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/random.h>
32 #include <trace/events/f2fs.h>
34 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
37 struct page *page = vmf->page;
38 struct inode *inode = file_inode(vma->vm_file);
39 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
40 struct dnode_of_data dn;
45 sb_start_pagefault(inode->i_sb);
47 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
49 /* block allocation */
51 set_new_dnode(&dn, inode, NULL, NULL, 0);
52 err = f2fs_reserve_block(&dn, page->index);
60 file_update_time(vma->vm_file);
62 if (unlikely(page->mapping != inode->i_mapping ||
63 page_offset(page) > i_size_read(inode) ||
64 !PageUptodate(page))) {
71 * check to see if the page is mapped already (no holes)
73 if (PageMappedToDisk(page))
76 /* page is wholly or partially inside EOF */
77 if (((loff_t)(page->index + 1) << PAGE_CACHE_SHIFT) >
80 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
81 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
84 SetPageUptodate(page);
86 trace_f2fs_vm_page_mkwrite(page, DATA);
89 f2fs_wait_on_page_writeback(page, DATA);
91 /* wait for GCed encrypted page writeback */
92 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
93 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
95 /* if gced page is attached, don't write to cold segment */
96 clear_cold_data(page);
98 sb_end_pagefault(inode->i_sb);
99 return block_page_mkwrite_return(err);
102 static const struct vm_operations_struct f2fs_file_vm_ops = {
103 .fault = filemap_fault,
104 .map_pages = filemap_map_pages,
105 .page_mkwrite = f2fs_vm_page_mkwrite,
108 static int get_parent_ino(struct inode *inode, nid_t *pino)
110 struct dentry *dentry;
112 inode = igrab(inode);
113 dentry = d_find_any_alias(inode);
118 if (update_dent_inode(inode, inode, &dentry->d_name)) {
123 *pino = parent_ino(dentry);
128 static inline bool need_do_checkpoint(struct inode *inode)
130 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
131 bool need_cp = false;
133 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
135 else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
137 else if (file_wrong_pino(inode))
139 else if (!space_for_roll_forward(sbi))
141 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
143 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
145 else if (test_opt(sbi, FASTBOOT))
147 else if (sbi->active_logs == 2)
153 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
155 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
157 /* But we need to avoid that there are some inode updates */
158 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
164 static void try_to_fix_pino(struct inode *inode)
166 struct f2fs_inode_info *fi = F2FS_I(inode);
169 down_write(&fi->i_sem);
171 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
172 get_parent_ino(inode, &pino)) {
174 file_got_pino(inode);
175 up_write(&fi->i_sem);
177 mark_inode_dirty_sync(inode);
178 f2fs_write_inode(inode, NULL);
180 up_write(&fi->i_sem);
184 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
186 struct inode *inode = file->f_mapping->host;
187 struct f2fs_inode_info *fi = F2FS_I(inode);
188 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
189 nid_t ino = inode->i_ino;
191 bool need_cp = false;
192 struct writeback_control wbc = {
193 .sync_mode = WB_SYNC_ALL,
194 .nr_to_write = LONG_MAX,
198 if (unlikely(f2fs_readonly(inode->i_sb)))
201 trace_f2fs_sync_file_enter(inode);
203 if (S_ISDIR(inode->i_mode))
206 /* if fdatasync is triggered, let's do in-place-update */
207 if (get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
208 set_inode_flag(fi, FI_NEED_IPU);
209 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
210 clear_inode_flag(fi, FI_NEED_IPU);
213 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
217 /* if the inode is dirty, let's recover all the time */
219 f2fs_write_inode(inode, NULL);
224 * if there is no written data, don't waste time to write recovery info.
226 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
227 !exist_written_data(sbi, ino, APPEND_INO)) {
229 /* it may call write_inode just prior to fsync */
230 if (need_inode_page_update(sbi, ino))
233 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
234 exist_written_data(sbi, ino, UPDATE_INO))
239 /* guarantee free sections for fsync */
240 f2fs_balance_fs(sbi);
243 * Both of fdatasync() and fsync() are able to be recovered from
246 down_read(&fi->i_sem);
247 need_cp = need_do_checkpoint(inode);
251 /* all the dirty node pages should be flushed for POR */
252 ret = f2fs_sync_fs(inode->i_sb, 1);
255 * We've secured consistency through sync_fs. Following pino
256 * will be used only for fsynced inodes after checkpoint.
