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 (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
79 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
80 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
83 SetPageUptodate(page);
85 trace_f2fs_vm_page_mkwrite(page, DATA);
88 f2fs_wait_on_page_writeback(page, DATA);
89 /* if gced page is attached, don't write to cold segment */
90 clear_cold_data(page);
92 sb_end_pagefault(inode->i_sb);
93 return block_page_mkwrite_return(err);
96 static const struct vm_operations_struct f2fs_file_vm_ops = {
97 .fault = filemap_fault,
98 .map_pages = filemap_map_pages,
99 .page_mkwrite = f2fs_vm_page_mkwrite,
102 static int get_parent_ino(struct inode *inode, nid_t *pino)
104 struct dentry *dentry;
106 inode = igrab(inode);
107 dentry = d_find_any_alias(inode);
112 if (update_dent_inode(inode, inode, &dentry->d_name)) {
117 *pino = parent_ino(dentry);
122 static inline bool need_do_checkpoint(struct inode *inode)
124 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
125 bool need_cp = false;
127 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
129 else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
131 else if (file_wrong_pino(inode))
133 else if (!space_for_roll_forward(sbi))
135 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
137 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
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);
165 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
166 get_parent_ino(inode, &pino)) {
168 file_got_pino(inode);
169 up_write(&fi->i_sem);
171 mark_inode_dirty_sync(inode);
172 f2fs_write_inode(inode, NULL);
174 up_write(&fi->i_sem);
178 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
180 struct inode *inode = file->f_mapping->host;
181 struct f2fs_inode_info *fi = F2FS_I(inode);
182 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
183 nid_t ino = inode->i_ino;
185 bool need_cp = false;
186 struct writeback_control wbc = {
187 .sync_mode = WB_SYNC_ALL,
188 .nr_to_write = LONG_MAX,
192 if (unlikely(f2fs_readonly(inode->i_sb)))
195 trace_f2fs_sync_file_enter(inode);
197 /* if fdatasync is triggered, let's do in-place-update */
198 if (get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
199 set_inode_flag(fi, FI_NEED_IPU);
200 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
201 clear_inode_flag(fi, FI_NEED_IPU);
204 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
208 /* if the inode is dirty, let's recover all the time */
210 f2fs_write_inode(inode, NULL);
215 * if there is no written data, don't waste time to write recovery info.
217 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
218 !exist_written_data(sbi, ino, APPEND_INO)) {
220 /* it may call write_inode just prior to fsync */
221 if (need_inode_page_update(sbi, ino))
224 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
225 exist_written_data(sbi, ino, UPDATE_INO))
230 /* guarantee free sections for fsync */
231 f2fs_balance_fs(sbi);
234 * Both of fdatasync() and fsync() are able to be recovered from
237 down_read(&fi->i_sem);
238 need_cp = need_do_checkpoint(inode);
242 /* all the dirty node pages should be flushed for POR */
243 ret = f2fs_sync_fs(inode->i_sb, 1);
246 * We've secured consistency through sync_fs. Following pino
247 * will be used only for fsynced inodes after checkpoint.
