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
);
37 struct dnode_of_data dn
;
42 sb_start_pagefault(inode
->i_sb
);
44 /* block allocation */
45 ilock
= mutex_lock_op(sbi
);
46 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
47 err
= get_dnode_of_data(&dn
, page
->index
, ALLOC_NODE
);
49 mutex_unlock_op(sbi
, ilock
);
53 old_blk_addr
= dn
.data_blkaddr
;
55 if (old_blk_addr
== NULL_ADDR
) {
56 err
= reserve_new_block(&dn
);
59 mutex_unlock_op(sbi
, ilock
);
64 mutex_unlock_op(sbi
, ilock
);
67 if (page
->mapping
!= inode
->i_mapping
||
68 page_offset(page
) >= i_size_read(inode
) ||
69 !PageUptodate(page
)) {
76 * check to see if the page is mapped already (no holes)
78 if (PageMappedToDisk(page
))
82 wait_on_page_writeback(page
);
84 /* page is wholly or partially inside EOF */
85 if (((page
->index
+ 1) << PAGE_CACHE_SHIFT
) > i_size_read(inode
)) {
87 offset
= i_size_read(inode
) & ~PAGE_CACHE_MASK
;
88 zero_user_segment(page
, offset
, PAGE_CACHE_SIZE
);
91 SetPageUptodate(page
);
93 file_update_time(vma
->vm_file
);
95 sb_end_pagefault(inode
->i_sb
);
96 return block_page_mkwrite_return(err
);
99 static const struct vm_operations_struct f2fs_file_vm_ops
= {
100 .fault
= filemap_fault
,
101 .page_mkwrite
= f2fs_vm_page_mkwrite
,
102 .remap_pages
= generic_file_remap_pages
,
105 int f2fs_sync_file(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
107 struct inode
*inode
= file
->f_mapping
->host
;
108 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
110 bool need_cp
= false;
111 struct writeback_control wbc
= {
112 .sync_mode
= WB_SYNC_ALL
,
113 .nr_to_write
= LONG_MAX
,
117 if (inode
->i_sb
->s_flags
& MS_RDONLY
)
120 trace_f2fs_sync_file_enter(inode
);
121 ret
= filemap_write_and_wait_range(inode
->i_mapping
, start
, end
);
123 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
127 /* guarantee free sections for fsync */
128 f2fs_balance_fs(sbi
);
130 mutex_lock(&inode
->i_mutex
);
132 if (datasync
&& !(inode
->i_state
& I_DIRTY_DATASYNC
))
135 if (!S_ISREG(inode
->i_mode
) || inode
->i_nlink
!= 1)
137 else if (is_cp_file(inode
))
139 else if (!space_for_roll_forward(sbi
))
141 else if (!is_checkpointed_node(sbi
, F2FS_I(inode
)->i_pino
))
145 /* all the dirty node pages should be flushed for POR */
146 ret
= f2fs_sync_fs(inode
->i_sb
, 1);
148 /* if there is no written node page, write its inode page */
149 while (!sync_node_pages(sbi
, inode
->i_ino
, &wbc
)) {
150 ret
= f2fs_write_inode(inode
, NULL
);
154 filemap_fdatawait_range(sbi
->node_inode
->i_mapping
,
156 ret
= blkdev_issue_flush(inode
->i_sb
->s_bdev
, GFP_KERNEL
, NULL
);
159 mutex_unlock(&inode
->i_mutex
);
160 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
164 static int f2fs_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
167 vma
->vm_ops
= &f2fs_file_vm_ops
;
171 static int truncate_data_blocks_range(struct dnode_of_data
*dn
, int count
)
173 int nr_free
= 0, ofs
= dn
->ofs_in_node
;
174 struct f2fs_sb_info
*sbi
= F2FS_SB(dn
->inode
->i_sb
);
175 struct f2fs_node
*raw_node
;
178 raw_node
= page_address(dn
->node_page
);
179 addr
= blkaddr_in_node(raw_node
) + ofs
;
181 for ( ; count
> 0; count
--, addr
++, dn
->ofs_in_node
++) {
182 block_t blkaddr
= le32_to_cpu(*addr
);
183 if (blkaddr
== NULL_ADDR
)
186 update_extent_cache(NULL_ADDR
, dn
);
187 invalidate_blocks(sbi
, blkaddr
);
188 dec_valid_block_count(sbi
, dn
->inode
, 1);
192 set_page_dirty(dn
