1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/buffer_head.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/types.h>
16 #include <linux/compat.h>
17 #include <linux/uaccess.h>
18 #include <linux/mount.h>
19 #include <linux/pagevec.h>
20 #include <linux/uio.h>
21 #include <linux/uuid.h>
22 #include <linux/file.h>
23 #include <linux/nls.h>
32 #include <trace/events/f2fs.h>
34 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
36 struct inode *inode = file_inode(vmf->vma->vm_file);
39 down_read(&F2FS_I(inode)->i_mmap_sem);
40 ret = filemap_fault(vmf);
41 up_read(&F2FS_I(inode)->i_mmap_sem);
43 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
48 static vm_fault_t 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 = { .node_changed = false };
56 if (unlikely(f2fs_cp_error(sbi))) {
61 if (!f2fs_is_checkpoint_ready(sbi)) {
66 sb_start_pagefault(inode->i_sb);
68 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
70 file_update_time(vmf->vma->vm_file);
71 down_read(&F2FS_I(inode)->i_mmap_sem);
73 if (unlikely(page->mapping != inode->i_mapping ||
74 page_offset(page) > i_size_read(inode) ||
75 !PageUptodate(page))) {
81 /* block allocation */
82 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
83 set_new_dnode(&dn, inode, NULL, NULL, 0);
84 err = f2fs_get_block(&dn, page->index);
86 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
93 f2fs_wait_on_page_writeback(page, DATA, false, true);
95 /* wait for GCed page writeback via META_MAPPING */
96 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
99 * check to see if the page is mapped already (no holes)
101 if (PageMappedToDisk(page))
104 /* page is wholly or partially inside EOF */
105 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
106 i_size_read(inode)) {
109 offset = i_size_read(inode) & ~PAGE_MASK;
110 zero_user_segment(page, offset, PAGE_SIZE);
112 set_page_dirty(page);
113 if (!PageUptodate(page))
114 SetPageUptodate(page);
116 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
117 f2fs_update_time(sbi, REQ_TIME);
119 trace_f2fs_vm_page_mkwrite(page, DATA);
121 up_read(&F2FS_I(inode)->i_mmap_sem);
123 f2fs_balance_fs(sbi, dn.node_changed);
125 sb_end_pagefault(inode->i_sb);
127 return block_page_mkwrite_return(err);
130 static const struct vm_operations_struct f2fs_file_vm_ops = {
131 .fault = f2fs_filemap_fault,
132 .map_pages = filemap_map_pages,
133 .page_mkwrite = f2fs_vm_page_mkwrite,
136 static int get_parent_ino(struct inode *inode, nid_t *pino)
138 struct dentry *dentry;
140 inode = igrab(inode);
141 dentry = d_find_any_alias(inode);
146 *pino = parent_ino(dentry);
151 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
153 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
154 enum cp_reason_type cp_reason = CP_NO_NEEDED;
156 if (!S_ISREG(inode->i_mode))
157 cp_reason = CP_NON_REGULAR;
158 else if (inode->i_nlink != 1)
159 cp_reason = CP_HARDLINK;
160 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
161 cp_reason = CP_SB_NEED_CP;
162 else if (file_wrong_pino(inode))
163 cp_reason = CP_WRONG_PINO;
164 else if (!f2fs_space_for_roll_forward(sbi))
165 cp_reason = CP_NO_SPC_ROLL;
166 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
167 cp_reason = CP_NODE_NEED_CP;
168 else if (test_opt(sbi, FASTBOOT))
169 cp_reason = CP_FASTBOOT_MODE;
170 else if (F2FS_OPTION(sbi).active_logs == 2)
171 cp_reason = CP_SPEC_LOG_NUM;
172 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
173 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
174 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
176 cp_reason = CP_RECOVER_DIR;
181 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
183 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
185 /* But we need to avoid that there are some inode updates */
186 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
192 static void try_to_fix_pino(struct inode *inode)
194 struct f2fs_inode_info *fi = F2FS_I(inode);
197 down_write(&fi->i_sem);
198 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
199 get_parent_ino(inode, &pino)) {
200 f2fs_i_pino_write(inode, pino);
201 file_got_pino(inode);
203 up_write(&fi->i_sem);
206 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
207 int datasync, bool atomic)
209 struct inode *inode = file->f_mapping->host;
210 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
211 nid_t ino = inode->i_ino;
213 enum cp_reason_type cp_reason = 0;
214 struct writeback_control wbc = {
215 .sync_mode = WB_SYNC_ALL,
216 .nr_to_write = LONG_MAX,
219 unsigned int seq_id = 0;
221 if (unlikely(f2fs_readonly(inode->i_sb) ||
222 is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
225 trace_f2fs_sync_file_enter(inode);
227 if (S_ISDIR(inode->i_mode))
230 /* if fdatasync is triggered, let's do in-place-update */
231 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
232 set_inode_flag(inode, FI_NEED_IPU);
233 ret = file_write_and_wait_range(file, start, end);
234 clear_inode_flag(inode, FI_NEED_IPU);
237 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
241 /* if the inode is dirty, let's recover all the time */
242 if (!f2fs_skip_inode_update(inode, datasync)) {
243 f2fs_write_inode(inode, NULL);
248 * if there is no written data, don't waste time to write recovery info.
250 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
251 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
253 /* it may call write_inode just prior to fsync */
254 if (need_inode_page_update(sbi, ino))
257 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
258 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
264 * Both of fdatasync() and fsync() are able to be recovered from
267 down_read(&F2FS_I(inode)->i_sem);
268 cp_reason = need_do_checkpoint(inode);
269 up_read(&F2FS_I(inode)->i_sem);
272 /* all the dirty node pages should be flushed for POR */
273 ret = f2fs_sync_fs(inode->i_sb, 1);
276 * We've secured consistency through sync_fs. Following pino
277 * will be used only for fsynced inodes after checkpoint.
