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 /* we don't need to use inline_data strictly */
419 err = f2fs_convert_inline_inode(inode);
424 vma->vm_ops = &f2fs_file_vm_ops;
428 static int f2fs_file_open(struct inode *inode, struct file *filp)
430 int ret = generic_file_open(inode, filp);
433 if (!ret && f2fs_encrypted_inode(inode)) {
434 ret = fscrypt_get_encryption_info(inode);
437 if (!fscrypt_has_encryption_key(inode))
440 dir = dget_parent(file_dentry(filp));
441 if (f2fs_encrypted_inode(d_inode(dir)) &&
442 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
450 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
452 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
453 struct f2fs_node *raw_node;
454 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
457 raw_node = F2FS_NODE(dn->node_page);
458 addr = blkaddr_in_node(raw_node) + ofs;
460 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
461 block_t blkaddr = le32_to_cpu(*addr);
462 if (blkaddr == NULL_ADDR)
465 dn->data_blkaddr = NULL_ADDR;
466 set_data_blkaddr(dn);
467 invalidate_blocks(sbi, blkaddr);
468 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
469 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
476 * once we invalidate valid blkaddr in range [ofs, ofs + count],
477 * we will invalidate all blkaddr in the whole range.
479 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
481 f2fs_update_extent_cache_range(dn, fofs, 0, len);
482 dec_valid_block_count(sbi, dn->inode, nr_free);
484 dn->ofs_in_node = ofs;
486 f2fs_update_time(sbi, REQ_TIME);
487 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
488 dn->ofs_in_node, nr_free);
492 void truncate_data_blocks(struct dnode_of_data *dn)
494 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
497 static int truncate_partial_data_page(struct inode *inode, u64 from,
500 unsigned offset = from & (PAGE_SIZE - 1);
501 pgoff_t index = from >> PAGE_SHIFT;
502 struct address_space *mapping = inode->i_mapping;
505 if (!offset && !cache_only)
509 page = find_lock_page(mapping, index);
510 if (page && PageUptodate(page))
512 f2fs_put_page(page, 1);
516 page = get_lock_data_page(inode, index, true);
518 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
520 f2fs_wait_on_page_writeback(page, DATA, true);
521 zero_user(page, offset, PAGE_SIZE - offset);
522 if (!cache_only || !f2fs_encrypted_inode(inode) ||
523 !S_ISREG(inode->i_mode))
524 set_page_dirty(page);
525 f2fs_put_page(page, 1);
529 int truncate_blocks(struct inode *inode, u64 from, bool lock)
531 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
532 unsigned int blocksize = inode->i_sb->s_blocksize;
533 struct dnode_of_data dn;
535 int count = 0, err = 0;
537 bool truncate_page = false;
539 trace_f2fs_truncate_blocks_enter(inode, from);
541 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
543 if (free_from >= sbi->max_file_blocks)
549 ipage = get_node_page(sbi, inode->i_ino);
551 err = PTR_ERR(ipage);
555 if (f2fs_has_inline_data(inode)) {
556 truncate_inline_inode(inode, ipage, from);
557 f2fs_put_page(ipage, 1);
558 truncate_page = true;
562 set_new_dnode(&dn, inode, ipage, NULL, 0);
563 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
570 count = ADDRS_PER_PAGE(dn.node_page, inode);
572 count -= dn.ofs_in_node;
573 f2fs_bug_on(sbi, count < 0);
575 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
576 truncate_data_blocks_range(&dn, count);
582 err = truncate_inode_blocks(inode, free_from);
587 /* lastly zero out the first data page */
589 err = truncate_partial_data_page(inode, from, truncate_page);
591 trace_f2fs_truncate_blocks_exit(inode, err);
595 int f2fs_truncate(struct inode *inode)
599 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
600 S_ISLNK(inode->i_mode)))
603 trace_f2fs_truncate(inode);
605 #ifdef CONFIG_F2FS_FAULT_INJECTION
606 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
607 f2fs_show_injection_info(FAULT_TRUNCATE);
611 /* we should check inline_data size */
612 if (!f2fs_may_inline_data(inode)) {
613 err = f2fs_convert_inline_inode(inode);
618 err = truncate_blocks(inode, i_size_read(inode), true);
622 inode->i_mtime = inode->i_ctime = current_time(inode);
623 f2fs_mark_inode_dirty_sync(inode, false);
627 int f2fs_getattr(const struct path *path, struct kstat *stat,
628 u32 request_mask, unsigned int query_flags)
630 struct inode *inode = d_inode(path->dentry);
631 struct f2fs_inode_info *fi = F2FS_I(inode);
634 flags = fi->i_flags & FS_FL_USER_VISIBLE;
635 if (flags & FS_APPEND_FL)
636 stat->attributes |= STATX_ATTR_APPEND;
637 if (flags & FS_COMPR_FL)
638 stat->attributes |= STATX_ATTR_COMPRESSED;
639 if (f2fs_encrypted_inode(inode))
640 stat->attributes |= STATX_ATTR_ENCRYPTED;
641 if (flags & FS_IMMUTABLE_FL)
642 stat->attributes |= STATX_ATTR_IMMUTABLE;
643 if (flags & FS_NODUMP_FL)
644 stat->attributes |= STATX_ATTR_NODUMP;
646 stat->attributes_mask |= (STATX_ATTR_APPEND |
647 STATX_ATTR_COMPRESSED |
648 STATX_ATTR_ENCRYPTED |
649 STATX_ATTR_IMMUTABLE |
652 generic_fillattr(inode, stat);
657 #ifdef CONFIG_F2FS_FS_POSIX_ACL
