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>
24 #include <linux/file.h>
33 #include <trace/events/f2fs.h>
35 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
38 struct page *page = vmf->page;
39 struct inode *inode = file_inode(vma->vm_file);
40 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
41 struct dnode_of_data dn;
44 sb_start_pagefault(inode->i_sb);
46 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
48 /* block allocation */
50 set_new_dnode(&dn, inode, NULL, NULL, 0);
51 err = f2fs_reserve_block(&dn, page->index);
59 f2fs_balance_fs(sbi, dn.node_changed);
61 file_update_time(vma->vm_file);
63 if (unlikely(page->mapping != inode->i_mapping ||
64 page_offset(page) > i_size_read(inode) ||
65 !PageUptodate(page))) {
72 * check to see if the page is mapped already (no holes)
74 if (PageMappedToDisk(page))
77 /* page is wholly or partially inside EOF */
78 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
81 offset = i_size_read(inode) & ~PAGE_MASK;
82 zero_user_segment(page, offset, PAGE_SIZE);
85 if (!PageUptodate(page))
86 SetPageUptodate(page);
88 trace_f2fs_vm_page_mkwrite(page, DATA);
91 f2fs_wait_on_page_writeback(page, DATA, false);
93 /* wait for GCed encrypted page writeback */
94 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
95 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
97 /* if gced page is attached, don't write to cold segment */
98 clear_cold_data(page);
100 sb_end_pagefault(inode->i_sb);
101 f2fs_update_time(sbi, REQ_TIME);
102 return block_page_mkwrite_return(err);
105 static const struct vm_operations_struct f2fs_file_vm_ops = {
106 .fault = filemap_fault,
107 .map_pages = filemap_map_pages,
108 .page_mkwrite = f2fs_vm_page_mkwrite,
111 static int get_parent_ino(struct inode *inode, nid_t *pino)
113 struct dentry *dentry;
115 inode = igrab(inode);
116 dentry = d_find_any_alias(inode);
121 if (update_dent_inode(inode, inode, &dentry->d_name)) {
126 *pino = parent_ino(dentry);
131 static inline bool need_do_checkpoint(struct inode *inode)
133 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
134 bool need_cp = false;
136 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
138 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
140 else if (file_wrong_pino(inode))
142 else if (!space_for_roll_forward(sbi))
144 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
146 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
148 else if (test_opt(sbi, FASTBOOT))
150 else if (sbi->active_logs == 2)
156 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
158 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
160 /* But we need to avoid that there are some inode updates */
161 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
167 static void try_to_fix_pino(struct inode *inode)
169 struct f2fs_inode_info *fi = F2FS_I(inode);
172 down_write(&fi->i_sem);
174 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
175 get_parent_ino(inode, &pino)) {
176 f2fs_i_pino_write(inode, pino);
177 file_got_pino(inode);
179 up_write(&fi->i_sem);
182 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
183 int datasync, bool atomic)
185 struct inode *inode = file->f_mapping->host;
186 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
187 nid_t ino = inode->i_ino;
189 bool need_cp = false;
190 struct writeback_control wbc = {
191 .sync_mode = WB_SYNC_ALL,
192 .nr_to_write = LONG_MAX,
196 if (unlikely(f2fs_readonly(inode->i_sb)))
199 trace_f2fs_sync_file_enter(inode);
201 if (S_ISDIR(inode->i_mode))
204 /* if fdatasync is triggered, let's do in-place-update */
205 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
206 set_inode_flag(inode, FI_NEED_IPU);
207 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
208 clear_inode_flag(inode, FI_NEED_IPU);
211 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
215 /* if the inode is dirty, let's recover all the time */
216 if (!datasync && !f2fs_skip_inode_update(inode)) {
217 f2fs_write_inode(inode, NULL);
222 * if there is no written data, don't waste time to write recovery info.
224 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
225 !exist_written_data(sbi, ino, APPEND_INO)) {
227 /* it may call write_inode just prior to fsync */
228 if (need_inode_page_update(sbi, ino))
231 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
232 exist_written_data(sbi, ino, UPDATE_INO))
238 * Both of fdatasync() and fsync() are able to be recovered from
241 down_read(&F2FS_I(inode)->i_sem);
242 need_cp = need_do_checkpoint(inode);
243 up_read(&F2FS_I(inode)->i_sem);
246 /* all the dirty node pages should be flushed for POR */
247 ret = f2fs_sync_fs(inode->i_sb, 1);
250 * We've secured consistency through sync_fs. Following pino
251 * will be used only for fsynced inodes after checkpoint.