279 try_to_fix_pino(inode);
280 clear_inode_flag(inode, FI_APPEND_WRITE);
281 clear_inode_flag(inode, FI_UPDATE_WRITE);
285 atomic_inc(&sbi->wb_sync_req[NODE]);
286 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
287 atomic_dec(&sbi->wb_sync_req[NODE]);
291 /* if cp_error was enabled, we should avoid infinite loop */
292 if (unlikely(f2fs_cp_error(sbi))) {
297 if (f2fs_need_inode_block_update(sbi, ino)) {
298 f2fs_mark_inode_dirty_sync(inode, true);
299 f2fs_write_inode(inode, NULL);
304 * If it's atomic_write, it's just fine to keep write ordering. So
305 * here we don't need to wait for node write completion, since we use
306 * node chain which serializes node blocks. If one of node writes are
307 * reordered, we can see simply broken chain, resulting in stopping
308 * roll-forward recovery. It means we'll recover all or none node blocks
312 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
317 /* once recovery info is written, don't need to tack this */
318 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
319 clear_inode_flag(inode, FI_APPEND_WRITE);
321 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
322 ret = f2fs_issue_flush(sbi, inode->i_ino);
324 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
325 clear_inode_flag(inode, FI_UPDATE_WRITE);
326 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
328 f2fs_update_time(sbi, REQ_TIME);
330 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
331 f2fs_trace_ios(NULL, 1);
335 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
337 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
339 return f2fs_do_sync_file(file, start, end, datasync, false);
342 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
343 pgoff_t pgofs, int whence)
348 if (whence != SEEK_DATA)
351 /* find first dirty page index */
352 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
361 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
362 pgoff_t dirty, pgoff_t pgofs, int whence)
366 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
367 __is_valid_data_blkaddr(blkaddr))
371 if (blkaddr == NULL_ADDR)
378 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
380 struct inode *inode = file->f_mapping->host;
381 loff_t maxbytes = inode->i_sb->s_maxbytes;
382 struct dnode_of_data dn;
383 pgoff_t pgofs, end_offset, dirty;
384 loff_t data_ofs = offset;
390 isize = i_size_read(inode);
394 /* handle inline data case */
395 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
396 if (whence == SEEK_HOLE)
401 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
403 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
405 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
406 set_new_dnode(&dn, inode, NULL, NULL, 0);
407 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
408 if (err && err != -ENOENT) {
410 } else if (err == -ENOENT) {
411 /* direct node does not exists */
412 if (whence == SEEK_DATA) {
413 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
420 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
422 /* find data/hole in dnode block */
423 for (; dn.ofs_in_node < end_offset;
424 dn.ofs_in_node++, pgofs++,
425 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
428 blkaddr = datablock_addr(dn.inode,
429 dn.node_page, dn.ofs_in_node);
431 if (__is_valid_data_blkaddr(blkaddr) &&
432 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
433 blkaddr, DATA_GENERIC_ENHANCE)) {
438 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
447 if (whence == SEEK_DATA)
450 if (whence == SEEK_HOLE && data_ofs > isize)
453 return vfs_setpos(file, data_ofs, maxbytes);
459 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
461 struct inode *inode = file->f_mapping->host;
462 loff_t maxbytes = inode->i_sb->s_maxbytes;
468 return generic_file_llseek_size(file, offset, whence,
469 maxbytes, i_size_read(inode));
474 return f2fs_seek_block(file, offset, whence);
480 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
482 struct inode *inode = file_inode(file);
485 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
488 /* we don't need to use inline_data strictly */
489 err = f2fs_convert_inline_inode(inode);
494 vma->vm_ops = &f2fs_file_vm_ops;
498 static int f2fs_file_open(struct inode *inode, struct file *filp)
500 int err = fscrypt_file_open(inode, filp);
505 filp->f_mode |= FMODE_NOWAIT;
507 return dquot_file_open(inode, filp);
510 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
512 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
513 struct f2fs_node *raw_node;
514 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
518 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
519 base = get_extra_isize(dn->inode);
521 raw_node = F2FS_NODE(dn->node_page);
522 addr = blkaddr_in_node(raw_node) + base + ofs;
524 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
525 block_t blkaddr = le32_to_cpu(*addr);
527 if (blkaddr == NULL_ADDR)
530 dn->data_blkaddr = NULL_ADDR;
531 f2fs_set_data_blkaddr(dn);
533 if (__is_valid_data_blkaddr(blkaddr) &&
534 !f2fs_is_valid_blkaddr(sbi, blkaddr,
535 DATA_GENERIC_ENHANCE))
538 f2fs_invalidate_blocks(sbi, blkaddr);
539 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
540 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
547 * once we invalidate valid blkaddr in range [ofs, ofs + count],
548 * we will invalidate all blkaddr in the whole range.
550 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
552 f2fs_update_extent_cache_range(dn, fofs, 0, len);
553 dec_valid_block_count(sbi, dn->inode, nr_free);
555 dn->ofs_in_node = ofs;
557 f2fs_update_time(sbi, REQ_TIME);
558 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
559 dn->ofs_in_node, nr_free);
562 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
564 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
567 static int truncate_partial_data_page(struct inode *inode, u64 from,
570 loff_t offset = from & (PAGE_SIZE - 1);
571 pgoff_t index = from >> PAGE_SHIFT;
572 struct address_space *mapping = inode->i_mapping;
575 if (!offset && !cache_only)
579 page = find_lock_page(mapping, index);
580 if (page && PageUptodate(page))
582 f2fs_put_page(page, 1);
586 page = f2fs_get_lock_data_page(inode, index, true);
588 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
590 f2fs_wait_on_page_writeback(page, DATA, true, true);
591 zero_user(page, offset, PAGE_SIZE - offset);
593 /* An encrypted inode should have a key and truncate the last page. */
594 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
596 set_page_dirty(page);
597 f2fs_put_page(page, 1);
601 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
603 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
604 struct dnode_of_data dn;
606 int count = 0, err = 0;
608 bool truncate_page = false;
610 trace_f2fs_truncate_blocks_enter(inode, from);
612 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
614 if (free_from >= sbi->max_file_blocks)
620 ipage = f2fs_get_node_page(sbi, inode->i_ino);
622 err = PTR_ERR(ipage);
626 if (f2fs_has_inline_data(inode)) {
627 f2fs_truncate_inline_inode(inode, ipage, from);
628 f2fs_put_page(ipage, 1);
629 truncate_page = true;
633 set_new_dnode(&dn, inode, ipage, NULL, 0);
634 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
641 count = ADDRS_PER_PAGE(dn.node_page, inode);
643 count -= dn.ofs_in_node;
644 f2fs_bug_on(sbi, count < 0);
646 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
647 f2fs_truncate_data_blocks_range(&dn, count);
653 err = f2fs_truncate_inode_blocks(inode, free_from);
658 /* lastly zero out the first data page */
660 err = truncate_partial_data_page(inode, from, truncate_page);
662 trace_f2fs_truncate_blocks_exit(inode, err);
666 int f2fs_truncate(struct inode *inode)
670 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
673 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
674 S_ISLNK(inode->i_mode)))
677 trace_f2fs_truncate(inode);
679 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
680 f2fs_show_injection_info(FAULT_TRUNCATE);
684 /* we should check inline_data size */
685 if (!f2fs_may_inline_data(inode)) {
686 err = f2fs_convert_inline_inode(inode);
691 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
695 inode->i_mtime = inode->i_ctime = current_time(inode);
696 f2fs_mark_inode_dirty_sync(inode, false);
700 int f2fs_getattr(const struct path *path, struct kstat *stat,
701 u32 request_mask, unsigned int query_flags)
703 struct inode *inode = d_inode(path->dentry);
704 struct f2fs_inode_info *fi = F2FS_I(inode);
705 struct f2fs_inode *ri;
708 if (f2fs_has_extra_attr(inode) &&
709 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
710 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
711 stat->result_mask |= STATX_BTIME;
712 stat->btime.tv_sec = fi->i_crtime.tv_sec;
713 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
717 if (flags & F2FS_APPEND_FL)
718 stat->attributes |= STATX_ATTR_APPEND;
719 if (IS_ENCRYPTED(inode))
720 stat->attributes |= STATX_ATTR_ENCRYPTED;
721 if (flags & F2FS_IMMUTABLE_FL)
722 stat->attributes |= STATX_ATTR_IMMUTABLE;
723 if (flags & F2FS_NODUMP_FL)
724 stat->attributes |= STATX_ATTR_NODUMP;
726 stat->attributes_mask |= (STATX_ATTR_APPEND |
727 STATX_ATTR_ENCRYPTED |
728 STATX_ATTR_IMMUTABLE |
731 generic_fillattr(inode, stat);
733 /* we need to show initial sectors used for inline_data/dentries */
734 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
735 f2fs_has_inline_dentry(inode))
736 stat->blocks += (stat->size + 511) >> 9;
741 #ifdef CONFIG_F2FS_FS_POSIX_ACL
742 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
744 unsigned int ia_valid = attr->ia_valid;
746 if (ia_valid & ATTR_UID)
747 inode->i_uid = attr->ia_uid;
748 if (ia_valid & ATTR_GID)
749 inode->i_gid = attr->ia_gid;
750 if (ia_valid & ATTR_ATIME)
751 inode->i_atime = timespec64_trunc(attr->ia_atime,
752 inode->i_sb->s_time_gran);
753 if (ia_valid & ATTR_MTIME)
754 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
755 inode->i_sb->s_time_gran);
756 if (ia_valid & ATTR_CTIME)
757 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
758 inode->i_sb->s_time_gran);
759 if (ia_valid & ATTR_MODE) {
760 umode_t mode = attr->ia_mode;
762 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
764 set_acl_inode(inode, mode);
768 #define __setattr_copy setattr_copy
771 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
773 struct inode *inode = d_inode(dentry);
776 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
779 err = setattr_prepare(dentry, attr);
783 err = fscrypt_prepare_setattr(dentry, attr);
787 if (is_quota_modification(inode, attr)) {
788 err = dquot_initialize(inode);
792 if ((attr->ia_valid & ATTR_UID &&
793 !uid_eq(attr->ia_uid, inode->i_uid)) ||
794 (attr->ia_valid & ATTR_GID &&
795 !gid_eq(attr->ia_gid, inode->i_gid))) {
796 f2fs_lock_op(F2FS_I_SB(inode));
797 err = dquot_transfer(inode, attr);
799 set_sbi_flag(F2FS_I_SB(inode),
800 SBI_QUOTA_NEED_REPAIR);
801 f2fs_unlock_op(F2FS_I_SB(inode));
805 * update uid/gid under lock_op(), so that dquot and inode can
806 * be updated atomically.