258 try_to_fix_pino(inode);
259 clear_inode_flag(fi, FI_APPEND_WRITE);
260 clear_inode_flag(fi, FI_UPDATE_WRITE);
264 sync_node_pages(sbi, ino, &wbc);
266 /* if cp_error was enabled, we should avoid infinite loop */
267 if (unlikely(f2fs_cp_error(sbi)))
270 if (need_inode_block_update(sbi, ino)) {
271 mark_inode_dirty_sync(inode);
272 f2fs_write_inode(inode, NULL);
276 ret = wait_on_node_pages_writeback(sbi, ino);
280 /* once recovery info is written, don't need to tack this */
281 remove_dirty_inode(sbi, ino, APPEND_INO);
282 clear_inode_flag(fi, FI_APPEND_WRITE);
284 remove_dirty_inode(sbi, ino, UPDATE_INO);
285 clear_inode_flag(fi, FI_UPDATE_WRITE);
286 ret = f2fs_issue_flush(sbi);
288 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
289 f2fs_trace_ios(NULL, 1);
293 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
294 pgoff_t pgofs, int whence)
299 if (whence != SEEK_DATA)
302 /* find first dirty page index */
303 pagevec_init(&pvec, 0);
304 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
305 PAGECACHE_TAG_DIRTY, 1);
306 pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
307 pagevec_release(&pvec);
311 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
312 pgoff_t dirty, pgoff_t pgofs, int whence)
316 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
317 is_valid_data_blkaddr(sbi, blkaddr))
321 if (blkaddr == NULL_ADDR)
328 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
330 struct inode *inode = file->f_mapping->host;
331 loff_t maxbytes = inode->i_sb->s_maxbytes;
332 struct dnode_of_data dn;
333 pgoff_t pgofs, end_offset, dirty;
334 loff_t data_ofs = offset;
338 mutex_lock(&inode->i_mutex);
340 isize = i_size_read(inode);
344 /* handle inline data case */
345 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
346 if (whence == SEEK_HOLE)
351 pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
353 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
355 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
356 set_new_dnode(&dn, inode, NULL, NULL, 0);
357 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
358 if (err && err != -ENOENT) {
360 } else if (err == -ENOENT) {
361 /* direct node does not exists */
362 if (whence == SEEK_DATA) {
363 pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
371 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
373 /* find data/hole in dnode block */
374 for (; dn.ofs_in_node < end_offset;
375 dn.ofs_in_node++, pgofs++,
376 data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
378 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
380 if (__is_valid_data_blkaddr(blkaddr) &&
381 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
382 blkaddr, DATA_GENERIC)) {
387 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
396 if (whence == SEEK_DATA)
399 if (whence == SEEK_HOLE && data_ofs > isize)
401 mutex_unlock(&inode->i_mutex);
402 return vfs_setpos(file, data_ofs, maxbytes);
404 mutex_unlock(&inode->i_mutex);
408 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
410 struct inode *inode = file->f_mapping->host;
411 loff_t maxbytes = inode->i_sb->s_maxbytes;
417 return generic_file_llseek_size(file, offset, whence,
418 maxbytes, i_size_read(inode));
423 return f2fs_seek_block(file, offset, whence);
429 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
431 struct inode *inode = file_inode(file);
433 if (f2fs_encrypted_inode(inode)) {
434 int err = f2fs_get_encryption_info(inode);
439 /* we don't need to use inline_data strictly */
440 if (f2fs_has_inline_data(inode)) {
441 int err = f2fs_convert_inline_inode(inode);
447 vma->vm_ops = &f2fs_file_vm_ops;
451 static int f2fs_file_open(struct inode *inode, struct file *filp)
453 int ret = generic_file_open(inode, filp);
455 if (!ret && f2fs_encrypted_inode(inode)) {
456 ret = f2fs_get_encryption_info(inode);
463 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
465 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
466 struct f2fs_node *raw_node;
467 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
470 raw_node = F2FS_NODE(dn->node_page);
471 addr = blkaddr_in_node(raw_node) + ofs;
473 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
474 block_t blkaddr = le32_to_cpu(*addr);
475 if (blkaddr == NULL_ADDR)
478 dn->data_blkaddr = NULL_ADDR;
479 set_data_blkaddr(dn);
481 if (__is_valid_data_blkaddr(blkaddr) &&
482 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
485 invalidate_blocks(sbi, blkaddr);
486 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
487 clear_inode_flag(F2FS_I(dn->inode),
488 FI_FIRST_BLOCK_WRITTEN);
495 * once we invalidate valid blkaddr in range [ofs, ofs + count],
496 * we will invalidate all blkaddr in the whole range.
