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/uio.h>
24 #include <linux/uuid.h>
25 #include <linux/file.h>
34 #include <trace/events/f2fs.h>
36 static int f2fs_vm_page_mkwrite(struct vm_fault *vmf)
38 struct page *page = vmf->page;
39 struct inode *inode = file_inode(vmf->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(vmf->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);
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 if (file_enc_name(inode))
116 inode = igrab(inode);
117 dentry = d_find_any_alias(inode);
122 if (update_dent_inode(inode, inode, &dentry->d_name)) {
127 *pino = parent_ino(dentry);
132 static inline bool need_do_checkpoint(struct inode *inode)
134 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
135 bool need_cp = false;
137 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
139 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
141 else if (file_wrong_pino(inode))
143 else if (!space_for_roll_forward(sbi))
145 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
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);
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 */
211 if (!f2fs_skip_inode_update(inode, datasync)) {
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, inode, &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 f2fs_mark_inode_dirty_sync(inode, true);
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);
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 = find_lock_page(mapping, index);
526 if (page && PageUptodate(page))
528 f2fs_put_page(page, 1);
532 page = get_lock_data_page(inode, index, true);
534 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
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 truncate_inline_inode(ipage, from);
574 clear_inode_flag(inode, FI_DATA_EXIST);
575 f2fs_put_page(ipage, 1);
576 truncate_page = true;
580 set_new_dnode(&dn, inode, ipage, NULL, 0);
581 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
588 count = ADDRS_PER_PAGE(dn.node_page, inode);
590 count -= dn.ofs_in_node;
591 f2fs_bug_on(sbi, count < 0);
593 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
594 truncate_data_blocks_range(&dn, count);
600 err = truncate_inode_blocks(inode, free_from);
605 /* lastly zero out the first data page */
607 err = truncate_partial_data_page(inode, from, truncate_page);
609 trace_f2fs_truncate_blocks_exit(inode, err);
613 int f2fs_truncate(struct inode *inode)
617 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
618 S_ISLNK(inode->i_mode)))
621 trace_f2fs_truncate(inode);
623 #ifdef CONFIG_F2FS_FAULT_INJECTION
624 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
625 f2fs_show_injection_info(FAULT_TRUNCATE);
629 /* we should check inline_data size */
630 if (!f2fs_may_inline_data(inode)) {
631 err = f2fs_convert_inline_inode(inode);
636 err = truncate_blocks(inode, i_size_read(inode), true);
640 inode->i_mtime = inode->i_ctime = current_time(inode);
641 f2fs_mark_inode_dirty_sync(inode, false);
645 int f2fs_getattr(const struct path *path, struct kstat *stat,
646 u32 request_mask, unsigned int flags)
648 struct inode *inode = d_inode(path->dentry);
649 generic_fillattr(inode, stat);
654 #ifdef CONFIG_F2FS_FS_POSIX_ACL
655 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
657 unsigned int ia_valid = attr->ia_valid;
659 if (ia_valid & ATTR_UID)
660 inode->i_uid = attr->ia_uid;
661 if (ia_valid & ATTR_GID)
662 inode->i_gid = attr->ia_gid;
663 if (ia_valid & ATTR_ATIME)
664 inode->i_atime = timespec_trunc(attr->ia_atime,
665 inode->i_sb->s_time_gran);
666 if (ia_valid & ATTR_MTIME)
667 inode->i_mtime = timespec_trunc(attr->ia_mtime,
668 inode->i_sb->s_time_gran);
669 if (ia_valid & ATTR_CTIME)
670 inode->i_ctime = timespec_trunc(attr->ia_ctime,
671 inode->i_sb->s_time_gran);
672 if (ia_valid & ATTR_MODE) {
673 umode_t mode = attr->ia_mode;
675 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
677 set_acl_inode(inode, mode);
681 #define __setattr_copy setattr_copy
684 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
686 struct inode *inode = d_inode(dentry);
688 bool size_changed = false;
690 err = setattr_prepare(dentry, attr);
694 if (attr->ia_valid & ATTR_SIZE) {
695 if (f2fs_encrypted_inode(inode) &&
696 fscrypt_get_encryption_info(inode))
699 if (attr->ia_size <= i_size_read(inode)) {
700 truncate_setsize(inode, attr->ia_size);
701 err = f2fs_truncate(inode);
706 * do not trim all blocks after i_size if target size is
707 * larger than i_size.
