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_filemap_fault(struct vm_fault *vmf)
38 struct inode *inode = file_inode(vmf->vma->vm_file);
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 err = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
48 static int f2fs_vm_page_mkwrite(struct vm_fault *vmf)
50 struct page *page = vmf->page;
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 struct dnode_of_data dn;
56 if (unlikely(f2fs_cp_error(sbi))) {
61 sb_start_pagefault(inode->i_sb);
63 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
65 /* block allocation */
67 set_new_dnode(&dn, inode, NULL, NULL, 0);
68 err = f2fs_reserve_block(&dn, page->index);
76 f2fs_balance_fs(sbi, dn.node_changed);
78 file_update_time(vmf->vma->vm_file);
79 down_read(&F2FS_I(inode)->i_mmap_sem);
81 if (unlikely(page->mapping != inode->i_mapping ||
82 page_offset(page) > i_size_read(inode) ||
83 !PageUptodate(page))) {
90 * check to see if the page is mapped already (no holes)
92 if (PageMappedToDisk(page))
95 /* page is wholly or partially inside EOF */
96 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
99 offset = i_size_read(inode) & ~PAGE_MASK;
100 zero_user_segment(page, offset, PAGE_SIZE);
102 set_page_dirty(page);
103 if (!PageUptodate(page))
104 SetPageUptodate(page);
106 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
108 trace_f2fs_vm_page_mkwrite(page, DATA);
111 f2fs_wait_on_page_writeback(page, DATA, false);
113 /* wait for GCed encrypted page writeback */
114 if (f2fs_encrypted_file(inode))
115 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
118 up_read(&F2FS_I(inode)->i_mmap_sem);
120 sb_end_pagefault(inode->i_sb);
121 f2fs_update_time(sbi, REQ_TIME);
123 return block_page_mkwrite_return(err);
126 static const struct vm_operations_struct f2fs_file_vm_ops = {
127 .fault = f2fs_filemap_fault,
128 .map_pages = filemap_map_pages,
129 .page_mkwrite = f2fs_vm_page_mkwrite,
132 static int get_parent_ino(struct inode *inode, nid_t *pino)
134 struct dentry *dentry;
136 inode = igrab(inode);
137 dentry = d_find_any_alias(inode);
142 *pino = parent_ino(dentry);
147 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
149 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
150 enum cp_reason_type cp_reason = CP_NO_NEEDED;
152 if (!S_ISREG(inode->i_mode))
153 cp_reason = CP_NON_REGULAR;
154 else if (inode->i_nlink != 1)
155 cp_reason = CP_HARDLINK;
156 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
157 cp_reason = CP_SB_NEED_CP;
158 else if (file_wrong_pino(inode))
159 cp_reason = CP_WRONG_PINO;
160 else if (!space_for_roll_forward(sbi))
161 cp_reason = CP_NO_SPC_ROLL;
162 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
163 cp_reason = CP_NODE_NEED_CP;
164 else if (test_opt(sbi, FASTBOOT))
165 cp_reason = CP_FASTBOOT_MODE;
166 else if (F2FS_OPTION(sbi).active_logs == 2)
167 cp_reason = CP_SPEC_LOG_NUM;
168 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
169 need_dentry_mark(sbi, inode->i_ino) &&
170 exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
171 cp_reason = CP_RECOVER_DIR;
176 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
178 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
180 /* But we need to avoid that there are some inode updates */
181 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
187 static void try_to_fix_pino(struct inode *inode)
189 struct f2fs_inode_info *fi = F2FS_I(inode);
192 down_write(&fi->i_sem);
193 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
194 get_parent_ino(inode, &pino)) {
195 f2fs_i_pino_write(inode, pino);
196 file_got_pino(inode);
198 up_write(&fi->i_sem);
201 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
202 int datasync, bool atomic)
204 struct inode *inode = file->f_mapping->host;
205 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
206 nid_t ino = inode->i_ino;
208 enum cp_reason_type cp_reason = 0;
209 struct writeback_control wbc = {
210 .sync_mode = WB_SYNC_ALL,
211 .nr_to_write = LONG_MAX,
215 if (unlikely(f2fs_readonly(inode->i_sb)))
218 trace_f2fs_sync_file_enter(inode);
220 /* if fdatasync is triggered, let's do in-place-update */
221 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
222 set_inode_flag(inode, FI_NEED_IPU);
223 ret = file_write_and_wait_range(file, start, end);
224 clear_inode_flag(inode, FI_NEED_IPU);
227 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
231 /* if the inode is dirty, let's recover all the time */
232 if (!f2fs_skip_inode_update(inode, datasync)) {
233 f2fs_write_inode(inode, NULL);
238 * if there is no written data, don't waste time to write recovery info.
240 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
241 !exist_written_data(sbi, ino, APPEND_INO)) {
243 /* it may call write_inode just prior to fsync */
244 if (need_inode_page_update(sbi, ino))
247 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
248 exist_written_data(sbi, ino, UPDATE_INO))
254 * Both of fdatasync() and fsync() are able to be recovered from
257 down_read(&F2FS_I(inode)->i_sem);
258 cp_reason = need_do_checkpoint(inode);
259 up_read(&F2FS_I(inode)->i_sem);
262 /* all the dirty node pages should be flushed for POR */
263 ret = f2fs_sync_fs(inode->i_sb, 1);
266 * We've secured consistency through sync_fs. Following pino
267 * will be used only for fsynced inodes after checkpoint.
269 try_to_fix_pino(inode);
270 clear_inode_flag(inode, FI_APPEND_WRITE);
271 clear_inode_flag(inode, FI_UPDATE_WRITE);
275 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
279 /* if cp_error was enabled, we should avoid infinite loop */
280 if (unlikely(f2fs_cp_error(sbi))) {
285 if (need_inode_block_update(sbi, ino)) {
286 f2fs_mark_inode_dirty_sync(inode, true);
287 f2fs_write_inode(inode, NULL);
292 * If it's atomic_write, it's just fine to keep write ordering. So
293 * here we don't need to wait for node write completion, since we use
294 * node chain which serializes node blocks. If one of node writes are
295 * reordered, we can see simply broken chain, resulting in stopping
296 * roll-forward recovery. It means we'll recover all or none node blocks
300 ret = wait_on_node_pages_writeback(sbi, ino);
305 /* once recovery info is written, don't need to tack this */
306 remove_ino_entry(sbi, ino, APPEND_INO);
307 clear_inode_flag(inode, FI_APPEND_WRITE);
310 ret = f2fs_issue_flush(sbi, inode->i_ino);
312 remove_ino_entry(sbi, ino, UPDATE_INO);
313 clear_inode_flag(inode, FI_UPDATE_WRITE);
314 remove_ino_entry(sbi, ino, FLUSH_INO);
316 f2fs_update_time(sbi, REQ_TIME);
318 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
319 f2fs_trace_ios(NULL, 1);
323 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
325 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
327 return f2fs_do_sync_file(file, start, end, datasync, false);
330 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
331 pgoff_t pgofs, int whence)
336 if (whence != SEEK_DATA)
339 /* find first dirty page index */
340 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
349 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
354 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
355 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
359 if (blkaddr == NULL_ADDR)
366 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
368 struct inode *inode = file->f_mapping->host;
369 loff_t maxbytes = inode->i_sb->s_maxbytes;
370 struct dnode_of_data dn;
371 pgoff_t pgofs, end_offset, dirty;
372 loff_t data_ofs = offset;
378 isize = i_size_read(inode);
382 /* handle inline data case */
383 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
384 if (whence == SEEK_HOLE)
389 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
391 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
393 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
394 set_new_dnode(&dn, inode, NULL, NULL, 0);
395 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
396 if (err && err != -ENOENT) {
398 } else if (err == -ENOENT) {
399 /* direct node does not exists */
400 if (whence == SEEK_DATA) {
401 pgofs = get_next_page_offset(&dn, pgofs);
408 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
410 /* find data/hole in dnode block */
411 for (; dn.ofs_in_node < end_offset;
412 dn.ofs_in_node++, pgofs++,
413 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
415 blkaddr = datablock_addr(dn.inode,
416 dn.node_page, dn.ofs_in_node);
418 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
426 if (whence == SEEK_DATA)
429 if (whence == SEEK_HOLE && data_ofs > isize)
432 return vfs_setpos(file, data_ofs, maxbytes);
438 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
440 struct inode *inode = file->f_mapping->host;
441 loff_t maxbytes = inode->i_sb->s_maxbytes;
447 return generic_file_llseek_size(file, offset, whence,
448 maxbytes, i_size_read(inode));
453 return f2fs_seek_block(file, offset, whence);
459 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
461 struct inode *inode = file_inode(file);
464 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
467 /* we don't need to use inline_data strictly */
468 err = f2fs_convert_inline_inode(inode);
473 vma->vm_ops = &f2fs_file_vm_ops;
477 static int f2fs_file_open(struct inode *inode, struct file *filp)
479 int err = fscrypt_file_open(inode, filp);
484 filp->f_mode |= FMODE_NOWAIT;
486 return dquot_file_open(inode, filp);
489 void truncate_data_blocks_range(struct dnode_of_data *dn, int count)
491 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
492 struct f2fs_node *raw_node;
493 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
497 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
498 base = get_extra_isize(dn->inode);
500 raw_node = F2FS_NODE(dn->node_page);
501 addr = blkaddr_in_node(raw_node) + base + ofs;
503 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
504 block_t blkaddr = le32_to_cpu(*addr);
505 if (blkaddr == NULL_ADDR)
508 dn->data_blkaddr = NULL_ADDR;
509 set_data_blkaddr(dn);
510 invalidate_blocks(sbi, blkaddr);
511 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
512 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
519 * once we invalidate valid blkaddr in range [ofs, ofs + count],
520 * we will invalidate all blkaddr in the whole range.
