xref: /openbmc/linux/fs/f2fs/inode.c (revision d699090510c3223641a23834b4710e2d4309a6ad)
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 	if (f2fs_is_atomic_file(inode))
39 		return;
40 
41 	mark_inode_dirty_sync(inode);
42 }
43 
f2fs_set_inode_flags(struct inode * inode)44 void f2fs_set_inode_flags(struct inode *inode)
45 {
46 	unsigned int flags = F2FS_I(inode)->i_flags;
47 	unsigned int new_fl = 0;
48 
49 	if (flags & F2FS_SYNC_FL)
50 		new_fl |= S_SYNC;
51 	if (flags & F2FS_APPEND_FL)
52 		new_fl |= S_APPEND;
53 	if (flags & F2FS_IMMUTABLE_FL)
54 		new_fl |= S_IMMUTABLE;
55 	if (flags & F2FS_NOATIME_FL)
56 		new_fl |= S_NOATIME;
57 	if (flags & F2FS_DIRSYNC_FL)
58 		new_fl |= S_DIRSYNC;
59 	if (file_is_encrypt(inode))
60 		new_fl |= S_ENCRYPTED;
61 	if (file_is_verity(inode))
62 		new_fl |= S_VERITY;
63 	if (flags & F2FS_CASEFOLD_FL)
64 		new_fl |= S_CASEFOLD;
65 	inode_set_flags(inode, new_fl,
66 			S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|
67 			S_ENCRYPTED|S_VERITY|S_CASEFOLD);
68 }
69 
__get_inode_rdev(struct inode * inode,struct page * node_page)70 static void __get_inode_rdev(struct inode *inode, struct page *node_page)
71 {
72 	__le32 *addr = get_dnode_addr(inode, node_page);
73 
74 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
75 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
76 		if (addr[0])
77 			inode->i_rdev = old_decode_dev(le32_to_cpu(addr[0]));
78 		else
79 			inode->i_rdev = new_decode_dev(le32_to_cpu(addr[1]));
80 	}
81 }
82 
__set_inode_rdev(struct inode * inode,struct page * node_page)83 static void __set_inode_rdev(struct inode *inode, struct page *node_page)
84 {
85 	__le32 *addr = get_dnode_addr(inode, node_page);
86 
87 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
88 		if (old_valid_dev(inode->i_rdev)) {
89 			addr[0] = cpu_to_le32(old_encode_dev(inode->i_rdev));
90 			addr[1] = 0;
91 		} else {
92 			addr[0] = 0;
93 			addr[1] = cpu_to_le32(new_encode_dev(inode->i_rdev));
94 			addr[2] = 0;
95 		}
96 	}
97 }
98 
__recover_inline_status(struct inode * inode,struct page * ipage)99 static void __recover_inline_status(struct inode *inode, struct page *ipage)
100 {
101 	void *inline_data = inline_data_addr(inode, ipage);
102 	__le32 *start = inline_data;
103 	__le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
104 
105 	while (start < end) {
106 		if (*start++) {
107 			f2fs_wait_on_page_writeback(ipage, NODE, true, true);
108 
109 			set_inode_flag(inode, FI_DATA_EXIST);
110 			set_raw_inline(inode, F2FS_INODE(ipage));
111 			set_page_dirty(ipage);
112 			return;
113 		}
114 	}
115 	return;
116 }
117 
f2fs_enable_inode_chksum(struct f2fs_sb_info * sbi,struct page * page)118 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
119 {
120 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
121 
122 	if (!f2fs_sb_has_inode_chksum(sbi))
123 		return false;
124 
125 	if (!IS_INODE(page) || !(ri->i_inline & F2FS_EXTRA_ATTR))
126 		return false;
127 
128 	if (!F2FS_FITS_IN_INODE(ri, le16_to_cpu(ri->i_extra_isize),
129 				i_inode_checksum))
130 		return false;
131 
132 	return true;
133 }
134 
f2fs_inode_chksum(struct f2fs_sb_info * sbi,struct page * page)135 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
136 {
137 	struct f2fs_node *node = F2FS_NODE(page);
138 	struct f2fs_inode *ri = &node->i;
139 	__le32 ino = node->footer.ino;
140 	__le32 gen = ri->i_generation;
141 	__u32 chksum, chksum_seed;
142 	__u32 dummy_cs = 0;
143 	unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
144 	unsigned int cs_size = sizeof(dummy_cs);
145 
146 	chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