709 truncate_setsize(inode, attr->ia_size);
711 /* should convert inline inode here */
712 if (!f2fs_may_inline_data(inode)) {
713 err = f2fs_convert_inline_inode(inode);
717 inode->i_mtime = inode->i_ctime = current_time(inode);
723 __setattr_copy(inode, attr);
725 if (attr->ia_valid & ATTR_MODE) {
726 err = posix_acl_chmod(inode, get_inode_mode(inode));
727 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
728 inode->i_mode = F2FS_I(inode)->i_acl_mode;
729 clear_inode_flag(inode, FI_ACL_MODE);
733 /* file size may changed here */
734 f2fs_mark_inode_dirty_sync(inode, size_changed);
736 /* inode change will produce dirty node pages flushed by checkpoint */
737 f2fs_balance_fs(F2FS_I_SB(inode), true);
742 const struct inode_operations f2fs_file_inode_operations = {
743 .getattr = f2fs_getattr,
744 .setattr = f2fs_setattr,
745 .get_acl = f2fs_get_acl,
746 .set_acl = f2fs_set_acl,
747 #ifdef CONFIG_F2FS_FS_XATTR
748 .listxattr = f2fs_listxattr,
750 .fiemap = f2fs_fiemap,
753 static int fill_zero(struct inode *inode, pgoff_t index,
754 loff_t start, loff_t len)
756 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
762 f2fs_balance_fs(sbi, true);
765 page = get_new_data_page(inode, NULL, index, false);
769 return PTR_ERR(page);
771 f2fs_wait_on_page_writeback(page, DATA, true);
772 zero_user(page, start, len);
773 set_page_dirty(page);
774 f2fs_put_page(page, 1);
778 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
782 while (pg_start < pg_end) {
783 struct dnode_of_data dn;
784 pgoff_t end_offset, count;
786 set_new_dnode(&dn, inode, NULL, NULL, 0);
787 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
789 if (err == -ENOENT) {
796 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
797 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
799 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
801 truncate_data_blocks_range(&dn, count);
809 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
811 pgoff_t pg_start, pg_end;
812 loff_t off_start, off_end;
815 ret = f2fs_convert_inline_inode(inode);
819 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
820 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
822 off_start = offset & (PAGE_SIZE - 1);
823 off_end = (offset + len) & (PAGE_SIZE - 1);
825 if (pg_start == pg_end) {
826 ret = fill_zero(inode, pg_start, off_start,
827 off_end - off_start);
832 ret = fill_zero(inode, pg_start++, off_start,
833 PAGE_SIZE - off_start);
838 ret = fill_zero(inode, pg_end, 0, off_end);
843 if (pg_start < pg_end) {
844 struct address_space *mapping = inode->i_mapping;
845 loff_t blk_start, blk_end;
846 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
848 f2fs_balance_fs(sbi, true);
850 blk_start = (loff_t)pg_start << PAGE_SHIFT;
851 blk_end = (loff_t)pg_end << PAGE_SHIFT;
852 truncate_inode_pages_range(mapping, blk_start,
856 ret = truncate_hole(inode, pg_start, pg_end);
864 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
865 int *do_replace, pgoff_t off, pgoff_t len)
867 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
868 struct dnode_of_data dn;
872 set_new_dnode(&dn, inode, NULL, NULL, 0);
873 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
874 if (ret && ret != -ENOENT) {
876 } else if (ret == -ENOENT) {
877 if (dn.max_level == 0)
879 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
885 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
886 dn.ofs_in_node, len);
887 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
888 *blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
889 if (!is_checkpointed_data(sbi, *blkaddr)) {
891 if (test_opt(sbi, LFS)) {
896 /* do not invalidate this block address */
897 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
910 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
911 int *do_replace, pgoff_t off, int len)
913 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
914 struct dnode_of_data dn;
917 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
918 if (*do_replace == 0)
921 set_new_dnode(&dn, inode, NULL, NULL, 0);
922 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
924 dec_valid_block_count(sbi, inode, 1);
925 invalidate_blocks(sbi, *blkaddr);
927 f2fs_update_data_blkaddr(&dn, *blkaddr);
934 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
935 block_t *blkaddr, int *do_replace,
936 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
938 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
943 if (blkaddr[i] == NULL_ADDR && !full) {
948 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
949 struct dnode_of_data dn;
954 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
955 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
