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 sb_start_pagefault(inode->i_sb);
58 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
60 /* block allocation */
62 set_new_dnode(&dn, inode, NULL, NULL, 0);
63 err = f2fs_reserve_block(&dn, page->index);
71 f2fs_balance_fs(sbi, dn.node_changed);
73 file_update_time(vmf->vma->vm_file);
74 down_read(&F2FS_I(inode)->i_mmap_sem);
76 if (unlikely(page->mapping != inode->i_mapping ||
77 page_offset(page) > i_size_read(inode) ||
78 !PageUptodate(page))) {
85 * check to see if the page is mapped already (no holes)
87 if (PageMappedToDisk(page))
90 /* page is wholly or partially inside EOF */
91 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
94 offset = i_size_read(inode) & ~PAGE_MASK;
95 zero_user_segment(page, offset, PAGE_SIZE);
98 if (!PageUptodate(page))
99 SetPageUptodate(page);
101 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
103 trace_f2fs_vm_page_mkwrite(page, DATA);
106 f2fs_wait_on_page_writeback(page, DATA, false);
108 /* wait for GCed encrypted page writeback */
109 if (f2fs_encrypted_file(inode))
110 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
113 up_read(&F2FS_I(inode)->i_mmap_sem);
115 sb_end_pagefault(inode->i_sb);
116 f2fs_update_time(sbi, REQ_TIME);
117 return block_page_mkwrite_return(err);
120 static const struct vm_operations_struct f2fs_file_vm_ops = {
121 .fault = f2fs_filemap_fault,
122 .map_pages = filemap_map_pages,
123 .page_mkwrite = f2fs_vm_page_mkwrite,
126 static int get_parent_ino(struct inode *inode, nid_t *pino)
128 struct dentry *dentry;
130 inode = igrab(inode);
131 dentry = d_find_any_alias(inode);
136 *pino = parent_ino(dentry);
141 static inline bool need_do_checkpoint(struct inode *inode)
143 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
144 bool need_cp = false;
146 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
148 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
150 else if (file_wrong_pino(inode))
152 else if (!space_for_roll_forward(sbi))
154 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
156 else if (test_opt(sbi, FASTBOOT))
158 else if (sbi->active_logs == 2)
164 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
166 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
168 /* But we need to avoid that there are some inode updates */
169 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
175 static void try_to_fix_pino(struct inode *inode)
177 struct f2fs_inode_info *fi = F2FS_I(inode);
180 down_write(&fi->i_sem);
181 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
182 get_parent_ino(inode, &pino)) {
183 f2fs_i_pino_write(inode, pino);
184 file_got_pino(inode);
186 up_write(&fi->i_sem);
189 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
190 int datasync, bool atomic)
192 struct inode *inode = file->f_mapping->host;
193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
194 nid_t ino = inode->i_ino;
196 bool need_cp = false;
197 struct writeback_control wbc = {
198 .sync_mode = WB_SYNC_ALL,
199 .nr_to_write = LONG_MAX,
203 if (unlikely(f2fs_readonly(inode->i_sb)))
206 trace_f2fs_sync_file_enter(inode);
208 if (S_ISDIR(inode->i_mode))
211 /* if fdatasync is triggered, let's do in-place-update */
212 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
213 set_inode_flag(inode, FI_NEED_IPU);
214 ret = file_write_and_wait_range(file, start, end);
215 clear_inode_flag(inode, FI_NEED_IPU);
218 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
222 /* if the inode is dirty, let's recover all the time */
223 if (!f2fs_skip_inode_update(inode, datasync)) {
224 f2fs_write_inode(inode, NULL);
229 * if there is no written data, don't waste time to write recovery info.
231 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
232 !exist_written_data(sbi, ino, APPEND_INO)) {
234 /* it may call write_inode just prior to fsync */
235 if (need_inode_page_update(sbi, ino))
238 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
239 exist_written_data(sbi, ino, UPDATE_INO))
245 * Both of fdatasync() and fsync() are able to be recovered from
248 down_read(&F2FS_I(inode)->i_sem);
249 need_cp = need_do_checkpoint(inode);
250 up_read(&F2FS_I(inode)->i_sem);
253 /* all the dirty node pages should be flushed for POR */
254 ret = f2fs_sync_fs(inode->i_sb, 1);
257 * We've secured consistency through sync_fs. Following pino
258 * will be used only for fsynced inodes after checkpoint.
260 try_to_fix_pino(inode);
261 clear_inode_flag(inode, FI_APPEND_WRITE);
262 clear_inode_flag(inode, FI_UPDATE_WRITE);
266 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
270 /* if cp_error was enabled, we should avoid infinite loop */
271 if (unlikely(f2fs_cp_error(sbi))) {
276 if (need_inode_block_update(sbi, ino)) {
277 f2fs_mark_inode_dirty_sync(inode, true);
278 f2fs_write_inode(inode, NULL);
283 * If it's atomic_write, it's just fine to keep write ordering. So
284 * here we don't need to wait for node write completion, since we use
285 * node chain which serializes node blocks. If one of node writes are
286 * reordered, we can see simply broken chain, resulting in stopping
287 * roll-forward recovery. It means we'll recover all or none node blocks
291 ret = wait_on_node_pages_writeback(sbi, ino);
296 /* once recovery info is written, don't need to tack this */
297 remove_ino_entry(sbi, ino, APPEND_INO);
298 clear_inode_flag(inode, FI_APPEND_WRITE);
300 remove_ino_entry(sbi, ino, UPDATE_INO);
301 clear_inode_flag(inode, FI_UPDATE_WRITE);
303 ret = f2fs_issue_flush(sbi);
304 f2fs_update_time(sbi, REQ_TIME);
306 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
307 f2fs_trace_ios(NULL, 1);
311 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
313 return f2fs_do_sync_file(file, start, end, datasync, false);
316 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
317 pgoff_t pgofs, int whence)
322 if (whence != SEEK_DATA)
325 /* find first dirty page index */
326 pagevec_init(&pvec, 0);
327 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
328 PAGECACHE_TAG_DIRTY, 1);
329 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
330 pagevec_release(&pvec);
334 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
335 pgoff_t dirty, pgoff_t pgofs, int whence)
339 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
340 is_valid_data_blkaddr(sbi, blkaddr))
344 if (blkaddr == NULL_ADDR)
351 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
353 struct inode *inode = file->f_mapping->host;
354 loff_t maxbytes = inode->i_sb->s_maxbytes;
355 struct dnode_of_data dn;
356 pgoff_t pgofs, end_offset, dirty;
357 loff_t data_ofs = offset;
363 isize = i_size_read(inode);
367 /* handle inline data case */
368 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
369 if (whence == SEEK_HOLE)
374 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
376 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
378 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
379 set_new_dnode(&dn, inode, NULL, NULL, 0);
380 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
381 if (err && err != -ENOENT) {
383 } else if (err == -ENOENT) {
384 /* direct node does not exists */
385 if (whence == SEEK_DATA) {
386 pgofs = get_next_page_offset(&dn, pgofs);
393 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
395 /* find data/hole in dnode block */
396 for (; dn.ofs_in_node < end_offset;
397 dn.ofs_in_node++, pgofs++,
398 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
400 blkaddr = datablock_addr(dn.inode,
401 dn.node_page, dn.ofs_in_node);
403 if (__is_valid_data_blkaddr(blkaddr) &&
404 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
405 blkaddr, DATA_GENERIC)) {
410 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