253 try_to_fix_pino(inode);
254 clear_inode_flag(inode, FI_APPEND_WRITE);
255 clear_inode_flag(inode, FI_UPDATE_WRITE);
259 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
263 /* if cp_error was enabled, we should avoid infinite loop */
264 if (unlikely(f2fs_cp_error(sbi))) {
269 if (need_inode_block_update(sbi, ino)) {
270 f2fs_mark_inode_dirty_sync(inode);
271 f2fs_write_inode(inode, NULL);
275 ret = wait_on_node_pages_writeback(sbi, ino);
279 /* once recovery info is written, don't need to tack this */
280 remove_ino_entry(sbi, ino, APPEND_INO);
281 clear_inode_flag(inode, FI_APPEND_WRITE);
283 remove_ino_entry(sbi, ino, UPDATE_INO);
284 clear_inode_flag(inode, FI_UPDATE_WRITE);
285 ret = f2fs_issue_flush(sbi);
286 f2fs_update_time(sbi, REQ_TIME);
288 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
289 f2fs_trace_ios(NULL, 1);
293 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
295 return f2fs_do_sync_file(file, start, end, datasync, false);
298 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
299 pgoff_t pgofs, int whence)
304 if (whence != SEEK_DATA)
307 /* find first dirty page index */
308 pagevec_init(&pvec, 0);
309 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
310 PAGECACHE_TAG_DIRTY, 1);
311 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
312 pagevec_release(&pvec);
316 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
317 pgoff_t dirty, pgoff_t pgofs, int whence)
321 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
322 is_valid_data_blkaddr(sbi, blkaddr))
326 if (blkaddr == NULL_ADDR)
333 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
335 struct inode *inode = file->f_mapping->host;
336 loff_t maxbytes = inode->i_sb->s_maxbytes;
337 struct dnode_of_data dn;
338 pgoff_t pgofs, end_offset, dirty;
339 loff_t data_ofs = offset;
345 isize = i_size_read(inode);
349 /* handle inline data case */
350 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
351 if (whence == SEEK_HOLE)
356 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
358 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
360 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
361 set_new_dnode(&dn, inode, NULL, NULL, 0);
362 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
363 if (err && err != -ENOENT) {
365 } else if (err == -ENOENT) {
366 /* direct node does not exists */
367 if (whence == SEEK_DATA) {
368 pgofs = get_next_page_offset(&dn, pgofs);
375 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
377 /* find data/hole in dnode block */
378 for (; dn.ofs_in_node < end_offset;
379 dn.ofs_in_node++, pgofs++,
380 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
382 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
384 if (__is_valid_data_blkaddr(blkaddr) &&
385 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
386 blkaddr, DATA_GENERIC)) {
391 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
400 if (whence == SEEK_DATA)
403 if (whence == SEEK_HOLE && data_ofs > isize)
406 return vfs_setpos(file, data_ofs, maxbytes);
412 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
414 struct inode *inode = file->f_mapping->host;
415 loff_t maxbytes = inode->i_sb->s_maxbytes;
421 return generic_file_llseek_size(file, offset, whence,
422 maxbytes, i_size_read(inode));
427 return f2fs_seek_block(file, offset, whence);
433 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
435 struct inode *inode = file_inode(file);
438 if (f2fs_encrypted_inode(inode)) {
439 err = fscrypt_get_encryption_info(inode);
442 if (!f2fs_encrypted_inode(inode))
446 /* we don't need to use inline_data strictly */
447 err = f2fs_convert_inline_inode(inode);
452 vma->vm_ops = &f2fs_file_vm_ops;
456 static int f2fs_file_open(struct inode *inode, struct file *filp)
458 int ret = generic_file_open(inode, filp);
461 if (!ret && f2fs_encrypted_inode(inode)) {
462 ret = fscrypt_get_encryption_info(inode);
465 if (!fscrypt_has_encryption_key(inode))
468 dir = dget_parent(file_dentry(filp));
469 if (f2fs_encrypted_inode(d_inode(dir)) &&
470 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
478 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
480 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
481 struct f2fs_node *raw_node;
482 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
485 raw_node = F2FS_NODE(dn->node_page);
486 addr = blkaddr_in_node(raw_node) + ofs;
488 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
489 block_t blkaddr = le32_to_cpu(*addr);
490 if (blkaddr == NULL_ADDR)
493 dn->data_blkaddr = NULL_ADDR;
494 set_data_blkaddr(dn);
496 if (__is_valid_data_blkaddr(blkaddr) &&
497 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
500 invalidate_blocks(sbi, blkaddr);
501 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
502 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
509 * once we invalidate valid blkaddr in range [ofs, ofs + count],
510 * we will invalidate all blkaddr in the whole range.
512 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
514 f2fs_update_extent_cache_range(dn, fofs, 0, len);
515 dec_valid_block_count(sbi, dn->inode, nr_free);
517 dn->ofs_in_node = ofs;
519 f2fs_update_time(sbi, REQ_TIME);
520 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
521 dn->ofs_in_node, nr_free);
525 void truncate_data_blocks(struct dnode_of_data *dn)
527 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
530 static int truncate_partial_data_page(struct inode *inode, u64 from,
533 unsigned offset = from & (PAGE_SIZE - 1);
534 pgoff_t index = from >> PAGE_SHIFT;
535 struct address_space *mapping = inode->i_mapping;
538 if (!offset && !cache_only)
542 page = find_lock_page(mapping, index);
543 if (page && PageUptodate(page))
545 f2fs_put_page(page, 1);
