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 inode = igrab(inode);
114 dentry = d_find_any_alias(inode);
119 *pino = parent_ino(dentry);
124 static inline bool need_do_checkpoint(struct inode *inode)
126 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
127 bool need_cp = false;
129 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
131 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
133 else if (file_wrong_pino(inode))
135 else if (!space_for_roll_forward(sbi))
137 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
139 else if (test_opt(sbi, FASTBOOT))
141 else if (sbi->active_logs == 2)
147 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
149 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
151 /* But we need to avoid that there are some inode updates */
152 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
158 static void try_to_fix_pino(struct inode *inode)
160 struct f2fs_inode_info *fi = F2FS_I(inode);
163 down_write(&fi->i_sem);
164 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
165 get_parent_ino(inode, &pino)) {
166 f2fs_i_pino_write(inode, pino);
167 file_got_pino(inode);
169 up_write(&fi->i_sem);
172 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
173 int datasync, bool atomic)
175 struct inode *inode = file->f_mapping->host;
176 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
177 nid_t ino = inode->i_ino;
179 bool need_cp = false;
180 struct writeback_control wbc = {
181 .sync_mode = WB_SYNC_ALL,
182 .nr_to_write = LONG_MAX,
186 if (unlikely(f2fs_readonly(inode->i_sb)))
189 trace_f2fs_sync_file_enter(inode);
191 /* if fdatasync is triggered, let's do in-place-update */
192 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
193 set_inode_flag(inode, FI_NEED_IPU);
194 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
195 clear_inode_flag(inode, FI_NEED_IPU);
198 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
202 /* if the inode is dirty, let's recover all the time */
203 if (!f2fs_skip_inode_update(inode, datasync)) {
204 f2fs_write_inode(inode, NULL);
209 * if there is no written data, don't waste time to write recovery info.
211 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
212 !exist_written_data(sbi, ino, APPEND_INO)) {
214 /* it may call write_inode just prior to fsync */
215 if (need_inode_page_update(sbi, ino))
218 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
219 exist_written_data(sbi, ino, UPDATE_INO))
225 * Both of fdatasync() and fsync() are able to be recovered from
228 down_read(&F2FS_I(inode)->i_sem);
229 need_cp = need_do_checkpoint(inode);
230 up_read(&F2FS_I(inode)->i_sem);
233 /* all the dirty node pages should be flushed for POR */
234 ret = f2fs_sync_fs(inode->i_sb, 1);
237 * We've secured consistency through sync_fs. Following pino
238 * will be used only for fsynced inodes after checkpoint.
240 try_to_fix_pino(inode);
241 clear_inode_flag(inode, FI_APPEND_WRITE);
242 clear_inode_flag(inode, FI_UPDATE_WRITE);
246 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
250 /* if cp_error was enabled, we should avoid infinite loop */
251 if (unlikely(f2fs_cp_error(sbi))) {
256 if (need_inode_block_update(sbi, ino)) {
257 f2fs_mark_inode_dirty_sync(inode, true);
258 f2fs_write_inode(inode, NULL);
262 ret = wait_on_node_pages_writeback(sbi, ino);
266 /* once recovery info is written, don't need to tack this */
267 remove_ino_entry(sbi, ino, APPEND_INO);
268 clear_inode_flag(inode, FI_APPEND_WRITE);
270 remove_ino_entry(sbi, ino, UPDATE_INO);
271 clear_inode_flag(inode, FI_UPDATE_WRITE);
273 ret = f2fs_issue_flush(sbi);
274 f2fs_update_time(sbi, REQ_TIME);
276 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
277 f2fs_trace_ios(NULL, 1);
281 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
283 return f2fs_do_sync_file(file, start, end, datasync, false);
286 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
287 pgoff_t pgofs, int whence)
292 if (whence != SEEK_DATA)
295 /* find first dirty page index */
296 pagevec_init(&pvec, 0);
297 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
298 PAGECACHE_TAG_DIRTY, 1);
299 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
300 pagevec_release(&pvec);
304 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
309 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
310 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
314 if (blkaddr == NULL_ADDR)
321 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
323 struct inode *inode = file->f_mapping->host;
324 loff_t maxbytes = inode->i_sb->s_maxbytes;
325 struct dnode_of_data dn;
326 pgoff_t pgofs, end_offset, dirty;
327 loff_t data_ofs = offset;
333 isize = i_size_read(inode);
337 /* handle inline data case */
338 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
339 if (whence == SEEK_HOLE)
344 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
346 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
348 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
349 set_new_dnode(&dn, inode, NULL, NULL, 0);
350 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
351 if (err && err != -ENOENT) {
353 } else if (err == -ENOENT) {
354 /* direct node does not exists */