419 if (whence == SEEK_DATA)
422 if (whence == SEEK_HOLE && data_ofs > isize)
425 return vfs_setpos(file, data_ofs, maxbytes);
431 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
433 struct inode *inode = file->f_mapping->host;
434 loff_t maxbytes = inode->i_sb->s_maxbytes;
440 return generic_file_llseek_size(file, offset, whence,
441 maxbytes, i_size_read(inode));
446 return f2fs_seek_block(file, offset, whence);
452 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
454 struct inode *inode = file_inode(file);
457 /* we don't need to use inline_data strictly */
458 err = f2fs_convert_inline_inode(inode);
463 vma->vm_ops = &f2fs_file_vm_ops;
467 static int f2fs_file_open(struct inode *inode, struct file *filp)
471 if (f2fs_encrypted_inode(inode)) {
472 int ret = fscrypt_get_encryption_info(inode);
475 if (!fscrypt_has_encryption_key(inode))
478 dir = dget_parent(file_dentry(filp));
479 if (f2fs_encrypted_inode(d_inode(dir)) &&
480 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
485 return dquot_file_open(inode, filp);
488 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
490 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
491 struct f2fs_node *raw_node;
492 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
496 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
497 base = get_extra_isize(dn->inode);
499 raw_node = F2FS_NODE(dn->node_page);
500 addr = blkaddr_in_node(raw_node) + base + ofs;
502 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
503 block_t blkaddr = le32_to_cpu(*addr);
504 if (blkaddr == NULL_ADDR)
507 dn->data_blkaddr = NULL_ADDR;
508 set_data_blkaddr(dn);
510 if (__is_valid_data_blkaddr(blkaddr) &&
511 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
514 invalidate_blocks(sbi, blkaddr);
515 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
516 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
523 * once we invalidate valid blkaddr in range [ofs, ofs + count],
524 * we will invalidate all blkaddr in the whole range.
526 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
528 f2fs_update_extent_cache_range(dn, fofs, 0, len);
529 dec_valid_block_count(sbi, dn->inode, nr_free);
531 dn->ofs_in_node = ofs;
533 f2fs_update_time(sbi, REQ_TIME);
534 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
535 dn->ofs_in_node, nr_free);
539 void truncate_data_blocks(struct dnode_of_data *dn)
541 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
544 static int truncate_partial_data_page(struct inode *inode, u64 from,
547 unsigned offset = from & (PAGE_SIZE - 1);
548 pgoff_t index = from >> PAGE_SHIFT;
549 struct address_space *mapping = inode->i_mapping;
552 if (!offset && !cache_only)
556 page = find_lock_page(mapping, index);
557 if (page && PageUptodate(page))
559 f2fs_put_page(page, 1);
563 page = get_lock_data_page(inode, index, true);
565 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
567 f2fs_wait_on_page_writeback(page, DATA, true);
568 zero_user(page, offset, PAGE_SIZE - offset);
570 /* An encrypted inode should have a key and truncate the last page. */
571 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
573 set_page_dirty(page);
574 f2fs_put_page(page, 1);
578 int truncate_blocks(struct inode *inode, u64 from, bool lock)
580 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
581 unsigned int blocksize = inode->i_sb->s_blocksize;
582 struct dnode_of_data dn;
584 int count = 0, err = 0;
586 bool truncate_page = false;
588 trace_f2fs_truncate_blocks_enter(inode, from);
590 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
592 if (free_from >= sbi->max_file_blocks)
598 ipage = get_node_page(sbi, inode->i_ino);
600 err = PTR_ERR(ipage);
604 if (f2fs_has_inline_data(inode)) {
605 truncate_inline_inode(inode, ipage, from);
606 f2fs_put_page(ipage, 1);
607 truncate_page = true;
611 set_new_dnode(&dn, inode, ipage, NULL, 0);
612 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
619 count = ADDRS_PER_PAGE(dn.node_page, inode);
621 count -= dn.ofs_in_node;
622 f2fs_bug_on(sbi, count < 0);
624 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
625 truncate_data_blocks_range(&dn, count);
631 err = truncate_inode_blocks(inode, free_from);
636 /* lastly zero out the first data page */
638 err = truncate_partial_data_page(inode, from, truncate_page);
640 trace_f2fs_truncate_blocks_exit(inode, err);
644 int f2fs_truncate(struct inode *inode)
648 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
649 S_ISLNK(inode->i_mode)))
652 trace_f2fs_truncate(inode);
654 #ifdef CONFIG_F2FS_FAULT_INJECTION
655 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
656 f2fs_show_injection_info(FAULT_TRUNCATE);
660 /* we should check inline_data size */
661 if (!f2fs_may_inline_data(inode)) {
662 err = f2fs_convert_inline_inode(inode);
667 err = truncate_blocks(inode, i_size_read(inode), true);
671 inode->i_mtime = inode->i_ctime = current_time(inode);
672 f2fs_mark_inode_dirty_sync(inode, false);
676 int f2fs_getattr(const struct path *path, struct kstat *stat,
677 u32 request_mask, unsigned int query_flags)
679 struct inode *inode = d_inode(path->dentry);
680 struct f2fs_inode_info *fi = F2FS_I(inode);
683 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
684 if (flags & FS_APPEND_FL)
685 stat->attributes |= STATX_ATTR_APPEND;
686 if (flags & FS_COMPR_FL)
687 stat->attributes |= STATX_ATTR_COMPRESSED;
688 if (f2fs_encrypted_inode(inode))
689 stat->attributes |= STATX_ATTR_ENCRYPTED;
690 if (flags & FS_IMMUTABLE_FL)
691 stat->attributes |= STATX_ATTR_IMMUTABLE;
692 if (flags & FS_NODUMP_FL)
693 stat->attributes |= STATX_ATTR_NODUMP;
695 stat->attributes_mask |= (STATX_ATTR_APPEND |
696 STATX_ATTR_COMPRESSED |
697 STATX_ATTR_ENCRYPTED |
698 STATX_ATTR_IMMUTABLE |
701 generic_fillattr(inode, stat);
703 /* we need to show initial sectors used for inline_data/dentries */
704 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
705 f2fs_has_inline_dentry(inode))
706 stat->blocks += (stat->size + 511) >> 9;
711 #ifdef CONFIG_F2FS_FS_POSIX_ACL
712 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
714 unsigned int ia_valid = attr->ia_valid;
716 if (ia_valid & ATTR_UID)
717 inode->i_uid = attr->ia_uid;
718 if (ia_valid & ATTR_GID)
719 inode->i_gid = attr->ia_gid;
720 if (ia_valid & ATTR_ATIME)
721 inode->i_atime = timespec_trunc(attr->ia_atime,
722 inode->i_sb->s_time_gran);
723 if (ia_valid & ATTR_MTIME)
724 inode->i_mtime = timespec_trunc(attr->ia_mtime,
725 inode->i_sb->s_time_gran);
726 if (ia_valid & ATTR_CTIME)
727 inode->i_ctime = timespec_trunc(attr->ia_ctime,
728 inode->i_sb->s_time_gran);
729 if (ia_valid & ATTR_MODE) {
730 umode_t mode = attr->ia_mode;
732 if (!in_group_p(inode->i_gid) &&
733 !capable_wrt_inode_uidgid(inode, CAP_FSETID))
735 set_acl_inode(inode, mode);
739 #define __setattr_copy setattr_copy
742 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
744 struct inode *inode = d_inode(dentry);
746 bool size_changed = false;
748 err = setattr_prepare(dentry, attr);
752 if (is_quota_modification(inode, attr)) {
753 err = dquot_initialize(inode);
757 if ((attr->ia_valid & ATTR_UID &&
758 !uid_eq(attr->ia_uid, inode->i_uid)) ||
759 (attr->ia_valid & ATTR_GID &&
760 !gid_eq(attr->ia_gid, inode->i_gid))) {
761 err = dquot_transfer(inode, attr);
766 if (attr->ia_valid & ATTR_SIZE) {
767 if (f2fs_encrypted_inode(inode)) {
768 err = fscrypt_get_encryption_info(inode);
771 if (!fscrypt_has_encryption_key(inode))
775 if (attr->ia_size <= i_size_read(inode)) {
776 down_write(&F2FS_I(inode)->i_mmap_sem);
777 truncate_setsize(inode, attr->ia_size);
778 err = f2fs_truncate(inode);
779 up_write(&F2FS_I(inode)->i_mmap_sem);
784 * do not trim all blocks after i_size if target size is
785 * larger than i_size.
