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/falloc.h>
17 #include <linux/types.h>
18 #include <linux/uaccess.h>
19 #include <linux/mount.h>
27 static int f2fs_vm_page_mkwrite(struct vm_area_struct
*vma
,
30 struct page
*page
= vmf
->page
;
31 struct inode
*inode
= vma
->vm_file
->f_path
.dentry
->d_inode
;
32 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
33 struct page
*node_page
;
35 struct dnode_of_data dn
;
40 sb_start_pagefault(inode
->i_sb
);
42 mutex_lock_op(sbi
, DATA_NEW
);
44 /* block allocation */
45 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
46 err
= get_dnode_of_data(&dn
, page
->index
, 0);
48 mutex_unlock_op(sbi
, DATA_NEW
);
52 old_blk_addr
= dn
.data_blkaddr
;
53 node_page
= dn
.node_page
;
55 if (old_blk_addr
== NULL_ADDR
) {
56 err
= reserve_new_block(&dn
);
59 mutex_unlock_op(sbi
, DATA_NEW
);
65 mutex_unlock_op(sbi
, DATA_NEW
);
68 if (page
->mapping
!= inode
->i_mapping
||
69 page_offset(page
) >= i_size_read(inode
) ||
70 !PageUptodate(page
)) {
77 * check to see if the page is mapped already (no holes)
79 if (PageMappedToDisk(page
))
83 wait_on_page_writeback(page
);
85 /* page is wholly or partially inside EOF */
86 if (((page
->index
+ 1) << PAGE_CACHE_SHIFT
) > i_size_read(inode
)) {
88 offset
= i_size_read(inode
) & ~PAGE_CACHE_MASK
;
89 zero_user_segment(page
, offset
, PAGE_CACHE_SIZE
);
92 SetPageUptodate(page
);
94 file_update_time(vma
->vm_file
);
96 sb_end_pagefault(inode
->i_sb
);
97 return block_page_mkwrite_return(err
);
100 static const struct vm_operations_struct f2fs_file_vm_ops
= {
101 .fault
= filemap_fault
,
102 .page_mkwrite
= f2fs_vm_page_mkwrite
,
105 static int need_to_sync_dir(struct f2fs_sb_info
*sbi
, struct inode
*inode
)
107 struct dentry
*dentry
;
110 inode
= igrab(inode
);
111 dentry
= d_find_any_alias(inode
);
116 pino
= dentry
->d_parent
->d_inode
->i_ino
;
119 return !is_checkpointed_node(sbi
, pino
);
122 int f2fs_sync_file(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
124 struct inode
*inode
= file
->f_mapping
->host
;
125 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
126 unsigned long long cur_version
;
128 bool need_cp
= false;
129 struct writeback_control wbc
= {
130 .sync_mode
= WB_SYNC_ALL
,
131 .nr_to_write
= LONG_MAX
,
135 ret
= filemap_write_and_wait_range(inode
->i_mapping
, start
, end
);
139 mutex_lock(&inode
->i_mutex
);
141 if (inode
->i_sb
->s_flags
& MS_RDONLY
)
143 if (datasync
&& !(inode
->i_state
& I_DIRTY_DATASYNC
))
146 mutex_lock(&sbi
->cp_mutex
);
147 cur_version
= le64_to_cpu(F2FS_CKPT(sbi
)->checkpoint_ver
);
148 mutex_unlock(&sbi
->cp_mutex
);
150 if (F2FS_I(inode
)->data_version
!= cur_version
&&
151 !(inode
->i_state
& I_DIRTY
))
153 F2FS_I(inode
)->data_version
--;
155 if (!S_ISREG(inode
->i_mode
) || inode
->i_nlink
!= 1)
157 if (is_inode_flag_set(F2FS_I(inode
), FI_NEED_CP
))
159 if (!space_for_roll_forward(sbi
))
161 if (need_to_sync_dir(sbi
, inode
))
164 f2fs_write_inode(inode
, NULL
);
167 /* all the dirty node pages should be flushed for POR */
168 ret
= f2fs_sync_fs(inode
->i_sb
, 1);
169 clear_inode_flag(F2FS_I(inode
), FI_NEED_CP
);
171 while (sync_node_pages(sbi
, inode
->i_ino
, &wbc
) == 0)
172 f2fs_write_inode(inode
, NULL
);
173 filemap_fdatawait_range(sbi
->node_inode
->i_mapping
,
177 mutex_unlock(&inode
->i_mutex
);
181 static int f2fs_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