959 get_node_info(sbi, dn.nid, &ni);
961 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
962 dn.ofs_in_node, len - i);
964 dn.data_blkaddr = datablock_addr(dn.node_page,
966 truncate_data_blocks_range(&dn, 1);
969 f2fs_i_blocks_write(src_inode,
971 f2fs_i_blocks_write(dst_inode,
973 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
974 blkaddr[i], ni.version, true, false);
980 new_size = (dst + i) << PAGE_SHIFT;
981 if (dst_inode->i_size < new_size)
982 f2fs_i_size_write(dst_inode, new_size);
983 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
987 struct page *psrc, *pdst;
989 psrc = get_lock_data_page(src_inode, src + i, true);
991 return PTR_ERR(psrc);
992 pdst = get_new_data_page(dst_inode, NULL, dst + i,
995 f2fs_put_page(psrc, 1);
996 return PTR_ERR(pdst);
998 f2fs_copy_page(psrc, pdst);
999 set_page_dirty(pdst);
1000 f2fs_put_page(pdst, 1);
1001 f2fs_put_page(psrc, 1);
1003 ret = truncate_hole(src_inode, src + i, src + i + 1);
1012 static int __exchange_data_block(struct inode *src_inode,
1013 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1014 pgoff_t len, bool full)
1016 block_t *src_blkaddr;
1022 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1024 src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1028 do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1030 kvfree(src_blkaddr);
1034 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1035 do_replace, src, olen);
1039 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1040 do_replace, src, dst, olen, full);
1048 kvfree(src_blkaddr);
1054 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1055 kvfree(src_blkaddr);
1060 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1062 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1063 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1066 f2fs_balance_fs(sbi, true);
1069 f2fs_drop_extent_tree(inode);
1071 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1072 f2fs_unlock_op(sbi);
1076 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1078 pgoff_t pg_start, pg_end;
1082 if (offset + len >= i_size_read(inode))
1085 /* collapse range should be aligned to block size of f2fs. */
1086 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1089 ret = f2fs_convert_inline_inode(inode);
1093 pg_start = offset >> PAGE_SHIFT;
1094 pg_end = (offset + len) >> PAGE_SHIFT;
1096 /* write out all dirty pages from offset */
1097 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1101 truncate_pagecache(inode, offset);
1103 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1107 /* write out all moved pages, if possible */
1108 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1109 truncate_pagecache(inode, offset);
1111 new_size = i_size_read(inode) - len;
1112 truncate_pagecache(inode, new_size);
1114 ret = truncate_blocks(inode, new_size, true);
1116 f2fs_i_size_write(inode, new_size);
1121 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1124 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1125 pgoff_t index = start;
1126 unsigned int ofs_in_node = dn->ofs_in_node;
1130 for (; index < end; index++, dn->ofs_in_node++) {
1131 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1135 dn->ofs_in_node = ofs_in_node;
1136 ret = reserve_new_blocks(dn, count);
1140 dn->ofs_in_node = ofs_in_node;
1141 for (index = start; index < end; index++, dn->ofs_in_node++) {
1143 datablock_addr(dn->node_page, dn->ofs_in_node);
1145 * reserve_new_blocks will not guarantee entire block
1148 if (dn->data_blkaddr == NULL_ADDR) {
1152 if (dn->data_blkaddr != NEW_ADDR) {
1153 invalidate_blocks(sbi, dn->data_blkaddr);
1154 dn->data_blkaddr = NEW_ADDR;
1155 set_data_blkaddr(dn);
1159 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1164 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1167 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1168 struct address_space *mapping = inode->i_mapping;
1169 pgoff_t index, pg_start, pg_end;
1170 loff_t new_size = i_size_read(inode);
1171 loff_t off_start, off_end;
1174 ret = inode_newsize_ok(inode, (len + offset));
1178 ret = f2fs_convert_inline_inode(inode);
1182 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1186 truncate_pagecache_range(inode, offset, offset + len - 1);
1188 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1189 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1191 off_start = offset & (PAGE_SIZE - 1);
1192 off_end = (offset + len) & (PAGE_SIZE - 1);
1194 if (pg_start == pg_end) {
1195 ret = fill_zero(inode, pg_start, off_start,
1196 off_end - off_start);
1200 if (offset + len > new_size)
1201 new_size = offset + len;
1202 new_size = max_t(loff_t, new_size, offset + len);