355 if (whence == SEEK_DATA) {
356 pgofs = get_next_page_offset(&dn, pgofs);
363 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
365 /* find data/hole in dnode block */
366 for (; dn.ofs_in_node < end_offset;
367 dn.ofs_in_node++, pgofs++,
368 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
370 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
372 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
380 if (whence == SEEK_DATA)
383 if (whence == SEEK_HOLE && data_ofs > isize)
386 return vfs_setpos(file, data_ofs, maxbytes);
392 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
394 struct inode *inode = file->f_mapping->host;
395 loff_t maxbytes = inode->i_sb->s_maxbytes;
401 return generic_file_llseek_size(file, offset, whence,
402 maxbytes, i_size_read(inode));
407 return f2fs_seek_block(file, offset, whence);
413 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
415 struct inode *inode = file_inode(file);
418 if (f2fs_encrypted_inode(inode)) {
419 err = fscrypt_get_encryption_info(inode);
422 if (!f2fs_encrypted_inode(inode))
426 /* we don't need to use inline_data strictly */
427 err = f2fs_convert_inline_inode(inode);
432 vma->vm_ops = &f2fs_file_vm_ops;
436 static int f2fs_file_open(struct inode *inode, struct file *filp)
438 int ret = generic_file_open(inode, filp);
441 if (!ret && f2fs_encrypted_inode(inode)) {
442 ret = fscrypt_get_encryption_info(inode);
445 if (!fscrypt_has_encryption_key(inode))
448 dir = dget_parent(file_dentry(filp));
449 if (f2fs_encrypted_inode(d_inode(dir)) &&
450 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
458 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
460 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
461 struct f2fs_node *raw_node;
462 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
465 raw_node = F2FS_NODE(dn->node_page);
466 addr = blkaddr_in_node(raw_node) + ofs;
468 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
469 block_t blkaddr = le32_to_cpu(*addr);
470 if (blkaddr == NULL_ADDR)
473 dn->data_blkaddr = NULL_ADDR;
474 set_data_blkaddr(dn);
475 invalidate_blocks(sbi, blkaddr);
476 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
477 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
484 * once we invalidate valid blkaddr in range [ofs, ofs + count],
485 * we will invalidate all blkaddr in the whole range.
487 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
489 f2fs_update_extent_cache_range(dn, fofs, 0, len);
490 dec_valid_block_count(sbi, dn->inode, nr_free);
492 dn->ofs_in_node = ofs;
494 f2fs_update_time(sbi, REQ_TIME);
495 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
496 dn->ofs_in_node, nr_free);
500 void truncate_data_blocks(struct dnode_of_data *dn)
502 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
505 static int truncate_partial_data_page(struct inode *inode, u64 from,
508 unsigned offset = from & (PAGE_SIZE - 1);
509 pgoff_t index = from >> PAGE_SHIFT;
510 struct address_space *mapping = inode->i_mapping;
513 if (!offset && !cache_only)
517 page = find_lock_page(mapping, index);
518 if (page && PageUptodate(page))
520 f2fs_put_page(page, 1);
524 page = get_lock_data_page(inode, index, true);
526 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
528 f2fs_wait_on_page_writeback(page, DATA, true);
529 zero_user(page, offset, PAGE_SIZE - offset);
530 if (!cache_only || !f2fs_encrypted_inode(inode) ||
531 !S_ISREG(inode->i_mode))
532 set_page_dirty(page);
533 f2fs_put_page(page, 1);
537 int truncate_blocks(struct inode *inode, u64 from, bool lock)
539 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
540 unsigned int blocksize = inode->i_sb->s_blocksize;
541 struct dnode_of_data dn;
543 int count = 0, err = 0;
545 bool truncate_page = false;
547 trace_f2fs_truncate_blocks_enter(inode, from);
549 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
551 if (free_from >= sbi->max_file_blocks)
557 ipage = get_node_page(sbi, inode->i_ino);
559 err = PTR_ERR(ipage);
563 if (f2fs_has_inline_data(inode)) {
564 truncate_inline_inode(inode, ipage, from);
565 f2fs_put_page(ipage, 1);
566 truncate_page = true;
570 set_new_dnode(&dn, inode, ipage, NULL, 0);
571 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
578 count = ADDRS_PER_PAGE(dn.node_page, inode);
580 count -= dn.ofs_in_node;
581 f2fs_bug_on(sbi, count < 0);
583 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
584 truncate_data_blocks_range(&dn, count);
590 err = truncate_inode_blocks(inode, free_from);
595 /* lastly zero out the first data page */
597 err = truncate_partial_data_page(inode, from, truncate_page);
599 trace_f2fs_truncate_blocks_exit(inode, err);
603 int f2fs_truncate(struct inode *inode)
607 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
608 S_ISLNK(inode->i_mode)))
611 trace_f2fs_truncate(inode);
613 #ifdef CONFIG_F2FS_FAULT_INJECTION
614 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
615 f2fs_show_injection_info(FAULT_TRUNCATE);
619 /* we should check inline_data size */
620 if (!f2fs_may_inline_data(inode)) {
621 err = f2fs_convert_inline_inode(inode);
626 err = truncate_blocks(inode, i_size_read(inode), true);
630 inode->i_mtime = inode->i_ctime = current_time(inode);