498 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
499 F2FS_I(dn->inode)) + ofs;
500 f2fs_update_extent_cache_range(dn, fofs, 0, len);
501 dec_valid_block_count(sbi, dn->inode, nr_free);
502 set_page_dirty(dn->node_page);
505 dn->ofs_in_node = ofs;
507 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
508 dn->ofs_in_node, nr_free);
512 void truncate_data_blocks(struct dnode_of_data *dn)
514 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
517 static int truncate_partial_data_page(struct inode *inode, u64 from,
520 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
521 pgoff_t index = from >> PAGE_CACHE_SHIFT;
522 struct address_space *mapping = inode->i_mapping;
525 if (!offset && !cache_only)
529 page = f2fs_grab_cache_page(mapping, index, false);
530 if (page && PageUptodate(page))
532 f2fs_put_page(page, 1);
536 page = get_lock_data_page(inode, index, true);
540 f2fs_wait_on_page_writeback(page, DATA);
541 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
542 if (!cache_only || !f2fs_encrypted_inode(inode) || !S_ISREG(inode->i_mode))
543 set_page_dirty(page);
544 f2fs_put_page(page, 1);
548 int truncate_blocks(struct inode *inode, u64 from, bool lock)
550 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
551 unsigned int blocksize = inode->i_sb->s_blocksize;
552 struct dnode_of_data dn;
554 int count = 0, err = 0;
556 bool truncate_page = false;
558 trace_f2fs_truncate_blocks_enter(inode, from);
560 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
565 ipage = get_node_page(sbi, inode->i_ino);
567 err = PTR_ERR(ipage);
571 if (f2fs_has_inline_data(inode)) {
572 if (truncate_inline_inode(ipage, from))
573 set_page_dirty(ipage);
574 f2fs_put_page(ipage, 1);
575 truncate_page = true;
579 set_new_dnode(&dn, inode, ipage, NULL, 0);
580 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
587 count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
589 count -= dn.ofs_in_node;
590 f2fs_bug_on(sbi, count < 0);
592 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
593 truncate_data_blocks_range(&dn, count);
599 err = truncate_inode_blocks(inode, free_from);
604 /* lastly zero out the first data page */
606 err = truncate_partial_data_page(inode, from, truncate_page);
608 trace_f2fs_truncate_blocks_exit(inode, err);
612 int f2fs_truncate(struct inode *inode, bool lock)
616 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
617 S_ISLNK(inode->i_mode)))
620 trace_f2fs_truncate(inode);
622 /* we should check inline_data size */
623 if (f2fs_has_inline_data(inode) && !f2fs_may_inline_data(inode)) {
624 err = f2fs_convert_inline_inode(inode);
629 err = truncate_blocks(inode, i_size_read(inode), lock);
633 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
634 mark_inode_dirty(inode);
638 int f2fs_getattr(struct vfsmount *mnt,
639 struct dentry *dentry, struct kstat *stat)
641 struct inode *inode = d_inode(dentry);
642 generic_fillattr(inode, stat);
647 #ifdef CONFIG_F2FS_FS_POSIX_ACL
648 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
650 struct f2fs_inode_info *fi = F2FS_I(inode);
651 unsigned int ia_valid = attr->ia_valid;
653 if (ia_valid & ATTR_UID)
654 inode->i_uid = attr->ia_uid;
655 if (ia_valid & ATTR_GID)
656 inode->i_gid = attr->ia_gid;
657 if (ia_valid & ATTR_ATIME)
658 inode->i_atime = timespec_trunc(attr->ia_atime,
659 inode->i_sb->s_time_gran);
660 if (ia_valid & ATTR_MTIME)
661 inode->i_mtime = timespec_trunc(attr->ia_mtime,
662 inode->i_sb->s_time_gran);
663 if (ia_valid & ATTR_CTIME)
664 inode->i_ctime = timespec_trunc(attr->ia_ctime,
665 inode->i_sb->s_time_gran);
666 if (ia_valid & ATTR_MODE) {
667 umode_t mode = attr->ia_mode;
669 if (!in_group_p(inode->i_gid) &&
670 !capable_wrt_inode_uidgid(inode, CAP_FSETID))
672 set_acl_inode(fi, mode);
676 #define __setattr_copy setattr_copy
679 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
681 struct inode *inode = d_inode(dentry);
682 struct f2fs_inode_info *fi = F2FS_I(inode);
685 err = inode_change_ok(inode, attr);
689 if (attr->ia_valid & ATTR_SIZE) {
690 if (f2fs_encrypted_inode(inode) &&
691 f2fs_get_encryption_info(inode))
694 if (attr->ia_size <= i_size_read(inode)) {
695 truncate_setsize(inode, attr->ia_size);
696 err = f2fs_truncate(inode, true);
699 f2fs_balance_fs(F2FS_I_SB(inode));
702 * do not trim all blocks after i_size if target size is
703 * larger than i_size.