249 try_to_fix_pino(inode);
250 clear_inode_flag(fi, FI_APPEND_WRITE);
251 clear_inode_flag(fi, FI_UPDATE_WRITE);
255 sync_node_pages(sbi, ino, &wbc);
257 /* if cp_error was enabled, we should avoid infinite loop */
258 if (unlikely(f2fs_cp_error(sbi)))
261 if (need_inode_block_update(sbi, ino)) {
262 mark_inode_dirty_sync(inode);
263 f2fs_write_inode(inode, NULL);
267 ret = wait_on_node_pages_writeback(sbi, ino);
271 /* once recovery info is written, don't need to tack this */
272 remove_dirty_inode(sbi, ino, APPEND_INO);
273 clear_inode_flag(fi, FI_APPEND_WRITE);
275 remove_dirty_inode(sbi, ino, UPDATE_INO);
276 clear_inode_flag(fi, FI_UPDATE_WRITE);
277 ret = f2fs_issue_flush(sbi);
279 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
280 f2fs_trace_ios(NULL, 1);
284 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
285 pgoff_t pgofs, int whence)
290 if (whence != SEEK_DATA)
293 /* find first dirty page index */
294 pagevec_init(&pvec, 0);
295 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
296 PAGECACHE_TAG_DIRTY, 1);
297 pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
298 pagevec_release(&pvec);
302 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
307 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
308 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
312 if (blkaddr == NULL_ADDR)
319 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
321 struct inode *inode = file->f_mapping->host;
322 loff_t maxbytes = inode->i_sb->s_maxbytes;
323 struct dnode_of_data dn;
324 pgoff_t pgofs, end_offset, dirty;
325 loff_t data_ofs = offset;
329 mutex_lock(&inode->i_mutex);
331 isize = i_size_read(inode);
335 /* handle inline data case */
336 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
337 if (whence == SEEK_HOLE)
342 pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
344 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
346 for (; data_ofs < isize; data_ofs = pgofs << PAGE_CACHE_SHIFT) {
347 set_new_dnode(&dn, inode, NULL, NULL, 0);
348 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
349 if (err && err != -ENOENT) {
351 } else if (err == -ENOENT) {
352 /* direct node does not exists */
353 if (whence == SEEK_DATA) {
354 pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
362 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
364 /* find data/hole in dnode block */
365 for (; dn.ofs_in_node < end_offset;
366 dn.ofs_in_node++, pgofs++,
367 data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
369 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
371 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
379 if (whence == SEEK_DATA)
382 if (whence == SEEK_HOLE && data_ofs > isize)
384 mutex_unlock(&inode->i_mutex);
385 return vfs_setpos(file, data_ofs, maxbytes);
387 mutex_unlock(&inode->i_mutex);
391 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
393 struct inode *inode = file->f_mapping->host;
394 loff_t maxbytes = inode->i_sb->s_maxbytes;
400 return generic_file_llseek_size(file, offset, whence,
401 maxbytes, i_size_read(inode));
406 return f2fs_seek_block(file, offset, whence);
412 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
414 struct inode *inode = file_inode(file);
416 if (f2fs_encrypted_inode(inode)) {
417 int err = f2fs_get_encryption_info(inode);
422 /* we don't need to use inline_data strictly */
423 if (f2fs_has_inline_data(inode)) {
424 int err = f2fs_convert_inline_inode(inode);
430 vma->vm_ops = &f2fs_file_vm_ops;
434 static int f2fs_file_open(struct inode *inode, struct file *filp)
436 int ret = generic_file_open(inode, filp);
438 if (!ret && f2fs_encrypted_inode(inode)) {
439 ret = f2fs_get_encryption_info(inode);
446 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
448 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
449 struct f2fs_node *raw_node;
450 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
453 raw_node = F2FS_NODE(dn->node_page);
454 addr = blkaddr_in_node(raw_node) + ofs;
456 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
457 block_t blkaddr = le32_to_cpu(*addr);
458 if (blkaddr == NULL_ADDR)
461 dn->data_blkaddr = NULL_ADDR;
462 set_data_blkaddr(dn);
463 invalidate_blocks(sbi, blkaddr);
464 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
465 clear_inode_flag(F2FS_I(dn->inode),
466 FI_FIRST_BLOCK_WRITTEN);
473 * once we invalidate valid blkaddr in range [ofs, ofs + count],
474 * we will invalidate all blkaddr in the whole range.
