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/uuid.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 if (!PageUptodate(page
))
85 SetPageUptodate(page
);
87 trace_f2fs_vm_page_mkwrite(page
, DATA
);
90 f2fs_wait_on_page_writeback(page
, DATA
, false);
92 /* wait for GCed encrypted page writeback */
93 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
))
94 f2fs_wait_on_encrypted_page_writeback(sbi
, dn
.data_blkaddr
);
96 /* if gced page is attached, don't write to cold segment */
97 clear_cold_data(page
);
99 sb_end_pagefault(inode
->i_sb
);
100 f2fs_update_time(sbi
, REQ_TIME
);
101 return block_page_mkwrite_return(err
);
104 static const struct vm_operations_struct f2fs_file_vm_ops
= {
105 .fault
= filemap_fault
,
106 .map_pages
= filemap_map_pages
,
107 .page_mkwrite
= f2fs_vm_page_mkwrite
,
110 static int get_parent_ino(struct inode
*inode
, nid_t
*pino
)
112 struct dentry
*dentry
;
114 inode
= igrab(inode
);
115 dentry
= d_find_any_alias(inode
);
120 if (update_dent_inode(inode
, inode
, &dentry
->d_name
)) {
125 *pino
= parent_ino(dentry
);
130 static inline bool need_do_checkpoint(struct inode
*inode
)
132 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
133 bool need_cp
= false;
135 if (!S_ISREG(inode
->i_mode
) || inode
->i_nlink
!= 1)
137 else if (file_enc_name(inode
) && need_dentry_mark(sbi
, inode
->i_ino
))
139 else if (file_wrong_pino(inode
))
141 else if (!space_for_roll_forward(sbi
))
143 else if (!is_checkpointed_node(sbi
, F2FS_I(inode
)->i_pino
))
145 else if (F2FS_I(inode
)->xattr_ver
== cur_cp_version(F2FS_CKPT(sbi
)))
147 else if (test_opt(sbi
, FASTBOOT
))
149 else if (sbi
->active_logs
== 2)
155 static bool need_inode_page_update(struct f2fs_sb_info
*sbi
, nid_t ino
)
157 struct page
*i
= find_get_page(NODE_MAPPING(sbi
), ino
);
159 /* But we need to avoid that there are some inode updates */
160 if ((i
&& PageDirty(i
)) || need_inode_block_update(sbi
, ino
))
166 static void try_to_fix_pino(struct inode
*inode
)
168 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
171 down_write(&fi
->i_sem
);
173 if (file_wrong_pino(inode
) && inode
->i_nlink
== 1 &&
174 get_parent_ino(inode
, &pino
)) {
175 f2fs_i_pino_write(inode
, pino
);
176 file_got_pino(inode
);
178 up_write(&fi
->i_sem
);
181 static int f2fs_do_sync_file(struct file
*file
, loff_t start
, loff_t end
,
182 int datasync
, bool atomic
)
184 struct inode
*inode
= file
->f_mapping
->host
;
185 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
186 nid_t ino
= inode
->i_ino
;
188 bool need_cp
= false;
189 struct writeback_control wbc
= {
190 .sync_mode
= WB_SYNC_ALL
,
191 .nr_to_write
= LONG_MAX
,
195 if (unlikely(f2fs_readonly(inode
->i_sb
)))
198 trace_f2fs_sync_file_enter(inode
);
200 /* if fdatasync is triggered, let's do in-place-update */
201 if (datasync
|| get_dirty_pages(inode
) <= SM_I(sbi
)->min_fsync_blocks
)
202 set_inode_flag(inode
, FI_NEED_IPU
);
203 ret
= filemap_write_and_wait_range(inode
->i_mapping
, start
, end
);
204 clear_inode_flag(inode
, FI_NEED_IPU
);
207 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
211 /* if the inode is dirty, let's recover all the time */
212 if (!datasync
&& !f2fs_skip_inode_update(inode
)) {
213 f2fs_write_inode(inode
, NULL
);
218 * if there is no written data, don't waste time to write recovery info.
220 if (!is_inode_flag_set(inode
, FI_APPEND_WRITE
) &&
221 !exist_written_data(sbi
, ino
, APPEND_INO
)) {
223 /* it may call write_inode just prior to fsync */
224 if (need_inode_page_update(sbi
, ino
))
227 if (is_inode_flag_set(inode
, FI_UPDATE_WRITE
) ||
228 exist_written_data(sbi
, ino
, UPDATE_INO
))
234 * Both of fdatasync() and fsync() are able to be recovered from
237 down_read(&F2FS_I(inode
)->i_sem
);
238 need_cp
= need_do_checkpoint(inode
);
239 up_read(&F2FS_I(inode
)->i_sem
);
242 /* all the dirty node pages should be flushed for POR */
243 ret
= f2fs_sync_fs(inode
->i_sb
, 1);
246 * We've secured consistency through sync_fs. Following pino
247 * will be used only for fsynced inodes after checkpoint.
249 try_to_fix_pino(inode
);
250 clear_inode_flag(inode
, FI_APPEND_WRITE
);
251 clear_inode_flag(inode
, FI_UPDATE_WRITE
);
255 ret
= fsync_node_pages(sbi
, inode
, &wbc
, atomic
);
259 /* if cp_error was enabled, we should avoid infinite loop */
260 if (unlikely(f2fs_cp_error(sbi
))) {
265 if (need_inode_block_update(sbi
, ino
)) {
266 f2fs_mark_inode_dirty_sync(inode
);
267 f2fs_write_inode(inode
, NULL
);
271 ret
= wait_on_node_pages_writeback(sbi
, ino
);
275 /* once recovery info is written, don't need to tack this */
276 remove_ino_entry(sbi
, ino
, APPEND_INO
);
277 clear_inode_flag(inode
, FI_APPEND_WRITE
);
279 remove_ino_entry(sbi
, ino
, UPDATE_INO
);
280 clear_inode_flag(inode
, FI_UPDATE_WRITE
);
281 ret
= f2fs_issue_flush(sbi
);
282 f2fs_update_time(sbi
, REQ_TIME
);
284 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
285 f2fs_trace_ios(NULL
, 1);
289 int f2fs_sync_file(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
291 return f2fs_do_sync_file(file
, start
, end
, datasync
, false);
294 static pgoff_t
__get_first_dirty_index(struct address_space
*mapping
,
295 pgoff_t pgofs
, int whence
)
300 if (whence
!= SEEK_DATA
)
303 /* find first dirty page index */
304 pagevec_init(&pvec
, 0);
305 nr_pages
= pagevec_lookup_tag(&pvec
, mapping
, &pgofs
,
306 PAGECACHE_TAG_DIRTY
, 1);
307 pgofs
= nr_pages
? pvec
.pages
[0]->index
: ULONG_MAX
;
308 pagevec_release(&pvec
);
312 static bool __found_offset(block_t blkaddr
, pgoff_t dirty
, pgoff_t pgofs
,
317 if ((blkaddr
== NEW_ADDR
&& dirty
== pgofs
) ||
318 (blkaddr
!= NEW_ADDR
&& blkaddr
!= NULL_ADDR
