4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
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.
12 #include <linux/f2fs_fs.h>
13 #include <linux/buffer_head.h>
14 #include <linux/backing-dev.h>
15 #include <linux/writeback.h>
21 #include <trace/events/f2fs.h>
23 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
25 if (f2fs_inode_dirtied(inode, sync))
28 mark_inode_dirty_sync(inode);
31 void f2fs_set_inode_flags(struct inode *inode)
33 unsigned int flags = F2FS_I(inode)->i_flags;
34 unsigned int new_fl = 0;
36 if (flags & FS_SYNC_FL)
38 if (flags & FS_APPEND_FL)
40 if (flags & FS_IMMUTABLE_FL)
41 new_fl |= S_IMMUTABLE;
42 if (flags & FS_NOATIME_FL)
44 if (flags & FS_DIRSYNC_FL)
46 inode_set_flags(inode, new_fl,
47 S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
50 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
52 int extra_size = get_extra_isize(inode);
54 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
55 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
56 if (ri->i_addr[extra_size])
57 inode->i_rdev = old_decode_dev(
58 le32_to_cpu(ri->i_addr[extra_size]));
60 inode->i_rdev = new_decode_dev(
61 le32_to_cpu(ri->i_addr[extra_size + 1]));
65 static int __written_first_block(struct f2fs_sb_info *sbi,
66 struct f2fs_inode *ri)
68 block_t addr = le32_to_cpu(ri->i_addr[offset_in_addr(ri)]);
70 if (!__is_valid_data_blkaddr(addr))
72 if (!f2fs_is_valid_blkaddr(sbi, addr, DATA_GENERIC))
77 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
79 int extra_size = get_extra_isize(inode);
81 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
82 if (old_valid_dev(inode->i_rdev)) {
83 ri->i_addr[extra_size] =
84 cpu_to_le32(old_encode_dev(inode->i_rdev));
85 ri->i_addr[extra_size + 1] = 0;
87 ri->i_addr[extra_size] = 0;
88 ri->i_addr[extra_size + 1] =
89 cpu_to_le32(new_encode_dev(inode->i_rdev));
90 ri->i_addr[extra_size + 2] = 0;
95 static void __recover_inline_status(struct inode *inode, struct page *ipage)
97 void *inline_data = inline_data_addr(inode, ipage);
98 __le32 *start = inline_data;
99 __le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
101 while (start < end) {
103 f2fs_wait_on_page_writeback(ipage, NODE, true);
105 set_inode_flag(inode, FI_DATA_EXIST);
106 set_raw_inline(inode, F2FS_INODE(ipage));
107 set_page_dirty(ipage);
114 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
116 struct f2fs_inode *ri = &F2FS_NODE(page)->i;
117 int extra_isize = le32_to_cpu(ri->i_extra_isize);
119 if (!f2fs_sb_has_inode_chksum(sbi->sb))
122 if (!RAW_IS_INODE(F2FS_NODE(page)) || !(ri->i_inline & F2FS_EXTRA_ATTR))
125 if (!F2FS_FITS_IN_INODE(ri, extra_isize, i_inode_checksum))
131 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
133 struct f2fs_node *node = F2FS_NODE(page);
134 struct f2fs_inode *ri = &node->i;
135 __le32 ino = node->footer.ino;
136 __le32 gen = ri->i_generation;
137 __u32 chksum, chksum_seed;
139 unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
140 unsigned int cs_size = sizeof(dummy_cs);
142 chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
144 chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
146 chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
147 chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
149 chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
150 F2FS_BLKSIZE - offset);
154 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
156 struct f2fs_inode *ri;
157 __u32 provided, calculated;
159 if (!f2fs_enable_inode_chksum(sbi, page) ||
160 PageDirty(page) || PageWriteback(page))
163 ri = &F2FS_NODE(page)->i;
164 provided = le32_to_cpu(ri->i_inode_checksum);
165 calculated = f2fs_inode_chksum(sbi, page);
167 if (provided != calculated)
168 f2fs_msg(sbi->sb, KERN_WARNING,
169 "checksum invalid, ino = %x, %x vs. %x",
170 ino_of_node(page), provided, calculated);
172 return provided == calculated;
175 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
177 struct f2fs_inode *ri = &F2FS_NODE(page)->i;
179 if (!f2fs_enable_inode_chksum(sbi, page))
182 ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
185 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
187 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
188 struct f2fs_inode_info *fi = F2FS_I(inode);
189 unsigned long long iblocks;
191 iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
193 set_sbi_flag(sbi, SBI_NEED_FSCK);
194 f2fs_msg(sbi->sb, KERN_WARNING,