476 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
477 F2FS_I(dn->inode)) + ofs;
478 f2fs_update_extent_cache_range(dn, fofs, 0, len);
479 dec_valid_block_count(sbi, dn->inode, nr_free);
480 set_page_dirty(dn->node_page);
483 dn->ofs_in_node = ofs;
485 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
486 dn->ofs_in_node, nr_free);
490 void truncate_data_blocks(struct dnode_of_data *dn)
492 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
495 static int truncate_partial_data_page(struct inode *inode, u64 from,
498 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
499 pgoff_t index = from >> PAGE_CACHE_SHIFT;
500 struct address_space *mapping = inode->i_mapping;
503 if (!offset && !cache_only)
507 page = grab_cache_page(mapping, index);
508 if (page && PageUptodate(page))
510 f2fs_put_page(page, 1);
514 page = get_lock_data_page(inode, index);
518 f2fs_wait_on_page_writeback(page, DATA);
519 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
520 if (!cache_only || !f2fs_encrypted_inode(inode) || !S_ISREG(inode->i_mode))
521 set_page_dirty(page);
522 f2fs_put_page(page, 1);
526 int truncate_blocks(struct inode *inode, u64 from, bool lock)
528 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
529 unsigned int blocksize = inode->i_sb->s_blocksize;
530 struct dnode_of_data dn;
532 int count = 0, err = 0;
534 bool truncate_page = false;
536 trace_f2fs_truncate_blocks_enter(inode, from);
538 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
543 ipage = get_node_page(sbi, inode->i_ino);
545 err = PTR_ERR(ipage);
549 if (f2fs_has_inline_data(inode)) {
550 if (truncate_inline_inode(ipage, from))
551 set_page_dirty(ipage);
552 f2fs_put_page(ipage, 1);
553 truncate_page = true;
557 set_new_dnode(&dn, inode, ipage, NULL, 0);
558 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
565 count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
567 count -= dn.ofs_in_node;
568 f2fs_bug_on(sbi, count < 0);
570 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
571 truncate_data_blocks_range(&dn, count);
577 err = truncate_inode_blocks(inode, free_from);
582 /* lastly zero out the first data page */
584 err = truncate_partial_data_page(inode, from, truncate_page);
586 trace_f2fs_truncate_blocks_exit(inode, err);
590 int f2fs_truncate(struct inode *inode, bool lock)
594 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
595 S_ISLNK(inode->i_mode)))
598 trace_f2fs_truncate(inode);
600 /* we should check inline_data size */
601 if (f2fs_has_inline_data(inode) && !f2fs_may_inline_data(inode)) {
602 err = f2fs_convert_inline_inode(inode);
607 err = truncate_blocks(inode, i_size_read(inode), lock);
611 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
612 mark_inode_dirty(inode);
616 int f2fs_getattr(struct vfsmount *mnt,
617 struct dentry *dentry, struct kstat *stat)
619 struct inode *inode = d_inode(dentry);
620 generic_fillattr(inode, stat);
625 #ifdef CONFIG_F2FS_FS_POSIX_ACL
626 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
628 struct f2fs_inode_info *fi = F2FS_I(inode);
629 unsigned int ia_valid = attr->ia_valid;
631 if (ia_valid & ATTR_UID)
632 inode->i_uid = attr->ia_uid;
633 if (ia_valid & ATTR_GID)
634 inode->i_gid = attr->ia_gid;
635 if (ia_valid & ATTR_ATIME)
636 inode->i_atime = timespec_trunc(attr->ia_atime,
637 inode->i_sb->s_time_gran);
638 if (ia_valid & ATTR_MTIME)
639 inode->i_mtime = timespec_trunc(attr->ia_mtime,
640 inode->i_sb->s_time_gran);
641 if (ia_valid & ATTR_CTIME)
642 inode->i_ctime = timespec_trunc(attr->ia_ctime,
643 inode->i_sb->s_time_gran);
644 if (ia_valid & ATTR_MODE) {
645 umode_t mode = attr->ia_mode;
647 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
649 set_acl_inode(fi, mode);
653 #define __setattr_copy setattr_copy
656 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
658 struct inode *inode = d_inode(dentry);
659 struct f2fs_inode_info *fi = F2FS_I(inode);
662 err = inode_change_ok(inode, attr);
666 if (attr->ia_valid & ATTR_SIZE) {
667 if (f2fs_encrypted_inode(inode) &&
668 f2fs_get_encryption_info(inode))
671 if (attr->ia_size <= i_size_read(inode)) {
672 truncate_setsize(inode, attr->ia_size);
673 err = f2fs_truncate(inode, true);
676 f2fs_balance_fs(F2FS_I_SB(inode));
679 * do not trim all blocks after i_size if target size is
680 * larger than i_size.