))
322 if (blkaddr
== NULL_ADDR
)
329 static loff_t
f2fs_seek_block(struct file
*file
, loff_t offset
, int whence
)
331 struct inode
*inode
= file
->f_mapping
->host
;
332 loff_t maxbytes
= inode
->i_sb
->s_maxbytes
;
333 struct dnode_of_data dn
;
334 pgoff_t pgofs
, end_offset
, dirty
;
335 loff_t data_ofs
= offset
;
341 isize
= i_size_read(inode
);
345 /* handle inline data case */
346 if (f2fs_has_inline_data(inode
) || f2fs_has_inline_dentry(inode
)) {
347 if (whence
== SEEK_HOLE
)
352 pgofs
= (pgoff_t
)(offset
>> PAGE_SHIFT
);
354 dirty
= __get_first_dirty_index(inode
->i_mapping
, pgofs
, whence
);
356 for (; data_ofs
< isize
; data_ofs
= (loff_t
)pgofs
<< PAGE_SHIFT
) {
357 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
358 err
= get_dnode_of_data(&dn
, pgofs
, LOOKUP_NODE
);
359 if (err
&& err
!= -ENOENT
) {
361 } else if (err
== -ENOENT
) {
362 /* direct node does not exists */
363 if (whence
== SEEK_DATA
) {
364 pgofs
= get_next_page_offset(&dn
, pgofs
);
371 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
373 /* find data/hole in dnode block */
374 for (; dn
.ofs_in_node
< end_offset
;
375 dn
.ofs_in_node
++, pgofs
++,
376 data_ofs
= (loff_t
)pgofs
<< PAGE_SHIFT
) {
378 blkaddr
= datablock_addr(dn
.node_page
, dn
.ofs_in_node
);
380 if (__found_offset(blkaddr
, dirty
, pgofs
, whence
)) {
388 if (whence
== SEEK_DATA
)
391 if (whence
== SEEK_HOLE
&& data_ofs
> isize
)
394 return vfs_setpos(file
, data_ofs
, maxbytes
);
400 static loff_t
f2fs_llseek(struct file
*file
, loff_t offset
, int whence
)
402 struct inode
*inode
= file
->f_mapping
->host
;
403 loff_t maxbytes
= inode
->i_sb
->s_maxbytes
;
409 return generic_file_llseek_size(file
, offset
, whence
,
410 maxbytes
, i_size_read(inode
));
415 return f2fs_seek_block(file
, offset
, whence
);
421 static int f2fs_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
423 struct inode
*inode
= file_inode(file
);
426 if (f2fs_encrypted_inode(inode
)) {
427 err
= fscrypt_get_encryption_info(inode
);
430 if (!f2fs_encrypted_inode(inode
))
434 /* we don't need to use inline_data strictly */
435 err
= f2fs_convert_inline_inode(inode
);
440 vma
->vm_ops
= &f2fs_file_vm_ops
;
444 static int f2fs_file_open(struct inode
*inode
, struct file
*filp
)
446 int ret
= generic_file_open(inode
, filp
);
449 if (!ret
&& f2fs_encrypted_inode(inode
)) {
450 ret
= fscrypt_get_encryption_info(inode
);
453 if (!fscrypt_has_encryption_key(inode
))
456 dir
= dget_parent(file_dentry(filp
));
457 if (f2fs_encrypted_inode(d_inode(dir
)) &&
458 !fscrypt_has_permitted_context(d_inode(dir
), inode
)) {
466 int truncate_data_blocks_range(struct dnode_of_data
*dn
, int count
)
468 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
469 struct f2fs_node
*raw_node
;
470 int nr_free
= 0, ofs
= dn
->ofs_in_node
, len
= count
;
473 raw_node
= F2FS_NODE(dn
->node_page
);
474 addr
= blkaddr_in_node(raw_node
) + ofs
;
476 for (; count
> 0; count
--, addr
++, dn
->ofs_in_node
++) {
477 block_t blkaddr
= le32_to_cpu(*addr
);
478 if (blkaddr
== NULL_ADDR
)
481 dn
->data_blkaddr
= NULL_ADDR
;
482 set_data_blkaddr(dn
);
483 invalidate_blocks(sbi
, blkaddr
);
484 if (dn
->ofs_in_node
== 0 && IS_INODE(dn
->node_page
))
485 clear_inode_flag(dn
->inode
, FI_FIRST_BLOCK_WRITTEN
);
492 * once we invalidate valid blkaddr in range [ofs, ofs + count],
493 * we will invalidate all blkaddr in the whole range.
495 fofs
= start_bidx_of_node(ofs_of_node(dn
->node_page
),
497 f2fs_update_extent_cache_range(dn
, fofs
, 0, len
);
498 dec_valid_block_count(sbi
, dn
->inode
, nr_free
);
500 dn
->ofs_in_node
= ofs
;
502 f2fs_update_time(sbi
, REQ_TIME
);
503 trace_f2fs_truncate_data_blocks_range(dn
->inode
, dn
->nid
,
504 dn
->ofs_in_node
, nr_free
);
508 void truncate_data_blocks(struct dnode_of_data
*dn
)
510 truncate_data_blocks_range(dn
, ADDRS_PER_BLOCK
);
513 static int truncate_partial_data_page(struct inode
*inode
, u64 from
,
516 unsigned offset
= from
& (PAGE_SIZE
- 1);
517 pgoff_t index
= from
>> PAGE_SHIFT
;
518 struct address_space
*mapping
= inode
->i_mapping
;
521 if (!offset
&& !cache_only
)
525 page
= f2fs_grab_cache_page(mapping
, index
, false);
526 if (page
&& PageUptodate(page
))
528 f2fs_put_page(page
, 1);
532 page
= get_lock_data_page(inode
, index
, true);
536 f2fs_wait_on_page_writeback(page
, DATA
, true);
537 zero_user(page
, offset
, PAGE_SIZE
- offset
);
538 if (!cache_only
|| !f2fs_encrypted_inode(inode
) ||
539 !S_ISREG(inode
->i_mode
))
540 set_page_dirty(page
);
541 f2fs_put_page(page
, 1);
545 int truncate_blocks(struct inode
*inode
, u64 from
, bool lock
)
547 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
548 unsigned int blocksize
= inode
->i_sb
->s_blocksize
;
549 struct dnode_of_data dn
;
551 int count
= 0, err
= 0;
553 bool truncate_page
= false;
555 trace_f2fs_truncate_blocks_enter(inode
, from
);
557 free_from
= (pgoff_t
)F2FS_BYTES_TO_BLK(from
+ blocksize
- 1);
559 if (free_from
>= sbi
->max_file_blocks
)
565 ipage
= get_node_page(sbi
, inode
->i_ino
);
567 err
= PTR_ERR(ipage
);
571 if (f2fs_has_inline_data(inode
)) {
572 if (truncate_inline_inode(ipage
, from
))
573 set_page_dirty(ipage
);
574 f2fs_put_page(ipage
, 1);
575 truncate_page
= true;
579 set_new_dnode(&dn
, inode
, ipage
, NULL
, 0);
580 err
= get_dnode_of_data(&dn
, free_from
, LOOKUP_NODE_RA
);
587 count
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
589 count
-= dn
.ofs_in_node
;
590 f2fs_bug_on(sbi
, count
< 0);
592 if (dn
.ofs_in_node
|| IS_INODE(dn
.node_page
)) {
593 truncate_data_blocks_range(&dn
, count
);
599 err
= truncate_inode_blocks(inode
, free_from
);
604 /* lastly zero out the first data page */