787 down_write(&F2FS_I(inode)->i_mmap_sem);
788 truncate_setsize(inode, attr->ia_size);
789 up_write(&F2FS_I(inode)->i_mmap_sem);
791 /* should convert inline inode here */
792 if (!f2fs_may_inline_data(inode)) {
793 err = f2fs_convert_inline_inode(inode);
797 inode->i_mtime = inode->i_ctime = current_time(inode);
803 __setattr_copy(inode, attr);
805 if (attr->ia_valid & ATTR_MODE) {
806 err = posix_acl_chmod(inode, get_inode_mode(inode));
807 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
808 inode->i_mode = F2FS_I(inode)->i_acl_mode;
809 clear_inode_flag(inode, FI_ACL_MODE);
813 /* file size may changed here */
814 f2fs_mark_inode_dirty_sync(inode, size_changed);
816 /* inode change will produce dirty node pages flushed by checkpoint */
817 f2fs_balance_fs(F2FS_I_SB(inode), true);
822 const struct inode_operations f2fs_file_inode_operations = {
823 .getattr = f2fs_getattr,
824 .setattr = f2fs_setattr,
825 .get_acl = f2fs_get_acl,
826 .set_acl = f2fs_set_acl,
827 #ifdef CONFIG_F2FS_FS_XATTR
828 .listxattr = f2fs_listxattr,
830 .fiemap = f2fs_fiemap,
833 static int fill_zero(struct inode *inode, pgoff_t index,
834 loff_t start, loff_t len)
836 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
842 f2fs_balance_fs(sbi, true);
845 page = get_new_data_page(inode, NULL, index, false);
849 return PTR_ERR(page);
851 f2fs_wait_on_page_writeback(page, DATA, true);
852 zero_user(page, start, len);
853 set_page_dirty(page);
854 f2fs_put_page(page, 1);
858 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
862 while (pg_start < pg_end) {
863 struct dnode_of_data dn;
864 pgoff_t end_offset, count;
866 set_new_dnode(&dn, inode, NULL, NULL, 0);
867 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
869 if (err == -ENOENT) {
876 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
877 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
879 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
881 truncate_data_blocks_range(&dn, count);
889 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
891 pgoff_t pg_start, pg_end;
892 loff_t off_start, off_end;
895 ret = f2fs_convert_inline_inode(inode);
899 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
900 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
902 off_start = offset & (PAGE_SIZE - 1);
903 off_end = (offset + len) & (PAGE_SIZE - 1);
905 if (pg_start == pg_end) {
906 ret = fill_zero(inode, pg_start, off_start,
907 off_end - off_start);
912 ret = fill_zero(inode, pg_start++, off_start,
913 PAGE_SIZE - off_start);
918 ret = fill_zero(inode, pg_end, 0, off_end);
923 if (pg_start < pg_end) {
924 struct address_space *mapping = inode->i_mapping;
925 loff_t blk_start, blk_end;
926 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
928 f2fs_balance_fs(sbi, true);
930 blk_start = (loff_t)pg_start << PAGE_SHIFT;
931 blk_end = (loff_t)pg_end << PAGE_SHIFT;
932 down_write(&F2FS_I(inode)->i_mmap_sem);
933 truncate_inode_pages_range(mapping, blk_start,
937 ret = truncate_hole(inode, pg_start, pg_end);
939 up_write(&F2FS_I(inode)->i_mmap_sem);
946 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
947 int *do_replace, pgoff_t off, pgoff_t len)
949 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
950 struct dnode_of_data dn;
954 set_new_dnode(&dn, inode, NULL, NULL, 0);
955 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
956 if (ret && ret != -ENOENT) {
958 } else if (ret == -ENOENT) {
959 if (dn.max_level == 0)
961 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
967 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
968 dn.ofs_in_node, len);
969 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
970 *blkaddr = datablock_addr(dn.inode,
971 dn.node_page, dn.ofs_in_node);
972 if (!is_checkpointed_data(sbi, *blkaddr)) {
974 if (test_opt(sbi, LFS)) {
979 /* do not invalidate this block address */
980 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
993 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
994 int *do_replace, pgoff_t off, int len)
996 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
997 struct dnode_of_data dn;
1000 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1001 if (*do_replace == 0)
1004 set_new_dnode(&dn, inode, NULL, NULL, 0);
1005 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1007 dec_valid_block_count(sbi, inode, 1);
1008 invalidate_blocks(sbi, *blkaddr);
1010 f2fs_update_data_blkaddr(&dn, *blkaddr);
1012 f2fs_put_dnode(&dn);
1017 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1018 block_t *blkaddr, int *do_replace,
1019 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1021 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1026 if (blkaddr[i] == NULL_ADDR && !full) {
1031 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1032 struct dnode_of_data dn;
1033 struct node_info ni;
1037 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1038 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1042 get_node_info(sbi, dn.nid, &ni);
1043 ilen = min((pgoff_t)
1044 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1045 dn.ofs_in_node, len - i);
1047 dn.data_blkaddr = datablock_addr(dn.inode,
1048 dn.node_page, dn.ofs_in_node);
1049 truncate_data_blocks_range(&dn, 1);
1051 if (do_replace[i]) {
1052 f2fs_i_blocks_write(src_inode,
1054 f2fs_i_blocks_write(dst_inode,
1056 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1057 blkaddr[i], ni.version, true, false);
1063 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1064 if (dst_inode->i_size < new_size)
1065 f2fs_i_size_write(dst_inode, new_size);
1066 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1068 f2fs_put_dnode(&dn);
1070 struct page *psrc, *pdst;
1072 psrc = get_lock_data_page(src_inode, src + i, true);
1074 return PTR_ERR(psrc);
1075 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1078 f2fs_put_page(psrc, 1);
1079 return PTR_ERR(pdst);
1081 f2fs_copy_page(psrc, pdst);
1082 set_page_dirty(pdst);
1083 f2fs_put_page(pdst, 1);
1084 f2fs_put_page(psrc, 1);
1086 ret = truncate_hole(src_inode, src + i, src + i + 1);
1095 static int __exchange_data_block(struct inode *src_inode,
1096 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1097 pgoff_t len, bool full)
1099 block_t *src_blkaddr;
1105 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1107 src_blkaddr = kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1111 do_replace = kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1113 kvfree(src_blkaddr);
1117 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1118 do_replace, src, olen);