1205 ret = fill_zero(inode, pg_start++, off_start,
1206 PAGE_SIZE - off_start);
1210 new_size = max_t(loff_t, new_size,
1211 (loff_t)pg_start << PAGE_SHIFT);
1214 for (index = pg_start; index < pg_end;) {
1215 struct dnode_of_data dn;
1216 unsigned int end_offset;
1221 set_new_dnode(&dn, inode, NULL, NULL, 0);
1222 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1224 f2fs_unlock_op(sbi);
1228 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1229 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1231 ret = f2fs_do_zero_range(&dn, index, end);
1232 f2fs_put_dnode(&dn);
1233 f2fs_unlock_op(sbi);
1235 f2fs_balance_fs(sbi, dn.node_changed);
1241 new_size = max_t(loff_t, new_size,
1242 (loff_t)index << PAGE_SHIFT);
1246 ret = fill_zero(inode, pg_end, 0, off_end);
1250 new_size = max_t(loff_t, new_size, offset + len);
1255 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1256 f2fs_i_size_write(inode, new_size);
1261 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1263 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1264 pgoff_t nr, pg_start, pg_end, delta, idx;
1268 new_size = i_size_read(inode) + len;
1269 if (new_size > inode->i_sb->s_maxbytes)
1272 if (offset >= i_size_read(inode))
1275 /* insert range should be aligned to block size of f2fs. */
1276 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1279 ret = f2fs_convert_inline_inode(inode);
1283 f2fs_balance_fs(sbi, true);
1285 ret = truncate_blocks(inode, i_size_read(inode), true);
1289 /* write out all dirty pages from offset */
1290 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1294 truncate_pagecache(inode, offset);
1296 pg_start = offset >> PAGE_SHIFT;
1297 pg_end = (offset + len) >> PAGE_SHIFT;
1298 delta = pg_end - pg_start;
1299 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1301 while (!ret && idx > pg_start) {
1302 nr = idx - pg_start;
1308 f2fs_drop_extent_tree(inode);
1310 ret = __exchange_data_block(inode, inode, idx,
1311 idx + delta, nr, false);
1312 f2fs_unlock_op(sbi);
1315 /* write out all moved pages, if possible */
1316 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1317 truncate_pagecache(inode, offset);
1320 f2fs_i_size_write(inode, new_size);
1324 static int expand_inode_data(struct inode *inode, loff_t offset,
1325 loff_t len, int mode)
1327 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1328 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1330 loff_t new_size = i_size_read(inode);
1334 err = inode_newsize_ok(inode, (len + offset));
1338 err = f2fs_convert_inline_inode(inode);
1342 f2fs_balance_fs(sbi, true);
1344 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1345 off_end = (offset + len) & (PAGE_SIZE - 1);
1347 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1348 map.m_len = pg_end - map.m_lblk;
1352 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1359 last_off = map.m_lblk + map.m_len - 1;
1361 /* update new size to the failed position */
1362 new_size = (last_off == pg_end) ? offset + len:
1363 (loff_t)(last_off + 1) << PAGE_SHIFT;
1365 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1368 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1369 f2fs_i_size_write(inode, new_size);
1374 static long f2fs_fallocate(struct file *file, int mode,
1375 loff_t offset, loff_t len)
1377 struct inode *inode = file_inode(file);
1380 /* f2fs only support ->fallocate for regular file */
1381 if (!S_ISREG(inode->i_mode))
1384 if (f2fs_encrypted_inode(inode) &&
1385 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1388 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1389 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1390 FALLOC_FL_INSERT_RANGE))
1395 if (mode & FALLOC_FL_PUNCH_HOLE) {
1396 if (offset >= inode->i_size)
1399 ret = punch_hole(inode, offset, len);
1400 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1401 ret = f2fs_collapse_range(inode, offset, len);
1402 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1403 ret = f2fs_zero_range(inode, offset, len, mode);
1404 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1405 ret = f2fs_insert_range(inode, offset, len);
1407 ret = expand_inode_data(inode, offset, len, mode);
1411 inode->i_mtime = inode->i_ctime = current_time(inode);
1412 f2fs_mark_inode_dirty_sync(inode, false);
1413 if (mode & FALLOC_FL_KEEP_SIZE)
1414 file_set_keep_isize(inode);
1415 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1419 inode_unlock(inode);
1421 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1425 static int f2fs_release_file(struct inode *inode, struct file *filp)
1428 * f2fs_relase_file is called at every close calls. So we should
1429 * not drop any inmemory pages by close called by other process.
