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_SB(inode->i_sb);
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 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
120 struct inode *inode = file->f_mapping->host;
121 struct f2fs_inode_info *fi = F2FS_I(inode);
122 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
124 bool need_cp = false;
125 struct writeback_control wbc = {
126 .sync_mode = WB_SYNC_ALL,
127 .nr_to_write = LONG_MAX,
131 if (unlikely(f2fs_readonly(inode->i_sb)))
134 trace_f2fs_sync_file_enter(inode);
136 /* if fdatasync is triggered, let's do in-place-update */
138 set_inode_flag(fi, FI_NEED_IPU);
140 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
142 clear_inode_flag(fi, FI_NEED_IPU);
144 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
149 * if there is no written data, don't waste time to write recovery info.
151 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
152 !exist_written_data(sbi, inode->i_ino, APPEND_INO)) {
153 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
154 exist_written_data(sbi, inode->i_ino, UPDATE_INO))
159 /* guarantee free sections for fsync */
160 f2fs_balance_fs(sbi);
162 down_read(&fi->i_sem);
165 * Both of fdatasync() and fsync() are able to be recovered from
168 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
170 else if (file_wrong_pino(inode))
172 else if (!space_for_roll_forward(sbi))
174 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
176 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
184 /* all the dirty node pages should be flushed for POR */
185 ret = f2fs_sync_fs(inode->i_sb, 1);
187 down_write(&fi->i_sem);
188 F2FS_I(inode)->xattr_ver = 0;
189 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
190 get_parent_ino(inode, &pino)) {
191 F2FS_I(inode)->i_pino = pino;
192 file_got_pino(inode);
193 up_write(&fi->i_sem);
194 mark_inode_dirty_sync(inode);
195 ret = f2fs_write_inode(inode, NULL);
199 up_write(&fi->i_sem);
202 /* if there is no written node page, write its inode page */
203 while (!sync_node_pages(sbi, inode->i_ino, &wbc)) {
204 if (fsync_mark_done(sbi, inode->i_ino))
206 mark_inode_dirty_sync(inode);
207 ret = f2fs_write_inode(inode, NULL);
211 ret = wait_on_node_pages_writeback(sbi, inode->i_ino);
215 /* once recovery info is written, don't need to tack this */
216 remove_dirty_inode(sbi, inode->i_ino, APPEND_INO);
217 clear_inode_flag(fi, FI_APPEND_WRITE);
219 remove_dirty_inode(sbi, inode->i_ino, UPDATE_INO);
220 clear_inode_flag(fi, FI_UPDATE_WRITE);
221 ret = f2fs_issue_flush(F2FS_SB(inode->i_sb));
224 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
228 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
229 pgoff_t pgofs, int whence)
234 if (whence != SEEK_DATA)
237 /* find first dirty page index */
238 pagevec_init(&pvec, 0);
239 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
240 PAGECACHE_TAG_DIRTY, 1);
241 pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
242 pagevec_release(&pvec);
246 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
251 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
252 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
256 if (blkaddr == NULL_ADDR)
263 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
265 struct inode *inode = file->f_mapping->host;
266 loff_t maxbytes = inode->i_sb->s_maxbytes;
267 struct dnode_of_data dn;
268 pgoff_t pgofs, end_offset, dirty;
269 loff_t data_ofs = offset;
273 mutex_lock(&inode->i_mutex);
275 isize = i_size_read(inode);
279 /* handle inline data case */
280 if (f2fs_has_inline_data(inode)) {
281 if (whence == SEEK_HOLE)
286 pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
288 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
290 for (; data_ofs < isize; data_ofs = pgofs << PAGE_CACHE_SHIFT) {