1122 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1123 do_replace, src, dst, olen, full);
1131 kvfree(src_blkaddr);
1137 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1138 kvfree(src_blkaddr);
1143 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1145 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1146 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1149 f2fs_balance_fs(sbi, true);
1152 f2fs_drop_extent_tree(inode);
1154 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1155 f2fs_unlock_op(sbi);
1159 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1161 pgoff_t pg_start, pg_end;
1165 if (offset + len >= i_size_read(inode))
1168 /* collapse range should be aligned to block size of f2fs. */
1169 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1172 ret = f2fs_convert_inline_inode(inode);
1176 pg_start = offset >> PAGE_SHIFT;
1177 pg_end = (offset + len) >> PAGE_SHIFT;
1179 down_write(&F2FS_I(inode)->i_mmap_sem);
1180 /* write out all dirty pages from offset */
1181 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1185 truncate_pagecache(inode, offset);
1187 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1191 /* write out all moved pages, if possible */
1192 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1193 truncate_pagecache(inode, offset);
1195 new_size = i_size_read(inode) - len;
1196 truncate_pagecache(inode, new_size);
1198 ret = truncate_blocks(inode, new_size, true);
1200 f2fs_i_size_write(inode, new_size);
1203 up_write(&F2FS_I(inode)->i_mmap_sem);
1207 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1210 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1211 pgoff_t index = start;
1212 unsigned int ofs_in_node = dn->ofs_in_node;
1216 for (; index < end; index++, dn->ofs_in_node++) {
1217 if (datablock_addr(dn->inode, dn->node_page,
1218 dn->ofs_in_node) == NULL_ADDR)
1222 dn->ofs_in_node = ofs_in_node;
1223 ret = reserve_new_blocks(dn, count);
1227 dn->ofs_in_node = ofs_in_node;
1228 for (index = start; index < end; index++, dn->ofs_in_node++) {
1229 dn->data_blkaddr = datablock_addr(dn->inode,
1230 dn->node_page, dn->ofs_in_node);
1232 * reserve_new_blocks will not guarantee entire block
1235 if (dn->data_blkaddr == NULL_ADDR) {
1239 if (dn->data_blkaddr != NEW_ADDR) {
1240 invalidate_blocks(sbi, dn->data_blkaddr);
1241 dn->data_blkaddr = NEW_ADDR;
1242 set_data_blkaddr(dn);
1246 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1251 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1254 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1255 struct address_space *mapping = inode->i_mapping;
1256 pgoff_t index, pg_start, pg_end;
1257 loff_t new_size = i_size_read(inode);
1258 loff_t off_start, off_end;
1261 ret = inode_newsize_ok(inode, (len + offset));
1265 ret = f2fs_convert_inline_inode(inode);
1269 down_write(&F2FS_I(inode)->i_mmap_sem);
1270 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1274 truncate_pagecache_range(inode, offset, offset + len - 1);
1276 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1277 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1279 off_start = offset & (PAGE_SIZE - 1);
1280 off_end = (offset + len) & (PAGE_SIZE - 1);
1282 if (pg_start == pg_end) {
1283 ret = fill_zero(inode, pg_start, off_start,
1284 off_end - off_start);
1288 new_size = max_t(loff_t, new_size, offset + len);
1291 ret = fill_zero(inode, pg_start++, off_start,
1292 PAGE_SIZE - off_start);
1296 new_size = max_t(loff_t, new_size,
1297 (loff_t)pg_start << PAGE_SHIFT);
1300 for (index = pg_start; index < pg_end;) {
1301 struct dnode_of_data dn;
1302 unsigned int end_offset;
1307 set_new_dnode(&dn, inode, NULL, NULL, 0);
1308 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1310 f2fs_unlock_op(sbi);
1314 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1315 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1317 ret = f2fs_do_zero_range(&dn, index, end);
1318 f2fs_put_dnode(&dn);
1319 f2fs_unlock_op(sbi);
1321 f2fs_balance_fs(sbi, dn.node_changed);
1327 new_size = max_t(loff_t, new_size,
1328 (loff_t)index << PAGE_SHIFT);
1332 ret = fill_zero(inode, pg_end, 0, off_end);
1336 new_size = max_t(loff_t, new_size, offset + len);
1341 if (new_size > i_size_read(inode)) {
1342 if (mode & FALLOC_FL_KEEP_SIZE)
1343 file_set_keep_isize(inode);
1345 f2fs_i_size_write(inode, new_size);
1348 up_write(&F2FS_I(inode)->i_mmap_sem);
1353 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1355 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1356 pgoff_t nr, pg_start, pg_end, delta, idx;
1360 new_size = i_size_read(inode) + len;
1361 ret = inode_newsize_ok(inode, new_size);
1365 if (offset >= i_size_read(inode))
1368 /* insert range should be aligned to block size of f2fs. */
1369 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1372 ret = f2fs_convert_inline_inode(inode);
1376 f2fs_balance_fs(sbi, true);
1378 down_write(&F2FS_I(inode)->i_mmap_sem);
1379 ret = truncate_blocks(inode, i_size_read(inode), true);
1383 /* write out all dirty pages from offset */
1384 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1388 truncate_pagecache(inode, offset);
1390 pg_start = offset >> PAGE_SHIFT;
1391 pg_end = (offset + len) >> PAGE_SHIFT;
1392 delta = pg_end - pg_start;
1393 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1395 while (!ret && idx > pg_start) {
1396 nr = idx - pg_start;
1402 f2fs_drop_extent_tree(inode);
1404 ret = __exchange_data_block(inode, inode, idx,
1405 idx + delta, nr, false);
1406 f2fs_unlock_op(sbi);
1409 /* write out all moved pages, if possible */
1410 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1411 truncate_pagecache(inode, offset);
1414 f2fs_i_size_write(inode, new_size);
1416 up_write(&F2FS_I(inode)->i_mmap_sem);
1420 static int expand_inode_data(struct inode *inode, loff_t offset,
1421 loff_t len, int mode)
1423 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1424 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1426 loff_t new_size = i_size_read(inode);
1430 err = inode_newsize_ok(inode, (len + offset));
1434 err = f2fs_convert_inline_inode(inode);
1438 f2fs_balance_fs(sbi, true);
1440 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1441 off_end = (offset + len) & (PAGE_SIZE - 1);
1443 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1444 map.m_len = pg_end - map.m_lblk;
1448 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1455 last_off = map.m_lblk + map.m_len - 1;
1457 /* update new size to the failed position */