808 if (attr->ia_valid & ATTR_UID)
809 inode->i_uid = attr->ia_uid;
810 if (attr->ia_valid & ATTR_GID)
811 inode->i_gid = attr->ia_gid;
812 f2fs_mark_inode_dirty_sync(inode, true);
813 f2fs_unlock_op(F2FS_I_SB(inode));
816 if (attr->ia_valid & ATTR_SIZE) {
817 bool to_smaller = (attr->ia_size <= i_size_read(inode));
819 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
820 down_write(&F2FS_I(inode)->i_mmap_sem);
822 truncate_setsize(inode, attr->ia_size);
825 err = f2fs_truncate(inode);
827 * do not trim all blocks after i_size if target size is
828 * larger than i_size.
830 up_write(&F2FS_I(inode)->i_mmap_sem);
831 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
837 /* should convert inline inode here */
838 if (!f2fs_may_inline_data(inode)) {
839 err = f2fs_convert_inline_inode(inode);
843 inode->i_mtime = inode->i_ctime = current_time(inode);
846 down_write(&F2FS_I(inode)->i_sem);
847 F2FS_I(inode)->last_disk_size = i_size_read(inode);
848 up_write(&F2FS_I(inode)->i_sem);
851 __setattr_copy(inode, attr);
853 if (attr->ia_valid & ATTR_MODE) {
854 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
855 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
856 inode->i_mode = F2FS_I(inode)->i_acl_mode;
857 clear_inode_flag(inode, FI_ACL_MODE);
861 /* file size may changed here */
862 f2fs_mark_inode_dirty_sync(inode, true);
864 /* inode change will produce dirty node pages flushed by checkpoint */
865 f2fs_balance_fs(F2FS_I_SB(inode), true);
870 const struct inode_operations f2fs_file_inode_operations = {
871 .getattr = f2fs_getattr,
872 .setattr = f2fs_setattr,
873 .get_acl = f2fs_get_acl,
874 .set_acl = f2fs_set_acl,
875 #ifdef CONFIG_F2FS_FS_XATTR
876 .listxattr = f2fs_listxattr,
878 .fiemap = f2fs_fiemap,
881 static int fill_zero(struct inode *inode, pgoff_t index,
882 loff_t start, loff_t len)
884 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
890 f2fs_balance_fs(sbi, true);
893 page = f2fs_get_new_data_page(inode, NULL, index, false);
897 return PTR_ERR(page);
899 f2fs_wait_on_page_writeback(page, DATA, true, true);
900 zero_user(page, start, len);
901 set_page_dirty(page);
902 f2fs_put_page(page, 1);
906 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
910 while (pg_start < pg_end) {
911 struct dnode_of_data dn;
912 pgoff_t end_offset, count;
914 set_new_dnode(&dn, inode, NULL, NULL, 0);
915 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
917 if (err == -ENOENT) {
918 pg_start = f2fs_get_next_page_offset(&dn,
925 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
926 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
928 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
930 f2fs_truncate_data_blocks_range(&dn, count);
938 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
940 pgoff_t pg_start, pg_end;
941 loff_t off_start, off_end;
944 ret = f2fs_convert_inline_inode(inode);
948 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
949 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
951 off_start = offset & (PAGE_SIZE - 1);
952 off_end = (offset + len) & (PAGE_SIZE - 1);
954 if (pg_start == pg_end) {
955 ret = fill_zero(inode, pg_start, off_start,
956 off_end - off_start);
961 ret = fill_zero(inode, pg_start++, off_start,
962 PAGE_SIZE - off_start);
967 ret = fill_zero(inode, pg_end, 0, off_end);
972 if (pg_start < pg_end) {
973 struct address_space *mapping = inode->i_mapping;
974 loff_t blk_start, blk_end;
975 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
977 f2fs_balance_fs(sbi, true);
979 blk_start = (loff_t)pg_start << PAGE_SHIFT;
980 blk_end = (loff_t)pg_end << PAGE_SHIFT;
982 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
983 down_write(&F2FS_I(inode)->i_mmap_sem);
985 truncate_inode_pages_range(mapping, blk_start,
989 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
992 up_write(&F2FS_I(inode)->i_mmap_sem);
993 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1000 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1001 int *do_replace, pgoff_t off, pgoff_t len)
1003 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1004 struct dnode_of_data dn;
1008 set_new_dnode(&dn, inode, NULL, NULL, 0);
1009 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1010 if (ret && ret != -ENOENT) {
1012 } else if (ret == -ENOENT) {
1013 if (dn.max_level == 0)
1015 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1022 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1023 dn.ofs_in_node, len);
1024 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1025 *blkaddr = datablock_addr(dn.inode,
1026 dn.node_page, dn.ofs_in_node);
1028 if (__is_valid_data_blkaddr(*blkaddr) &&
1029 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1030 DATA_GENERIC_ENHANCE)) {
1031 f2fs_put_dnode(&dn);
1032 return -EFSCORRUPTED;
1035 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1037 if (test_opt(sbi, LFS)) {
1038 f2fs_put_dnode(&dn);
1042 /* do not invalidate this block address */
1043 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1047 f2fs_put_dnode(&dn);
1056 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1057 int *do_replace, pgoff_t off, int len)
1059 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1060 struct dnode_of_data dn;
1063 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1064 if (*do_replace == 0)
1067 set_new_dnode(&dn, inode, NULL, NULL, 0);
1068 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1070 dec_valid_block_count(sbi, inode, 1);
1071 f2fs_invalidate_blocks(sbi, *blkaddr);
1073 f2fs_update_data_blkaddr(&dn, *blkaddr);
1075 f2fs_put_dnode(&dn);
1080 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1081 block_t *blkaddr, int *do_replace,
1082 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1084 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1089 if (blkaddr[i] == NULL_ADDR && !full) {
1094 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1095 struct dnode_of_data dn;
1096 struct node_info ni;
1100 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1101 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1105 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1107 f2fs_put_dnode(&dn);
1111 ilen = min((pgoff_t)
1112 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1113 dn.ofs_in_node, len - i);
1115 dn.data_blkaddr = datablock_addr(dn.inode,
1116 dn.node_page, dn.ofs_in_node);
1117 f2fs_truncate_data_blocks_range(&dn, 1);
1119 if (do_replace[i]) {
1120 f2fs_i_blocks_write(src_inode,
1122 f2fs_i_blocks_write(dst_inode,
1124 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1125 blkaddr[i], ni.version, true, false);
1131 new_size = (dst + i) << PAGE_SHIFT;
1132 if (dst_inode->i_size < new_size)
1133 f2fs_i_size_write(dst_inode, new_size);
1134 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1136 f2fs_put_dnode(&dn);
1138 struct page *psrc, *pdst;
1140 psrc = f2fs_get_lock_data_page(src_inode,
1143 return PTR_ERR(psrc);
1144 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1147 f2fs_put_page(psrc, 1);
1148 return PTR_ERR(pdst);
1150 f2fs_copy_page(psrc, pdst);
1151 set_page_dirty(pdst);
1152 f2fs_put_page(pdst, 1);
1153 f2fs_put_page(psrc, 1);
1155 ret = f2fs_truncate_hole(src_inode,
1156 src + i, src + i + 1);
1165 static int __exchange_data_block(struct inode *src_inode,
1166 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1167 pgoff_t len, bool full)
1169 block_t *src_blkaddr;
1175 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1177 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1178 array_size(olen, sizeof(block_t)),
1183 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1184 array_size(olen, sizeof(int)),
1187 kvfree(src_blkaddr);
1191 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1192 do_replace, src, olen);
1196 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1197 do_replace, src, dst, olen, full);
1205 kvfree(src_blkaddr);
1211 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1212 kvfree(src_blkaddr);
1217 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1219 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1220 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1221 pgoff_t start = offset >> PAGE_SHIFT;
1222 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1225 f2fs_balance_fs(sbi, true);
1227 /* avoid gc operation during block exchange */
1228 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1229 down_write(&F2FS_I(inode)->i_mmap_sem);
1232 f2fs_drop_extent_tree(inode);
1233 truncate_pagecache(inode, offset);
1234 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1235 f2fs_unlock_op(sbi);
1237 up_write(&F2FS_I(inode)->i_mmap_sem);