147 							sizeof(ino));
148 	chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
149 
150 	chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
151 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
152 	offset += cs_size;
153 	chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
154 						F2FS_BLKSIZE - offset);
155 	return chksum;
156 }
157 
f2fs_inode_chksum_verify(struct f2fs_sb_info * sbi,struct page * page)158 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
159 {
160 	struct f2fs_inode *ri;
161 	__u32 provided, calculated;
162 
163 	if (unlikely(is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)))
164 		return true;
165 
166 #ifdef CONFIG_F2FS_CHECK_FS
167 	if (!f2fs_enable_inode_chksum(sbi, page))
168 #else
169 	if (!f2fs_enable_inode_chksum(sbi, page) ||
170 			PageDirty(page) || PageWriteback(page))
171 #endif
172 		return true;
173 
174 	ri = &F2FS_NODE(page)->i;
175 	provided = le32_to_cpu(ri->i_inode_checksum);
176 	calculated = f2fs_inode_chksum(sbi, page);
177 
178 	if (provided != calculated)
179 		f2fs_warn(sbi, "checksum invalid, nid = %lu, ino_of_node = %x, %x vs. %x",
180 			  page->index, ino_of_node(page), provided, calculated);
181 
182 	return provided == calculated;
183 }
184 
f2fs_inode_chksum_set(struct f2fs_sb_info * sbi,struct page * page)185 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
186 {
187 	struct f2fs_inode *ri = &F2FS_NODE(page)->i;
188 
189 	if (!f2fs_enable_inode_chksum(sbi, page))
190 		return;
191 
192 	ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
193 }
194 
sanity_check_compress_inode(struct inode * inode,struct f2fs_inode * ri)195 static bool sanity_check_compress_inode(struct inode *inode,
196 			struct f2fs_inode *ri)
197 {
198 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
199 	unsigned char clevel;
200 
201 	if (ri->i_compress_algorithm >= COMPRESS_MAX) {
202 		f2fs_warn(sbi,
203 			"%s: inode (ino=%lx) has unsupported compress algorithm: %u, run fsck to fix",
204 			__func__, inode->i_ino, ri->i_compress_algorithm);
205 		return false;
206 	}
207 	if (le64_to_cpu(ri->i_compr_blocks) >
208 			SECTOR_TO_BLOCK(inode->i_blocks)) {
209 		f2fs_warn(sbi,
210 			"%s: inode (ino=%lx) has inconsistent i_compr_blocks:%llu, i_blocks:%llu, run fsck to fix",
211 			__func__, inode->i_ino, le64_to_cpu(ri->i_compr_blocks),
212 			SECTOR_TO_BLOCK(inode->i_blocks));
213 		return false;
214 	}
215 	if (ri->i_log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
216 		ri->i_log_cluster_size > MAX_COMPRESS_LOG_SIZE) {
217 		f2fs_warn(sbi,
218 			"%s: inode (ino=%lx) has unsupported log cluster size: %u, run fsck to fix",
219 			__func__, inode->i_ino, ri->i_log_cluster_size);
220 		return false;
221 	}
222 
223 	clevel = le16_to_cpu(ri->i_compress_flag) >>
224 				COMPRESS_LEVEL_OFFSET;
225 	switch (ri->i_compress_algorithm) {
226 	case COMPRESS_LZO:
227 #ifdef CONFIG_F2FS_FS_LZO
228 		if (clevel)
229 			goto err_level;
230 #endif
231 		break;
232 	case COMPRESS_LZORLE:
233 #ifdef CONFIG_F2FS_FS_LZORLE
234 		if (clevel)
235 			goto err_level;
236 #endif
237 		break;
238 	case COMPRESS_LZ4:
239 #ifdef CONFIG_F2FS_FS_LZ4
240 #ifdef CONFIG_F2FS_FS_LZ4HC
241 		if (clevel &&
242 		   (clevel < LZ4HC_MIN_CLEVEL || clevel > LZ4HC_MAX_CLEVEL))
243 			goto err_level;
244 #else
245 		if (clevel)
246 			goto err_level;
247 #endif
248 #endif
249 		break;
250 	case COMPRESS_ZSTD:
251 #ifdef CONFIG_F2FS_FS_ZSTD
252 		if (clevel < zstd_min_clevel() || clevel > zstd_max_clevel())
253 			goto err_level;
254 #endif
255 		break;
256 	default:
257 		goto err_level;
258 	}
259 
260 	return true;
261 err_level:
