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/stat.h>
14 #include <linux/buffer_head.h>
15 #include <linux/writeback.h>
16 #include <linux/blkdev.h>
17 #include <linux/falloc.h>
18 #include <linux/types.h>
19 #include <linux/compat.h>
20 #include <linux/uaccess.h>
21 #include <linux/mount.h>
22 #include <linux/pagevec.h>
29 #include <trace/events/f2fs.h>
31 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
34 struct page *page = vmf->page;
35 struct inode *inode = file_inode(vma->vm_file);
36 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
37 struct dnode_of_data dn;
42 sb_start_pagefault(inode->i_sb);
44 /* force to convert with normal data indices */
45 err = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1, page);
49 /* block allocation */
51 set_new_dnode(&dn, inode, NULL, NULL, 0);
52 err = f2fs_reserve_block(&dn, page->index);
57 file_update_time(vma->vm_file);
59 if (unlikely(page->mapping != inode->i_mapping ||
60 page_offset(page) > i_size_read(inode) ||
61 !PageUptodate(page))) {
68 * check to see if the page is mapped already (no holes)
70 if (PageMappedToDisk(page))
73 /* page is wholly or partially inside EOF */
74 if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
76 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
77 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
80 SetPageUptodate(page);
82 trace_f2fs_vm_page_mkwrite(page, DATA);
85 f2fs_wait_on_page_writeback(page, DATA);
87 sb_end_pagefault(inode->i_sb);
88 return block_page_mkwrite_return(err);
91 static const struct vm_operations_struct f2fs_file_vm_ops = {
92 .fault = filemap_fault,
93 .map_pages = filemap_map_pages,
94 .page_mkwrite = f2fs_vm_page_mkwrite,
95 .remap_pages = generic_file_remap_pages,
98 static int get_parent_ino(struct inode *inode, nid_t *pino)
100 struct dentry *dentry;
102 inode = igrab(inode);
103 dentry = d_find_any_alias(inode);
108 if (update_dent_inode(inode, &dentry->d_name)) {
113 *pino = parent_ino(dentry);
118 static inline bool need_do_checkpoint(struct inode *inode)
120 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
121 bool need_cp = false;
123 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
125 else if (file_wrong_pino(inode))
127 else if (!space_for_roll_forward(sbi))
129 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
131 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
137 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
139 struct inode *inode = file->f_mapping->host;
140 struct f2fs_inode_info *fi = F2FS_I(inode);
141 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
142 nid_t ino = inode->i_ino;
144 bool need_cp = false;
145 struct writeback_control wbc = {
146 .sync_mode = WB_SYNC_ALL,
147 .nr_to_write = LONG_MAX,
151 if (unlikely(f2fs_readonly(inode->i_sb)))
154 trace_f2fs_sync_file_enter(inode);
156 /* if fdatasync is triggered, let's do in-place-update */
157 if (get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
158 set_inode_flag(fi, FI_NEED_IPU);
159 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
160 clear_inode_flag(fi, FI_NEED_IPU);
163 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
168 * if there is no written data, don't waste time to write recovery info.
170 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
171 !exist_written_data(sbi, ino, APPEND_INO)) {
172 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
173 exist_written_data(sbi, ino, UPDATE_INO))
178 /* guarantee free sections for fsync */
179 f2fs_balance_fs(sbi);
182 * Both of fdatasync() and fsync() are able to be recovered from
185 down_read(&fi->i_sem);
186 need_cp = need_do_checkpoint(inode);
192 /* all the dirty node pages should be flushed for POR */
193 ret = f2fs_sync_fs(inode->i_sb, 1);
195 down_write(&fi->i_sem);
196 F2FS_I(inode)->xattr_ver = 0;
197 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
198 get_parent_ino(inode, &pino)) {
199 F2FS_I(inode)->i_pino = pino;
200 file_got_pino(inode);
201 up_write(&fi->i_sem);
202 mark_inode_dirty_sync(inode);
203 ret = f2fs_write_inode(inode, NULL);
207 up_write(&fi->i_sem);
210 /* if there is no written node page, write its inode page */
211 while (!sync_node_pages(sbi, ino, &wbc)) {
212 if (fsync_mark_done(sbi, ino))
214 mark_inode_dirty_sync(inode);
215 ret = f2fs_write_inode(inode, NULL);
219 ret = wait_on_node_pages_writeback(sbi, ino);