631 f2fs_mark_inode_dirty_sync(inode, false);
635 int f2fs_getattr(const struct path *path, struct kstat *stat,
636 u32 request_mask, unsigned int flags)
638 struct inode *inode = d_inode(path->dentry);
639 generic_fillattr(inode, stat);
644 #ifdef CONFIG_F2FS_FS_POSIX_ACL
645 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
647 unsigned int ia_valid = attr->ia_valid;
649 if (ia_valid & ATTR_UID)
650 inode->i_uid = attr->ia_uid;
651 if (ia_valid & ATTR_GID)
652 inode->i_gid = attr->ia_gid;
653 if (ia_valid & ATTR_ATIME)
654 inode->i_atime = timespec_trunc(attr->ia_atime,
655 inode->i_sb->s_time_gran);
656 if (ia_valid & ATTR_MTIME)
657 inode->i_mtime = timespec_trunc(attr->ia_mtime,
658 inode->i_sb->s_time_gran);
659 if (ia_valid & ATTR_CTIME)
660 inode->i_ctime = timespec_trunc(attr->ia_ctime,
661 inode->i_sb->s_time_gran);
662 if (ia_valid & ATTR_MODE) {
663 umode_t mode = attr->ia_mode;
665 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
667 set_acl_inode(inode, mode);
671 #define __setattr_copy setattr_copy
674 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
676 struct inode *inode = d_inode(dentry);
678 bool size_changed = false;
680 err = setattr_prepare(dentry, attr);
684 if (attr->ia_valid & ATTR_SIZE) {
685 if (f2fs_encrypted_inode(inode) &&
686 fscrypt_get_encryption_info(inode))
689 if (attr->ia_size <= i_size_read(inode)) {
690 truncate_setsize(inode, attr->ia_size);
691 err = f2fs_truncate(inode);
696 * do not trim all blocks after i_size if target size is
697 * larger than i_size.
699 truncate_setsize(inode, attr->ia_size);
701 /* should convert inline inode here */
702 if (!f2fs_may_inline_data(inode)) {
703 err = f2fs_convert_inline_inode(inode);
707 inode->i_mtime = inode->i_ctime = current_time(inode);
713 __setattr_copy(inode, attr);
715 if (attr->ia_valid & ATTR_MODE) {
716 err = posix_acl_chmod(inode, get_inode_mode(inode));
717 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
718 inode->i_mode = F2FS_I(inode)->i_acl_mode;
719 clear_inode_flag(inode, FI_ACL_MODE);
723 /* file size may changed here */
724 f2fs_mark_inode_dirty_sync(inode, size_changed);
726 /* inode change will produce dirty node pages flushed by checkpoint */
727 f2fs_balance_fs(F2FS_I_SB(inode), true);
732 const struct inode_operations f2fs_file_inode_operations = {
733 .getattr = f2fs_getattr,
734 .setattr = f2fs_setattr,
735 .get_acl = f2fs_get_acl,
736 .set_acl = f2fs_set_acl,
737 #ifdef CONFIG_F2FS_FS_XATTR
738 .listxattr = f2fs_listxattr,
740 .fiemap = f2fs_fiemap,
743 static int fill_zero(struct inode *inode, pgoff_t index,
744 loff_t start, loff_t len)
746 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
752 f2fs_balance_fs(sbi, true);
755 page = get_new_data_page(inode, NULL, index, false);
759 return PTR_ERR(page);
761 f2fs_wait_on_page_writeback(page, DATA, true);
762 zero_user(page, start, len);
763 set_page_dirty(page);
764 f2fs_put_page(page, 1);
768 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
772 while (pg_start < pg_end) {
773 struct dnode_of_data dn;
774 pgoff_t end_offset, count;
776 set_new_dnode(&dn, inode, NULL, NULL, 0);
777 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
779 if (err == -ENOENT) {
786 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
787 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
789 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
791 truncate_data_blocks_range(&dn, count);
799 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
801 pgoff_t pg_start, pg_end;
802 loff_t off_start, off_end;
805 ret = f2fs_convert_inline_inode(inode);
809 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
810 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
812 off_start = offset & (PAGE_SIZE - 1);
813 off_end = (offset + len) & (PAGE_SIZE - 1);
815 if (pg_start == pg_end) {
816 ret = fill_zero(inode, pg_start, off_start,
817 off_end - off_start);
822 ret = fill_zero(inode, pg_start++, off_start,
823 PAGE_SIZE - off_start);
828 ret = fill_zero(inode, pg_end, 0, off_end);
833 if (pg_start < pg_end) {
834 struct address_space *mapping = inode->i_mapping;
835 loff_t blk_start, blk_end;
836 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
838 f2fs_balance_fs(sbi, true);
840 blk_start = (loff_t)pg_start << PAGE_SHIFT;
841 blk_end = (loff_t)pg_end << PAGE_SHIFT;
842 truncate_inode_pages_range(mapping, blk_start,
846 ret = truncate_hole(inode, pg_start, pg_end);
854 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
855 int *do_replace, pgoff_t off, pgoff_t len)
857 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
858 struct dnode_of_data dn;
862 set_new_dnode(&dn, inode, NULL, NULL, 0);
863 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
864 if (ret && ret != -ENOENT) {
866 } else if (ret == -ENOENT) {
867 if (dn.max_level == 0)
869 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
875 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
876 dn.ofs_in_node, len);
877 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
878 *blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
879 if (!is_checkpointed_data(sbi, *blkaddr)) {
881 if (test_opt(sbi, LFS)) {