262 	f2fs_warn(sbi, "%s: inode (ino=%lx) has unsupported compress level: %u, run fsck to fix",
263 		  __func__, inode->i_ino, clevel);
264 	return false;
265 }
266 
sanity_check_inode(struct inode * inode,struct page * node_page)267 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
268 {
269 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
270 	struct f2fs_inode_info *fi = F2FS_I(inode);
271 	struct f2fs_inode *ri = F2FS_INODE(node_page);
272 	unsigned long long iblocks;
273 
274 	iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
275 	if (!iblocks) {
276 		f2fs_warn(sbi, "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, run fsck to fix.",
277 			  __func__, inode->i_ino, iblocks);
278 		return false;
279 	}
280 
281 	if (ino_of_node(node_page) != nid_of_node(node_page)) {
282 		f2fs_warn(sbi, "%s: corrupted inode footer i_ino=%lx, ino,nid: [%u, %u] run fsck to fix.",
283 			  __func__, inode->i_ino,
284 			  ino_of_node(node_page), nid_of_node(node_page));
285 		return false;
286 	}
287 
288 	if (f2fs_has_extra_attr(inode)) {
289 		if (!f2fs_sb_has_extra_attr(sbi)) {
290 			f2fs_warn(sbi, "%s: inode (ino=%lx) is with extra_attr, but extra_attr feature is off",
291 				  __func__, inode->i_ino);
292 			return false;
293 		}
294 		if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
295 			fi->i_extra_isize < F2FS_MIN_EXTRA_ATTR_SIZE ||
296 			fi->i_extra_isize % sizeof(__le32)) {
297 			f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, max: %zu",
298 				  __func__, inode->i_ino, fi->i_extra_isize,
299 				  F2FS_TOTAL_EXTRA_ATTR_SIZE);
300 			return false;
301 		}
302 		if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
303 			f2fs_has_inline_xattr(inode) &&
304 			(!fi->i_inline_xattr_size ||
305 			fi->i_inline_xattr_size > MAX_INLINE_XATTR_SIZE)) {
306 			f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %zu",
307 				  __func__, inode->i_ino, fi->i_inline_xattr_size,
308 				  MAX_INLINE_XATTR_SIZE);
309 			return false;
310 		}
311 		if (f2fs_sb_has_compression(sbi) &&
312 			fi->i_flags & F2FS_COMPR_FL &&
313 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
314 						i_compress_flag)) {
315 			if (!sanity_check_compress_inode(inode, ri))
316 				return false;
317 		}
318 	} else if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
319 		f2fs_warn(sbi, "%s: corrupted inode ino=%lx, run fsck to fix.",
320 			  __func__, inode->i_ino);
321 		return false;
322 	}
323 
324 	if (!f2fs_sb_has_extra_attr(sbi)) {
325 		if (f2fs_sb_has_project_quota(sbi)) {
326 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
327 				  __func__, inode->i_ino, F2FS_FEATURE_PRJQUOTA);
328 			return false;
329 		}
330 		if (f2fs_sb_has_inode_chksum(sbi)) {
331 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
332 				  __func__, inode->i_ino, F2FS_FEATURE_INODE_CHKSUM);
333 			return false;
334 		}
335 		if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
336 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
337 				  __func__, inode->i_ino, F2FS_FEATURE_FLEXIBLE_INLINE_XATTR);
338 			return false;
339 		}
340 		if (f2fs_sb_has_inode_crtime(sbi)) {
341 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
342 				  __func__, inode->i_ino, F2FS_FEATURE_INODE_CRTIME);
343 			return false;
344 		}
345 		if (f2fs_sb_has_compression(sbi)) {
346 			f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
347 				  __func__, inode->i_ino, F2FS_FEATURE_COMPRESSION);
348 			return false;
349 		}
350 	}
351 
352 	if (f2fs_sanity_check_inline_data(inode, node_page)) {
353 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_data, run fsck to fix",
354 			  __func__, inode->i_ino, inode->i_mode);
