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>
23 #include <linux/random.h>
32 #include <trace/events/f2fs.h>
34 static int f2fs_vm_page_mkwrite(struct vm_area_struct
*vma
,
37 struct page
*page
= vmf
->page
;
38 struct inode
*inode
= file_inode(vma
->vm_file
);
39 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
40 struct dnode_of_data dn
;
43 sb_start_pagefault(inode
->i_sb
);
45 f2fs_bug_on(sbi
, f2fs_has_inline_data(inode
));
47 /* block allocation */
49 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
50 err
= f2fs_reserve_block(&dn
, page
->index
);
58 f2fs_balance_fs(sbi
, dn
.node_changed
);
60 file_update_time(vma
->vm_file
);
62 if (unlikely(page
->mapping
!= inode
->i_mapping
||
63 page_offset(page
) > i_size_read(inode
) ||
64 !PageUptodate(page
))) {
71 * check to see if the page is mapped already (no holes)
73 if (PageMappedToDisk(page
))
76 /* page is wholly or partially inside EOF */
77 if (((loff_t
)(page
->index
+ 1) << PAGE_SHIFT
) >
80 offset
= i_size_read(inode
) & ~PAGE_MASK
;
81 zero_user_segment(page
, offset
, PAGE_SIZE
);
84 SetPageUptodate(page
);
86 trace_f2fs_vm_page_mkwrite(page
, DATA
);
89 f2fs_wait_on_page_writeback(page
, DATA
, false);
91 /* wait for GCed encrypted page writeback */
92 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
))
93 f2fs_wait_on_encrypted_page_writeback(sbi
, dn
.data_blkaddr
);
95 /* if gced page is attached, don't write to cold segment */
96 clear_cold_data(page
);
98 sb_end_pagefault(inode
->i_sb
);
99 f2fs_update_time(sbi
, REQ_TIME
);
100 return block_page_mkwrite_return(err
);
103 static const struct vm_operations_struct f2fs_file_vm_ops
= {
104 .fault
= filemap_fault
,
105 .map_pages
= filemap_map_pages
,
106 .page_mkwrite
= f2fs_vm_page_mkwrite
,
109 static int get_parent_ino(struct inode
*inode
, nid_t
*pino
)
111 struct dentry
*dentry
;
113 inode
= igrab(inode
);
114 dentry
= d_find_any_alias(inode
);
119 if (update_dent_inode(inode
, inode
, &dentry
->d_name
)) {
124 *pino
= parent_ino(dentry
);
129 static inline bool need_do_checkpoint(struct inode
*inode
)
131 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
132 bool need_cp
= false;
134 if (!S_ISREG(inode
->i_mode
) || inode
->i_nlink
!= 1)
136 else if (file_enc_name(inode
) && need_dentry_mark(sbi
, inode
->i_ino
))
138 else if (file_wrong_pino(inode
))
140 else if (!space_for_roll_forward(sbi
))
142 else if (!is_checkpointed_node(sbi
, F2FS_I(inode
)->i_pino
))
144 else if (F2FS_I(inode
)->xattr_ver
== cur_cp_version(F2FS_CKPT(sbi
)))
146 else if (test_opt(sbi
, FASTBOOT
))
148 else if (sbi
->active_logs
== 2)
154 static bool need_inode_page_update(struct f2fs_sb_info
*sbi
, nid_t ino
)
156 struct page
*i
= find_get_page(NODE_MAPPING(sbi
), ino
);
158 /* But we need to avoid that there are some inode updates */
159 if ((i
&& PageDirty(i
)) || need_inode_block_update(sbi
, ino
))
165 static void try_to_fix_pino(struct inode
*inode
)
167 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
170 down_write(&fi
->i_sem
);
172 if (file_wrong_pino(inode
) && inode
->i_nlink
== 1 &&
173 get_parent_ino(inode
, &pino
)) {
175 file_got_pino(inode
);
176 up_write(&fi
->i_sem
);
178 mark_inode_dirty_sync(inode
);
179 f2fs_write_inode(inode
, NULL
);
181 up_write(&fi
->i_sem
);
185 static int f2fs_do_sync_file(struct file
*file
, loff_t start
, loff_t end
,
186 int datasync
, bool atomic
)
188 struct inode
*inode
= file
->f_mapping
->host
;
189 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
190 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
191 nid_t ino
= inode
->i_ino
;
193 bool need_cp
= false;
194 struct writeback_control wbc
= {
195 .sync_mode
= WB_SYNC_ALL
,
196 .nr_to_write
= LONG_MAX
,
200 if (unlikely(f2fs_readonly(inode
->i_sb
)))
203 trace_f2fs_sync_file_enter(inode
);
205 /* if fdatasync is triggered, let's do in-place-update */
206 if (datasync
|| get_dirty_pages(inode
) <= SM_I(sbi
)->min_fsync_blocks
)
207 set_inode_flag(fi
, FI_NEED_IPU
);
208 ret
= filemap_write_and_wait_range(inode
->i_mapping
, start
, end
);
209 clear_inode_flag(fi
, FI_NEED_IPU
);
212 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
216 /* if the inode is dirty, let's recover all the time */
218 f2fs_write_inode(inode
, NULL
);
223 * if there is no written data, don't waste time to write recovery info.
225 if (!is_inode_flag_set(fi
, FI_APPEND_WRITE
) &&
226 !exist_written_data(sbi
, ino
, APPEND_INO
)) {
228 /* it may call write_inode just prior to fsync */
229 if (need_inode_page_update(sbi
, ino
))
232 if (is_inode_flag_set(fi
, FI_UPDATE_WRITE
) ||
233 exist_written_data(sbi
, ino
, UPDATE_INO
))
239 * Both of fdatasync() and fsync() are able to be recovered from
242 down_read(&fi
->i_sem
);
243 need_cp
= need_do_checkpoint(inode
);
247 /* all the dirty node pages should be flushed for POR */
248 ret
= f2fs_sync_fs(inode
->i_sb
, 1);
251 * We've secured consistency through sync_fs. Following pino
252 * will be used only for fsynced inodes after checkpoint.
