xref: /openbmc/linux/fs/f2fs/inode.c (revision bc5aa3a0)
1 /*
2  * fs/f2fs/inode.c
3  *
4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5  *             http://www.samsung.com/
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  */
11 #include <linux/fs.h>
12 #include <linux/f2fs_fs.h>
13 #include <linux/buffer_head.h>
14 #include <linux/writeback.h>
15 
16 #include "f2fs.h"
17 #include "node.h"
18 
19 #include <trace/events/f2fs.h>
20 
21 void f2fs_mark_inode_dirty_sync(struct inode *inode)
22 {
23 	if (f2fs_inode_dirtied(inode))
24 		return;
25 	mark_inode_dirty_sync(inode);
26 }
27 
28 void f2fs_set_inode_flags(struct inode *inode)
29 {
30 	unsigned int flags = F2FS_I(inode)->i_flags;
31 	unsigned int new_fl = 0;
32 
33 	if (flags & FS_SYNC_FL)
34 		new_fl |= S_SYNC;
35 	if (flags & FS_APPEND_FL)
36 		new_fl |= S_APPEND;
37 	if (flags & FS_IMMUTABLE_FL)
38 		new_fl |= S_IMMUTABLE;
39 	if (flags & FS_NOATIME_FL)
40 		new_fl |= S_NOATIME;
41 	if (flags & FS_DIRSYNC_FL)
42 		new_fl |= S_DIRSYNC;
43 	inode_set_flags(inode, new_fl,
44 			S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
45 	f2fs_mark_inode_dirty_sync(inode);
46 }
47 
48 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
49 {
50 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
51 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
52 		if (ri->i_addr[0])
53 			inode->i_rdev =
54 				old_decode_dev(le32_to_cpu(ri->i_addr[0]));
55 		else
56 			inode->i_rdev =
57 				new_decode_dev(le32_to_cpu(ri->i_addr[1]));
58 	}
59 }
60 
61 static bool __written_first_block(struct f2fs_inode *ri)
62 {
63 	block_t addr = le32_to_cpu(ri->i_addr[0]);
64 
65 	if (addr != NEW_ADDR && addr != NULL_ADDR)
66 		return true;
67 	return false;
68 }
69 
70 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
71 {
72 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
73 		if (old_valid_dev(inode->i_rdev)) {
74 			ri->i_addr[0] =
75 				cpu_to_le32(old_encode_dev(inode->i_rdev));
76 			ri->i_addr[1] = 0;
77 		} else {
78 			ri->i_addr[0] = 0;
79 			ri->i_addr[1] =
80 				cpu_to_le32(new_encode_dev(inode->i_rdev));
81 			ri->i_addr[2] = 0;
82 		}
83 	}
84 }
85 
86 static void __recover_inline_status(struct inode *inode, struct page *ipage)
87 {
88 	void *inline_data = inline_data_addr(ipage);
89 	__le32 *start = inline_data;
90 	__le32 *end = start + MAX_INLINE_DATA / sizeof(__le32);
91 
92 	while (start < end) {
93 		if (*start++) {
94 			f2fs_wait_on_page_writeback(ipage, NODE, true);
95 
96 			set_inode_flag(inode, FI_DATA_EXIST);
97 			set_raw_inline(inode, F2FS_INODE(ipage));
98 			set_page_dirty(ipage);
99 			return;
100 		}
101 	}
102 	return;
103 }
104 
105 static int do_read_inode(struct inode *inode)
106 {
107 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
108 	struct f2fs_inode_info *fi = F2FS_I(inode);
109 	struct page *node_page;
110 	struct f2fs_inode *ri;
111 
112 	/* Check if ino is within scope */
113 	if (check_nid_range(sbi, inode->i_ino)) {
114 		f2fs_msg(inode->i_sb, KERN_ERR, "bad inode number: %lu",
115 			 (unsigned long) inode->i_ino);
116 		WARN_ON(1);
117 		return -EINVAL;
118 	}
119 
120 	node_page = get_node_page(sbi, inode->i_ino);
121 	if (IS_ERR(node_page))
122 		return PTR_ERR(node_page);
123 
124 	ri = F2FS_INODE(node_page);
125 
126 	inode->i_mode = le16_to_cpu(ri->i_mode);
127 	i_uid_write(inode, le32_to_cpu(ri->i_uid));
128 	i_gid_write(inode, le32_to_cpu(ri->i_gid));
129 	set_nlink(inode, le32_to_cpu(ri->i_links));
130 	inode->i_size = le64_to_cpu(ri->i_size);
131 	inode->i_blocks = le64_to_cpu(ri->i_blocks);
132 
133 	inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
134 	inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
135 	inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
136 	inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
137 	inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
138 	inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
139 	inode->i_generation = le32_to_cpu(ri->i_generation);
140 
141 	fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
142 	fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
143 	fi->i_flags = le32_to_cpu(ri->i_flags);
144 	fi->flags = 0;
145 	fi->i_advise = ri->i_advise;
146 	fi->i_pino = le32_to_cpu(ri->i_pino);
147 	fi->i_dir_level = ri->i_dir_level;
148 
149 	if (f2fs_init_extent_tree(inode, &ri->i_ext))
150 		set_page_dirty(node_page);
151 
152 	get_inline_info(inode, ri);
153 
154 	/* check data exist */
155 	if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
156 		__recover_inline_status(inode, node_page);
157 
158 	/* get rdev by using inline_info */
159 	__get_inode_rdev(inode, ri);
160 
161 	if (__written_first_block(ri))
162 		set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
163 
164 	if (!need_inode_block_update(sbi, inode->i_ino))
165 		fi->last_disk_size = inode->i_size;
166 
167 	f2fs_put_page(node_page, 1);
168 
169 	stat_inc_inline_xattr(inode);
170 	stat_inc_inline_inode(inode);