355 		return false;
356 	}
357 
358 	if (f2fs_has_inline_dentry(inode) && !S_ISDIR(inode->i_mode)) {
359 		f2fs_warn(sbi, "%s: inode (ino=%lx, mode=%u) should not have inline_dentry, run fsck to fix",
360 			  __func__, inode->i_ino, inode->i_mode);
361 		return false;
362 	}
363 
364 	if ((fi->i_flags & F2FS_CASEFOLD_FL) && !f2fs_sb_has_casefold(sbi)) {
365 		f2fs_warn(sbi, "%s: inode (ino=%lx) has casefold flag, but casefold feature is off",
366 			  __func__, inode->i_ino);
367 		return false;
368 	}
369 
370 	if (fi->i_xattr_nid && f2fs_check_nid_range(sbi, fi->i_xattr_nid)) {
371 		f2fs_warn(sbi, "%s: inode (ino=%lx) has corrupted i_xattr_nid: %u, run fsck to fix.",
372 			  __func__, inode->i_ino, fi->i_xattr_nid);
373 		return false;
374 	}
375 
376 	return true;
377 }
378 
init_idisk_time(struct inode * inode)379 static void init_idisk_time(struct inode *inode)
380 {
381 	struct f2fs_inode_info *fi = F2FS_I(inode);
382 
383 	fi->i_disk_time[0] = inode->i_atime;
384 	fi->i_disk_time[1] = inode_get_ctime(inode);
385 	fi->i_disk_time[2] = inode->i_mtime;
386 }
387 
do_read_inode(struct inode * inode)388 static int do_read_inode(struct inode *inode)
389 {
390 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
391 	struct f2fs_inode_info *fi = F2FS_I(inode);
392 	struct page *node_page;
393 	struct f2fs_inode *ri;
394 	projid_t i_projid;
395 
396 	/* Check if ino is within scope */
397 	if (f2fs_check_nid_range(sbi, inode->i_ino))
398 		return -EINVAL;
399 
400 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
401 	if (IS_ERR(node_page))
402 		return PTR_ERR(node_page);
403 
404 	ri = F2FS_INODE(node_page);
405 
406 	inode->i_mode = le16_to_cpu(ri->i_mode);
407 	i_uid_write(inode, le32_to_cpu(ri->i_uid));
408 	i_gid_write(inode, le32_to_cpu(ri->i_gid));
409 	set_nlink(inode, le32_to_cpu(ri->i_links));
410 	inode->i_size = le64_to_cpu(ri->i_size);
411 	inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
412 
413 	inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
414 	inode_set_ctime(inode, le64_to_cpu(ri->i_ctime),
415 			le32_to_cpu(ri->i_ctime_nsec));
416 	inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
417 	inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
418 	inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
419 	inode->i_generation = le32_to_cpu(ri->i_generation);
420 	if (S_ISDIR(inode->i_mode))
421 		fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
422 	else if (S_ISREG(inode->i_mode))
423 		fi->i_gc_failures[GC_FAILURE_PIN] =
424 					le16_to_cpu(ri->i_gc_failures);
425 	fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
426 	fi->i_flags = le32_to_cpu(ri->i_flags);
427 	if (S_ISREG(inode->i_mode))
428 		fi->i_flags &= ~F2FS_PROJINHERIT_FL;
429 	bitmap_zero(fi->flags, FI_MAX);
430 	fi->i_advise = ri->i_advise;
431 	fi->i_pino = le32_to_cpu(ri->i_pino);
432 	fi->i_dir_level = ri->i_dir_level;
433 
434 	get_inline_info(inode, ri);
435 
436 	fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
437 					le16_to_cpu(ri->i_extra_isize) : 0;
438 
439 	if (f2fs_sb_has_flexible_inline_xattr(sbi)) {
440 		fi->i_inline_xattr_size = le16_to_cpu(ri->i_inline_xattr_size);
441 	} else if (f2fs_has_inline_xattr(inode) ||
442 				f2fs_has_inline_dentry(inode)) {
443 		fi->i_inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
444 	} else {
445 
446 		/*
447 		 * Previous inline data or directory always reserved 200 bytes
448 		 * in inode layout, even if inline_xattr is disabled. In order
449 		 * to keep inline_dentry's structure for backward compatibility,
450 		 * we get the space back only from inline_data.