1431 if (!(filp->f_mode & FMODE_WRITE) ||
1432 atomic_read(&inode->i_writecount) != 1)
1435 /* some remained atomic pages should discarded */
1436 if (f2fs_is_atomic_file(inode))
1437 drop_inmem_pages(inode);
1438 if (f2fs_is_volatile_file(inode)) {
1439 clear_inode_flag(inode, FI_VOLATILE_FILE);
1440 set_inode_flag(inode, FI_DROP_CACHE);
1441 filemap_fdatawrite(inode->i_mapping);
1442 clear_inode_flag(inode, FI_DROP_CACHE);
1447 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1448 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1450 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1454 else if (S_ISREG(mode))
1455 return flags & F2FS_REG_FLMASK;
1457 return flags & F2FS_OTHER_FLMASK;
1460 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1462 struct inode *inode = file_inode(filp);
1463 struct f2fs_inode_info *fi = F2FS_I(inode);
1464 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1465 return put_user(flags, (int __user *)arg);
1468 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1470 struct inode *inode = file_inode(filp);
1471 struct f2fs_inode_info *fi = F2FS_I(inode);
1473 unsigned int oldflags;
1476 if (!inode_owner_or_capable(inode))
1479 if (get_user(flags, (int __user *)arg))
1482 ret = mnt_want_write_file(filp);
1486 flags = f2fs_mask_flags(inode->i_mode, flags);
1490 oldflags = fi->i_flags;
1492 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1493 if (!capable(CAP_LINUX_IMMUTABLE)) {
1494 inode_unlock(inode);
1500 flags = flags & FS_FL_USER_MODIFIABLE;
1501 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1502 fi->i_flags = flags;
1503 inode_unlock(inode);
1505 inode->i_ctime = current_time(inode);
1506 f2fs_set_inode_flags(inode);
1508 mnt_drop_write_file(filp);
1512 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1514 struct inode *inode = file_inode(filp);
1516 return put_user(inode->i_generation, (int __user *)arg);
1519 static int f2fs_ioc_start_atomic_write(struct file *filp)
1521 struct inode *inode = file_inode(filp);
1524 if (!inode_owner_or_capable(inode))
1527 if (!S_ISREG(inode->i_mode))
1530 ret = mnt_want_write_file(filp);
1536 if (f2fs_is_atomic_file(inode))
1539 ret = f2fs_convert_inline_inode(inode);
1543 set_inode_flag(inode, FI_ATOMIC_FILE);
1544 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1546 if (!get_dirty_pages(inode))
1549 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1550 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1551 inode->i_ino, get_dirty_pages(inode));
1552 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1554 clear_inode_flag(inode, FI_ATOMIC_FILE);
1556 stat_inc_atomic_write(inode);
1557 stat_update_max_atomic_write(inode);
1558 inode_unlock(inode);
1559 mnt_drop_write_file(filp);
1563 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1565 struct inode *inode = file_inode(filp);
1568 if (!inode_owner_or_capable(inode))
1571 ret = mnt_want_write_file(filp);
1577 if (f2fs_is_volatile_file(inode))
1580 if (f2fs_is_atomic_file(inode)) {
1581 ret = commit_inmem_pages(inode);
1585 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1587 clear_inode_flag(inode, FI_ATOMIC_FILE);
1588 stat_dec_atomic_write(inode);
1591 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1594 inode_unlock(inode);
1595 mnt_drop_write_file(filp);
1599 static int f2fs_ioc_start_volatile_write(struct file *filp)
1601 struct inode *inode = file_inode(filp);
1604 if (!inode_owner_or_capable(inode))
1607 if (!S_ISREG(inode->i_mode))
1610 ret = mnt_want_write_file(filp);
1616 if (f2fs_is_volatile_file(inode))
1619 ret = f2fs_convert_inline_inode(inode);
1623 set_inode_flag(inode, FI_VOLATILE_FILE);
1624 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1626 inode_unlock(inode);
1627 mnt_drop_write_file(filp);
1631 static int f2fs_ioc_release_volatile_write(struct file *filp)
1633 struct inode *inode = file_inode(filp);
1636 if (!inode_owner_or_capable(inode))
1639 ret = mnt_want_write_file(filp);
1645 if (!f2fs_is_volatile_file(inode))
1648 if (!f2fs_is_first_block_written(inode)) {
1649 ret = truncate_partial_data_page(inode, 0, true);
1653 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1655 inode_unlock(inode);
1656 mnt_drop_write_file(filp);
1660 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1662 struct inode *inode = file_inode(filp);
1665 if (!inode_owner_or_capable(inode))
1668 ret = mnt_want_write_file(filp);
1674 if (f2fs_is_atomic_file(inode))
1675 drop_inmem_pages(inode);
1676 if (f2fs_is_volatile_file(inode)) {
1677 clear_inode_flag(inode, FI_VOLATILE_FILE);