1458 new_size = (last_off == pg_end) ? offset + len:
1459 (loff_t)(last_off + 1) << PAGE_SHIFT;
1461 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1464 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1465 f2fs_i_size_write(inode, new_size);
1470 static long f2fs_fallocate(struct file *file, int mode,
1471 loff_t offset, loff_t len)
1473 struct inode *inode = file_inode(file);
1476 /* f2fs only support ->fallocate for regular file */
1477 if (!S_ISREG(inode->i_mode))
1480 if (f2fs_encrypted_inode(inode) &&
1481 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1484 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1485 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1486 FALLOC_FL_INSERT_RANGE))
1491 if (mode & FALLOC_FL_PUNCH_HOLE) {
1492 if (offset >= inode->i_size)
1495 ret = punch_hole(inode, offset, len);
1496 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1497 ret = f2fs_collapse_range(inode, offset, len);
1498 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1499 ret = f2fs_zero_range(inode, offset, len, mode);
1500 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1501 ret = f2fs_insert_range(inode, offset, len);
1503 ret = expand_inode_data(inode, offset, len, mode);
1507 inode->i_mtime = inode->i_ctime = current_time(inode);
1508 f2fs_mark_inode_dirty_sync(inode, false);
1509 if (mode & FALLOC_FL_KEEP_SIZE)
1510 file_set_keep_isize(inode);
1511 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1515 inode_unlock(inode);
1517 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1521 static int f2fs_release_file(struct inode *inode, struct file *filp)
1524 * f2fs_relase_file is called at every close calls. So we should
1525 * not drop any inmemory pages by close called by other process.
1527 if (!(filp->f_mode & FMODE_WRITE) ||
1528 atomic_read(&inode->i_writecount) != 1)
1531 /* some remained atomic pages should discarded */
1532 if (f2fs_is_atomic_file(inode))
1533 drop_inmem_pages(inode);
1534 if (f2fs_is_volatile_file(inode)) {
1535 clear_inode_flag(inode, FI_VOLATILE_FILE);
1536 stat_dec_volatile_write(inode);
1537 set_inode_flag(inode, FI_DROP_CACHE);
1538 filemap_fdatawrite(inode->i_mapping);
1539 clear_inode_flag(inode, FI_DROP_CACHE);
1544 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1546 struct inode *inode = file_inode(file);
1549 * If the process doing a transaction is crashed, we should do
1550 * roll-back. Otherwise, other reader/write can see corrupted database
1551 * until all the writers close its file. Since this should be done
1552 * before dropping file lock, it needs to do in ->flush.
1554 if (f2fs_is_atomic_file(inode) &&
1555 F2FS_I(inode)->inmem_task == current)
1556 drop_inmem_pages(inode);
1560 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1562 struct inode *inode = file_inode(filp);
1563 struct f2fs_inode_info *fi = F2FS_I(inode);
1564 unsigned int flags = fi->i_flags &
1565 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1566 return put_user(flags, (int __user *)arg);
1569 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1571 struct f2fs_inode_info *fi = F2FS_I(inode);
1572 unsigned int oldflags;
1574 /* Is it quota file? Do not allow user to mess with it */
1575 if (IS_NOQUOTA(inode))
1578 flags = f2fs_mask_flags(inode->i_mode, flags);
1580 oldflags = fi->i_flags;
1582 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1583 if (!capable(CAP_LINUX_IMMUTABLE))
1586 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1587 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1588 fi->i_flags = flags;
1590 if (fi->i_flags & FS_PROJINHERIT_FL)
1591 set_inode_flag(inode, FI_PROJ_INHERIT);
1593 clear_inode_flag(inode, FI_PROJ_INHERIT);
1595 inode->i_ctime = current_time(inode);
1596 f2fs_set_inode_flags(inode);
1597 f2fs_mark_inode_dirty_sync(inode, true);
1601 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1603 struct inode *inode = file_inode(filp);
1607 if (!inode_owner_or_capable(inode))
1610 if (get_user(flags, (int __user *)arg))
1613 ret = mnt_want_write_file(filp);
1619 ret = __f2fs_ioc_setflags(inode, flags);
1621 inode_unlock(inode);
1622 mnt_drop_write_file(filp);
1626 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1628 struct inode *inode = file_inode(filp);
1630 return put_user(inode->i_generation, (int __user *)arg);
1633 static int f2fs_ioc_start_atomic_write(struct file *filp)
1635 struct inode *inode = file_inode(filp);
1638 if (!inode_owner_or_capable(inode))
1641 if (!S_ISREG(inode->i_mode))
1644 ret = mnt_want_write_file(filp);
1650 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1652 if (f2fs_is_atomic_file(inode))
1655 ret = f2fs_convert_inline_inode(inode);
1659 set_inode_flag(inode, FI_ATOMIC_FILE);
1660 set_inode_flag(inode, FI_HOT_DATA);
1661 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1663 if (!get_dirty_pages(inode))
1666 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1667 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1668 inode->i_ino, get_dirty_pages(inode));
1669 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1671 clear_inode_flag(inode, FI_ATOMIC_FILE);
1672 clear_inode_flag(inode, FI_HOT_DATA);
1677 F2FS_I(inode)->inmem_task = current;
1678 stat_inc_atomic_write(inode);
1679 stat_update_max_atomic_write(inode);
1681 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1682 inode_unlock(inode);
1683 mnt_drop_write_file(filp);
1687 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1689 struct inode *inode = file_inode(filp);
1692 if (!inode_owner_or_capable(inode))
1695 ret = mnt_want_write_file(filp);
1701 if (f2fs_is_volatile_file(inode))
1704 if (f2fs_is_atomic_file(inode)) {
1705 ret = commit_inmem_pages(inode);
1709 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1711 clear_inode_flag(inode, FI_ATOMIC_FILE);
1712 clear_inode_flag(inode, FI_HOT_DATA);
1713 stat_dec_atomic_write(inode);
1716 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1719 inode_unlock(inode);
1720 mnt_drop_write_file(filp);
1724 static int f2fs_ioc_start_volatile_write(struct file *filp)
1726 struct inode *inode = file_inode(filp);
1729 if (!inode_owner_or_capable(inode))
1732 if (!S_ISREG(inode->i_mode))
1735 ret = mnt_want_write_file(filp);
1741 if (f2fs_is_volatile_file(inode))
1744 ret = f2fs_convert_inline_inode(inode);
1748 stat_inc_volatile_write(inode);
1749 stat_update_max_volatile_write(inode);
1751 set_inode_flag(inode, FI_VOLATILE_FILE);
1752 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1754 inode_unlock(inode);
