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 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
)) {
174 f2fs_i_pino_write(inode
, pino
);
175 file_got_pino(inode
);
177 up_write(&fi
->i_sem
);
180 static int f2fs_do_sync_file(struct file
*file
, loff_t start
, loff_t end
,
181 int datasync
, bool atomic
)
183 struct inode
*inode
= file
->f_mapping
->host
;
184 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
185 nid_t ino
= inode
->i_ino
;
187 bool need_cp
= false;
188 struct writeback_control wbc
= {
189 .sync_mode
= WB_SYNC_ALL
,
190 .nr_to_write
= LONG_MAX
,
194 if (unlikely(f2fs_readonly(inode
->i_sb
)))
197 trace_f2fs_sync_file_enter(inode
);
199 /* if fdatasync is triggered, let's do in-place-update */
200 if (datasync
|| get_dirty_pages(inode
) <= SM_I(sbi
)->min_fsync_blocks
)
201 set_inode_flag(inode
, FI_NEED_IPU
);
202 ret
= filemap_write_and_wait_range(inode
->i_mapping
, start
, end
);
203 clear_inode_flag(inode
, FI_NEED_IPU
);
206 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
210 /* if the inode is dirty, let's recover all the time */
212 f2fs_write_inode(inode
, NULL
);
217 * if there is no written data, don't waste time to write recovery info.
219 if (!is_inode_flag_set(inode
, FI_APPEND_WRITE
) &&
220 !exist_written_data(sbi
, ino
, APPEND_INO
)) {
222 /* it may call write_inode just prior to fsync */
223 if (need_inode_page_update(sbi
, ino
))
226 if (is_inode_flag_set(inode
, FI_UPDATE_WRITE
) ||
227 exist_written_data(sbi
, ino
, UPDATE_INO
))
233 * Both of fdatasync() and fsync() are able to be recovered from
236 down_read(&F2FS_I(inode
)->i_sem
);
237 need_cp
= need_do_checkpoint(inode
);
238 up_read(&F2FS_I(inode
)->i_sem
);
241 /* all the dirty node pages should be flushed for POR */
242 ret
= f2fs_sync_fs(inode
->i_sb
, 1);
245 * We've secured consistency through sync_fs. Following pino
246 * will be used only for fsynced inodes after checkpoint.
248 try_to_fix_pino(inode
);
249 clear_inode_flag(inode
, FI_APPEND_WRITE
);
250 clear_inode_flag(inode
, FI_UPDATE_WRITE
);
254 ret
= fsync_node_pages(sbi
, ino
, &wbc
, atomic
);
258 /* if cp_error was enabled, we should avoid infinite loop */
259 if (unlikely(f2fs_cp_error(sbi
))) {
264 if (need_inode_block_update(sbi
, ino
)) {
265 mark_inode_dirty_sync(inode
);
266 f2fs_write_inode(inode
, NULL
);
270 ret
= wait_on_node_pages_writeback(sbi
, ino
);
274 /* once recovery info is written, don't need to tack this */
275 remove_ino_entry(sbi
, ino
, APPEND_INO
);
276 clear_inode_flag(inode
, FI_APPEND_WRITE
);
278 remove_ino_entry(sbi
, ino
, UPDATE_INO
);
279 clear_inode_flag(inode
, FI_UPDATE_WRITE
);
280 ret
= f2fs_issue_flush(sbi
);
281 f2fs_update_time(sbi
, REQ_TIME
);
283 trace_f2fs_sync_file_exit(inode
, need_cp
, datasync
, ret
);
284 f2fs_trace_ios(NULL
, 1);
288 int f2fs_sync_file(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
290 return f2fs_do_sync_file(file
, start
, end
, datasync
, false);
293 static pgoff_t
__get_first_dirty_index(struct address_space
*mapping
,
294 pgoff_t pgofs
, int whence
)
299 if (whence
!= SEEK_DATA
)
302 /* find first dirty page index */
303 pagevec_init(&pvec
, 0);
304 nr_pages
= pagevec_lookup_tag(&pvec
, mapping
, &pgofs
,
305 PAGECACHE_TAG_DIRTY
, 1);
306 pgofs
= nr_pages
? pvec
.pages
[0]->index
: ULONG_MAX
;
307 pagevec_release(&pvec
);
311 static bool __found_offset(block_t blkaddr
, pgoff_t dirty
, pgoff_t pgofs
,
316 if ((blkaddr
== NEW_ADDR
&& dirty
== pgofs
) ||
317 (blkaddr
!= NEW_ADDR
&& blkaddr
!= NULL_ADDR
))
321 if (blkaddr
== NULL_ADDR
)
328 static loff_t
f2fs_seek_block(struct file
*file
, loff_t offset
, int whence
)
330 struct inode
*inode
= file
->f_mapping
->host
;
331 loff_t maxbytes
= inode
->i_sb
->s_maxbytes
;
332 struct dnode_of_data dn
;
333 pgoff_t pgofs
, end_offset
, dirty
;
334 loff_t data_ofs
= offset
;
340 isize
= i_size_read(inode
);
344 /* handle inline data case */
345 if (f2fs_has_inline_data(inode
) || f2fs_has_inline_dentry(inode
)) {
346 if (whence
== SEEK_HOLE
)
351 pgofs
= (pgoff_t
)(offset
>> PAGE_SHIFT
);
353 dirty
= __get_first_dirty_index(inode
->i_mapping
, pgofs
, whence
);
355 for (; data_ofs
< isize
; data_ofs
= (loff_t
)pgofs
<< PAGE_SHIFT
) {
356 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
357 err
= get_dnode_of_data(&dn
, pgofs
, LOOKUP_NODE_RA
);
358 if (err
&& err
!= -ENOENT
) {
360 } else if (err
== -ENOENT
) {
361 /* direct node does not exists */
362 if (whence
== SEEK_DATA
) {
363 pgofs
= get_next_page_offset(&dn
, pgofs
);
370 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
372 /* find data/hole in dnode block */
373 for (; dn
.ofs_in_node
< end_offset
;
374 dn
.ofs_in_node
++, pgofs
++,
375 data_ofs
= (loff_t
)pgofs
<< PAGE_SHIFT
) {
377 blkaddr
= datablock_addr(dn
.node_page
, dn
.ofs_in_node
);
379 if (__found_offset(blkaddr
, dirty
, pgofs
, whence
)) {
387 if (whence
== SEEK_DATA
)
390 if (whence
== SEEK_HOLE
&& data_ofs
> isize
)
393 return vfs_setpos(file
, data_ofs
, maxbytes
);
399 static loff_t
f2fs_llseek(struct file
*file
, loff_t offset
, int whence
)
401 struct inode
*inode
= file
->f_mapping
->host
;
402 loff_t maxbytes
= inode
->i_sb
->s_maxbytes
;
408 return generic_file_llseek_size(file
, offset
, whence
,
409 maxbytes
, i_size_read(inode
));
414 return f2fs_seek_block(file
, offset
, whence
);
420 static int f2fs_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
422 struct inode
*inode
= file_inode(file
);
425 if (f2fs_encrypted_inode(inode
)) {
426 err
= fscrypt_get_encryption_info(inode
);
429 if (!f2fs_encrypted_inode(inode
))
433 /* we don't need to use inline_data strictly */
434 err
= f2fs_convert_inline_inode(inode
);