522 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
524 f2fs_update_extent_cache_range(dn, fofs, 0, len);
525 dec_valid_block_count(sbi, dn->inode, nr_free);
527 dn->ofs_in_node = ofs;
529 f2fs_update_time(sbi, REQ_TIME);
530 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
531 dn->ofs_in_node, nr_free);
534 void truncate_data_blocks(struct dnode_of_data *dn)
536 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
539 static int truncate_partial_data_page(struct inode *inode, u64 from,
542 unsigned offset = from & (PAGE_SIZE - 1);
543 pgoff_t index = from >> PAGE_SHIFT;
544 struct address_space *mapping = inode->i_mapping;
547 if (!offset && !cache_only)
551 page = find_lock_page(mapping, index);
552 if (page && PageUptodate(page))
554 f2fs_put_page(page, 1);
558 page = get_lock_data_page(inode, index, true);
560 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
562 f2fs_wait_on_page_writeback(page, DATA, true);
563 zero_user(page, offset, PAGE_SIZE - offset);
565 /* An encrypted inode should have a key and truncate the last page. */
566 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
568 set_page_dirty(page);
569 f2fs_put_page(page, 1);
573 int truncate_blocks(struct inode *inode, u64 from, bool lock)
575 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
576 struct dnode_of_data dn;
578 int count = 0, err = 0;
580 bool truncate_page = false;
582 trace_f2fs_truncate_blocks_enter(inode, from);
584 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
586 if (free_from >= sbi->max_file_blocks)
592 ipage = get_node_page(sbi, inode->i_ino);
594 err = PTR_ERR(ipage);
598 if (f2fs_has_inline_data(inode)) {
599 truncate_inline_inode(inode, ipage, from);
600 f2fs_put_page(ipage, 1);
601 truncate_page = true;
605 set_new_dnode(&dn, inode, ipage, NULL, 0);
606 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
613 count = ADDRS_PER_PAGE(dn.node_page, inode);
615 count -= dn.ofs_in_node;
616 f2fs_bug_on(sbi, count < 0);
618 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
619 truncate_data_blocks_range(&dn, count);
625 err = truncate_inode_blocks(inode, free_from);
630 /* lastly zero out the first data page */
632 err = truncate_partial_data_page(inode, from, truncate_page);
634 trace_f2fs_truncate_blocks_exit(inode, err);
638 int f2fs_truncate(struct inode *inode)
642 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
645 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
646 S_ISLNK(inode->i_mode)))
649 trace_f2fs_truncate(inode);
651 #ifdef CONFIG_F2FS_FAULT_INJECTION
652 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
653 f2fs_show_injection_info(FAULT_TRUNCATE);
657 /* we should check inline_data size */
658 if (!f2fs_may_inline_data(inode)) {
659 err = f2fs_convert_inline_inode(inode);
664 err = truncate_blocks(inode, i_size_read(inode), true);
668 inode->i_mtime = inode->i_ctime = current_time(inode);
669 f2fs_mark_inode_dirty_sync(inode, false);
673 int f2fs_getattr(const struct path *path, struct kstat *stat,
674 u32 request_mask, unsigned int query_flags)
676 struct inode *inode = d_inode(path->dentry);
677 struct f2fs_inode_info *fi = F2FS_I(inode);
678 struct f2fs_inode *ri;
681 if (f2fs_has_extra_attr(inode) &&
682 f2fs_sb_has_inode_crtime(inode->i_sb) &&
683 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
684 stat->result_mask |= STATX_BTIME;
685 stat->btime.tv_sec = fi->i_crtime.tv_sec;
686 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
689 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
690 if (flags & FS_APPEND_FL)
691 stat->attributes |= STATX_ATTR_APPEND;
692 if (flags & FS_COMPR_FL)
693 stat->attributes |= STATX_ATTR_COMPRESSED;
694 if (f2fs_encrypted_inode(inode))
695 stat->attributes |= STATX_ATTR_ENCRYPTED;
696 if (flags & FS_IMMUTABLE_FL)
697 stat->attributes |= STATX_ATTR_IMMUTABLE;
698 if (flags & FS_NODUMP_FL)
699 stat->attributes |= STATX_ATTR_NODUMP;
701 stat->attributes_mask |= (STATX_ATTR_APPEND |
702 STATX_ATTR_COMPRESSED |
703 STATX_ATTR_ENCRYPTED |
704 STATX_ATTR_IMMUTABLE |
707 generic_fillattr(inode, stat);
709 /* we need to show initial sectors used for inline_data/dentries */
710 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
711 f2fs_has_inline_dentry(inode))
712 stat->blocks += (stat->size + 511) >> 9;
717 #ifdef CONFIG_F2FS_FS_POSIX_ACL
718 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
720 unsigned int ia_valid = attr->ia_valid;
722 if (ia_valid & ATTR_UID)
723 inode->i_uid = attr->ia_uid;
724 if (ia_valid & ATTR_GID)
725 inode->i_gid = attr->ia_gid;
726 if (ia_valid & ATTR_ATIME)
727 inode->i_atime = timespec_trunc(attr->ia_atime,
728 inode->i_sb->s_time_gran);
729 if (ia_valid & ATTR_MTIME)
730 inode->i_mtime = timespec_trunc(attr->ia_mtime,
731 inode->i_sb->s_time_gran);
732 if (ia_valid & ATTR_CTIME)
733 inode->i_ctime = timespec_trunc(attr->ia_ctime,
734 inode->i_sb->s_time_gran);
735 if (ia_valid & ATTR_MODE) {
736 umode_t mode = attr->ia_mode;
738 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
740 set_acl_inode(inode, mode);
744 #define __setattr_copy setattr_copy
747 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
749 struct inode *inode = d_inode(dentry);
751 bool size_changed = false;
753 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
756 err = setattr_prepare(dentry, attr);
760 err = fscrypt_prepare_setattr(dentry, attr);
764 if (is_quota_modification(inode, attr)) {
765 err = dquot_initialize(inode);
769 if ((attr->ia_valid & ATTR_UID &&
770 !uid_eq(attr->ia_uid, inode->i_uid)) ||
771 (attr->ia_valid & ATTR_GID &&
772 !gid_eq(attr->ia_gid, inode->i_gid))) {
773 err = dquot_transfer(inode, attr);
778 if (attr->ia_valid & ATTR_SIZE) {
779 if (attr->ia_size <= i_size_read(inode)) {
780 down_write(&F2FS_I(inode)->i_mmap_sem);
781 truncate_setsize(inode, attr->ia_size);
782 err = f2fs_truncate(inode);
783 up_write(&F2FS_I(inode)->i_mmap_sem);
788 * do not trim all blocks after i_size if target size is
789 * larger than i_size.