549 page = get_lock_data_page(inode, index, true);
553 f2fs_wait_on_page_writeback(page, DATA, true);
554 zero_user(page, offset, PAGE_SIZE - offset);
555 if (!cache_only || !f2fs_encrypted_inode(inode) ||
556 !S_ISREG(inode->i_mode))
557 set_page_dirty(page);
558 f2fs_put_page(page, 1);
562 int truncate_blocks(struct inode *inode, u64 from, bool lock)
564 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
565 unsigned int blocksize = inode->i_sb->s_blocksize;
566 struct dnode_of_data dn;
568 int count = 0, err = 0;
570 bool truncate_page = false;
572 trace_f2fs_truncate_blocks_enter(inode, from);
574 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
576 if (free_from >= sbi->max_file_blocks)
582 ipage = get_node_page(sbi, inode->i_ino);
584 err = PTR_ERR(ipage);
588 if (f2fs_has_inline_data(inode)) {
589 if (truncate_inline_inode(ipage, from))
590 set_page_dirty(ipage);
591 f2fs_put_page(ipage, 1);
592 truncate_page = true;
596 set_new_dnode(&dn, inode, ipage, NULL, 0);
597 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
604 count = ADDRS_PER_PAGE(dn.node_page, inode);
606 count -= dn.ofs_in_node;
607 f2fs_bug_on(sbi, count < 0);
609 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
610 truncate_data_blocks_range(&dn, count);
616 err = truncate_inode_blocks(inode, free_from);
621 /* lastly zero out the first data page */
623 err = truncate_partial_data_page(inode, from, truncate_page);
625 trace_f2fs_truncate_blocks_exit(inode, err);
629 int f2fs_truncate(struct inode *inode)
633 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
634 S_ISLNK(inode->i_mode)))
637 trace_f2fs_truncate(inode);
639 /* we should check inline_data size */
640 if (!f2fs_may_inline_data(inode)) {
641 err = f2fs_convert_inline_inode(inode);
646 err = truncate_blocks(inode, i_size_read(inode), true);
650 inode->i_mtime = inode->i_ctime = current_time(inode);
651 f2fs_mark_inode_dirty_sync(inode);
655 int f2fs_getattr(struct vfsmount *mnt,
656 struct dentry *dentry, struct kstat *stat)
658 struct inode *inode = d_inode(dentry);
659 generic_fillattr(inode, stat);
664 #ifdef CONFIG_F2FS_FS_POSIX_ACL
665 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
667 unsigned int ia_valid = attr->ia_valid;
669 if (ia_valid & ATTR_UID)
670 inode->i_uid = attr->ia_uid;
671 if (ia_valid & ATTR_GID)
672 inode->i_gid = attr->ia_gid;
673 if (ia_valid & ATTR_ATIME)
674 inode->i_atime = timespec_trunc(attr->ia_atime,
675 inode->i_sb->s_time_gran);
676 if (ia_valid & ATTR_MTIME)
677 inode->i_mtime = timespec_trunc(attr->ia_mtime,
678 inode->i_sb->s_time_gran);
679 if (ia_valid & ATTR_CTIME)
680 inode->i_ctime = timespec_trunc(attr->ia_ctime,
681 inode->i_sb->s_time_gran);
682 if (ia_valid & ATTR_MODE) {
683 umode_t mode = attr->ia_mode;
685 if (!in_group_p(inode->i_gid) &&
686 !capable_wrt_inode_uidgid(inode, CAP_FSETID))
688 set_acl_inode(inode, mode);
692 #define __setattr_copy setattr_copy
695 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
697 struct inode *inode = d_inode(dentry);
700 err = setattr_prepare(dentry, attr);
704 if (attr->ia_valid & ATTR_SIZE) {
705 if (f2fs_encrypted_inode(inode) &&
706 fscrypt_get_encryption_info(inode))
709 if (attr->ia_size <= i_size_read(inode)) {
710 truncate_setsize(inode, attr->ia_size);
711 err = f2fs_truncate(inode);
714 f2fs_balance_fs(F2FS_I_SB(inode), true);
717 * do not trim all blocks after i_size if target size is
718 * larger than i_size.
720 truncate_setsize(inode, attr->ia_size);
722 /* should convert inline inode here */
723 if (!f2fs_may_inline_data(inode)) {
724 err = f2fs_convert_inline_inode(inode);
728 inode->i_mtime = inode->i_ctime = current_time(inode);
732 __setattr_copy(inode, attr);
734 if (attr->ia_valid & ATTR_MODE) {
735 err = posix_acl_chmod(inode, get_inode_mode(inode));
736 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
737 inode->i_mode = F2FS_I(inode)->i_acl_mode;
738 clear_inode_flag(inode, FI_ACL_MODE);
742 f2fs_mark_inode_dirty_sync(inode);
746 const struct inode_operations f2fs_file_inode_operations = {
747 .getattr = f2fs_getattr,
748 .setattr = f2fs_setattr,
749 .get_acl = f2fs_get_acl,
750 .set_acl = f2fs_set_acl,
751 #ifdef CONFIG_F2FS_FS_XATTR
752 .listxattr = f2fs_listxattr,
754 .fiemap = f2fs_fiemap,
757 static int fill_zero(struct inode *inode, pgoff_t index,
758 loff_t start, loff_t len)
760 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
766 f2fs_balance_fs(sbi, true);
769 page = get_new_data_page(inode, NULL, index, false);
773 return PTR_ERR(page);
775 f2fs_wait_on_page_writeback(page, DATA, true);
776 zero_user(page, start, len);
777 set_page_dirty(page);
778 f2fs_put_page(page, 1);
782 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
786 while (pg_start < pg_end) {
787 struct dnode_of_data dn;
788 pgoff_t end_offset, count;
790 set_new_dnode(&dn, inode, NULL, NULL, 0);
791 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
793 if (err == -ENOENT) {
800 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
801 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
803 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
805 truncate_data_blocks_range(&dn, count);
813 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
815 pgoff_t pg_start, pg_end;
816 loff_t off_start, off_end;
819 ret = f2fs_convert_inline_inode(inode);
823 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
824 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
826 off_start = offset & (PAGE_SIZE - 1);
827 off_end = (offset + len) & (PAGE_SIZE - 1);
829 if (pg_start == pg_end) {
830 ret = fill_zero(inode, pg_start, off_start,
831 off_end - off_start);
836 ret = fill_zero(inode, pg_start++, off_start,
837 PAGE_SIZE - off_start);
842 ret = fill_zero(inode, pg_end, 0, off_end);