->node_page
);
195 dn
->ofs_in_node
= ofs
;
197 trace_f2fs_truncate_data_blocks_range(dn
->inode
, dn
->nid
,
198 dn
->ofs_in_node
, nr_free
);
202 void truncate_data_blocks(struct dnode_of_data
*dn
)
204 truncate_data_blocks_range(dn
, ADDRS_PER_BLOCK
);
207 static void truncate_partial_data_page(struct inode
*inode
, u64 from
)
209 unsigned offset
= from
& (PAGE_CACHE_SIZE
- 1);
215 page
= find_data_page(inode
, from
>> PAGE_CACHE_SHIFT
, false);
220 if (page
->mapping
!= inode
->i_mapping
) {
221 f2fs_put_page(page
, 1);
224 wait_on_page_writeback(page
);
225 zero_user(page
, offset
, PAGE_CACHE_SIZE
- offset
);
226 set_page_dirty(page
);
227 f2fs_put_page(page
, 1);
230 static int truncate_blocks(struct inode
*inode
, u64 from
)
232 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
233 unsigned int blocksize
= inode
->i_sb
->s_blocksize
;
234 struct dnode_of_data dn
;
236 int count
= 0, ilock
= -1;
239 trace_f2fs_truncate_blocks_enter(inode
, from
);
241 free_from
= (pgoff_t
)
242 ((from
+ blocksize
- 1) >> (sbi
->log_blocksize
));
244 ilock
= mutex_lock_op(sbi
);
245 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
246 err
= get_dnode_of_data(&dn
, free_from
, LOOKUP_NODE
);
250 mutex_unlock_op(sbi
, ilock
);
251 trace_f2fs_truncate_blocks_exit(inode
, err
);
255 if (IS_INODE(dn
.node_page
))
256 count
= ADDRS_PER_INODE
;
258 count
= ADDRS_PER_BLOCK
;
260 count
-= dn
.ofs_in_node
;
263 if (dn
.ofs_in_node
|| IS_INODE(dn
.node_page
)) {
264 truncate_data_blocks_range(&dn
, count
);
270 err
= truncate_inode_blocks(inode
, free_from
);
271 mutex_unlock_op(sbi
, ilock
);
273 /* lastly zero out the first data page */
274 truncate_partial_data_page(inode
, from
);
276 trace_f2fs_truncate_blocks_exit(inode
, err
);
280 void f2fs_truncate(struct inode
*inode
)
282 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
283 S_ISLNK(inode
->i_mode
)))
286 trace_f2fs_truncate(inode
);
288 if (!truncate_blocks(inode
, i_size_read(inode
))) {
289 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
290 mark_inode_dirty(inode
);
294 static int f2fs_getattr(struct vfsmount
*mnt
,
295 struct dentry
*dentry
, struct kstat
*stat
)
297 struct inode
*inode
= dentry
->d_inode
;
298 generic_fillattr(inode
, stat
);
303 #ifdef CONFIG_F2FS_FS_POSIX_ACL
304 static void __setattr_copy(struct inode
*inode
, const struct iattr
*attr
)
306 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
307 unsigned int ia_valid
= attr
->ia_valid
;
309 if (ia_valid
& ATTR_UID
)
310 inode
->i_uid
= attr
->ia_uid
;
311 if (ia_valid
& ATTR_GID
)
312 inode
->i_gid
= attr
->ia_gid
;
313 if (ia_valid
& ATTR_ATIME
)
314 inode
->i_atime
= timespec_trunc(attr
->ia_atime
,
315 inode
->i_sb
->s_time_gran
);
316 if (ia_valid
& ATTR_MTIME
)
317 inode
->i_mtime
= timespec_trunc(attr
->ia_mtime
,
318 inode
->i_sb
->s_time_gran
);
319 if (ia_valid
& ATTR_CTIME
)
320 inode
->i_ctime
= timespec_trunc(attr
->ia_ctime
,
321 inode
->i_sb
->s_time_gran
);
322 if (ia_valid
& ATTR_MODE
) {
323 umode_t mode
= attr
->ia_mode
;
325 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
327 set_acl_inode(fi
, mode
);
331 #define __setattr_copy setattr_copy
334 int f2fs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
336 struct inode
*inode
= dentry
->d_inode
;
337 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
340 err
= inode_change_ok(inode
, attr
);
344 if ((attr
->ia_valid
& ATTR_SIZE
) &&
