode);
634 	trace_f2fs_iget_exit(inode, ret);
635 	return ERR_PTR(ret);
636 }
637 
f2fs_iget_retry(struct super_block * sb,unsigned long ino)638 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
639 {
640 	struct inode *inode;
641 retry:
642 	inode = f2fs_iget(sb, ino);
643 	if (IS_ERR(inode)) {
644 		if (PTR_ERR(inode) == -ENOMEM) {
645 			memalloc_retry_wait(GFP_NOFS);
646 			goto retry;
647 		}
648 	}
649 	return inode;
650 }
651 
f2fs_update_inode(struct inode * inode,struct page * node_page)652 void f2fs_update_inode(struct inode *inode, struct page *node_page)
653 {
654 	struct f2fs_inode *ri;
655 	struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
656 
657 	f2fs_wait_on_page_writeback(node_page, NODE, true, true);
658 	set_page_dirty(node_page);
659 
660 	f2fs_inode_synced(inode);
661 
662 	ri = F2FS_INODE(node_page);
663 
664 	ri->i_mode = cpu_to_le16(inode->i_mode);
665 	ri->i_advise = F2FS_I(inode)->i_advise;
666 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
667 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
668 	ri->i_links = cpu_to_le32(inode->i_nlink);
669 	ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
670 
671 	if (!f2fs_is_atomic_file(inode) ||
672 			is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
673 		ri->i_size = cpu_to_le64(i_size_read(inode));
674 
675 	if (et) {
676 		read_lock(&et->lock);
677 		set_raw_read_extent(&et->largest, &ri->i_ext);
678 		read_unlock(&et->lock);
679 	} else {
680 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
681 	}
682 	set_raw_inline(inode, ri);
683 
684 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
685 	ri->i_ctime = cpu_to_le64(inode_get_ctime(inode).tv_sec);
686 	ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
687 	ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
688 	ri->i_ctime_nsec = cpu_to_le32(inode_get_ctime(inode).tv_nsec);
689 	ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
690 	if (S_ISDIR(inode->i_mode))
691 		ri->i_current_depth =
692 			cpu_to_le32(F2FS_I(inode)->i_current_depth);
693 	else if (S_ISREG(inode->i_mode))
694 		ri->i_gc_failures =
695 			cpu_to_le16(F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]);
696 	ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
697 	ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
698 	ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
699 	ri->i_generation = cpu_to_le32(inode->i_generation);
700 	ri->i_dir_level = F2FS_I(inode)->i_dir_level;
701 
702 	if (f2fs_has_extra_attr(inode)) {
703 		ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
704 
705 		if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
706 			ri->i_inline_xattr_size =
707 				cpu_to_le16(F2FS_I(inode)->i_inline_xattr_size);
708 
709 		if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
710 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
711 								i_projid)) {
712 			projid_t i_projid;
713 
714 			i_projid = from_kprojid(&init_user_ns,
715 						F2FS_I(inode)->i_projid);
716 			ri->i_projid = cpu_to_le32(i_projid);
717 		}
718 
719 		if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
720 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
721 								i_crtime)) {
722 			ri->i_crtime =
723 				cpu_to_le64(F2FS_I(inode)->i_crtime.tv_sec);
724 			ri->i_crtime_nsec =
725 				cpu_to_le32(F2FS_I(inode)->i_crtime.tv_nsec);
726 		}
727 
728 		if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
729 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
730 							i_compress_flag)) {
731 			unsigned short compress_flag;
732 
733 			ri->i_compr_blocks =
734 				cpu_to_le64(atomic_read(
735 					&F2FS_I(inode)->i_compr_blocks));
736 			ri->i_compress_algorithm =
737 				F2FS_I(inode)->i_compress_algorithm;
738 			compress_flag = F2FS_I(inode)->i_compress_flag |
739 				F2FS_I(inode)->i_compress_level <<
740 						COMPRESS_LEVEL_OFFSET;
741 			ri->i_compress_flag = cpu_to_le16(compress_flag);
742 			ri->i_log_cluster_size =
743 				F2FS_I(inode)->i_log_cluster_size;
744 		}
745 	}
746 
747 	__set_inode_rdev(inode, node_page);
748 
749 	/* deleted inode */
750 	if (inode->i_nlink == 0)
751 		clear_page_private_inline(node_page);