791 down_write(&F2FS_I(inode)->i_mmap_sem);
792 truncate_setsize(inode, attr->ia_size);
793 up_write(&F2FS_I(inode)->i_mmap_sem);
795 /* should convert inline inode here */
796 if (!f2fs_may_inline_data(inode)) {
797 err = f2fs_convert_inline_inode(inode);
801 inode->i_mtime = inode->i_ctime = current_time(inode);
804 down_write(&F2FS_I(inode)->i_sem);
805 F2FS_I(inode)->last_disk_size = i_size_read(inode);
806 up_write(&F2FS_I(inode)->i_sem);
811 __setattr_copy(inode, attr);
813 if (attr->ia_valid & ATTR_MODE) {
814 err = posix_acl_chmod(inode, get_inode_mode(inode));
815 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
816 inode->i_mode = F2FS_I(inode)->i_acl_mode;
817 clear_inode_flag(inode, FI_ACL_MODE);
821 /* file size may changed here */
822 f2fs_mark_inode_dirty_sync(inode, size_changed);
824 /* inode change will produce dirty node pages flushed by checkpoint */
825 f2fs_balance_fs(F2FS_I_SB(inode), true);
830 const struct inode_operations f2fs_file_inode_operations = {
831 .getattr = f2fs_getattr,
832 .setattr = f2fs_setattr,
833 .get_acl = f2fs_get_acl,
834 .set_acl = f2fs_set_acl,
835 #ifdef CONFIG_F2FS_FS_XATTR
836 .listxattr = f2fs_listxattr,
838 .fiemap = f2fs_fiemap,
841 static int fill_zero(struct inode *inode, pgoff_t index,
842 loff_t start, loff_t len)
844 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
850 f2fs_balance_fs(sbi, true);
853 page = get_new_data_page(inode, NULL, index, false);
857 return PTR_ERR(page);
859 f2fs_wait_on_page_writeback(page, DATA, true);
860 zero_user(page, start, len);
861 set_page_dirty(page);
862 f2fs_put_page(page, 1);
866 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
870 while (pg_start < pg_end) {
871 struct dnode_of_data dn;
872 pgoff_t end_offset, count;
874 set_new_dnode(&dn, inode, NULL, NULL, 0);
875 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
877 if (err == -ENOENT) {
878 pg_start = get_next_page_offset(&dn, pg_start);
884 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
885 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
887 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
889 truncate_data_blocks_range(&dn, count);
897 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
899 pgoff_t pg_start, pg_end;
900 loff_t off_start, off_end;
903 ret = f2fs_convert_inline_inode(inode);
907 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
908 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
910 off_start = offset & (PAGE_SIZE - 1);
911 off_end = (offset + len) & (PAGE_SIZE - 1);
913 if (pg_start == pg_end) {
914 ret = fill_zero(inode, pg_start, off_start,
915 off_end - off_start);
920 ret = fill_zero(inode, pg_start++, off_start,
921 PAGE_SIZE - off_start);
926 ret = fill_zero(inode, pg_end, 0, off_end);
931 if (pg_start < pg_end) {
932 struct address_space *mapping = inode->i_mapping;
933 loff_t blk_start, blk_end;
934 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
936 f2fs_balance_fs(sbi, true);
938 blk_start = (loff_t)pg_start << PAGE_SHIFT;
939 blk_end = (loff_t)pg_end << PAGE_SHIFT;
940 down_write(&F2FS_I(inode)->i_mmap_sem);
941 truncate_inode_pages_range(mapping, blk_start,
945 ret = truncate_hole(inode, pg_start, pg_end);
947 up_write(&F2FS_I(inode)->i_mmap_sem);
954 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
955 int *do_replace, pgoff_t off, pgoff_t len)
957 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
958 struct dnode_of_data dn;
962 set_new_dnode(&dn, inode, NULL, NULL, 0);
963 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
964 if (ret && ret != -ENOENT) {
966 } else if (ret == -ENOENT) {
967 if (dn.max_level == 0)
969 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
975 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
976 dn.ofs_in_node, len);
977 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
978 *blkaddr = datablock_addr(dn.inode,
979 dn.node_page, dn.ofs_in_node);
980 if (!is_checkpointed_data(sbi, *blkaddr)) {
982 if (test_opt(sbi, LFS)) {
987 /* do not invalidate this block address */
988 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1001 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1002 int *do_replace, pgoff_t off, int len)
1004 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1005 struct dnode_of_data dn;
1008 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1009 if (*do_replace == 0)
1012 set_new_dnode(&dn, inode, NULL, NULL, 0);
1013 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1015 dec_valid_block_count(sbi, inode, 1);
1016 invalidate_blocks(sbi, *blkaddr);
1018 f2fs_update_data_blkaddr(&dn, *blkaddr);
1020 f2fs_put_dnode(&dn);
1025 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1026 block_t *blkaddr, int *do_replace,
1027 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1029 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1034 if (blkaddr[i] == NULL_ADDR && !full) {
1039 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1040 struct dnode_of_data dn;
1041 struct node_info ni;
1045 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1046 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1050 get_node_info(sbi, dn.nid, &ni);
1051 ilen = min((pgoff_t)
1052 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1053 dn.ofs_in_node, len - i);
1055 dn.data_blkaddr = datablock_addr(dn.inode,
1056 dn.node_page, dn.ofs_in_node);
1057 truncate_data_blocks_range(&dn, 1);
1059 if (do_replace[i]) {
1060 f2fs_i_blocks_write(src_inode,
1062 f2fs_i_blocks_write(dst_inode,
1064 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1065 blkaddr[i], ni.version, true, false);
1071 new_size = (dst + i) << PAGE_SHIFT;
1072 if (dst_inode->i_size < new_size)
1073 f2fs_i_size_write(dst_inode, new_size);
1074 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1076 f2fs_put_dnode(&dn);
1078 struct page *psrc, *pdst;
1080 psrc = get_lock_data_page(src_inode, src + i, true);
1082 return PTR_ERR(psrc);
1083 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1086 f2fs_put_page(psrc, 1);
1087 return PTR_ERR(pdst);
1089 f2fs_copy_page(psrc, pdst);
1090 set_page_dirty(pdst);
1091 f2fs_put_page(pdst, 1);
1092 f2fs_put_page(psrc, 1);
1094 ret = truncate_hole(src_inode, src + i, src + i + 1);
1103 static int __exchange_data_block(struct inode *src_inode,
1104 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1105 pgoff_t len, bool full)
1107 block_t *src_blkaddr;
1113 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1115 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1116 sizeof(block_t) * olen, GFP_KERNEL);
1120 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1121 sizeof(int) * olen, GFP_KERNEL);
1123 kvfree(src_blkaddr);
1127 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1128 do_replace, src, olen);
1132 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1133 do_replace, src, dst, olen, full);
1141 kvfree(src_blkaddr);
1147 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1148 kvfree(src_blkaddr);
1153 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1155 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1156 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1159 f2fs_balance_fs(sbi, true);
1162 f2fs_drop_extent_tree(inode);
1164 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1165 f2fs_unlock_op(sbi);
1169 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1171 pgoff_t pg_start, pg_end;
1175 if (offset + len >= i_size_read(inode))
1178 /* collapse range should be aligned to block size of f2fs. */
1179 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1182 ret = f2fs_convert_inline_inode(inode);
1186 pg_start = offset >> PAGE_SHIFT;
1187 pg_end = (offset + len) >> PAGE_SHIFT;
1189 /* avoid gc operation during block exchange */
