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 */
158 set_inode_flag(fi, FI_NEED_IPU);
160 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
162 clear_inode_flag(fi, FI_NEED_IPU);
164 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
169 * if there is no written data, don't waste time to write recovery info.
171 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
172 !exist_written_data(sbi, ino, APPEND_INO)) {
173 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
174 exist_written_data(sbi, ino, UPDATE_INO))
179 /* guarantee free sections for fsync */
180 f2fs_balance_fs(sbi);
183 * Both of fdatasync() and fsync() are able to be recovered from
186 down_read(&fi->i_sem);
187 need_cp = need_do_checkpoint(inode);
193 /* all the dirty node pages should be flushed for POR */
194 ret = f2fs_sync_fs(inode->i_sb, 1);
196 down_write(&fi->i_sem);
197 F2FS_I(inode)->xattr_ver = 0;
198 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
199 get_parent_ino(inode, &pino)) {
200 F2FS_I(inode)->i_pino = pino;
201 file_got_pino(inode);
202 up_write(&fi->i_sem);
203 mark_inode_dirty_sync(inode);
204 ret = f2fs_write_inode(inode, NULL);
208 up_write(&fi->i_sem);
211 /* if there is no written node page, write its inode page */
212 while (!sync_node_pages(sbi, ino, &wbc)) {
213 if (fsync_mark_done(sbi, ino))
215 mark_inode_dirty_sync(inode);
216 ret = f2fs_write_inode(inode, NULL);
220 ret = wait_on_node_pages_writeback(sbi, ino);
224 /* once recovery info is written, don't need to tack this */
225 remove_dirty_inode(sbi, ino, APPEND_INO);
226 clear_inode_flag(fi, FI_APPEND_WRITE);
228 remove_dirty_inode(sbi, ino, UPDATE_INO);
229 clear_inode_flag(fi, FI_UPDATE_WRITE);
230 ret = f2fs_issue_flush(F2FS_I_SB(inode));
233 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
237 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
238 pgoff_t pgofs, int whence)
243 if (whence != SEEK_DATA)
246 /* find first dirty page index */
247 pagevec_init(&pvec, 0);
248 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
249 PAGECACHE_TAG_DIRTY, 1);
250 pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
251 pagevec_release(&pvec);
255 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
260 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
261 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
265 if (blkaddr == NULL_ADDR)
272 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
274 struct inode *inode = file->f_mapping->host;
275 loff_t maxbytes = inode->i_sb->s_maxbytes;
276 struct dnode_of_data dn;
277 pgoff_t pgofs, end_offset, dirty;
278 loff_t data_ofs = offset;
282 mutex_lock(&inode->i_mutex);
284 isize = i_size_read(inode);
288 /* handle inline data case */
289 if (f2fs_has_inline_data(inode)) {
290 if (whence == SEEK_HOLE)
295 pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
297 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
299 for (; data_ofs < isize; data_ofs = pgofs << PAGE_CACHE_SHIFT) {
300 set_new_dnode(&dn, inode, NULL, NULL, 0);
301 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
302 if (err && err != -ENOENT) {
304 } else if (err == -ENOENT) {
305 /* direct node does not exists */
306 if (whence == SEEK_DATA) {
307 pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
315 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
317 /* find data/hole in dnode block */
318 for (; dn.ofs_in_node < end_offset;
319 dn.ofs_in_node++, pgofs++,
320 data_ofs = pgofs << PAGE_CACHE_SHIFT) {
322 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
324 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
332 if (whence == SEEK_DATA)
335 if (whence == SEEK_HOLE && data_ofs > isize)
337 mutex_unlock(&inode->i_mutex);
338 return vfs_setpos(file, data_ofs, maxbytes);
340 mutex_unlock(&inode->i_mutex);
344 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
346 struct inode *inode = file->f_mapping->host;
347 loff_t maxbytes = inode->i_sb->s_maxbytes;
353 return generic_file_llseek_size(file, offset, whence,
354 maxbytes, i_size_read(inode));
359 return f2fs_seek_block(file, offset, whence);
365 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
368 vma->vm_ops = &f2fs_file_vm_ops;
372 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
374 int nr_free = 0, ofs = dn->ofs_in_node;
375 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
376 struct f2fs_node *raw_node;
379 raw_node = F2FS_NODE(dn->node_page);
380 addr = blkaddr_in_node(raw_node) + ofs;
382 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
383 block_t blkaddr = le32_to_cpu(*addr);
384 if (blkaddr == NULL_ADDR)
387 update_extent_cache(NULL_ADDR, dn);
388 invalidate_blocks(sbi, blkaddr);
392 dec_valid_block_count(sbi, dn->inode, nr_free);
393 set_page_dirty(dn->node_page);
396 dn->ofs_in_node = ofs;