606 err
= truncate_partial_data_page(inode
, from
, truncate_page
);
608 trace_f2fs_truncate_blocks_exit(inode
, err
);
612 int f2fs_truncate(struct inode
*inode
)
616 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
617 S_ISLNK(inode
->i_mode
)))
620 trace_f2fs_truncate(inode
);
622 /* we should check inline_data size */
623 if (!f2fs_may_inline_data(inode
)) {
624 err
= f2fs_convert_inline_inode(inode
);
629 err
= truncate_blocks(inode
, i_size_read(inode
), true);
633 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
634 f2fs_mark_inode_dirty_sync(inode
);
638 int f2fs_getattr(struct vfsmount
*mnt
,
639 struct dentry
*dentry
, struct kstat
*stat
)
641 struct inode
*inode
= d_inode(dentry
);
642 generic_fillattr(inode
, stat
);
647 #ifdef CONFIG_F2FS_FS_POSIX_ACL
648 static void __setattr_copy(struct inode
*inode
, const struct iattr
*attr
)
650 unsigned int ia_valid
= attr
->ia_valid
;
652 if (ia_valid
& ATTR_UID
)
653 inode
->i_uid
= attr
->ia_uid
;
654 if (ia_valid
& ATTR_GID
)
655 inode
->i_gid
= attr
->ia_gid
;
656 if (ia_valid
& ATTR_ATIME
)
657 inode
->i_atime
= timespec_trunc(attr
->ia_atime
,
658 inode
->i_sb
->s_time_gran
);
659 if (ia_valid
& ATTR_MTIME
)
660 inode
->i_mtime
= timespec_trunc(attr
->ia_mtime
,
661 inode
->i_sb
->s_time_gran
);
662 if (ia_valid
& ATTR_CTIME
)
663 inode
->i_ctime
= timespec_trunc(attr
->ia_ctime
,
664 inode
->i_sb
->s_time_gran
);
665 if (ia_valid
& ATTR_MODE
) {
666 umode_t mode
= attr
->ia_mode
;
668 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
670 set_acl_inode(inode
, mode
);
674 #define __setattr_copy setattr_copy
677 int f2fs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
679 struct inode
*inode
= d_inode(dentry
);
682 err
= inode_change_ok(inode
, attr
);
686 if (attr
->ia_valid
& ATTR_SIZE
) {
687 if (f2fs_encrypted_inode(inode
) &&
688 fscrypt_get_encryption_info(inode
))
691 if (attr
->ia_size
<= i_size_read(inode
)) {
692 truncate_setsize(inode
, attr
->ia_size
);
693 err
= f2fs_truncate(inode
);
696 f2fs_balance_fs(F2FS_I_SB(inode
), true);
699 * do not trim all blocks after i_size if target size is
700 * larger than i_size.
702 truncate_setsize(inode
, attr
->ia_size
);
704 /* should convert inline inode here */
705 if (!f2fs_may_inline_data(inode
)) {
706 err
= f2fs_convert_inline_inode(inode
);
710 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
714 __setattr_copy(inode
, attr
);
716 if (attr
->ia_valid
& ATTR_MODE
) {
717 err
= posix_acl_chmod(inode
, get_inode_mode(inode
));
718 if (err
|| is_inode_flag_set(inode
, FI_ACL_MODE
)) {
719 inode
->i_mode
= F2FS_I(inode
)->i_acl_mode
;
720 clear_inode_flag(inode
, FI_ACL_MODE
);
724 f2fs_mark_inode_dirty_sync(inode
);
728 const struct inode_operations f2fs_file_inode_operations
= {
729 .getattr
= f2fs_getattr
,
730 .setattr
= f2fs_setattr
,
731 .get_acl
= f2fs_get_acl
,
732 .set_acl
= f2fs_set_acl
,
733 #ifdef CONFIG_F2FS_FS_XATTR
734 .setxattr
= generic_setxattr
,
735 .getxattr
= generic_getxattr
,
736 .listxattr
= f2fs_listxattr
,
737 .removexattr
= generic_removexattr
,
739 .fiemap
= f2fs_fiemap
,
742 static int fill_zero(struct inode
*inode
, pgoff_t index
,
743 loff_t start
, loff_t len
)
745 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
751 f2fs_balance_fs(sbi
, true);
754 page
= get_new_data_page(inode
, NULL
, index
, false);
758 return PTR_ERR(page
);
760 f2fs_wait_on_page_writeback(page
, DATA
, true);
761 zero_user(page
, start
, len
);
762 set_page_dirty(page
);
763 f2fs_put_page(page
, 1);
767 int truncate_hole(struct inode
*inode
, pgoff_t pg_start
, pgoff_t pg_end
)
771 while (pg_start
< pg_end
) {
772 struct dnode_of_data dn
;
773 pgoff_t end_offset
, count
;
775 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
776 err
= get_dnode_of_data(&dn
, pg_start
, LOOKUP_NODE
);
778 if (err
== -ENOENT
) {
785 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
786 count
= min(end_offset
- dn
.ofs_in_node
, pg_end
- pg_start
);
788 f2fs_bug_on(F2FS_I_SB(inode
), count
== 0 || count
> end_offset
);
790 truncate_data_blocks_range(&dn
, count
);
798 static int punch_hole(struct inode
*inode
, loff_t offset
, loff_t len
)
800 pgoff_t pg_start
, pg_end
;
801 loff_t off_start
, off_end
;
804 ret
= f2fs_convert_inline_inode(inode
);
808 pg_start
= ((unsigned long long) offset
) >> PAGE_SHIFT
;
809 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_SHIFT
;
811 off_start
= offset
& (PAGE_SIZE
- 1);
812 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
814 if (pg_start
== pg_end
) {
815 ret
= fill_zero(inode
, pg_start
, off_start
,
816 off_end
- off_start
);
821 ret
= fill_zero(inode
, pg_start
++, off_start
,
822 PAGE_SIZE
- off_start
);
827 ret
= fill_zero(inode
, pg_end
, 0, off_end
);
832 if (pg_start
< pg_end
) {
833 struct address_space
*mapping
= inode
->i_mapping
;
834 loff_t blk_start
, blk_end
;
835 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
837 f2fs_balance_fs(sbi
, true);
839 blk_start
= (loff_t
)pg_start
<< PAGE_SHIFT
;
840 blk_end
= (loff_t
)pg_end
<< PAGE_SHIFT
;
841 truncate_inode_pages_range(mapping
, blk_start
,
845 ret
= truncate_hole(inode
, pg_start
, pg_end
);
853 static int __read_out_blkaddrs(struct inode
*inode
, block_t
*blkaddr
,
854 int *do_replace
, pgoff_t off
, pgoff_t len
)
856 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
857 struct dnode_of_data dn
;
861 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
862 ret
= get_dnode_of_data(&dn
, off
, LOOKUP_NODE_RA
);
863 if (ret
&& ret
!= -ENOENT
) {
865 } else if (ret
== -ENOENT
) {
866 if (dn
.max_level
== 0)
868 done
= min((pgoff_t
)ADDRS_PER_BLOCK
- dn
.ofs_in_node
, len