886 /* do not invalidate this block address */
887 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
900 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
901 int *do_replace, pgoff_t off, int len)
903 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
904 struct dnode_of_data dn;
907 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
908 if (*do_replace == 0)
911 set_new_dnode(&dn, inode, NULL, NULL, 0);
912 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
914 dec_valid_block_count(sbi, inode, 1);
915 invalidate_blocks(sbi, *blkaddr);
917 f2fs_update_data_blkaddr(&dn, *blkaddr);
924 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
925 block_t *blkaddr, int *do_replace,
926 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
928 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
933 if (blkaddr[i] == NULL_ADDR && !full) {
938 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
939 struct dnode_of_data dn;
944 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
945 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
949 get_node_info(sbi, dn.nid, &ni);
951 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
952 dn.ofs_in_node, len - i);
954 dn.data_blkaddr = datablock_addr(dn.node_page,
956 truncate_data_blocks_range(&dn, 1);
959 f2fs_i_blocks_write(src_inode,
961 f2fs_i_blocks_write(dst_inode,
963 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
964 blkaddr[i], ni.version, true, false);
970 new_size = (dst + i) << PAGE_SHIFT;
971 if (dst_inode->i_size < new_size)
972 f2fs_i_size_write(dst_inode, new_size);
973 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
977 struct page *psrc, *pdst;
979 psrc = get_lock_data_page(src_inode, src + i, true);
981 return PTR_ERR(psrc);
982 pdst = get_new_data_page(dst_inode, NULL, dst + i,
985 f2fs_put_page(psrc, 1);
986 return PTR_ERR(pdst);
988 f2fs_copy_page(psrc, pdst);
989 set_page_dirty(pdst);
990 f2fs_put_page(pdst, 1);
991 f2fs_put_page(psrc, 1);
993 ret = truncate_hole(src_inode, src + i, src + i + 1);
1002 static int __exchange_data_block(struct inode *src_inode,
1003 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1004 pgoff_t len, bool full)
1006 block_t *src_blkaddr;
1012 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1014 src_blkaddr = kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1018 do_replace = kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1020 kvfree(src_blkaddr);
1024 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1025 do_replace, src, olen);
1029 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1030 do_replace, src, dst, olen, full);
1038 kvfree(src_blkaddr);
1044 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1045 kvfree(src_blkaddr);
1050 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1052 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1053 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1056 f2fs_balance_fs(sbi, true);
1059 f2fs_drop_extent_tree(inode);
1061 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1062 f2fs_unlock_op(sbi);
1066 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1068 pgoff_t pg_start, pg_end;
1072 if (offset + len >= i_size_read(inode))
1075 /* collapse range should be aligned to block size of f2fs. */
1076 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1079 ret = f2fs_convert_inline_inode(inode);
1083 pg_start = offset >> PAGE_SHIFT;
1084 pg_end = (offset + len) >> PAGE_SHIFT;
1086 /* write out all dirty pages from offset */
1087 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1091 truncate_pagecache(inode, offset);
1093 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1097 /* write out all moved pages, if possible */
1098 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1099 truncate_pagecache(inode, offset);
1101 new_size = i_size_read(inode) - len;
1102 truncate_pagecache(inode, new_size);
1104 ret = truncate_blocks(inode, new_size, true);
1106 f2fs_i_size_write(inode, new_size);
1111 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1114 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1115 pgoff_t index = start;
1116 unsigned int ofs_in_node = dn->ofs_in_node;
1120 for (; index < end; index++, dn->ofs_in_node++) {
1121 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1125 dn->ofs_in_node = ofs_in_node;
1126 ret = reserve_new_blocks(dn, count);
1130 dn->ofs_in_node = ofs_in_node;
1131 for (index = start; index < end; index++, dn->ofs_in_node++) {
1133 datablock_addr(dn->node_page, dn->ofs_in_node);
1135 * reserve_new_blocks will not guarantee entire block
1138 if (dn->data_blkaddr == NULL_ADDR) {
1142 if (dn->data_blkaddr != NEW_ADDR) {
1143 invalidate_blocks(sbi, dn->data_blkaddr);
1144 dn->data_blkaddr = NEW_ADDR;
1145 set_data_blkaddr(dn);
1149 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1154 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1157 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1158 struct address_space *mapping = inode->i_mapping;
1159 pgoff_t index, pg_start, pg_end;
1160 loff_t new_size = i_size_read(inode);
1161 loff_t off_start, off_end;
1164 ret = inode_newsize_ok(inode, (len + offset));
1168 ret = f2fs_convert_inline_inode(inode);
