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
);
66 file_update_time(vma
->vm_file
);
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
))
82 /* page is wholly or partially inside EOF */
83 if (((page
->index
+ 1) << PAGE_CACHE_SHIFT
) > i_size_read(inode
)) {
85 offset
= i_size_read(inode
) & ~PAGE_CACHE_MASK
;
86 zero_user_segment(page
, offset
, PAGE_CACHE_SIZE
);
89 SetPageUptodate(page
);
93 wait_on_page_writeback(page
);
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 static int get_parent_ino(struct inode
*inode
, nid_t
*pino
)
107 struct dentry
*dentry
;
109 inode
= igrab(inode
);
110 dentry
= d_find_any_alias(inode
);
115 if (update_dent_inode(inode
, &dentry
->d_name
)) {
120 *pino
= parent_ino(dentry
);
125 int f2fs_sync_file(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
127 struct inode
*inode
= file
->f_mapping
->host
;
128 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
130 bool need_cp
= false;
131 struct writeback_control wbc
= {
132 .sync_mode
= WB_SYNC_ALL
,
133 .nr_to_write
= LONG_MAX
,
137 if (f2fs_readonly(inode
->i_sb
))
140 trace_f2fs_sync_file_enter(inode
);
141 ret
= filemap_write_and_wait_range(inode
->i_mapping
, start
, end
);
143 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
147 /* guarantee free sections for fsync */
148 f2fs_balance_fs(sbi
);
150 mutex_lock(&inode
->i_mutex
);
153 * Both of fdatasync() and fsync() are able to be recovered from
156 if (!S_ISREG(inode
->i_mode
) || inode
->i_nlink
!= 1)
158 else if (file_wrong_pino(inode
))
160 else if (!space_for_roll_forward(sbi
))
162 else if (!is_checkpointed_node(sbi
, F2FS_I(inode
)->i_pino
))
164 else if (F2FS_I(inode
)->xattr_ver
==
165 le64_to_cpu(F2FS_CKPT(sbi
)->checkpoint_ver
))
171 F2FS_I(inode
)->xattr_ver
= 0;
173 /* all the dirty node pages should be flushed for POR */
174 ret
= f2fs_sync_fs(inode
->i_sb
, 1);
175 if (file_wrong_pino(inode
) && inode
->i_nlink
== 1 &&
176 get_parent_ino(inode
, &pino
)) {
177 F2FS_I(inode
)->i_pino
= pino
;
178 file_got_pino(inode
);
179 mark_inode_dirty_sync(inode
);
180 ret
= f2fs_write_inode(inode
, NULL
);
185 /* if there is no written node page, write its inode page */
186 while (!sync_node_pages(sbi
, inode
->i_ino
, &wbc
)) {
187 mark_inode_dirty_sync(inode
);
188 ret
= f2fs_write_inode(inode
, NULL
);
192 filemap_fdatawait_range(sbi
->node_inode
->i_mapping
,
194 ret
= blkdev_issue_flush(inode
->i_sb
->s_bdev
, GFP_KERNEL
, NULL
);
197 mutex_unlock(&inode
->i_mutex
);
198 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
202 static int f2fs_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
205 vma
->vm_ops
= &f2fs_file_vm_ops
;
209 int truncate_data_blocks_range(struct dnode_of_data
*dn
, int count
)
211 int nr_free
= 0, ofs
= dn
->ofs_in_node
;
212 struct f2fs_sb_info
*sbi
= F2FS_SB(dn
->inode
->i_sb
);
213 struct f2fs_node
*raw_node
;
216 raw_node
= F2FS_NODE(dn
->node_page
);
217 addr
= blkaddr_in_node(raw_node
) + ofs
;
219 for ( ; count
> 0; count
--, addr
++, dn
->ofs_in_node
++) {
220 block_t blkaddr
= le32_to_cpu(*addr
);
221 if (blkaddr
== NULL_ADDR
)
224 update_extent_cache(NULL_ADDR
, dn
);
225 invalidate_blocks(sbi
, blkaddr
);
229 dec_valid_block_count(sbi
, dn
->inode
, nr_free
);
230 set_page_dirty(dn
->node_page
);
233 dn
->ofs_in_node
= ofs
;
235 trace_f2fs_truncate_data_blocks_range(dn
->inode
, dn
->nid
,
236 dn