658 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
660 unsigned int ia_valid = attr->ia_valid;
662 if (ia_valid & ATTR_UID)
663 inode->i_uid = attr->ia_uid;
664 if (ia_valid & ATTR_GID)
665 inode->i_gid = attr->ia_gid;
666 if (ia_valid & ATTR_ATIME)
667 inode->i_atime = timespec_trunc(attr->ia_atime,
668 inode->i_sb->s_time_gran);
669 if (ia_valid & ATTR_MTIME)
670 inode->i_mtime = timespec_trunc(attr->ia_mtime,
671 inode->i_sb->s_time_gran);
672 if (ia_valid & ATTR_CTIME)
673 inode->i_ctime = timespec_trunc(attr->ia_ctime,
674 inode->i_sb->s_time_gran);
675 if (ia_valid & ATTR_MODE) {
676 umode_t mode = attr->ia_mode;
678 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
680 set_acl_inode(inode, mode);
684 #define __setattr_copy setattr_copy
687 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
689 struct inode *inode = d_inode(dentry);
691 bool size_changed = false;
693 err = setattr_prepare(dentry, attr);
697 if (attr->ia_valid & ATTR_SIZE) {
698 if (f2fs_encrypted_inode(inode) &&
699 fscrypt_get_encryption_info(inode))
702 if (attr->ia_size <= i_size_read(inode)) {
703 truncate_setsize(inode, attr->ia_size);
704 err = f2fs_truncate(inode);
709 * do not trim all blocks after i_size if target size is
710 * larger than i_size.
712 truncate_setsize(inode, attr->ia_size);
714 /* should convert inline inode here */
715 if (!f2fs_may_inline_data(inode)) {
716 err = f2fs_convert_inline_inode(inode);
720 inode->i_mtime = inode->i_ctime = current_time(inode);
726 __setattr_copy(inode, attr);
728 if (attr->ia_valid & ATTR_MODE) {
729 err = posix_acl_chmod(inode, get_inode_mode(inode));
730 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
731 inode->i_mode = F2FS_I(inode)->i_acl_mode;
732 clear_inode_flag(inode, FI_ACL_MODE);
736 /* file size may changed here */
737 f2fs_mark_inode_dirty_sync(inode, size_changed);
739 /* inode change will produce dirty node pages flushed by checkpoint */
740 f2fs_balance_fs(F2FS_I_SB(inode), true);
745 const struct inode_operations f2fs_file_inode_operations = {
746 .getattr = f2fs_getattr,
747 .setattr = f2fs_setattr,
748 .get_acl = f2fs_get_acl,
749 .set_acl = f2fs_set_acl,
750 #ifdef CONFIG_F2FS_FS_XATTR
751 .listxattr = f2fs_listxattr,
753 .fiemap = f2fs_fiemap,
756 static int fill_zero(struct inode *inode, pgoff_t index,
757 loff_t start, loff_t len)
759 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
765 f2fs_balance_fs(sbi, true);
768 page = get_new_data_page(inode, NULL, index, false);
772 return PTR_ERR(page);
774 f2fs_wait_on_page_writeback(page, DATA, true);
775 zero_user(page, start, len);
776 set_page_dirty(page);
777 f2fs_put_page(page, 1);
781 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
785 while (pg_start < pg_end) {
786 struct dnode_of_data dn;
787 pgoff_t end_offset, count;
789 set_new_dnode(&dn, inode, NULL, NULL, 0);
790 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
792 if (err == -ENOENT) {
799 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
800 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
802 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
804 truncate_data_blocks_range(&dn, count);
812 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
814 pgoff_t pg_start, pg_end;
815 loff_t off_start, off_end;
818 ret = f2fs_convert_inline_inode(inode);
822 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
823 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
825 off_start = offset & (PAGE_SIZE - 1);
826 off_end = (offset + len) & (PAGE_SIZE - 1);
828 if (pg_start == pg_end) {
829 ret = fill_zero(inode, pg_start, off_start,
830 off_end - off_start);
835 ret = fill_zero(inode, pg_start++, off_start,
836 PAGE_SIZE - off_start);
841 ret = fill_zero(inode, pg_end, 0, off_end);
846 if (pg_start < pg_end) {
847 struct address_space *mapping = inode->i_mapping;
848 loff_t blk_start, blk_end;
849 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
851 f2fs_balance_fs(sbi, true);
853 blk_start = (loff_t)pg_start << PAGE_SHIFT;
854 blk_end = (loff_t)pg_end << PAGE_SHIFT;
855 truncate_inode_pages_range(mapping, blk_start,
859 ret = truncate_hole(inode, pg_start, pg_end);
867 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
868 int *do_replace, pgoff_t off, pgoff_t len)
870 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
871 struct dnode_of_data dn;
875 set_new_dnode(&dn, inode, NULL, NULL, 0);
876 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
877 if (ret && ret != -ENOENT) {
879 } else if (ret == -ENOENT) {
880 if (dn.max_level == 0)
882 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
888 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
889 dn.ofs_in_node, len);
890 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
891 *blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
892 if (!is_checkpointed_data(sbi, *blkaddr)) {
894 if (test_opt(sbi, LFS)) {
899 /* do not invalidate this block address */
900 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
913 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
914 int *do_replace, pgoff_t off, int len)