1238 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1242 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1247 if (offset + len >= i_size_read(inode))
1250 /* collapse range should be aligned to block size of f2fs. */
1251 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1254 ret = f2fs_convert_inline_inode(inode);
1258 /* write out all dirty pages from offset */
1259 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1263 ret = f2fs_do_collapse(inode, offset, len);
1267 /* write out all moved pages, if possible */
1268 down_write(&F2FS_I(inode)->i_mmap_sem);
1269 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1270 truncate_pagecache(inode, offset);
1272 new_size = i_size_read(inode) - len;
1273 truncate_pagecache(inode, new_size);
1275 ret = f2fs_truncate_blocks(inode, new_size, true);
1276 up_write(&F2FS_I(inode)->i_mmap_sem);
1278 f2fs_i_size_write(inode, new_size);
1282 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1285 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1286 pgoff_t index = start;
1287 unsigned int ofs_in_node = dn->ofs_in_node;
1291 for (; index < end; index++, dn->ofs_in_node++) {
1292 if (datablock_addr(dn->inode, dn->node_page,
1293 dn->ofs_in_node) == NULL_ADDR)
1297 dn->ofs_in_node = ofs_in_node;
1298 ret = f2fs_reserve_new_blocks(dn, count);
1302 dn->ofs_in_node = ofs_in_node;
1303 for (index = start; index < end; index++, dn->ofs_in_node++) {
1304 dn->data_blkaddr = datablock_addr(dn->inode,
1305 dn->node_page, dn->ofs_in_node);
1307 * f2fs_reserve_new_blocks will not guarantee entire block
1310 if (dn->data_blkaddr == NULL_ADDR) {
1314 if (dn->data_blkaddr != NEW_ADDR) {
1315 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1316 dn->data_blkaddr = NEW_ADDR;
1317 f2fs_set_data_blkaddr(dn);
1321 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1326 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1329 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1330 struct address_space *mapping = inode->i_mapping;
1331 pgoff_t index, pg_start, pg_end;
1332 loff_t new_size = i_size_read(inode);
1333 loff_t off_start, off_end;
1336 ret = inode_newsize_ok(inode, (len + offset));
1340 ret = f2fs_convert_inline_inode(inode);
1344 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1348 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1349 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1351 off_start = offset & (PAGE_SIZE - 1);
1352 off_end = (offset + len) & (PAGE_SIZE - 1);
1354 if (pg_start == pg_end) {
1355 ret = fill_zero(inode, pg_start, off_start,
1356 off_end - off_start);
1360 new_size = max_t(loff_t, new_size, offset + len);
1363 ret = fill_zero(inode, pg_start++, off_start,
1364 PAGE_SIZE - off_start);
1368 new_size = max_t(loff_t, new_size,
1369 (loff_t)pg_start << PAGE_SHIFT);
1372 for (index = pg_start; index < pg_end;) {
1373 struct dnode_of_data dn;
1374 unsigned int end_offset;
1377 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1378 down_write(&F2FS_I(inode)->i_mmap_sem);
1380 truncate_pagecache_range(inode,
1381 (loff_t)index << PAGE_SHIFT,
1382 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1386 set_new_dnode(&dn, inode, NULL, NULL, 0);
1387 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1389 f2fs_unlock_op(sbi);
1390 up_write(&F2FS_I(inode)->i_mmap_sem);
1391 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1395 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1396 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1398 ret = f2fs_do_zero_range(&dn, index, end);
1399 f2fs_put_dnode(&dn);
1401 f2fs_unlock_op(sbi);
1402 up_write(&F2FS_I(inode)->i_mmap_sem);
1403 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1405 f2fs_balance_fs(sbi, dn.node_changed);
1411 new_size = max_t(loff_t, new_size,
1412 (loff_t)index << PAGE_SHIFT);
1416 ret = fill_zero(inode, pg_end, 0, off_end);
1420 new_size = max_t(loff_t, new_size, offset + len);
1425 if (new_size > i_size_read(inode)) {
1426 if (mode & FALLOC_FL_KEEP_SIZE)
1427 file_set_keep_isize(inode);
1429 f2fs_i_size_write(inode, new_size);
1434 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1436 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1437 pgoff_t nr, pg_start, pg_end, delta, idx;
1441 new_size = i_size_read(inode) + len;
1442 ret = inode_newsize_ok(inode, new_size);
1446 if (offset >= i_size_read(inode))
1449 /* insert range should be aligned to block size of f2fs. */
1450 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1453 ret = f2fs_convert_inline_inode(inode);
1457 f2fs_balance_fs(sbi, true);
1459 down_write(&F2FS_I(inode)->i_mmap_sem);
1460 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1461 up_write(&F2FS_I(inode)->i_mmap_sem);
1465 /* write out all dirty pages from offset */
1466 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1470 pg_start = offset >> PAGE_SHIFT;
1471 pg_end = (offset + len) >> PAGE_SHIFT;
1472 delta = pg_end - pg_start;
1473 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1475 /* avoid gc operation during block exchange */
1476 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1477 down_write(&F2FS_I(inode)->i_mmap_sem);
1478 truncate_pagecache(inode, offset);
1480 while (!ret && idx > pg_start) {
1481 nr = idx - pg_start;
1487 f2fs_drop_extent_tree(inode);
1489 ret = __exchange_data_block(inode, inode, idx,
1490 idx + delta, nr, false);
1491 f2fs_unlock_op(sbi);
1493 up_write(&F2FS_I(inode)->i_mmap_sem);
1494 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1496 /* write out all moved pages, if possible */
1497 down_write(&F2FS_I(inode)->i_mmap_sem);
1498 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1499 truncate_pagecache(inode, offset);
1500 up_write(&F2FS_I(inode)->i_mmap_sem);
1503 f2fs_i_size_write(inode, new_size);
1507 static int expand_inode_data(struct inode *inode, loff_t offset,
1508 loff_t len, int mode)
1510 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1511 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1512 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1513 .m_may_create = true };
1515 loff_t new_size = i_size_read(inode);
1519 err = inode_newsize_ok(inode, (len + offset));
1523 err = f2fs_convert_inline_inode(inode);
1527 f2fs_balance_fs(sbi, true);
1529 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1530 off_end = (offset + len) & (PAGE_SIZE - 1);
1532 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1533 map.m_len = pg_end - map.m_lblk;
1537 if (f2fs_is_pinned_file(inode))
1538 map.m_seg_type = CURSEG_COLD_DATA;
1540 err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
1541 F2FS_GET_BLOCK_PRE_DIO :
1542 F2FS_GET_BLOCK_PRE_AIO));
1549 last_off = map.m_lblk + map.m_len - 1;
1551 /* update new size to the failed position */
1552 new_size = (last_off == pg_end) ? offset + len :
1553 (loff_t)(last_off + 1) << PAGE_SHIFT;
1555 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1558 if (new_size > i_size_read(inode)) {
1559 if (mode & FALLOC_FL_KEEP_SIZE)
1560 file_set_keep_isize(inode);
1562 f2fs_i_size_write(inode, new_size);
1568 static long f2fs_fallocate(struct file *file, int mode,
1569 loff_t offset, loff_t len)
1571 struct inode *inode = file_inode(file);
1574 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1576 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1579 /* f2fs only support ->fallocate for regular file */
1580 if (!S_ISREG(inode->i_mode))
1583 if (IS_ENCRYPTED(inode) &&
1584 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1587 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1588 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1589 FALLOC_FL_INSERT_RANGE))
1594 if (mode & FALLOC_FL_PUNCH_HOLE) {
1595 if (offset >= inode->i_size)
1598 ret = punch_hole(inode, offset, len);
1599 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1600 ret = f2fs_collapse_range(inode, offset, len);
1601 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1602 ret = f2fs_zero_range(inode, offset, len, mode);
1603 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1604 ret = f2fs_insert_range(inode, offset, len);
1606 ret = expand_inode_data(inode, offset, len, mode);
1610 inode->i_mtime = inode->i_ctime = current_time(inode);
1611 f2fs_mark_inode_dirty_sync(inode, false);
1612 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1616 inode_unlock(inode);
1618 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1622 static int f2fs_release_file(struct inode *inode, struct file *filp)
1625 * f2fs_relase_file is called at every close calls. So we should
1626 * not drop any inmemory pages by close called by other process.