847 if (pg_start < pg_end) {
848 struct address_space *mapping = inode->i_mapping;
849 loff_t blk_start, blk_end;
850 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
852 f2fs_balance_fs(sbi, true);
854 blk_start = (loff_t)pg_start << PAGE_SHIFT;
855 blk_end = (loff_t)pg_end << PAGE_SHIFT;
856 truncate_inode_pages_range(mapping, blk_start,
860 ret = truncate_hole(inode, pg_start, pg_end);
868 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
869 int *do_replace, pgoff_t off, pgoff_t len)
871 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
872 struct dnode_of_data dn;
876 set_new_dnode(&dn, inode, NULL, NULL, 0);
877 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
878 if (ret && ret != -ENOENT) {
880 } else if (ret == -ENOENT) {
881 if (dn.max_level == 0)
883 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
889 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
890 dn.ofs_in_node, len);
891 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
892 *blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
893 if (!is_checkpointed_data(sbi, *blkaddr)) {
895 if (test_opt(sbi, LFS)) {
900 /* do not invalidate this block address */
901 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
914 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
915 int *do_replace, pgoff_t off, int len)
917 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
918 struct dnode_of_data dn;
921 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
922 if (*do_replace == 0)
925 set_new_dnode(&dn, inode, NULL, NULL, 0);
926 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
928 dec_valid_block_count(sbi, inode, 1);
929 invalidate_blocks(sbi, *blkaddr);
931 f2fs_update_data_blkaddr(&dn, *blkaddr);
938 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
939 block_t *blkaddr, int *do_replace,
940 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
942 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
947 if (blkaddr[i] == NULL_ADDR && !full) {
952 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
953 struct dnode_of_data dn;
958 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
959 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
963 get_node_info(sbi, dn.nid, &ni);
965 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
966 dn.ofs_in_node, len - i);
968 dn.data_blkaddr = datablock_addr(dn.node_page,
970 truncate_data_blocks_range(&dn, 1);
973 f2fs_i_blocks_write(src_inode,
975 f2fs_i_blocks_write(dst_inode,
977 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
978 blkaddr[i], ni.version, true, false);
984 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
985 if (dst_inode->i_size < new_size)
986 f2fs_i_size_write(dst_inode, new_size);
987 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
991 struct page *psrc, *pdst;
993 psrc = get_lock_data_page(src_inode, src + i, true);
995 return PTR_ERR(psrc);
996 pdst = get_new_data_page(dst_inode, NULL, dst + i,
999 f2fs_put_page(psrc, 1);
1000 return PTR_ERR(pdst);
1002 f2fs_copy_page(psrc, pdst);
1003 set_page_dirty(pdst);
1004 f2fs_put_page(pdst, 1);
1005 f2fs_put_page(psrc, 1);
1007 ret = truncate_hole(src_inode, src + i, src + i + 1);
1016 static int __exchange_data_block(struct inode *src_inode,
1017 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1018 pgoff_t len, bool full)
1020 block_t *src_blkaddr;
1026 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1028 src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1032 do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1034 kvfree(src_blkaddr);
1038 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1039 do_replace, src, olen);
1043 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1044 do_replace, src, dst, olen, full);
1052 kvfree(src_blkaddr);
1058 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1059 kvfree(src_blkaddr);
1064 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1066 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1067 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1070 f2fs_balance_fs(sbi, true);
1073 f2fs_drop_extent_tree(inode);
1075 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1076 f2fs_unlock_op(sbi);
1080 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1082 pgoff_t pg_start, pg_end;
1086 if (offset + len >= i_size_read(inode))
1089 /* collapse range should be aligned to block size of f2fs. */
1090 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1093 ret = f2fs_convert_inline_inode(inode);
1097 pg_start = offset >> PAGE_SHIFT;
1098 pg_end = (offset + len) >> PAGE_SHIFT;
1100 /* write out all dirty pages from offset */
1101 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1105 truncate_pagecache(inode, offset);
1107 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1111 /* write out all moved pages, if possible */
1112 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1113 truncate_pagecache(inode, offset);
1115 new_size = i_size_read(inode) - len;
1116 truncate_pagecache(inode, new_size);
1118 ret = truncate_blocks(inode, new_size, true);
1120 f2fs_i_size_write(inode, new_size);
1125 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1128 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1129 pgoff_t index = start;
1130 unsigned int ofs_in_node = dn->ofs_in_node;
1134 for (; index < end; index++, dn->ofs_in_node++) {
1135 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1139 dn->ofs_in_node = ofs_in_node;
1140 ret = reserve_new_blocks(dn, count);
1144 dn->ofs_in_node = ofs_in_node;
1145 for (index = start; index < end; index++, dn->ofs_in_node++) {
1147 datablock_addr(dn->node_page, dn->ofs_in_node);
1149 * reserve_new_blocks will not guarantee entire block
1152 if (dn->data_blkaddr == NULL_ADDR) {