195 "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, "
197 __func__, inode->i_ino, iblocks);
201 if (ino_of_node(node_page) != nid_of_node(node_page)) {
202 set_sbi_flag(sbi, SBI_NEED_FSCK);
203 f2fs_msg(sbi->sb, KERN_WARNING,
204 "%s: corrupted inode footer i_ino=%lx, ino,nid: "
205 "[%u, %u] run fsck to fix.",
206 __func__, inode->i_ino,
207 ino_of_node(node_page), nid_of_node(node_page));
211 if (f2fs_has_extra_attr(inode) &&
212 !f2fs_sb_has_extra_attr(sbi->sb)) {
213 set_sbi_flag(sbi, SBI_NEED_FSCK);
214 f2fs_msg(sbi->sb, KERN_WARNING,
215 "%s: inode (ino=%lx) is with extra_attr, "
216 "but extra_attr feature is off",
217 __func__, inode->i_ino);
221 if (fi->i_extra_isize > F2FS_TOTAL_EXTRA_ATTR_SIZE ||
222 fi->i_extra_isize % sizeof(__le32)) {
223 set_sbi_flag(sbi, SBI_NEED_FSCK);
224 f2fs_msg(sbi->sb, KERN_WARNING,
225 "%s: inode (ino=%lx) has corrupted i_extra_isize: %d, "
227 __func__, inode->i_ino, fi->i_extra_isize,
228 F2FS_TOTAL_EXTRA_ATTR_SIZE);
232 if (F2FS_I(inode)->extent_tree) {
233 struct extent_info *ei = &F2FS_I(inode)->extent_tree->largest;
236 (!f2fs_is_valid_blkaddr(sbi, ei->blk, DATA_GENERIC) ||
237 !f2fs_is_valid_blkaddr(sbi, ei->blk + ei->len - 1,
239 set_sbi_flag(sbi, SBI_NEED_FSCK);
240 f2fs_msg(sbi->sb, KERN_WARNING,
241 "%s: inode (ino=%lx) extent info [%u, %u, %u] "
242 "is incorrect, run fsck to fix",
243 __func__, inode->i_ino,
244 ei->blk, ei->fofs, ei->len);
251 static int do_read_inode(struct inode *inode)
253 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
254 struct f2fs_inode_info *fi = F2FS_I(inode);
255 struct page *node_page;
256 struct f2fs_inode *ri;
260 /* Check if ino is within scope */
261 if (check_nid_range(sbi, inode->i_ino))
264 node_page = get_node_page(sbi, inode->i_ino);
265 if (IS_ERR(node_page))
266 return PTR_ERR(node_page);
268 ri = F2FS_INODE(node_page);
270 inode->i_mode = le16_to_cpu(ri->i_mode);
271 i_uid_write(inode, le32_to_cpu(ri->i_uid));
272 i_gid_write(inode, le32_to_cpu(ri->i_gid));
273 set_nlink(inode, le32_to_cpu(ri->i_links));
274 inode->i_size = le64_to_cpu(ri->i_size);
275 inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
277 inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
278 inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
279 inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
280 inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
281 inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
282 inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
283 inode->i_generation = le32_to_cpu(ri->i_generation);
285 fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
286 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
287 fi->i_flags = le32_to_cpu(ri->i_flags);
289 fi->i_advise = ri->i_advise;
290 fi->i_pino = le32_to_cpu(ri->i_pino);
291 fi->i_dir_level = ri->i_dir_level;
293 if (f2fs_init_extent_tree(inode, &ri->i_ext))
294 set_page_dirty(node_page);
296 get_inline_info(inode, ri);
298 fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
299 le16_to_cpu(ri->i_extra_isize) : 0;
301 if (!sanity_check_inode(inode, node_page)) {
302 f2fs_put_page(node_page, 1);
303 return -EFSCORRUPTED;
306 /* check data exist */
307 if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
308 __recover_inline_status(inode, node_page);
310 /* get rdev by using inline_info */
311 __get_inode_rdev(inode, ri);
313 err = __written_first_block(sbi, ri);
315 f2fs_put_page(node_page, 1);
319 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
321 if (!need_inode_block_update(sbi, inode->i_ino))
322 fi->last_disk_size = inode->i_size;
324 if (fi->i_flags & FS_PROJINHERIT_FL)
325 set_inode_flag(inode, FI_PROJ_INHERIT);
327 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi->sb) &&
328 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
329 i_projid = (projid_t)le32_to_cpu(ri->i_projid);
331 i_projid = F2FS_DEF_PROJID;
332 fi->i_projid = make_kprojid(&init_user_ns, i_projid);
334 f2fs_put_page(node_page, 1);
336 stat_inc_inline_xattr(inode);
337 stat_inc_inline_inode(inode);
338 stat_inc_inline_dir(inode);
343 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
345 struct f2fs_sb_info *sbi = F2FS_SB(sb);
349 inode = iget_locked(sb, ino);
351 return ERR_PTR(-ENOMEM);
353 if (!(inode->i_state & I_NEW)) {
354 trace_f2fs_iget(inode);
357 if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