451 		 */
452 		fi->i_inline_xattr_size = 0;
453 	}
454 
455 	if (!sanity_check_inode(inode, node_page)) {
456 		f2fs_put_page(node_page, 1);
457 		set_sbi_flag(sbi, SBI_NEED_FSCK);
458 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
459 		return -EFSCORRUPTED;
460 	}
461 
462 	/* check data exist */
463 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
464 		__recover_inline_status(inode, node_page);
465 
466 	/* try to recover cold bit for non-dir inode */
467 	if (!S_ISDIR(inode->i_mode) && !is_cold_node(node_page)) {
468 		f2fs_wait_on_page_writeback(node_page, NODE, true, true);
469 		set_cold_node(node_page, false);
470 		set_page_dirty(node_page);
471 	}
472 
473 	/* get rdev by using inline_info */
474 	__get_inode_rdev(inode, node_page);
475 
476 	if (!f2fs_need_inode_block_update(sbi, inode->i_ino))
477 		fi->last_disk_size = inode->i_size;
478 
479 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
480 		set_inode_flag(inode, FI_PROJ_INHERIT);
481 
482 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi) &&
483 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
484 		i_projid = (projid_t)le32_to_cpu(ri->i_projid);
485 	else
486 		i_projid = F2FS_DEF_PROJID;
487 	fi->i_projid = make_kprojid(&init_user_ns, i_projid);
488 
489 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_inode_crtime(sbi) &&
490 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
491 		fi->i_crtime.tv_sec = le64_to_cpu(ri->i_crtime);
492 		fi->i_crtime.tv_nsec = le32_to_cpu(ri->i_crtime_nsec);
493 	}
494 
495 	if (f2fs_has_extra_attr(inode) && f2fs_sb_has_compression(sbi) &&
496 					(fi->i_flags & F2FS_COMPR_FL)) {
497 		if (F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
498 					i_compress_flag)) {
499 			unsigned short compress_flag;
500 
501 			atomic_set(&fi->i_compr_blocks,
502 					le64_to_cpu(ri->i_compr_blocks));
503 			fi->i_compress_algorithm = ri->i_compress_algorithm;
504 			fi->i_log_cluster_size = ri->i_log_cluster_size;
505 			compress_flag = le16_to_cpu(ri->i_compress_flag);
506 			fi->i_compress_level = compress_flag >>
507 						COMPRESS_LEVEL_OFFSET;
508 			fi->i_compress_flag = compress_flag &
509 					GENMASK(COMPRESS_LEVEL_OFFSET - 1, 0);
510 			fi->i_cluster_size = BIT(fi->i_log_cluster_size);
511 			set_inode_flag(inode, FI_COMPRESSED_FILE);
512 		}
513 	}
514 
515 	init_idisk_time(inode);
516 
517 	if (!sanity_check_extent_cache(inode, node_page)) {
518 		f2fs_put_page(node_page, 1);
519 		f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
520 		return -EFSCORRUPTED;
521 	}
522 
523 	/* Need all the flag bits */
524 	f2fs_init_read_extent_tree(inode, node_page);
525 	f2fs_init_age_extent_tree(inode);
526 
527 	f2fs_put_page(node_page, 1);
528 
529 	stat_inc_inline_xattr(inode);
530 	stat_inc_inline_inode(inode);
531 	stat_inc_inline_dir(inode);
532 	stat_inc_compr_inode(inode);
533 	stat_add_compr_blocks(inode, atomic_read(&fi->i_compr_blocks));
534 
535 	return 0;
536 }
537 
is_meta_ino(struct f2fs_sb_info * sbi,unsigned int ino)538 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
539 {
540 	return ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi) ||
541 		ino == F2FS_COMPRESS_INO(sbi);
542 }
543 
f2fs_iget(struct super_block * sb,unsigned long ino)544 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
545 {
546 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
547 	struct inode *inode;
548 	int ret = 0;
549 
550 	inode = iget_locked(sb, ino);
551 	if (!inode)
552 		return ERR_PTR(-ENOMEM);
553 
554 	if (!(inode->i_state & I_NEW)) {
555 		if (is_meta_ino(sbi, ino)) {
556 			f2fs_err(sbi, "inaccessible inode: %lu, run fsck to repair", ino);
557 			set_sbi_flag(sbi, SBI_NEED_FSCK);
558 			ret = -EFSCORRUPTED;
559 			trace_f2fs_iget_exit(inode, ret);
560 			iput(inode);
561 			f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
562 			return ERR_PTR(ret);
563 		}
564 
565 		trace_f2fs_iget(inode);
566 		return inode;
567 	}
568 
569 	if (is_meta_ino(sbi, ino))
570 		goto make_now;
571 
572 	ret = do_read_inode(inode);
573 	if (ret)
574 		goto bad_inode;
575 make_now:
576 	if (ino == F2FS_NODE_INO(sbi)) {
577 		inode->i_mapping->a_ops = &f2fs_node_aops;
578 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
579 	} else if (ino == F2FS_META_INO(sbi)) {
580 		inode->i_mapping->a_ops = &f2fs_meta_aops;
581 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
582 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
583 #ifdef CONFIG_F2FS_FS_COMPRESSION
584 		inode->i_mapping->a_ops = &f2fs_compress_aops;
585 		/*
586 		 * generic_error_remove_page only truncates pages of regular
587 		 * inode
588 		 */
589 		inode->i_mode |= S_IFREG;
590 #endif
591 		mapping_set_gfp_mask(inode->i_mapping,
592 			GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