1156 if (dn->data_blkaddr != NEW_ADDR) {
1157 invalidate_blocks(sbi, dn->data_blkaddr);
1158 dn->data_blkaddr = NEW_ADDR;
1159 set_data_blkaddr(dn);
1163 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1168 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1171 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1172 struct address_space *mapping = inode->i_mapping;
1173 pgoff_t index, pg_start, pg_end;
1174 loff_t new_size = i_size_read(inode);
1175 loff_t off_start, off_end;
1178 ret = inode_newsize_ok(inode, (len + offset));
1182 ret = f2fs_convert_inline_inode(inode);
1186 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1190 truncate_pagecache_range(inode, offset, offset + len - 1);
1192 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1193 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1195 off_start = offset & (PAGE_SIZE - 1);
1196 off_end = (offset + len) & (PAGE_SIZE - 1);
1198 if (pg_start == pg_end) {
1199 ret = fill_zero(inode, pg_start, off_start,
1200 off_end - off_start);
1204 if (offset + len > new_size)
1205 new_size = offset + len;
1206 new_size = max_t(loff_t, new_size, offset + len);
1209 ret = fill_zero(inode, pg_start++, off_start,
1210 PAGE_SIZE - off_start);
1214 new_size = max_t(loff_t, new_size,
1215 (loff_t)pg_start << PAGE_SHIFT);
1218 for (index = pg_start; index < pg_end;) {
1219 struct dnode_of_data dn;
1220 unsigned int end_offset;
1225 set_new_dnode(&dn, inode, NULL, NULL, 0);
1226 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1228 f2fs_unlock_op(sbi);
1232 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1233 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1235 ret = f2fs_do_zero_range(&dn, index, end);
1236 f2fs_put_dnode(&dn);
1237 f2fs_unlock_op(sbi);
1242 new_size = max_t(loff_t, new_size,
1243 (loff_t)index << PAGE_SHIFT);
1247 ret = fill_zero(inode, pg_end, 0, off_end);
1251 new_size = max_t(loff_t, new_size, offset + len);
1256 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1257 f2fs_i_size_write(inode, new_size);
1262 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1264 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1265 pgoff_t nr, pg_start, pg_end, delta, idx;
1269 new_size = i_size_read(inode) + len;
1270 if (new_size > inode->i_sb->s_maxbytes)
1273 if (offset >= i_size_read(inode))
1276 /* insert range should be aligned to block size of f2fs. */
1277 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1280 ret = f2fs_convert_inline_inode(inode);
1284 f2fs_balance_fs(sbi, true);
1286 ret = truncate_blocks(inode, i_size_read(inode), true);
1290 /* write out all dirty pages from offset */
1291 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1295 truncate_pagecache(inode, offset);
1297 pg_start = offset >> PAGE_SHIFT;
1298 pg_end = (offset + len) >> PAGE_SHIFT;
1299 delta = pg_end - pg_start;
1300 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1302 while (!ret && idx > pg_start) {
1303 nr = idx - pg_start;
1309 f2fs_drop_extent_tree(inode);
1311 ret = __exchange_data_block(inode, inode, idx,
1312 idx + delta, nr, false);
1313 f2fs_unlock_op(sbi);
1316 /* write out all moved pages, if possible */
1317 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1318 truncate_pagecache(inode, offset);
1321 f2fs_i_size_write(inode, new_size);
1325 static int expand_inode_data(struct inode *inode, loff_t offset,
1326 loff_t len, int mode)
1328 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1329 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1331 loff_t new_size = i_size_read(inode);
1335 ret = inode_newsize_ok(inode, (len + offset));
1339 ret = f2fs_convert_inline_inode(inode);
1343 f2fs_balance_fs(sbi, true);
1345 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1346 off_end = (offset + len) & (PAGE_SIZE - 1);
1348 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1349 map.m_len = pg_end - map.m_lblk;
1353 ret = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1360 last_off = map.m_lblk + map.m_len - 1;
1362 /* update new size to the failed position */
1363 new_size = (last_off == pg_end) ? offset + len:
1364 (loff_t)(last_off + 1) << PAGE_SHIFT;
1366 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1369 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1370 f2fs_i_size_write(inode, new_size);
1375 static long f2fs_fallocate(struct file *file, int mode,
1376 loff_t offset, loff_t len)
1378 struct inode *inode = file_inode(file);
1381 /* f2fs only support ->fallocate for regular file */
1382 if (!S_ISREG(inode->i_mode))
1385 if (f2fs_encrypted_inode(inode) &&
1386 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1389 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1390 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1391 FALLOC_FL_INSERT_RANGE))
1396 if (mode & FALLOC_FL_PUNCH_HOLE) {
1397 if (offset >= inode->i_size)
1400 ret = punch_hole(inode, offset, len);
1401 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1402 ret = f2fs_collapse_range(inode, offset, len);
1403 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1404 ret = f2fs_zero_range(inode, offset, len, mode);
1405 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1406 ret = f2fs_insert_range(inode, offset, len);
1408 ret = expand_inode_data(inode, offset, len, mode);
1412 inode->i_mtime = inode->i_ctime = current_time(inode);
1413 f2fs_mark_inode_dirty_sync(inode);
1414 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1418 inode_unlock(inode);
1420 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1424 static int f2fs_release_file(struct inode *inode, struct file *filp)
1427 * f2fs_relase_file is called at every close calls. So we should
1428 * not drop any inmemory pages by close called by other process.