916 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
917 struct dnode_of_data dn;
920 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
921 if (*do_replace == 0)
924 set_new_dnode(&dn, inode, NULL, NULL, 0);
925 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
927 dec_valid_block_count(sbi, inode, 1);
928 invalidate_blocks(sbi, *blkaddr);
930 f2fs_update_data_blkaddr(&dn, *blkaddr);
937 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
938 block_t *blkaddr, int *do_replace,
939 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
941 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
946 if (blkaddr[i] == NULL_ADDR && !full) {
951 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
952 struct dnode_of_data dn;
957 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
958 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
962 get_node_info(sbi, dn.nid, &ni);
964 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
965 dn.ofs_in_node, len - i);
967 dn.data_blkaddr = datablock_addr(dn.node_page,
969 truncate_data_blocks_range(&dn, 1);
972 f2fs_i_blocks_write(src_inode,
974 f2fs_i_blocks_write(dst_inode,
976 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
977 blkaddr[i], ni.version, true, false);
983 new_size = (dst + i) << PAGE_SHIFT;
984 if (dst_inode->i_size < new_size)
985 f2fs_i_size_write(dst_inode, new_size);
986 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
990 struct page *psrc, *pdst;
992 psrc = get_lock_data_page(src_inode, src + i, true);
994 return PTR_ERR(psrc);
995 pdst = get_new_data_page(dst_inode, NULL, dst + i,
998 f2fs_put_page(psrc, 1);
999 return PTR_ERR(pdst);
1001 f2fs_copy_page(psrc, pdst);
1002 set_page_dirty(pdst);
1003 f2fs_put_page(pdst, 1);
1004 f2fs_put_page(psrc, 1);
1006 ret = truncate_hole(src_inode, src + i, src + i + 1);
1015 static int __exchange_data_block(struct inode *src_inode,
1016 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1017 pgoff_t len, bool full)
1019 block_t *src_blkaddr;
1025 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1027 src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1031 do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1033 kvfree(src_blkaddr);
1037 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1038 do_replace, src, olen);
1042 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1043 do_replace, src, dst, olen, full);
1051 kvfree(src_blkaddr);
1057 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1058 kvfree(src_blkaddr);
1063 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1065 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1066 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1069 f2fs_balance_fs(sbi, true);
1072 f2fs_drop_extent_tree(inode);
1074 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1075 f2fs_unlock_op(sbi);
1079 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1081 pgoff_t pg_start, pg_end;
1085 if (offset + len >= i_size_read(inode))
1088 /* collapse range should be aligned to block size of f2fs. */
1089 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1092 ret = f2fs_convert_inline_inode(inode);
1096 pg_start = offset >> PAGE_SHIFT;
1097 pg_end = (offset + len) >> PAGE_SHIFT;
1099 /* write out all dirty pages from offset */
1100 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1104 truncate_pagecache(inode, offset);
1106 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1110 /* write out all moved pages, if possible */
1111 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1112 truncate_pagecache(inode, offset);
1114 new_size = i_size_read(inode) - len;
1115 truncate_pagecache(inode, new_size);
1117 ret = truncate_blocks(inode, new_size, true);
1119 f2fs_i_size_write(inode, new_size);
1124 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1127 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1128 pgoff_t index = start;
1129 unsigned int ofs_in_node = dn->ofs_in_node;
1133 for (; index < end; index++, dn->ofs_in_node++) {
1134 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1138 dn->ofs_in_node = ofs_in_node;
1139 ret = reserve_new_blocks(dn, count);
1143 dn->ofs_in_node = ofs_in_node;
1144 for (index = start; index < end; index++, dn->ofs_in_node++) {
1146 datablock_addr(dn->node_page, dn->ofs_in_node);
1148 * reserve_new_blocks will not guarantee entire block
1151 if (dn->data_blkaddr == NULL_ADDR) {
1155 if (dn->data_blkaddr != NEW_ADDR) {
1156 invalidate_blocks(sbi, dn->data_blkaddr);
1157 dn->data_blkaddr = NEW_ADDR;
1158 set_data_blkaddr(dn);
1162 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1167 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1170 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1171 struct address_space *mapping = inode->i_mapping;
1172 pgoff_t index, pg_start, pg_end;
1173 loff_t new_size = i_size_read(inode);
1174 loff_t off_start, off_end;
1177 ret = inode_newsize_ok(inode, (len + offset));
1181 ret = f2fs_convert_inline_inode(inode);
1185 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1189 truncate_pagecache_range(inode, offset, offset + len - 1);