1678 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1681 inode_unlock(inode);
1683 mnt_drop_write_file(filp);
1684 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1688 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1690 struct inode *inode = file_inode(filp);
1691 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1692 struct super_block *sb = sbi->sb;
1696 if (!capable(CAP_SYS_ADMIN))
1699 if (get_user(in, (__u32 __user *)arg))
1702 ret = mnt_want_write_file(filp);
1707 case F2FS_GOING_DOWN_FULLSYNC:
1708 sb = freeze_bdev(sb->s_bdev);
1709 if (sb && !IS_ERR(sb)) {
1710 f2fs_stop_checkpoint(sbi, false);
1711 thaw_bdev(sb->s_bdev, sb);
1714 case F2FS_GOING_DOWN_METASYNC:
1715 /* do checkpoint only */
1716 f2fs_sync_fs(sb, 1);
1717 f2fs_stop_checkpoint(sbi, false);
1719 case F2FS_GOING_DOWN_NOSYNC:
1720 f2fs_stop_checkpoint(sbi, false);
1722 case F2FS_GOING_DOWN_METAFLUSH:
1723 sync_meta_pages(sbi, META, LONG_MAX);
1724 f2fs_stop_checkpoint(sbi, false);
1730 f2fs_update_time(sbi, REQ_TIME);
1732 mnt_drop_write_file(filp);
1736 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1738 struct inode *inode = file_inode(filp);
1739 struct super_block *sb = inode->i_sb;
1740 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1741 struct fstrim_range range;
1744 if (!capable(CAP_SYS_ADMIN))
1747 if (!blk_queue_discard(q))
1750 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1754 ret = mnt_want_write_file(filp);
1758 range.minlen = max((unsigned int)range.minlen,
1759 q->limits.discard_granularity);
1760 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1761 mnt_drop_write_file(filp);
1765 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1768 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1772 static bool uuid_is_nonzero(__u8 u[16])
1776 for (i = 0; i < 16; i++)
1782 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1784 struct inode *inode = file_inode(filp);
1786 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1788 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1791 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1793 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1796 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1798 struct inode *inode = file_inode(filp);
1799 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1802 if (!f2fs_sb_has_crypto(inode->i_sb))
1805 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1808 err = mnt_want_write_file(filp);
1812 /* update superblock with uuid */
1813 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1815 err = f2fs_commit_super(sbi, false);
1818 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1819 mnt_drop_write_file(filp);
1822 mnt_drop_write_file(filp);
1824 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1830 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1832 struct inode *inode = file_inode(filp);
1833 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1837 if (!capable(CAP_SYS_ADMIN))
1840 if (get_user(sync, (__u32 __user *)arg))
1843 if (f2fs_readonly(sbi->sb))
1846 ret = mnt_want_write_file(filp);
1851 if (!mutex_trylock(&sbi->gc_mutex)) {
1856 mutex_lock(&sbi->gc_mutex);
1859 ret = f2fs_gc(sbi, sync, true);
1861 mnt_drop_write_file(filp);
1865 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1867 struct inode *inode = file_inode(filp);
1868 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1871 if (!capable(CAP_SYS_ADMIN))
1874 if (f2fs_readonly(sbi->sb))
1877 ret = mnt_want_write_file(filp);
1881 ret = f2fs_sync_fs(sbi->sb, 1);
1883 mnt_drop_write_file(filp);
1887 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1889 struct f2fs_defragment *range)
1891 struct inode *inode = file_inode(filp);
1892 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1893 struct extent_info ei = {0,0,0};
1894 pgoff_t pg_start, pg_end;
1895 unsigned int blk_per_seg = sbi->blocks_per_seg;
1896 unsigned int total = 0, sec_num;
1897 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1898 block_t blk_end = 0;
1899 bool fragmented = false;
1902 /* if in-place-update policy is enabled, don't waste time here */
1903 if (need_inplace_update(inode))
1906 pg_start = range->start >> PAGE_SHIFT;
1907 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1909 f2fs_balance_fs(sbi, true);
1913 /* writeback all dirty pages in the range */
1914 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1915 range->start + range->len - 1);
1920 * lookup mapping info in extent cache, skip defragmenting if physical
1921 * block addresses are continuous.