1190 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1192 down_write(&F2FS_I(inode)->i_mmap_sem);
1193 /* write out all dirty pages from offset */
1194 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1198 truncate_pagecache(inode, offset);
1200 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1204 /* write out all moved pages, if possible */
1205 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1206 truncate_pagecache(inode, offset);
1208 new_size = i_size_read(inode) - len;
1209 truncate_pagecache(inode, new_size);
1211 ret = truncate_blocks(inode, new_size, true);
1213 f2fs_i_size_write(inode, new_size);
1215 up_write(&F2FS_I(inode)->i_mmap_sem);
1216 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1220 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1223 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1224 pgoff_t index = start;
1225 unsigned int ofs_in_node = dn->ofs_in_node;
1229 for (; index < end; index++, dn->ofs_in_node++) {
1230 if (datablock_addr(dn->inode, dn->node_page,
1231 dn->ofs_in_node) == NULL_ADDR)
1235 dn->ofs_in_node = ofs_in_node;
1236 ret = reserve_new_blocks(dn, count);
1240 dn->ofs_in_node = ofs_in_node;
1241 for (index = start; index < end; index++, dn->ofs_in_node++) {
1242 dn->data_blkaddr = datablock_addr(dn->inode,
1243 dn->node_page, dn->ofs_in_node);
1245 * reserve_new_blocks will not guarantee entire block
1248 if (dn->data_blkaddr == NULL_ADDR) {
1252 if (dn->data_blkaddr != NEW_ADDR) {
1253 invalidate_blocks(sbi, dn->data_blkaddr);
1254 dn->data_blkaddr = NEW_ADDR;
1255 set_data_blkaddr(dn);
1259 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1264 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1267 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1268 struct address_space *mapping = inode->i_mapping;
1269 pgoff_t index, pg_start, pg_end;
1270 loff_t new_size = i_size_read(inode);
1271 loff_t off_start, off_end;
1274 ret = inode_newsize_ok(inode, (len + offset));
1278 ret = f2fs_convert_inline_inode(inode);
1282 down_write(&F2FS_I(inode)->i_mmap_sem);
1283 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1287 truncate_pagecache_range(inode, offset, offset + len - 1);
1289 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1290 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1292 off_start = offset & (PAGE_SIZE - 1);
1293 off_end = (offset + len) & (PAGE_SIZE - 1);
1295 if (pg_start == pg_end) {
1296 ret = fill_zero(inode, pg_start, off_start,
1297 off_end - off_start);
1301 new_size = max_t(loff_t, new_size, offset + len);
1304 ret = fill_zero(inode, pg_start++, off_start,
1305 PAGE_SIZE - off_start);
1309 new_size = max_t(loff_t, new_size,
1310 (loff_t)pg_start << PAGE_SHIFT);
1313 for (index = pg_start; index < pg_end;) {
1314 struct dnode_of_data dn;
1315 unsigned int end_offset;
1320 set_new_dnode(&dn, inode, NULL, NULL, 0);
1321 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1323 f2fs_unlock_op(sbi);
1327 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1328 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1330 ret = f2fs_do_zero_range(&dn, index, end);
1331 f2fs_put_dnode(&dn);
1332 f2fs_unlock_op(sbi);
1334 f2fs_balance_fs(sbi, dn.node_changed);
1340 new_size = max_t(loff_t, new_size,
1341 (loff_t)index << PAGE_SHIFT);
1345 ret = fill_zero(inode, pg_end, 0, off_end);
1349 new_size = max_t(loff_t, new_size, offset + len);
1354 if (new_size > i_size_read(inode)) {
1355 if (mode & FALLOC_FL_KEEP_SIZE)
1356 file_set_keep_isize(inode);
1358 f2fs_i_size_write(inode, new_size);
1361 up_write(&F2FS_I(inode)->i_mmap_sem);
1366 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1368 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1369 pgoff_t nr, pg_start, pg_end, delta, idx;
1373 new_size = i_size_read(inode) + len;
1374 ret = inode_newsize_ok(inode, new_size);
1378 if (offset >= i_size_read(inode))
1381 /* insert range should be aligned to block size of f2fs. */
1382 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1385 ret = f2fs_convert_inline_inode(inode);
1389 f2fs_balance_fs(sbi, true);
1391 /* avoid gc operation during block exchange */
1392 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1394 down_write(&F2FS_I(inode)->i_mmap_sem);
1395 ret = truncate_blocks(inode, i_size_read(inode), true);
1399 /* write out all dirty pages from offset */
1400 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1404 truncate_pagecache(inode, offset);
1406 pg_start = offset >> PAGE_SHIFT;
1407 pg_end = (offset + len) >> PAGE_SHIFT;
1408 delta = pg_end - pg_start;
1409 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1411 while (!ret && idx > pg_start) {
1412 nr = idx - pg_start;
1418 f2fs_drop_extent_tree(inode);
1420 ret = __exchange_data_block(inode, inode, idx,
1421 idx + delta, nr, false);
1422 f2fs_unlock_op(sbi);
1425 /* write out all moved pages, if possible */
1426 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1427 truncate_pagecache(inode, offset);
1430 f2fs_i_size_write(inode, new_size);
1432 up_write(&F2FS_I(inode)->i_mmap_sem);
1433 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1437 static int expand_inode_data(struct inode *inode, loff_t offset,
1438 loff_t len, int mode)
1440 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1441 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1442 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1444 loff_t new_size = i_size_read(inode);
1448 err = inode_newsize_ok(inode, (len + offset));
1452 err = f2fs_convert_inline_inode(inode);
1456 f2fs_balance_fs(sbi, true);
1458 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1459 off_end = (offset + len) & (PAGE_SIZE - 1);
1461 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1462 map.m_len = pg_end - map.m_lblk;
1466 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1473 last_off = map.m_lblk + map.m_len - 1;
1475 /* update new size to the failed position */
1476 new_size = (last_off == pg_end) ? offset + len:
1477 (loff_t)(last_off + 1) << PAGE_SHIFT;
1479 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1482 if (new_size > i_size_read(inode)) {
1483 if (mode & FALLOC_FL_KEEP_SIZE)
1484 file_set_keep_isize(inode);
1486 f2fs_i_size_write(inode, new_size);
1492 static long f2fs_fallocate(struct file *file, int mode,
1493 loff_t offset, loff_t len)
1495 struct inode *inode = file_inode(file);
1498 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1501 /* f2fs only support ->fallocate for regular file */
1502 if (!S_ISREG(inode->i_mode))
1505 if (f2fs_encrypted_inode(inode) &&
1506 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1509 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1510 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1511 FALLOC_FL_INSERT_RANGE))
1516 if (mode & FALLOC_FL_PUNCH_HOLE) {
1517 if (offset >= inode->i_size)
1520 ret = punch_hole(inode, offset, len);
1521 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1522 ret = f2fs_collapse_range(inode, offset, len);
1523 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1524 ret = f2fs_zero_range(inode, offset, len, mode);
1525 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1526 ret = f2fs_insert_range(inode, offset, len);
1528 ret = expand_inode_data(inode, offset, len, mode);
1532 inode->i_mtime = inode->i_ctime = current_time(inode);
1533 f2fs_mark_inode_dirty_sync(inode, false);
1534 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1538 inode_unlock(inode);
1540 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1544 static int f2fs_release_file(struct inode *inode, struct file *filp)
1547 * f2fs_relase_file is called at every close calls. So we should
1548 * not drop any inmemory pages by close called by other process.