360 ret = do_read_inode(inode);
364 if (ino == F2FS_NODE_INO(sbi)) {
365 inode->i_mapping->a_ops = &f2fs_node_aops;
366 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
367 } else if (ino == F2FS_META_INO(sbi)) {
368 inode->i_mapping->a_ops = &f2fs_meta_aops;
369 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
370 } else if (S_ISREG(inode->i_mode)) {
371 inode->i_op = &f2fs_file_inode_operations;
372 inode->i_fop = &f2fs_file_operations;
373 inode->i_mapping->a_ops = &f2fs_dblock_aops;
374 } else if (S_ISDIR(inode->i_mode)) {
375 inode->i_op = &f2fs_dir_inode_operations;
376 inode->i_fop = &f2fs_dir_operations;
377 inode->i_mapping->a_ops = &f2fs_dblock_aops;
378 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_HIGH_ZERO);
379 } else if (S_ISLNK(inode->i_mode)) {
380 if (f2fs_encrypted_inode(inode))
381 inode->i_op = &f2fs_encrypted_symlink_inode_operations;
383 inode->i_op = &f2fs_symlink_inode_operations;
384 inode_nohighmem(inode);
385 inode->i_mapping->a_ops = &f2fs_dblock_aops;
386 } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
387 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
388 inode->i_op = &f2fs_special_inode_operations;
389 init_special_inode(inode, inode->i_mode, inode->i_rdev);
394 f2fs_set_inode_flags(inode);
395 unlock_new_inode(inode);
396 trace_f2fs_iget(inode);
400 f2fs_inode_synced(inode);
402 trace_f2fs_iget_exit(inode, ret);
406 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
410 inode = f2fs_iget(sb, ino);
412 if (PTR_ERR(inode) == -ENOMEM) {
413 congestion_wait(BLK_RW_ASYNC, HZ/50);
420 int update_inode(struct inode *inode, struct page *node_page)
422 struct f2fs_inode *ri;
423 struct extent_tree *et = F2FS_I(inode)->extent_tree;
425 f2fs_inode_synced(inode);
427 f2fs_wait_on_page_writeback(node_page, NODE, true);
429 ri = F2FS_INODE(node_page);
431 ri->i_mode = cpu_to_le16(inode->i_mode);
432 ri->i_advise = F2FS_I(inode)->i_advise;
433 ri->i_uid = cpu_to_le32(i_uid_read(inode));
434 ri->i_gid = cpu_to_le32(i_gid_read(inode));
435 ri->i_links = cpu_to_le32(inode->i_nlink);
436 ri->i_size = cpu_to_le64(i_size_read(inode));
437 ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
440 read_lock(&et->lock);
441 set_raw_extent(&et->largest, &ri->i_ext);
442 read_unlock(&et->lock);
444 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
446 set_raw_inline(inode, ri);
448 ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
449 ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
450 ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
451 ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
452 ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
453 ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
454 ri->i_current_depth = cpu_to_le32(F2FS_I(inode)->i_current_depth);
455 ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
456 ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
457 ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
458 ri->i_generation = cpu_to_le32(inode->i_generation);
459 ri->i_dir_level = F2FS_I(inode)->i_dir_level;
461 if (f2fs_has_extra_attr(inode)) {
462 ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
464 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)->sb) &&
465 F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
469 i_projid = from_kprojid(&init_user_ns,
470 F2FS_I(inode)->i_projid);
471 ri->i_projid = cpu_to_le32(i_projid);
475 __set_inode_rdev(inode, ri);
476 set_cold_node(inode, node_page);
479 if (inode->i_nlink == 0)
480 clear_inline_node(node_page);
482 return set_page_dirty(node_page);
485 int update_inode_page(struct inode *inode)
487 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
488 struct page *node_page;
491 node_page = get_node_page(sbi, inode->i_ino);
492 if (IS_ERR(node_page)) {
493 int err = PTR_ERR(node_page);
494 if (err == -ENOMEM) {
497 } else if (err != -ENOENT) {
498 f2fs_stop_checkpoint(sbi, false);
502 ret = update_inode(inode, node_page);
503 f2fs_put_page(node_page, 1);
507 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
509 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
511 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
512 inode->i_ino == F2FS_META_INO(sbi))
515 if (!is_inode_flag_set(inode, FI_DIRTY_INODE))
519 * We need to balance fs here to prevent from producing dirty node pages
520 * during the urgent cleaning time when runing out of free sections.