1628 if (!(filp->f_mode & FMODE_WRITE) ||
1629 atomic_read(&inode->i_writecount) != 1)
1632 /* some remained atomic pages should discarded */
1633 if (f2fs_is_atomic_file(inode))
1634 f2fs_drop_inmem_pages(inode);
1635 if (f2fs_is_volatile_file(inode)) {
1636 set_inode_flag(inode, FI_DROP_CACHE);
1637 filemap_fdatawrite(inode->i_mapping);
1638 clear_inode_flag(inode, FI_DROP_CACHE);
1639 clear_inode_flag(inode, FI_VOLATILE_FILE);
1640 stat_dec_volatile_write(inode);
1645 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1647 struct inode *inode = file_inode(file);
1650 * If the process doing a transaction is crashed, we should do
1651 * roll-back. Otherwise, other reader/write can see corrupted database
1652 * until all the writers close its file. Since this should be done
1653 * before dropping file lock, it needs to do in ->flush.
1655 if (f2fs_is_atomic_file(inode) &&
1656 F2FS_I(inode)->inmem_task == current)
1657 f2fs_drop_inmem_pages(inode);
1661 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1663 struct f2fs_inode_info *fi = F2FS_I(inode);
1665 /* Is it quota file? Do not allow user to mess with it */
1666 if (IS_NOQUOTA(inode))
1669 if ((iflags ^ fi->i_flags) & F2FS_CASEFOLD_FL) {
1670 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1672 if (!f2fs_empty_dir(inode))
1676 fi->i_flags = iflags | (fi->i_flags & ~mask);
1678 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1679 set_inode_flag(inode, FI_PROJ_INHERIT);
1681 clear_inode_flag(inode, FI_PROJ_INHERIT);
1683 inode->i_ctime = current_time(inode);
1684 f2fs_set_inode_flags(inode);
1685 f2fs_mark_inode_dirty_sync(inode, true);
1689 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1692 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1693 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1694 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1695 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1698 static const struct {
1701 } f2fs_fsflags_map[] = {
1702 { F2FS_SYNC_FL, FS_SYNC_FL },
1703 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1704 { F2FS_APPEND_FL, FS_APPEND_FL },
1705 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1706 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1707 { F2FS_INDEX_FL, FS_INDEX_FL },
1708 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1709 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1710 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1713 #define F2FS_GETTABLE_FS_FL ( \
1721 FS_PROJINHERIT_FL | \
1723 FS_INLINE_DATA_FL | \
1727 #define F2FS_SETTABLE_FS_FL ( \
1734 FS_PROJINHERIT_FL | \
1737 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1738 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1743 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1744 if (iflags & f2fs_fsflags_map[i].iflag)
1745 fsflags |= f2fs_fsflags_map[i].fsflag;
1750 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1751 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1756 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1757 if (fsflags & f2fs_fsflags_map[i].fsflag)
1758 iflags |= f2fs_fsflags_map[i].iflag;
1763 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1765 struct inode *inode = file_inode(filp);
1766 struct f2fs_inode_info *fi = F2FS_I(inode);
1767 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1769 if (IS_ENCRYPTED(inode))
1770 fsflags |= FS_ENCRYPT_FL;
1771 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1772 fsflags |= FS_INLINE_DATA_FL;
1773 if (is_inode_flag_set(inode, FI_PIN_FILE))
1774 fsflags |= FS_NOCOW_FL;
1776 fsflags &= F2FS_GETTABLE_FS_FL;
1778 return put_user(fsflags, (int __user *)arg);
1781 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1783 struct inode *inode = file_inode(filp);
1784 struct f2fs_inode_info *fi = F2FS_I(inode);
1785 u32 fsflags, old_fsflags;
1789 if (!inode_owner_or_capable(inode))
1792 if (get_user(fsflags, (int __user *)arg))
1795 if (fsflags & ~F2FS_GETTABLE_FS_FL)
1797 fsflags &= F2FS_SETTABLE_FS_FL;
1799 iflags = f2fs_fsflags_to_iflags(fsflags);
1800 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1803 ret = mnt_want_write_file(filp);
1809 old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1810 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1814 ret = f2fs_setflags_common(inode, iflags,
1815 f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1817 inode_unlock(inode);
1818 mnt_drop_write_file(filp);
1822 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1824 struct inode *inode = file_inode(filp);
1826 return put_user(inode->i_generation, (int __user *)arg);
1829 static int f2fs_ioc_start_atomic_write(struct file *filp)
1831 struct inode *inode = file_inode(filp);
1834 if (!inode_owner_or_capable(inode))
1837 if (!S_ISREG(inode->i_mode))
1840 if (filp->f_flags & O_DIRECT)
1843 ret = mnt_want_write_file(filp);
1849 if (f2fs_is_atomic_file(inode)) {
1850 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1855 ret = f2fs_convert_inline_inode(inode);
1859 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1862 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1863 * f2fs_is_atomic_file.
1865 if (get_dirty_pages(inode))
1866 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1867 inode->i_ino, get_dirty_pages(inode));
1868 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1870 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1874 set_inode_flag(inode, FI_ATOMIC_FILE);
1875 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1876 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1878 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1879 F2FS_I(inode)->inmem_task = current;
1880 stat_inc_atomic_write(inode);
1881 stat_update_max_atomic_write(inode);
1883 inode_unlock(inode);
1884 mnt_drop_write_file(filp);
1888 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1890 struct inode *inode = file_inode(filp);
1893 if (!inode_owner_or_capable(inode))
1896 ret = mnt_want_write_file(filp);
1900 f2fs_balance_fs(F2FS_I_SB(inode), true);
1904 if (f2fs_is_volatile_file(inode)) {
1909 if (f2fs_is_atomic_file(inode)) {
1910 ret = f2fs_commit_inmem_pages(inode);
1914 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1916 clear_inode_flag(inode, FI_ATOMIC_FILE);
1917 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1918 stat_dec_atomic_write(inode);
1921 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1924 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1925 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1928 inode_unlock(inode);
1929 mnt_drop_write_file(filp);
1933 static int f2fs_ioc_start_volatile_write(struct file *filp)
1935 struct inode *inode = file_inode(filp);
1938 if (!inode_owner_or_capable(inode))
1941 if (!S_ISREG(inode->i_mode))
1944 ret = mnt_want_write_file(filp);
1950 if (f2fs_is_volatile_file(inode))
1953 ret = f2fs_convert_inline_inode(inode);
1957 stat_inc_volatile_write(inode);
1958 stat_update_max_volatile_write(inode);
1960 set_inode_flag(inode, FI_VOLATILE_FILE);
1961 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1963 inode_unlock(inode);
1964 mnt_drop_write_file(filp);
1968 static int f2fs_ioc_release_volatile_write(struct file *filp)
1970 struct inode *inode = file_inode(filp);
1973 if (!inode_owner_or_capable(inode))
1976 ret = mnt_want_write_file(filp);
