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>
28 #include <trace/events/f2fs.h>
30 static int f2fs_vm_page_mkwrite(struct vm_area_struct
*vma
,
33 struct page
*page
= vmf
->page
;
34 struct inode
*inode
= file_inode(vma
->vm_file
);
35 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
36 struct dnode_of_data dn
;
41 sb_start_pagefault(inode
->i_sb
);
43 /* block allocation */
45 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
46 err
= f2fs_reserve_block(&dn
, page
->index
);
51 file_update_time(vma
->vm_file
);
53 if (unlikely(page
->mapping
!= inode
->i_mapping
||
54 page_offset(page
) > i_size_read(inode
) ||
55 !PageUptodate(page
))) {
62 * check to see if the page is mapped already (no holes)
64 if (PageMappedToDisk(page
))
67 /* page is wholly or partially inside EOF */
68 if (((page
->index
+ 1) << PAGE_CACHE_SHIFT
) > i_size_read(inode
)) {
70 offset
= i_size_read(inode
) & ~PAGE_CACHE_MASK
;
71 zero_user_segment(page
, offset
, PAGE_CACHE_SIZE
);
74 SetPageUptodate(page
);
76 trace_f2fs_vm_page_mkwrite(page
, DATA
);
79 wait_on_page_writeback(page
);
81 sb_end_pagefault(inode
->i_sb
);
82 return block_page_mkwrite_return(err
);
85 static const struct vm_operations_struct f2fs_file_vm_ops
= {
86 .fault
= filemap_fault
,
87 .page_mkwrite
= f2fs_vm_page_mkwrite
,
88 .remap_pages
= generic_file_remap_pages
,
91 static int get_parent_ino(struct inode
*inode
, nid_t
*pino
)
93 struct dentry
*dentry
;
96 dentry
= d_find_any_alias(inode
);
101 if (update_dent_inode(inode
, &dentry
->d_name
)) {
106 *pino
= parent_ino(dentry
);
111 int f2fs_sync_file(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
113 struct inode
*inode
= file
->f_mapping
->host
;
114 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
116 bool need_cp
= false;
117 struct writeback_control wbc
= {
118 .sync_mode
= WB_SYNC_NONE
,
119 .nr_to_write
= LONG_MAX
,
123 if (unlikely(f2fs_readonly(inode
->i_sb
)))
126 trace_f2fs_sync_file_enter(inode
);
127 ret
= filemap_write_and_wait_range(inode
->i_mapping
, start
, end
);
129 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
133 /* guarantee free sections for fsync */
134 f2fs_balance_fs(sbi
);
136 mutex_lock(&inode
->i_mutex
);
139 * Both of fdatasync() and fsync() are able to be recovered from
142 if (!S_ISREG(inode
->i_mode
) || inode
->i_nlink
!= 1)
144 else if (file_wrong_pino(inode
))
146 else if (!space_for_roll_forward(sbi
))
148 else if (!is_checkpointed_node(sbi
, F2FS_I(inode
)->i_pino
))
150 else if (F2FS_I(inode
)->xattr_ver
== cur_cp_version(F2FS_CKPT(sbi
)))
156 F2FS_I(inode
)->xattr_ver
= 0;
158 /* all the dirty node pages should be flushed for POR */
159 ret
= f2fs_sync_fs(inode
->i_sb
, 1);
160 if (file_wrong_pino(inode
) && inode
->i_nlink
== 1 &&
161 get_parent_ino(inode
, &pino
)) {
162 F2FS_I(inode
)->i_pino
= pino
;
163 file_got_pino(inode
);
164 mark_inode_dirty_sync(inode
);
165 ret
= f2fs_write_inode(inode
, NULL
);
170 /* if there is no written node page, write its inode page */
171 while (!sync_node_pages(sbi
, inode
->i_ino
, &wbc
)) {
172 mark_inode_dirty_sync(inode
);
173 ret
= f2fs_write_inode(inode
, NULL
);
177 ret
= wait_on_node_pages_writeback(sbi
, inode
->i_ino
);
180 ret
= blkdev_issue_flush(inode
->i_sb
->s_bdev
, GFP_KERNEL
, NULL
);
183 mutex_unlock(&inode
->i_mutex
);
184 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
188 static int f2fs_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
191 vma
->vm_ops
= &f2fs_file_vm_ops
;