439 vma
->vm_ops
= &f2fs_file_vm_ops
;
443 static int f2fs_file_open(struct inode
*inode
, struct file
*filp
)
445 int ret
= generic_file_open(inode
, filp
);
448 if (!ret
&& f2fs_encrypted_inode(inode
)) {
449 ret
= fscrypt_get_encryption_info(inode
);
452 if (!fscrypt_has_encryption_key(inode
))
455 dir
= dget_parent(file_dentry(filp
));
456 if (f2fs_encrypted_inode(d_inode(dir
)) &&
457 !fscrypt_has_permitted_context(d_inode(dir
), inode
)) {
465 int truncate_data_blocks_range(struct dnode_of_data
*dn
, int count
)
467 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
468 struct f2fs_node
*raw_node
;
469 int nr_free
= 0, ofs
= dn
->ofs_in_node
, len
= count
;
472 raw_node
= F2FS_NODE(dn
->node_page
);
473 addr
= blkaddr_in_node(raw_node
) + ofs
;
475 for (; count
> 0; count
--, addr
++, dn
->ofs_in_node
++) {
476 block_t blkaddr
= le32_to_cpu(*addr
);
477 if (blkaddr
== NULL_ADDR
)
480 dn
->data_blkaddr
= NULL_ADDR
;
481 set_data_blkaddr(dn
);
482 invalidate_blocks(sbi
, blkaddr
);
483 if (dn
->ofs_in_node
== 0 && IS_INODE(dn
->node_page
))
484 clear_inode_flag(dn
->inode
, FI_FIRST_BLOCK_WRITTEN
);
491 * once we invalidate valid blkaddr in range [ofs, ofs + count],
492 * we will invalidate all blkaddr in the whole range.
494 fofs
= start_bidx_of_node(ofs_of_node(dn
->node_page
),
496 f2fs_update_extent_cache_range(dn
, fofs
, 0, len
);
497 dec_valid_block_count(sbi
, dn
->inode
, nr_free
);
499 dn
->ofs_in_node
= ofs
;
501 f2fs_update_time(sbi
, REQ_TIME
);
502 trace_f2fs_truncate_data_blocks_range(dn
->inode
, dn
->nid
,
503 dn
->ofs_in_node
, nr_free
);
507 void truncate_data_blocks(struct dnode_of_data
*dn
)
509 truncate_data_blocks_range(dn
, ADDRS_PER_BLOCK
);
512 static int truncate_partial_data_page(struct inode
*inode
, u64 from
,
515 unsigned offset
= from
& (PAGE_SIZE
- 1);
516 pgoff_t index
= from
>> PAGE_SHIFT
;
517 struct address_space
*mapping
= inode
->i_mapping
;
520 if (!offset
&& !cache_only
)
524 page
= f2fs_grab_cache_page(mapping
, index
, false);
525 if (page
&& PageUptodate(page
))
527 f2fs_put_page(page
, 1);
531 page
= get_lock_data_page(inode
, index
, true);
535 f2fs_wait_on_page_writeback(page
, DATA
, true);
536 zero_user(page
, offset
, PAGE_SIZE
- offset
);
537 if (!cache_only
|| !f2fs_encrypted_inode(inode
) ||
538 !S_ISREG(inode
->i_mode
))
539 set_page_dirty(page
);
540 f2fs_put_page(page
, 1);
544 int truncate_blocks(struct inode
*inode
, u64 from
, bool lock
)
546 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
547 unsigned int blocksize
= inode
->i_sb
->s_blocksize
;
548 struct dnode_of_data dn
;
550 int count
= 0, err
= 0;
552 bool truncate_page
= false;
554 trace_f2fs_truncate_blocks_enter(inode
, from
);
556 free_from
= (pgoff_t
)F2FS_BYTES_TO_BLK(from
+ blocksize
- 1);
558 if (free_from
>= sbi
->max_file_blocks
)
564 ipage
= get_node_page(sbi
, inode
->i_ino
);
566 err
= PTR_ERR(ipage
);
570 if (f2fs_has_inline_data(inode
)) {
571 if (truncate_inline_inode(ipage
, from
))
572 set_page_dirty(ipage
);
573 f2fs_put_page(ipage
, 1);
574 truncate_page
= true;
578 set_new_dnode(&dn
, inode
, ipage
, NULL
, 0);
579 err
= get_dnode_of_data(&dn
, free_from
, LOOKUP_NODE_RA
);
586 count
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
588 count
-= dn
.ofs_in_node
;
589 f2fs_bug_on(sbi
, count
< 0);
591 if (dn
.ofs_in_node
|| IS_INODE(dn
.node_page
)) {
592 truncate_data_blocks_range(&dn
, count
);
598 err
= truncate_inode_blocks(inode
, free_from
);
603 /* lastly zero out the first data page */
605 err
= truncate_partial_data_page(inode
, from
, truncate_page
);
607 trace_f2fs_truncate_blocks_exit(inode
, err
);
611 int f2fs_truncate(struct inode
*inode
, bool lock
)
615 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
) ||
616 S_ISLNK(inode
->i_mode
)))
619 trace_f2fs_truncate(inode
);
621 /* we should check inline_data size */
622 if (!f2fs_may_inline_data(inode
)) {
623 err
= f2fs_convert_inline_inode(inode
);
628 err
= truncate_blocks(inode
, i_size_read(inode
), lock
);
632 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
633 mark_inode_dirty_sync(inode
);
637 int f2fs_getattr(struct vfsmount
*mnt
,
638 struct dentry
*dentry
, struct kstat
*stat
)
640 struct inode
*inode
= d_inode(dentry
);
641 generic_fillattr(inode
, stat
);
646 #ifdef CONFIG_F2FS_FS_POSIX_ACL
647 static void __setattr_copy(struct inode
*inode
, const struct iattr
*attr
)
649 unsigned int ia_valid
= attr
->ia_valid
;
651 if (ia_valid
& ATTR_UID
)
652 inode
->i_uid
= attr
->ia_uid
;
653 if (ia_valid
& ATTR_GID
)
654 inode
->i_gid
= attr
->ia_gid
;
655 if (ia_valid
& ATTR_ATIME
)
656 inode
->i_atime
= timespec_trunc(attr
->ia_atime
,
657 inode
->i_sb
->s_time_gran
);
658 if (ia_valid
& ATTR_MTIME
)
659 inode
->i_mtime
= timespec_trunc(attr
->ia_mtime
,
660 inode
->i_sb
->s_time_gran
);
661 if (ia_valid
& ATTR_CTIME
)
662 inode
->i_ctime
= timespec_trunc(attr
->ia_ctime
,
663 inode
->i_sb
->s_time_gran
);
664 if (ia_valid
& ATTR_MODE
) {
665 umode_t mode
= attr
->ia_mode
;
667 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
669 set_acl_inode(inode
, mode
);
673 #define __setattr_copy setattr_copy
676 int f2fs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
678 struct inode
*inode
= d_inode(dentry
);
681 err
= inode_change_ok(inode
, attr
);
685 if (attr
->ia_valid
& ATTR_SIZE
) {
686 if (f2fs_encrypted_inode(inode
) &&
687 fscrypt_get_encryption_info(inode
))
690 if (attr
->ia_size
<= i_size_read(inode
)) {
691 truncate_setsize(inode
, attr
->ia_size
);
692 err
= f2fs_truncate(inode
, true);
695 f2fs_balance_fs(F2FS_I_SB(inode
), true);
698 * do not trim all blocks after i_size if target size is
699 * larger than i_size.