705 truncate_setsize(inode, attr->ia_size);
706 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
710 __setattr_copy(inode, attr);
712 if (attr->ia_valid & ATTR_MODE) {
713 err = posix_acl_chmod(inode, get_inode_mode(inode));
714 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
715 inode->i_mode = fi->i_acl_mode;
716 clear_inode_flag(fi, FI_ACL_MODE);
720 mark_inode_dirty(inode);
724 const struct inode_operations f2fs_file_inode_operations = {
725 .getattr = f2fs_getattr,
726 .setattr = f2fs_setattr,
727 .get_acl = f2fs_get_acl,
728 .set_acl = f2fs_set_acl,
729 #ifdef CONFIG_F2FS_FS_XATTR
730 .setxattr = generic_setxattr,
731 .getxattr = generic_getxattr,
732 .listxattr = f2fs_listxattr,
733 .removexattr = generic_removexattr,
735 .fiemap = f2fs_fiemap,
738 static int fill_zero(struct inode *inode, pgoff_t index,
739 loff_t start, loff_t len)
741 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
747 f2fs_balance_fs(sbi);
750 page = get_new_data_page(inode, NULL, index, false);
754 return PTR_ERR(page);
756 f2fs_wait_on_page_writeback(page, DATA);
757 zero_user(page, start, len);
758 set_page_dirty(page);
759 f2fs_put_page(page, 1);
763 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
767 while (pg_start < pg_end) {
768 struct dnode_of_data dn;
769 pgoff_t end_offset, count;
771 set_new_dnode(&dn, inode, NULL, NULL, 0);
772 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
774 if (err == -ENOENT) {
781 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
782 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
784 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
786 truncate_data_blocks_range(&dn, count);
794 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
796 pgoff_t pg_start, pg_end;
797 loff_t off_start, off_end;
800 if (f2fs_has_inline_data(inode)) {
801 ret = f2fs_convert_inline_inode(inode);
806 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
807 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
809 off_start = offset & (PAGE_CACHE_SIZE - 1);
810 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
812 if (pg_start == pg_end) {
813 ret = fill_zero(inode, pg_start, off_start,
814 off_end - off_start);
819 ret = fill_zero(inode, pg_start++, off_start,
820 PAGE_CACHE_SIZE - off_start);
825 ret = fill_zero(inode, pg_end, 0, off_end);
830 if (pg_start < pg_end) {
831 struct address_space *mapping = inode->i_mapping;
832 loff_t blk_start, blk_end;
833 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
835 f2fs_balance_fs(sbi);
837 blk_start = (loff_t)pg_start << PAGE_CACHE_SHIFT;
838 blk_end = (loff_t)pg_end << PAGE_CACHE_SHIFT;
839 truncate_inode_pages_range(mapping, blk_start,
843 ret = truncate_hole(inode, pg_start, pg_end);
851 static int __exchange_data_block(struct inode *inode, pgoff_t src,
852 pgoff_t dst, bool full)
854 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
855 struct dnode_of_data dn;
857 bool do_replace = false;
860 set_new_dnode(&dn, inode, NULL, NULL, 0);
861 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
862 if (ret && ret != -ENOENT) {
864 } else if (ret == -ENOENT) {
865 new_addr = NULL_ADDR;
867 new_addr = dn.data_blkaddr;
868 if (!is_checkpointed_data(sbi, new_addr)) {
869 dn.data_blkaddr = NULL_ADDR;
870 /* do not invalidate this block address */
871 set_data_blkaddr(&dn);
872 f2fs_update_extent_cache(&dn);
878 if (new_addr == NULL_ADDR)
879 return full ? truncate_hole(inode, dst, dst + 1) : 0;
882 struct page *ipage = get_node_page(sbi, inode->i_ino);
886 ret = PTR_ERR(ipage);
890 set_new_dnode(&dn, inode, ipage, NULL, 0);
891 ret = f2fs_reserve_block(&dn, dst);
895 truncate_data_blocks_range(&dn, 1);
897 get_node_info(sbi, dn.nid, &ni);
898 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
902 struct page *psrc, *pdst;
904 psrc = get_lock_data_page(inode, src, true);
906 return PTR_ERR(psrc);
907 pdst = get_new_data_page(inode, NULL, dst, false);
909 f2fs_put_page(psrc, 1);
910 return PTR_ERR(pdst);
912 f2fs_copy_page(psrc, pdst);
913 set_page_dirty(pdst);
914 f2fs_put_page(pdst, 1);
915 f2fs_put_page(psrc, 1);
917 return truncate_hole(inode, src, src + 1);
922 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
923 dn.data_blkaddr = new_addr;
924 set_data_blkaddr(&dn);
925 f2fs_update_extent_cache(&dn);