1755 mnt_drop_write_file(filp);
1759 static int f2fs_ioc_release_volatile_write(struct file *filp)
1761 struct inode *inode = file_inode(filp);
1764 if (!inode_owner_or_capable(inode))
1767 ret = mnt_want_write_file(filp);
1773 if (!f2fs_is_volatile_file(inode))
1776 if (!f2fs_is_first_block_written(inode)) {
1777 ret = truncate_partial_data_page(inode, 0, true);
1781 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1783 inode_unlock(inode);
1784 mnt_drop_write_file(filp);
1788 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1790 struct inode *inode = file_inode(filp);
1793 if (!inode_owner_or_capable(inode))
1796 ret = mnt_want_write_file(filp);
1802 if (f2fs_is_atomic_file(inode))
1803 drop_inmem_pages(inode);
1804 if (f2fs_is_volatile_file(inode)) {
1805 clear_inode_flag(inode, FI_VOLATILE_FILE);
1806 stat_dec_volatile_write(inode);
1807 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1810 inode_unlock(inode);
1812 mnt_drop_write_file(filp);
1813 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1817 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1819 struct inode *inode = file_inode(filp);
1820 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1821 struct super_block *sb = sbi->sb;
1825 if (!capable(CAP_SYS_ADMIN))
1828 if (get_user(in, (__u32 __user *)arg))
1831 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1832 ret = mnt_want_write_file(filp);
1838 case F2FS_GOING_DOWN_FULLSYNC:
1839 sb = freeze_bdev(sb->s_bdev);
1840 if (sb && !IS_ERR(sb)) {
1841 f2fs_stop_checkpoint(sbi, false);
1842 thaw_bdev(sb->s_bdev, sb);
1845 case F2FS_GOING_DOWN_METASYNC:
1846 /* do checkpoint only */
1847 f2fs_sync_fs(sb, 1);
1848 f2fs_stop_checkpoint(sbi, false);
1850 case F2FS_GOING_DOWN_NOSYNC:
1851 f2fs_stop_checkpoint(sbi, false);
1853 case F2FS_GOING_DOWN_METAFLUSH:
1854 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1855 f2fs_stop_checkpoint(sbi, false);
1861 f2fs_update_time(sbi, REQ_TIME);
1863 if (in != F2FS_GOING_DOWN_FULLSYNC)
1864 mnt_drop_write_file(filp);
1868 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1870 struct inode *inode = file_inode(filp);
1871 struct super_block *sb = inode->i_sb;
1872 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1873 struct fstrim_range range;
1876 if (!capable(CAP_SYS_ADMIN))
1879 if (!blk_queue_discard(q))
1882 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1886 ret = mnt_want_write_file(filp);
1890 range.minlen = max((unsigned int)range.minlen,
1891 q->limits.discard_granularity);
1892 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1893 mnt_drop_write_file(filp);
1897 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1900 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1904 static bool uuid_is_nonzero(__u8 u[16])
1908 for (i = 0; i < 16; i++)
1914 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1916 struct inode *inode = file_inode(filp);
1918 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1920 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1923 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1925 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1928 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1930 struct inode *inode = file_inode(filp);
1931 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1934 if (!f2fs_sb_has_crypto(inode->i_sb))
1937 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1940 err = mnt_want_write_file(filp);
1944 /* update superblock with uuid */
1945 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1947 err = f2fs_commit_super(sbi, false);
1950 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1951 mnt_drop_write_file(filp);
1954 mnt_drop_write_file(filp);
1956 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1962 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1964 struct inode *inode = file_inode(filp);
1965 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1969 if (!capable(CAP_SYS_ADMIN))
1972 if (get_user(sync, (__u32 __user *)arg))
1975 if (f2fs_readonly(sbi->sb))
1978 ret = mnt_want_write_file(filp);
1983 if (!mutex_trylock(&sbi->gc_mutex)) {
1988 mutex_lock(&sbi->gc_mutex);
1991 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
1993 mnt_drop_write_file(filp);
1997 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
1999 struct inode *inode = file_inode(filp);
2000 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2001 struct f2fs_gc_range range;
2005 if (!capable(CAP_SYS_ADMIN))
2008 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2012 if (f2fs_readonly(sbi->sb))
2015 ret = mnt_want_write_file(filp);
2019 end = range.start + range.len;
2020 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi))
2024 if (!mutex_trylock(&sbi->gc_mutex)) {
2029 mutex_lock(&sbi->gc_mutex);
2032 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2033 range.start += BLKS_PER_SEC(sbi);
2034 if (range.start <= end)
2037 mnt_drop_write_file(filp);
2041 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2043 struct inode *inode = file_inode(filp);
2044 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2047 if (!capable(CAP_SYS_ADMIN))
2050 if (f2fs_readonly(sbi->sb))
2053 ret = mnt_want_write_file(filp);
2057 ret = f2fs_sync_fs(sbi->sb, 1);
2059 mnt_drop_write_file(filp);
2063 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2065 struct f2fs_defragment *range)
2067 struct inode *inode = file_inode(filp);
2068 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
2069 struct extent_info ei = {0,0,0};
2070 pgoff_t pg_start, pg_end;
2071 unsigned int blk_per_seg = sbi->blocks_per_seg;
2072 unsigned int total = 0, sec_num;
2073 block_t blk_end = 0;
2074 bool fragmented = false;
2077 /* if in-place-update policy is enabled, don't waste time here */
2078 if (need_inplace_update_policy(inode, NULL))
2081 pg_start = range->start >> PAGE_SHIFT;
2082 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2084 f2fs_balance_fs(sbi, true);
2088 /* writeback all dirty pages in the range */
2089 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2090 range->start + range->len - 1);
2095 * lookup mapping info in extent cache, skip defragmenting if physical
2096 * block addresses are continuous.
2098 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2099 if (ei.fofs + ei.len >= pg_end)
2103 map.m_lblk = pg_start;
2106 * lookup mapping info in dnode page cache, skip defragmenting if all
2107 * physical block addresses are continuous even if there are hole(s)
2108 * in logical blocks.