1982 if (!f2fs_is_volatile_file(inode))
1985 if (!f2fs_is_first_block_written(inode)) {
1986 ret = truncate_partial_data_page(inode, 0, true);
1990 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1992 inode_unlock(inode);
1993 mnt_drop_write_file(filp);
1997 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1999 struct inode *inode = file_inode(filp);
2002 if (!inode_owner_or_capable(inode))
2005 ret = mnt_want_write_file(filp);
2011 if (f2fs_is_atomic_file(inode))
2012 f2fs_drop_inmem_pages(inode);
2013 if (f2fs_is_volatile_file(inode)) {
2014 clear_inode_flag(inode, FI_VOLATILE_FILE);
2015 stat_dec_volatile_write(inode);
2016 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2019 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2021 inode_unlock(inode);
2023 mnt_drop_write_file(filp);
2024 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2028 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2030 struct inode *inode = file_inode(filp);
2031 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2032 struct super_block *sb = sbi->sb;
2036 if (!capable(CAP_SYS_ADMIN))
2039 if (get_user(in, (__u32 __user *)arg))
2042 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2043 ret = mnt_want_write_file(filp);
2049 case F2FS_GOING_DOWN_FULLSYNC:
2050 sb = freeze_bdev(sb->s_bdev);
2056 f2fs_stop_checkpoint(sbi, false);
2057 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2058 thaw_bdev(sb->s_bdev, sb);
2061 case F2FS_GOING_DOWN_METASYNC:
2062 /* do checkpoint only */
2063 ret = f2fs_sync_fs(sb, 1);
2066 f2fs_stop_checkpoint(sbi, false);
2067 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2069 case F2FS_GOING_DOWN_NOSYNC:
2070 f2fs_stop_checkpoint(sbi, false);
2071 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2073 case F2FS_GOING_DOWN_METAFLUSH:
2074 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2075 f2fs_stop_checkpoint(sbi, false);
2076 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2078 case F2FS_GOING_DOWN_NEED_FSCK:
2079 set_sbi_flag(sbi, SBI_NEED_FSCK);
2080 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2081 set_sbi_flag(sbi, SBI_IS_DIRTY);
2082 /* do checkpoint only */
2083 ret = f2fs_sync_fs(sb, 1);
2090 f2fs_stop_gc_thread(sbi);
2091 f2fs_stop_discard_thread(sbi);
2093 f2fs_drop_discard_cmd(sbi);
2094 clear_opt(sbi, DISCARD);
2096 f2fs_update_time(sbi, REQ_TIME);
2098 if (in != F2FS_GOING_DOWN_FULLSYNC)
2099 mnt_drop_write_file(filp);
2101 trace_f2fs_shutdown(sbi, in, ret);
2106 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2108 struct inode *inode = file_inode(filp);
2109 struct super_block *sb = inode->i_sb;
2110 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2111 struct fstrim_range range;
2114 if (!capable(CAP_SYS_ADMIN))
2117 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2120 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2124 ret = mnt_want_write_file(filp);
2128 range.minlen = max((unsigned int)range.minlen,
2129 q->limits.discard_granularity);
2130 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2131 mnt_drop_write_file(filp);
2135 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2138 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2142 static bool uuid_is_nonzero(__u8 u[16])
2146 for (i = 0; i < 16; i++)
2152 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2154 struct inode *inode = file_inode(filp);
2156 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2159 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2161 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2164 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2166 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2168 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2171 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2173 struct inode *inode = file_inode(filp);
2174 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2177 if (!f2fs_sb_has_encrypt(sbi))
2180 err = mnt_want_write_file(filp);
2184 down_write(&sbi->sb_lock);
2186 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2189 /* update superblock with uuid */
2190 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2192 err = f2fs_commit_super(sbi, false);
2195 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2199 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2203 up_write(&sbi->sb_lock);
2204 mnt_drop_write_file(filp);
2208 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2210 struct inode *inode = file_inode(filp);
2211 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2215 if (!capable(CAP_SYS_ADMIN))
2218 if (get_user(sync, (__u32 __user *)arg))
2221 if (f2fs_readonly(sbi->sb))
2224 ret = mnt_want_write_file(filp);
2229 if (!mutex_trylock(&sbi->gc_mutex)) {
2234 mutex_lock(&sbi->gc_mutex);
2237 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2239 mnt_drop_write_file(filp);
2243 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2245 struct inode *inode = file_inode(filp);
2246 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2247 struct f2fs_gc_range range;
2251 if (!capable(CAP_SYS_ADMIN))
2254 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2258 if (f2fs_readonly(sbi->sb))
2261 end = range.start + range.len;
2262 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2266 ret = mnt_want_write_file(filp);
2272 if (!mutex_trylock(&sbi->gc_mutex)) {
2277 mutex_lock(&sbi->gc_mutex);
2280 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2281 range.start += BLKS_PER_SEC(sbi);
2282 if (range.start <= end)
2285 mnt_drop_write_file(filp);
2289 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2291 struct inode *inode = file_inode(filp);
2292 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2295 if (!capable(CAP_SYS_ADMIN))
2298 if (f2fs_readonly(sbi->sb))
2301 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2302 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2306 ret = mnt_want_write_file(filp);
2310 ret = f2fs_sync_fs(sbi->sb, 1);
2312 mnt_drop_write_file(filp);
2316 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2318 struct f2fs_defragment *range)
2320 struct inode *inode = file_inode(filp);
2321 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2322 .m_seg_type = NO_CHECK_TYPE ,
2323 .m_may_create = false };
2324 struct extent_info ei = {0, 0, 0};
2325 pgoff_t pg_start, pg_end, next_pgofs;
2326 unsigned int blk_per_seg = sbi->blocks_per_seg;
2327 unsigned int total = 0, sec_num;
2328 block_t blk_end = 0;
2329 bool fragmented = false;
2332 /* if in-place-update policy is enabled, don't waste time here */
2333 if (f2fs_should_update_inplace(inode, NULL))
2336 pg_start = range->start >> PAGE_SHIFT;
2337 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2339 f2fs_balance_fs(sbi, true);
2343 /* writeback all dirty pages in the range */
2344 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2345 range->start + range->len - 1);
2350 * lookup mapping info in extent cache, skip defragmenting if physical
2351 * block addresses are continuous.
2353 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2354 if (ei.fofs + ei.len >= pg_end)
2358 map.m_lblk = pg_start;
2359 map.m_next_pgofs = &next_pgofs;
2362 * lookup mapping info in dnode page cache, skip defragmenting if all
2363 * physical block addresses are continuous even if there are hole(s)
2364 * in logical blocks.