701 truncate_setsize(inode
, attr
->ia_size
);
703 /* should convert inline inode here */
704 if (!f2fs_may_inline_data(inode
)) {
705 err
= f2fs_convert_inline_inode(inode
);
709 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
713 __setattr_copy(inode
, attr
);
715 if (attr
->ia_valid
& ATTR_MODE
) {
716 err
= posix_acl_chmod(inode
, get_inode_mode(inode
));
717 if (err
|| is_inode_flag_set(inode
, FI_ACL_MODE
)) {
718 inode
->i_mode
= F2FS_I(inode
)->i_acl_mode
;
719 clear_inode_flag(inode
, FI_ACL_MODE
);
723 mark_inode_dirty_sync(inode
);
727 const struct inode_operations f2fs_file_inode_operations
= {
728 .getattr
= f2fs_getattr
,
729 .setattr
= f2fs_setattr
,
730 .get_acl
= f2fs_get_acl
,
731 .set_acl
= f2fs_set_acl
,
732 #ifdef CONFIG_F2FS_FS_XATTR
733 .setxattr
= generic_setxattr
,
734 .getxattr
= generic_getxattr
,
735 .listxattr
= f2fs_listxattr
,
736 .removexattr
= generic_removexattr
,
738 .fiemap
= f2fs_fiemap
,
741 static int fill_zero(struct inode
*inode
, pgoff_t index
,
742 loff_t start
, loff_t len
)
744 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
750 f2fs_balance_fs(sbi
, true);
753 page
= get_new_data_page(inode
, NULL
, index
, false);
757 return PTR_ERR(page
);
759 f2fs_wait_on_page_writeback(page
, DATA
, true);
760 zero_user(page
, start
, len
);
761 set_page_dirty(page
);
762 f2fs_put_page(page
, 1);
766 int truncate_hole(struct inode
*inode
, pgoff_t pg_start
, pgoff_t pg_end
)
770 while (pg_start
< pg_end
) {
771 struct dnode_of_data dn
;
772 pgoff_t end_offset
, count
;
774 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
775 err
= get_dnode_of_data(&dn
, pg_start
, LOOKUP_NODE
);
777 if (err
== -ENOENT
) {
784 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
785 count
= min(end_offset
- dn
.ofs_in_node
, pg_end
- pg_start
);
787 f2fs_bug_on(F2FS_I_SB(inode
), count
== 0 || count
> end_offset
);
789 truncate_data_blocks_range(&dn
, count
);
797 static int punch_hole(struct inode
*inode
, loff_t offset
, loff_t len
)
799 pgoff_t pg_start
, pg_end
;
800 loff_t off_start
, off_end
;
803 ret
= f2fs_convert_inline_inode(inode
);
807 pg_start
= ((unsigned long long) offset
) >> PAGE_SHIFT
;
808 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_SHIFT
;
810 off_start
= offset
& (PAGE_SIZE
- 1);
811 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
813 if (pg_start
== pg_end
) {
814 ret
= fill_zero(inode
, pg_start
, off_start
,
815 off_end
- off_start
);
820 ret
= fill_zero(inode
, pg_start
++, off_start
,
821 PAGE_SIZE
- off_start
);
826 ret
= fill_zero(inode
, pg_end
, 0, off_end
);
831 if (pg_start
< pg_end
) {
832 struct address_space
*mapping
= inode
->i_mapping
;
833 loff_t blk_start
, blk_end
;
834 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
836 f2fs_balance_fs(sbi
, true);
838 blk_start
= (loff_t
)pg_start
<< PAGE_SHIFT
;
839 blk_end
= (loff_t
)pg_end
<< PAGE_SHIFT
;
840 truncate_inode_pages_range(mapping
, blk_start
,
844 ret
= truncate_hole(inode
, pg_start
, pg_end
);
852 static int __exchange_data_block(struct inode
*inode
, pgoff_t src
,
853 pgoff_t dst
, bool full
)
855 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
856 struct dnode_of_data dn
;
858 bool do_replace
= false;
861 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
862 ret
= get_dnode_of_data(&dn
, src
, LOOKUP_NODE_RA
);
863 if (ret
&& ret
!= -ENOENT
) {
865 } else if (ret
== -ENOENT
) {
866 new_addr
= NULL_ADDR
;
868 new_addr
= dn
.data_blkaddr
;
869 if (!is_checkpointed_data(sbi
, new_addr
)) {
870 /* do not invalidate this block address */
871 f2fs_update_data_blkaddr(&dn
, NULL_ADDR
);
877 if (new_addr
== NULL_ADDR
)
878 return full
? truncate_hole(inode
, dst
, dst
+ 1) : 0;
881 struct page
*ipage
= get_node_page(sbi
, inode
->i_ino
);
885 ret
= PTR_ERR(ipage
);
889 set_new_dnode(&dn
, inode
, ipage
, NULL
, 0);
890 ret
= f2fs_reserve_block(&dn
, dst
);
894 truncate_data_blocks_range(&dn
, 1);
896 get_node_info(sbi
, dn
.nid
, &ni
);
897 f2fs_replace_block(sbi
, &dn
, dn
.data_blkaddr
, new_addr
,
898 ni
.version
, true, false);
901 struct page
*psrc
, *pdst
;
903 psrc
= get_lock_data_page(inode
, src
, true);
905 return PTR_ERR(psrc
);
906 pdst
= get_new_data_page(inode
, NULL