2110 while (map.m_lblk < pg_end) {
2111 map.m_len = pg_end - map.m_lblk;
2112 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2116 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2121 if (blk_end && blk_end != map.m_pblk) {
2125 blk_end = map.m_pblk + map.m_len;
2127 map.m_lblk += map.m_len;
2133 map.m_lblk = pg_start;
2134 map.m_len = pg_end - pg_start;
2136 sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2139 * make sure there are enough free section for LFS allocation, this can
2140 * avoid defragment running in SSR mode when free section are allocated
2143 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2148 while (map.m_lblk < pg_end) {
2153 map.m_len = pg_end - map.m_lblk;
2154 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2158 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2163 set_inode_flag(inode, FI_DO_DEFRAG);
2166 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2169 page = get_lock_data_page(inode, idx, true);
2171 err = PTR_ERR(page);
2175 set_page_dirty(page);
2176 f2fs_put_page(page, 1);
2185 if (idx < pg_end && cnt < blk_per_seg)
2188 clear_inode_flag(inode, FI_DO_DEFRAG);
2190 err = filemap_fdatawrite(inode->i_mapping);
2195 clear_inode_flag(inode, FI_DO_DEFRAG);
2197 inode_unlock(inode);
2199 range->len = (u64)total << PAGE_SHIFT;
2203 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2205 struct inode *inode = file_inode(filp);
2206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2207 struct f2fs_defragment range;
2210 if (!capable(CAP_SYS_ADMIN))
2213 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2216 if (f2fs_readonly(sbi->sb))
2219 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2223 /* verify alignment of offset & size */
2224 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2227 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2228 sbi->max_file_blocks))
2231 err = mnt_want_write_file(filp);
2235 err = f2fs_defragment_range(sbi, filp, &range);
2236 mnt_drop_write_file(filp);
2238 f2fs_update_time(sbi, REQ_TIME);
2242 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2249 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2250 struct file *file_out, loff_t pos_out, size_t len)
2252 struct inode *src = file_inode(file_in);
2253 struct inode *dst = file_inode(file_out);
2254 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2255 size_t olen = len, dst_max_i_size = 0;
2259 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2260 src->i_sb != dst->i_sb)
2263 if (unlikely(f2fs_readonly(src->i_sb)))
2266 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2269 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2273 if (pos_in == pos_out)
2275 if (pos_out > pos_in && pos_out < pos_in + len)
2281 if (!inode_trylock(dst)) {
2288 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2291 olen = len = src->i_size - pos_in;
2292 if (pos_in + len == src->i_size)
2293 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2299 dst_osize = dst->i_size;
2300 if (pos_out + olen > dst->i_size)
2301 dst_max_i_size = pos_out + olen;
2303 /* verify the end result is block aligned */
2304 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2305 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2306 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2309 ret = f2fs_convert_inline_inode(src);
2313 ret = f2fs_convert_inline_inode(dst);
2317 /* write out all dirty pages from offset */
2318 ret = filemap_write_and_wait_range(src->i_mapping,
2319 pos_in, pos_in + len);
2323 ret = filemap_write_and_wait_range(dst->i_mapping,
2324 pos_out, pos_out + len);
2328 f2fs_balance_fs(sbi, true);
2330 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2331 pos_out >> F2FS_BLKSIZE_BITS,
2332 len >> F2FS_BLKSIZE_BITS, false);
2336 f2fs_i_size_write(dst, dst_max_i_size);
2337 else if (dst_osize != dst->i_size)
2338 f2fs_i_size_write(dst, dst_osize);
2340 f2fs_unlock_op(sbi);
2349 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2351 struct f2fs_move_range range;
2355 if (!(filp->f_mode & FMODE_READ) ||
2356 !(filp->f_mode & FMODE_WRITE))
2359 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2363 dst = fdget(range.dst_fd);
2367 if (!(dst.file->f_mode & FMODE_WRITE)) {
2372 err = mnt_want_write_file(filp);
2376 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2377 range.pos_out, range.len);
2379 mnt_drop_write_file(filp);
2383 if (copy_to_user((struct f2fs_move_range __user *)arg,
2384 &range, sizeof(range)))
2391 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2393 struct inode *inode = file_inode(filp);
2394 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2395 struct sit_info *sm = SIT_I(sbi);
2396 unsigned int start_segno = 0, end_segno = 0;
2397 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2398 struct f2fs_flush_device range;
2401 if (!capable(CAP_SYS_ADMIN))
2404 if (f2fs_readonly(sbi->sb))
2407 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2411 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2412 sbi->segs_per_sec != 1) {
2413 f2fs_msg(sbi->sb, KERN_WARNING,
2414 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2415 range.dev_num, sbi->s_ndevs,
2420 ret = mnt_want_write_file(filp);
2424 if (range.dev_num != 0)
2425 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2426 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2428 start_segno = sm->last_victim[FLUSH_DEVICE];
2429 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2430 start_segno = dev_start_segno;
2431 end_segno = min(start_segno + range.segments, dev_end_segno);
2433 while (start_segno < end_segno) {
2434 if (!mutex_trylock(&sbi->gc_mutex)) {
2438 sm->last_victim[GC_CB] = end_segno + 1;
2439 sm->last_victim[GC_GREEDY] = end_segno + 1;
2440 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2441 ret = f2fs_gc(sbi, true, true, start_segno);
2449 mnt_drop_write_file(filp);
2453 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2455 struct inode *inode = file_inode(filp);
2456 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2458 /* Must validate to set it with SQLite behavior in Android. */
2459 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2461 return put_user(sb_feature, (u32 __user *)arg);
2465 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2467 struct inode *inode = file_inode(filp);
2468 struct f2fs_inode_info *fi = F2FS_I(inode);
2469 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2470 struct super_block *sb = sbi->sb;
2471 struct dquot *transfer_to[MAXQUOTAS] = {};
2476 if (!f2fs_sb_has_project_quota(sb)) {
2477 if (projid != F2FS_DEF_PROJID)
2483 if (!f2fs_has_extra_attr(inode))
2486 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2488 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2491 err = mnt_want_write_file(filp);
2498 /* Is it quota file? Do not allow user to mess with it */
2499 if (IS_NOQUOTA(inode))
2502 ipage = get_node_page(sbi, inode->i_ino);
2503 if (IS_ERR(ipage)) {
2504 err = PTR_ERR(ipage);
2508 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2511 f2fs_put_page(ipage, 1);
2514 f2fs_put_page(ipage, 1);
2516 err = dquot_initialize(inode);
2520 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2521 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2522 err = __dquot_transfer(inode, transfer_to);
2523 dqput(transfer_to[PRJQUOTA]);
2528 F2FS_I(inode)->i_projid = kprojid;
2529 inode->i_ctime = current_time(inode);