->ofs_in_node
, nr_free
);
240 void truncate_data_blocks(struct dnode_of_data
*dn
)
242 truncate_data_blocks_range(dn
, ADDRS_PER_BLOCK
);
245 static void truncate_partial_data_page(struct inode
*inode
, u64 from
)
247 unsigned offset
= from
& (PAGE_CACHE_SIZE
- 1);
253 page
= find_data_page(inode
, from
>> PAGE_CACHE_SHIFT
, false);
258 if (page
->mapping
!= inode
->i_mapping
) {
259 f2fs_put_page(page
, 1);
262 wait_on_page_writeback(page
);
263 zero_user(page
, offset
, PAGE_CACHE_SIZE
- offset
);
264 set_page_dirty(page
);
265 f2fs_put_page(page
, 1);
268 static int truncate_blocks(struct inode
*inode
, u64 from
)
270 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
271 unsigned int blocksize
= inode
->i_sb
->s_blocksize
;
272 struct dnode_of_data dn
;
274 int count
= 0, ilock
= -1;
277 trace_f2fs_truncate_blocks_enter(inode
, from
);
279 free_from
= (pgoff_t
)
280 ((from
+ blocksize
- 1) >> (sbi
->log_blocksize
));
282 ilock
= mutex_lock_op(sbi
);
283 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
284 err
= get_dnode_of_data(&dn
, free_from
, LOOKUP_NODE
);
288 mutex_unlock_op(sbi
, ilock
);
289 trace_f2fs_truncate_blocks_exit(inode
, err
);
293 if (IS_INODE(dn
.node_page
))
294 count
= ADDRS_PER_INODE
;
296 count
= ADDRS_PER_BLOCK
;
298 count
-= dn
.ofs_in_node
;
301 if (dn
.ofs_in_node
|| IS_INODE(dn
.node_page
)) {
302 truncate_data_blocks_range(&dn
, count
);
308 err
= truncate_inode_blocks(inode
, free_from
);
309 mutex_unlock_op(sbi
, ilock
);
311 /* lastly zero out the first data page */
312 truncate_partial_data_page(inode
, from
);
314 trace_f2fs_truncate_blocks_exit(inode
, err
);
318 void f2fs_truncate(struct inode
*inode
)
320 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
321 S_ISLNK(inode
->i_mode
)))
324 trace_f2fs_truncate(inode
);
326 if (!truncate_blocks(inode
, i_size_read(inode
))) {
327 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
328 mark_inode_dirty(inode
);
332 int f2fs_getattr(struct vfsmount
*mnt
,
333 struct dentry
*dentry
, struct kstat
*stat
)
335 struct inode
*inode
= dentry
->d_inode
;
336 generic_fillattr(inode
, stat
);
341 #ifdef CONFIG_F2FS_FS_POSIX_ACL
342 static void __setattr_copy(struct inode
*inode
, const struct iattr
*attr
)
344 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
345 unsigned int ia_valid
= attr
->ia_valid
;
347 if (ia_valid
& ATTR_UID
)
348 inode
->i_uid
= attr
->ia_uid
;
349 if (ia_valid
& ATTR_GID
)
350 inode
->i_gid
= attr
->ia_gid
;
351 if (ia_valid
& ATTR_ATIME
)
352 inode
->i_atime
= timespec_trunc(attr
->ia_atime
,
353 inode
->i_sb
->s_time_gran
);
354 if (ia_valid
& ATTR_MTIME
)
355 inode
->i_mtime
= timespec_trunc(attr
->ia_mtime
,
356 inode
->i_sb
->s_time_gran
);
357 if (ia_valid
& ATTR_CTIME
)
358 inode
->i_ctime
= timespec_trunc(attr
->ia_ctime
,
359 inode
->i_sb
->s_time_gran
);
360 if (ia_valid
& ATTR_MODE
) {
361 umode_t mode
= attr
->ia_mode
;
363 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
365 set_acl_inode(fi
, mode
);
369 #define __setattr_copy setattr_copy
372 int f2fs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
374 struct inode
*inode
= dentry
->d_inode
;
375 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
378 err
= inode_change_ok(inode
, attr
);
382 if ((attr
->ia_valid
& ATTR_SIZE
) &&
383 attr
->ia_size
!= i_size_read(inode
)) {
384 truncate_setsize(inode
, attr
->ia_size
);