, dst
, true);
908 f2fs_put_page(psrc
, 1);
909 return PTR_ERR(pdst
);
911 f2fs_copy_page(psrc
, pdst
);
912 set_page_dirty(pdst
);
913 f2fs_put_page(pdst
, 1);
914 f2fs_put_page(psrc
, 1);
916 return truncate_hole(inode
, src
, src
+ 1);
921 if (!get_dnode_of_data(&dn
, src
, LOOKUP_NODE
)) {
922 f2fs_update_data_blkaddr(&dn
, new_addr
);
928 static int f2fs_do_collapse(struct inode
*inode
, pgoff_t start
, pgoff_t end
)
930 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
931 pgoff_t nrpages
= (i_size_read(inode
) + PAGE_SIZE
- 1) / PAGE_SIZE
;
934 for (; end
< nrpages
; start
++, end
++) {
935 f2fs_balance_fs(sbi
, true);
937 ret
= __exchange_data_block(inode
, end
, start
, true);
945 static int f2fs_collapse_range(struct inode
*inode
, loff_t offset
, loff_t len
)
947 pgoff_t pg_start
, pg_end
;
951 if (offset
+ len
>= i_size_read(inode
))
954 /* collapse range should be aligned to block size of f2fs. */
955 if (offset
& (F2FS_BLKSIZE
- 1) || len
& (F2FS_BLKSIZE
- 1))
958 ret
= f2fs_convert_inline_inode(inode
);
962 pg_start
= offset
>> PAGE_SHIFT
;
963 pg_end
= (offset
+ len
) >> PAGE_SHIFT
;
965 /* write out all dirty pages from offset */
966 ret
= filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
970 truncate_pagecache(inode
, offset
);
972 ret
= f2fs_do_collapse(inode
, pg_start
, pg_end
);
976 /* write out all moved pages, if possible */
977 filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
978 truncate_pagecache(inode
, offset
);
980 new_size
= i_size_read(inode
) - len
;
981 truncate_pagecache(inode
, new_size
);
983 ret
= truncate_blocks(inode
, new_size
, true);
985 f2fs_i_size_write(inode
, new_size
);
990 static int f2fs_do_zero_range(struct dnode_of_data
*dn
, pgoff_t start
,
993 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
994 pgoff_t index
= start
;
995 unsigned int ofs_in_node
= dn
->ofs_in_node
;
999 for (; index
< end
; index
++, dn
->ofs_in_node
++) {
1000 if (datablock_addr(dn
->node_page
, dn
->ofs_in_node
) == NULL_ADDR
)
1004 dn
->ofs_in_node
= ofs_in_node
;
1005 ret
= reserve_new_blocks(dn
, count
);
1009 dn
->ofs_in_node
= ofs_in_node
;
1010 for (index
= start
; index
< end
; index
++, dn
->ofs_in_node
++) {
1012 datablock_addr(dn
->node_page
, dn
->ofs_in_node
);
1014 * reserve_new_blocks will not guarantee entire block
1017 if (dn
->data_blkaddr
== NULL_ADDR
) {
1021 if (dn
->data_blkaddr
!= NEW_ADDR
) {
1022 invalidate_blocks(sbi
, dn
->data_blkaddr
);
1023 dn
->data_blkaddr
= NEW_ADDR
;
1024 set_data_blkaddr(dn
);
1028 f2fs_update_extent_cache_range(dn
, start
, 0, index
- start
);
1033 static int f2fs_zero_range(struct inode
*inode
, loff_t offset
, loff_t len
,
1036 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1037 struct address_space
*mapping
= inode
->i_mapping
;
1038 pgoff_t index
, pg_start
, pg_end
;
1039 loff_t new_size
= i_size_read(inode
);
1040 loff_t off_start
, off_end
;
1043 ret
= inode_newsize_ok(inode
, (len
+ offset
));
1047 ret
= f2fs_convert_inline_inode(inode
);
1051 ret
= filemap_write_and_wait_range(mapping
, offset
, offset
+ len
- 1);
1055 truncate_pagecache_range(inode
, offset
, offset
+ len
- 1);
1057 pg_start
= ((unsigned long long) offset
) >> PAGE_SHIFT
;
1058 pg_end
= ((unsigned long long) offset
+ len
) >> PAGE_SHIFT
;
1060 off_start
= offset
& (PAGE_SIZE
- 1);
1061 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
1063 if (pg_start
== pg_end
) {
1064 ret
= fill_zero(inode
, pg_start
, off_start
,
1065 off_end
- off_start
);
1069 if (offset
+ len
> new_size
)
1070 new_size
= offset
+ len
;
1071 new_size
= max_t(loff_t
, new_size
, offset
+ len
);
1074 ret
= fill_zero(inode
, pg_start
++, off_start
,
1075 PAGE_SIZE
- off_start
);
1079 new_size
= max_t(loff_t
, new_size
,
1080 (loff_t
)pg_start
<< PAGE_SHIFT
);
1083 for (index
= pg_start
; index
< pg_end
;) {
1084 struct dnode_of_data dn
;
1085 unsigned int end_offset
;
1090 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
1091 ret
= get_dnode_of_data(&dn
, index
, ALLOC_NODE
);
1093 f2fs_unlock_op(sbi
);
1097 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
1098 end
= min(pg_end
, end_offset
- dn
.ofs_in_node
+ index
);
1100 ret
= f2fs_do_zero_range(&dn
, index
, end