1191 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1192 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1194 off_start = offset & (PAGE_SIZE - 1);
1195 off_end = (offset + len) & (PAGE_SIZE - 1);
1197 if (pg_start == pg_end) {
1198 ret = fill_zero(inode, pg_start, off_start,
1199 off_end - off_start);
1203 new_size = max_t(loff_t, new_size, offset + len);
1206 ret = fill_zero(inode, pg_start++, off_start,
1207 PAGE_SIZE - off_start);
1211 new_size = max_t(loff_t, new_size,
1212 (loff_t)pg_start << PAGE_SHIFT);
1215 for (index = pg_start; index < pg_end;) {
1216 struct dnode_of_data dn;
1217 unsigned int end_offset;
1222 set_new_dnode(&dn, inode, NULL, NULL, 0);
1223 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1225 f2fs_unlock_op(sbi);
1229 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1230 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1232 ret = f2fs_do_zero_range(&dn, index, end);
1233 f2fs_put_dnode(&dn);
1234 f2fs_unlock_op(sbi);
1236 f2fs_balance_fs(sbi, dn.node_changed);
1242 new_size = max_t(loff_t, new_size,
1243 (loff_t)index << PAGE_SHIFT);
1247 ret = fill_zero(inode, pg_end, 0, off_end);
1251 new_size = max_t(loff_t, new_size, offset + len);
1256 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1257 f2fs_i_size_write(inode, new_size);
1262 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1264 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1265 pgoff_t nr, pg_start, pg_end, delta, idx;
1269 new_size = i_size_read(inode) + len;
1270 ret = inode_newsize_ok(inode, new_size);
1274 if (offset >= i_size_read(inode))
1277 /* insert range should be aligned to block size of f2fs. */
1278 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1281 ret = f2fs_convert_inline_inode(inode);
1285 f2fs_balance_fs(sbi, true);
1287 ret = truncate_blocks(inode, i_size_read(inode), true);
1291 /* write out all dirty pages from offset */
1292 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1296 truncate_pagecache(inode, offset);
1298 pg_start = offset >> PAGE_SHIFT;
1299 pg_end = (offset + len) >> PAGE_SHIFT;
1300 delta = pg_end - pg_start;
1301 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1303 while (!ret && idx > pg_start) {
1304 nr = idx - pg_start;
1310 f2fs_drop_extent_tree(inode);
1312 ret = __exchange_data_block(inode, inode, idx,
1313 idx + delta, nr, false);
1314 f2fs_unlock_op(sbi);
1317 /* write out all moved pages, if possible */
1318 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1319 truncate_pagecache(inode, offset);
1322 f2fs_i_size_write(inode, new_size);
1326 static int expand_inode_data(struct inode *inode, loff_t offset,
1327 loff_t len, int mode)
1329 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1330 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1332 loff_t new_size = i_size_read(inode);
1336 err = inode_newsize_ok(inode, (len + offset));
1340 err = f2fs_convert_inline_inode(inode);
1344 f2fs_balance_fs(sbi, true);
1346 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1347 off_end = (offset + len) & (PAGE_SIZE - 1);
1349 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1350 map.m_len = pg_end - map.m_lblk;
1354 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1361 last_off = map.m_lblk + map.m_len - 1;
1363 /* update new size to the failed position */
1364 new_size = (last_off == pg_end) ? offset + len:
1365 (loff_t)(last_off + 1) << PAGE_SHIFT;
1367 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1370 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1371 f2fs_i_size_write(inode, new_size);
1376 static long f2fs_fallocate(struct file *file, int mode,
1377 loff_t offset, loff_t len)
1379 struct inode *inode = file_inode(file);
1382 /* f2fs only support ->fallocate for regular file */
1383 if (!S_ISREG(inode->i_mode))
1386 if (f2fs_encrypted_inode(inode) &&
1387 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1390 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1391 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1392 FALLOC_FL_INSERT_RANGE))
1397 if (mode & FALLOC_FL_PUNCH_HOLE) {
1398 if (offset >= inode->i_size)
1401 ret = punch_hole(inode, offset, len);
1402 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1403 ret = f2fs_collapse_range(inode, offset, len);
1404 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1405 ret = f2fs_zero_range(inode, offset, len, mode);
1406 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1407 ret = f2fs_insert_range(inode, offset, len);
1409 ret = expand_inode_data(inode, offset, len, mode);
1413 inode->i_mtime = inode->i_ctime = current_time(inode);
1414 f2fs_mark_inode_dirty_sync(inode, false);
1415 if (mode & FALLOC_FL_KEEP_SIZE)
1416 file_set_keep_isize(inode);
1417 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1421 inode_unlock(inode);
1423 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1427 static int f2fs_release_file(struct inode *inode, struct file *filp)
1430 * f2fs_relase_file is called at every close calls. So we should
1431 * not drop any inmemory pages by close called by other process.