385 f2fs_truncate(inode
);
386 f2fs_balance_fs(F2FS_SB(inode
->i_sb
));
389 __setattr_copy(inode
, attr
);
391 if (attr
->ia_valid
& ATTR_MODE
) {
392 err
= f2fs_acl_chmod(inode
);
393 if (err
|| is_inode_flag_set(fi
, FI_ACL_MODE
)) {
394 inode
->i_mode
= fi
->i_acl_mode
;
395 clear_inode_flag(fi
, FI_ACL_MODE
);
399 mark_inode_dirty(inode
);
403 const struct inode_operations f2fs_file_inode_operations
= {
404 .getattr
= f2fs_getattr
,
405 .setattr
= f2fs_setattr
,
406 .get_acl
= f2fs_get_acl
,
407 #ifdef CONFIG_F2FS_FS_XATTR
408 .setxattr
= generic_setxattr
,
409 .getxattr
= generic_getxattr
,
410 .listxattr
= f2fs_listxattr
,
411 .removexattr
= generic_removexattr
,
415 static void fill_zero(struct inode
*inode
, pgoff_t index
,
416 loff_t start
, loff_t len
)
418 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
425 f2fs_balance_fs(sbi
);
427 ilock
= mutex_lock_op(sbi
);
428 page
= get_new_data_page(inode
, NULL
, index
, false);
429 mutex_unlock_op(sbi
, ilock
);
432 wait_on_page_writeback(page
);
433 zero_user(page
, start
, len
);
434 set_page_dirty(page
);
435 f2fs_put_page(page
, 1);
439 int truncate_hole(struct inode
*inode
, pgoff_t pg_start
, pgoff_t pg_end
)
444 for (index
= pg_start
; index
< pg_end
; index
++) {
445 struct dnode_of_data dn
;
447 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
448 err
= get_dnode_of_data(&dn
, index
, LOOKUP_NODE
);
455 if (dn
.data_blkaddr
!= NULL_ADDR
)
456 truncate_data_blocks_range(&dn
, 1);
462 static int punch_hole(struct inode
*inode
, loff_t offset
, loff_t len
, int mode
)
464 pgoff_t pg_start
, pg_end
;
465 loff_t off_start
, off_end
;
468 pg_start
= ((unsigned long long) offset
) >> PAGE_CACHE_SHIFT
;
469 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_CACHE_SHIFT
;
471 off_start
= offset
& (PAGE_CACHE_SIZE
- 1);
472 off_end
= (offset
+ len
) & (PAGE_CACHE_SIZE
- 1);
474 if (pg_start
== pg_end
) {
475 fill_zero(inode
, pg_start
, off_start
,
476 off_end
- off_start
);
479 fill_zero(inode
, pg_start
++, off_start
,
480 PAGE_CACHE_SIZE
- off_start
);
482 fill_zero(inode
, pg_end
, 0, off_end
);
484 if (pg_start
< pg_end
) {
485 struct address_space
*mapping
= inode
->i_mapping
;
486 loff_t blk_start
, blk_end
;
487 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
490 f2fs_balance_fs(sbi
);
492 blk_start
= pg_start
<< PAGE_CACHE_SHIFT
;
493 blk_end
= pg_end
<< PAGE_CACHE_SHIFT
;
494 truncate_inode_pages_range(mapping
, blk_start
,
497 ilock
= mutex_lock_op(sbi
);
498 ret
= truncate_hole(inode
, pg_start
, pg_end
);
499 mutex_unlock_op(sbi
, ilock
);
503 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
504 i_size_read(inode
) <= (offset
+ len
)) {
505 i_size_write(inode
, offset
);
506 mark_inode_dirty(inode
);
512 static int expand_inode_data(struct inode
*inode
, loff_t offset
,
513 loff_t len
, int mode
)
515 struct f2fs_sb_info
*sbi
= F2FS_SB(inode
->i_sb
);
516 pgoff_t index
, pg_start
, pg_end
;
517 loff_t new_size
= i_size_read(inode
);
518 loff_t off_start
, off_end
;
521 ret
= inode_newsize_ok(inode
, (len
+ offset
));
525 pg_start
= ((unsigned long long) offset
) >> PAGE_CACHE_SHIFT
;
526 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_CACHE_SHIFT
;
528 off_start
= offset
& (PAGE_CACHE_SIZE
- 1);
529 off_end
= (offset
+ len
) & (PAGE_CACHE_SIZE
- 1);
531 for (index
= pg_start
; index
<= pg_end
; index
++) {
532 struct dnode_of_data dn