593 	} else if (S_ISREG(inode->i_mode)) {
594 		inode->i_op = &f2fs_file_inode_operations;
595 		inode->i_fop = &f2fs_file_operations;
596 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
597 	} else if (S_ISDIR(inode->i_mode)) {
598 		inode->i_op = &f2fs_dir_inode_operations;
599 		inode->i_fop = &f2fs_dir_operations;
600 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
601 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
602 	} else if (S_ISLNK(inode->i_mode)) {
603 		if (file_is_encrypt(inode))
604 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
605 		else
606 			inode->i_op = &f2fs_symlink_inode_operations;
607 		inode_nohighmem(inode);
608 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
609 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
610 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
611 		inode->i_op = &f2fs_special_inode_operations;
612 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
613 	} else {
614 		ret = -EIO;
615 		goto bad_inode;
616 	}
617 	f2fs_set_inode_flags(inode);
618 
619 	unlock_new_inode(inode);
620 	trace_f2fs_iget(inode);
621 	return inode;
622 
623 bad_inode:
624 	f2fs_inode_synced(inode);
625 	iget_failed(inode);
626 	trace_f2fs_iget_exit(inode, ret);
627 	return ERR_PTR(ret);
628 }
629 
f2fs_iget_retry(struct super_block * sb,unsigned long ino)630 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
631 {
632 	struct inode *inode;
633 retry:
634 	inode = f2fs_iget(sb, ino);
635 	if (IS_ERR(inode)) {
636 		if (PTR_ERR(inode) == -ENOMEM) {
637 			memalloc_retry_wait(GFP_NOFS);
638 			goto retry;
639 		}
640 	}
641 	return inode;
642 }
643 
f2fs_update_inode(struct inode * inode,struct page * node_page)644 void f2fs_update_inode(struct inode *inode, struct page *node_page)
645 {
646 	struct f2fs_inode *ri;
647 	struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
648 
649 	f2fs_wait_on_page_writeback(node_page, NODE, true, true);
650 	set_page_dirty(node_page);
651 
652 	f2fs_inode_synced(inode);
653 
654 	ri = F2FS_INODE(node_page);
655 
656 	ri->i_mode = cpu_to_le16(inode->i_mode);
657 	ri->i_advise = F2FS_I(inode)->i_advise;
658 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
659 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
660 	ri->i_links = cpu_to_le32(inode->i_nlink);
661 	ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
662 
663 	if (!f2fs_is_atomic_file(inode) ||
664 			is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
665 		ri->i_size = cpu_to_le64(i_size_read(inode));
666 
667 	if (et) {
668 		read_lock(&et->lock);
669 		set_raw_read_extent(&et->largest, &ri->i_ext);
670 		read_unlock(&et->lock);
671 	} else {
672 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
673 	}
674 	set_raw_inline(inode, ri);
675 
676 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
677 	ri->i_ctime = cpu_to_le64(inode_get_ctime(inode).tv_sec);
678 	ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
679 	ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
680 	ri->i_ctime_nsec = cpu_to_le32(inode_get_ctime(inode).tv_nsec);
681 	ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
682 	if (S_ISDIR(inode->i_mode))
683 		ri->i_current_depth =
684 			cpu_to_le32(F2FS_I(inode)->i_current_depth);
685 	else if (S_ISREG(inode->i_mode))
686 		ri->i_gc_failures =
687 			cpu_to_le16(F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]);
688 	ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
689 	ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
690 	ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
691 	ri->i_generation = cpu_to_le32(inode->i_generation);
692 	ri->i_dir_level = F2FS_I(inode)->i_dir_level;
693 
694 	if (f2fs_has_extra_attr(inode)) {
695 		ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
696 
697 		if (f2fs_sb_has_flexible_inline_xattr(F2FS_I_SB(inode)))
698 			ri->i_inline_xattr_size =
699 				cpu_to_le16(F2FS_I(inode)->i_inline_xattr_size);
700 
701 		if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)) &&
702 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
703 								i_projid)) {
704 			projid_t i_projid;
705 
706 			i_projid = from_kprojid(&init_user_ns,
707 						F2FS_I(inode)->i_projid);
708 			ri->i_projid = cpu_to_le32(i_projid);
709 		}
710 
711 		if (f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
712 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
713 								i_crtime)) {
714 			ri->i_crtime =
715 				cpu_to_le64(F2FS_I(inode)->i_crtime.tv_sec);
716 			ri->i_crtime_nsec =
717 				cpu_to_le32(F2FS_I(inode)->i_crtime.tv_nsec);
718 		}
719 
720 		if (f2fs_sb_has_compression(F2FS_I_SB(inode)) &&
721 			F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
722 							i_compress_flag)) {