254 try_to_fix_pino(inode
);
255 clear_inode_flag(fi
, FI_APPEND_WRITE
);
256 clear_inode_flag(fi
, FI_UPDATE_WRITE
);
260 ret
= fsync_node_pages(sbi
, ino
, &wbc
, atomic
);
264 /* if cp_error was enabled, we should avoid infinite loop */
265 if (unlikely(f2fs_cp_error(sbi
))) {
270 if (need_inode_block_update(sbi
, ino
)) {
271 mark_inode_dirty_sync(inode
);
272 f2fs_write_inode(inode
, NULL
);
276 ret
= wait_on_node_pages_writeback(sbi
, ino
);
280 /* once recovery info is written, don't need to tack this */
281 remove_ino_entry(sbi
, ino
, APPEND_INO
);
282 clear_inode_flag(fi
, FI_APPEND_WRITE
);
284 remove_ino_entry(sbi
, ino
, UPDATE_INO
);
285 clear_inode_flag(fi
, FI_UPDATE_WRITE
);
286 ret
= f2fs_issue_flush(sbi
);
287 f2fs_update_time(sbi
, REQ_TIME
);
289 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
290 f2fs_trace_ios(NULL
, 1);
294 int f2fs_sync_file(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
296 return f2fs_do_sync_file(file
, start
, end
, datasync
, false);
299 static pgoff_t
__get_first_dirty_index(struct address_space
*mapping
,
300 pgoff_t pgofs
, int whence
)
305 if (whence
!= SEEK_DATA
)
308 /* find first dirty page index */
309 pagevec_init(&pvec
, 0);
310 nr_pages
= pagevec_lookup_tag(&pvec
, mapping
, &pgofs
,
311 PAGECACHE_TAG_DIRTY
, 1);
312 pgofs
= nr_pages
? pvec
.pages
[0]->index
: ULONG_MAX
;
313 pagevec_release(&pvec
);
317 static bool __found_offset(block_t blkaddr
, pgoff_t dirty
, pgoff_t pgofs
,
322 if ((blkaddr
== NEW_ADDR
&& dirty
== pgofs
) ||
323 (blkaddr
!= NEW_ADDR
&& blkaddr
!= NULL_ADDR
))
327 if (blkaddr
== NULL_ADDR
)
334 static loff_t
f2fs_seek_block(struct file
*file
, loff_t offset
, int whence
)
336 struct inode
*inode
= file
->f_mapping
->host
;
337 loff_t maxbytes
= inode
->i_sb
->s_maxbytes
;
338 struct dnode_of_data dn
;
339 pgoff_t pgofs
, end_offset
, dirty
;
340 loff_t data_ofs
= offset
;
346 isize
= i_size_read(inode
);
350 /* handle inline data case */
351 if (f2fs_has_inline_data(inode
) || f2fs_has_inline_dentry(inode
)) {
352 if (whence
== SEEK_HOLE
)
357 pgofs
= (pgoff_t
)(offset
>> PAGE_SHIFT
);
359 dirty
= __get_first_dirty_index(inode
->i_mapping
, pgofs
, whence
);
361 for (; data_ofs
< isize
; data_ofs
= (loff_t
)pgofs
<< PAGE_SHIFT
) {
362 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
363 err
= get_dnode_of_data(&dn
, pgofs
, LOOKUP_NODE_RA
);
364 if (err
&& err
!= -ENOENT
) {
366 } else if (err
== -ENOENT
) {
367 /* direct node does not exists */
368 if (whence
== SEEK_DATA
) {
369 pgofs
= get_next_page_offset(&dn
, pgofs
);
376 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
378 /* find data/hole in dnode block */
379 for (; dn
.ofs_in_node
< end_offset
;
380 dn
.ofs_in_node
++, pgofs
++,
381 data_ofs
= (loff_t
)pgofs
<< PAGE_SHIFT
) {
383 blkaddr
= datablock_addr(dn
.node_page
, dn
.ofs_in_node
);
385 if (__found_offset(blkaddr
, dirty
, pgofs
, whence
)) {
393 if (whence
== SEEK_DATA
)
396 if (whence
== SEEK_HOLE
&& data_ofs
> isize
)
399 return vfs_setpos(file
, data_ofs
, maxbytes
);
405 static loff_t
f2fs_llseek(struct file
*file
, loff_t offset
, int whence
)
407 struct inode
*inode
= file
->f_mapping
->host
;
408 loff_t maxbytes
= inode
->i_sb
->s_maxbytes
;
414 return generic_file_llseek_size(file
, offset
, whence
,
415 maxbytes
, i_size_read(inode
));
420 return f2fs_seek_block(file
, offset
, whence
);
426 static int f2fs_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
428 struct inode
*inode
= file_inode(file
);
431 if (f2fs_encrypted_inode(inode
)) {
432 err
= fscrypt_get_encryption_info(inode
);
435 if (!f2fs_encrypted_inode(inode
))
439 /* we don't need to use inline_data strictly */
440 err
= f2fs_convert_inline_inode(inode
);
445 vma
->vm_ops
= &f2fs_file_vm_ops
;
449 static int f2fs_file_open(struct inode
*inode
, struct file
*filp
)
451 int ret
= generic_file_open(inode
, filp
);
454 if (!ret
&& f2fs_encrypted_inode(inode
)) {
455 ret
= fscrypt_get_encryption_info(inode
);
458 if (!fscrypt_has_encryption_key(inode
))
461 dir
= dget_parent(file_dentry(filp
));
462 if (f2fs_encrypted_inode(d_inode(dir
)) &&
463 !fscrypt_has_permitted_context(d_inode(dir
), inode
)) {
471 int truncate_data_blocks_range(struct dnode_of_data
*dn
, int count
)
473 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
474 struct f2fs_node
*raw_node
;
475 int nr_free
= 0, ofs
= dn
->ofs_in_node
, len
= count
;
478 raw_node
= F2FS_NODE(dn
->node_page
);
479 addr
= blkaddr_in_node(raw_node
) + ofs
;
481 for (; count
> 0; count
--, addr
++, dn
->ofs_in_node
++) {
482 block_t blkaddr
= le32_to_cpu(*addr
);
483 if (blkaddr
== NULL_ADDR
)
486 dn
->data_blkaddr
= NULL_ADDR
;
487 set_data_blkaddr(dn
);
488 invalidate_blocks(sbi
, blkaddr
);
489 if (dn
->ofs_in_node
== 0 && IS_INODE(dn
->node_page
))
490 clear_inode_flag(F2FS_I(dn
->inode
),
491 FI_FIRST_BLOCK_WRITTEN
);
498 * once we invalidate valid blkaddr in range [ofs, ofs + count],
499 * we will invalidate all blkaddr in the whole range.