;
535 ilock
= mutex_lock_op(sbi
);
536 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
537 ret
= get_dnode_of_data(&dn
, index
, ALLOC_NODE
);
539 mutex_unlock_op(sbi
, ilock
);
543 if (dn
.data_blkaddr
== NULL_ADDR
) {
544 ret
= reserve_new_block(&dn
);
547 mutex_unlock_op(sbi
, ilock
);
552 mutex_unlock_op(sbi
, ilock
);
554 if (pg_start
== pg_end
)
555 new_size
= offset
+ len
;
556 else if (index
== pg_start
&& off_start
)
557 new_size
= (index
+ 1) << PAGE_CACHE_SHIFT
;
558 else if (index
== pg_end
)
559 new_size
= (index
<< PAGE_CACHE_SHIFT
) + off_end
;
561 new_size
+= PAGE_CACHE_SIZE
;
564 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
565 i_size_read(inode
) < new_size
) {
566 i_size_write(inode
, new_size
);
567 mark_inode_dirty(inode
);
573 static long f2fs_fallocate(struct file
*file
, int mode
,
574 loff_t offset
, loff_t len
)
576 struct inode
*inode
= file_inode(file
);
579 if (mode
& ~(FALLOC_FL_KEEP_SIZE
| FALLOC_FL_PUNCH_HOLE
))
582 if (mode
& FALLOC_FL_PUNCH_HOLE
)
583 ret
= punch_hole(inode
, offset
, len
, mode
);
585 ret
= expand_inode_data(inode
, offset
, len
, mode
);
588 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
589 mark_inode_dirty(inode
);
591 trace_f2fs_fallocate(inode
, mode
, offset
, len
, ret
);
595 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
596 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
598 static inline __u32
f2fs_mask_flags(umode_t mode
, __u32 flags
)
602 else if (S_ISREG(mode
))
603 return flags
& F2FS_REG_FLMASK
;
605 return flags
& F2FS_OTHER_FLMASK
;
608 long f2fs_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
610 struct inode
*inode
= file_inode(filp
);
611 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
616 case F2FS_IOC_GETFLAGS
:
617 flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
618 return put_user(flags
, (int __user
*) arg
);
619 case F2FS_IOC_SETFLAGS
:
621 unsigned int oldflags
;
623 ret
= mnt_want_write_file(filp
);
627 if (!inode_owner_or_capable(inode
)) {
632 if (get_user(flags
, (int __user
*) arg
)) {
637 flags
= f2fs_mask_flags(inode
->i_mode
, flags
);
639 mutex_lock(&inode
->i_mutex
);
641 oldflags
= fi
->i_flags
;
643 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
644 if (!capable(CAP_LINUX_IMMUTABLE
)) {
645 mutex_unlock(&inode
->i_mutex
);
651 flags
= flags
& FS_FL_USER_MODIFIABLE
;
652 flags
|= oldflags
& ~FS_FL_USER_MODIFIABLE
;
654 mutex_unlock(&inode
->i_mutex
);
656 f2fs_set_inode_flags(inode
);
657 inode
->i_ctime
= CURRENT_TIME
;
658 mark_inode_dirty(inode
);
660 mnt_drop_write_file(filp
);
669 long f2fs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
672 case F2FS_IOC32_GETFLAGS
:
673 cmd
= F2FS_IOC_GETFLAGS
;
675 case F2FS_IOC32_SETFLAGS
:
676 cmd
= F2FS_IOC_SETFLAGS
;
681 return f2fs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));
685 const struct file_operations f2fs_file_operations
= {
686 .llseek
= generic_file_llseek
,
687 .read
= do_sync_read
,
688 .write
= do_sync_write
,
689 .aio_read
= generic_file_aio_read
,
690 .aio_write
= generic_file_aio_write
,
691 .open
= generic_file_open
,
692 .mmap
= f2fs_file_mmap
,
693 .fsync
= f2fs_sync_file
,
694 .fallocate
= f2fs_fallocate
,
695 .unlocked_ioctl
= f2fs_ioctl
,
697 .compat_ioctl
= f2fs_compat_ioctl
,
699 .splice_read
= generic_file_splice_read
,
700 .splice_write
= generic_file_splice_write
,