1433 if (!(filp->f_mode & FMODE_WRITE) ||
1434 atomic_read(&inode->i_writecount) != 1)
1437 /* some remained atomic pages should discarded */
1438 if (f2fs_is_atomic_file(inode))
1439 drop_inmem_pages(inode);
1440 if (f2fs_is_volatile_file(inode)) {
1441 clear_inode_flag(inode, FI_VOLATILE_FILE);
1442 stat_dec_volatile_write(inode);
1443 set_inode_flag(inode, FI_DROP_CACHE);
1444 filemap_fdatawrite(inode->i_mapping);
1445 clear_inode_flag(inode, FI_DROP_CACHE);
1450 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1451 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1453 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1457 else if (S_ISREG(mode))
1458 return flags & F2FS_REG_FLMASK;
1460 return flags & F2FS_OTHER_FLMASK;
1463 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1465 struct inode *inode = file_inode(filp);
1466 struct f2fs_inode_info *fi = F2FS_I(inode);
1467 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1468 return put_user(flags, (int __user *)arg);
1471 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1473 struct inode *inode = file_inode(filp);
1474 struct f2fs_inode_info *fi = F2FS_I(inode);
1476 unsigned int oldflags;
1479 if (!inode_owner_or_capable(inode))
1482 if (get_user(flags, (int __user *)arg))
1485 ret = mnt_want_write_file(filp);
1491 flags = f2fs_mask_flags(inode->i_mode, flags);
1493 oldflags = fi->i_flags;
1495 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1496 if (!capable(CAP_LINUX_IMMUTABLE)) {
1497 inode_unlock(inode);
1503 flags = flags & FS_FL_USER_MODIFIABLE;
1504 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1505 fi->i_flags = flags;
1507 inode->i_ctime = current_time(inode);
1508 f2fs_set_inode_flags(inode);
1509 f2fs_mark_inode_dirty_sync(inode, false);
1511 inode_unlock(inode);
1513 mnt_drop_write_file(filp);
1517 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1519 struct inode *inode = file_inode(filp);
1521 return put_user(inode->i_generation, (int __user *)arg);
1524 static int f2fs_ioc_start_atomic_write(struct file *filp)
1526 struct inode *inode = file_inode(filp);
1529 if (!inode_owner_or_capable(inode))
1532 if (!S_ISREG(inode->i_mode))
1535 ret = mnt_want_write_file(filp);
1541 if (f2fs_is_atomic_file(inode))
1544 ret = f2fs_convert_inline_inode(inode);
1548 set_inode_flag(inode, FI_ATOMIC_FILE);
1549 set_inode_flag(inode, FI_HOT_DATA);
1550 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1552 if (!get_dirty_pages(inode))
1555 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1556 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1557 inode->i_ino, get_dirty_pages(inode));
1558 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1560 clear_inode_flag(inode, FI_ATOMIC_FILE);
1565 stat_inc_atomic_write(inode);
1566 stat_update_max_atomic_write(inode);
1568 inode_unlock(inode);
1569 mnt_drop_write_file(filp);
1573 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1575 struct inode *inode = file_inode(filp);
1578 if (!inode_owner_or_capable(inode))
1581 ret = mnt_want_write_file(filp);
1587 if (f2fs_is_volatile_file(inode))
1590 if (f2fs_is_atomic_file(inode)) {
1591 ret = commit_inmem_pages(inode);
1595 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1597 clear_inode_flag(inode, FI_ATOMIC_FILE);
1598 stat_dec_atomic_write(inode);
1601 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1604 inode_unlock(inode);
1605 mnt_drop_write_file(filp);
1609 static int f2fs_ioc_start_volatile_write(struct file *filp)
1611 struct inode *inode = file_inode(filp);
1614 if (!inode_owner_or_capable(inode))
1617 if (!S_ISREG(inode->i_mode))
1620 ret = mnt_want_write_file(filp);
1626 if (f2fs_is_volatile_file(inode))
1629 ret = f2fs_convert_inline_inode(inode);
1633 stat_inc_volatile_write(inode);
1634 stat_update_max_volatile_write(inode);
1636 set_inode_flag(inode, FI_VOLATILE_FILE);
1637 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1639 inode_unlock(inode);
1640 mnt_drop_write_file(filp);
1644 static int f2fs_ioc_release_volatile_write(struct file *filp)
1646 struct inode *inode = file_inode(filp);
1649 if (!inode_owner_or_capable(inode))
1652 ret = mnt_want_write_file(filp);
1658 if (!f2fs_is_volatile_file(inode))
1661 if (!f2fs_is_first_block_written(inode)) {
1662 ret = truncate_partial_data_page(inode, 0, true);
1666 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1668 inode_unlock(inode);
1669 mnt_drop_write_file(filp);
1673 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1675 struct inode *inode = file_inode(filp);
1678 if (!inode_owner_or_capable(inode))
1681 ret = mnt_want_write_file(filp);
1687 if (f2fs_is_atomic_file(inode))
1688 drop_inmem_pages(inode);
1689 if (f2fs_is_volatile_file(inode)) {
1690 clear_inode_flag(inode, FI_VOLATILE_FILE);
1691 stat_dec_volatile_write(inode);