291 set_new_dnode(&dn, inode, NULL, NULL, 0);
292 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
293 if (err && err != -ENOENT) {
295 } else if (err == -ENOENT) {
296 /* direct node does not exists */
297 if (whence == SEEK_DATA) {
298 pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
306 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
308 /* find data/hole in dnode block */
309 for (; dn.ofs_in_node < end_offset;
310 dn.ofs_in_node++, pgofs++,
311 data_ofs = pgofs << PAGE_CACHE_SHIFT) {
313 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
315 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
323 if (whence == SEEK_DATA)
326 if (whence == SEEK_HOLE && data_ofs > isize)
328 mutex_unlock(&inode->i_mutex);
329 return vfs_setpos(file, data_ofs, maxbytes);
331 mutex_unlock(&inode->i_mutex);
335 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
337 struct inode *inode = file->f_mapping->host;
338 loff_t maxbytes = inode->i_sb->s_maxbytes;
344 return generic_file_llseek_size(file, offset, whence,
345 maxbytes, i_size_read(inode));
348 return f2fs_seek_block(file, offset, whence);
354 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
357 vma->vm_ops = &f2fs_file_vm_ops;
361 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
363 int nr_free = 0, ofs = dn->ofs_in_node;
364 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
365 struct f2fs_node *raw_node;
368 raw_node = F2FS_NODE(dn->node_page);
369 addr = blkaddr_in_node(raw_node) + ofs;
371 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
372 block_t blkaddr = le32_to_cpu(*addr);
373 if (blkaddr == NULL_ADDR)
376 update_extent_cache(NULL_ADDR, dn);
377 invalidate_blocks(sbi, blkaddr);
381 dec_valid_block_count(sbi, dn->inode, nr_free);
382 set_page_dirty(dn->node_page);
385 dn->ofs_in_node = ofs;
387 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
388 dn->ofs_in_node, nr_free);
392 void truncate_data_blocks(struct dnode_of_data *dn)
394 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
397 static void truncate_partial_data_page(struct inode *inode, u64 from)
399 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
402 if (f2fs_has_inline_data(inode))
403 return truncate_inline_data(inode, from);
408 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false);
413 if (unlikely(!PageUptodate(page) ||
414 page->mapping != inode->i_mapping))
417 f2fs_wait_on_page_writeback(page, DATA);
418 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
419 set_page_dirty(page);
422 f2fs_put_page(page, 1);
425 int truncate_blocks(struct inode *inode, u64 from)
427 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
428 unsigned int blocksize = inode->i_sb->s_blocksize;
429 struct dnode_of_data dn;
431 int count = 0, err = 0;
433 trace_f2fs_truncate_blocks_enter(inode, from);
435 if (f2fs_has_inline_data(inode))
438 free_from = (pgoff_t)
439 ((from + blocksize - 1) >> (sbi->log_blocksize));
443 set_new_dnode(&dn, inode, NULL, NULL, 0);
444 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
449 trace_f2fs_truncate_blocks_exit(inode, err);
453 count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
455 count -= dn.ofs_in_node;
456 f2fs_bug_on(count < 0);
458 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
459 truncate_data_blocks_range(&dn, count);
465 err = truncate_inode_blocks(inode, free_from);
468 /* lastly zero out the first data page */
469 truncate_partial_data_page(inode, from);
471 trace_f2fs_truncate_blocks_exit(inode, err);
475 void f2fs_truncate(struct inode *inode)
477 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
478 S_ISLNK(inode->i_mode)))
481 trace_f2fs_truncate(inode);
483 if (!truncate_blocks(inode, i_size_read(inode))) {
484 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
485 mark_inode_dirty(inode);
489 int f2fs_getattr(struct vfsmount *mnt,