1172 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1176 truncate_pagecache_range(inode, offset, offset + len - 1);
1178 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1179 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1181 off_start = offset & (PAGE_SIZE - 1);
1182 off_end = (offset + len) & (PAGE_SIZE - 1);
1184 if (pg_start == pg_end) {
1185 ret = fill_zero(inode, pg_start, off_start,
1186 off_end - off_start);
1190 new_size = max_t(loff_t, new_size, offset + len);
1193 ret = fill_zero(inode, pg_start++, off_start,
1194 PAGE_SIZE - off_start);
1198 new_size = max_t(loff_t, new_size,
1199 (loff_t)pg_start << PAGE_SHIFT);
1202 for (index = pg_start; index < pg_end;) {
1203 struct dnode_of_data dn;
1204 unsigned int end_offset;
1209 set_new_dnode(&dn, inode, NULL, NULL, 0);
1210 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1212 f2fs_unlock_op(sbi);
1216 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1217 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1219 ret = f2fs_do_zero_range(&dn, index, end);
1220 f2fs_put_dnode(&dn);
1221 f2fs_unlock_op(sbi);
1223 f2fs_balance_fs(sbi, dn.node_changed);
1229 new_size = max_t(loff_t, new_size,
1230 (loff_t)index << PAGE_SHIFT);
1234 ret = fill_zero(inode, pg_end, 0, off_end);
1238 new_size = max_t(loff_t, new_size, offset + len);
1243 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1244 f2fs_i_size_write(inode, new_size);
1249 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1251 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1252 pgoff_t nr, pg_start, pg_end, delta, idx;
1256 new_size = i_size_read(inode) + len;
1257 ret = inode_newsize_ok(inode, new_size);
1261 if (offset >= i_size_read(inode))
1264 /* insert range should be aligned to block size of f2fs. */
1265 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1268 ret = f2fs_convert_inline_inode(inode);
1272 f2fs_balance_fs(sbi, true);
1274 ret = truncate_blocks(inode, i_size_read(inode), true);
1278 /* write out all dirty pages from offset */
1279 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1283 truncate_pagecache(inode, offset);
1285 pg_start = offset >> PAGE_SHIFT;
1286 pg_end = (offset + len) >> PAGE_SHIFT;
1287 delta = pg_end - pg_start;
1288 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1290 while (!ret && idx > pg_start) {
1291 nr = idx - pg_start;
1297 f2fs_drop_extent_tree(inode);
1299 ret = __exchange_data_block(inode, inode, idx,
1300 idx + delta, nr, false);
1301 f2fs_unlock_op(sbi);
1304 /* write out all moved pages, if possible */
1305 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1306 truncate_pagecache(inode, offset);
1309 f2fs_i_size_write(inode, new_size);
1313 static int expand_inode_data(struct inode *inode, loff_t offset,
1314 loff_t len, int mode)
1316 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1317 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1319 loff_t new_size = i_size_read(inode);
1323 err = inode_newsize_ok(inode, (len + offset));
1327 err = f2fs_convert_inline_inode(inode);
1331 f2fs_balance_fs(sbi, true);
1333 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1334 off_end = (offset + len) & (PAGE_SIZE - 1);
1336 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1337 map.m_len = pg_end - map.m_lblk;
1341 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1348 last_off = map.m_lblk + map.m_len - 1;
1350 /* update new size to the failed position */
1351 new_size = (last_off == pg_end) ? offset + len:
1352 (loff_t)(last_off + 1) << PAGE_SHIFT;
1354 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1357 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1358 f2fs_i_size_write(inode, new_size);
1363 static long f2fs_fallocate(struct file *file, int mode,
1364 loff_t offset, loff_t len)
1366 struct inode *inode = file_inode(file);
1369 /* f2fs only support ->fallocate for regular file */
1370 if (!S_ISREG(inode->i_mode))
1373 if (f2fs_encrypted_inode(inode) &&
1374 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1377 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1378 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1379 FALLOC_FL_INSERT_RANGE))
1384 if (mode & FALLOC_FL_PUNCH_HOLE) {
1385 if (offset >= inode->i_size)
1388 ret = punch_hole(inode, offset, len);
1389 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1390 ret = f2fs_collapse_range(inode, offset, len);
1391 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1392 ret = f2fs_zero_range(inode, offset, len, mode);
1393 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1394 ret = f2fs_insert_range(inode, offset, len);
1396 ret = expand_inode_data(inode, offset, len, mode);
1400 inode->i_mtime = inode->i_ctime = current_time(inode);
1401 f2fs_mark_inode_dirty_sync(inode, false);
1402 if (mode & FALLOC_FL_KEEP_SIZE)
1403 file_set_keep_isize(inode);
1404 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1408 inode_unlock(inode);
1410 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1414 static int f2fs_release_file(struct inode *inode, struct file *filp)
1417 * f2fs_relase_file is called at every close calls. So we should
1418 * not drop any inmemory pages by close called by other process.