682 truncate_setsize(inode, attr->ia_size);
686 __setattr_copy(inode, attr);
688 if (attr->ia_valid & ATTR_MODE) {
689 err = posix_acl_chmod(inode, get_inode_mode(inode));
690 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
691 inode->i_mode = fi->i_acl_mode;
692 clear_inode_flag(fi, FI_ACL_MODE);
696 mark_inode_dirty(inode);
700 const struct inode_operations f2fs_file_inode_operations = {
701 .getattr = f2fs_getattr,
702 .setattr = f2fs_setattr,
703 .get_acl = f2fs_get_acl,
704 .set_acl = f2fs_set_acl,
705 #ifdef CONFIG_F2FS_FS_XATTR
706 .setxattr = generic_setxattr,
707 .getxattr = generic_getxattr,
708 .listxattr = f2fs_listxattr,
709 .removexattr = generic_removexattr,
711 .fiemap = f2fs_fiemap,
714 static int fill_zero(struct inode *inode, pgoff_t index,
715 loff_t start, loff_t len)
717 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
723 f2fs_balance_fs(sbi);
726 page = get_new_data_page(inode, NULL, index, false);
730 return PTR_ERR(page);
732 f2fs_wait_on_page_writeback(page, DATA);
733 zero_user(page, start, len);
734 set_page_dirty(page);
735 f2fs_put_page(page, 1);
739 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
744 for (index = pg_start; index < pg_end; index++) {
745 struct dnode_of_data dn;
747 set_new_dnode(&dn, inode, NULL, NULL, 0);
748 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
755 if (dn.data_blkaddr != NULL_ADDR)
756 truncate_data_blocks_range(&dn, 1);
762 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
764 pgoff_t pg_start, pg_end;
765 loff_t off_start, off_end;
768 if (!S_ISREG(inode->i_mode))
771 if (f2fs_has_inline_data(inode)) {
772 ret = f2fs_convert_inline_inode(inode);
777 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
778 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
780 off_start = offset & (PAGE_CACHE_SIZE - 1);
781 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
783 if (pg_start == pg_end) {
784 ret = fill_zero(inode, pg_start, off_start,
785 off_end - off_start);
790 ret = fill_zero(inode, pg_start++, off_start,
791 PAGE_CACHE_SIZE - off_start);
796 ret = fill_zero(inode, pg_end, 0, off_end);
801 if (pg_start < pg_end) {
802 struct address_space *mapping = inode->i_mapping;
803 loff_t blk_start, blk_end;
804 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
806 f2fs_balance_fs(sbi);
808 blk_start = pg_start << PAGE_CACHE_SHIFT;
809 blk_end = pg_end << PAGE_CACHE_SHIFT;
810 truncate_inode_pages_range(mapping, blk_start,
814 ret = truncate_hole(inode, pg_start, pg_end);
822 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
824 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
825 struct dnode_of_data dn;
826 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
829 for (; end < nrpages; start++, end++) {
830 block_t new_addr, old_addr;
834 set_new_dnode(&dn, inode, NULL, NULL, 0);
835 ret = get_dnode_of_data(&dn, end, LOOKUP_NODE_RA);
836 if (ret && ret != -ENOENT) {
838 } else if (ret == -ENOENT) {
839 new_addr = NULL_ADDR;
841 new_addr = dn.data_blkaddr;
842 truncate_data_blocks_range(&dn, 1);
846 if (new_addr == NULL_ADDR) {
847 set_new_dnode(&dn, inode, NULL, NULL, 0);
848 ret = get_dnode_of_data(&dn, start, LOOKUP_NODE_RA);
849 if (ret && ret != -ENOENT) {
851 } else if (ret == -ENOENT) {
856 if (dn.data_blkaddr == NULL_ADDR) {
861 truncate_data_blocks_range(&dn, 1);
868 ipage = get_node_page(sbi, inode->i_ino);
870 ret = PTR_ERR(ipage);
874 set_new_dnode(&dn, inode, ipage, NULL, 0);
875 ret = f2fs_reserve_block(&dn, start);
879 old_addr = dn.data_blkaddr;
880 if (old_addr != NEW_ADDR && new_addr == NEW_ADDR) {
881 dn.data_blkaddr = NULL_ADDR;
882 f2fs_update_extent_cache(&dn);
883 invalidate_blocks(sbi, old_addr);
885 dn.data_blkaddr = new_addr;
886 set_data_blkaddr(&dn);
887 } else if (new_addr != NEW_ADDR) {
890 get_node_info(sbi, dn.nid, &ni);
891 f2fs_replace_block(sbi, &dn, old_addr, new_addr,
905 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
907 pgoff_t pg_start, pg_end;
911 if (!S_ISREG(inode->i_mode))
914 if (offset + len >= i_size_read(inode))
917 /* collapse range should be aligned to block size of f2fs. */