345 attr
->ia_size
!= i_size_read(inode
)) {
346 truncate_setsize(inode
, attr
->ia_size
);
347 f2fs_truncate(inode
);
348 f2fs_balance_fs(F2FS_SB(inode
->i_sb
));
351 __setattr_copy(inode
, attr
);
353 if (attr
->ia_valid
& ATTR_MODE
) {
354 err
= f2fs_acl_chmod(inode
);
355 if (err
|| is_inode_flag_set(fi
, FI_ACL_MODE
)) {
356 inode
->i_mode
= fi
->i_acl_mode
;
357 clear_inode_flag(fi
, FI_ACL_MODE
);
361 mark_inode_dirty(inode
);
365 const struct inode_operations f2fs_file_inode_operations
= {
366 .getattr
= f2fs_getattr
,
367 .setattr
= f2fs_setattr
,
368 .get_acl
= f2fs_get_acl
,
369 #ifdef CONFIG_F2FS_FS_XATTR
370 .setxattr
= generic_setxattr
,
371 .getxattr
= generic_getxattr
,
372 .listxattr
= f2fs_listxattr
,
373 .removexattr
= generic_removexattr
,
377 static void fill_zero(struct inode
*inode
, pgoff_t index
,
378 loff_t start
, loff_t len
)
380 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
387 f2fs_balance_fs(sbi
);
389 ilock
= mutex_lock_op(sbi
);
390 page
= get_new_data_page(inode
, index
, false);
391 mutex_unlock_op(sbi
, ilock
);
394 wait_on_page_writeback(page
);
395 zero_user(page
, start
, len
);
396 set_page_dirty(page
);
397 f2fs_put_page(page
, 1);
401 int truncate_hole(struct inode
*inode
, pgoff_t pg_start
, pgoff_t pg_end
)
406 for (index
= pg_start
; index
< pg_end
; index
++) {
407 struct dnode_of_data dn
;
409 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
410 err
= get_dnode_of_data(&dn
, index
, LOOKUP_NODE
);
417 if (dn
.data_blkaddr
!= NULL_ADDR
)
418 truncate_data_blocks_range(&dn
, 1);
424 static int punch_hole(struct inode
*inode
, loff_t offset
, loff_t len
, int mode
)
426 pgoff_t pg_start
, pg_end
;
427 loff_t off_start
, off_end
;
430 pg_start
= ((unsigned long long) offset
) >> PAGE_CACHE_SHIFT
;
431 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_CACHE_SHIFT
;
433 off_start
= offset
& (PAGE_CACHE_SIZE
- 1);
434 off_end
= (offset
+ len
) & (PAGE_CACHE_SIZE
- 1);
436 if (pg_start
== pg_end
) {
437 fill_zero(inode
, pg_start
, off_start
,
438 off_end
- off_start
);
441 fill_zero(inode
, pg_start
++, off_start
,
442 PAGE_CACHE_SIZE
- off_start
);
444 fill_zero(inode
, pg_end
, 0, off_end
);
446 if (pg_start
< pg_end
) {
447 struct address_space
*mapping
= inode
->i_mapping
;
448 loff_t blk_start
, blk_end
;
449 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
452 f2fs_balance_fs(sbi
);
454 blk_start
= pg_start
<< PAGE_CACHE_SHIFT
;
455 blk_end
= pg_end
<< PAGE_CACHE_SHIFT
;
456 truncate_inode_pages_range(mapping
, blk_start
,
459 ilock
= mutex_lock_op(sbi
);
460 ret
= truncate_hole(inode
, pg_start
, pg_end
);
461 mutex_unlock_op(sbi
, ilock
);
465 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
466 i_size_read(inode
) <= (offset
+ len
)) {
467 i_size_write(inode
, offset
);
468 mark_inode_dirty(inode
);
474 static int expand_inode_data(struct inode
*inode
, loff_t offset
,
475 loff_t len
, int mode
)
477 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
478 pgoff_t index
, pg_start
, pg_end
;
479 loff_t new_size
= i_size_read(inode
);
480 loff_t off_start
, off_end
;
483 ret
= inode_newsize_ok(inode
, (len
+ offset
));
487 pg_start
= ((unsigned long long) offset
) >> PAGE_CACHE_SHIFT
;
488 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_CACHE_SHIFT
;
490 off_start
= offset
& (PAGE_CACHE_SIZE
- 1);
491 off_end
= (offset
+ len
) & (PAGE_CACHE_SIZE
- 1);
493 for (index
= pg_start