184 vma
->vm_ops
= &f2fs_file_vm_ops
;
188 static int truncate_data_blocks_range(struct dnode_of_data
*dn
, int count
)
190 int nr_free
= 0, ofs
= dn
->ofs_in_node
;
191 struct f2fs_sb_info
*sbi
= F2FS_SB(dn
->inode
->i_sb
);
192 struct f2fs_node
*raw_node
;
195 raw_node
= page_address(dn
->node_page
);
196 addr
= blkaddr_in_node(raw_node
) + ofs
;
198 for ( ; count
> 0; count
--, addr
++, dn
->ofs_in_node
++) {
199 block_t blkaddr
= le32_to_cpu(*addr
);
200 if (blkaddr
== NULL_ADDR
)
203 update_extent_cache(NULL_ADDR
, dn
);
204 invalidate_blocks(sbi
, blkaddr
);
205 dec_valid_block_count(sbi
, dn
->inode
, 1);
209 set_page_dirty(dn
->node_page
);
212 dn
->ofs_in_node
= ofs
;
216 void truncate_data_blocks(struct dnode_of_data
*dn
)
218 truncate_data_blocks_range(dn
, ADDRS_PER_BLOCK
);
221 static void truncate_partial_data_page(struct inode
*inode
, u64 from
)
223 unsigned offset
= from
& (PAGE_CACHE_SIZE
- 1);
229 page
= find_data_page(inode
, from
>> PAGE_CACHE_SHIFT
);
234 wait_on_page_writeback(page
);
235 zero_user(page
, offset
, PAGE_CACHE_SIZE
- offset
);
236 set_page_dirty(page
);
237 f2fs_put_page(page
, 1);
240 static int truncate_blocks(struct inode
*inode
, u64 from
)
242 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
243 unsigned int blocksize
= inode
->i_sb
->s_blocksize
;
244 struct dnode_of_data dn
;
249 free_from
= (pgoff_t
)
250 ((from
+ blocksize
- 1) >> (sbi
->log_blocksize
));
252 mutex_lock_op(sbi
, DATA_TRUNC
);
254 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
255 err
= get_dnode_of_data(&dn
, free_from
, RDONLY_NODE
);
259 mutex_unlock_op(sbi
, DATA_TRUNC
);
263 if (IS_INODE(dn
.node_page
))
264 count
= ADDRS_PER_INODE
;
266 count
= ADDRS_PER_BLOCK
;
268 count
-= dn
.ofs_in_node
;
270 if (dn
.ofs_in_node
|| IS_INODE(dn
.node_page
)) {
271 truncate_data_blocks_range(&dn
, count
);
277 err
= truncate_inode_blocks(inode
, free_from
);
278 mutex_unlock_op(sbi
, DATA_TRUNC
);
280 /* lastly zero out the first data page */
281 truncate_partial_data_page(inode
, from
);
286 void f2fs_truncate(struct inode
*inode
)
288 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
289 S_ISLNK(inode
->i_mode
)))
292 if (!truncate_blocks(inode
, i_size_read(inode
))) {
293 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
294 mark_inode_dirty(inode
);
297 f2fs_balance_fs(F2FS_SB(inode
->i_sb
));
300 static int f2fs_getattr(struct vfsmount
*mnt
,
301 struct dentry
*dentry
, struct kstat
*stat
)
303 struct inode
*inode
= dentry
->d_inode
;
304 generic_fillattr(inode
, stat
);
309 #ifdef CONFIG_F2FS_FS_POSIX_ACL
310 static void __setattr_copy(struct inode
*inode
, const struct iattr
*attr
)
312 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
313 unsigned int ia_valid
= attr
->ia_valid
;
315 if (ia_valid
& ATTR_UID
)
316 inode
->i_uid
= attr
->ia_uid
;
317 if (ia_valid
& ATTR_GID
)
318 inode
->i_gid
= attr
->ia_gid
;
319 if (ia_valid
& ATTR_ATIME
)
320 inode
->i_atime
= timespec_trunc(attr
->ia_atime
,
321 inode
->i_sb
->s_time_gran
);
322 if (ia_valid
& ATTR_MTIME
)
323 inode
->i_mtime
= timespec_trunc(attr
->ia_mtime
,
324 inode
->i_sb
->s_time_gran
);
325 if (ia_valid
& ATTR_CTIME
)
326 inode
->i_ctime
= timespec_trunc(attr
->ia_ctime
,
327 inode
->i_sb
->s_time_gran
);
328 if (ia_valid
& ATTR_MODE
) {
329 umode_t mode
= attr
->ia_mode