501 fofs
= start_bidx_of_node(ofs_of_node(dn
->node_page
),
503 f2fs_update_extent_cache_range(dn
, fofs
, 0, len
);
504 dec_valid_block_count(sbi
, dn
->inode
, nr_free
);
507 dn
->ofs_in_node
= ofs
;
509 f2fs_update_time(sbi
, REQ_TIME
);
510 trace_f2fs_truncate_data_blocks_range(dn
->inode
, dn
->nid
,
511 dn
->ofs_in_node
, nr_free
);
515 void truncate_data_blocks(struct dnode_of_data
*dn
)
517 truncate_data_blocks_range(dn
, ADDRS_PER_BLOCK
);
520 static int truncate_partial_data_page(struct inode
*inode
, u64 from
,
523 unsigned offset
= from
& (PAGE_SIZE
- 1);
524 pgoff_t index
= from
>> PAGE_SHIFT
;
525 struct address_space
*mapping
= inode
->i_mapping
;
528 if (!offset
&& !cache_only
)
532 page
= f2fs_grab_cache_page(mapping
, index
, false);
533 if (page
&& PageUptodate(page
))
535 f2fs_put_page(page
, 1);
539 page
= get_lock_data_page(inode
, index
, true);
543 f2fs_wait_on_page_writeback(page
, DATA
, true);
544 zero_user(page
, offset
, PAGE_SIZE
- offset
);
545 if (!cache_only
|| !f2fs_encrypted_inode(inode
) ||
546 !S_ISREG(inode
->i_mode
))
547 set_page_dirty(page
);
548 f2fs_put_page(page
, 1);
552 int truncate_blocks(struct inode
*inode
, u64 from
, bool lock
)
554 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
555 unsigned int blocksize
= inode
->i_sb
->s_blocksize
;
556 struct dnode_of_data dn
;
558 int count
= 0, err
= 0;
560 bool truncate_page
= false;
562 trace_f2fs_truncate_blocks_enter(inode
, from
);
564 free_from
= (pgoff_t
)F2FS_BYTES_TO_BLK(from
+ blocksize
- 1);
569 ipage
= get_node_page(sbi
, inode
->i_ino
);
571 err
= PTR_ERR(ipage
);
575 if (f2fs_has_inline_data(inode
)) {
576 if (truncate_inline_inode(ipage
, from
))
577 set_page_dirty(ipage
);
578 f2fs_put_page(ipage
, 1);
579 truncate_page
= true;
583 set_new_dnode(&dn
, inode
, ipage
, NULL
, 0);
584 err
= get_dnode_of_data(&dn
, free_from
, LOOKUP_NODE
);
591 count
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
593 count
-= dn
.ofs_in_node
;
594 f2fs_bug_on(sbi
, count
< 0);
596 if (dn
.ofs_in_node
|| IS_INODE(dn
.node_page
)) {
597 truncate_data_blocks_range(&dn
, count
);
603 err
= truncate_inode_blocks(inode
, free_from
);
608 /* lastly zero out the first data page */
610 err
= truncate_partial_data_page(inode
, from
, truncate_page
);
612 trace_f2fs_truncate_blocks_exit(inode
, err
);
616 int f2fs_truncate(struct inode
*inode
, bool lock
)
620 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
621 S_ISLNK(inode
->i_mode
)))
624 trace_f2fs_truncate(inode
);
626 /* we should check inline_data size */
627 if (!f2fs_may_inline_data(inode
)) {
628 err
= f2fs_convert_inline_inode(inode
);
633 err
= truncate_blocks(inode
, i_size_read(inode
), lock
);
637 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
638 mark_inode_dirty(inode
);
642 int f2fs_getattr(struct vfsmount
*mnt
,
643 struct dentry
*dentry
, struct kstat
*stat
)
645 struct inode
*inode
= d_inode(dentry
);
646 generic_fillattr(inode
, stat
);
651 #ifdef CONFIG_F2FS_FS_POSIX_ACL
652 static void __setattr_copy(struct inode
*inode
, const struct iattr
*attr
)
654 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
655 unsigned int ia_valid
= attr
->ia_valid
;
657 if (ia_valid
& ATTR_UID
)
658 inode
->i_uid
= attr
->ia_uid
;
659 if (ia_valid
& ATTR_GID
)
660 inode
->i_gid
= attr
->ia_gid
;
661 if (ia_valid
& ATTR_ATIME
)
662 inode
->i_atime
= timespec_trunc(attr
->ia_atime
,
663 inode
->i_sb
->s_time_gran
);
664 if (ia_valid
& ATTR_MTIME
)
665 inode
->i_mtime
= timespec_trunc(attr
->ia_mtime
,
666 inode
->i_sb
->s_time_gran
);
667 if (ia_valid
& ATTR_CTIME
)
668 inode
->i_ctime
= timespec_trunc(attr
->ia_ctime
,
669 inode
->i_sb
->s_time_gran
);
670 if (ia_valid
& ATTR_MODE
) {
671 umode_t mode
= attr
->ia_mode
;
673 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
675 set_acl_inode(fi
, mode
);
679 #define __setattr_copy setattr_copy
682 int f2fs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
684 struct inode
*inode
= d_inode(dentry
);
685 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
688 err
= inode_change_ok(inode
, attr
);
692 if (attr
->ia_valid
& ATTR_SIZE
) {
693 if (f2fs_encrypted_inode(inode
) &&
694 fscrypt_get_encryption_info(inode
))
697 if (attr
->ia_size
<= i_size_read(inode
)) {
698 truncate_setsize(inode
, attr
->ia_size
);
699 err
= f2fs_truncate(inode
, true);
702 f2fs_balance_fs(F2FS_I_SB(inode
), true);
705 * do not trim all blocks after i_size if target size is
706 * larger than i_size.