490 struct dentry *dentry, struct kstat *stat)
492 struct inode *inode = dentry->d_inode;
493 generic_fillattr(inode, stat);
498 #ifdef CONFIG_F2FS_FS_POSIX_ACL
499 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
501 struct f2fs_inode_info *fi = F2FS_I(inode);
502 unsigned int ia_valid = attr->ia_valid;
504 if (ia_valid & ATTR_UID)
505 inode->i_uid = attr->ia_uid;
506 if (ia_valid & ATTR_GID)
507 inode->i_gid = attr->ia_gid;
508 if (ia_valid & ATTR_ATIME)
509 inode->i_atime = timespec_trunc(attr->ia_atime,
510 inode->i_sb->s_time_gran);
511 if (ia_valid & ATTR_MTIME)
512 inode->i_mtime = timespec_trunc(attr->ia_mtime,
513 inode->i_sb->s_time_gran);
514 if (ia_valid & ATTR_CTIME)
515 inode->i_ctime = timespec_trunc(attr->ia_ctime,
516 inode->i_sb->s_time_gran);
517 if (ia_valid & ATTR_MODE) {
518 umode_t mode = attr->ia_mode;
520 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
522 set_acl_inode(fi, mode);
526 #define __setattr_copy setattr_copy
529 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
531 struct inode *inode = dentry->d_inode;
532 struct f2fs_inode_info *fi = F2FS_I(inode);
535 err = inode_change_ok(inode, attr);
539 if ((attr->ia_valid & ATTR_SIZE) &&
540 attr->ia_size != i_size_read(inode)) {
541 err = f2fs_convert_inline_data(inode, attr->ia_size, NULL);
545 truncate_setsize(inode, attr->ia_size);
546 f2fs_truncate(inode);
547 f2fs_balance_fs(F2FS_SB(inode->i_sb));
550 __setattr_copy(inode, attr);
552 if (attr->ia_valid & ATTR_MODE) {
553 err = posix_acl_chmod(inode, get_inode_mode(inode));
554 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
555 inode->i_mode = fi->i_acl_mode;
556 clear_inode_flag(fi, FI_ACL_MODE);
560 mark_inode_dirty(inode);
564 const struct inode_operations f2fs_file_inode_operations = {
565 .getattr = f2fs_getattr,
566 .setattr = f2fs_setattr,
567 .get_acl = f2fs_get_acl,
568 .set_acl = f2fs_set_acl,
569 #ifdef CONFIG_F2FS_FS_XATTR
570 .setxattr = generic_setxattr,
571 .getxattr = generic_getxattr,
572 .listxattr = f2fs_listxattr,
573 .removexattr = generic_removexattr,
575 .fiemap = f2fs_fiemap,
578 static void fill_zero(struct inode *inode, pgoff_t index,
579 loff_t start, loff_t len)
581 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
587 f2fs_balance_fs(sbi);
590 page = get_new_data_page(inode, NULL, index, false);
594 f2fs_wait_on_page_writeback(page, DATA);
595 zero_user(page, start, len);
596 set_page_dirty(page);
597 f2fs_put_page(page, 1);
601 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
606 for (index = pg_start; index < pg_end; index++) {
607 struct dnode_of_data dn;
609 set_new_dnode(&dn, inode, NULL, NULL, 0);
610 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
617 if (dn.data_blkaddr != NULL_ADDR)
618 truncate_data_blocks_range(&dn, 1);
624 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
626 pgoff_t pg_start, pg_end;
627 loff_t off_start, off_end;
630 ret = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1, NULL);
634 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
635 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
637 off_start = offset & (PAGE_CACHE_SIZE - 1);
638 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
640 if (pg_start == pg_end) {
641 fill_zero(inode, pg_start, off_start,
642 off_end - off_start);
645 fill_zero(inode, pg_start++, off_start,
646 PAGE_CACHE_SIZE - off_start);
648 fill_zero(inode, pg_end, 0, off_end);
650 if (pg_start < pg_end) {
651 struct address_space *mapping = inode->i_mapping;
652 loff_t blk_start, blk_end;
653 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
655 f2fs_balance_fs(sbi);
657 blk_start = pg_start << PAGE_CACHE_SHIFT;
658 blk_end = pg_end << PAGE_CACHE_SHIFT;
659 truncate_inode_pages_range(mapping, blk_start,
663 ret = truncate_hole(inode, pg_start, pg_end);