1550 if (!(filp->f_mode & FMODE_WRITE) ||
1551 atomic_read(&inode->i_writecount) != 1)
1554 /* some remained atomic pages should discarded */
1555 if (f2fs_is_atomic_file(inode))
1556 drop_inmem_pages(inode);
1557 if (f2fs_is_volatile_file(inode)) {
1558 clear_inode_flag(inode, FI_VOLATILE_FILE);
1559 stat_dec_volatile_write(inode);
1560 set_inode_flag(inode, FI_DROP_CACHE);
1561 filemap_fdatawrite(inode->i_mapping);
1562 clear_inode_flag(inode, FI_DROP_CACHE);
1567 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1569 struct inode *inode = file_inode(file);
1572 * If the process doing a transaction is crashed, we should do
1573 * roll-back. Otherwise, other reader/write can see corrupted database
1574 * until all the writers close its file. Since this should be done
1575 * before dropping file lock, it needs to do in ->flush.
1577 if (f2fs_is_atomic_file(inode) &&
1578 F2FS_I(inode)->inmem_task == current)
1579 drop_inmem_pages(inode);
1583 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1585 struct inode *inode = file_inode(filp);
1586 struct f2fs_inode_info *fi = F2FS_I(inode);
1587 unsigned int flags = fi->i_flags &
1588 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1589 return put_user(flags, (int __user *)arg);
1592 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1594 struct f2fs_inode_info *fi = F2FS_I(inode);
1595 unsigned int oldflags;
1597 /* Is it quota file? Do not allow user to mess with it */
1598 if (IS_NOQUOTA(inode))
1601 flags = f2fs_mask_flags(inode->i_mode, flags);
1603 oldflags = fi->i_flags;
1605 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1606 if (!capable(CAP_LINUX_IMMUTABLE))
1609 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1610 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1611 fi->i_flags = flags;
1613 if (fi->i_flags & FS_PROJINHERIT_FL)
1614 set_inode_flag(inode, FI_PROJ_INHERIT);
1616 clear_inode_flag(inode, FI_PROJ_INHERIT);
1618 inode->i_ctime = current_time(inode);
1619 f2fs_set_inode_flags(inode);
1620 f2fs_mark_inode_dirty_sync(inode, false);
1624 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1626 struct inode *inode = file_inode(filp);
1630 if (!inode_owner_or_capable(inode))
1633 if (get_user(flags, (int __user *)arg))
1636 ret = mnt_want_write_file(filp);
1642 ret = __f2fs_ioc_setflags(inode, flags);
1644 inode_unlock(inode);
1645 mnt_drop_write_file(filp);
1649 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1651 struct inode *inode = file_inode(filp);
1653 return put_user(inode->i_generation, (int __user *)arg);
1656 static int f2fs_ioc_start_atomic_write(struct file *filp)
1658 struct inode *inode = file_inode(filp);
1661 if (!inode_owner_or_capable(inode))
1664 if (!S_ISREG(inode->i_mode))
1667 ret = mnt_want_write_file(filp);
1673 if (f2fs_is_atomic_file(inode))
1676 ret = f2fs_convert_inline_inode(inode);
1680 set_inode_flag(inode, FI_ATOMIC_FILE);
1681 set_inode_flag(inode, FI_HOT_DATA);
1682 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1684 if (!get_dirty_pages(inode))
1687 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1688 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1689 inode->i_ino, get_dirty_pages(inode));
1690 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1692 clear_inode_flag(inode, FI_ATOMIC_FILE);
1693 clear_inode_flag(inode, FI_HOT_DATA);
1698 F2FS_I(inode)->inmem_task = current;
1699 stat_inc_atomic_write(inode);
1700 stat_update_max_atomic_write(inode);
1702 inode_unlock(inode);
1703 mnt_drop_write_file(filp);
1707 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1709 struct inode *inode = file_inode(filp);
1712 if (!inode_owner_or_capable(inode))
1715 ret = mnt_want_write_file(filp);
1721 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1723 if (f2fs_is_volatile_file(inode))
1726 if (f2fs_is_atomic_file(inode)) {
1727 ret = commit_inmem_pages(inode);
1731 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1733 clear_inode_flag(inode, FI_ATOMIC_FILE);
1734 clear_inode_flag(inode, FI_HOT_DATA);
1735 stat_dec_atomic_write(inode);
1738 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1741 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1742 inode_unlock(inode);
1743 mnt_drop_write_file(filp);
1747 static int f2fs_ioc_start_volatile_write(struct file *filp)
1749 struct inode *inode = file_inode(filp);
1752 if (!inode_owner_or_capable(inode))
1755 if (!S_ISREG(inode->i_mode))
1758 ret = mnt_want_write_file(filp);
1764 if (f2fs_is_volatile_file(inode))
1767 ret = f2fs_convert_inline_inode(inode);
1771 stat_inc_volatile_write(inode);
1772 stat_update_max_volatile_write(inode);
1774 set_inode_flag(inode, FI_VOLATILE_FILE);
1775 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1777 inode_unlock(inode);
1778 mnt_drop_write_file(filp);
1782 static int f2fs_ioc_release_volatile_write(struct file *filp)
1784 struct inode *inode = file_inode(filp);
1787 if (!inode_owner_or_capable(inode))
1790 ret = mnt_want_write_file(filp);
1796 if (!f2fs_is_volatile_file(inode))
1799 if (!f2fs_is_first_block_written(inode)) {
1800 ret = truncate_partial_data_page(inode, 0, true);
1804 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1806 inode_unlock(inode);
1807 mnt_drop_write_file(filp);
1811 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1813 struct inode *inode = file_inode(filp);
1816 if (!inode_owner_or_capable(inode))
1819 ret = mnt_want_write_file(filp);
1825 if (f2fs_is_atomic_file(inode))
1826 drop_inmem_pages(inode);
1827 if (f2fs_is_volatile_file(inode)) {
1828 clear_inode_flag(inode, FI_VOLATILE_FILE);
1829 stat_dec_volatile_write(inode);
1830 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1833 inode_unlock(inode);
1835 mnt_drop_write_file(filp);
1836 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1840 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1842 struct inode *inode = file_inode(filp);
1843 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1844 struct super_block *sb = sbi->sb;
1848 if (!capable(CAP_SYS_ADMIN))
1851 if (get_user(in, (__u32 __user *)arg))
1854 ret = mnt_want_write_file(filp);
1859 case F2FS_GOING_DOWN_FULLSYNC:
1860 sb = freeze_bdev(sb->s_bdev);
1866 f2fs_stop_checkpoint(sbi, false);
1867 thaw_bdev(sb->s_bdev, sb);
1870 case F2FS_GOING_DOWN_METASYNC:
1871 /* do checkpoint only */
1872 ret = f2fs_sync_fs(sb, 1);
1875 f2fs_stop_checkpoint(sbi, false);
1877 case F2FS_GOING_DOWN_NOSYNC:
1878 f2fs_stop_checkpoint(sbi, false);
1880 case F2FS_GOING_DOWN_METAFLUSH:
1881 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1882 f2fs_stop_checkpoint(sbi, false);
1889 stop_gc_thread(sbi);
1890 stop_discard_thread(sbi);
1892 drop_discard_cmd(sbi);
1893 clear_opt(sbi, DISCARD);
1895 f2fs_update_time(sbi, REQ_TIME);
1897 mnt_drop_write_file(filp);
1901 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1903 struct inode *inode = file_inode(filp);
1904 struct super_block *sb = inode->i_sb;
1905 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1906 struct fstrim_range range;
1909 if (!capable(CAP_SYS_ADMIN))
1912 if (!blk_queue_discard(q))
1915 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1919 ret = mnt_want_write_file(filp);
1923 range.minlen = max((unsigned int)range.minlen,
1924 q->limits.discard_granularity);
1925 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1926 mnt_drop_write_file(filp);