1430 if (!(filp->f_mode & FMODE_WRITE) ||
1431 atomic_read(&inode->i_writecount) != 1)
1434 /* some remained atomic pages should discarded */
1435 if (f2fs_is_atomic_file(inode))
1436 drop_inmem_pages(inode);
1437 if (f2fs_is_volatile_file(inode)) {
1438 clear_inode_flag(inode, FI_VOLATILE_FILE);
1439 set_inode_flag(inode, FI_DROP_CACHE);
1440 filemap_fdatawrite(inode->i_mapping);
1441 clear_inode_flag(inode, FI_DROP_CACHE);
1446 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1447 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1449 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1453 else if (S_ISREG(mode))
1454 return flags & F2FS_REG_FLMASK;
1456 return flags & F2FS_OTHER_FLMASK;
1459 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1461 struct inode *inode = file_inode(filp);
1462 struct f2fs_inode_info *fi = F2FS_I(inode);
1463 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1464 return put_user(flags, (int __user *)arg);
1467 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1469 struct inode *inode = file_inode(filp);
1470 struct f2fs_inode_info *fi = F2FS_I(inode);
1472 unsigned int oldflags;
1475 if (!inode_owner_or_capable(inode))
1478 if (get_user(flags, (int __user *)arg))
1481 ret = mnt_want_write_file(filp);
1485 flags = f2fs_mask_flags(inode->i_mode, flags);
1489 oldflags = fi->i_flags;
1491 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1492 if (!capable(CAP_LINUX_IMMUTABLE)) {
1493 inode_unlock(inode);
1499 flags = flags & FS_FL_USER_MODIFIABLE;
1500 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1501 fi->i_flags = flags;
1502 inode_unlock(inode);
1504 inode->i_ctime = current_time(inode);
1505 f2fs_set_inode_flags(inode);
1507 mnt_drop_write_file(filp);
1511 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1513 struct inode *inode = file_inode(filp);
1515 return put_user(inode->i_generation, (int __user *)arg);
1518 static int f2fs_ioc_start_atomic_write(struct file *filp)
1520 struct inode *inode = file_inode(filp);
1523 if (!inode_owner_or_capable(inode))
1526 ret = mnt_want_write_file(filp);
1532 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1534 if (f2fs_is_atomic_file(inode))
1537 ret = f2fs_convert_inline_inode(inode);
1541 set_inode_flag(inode, FI_ATOMIC_FILE);
1542 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1544 if (!get_dirty_pages(inode))
1547 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1548 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1549 inode->i_ino, get_dirty_pages(inode));
1550 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1552 clear_inode_flag(inode, FI_ATOMIC_FILE);
1554 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1555 inode_unlock(inode);
1556 mnt_drop_write_file(filp);
1560 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1562 struct inode *inode = file_inode(filp);
1565 if (!inode_owner_or_capable(inode))
1568 ret = mnt_want_write_file(filp);
1574 if (f2fs_is_volatile_file(inode))
1577 if (f2fs_is_atomic_file(inode)) {
1578 clear_inode_flag(inode, FI_ATOMIC_FILE);
1579 ret = commit_inmem_pages(inode);
1581 set_inode_flag(inode, FI_ATOMIC_FILE);
1586 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1588 inode_unlock(inode);
1589 mnt_drop_write_file(filp);
1593 static int f2fs_ioc_start_volatile_write(struct file *filp)
1595 struct inode *inode = file_inode(filp);
1598 if (!inode_owner_or_capable(inode))
1601 ret = mnt_want_write_file(filp);
1607 if (f2fs_is_volatile_file(inode))
1610 ret = f2fs_convert_inline_inode(inode);
1614 set_inode_flag(inode, FI_VOLATILE_FILE);
1615 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1617 inode_unlock(inode);
1618 mnt_drop_write_file(filp);
1622 static int f2fs_ioc_release_volatile_write(struct file *filp)
1624 struct inode *inode = file_inode(filp);
1627 if (!inode_owner_or_capable(inode))
1630 ret = mnt_want_write_file(filp);
1636 if (!f2fs_is_volatile_file(inode))
1639 if (!f2fs_is_first_block_written(inode)) {
1640 ret = truncate_partial_data_page(inode, 0, true);
1644 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1646 inode_unlock(inode);
1647 mnt_drop_write_file(filp);
1651 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1653 struct inode *inode = file_inode(filp);
1656 if (!inode_owner_or_capable(inode))
1659 ret = mnt_want_write_file(filp);
1665 if (f2fs_is_atomic_file(inode))
1666 drop_inmem_pages(inode);
1667 if (f2fs_is_volatile_file(inode)) {
1668 clear_inode_flag(inode, FI_VOLATILE_FILE);
1669 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1672 inode_unlock(inode);
1674 mnt_drop_write_file(filp);
1675 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1679 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1681 struct inode *inode = file_inode(filp);
1682 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1683 struct super_block *sb = sbi->sb;
1687 if (!capable(CAP_SYS_ADMIN))
1690 if (get_user(in, (__u32 __user *)arg))
1693 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1694 ret = mnt_want_write_file(filp);
1700 case F2FS_GOING_DOWN_FULLSYNC:
1701 sb = freeze_bdev(sb->s_bdev);
1702 if (sb && !IS_ERR(sb)) {
1703 f2fs_stop_checkpoint(sbi, false);
1704 thaw_bdev(sb->s_bdev, sb);
1707 case F2FS_GOING_DOWN_METASYNC:
1708 /* do checkpoint only */
1709 f2fs_sync_fs(sb, 1);
1710 f2fs_stop_checkpoint(sbi, false);
1712 case F2FS_GOING_DOWN_NOSYNC:
1713 f2fs_stop_checkpoint(sbi, false);
1715 case F2FS_GOING_DOWN_METAFLUSH:
1716 sync_meta_pages(sbi, META, LONG_MAX);
1717 f2fs_stop_checkpoint(sbi, false);
1723 f2fs_update_time(sbi, REQ_TIME);
1725 if (in != F2FS_GOING_DOWN_FULLSYNC)
1726 mnt_drop_write_file(filp);
1730 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1732 struct inode *inode = file_inode(filp);
1733 struct super_block *sb = inode->i_sb;