671 static int expand_inode_data(struct inode *inode, loff_t offset,
672 loff_t len, int mode)
674 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
675 pgoff_t index, pg_start, pg_end;
676 loff_t new_size = i_size_read(inode);
677 loff_t off_start, off_end;
680 f2fs_balance_fs(sbi);
682 ret = inode_newsize_ok(inode, (len + offset));
686 ret = f2fs_convert_inline_data(inode, offset + len, NULL);
690 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
691 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
693 off_start = offset & (PAGE_CACHE_SIZE - 1);
694 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
698 for (index = pg_start; index <= pg_end; index++) {
699 struct dnode_of_data dn;
701 if (index == pg_end && !off_end)
704 set_new_dnode(&dn, inode, NULL, NULL, 0);
705 ret = f2fs_reserve_block(&dn, index);
709 if (pg_start == pg_end)
710 new_size = offset + len;
711 else if (index == pg_start && off_start)
712 new_size = (index + 1) << PAGE_CACHE_SHIFT;
713 else if (index == pg_end)
714 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
716 new_size += PAGE_CACHE_SIZE;
719 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
720 i_size_read(inode) < new_size) {
721 i_size_write(inode, new_size);
722 mark_inode_dirty(inode);
723 update_inode_page(inode);
730 static long f2fs_fallocate(struct file *file, int mode,
731 loff_t offset, loff_t len)
733 struct inode *inode = file_inode(file);
736 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
739 mutex_lock(&inode->i_mutex);
741 if (mode & FALLOC_FL_PUNCH_HOLE)
742 ret = punch_hole(inode, offset, len);
744 ret = expand_inode_data(inode, offset, len, mode);
747 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
748 mark_inode_dirty(inode);
751 mutex_unlock(&inode->i_mutex);
753 trace_f2fs_fallocate(inode, mode, offset, len, ret);
757 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
758 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
760 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
764 else if (S_ISREG(mode))
765 return flags & F2FS_REG_FLMASK;
767 return flags & F2FS_OTHER_FLMASK;
770 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
772 struct inode *inode = file_inode(filp);
773 struct f2fs_inode_info *fi = F2FS_I(inode);
778 case F2FS_IOC_GETFLAGS:
779 flags = fi->i_flags & FS_FL_USER_VISIBLE;
780 return put_user(flags, (int __user *) arg);
781 case F2FS_IOC_SETFLAGS:
783 unsigned int oldflags;
785 ret = mnt_want_write_file(filp);
789 if (!inode_owner_or_capable(inode)) {
794 if (get_user(flags, (int __user *) arg)) {
799 flags = f2fs_mask_flags(inode->i_mode, flags);
801 mutex_lock(&inode->i_mutex);
803 oldflags = fi->i_flags;
805 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
806 if (!capable(CAP_LINUX_IMMUTABLE)) {
807 mutex_unlock(&inode->i_mutex);
813 flags = flags & FS_FL_USER_MODIFIABLE;
814 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
816 mutex_unlock(&inode->i_mutex);
818 f2fs_set_inode_flags(inode);
819 inode->i_ctime = CURRENT_TIME;
820 mark_inode_dirty(inode);
822 mnt_drop_write_file(filp);
831 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
834 case F2FS_IOC32_GETFLAGS:
835 cmd = F2FS_IOC_GETFLAGS;
837 case F2FS_IOC32_SETFLAGS:
838 cmd = F2FS_IOC_SETFLAGS;
843 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
847 const struct file_operations f2fs_file_operations = {
848 .llseek = f2fs_llseek,
849 .read = new_sync_read,
850 .write = new_sync_write,
851 .read_iter = generic_file_read_iter,
852 .write_iter = generic_file_write_iter,
853 .open = generic_file_open,
854 .mmap = f2fs_file_mmap,
855 .fsync = f2fs_sync_file,
856 .fallocate = f2fs_fallocate,
857 .unlocked_ioctl = f2fs_ioctl,
859 .compat_ioctl = f2fs_compat_ioctl,
861 .splice_read = generic_file_splice_read,
862 .splice_write = iter_file_splice_write,