);
874 done
= min((pgoff_t
)ADDRS_PER_PAGE(dn
.node_page
, inode
) -
875 dn
.ofs_in_node
, len
);
876 for (i
= 0; i
< done
; i
++, blkaddr
++, do_replace
++, dn
.ofs_in_node
++) {
877 *blkaddr
= datablock_addr(dn
.node_page
, dn
.ofs_in_node
);
878 if (!is_checkpointed_data(sbi
, *blkaddr
)) {
880 if (test_opt(sbi
, LFS
)) {
885 /* do not invalidate this block address */
886 f2fs_update_data_blkaddr(&dn
, NULL_ADDR
);
899 static int __roll_back_blkaddrs(struct inode
*inode
, block_t
*blkaddr
,
900 int *do_replace
, pgoff_t off
, int len
)
902 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
903 struct dnode_of_data dn
;
906 for (i
= 0; i
< len
; i
++, do_replace
++, blkaddr
++) {
907 if (*do_replace
== 0)
910 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
911 ret
= get_dnode_of_data(&dn
, off
+ i
, LOOKUP_NODE_RA
);
913 dec_valid_block_count(sbi
, inode
, 1);
914 invalidate_blocks(sbi
, *blkaddr
);
916 f2fs_update_data_blkaddr(&dn
, *blkaddr
);
923 static int __clone_blkaddrs(struct inode
*src_inode
, struct inode
*dst_inode
,
924 block_t
*blkaddr
, int *do_replace
,
925 pgoff_t src
, pgoff_t dst
, pgoff_t len
, bool full
)
927 struct f2fs_sb_info
*sbi
= F2FS_I_SB(src_inode
);
932 if (blkaddr
[i
] == NULL_ADDR
&& !full
) {
937 if (do_replace
[i
] || blkaddr
[i
] == NULL_ADDR
) {
938 struct dnode_of_data dn
;
943 set_new_dnode(&dn
, dst_inode
, NULL
, NULL
, 0);
944 ret
= get_dnode_of_data(&dn
, dst
+ i
, ALLOC_NODE
);
948 get_node_info(sbi
, dn
.nid
, &ni
);
950 ADDRS_PER_PAGE(dn
.node_page
, dst_inode
) -
951 dn
.ofs_in_node
, len
- i
);
953 dn
.data_blkaddr
= datablock_addr(dn
.node_page
,
955 truncate_data_blocks_range(&dn
, 1);
958 f2fs_i_blocks_write(src_inode
,
960 f2fs_i_blocks_write(dst_inode
,
962 f2fs_replace_block(sbi
, &dn
, dn
.data_blkaddr
,
963 blkaddr
[i
], ni
.version
, true, false);
969 new_size
= (dst
+ i
) << PAGE_SHIFT
;
970 if (dst_inode
->i_size
< new_size
)
971 f2fs_i_size_write(dst_inode
, new_size
);
972 } while ((do_replace
[i
] || blkaddr
[i
] == NULL_ADDR
) && --ilen
);
976 struct page
*psrc
, *pdst
;
978 psrc
= get_lock_data_page(src_inode
, src
+ i
, true);
980 return PTR_ERR(psrc
);
981 pdst
= get_new_data_page(dst_inode
, NULL
, dst
+ i
,
984 f2fs_put_page(psrc
, 1);
985 return PTR_ERR(pdst
);
987 f2fs_copy_page(psrc
, pdst
);
988 set_page_dirty(pdst
);
989 f2fs_put_page(pdst
, 1);
990 f2fs_put_page(psrc
, 1);
992 ret
= truncate_hole(src_inode
, src
+ i
, src
+ i
+ 1);
1001 static int __exchange_data_block(struct inode
*src_inode
,
1002 struct inode
*dst_inode
, pgoff_t src
, pgoff_t dst
,
1005 block_t
*src_blkaddr
;
1009 src_blkaddr
= f2fs_kvzalloc(sizeof(block_t
) * len
, GFP_KERNEL
);
1013 do_replace
= f2fs_kvzalloc(sizeof(int) * len
, GFP_KERNEL
);
1015 kvfree(src_blkaddr
);
1019 ret
= __read_out_blkaddrs(src_inode
, src_blkaddr
, do_replace
, src
, len
);
1023 ret
= __clone_blkaddrs(src_inode
, dst_inode
, src_blkaddr
,
1024 do_replace
, src
, dst
, len
, full
);
1028 kvfree(src_blkaddr
);
1033 __roll_back_blkaddrs(src_inode
, src_blkaddr
, do_replace
, src
, len
);
1034 kvfree(src_blkaddr
);
1039 static int f2fs_do_collapse(struct inode
*inode
, pgoff_t start
, pgoff_t end
)
1041 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1042 pgoff_t nrpages
= (i_size_read(inode
) + PAGE_SIZE
- 1) / PAGE_SIZE
;
1045 f2fs_balance_fs(sbi
, true);
1048 f2fs_drop_extent_tree(inode
);
1050 ret
= __exchange_data_block(inode
, inode
, end
, start
, nrpages
- end
, true);
1051 f2fs_unlock_op(sbi
);
1055 static int f2fs_collapse_range(struct inode
*inode
, loff_t offset
, loff_t len
)
1057 pgoff_t pg_start
, pg_end
;
1061 if (offset
+ len
>= i_size_read(inode
))
1064 /* collapse range should be aligned to block size of f2fs. */
1065 if (offset
& (F2FS_BLKSIZE
- 1) || len
& (F2FS_BLKSIZE
- 1))
1068 ret
= f2fs_convert_inline_inode(inode
);
1072 pg_start
= offset
>> PAGE_SHIFT
;
1073 pg_end
= (offset
+ len
) >> PAGE_SHIFT
;
1075 /* write out all dirty pages from offset */
1076 ret
= filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
1080 truncate_pagecache(inode
, offset
);
1082 ret
= f2fs_do_collapse(inode
, pg_start
, pg_end
);
1086 /* write out all moved pages, if possible */
1087 filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
1088 truncate_pagecache(inode
, offset
);
1090 new_size
= i_size_read(inode
) - len
;
1091 truncate_pagecache(inode
, new_size
);
1093 ret
= truncate_blocks(inode
, new_size
, true);
1095 f2fs_i_size_write(inode
, new_size
);
1100 static int f2fs_do_zero_range(struct dnode_of_data
*dn
, pgoff_t start
,
1103 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
1104 pgoff_t index
= start
;
1105 unsigned int ofs_in_node
= dn
->ofs_in_node
;
1109 for (; index
< end
; index
++, dn
->ofs_in_node
++) {
1110 if (datablock_addr(dn
->node_page
, dn
->ofs_in_node
) == NULL_ADDR
)
1114 dn
->ofs_in_node
= ofs_in_node
;
1115 ret
= reserve_new_blocks(dn
, count
);
1119 dn
->ofs_in_node
= ofs_in_node
;
1120 for (index
= start
; index
< end
; index
++, dn
->ofs_in_node
++) {
1122 datablock_addr(dn
->node_page
, dn
->ofs_in_node
);
1124 * reserve_new_blocks will not guarantee entire block
1127 if (dn
->data_blkaddr
== NULL_ADDR
) {
1131 if (dn
->data_blkaddr
!= NEW_ADDR
) {
1132 invalidate_blocks(sbi
, dn
->data_blkaddr
);
1133 dn
->data_blkaddr
= NEW_ADDR
;
1134 set_data_blkaddr(dn
);
1138 f2fs_update_extent_cache_range(dn
, start
, 0, index
- start
);
1143 static int f2fs_zero_range(struct inode
*inode
, loff_t offset
, loff_t len
,
1146 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1147 struct address_space
*mapping
= inode
->i_mapping
;
1148 pgoff_t index
, pg_start