931 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
933 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
934 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
937 for (; end < nrpages; start++, end++) {
938 f2fs_balance_fs(sbi);
940 ret = __exchange_data_block(inode, end, start, true);
948 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
950 pgoff_t pg_start, pg_end;
954 if (offset + len >= i_size_read(inode))
957 /* collapse range should be aligned to block size of f2fs. */
958 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
961 f2fs_balance_fs(F2FS_I_SB(inode));
963 if (f2fs_has_inline_data(inode)) {
964 ret = f2fs_convert_inline_inode(inode);
969 pg_start = offset >> PAGE_CACHE_SHIFT;
970 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
972 /* write out all dirty pages from offset */
973 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
977 truncate_pagecache(inode, offset);
979 ret = f2fs_do_collapse(inode, pg_start, pg_end);
983 /* write out all moved pages, if possible */
984 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
985 truncate_pagecache(inode, offset);
987 new_size = i_size_read(inode) - len;
988 truncate_pagecache(inode, new_size);
990 ret = truncate_blocks(inode, new_size, true);
992 i_size_write(inode, new_size);
997 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1000 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1001 struct address_space *mapping = inode->i_mapping;
1002 pgoff_t index, pg_start, pg_end;
1003 loff_t new_size = i_size_read(inode);
1004 loff_t off_start, off_end;
1007 ret = inode_newsize_ok(inode, (len + offset));
1011 f2fs_balance_fs(sbi);
1013 if (f2fs_has_inline_data(inode)) {
1014 ret = f2fs_convert_inline_inode(inode);
1019 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1023 truncate_pagecache_range(inode, offset, offset + len - 1);
1025 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1026 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1028 off_start = offset & (PAGE_CACHE_SIZE - 1);
1029 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1031 if (pg_start == pg_end) {
1032 ret = fill_zero(inode, pg_start, off_start,
1033 off_end - off_start);
1037 if (offset + len > new_size)
1038 new_size = offset + len;
1039 new_size = max_t(loff_t, new_size, offset + len);
1042 ret = fill_zero(inode, pg_start++, off_start,
1043 PAGE_CACHE_SIZE - off_start);
1047 new_size = max_t(loff_t, new_size,
1048 (loff_t)pg_start << PAGE_CACHE_SHIFT);
1051 for (index = pg_start; index < pg_end; index++) {
1052 struct dnode_of_data dn;
1057 ipage = get_node_page(sbi, inode->i_ino);
1058 if (IS_ERR(ipage)) {
1059 ret = PTR_ERR(ipage);
1060 f2fs_unlock_op(sbi);
1064 set_new_dnode(&dn, inode, ipage, NULL, 0);
1065 ret = f2fs_reserve_block(&dn, index);
1067 f2fs_unlock_op(sbi);
1071 if (dn.data_blkaddr != NEW_ADDR) {
1072 invalidate_blocks(sbi, dn.data_blkaddr);
1074 dn.data_blkaddr = NEW_ADDR;
1075 set_data_blkaddr(&dn);
1077 dn.data_blkaddr = NULL_ADDR;
1078 f2fs_update_extent_cache(&dn);
1080 f2fs_put_dnode(&dn);
1081 f2fs_unlock_op(sbi);
1083 new_size = max_t(loff_t, new_size,
1084 (loff_t)(index + 1) << PAGE_CACHE_SHIFT);
1088 ret = fill_zero(inode, pg_end, 0, off_end);
1092 new_size = max_t(loff_t, new_size, offset + len);
1097 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1098 i_size_write(inode, new_size);
1099 mark_inode_dirty(inode);
1100 update_inode_page(inode);
1106 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1108 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1109 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1113 new_size = i_size_read(inode) + len;
1114 if (new_size > inode->i_sb->s_maxbytes)
1117 if (offset >= i_size_read(inode))
1120 /* insert range should be aligned to block size of f2fs. */
1121 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1124 f2fs_balance_fs(sbi);
1126 if (f2fs_has_inline_data(inode)) {
1127 ret = f2fs_convert_inline_inode(inode);
1132 ret = truncate_blocks(inode, i_size_read(inode), true);
1136 /* write out all dirty pages from offset */
1137 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1141 truncate_pagecache(inode, offset);
1143 pg_start = offset >> PAGE_CACHE_SHIFT;
1144 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