708 truncate_setsize(inode
, attr
->ia_size
);
710 /* should convert inline inode here */
711 if (!f2fs_may_inline_data(inode
)) {
712 err
= f2fs_convert_inline_inode(inode
);
716 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
720 __setattr_copy(inode
, attr
);
722 if (attr
->ia_valid
& ATTR_MODE
) {
723 err
= posix_acl_chmod(inode
, get_inode_mode(inode
));
724 if (err
|| is_inode_flag_set(fi
, FI_ACL_MODE
)) {
725 inode
->i_mode
= fi
->i_acl_mode
;
726 clear_inode_flag(fi
, FI_ACL_MODE
);
730 mark_inode_dirty(inode
);
734 const struct inode_operations f2fs_file_inode_operations
= {
735 .getattr
= f2fs_getattr
,
736 .setattr
= f2fs_setattr
,
737 .get_acl
= f2fs_get_acl
,
738 .set_acl
= f2fs_set_acl
,
739 #ifdef CONFIG_F2FS_FS_XATTR
740 .setxattr
= generic_setxattr
,
741 .getxattr
= generic_getxattr
,
742 .listxattr
= f2fs_listxattr
,
743 .removexattr
= generic_removexattr
,
745 .fiemap
= f2fs_fiemap
,
748 static int fill_zero(struct inode
*inode
, pgoff_t index
,
749 loff_t start
, loff_t len
)
751 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
757 f2fs_balance_fs(sbi
, true);
760 page
= get_new_data_page(inode
, NULL
, index
, false);
764 return PTR_ERR(page
);
766 f2fs_wait_on_page_writeback(page
, DATA
, true);
767 zero_user(page
, start
, len
);
768 set_page_dirty(page
);
769 f2fs_put_page(page
, 1);
773 int truncate_hole(struct inode
*inode
, pgoff_t pg_start
, pgoff_t pg_end
)
777 while (pg_start
< pg_end
) {
778 struct dnode_of_data dn
;
779 pgoff_t end_offset
, count
;
781 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
782 err
= get_dnode_of_data(&dn
, pg_start
, LOOKUP_NODE
);
784 if (err
== -ENOENT
) {
791 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
792 count
= min(end_offset
- dn
.ofs_in_node
, pg_end
- pg_start
);
794 f2fs_bug_on(F2FS_I_SB(inode
), count
== 0 || count
> end_offset
);
796 truncate_data_blocks_range(&dn
, count
);
804 static int punch_hole(struct inode
*inode
, loff_t offset
, loff_t len
)
806 pgoff_t pg_start
, pg_end
;
807 loff_t off_start
, off_end
;
810 ret
= f2fs_convert_inline_inode(inode
);
814 pg_start
= ((unsigned long long) offset
) >> PAGE_SHIFT
;
815 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_SHIFT
;
817 off_start
= offset
& (PAGE_SIZE
- 1);
818 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
820 if (pg_start
== pg_end
) {
821 ret
= fill_zero(inode
, pg_start
, off_start
,
822 off_end
- off_start
);
827 ret
= fill_zero(inode
, pg_start
++, off_start
,
828 PAGE_SIZE
- off_start
);
833 ret
= fill_zero(inode
, pg_end
, 0, off_end
);
838 if (pg_start
< pg_end
) {
839 struct address_space
*mapping
= inode
->i_mapping
;
840 loff_t blk_start
, blk_end
;
841 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
843 f2fs_balance_fs(sbi
, true);
845 blk_start
= (loff_t
)pg_start
<< PAGE_SHIFT
;
846 blk_end
= (loff_t
)pg_end
<< PAGE_SHIFT
;
847 truncate_inode_pages_range(mapping
, blk_start
,
851 ret
= truncate_hole(inode
, pg_start
, pg_end
);
859 static int __exchange_data_block(struct inode
*inode
, pgoff_t src
,
860 pgoff_t dst
, bool full
)
862 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
863 struct dnode_of_data dn
;
865 bool do_replace
= false;
868 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
869 ret
= get_dnode_of_data(&dn
, src
, LOOKUP_NODE_RA
);
870 if (ret
&& ret
!= -ENOENT
) {
872 } else if (ret
== -ENOENT
) {
873 new_addr
= NULL_ADDR
;
875 new_addr
= dn
.data_blkaddr
;
876 if (!is_checkpointed_data(sbi
, new_addr
)) {
877 /* do not invalidate this block address */
878 f2fs_update_data_blkaddr(&dn
, NULL_ADDR
);
884 if (new_addr
== NULL_ADDR
)
885 return full
? truncate_hole(inode
, dst
, dst
+ 1) : 0;
888 struct page
*ipage
= get_node_page(sbi
, inode
->i_ino
);
892 ret
= PTR_ERR(ipage
);
896 set_new_dnode(&dn
, inode
, ipage
, NULL
, 0);
897 ret
= f2fs_reserve_block(&dn
, dst
);
901 truncate_data_blocks_range(&dn
, 1);
903 get_node_info(sbi
, dn
.nid
, &ni
);
904 f2fs_replace_block(sbi
, &dn
, dn
.data_blkaddr
, new_addr
,
905 ni
.version
, true, false);
908 struct page
*psrc
, *pdst
;
910 psrc
= get_lock_data_page(inode
, src
, true);
912 return PTR_ERR(psrc
);
913 pdst
= get_new_data_page(inode
, NULL
, dst
, true);
915 f2fs_put_page(psrc
, 1);
916 return PTR_ERR(pdst
);
918 f2fs_copy_page(psrc
, pdst
);
919 set_page_dirty(pdst
);
920 f2fs_put_page(pdst
, 1);
921 f2fs_put_page(psrc
, 1);
923 return truncate_hole(inode
, src
, src
+ 1);
928 if (!get_dnode_of_data(&dn
, src
, LOOKUP_NODE
)) {
929 f2fs_update_data_blkaddr(&dn
, new_addr
);
935 static int f2fs_do_collapse(struct inode
*inode
, pgoff_t start
, pgoff_t end
)
937 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
938 pgoff_t nrpages
= (i_size_read(inode
) + PAGE_SIZE
- 1) / PAGE_SIZE
;
941 for (; end
< nrpages
; start
++, end
++) {
942 f2fs_balance_fs(sbi
, true);
944 ret
= __exchange_data_block(inode
, end
, start
, true);
952 static int f2fs_collapse_range(struct inode
*inode
, loff_t offset
, loff_t len
)
954 pgoff_t pg_start
, pg_end
;
958 if (offset
+ len
>= i_size_read(inode
))
961 /* collapse range should be aligned to block size of f2fs. */
962 if (offset
& (F2FS_BLKSIZE
- 1) || len
& (F2FS_BLKSIZE
- 1))
965 ret
= f2fs_convert_inline_inode(inode
);
969 pg_start
= offset
>> PAGE_SHIFT
;
970 pg_end
= (offset
+ len
) >> PAGE_SHIFT
;
972 /* write out all dirty pages from offset */
973 ret
= filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
977 truncate_pagecache(inode
, offset
);
979 ret
= f2fs_do_collapse(inode
, pg_start
, pg_end
);
983 /* write out all moved pages, if possible */
984 filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
985 truncate_pagecache(inode
, offset
);
987 new_size
= i_size_read(inode
) - len
;
988 truncate_pagecache(inode
, new_size
);
990 ret
= truncate_blocks(inode
, new_size
, true);
992 i_size_write(inode
, new_size
);
997 static int f2fs_zero_range(struct inode
*inode
, loff_t offset
, loff_t len
,
1000 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1001 struct address_space
*mapping
= inode
->i_mapping