1734 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1735 struct fstrim_range range;
1738 if (!capable(CAP_SYS_ADMIN))
1741 if (!blk_queue_discard(q))
1744 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1748 ret = mnt_want_write_file(filp);
1752 range.minlen = max((unsigned int)range.minlen,
1753 q->limits.discard_granularity);
1754 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1755 mnt_drop_write_file(filp);
1759 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1762 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1766 static bool uuid_is_nonzero(__u8 u[16])
1770 for (i = 0; i < 16; i++)
1776 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1778 struct inode *inode = file_inode(filp);
1780 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1782 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1785 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1787 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1790 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1792 struct inode *inode = file_inode(filp);
1793 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1796 if (!f2fs_sb_has_crypto(inode->i_sb))
1799 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1802 err = mnt_want_write_file(filp);
1806 /* update superblock with uuid */
1807 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1809 err = f2fs_commit_super(sbi, false);
1812 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1813 mnt_drop_write_file(filp);
1816 mnt_drop_write_file(filp);
1818 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1824 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1826 struct inode *inode = file_inode(filp);
1827 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1831 if (!capable(CAP_SYS_ADMIN))
1834 if (get_user(sync, (__u32 __user *)arg))
1837 if (f2fs_readonly(sbi->sb))
1840 ret = mnt_want_write_file(filp);
1845 if (!mutex_trylock(&sbi->gc_mutex)) {
1850 mutex_lock(&sbi->gc_mutex);
1853 ret = f2fs_gc(sbi, sync);
1855 mnt_drop_write_file(filp);
1859 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1861 struct inode *inode = file_inode(filp);
1862 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1865 if (!capable(CAP_SYS_ADMIN))
1868 if (f2fs_readonly(sbi->sb))
1871 ret = mnt_want_write_file(filp);
1875 ret = f2fs_sync_fs(sbi->sb, 1);
1877 mnt_drop_write_file(filp);
1881 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1883 struct f2fs_defragment *range)
1885 struct inode *inode = file_inode(filp);
1886 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1887 struct extent_info ei;
1888 pgoff_t pg_start, pg_end;
1889 unsigned int blk_per_seg = sbi->blocks_per_seg;
1890 unsigned int total = 0, sec_num;
1891 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1892 block_t blk_end = 0;
1893 bool fragmented = false;
1896 /* if in-place-update policy is enabled, don't waste time here */
1897 if (need_inplace_update(inode))
1900 pg_start = range->start >> PAGE_SHIFT;
1901 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1903 f2fs_balance_fs(sbi, true);
1907 /* writeback all dirty pages in the range */
1908 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1909 range->start + range->len - 1);
1914 * lookup mapping info in extent cache, skip defragmenting if physical
1915 * block addresses are continuous.
1917 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1918 if (ei.fofs + ei.len >= pg_end)
1922 map.m_lblk = pg_start;
1925 * lookup mapping info in dnode page cache, skip defragmenting if all
1926 * physical block addresses are continuous even if there are hole(s)
1927 * in logical blocks.
1929 while (map.m_lblk < pg_end) {
1930 map.m_len = pg_end - map.m_lblk;
1931 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1935 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1940 if (blk_end && blk_end != map.m_pblk) {
1944 blk_end = map.m_pblk + map.m_len;
1946 map.m_lblk += map.m_len;
1952 map.m_lblk = pg_start;
1953 map.m_len = pg_end - pg_start;
1955 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1958 * make sure there are enough free section for LFS allocation, this can
1959 * avoid defragment running in SSR mode when free section are allocated
1962 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
1967 while (map.m_lblk < pg_end) {
1972 map.m_len = pg_end - map.m_lblk;
1973 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1977 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1982 set_inode_flag(inode, FI_DO_DEFRAG);
1985 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1988 page = get_lock_data_page(inode, idx, true);
1990 err = PTR_ERR(page);
1994 set_page_dirty(page);
1995 f2fs_put_page(page, 1);
2004 if (idx < pg_end && cnt < blk_per_seg)
2007 clear_inode_flag(inode, FI_DO_DEFRAG);
2009 err = filemap_fdatawrite(inode->i_mapping);
2014 clear_inode_flag(inode, FI_DO_DEFRAG);
2016 inode_unlock(inode);
2018 range->len = (u64)total << PAGE_SHIFT;
2022 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2024 struct inode *inode = file_inode(filp);
2025 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2026 struct f2fs_defragment range;
2029 if (!capable(CAP_SYS_ADMIN))
2032 if (!S_ISREG(inode->i_mode))
2035 err = mnt_want_write_file(filp);
2039 if (f2fs_readonly(sbi->sb)) {
2044 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2050 /* verify alignment of offset & size */
2051 if (range.start & (F2FS_BLKSIZE - 1) ||
2052 range.len & (F2FS_BLKSIZE - 1)) {
2057 err = f2fs_defragment_range(sbi, filp, &range);
2058 f2fs_update_time(sbi, REQ_TIME);
2062 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2066 mnt_drop_write_file(filp);
2070 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2071 struct file *file_out, loff_t pos_out, size_t len)
2073 struct inode *src = file_inode(file_in);