1923 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1924 if (ei.fofs + ei.len >= pg_end)
1928 map.m_lblk = pg_start;
1931 * lookup mapping info in dnode page cache, skip defragmenting if all
1932 * physical block addresses are continuous even if there are hole(s)
1933 * in logical blocks.
1935 while (map.m_lblk < pg_end) {
1936 map.m_len = pg_end - map.m_lblk;
1937 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1941 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1946 if (blk_end && blk_end != map.m_pblk) {
1950 blk_end = map.m_pblk + map.m_len;
1952 map.m_lblk += map.m_len;
1958 map.m_lblk = pg_start;
1959 map.m_len = pg_end - pg_start;
1961 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1964 * make sure there are enough free section for LFS allocation, this can
1965 * avoid defragment running in SSR mode when free section are allocated
1968 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
1973 while (map.m_lblk < pg_end) {
1978 map.m_len = pg_end - map.m_lblk;
1979 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1983 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1988 set_inode_flag(inode, FI_DO_DEFRAG);
1991 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1994 page = get_lock_data_page(inode, idx, true);
1996 err = PTR_ERR(page);
2000 set_page_dirty(page);
2001 f2fs_put_page(page, 1);
2010 if (idx < pg_end && cnt < blk_per_seg)
2013 clear_inode_flag(inode, FI_DO_DEFRAG);
2015 err = filemap_fdatawrite(inode->i_mapping);
2020 clear_inode_flag(inode, FI_DO_DEFRAG);
2022 inode_unlock(inode);
2024 range->len = (u64)total << PAGE_SHIFT;
2028 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2030 struct inode *inode = file_inode(filp);
2031 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2032 struct f2fs_defragment range;
2035 if (!capable(CAP_SYS_ADMIN))
2038 if (!S_ISREG(inode->i_mode))
2041 err = mnt_want_write_file(filp);
2045 if (f2fs_readonly(sbi->sb)) {
2050 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2056 /* verify alignment of offset & size */
2057 if (range.start & (F2FS_BLKSIZE - 1) ||
2058 range.len & (F2FS_BLKSIZE - 1)) {
2063 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2064 sbi->max_file_blocks)) {
2069 err = f2fs_defragment_range(sbi, filp, &range);
2070 f2fs_update_time(sbi, REQ_TIME);
2074 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2078 mnt_drop_write_file(filp);
2082 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2083 struct file *file_out, loff_t pos_out, size_t len)
2085 struct inode *src = file_inode(file_in);
2086 struct inode *dst = file_inode(file_out);
2087 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2088 size_t olen = len, dst_max_i_size = 0;
2092 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2093 src->i_sb != dst->i_sb)
2096 if (unlikely(f2fs_readonly(src->i_sb)))
2099 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2102 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2106 if (pos_in == pos_out)
2108 if (pos_out > pos_in && pos_out < pos_in + len)
2114 if (!inode_trylock(dst)) {
2121 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2124 olen = len = src->i_size - pos_in;
2125 if (pos_in + len == src->i_size)
2126 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2132 dst_osize = dst->i_size;
2133 if (pos_out + olen > dst->i_size)
2134 dst_max_i_size = pos_out + olen;
2136 /* verify the end result is block aligned */
2137 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2138 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2139 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2142 ret = f2fs_convert_inline_inode(src);
2146 ret = f2fs_convert_inline_inode(dst);
2150 /* write out all dirty pages from offset */
2151 ret = filemap_write_and_wait_range(src->i_mapping,
2152 pos_in, pos_in + len);
2156 ret = filemap_write_and_wait_range(dst->i_mapping,
2157 pos_out, pos_out + len);
2161 f2fs_balance_fs(sbi, true);
2163 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2164 pos_out >> F2FS_BLKSIZE_BITS,
2165 len >> F2FS_BLKSIZE_BITS, false);
2169 f2fs_i_size_write(dst, dst_max_i_size);
2170 else if (dst_osize != dst->i_size)
2171 f2fs_i_size_write(dst, dst_osize);
2173 f2fs_unlock_op(sbi);