723 			unsigned short compress_flag;
724 
725 			ri->i_compr_blocks =
726 				cpu_to_le64(atomic_read(
727 					&F2FS_I(inode)->i_compr_blocks));
728 			ri->i_compress_algorithm =
729 				F2FS_I(inode)->i_compress_algorithm;
730 			compress_flag = F2FS_I(inode)->i_compress_flag |
731 				F2FS_I(inode)->i_compress_level <<
732 						COMPRESS_LEVEL_OFFSET;
733 			ri->i_compress_flag = cpu_to_le16(compress_flag);
734 			ri->i_log_cluster_size =
735 				F2FS_I(inode)->i_log_cluster_size;
736 		}
737 	}
738 
739 	__set_inode_rdev(inode, node_page);
740 
741 	/* deleted inode */
742 	if (inode->i_nlink == 0)
743 		clear_page_private_inline(node_page);
744 
745 	init_idisk_time(inode);
746 #ifdef CONFIG_F2FS_CHECK_FS
747 	f2fs_inode_chksum_set(F2FS_I_SB(inode), node_page);
748 #endif
749 }
750 
f2fs_update_inode_page(struct inode * inode)751 void f2fs_update_inode_page(struct inode *inode)
752 {
753 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
754 	struct page *node_page;
755 	int count = 0;
756 retry:
757 	node_page = f2fs_get_node_page(sbi, inode->i_ino);
758 	if (IS_ERR(node_page)) {
759 		int err = PTR_ERR(node_page);
760 
761 		/* The node block was truncated. */
762 		if (err == -ENOENT)
763 			return;
764 
765 		if (err == -EFSCORRUPTED)
766 			goto stop_checkpoint;
767 
768 		if (err == -ENOMEM || ++count <= DEFAULT_RETRY_IO_COUNT)
769 			goto retry;
770 stop_checkpoint:
771 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_UPDATE_INODE);
772 		return;
773 	}
774 	f2fs_update_inode(inode, node_page);
775 	f2fs_put_page(node_page, 1);
776 }
777 
f2fs_write_inode(struct inode * inode,struct writeback_control * wbc)778 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
779 {
780 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
781 
782 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
783 			inode->i_ino == F2FS_META_INO(sbi))
784 		return 0;
785 
786 	/*
787 	 * atime could be updated without dirtying f2fs inode in lazytime mode
788 	 */
789 	if (f2fs_is_time_consistent(inode) &&
790 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
791 		return 0;
792 
793 	if (!f2fs_is_checkpoint_ready(sbi)) {
794 		f2fs_mark_inode_dirty_sync(inode, true);
795 		return -ENOSPC;
796 	}
797 
798 	/*
799 	 * We need to balance fs here to prevent from producing dirty node pages
800 	 * during the urgent cleaning time when running out of free sections.
801 	 */
802 	f2fs_update_inode_page(inode);
803 	if (wbc && wbc->nr_to_write)
804 		f2fs_balance_fs(sbi, true);
805 	return 0;
806 }
807 
808 /*
809  * Called at the last iput() if i_nlink is zero
810  */
f2fs_evict_inode(struct inode * inode)811 void f2fs_evict_inode(struct inode *inode)
812 {
813 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
814 	struct f2fs_inode_info *fi = F2FS_I(inode);
815 	nid_t xnid = fi->i_xattr_nid;
816 	int err = 0;
817 
818 	f2fs_abort_atomic_write(inode, true);
819 
820 	if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
821 		clear_inode_flag(fi->cow_inode, FI_COW_FILE);
822 		F2FS_I(fi->cow_inode)->atomic_inode = NULL;
823 		iput(fi->cow_inode);
824 		fi->cow_inode = NULL;
825 	}
826 
827 	trace_f2fs_evict_inode(inode);
828 	truncate_inode_pages_final(&inode->i_data);
829 
830 	if ((inode->i_nlink || is_bad_inode(inode)) &&
831 		test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
832 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
833 
834 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
835 			inode->i_ino == F2FS_META_INO(sbi) ||
836 			inode->i_ino == F2FS_COMPRESS_INO(sbi))
837 		goto out_clear;
838 
839 	f2fs_bug_on(sbi, get_dirty_pages(inode));
840 	f2fs_remove_dirty_inode(inode);
841 
842 	f2fs_destroy_extent_tree(inode);
843 
844 	if (inode->i_nlink || is_bad_inode(inode))
845 		goto no_delete;
846 
847 	err = f2fs_dquot_initialize(inode);
848 	if (err) {
849 		err = 0;
850 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
851 	}
852 
853 	f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
854 	f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
855 	f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
856 
857 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
858 		sb_start_intwrite(inode->i_sb);
859 	set_inode_flag(inode, FI_NO_ALLOC);
860 	i_size_write(inode, 0);
861 retry:
862 	if (F2FS_HAS_BLOCKS(inode))
863 		err = f2fs_truncate(inode);
864 
865 	if (time_to_inject(sbi, FAULT_EVICT_INODE))
866 		err = -EIO;
867 
868 	if (!err) {
869 		f2fs_lock_op(sbi);
870 		err = f2fs_remove_inode_page(inode);
871 		f2fs_unlock_op(sbi);
872 		if (err == -ENOENT) {
873 			err = 0;
874 
875 			/*
876 			 * in fuzzed image, another node may has the same
877 			 * block address as inode's, if it was truncated
878 			 * previously, truncation of inode node will fail.