398 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
399 dn->ofs_in_node, nr_free);
403 void truncate_data_blocks(struct dnode_of_data *dn)
405 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
408 static void truncate_partial_data_page(struct inode *inode, u64 from)
410 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
413 if (f2fs_has_inline_data(inode))
414 return truncate_inline_data(inode, from);
419 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false);
424 if (unlikely(!PageUptodate(page) ||
425 page->mapping != inode->i_mapping))
428 f2fs_wait_on_page_writeback(page, DATA);
429 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
430 set_page_dirty(page);
433 f2fs_put_page(page, 1);
436 int truncate_blocks(struct inode *inode, u64 from, bool lock)
438 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
439 unsigned int blocksize = inode->i_sb->s_blocksize;
440 struct dnode_of_data dn;
442 int count = 0, err = 0;
444 trace_f2fs_truncate_blocks_enter(inode, from);
446 if (f2fs_has_inline_data(inode))
449 free_from = (pgoff_t)
450 ((from + blocksize - 1) >> (sbi->log_blocksize));
455 set_new_dnode(&dn, inode, NULL, NULL, 0);
456 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
462 trace_f2fs_truncate_blocks_exit(inode, err);
466 count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
468 count -= dn.ofs_in_node;
469 f2fs_bug_on(sbi, count < 0);
471 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
472 truncate_data_blocks_range(&dn, count);
478 err = truncate_inode_blocks(inode, free_from);
482 /* lastly zero out the first data page */
483 truncate_partial_data_page(inode, from);
485 trace_f2fs_truncate_blocks_exit(inode, err);
489 void f2fs_truncate(struct inode *inode)
491 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
492 S_ISLNK(inode->i_mode)))
495 trace_f2fs_truncate(inode);
497 if (!truncate_blocks(inode, i_size_read(inode), true)) {
498 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
499 mark_inode_dirty(inode);
503 int f2fs_getattr(struct vfsmount *mnt,
504 struct dentry *dentry, struct kstat *stat)
506 struct inode *inode = dentry->d_inode;
507 generic_fillattr(inode, stat);
512 #ifdef CONFIG_F2FS_FS_POSIX_ACL
513 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
515 struct f2fs_inode_info *fi = F2FS_I(inode);
516 unsigned int ia_valid = attr->ia_valid;
518 if (ia_valid & ATTR_UID)
519 inode->i_uid = attr->ia_uid;
520 if (ia_valid & ATTR_GID)
521 inode->i_gid = attr->ia_gid;
522 if (ia_valid & ATTR_ATIME)
523 inode->i_atime = timespec_trunc(attr->ia_atime,
524 inode->i_sb->s_time_gran);
525 if (ia_valid & ATTR_MTIME)
526 inode->i_mtime = timespec_trunc(attr->ia_mtime,
527 inode->i_sb->s_time_gran);
528 if (ia_valid & ATTR_CTIME)
529 inode->i_ctime = timespec_trunc(attr->ia_ctime,
530 inode->i_sb->s_time_gran);
531 if (ia_valid & ATTR_MODE) {
532 umode_t mode = attr->ia_mode;
534 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
536 set_acl_inode(fi, mode);
540 #define __setattr_copy setattr_copy
543 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
545 struct inode *inode = dentry->d_inode;
546 struct f2fs_inode_info *fi = F2FS_I(inode);
549 err = inode_change_ok(inode, attr);
553 if ((attr->ia_valid & ATTR_SIZE) &&
554 attr->ia_size != i_size_read(inode)) {
555 err = f2fs_convert_inline_data(inode, attr->ia_size, NULL);
559 truncate_setsize(inode, attr->ia_size);
560 f2fs_truncate(inode);
561 f2fs_balance_fs(F2FS_I_SB(inode));
564 __setattr_copy(inode, attr);
566 if (attr->ia_valid & ATTR_MODE) {
567 err = posix_acl_chmod(inode, get_inode_mode(inode));
568 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
569 inode->i_mode = fi->i_acl_mode;
570 clear_inode_flag(fi, FI_ACL_MODE);
574 mark_inode_dirty(inode);
578 const struct inode_operations f2fs_file_inode_operations = {
579 .getattr = f2fs_getattr,
580 .setattr = f2fs_setattr,
581 .get_acl = f2fs_get_acl,
582 .set_acl = f2fs_set_acl,
583 #ifdef CONFIG_F2FS_FS_XATTR
584 .setxattr = generic_setxattr,
585 .getxattr = generic_getxattr,
586 .listxattr = f2fs_listxattr,
587 .removexattr = generic_removexattr,
589 .fiemap = f2fs_fiemap,
592 static void fill_zero(struct inode *inode, pgoff_t index,
593 loff_t start, loff_t len)
595 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
601 f2fs_balance_fs(sbi);
604 page = get_new_data_page(inode, NULL, index, false);
608 f2fs_wait_on_page_writeback(page, DATA);
609 zero_user(page, start, len);
610 set_page_dirty(page);
611 f2fs_put_page(page, 1);
615 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
620 for (index = pg_start; index < pg_end; index++) {
621 struct dnode_of_data dn;
623 set_new_dnode(&dn, inode, NULL, NULL, 0);
624 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
631 if (dn.data_blkaddr != NULL_ADDR)
632 truncate_data_blocks_range(&dn, 1);