);
1101 f2fs_put_dnode(&dn
);
1102 f2fs_unlock_op(sbi
);
1107 new_size
= max_t(loff_t
, new_size
,
1108 (loff_t
)index
<< PAGE_SHIFT
);
1112 ret
= fill_zero(inode
, pg_end
, 0, off_end
);
1116 new_size
= max_t(loff_t
, new_size
, offset
+ len
);
1121 if (!(mode
& FALLOC_FL_KEEP_SIZE
) && i_size_read(inode
) < new_size
)
1122 f2fs_i_size_write(inode
, new_size
);
1127 static int f2fs_insert_range(struct inode
*inode
, loff_t offset
, loff_t len
)
1129 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1130 pgoff_t pg_start
, pg_end
, delta
, nrpages
, idx
;
1134 new_size
= i_size_read(inode
) + len
;
1135 if (new_size
> inode
->i_sb
->s_maxbytes
)
1138 if (offset
>= i_size_read(inode
))
1141 /* insert range should be aligned to block size of f2fs. */
1142 if (offset
& (F2FS_BLKSIZE
- 1) || len
& (F2FS_BLKSIZE
- 1))
1145 ret
= f2fs_convert_inline_inode(inode
);
1149 f2fs_balance_fs(sbi
, true);
1151 ret
= truncate_blocks(inode
, i_size_read(inode
), true);
1155 /* write out all dirty pages from offset */
1156 ret
= filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
1160 truncate_pagecache(inode
, offset
);
1162 pg_start
= offset
>> PAGE_SHIFT
;
1163 pg_end
= (offset
+ len
) >> PAGE_SHIFT
;
1164 delta
= pg_end
- pg_start
;
1165 nrpages
= (i_size_read(inode
) + PAGE_SIZE
- 1) / PAGE_SIZE
;
1167 for (idx
= nrpages
- 1; idx
>= pg_start
&& idx
!= -1; idx
--) {
1169 ret
= __exchange_data_block(inode
, idx
, idx
+ delta
, false);
1170 f2fs_unlock_op(sbi
);
1175 /* write out all moved pages, if possible */
1176 filemap_write_and_wait_range(inode
->i_mapping
, offset
, LLONG_MAX
);
1177 truncate_pagecache(inode
, offset
);
1180 f2fs_i_size_write(inode
, new_size
);
1184 static int expand_inode_data(struct inode
*inode
, loff_t offset
,
1185 loff_t len
, int mode
)
1187 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1188 struct f2fs_map_blocks map
= { .m_next_pgofs
= NULL
};
1190 loff_t new_size
= i_size_read(inode
);
1194 ret
= inode_newsize_ok(inode
, (len
+ offset
));
1198 ret
= f2fs_convert_inline_inode(inode
);
1202 f2fs_balance_fs(sbi
, true);
1204 pg_end
= ((unsigned long long)offset
+ len
) >> PAGE_SHIFT
;
1205 off_end
= (offset
+ len
) & (PAGE_SIZE
- 1);
1207 map
.m_lblk
= ((unsigned long long)offset
) >> PAGE_SHIFT
;
1208 map
.m_len
= pg_end
- map
.m_lblk
;
1212 ret
= f2fs_map_blocks(inode
, &map
, 1, F2FS_GET_BLOCK_PRE_AIO
);
1219 last_off
= map
.m_lblk
+ map
.m_len
- 1;
1221 /* update new size to the failed position */
1222 new_size
= (last_off
== pg_end
) ? offset
+ len
:
1223 (loff_t
)(last_off
+ 1) << PAGE_SHIFT
;
1225 new_size
= ((loff_t
)pg_end
<< PAGE_SHIFT
) + off_end
;
1228 if (!(mode
& FALLOC_FL_KEEP_SIZE
) && i_size_read(inode
) < new_size
)
1229 f2fs_i_size_write(inode
, new_size
);
1234 static long f2fs_fallocate(struct file
*file
, int mode
,
1235 loff_t offset
, loff_t len
)
1237 struct inode
*inode
= file_inode(file
);
1240 /* f2fs only support ->fallocate for regular file */
1241 if (!S_ISREG(inode
->i_mode
))
1244 if (f2fs_encrypted_inode(inode
) &&
1245 (mode
& (FALLOC_FL_COLLAPSE_RANGE
| FALLOC_FL_INSERT_RANGE
)))
1248 if (mode
& ~(FALLOC_FL_KEEP_SIZE
| FALLOC_FL_PUNCH_HOLE
|
1249 FALLOC_FL_COLLAPSE_RANGE
| FALLOC_FL_ZERO_RANGE
|
1250 FALLOC_FL_INSERT_RANGE
))
1255 if (mode
& FALLOC_FL_PUNCH_HOLE
) {
1256 if (offset
>= inode
->i_size
)
1259 ret
= punch_hole(inode
, offset
, len
);
1260 } else if (mode
& FALLOC_FL_COLLAPSE_RANGE
) {
1261 ret
= f2fs_collapse_range(inode
, offset
, len
);
1262 } else if (mode
& FALLOC_FL_ZERO_RANGE
) {
1263 ret
= f2fs_zero_range(inode
, offset
, len
, mode
);
1264 } else if (mode
& FALLOC_FL_INSERT_RANGE
) {
1265 ret
= f2fs_insert_range(inode
, offset
, len
);
1267 ret
= expand_inode_data(inode
, offset
, len
, mode
);
1271 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
1272 mark_inode_dirty_sync(inode
);
1273 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1277 inode_unlock(inode
);
1279 trace_f2fs_fallocate(inode
, mode
, offset
, len
, ret
);
1283 static int f2fs_release_file(struct inode
*inode
, struct file
*filp
)
1286 * f2fs_relase_file is called at every close calls. So we should
1287 * not drop any inmemory pages by close called by other process.