2074 struct inode *dst = file_inode(file_out);
2075 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2076 size_t olen = len, dst_max_i_size = 0;
2080 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2081 src->i_sb != dst->i_sb)
2084 if (unlikely(f2fs_readonly(src->i_sb)))
2087 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2090 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2094 if (pos_in == pos_out)
2096 if (pos_out > pos_in && pos_out < pos_in + len)
2102 if (!inode_trylock(dst)) {
2109 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2112 olen = len = src->i_size - pos_in;
2113 if (pos_in + len == src->i_size)
2114 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2120 dst_osize = dst->i_size;
2121 if (pos_out + olen > dst->i_size)
2122 dst_max_i_size = pos_out + olen;
2124 /* verify the end result is block aligned */
2125 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2126 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2127 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2130 ret = f2fs_convert_inline_inode(src);
2134 ret = f2fs_convert_inline_inode(dst);
2138 /* write out all dirty pages from offset */
2139 ret = filemap_write_and_wait_range(src->i_mapping,
2140 pos_in, pos_in + len);
2144 ret = filemap_write_and_wait_range(dst->i_mapping,
2145 pos_out, pos_out + len);
2149 f2fs_balance_fs(sbi, true);
2151 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2152 pos_out >> F2FS_BLKSIZE_BITS,
2153 len >> F2FS_BLKSIZE_BITS, false);
2157 f2fs_i_size_write(dst, dst_max_i_size);
2158 else if (dst_osize != dst->i_size)
2159 f2fs_i_size_write(dst, dst_osize);
2161 f2fs_unlock_op(sbi);
2170 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2172 struct f2fs_move_range range;
2176 if (!(filp->f_mode & FMODE_READ) ||
2177 !(filp->f_mode & FMODE_WRITE))
2180 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2184 dst = fdget(range.dst_fd);
2188 if (!(dst.file->f_mode & FMODE_WRITE)) {
2193 err = mnt_want_write_file(filp);
2197 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2198 range.pos_out, range.len);
2200 mnt_drop_write_file(filp);
2202 if (copy_to_user((struct f2fs_move_range __user *)arg,
2203 &range, sizeof(range)))
2210 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2213 case F2FS_IOC_GETFLAGS:
2214 return f2fs_ioc_getflags(filp, arg);
2215 case F2FS_IOC_SETFLAGS:
2216 return f2fs_ioc_setflags(filp, arg);
2217 case F2FS_IOC_GETVERSION:
2218 return f2fs_ioc_getversion(filp, arg);
2219 case F2FS_IOC_START_ATOMIC_WRITE:
2220 return f2fs_ioc_start_atomic_write(filp);
2221 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2222 return f2fs_ioc_commit_atomic_write(filp);
2223 case F2FS_IOC_START_VOLATILE_WRITE:
2224 return f2fs_ioc_start_volatile_write(filp);
2225 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2226 return f2fs_ioc_release_volatile_write(filp);
2227 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2228 return f2fs_ioc_abort_volatile_write(filp);
2229 case F2FS_IOC_SHUTDOWN:
2230 return f2fs_ioc_shutdown(filp, arg);
2232 return f2fs_ioc_fitrim(filp, arg);
2233 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2234 return f2fs_ioc_set_encryption_policy(filp, arg);
2235 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2236 return f2fs_ioc_get_encryption_policy(filp, arg);
2237 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2238 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2239 case F2FS_IOC_GARBAGE_COLLECT:
2240 return f2fs_ioc_gc(filp, arg);
2241 case F2FS_IOC_WRITE_CHECKPOINT:
2242 return f2fs_ioc_write_checkpoint(filp, arg);
2243 case F2FS_IOC_DEFRAGMENT:
2244 return f2fs_ioc_defragment(filp, arg);
2245 case F2FS_IOC_MOVE_RANGE:
2246 return f2fs_ioc_move_range(filp, arg);
2252 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2254 struct file *file = iocb->ki_filp;
2255 struct inode *inode = file_inode(file);
2256 struct blk_plug plug;
2259 if (f2fs_encrypted_inode(inode) &&
2260 !fscrypt_has_encryption_key(inode) &&
2261 fscrypt_get_encryption_info(inode))
2265 ret = generic_write_checks(iocb, from);
2267 ret = f2fs_preallocate_blocks(iocb, from);
2269 blk_start_plug(&plug);
2270 ret = __generic_file_write_iter(iocb, from);
2271 blk_finish_plug(&plug);
2274 inode_unlock(inode);
2277 ret = generic_write_sync(iocb, ret);
2281 #ifdef CONFIG_COMPAT
2282 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2285 case F2FS_IOC32_GETFLAGS:
2286 cmd = F2FS_IOC_GETFLAGS;
2288 case F2FS_IOC32_SETFLAGS:
2289 cmd = F2FS_IOC_SETFLAGS;
2291 case F2FS_IOC32_GETVERSION:
2292 cmd = F2FS_IOC_GETVERSION;
2294 case F2FS_IOC_START_ATOMIC_WRITE:
2295 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2296 case F2FS_IOC_START_VOLATILE_WRITE:
2297 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2298 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2299 case F2FS_IOC_SHUTDOWN:
2300 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2301 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2302 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2303 case F2FS_IOC_GARBAGE_COLLECT:
2304 case F2FS_IOC_WRITE_CHECKPOINT:
2305 case F2FS_IOC_DEFRAGMENT:
2307 case F2FS_IOC_MOVE_RANGE:
2310 return -ENOIOCTLCMD;
2312 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2316 const struct file_operations f2fs_file_operations = {
2317 .llseek = f2fs_llseek,
2318 .read_iter = generic_file_read_iter,
2319 .write_iter = f2fs_file_write_iter,
2320 .open = f2fs_file_open,
2321 .release = f2fs_release_file,
2322 .mmap = f2fs_file_mmap,
2323 .fsync = f2fs_sync_file,
2324 .fallocate = f2fs_fallocate,
2325 .unlocked_ioctl = f2fs_ioctl,
2326 #ifdef CONFIG_COMPAT
2327 .compat_ioctl = f2fs_compat_ioctl,
2329 .splice_read = generic_file_splice_read,
2330 .splice_write = iter_file_splice_write,