, pg_end
;
1149 loff_t new_size
= i_size_read(inode
);
1150 loff_t off_start
, off_end
;
1153 ret
= inode_newsize_ok(inode
, (len
+ offset
));
1157 ret
= f2fs_convert_inline_inode(inode
);
1161 ret
= filemap_write_and_wait_range(mapping
, offset
, offset
+ len
- 1);
1165 truncate_pagecache_range(inode
, offset
, offset
+ len
- 1);
1167 pg_start
= ((unsigned long long) offset
) >> PAGE_SHIFT
;
1168 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_SHIFT
;
1170 off_start
= offset
& (PAGE_SIZE
- 1);
1171 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
1173 if (pg_start
== pg_end
) {
1174 ret
= fill_zero(inode
, pg_start
, off_start
,
1175 off_end
- off_start
);
1179 if (offset
+ len
> new_size
)
1180 new_size
= offset
+ len
;
1181 new_size
= max_t(loff_t
, new_size
, offset
+ len
);
1184 ret
= fill_zero(inode
, pg_start
++, off_start
,
1185 PAGE_SIZE
- off_start
);
1189 new_size
= max_t(loff_t
, new_size
,
1190 (loff_t
)pg_start
<< PAGE_SHIFT
);
1193 for (index
= pg_start
; index
< pg_end
;) {
1194 struct dnode_of_data dn
;
1195 unsigned int end_offset
;
1200 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
1201 ret
= get_dnode_of_data(&dn
, index
, ALLOC_NODE
);
1203 f2fs_unlock_op(sbi
);
1207 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
1208 end
= min(pg_end
, end_offset
- dn
.ofs_in_node
+ index
);
1210 ret
= f2fs_do_zero_range(&dn
, index
, end
);
1211 f2fs_put_dnode(&dn
);
1212 f2fs_unlock_op(sbi
);
1217 new_size
= max_t(loff_t
, new_size
,
1218 (loff_t
)index
<< PAGE_SHIFT
);
1222 ret
= fill_zero(inode
, pg_end
, 0, off_end
);
1226 new_size
= max_t(loff_t
, new_size
, offset
+ len
);
1231 if (!(mode
& FALLOC_FL_KEEP_SIZE
) && i_size_read(inode
) < new_size
)
1232 f2fs_i_size_write(inode
, new_size
);
1237 static int f2fs_insert_range(struct inode
*inode
, loff_t offset
, loff_t len
)
1239 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1240 pgoff_t nr
, pg_start
, pg_end
, delta
, idx
;
1244 new_size
= i_size_read(inode
) + len
;
1245 if (new_size
> inode
->i_sb
->s_maxbytes
)
1248 if (offset
>= i_size_read(inode
))
1251 /* insert range should be aligned to block size of f2fs. */
1252 if (offset
& (F2FS_BLKSIZE
- 1) || len
& (F2FS_BLKSIZE
- 1))
1255 ret
= f2fs_convert_inline_inode(inode
);
1259 f2fs_balance_fs(sbi
, true);
1261 ret
= truncate_blocks(inode
, i_size_read(inode
), true);
1265 /* write out all dirty pages from offset */
1266 ret
= filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
1270 truncate_pagecache(inode
, offset
);
1272 pg_start
= offset
>> PAGE_SHIFT
;
1273 pg_end
= (offset
+ len
) >> PAGE_SHIFT
;
1274 delta
= pg_end
- pg_start
;
1275 idx
= (i_size_read(inode
) + PAGE_SIZE
- 1) / PAGE_SIZE
;
1277 while (!ret
&& idx
> pg_start
) {
1278 nr
= idx
- pg_start
;
1284 f2fs_drop_extent_tree(inode
);
1286 ret
= __exchange_data_block(inode
, inode
, idx
,
1287 idx
+ delta
, nr
, false);
1288 f2fs_unlock_op(sbi
);
1291 /* write out all moved pages, if possible */
1292 filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
1293 truncate_pagecache(inode
, offset
);
1296 f2fs_i_size_write(inode
, new_size
);
1300 static int expand_inode_data(struct inode
*inode
, loff_t offset
,
1301 loff_t len
, int mode
)
1303 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1304 struct f2fs_map_blocks map
= { .m_next_pgofs
= NULL
};
1306 loff_t new_size
= i_size_read(inode
);
1310 ret
= inode_newsize_ok(inode
, (len
+ offset
));
1314 ret
= f2fs_convert_inline_inode(inode
);
1318 f2fs_balance_fs(sbi
, true);
1320 pg_end
= ((unsigned long long)offset
+ len
) >> PAGE_SHIFT
;
1321 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
1323 map
.m_lblk
= ((unsigned long long)offset
) >> PAGE_SHIFT
;
1324 map
.m_len
= pg_end
- map
.m_lblk
;
1328 ret
= f2fs_map_blocks(inode
, &map
, 1, F2FS_GET_BLOCK_PRE_AIO
);
1335 last_off
= map
.m_lblk
+ map
.m_len
- 1;
1337 /* update new size to the failed position */
1338 new_size
= (last_off
== pg_end
) ? offset
+ len
:
1339 (loff_t
)(last_off
+ 1) << PAGE_SHIFT
;
1341 new_size
= ((loff_t
)pg_end
<< PAGE_SHIFT
) + off_end
;
1344 if (!(mode
& FALLOC_FL_KEEP_SIZE
) && i_size_read(inode
) < new_size
)
1345 f2fs_i_size_write(inode
, new_size
);
1350 static long f2fs_fallocate(struct file
*file
, int mode
,
1351 loff_t offset
, loff_t len
)
1353 struct inode
*inode
= file_inode(file
);
1356 /* f2fs only support ->fallocate for regular file */
1357 if (!S_ISREG(inode
->i_mode
))
1360 if (f2fs_encrypted_inode(inode
) &&
1361 (mode
& (FALLOC_FL_COLLAPSE_RANGE
| FALLOC_FL_INSERT_RANGE
)))
1364 if (mode
& ~(FALLOC_FL_KEEP_SIZE
| FALLOC_FL_PUNCH_HOLE
|
1365 FALLOC_FL_COLLAPSE_RANGE
| FALLOC_FL_ZERO_RANGE
|
1366 FALLOC_FL_INSERT_RANGE
))
1371 if (mode
& FALLOC_FL_PUNCH_HOLE
) {
1372 if (offset
>= inode
->i_size
)
1375 ret
= punch_hole(inode
, offset
, len
);
1376 } else if (mode
& FALLOC_FL_COLLAPSE_RANGE
) {
1377 ret
= f2fs_collapse_range(inode
, offset
, len
);
1378 } else if (mode
& FALLOC_FL_ZERO_RANGE
) {
1379 ret
= f2fs_zero_range(inode
, offset
, len
, mode
);
1380 } else if (mode
& FALLOC_FL_INSERT_RANGE
) {
1381 ret
= f2fs_insert_range(inode
, offset
, len
);
1383 ret
= expand_inode_data(inode
, offset
, len
, mode
);
1387 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
1388 f2fs_mark_inode_dirty_sync(inode
);
1389 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1393 inode_unlock(inode
);
1395 trace_f2fs_fallocate(inode
, mode
, offset
, len
, ret
);
1399 static int f2fs_release_file(struct inode
*inode
, struct file
*filp
)
1402 * f2fs_relase_file is called at every close calls. So we should
1403 * not drop any inmemory pages by close called by other process.