;
1002 pgoff_t index
, pg_start
, pg_end
;
1003 loff_t new_size
= i_size_read(inode
);
1004 loff_t off_start
, off_end
;
1007 ret
= inode_newsize_ok(inode
, (len
+ offset
));
1011 ret
= f2fs_convert_inline_inode(inode
);
1015 ret
= filemap_write_and_wait_range(mapping
, offset
, offset
+ len
- 1);
1019 truncate_pagecache_range(inode
, offset
, offset
+ len
- 1);
1021 pg_start
= ((unsigned long long) offset
) >> PAGE_SHIFT
;
1022 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_SHIFT
;
1024 off_start
= offset
& (PAGE_SIZE
- 1);
1025 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
1027 if (pg_start
== pg_end
) {
1028 ret
= fill_zero(inode
, pg_start
, off_start
,
1029 off_end
- off_start
);
1033 if (offset
+ len
> new_size
)
1034 new_size
= offset
+ len
;
1035 new_size
= max_t(loff_t
, new_size
, offset
+ len
);
1038 ret
= fill_zero(inode
, pg_start
++, off_start
,
1039 PAGE_SIZE
- off_start
);
1043 new_size
= max_t(loff_t
, new_size
,
1044 (loff_t
)pg_start
<< PAGE_SHIFT
);
1047 for (index
= pg_start
; index
< pg_end
; index
++) {
1048 struct dnode_of_data dn
;
1053 ipage
= get_node_page(sbi
, inode
->i_ino
);
1054 if (IS_ERR(ipage
)) {
1055 ret
= PTR_ERR(ipage
);
1056 f2fs_unlock_op(sbi
);
1060 set_new_dnode(&dn
, inode
, ipage
, NULL
, 0);
1061 ret
= f2fs_reserve_block(&dn
, index
);
1063 f2fs_unlock_op(sbi
);
1067 if (dn
.data_blkaddr
!= NEW_ADDR
) {
1068 invalidate_blocks(sbi
, dn
.data_blkaddr
);
1069 f2fs_update_data_blkaddr(&dn
, NEW_ADDR
);
1071 f2fs_put_dnode(&dn
);
1072 f2fs_unlock_op(sbi
);
1074 new_size
= max_t(loff_t
, new_size
,
1075 (loff_t
)(index
+ 1) << PAGE_SHIFT
);
1079 ret
= fill_zero(inode
, pg_end
, 0, off_end
);
1083 new_size
= max_t(loff_t
, new_size
, offset
+ len
);
1088 if (!(mode
& FALLOC_FL_KEEP_SIZE
) && i_size_read(inode
) < new_size
) {
1089 i_size_write(inode
, new_size
);
1090 mark_inode_dirty(inode
);
1091 update_inode_page(inode
);
1097 static int f2fs_insert_range(struct inode
*inode
, loff_t offset
, loff_t len
)
1099 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1100 pgoff_t pg_start
, pg_end
, delta
, nrpages
, idx
;
1104 new_size
= i_size_read(inode
) + len
;
1105 if (new_size
> inode
->i_sb
->s_maxbytes
)
1108 if (offset
>= i_size_read(inode
))
1111 /* insert range should be aligned to block size of f2fs. */
1112 if (offset
& (F2FS_BLKSIZE
- 1) || len
& (F2FS_BLKSIZE
- 1))
1115 ret
= f2fs_convert_inline_inode(inode
);
1119 f2fs_balance_fs(sbi
, true);
1121 ret
= truncate_blocks(inode
, i_size_read(inode
), true);
1125 /* write out all dirty pages from offset */
1126 ret
= filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
1130 truncate_pagecache(inode
, offset
);
1132 pg_start
= offset
>> PAGE_SHIFT
;
1133 pg_end
= (offset
+ len
) >> PAGE_SHIFT
;
1134 delta
= pg_end
- pg_start
;
1135 nrpages
= (i_size_read(inode
) + PAGE_SIZE
- 1) / PAGE_SIZE
;
1137 for (idx
= nrpages
- 1; idx
>= pg_start
&& idx
!= -1; idx
--) {
1139 ret
= __exchange_data_block(inode
, idx
, idx
+ delta
, false);
1140 f2fs_unlock_op(sbi
);
1145 /* write out all moved pages, if possible */
1146 filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
1147 truncate_pagecache(inode
, offset
);
1150 i_size_write(inode
, new_size
);
1154 static int expand_inode_data(struct inode
*inode
, loff_t offset
,
1155 loff_t len
, int mode
)
1157 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1158 pgoff_t index
, pg_start
, pg_end
;
1159 loff_t new_size
= i_size_read(inode
);
1160 loff_t off_start
, off_end
;
1163 ret
= inode_newsize_ok(inode
, (len
+ offset
));
1167 ret
= f2fs_convert_inline_inode(inode
);
1171 f2fs_balance_fs(sbi
, true);
1173 pg_start
= ((unsigned long long) offset
) >> PAGE_SHIFT
;
1174 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_SHIFT
;
1176 off_start
= offset
& (PAGE_SIZE
- 1);
1177 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
1181 for (index
= pg_start
; index
<= pg_end
; index
++) {
1182 struct dnode_of_data dn
;
1184 if (index
== pg_end
&& !off_end
)
1187 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
1188 ret
= f2fs_reserve_block(&dn
, index
);
1192 if (pg_start
== pg_end
)
1193 new_size
= offset
+ len
;
1194 else if (index
== pg_start
&& off_start
)
1195 new_size
= (loff_t
)(index
+ 1) << PAGE_SHIFT
;
1196 else if (index
== pg_end
)
1197 new_size
= ((loff_t
)index
<< PAGE_SHIFT
) +
1200 new_size
+= PAGE_SIZE
;
1203 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
1204 i_size_read(inode
) < new_size
) {
1205 i_size_write(inode
, new_size
);
1206 mark_inode_dirty(inode
);
1207 update_inode_page(inode
);
1209 f2fs_unlock_op(sbi
);
1214 static long f2fs_fallocate(struct file
*file
, int mode
,
1215 loff_t offset
, loff_t len
)
1217 struct inode
*inode
= file_inode(file
);
1220 /* f2fs only support ->fallocate for regular file */
1221 if (!S_ISREG(inode
->i_mode
))
1224 if (f2fs_encrypted_inode(inode
) &&
1225 (mode
& (FALLOC_FL_COLLAPSE_RANGE
| FALLOC_FL_INSERT_RANGE
)))
1228 if (mode
& ~(FALLOC_FL_KEEP_SIZE
| FALLOC_FL_PUNCH_HOLE
|
1229 FALLOC_FL_COLLAPSE_RANGE
| FALLOC_FL_ZERO_RANGE
|
1230 FALLOC_FL_INSERT_RANGE
))
1235 if (mode
& FALLOC_FL_PUNCH_HOLE
) {
1236 if (offset
>= inode
->i_size
)
1239 ret
= punch_hole(inode
, offset
, len
);
1240 } else if (mode
& FALLOC_FL_COLLAPSE_RANGE
) {
1241 ret
= f2fs_collapse_range(inode
, offset
, len
);
1242 } else if (mode
& FALLOC_FL_ZERO_RANGE
) {
1243 ret
= f2fs_zero_range(inode
, offset
, len
, mode
);
1244 } else if (mode
& FALLOC_FL_INSERT_RANGE
) {
1245 ret
= f2fs_insert_range(inode
, offset
, len
);
1247 ret
= expand_inode_data(inode
, offset
, len
, mode
);
1251 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
1252 mark_inode_dirty(inode
);
1253 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1257 inode_unlock(inode
);
1259 trace_f2fs_fallocate(inode
, mode
, offset
, len
, ret
);
1263 static int f2fs_release_file(struct inode
*inode
, struct file
*filp
)
1266 * f2fs_relase_file is called at every close calls. So we should
1267 * not drop any inmemory pages by close called by other process.