1289 if (!(filp
->f_mode
& FMODE_WRITE
) ||
1290 atomic_read(&inode
->i_writecount
) != 1)
1293 /* some remained atomic pages should discarded */
1294 if (f2fs_is_atomic_file(inode
))
1295 drop_inmem_pages(inode
);
1296 if (f2fs_is_volatile_file(inode
)) {
1297 clear_inode_flag(inode
, FI_VOLATILE_FILE
);
1298 set_inode_flag(inode
, FI_DROP_CACHE
);
1299 filemap_fdatawrite(inode
->i_mapping
);
1300 clear_inode_flag(inode
, FI_DROP_CACHE
);
1305 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1306 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1308 static inline __u32
f2fs_mask_flags(umode_t mode
, __u32 flags
)
1312 else if (S_ISREG(mode
))
1313 return flags
& F2FS_REG_FLMASK
;
1315 return flags
& F2FS_OTHER_FLMASK
;
1318 static int f2fs_ioc_getflags(struct file
*filp
, unsigned long arg
)
1320 struct inode
*inode
= file_inode(filp
);
1321 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
1322 unsigned int flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
1323 return put_user(flags
, (int __user
*)arg
);
1326 static int f2fs_ioc_setflags(struct file
*filp
, unsigned long arg
)
1328 struct inode
*inode
= file_inode(filp
);
1329 struct f2fs_inode_info
*fi
= F2FS_I(inode
);
1330 unsigned int flags
= fi
->i_flags
& FS_FL_USER_VISIBLE
;
1331 unsigned int oldflags
;
1334 if (!inode_owner_or_capable(inode
))
1337 if (get_user(flags
, (int __user
*)arg
))
1340 ret
= mnt_want_write_file(filp
);
1344 flags
= f2fs_mask_flags(inode
->i_mode
, flags
);
1348 oldflags
= fi
->i_flags
;
1350 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
1351 if (!capable(CAP_LINUX_IMMUTABLE
)) {
1352 inode_unlock(inode
);
1358 flags
= flags
& FS_FL_USER_MODIFIABLE
;
1359 flags
|= oldflags
& ~FS_FL_USER_MODIFIABLE
;
1360 fi
->i_flags
= flags
;
1361 inode_unlock(inode
);
1363 inode
->i_ctime
= CURRENT_TIME
;
1364 f2fs_set_inode_flags(inode
);
1366 mnt_drop_write_file(filp
);
1370 static int f2fs_ioc_getversion(struct file
*filp
, unsigned long arg
)
1372 struct inode
*inode
= file_inode(filp
);
1374 return put_user(inode
->i_generation
, (int __user
*)arg
);
1377 static int f2fs_ioc_start_atomic_write(struct file
*filp
)
1379 struct inode
*inode
= file_inode(filp
);
1382 if (!inode_owner_or_capable(inode
))
1385 ret
= mnt_want_write_file(filp
);
1391 if (f2fs_is_atomic_file(inode
))
1394 ret
= f2fs_convert_inline_inode(inode
);
1398 set_inode_flag(inode
, FI_ATOMIC_FILE
);
1399 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1401 if (!get_dirty_pages(inode
))
1404 f2fs_msg(F2FS_I_SB(inode
)->sb
, KERN_WARNING
,
1405 "Unexpected flush for atomic writes: ino=%lu, npages=%lld",
1406 inode
->i_ino
, get_dirty_pages(inode
));
1407 ret
= filemap_write_and_wait_range(inode
->i_mapping
, 0, LLONG_MAX
);
1409 clear_inode_flag(inode
, FI_ATOMIC_FILE
);
1411 inode_unlock(inode
);
1412 mnt_drop_write_file(filp
);
1416 static int f2fs_ioc_commit_atomic_write(struct file
*filp
)
1418 struct inode
*inode
= file_inode(filp
);
1421 if (!inode_owner_or_capable(inode
))
1424 ret
= mnt_want_write_file(filp
);
1430 if (f2fs_is_volatile_file(inode
))
1433 if (f2fs_is_atomic_file(inode
)) {
1434 clear_inode_flag(inode
, FI_ATOMIC_FILE
);
1435 ret
= commit_inmem_pages(inode
);
1437 set_inode_flag(inode
, FI_ATOMIC_FILE
);
1442 ret
= f2fs_do_sync_file(filp
, 0, LLONG_MAX
, 0, true);
1444 inode_unlock(inode
);
1445 mnt_drop_write_file(filp
);
1449 static int f2fs_ioc_start_volatile_write(struct file
*filp
)
1451 struct inode
*inode
= file_inode(filp
);
1454 if (!inode_owner_or_capable(inode
))
1457 ret
= mnt_want_write_file(filp
);
1463 if (f2fs_is_volatile_file(inode
))
1466 ret
= f2fs_convert_inline_inode(inode
);
1470 set_inode_flag(inode
, FI_VOLATILE_FILE
);
1471 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1473 inode_unlock(inode
);
1474 mnt_drop_write_file(filp
);
1478 static int f2fs_ioc_release_volatile_write(struct file
*filp
)
1480 struct inode
*inode
= file_inode(filp
);
1483 if (!inode_owner_or_capable(inode
))
1486 ret
= mnt_want_write_file(filp
);
1492 if (!f2fs_is_volatile_file(inode
))
1495 if (!f2fs_is_first_block_written(inode
)) {
1496 ret
= truncate_partial_data_page(inode
, 0, true);
1500 ret
= punch_hole(inode
, 0, F2FS_BLKSIZE
);
1502 inode_unlock(inode
);
1503 mnt_drop_write_file(filp
);
1507 static int f2fs_ioc_abort_volatile_write(struct file
*filp
)
1509 struct inode
*inode
= file_inode(filp
);
1512 if (!inode_owner_or_capable(inode
))
1515 ret
= mnt_want_write_file(filp
);
1521 if (f2fs_is_atomic_file(inode
))
1522 drop_inmem_pages(inode
);
1523 if (f2fs_is_volatile_file(inode
)) {
1524 clear_inode_flag(inode
, FI_VOLATILE_FILE
);
1525 ret
= f2fs_do_sync_file(filp
, 0, LLONG_MAX
, 0, true);
1528 inode_unlock(inode
);
1530 mnt_drop_write_file(filp
);
1531 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1535 static int f2fs_ioc_shutdown(struct file
*filp
, unsigned long arg
)
1537 struct inode
*inode
= file_inode(filp
);
1538 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1539 struct super_block
*sb
= sbi
->sb
;
1543 if (!capable(CAP_SYS_ADMIN
))