195 int truncate_data_blocks_range(struct dnode_of_data
*dn
, int count
)
197 int nr_free
= 0, ofs
= dn
->ofs_in_node
;
198 struct f2fs_sb_info
*sbi
= F2FS_SB(dn
->inode
->i_sb
);
199 struct f2fs_node
*raw_node
;
202 raw_node
= F2FS_NODE(dn
->node_page
);
203 addr
= blkaddr_in_node(raw_node
) + ofs
;
205 for (; count
> 0; count
--, addr
++, dn
->ofs_in_node
++) {
206 block_t blkaddr
= le32_to_cpu(*addr
);
207 if (blkaddr
== NULL_ADDR
)
210 update_extent_cache(NULL_ADDR
, dn
);
211 invalidate_blocks(sbi
, blkaddr
);
215 dec_valid_block_count(sbi
, dn
->inode
, nr_free
);
216 set_page_dirty(dn
->node_page
);
219 dn
->ofs_in_node
= ofs
;
221 trace_f2fs_truncate_data_blocks_range(dn
->inode
, dn
->nid
,
222 dn
->ofs_in_node
, nr_free
);
226 void truncate_data_blocks(struct dnode_of_data
*dn
)
228 truncate_data_blocks_range(dn
, ADDRS_PER_BLOCK
);
231 static void truncate_partial_data_page(struct inode
*inode
, u64 from
)
233 unsigned offset
= from
& (PAGE_CACHE_SIZE
- 1);
239 page
= find_data_page(inode
, from
>> PAGE_CACHE_SHIFT
, false);
244 if (unlikely(page
->mapping
!= inode
->i_mapping
)) {
245 f2fs_put_page(page
, 1);
248 wait_on_page_writeback(page
);
249 zero_user(page
, offset
, PAGE_CACHE_SIZE
- offset
);
250 set_page_dirty(page
);
251 f2fs_put_page(page
, 1);
254 int truncate_blocks(struct inode
*inode
, u64 from
)
256 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
257 unsigned int blocksize
= inode
->i_sb
->s_blocksize
;
258 struct dnode_of_data dn
;
260 int count
= 0, err
= 0;
262 trace_f2fs_truncate_blocks_enter(inode
, from
);
264 if (f2fs_has_inline_data(inode
))
267 free_from
= (pgoff_t
)
268 ((from
+ blocksize
- 1) >> (sbi
->log_blocksize
));
272 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
273 err
= get_dnode_of_data(&dn
, free_from
, LOOKUP_NODE
);
278 trace_f2fs_truncate_blocks_exit(inode
, err
);
282 if (IS_INODE(dn
.node_page
))
283 count
= ADDRS_PER_INODE(F2FS_I(inode
));
285 count
= ADDRS_PER_BLOCK
;
287 count
-= dn
.ofs_in_node
;
288 f2fs_bug_on(count
< 0);
290 if (dn
.ofs_in_node
|| IS_INODE(dn
.node_page
)) {
291 truncate_data_blocks_range(&dn
, count
);
297 err
= truncate_inode_blocks(inode
, free_from
);
300 /* lastly zero out the first data page */
301 truncate_partial_data_page(inode
, from
);
303 trace_f2fs_truncate_blocks_exit(inode
, err
);
307 void f2fs_truncate(struct inode
*inode
)
309 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
310 S_ISLNK(inode
->i_mode
)))
313 trace_f2fs_truncate(inode
);
315 if (!truncate_blocks(inode
, i_size_read(inode
))) {
316 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
317 mark_inode_dirty(inode
);
321 int f2fs_getattr(struct vfsmount
*mnt
,
322 struct dentry
*dentry
, struct kstat
*stat
)
324 struct inode
*inode
= dentry
->d_inode
;
325 generic_fillattr(inode
, stat
);
330 #ifdef CONFIG_F2FS_FS_POSIX_ACL
331 static void __setattr_copy(struct inode
*inode
, const struct iattr
*attr
)
333 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
334 unsigned int ia_valid
= attr
->ia_valid
;
336 if (ia_valid
& ATTR_UID
)
337 inode
->i_uid
= attr
->ia_uid
;
338 if (ia_valid
& ATTR_GID
)
339 inode
->i_gid
= attr
->ia_gid
;
340 if (ia_valid
& ATTR_ATIME
)
341 inode
->i_atime
= timespec_trunc(attr
->ia_atime
,
342 inode
->i_sb
->s_time_gran
);
343 if (ia_valid
& ATTR_MTIME
)
344 inode
->i_mtime
= timespec_trunc(attr
->ia_mtime
,
345 inode
->i_sb
->s_time_gran
);
346 if (ia_valid
& ATTR_CTIME
)
347 inode
->i_ctime
= timespec_trunc(attr
->ia_ctime