;
331 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
333 set_acl_inode(fi
, mode
);
337 #define __setattr_copy setattr_copy
340 int f2fs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
342 struct inode
*inode
= dentry
->d_inode
;
343 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
346 err
= inode_change_ok(inode
, attr
);
350 if ((attr
->ia_valid
& ATTR_SIZE
) &&
351 attr
->ia_size
!= i_size_read(inode
)) {
352 truncate_setsize(inode
, attr
->ia_size
);
353 f2fs_truncate(inode
);
356 __setattr_copy(inode
, attr
);
358 if (attr
->ia_valid
& ATTR_MODE
) {
359 err
= f2fs_acl_chmod(inode
);
360 if (err
|| is_inode_flag_set(fi
, FI_ACL_MODE
)) {
361 inode
->i_mode
= fi
->i_acl_mode
;
362 clear_inode_flag(fi
, FI_ACL_MODE
);
366 mark_inode_dirty(inode
);
370 const struct inode_operations f2fs_file_inode_operations
= {
371 .getattr
= f2fs_getattr
,
372 .setattr
= f2fs_setattr
,
373 .get_acl
= f2fs_get_acl
,
374 #ifdef CONFIG_F2FS_FS_XATTR
375 .setxattr
= generic_setxattr
,
376 .getxattr
= generic_getxattr
,
377 .listxattr
= f2fs_listxattr
,
378 .removexattr
= generic_removexattr
,
382 static void fill_zero(struct inode
*inode
, pgoff_t index
,
383 loff_t start
, loff_t len
)
390 page
= get_new_data_page(inode
, index
, false);
393 wait_on_page_writeback(page
);
394 zero_user(page
, start
, len
);
395 set_page_dirty(page
);
396 f2fs_put_page(page
, 1);
400 int truncate_hole(struct inode
*inode
, pgoff_t pg_start
, pgoff_t pg_end
)
405 for (index
= pg_start
; index
< pg_end
; index
++) {
406 struct dnode_of_data dn
;
407 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
409 mutex_lock_op(sbi
, DATA_TRUNC
);
410 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
411 err
= get_dnode_of_data(&dn
, index
, RDONLY_NODE
);
413 mutex_unlock_op(sbi
, DATA_TRUNC
);
419 if (dn
.data_blkaddr
!= NULL_ADDR
)
420 truncate_data_blocks_range(&dn
, 1);
422 mutex_unlock_op(sbi
, DATA_TRUNC
);
427 static int punch_hole(struct inode
*inode
, loff_t offset
, loff_t len
, int mode
)
429 pgoff_t pg_start
, pg_end
;
430 loff_t off_start
, off_end
;
433 pg_start
= ((unsigned long long) offset
) >> PAGE_CACHE_SHIFT
;
434 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_CACHE_SHIFT
;
436 off_start
= offset
& (PAGE_CACHE_SIZE
- 1);
437 off_end
= (offset
+ len
) & (PAGE_CACHE_SIZE
- 1);
439 if (pg_start
== pg_end
) {
440 fill_zero(inode
, pg_start
, off_start
,
441 off_end
- off_start
);
444 fill_zero(inode
, pg_start
++, off_start
,
445 PAGE_CACHE_SIZE
- off_start
);
447 fill_zero(inode
, pg_end
, 0, off_end
);
449 if (pg_start
< pg_end
) {
450 struct address_space
*mapping
= inode
->i_mapping
;
451 loff_t blk_start
, blk_end
;
453 blk_start
= pg_start
<< PAGE_CACHE_SHIFT
;
454 blk_end
= pg_end
<< PAGE_CACHE_SHIFT
;
455 truncate_inode_pages_range(mapping
, blk_start
,
457 ret
= truncate_hole(inode
, pg_start
, pg_end
);
461 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
462 i_size_read(inode
) <= (offset
+ len
)) {
463 i_size_write(inode
, offset
);
464 mark_inode_dirty(inode
);
470 static int expand_inode_data(struct inode
*inode
, loff_t offset
,
471 loff_t len
, int mode
)
473 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
474 pgoff_t index
, pg_start
, pg_end
;
475 loff_t new_size
= i_size_read(inode
);
476 loff_t off_start
, off_end
;
479 ret
= inode_newsize_ok(inode
, (len
+ offset
));