223 /* once recovery info is written, don't need to tack this */
224 remove_dirty_inode(sbi, ino, APPEND_INO);
225 clear_inode_flag(fi, FI_APPEND_WRITE);
227 remove_dirty_inode(sbi, ino, UPDATE_INO);
228 clear_inode_flag(fi, FI_UPDATE_WRITE);
229 ret = f2fs_issue_flush(F2FS_I_SB(inode));
232 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
236 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
237 pgoff_t pgofs, int whence)
242 if (whence != SEEK_DATA)
245 /* find first dirty page index */
246 pagevec_init(&pvec, 0);
247 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
248 PAGECACHE_TAG_DIRTY, 1);
249 pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
250 pagevec_release(&pvec);
254 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
259 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
260 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
264 if (blkaddr == NULL_ADDR)
271 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
273 struct inode *inode = file->f_mapping->host;
274 loff_t maxbytes = inode->i_sb->s_maxbytes;
275 struct dnode_of_data dn;
276 pgoff_t pgofs, end_offset, dirty;
277 loff_t data_ofs = offset;
281 mutex_lock(&inode->i_mutex);
283 isize = i_size_read(inode);
287 /* handle inline data case */
288 if (f2fs_has_inline_data(inode)) {
289 if (whence == SEEK_HOLE)
294 pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
296 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
298 for (; data_ofs < isize; data_ofs = pgofs << PAGE_CACHE_SHIFT) {
299 set_new_dnode(&dn, inode, NULL, NULL, 0);
300 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
301 if (err && err != -ENOENT) {
303 } else if (err == -ENOENT) {
304 /* direct node does not exists */
305 if (whence == SEEK_DATA) {
306 pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
314 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
316 /* find data/hole in dnode block */
317 for (; dn.ofs_in_node < end_offset;
318 dn.ofs_in_node++, pgofs++,
319 data_ofs = pgofs << PAGE_CACHE_SHIFT) {
321 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
323 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
331 if (whence == SEEK_DATA)
334 if (whence == SEEK_HOLE && data_ofs > isize)
336 mutex_unlock(&inode->i_mutex);
337 return vfs_setpos(file, data_ofs, maxbytes);
339 mutex_unlock(&inode->i_mutex);
343 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
345 struct inode *inode = file->f_mapping->host;
346 loff_t maxbytes = inode->i_sb->s_maxbytes;
352 return generic_file_llseek_size(file, offset, whence,
353 maxbytes, i_size_read(inode));
358 return f2fs_seek_block(file, offset, whence);
364 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
367 vma->vm_ops = &f2fs_file_vm_ops;
371 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
373 int nr_free = 0, ofs = dn->ofs_in_node;
374 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
375 struct f2fs_node *raw_node;
378 raw_node = F2FS_NODE(dn->node_page);
379 addr = blkaddr_in_node(raw_node) + ofs;
381 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
382 block_t blkaddr = le32_to_cpu(*addr);
383 if (blkaddr == NULL_ADDR)
386 update_extent_cache(NULL_ADDR, dn);
387 invalidate_blocks(sbi, blkaddr);
391 dec_valid_block_count(sbi, dn->inode, nr_free);
392 set_page_dirty(dn->node_page);
395 dn->ofs_in_node = ofs;
397 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
398 dn->ofs_in_node, nr_free);
402 void truncate_data_blocks(struct dnode_of_data *dn)
404 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
407 static void truncate_partial_data_page(struct inode *inode, u64 from)
409 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
412 if (f2fs_has_inline_data(inode))
413 return truncate_inline_data(inode, from);
418 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false);
423 if (unlikely(!PageUptodate(page) ||
424 page->mapping != inode->i_mapping))
427 f2fs_wait_on_page_writeback(page, DATA);
428 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
429 set_page_dirty(page);
432 f2fs_put_page(page, 1);
435 int truncate_blocks(struct inode *inode, u64 from, bool lock)
437 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
438 unsigned int blocksize = inode->i_sb->s_blocksize;
439 struct dnode_of_data dn;
441 int count = 0, err = 0;
443 trace_f2fs_truncate_blocks_enter(inode, from);