918 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
921 f2fs_balance_fs(F2FS_I_SB(inode));
923 if (f2fs_has_inline_data(inode)) {
924 ret = f2fs_convert_inline_inode(inode);
929 pg_start = offset >> PAGE_CACHE_SHIFT;
930 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
932 /* write out all dirty pages from offset */
933 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
937 truncate_pagecache(inode, offset);
939 ret = f2fs_do_collapse(inode, pg_start, pg_end);
943 new_size = i_size_read(inode) - len;
945 ret = truncate_blocks(inode, new_size, true);
947 i_size_write(inode, new_size);
952 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
955 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
956 struct address_space *mapping = inode->i_mapping;
957 pgoff_t index, pg_start, pg_end;
958 loff_t new_size = i_size_read(inode);
959 loff_t off_start, off_end;
962 if (!S_ISREG(inode->i_mode))
965 ret = inode_newsize_ok(inode, (len + offset));
969 f2fs_balance_fs(sbi);
971 if (f2fs_has_inline_data(inode)) {
972 ret = f2fs_convert_inline_inode(inode);
977 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
981 truncate_pagecache_range(inode, offset, offset + len - 1);
983 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
984 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
986 off_start = offset & (PAGE_CACHE_SIZE - 1);
987 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
989 if (pg_start == pg_end) {
990 ret = fill_zero(inode, pg_start, off_start,
991 off_end - off_start);
995 if (offset + len > new_size)
996 new_size = offset + len;
997 new_size = max_t(loff_t, new_size, offset + len);
1000 ret = fill_zero(inode, pg_start++, off_start,
1001 PAGE_CACHE_SIZE - off_start);
1005 new_size = max_t(loff_t, new_size,
1006 pg_start << PAGE_CACHE_SHIFT);
1009 for (index = pg_start; index < pg_end; index++) {
1010 struct dnode_of_data dn;
1015 ipage = get_node_page(sbi, inode->i_ino);
1016 if (IS_ERR(ipage)) {
1017 ret = PTR_ERR(ipage);
1018 f2fs_unlock_op(sbi);
1022 set_new_dnode(&dn, inode, ipage, NULL, 0);
1023 ret = f2fs_reserve_block(&dn, index);
1025 f2fs_unlock_op(sbi);
1029 if (dn.data_blkaddr != NEW_ADDR) {
1030 invalidate_blocks(sbi, dn.data_blkaddr);
1032 dn.data_blkaddr = NEW_ADDR;
1033 set_data_blkaddr(&dn);
1035 dn.data_blkaddr = NULL_ADDR;
1036 f2fs_update_extent_cache(&dn);
1038 f2fs_put_dnode(&dn);
1039 f2fs_unlock_op(sbi);
1041 new_size = max_t(loff_t, new_size,
1042 (index + 1) << PAGE_CACHE_SHIFT);
1046 ret = fill_zero(inode, pg_end, 0, off_end);
1050 new_size = max_t(loff_t, new_size, offset + len);
1055 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1056 i_size_write(inode, new_size);
1057 mark_inode_dirty(inode);
1058 update_inode_page(inode);
1064 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1066 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1067 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1071 if (!S_ISREG(inode->i_mode))
1074 new_size = i_size_read(inode) + len;
1075 if (new_size > inode->i_sb->s_maxbytes)
1078 if (offset >= i_size_read(inode))
1081 /* insert range should be aligned to block size of f2fs. */
1082 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1085 f2fs_balance_fs(sbi);
1087 if (f2fs_has_inline_data(inode)) {
1088 ret = f2fs_convert_inline_inode(inode);
1093 ret = truncate_blocks(inode, i_size_read(inode), true);
1097 /* write out all dirty pages from offset */
1098 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1102 truncate_pagecache(inode, offset);
1104 pg_start = offset >> PAGE_CACHE_SHIFT;
1105 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
1106 delta = pg_end - pg_start;
1107 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1109 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1110 struct dnode_of_data dn;
1112 block_t new_addr, old_addr;
1116 set_new_dnode(&dn, inode, NULL, NULL, 0);
1117 ret = get_dnode_of_data(&dn, idx, LOOKUP_NODE_RA);
1118 if (ret && ret != -ENOENT) {
1120 } else if (ret == -ENOENT) {
1122 } else if (dn.data_blkaddr == NULL_ADDR) {
1123 f2fs_put_dnode(&dn);
1126 new_addr = dn.data_blkaddr;