1930 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1933 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1937 static bool uuid_is_nonzero(__u8 u[16])
1941 for (i = 0; i < 16; i++)
1947 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1949 struct inode *inode = file_inode(filp);
1951 if (!f2fs_sb_has_encrypt(inode->i_sb))
1954 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1956 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1959 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1961 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
1963 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1966 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1968 struct inode *inode = file_inode(filp);
1969 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1972 if (!f2fs_sb_has_encrypt(inode->i_sb))
1975 err = mnt_want_write_file(filp);
1979 down_write(&sbi->sb_lock);
1981 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1984 /* update superblock with uuid */
1985 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1987 err = f2fs_commit_super(sbi, false);
1990 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1994 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1998 up_write(&sbi->sb_lock);
1999 mnt_drop_write_file(filp);
2003 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2005 struct inode *inode = file_inode(filp);
2006 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2010 if (!capable(CAP_SYS_ADMIN))
2013 if (get_user(sync, (__u32 __user *)arg))
2016 if (f2fs_readonly(sbi->sb))
2019 ret = mnt_want_write_file(filp);
2024 if (!mutex_trylock(&sbi->gc_mutex)) {
2029 mutex_lock(&sbi->gc_mutex);
2032 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2034 mnt_drop_write_file(filp);
2038 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2040 struct inode *inode = file_inode(filp);
2041 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2042 struct f2fs_gc_range range;
2046 if (!capable(CAP_SYS_ADMIN))
2049 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2053 if (f2fs_readonly(sbi->sb))
2056 ret = mnt_want_write_file(filp);
2060 end = range.start + range.len;
2061 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2067 if (!mutex_trylock(&sbi->gc_mutex)) {
2072 mutex_lock(&sbi->gc_mutex);
2075 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2076 range.start += sbi->blocks_per_seg;
2077 if (range.start <= end)
2080 mnt_drop_write_file(filp);
2084 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2086 struct inode *inode = file_inode(filp);
2087 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2090 if (!capable(CAP_SYS_ADMIN))
2093 if (f2fs_readonly(sbi->sb))
2096 ret = mnt_want_write_file(filp);
2100 ret = f2fs_sync_fs(sbi->sb, 1);
2102 mnt_drop_write_file(filp);
2106 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2108 struct f2fs_defragment *range)
2110 struct inode *inode = file_inode(filp);
2111 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2112 .m_seg_type = NO_CHECK_TYPE };
2113 struct extent_info ei = {0,0,0};
2114 pgoff_t pg_start, pg_end, next_pgofs;
2115 unsigned int blk_per_seg = sbi->blocks_per_seg;
2116 unsigned int total = 0, sec_num;
2117 block_t blk_end = 0;
2118 bool fragmented = false;
2121 /* if in-place-update policy is enabled, don't waste time here */
2122 if (should_update_inplace(inode, NULL))
2125 pg_start = range->start >> PAGE_SHIFT;
2126 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2128 f2fs_balance_fs(sbi, true);
2132 /* writeback all dirty pages in the range */
2133 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2134 range->start + range->len - 1);
2139 * lookup mapping info in extent cache, skip defragmenting if physical
2140 * block addresses are continuous.
2142 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2143 if (ei.fofs + ei.len >= pg_end)
2147 map.m_lblk = pg_start;
2148 map.m_next_pgofs = &next_pgofs;
2151 * lookup mapping info in dnode page cache, skip defragmenting if all
2152 * physical block addresses are continuous even if there are hole(s)
2153 * in logical blocks.
2155 while (map.m_lblk < pg_end) {
2156 map.m_len = pg_end - map.m_lblk;
2157 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2161 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2162 map.m_lblk = next_pgofs;
2166 if (blk_end && blk_end != map.m_pblk)
2169 /* record total count of block that we're going to move */
2172 blk_end = map.m_pblk + map.m_len;
2174 map.m_lblk += map.m_len;
2180 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2183 * make sure there are enough free section for LFS allocation, this can
2184 * avoid defragment running in SSR mode when free section are allocated
2187 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2192 map.m_lblk = pg_start;
2193 map.m_len = pg_end - pg_start;
2196 while (map.m_lblk < pg_end) {
2201 map.m_len = pg_end - map.m_lblk;
2202 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2206 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2207 map.m_lblk = next_pgofs;
2211 set_inode_flag(inode, FI_DO_DEFRAG);
2214 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2217 page = get_lock_data_page(inode, idx, true);
2219 err = PTR_ERR(page);
2223 set_page_dirty(page);
2224 f2fs_put_page(page, 1);
2233 if (idx < pg_end && cnt < blk_per_seg)
2236 clear_inode_flag(inode, FI_DO_DEFRAG);
2238 err = filemap_fdatawrite(inode->i_mapping);
2243 clear_inode_flag(inode, FI_DO_DEFRAG);
2245 inode_unlock(inode);
2247 range->len = (u64)total << PAGE_SHIFT;
2251 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2253 struct inode *inode = file_inode(filp);
2254 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2255 struct f2fs_defragment range;
2258 if (!capable(CAP_SYS_ADMIN))
2261 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2264 if (f2fs_readonly(sbi->sb))
2267 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2271 /* verify alignment of offset & size */
2272 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2275 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2276 sbi->max_file_blocks))
2279 err = mnt_want_write_file(filp);
2283 err = f2fs_defragment_range(sbi, filp, &range);
2284 mnt_drop_write_file(filp);
2286 f2fs_update_time(sbi, REQ_TIME);
2290 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2297 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2298 struct file *file_out, loff_t pos_out, size_t len)
2300 struct inode *src = file_inode(file_in);
2301 struct inode *dst = file_inode(file_out);
2302 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2303 size_t olen = len, dst_max_i_size = 0;
2307 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2308 src->i_sb != dst->i_sb)
2311 if (unlikely(f2fs_readonly(src->i_sb)))
2314 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2317 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2321 if (pos_in == pos_out)
2323 if (pos_out > pos_in && pos_out < pos_in + len)
2328 down_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2331 if (!inode_trylock(dst))
2333 if (!down_write_trylock(&F2FS_I(dst)->dio_rwsem[WRITE])) {
2340 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2343 olen = len = src->i_size - pos_in;
2344 if (pos_in + len == src->i_size)
2345 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2351 dst_osize = dst->i_size;
2352 if (pos_out + olen > dst->i_size)