1692 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1695 inode_unlock(inode);
1697 mnt_drop_write_file(filp);
1698 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1702 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1704 struct inode *inode = file_inode(filp);
1705 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1706 struct super_block *sb = sbi->sb;
1710 if (!capable(CAP_SYS_ADMIN))
1713 if (get_user(in, (__u32 __user *)arg))
1716 ret = mnt_want_write_file(filp);
1721 case F2FS_GOING_DOWN_FULLSYNC:
1722 sb = freeze_bdev(sb->s_bdev);
1723 if (sb && !IS_ERR(sb)) {
1724 f2fs_stop_checkpoint(sbi, false);
1725 thaw_bdev(sb->s_bdev, sb);
1728 case F2FS_GOING_DOWN_METASYNC:
1729 /* do checkpoint only */
1730 f2fs_sync_fs(sb, 1);
1731 f2fs_stop_checkpoint(sbi, false);
1733 case F2FS_GOING_DOWN_NOSYNC:
1734 f2fs_stop_checkpoint(sbi, false);
1736 case F2FS_GOING_DOWN_METAFLUSH:
1737 sync_meta_pages(sbi, META, LONG_MAX);
1738 f2fs_stop_checkpoint(sbi, false);
1744 f2fs_update_time(sbi, REQ_TIME);
1746 mnt_drop_write_file(filp);
1750 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1752 struct inode *inode = file_inode(filp);
1753 struct super_block *sb = inode->i_sb;
1754 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1755 struct fstrim_range range;
1758 if (!capable(CAP_SYS_ADMIN))
1761 if (!blk_queue_discard(q))
1764 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1768 ret = mnt_want_write_file(filp);
1772 range.minlen = max((unsigned int)range.minlen,
1773 q->limits.discard_granularity);
1774 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1775 mnt_drop_write_file(filp);
1779 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1782 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1786 static bool uuid_is_nonzero(__u8 u[16])
1790 for (i = 0; i < 16; i++)
1796 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1798 struct inode *inode = file_inode(filp);
1800 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1802 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1805 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1807 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1810 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1812 struct inode *inode = file_inode(filp);
1813 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1816 if (!f2fs_sb_has_crypto(inode->i_sb))
1819 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1822 err = mnt_want_write_file(filp);
1826 /* update superblock with uuid */
1827 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1829 err = f2fs_commit_super(sbi, false);
1832 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1833 mnt_drop_write_file(filp);
1836 mnt_drop_write_file(filp);
1838 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1844 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1846 struct inode *inode = file_inode(filp);
1847 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1851 if (!capable(CAP_SYS_ADMIN))
1854 if (get_user(sync, (__u32 __user *)arg))
1857 if (f2fs_readonly(sbi->sb))
1860 ret = mnt_want_write_file(filp);
1865 if (!mutex_trylock(&sbi->gc_mutex)) {
1870 mutex_lock(&sbi->gc_mutex);
1873 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
1875 mnt_drop_write_file(filp);
1879 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1881 struct inode *inode = file_inode(filp);
1882 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1885 if (!capable(CAP_SYS_ADMIN))
1888 if (f2fs_readonly(sbi->sb))
1891 ret = mnt_want_write_file(filp);
1895 ret = f2fs_sync_fs(sbi->sb, 1);
1897 mnt_drop_write_file(filp);
1901 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1903 struct f2fs_defragment *range)
1905 struct inode *inode = file_inode(filp);
1906 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1907 struct extent_info ei = {0,0,0};
1908 pgoff_t pg_start, pg_end;
1909 unsigned int blk_per_seg = sbi->blocks_per_seg;
1910 unsigned int total = 0, sec_num;
1911 block_t blk_end = 0;
1912 bool fragmented = false;
1915 /* if in-place-update policy is enabled, don't waste time here */
1916 if (need_inplace_update_policy(inode, NULL))
1919 pg_start = range->start >> PAGE_SHIFT;
1920 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1922 f2fs_balance_fs(sbi, true);
1926 /* writeback all dirty pages in the range */
1927 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1928 range->start + range->len - 1);
1933 * lookup mapping info in extent cache, skip defragmenting if physical
1934 * block addresses are continuous.
1936 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1937 if (ei.fofs + ei.len >= pg_end)
1941 map.m_lblk = pg_start;
1944 * lookup mapping info in dnode page cache, skip defragmenting if all
1945 * physical block addresses are continuous even if there are hole(s)
1946 * in logical blocks.