2366 while (map.m_lblk < pg_end) {
2367 map.m_len = pg_end - map.m_lblk;
2368 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2372 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2373 map.m_lblk = next_pgofs;
2377 if (blk_end && blk_end != map.m_pblk)
2380 /* record total count of block that we're going to move */
2383 blk_end = map.m_pblk + map.m_len;
2385 map.m_lblk += map.m_len;
2393 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2396 * make sure there are enough free section for LFS allocation, this can
2397 * avoid defragment running in SSR mode when free section are allocated
2400 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2405 map.m_lblk = pg_start;
2406 map.m_len = pg_end - pg_start;
2409 while (map.m_lblk < pg_end) {
2414 map.m_len = pg_end - map.m_lblk;
2415 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2419 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2420 map.m_lblk = next_pgofs;
2424 set_inode_flag(inode, FI_DO_DEFRAG);
2427 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2430 page = f2fs_get_lock_data_page(inode, idx, true);
2432 err = PTR_ERR(page);
2436 set_page_dirty(page);
2437 f2fs_put_page(page, 1);
2446 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2449 clear_inode_flag(inode, FI_DO_DEFRAG);
2451 err = filemap_fdatawrite(inode->i_mapping);
2456 clear_inode_flag(inode, FI_DO_DEFRAG);
2458 inode_unlock(inode);
2460 range->len = (u64)total << PAGE_SHIFT;
2464 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2466 struct inode *inode = file_inode(filp);
2467 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2468 struct f2fs_defragment range;
2471 if (!capable(CAP_SYS_ADMIN))
2474 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2477 if (f2fs_readonly(sbi->sb))
2480 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2484 /* verify alignment of offset & size */
2485 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2488 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2489 sbi->max_file_blocks))
2492 err = mnt_want_write_file(filp);
2496 err = f2fs_defragment_range(sbi, filp, &range);
2497 mnt_drop_write_file(filp);
2499 f2fs_update_time(sbi, REQ_TIME);
2503 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2510 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2511 struct file *file_out, loff_t pos_out, size_t len)
2513 struct inode *src = file_inode(file_in);
2514 struct inode *dst = file_inode(file_out);
2515 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2516 size_t olen = len, dst_max_i_size = 0;
2520 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2521 src->i_sb != dst->i_sb)
2524 if (unlikely(f2fs_readonly(src->i_sb)))
2527 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2530 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2534 if (pos_in == pos_out)
2536 if (pos_out > pos_in && pos_out < pos_in + len)
2543 if (!inode_trylock(dst))
2548 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2551 olen = len = src->i_size - pos_in;
2552 if (pos_in + len == src->i_size)
2553 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2559 dst_osize = dst->i_size;
2560 if (pos_out + olen > dst->i_size)
2561 dst_max_i_size = pos_out + olen;
2563 /* verify the end result is block aligned */
2564 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2565 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2566 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2569 ret = f2fs_convert_inline_inode(src);
2573 ret = f2fs_convert_inline_inode(dst);
2577 /* write out all dirty pages from offset */
2578 ret = filemap_write_and_wait_range(src->i_mapping,
2579 pos_in, pos_in + len);
2583 ret = filemap_write_and_wait_range(dst->i_mapping,
2584 pos_out, pos_out + len);
2588 f2fs_balance_fs(sbi, true);
2590 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2593 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2598 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2599 pos_out >> F2FS_BLKSIZE_BITS,
2600 len >> F2FS_BLKSIZE_BITS, false);
2604 f2fs_i_size_write(dst, dst_max_i_size);
2605 else if (dst_osize != dst->i_size)
2606 f2fs_i_size_write(dst, dst_osize);
2608 f2fs_unlock_op(sbi);
2611 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2613 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2622 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2624 struct f2fs_move_range range;
2628 if (!(filp->f_mode & FMODE_READ) ||
2629 !(filp->f_mode & FMODE_WRITE))
2632 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2636 dst = fdget(range.dst_fd);
2640 if (!(dst.file->f_mode & FMODE_WRITE)) {
2645 err = mnt_want_write_file(filp);
2649 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2650 range.pos_out, range.len);
2652 mnt_drop_write_file(filp);
2656 if (copy_to_user((struct f2fs_move_range __user *)arg,
2657 &range, sizeof(range)))
2664 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2666 struct inode *inode = file_inode(filp);
2667 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2668 struct sit_info *sm = SIT_I(sbi);
2669 unsigned int start_segno = 0, end_segno = 0;
2670 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2671 struct f2fs_flush_device range;
2674 if (!capable(CAP_SYS_ADMIN))
2677 if (f2fs_readonly(sbi->sb))
2680 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2683 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2687 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2688 __is_large_section(sbi)) {
2689 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2690 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2694 ret = mnt_want_write_file(filp);
2698 if (range.dev_num != 0)
2699 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2700 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2702 start_segno = sm->last_victim[FLUSH_DEVICE];
2703 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2704 start_segno = dev_start_segno;
2705 end_segno = min(start_segno + range.segments, dev_end_segno);
2707 while (start_segno < end_segno) {
2708 if (!mutex_trylock(&sbi->gc_mutex)) {
2712 sm->last_victim[GC_CB] = end_segno + 1;
2713 sm->last_victim[GC_GREEDY] = end_segno + 1;
2714 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2715 ret = f2fs_gc(sbi, true, true, start_segno);
2723 mnt_drop_write_file(filp);
2727 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2729 struct inode *inode = file_inode(filp);
2730 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2732 /* Must validate to set it with SQLite behavior in Android. */
2733 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2735 return put_user(sb_feature, (u32 __user *)arg);
2739 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2741 struct dquot *transfer_to[MAXQUOTAS] = {};
2742 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2743 struct super_block *sb = sbi->sb;
2746 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2747 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2748 err = __dquot_transfer(inode, transfer_to);
2750 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2751 dqput(transfer_to[PRJQUOTA]);
2756 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2758 struct inode *inode = file_inode(filp);
2759 struct f2fs_inode_info *fi = F2FS_I(inode);
2760 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2765 if (!f2fs_sb_has_project_quota(sbi)) {
2766 if (projid != F2FS_DEF_PROJID)
2772 if (!f2fs_has_extra_attr(inode))
2775 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2777 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2781 /* Is it quota file? Do not allow user to mess with it */
2782 if (IS_NOQUOTA(inode))
2785 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2787 return PTR_ERR(ipage);
2789 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2792 f2fs_put_page(ipage, 1);
2795 f2fs_put_page(ipage, 1);
2797 err = dquot_initialize(inode);
2802 err = f2fs_transfer_project_quota(inode, kprojid);
2806 F2FS_I(inode)->i_projid = kprojid;
2807 inode->i_ctime = current_time(inode);
2808 f2fs_mark_inode_dirty_sync(inode, true);
2810 f2fs_unlock_op(sbi);
2814 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2819 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2821 if (projid != F2FS_DEF_PROJID)
2827 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
2830 * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
2831 * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
2832 * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
2835 static const struct {
2838 } f2fs_xflags_map[] = {
2839 { F2FS_SYNC_FL, FS_XFLAG_SYNC },
2840 { F2FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE },
2841 { F2FS_APPEND_FL, FS_XFLAG_APPEND },
2842 { F2FS_NODUMP_FL, FS_XFLAG_NODUMP },
2843 { F2FS_NOATIME_FL, FS_XFLAG_NOATIME },
2844 { F2FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT },
2847 #define F2FS_SUPPORTED_XFLAGS ( \
2849 FS_XFLAG_IMMUTABLE | \
2852 FS_XFLAG_NOATIME | \
2853 FS_XFLAG_PROJINHERIT)
2855 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
2856 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
2861 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2862 if (iflags & f2fs_xflags_map[i].iflag)
2863 xflags |= f2fs_xflags_map[i].xflag;
2868 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
2869 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
2874 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2875 if (xflags & f2fs_xflags_map[i].xflag)
2876 iflags |= f2fs_xflags_map[i].iflag;
2881 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
2883 struct f2fs_inode_info *fi = F2FS_I(inode);
2885 simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
2887 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2888 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2891 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2893 struct inode *inode = file_inode(filp);
2896 f2fs_fill_fsxattr(inode, &fa);