1420 if (!(filp->f_mode & FMODE_WRITE) ||
1421 atomic_read(&inode->i_writecount) != 1)
1424 /* some remained atomic pages should discarded */
1425 if (f2fs_is_atomic_file(inode))
1426 drop_inmem_pages(inode);
1427 if (f2fs_is_volatile_file(inode)) {
1428 clear_inode_flag(inode, FI_VOLATILE_FILE);
1429 stat_dec_volatile_write(inode);
1430 set_inode_flag(inode, FI_DROP_CACHE);
1431 filemap_fdatawrite(inode->i_mapping);
1432 clear_inode_flag(inode, FI_DROP_CACHE);
1437 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1438 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1440 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1444 else if (S_ISREG(mode))
1445 return flags & F2FS_REG_FLMASK;
1447 return flags & F2FS_OTHER_FLMASK;
1450 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1452 struct inode *inode = file_inode(filp);
1453 struct f2fs_inode_info *fi = F2FS_I(inode);
1454 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1455 return put_user(flags, (int __user *)arg);
1458 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1460 struct inode *inode = file_inode(filp);
1461 struct f2fs_inode_info *fi = F2FS_I(inode);
1463 unsigned int oldflags;
1466 if (!inode_owner_or_capable(inode))
1469 if (get_user(flags, (int __user *)arg))
1472 ret = mnt_want_write_file(filp);
1478 flags = f2fs_mask_flags(inode->i_mode, flags);
1480 oldflags = fi->i_flags;
1482 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1483 if (!capable(CAP_LINUX_IMMUTABLE)) {
1484 inode_unlock(inode);
1490 flags = flags & FS_FL_USER_MODIFIABLE;
1491 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1492 fi->i_flags = flags;
1494 inode->i_ctime = current_time(inode);
1495 f2fs_set_inode_flags(inode);
1497 inode_unlock(inode);
1499 mnt_drop_write_file(filp);
1503 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1505 struct inode *inode = file_inode(filp);
1507 return put_user(inode->i_generation, (int __user *)arg);
1510 static int f2fs_ioc_start_atomic_write(struct file *filp)
1512 struct inode *inode = file_inode(filp);
1515 if (!inode_owner_or_capable(inode))
1518 if (!S_ISREG(inode->i_mode))
1521 ret = mnt_want_write_file(filp);
1527 if (f2fs_is_atomic_file(inode))
1530 ret = f2fs_convert_inline_inode(inode);
1534 set_inode_flag(inode, FI_ATOMIC_FILE);
1535 set_inode_flag(inode, FI_HOT_DATA);
1536 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1538 if (!get_dirty_pages(inode))
1541 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1542 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1543 inode->i_ino, get_dirty_pages(inode));
1544 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1546 clear_inode_flag(inode, FI_ATOMIC_FILE);
1551 stat_inc_atomic_write(inode);
1552 stat_update_max_atomic_write(inode);
1554 inode_unlock(inode);
1555 mnt_drop_write_file(filp);
1559 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1561 struct inode *inode = file_inode(filp);
1564 if (!inode_owner_or_capable(inode))
1567 ret = mnt_want_write_file(filp);
1573 if (f2fs_is_volatile_file(inode))
1576 if (f2fs_is_atomic_file(inode)) {
1577 ret = commit_inmem_pages(inode);
1581 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1583 clear_inode_flag(inode, FI_ATOMIC_FILE);
1584 stat_dec_atomic_write(inode);
1587 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1590 inode_unlock(inode);
1591 mnt_drop_write_file(filp);
1595 static int f2fs_ioc_start_volatile_write(struct file *filp)
1597 struct inode *inode = file_inode(filp);
1600 if (!inode_owner_or_capable(inode))
1603 if (!S_ISREG(inode->i_mode))
1606 ret = mnt_want_write_file(filp);
1612 if (f2fs_is_volatile_file(inode))
1615 ret = f2fs_convert_inline_inode(inode);
1619 stat_inc_volatile_write(inode);
1620 stat_update_max_volatile_write(inode);
1622 set_inode_flag(inode, FI_VOLATILE_FILE);
1623 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1625 inode_unlock(inode);
1626 mnt_drop_write_file(filp);
1630 static int f2fs_ioc_release_volatile_write(struct file *filp)
1632 struct inode *inode = file_inode(filp);
1635 if (!inode_owner_or_capable(inode))
1638 ret = mnt_want_write_file(filp);
1644 if (!f2fs_is_volatile_file(inode))
1647 if (!f2fs_is_first_block_written(inode)) {
1648 ret = truncate_partial_data_page(inode, 0, true);
1652 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1654 inode_unlock(inode);
1655 mnt_drop_write_file(filp);
1659 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1661 struct inode *inode = file_inode(filp);
1664 if (!inode_owner_or_capable(inode))
1667 ret = mnt_want_write_file(filp);
1673 if (f2fs_is_atomic_file(inode))
1674 drop_inmem_pages(inode);
1675 if (f2fs_is_volatile_file(inode)) {
1676 clear_inode_flag(inode, FI_VOLATILE_FILE);
1677 stat_dec_volatile_write(inode);
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, NULL_SEGNO);
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 block_t blk_end = 0;
1898 bool fragmented = false;
1901 /* if in-place-update policy is enabled, don't waste time here */
1902 if (need_inplace_update_policy(inode, NULL))
1905 pg_start = range->start >> PAGE_SHIFT;
1906 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1908 f2fs_balance_fs(sbi, true);
1912 /* writeback all dirty pages in the range */
1913 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1914 range->start + range->len - 1);
1919 * lookup mapping info in extent cache, skip defragmenting if physical
1920 * block addresses are continuous.