1145 delta = pg_end - pg_start;
1146 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1148 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1150 ret = __exchange_data_block(inode, idx, idx + delta, false);
1151 f2fs_unlock_op(sbi);
1156 /* write out all moved pages, if possible */
1157 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1158 truncate_pagecache(inode, offset);
1161 i_size_write(inode, new_size);
1165 static int expand_inode_data(struct inode *inode, loff_t offset,
1166 loff_t len, int mode)
1168 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1169 pgoff_t index, pg_start, pg_end;
1170 loff_t new_size = i_size_read(inode);
1171 loff_t off_start, off_end;
1174 f2fs_balance_fs(sbi);
1176 ret = inode_newsize_ok(inode, (len + offset));
1180 if (f2fs_has_inline_data(inode)) {
1181 ret = f2fs_convert_inline_inode(inode);
1186 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1187 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1189 off_start = offset & (PAGE_CACHE_SIZE - 1);
1190 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1194 for (index = pg_start; index <= pg_end; index++) {
1195 struct dnode_of_data dn;
1197 if (index == pg_end && !off_end)
1200 set_new_dnode(&dn, inode, NULL, NULL, 0);
1201 ret = f2fs_reserve_block(&dn, index);
1205 if (pg_start == pg_end)
1206 new_size = offset + len;
1207 else if (index == pg_start && off_start)
1208 new_size = (loff_t)(index + 1) << PAGE_CACHE_SHIFT;
1209 else if (index == pg_end)
1210 new_size = ((loff_t)index << PAGE_CACHE_SHIFT) +
1213 new_size += PAGE_CACHE_SIZE;
1216 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1217 i_size_read(inode) < new_size) {
1218 i_size_write(inode, new_size);
1219 mark_inode_dirty(inode);
1220 update_inode_page(inode);
1222 f2fs_unlock_op(sbi);
1227 static long f2fs_fallocate(struct file *file, int mode,
1228 loff_t offset, loff_t len)
1230 struct inode *inode = file_inode(file);
1233 /* f2fs only support ->fallocate for regular file */
1234 if (!S_ISREG(inode->i_mode))
1237 if (f2fs_encrypted_inode(inode) &&
1238 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1241 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1242 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1243 FALLOC_FL_INSERT_RANGE))
1246 mutex_lock(&inode->i_mutex);
1248 if (mode & FALLOC_FL_PUNCH_HOLE) {
1249 if (offset >= inode->i_size)
1252 ret = punch_hole(inode, offset, len);
1253 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1254 ret = f2fs_collapse_range(inode, offset, len);
1255 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1256 ret = f2fs_zero_range(inode, offset, len, mode);
1257 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1258 ret = f2fs_insert_range(inode, offset, len);
1260 ret = expand_inode_data(inode, offset, len, mode);
1264 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1265 mark_inode_dirty(inode);
1269 mutex_unlock(&inode->i_mutex);
1271 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1275 static int f2fs_release_file(struct inode *inode, struct file *filp)
1277 /* some remained atomic pages should discarded */
1278 if (f2fs_is_atomic_file(inode))
1279 commit_inmem_pages(inode, true);
1280 if (f2fs_is_volatile_file(inode)) {
1281 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1282 filemap_fdatawrite(inode->i_mapping);
1283 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1288 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1289 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1291 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1295 else if (S_ISREG(mode))
1296 return flags & F2FS_REG_FLMASK;
1298 return flags & F2FS_OTHER_FLMASK;
1301 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1303 struct inode *inode = file_inode(filp);
1304 struct f2fs_inode_info *fi = F2FS_I(inode);
1305 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1306 return put_user(flags, (int __user *)arg);
1309 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1311 struct inode *inode = file_inode(filp);
1312 struct f2fs_inode_info *fi = F2FS_I(inode);
1313 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1314 unsigned int oldflags;
1317 ret = mnt_want_write_file(filp);
1321 if (!inode_owner_or_capable(inode)) {
1326 if (get_user(flags, (int __user *)arg)) {