1269 if (!(filp
->f_mode
& FMODE_WRITE
) ||
1270 atomic_read(&inode
->i_writecount
) != 1)
1273 /* some remained atomic pages should discarded */
1274 if (f2fs_is_atomic_file(inode
))
1275 drop_inmem_pages(inode
);
1276 if (f2fs_is_volatile_file(inode
)) {
1277 clear_inode_flag(F2FS_I(inode
), FI_VOLATILE_FILE
);
1278 set_inode_flag(F2FS_I(inode
), FI_DROP_CACHE
);
1279 filemap_fdatawrite(inode
->i_mapping
);
1280 clear_inode_flag(F2FS_I(inode
), FI_DROP_CACHE
);
1285 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1286 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1288 static inline __u32
f2fs_mask_flags(umode_t mode
, __u32 flags
)
1292 else if (S_ISREG(mode
))
1293 return flags
& F2FS_REG_FLMASK
;
1295 return flags
& F2FS_OTHER_FLMASK
;
1298 static int f2fs_ioc_getflags(struct file
*filp
, unsigned long arg
)
1300 struct inode
*inode
= file_inode(filp
);
1301 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
1302 unsigned int flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
1303 return put_user(flags
, (int __user
*)arg
);
1306 static int f2fs_ioc_setflags(struct file
*filp
, unsigned long arg
)
1308 struct inode
*inode
= file_inode(filp
);
1309 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
1310 unsigned int flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
1311 unsigned int oldflags
;
1314 ret
= mnt_want_write_file(filp
);
1318 if (!inode_owner_or_capable(inode
)) {
1323 if (get_user(flags
, (int __user
*)arg
)) {
1328 flags
= f2fs_mask_flags(inode
->i_mode
, flags
);
1332 oldflags
= fi
->i_flags
;
1334 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
1335 if (!capable(CAP_LINUX_IMMUTABLE
)) {
1336 inode_unlock(inode
);
1342 flags
= flags
& FS_FL_USER_MODIFIABLE
;
1343 flags
|= oldflags
& ~FS_FL_USER_MODIFIABLE
;
1344 fi
->i_flags
= flags
;
1345 inode_unlock(inode
);
1347 f2fs_set_inode_flags(inode
);
1348 inode
->i_ctime
= CURRENT_TIME
;
1349 mark_inode_dirty(inode
);
1351 mnt_drop_write_file(filp
);
1355 static int f2fs_ioc_getversion(struct file
*filp
, unsigned long arg
)
1357 struct inode
*inode
= file_inode(filp
);
1359 return put_user(inode
->i_generation
, (int __user
*)arg
);
1362 static int f2fs_ioc_start_atomic_write(struct file
*filp
)
1364 struct inode
*inode
= file_inode(filp
);
1367 if (!inode_owner_or_capable(inode
))
1370 if (f2fs_is_atomic_file(inode
))
1373 ret
= f2fs_convert_inline_inode(inode
);
1377 set_inode_flag(F2FS_I(inode
), FI_ATOMIC_FILE
);
1378 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1380 if (!get_dirty_pages(inode
))
1383 f2fs_msg(F2FS_I_SB(inode
)->sb
, KERN_WARNING
,
1384 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1385 inode
->i_ino
, get_dirty_pages(inode
));
1386 ret
= filemap_write_and_wait_range(inode
->i_mapping
, 0, LLONG_MAX
);
1388 clear_inode_flag(F2FS_I(inode
), FI_ATOMIC_FILE
);
1392 static int f2fs_ioc_commit_atomic_write(struct file
*filp
)
1394 struct inode
*inode
= file_inode(filp
);
1397 if (!inode_owner_or_capable(inode
))
1400 if (f2fs_is_volatile_file(inode
))
1403 ret
= mnt_want_write_file(filp
);
1407 if (f2fs_is_atomic_file(inode
)) {
1408 clear_inode_flag(F2FS_I(inode
), FI_ATOMIC_FILE
);
1409 ret
= commit_inmem_pages(inode
);
1411 set_inode_flag(F2FS_I(inode
), FI_ATOMIC_FILE
);
1416 ret
= f2fs_do_sync_file(filp
, 0, LLONG_MAX
, 0, true);
1418 mnt_drop_write_file(filp
);
1422 static int f2fs_ioc_start_volatile_write(struct file
*filp
)
1424 struct inode
*inode
= file_inode(filp
);
1427 if (!inode_owner_or_capable(inode
))
1430 if (f2fs_is_volatile_file(inode
))
1433 ret
= f2fs_convert_inline_inode(inode
);
1437 set_inode_flag(F2FS_I(inode
), FI_VOLATILE_FILE
);
1438 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1442 static int f2fs_ioc_release_volatile_write(struct file
*filp
)
1444 struct inode
*inode
= file_inode(filp
);
1446 if (!inode_owner_or_capable(inode
))
1449 if (!f2fs_is_volatile_file(inode
))
1452 if (!f2fs_is_first_block_written(inode
))
1453 return truncate_partial_data_page(inode
, 0, true);
1455 return punch_hole(inode
, 0, F2FS_BLKSIZE
);
1458 static int f2fs_ioc_abort_volatile_write(struct file
*filp
)
1460 struct inode
*inode
= file_inode(filp
);
1463 if (!inode_owner_or_capable(inode
))
1466 ret
= mnt_want_write_file(filp
);
1470 if (f2fs_is_atomic_file(inode
))
1471 drop_inmem_pages(inode
);
1472 if (f2fs_is_volatile_file(inode
)) {
1473 clear_inode_flag(F2FS_I(inode
), FI_VOLATILE_FILE
);
1474 ret
= f2fs_do_sync_file(filp
, 0, LLONG_MAX
, 0, true);
1477 mnt_drop_write_file(filp
);
1478 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1482 static int f2fs_ioc_shutdown(struct file
*filp
, unsigned long arg
)
1484 struct inode
*inode
= file_inode(filp
);
1485 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1486 struct super_block