,
348 inode
->i_sb
->s_time_gran
);
349 if (ia_valid
& ATTR_MODE
) {
350 umode_t mode
= attr
->ia_mode
;
352 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
354 set_acl_inode(fi
, mode
);
358 #define __setattr_copy setattr_copy
361 int f2fs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
363 struct inode
*inode
= dentry
->d_inode
;
364 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
367 err
= inode_change_ok(inode
, attr
);
371 if ((attr
->ia_valid
& ATTR_SIZE
) &&
372 attr
->ia_size
!= i_size_read(inode
)) {
373 err
= f2fs_convert_inline_data(inode
, attr
->ia_size
);
377 truncate_setsize(inode
, attr
->ia_size
);
378 f2fs_truncate(inode
);
379 f2fs_balance_fs(F2FS_SB(inode
->i_sb
));
382 __setattr_copy(inode
, attr
);
384 if (attr
->ia_valid
& ATTR_MODE
) {
385 err
= f2fs_acl_chmod(inode
);
386 if (err
|| is_inode_flag_set(fi
, FI_ACL_MODE
)) {
387 inode
->i_mode
= fi
->i_acl_mode
;
388 clear_inode_flag(fi
, FI_ACL_MODE
);
392 mark_inode_dirty(inode
);
396 const struct inode_operations f2fs_file_inode_operations
= {
397 .getattr
= f2fs_getattr
,
398 .setattr
= f2fs_setattr
,
399 .get_acl
= f2fs_get_acl
,
400 #ifdef CONFIG_F2FS_FS_XATTR
401 .setxattr
= generic_setxattr
,
402 .getxattr
= generic_getxattr
,
403 .listxattr
= f2fs_listxattr
,
404 .removexattr
= generic_removexattr
,
408 static void fill_zero(struct inode
*inode
, pgoff_t index
,
409 loff_t start
, loff_t len
)
411 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
417 f2fs_balance_fs(sbi
);
420 page
= get_new_data_page(inode
, NULL
, index
, false);
424 wait_on_page_writeback(page
);
425 zero_user(page
, start
, len
);
426 set_page_dirty(page
);
427 f2fs_put_page(page
, 1);
431 int truncate_hole(struct inode
*inode
, pgoff_t pg_start
, pgoff_t pg_end
)
436 for (index
= pg_start
; index
< pg_end
; index
++) {
437 struct dnode_of_data dn
;
439 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
440 err
= get_dnode_of_data(&dn
, index
, LOOKUP_NODE
);
447 if (dn
.data_blkaddr
!= NULL_ADDR
)
448 truncate_data_blocks_range(&dn
, 1);
454 static int punch_hole(struct inode
*inode
, loff_t offset
, loff_t len
)
456 pgoff_t pg_start
, pg_end
;
457 loff_t off_start
, off_end
;
460 ret
= f2fs_convert_inline_data(inode
, MAX_INLINE_DATA
+ 1);
464 pg_start
= ((unsigned long long) offset
) >> PAGE_CACHE_SHIFT
;
465 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_CACHE_SHIFT
;
467 off_start
= offset
& (PAGE_CACHE_SIZE
- 1);
468 off_end
= (offset
+ len
) & (PAGE_CACHE_SIZE
- 1);
470 if (pg_start
== pg_end
) {
471 fill_zero(inode
, pg_start
, off_start
,
472 off_end
- off_start
);
475 fill_zero(inode
, pg_start
++, off_start
,
476 PAGE_CACHE_SIZE
- off_start
);
478 fill_zero(inode
, pg_end
, 0, off_end
);
480 if (pg_start
< pg_end
) {
481 struct address_space
*mapping
= inode
->i_mapping
;
482 loff_t blk_start
, blk_end
;
483 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
485 f2fs_balance_fs(sbi
);
487 blk_start
= pg_start
<< PAGE_CACHE_SHIFT
;
488 blk_end
= pg_end
<< PAGE_CACHE_SHIFT
;
489 truncate_inode_pages_range(mapping
, blk_start
,
493 ret
= truncate_hole(inode
, pg_start
, pg_end
);
501 static int expand_inode_data(struct inode
*inode
, loff_t offset
,
502 loff_t len
, int mode
)
504 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
505 pgoff_t index
, pg_start
, pg_end
;
506 loff_t new_size
= i_size_read(inode
);
507 loff_t off_start
, off_end
;
510 ret
= inode_newsize_ok(inode
, (len
+ offset
));
514 ret
= f2fs_convert_inline_data(inode