1331 flags = f2fs_mask_flags(inode->i_mode, flags);
1333 mutex_lock(&inode->i_mutex);
1335 oldflags = fi->i_flags;
1337 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1338 if (!capable(CAP_LINUX_IMMUTABLE)) {
1339 mutex_unlock(&inode->i_mutex);
1345 flags = flags & FS_FL_USER_MODIFIABLE;
1346 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1347 fi->i_flags = flags;
1348 mutex_unlock(&inode->i_mutex);
1350 f2fs_set_inode_flags(inode);
1351 inode->i_ctime = CURRENT_TIME;
1352 mark_inode_dirty(inode);
1354 mnt_drop_write_file(filp);
1358 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1360 struct inode *inode = file_inode(filp);
1362 return put_user(inode->i_generation, (int __user *)arg);
1365 static int f2fs_ioc_start_atomic_write(struct file *filp)
1367 struct inode *inode = file_inode(filp);
1370 if (!inode_owner_or_capable(inode))
1373 f2fs_balance_fs(F2FS_I_SB(inode));
1375 if (f2fs_is_atomic_file(inode))
1378 ret = f2fs_convert_inline_inode(inode);
1382 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1386 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1388 struct inode *inode = file_inode(filp);
1391 if (!inode_owner_or_capable(inode))
1394 if (f2fs_is_volatile_file(inode))
1397 ret = mnt_want_write_file(filp);
1401 if (f2fs_is_atomic_file(inode)) {
1402 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1403 ret = commit_inmem_pages(inode, false);
1408 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1410 mnt_drop_write_file(filp);
1414 static int f2fs_ioc_start_volatile_write(struct file *filp)
1416 struct inode *inode = file_inode(filp);
1419 if (!inode_owner_or_capable(inode))
1422 if (f2fs_is_volatile_file(inode))
1425 ret = f2fs_convert_inline_inode(inode);
1429 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1433 static int f2fs_ioc_release_volatile_write(struct file *filp)
1435 struct inode *inode = file_inode(filp);
1437 if (!inode_owner_or_capable(inode))
1440 if (!f2fs_is_volatile_file(inode))
1443 if (!f2fs_is_first_block_written(inode))
1444 return truncate_partial_data_page(inode, 0, true);
1446 return punch_hole(inode, 0, F2FS_BLKSIZE);
1449 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1451 struct inode *inode = file_inode(filp);
1454 if (!inode_owner_or_capable(inode))
1457 ret = mnt_want_write_file(filp);
1461 f2fs_balance_fs(F2FS_I_SB(inode));
1463 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1464 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1465 commit_inmem_pages(inode, true);
1467 mnt_drop_write_file(filp);
1471 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1473 struct inode *inode = file_inode(filp);
1474 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1475 struct super_block *sb = sbi->sb;
1478 if (!capable(CAP_SYS_ADMIN))
1481 if (get_user(in, (__u32 __user *)arg))
1485 case F2FS_GOING_DOWN_FULLSYNC:
1486 sb = freeze_bdev(sb->s_bdev);
1487 if (sb && !IS_ERR(sb)) {
1488 f2fs_stop_checkpoint(sbi);
1489 thaw_bdev(sb->s_bdev, sb);
1492 case F2FS_GOING_DOWN_METASYNC:
1493 /* do checkpoint only */
1494 f2fs_sync_fs(sb, 1);
1495 f2fs_stop_checkpoint(sbi);
1497 case F2FS_GOING_DOWN_NOSYNC:
1498 f2fs_stop_checkpoint(sbi);
1500 case F2FS_GOING_DOWN_METAFLUSH:
1501 sync_meta_pages(sbi, META, LONG_MAX);
1502 f2fs_stop_checkpoint(sbi);
1510 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1512 struct inode *inode = file_inode(filp);
1513 struct super_block *sb = inode->i_sb;
1514 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1515 struct fstrim_range range;
1518 if (!capable(CAP_SYS_ADMIN))
1521 if (!blk_queue_discard(q))
1524 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1528 range.minlen = max((unsigned int)range.minlen,
1529 q->limits.discard_granularity);
1530 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1534 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1540 static bool uuid_is_nonzero(__u8 u[16])
1544 for (i = 0; i < 16; i++)
1550 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1552 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1553 struct f2fs_encryption_policy policy;
1554 struct inode *inode = file_inode(filp);
1557 if (copy_from_user(&policy, (struct f2fs_encryption_policy __user *)arg,
1561 err = mnt_want_write_file(filp);
1565 mutex_lock(&inode->i_mutex);