1546 if (get_user(in
, (__u32 __user
*)arg
))
1549 ret
= mnt_want_write_file(filp
);
1554 case F2FS_GOING_DOWN_FULLSYNC
:
1555 sb
= freeze_bdev(sb
->s_bdev
);
1556 if (sb
&& !IS_ERR(sb
)) {
1557 f2fs_stop_checkpoint(sbi
, false);
1558 thaw_bdev(sb
->s_bdev
, sb
);
1561 case F2FS_GOING_DOWN_METASYNC
:
1562 /* do checkpoint only */
1563 f2fs_sync_fs(sb
, 1);
1564 f2fs_stop_checkpoint(sbi
, false);
1566 case F2FS_GOING_DOWN_NOSYNC
:
1567 f2fs_stop_checkpoint(sbi
, false);
1569 case F2FS_GOING_DOWN_METAFLUSH
:
1570 sync_meta_pages(sbi
, META
, LONG_MAX
);
1571 f2fs_stop_checkpoint(sbi
, false);
1577 f2fs_update_time(sbi
, REQ_TIME
);
1579 mnt_drop_write_file(filp
);
1583 static int f2fs_ioc_fitrim(struct file
*filp
, unsigned long arg
)
1585 struct inode
*inode
= file_inode(filp
);
1586 struct super_block
*sb
= inode
->i_sb
;
1587 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
1588 struct fstrim_range range
;
1591 if (!capable(CAP_SYS_ADMIN
))
1594 if (!blk_queue_discard(q
))
1597 if (copy_from_user(&range
, (struct fstrim_range __user
*)arg
,
1601 ret
= mnt_want_write_file(filp
);
1605 range
.minlen
= max((unsigned int)range
.minlen
,
1606 q
->limits
.discard_granularity
);
1607 ret
= f2fs_trim_fs(F2FS_SB(sb
), &range
);
1608 mnt_drop_write_file(filp
);
1612 if (copy_to_user((struct fstrim_range __user
*)arg
, &range
,
1615 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1619 static bool uuid_is_nonzero(__u8 u
[16])
1623 for (i
= 0; i
< 16; i
++)
1629 static int f2fs_ioc_set_encryption_policy(struct file
*filp
, unsigned long arg
)
1631 struct fscrypt_policy policy
;
1632 struct inode
*inode
= file_inode(filp
);
1635 if (copy_from_user(&policy
, (struct fscrypt_policy __user
*)arg
,
1639 ret
= mnt_want_write_file(filp
);
1643 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1644 ret
= fscrypt_process_policy(inode
, &policy
);
1646 mnt_drop_write_file(filp
);
1650 static int f2fs_ioc_get_encryption_policy(struct file
*filp
, unsigned long arg
)
1652 struct fscrypt_policy policy
;
1653 struct inode
*inode
= file_inode(filp
);
1656 err
= fscrypt_get_policy(inode
, &policy
);
1660 if (copy_to_user((struct fscrypt_policy __user
*)arg
, &policy
, sizeof(policy
)))
1665 static int f2fs_ioc_get_encryption_pwsalt(struct file
*filp
, unsigned long arg
)
1667 struct inode
*inode
= file_inode(filp
);
1668 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1671 if (!f2fs_sb_has_crypto(inode
->i_sb
))
1674 if (uuid_is_nonzero(sbi
->raw_super
->encrypt_pw_salt
))
1677 err
= mnt_want_write_file(filp
);
1681 /* update superblock with uuid */
1682 generate_random_uuid(sbi
->raw_super
->encrypt_pw_salt
);
1684 err
= f2fs_commit_super(sbi
, false);
1687 memset(sbi
->raw_super
->encrypt_pw_salt
, 0, 16);
1688 mnt_drop_write_file(filp
);
1691 mnt_drop_write_file(filp
);
1693 if (copy_to_user((__u8 __user
*)arg
, sbi
->raw_super
->encrypt_pw_salt
,
1699 static int f2fs_ioc_gc(struct file
*filp
, unsigned long arg
)
1701 struct inode
*inode
= file_inode(filp
);
1702 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1706 if (!capable(CAP_SYS_ADMIN
))
1709 if (get_user(sync
, (__u32 __user
*)arg
))
1712 if (f2fs_readonly(sbi
->sb
))
1715 ret
= mnt_want_write_file(filp
);
1720 if (!mutex_trylock(&sbi
->gc_mutex
)) {
1725 mutex_lock(&sbi
->gc_mutex
);
1728 ret
= f2fs_gc(sbi
, sync
);
1730 mnt_drop_write_file(filp
);
1734 static int f2fs_ioc_write_checkpoint(struct file
*filp
, unsigned long arg
)
1736 struct inode
*inode
= file_inode(filp
);
1737 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1740 if (!capable(CAP_SYS_ADMIN
))
1743 if (f2fs_readonly(sbi
->sb
))
1746 ret
= mnt_want_write_file(filp
);
1750 ret
= f2fs_sync_fs(sbi
->sb
, 1);
1752 mnt_drop_write_file(filp
);
1756 static int f2fs_defragment_range(struct f2fs_sb_info
*sbi
,
1758 struct f2fs_defragment
*range
)
1760 struct inode
*inode
= file_inode(filp
);
1761 struct f2fs_map_blocks map
= { .m_next_pgofs
= NULL
};
1762 struct extent_info ei
;
1763 pgoff_t pg_start
, pg_end
;
1764 unsigned int blk_per_seg
= sbi
->blocks_per_seg
;
1765 unsigned int total
= 0, sec_num
;
1766 unsigned int pages_per_sec
= sbi
->segs_per_sec
* blk_per_seg
;
1767 block_t blk_end
= 0;
1768 bool fragmented
= false;
1771 /* if in-place-update policy is enabled, don't waste time here */
1772 if (need_inplace_update(inode
))
1775 pg_start
= range
->start
>> PAGE_SHIFT
;
1776 pg_end
= (range
->start
+ range
->len
) >> PAGE_SHIFT
;
1778 f2fs_balance_fs(sbi
, true);
1782 /* writeback all dirty pages in the range */
1783 err
= filemap_write_and_wait_range(inode
->i_mapping
, range
->start
,
1784 range
->start
+ range
->len
- 1);
1789 * lookup mapping info in extent cache, skip defragmenting if physical
1790 * block addresses are continuous.
1792 if (f2fs_lookup_extent_cache(inode
, pg_start
, &ei
)) {
1793 if (ei
.fofs
+ ei
.len
>= pg_end
)
1797 map
.m_lblk
= pg_start
;
1800 * lookup mapping info in dnode page cache, skip defragmenting if all
1801 * physical block addresses are continuous even if there are hole(s)
1802 * in logical blocks.