2182 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2184 struct f2fs_move_range range;
2188 if (!(filp->f_mode & FMODE_READ) ||
2189 !(filp->f_mode & FMODE_WRITE))
2192 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2196 dst = fdget(range.dst_fd);
2200 if (!(dst.file->f_mode & FMODE_WRITE)) {
2205 err = mnt_want_write_file(filp);
2209 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2210 range.pos_out, range.len);
2212 mnt_drop_write_file(filp);
2214 if (copy_to_user((struct f2fs_move_range __user *)arg,
2215 &range, sizeof(range)))
2222 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2225 case F2FS_IOC_GETFLAGS:
2226 return f2fs_ioc_getflags(filp, arg);
2227 case F2FS_IOC_SETFLAGS:
2228 return f2fs_ioc_setflags(filp, arg);
2229 case F2FS_IOC_GETVERSION:
2230 return f2fs_ioc_getversion(filp, arg);
2231 case F2FS_IOC_START_ATOMIC_WRITE:
2232 return f2fs_ioc_start_atomic_write(filp);
2233 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2234 return f2fs_ioc_commit_atomic_write(filp);
2235 case F2FS_IOC_START_VOLATILE_WRITE:
2236 return f2fs_ioc_start_volatile_write(filp);
2237 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2238 return f2fs_ioc_release_volatile_write(filp);
2239 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2240 return f2fs_ioc_abort_volatile_write(filp);
2241 case F2FS_IOC_SHUTDOWN:
2242 return f2fs_ioc_shutdown(filp, arg);
2244 return f2fs_ioc_fitrim(filp, arg);
2245 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2246 return f2fs_ioc_set_encryption_policy(filp, arg);
2247 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2248 return f2fs_ioc_get_encryption_policy(filp, arg);
2249 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2250 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2251 case F2FS_IOC_GARBAGE_COLLECT:
2252 return f2fs_ioc_gc(filp, arg);
2253 case F2FS_IOC_WRITE_CHECKPOINT:
2254 return f2fs_ioc_write_checkpoint(filp, arg);
2255 case F2FS_IOC_DEFRAGMENT:
2256 return f2fs_ioc_defragment(filp, arg);
2257 case F2FS_IOC_MOVE_RANGE:
2258 return f2fs_ioc_move_range(filp, arg);
2264 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2266 struct file *file = iocb->ki_filp;
2267 struct inode *inode = file_inode(file);
2268 struct blk_plug plug;
2271 if (f2fs_encrypted_inode(inode) &&
2272 !fscrypt_has_encryption_key(inode) &&
2273 fscrypt_get_encryption_info(inode))
2277 ret = generic_write_checks(iocb, from);
2281 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2282 set_inode_flag(inode, FI_NO_PREALLOC);
2284 err = f2fs_preallocate_blocks(iocb, from);
2286 inode_unlock(inode);
2289 blk_start_plug(&plug);
2290 ret = __generic_file_write_iter(iocb, from);
2291 blk_finish_plug(&plug);
2292 clear_inode_flag(inode, FI_NO_PREALLOC);
2294 inode_unlock(inode);
2297 ret = generic_write_sync(iocb, ret);
2301 #ifdef CONFIG_COMPAT
2302 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2305 case F2FS_IOC32_GETFLAGS:
2306 cmd = F2FS_IOC_GETFLAGS;
2308 case F2FS_IOC32_SETFLAGS:
2309 cmd = F2FS_IOC_SETFLAGS;
2311 case F2FS_IOC32_GETVERSION:
2312 cmd = F2FS_IOC_GETVERSION;
2314 case F2FS_IOC_START_ATOMIC_WRITE:
2315 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2316 case F2FS_IOC_START_VOLATILE_WRITE:
2317 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2318 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2319 case F2FS_IOC_SHUTDOWN:
2320 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2321 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2322 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2323 case F2FS_IOC_GARBAGE_COLLECT:
2324 case F2FS_IOC_WRITE_CHECKPOINT:
2325 case F2FS_IOC_DEFRAGMENT:
2327 case F2FS_IOC_MOVE_RANGE:
2330 return -ENOIOCTLCMD;
2332 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2336 const struct file_operations f2fs_file_operations = {
2337 .llseek = f2fs_llseek,
2338 .read_iter = generic_file_read_iter,
2339 .write_iter = f2fs_file_write_iter,
2340 .open = f2fs_file_open,
2341 .release = f2fs_release_file,
2342 .mmap = f2fs_file_mmap,
2343 .fsync = f2fs_sync_file,
2344 .fallocate = f2fs_fallocate,
2345 .unlocked_ioctl = f2fs_ioctl,
2346 #ifdef CONFIG_COMPAT
2347 .compat_ioctl = f2fs_compat_ioctl,
2349 .splice_read = generic_file_splice_read,
2350 .splice_write = iter_file_splice_write,