2353 dst_max_i_size = pos_out + olen;
2355 /* verify the end result is block aligned */
2356 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2357 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2358 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2361 ret = f2fs_convert_inline_inode(src);
2365 ret = f2fs_convert_inline_inode(dst);
2369 /* write out all dirty pages from offset */
2370 ret = filemap_write_and_wait_range(src->i_mapping,
2371 pos_in, pos_in + len);
2375 ret = filemap_write_and_wait_range(dst->i_mapping,
2376 pos_out, pos_out + len);
2380 f2fs_balance_fs(sbi, true);
2382 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2383 pos_out >> F2FS_BLKSIZE_BITS,
2384 len >> F2FS_BLKSIZE_BITS, false);
2388 f2fs_i_size_write(dst, dst_max_i_size);
2389 else if (dst_osize != dst->i_size)
2390 f2fs_i_size_write(dst, dst_osize);
2392 f2fs_unlock_op(sbi);
2395 up_write(&F2FS_I(dst)->dio_rwsem[WRITE]);
2399 up_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2404 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2406 struct f2fs_move_range range;
2410 if (!(filp->f_mode & FMODE_READ) ||
2411 !(filp->f_mode & FMODE_WRITE))
2414 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2418 dst = fdget(range.dst_fd);
2422 if (!(dst.file->f_mode & FMODE_WRITE)) {
2427 err = mnt_want_write_file(filp);
2431 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2432 range.pos_out, range.len);
2434 mnt_drop_write_file(filp);
2438 if (copy_to_user((struct f2fs_move_range __user *)arg,
2439 &range, sizeof(range)))
2446 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2448 struct inode *inode = file_inode(filp);
2449 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2450 struct sit_info *sm = SIT_I(sbi);
2451 unsigned int start_segno = 0, end_segno = 0;
2452 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2453 struct f2fs_flush_device range;
2456 if (!capable(CAP_SYS_ADMIN))
2459 if (f2fs_readonly(sbi->sb))
2462 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2466 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2467 sbi->segs_per_sec != 1) {
2468 f2fs_msg(sbi->sb, KERN_WARNING,
2469 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2470 range.dev_num, sbi->s_ndevs,
2475 ret = mnt_want_write_file(filp);
2479 if (range.dev_num != 0)
2480 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2481 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2483 start_segno = sm->last_victim[FLUSH_DEVICE];
2484 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2485 start_segno = dev_start_segno;
2486 end_segno = min(start_segno + range.segments, dev_end_segno);
2488 while (start_segno < end_segno) {
2489 if (!mutex_trylock(&sbi->gc_mutex)) {
2493 sm->last_victim[GC_CB] = end_segno + 1;
2494 sm->last_victim[GC_GREEDY] = end_segno + 1;
2495 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2496 ret = f2fs_gc(sbi, true, true, start_segno);
2504 mnt_drop_write_file(filp);
2508 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2510 struct inode *inode = file_inode(filp);
2511 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2513 /* Must validate to set it with SQLite behavior in Android. */
2514 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2516 return put_user(sb_feature, (u32 __user *)arg);
2520 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2522 struct inode *inode = file_inode(filp);
2523 struct f2fs_inode_info *fi = F2FS_I(inode);
2524 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2525 struct super_block *sb = sbi->sb;
2526 struct dquot *transfer_to[MAXQUOTAS] = {};
2531 if (!f2fs_sb_has_project_quota(sb)) {
2532 if (projid != F2FS_DEF_PROJID)
2538 if (!f2fs_has_extra_attr(inode))
2541 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2543 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2546 err = mnt_want_write_file(filp);
2553 /* Is it quota file? Do not allow user to mess with it */
2554 if (IS_NOQUOTA(inode))
2557 ipage = get_node_page(sbi, inode->i_ino);
2558 if (IS_ERR(ipage)) {
2559 err = PTR_ERR(ipage);
2563 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2566 f2fs_put_page(ipage, 1);
2569 f2fs_put_page(ipage, 1);
2571 dquot_initialize(inode);
2573 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2574 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2575 err = __dquot_transfer(inode, transfer_to);
2576 dqput(transfer_to[PRJQUOTA]);
2581 F2FS_I(inode)->i_projid = kprojid;
2582 inode->i_ctime = current_time(inode);
2584 f2fs_mark_inode_dirty_sync(inode, true);
2586 inode_unlock(inode);
2587 mnt_drop_write_file(filp);
2591 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2593 if (projid != F2FS_DEF_PROJID)
2599 /* Transfer internal flags to xflags */
2600 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2604 if (iflags & FS_SYNC_FL)
2605 xflags |= FS_XFLAG_SYNC;
2606 if (iflags & FS_IMMUTABLE_FL)
2607 xflags |= FS_XFLAG_IMMUTABLE;
2608 if (iflags & FS_APPEND_FL)
2609 xflags |= FS_XFLAG_APPEND;
2610 if (iflags & FS_NODUMP_FL)
2611 xflags |= FS_XFLAG_NODUMP;
2612 if (iflags & FS_NOATIME_FL)
2613 xflags |= FS_XFLAG_NOATIME;
2614 if (iflags & FS_PROJINHERIT_FL)
2615 xflags |= FS_XFLAG_PROJINHERIT;
2619 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2620 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2621 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2623 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2624 #define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2631 /* Transfer xflags flags to internal */
2632 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2634 unsigned long iflags = 0;
2636 if (xflags & FS_XFLAG_SYNC)
2637 iflags |= FS_SYNC_FL;
2638 if (xflags & FS_XFLAG_IMMUTABLE)
2639 iflags |= FS_IMMUTABLE_FL;
2640 if (xflags & FS_XFLAG_APPEND)
2641 iflags |= FS_APPEND_FL;
2642 if (xflags & FS_XFLAG_NODUMP)
2643 iflags |= FS_NODUMP_FL;
2644 if (xflags & FS_XFLAG_NOATIME)
2645 iflags |= FS_NOATIME_FL;
2646 if (xflags & FS_XFLAG_PROJINHERIT)
2647 iflags |= FS_PROJINHERIT_FL;
2652 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2654 struct inode *inode = file_inode(filp);
2655 struct f2fs_inode_info *fi = F2FS_I(inode);
2658 memset(&fa, 0, sizeof(struct fsxattr));
2659 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2660 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2662 if (f2fs_sb_has_project_quota(inode->i_sb))
2663 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2666 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2671 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2673 struct inode *inode = file_inode(filp);
2674 struct f2fs_inode_info *fi = F2FS_I(inode);
2679 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2682 /* Make sure caller has proper permission */
2683 if (!inode_owner_or_capable(inode))
2686 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2689 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2690 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2693 err = mnt_want_write_file(filp);
2698 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2699 (flags & F2FS_FL_XFLAG_VISIBLE);
2700 err = __f2fs_ioc_setflags(inode, flags);
2701 inode_unlock(inode);
2702 mnt_drop_write_file(filp);
2706 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2713 int f2fs_pin_file_control(struct inode *inode, bool inc)
2715 struct f2fs_inode_info *fi = F2FS_I(inode);