1804 while (map
.m_lblk
< pg_end
) {
1805 map
.m_len
= pg_end
- map
.m_lblk
;
1806 err
= f2fs_map_blocks(inode
, &map
, 0, F2FS_GET_BLOCK_READ
);
1810 if (!(map
.m_flags
& F2FS_MAP_FLAGS
)) {
1815 if (blk_end
&& blk_end
!= map
.m_pblk
) {
1819 blk_end
= map
.m_pblk
+ map
.m_len
;
1821 map
.m_lblk
+= map
.m_len
;
1827 map
.m_lblk
= pg_start
;
1828 map
.m_len
= pg_end
- pg_start
;
1830 sec_num
= (map
.m_len
+ pages_per_sec
- 1) / pages_per_sec
;
1833 * make sure there are enough free section for LFS allocation, this can
1834 * avoid defragment running in SSR mode when free section are allocated
1837 if (has_not_enough_free_secs(sbi
, sec_num
)) {
1842 while (map
.m_lblk
< pg_end
) {
1847 map
.m_len
= pg_end
- map
.m_lblk
;
1848 err
= f2fs_map_blocks(inode
, &map
, 0, F2FS_GET_BLOCK_READ
);
1852 if (!(map
.m_flags
& F2FS_MAP_FLAGS
)) {
1857 set_inode_flag(inode
, FI_DO_DEFRAG
);
1860 while (idx
< map
.m_lblk
+ map
.m_len
&& cnt
< blk_per_seg
) {
1863 page
= get_lock_data_page(inode
, idx
, true);
1865 err
= PTR_ERR(page
);
1869 set_page_dirty(page
);
1870 f2fs_put_page(page
, 1);
1879 if (idx
< pg_end
&& cnt
< blk_per_seg
)
1882 clear_inode_flag(inode
, FI_DO_DEFRAG
);
1884 err
= filemap_fdatawrite(inode
->i_mapping
);
1889 clear_inode_flag(inode
, FI_DO_DEFRAG
);
1891 inode_unlock(inode
);
1893 range
->len
= (u64
)total
<< PAGE_SHIFT
;
1897 static int f2fs_ioc_defragment(struct file
*filp
, unsigned long arg
)
1899 struct inode
*inode
= file_inode(filp
);
1900 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1901 struct f2fs_defragment range
;
1904 if (!capable(CAP_SYS_ADMIN
))
1907 if (!S_ISREG(inode
->i_mode
))
1910 err
= mnt_want_write_file(filp
);
1914 if (f2fs_readonly(sbi
->sb
)) {
1919 if (copy_from_user(&range
, (struct f2fs_defragment __user
*)arg
,
1925 /* verify alignment of offset & size */
1926 if (range
.start
& (F2FS_BLKSIZE
- 1) ||
1927 range
.len
& (F2FS_BLKSIZE
- 1)) {
1932 err
= f2fs_defragment_range(sbi
, filp
, &range
);
1933 f2fs_update_time(sbi
, REQ_TIME
);
1937 if (copy_to_user((struct f2fs_defragment __user
*)arg
, &range
,
1941 mnt_drop_write_file(filp
);
1945 long f2fs_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
1948 case F2FS_IOC_GETFLAGS
:
1949 return f2fs_ioc_getflags(filp
, arg
);
1950 case F2FS_IOC_SETFLAGS
:
1951 return f2fs_ioc_setflags(filp
, arg
);
1952 case F2FS_IOC_GETVERSION
:
1953 return f2fs_ioc_getversion(filp
, arg
);
1954 case F2FS_IOC_START_ATOMIC_WRITE
:
1955 return f2fs_ioc_start_atomic_write(filp
);
1956 case F2FS_IOC_COMMIT_ATOMIC_WRITE
:
1957 return f2fs_ioc_commit_atomic_write(filp
);
1958 case F2FS_IOC_START_VOLATILE_WRITE
:
1959 return f2fs_ioc_start_volatile_write(filp
);
1960 case F2FS_IOC_RELEASE_VOLATILE_WRITE
:
1961 return f2fs_ioc_release_volatile_write(filp
);
1962 case F2FS_IOC_ABORT_VOLATILE_WRITE
:
1963 return f2fs_ioc_abort_volatile_write(filp
);
1964 case F2FS_IOC_SHUTDOWN
:
1965 return f2fs_ioc_shutdown(filp
, arg
);
1967 return f2fs_ioc_fitrim(filp
, arg
);
1968 case F2FS_IOC_SET_ENCRYPTION_POLICY
:
1969 return f2fs_ioc_set_encryption_policy(filp
, arg
);
1970 case F2FS_IOC_GET_ENCRYPTION_POLICY
:
1971 return f2fs_ioc_get_encryption_policy(filp
, arg
);
1972 case F2FS_IOC_GET_ENCRYPTION_PWSALT
:
1973 return f2fs_ioc_get_encryption_pwsalt(filp
, arg
);
1974 case F2FS_IOC_GARBAGE_COLLECT
:
1975 return f2fs_ioc_gc(filp
, arg
);
1976 case F2FS_IOC_WRITE_CHECKPOINT
:
1977 return f2fs_ioc_write_checkpoint(filp
, arg
);
1978 case F2FS_IOC_DEFRAGMENT
:
1979 return f2fs_ioc_defragment(filp
, arg
);
1985 static ssize_t
f2fs_file_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1987 struct file
*file
= iocb
->ki_filp
;
1988 struct inode
*inode
= file_inode(file
);
1991 if (f2fs_encrypted_inode(inode
) &&
1992 !fscrypt_has_encryption_key(inode
) &&
1993 fscrypt_get_encryption_info(inode
))
1997 ret
= generic_write_checks(iocb
, from
);
1999 ret
= f2fs_preallocate_blocks(iocb
, from
);
2001 ret
= __generic_file_write_iter(iocb
, from
);
2003 inode_unlock(inode
);
2006 ret
= generic_write_sync(iocb
, ret
);
2010 #ifdef CONFIG_COMPAT
2011 long f2fs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
2014 case F2FS_IOC32_GETFLAGS
:
2015 cmd
= F2FS_IOC_GETFLAGS
;
2017 case F2FS_IOC32_SETFLAGS
:
2018 cmd
= F2FS_IOC_SETFLAGS
;
2020 case F2FS_IOC32_GETVERSION
:
2021 cmd
= F2FS_IOC_GETVERSION
;
2023 case F2FS_IOC_START_ATOMIC_WRITE
:
2024 case F2FS_IOC_COMMIT_ATOMIC_WRITE
:
2025 case F2FS_IOC_START_VOLATILE_WRITE
:
2026 case F2FS_IOC_RELEASE_VOLATILE_WRITE
:
2027 case F2FS_IOC_ABORT_VOLATILE_WRITE
:
2028 case F2FS_IOC_SHUTDOWN
:
2029 case F2FS_IOC_SET_ENCRYPTION_POLICY
:
2030 case F2FS_IOC_GET_ENCRYPTION_PWSALT
:
2031 case F2FS_IOC_GET_ENCRYPTION_POLICY
:
2032 case F2FS_IOC_GARBAGE_COLLECT
:
2033 case F2FS_IOC_WRITE_CHECKPOINT
:
2034 case F2FS_IOC_DEFRAGMENT
:
2037 return -ENOIOCTLCMD
;
2039 return f2fs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));
2043 const struct file_operations f2fs_file_operations
= {
2044 .llseek
= f2fs_llseek
,
2045 .read_iter
= generic_file_read_iter
,
2046 .write_iter
= f2fs_file_write_iter
,
2047 .open
= f2fs_file_open
,
2048 .release
= f2fs_release_file
,
2049 .mmap
= f2fs_file_mmap
,
2050 .fsync
= f2fs_sync_file
,
2051 .fallocate
= f2fs_fallocate
,
2052 .unlocked_ioctl
= f2fs_ioctl
,
2053 #ifdef CONFIG_COMPAT
2054 .compat_ioctl
= f2fs_compat_ioctl
,
2056 .splice_read
= generic_file_splice_read
,
2057 .splice_write
= iter_file_splice_write
,