522 update_inode_page(inode);
523 if (wbc && wbc->nr_to_write)
524 f2fs_balance_fs(sbi, true);
529 * Called at the last iput() if i_nlink is zero
531 void f2fs_evict_inode(struct inode *inode)
533 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
534 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
537 /* some remained atomic pages should discarded */
538 if (f2fs_is_atomic_file(inode))
539 drop_inmem_pages(inode);
541 trace_f2fs_evict_inode(inode);
542 truncate_inode_pages_final(&inode->i_data);
544 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
545 inode->i_ino == F2FS_META_INO(sbi))
548 f2fs_bug_on(sbi, get_dirty_pages(inode));
549 remove_dirty_inode(inode);
551 f2fs_destroy_extent_tree(inode);
553 if (inode->i_nlink || is_bad_inode(inode))
556 dquot_initialize(inode);
558 remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
559 remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
561 sb_start_intwrite(inode->i_sb);
562 set_inode_flag(inode, FI_NO_ALLOC);
563 i_size_write(inode, 0);
565 if (F2FS_HAS_BLOCKS(inode))
566 err = f2fs_truncate(inode);
568 #ifdef CONFIG_F2FS_FAULT_INJECTION
569 if (time_to_inject(sbi, FAULT_EVICT_INODE)) {
570 f2fs_show_injection_info(FAULT_EVICT_INODE);
576 err = remove_inode_page(inode);
582 /* give more chances, if ENOMEM case */
583 if (err == -ENOMEM) {
589 update_inode_page(inode);
590 dquot_free_inode(inode);
591 sb_end_intwrite(inode->i_sb);
595 stat_dec_inline_xattr(inode);
596 stat_dec_inline_dir(inode);
597 stat_dec_inline_inode(inode);
599 if (!is_set_ckpt_flags(sbi, CP_ERROR_FLAG))
600 f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
602 /* ino == 0, if f2fs_new_inode() was failed t*/
604 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
607 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
608 if (inode->i_nlink) {
609 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
610 add_ino_entry(sbi, inode->i_ino, APPEND_INO);
611 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
612 add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
614 if (is_inode_flag_set(inode, FI_FREE_NID)) {
615 alloc_nid_failed(sbi, inode->i_ino);
616 clear_inode_flag(inode, FI_FREE_NID);
619 * If xattr nid is corrupted, we can reach out error condition,
620 * err & !exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
621 * In that case, check_nid_range() is enough to give a clue.
625 fscrypt_put_encryption_info(inode, NULL);
629 /* caller should call f2fs_lock_op() */
630 void handle_failed_inode(struct inode *inode)
632 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
636 * clear nlink of inode in order to release resource of inode
642 * we must call this to avoid inode being remained as dirty, resulting
643 * in a panic when flushing dirty inodes in gdirty_list.
645 update_inode_page(inode);
646 f2fs_inode_synced(inode);
648 /* don't make bad inode, since it becomes a regular file. */
649 unlock_new_inode(inode);
652 * Note: we should add inode to orphan list before f2fs_unlock_op()
653 * so we can prevent losing this orphan when encoutering checkpoint
654 * and following suddenly power-off.
656 get_node_info(sbi, inode->i_ino, &ni);
658 if (ni.blk_addr != NULL_ADDR) {
659 int err = acquire_orphan_inode(sbi);
661 set_sbi_flag(sbi, SBI_NEED_FSCK);
662 f2fs_msg(sbi->sb, KERN_WARNING,
663 "Too many orphan inodes, run fsck to fix.");
665 add_orphan_inode(inode);
667 alloc_nid_done(sbi, inode->i_ino);
669 set_inode_flag(inode, FI_FREE_NID);
674 /* iput will drop the inode object */