1405 if (!(filp
->f_mode
& FMODE_WRITE
) ||
1406 atomic_read(&inode
->i_writecount
) != 1)
1409 /* some remained atomic pages should discarded */
1410 if (f2fs_is_atomic_file(inode
))
1411 drop_inmem_pages(inode
);
1412 if (f2fs_is_volatile_file(inode
)) {
1413 clear_inode_flag(inode
, FI_VOLATILE_FILE
);
1414 set_inode_flag(inode
, FI_DROP_CACHE
);
1415 filemap_fdatawrite(inode
->i_mapping
);
1416 clear_inode_flag(inode
, FI_DROP_CACHE
);
1421 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1422 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1424 static inline __u32
f2fs_mask_flags(umode_t mode
, __u32 flags
)
1428 else if (S_ISREG(mode
))
1429 return flags
& F2FS_REG_FLMASK
;
1431 return flags
& F2FS_OTHER_FLMASK
;
1434 static int f2fs_ioc_getflags(struct file
*filp
, unsigned long arg
)
1436 struct inode
*inode
= file_inode(filp
);
1437 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
1438 unsigned int flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
1439 return put_user(flags
, (int __user
*)arg
);
1442 static int f2fs_ioc_setflags(struct file
*filp
, unsigned long arg
)
1444 struct inode
*inode
= file_inode(filp
);
1445 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
1446 unsigned int flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
1447 unsigned int oldflags
;
1450 if (!inode_owner_or_capable(inode
))
1453 if (get_user(flags
, (int __user
*)arg
))
1456 ret
= mnt_want_write_file(filp
);
1460 flags
= f2fs_mask_flags(inode
->i_mode
, flags
);
1464 oldflags
= fi
->i_flags
;
1466 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
1467 if (!capable(CAP_LINUX_IMMUTABLE
)) {
1468 inode_unlock(inode
);
1474 flags
= flags
& FS_FL_USER_MODIFIABLE
;
1475 flags
|= oldflags
& ~FS_FL_USER_MODIFIABLE
;
1476 fi
->i_flags
= flags
;
1477 inode_unlock(inode
);
1479 inode
->i_ctime
= CURRENT_TIME
;
1480 f2fs_set_inode_flags(inode
);
1482 mnt_drop_write_file(filp
);
1486 static int f2fs_ioc_getversion(struct file
*filp
, unsigned long arg
)
1488 struct inode
*inode
= file_inode(filp
);
1490 return put_user(inode
->i_generation
, (int __user
*)arg
);
1493 static int f2fs_ioc_start_atomic_write(struct file
*filp
)
1495 struct inode
*inode
= file_inode(filp
);
1498 if (!inode_owner_or_capable(inode
))
1501 ret
= mnt_want_write_file(filp
);
1507 if (f2fs_is_atomic_file(inode
))
1510 ret
= f2fs_convert_inline_inode(inode
);
1514 set_inode_flag(inode
, FI_ATOMIC_FILE
);
1515 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1517 if (!get_dirty_pages(inode
))
1520 f2fs_msg(F2FS_I_SB(inode
)->sb
, KERN_WARNING
,
1521 "Unexpected flush for atomic writes: ino=%lu, npages=%lld",
1522 inode
->i_ino
, get_dirty_pages(inode
));
1523 ret
= filemap_write_and_wait_range(inode
->i_mapping
, 0, LLONG_MAX
);
1525 clear_inode_flag(inode
, FI_ATOMIC_FILE
);
1527 inode_unlock(inode
);
1528 mnt_drop_write_file(filp
);
1532 static int f2fs_ioc_commit_atomic_write(struct file
*filp
)
1534 struct inode
*inode
= file_inode(filp
);
1537 if (!inode_owner_or_capable(inode
))
1540 ret
= mnt_want_write_file(filp
);
1546 if (f2fs_is_volatile_file(inode
))
1549 if (f2fs_is_atomic_file(inode
)) {
1550 clear_inode_flag(inode
, FI_ATOMIC_FILE
);
1551 ret
= commit_inmem_pages(inode
);
1553 set_inode_flag(inode
, FI_ATOMIC_FILE
);
1558 ret
= f2fs_do_sync_file(filp
, 0, LLONG_MAX
, 0, true);
1560 inode_unlock(inode
);
1561 mnt_drop_write_file(filp
);
1565 static int f2fs_ioc_start_volatile_write(struct file
*filp
)
1567 struct inode
*inode
= file_inode(filp
);
1570 if (!inode_owner_or_capable(inode
))
1573 ret
= mnt_want_write_file(filp
);
1579 if (f2fs_is_volatile_file(inode
))
1582 ret
= f2fs_convert_inline_inode(inode
);
1586 set_inode_flag(inode
, FI_VOLATILE_FILE
);
1587 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1589 inode_unlock(inode
);
1590 mnt_drop_write_file(filp
);
1594 static int f2fs_ioc_release_volatile_write(struct file
*filp
)
1596 struct inode
*inode
= file_inode(filp
);
1599 if (!inode_owner_or_capable(inode
))
1602 ret
= mnt_want_write_file(filp
);
1608 if (!f2fs_is_volatile_file(inode
))
1611 if (!f2fs_is_first_block_written(inode
)) {
1612 ret
= truncate_partial_data_page(inode
, 0, true);
1616 ret
= punch_hole(inode
, 0, F2FS_BLKSIZE
);
1618 inode_unlock(inode
);
1619 mnt_drop_write_file(filp
);
1623 static int f2fs_ioc_abort_volatile_write(struct file
*filp
)
1625 struct inode
*inode
= file_inode(filp
);
1628 if (!inode_owner_or_capable(inode
))
1631 ret
= mnt_want_write_file(filp
);
1637 if (f2fs_is_atomic_file(inode
))
1638 drop_inmem_pages(inode
);
1639 if (f2fs_is_volatile_file(inode
)) {
1640 clear_inode_flag(inode
, FI_VOLATILE_FILE
);
1641 ret
= f2fs_do_sync_file(filp
, 0, LLONG_MAX
, 0, true);
1644 inode_unlock(inode
);
1646 mnt_drop_write_file(filp
);
1647 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1651 static int f2fs_ioc_shutdown(struct file
*filp
, unsigned long arg
)
1653 struct inode
*inode
= file_inode(filp
);
1654 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1655 struct super_block
*sb
= sbi
->sb
;
1659 if (!capable(CAP_SYS_ADMIN