, offset
+ len
);
518 pg_start
= ((unsigned long long) offset
) >> PAGE_CACHE_SHIFT
;
519 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_CACHE_SHIFT
;
521 off_start
= offset
& (PAGE_CACHE_SIZE
- 1);
522 off_end
= (offset
+ len
) & (PAGE_CACHE_SIZE
- 1);
524 for (index
= pg_start
; index
<= pg_end
; index
++) {
525 struct dnode_of_data dn
;
528 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
529 ret
= f2fs_reserve_block(&dn
, index
);
534 if (pg_start
== pg_end
)
535 new_size
= offset
+ len
;
536 else if (index
== pg_start
&& off_start
)
537 new_size
= (index
+ 1) << PAGE_CACHE_SHIFT
;
538 else if (index
== pg_end
)
539 new_size
= (index
<< PAGE_CACHE_SHIFT
) + off_end
;
541 new_size
+= PAGE_CACHE_SIZE
;
544 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
545 i_size_read(inode
) < new_size
) {
546 i_size_write(inode
, new_size
);
547 mark_inode_dirty(inode
);
553 static long f2fs_fallocate(struct file
*file
, int mode
,
554 loff_t offset
, loff_t len
)
556 struct inode
*inode
= file_inode(file
);
559 if (mode
& ~(FALLOC_FL_KEEP_SIZE
| FALLOC_FL_PUNCH_HOLE
))
562 if (mode
& FALLOC_FL_PUNCH_HOLE
)
563 ret
= punch_hole(inode
, offset
, len
);
565 ret
= expand_inode_data(inode
, offset
, len
, mode
);
568 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
569 mark_inode_dirty(inode
);
571 trace_f2fs_fallocate(inode
, mode
, offset
, len
, ret
);
575 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
576 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
578 static inline __u32
f2fs_mask_flags(umode_t mode
, __u32 flags
)
582 else if (S_ISREG(mode
))
583 return flags
& F2FS_REG_FLMASK
;
585 return flags
& F2FS_OTHER_FLMASK
;
588 long f2fs_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
590 struct inode
*inode
= file_inode(filp
);
591 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
596 case F2FS_IOC_GETFLAGS
:
597 flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
598 return put_user(flags
, (int __user
*) arg
);
599 case F2FS_IOC_SETFLAGS
:
601 unsigned int oldflags
;
603 ret
= mnt_want_write_file(filp
);
607 if (!inode_owner_or_capable(inode
)) {
612 if (get_user(flags
, (int __user
*) arg
)) {
617 flags
= f2fs_mask_flags(inode
->i_mode
, flags
);
619 mutex_lock(&inode
->i_mutex
);
621 oldflags
= fi
->i_flags
;
623 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
624 if (!capable(CAP_LINUX_IMMUTABLE
)) {
625 mutex_unlock(&inode
->i_mutex
);
631 flags
= flags
& FS_FL_USER_MODIFIABLE
;
632 flags
|= oldflags
& ~FS_FL_USER_MODIFIABLE
;
634 mutex_unlock(&inode
->i_mutex
);
636 f2fs_set_inode_flags(inode
);
637 inode
->i_ctime
= CURRENT_TIME
;
638 mark_inode_dirty(inode
);
640 mnt_drop_write_file(filp
);
649 long f2fs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
652 case F2FS_IOC32_GETFLAGS
:
653 cmd
= F2FS_IOC_GETFLAGS
;
655 case F2FS_IOC32_SETFLAGS
:
656 cmd
= F2FS_IOC_SETFLAGS
;
661 return f2fs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));
665 const struct file_operations f2fs_file_operations
= {
666 .llseek
= generic_file_llseek
,
667 .read
= do_sync_read
,
668 .write
= do_sync_write
,
669 .aio_read
= generic_file_aio_read
,
670 .aio_write
= generic_file_aio_write
,
671 .open
= generic_file_open
,
672 .mmap
= f2fs_file_mmap
,
673 .fsync
= f2fs_sync_file
,
674 .fallocate
= f2fs_fallocate
,
675 .unlocked_ioctl
= f2fs_ioctl
,
677 .compat_ioctl
= f2fs_compat_ioctl
,
679 .splice_read
= generic_file_splice_read
,
680 .splice_write
= generic_file_splice_write
,