1922 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1923 if (ei.fofs + ei.len >= pg_end)
1927 map.m_lblk = pg_start;
1930 * lookup mapping info in dnode page cache, skip defragmenting if all
1931 * physical block addresses are continuous even if there are hole(s)
1932 * in logical blocks.
1934 while (map.m_lblk < pg_end) {
1935 map.m_len = pg_end - map.m_lblk;
1936 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1940 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1945 if (blk_end && blk_end != map.m_pblk) {
1949 blk_end = map.m_pblk + map.m_len;
1951 map.m_lblk += map.m_len;
1957 map.m_lblk = pg_start;
1958 map.m_len = pg_end - pg_start;
1960 sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
1963 * make sure there are enough free section for LFS allocation, this can
1964 * avoid defragment running in SSR mode when free section are allocated
1967 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
1972 while (map.m_lblk < pg_end) {
1977 map.m_len = pg_end - map.m_lblk;
1978 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1982 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1987 set_inode_flag(inode, FI_DO_DEFRAG);
1990 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1993 page = get_lock_data_page(inode, idx, true);
1995 err = PTR_ERR(page);
1999 set_page_dirty(page);
2000 f2fs_put_page(page, 1);
2009 if (idx < pg_end && cnt < blk_per_seg)
2012 clear_inode_flag(inode, FI_DO_DEFRAG);
2014 err = filemap_fdatawrite(inode->i_mapping);
2019 clear_inode_flag(inode, FI_DO_DEFRAG);
2021 inode_unlock(inode);
2023 range->len = (u64)total << PAGE_SHIFT;
2027 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2029 struct inode *inode = file_inode(filp);
2030 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2031 struct f2fs_defragment range;
2034 if (!capable(CAP_SYS_ADMIN))
2037 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2040 if (f2fs_readonly(sbi->sb))
2043 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2047 /* verify alignment of offset & size */
2048 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2051 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2052 sbi->max_file_blocks))
2055 err = mnt_want_write_file(filp);
2059 err = f2fs_defragment_range(sbi, filp, &range);
2060 mnt_drop_write_file(filp);
2062 f2fs_update_time(sbi, REQ_TIME);
2066 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2073 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2074 struct file *file_out, loff_t pos_out, size_t len)
2076 struct inode *src = file_inode(file_in);
2077 struct inode *dst = file_inode(file_out);
2078 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2079 size_t olen = len, dst_max_i_size = 0;
2083 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2084 src->i_sb != dst->i_sb)
2087 if (unlikely(f2fs_readonly(src->i_sb)))
2090 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2093 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2097 if (pos_in == pos_out)
2099 if (pos_out > pos_in && pos_out < pos_in + len)
2105 if (!inode_trylock(dst)) {
2112 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2115 olen = len = src->i_size - pos_in;
2116 if (pos_in + len == src->i_size)
2117 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2123 dst_osize = dst->i_size;
2124 if (pos_out + olen > dst->i_size)
2125 dst_max_i_size = pos_out + olen;
2127 /* verify the end result is block aligned */
2128 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2129 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2130 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2133 ret = f2fs_convert_inline_inode(src);
2137 ret = f2fs_convert_inline_inode(dst);
2141 /* write out all dirty pages from offset */
2142 ret = filemap_write_and_wait_range(src->i_mapping,
2143 pos_in, pos_in + len);
2147 ret = filemap_write_and_wait_range(dst->i_mapping,
2148 pos_out, pos_out + len);
2152 f2fs_balance_fs(sbi, true);
2154 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2155 pos_out >> F2FS_BLKSIZE_BITS,
2156 len >> F2FS_BLKSIZE_BITS, false);
2160 f2fs_i_size_write(dst, dst_max_i_size);
2161 else if (dst_osize != dst->i_size)
2162 f2fs_i_size_write(dst, dst_osize);
2164 f2fs_unlock_op(sbi);
2173 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2175 struct f2fs_move_range range;
2179 if (!(filp->f_mode & FMODE_READ) ||
2180 !(filp->f_mode & FMODE_WRITE))
2183 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2187 dst = fdget(range.dst_fd);
2191 if (!(dst.file->f_mode & FMODE_WRITE)) {
2196 err = mnt_want_write_file(filp);
2200 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2201 range.pos_out, range.len);
2203 mnt_drop_write_file(filp);
2207 if (copy_to_user((struct f2fs_move_range __user *)arg,
2208 &range, sizeof(range)))
2215 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2217 struct inode *inode = file_inode(filp);
2218 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2219 struct sit_info *sm = SIT_I(sbi);
2220 unsigned int start_segno = 0, end_segno = 0;