))
1662 if (get_user(in
, (__u32 __user
*)arg
))
1665 ret
= mnt_want_write_file(filp
);
1670 case F2FS_GOING_DOWN_FULLSYNC
:
1671 sb
= freeze_bdev(sb
->s_bdev
);
1672 if (sb
&& !IS_ERR(sb
)) {
1673 f2fs_stop_checkpoint(sbi
, false);
1674 thaw_bdev(sb
->s_bdev
, sb
);
1677 case F2FS_GOING_DOWN_METASYNC
:
1678 /* do checkpoint only */
1679 f2fs_sync_fs(sb
, 1);
1680 f2fs_stop_checkpoint(sbi
, false);
1682 case F2FS_GOING_DOWN_NOSYNC
:
1683 f2fs_stop_checkpoint(sbi
, false);
1685 case F2FS_GOING_DOWN_METAFLUSH
:
1686 sync_meta_pages(sbi
, META
, LONG_MAX
);
1687 f2fs_stop_checkpoint(sbi
, false);
1693 f2fs_update_time(sbi
, REQ_TIME
);
1695 mnt_drop_write_file(filp
);
1699 static int f2fs_ioc_fitrim(struct file
*filp
, unsigned long arg
)
1701 struct inode
*inode
= file_inode(filp
);
1702 struct super_block
*sb
= inode
->i_sb
;
1703 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
1704 struct fstrim_range range
;
1707 if (!capable(CAP_SYS_ADMIN
))
1710 if (!blk_queue_discard(q
))
1713 if (copy_from_user(&range
, (struct fstrim_range __user
*)arg
,
1717 ret
= mnt_want_write_file(filp
);
1721 range
.minlen
= max((unsigned int)range
.minlen
,
1722 q
->limits
.discard_granularity
);
1723 ret
= f2fs_trim_fs(F2FS_SB(sb
), &range
);
1724 mnt_drop_write_file(filp
);
1728 if (copy_to_user((struct fstrim_range __user
*)arg
, &range
,
1731 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1735 static bool uuid_is_nonzero(__u8 u
[16])
1739 for (i
= 0; i
< 16; i
++)
1745 static int f2fs_ioc_set_encryption_policy(struct file
*filp
, unsigned long arg
)
1747 struct fscrypt_policy policy
;
1748 struct inode
*inode
= file_inode(filp
);
1751 if (copy_from_user(&policy
, (struct fscrypt_policy __user
*)arg
,
1755 ret
= mnt_want_write_file(filp
);
1759 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1760 ret
= fscrypt_process_policy(inode
, &policy
);
1762 mnt_drop_write_file(filp
);
1766 static int f2fs_ioc_get_encryption_policy(struct file
*filp
, unsigned long arg
)
1768 struct fscrypt_policy policy
;
1769 struct inode
*inode
= file_inode(filp
);
1772 err
= fscrypt_get_policy(inode
, &policy
);
1776 if (copy_to_user((struct fscrypt_policy __user
*)arg
, &policy
, sizeof(policy
)))
1781 static int f2fs_ioc_get_encryption_pwsalt(struct file
*filp
, unsigned long arg
)
1783 struct inode
*inode
= file_inode(filp
);
1784 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1787 if (!f2fs_sb_has_crypto(inode
->i_sb
))
1790 if (uuid_is_nonzero(sbi
->raw_super
->encrypt_pw_salt
))
1793 err
= mnt_want_write_file(filp
);
1797 /* update superblock with uuid */
1798 generate_random_uuid(sbi
->raw_super
->encrypt_pw_salt
);
1800 err
= f2fs_commit_super(sbi
, false);
1803 memset(sbi
->raw_super
->encrypt_pw_salt
, 0, 16);
1804 mnt_drop_write_file(filp
);
1807 mnt_drop_write_file(filp
);
1809 if (copy_to_user((__u8 __user
*)arg
, sbi
->raw_super
->encrypt_pw_salt
,
1815 static int f2fs_ioc_gc(struct file
*filp
, unsigned long arg
)
1817 struct inode
*inode
= file_inode(filp
);
1818 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1822 if (!capable(CAP_SYS_ADMIN
))
1825 if (get_user(sync
, (__u32 __user
*)arg
))
1828 if (f2fs_readonly(sbi
->sb
))
1831 ret
= mnt_want_write_file(filp
);
1836 if (!mutex_trylock(&sbi
->gc_mutex
)) {
1841 mutex_lock(&sbi
->gc_mutex
);
1844 ret
= f2fs_gc(sbi
, sync
);
1846 mnt_drop_write_file(filp
);
1850 static int f2fs_ioc_write_checkpoint(struct file
*filp
, unsigned long arg
)
1852 struct inode
*inode
= file_inode(filp
);
1853 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1856 if (!capable(CAP_SYS_ADMIN
))
1859 if (f2fs_readonly(sbi
->sb
))
1862 ret
= mnt_want_write_file(filp
);
1866 ret
= f2fs_sync_fs(sbi
->sb
, 1);
1868 mnt_drop_write_file(filp
);
1872 static int f2fs_defragment_range(struct f2fs_sb_info
*sbi
,
1874 struct f2fs_defragment
*range
)
1876 struct inode
*inode
= file_inode(filp
);
1877 struct f2fs_map_blocks map
= { .m_next_pgofs
= NULL
};
1878 struct extent_info ei
;
1879 pgoff_t pg_start
, pg_end
;
1880 unsigned int blk_per_seg
= sbi
->blocks_per_seg
;
1881 unsigned int total
= 0, sec_num
;
1882 unsigned int pages_per_sec
= sbi
->segs_per_sec
* blk_per_seg
;
1883 block_t blk_end
= 0;
1884 bool fragmented
= false;
1887 /* if in-place-update policy is enabled, don't waste time here */
1888 if (need_inplace_update(inode
))
1891 pg_start
= range
->start
>> PAGE_SHIFT
;
1892 pg_end
= (range
->start
+ range
->len
) >> PAGE_SHIFT
;
1894 f2fs_balance_fs(sbi
, true);
1898 /* writeback all dirty pages in the range */
1899 err
= filemap_write_and_wait_range(inode
->i_mapping
, range
->start
,
1900 range
->start
+ range
->len
- 1);
1905 * lookup mapping info in extent cache, skip defragmenting if physical
1906 * block addresses are continuous.
1908 if (f2fs_lookup_extent_cache(inode
, pg_start
, &ei
)) {
1909 if (ei
.fofs
+ ei
.len
>= pg_end
)
1913 map
.m_lblk
= pg_start
;
1916 * lookup mapping info in dnode page cache, skip defragmenting if all
1917 * physical block addresses are continuous even if there are hole(s)
1918 * in logical blocks.