752 
753 	init_idisk_time(inode);
754 #ifdef CONFIG_F2FS_CHECK_FS
755 	f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
756 #endif
757 }
758 
f2fs_update_inode_page(struct inode * inode)759 void f2fs_update_inode_page(struct inode *inode)
760 {
761 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
762 	struct page *node_page;
763 	int count = 0;
764 retry:
765 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
766 	if (IS_ERR(node_page)) {
767 		int err = PTR_ERR(node_page);
768 
769 		/* The node block was truncated. */
770 		if (err == -ENOENT)
771 			return;
772 
773 		if (err == -EFSCORRUPTED)
774 			goto stop_checkpoint;
775 
776 		if (err == -ENOMEM || ++count <= DEFAULT_RETRY_IO_COUNT)
777 			goto retry;
778 stop_checkpoint:
779 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_UPDATE_INODE);
780 		return;
781 	}
782 	f2fs_update_inode(inode, node_page);
783 	f2fs_put_page(node_page, 1);
784 }
785 
f2fs_write_inode(struct inode * inode,struct writeback_control * wbc)786 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
787 {
788 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
789 
790 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
791 			inode->i_ino == F2FS_META_INO(sbi))
792 		return 0;
793 
794 	/*
795 	 * atime could be updated without dirtying f2fs inode in lazytime mode
796 	 */
797 	if (f2fs_is_time_consistent(inode) &&
798 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
799 		return 0;
800 
801 	if (!f2fs_is_checkpoint_ready(sbi)) {
802 		f2fs_mark_inode_dirty_sync(inode, true);
803 		return -ENOSPC;
804 	}
805 
806 	/*
807 	 * We need to balance fs here to prevent from producing dirty node pages
808 	 * during the urgent cleaning time when running out of free sections.
809 	 */
810 	f2fs_update_inode_page(inode);
811 	if (wbc && wbc->nr_to_write)
812 		f2fs_balance_fs(sbi, true);
813 	return 0;
814 }
815 
816 /*
817  * Called at the last iput() if i_nlink is zero
818  */
f2fs_evict_inode(struct inode * inode)819 void f2fs_evict_inode(struct inode *inode)
820 {
821 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
822 	struct f2fs_inode_info *fi = F2FS_I(inode);
823 	nid_t xnid = fi->i_xattr_nid;
824 	int err = 0;
825 
826 	f2fs_abort_atomic_write(inode, true);
827 
828 	if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
829 		clear_inode_flag(fi->cow_inode, FI_COW_FILE);
830 		F2FS_I(fi->cow_inode)->atomic_inode = NULL;
831 		iput(fi->cow_inode);
832 		fi->cow_inode = NULL;
833 	}
834 
835 	trace_f2fs_evict_inode(inode);
836 	truncate_inode_pages_final(&inode->i_data);
837 
838 	if ((inode->i_nlink || is_bad_inode(inode)) &&
839 		test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
840 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
841 
842 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
843 			inode->i_ino == F2FS_META_INO(sbi) ||
844 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
845 		goto out_clear;
846 
847 	f2fs_bug_on(sbi, get_dirty_pages(inode));
848 	f2fs_remove_dirty_inode(inode);
849 
850 	f2fs_destroy_extent_tree(inode);
851 
852 	if (inode->i_nlink || is_bad_inode(inode))
853 		goto no_delete;
854 
855 	err = f2fs_dquot_initialize(inode);
856 	if (err) {
857 		err = 0;
858 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
859 	}
860 
861 	f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
862 	f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
863 	f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
864 
865 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
866 		sb_start_intwrite(inode->i_sb);
867 	set_inode_flag(inode, FI_NO_ALLOC);
868 	i_size_write(inode, 0);
869 retry:
870 	if (F2FS_HAS_BLOCKS(inode))
871 		err = f2fs_truncate(inode);
872 
873 	if (time_to_inject(sbi, FAULT_EVICT_INODE))
874 		err = -EIO;
875 
876 	if (!err) {
877 		f2fs_lock_op(sbi);
878 		err = f2fs_remove_inode_page(inode);
879 		f2fs_unlock_op(sbi);
880 		if (err == -ENOENT) {
881 			err = 0;
882 
883 			/*
884 			 * in fuzzed image, another node may has the same
885 			 * block address as inode's, if it was truncated
886 			 * previously, truncation of inode node will fail.