2531 f2fs_mark_inode_dirty_sync(inode, true);
2533 inode_unlock(inode);
2534 mnt_drop_write_file(filp);
2538 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2540 if (projid != F2FS_DEF_PROJID)
2546 /* Transfer internal flags to xflags */
2547 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2551 if (iflags & FS_SYNC_FL)
2552 xflags |= FS_XFLAG_SYNC;
2553 if (iflags & FS_IMMUTABLE_FL)
2554 xflags |= FS_XFLAG_IMMUTABLE;
2555 if (iflags & FS_APPEND_FL)
2556 xflags |= FS_XFLAG_APPEND;
2557 if (iflags & FS_NODUMP_FL)
2558 xflags |= FS_XFLAG_NODUMP;
2559 if (iflags & FS_NOATIME_FL)
2560 xflags |= FS_XFLAG_NOATIME;
2561 if (iflags & FS_PROJINHERIT_FL)
2562 xflags |= FS_XFLAG_PROJINHERIT;
2566 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2567 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2568 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2570 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2571 #define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2578 /* Transfer xflags flags to internal */
2579 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2581 unsigned long iflags = 0;
2583 if (xflags & FS_XFLAG_SYNC)
2584 iflags |= FS_SYNC_FL;
2585 if (xflags & FS_XFLAG_IMMUTABLE)
2586 iflags |= FS_IMMUTABLE_FL;
2587 if (xflags & FS_XFLAG_APPEND)
2588 iflags |= FS_APPEND_FL;
2589 if (xflags & FS_XFLAG_NODUMP)
2590 iflags |= FS_NODUMP_FL;
2591 if (xflags & FS_XFLAG_NOATIME)
2592 iflags |= FS_NOATIME_FL;
2593 if (xflags & FS_XFLAG_PROJINHERIT)
2594 iflags |= FS_PROJINHERIT_FL;
2599 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2601 struct inode *inode = file_inode(filp);
2602 struct f2fs_inode_info *fi = F2FS_I(inode);
2605 memset(&fa, 0, sizeof(struct fsxattr));
2606 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2607 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2609 if (f2fs_sb_has_project_quota(inode->i_sb))
2610 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2613 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2618 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2620 struct inode *inode = file_inode(filp);
2621 struct f2fs_inode_info *fi = F2FS_I(inode);
2626 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2629 /* Make sure caller has proper permission */
2630 if (!inode_owner_or_capable(inode))
2633 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2636 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2637 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2640 err = mnt_want_write_file(filp);
2645 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2646 (flags & F2FS_FL_XFLAG_VISIBLE);
2647 err = __f2fs_ioc_setflags(inode, flags);
2648 inode_unlock(inode);
2649 mnt_drop_write_file(filp);
2653 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2660 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2663 case F2FS_IOC_GETFLAGS:
2664 return f2fs_ioc_getflags(filp, arg);
2665 case F2FS_IOC_SETFLAGS:
2666 return f2fs_ioc_setflags(filp, arg);
2667 case F2FS_IOC_GETVERSION:
2668 return f2fs_ioc_getversion(filp, arg);
2669 case F2FS_IOC_START_ATOMIC_WRITE:
2670 return f2fs_ioc_start_atomic_write(filp);
2671 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2672 return f2fs_ioc_commit_atomic_write(filp);
2673 case F2FS_IOC_START_VOLATILE_WRITE:
2674 return f2fs_ioc_start_volatile_write(filp);
2675 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2676 return f2fs_ioc_release_volatile_write(filp);
2677 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2678 return f2fs_ioc_abort_volatile_write(filp);
2679 case F2FS_IOC_SHUTDOWN:
2680 return f2fs_ioc_shutdown(filp, arg);
2682 return f2fs_ioc_fitrim(filp, arg);
2683 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2684 return f2fs_ioc_set_encryption_policy(filp, arg);
2685 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2686 return f2fs_ioc_get_encryption_policy(filp, arg);
2687 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2688 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2689 case F2FS_IOC_GARBAGE_COLLECT:
2690 return f2fs_ioc_gc(filp, arg);
2691 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2692 return f2fs_ioc_gc_range(filp, arg);
2693 case F2FS_IOC_WRITE_CHECKPOINT:
2694 return f2fs_ioc_write_checkpoint(filp, arg);
2695 case F2FS_IOC_DEFRAGMENT:
2696 return f2fs_ioc_defragment(filp, arg);
2697 case F2FS_IOC_MOVE_RANGE:
2698 return f2fs_ioc_move_range(filp, arg);
2699 case F2FS_IOC_FLUSH_DEVICE:
2700 return f2fs_ioc_flush_device(filp, arg);
2701 case F2FS_IOC_GET_FEATURES:
2702 return f2fs_ioc_get_features(filp, arg);
2703 case F2FS_IOC_FSGETXATTR:
2704 return f2fs_ioc_fsgetxattr(filp, arg);
2705 case F2FS_IOC_FSSETXATTR:
2706 return f2fs_ioc_fssetxattr(filp, arg);
2712 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2714 struct file *file = iocb->ki_filp;
2715 struct inode *inode = file_inode(file);
2716 struct blk_plug plug;
2720 ret = generic_write_checks(iocb, from);
2722 bool preallocated = false;
2723 size_t target_size = 0;
2726 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2727 set_inode_flag(inode, FI_NO_PREALLOC);
2729 preallocated = true;
2730 target_size = iocb->ki_pos + iov_iter_count(from);
2732 err = f2fs_preallocate_blocks(iocb, from);
2734 clear_inode_flag(inode, FI_NO_PREALLOC);
2735 inode_unlock(inode);
2738 blk_start_plug(&plug);
2739 ret = __generic_file_write_iter(iocb, from);
2740 blk_finish_plug(&plug);
2741 clear_inode_flag(inode, FI_NO_PREALLOC);
2743 /* if we couldn't write data, we should deallocate blocks. */
2744 if (preallocated && i_size_read(inode) < target_size)
2745 f2fs_truncate(inode);
2748 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2750 inode_unlock(inode);
2753 ret = generic_write_sync(iocb, ret);
2757 #ifdef CONFIG_COMPAT
2758 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2761 case F2FS_IOC32_GETFLAGS:
2762 cmd = F2FS_IOC_GETFLAGS;
2764 case F2FS_IOC32_SETFLAGS:
2765 cmd = F2FS_IOC_SETFLAGS;
2767 case F2FS_IOC32_GETVERSION:
2768 cmd = F2FS_IOC_GETVERSION;
2770 case F2FS_IOC_START_ATOMIC_WRITE:
2771 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2772 case F2FS_IOC_START_VOLATILE_WRITE:
2773 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2774 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2775 case F2FS_IOC_SHUTDOWN:
2776 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2777 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2778 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2779 case F2FS_IOC_GARBAGE_COLLECT:
2780 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2781 case F2FS_IOC_WRITE_CHECKPOINT:
2782 case F2FS_IOC_DEFRAGMENT:
2783 case F2FS_IOC_MOVE_RANGE:
2784 case F2FS_IOC_FLUSH_DEVICE:
2785 case F2FS_IOC_GET_FEATURES:
2786 case F2FS_IOC_FSGETXATTR:
2787 case F2FS_IOC_FSSETXATTR:
2790 return -ENOIOCTLCMD;
2792 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2796 const struct file_operations f2fs_file_operations = {
2797 .llseek = f2fs_llseek,
2798 .read_iter = generic_file_read_iter,
2799 .write_iter = f2fs_file_write_iter,
2800 .open = f2fs_file_open,
2801 .release = f2fs_release_file,
2802 .mmap = f2fs_file_mmap,
2803 .flush = f2fs_file_flush,
2804 .fsync = f2fs_sync_file,
2805 .fallocate = f2fs_fallocate,
2806 .unlocked_ioctl = f2fs_ioctl,
2807 #ifdef CONFIG_COMPAT
2808 .compat_ioctl = f2fs_compat_ioctl,
2810 .splice_read = generic_file_splice_read,
2811 .splice_write = iter_file_splice_write,