*sb
= sbi
->sb
;
1489 if (!capable(CAP_SYS_ADMIN
))
1492 if (get_user(in
, (__u32 __user
*)arg
))
1496 case F2FS_GOING_DOWN_FULLSYNC
:
1497 sb
= freeze_bdev(sb
->s_bdev
);
1498 if (sb
&& !IS_ERR(sb
)) {
1499 f2fs_stop_checkpoint(sbi
);
1500 thaw_bdev(sb
->s_bdev
, sb
);
1503 case F2FS_GOING_DOWN_METASYNC
:
1504 /* do checkpoint only */
1505 f2fs_sync_fs(sb
, 1);
1506 f2fs_stop_checkpoint(sbi
);
1508 case F2FS_GOING_DOWN_NOSYNC
:
1509 f2fs_stop_checkpoint(sbi
);
1511 case F2FS_GOING_DOWN_METAFLUSH
:
1512 sync_meta_pages(sbi
, META
, LONG_MAX
);
1513 f2fs_stop_checkpoint(sbi
);
1518 f2fs_update_time(sbi
, REQ_TIME
);
1522 static int f2fs_ioc_fitrim(struct file
*filp
, unsigned long arg
)
1524 struct inode
*inode
= file_inode(filp
);
1525 struct super_block
*sb
= inode
->i_sb
;
1526 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
1527 struct fstrim_range range
;
1530 if (!capable(CAP_SYS_ADMIN
))
1533 if (!blk_queue_discard(q
))
1536 if (copy_from_user(&range
, (struct fstrim_range __user
*)arg
,
1540 range
.minlen
= max((unsigned int)range
.minlen
,
1541 q
->limits
.discard_granularity
);
1542 ret
= f2fs_trim_fs(F2FS_SB(sb
), &range
);
1546 if (copy_to_user((struct fstrim_range __user
*)arg
, &range
,
1549 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1553 static bool uuid_is_nonzero(__u8 u
[16])
1557 for (i
= 0; i
< 16; i
++)
1563 static int f2fs_ioc_set_encryption_policy(struct file
*filp
, unsigned long arg
)
1565 struct fscrypt_policy policy
;
1566 struct inode
*inode
= file_inode(filp
);
1568 if (copy_from_user(&policy
, (struct fscrypt_policy __user
*)arg
,
1572 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1573 return fscrypt_process_policy(inode
, &policy
);
1576 static int f2fs_ioc_get_encryption_policy(struct file
*filp
, unsigned long arg
)
1578 struct fscrypt_policy policy
;
1579 struct inode
*inode
= file_inode(filp
);
1582 err
= fscrypt_get_policy(inode
, &policy
);
1586 if (copy_to_user((struct fscrypt_policy __user
*)arg
, &policy
, sizeof(policy
)))
1591 static int f2fs_ioc_get_encryption_pwsalt(struct file
*filp
, unsigned long arg
)
1593 struct inode
*inode
= file_inode(filp
);
1594 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1597 if (!f2fs_sb_has_crypto(inode
->i_sb
))
1600 if (uuid_is_nonzero(sbi
->raw_super
->encrypt_pw_salt
))
1603 err
= mnt_want_write_file(filp
);
1607 /* update superblock with uuid */
1608 generate_random_uuid(sbi
->raw_super
->encrypt_pw_salt
);
1610 err
= f2fs_commit_super(sbi
, false);
1613 memset(sbi
->raw_super
->encrypt_pw_salt
, 0, 16);
1614 mnt_drop_write_file(filp
);
1617 mnt_drop_write_file(filp
);
1619 if (copy_to_user((__u8 __user
*)arg
, sbi
->raw_super
->encrypt_pw_salt
,
1625 static int f2fs_ioc_gc(struct file
*filp
, unsigned long arg
)
1627 struct inode
*inode
= file_inode(filp
);
1628 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1631 if (!capable(CAP_SYS_ADMIN
))
1634 if (get_user(sync
, (__u32 __user
*)arg
))
1637 if (f2fs_readonly(sbi
->sb
))
1641 if (!mutex_trylock(&sbi
->gc_mutex
))
1644 mutex_lock(&sbi
->gc_mutex
);
1647 return f2fs_gc(sbi
, sync
);
1650 static int f2fs_ioc_write_checkpoint(struct file
*filp
, unsigned long arg
)
1652 struct inode
*inode
= file_inode(filp
);
1653 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1655 if (!capable(CAP_SYS_ADMIN
))
1658 if (f2fs_readonly(sbi
->sb
))
1661 return f2fs_sync_fs(sbi
->sb
, 1);
1664 static int f2fs_defragment_range(struct f2fs_sb_info
*sbi
,
1666 struct f2fs_defragment
*range
)
1668 struct inode
*inode
= file_inode(filp
);
1669 struct f2fs_map_blocks map
= { .m_next_pgofs
= NULL
};
1670 struct extent_info ei
;
1671 pgoff_t pg_start
, pg_end
;
1672 unsigned int blk_per_seg
= sbi
->blocks_per_seg
;
1673 unsigned int total
= 0, sec_num
;
1674 unsigned int pages_per_sec
= sbi
->segs_per_sec
* blk_per_seg
;
1675 block_t blk_end
= 0;
1676 bool fragmented
= false;
1679 /* if in-place-update policy is enabled, don't waste time here */
1680 if (need_inplace_update(inode
))
1683 pg_start
= range
->start
>> PAGE_SHIFT
;
1684 pg_end
= (range
->start
+ range
->len
) >> PAGE_SHIFT
;
1686 f2fs_balance_fs(sbi
, true);
1690 /* writeback all dirty pages in the range */
1691 err
= filemap_write_and_wait_range(inode
->i_mapping
, range
->start
,
1692 range
->start
+ range
->len
- 1);
1697 * lookup mapping info in extent cache, skip defragmenting if physical
1698 * block addresses are continuous.
1700 if (f2fs_lookup_extent_cache(inode
, pg_start
, &ei
)) {
1701 if (ei
.fofs
+ ei
.len
>= pg_end
)
1705 map
.m_lblk
= pg_start
;
1708 * lookup mapping info in dnode page cache, skip defragmenting if all
1709 * physical block addresses are continuous even if there are hole(s)
1710 * in logical blocks.