1127 truncate_data_blocks_range(&dn, 1);
1128 f2fs_put_dnode(&dn);
1131 ipage = get_node_page(sbi, inode->i_ino);
1132 if (IS_ERR(ipage)) {
1133 ret = PTR_ERR(ipage);
1137 set_new_dnode(&dn, inode, ipage, NULL, 0);
1138 ret = f2fs_reserve_block(&dn, idx + delta);
1142 old_addr = dn.data_blkaddr;
1143 f2fs_bug_on(sbi, old_addr != NEW_ADDR);
1145 if (new_addr != NEW_ADDR) {
1146 struct node_info ni;
1148 get_node_info(sbi, dn.nid, &ni);
1149 f2fs_replace_block(sbi, &dn, old_addr, new_addr,
1152 f2fs_put_dnode(&dn);
1154 f2fs_unlock_op(sbi);
1157 i_size_write(inode, new_size);
1160 f2fs_unlock_op(sbi);
1164 static int expand_inode_data(struct inode *inode, loff_t offset,
1165 loff_t len, int mode)
1167 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1168 pgoff_t index, pg_start, pg_end;
1169 loff_t new_size = i_size_read(inode);
1170 loff_t off_start, off_end;
1173 f2fs_balance_fs(sbi);
1175 ret = inode_newsize_ok(inode, (len + offset));
1179 if (f2fs_has_inline_data(inode)) {
1180 ret = f2fs_convert_inline_inode(inode);
1185 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1186 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1188 off_start = offset & (PAGE_CACHE_SIZE - 1);
1189 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1193 for (index = pg_start; index <= pg_end; index++) {
1194 struct dnode_of_data dn;
1196 if (index == pg_end && !off_end)
1199 set_new_dnode(&dn, inode, NULL, NULL, 0);
1200 ret = f2fs_reserve_block(&dn, index);
1204 if (pg_start == pg_end)
1205 new_size = offset + len;
1206 else if (index == pg_start && off_start)
1207 new_size = (index + 1) << PAGE_CACHE_SHIFT;
1208 else if (index == pg_end)
1209 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
1211 new_size += PAGE_CACHE_SIZE;
1214 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1215 i_size_read(inode) < new_size) {
1216 i_size_write(inode, new_size);
1217 mark_inode_dirty(inode);
1218 update_inode_page(inode);
1220 f2fs_unlock_op(sbi);
1225 static long f2fs_fallocate(struct file *file, int mode,
1226 loff_t offset, loff_t len)
1228 struct inode *inode = file_inode(file);
1231 if (f2fs_encrypted_inode(inode) &&
1232 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1235 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1236 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1237 FALLOC_FL_INSERT_RANGE))
1240 mutex_lock(&inode->i_mutex);
1242 if (mode & FALLOC_FL_PUNCH_HOLE) {
1243 if (offset >= inode->i_size)
1246 ret = punch_hole(inode, offset, len);
1247 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1248 ret = f2fs_collapse_range(inode, offset, len);
1249 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1250 ret = f2fs_zero_range(inode, offset, len, mode);
1251 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1252 ret = f2fs_insert_range(inode, offset, len);
1254 ret = expand_inode_data(inode, offset, len, mode);
1258 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1259 mark_inode_dirty(inode);
1263 mutex_unlock(&inode->i_mutex);
1265 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1269 static int f2fs_release_file(struct inode *inode, struct file *filp)
1271 /* some remained atomic pages should discarded */
1272 if (f2fs_is_atomic_file(inode))
1273 commit_inmem_pages(inode, true);
1274 if (f2fs_is_volatile_file(inode)) {
1275 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1276 filemap_fdatawrite(inode->i_mapping);
1277 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1282 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1283 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1285 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1289 else if (S_ISREG(mode))
1290 return flags & F2FS_REG_FLMASK;
1292 return flags & F2FS_OTHER_FLMASK;
1295 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1297 struct inode *inode = file_inode(filp);
1298 struct f2fs_inode_info *fi = F2FS_I(inode);
1299 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1300 return put_user(flags, (int __user *)arg);
1303 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1305 struct inode *inode = file_inode(filp);