171 	stat_inc_inline_dir(inode);
172 
173 	return 0;
174 }
175 
176 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
177 {
178 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
179 	struct inode *inode;
180 	int ret = 0;
181 
182 	inode = iget_locked(sb, ino);
183 	if (!inode)
184 		return ERR_PTR(-ENOMEM);
185 
186 	if (!(inode->i_state & I_NEW)) {
187 		trace_f2fs_iget(inode);
188 		return inode;
189 	}
190 	if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
191 		goto make_now;
192 
193 	ret = do_read_inode(inode);
194 	if (ret)
195 		goto bad_inode;
196 make_now:
197 	if (ino == F2FS_NODE_INO(sbi)) {
198 		inode->i_mapping->a_ops = &f2fs_node_aops;
199 		mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
200 	} else if (ino == F2FS_META_INO(sbi)) {
201 		inode->i_mapping->a_ops = &f2fs_meta_aops;
202 		mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
203 	} else if (S_ISREG(inode->i_mode)) {
204 		inode->i_op = &f2fs_file_inode_operations;
205 		inode->i_fop = &f2fs_file_operations;
206 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
207 	} else if (S_ISDIR(inode->i_mode)) {
208 		inode->i_op = &f2fs_dir_inode_operations;
209 		inode->i_fop = &f2fs_dir_operations;
210 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
211 		mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_HIGH_ZERO);
212 	} else if (S_ISLNK(inode->i_mode)) {
213 		if (f2fs_encrypted_inode(inode))
214 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
215 		else
216 			inode->i_op = &f2fs_symlink_inode_operations;
217 		inode_nohighmem(inode);
218 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
219 	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
220 			S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
221 		inode->i_op = &f2fs_special_inode_operations;
222 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
223 	} else {
224 		ret = -EIO;
225 		goto bad_inode;
226 	}
227 	unlock_new_inode(inode);
228 	trace_f2fs_iget(inode);
229 	return inode;
230 
231 bad_inode:
232 	iget_failed(inode);
233 	trace_f2fs_iget_exit(inode, ret);
234 	return ERR_PTR(ret);
235 }
236 
237 int update_inode(struct inode *inode, struct page *node_page)
238 {
239 	struct f2fs_inode *ri;
240 
241 	f2fs_inode_synced(inode);
242 
243 	f2fs_wait_on_page_writeback(node_page, NODE, true);
244 
245 	ri = F2FS_INODE(node_page);
246 
247 	ri->i_mode = cpu_to_le16(inode->i_mode);
248 	ri->i_advise = F2FS_I(inode)->i_advise;
249 	ri->i_uid = cpu_to_le32(i_uid_read(inode));
250 	ri->i_gid = cpu_to_le32(i_gid_read(inode));
251 	ri->i_links = cpu_to_le32(inode->i_nlink);
252 	ri->i_size = cpu_to_le64(i_size_read(inode));
253 	ri->i_blocks = cpu_to_le64(inode->i_blocks);
254 
255 	if (F2FS_I(inode)->extent_tree)
256 		set_raw_extent(&F2FS_I(inode)->extent_tree->largest,
257 							&ri->i_ext);
258 	else
259 		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
260 	set_raw_inline(inode, ri);
261 
262 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
263 	ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
264 	ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
265 	ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
266 	ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
267 	ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
268 	ri->i_current_depth = cpu_to_le32(F2FS_I(inode)->i_current_depth);
269 	ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
270 	ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
271 	ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
272 	ri->i_generation = cpu_to_le32(inode->i_generation);
273 	ri->i_dir_level = F2FS_I(inode)->i_dir_level;
274 
275 	__set_inode_rdev(inode, ri);
276 	set_cold_node(inode, node_page);
277 
278 	/* deleted inode */
279 	if (inode->i_nlink == 0)
280 		clear_inline_node(node_page);
281 
282 	return set_page_dirty(node_page);
283 }
284 
285 int update_inode_page(struct inode *inode)
286 {
287 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
288 	struct page *node_page;
289 	int ret = 0;
290 retry:
291 	node_page = get_node_page(sbi, inode->i_ino);
292 	if (IS_ERR(node_page)) {
293 		int err = PTR_ERR(node_page);
294 		if (err == -ENOMEM) {
295 			cond_resched();
296 			goto retry;
297 		} else if (err != -ENOENT) {
298 			f2fs_stop_checkpoint(sbi, false);
299 		}
300 		f2fs_inode_synced(inode);
301 		return 0;
302 	}
303 	ret = update_inode(inode, node_page);
304 	f2fs_put_page(node_page, 1);
305 	return ret;
306 }
307 
308 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
309 {
310 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
311 
312 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
313 			inode->i_ino == F2FS_META_INO(sbi))
314 		return 0;
315 
316 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE))
317 		return 0;
318 
319 	/*
320 	 * We need to balance fs here to prevent from producing dirty node pages
321 	 * during the urgent cleaning time when runing out of free sections.