1948 while (map.m_lblk < pg_end) {
1949 map.m_len = pg_end - map.m_lblk;
1950 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1954 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1959 if (blk_end && blk_end != map.m_pblk) {
1963 blk_end = map.m_pblk + map.m_len;
1965 map.m_lblk += map.m_len;
1971 map.m_lblk = pg_start;
1972 map.m_len = pg_end - pg_start;
1974 sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
1977 * make sure there are enough free section for LFS allocation, this can
1978 * avoid defragment running in SSR mode when free section are allocated
1981 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
1986 while (map.m_lblk < pg_end) {
1991 map.m_len = pg_end - map.m_lblk;
1992 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1996 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2001 set_inode_flag(inode, FI_DO_DEFRAG);
2004 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2007 page = get_lock_data_page(inode, idx, true);
2009 err = PTR_ERR(page);
2013 set_page_dirty(page);
2014 f2fs_put_page(page, 1);
2023 if (idx < pg_end && cnt < blk_per_seg)
2026 clear_inode_flag(inode, FI_DO_DEFRAG);
2028 err = filemap_fdatawrite(inode->i_mapping);
2033 clear_inode_flag(inode, FI_DO_DEFRAG);
2035 inode_unlock(inode);
2037 range->len = (u64)total << PAGE_SHIFT;
2041 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2043 struct inode *inode = file_inode(filp);
2044 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2045 struct f2fs_defragment range;
2048 if (!capable(CAP_SYS_ADMIN))
2051 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2054 if (f2fs_readonly(sbi->sb))
2057 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2061 /* verify alignment of offset & size */
2062 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2065 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2066 sbi->max_file_blocks))
2069 err = mnt_want_write_file(filp);
2073 err = f2fs_defragment_range(sbi, filp, &range);
2074 mnt_drop_write_file(filp);
2076 f2fs_update_time(sbi, REQ_TIME);
2080 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2087 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2088 struct file *file_out, loff_t pos_out, size_t len)
2090 struct inode *src = file_inode(file_in);
2091 struct inode *dst = file_inode(file_out);
2092 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2093 size_t olen = len, dst_max_i_size = 0;
2097 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2098 src->i_sb != dst->i_sb)
2101 if (unlikely(f2fs_readonly(src->i_sb)))
2104 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2107 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2111 if (pos_in == pos_out)
2113 if (pos_out > pos_in && pos_out < pos_in + len)
2119 if (!inode_trylock(dst)) {
2126 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2129 olen = len = src->i_size - pos_in;
2130 if (pos_in + len == src->i_size)
2131 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2137 dst_osize = dst->i_size;
2138 if (pos_out + olen > dst->i_size)
2139 dst_max_i_size = pos_out + olen;
2141 /* verify the end result is block aligned */
2142 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2143 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2144 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2147 ret = f2fs_convert_inline_inode(src);
2151 ret = f2fs_convert_inline_inode(dst);
2155 /* write out all dirty pages from offset */
2156 ret = filemap_write_and_wait_range(src->i_mapping,
2157 pos_in, pos_in + len);
2161 ret = filemap_write_and_wait_range(dst->i_mapping,
2162 pos_out, pos_out + len);
2166 f2fs_balance_fs(sbi, true);
2168 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2169 pos_out >> F2FS_BLKSIZE_BITS,
2170 len >> F2FS_BLKSIZE_BITS, false);
2174 f2fs_i_size_write(dst, dst_max_i_size);
2175 else if (dst_osize != dst->i_size)
2176 f2fs_i_size_write(dst, dst_osize);
2178 f2fs_unlock_op(sbi);
2187 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2189 struct f2fs_move_range range;
2193 if (!(filp->f_mode & FMODE_READ) ||
2194 !(filp->f_mode & FMODE_WRITE))
2197 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2201 dst = fdget(range.dst_fd);
2205 if (!(dst.file->f_mode & FMODE_WRITE)) {
2210 err = mnt_want_write_file(filp);
2214 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2215 range.pos_out, range.len);
2217 mnt_drop_write_file(filp);
2221 if (copy_to_user((struct f2fs_move_range __user *)arg,
2222 &range, sizeof(range)))
2229 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2231 struct inode *inode = file_inode(filp);
2232 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2233 struct sit_info *sm = SIT_I(sbi);
2234 unsigned int start_segno = 0, end_segno = 0;
2235 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2236 struct f2fs_flush_device range;
2239 if (!capable(CAP_SYS_ADMIN))