1306 struct f2fs_inode_info *fi = F2FS_I(inode);
1307 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1308 unsigned int oldflags;
1311 ret = mnt_want_write_file(filp);
1315 if (!inode_owner_or_capable(inode)) {
1320 if (get_user(flags, (int __user *)arg)) {
1325 flags = f2fs_mask_flags(inode->i_mode, flags);
1327 mutex_lock(&inode->i_mutex);
1329 oldflags = fi->i_flags;
1331 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1332 if (!capable(CAP_LINUX_IMMUTABLE)) {
1333 mutex_unlock(&inode->i_mutex);
1339 flags = flags & FS_FL_USER_MODIFIABLE;
1340 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1341 fi->i_flags = flags;
1342 mutex_unlock(&inode->i_mutex);
1344 f2fs_set_inode_flags(inode);
1345 inode->i_ctime = CURRENT_TIME;
1346 mark_inode_dirty(inode);
1348 mnt_drop_write_file(filp);
1352 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1354 struct inode *inode = file_inode(filp);
1356 return put_user(inode->i_generation, (int __user *)arg);
1359 static int f2fs_ioc_start_atomic_write(struct file *filp)
1361 struct inode *inode = file_inode(filp);
1364 if (!inode_owner_or_capable(inode))
1367 f2fs_balance_fs(F2FS_I_SB(inode));
1369 if (f2fs_is_atomic_file(inode))
1372 ret = f2fs_convert_inline_inode(inode);
1376 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1380 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1382 struct inode *inode = file_inode(filp);
1385 if (!inode_owner_or_capable(inode))
1388 if (f2fs_is_volatile_file(inode))
1391 ret = mnt_want_write_file(filp);
1395 if (f2fs_is_atomic_file(inode)) {
1396 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1397 ret = commit_inmem_pages(inode, false);
1402 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1404 mnt_drop_write_file(filp);
1408 static int f2fs_ioc_start_volatile_write(struct file *filp)
1410 struct inode *inode = file_inode(filp);
1413 if (!inode_owner_or_capable(inode))
1416 if (f2fs_is_volatile_file(inode))
1419 ret = f2fs_convert_inline_inode(inode);
1423 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1427 static int f2fs_ioc_release_volatile_write(struct file *filp)
1429 struct inode *inode = file_inode(filp);
1431 if (!inode_owner_or_capable(inode))
1434 if (!f2fs_is_volatile_file(inode))
1437 if (!f2fs_is_first_block_written(inode))
1438 return truncate_partial_data_page(inode, 0, true);
1440 punch_hole(inode, 0, F2FS_BLKSIZE);
1444 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1446 struct inode *inode = file_inode(filp);
1449 if (!inode_owner_or_capable(inode))
1452 ret = mnt_want_write_file(filp);
1456 f2fs_balance_fs(F2FS_I_SB(inode));
1458 if (f2fs_is_atomic_file(inode)) {
1459 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1460 commit_inmem_pages(inode, true);
1463 if (f2fs_is_volatile_file(inode))
1464 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1466 mnt_drop_write_file(filp);
1470 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1472 struct inode *inode = file_inode(filp);
1473 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1474 struct super_block *sb = sbi->sb;
1477 if (!capable(CAP_SYS_ADMIN))
1480 if (get_user(in, (__u32 __user *)arg))
1484 case F2FS_GOING_DOWN_FULLSYNC:
1485 sb = freeze_bdev(sb->s_bdev);
1486 if (sb && !IS_ERR(sb)) {
1487 f2fs_stop_checkpoint(sbi);
1488 thaw_bdev(sb->s_bdev, sb);
1491 case F2FS_GOING_DOWN_METASYNC:
1492 /* do checkpoint only */
1493 f2fs_sync_fs(sb, 1);
1494 f2fs_stop_checkpoint(sbi);
1496 case F2FS_GOING_DOWN_NOSYNC:
1497 f2fs_stop_checkpoint(sbi);
1505 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1507 struct inode *inode = file_inode(filp);
1508 struct super_block *sb = inode->i_sb;
1509 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1510 struct fstrim_range range;
1513 if (!capable(CAP_SYS_ADMIN))
1516 if (!blk_queue_discard(q))
1519 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1523 range.minlen = max((unsigned int)range.minlen,
1524 q->limits.discard_granularity);
1525 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1529 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1535 static bool uuid_is_nonzero(__u8 u[16])