1712 while (map
.m_lblk
< pg_end
) {
1713 map
.m_len
= pg_end
- map
.m_lblk
;
1714 err
= f2fs_map_blocks(inode
, &map
, 0, F2FS_GET_BLOCK_READ
);
1718 if (!(map
.m_flags
& F2FS_MAP_FLAGS
)) {
1723 if (blk_end
&& blk_end
!= map
.m_pblk
) {
1727 blk_end
= map
.m_pblk
+ map
.m_len
;
1729 map
.m_lblk
+= map
.m_len
;
1735 map
.m_lblk
= pg_start
;
1736 map
.m_len
= pg_end
- pg_start
;
1738 sec_num
= (map
.m_len
+ pages_per_sec
- 1) / pages_per_sec
;
1741 * make sure there are enough free section for LFS allocation, this can
1742 * avoid defragment running in SSR mode when free section are allocated
1745 if (has_not_enough_free_secs(sbi
, sec_num
)) {
1750 while (map
.m_lblk
< pg_end
) {
1755 map
.m_len
= pg_end
- map
.m_lblk
;
1756 err
= f2fs_map_blocks(inode
, &map
, 0, F2FS_GET_BLOCK_READ
);
1760 if (!(map
.m_flags
& F2FS_MAP_FLAGS
)) {
1765 set_inode_flag(F2FS_I(inode
), FI_DO_DEFRAG
);
1768 while (idx
< map
.m_lblk
+ map
.m_len
&& cnt
< blk_per_seg
) {
1771 page
= get_lock_data_page(inode
, idx
, true);
1773 err
= PTR_ERR(page
);
1777 set_page_dirty(page
);
1778 f2fs_put_page(page
, 1);
1787 if (idx
< pg_end
&& cnt
< blk_per_seg
)
1790 clear_inode_flag(F2FS_I(inode
), FI_DO_DEFRAG
);
1792 err
= filemap_fdatawrite(inode
->i_mapping
);
1797 clear_inode_flag(F2FS_I(inode
), FI_DO_DEFRAG
);
1799 inode_unlock(inode
);
1801 range
->len
= (u64
)total
<< PAGE_SHIFT
;
1805 static int f2fs_ioc_defragment(struct file
*filp
, unsigned long arg
)
1807 struct inode
*inode
= file_inode(filp
);
1808 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1809 struct f2fs_defragment range
;
1812 if (!capable(CAP_SYS_ADMIN
))
1815 if (!S_ISREG(inode
->i_mode
))
1818 err
= mnt_want_write_file(filp
);
1822 if (f2fs_readonly(sbi
->sb
)) {
1827 if (copy_from_user(&range
, (struct f2fs_defragment __user
*)arg
,
1833 /* verify alignment of offset & size */
1834 if (range
.start
& (F2FS_BLKSIZE
- 1) ||
1835 range
.len
& (F2FS_BLKSIZE
- 1)) {
1840 err
= f2fs_defragment_range(sbi
, filp
, &range
);
1841 f2fs_update_time(sbi
, REQ_TIME
);
1845 if (copy_to_user((struct f2fs_defragment __user
*)arg
, &range
,
1849 mnt_drop_write_file(filp
);
1853 long f2fs_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
1856 case F2FS_IOC_GETFLAGS
:
1857 return f2fs_ioc_getflags(filp
, arg
);
1858 case F2FS_IOC_SETFLAGS
:
1859 return f2fs_ioc_setflags(filp
, arg
);
1860 case F2FS_IOC_GETVERSION
:
1861 return f2fs_ioc_getversion(filp
, arg
);
1862 case F2FS_IOC_START_ATOMIC_WRITE
:
1863 return f2fs_ioc_start_atomic_write(filp
);
1864 case F2FS_IOC_COMMIT_ATOMIC_WRITE
:
1865 return f2fs_ioc_commit_atomic_write(filp
);
1866 case F2FS_IOC_START_VOLATILE_WRITE
:
1867 return f2fs_ioc_start_volatile_write(filp
);
1868 case F2FS_IOC_RELEASE_VOLATILE_WRITE
:
1869 return f2fs_ioc_release_volatile_write(filp
);
1870 case F2FS_IOC_ABORT_VOLATILE_WRITE
:
1871 return f2fs_ioc_abort_volatile_write(filp
);
1872 case F2FS_IOC_SHUTDOWN
:
1873 return f2fs_ioc_shutdown(filp
, arg
);
1875 return f2fs_ioc_fitrim(filp
, arg
);
1876 case F2FS_IOC_SET_ENCRYPTION_POLICY
:
1877 return f2fs_ioc_set_encryption_policy(filp
, arg
);
1878 case F2FS_IOC_GET_ENCRYPTION_POLICY
:
1879 return f2fs_ioc_get_encryption_policy(filp
, arg
);
1880 case F2FS_IOC_GET_ENCRYPTION_PWSALT
:
1881 return f2fs_ioc_get_encryption_pwsalt(filp
, arg
);
1882 case F2FS_IOC_GARBAGE_COLLECT
:
1883 return f2fs_ioc_gc(filp
, arg
);
1884 case F2FS_IOC_WRITE_CHECKPOINT
:
1885 return f2fs_ioc_write_checkpoint(filp
, arg
);
1886 case F2FS_IOC_DEFRAGMENT
:
1887 return f2fs_ioc_defragment(filp
, arg
);
1893 static ssize_t
f2fs_file_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1895 struct file
*file
= iocb
->ki_filp
;
1896 struct inode
*inode
= file_inode(file
);
1899 if (f2fs_encrypted_inode(inode
) &&
1900 !fscrypt_has_encryption_key(inode
) &&
1901 fscrypt_get_encryption_info(inode
))
1905 ret
= generic_write_checks(iocb
, from
);
1907 ret
= f2fs_preallocate_blocks(iocb
, from
);
1909 ret
= __generic_file_write_iter(iocb
, from
);
1911 inode_unlock(inode
);
1916 err
= generic_write_sync(file
, iocb
->ki_pos
- ret
, ret
);
1923 #ifdef CONFIG_COMPAT
1924 long f2fs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
1927 case F2FS_IOC32_GETFLAGS
:
1928 cmd
= F2FS_IOC_GETFLAGS
;
1930 case F2FS_IOC32_SETFLAGS
:
1931 cmd
= F2FS_IOC_SETFLAGS
;
1933 case F2FS_IOC32_GETVERSION
:
1934 cmd
= F2FS_IOC_GETVERSION
;
1936 case F2FS_IOC_START_ATOMIC_WRITE
:
1937 case F2FS_IOC_COMMIT_ATOMIC_WRITE
:
1938 case F2FS_IOC_START_VOLATILE_WRITE
:
1939 case F2FS_IOC_RELEASE_VOLATILE_WRITE
:
1940 case F2FS_IOC_ABORT_VOLATILE_WRITE
:
1941 case F2FS_IOC_SHUTDOWN
:
1942 case F2FS_IOC_SET_ENCRYPTION_POLICY
:
1943 case F2FS_IOC_GET_ENCRYPTION_PWSALT
:
1944 case F2FS_IOC_GET_ENCRYPTION_POLICY
:
1945 case F2FS_IOC_GARBAGE_COLLECT
:
1946 case F2FS_IOC_WRITE_CHECKPOINT
:
1947 case F2FS_IOC_DEFRAGMENT
:
1950 return -ENOIOCTLCMD
;
1952 return f2fs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));
1956 const struct file_operations f2fs_file_operations
= {
1957 .llseek
= f2fs_llseek
,
1958 .read_iter
= generic_file_read_iter
,
1959 .write_iter
= f2fs_file_write_iter
,
1960 .open
= f2fs_file_open
,
1961 .release
= f2fs_release_file
,
1962 .mmap
= f2fs_file_mmap
,
1963 .fsync
= f2fs_sync_file
,
1964 .fallocate
= f2fs_fallocate
,
1965 .unlocked_ioctl
= f2fs_ioctl
,
1966 #ifdef CONFIG_COMPAT
1967 .compat_ioctl
= f2fs_compat_ioctl
,
1969 .splice_read
= generic_file_splice_read
,
1970 .splice_write
= iter_file_splice_write
,