887 			 */
888 			if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
889 				f2fs_warn(F2FS_I_SB(inode),
890 					"f2fs_evict_inode: inconsistent node id, ino:%lu",
891 					inode->i_ino);
892 				f2fs_inode_synced(inode);
893 				set_sbi_flag(sbi, SBI_NEED_FSCK);
894 			}
895 		}
896 	}
897 
898 	/* give more chances, if ENOMEM case */
899 	if (err == -ENOMEM) {
900 		err = 0;
901 		goto retry;
902 	}
903 
904 	if (err) {
905 		f2fs_update_inode_page(inode);
906 		if (dquot_initialize_needed(inode))
907 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
908 	}
909 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
910 		sb_end_intwrite(inode->i_sb);
911 no_delete:
912 	dquot_drop(inode);
913 
914 	stat_dec_inline_xattr(inode);
915 	stat_dec_inline_dir(inode);
916 	stat_dec_inline_inode(inode);
917 	stat_dec_compr_inode(inode);
918 	stat_sub_compr_blocks(inode,
919 			atomic_read(&fi->i_compr_blocks));
920 
921 	if (likely(!f2fs_cp_error(sbi) &&
922 				!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
923 		f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
924 	else
925 		f2fs_inode_synced(inode);
926 
927 	/* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
928 	if (inode->i_ino)
929 		invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
930 							inode->i_ino);
931 	if (xnid)
932 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
933 	if (inode->i_nlink) {
934 		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
935 			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
936 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
937 			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
938 	}
939 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
940 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
941 		clear_inode_flag(inode, FI_FREE_NID);
942 	} else {
943 		/*
944 		 * If xattr nid is corrupted, we can reach out error condition,
945 		 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
946 		 * In that case, f2fs_check_nid_range() is enough to give a clue.
947 		 */
948 	}
949 out_clear:
950 	fscrypt_put_encryption_info(inode);
951 	fsverity_cleanup_inode(inode);
952 	clear_inode(inode);
953 }
954 
955 /* caller should call f2fs_lock_op() */
f2fs_handle_failed_inode(struct inode * inode)956 void f2fs_handle_failed_inode(struct inode *inode)
957 {
958 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
959 	struct node_info ni;
960 	int err;
961 
962 	/*
963 	 * clear nlink of inode in order to release resource of inode
964 	 * immediately.
965 	 */
966 	clear_nlink(inode);
967 
968 	/*
969 	 * we must call this to avoid inode being remained as dirty, resulting
970 	 * in a panic when flushing dirty inodes in gdirty_list.
971 	 */
972 	f2fs_update_inode_page(inode);
973 	f2fs_inode_synced(inode);
974 
975 	/* don't make bad inode, since it becomes a regular file. */
976 	unlock_new_inode(inode);
977 
978 	/*
979 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
980 	 * so we can prevent losing this orphan when encoutering checkpoint
981 	 * and following suddenly power-off.
982 	 */
983 	err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
984 	if (err) {
985 		set_sbi_flag(sbi, SBI_NEED_FSCK);
986 		set_inode_flag(inode, FI_FREE_NID);
987 		f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
988 		goto out;
989 	}
990 
991 	if (ni.blk_addr != NULL_ADDR) {
992 		err = f2fs_acquire_orphan_inode(sbi);
993 		if (err) {
994 			set_sbi_flag(sbi, SBI_NEED_FSCK);
995 			f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
996 		} else {
997 			f2fs_add_orphan_inode(inode);
998 		}
999 		f2fs_alloc_nid_done(sbi, inode->i_ino);
1000 	} else {
1001 		set_inode_flag(inode, FI_FREE_NID);
1002 	}
1003 
1004 out:
1005 	f2fs_unlock_op(sbi);
1006 
1007 	/* iput will drop the inode object */
1008 	iput(inode);
1009 }
1010