2716 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2718 /* Use i_gc_failures for normal file as a risk signal. */
2720 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
2722 if (fi->i_gc_failures > sbi->gc_pin_file_threshold) {
2723 f2fs_msg(sbi->sb, KERN_WARNING,
2724 "%s: Enable GC = ino %lx after %x GC trials\n",
2725 __func__, inode->i_ino, fi->i_gc_failures);
2726 clear_inode_flag(inode, FI_PIN_FILE);
2732 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2734 struct inode *inode = file_inode(filp);
2738 if (!inode_owner_or_capable(inode))
2741 if (get_user(pin, (__u32 __user *)arg))
2744 if (!S_ISREG(inode->i_mode))
2747 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2750 ret = mnt_want_write_file(filp);
2756 if (should_update_outplace(inode, NULL)) {
2762 clear_inode_flag(inode, FI_PIN_FILE);
2763 F2FS_I(inode)->i_gc_failures = 1;
2767 if (f2fs_pin_file_control(inode, false)) {
2771 ret = f2fs_convert_inline_inode(inode);
2775 set_inode_flag(inode, FI_PIN_FILE);
2776 ret = F2FS_I(inode)->i_gc_failures;
2778 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2780 inode_unlock(inode);
2781 mnt_drop_write_file(filp);
2785 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2787 struct inode *inode = file_inode(filp);
2790 if (is_inode_flag_set(inode, FI_PIN_FILE))
2791 pin = F2FS_I(inode)->i_gc_failures;
2792 return put_user(pin, (u32 __user *)arg);
2795 int f2fs_precache_extents(struct inode *inode)
2797 struct f2fs_inode_info *fi = F2FS_I(inode);
2798 struct f2fs_map_blocks map;
2799 pgoff_t m_next_extent;
2803 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2807 map.m_next_pgofs = NULL;
2808 map.m_next_extent = &m_next_extent;
2809 map.m_seg_type = NO_CHECK_TYPE;
2810 end = F2FS_I_SB(inode)->max_file_blocks;
2812 while (map.m_lblk < end) {
2813 map.m_len = end - map.m_lblk;
2815 down_write(&fi->dio_rwsem[WRITE]);
2816 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2817 up_write(&fi->dio_rwsem[WRITE]);
2821 map.m_lblk = m_next_extent;
2827 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2829 return f2fs_precache_extents(file_inode(filp));
2832 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2834 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2838 case F2FS_IOC_GETFLAGS:
2839 return f2fs_ioc_getflags(filp, arg);
2840 case F2FS_IOC_SETFLAGS:
2841 return f2fs_ioc_setflags(filp, arg);
2842 case F2FS_IOC_GETVERSION:
2843 return f2fs_ioc_getversion(filp, arg);
2844 case F2FS_IOC_START_ATOMIC_WRITE:
2845 return f2fs_ioc_start_atomic_write(filp);
2846 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2847 return f2fs_ioc_commit_atomic_write(filp);
2848 case F2FS_IOC_START_VOLATILE_WRITE:
2849 return f2fs_ioc_start_volatile_write(filp);
2850 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2851 return f2fs_ioc_release_volatile_write(filp);
2852 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2853 return f2fs_ioc_abort_volatile_write(filp);
2854 case F2FS_IOC_SHUTDOWN:
2855 return f2fs_ioc_shutdown(filp, arg);
2857 return f2fs_ioc_fitrim(filp, arg);
2858 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2859 return f2fs_ioc_set_encryption_policy(filp, arg);
2860 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2861 return f2fs_ioc_get_encryption_policy(filp, arg);
2862 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2863 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2864 case F2FS_IOC_GARBAGE_COLLECT:
2865 return f2fs_ioc_gc(filp, arg);
2866 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2867 return f2fs_ioc_gc_range(filp, arg);
2868 case F2FS_IOC_WRITE_CHECKPOINT:
2869 return f2fs_ioc_write_checkpoint(filp, arg);
2870 case F2FS_IOC_DEFRAGMENT:
2871 return f2fs_ioc_defragment(filp, arg);
2872 case F2FS_IOC_MOVE_RANGE:
2873 return f2fs_ioc_move_range(filp, arg);
2874 case F2FS_IOC_FLUSH_DEVICE:
2875 return f2fs_ioc_flush_device(filp, arg);
2876 case F2FS_IOC_GET_FEATURES:
2877 return f2fs_ioc_get_features(filp, arg);
2878 case F2FS_IOC_FSGETXATTR:
2879 return f2fs_ioc_fsgetxattr(filp, arg);
2880 case F2FS_IOC_FSSETXATTR:
2881 return f2fs_ioc_fssetxattr(filp, arg);
2882 case F2FS_IOC_GET_PIN_FILE:
2883 return f2fs_ioc_get_pin_file(filp, arg);
2884 case F2FS_IOC_SET_PIN_FILE:
2885 return f2fs_ioc_set_pin_file(filp, arg);
2886 case F2FS_IOC_PRECACHE_EXTENTS:
2887 return f2fs_ioc_precache_extents(filp, arg);
2893 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2895 struct file *file = iocb->ki_filp;
2896 struct inode *inode = file_inode(file);
2897 struct blk_plug plug;
2900 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2903 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
2906 if (!inode_trylock(inode)) {
2907 if (iocb->ki_flags & IOCB_NOWAIT)
2912 ret = generic_write_checks(iocb, from);
2916 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2917 set_inode_flag(inode, FI_NO_PREALLOC);
2919 if ((iocb->ki_flags & IOCB_NOWAIT) &&
2920 (iocb->ki_flags & IOCB_DIRECT)) {
2921 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
2922 iov_iter_count(from)) ||
2923 f2fs_has_inline_data(inode) ||
2924 f2fs_force_buffered_io(inode, WRITE)) {
2925 inode_unlock(inode);
2930 err = f2fs_preallocate_blocks(iocb, from);
2932 clear_inode_flag(inode, FI_NO_PREALLOC);
2933 inode_unlock(inode);
2937 blk_start_plug(&plug);
2938 ret = __generic_file_write_iter(iocb, from);
2939 blk_finish_plug(&plug);
2940 clear_inode_flag(inode, FI_NO_PREALLOC);
2943 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2945 inode_unlock(inode);
2948 ret = generic_write_sync(iocb, ret);
2952 #ifdef CONFIG_COMPAT
2953 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2956 case F2FS_IOC32_GETFLAGS:
2957 cmd = F2FS_IOC_GETFLAGS;
2959 case F2FS_IOC32_SETFLAGS:
2960 cmd = F2FS_IOC_SETFLAGS;
2962 case F2FS_IOC32_GETVERSION:
2963 cmd = F2FS_IOC_GETVERSION;
2965 case F2FS_IOC_START_ATOMIC_WRITE:
2966 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2967 case F2FS_IOC_START_VOLATILE_WRITE:
2968 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2969 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2970 case F2FS_IOC_SHUTDOWN:
2971 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2972 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2973 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2974 case F2FS_IOC_GARBAGE_COLLECT:
2975 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2976 case F2FS_IOC_WRITE_CHECKPOINT:
2977 case F2FS_IOC_DEFRAGMENT:
2978 case F2FS_IOC_MOVE_RANGE:
2979 case F2FS_IOC_FLUSH_DEVICE:
2980 case F2FS_IOC_GET_FEATURES:
2981 case F2FS_IOC_FSGETXATTR:
2982 case F2FS_IOC_FSSETXATTR:
2983 case F2FS_IOC_GET_PIN_FILE:
2984 case F2FS_IOC_SET_PIN_FILE:
2985 case F2FS_IOC_PRECACHE_EXTENTS:
2988 return -ENOIOCTLCMD;
2990 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2994 const struct file_operations f2fs_file_operations = {
2995 .llseek = f2fs_llseek,
2996 .read_iter = generic_file_read_iter,
2997 .write_iter = f2fs_file_write_iter,
2998 .open = f2fs_file_open,
2999 .release = f2fs_release_file,
3000 .mmap = f2fs_file_mmap,
3001 .flush = f2fs_file_flush,
3002 .fsync = f2fs_sync_file,
3003 .fallocate = f2fs_fallocate,
3004 .unlocked_ioctl = f2fs_ioctl,
3005 #ifdef CONFIG_COMPAT
3006 .compat_ioctl = f2fs_compat_ioctl,
3008 .splice_read = generic_file_splice_read,
3009 .splice_write = iter_file_splice_write,