1539 for (i = 0; i < 16; i++)
1545 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1547 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1548 struct f2fs_encryption_policy policy;
1549 struct inode *inode = file_inode(filp);
1551 if (copy_from_user(&policy, (struct f2fs_encryption_policy __user *)arg,
1555 return f2fs_process_policy(&policy, inode);
1561 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1563 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1564 struct f2fs_encryption_policy policy;
1565 struct inode *inode = file_inode(filp);
1568 err = f2fs_get_policy(inode, &policy);
1572 if (copy_to_user((struct f2fs_encryption_policy __user *)arg, &policy,
1581 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1583 struct inode *inode = file_inode(filp);
1584 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1587 if (!f2fs_sb_has_crypto(inode->i_sb))
1590 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1593 err = mnt_want_write_file(filp);
1597 /* update superblock with uuid */
1598 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1600 err = f2fs_commit_super(sbi, false);
1602 mnt_drop_write_file(filp);
1605 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1609 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1615 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1617 struct inode *inode = file_inode(filp);
1618 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1621 if (!capable(CAP_SYS_ADMIN))
1624 if (get_user(count, (__u32 __user *)arg))
1627 if (!count || count > F2FS_BATCH_GC_MAX_NUM)
1630 for (i = 0; i < count; i++) {
1631 if (!mutex_trylock(&sbi->gc_mutex))
1638 if (put_user(i, (__u32 __user *)arg))
1644 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1647 case F2FS_IOC_GETFLAGS:
1648 return f2fs_ioc_getflags(filp, arg);
1649 case F2FS_IOC_SETFLAGS:
1650 return f2fs_ioc_setflags(filp, arg);
1651 case F2FS_IOC_GETVERSION:
1652 return f2fs_ioc_getversion(filp, arg);
1653 case F2FS_IOC_START_ATOMIC_WRITE:
1654 return f2fs_ioc_start_atomic_write(filp);
1655 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1656 return f2fs_ioc_commit_atomic_write(filp);
1657 case F2FS_IOC_START_VOLATILE_WRITE:
1658 return f2fs_ioc_start_volatile_write(filp);
1659 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1660 return f2fs_ioc_release_volatile_write(filp);
1661 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1662 return f2fs_ioc_abort_volatile_write(filp);
1663 case F2FS_IOC_SHUTDOWN:
1664 return f2fs_ioc_shutdown(filp, arg);
1666 return f2fs_ioc_fitrim(filp, arg);
1667 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1668 return f2fs_ioc_set_encryption_policy(filp, arg);
1669 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1670 return f2fs_ioc_get_encryption_policy(filp, arg);
1671 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1672 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1673 case F2FS_IOC_GARBAGE_COLLECT:
1674 return f2fs_ioc_gc(filp, arg);
1680 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1682 struct inode *inode = file_inode(iocb->ki_filp);
1684 if (f2fs_encrypted_inode(inode) &&
1685 !f2fs_has_encryption_key(inode) &&
1686 f2fs_get_encryption_info(inode))
1689 return generic_file_write_iter(iocb, from);
1692 #ifdef CONFIG_COMPAT
1693 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1696 case F2FS_IOC32_GETFLAGS:
1697 cmd = F2FS_IOC_GETFLAGS;
1699 case F2FS_IOC32_SETFLAGS:
1700 cmd = F2FS_IOC_SETFLAGS;
1703 return -ENOIOCTLCMD;
1705 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1709 const struct file_operations f2fs_file_operations = {
1710 .llseek = f2fs_llseek,
1711 .read_iter = generic_file_read_iter,
1712 .write_iter = f2fs_file_write_iter,
1713 .open = f2fs_file_open,
1714 .release = f2fs_release_file,
1715 .mmap = f2fs_file_mmap,
1716 .fsync = f2fs_sync_file,
1717 .fallocate = f2fs_fallocate,
1718 .unlocked_ioctl = f2fs_ioctl,
1719 #ifdef CONFIG_COMPAT
1720 .compat_ioctl = f2fs_compat_ioctl,
1722 .splice_read = generic_file_splice_read,
1723 .splice_write = iter_file_splice_write,