445 if (f2fs_has_inline_data(inode))
448 free_from = (pgoff_t)
449 ((from + blocksize - 1) >> (sbi->log_blocksize));
454 set_new_dnode(&dn, inode, NULL, NULL, 0);
455 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
461 trace_f2fs_truncate_blocks_exit(inode, err);
465 count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
467 count -= dn.ofs_in_node;
468 f2fs_bug_on(sbi, count < 0);
470 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
471 truncate_data_blocks_range(&dn, count);
477 err = truncate_inode_blocks(inode, free_from);
481 /* lastly zero out the first data page */
482 truncate_partial_data_page(inode, from);
484 trace_f2fs_truncate_blocks_exit(inode, err);
488 void f2fs_truncate(struct inode *inode)
490 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
491 S_ISLNK(inode->i_mode)))
494 trace_f2fs_truncate(inode);
496 if (!truncate_blocks(inode, i_size_read(inode), true)) {
497 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
498 mark_inode_dirty(inode);
502 int f2fs_getattr(struct vfsmount *mnt,
503 struct dentry *dentry, struct kstat *stat)
505 struct inode *inode = dentry->d_inode;
506 generic_fillattr(inode, stat);
511 #ifdef CONFIG_F2FS_FS_POSIX_ACL
512 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
514 struct f2fs_inode_info *fi = F2FS_I(inode);
515 unsigned int ia_valid = attr->ia_valid;
517 if (ia_valid & ATTR_UID)
518 inode->i_uid = attr->ia_uid;
519 if (ia_valid & ATTR_GID)
520 inode->i_gid = attr->ia_gid;
521 if (ia_valid & ATTR_ATIME)
522 inode->i_atime = timespec_trunc(attr->ia_atime,
523 inode->i_sb->s_time_gran);
524 if (ia_valid & ATTR_MTIME)
525 inode->i_mtime = timespec_trunc(attr->ia_mtime,
526 inode->i_sb->s_time_gran);
527 if (ia_valid & ATTR_CTIME)
528 inode->i_ctime = timespec_trunc(attr->ia_ctime,
529 inode->i_sb->s_time_gran);
530 if (ia_valid & ATTR_MODE) {
531 umode_t mode = attr->ia_mode;
533 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
535 set_acl_inode(fi, mode);
539 #define __setattr_copy setattr_copy
542 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
544 struct inode *inode = dentry->d_inode;
545 struct f2fs_inode_info *fi = F2FS_I(inode);
548 err = inode_change_ok(inode, attr);
552 if ((attr->ia_valid & ATTR_SIZE) &&
553 attr->ia_size != i_size_read(inode)) {
554 err = f2fs_convert_inline_data(inode, attr->ia_size, NULL);
558 truncate_setsize(inode, attr->ia_size);
559 f2fs_truncate(inode);
560 f2fs_balance_fs(F2FS_I_SB(inode));
563 __setattr_copy(inode, attr);
565 if (attr->ia_valid & ATTR_MODE) {
566 err = posix_acl_chmod(inode, get_inode_mode(inode));
567 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
568 inode->i_mode = fi->i_acl_mode;
569 clear_inode_flag(fi, FI_ACL_MODE);
573 mark_inode_dirty(inode);
577 const struct inode_operations f2fs_file_inode_operations = {
578 .getattr = f2fs_getattr,
579 .setattr = f2fs_setattr,
580 .get_acl = f2fs_get_acl,
581 .set_acl = f2fs_set_acl,
582 #ifdef CONFIG_F2FS_FS_XATTR
583 .setxattr = generic_setxattr,
584 .getxattr = generic_getxattr,
585 .listxattr = f2fs_listxattr,
586 .removexattr = generic_removexattr,
588 .fiemap = f2fs_fiemap,
591 static void fill_zero(struct inode *inode, pgoff_t index,
592 loff_t start, loff_t len)
594 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
600 f2fs_balance_fs(sbi);
603 page = get_new_data_page(inode, NULL, index, false);
607 f2fs_wait_on_page_writeback(page, DATA);
608 zero_user(page, start, len);
609 set_page_dirty(page);
610 f2fs_put_page(page, 1);
614 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
619 for (index = pg_start; index < pg_end; index++) {
620 struct dnode_of_data dn;
622 set_new_dnode(&dn, inode, NULL, NULL, 0);
623 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
630 if (dn.data_blkaddr != NULL_ADDR)
631 truncate_data_blocks_range(&dn, 1);
637 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
639 pgoff_t pg_start, pg_end;
640 loff_t off_start, off_end;
643 ret = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1, NULL);
647 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
648 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
650 off_start = offset & (PAGE_CACHE_SIZE - 1);
651 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
653 if (pg_start == pg_end) {
654 fill_zero(inode, pg_start, off_start,