2242 if (f2fs_readonly(sbi->sb))
2245 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2249 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2250 sbi->segs_per_sec != 1) {
2251 f2fs_msg(sbi->sb, KERN_WARNING,
2252 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2253 range.dev_num, sbi->s_ndevs,
2258 ret = mnt_want_write_file(filp);
2262 if (range.dev_num != 0)
2263 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2264 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2266 start_segno = sm->last_victim[FLUSH_DEVICE];
2267 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2268 start_segno = dev_start_segno;
2269 end_segno = min(start_segno + range.segments, dev_end_segno);
2271 while (start_segno < end_segno) {
2272 if (!mutex_trylock(&sbi->gc_mutex)) {
2276 sm->last_victim[GC_CB] = end_segno + 1;
2277 sm->last_victim[GC_GREEDY] = end_segno + 1;
2278 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2279 ret = f2fs_gc(sbi, true, true, start_segno);
2287 mnt_drop_write_file(filp);
2292 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2295 case F2FS_IOC_GETFLAGS:
2296 return f2fs_ioc_getflags(filp, arg);
2297 case F2FS_IOC_SETFLAGS:
2298 return f2fs_ioc_setflags(filp, arg);
2299 case F2FS_IOC_GETVERSION:
2300 return f2fs_ioc_getversion(filp, arg);
2301 case F2FS_IOC_START_ATOMIC_WRITE:
2302 return f2fs_ioc_start_atomic_write(filp);
2303 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2304 return f2fs_ioc_commit_atomic_write(filp);
2305 case F2FS_IOC_START_VOLATILE_WRITE:
2306 return f2fs_ioc_start_volatile_write(filp);
2307 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2308 return f2fs_ioc_release_volatile_write(filp);
2309 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2310 return f2fs_ioc_abort_volatile_write(filp);
2311 case F2FS_IOC_SHUTDOWN:
2312 return f2fs_ioc_shutdown(filp, arg);
2314 return f2fs_ioc_fitrim(filp, arg);
2315 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2316 return f2fs_ioc_set_encryption_policy(filp, arg);
2317 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2318 return f2fs_ioc_get_encryption_policy(filp, arg);
2319 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2320 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2321 case F2FS_IOC_GARBAGE_COLLECT:
2322 return f2fs_ioc_gc(filp, arg);
2323 case F2FS_IOC_WRITE_CHECKPOINT:
2324 return f2fs_ioc_write_checkpoint(filp, arg);
2325 case F2FS_IOC_DEFRAGMENT:
2326 return f2fs_ioc_defragment(filp, arg);
2327 case F2FS_IOC_MOVE_RANGE:
2328 return f2fs_ioc_move_range(filp, arg);
2329 case F2FS_IOC_FLUSH_DEVICE:
2330 return f2fs_ioc_flush_device(filp, arg);
2336 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2338 struct file *file = iocb->ki_filp;
2339 struct inode *inode = file_inode(file);
2340 struct blk_plug plug;
2343 if (f2fs_encrypted_inode(inode) &&
2344 !fscrypt_has_encryption_key(inode) &&
2345 fscrypt_get_encryption_info(inode))
2349 ret = generic_write_checks(iocb, from);
2353 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2354 set_inode_flag(inode, FI_NO_PREALLOC);
2356 err = f2fs_preallocate_blocks(iocb, from);
2358 inode_unlock(inode);
2361 blk_start_plug(&plug);
2362 ret = __generic_file_write_iter(iocb, from);
2363 blk_finish_plug(&plug);
2364 clear_inode_flag(inode, FI_NO_PREALLOC);
2366 inode_unlock(inode);
2369 ret = generic_write_sync(iocb, ret);
2373 #ifdef CONFIG_COMPAT
2374 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2377 case F2FS_IOC32_GETFLAGS:
2378 cmd = F2FS_IOC_GETFLAGS;
2380 case F2FS_IOC32_SETFLAGS:
2381 cmd = F2FS_IOC_SETFLAGS;
2383 case F2FS_IOC32_GETVERSION:
2384 cmd = F2FS_IOC_GETVERSION;
2386 case F2FS_IOC_START_ATOMIC_WRITE:
2387 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2388 case F2FS_IOC_START_VOLATILE_WRITE:
2389 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2390 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2391 case F2FS_IOC_SHUTDOWN:
2392 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2393 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2394 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2395 case F2FS_IOC_GARBAGE_COLLECT:
2396 case F2FS_IOC_WRITE_CHECKPOINT:
2397 case F2FS_IOC_DEFRAGMENT:
2398 case F2FS_IOC_MOVE_RANGE:
2399 case F2FS_IOC_FLUSH_DEVICE:
2402 return -ENOIOCTLCMD;
2404 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2408 const struct file_operations f2fs_file_operations = {
2409 .llseek = f2fs_llseek,
2410 .read_iter = generic_file_read_iter,
2411 .write_iter = f2fs_file_write_iter,
2412 .open = f2fs_file_open,
2413 .release = f2fs_release_file,
2414 .mmap = f2fs_file_mmap,
2415 .fsync = f2fs_sync_file,
2416 .fallocate = f2fs_fallocate,
2417 .unlocked_ioctl = f2fs_ioctl,
2418 #ifdef CONFIG_COMPAT
2419 .compat_ioctl = f2fs_compat_ioctl,
2421 .splice_read = generic_file_splice_read,
2422 .splice_write = iter_file_splice_write,