1920 while (map
.m_lblk
< pg_end
) {
1921 map
.m_len
= pg_end
- map
.m_lblk
;
1922 err
= f2fs_map_blocks(inode
, &map
, 0, F2FS_GET_BLOCK_READ
);
1926 if (!(map
.m_flags
& F2FS_MAP_FLAGS
)) {
1931 if (blk_end
&& blk_end
!= map
.m_pblk
) {
1935 blk_end
= map
.m_pblk
+ map
.m_len
;
1937 map
.m_lblk
+= map
.m_len
;
1943 map
.m_lblk
= pg_start
;
1944 map
.m_len
= pg_end
- pg_start
;
1946 sec_num
= (map
.m_len
+ pages_per_sec
- 1) / pages_per_sec
;
1949 * make sure there are enough free section for LFS allocation, this can
1950 * avoid defragment running in SSR mode when free section are allocated
1953 if (has_not_enough_free_secs(sbi
, sec_num
)) {
1958 while (map
.m_lblk
< pg_end
) {
1963 map
.m_len
= pg_end
- map
.m_lblk
;
1964 err
= f2fs_map_blocks(inode
, &map
, 0, F2FS_GET_BLOCK_READ
);
1968 if (!(map
.m_flags
& F2FS_MAP_FLAGS
)) {
1973 set_inode_flag(inode
, FI_DO_DEFRAG
);
1976 while (idx
< map
.m_lblk
+ map
.m_len
&& cnt
< blk_per_seg
) {
1979 page
= get_lock_data_page(inode
, idx
, true);
1981 err
= PTR_ERR(page
);
1985 set_page_dirty(page
);
1986 f2fs_put_page(page
, 1);
1995 if (idx
< pg_end
&& cnt
< blk_per_seg
)
1998 clear_inode_flag(inode
, FI_DO_DEFRAG
);
2000 err
= filemap_fdatawrite(inode
->i_mapping
);
2005 clear_inode_flag(inode
, FI_DO_DEFRAG
);
2007 inode_unlock(inode
);
2009 range
->len
= (u64
)total
<< PAGE_SHIFT
;
2013 static int f2fs_ioc_defragment(struct file
*filp
, unsigned long arg
)
2015 struct inode
*inode
= file_inode(filp
);
2016 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
2017 struct f2fs_defragment range
;
2020 if (!capable(CAP_SYS_ADMIN
))
2023 if (!S_ISREG(inode
->i_mode
))
2026 err
= mnt_want_write_file(filp
);
2030 if (f2fs_readonly(sbi
->sb
)) {
2035 if (copy_from_user(&range
, (struct f2fs_defragment __user
*)arg
,
2041 /* verify alignment of offset & size */
2042 if (range
.start
& (F2FS_BLKSIZE
- 1) ||
2043 range
.len
& (F2FS_BLKSIZE
- 1)) {
2048 err
= f2fs_defragment_range(sbi
, filp
, &range
);
2049 f2fs_update_time(sbi
, REQ_TIME
);
2053 if (copy_to_user((struct f2fs_defragment __user
*)arg
, &range
,
2057 mnt_drop_write_file(filp
);
2061 long f2fs_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
2064 case F2FS_IOC_GETFLAGS
:
2065 return f2fs_ioc_getflags(filp
, arg
);
2066 case F2FS_IOC_SETFLAGS
:
2067 return f2fs_ioc_setflags(filp
, arg
);
2068 case F2FS_IOC_GETVERSION
:
2069 return f2fs_ioc_getversion(filp
, arg
);
2070 case F2FS_IOC_START_ATOMIC_WRITE
:
2071 return f2fs_ioc_start_atomic_write(filp
);
2072 case F2FS_IOC_COMMIT_ATOMIC_WRITE
:
2073 return f2fs_ioc_commit_atomic_write(filp
);
2074 case F2FS_IOC_START_VOLATILE_WRITE
:
2075 return f2fs_ioc_start_volatile_write(filp
);
2076 case F2FS_IOC_RELEASE_VOLATILE_WRITE
:
2077 return f2fs_ioc_release_volatile_write(filp
);
2078 case F2FS_IOC_ABORT_VOLATILE_WRITE
:
2079 return f2fs_ioc_abort_volatile_write(filp
);
2080 case F2FS_IOC_SHUTDOWN
:
2081 return f2fs_ioc_shutdown(filp
, arg
);
2083 return f2fs_ioc_fitrim(filp
, arg
);
2084 case F2FS_IOC_SET_ENCRYPTION_POLICY
:
2085 return f2fs_ioc_set_encryption_policy(filp
, arg
);
2086 case F2FS_IOC_GET_ENCRYPTION_POLICY
:
2087 return f2fs_ioc_get_encryption_policy(filp
, arg
);
2088 case F2FS_IOC_GET_ENCRYPTION_PWSALT
:
2089 return f2fs_ioc_get_encryption_pwsalt(filp
, arg
);
2090 case F2FS_IOC_GARBAGE_COLLECT
:
2091 return f2fs_ioc_gc(filp
, arg
);
2092 case F2FS_IOC_WRITE_CHECKPOINT
:
2093 return f2fs_ioc_write_checkpoint(filp
, arg
);
2094 case F2FS_IOC_DEFRAGMENT
:
2095 return f2fs_ioc_defragment(filp
, arg
);
2101 static ssize_t
f2fs_file_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
2103 struct file
*file
= iocb
->ki_filp
;
2104 struct inode
*inode
= file_inode(file
);
2105 struct blk_plug plug
;
2108 if (f2fs_encrypted_inode(inode
) &&
2109 !fscrypt_has_encryption_key(inode
) &&
2110 fscrypt_get_encryption_info(inode
))
2114 ret
= generic_write_checks(iocb
, from
);
2116 ret
= f2fs_preallocate_blocks(iocb
, from
);
2118 blk_start_plug(&plug
);
2119 ret
= __generic_file_write_iter(iocb
, from
);
2120 blk_finish_plug(&plug
);
2123 inode_unlock(inode
);
2126 ret
= generic_write_sync(iocb
, ret
);
2130 #ifdef CONFIG_COMPAT
2131 long f2fs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
2134 case F2FS_IOC32_GETFLAGS
:
2135 cmd
= F2FS_IOC_GETFLAGS
;
2137 case F2FS_IOC32_SETFLAGS
:
2138 cmd
= F2FS_IOC_SETFLAGS
;
2140 case F2FS_IOC32_GETVERSION
:
2141 cmd
= F2FS_IOC_GETVERSION
;
2143 case F2FS_IOC_START_ATOMIC_WRITE
:
2144 case F2FS_IOC_COMMIT_ATOMIC_WRITE
:
2145 case F2FS_IOC_START_VOLATILE_WRITE
:
2146 case F2FS_IOC_RELEASE_VOLATILE_WRITE
:
2147 case F2FS_IOC_ABORT_VOLATILE_WRITE
:
2148 case F2FS_IOC_SHUTDOWN
:
2149 case F2FS_IOC_SET_ENCRYPTION_POLICY
:
2150 case F2FS_IOC_GET_ENCRYPTION_PWSALT
:
2151 case F2FS_IOC_GET_ENCRYPTION_POLICY
:
2152 case F2FS_IOC_GARBAGE_COLLECT
:
2153 case F2FS_IOC_WRITE_CHECKPOINT
:
2154 case F2FS_IOC_DEFRAGMENT
:
2157 return -ENOIOCTLCMD
;
2159 return f2fs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));
2163 const struct file_operations f2fs_file_operations
= {
2164 .llseek
= f2fs_llseek
,
2165 .read_iter
= generic_file_read_iter
,
2166 .write_iter
= f2fs_file_write_iter
,
2167 .open
= f2fs_file_open
,
2168 .release
= f2fs_release_file
,
2169 .mmap
= f2fs_file_mmap
,
2170 .fsync
= f2fs_sync_file
,
2171 .fallocate
= f2fs_fallocate
,
2172 .unlocked_ioctl
= f2fs_ioctl
,
2173 #ifdef CONFIG_COMPAT
2174 .compat_ioctl
= f2fs_compat_ioctl
,
2176 .splice_read
= generic_file_splice_read
,
2177 .splice_write
= iter_file_splice_write
,