483 pg_start
= ((unsigned long long) offset
) >> PAGE_CACHE_SHIFT
;
484 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_CACHE_SHIFT
;
486 off_start
= offset
& (PAGE_CACHE_SIZE
- 1);
487 off_end
= (offset
+ len
) & (PAGE_CACHE_SIZE
- 1);
489 for (index
= pg_start
; index
<= pg_end
; index
++) {
490 struct dnode_of_data dn
;
492 mutex_lock_op(sbi
, DATA_NEW
);
494 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
495 ret
= get_dnode_of_data(&dn
, index
, 0);
497 mutex_unlock_op(sbi
, DATA_NEW
);
501 if (dn
.data_blkaddr
== NULL_ADDR
) {
502 ret
= reserve_new_block(&dn
);
505 mutex_unlock_op(sbi
, DATA_NEW
);
511 mutex_unlock_op(sbi
, DATA_NEW
);
513 if (pg_start
== pg_end
)
514 new_size
= offset
+ len
;
515 else if (index
== pg_start
&& off_start
)
516 new_size
= (index
+ 1) << PAGE_CACHE_SHIFT
;
517 else if (index
== pg_end
)
518 new_size
= (index
<< PAGE_CACHE_SHIFT
) + off_end
;
520 new_size
+= PAGE_CACHE_SIZE
;
523 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
524 i_size_read(inode
) < new_size
) {
525 i_size_write(inode
, new_size
);
526 mark_inode_dirty(inode
);
532 static long f2fs_fallocate(struct file
*file
, int mode
,
533 loff_t offset
, loff_t len
)
535 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
536 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
539 if (mode
& ~(FALLOC_FL_KEEP_SIZE
| FALLOC_FL_PUNCH_HOLE
))
542 if (mode
& FALLOC_FL_PUNCH_HOLE
)
543 ret
= punch_hole(inode
, offset
, len
, mode
);
545 ret
= expand_inode_data(inode
, offset
, len
, mode
);
547 f2fs_balance_fs(sbi
);
551 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
552 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
554 static inline __u32
f2fs_mask_flags(umode_t mode
, __u32 flags
)
558 else if (S_ISREG(mode
))
559 return flags
& F2FS_REG_FLMASK
;
561 return flags
& F2FS_OTHER_FLMASK
;
564 long f2fs_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
566 struct inode
*inode
= filp
->f_dentry
->d_inode
;
567 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
572 case FS_IOC_GETFLAGS
:
573 flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
574 return put_user(flags
, (int __user
*) arg
);
575 case FS_IOC_SETFLAGS
:
577 unsigned int oldflags
;
579 ret
= mnt_want_write(filp
->f_path
.mnt
);
583 if (!inode_owner_or_capable(inode
)) {
588 if (get_user(flags
, (int __user
*) arg
)) {
593 flags
= f2fs_mask_flags(inode
->i_mode
, flags
);
595 mutex_lock(&inode
->i_mutex
);
597 oldflags
= fi
->i_flags
;
599 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
600 if (!capable(CAP_LINUX_IMMUTABLE
)) {
601 mutex_unlock(&inode
->i_mutex
);
607 flags
= flags
& FS_FL_USER_MODIFIABLE
;
608 flags
|= oldflags
& ~FS_FL_USER_MODIFIABLE
;
610 mutex_unlock(&inode
->i_mutex
);
612 f2fs_set_inode_flags(inode
);
613 inode
->i_ctime
= CURRENT_TIME
;
614 mark_inode_dirty(inode
);
616 mnt_drop_write(filp
->f_path
.mnt
);
624 const struct file_operations f2fs_file_operations
= {
625 .llseek
= generic_file_llseek
,
626 .read
= do_sync_read
,
627 .write
= do_sync_write
,
628 .aio_read
= generic_file_aio_read
,
629 .aio_write
= generic_file_aio_write
,
630 .open
= generic_file_open
,
631 .mmap
= f2fs_file_mmap
,
632 .fsync
= f2fs_sync_file
,
633 .fallocate
= f2fs_fallocate
,
634 .unlocked_ioctl
= f2fs_ioctl
,
635 .splice_read
= generic_file_splice_read
,
636 .splice_write
= generic_file_splice_write
,