638 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
640 pgoff_t pg_start, pg_end;
641 loff_t off_start, off_end;
644 ret = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1, NULL);
648 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
649 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
651 off_start = offset & (PAGE_CACHE_SIZE - 1);
652 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
654 if (pg_start == pg_end) {
655 fill_zero(inode, pg_start, off_start,
656 off_end - off_start);
659 fill_zero(inode, pg_start++, off_start,
660 PAGE_CACHE_SIZE - off_start);
662 fill_zero(inode, pg_end, 0, off_end);
664 if (pg_start < pg_end) {
665 struct address_space *mapping = inode->i_mapping;
666 loff_t blk_start, blk_end;
667 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
669 f2fs_balance_fs(sbi);
671 blk_start = pg_start << PAGE_CACHE_SHIFT;
672 blk_end = pg_end << PAGE_CACHE_SHIFT;
673 truncate_inode_pages_range(mapping, blk_start,
677 ret = truncate_hole(inode, pg_start, pg_end);
685 static int expand_inode_data(struct inode *inode, loff_t offset,
686 loff_t len, int mode)
688 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
689 pgoff_t index, pg_start, pg_end;
690 loff_t new_size = i_size_read(inode);
691 loff_t off_start, off_end;
694 f2fs_balance_fs(sbi);
696 ret = inode_newsize_ok(inode, (len + offset));
700 ret = f2fs_convert_inline_data(inode, offset + len, NULL);
704 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
705 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
707 off_start = offset & (PAGE_CACHE_SIZE - 1);
708 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
712 for (index = pg_start; index <= pg_end; index++) {
713 struct dnode_of_data dn;
715 if (index == pg_end && !off_end)
718 set_new_dnode(&dn, inode, NULL, NULL, 0);
719 ret = f2fs_reserve_block(&dn, index);
723 if (pg_start == pg_end)
724 new_size = offset + len;
725 else if (index == pg_start && off_start)
726 new_size = (index + 1) << PAGE_CACHE_SHIFT;
727 else if (index == pg_end)
728 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
730 new_size += PAGE_CACHE_SIZE;
733 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
734 i_size_read(inode) < new_size) {
735 i_size_write(inode, new_size);
736 mark_inode_dirty(inode);
737 update_inode_page(inode);
744 static long f2fs_fallocate(struct file *file, int mode,
745 loff_t offset, loff_t len)
747 struct inode *inode = file_inode(file);
750 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
753 mutex_lock(&inode->i_mutex);
755 if (mode & FALLOC_FL_PUNCH_HOLE)
756 ret = punch_hole(inode, offset, len);
758 ret = expand_inode_data(inode, offset, len, mode);
761 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
762 mark_inode_dirty(inode);
765 mutex_unlock(&inode->i_mutex);
767 trace_f2fs_fallocate(inode, mode, offset, len, ret);
771 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
772 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
774 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
778 else if (S_ISREG(mode))
779 return flags & F2FS_REG_FLMASK;
781 return flags & F2FS_OTHER_FLMASK;
784 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
786 struct inode *inode = file_inode(filp);
787 struct f2fs_inode_info *fi = F2FS_I(inode);
792 case F2FS_IOC_GETFLAGS:
793 flags = fi->i_flags & FS_FL_USER_VISIBLE;
794 return put_user(flags, (int __user *) arg);
795 case F2FS_IOC_SETFLAGS:
797 unsigned int oldflags;
799 ret = mnt_want_write_file(filp);
803 if (!inode_owner_or_capable(inode)) {
808 if (get_user(flags, (int __user *) arg)) {
813 flags = f2fs_mask_flags(inode->i_mode, flags);
815 mutex_lock(&inode->i_mutex);
817 oldflags = fi->i_flags;
819 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
820 if (!capable(CAP_LINUX_IMMUTABLE)) {
821 mutex_unlock(&inode->i_mutex);
827 flags = flags & FS_FL_USER_MODIFIABLE;
828 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
830 mutex_unlock(&inode->i_mutex);
832 f2fs_set_inode_flags(inode);
833 inode->i_ctime = CURRENT_TIME;
834 mark_inode_dirty(inode);
836 mnt_drop_write_file(filp);
845 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
848 case F2FS_IOC32_GETFLAGS:
849 cmd = F2FS_IOC_GETFLAGS;
851 case F2FS_IOC32_SETFLAGS:
852 cmd = F2FS_IOC_SETFLAGS;
857 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
861 const struct file_operations f2fs_file_operations = {
862 .llseek = f2fs_llseek,
863 .read = new_sync_read,
864 .write = new_sync_write,
865 .read_iter = generic_file_read_iter,
866 .write_iter = generic_file_write_iter,
867 .open = generic_file_open,
868 .mmap = f2fs_file_mmap,
869 .fsync = f2fs_sync_file,
870 .fallocate = f2fs_fallocate,
871 .unlocked_ioctl = f2fs_ioctl,
873 .compat_ioctl = f2fs_compat_ioctl,
875 .splice_read = generic_file_splice_read,
876 .splice_write = iter_file_splice_write,