1567 err = f2fs_process_policy(&policy, inode);
1569 mutex_unlock(&inode->i_mutex);
1571 mnt_drop_write_file(filp);
1579 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1581 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1582 struct f2fs_encryption_policy policy;
1583 struct inode *inode = file_inode(filp);
1586 err = f2fs_get_policy(inode, &policy);
1590 if (copy_to_user((struct f2fs_encryption_policy __user *)arg, &policy,
1599 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1601 struct inode *inode = file_inode(filp);
1602 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1605 if (!f2fs_sb_has_crypto(inode->i_sb))
1608 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1611 err = mnt_want_write_file(filp);
1615 /* update superblock with uuid */
1616 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1618 err = f2fs_commit_super(sbi, false);
1620 mnt_drop_write_file(filp);
1623 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1627 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1633 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1635 struct inode *inode = file_inode(filp);
1636 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1639 if (!capable(CAP_SYS_ADMIN))
1642 if (get_user(sync, (__u32 __user *)arg))
1645 if (f2fs_readonly(sbi->sb))
1649 if (!mutex_trylock(&sbi->gc_mutex))
1652 mutex_lock(&sbi->gc_mutex);
1655 return f2fs_gc(sbi, sync);
1658 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1660 struct inode *inode = file_inode(filp);
1661 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1662 struct cp_control cpc;
1664 if (!capable(CAP_SYS_ADMIN))
1667 if (f2fs_readonly(sbi->sb))
1670 cpc.reason = __get_cp_reason(sbi);
1672 mutex_lock(&sbi->gc_mutex);
1673 write_checkpoint(sbi, &cpc);
1674 mutex_unlock(&sbi->gc_mutex);
1679 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1682 case F2FS_IOC_GETFLAGS:
1683 return f2fs_ioc_getflags(filp, arg);
1684 case F2FS_IOC_SETFLAGS:
1685 return f2fs_ioc_setflags(filp, arg);
1686 case F2FS_IOC_GETVERSION:
1687 return f2fs_ioc_getversion(filp, arg);
1688 case F2FS_IOC_START_ATOMIC_WRITE:
1689 return f2fs_ioc_start_atomic_write(filp);
1690 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1691 return f2fs_ioc_commit_atomic_write(filp);
1692 case F2FS_IOC_START_VOLATILE_WRITE:
1693 return f2fs_ioc_start_volatile_write(filp);
1694 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1695 return f2fs_ioc_release_volatile_write(filp);
1696 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1697 return f2fs_ioc_abort_volatile_write(filp);
1698 case F2FS_IOC_SHUTDOWN:
1699 return f2fs_ioc_shutdown(filp, arg);
1701 return f2fs_ioc_fitrim(filp, arg);
1702 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1703 return f2fs_ioc_set_encryption_policy(filp, arg);
1704 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1705 return f2fs_ioc_get_encryption_policy(filp, arg);
1706 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1707 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1708 case F2FS_IOC_GARBAGE_COLLECT:
1709 return f2fs_ioc_gc(filp, arg);
1710 case F2FS_IOC_WRITE_CHECKPOINT:
1711 return f2fs_ioc_write_checkpoint(filp, arg);
1717 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1719 struct inode *inode = file_inode(iocb->ki_filp);
1721 if (f2fs_encrypted_inode(inode) &&
1722 !f2fs_has_encryption_key(inode) &&
1723 f2fs_get_encryption_info(inode))
1726 return generic_file_write_iter(iocb, from);
1729 #ifdef CONFIG_COMPAT
1730 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1733 case F2FS_IOC32_GETFLAGS:
1734 cmd = F2FS_IOC_GETFLAGS;
1736 case F2FS_IOC32_SETFLAGS:
1737 cmd = F2FS_IOC_SETFLAGS;
1740 return -ENOIOCTLCMD;
1742 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1746 const struct file_operations f2fs_file_operations = {
1747 .llseek = f2fs_llseek,
1748 .read_iter = generic_file_read_iter,
1749 .write_iter = f2fs_file_write_iter,
1750 .open = f2fs_file_open,
1751 .release = f2fs_release_file,
1752 .mmap = f2fs_file_mmap,
1753 .fsync = f2fs_sync_file,
1754 .fallocate = f2fs_fallocate,
1755 .unlocked_ioctl = f2fs_ioctl,
1756 #ifdef CONFIG_COMPAT
1757 .compat_ioctl = f2fs_compat_ioctl,
1759 .splice_read = generic_file_splice_read,
1760 .splice_write = iter_file_splice_write,