; index
<= pg_end
; index
++) {
494 struct dnode_of_data dn
;
497 ilock
= mutex_lock_op(sbi
);
498 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
499 ret
= get_dnode_of_data(&dn
, index
, ALLOC_NODE
);
501 mutex_unlock_op(sbi
, ilock
);
505 if (dn
.data_blkaddr
== NULL_ADDR
) {
506 ret
= reserve_new_block(&dn
);
509 mutex_unlock_op(sbi
, ilock
);
514 mutex_unlock_op(sbi
, ilock
);
516 if (pg_start
== pg_end
)
517 new_size
= offset
+ len
;
518 else if (index
== pg_start
&& off_start
)
519 new_size
= (index
+ 1) << PAGE_CACHE_SHIFT
;
520 else if (index
== pg_end
)
521 new_size
= (index
<< PAGE_CACHE_SHIFT
) + off_end
;
523 new_size
+= PAGE_CACHE_SIZE
;
526 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
527 i_size_read(inode
) < new_size
) {
528 i_size_write(inode
, new_size
);
529 mark_inode_dirty(inode
);
535 static long f2fs_fallocate(struct file
*file
, int mode
,
536 loff_t offset
, loff_t len
)
538 struct inode
*inode
= file_inode(file
);
541 if (mode
& ~(FALLOC_FL_KEEP_SIZE
| FALLOC_FL_PUNCH_HOLE
))
544 if (mode
& FALLOC_FL_PUNCH_HOLE
)
545 ret
= punch_hole(inode
, offset
, len
, mode
);
547 ret
= expand_inode_data(inode
, offset
, len
, mode
);
550 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
551 mark_inode_dirty(inode
);
553 trace_f2fs_fallocate(inode
, mode
, offset
, len
, ret
);
557 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
558 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
560 static inline __u32
f2fs_mask_flags(umode_t mode
, __u32 flags
)
564 else if (S_ISREG(mode
))
565 return flags
& F2FS_REG_FLMASK
;
567 return flags
& F2FS_OTHER_FLMASK
;
570 long f2fs_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
572 struct inode
*inode
= file_inode(filp
);
573 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
578 case FS_IOC_GETFLAGS
:
579 flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
580 return put_user(flags
, (int __user
*) arg
);
581 case FS_IOC_SETFLAGS
:
583 unsigned int oldflags
;
585 ret
= mnt_want_write_file(filp
);
589 if (!inode_owner_or_capable(inode
)) {
594 if (get_user(flags
, (int __user
*) arg
)) {
599 flags
= f2fs_mask_flags(inode
->i_mode
, flags
);
601 mutex_lock(&inode
->i_mutex
);
603 oldflags
= fi
->i_flags
;
605 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
606 if (!capable(CAP_LINUX_IMMUTABLE
)) {
607 mutex_unlock(&inode
->i_mutex
);
613 flags
= flags
& FS_FL_USER_MODIFIABLE
;
614 flags
|= oldflags
& ~FS_FL_USER_MODIFIABLE
;
616 mutex_unlock(&inode
->i_mutex
);
618 f2fs_set_inode_flags(inode
);
619 inode
->i_ctime
= CURRENT_TIME
;
620 mark_inode_dirty(inode
);
622 mnt_drop_write_file(filp
);
631 long f2fs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
634 case F2FS_IOC32_GETFLAGS
:
635 cmd
= F2FS_IOC_GETFLAGS
;
637 case F2FS_IOC32_SETFLAGS
:
638 cmd
= F2FS_IOC_SETFLAGS
;
643 return f2fs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));
647 const struct file_operations f2fs_file_operations
= {
648 .llseek
= generic_file_llseek
,
649 .read
= do_sync_read
,
650 .write
= do_sync_write
,
651 .aio_read
= generic_file_aio_read
,
652 .aio_write
= generic_file_aio_write
,
653 .open
= generic_file_open
,
654 .mmap
= f2fs_file_mmap
,
655 .fsync
= f2fs_sync_file
,
656 .fallocate
= f2fs_fallocate
,
657 .unlocked_ioctl
= f2fs_ioctl
,
659 .compat_ioctl
= f2fs_compat_ioctl
,
661 .splice_read
= generic_file_splice_read
,
662 .splice_write
= generic_file_splice_write
,