655 off_end - off_start);
658 fill_zero(inode, pg_start++, off_start,
659 PAGE_CACHE_SIZE - off_start);
661 fill_zero(inode, pg_end, 0, off_end);
663 if (pg_start < pg_end) {
664 struct address_space *mapping = inode->i_mapping;
665 loff_t blk_start, blk_end;
666 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
668 f2fs_balance_fs(sbi);
670 blk_start = pg_start << PAGE_CACHE_SHIFT;
671 blk_end = pg_end << PAGE_CACHE_SHIFT;
672 truncate_inode_pages_range(mapping, blk_start,
676 ret = truncate_hole(inode, pg_start, pg_end);
684 static int expand_inode_data(struct inode *inode, loff_t offset,
685 loff_t len, int mode)
687 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
688 pgoff_t index, pg_start, pg_end;
689 loff_t new_size = i_size_read(inode);
690 loff_t off_start, off_end;
693 f2fs_balance_fs(sbi);
695 ret = inode_newsize_ok(inode, (len + offset));
699 ret = f2fs_convert_inline_data(inode, offset + len, NULL);
703 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
704 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
706 off_start = offset & (PAGE_CACHE_SIZE - 1);
707 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
711 for (index = pg_start; index <= pg_end; index++) {
712 struct dnode_of_data dn;
714 if (index == pg_end && !off_end)
717 set_new_dnode(&dn, inode, NULL, NULL, 0);
718 ret = f2fs_reserve_block(&dn, index);
722 if (pg_start == pg_end)
723 new_size = offset + len;
724 else if (index == pg_start && off_start)
725 new_size = (index + 1) << PAGE_CACHE_SHIFT;
726 else if (index == pg_end)
727 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
729 new_size += PAGE_CACHE_SIZE;
732 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
733 i_size_read(inode) < new_size) {
734 i_size_write(inode, new_size);
735 mark_inode_dirty(inode);
736 update_inode_page(inode);
743 static long f2fs_fallocate(struct file *file, int mode,
744 loff_t offset, loff_t len)
746 struct inode *inode = file_inode(file);
749 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
752 mutex_lock(&inode->i_mutex);
754 if (mode & FALLOC_FL_PUNCH_HOLE)
755 ret = punch_hole(inode, offset, len);
757 ret = expand_inode_data(inode, offset, len, mode);
760 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
761 mark_inode_dirty(inode);
764 mutex_unlock(&inode->i_mutex);
766 trace_f2fs_fallocate(inode, mode, offset, len, ret);
770 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
771 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
773 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
777 else if (S_ISREG(mode))
778 return flags & F2FS_REG_FLMASK;
780 return flags & F2FS_OTHER_FLMASK;
783 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
785 struct inode *inode = file_inode(filp);
786 struct f2fs_inode_info *fi = F2FS_I(inode);
791 case F2FS_IOC_GETFLAGS:
792 flags = fi->i_flags & FS_FL_USER_VISIBLE;
793 return put_user(flags, (int __user *) arg);
794 case F2FS_IOC_SETFLAGS:
796 unsigned int oldflags;
798 ret = mnt_want_write_file(filp);
802 if (!inode_owner_or_capable(inode)) {
807 if (get_user(flags, (int __user *) arg)) {
812 flags = f2fs_mask_flags(inode->i_mode, flags);
814 mutex_lock(&inode->i_mutex);
816 oldflags = fi->i_flags;
818 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
819 if (!capable(CAP_LINUX_IMMUTABLE)) {
820 mutex_unlock(&inode->i_mutex);
826 flags = flags & FS_FL_USER_MODIFIABLE;
827 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
829 mutex_unlock(&inode->i_mutex);
831 f2fs_set_inode_flags(inode);
832 inode->i_ctime = CURRENT_TIME;
833 mark_inode_dirty(inode);
835 mnt_drop_write_file(filp);
844 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
847 case F2FS_IOC32_GETFLAGS:
848 cmd = F2FS_IOC_GETFLAGS;
850 case F2FS_IOC32_SETFLAGS:
851 cmd = F2FS_IOC_SETFLAGS;
856 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
860 const struct file_operations f2fs_file_operations = {
861 .llseek = f2fs_llseek,
862 .read = new_sync_read,
863 .write = new_sync_write,
864 .read_iter = generic_file_read_iter,
865 .write_iter = generic_file_write_iter,
866 .open = generic_file_open,
867 .mmap = f2fs_file_mmap,
868 .fsync = f2fs_sync_file,
869 .fallocate = f2fs_fallocate,
870 .unlocked_ioctl = f2fs_ioctl,
872 .compat_ioctl = f2fs_compat_ioctl,
874 .splice_read = generic_file_splice_read,
875 .splice_write = iter_file_splice_write,