879 			 */
880 			if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
881 				f2fs_warn(F2FS_I_SB(inode),
882 					"f2fs_evict_inode: inconsistent node id, ino:%lu",
883 					inode->i_ino);
884 				f2fs_inode_synced(inode);
885 				set_sbi_flag(sbi, SBI_NEED_FSCK);
886 			}
887 		}
888 	}
889 
890 	/* give more chances, if ENOMEM case */
891 	if (err == -ENOMEM) {
892 		err = 0;
893 		goto retry;
894 	}
895 
896 	if (err) {
897 		f2fs_update_inode_page(inode);
898 		if (dquot_initialize_needed(inode))
899 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
900 	}
901 	if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING))
902 		sb_end_intwrite(inode->i_sb);
903 no_delete:
904 	dquot_drop(inode);
905 
906 	stat_dec_inline_xattr(inode);
907 	stat_dec_inline_dir(inode);
908 	stat_dec_inline_inode(inode);
909 	stat_dec_compr_inode(inode);
910 	stat_sub_compr_blocks(inode,
911 			atomic_read(&fi->i_compr_blocks));
912 
913 	if (likely(!f2fs_cp_error(sbi) &&
914 				!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
915 		f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
916 	else
917 		f2fs_inode_synced(inode);
918 
919 	/* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
920 	if (inode->i_ino)
921 		invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
922 							inode->i_ino);
923 	if (xnid)
924 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
925 	if (inode->i_nlink) {
926 		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
927 			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
928 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
929 			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
930 	}
931 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
932 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
933 		clear_inode_flag(inode, FI_FREE_NID);
934 	} else {
935 		/*
936 		 * If xattr nid is corrupted, we can reach out error condition,
937 		 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
938 		 * In that case, f2fs_check_nid_range() is enough to give a clue.
939 		 */
940 	}
941 out_clear:
942 	fscrypt_put_encryption_info(inode);
943 	fsverity_cleanup_inode(inode);
944 	clear_inode(inode);
945 }
946 
947 /* caller should call f2fs_lock_op() */
f2fs_handle_failed_inode(struct inode * inode)948 void f2fs_handle_failed_inode(struct inode *inode)
949 {
950 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
951 	struct node_info ni;
952 	int err;
953 
954 	/*
955 	 * clear nlink of inode in order to release resource of inode
956 	 * immediately.
957 	 */
958 	clear_nlink(inode);
959 
960 	/*
961 	 * we must call this to avoid inode being remained as dirty, resulting
962 	 * in a panic when flushing dirty inodes in gdirty_list.
963 	 */
964 	f2fs_update_inode_page(inode);
965 	f2fs_inode_synced(inode);
966 
967 	/* don't make bad inode, since it becomes a regular file. */
968 	unlock_new_inode(inode);
969 
970 	/*
971 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
972 	 * so we can prevent losing this orphan when encoutering checkpoint
973 	 * and following suddenly power-off.
974 	 */
975 	err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
976 	if (err) {
977 		set_sbi_flag(sbi, SBI_NEED_FSCK);
978 		set_inode_flag(inode, FI_FREE_NID);
979 		f2fs_warn(sbi, "May loss orphan inode, run fsck to fix.");
980 		goto out;
981 	}
982 
983 	if (ni.blk_addr != NULL_ADDR) {
984 		err = f2fs_acquire_orphan_inode(sbi);
985 		if (err) {
986 			set_sbi_flag(sbi, SBI_NEED_FSCK);
987 			f2fs_warn(sbi, "Too many orphan inodes, run fsck to fix.");
988 		} else {
989 			f2fs_add_orphan_inode(inode);
990 		}
991 		f2fs_alloc_nid_done(sbi, inode->i_ino);
992 	} else {
993 		set_inode_flag(inode, FI_FREE_NID);
994 	}
995 
996 out:
997 	f2fs_unlock_op(sbi);
998 
999 	/* iput will drop the inode object */
1000 	iput(inode);
1001 }
1002