322 	 */
323 	if (update_inode_page(inode))
324 		f2fs_balance_fs(sbi, true);
325 	return 0;
326 }
327 
328 /*
329  * Called at the last iput() if i_nlink is zero
330  */
331 void f2fs_evict_inode(struct inode *inode)
332 {
333 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
334 	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
335 	int err = 0;
336 
337 	/* some remained atomic pages should discarded */
338 	if (f2fs_is_atomic_file(inode))
339 		drop_inmem_pages(inode);
340 
341 	trace_f2fs_evict_inode(inode);
342 	truncate_inode_pages_final(&inode->i_data);
343 
344 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
345 			inode->i_ino == F2FS_META_INO(sbi))
346 		goto out_clear;
347 
348 	f2fs_bug_on(sbi, get_dirty_pages(inode));
349 	remove_dirty_inode(inode);
350 
351 	f2fs_destroy_extent_tree(inode);
352 
353 	if (inode->i_nlink || is_bad_inode(inode))
354 		goto no_delete;
355 
356 #ifdef CONFIG_F2FS_FAULT_INJECTION
357 	if (time_to_inject(FAULT_EVICT_INODE))
358 		goto no_delete;
359 #endif
360 
361 	sb_start_intwrite(inode->i_sb);
362 	set_inode_flag(inode, FI_NO_ALLOC);
363 	i_size_write(inode, 0);
364 retry:
365 	if (F2FS_HAS_BLOCKS(inode))
366 		err = f2fs_truncate(inode);
367 
368 	if (!err) {
369 		f2fs_lock_op(sbi);
370 		err = remove_inode_page(inode);
371 		f2fs_unlock_op(sbi);
372 	}
373 
374 	/* give more chances, if ENOMEM case */
375 	if (err == -ENOMEM) {
376 		err = 0;
377 		goto retry;
378 	}
379 
380 	if (err)
381 		update_inode_page(inode);
382 	sb_end_intwrite(inode->i_sb);
383 no_delete:
384 	stat_dec_inline_xattr(inode);
385 	stat_dec_inline_dir(inode);
386 	stat_dec_inline_inode(inode);
387 
388 	invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino, inode->i_ino);
389 	if (xnid)
390 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
391 	if (is_inode_flag_set(inode, FI_APPEND_WRITE))
392 		add_ino_entry(sbi, inode->i_ino, APPEND_INO);
393 	if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
394 		add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
395 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
396 		alloc_nid_failed(sbi, inode->i_ino);
397 		clear_inode_flag(inode, FI_FREE_NID);
398 	}
399 	f2fs_bug_on(sbi, err &&
400 		!exist_written_data(sbi, inode->i_ino, ORPHAN_INO));
401 out_clear:
402 	fscrypt_put_encryption_info(inode, NULL);
403 	clear_inode(inode);
404 }
405 
406 /* caller should call f2fs_lock_op() */
407 void handle_failed_inode(struct inode *inode)
408 {
409 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
410 	struct node_info ni;
411 
412 	/* don't make bad inode, since it becomes a regular file. */
413 	unlock_new_inode(inode);
414 
415 	/*
416 	 * Note: we should add inode to orphan list before f2fs_unlock_op()
417 	 * so we can prevent losing this orphan when encoutering checkpoint
418 	 * and following suddenly power-off.
419 	 */
420 	get_node_info(sbi, inode->i_ino, &ni);
421 
422 	if (ni.blk_addr != NULL_ADDR) {
423 		int err = acquire_orphan_inode(sbi);
424 		if (err) {
425 			set_sbi_flag(sbi, SBI_NEED_FSCK);
426 			f2fs_msg(sbi->sb, KERN_WARNING,
427 				"Too many orphan inodes, run fsck to fix.");
428 		} else {
429 			add_orphan_inode(inode);
430 		}
431 		alloc_nid_done(sbi, inode->i_ino);
432 	} else {
433 		set_inode_flag(inode, FI_FREE_NID);
434 	}
435 
436 	f2fs_unlock_op(sbi);
437 
438 	/* iput will drop the inode object */
439 	iput(inode);
440 }
441