2221 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2222 struct f2fs_flush_device range;
2225 if (!capable(CAP_SYS_ADMIN))
2228 if (f2fs_readonly(sbi->sb))
2231 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2235 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2236 sbi->segs_per_sec != 1) {
2237 f2fs_msg(sbi->sb, KERN_WARNING,
2238 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2239 range.dev_num, sbi->s_ndevs,
2244 ret = mnt_want_write_file(filp);
2248 if (range.dev_num != 0)
2249 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2250 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2252 start_segno = sm->last_victim[FLUSH_DEVICE];
2253 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2254 start_segno = dev_start_segno;
2255 end_segno = min(start_segno + range.segments, dev_end_segno);
2257 while (start_segno < end_segno) {
2258 if (!mutex_trylock(&sbi->gc_mutex)) {
2262 sm->last_victim[GC_CB] = end_segno + 1;
2263 sm->last_victim[GC_GREEDY] = end_segno + 1;
2264 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2265 ret = f2fs_gc(sbi, true, true, start_segno);
2273 mnt_drop_write_file(filp);
2278 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2281 case F2FS_IOC_GETFLAGS:
2282 return f2fs_ioc_getflags(filp, arg);
2283 case F2FS_IOC_SETFLAGS:
2284 return f2fs_ioc_setflags(filp, arg);
2285 case F2FS_IOC_GETVERSION:
2286 return f2fs_ioc_getversion(filp, arg);
2287 case F2FS_IOC_START_ATOMIC_WRITE:
2288 return f2fs_ioc_start_atomic_write(filp);
2289 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2290 return f2fs_ioc_commit_atomic_write(filp);
2291 case F2FS_IOC_START_VOLATILE_WRITE:
2292 return f2fs_ioc_start_volatile_write(filp);
2293 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2294 return f2fs_ioc_release_volatile_write(filp);
2295 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2296 return f2fs_ioc_abort_volatile_write(filp);
2297 case F2FS_IOC_SHUTDOWN:
2298 return f2fs_ioc_shutdown(filp, arg);
2300 return f2fs_ioc_fitrim(filp, arg);
2301 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2302 return f2fs_ioc_set_encryption_policy(filp, arg);
2303 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2304 return f2fs_ioc_get_encryption_policy(filp, arg);
2305 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2306 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2307 case F2FS_IOC_GARBAGE_COLLECT:
2308 return f2fs_ioc_gc(filp, arg);
2309 case F2FS_IOC_WRITE_CHECKPOINT:
2310 return f2fs_ioc_write_checkpoint(filp, arg);
2311 case F2FS_IOC_DEFRAGMENT:
2312 return f2fs_ioc_defragment(filp, arg);
2313 case F2FS_IOC_MOVE_RANGE:
2314 return f2fs_ioc_move_range(filp, arg);
2315 case F2FS_IOC_FLUSH_DEVICE:
2316 return f2fs_ioc_flush_device(filp, arg);
2322 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2324 struct file *file = iocb->ki_filp;
2325 struct inode *inode = file_inode(file);
2326 struct blk_plug plug;
2329 if (f2fs_encrypted_inode(inode) &&
2330 !fscrypt_has_encryption_key(inode) &&
2331 fscrypt_get_encryption_info(inode))
2335 ret = generic_write_checks(iocb, from);
2339 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2340 set_inode_flag(inode, FI_NO_PREALLOC);
2342 err = f2fs_preallocate_blocks(iocb, from);
2344 inode_unlock(inode);
2347 blk_start_plug(&plug);
2348 ret = __generic_file_write_iter(iocb, from);
2349 blk_finish_plug(&plug);
2350 clear_inode_flag(inode, FI_NO_PREALLOC);
2352 inode_unlock(inode);
2355 ret = generic_write_sync(iocb, ret);
2359 #ifdef CONFIG_COMPAT
2360 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2363 case F2FS_IOC32_GETFLAGS:
2364 cmd = F2FS_IOC_GETFLAGS;
2366 case F2FS_IOC32_SETFLAGS:
2367 cmd = F2FS_IOC_SETFLAGS;
2369 case F2FS_IOC32_GETVERSION:
2370 cmd = F2FS_IOC_GETVERSION;
2372 case F2FS_IOC_START_ATOMIC_WRITE:
2373 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2374 case F2FS_IOC_START_VOLATILE_WRITE:
2375 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2376 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2377 case F2FS_IOC_SHUTDOWN:
2378 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2379 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2380 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2381 case F2FS_IOC_GARBAGE_COLLECT:
2382 case F2FS_IOC_WRITE_CHECKPOINT:
2383 case F2FS_IOC_DEFRAGMENT:
2384 case F2FS_IOC_MOVE_RANGE:
2385 case F2FS_IOC_FLUSH_DEVICE:
2388 return -ENOIOCTLCMD;
2390 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2394 const struct file_operations f2fs_file_operations = {
2395 .llseek = f2fs_llseek,
2396 .read_iter = generic_file_read_iter,
2397 .write_iter = f2fs_file_write_iter,
2398 .open = f2fs_file_open,
2399 .release = f2fs_release_file,
2400 .mmap = f2fs_file_mmap,
2401 .fsync = f2fs_sync_file,
2402 .fallocate = f2fs_fallocate,
2403 .unlocked_ioctl = f2fs_ioctl,
2404 #ifdef CONFIG_COMPAT
2405 .compat_ioctl = f2fs_compat_ioctl,
2407 .splice_read = generic_file_splice_read,
2408 .splice_write = iter_file_splice_write,