2898 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2903 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2905 struct inode *inode = file_inode(filp);
2906 struct fsxattr fa, old_fa;
2910 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2913 /* Make sure caller has proper permission */
2914 if (!inode_owner_or_capable(inode))
2917 if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
2920 iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2921 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2924 err = mnt_want_write_file(filp);
2930 f2fs_fill_fsxattr(inode, &old_fa);
2931 err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
2935 err = f2fs_setflags_common(inode, iflags,
2936 f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
2940 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2942 inode_unlock(inode);
2943 mnt_drop_write_file(filp);
2947 int f2fs_pin_file_control(struct inode *inode, bool inc)
2949 struct f2fs_inode_info *fi = F2FS_I(inode);
2950 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2952 /* Use i_gc_failures for normal file as a risk signal. */
2954 f2fs_i_gc_failures_write(inode,
2955 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2957 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2958 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
2959 __func__, inode->i_ino,
2960 fi->i_gc_failures[GC_FAILURE_PIN]);
2961 clear_inode_flag(inode, FI_PIN_FILE);
2967 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2969 struct inode *inode = file_inode(filp);
2973 if (get_user(pin, (__u32 __user *)arg))
2976 if (!S_ISREG(inode->i_mode))
2979 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2982 ret = mnt_want_write_file(filp);
2988 if (f2fs_should_update_outplace(inode, NULL)) {
2994 clear_inode_flag(inode, FI_PIN_FILE);
2995 f2fs_i_gc_failures_write(inode, 0);
2999 if (f2fs_pin_file_control(inode, false)) {
3003 ret = f2fs_convert_inline_inode(inode);
3007 set_inode_flag(inode, FI_PIN_FILE);
3008 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3010 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3012 inode_unlock(inode);
3013 mnt_drop_write_file(filp);
3017 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3019 struct inode *inode = file_inode(filp);
3022 if (is_inode_flag_set(inode, FI_PIN_FILE))
3023 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3024 return put_user(pin, (u32 __user *)arg);
3027 int f2fs_precache_extents(struct inode *inode)
3029 struct f2fs_inode_info *fi = F2FS_I(inode);
3030 struct f2fs_map_blocks map;
3031 pgoff_t m_next_extent;
3035 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3039 map.m_next_pgofs = NULL;
3040 map.m_next_extent = &m_next_extent;
3041 map.m_seg_type = NO_CHECK_TYPE;
3042 map.m_may_create = false;
3043 end = F2FS_I_SB(inode)->max_file_blocks;
3045 while (map.m_lblk < end) {
3046 map.m_len = end - map.m_lblk;
3048 down_write(&fi->i_gc_rwsem[WRITE]);
3049 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3050 up_write(&fi->i_gc_rwsem[WRITE]);
3054 map.m_lblk = m_next_extent;
3060 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3062 return f2fs_precache_extents(file_inode(filp));
3065 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3067 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3071 if (!capable(CAP_SYS_ADMIN))
3074 if (f2fs_readonly(sbi->sb))
3077 if (copy_from_user(&block_count, (void __user *)arg,
3078 sizeof(block_count)))
3081 ret = f2fs_resize_fs(sbi, block_count);
3086 static int f2fs_get_volume_name(struct file *filp, unsigned long arg)
3088 struct inode *inode = file_inode(filp);
3089 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3094 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3098 down_read(&sbi->sb_lock);
3099 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3100 ARRAY_SIZE(sbi->raw_super->volume_name),
3101 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3102 up_read(&sbi->sb_lock);
3104 if (copy_to_user((char __user *)arg, vbuf,
3105 min(FSLABEL_MAX, count)))
3112 static int f2fs_set_volume_name(struct file *filp, unsigned long arg)
3114 struct inode *inode = file_inode(filp);
3115 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3119 if (!capable(CAP_SYS_ADMIN))
3122 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3124 return PTR_ERR(vbuf);
3126 err = mnt_want_write_file(filp);
3130 down_write(&sbi->sb_lock);
3132 memset(sbi->raw_super->volume_name, 0,
3133 sizeof(sbi->raw_super->volume_name));
3134 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3135 sbi->raw_super->volume_name,
3136 ARRAY_SIZE(sbi->raw_super->volume_name));
3138 err = f2fs_commit_super(sbi, false);
3140 up_write(&sbi->sb_lock);
3142 mnt_drop_write_file(filp);
3148 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3150 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3152 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
3156 case F2FS_IOC_GETFLAGS:
3157 return f2fs_ioc_getflags(filp, arg);
3158 case F2FS_IOC_SETFLAGS:
3159 return f2fs_ioc_setflags(filp, arg);
3160 case F2FS_IOC_GETVERSION:
3161 return f2fs_ioc_getversion(filp, arg);
3162 case F2FS_IOC_START_ATOMIC_WRITE:
3163 return f2fs_ioc_start_atomic_write(filp);
3164 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3165 return f2fs_ioc_commit_atomic_write(filp);
3166 case F2FS_IOC_START_VOLATILE_WRITE:
3167 return f2fs_ioc_start_volatile_write(filp);
3168 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3169 return f2fs_ioc_release_volatile_write(filp);
3170 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3171 return f2fs_ioc_abort_volatile_write(filp);
3172 case F2FS_IOC_SHUTDOWN:
3173 return f2fs_ioc_shutdown(filp, arg);
3175 return f2fs_ioc_fitrim(filp, arg);
3176 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3177 return f2fs_ioc_set_encryption_policy(filp, arg);
3178 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3179 return f2fs_ioc_get_encryption_policy(filp, arg);
3180 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3181 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3182 case F2FS_IOC_GARBAGE_COLLECT:
3183 return f2fs_ioc_gc(filp, arg);
3184 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3185 return f2fs_ioc_gc_range(filp, arg);
3186 case F2FS_IOC_WRITE_CHECKPOINT:
3187 return f2fs_ioc_write_checkpoint(filp, arg);
3188 case F2FS_IOC_DEFRAGMENT:
3189 return f2fs_ioc_defragment(filp, arg);
3190 case F2FS_IOC_MOVE_RANGE:
3191 return f2fs_ioc_move_range(filp, arg);
3192 case F2FS_IOC_FLUSH_DEVICE:
3193 return f2fs_ioc_flush_device(filp, arg);
3194 case F2FS_IOC_GET_FEATURES:
3195 return f2fs_ioc_get_features(filp, arg);
3196 case F2FS_IOC_FSGETXATTR:
3197 return f2fs_ioc_fsgetxattr(filp, arg);
3198 case F2FS_IOC_FSSETXATTR:
3199 return f2fs_ioc_fssetxattr(filp, arg);
3200 case F2FS_IOC_GET_PIN_FILE:
3201 return f2fs_ioc_get_pin_file(filp, arg);
3202 case F2FS_IOC_SET_PIN_FILE:
3203 return f2fs_ioc_set_pin_file(filp, arg);
3204 case F2FS_IOC_PRECACHE_EXTENTS:
3205 return f2fs_ioc_precache_extents(filp, arg);
3206 case F2FS_IOC_RESIZE_FS:
3207 return f2fs_ioc_resize_fs(filp, arg);
3208 case F2FS_IOC_GET_VOLUME_NAME:
3209 return f2fs_get_volume_name(filp, arg);
3210 case F2FS_IOC_SET_VOLUME_NAME:
3211 return f2fs_set_volume_name(filp, arg);
3217 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3219 struct file *file = iocb->ki_filp;
3220 struct inode *inode = file_inode(file);
3223 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3228 if (!inode_trylock(inode)) {
3229 if (iocb->ki_flags & IOCB_NOWAIT) {
3236 ret = generic_write_checks(iocb, from);
3238 bool preallocated = false;
3239 size_t target_size = 0;
3242 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3243 set_inode_flag(inode, FI_NO_PREALLOC);
3245 if ((iocb->ki_flags & IOCB_NOWAIT)) {
3246 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3247 iov_iter_count(from)) ||
3248 f2fs_has_inline_data(inode) ||
3249 f2fs_force_buffered_io(inode, iocb, from)) {
3250 clear_inode_flag(inode, FI_NO_PREALLOC);
3251 inode_unlock(inode);
3256 preallocated = true;
3257 target_size = iocb->ki_pos + iov_iter_count(from);
3259 err = f2fs_preallocate_blocks(iocb, from);
3261 clear_inode_flag(inode, FI_NO_PREALLOC);
3262 inode_unlock(inode);
3267 ret = __generic_file_write_iter(iocb, from);
3268 clear_inode_flag(inode, FI_NO_PREALLOC);
3270 /* if we couldn't write data, we should deallocate blocks. */
3271 if (preallocated && i_size_read(inode) < target_size)
3272 f2fs_truncate(inode);
3275 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3277 inode_unlock(inode);
3279 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3280 iov_iter_count(from), ret);
3282 ret = generic_write_sync(iocb, ret);
3286 #ifdef CONFIG_COMPAT
3287 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3290 case F2FS_IOC32_GETFLAGS:
3291 cmd = F2FS_IOC_GETFLAGS;
3293 case F2FS_IOC32_SETFLAGS:
3294 cmd = F2FS_IOC_SETFLAGS;
3296 case F2FS_IOC32_GETVERSION:
3297 cmd = F2FS_IOC_GETVERSION;
3299 case F2FS_IOC_START_ATOMIC_WRITE:
3300 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3301 case F2FS_IOC_START_VOLATILE_WRITE:
3302 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3303 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3304 case F2FS_IOC_SHUTDOWN:
3305 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3306 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3307 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3308 case F2FS_IOC_GARBAGE_COLLECT:
3309 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3310 case F2FS_IOC_WRITE_CHECKPOINT:
3311 case F2FS_IOC_DEFRAGMENT:
3312 case F2FS_IOC_MOVE_RANGE:
3313 case F2FS_IOC_FLUSH_DEVICE:
3314 case F2FS_IOC_GET_FEATURES:
3315 case F2FS_IOC_FSGETXATTR:
3316 case F2FS_IOC_FSSETXATTR:
3317 case F2FS_IOC_GET_PIN_FILE:
3318 case F2FS_IOC_SET_PIN_FILE:
3319 case F2FS_IOC_PRECACHE_EXTENTS:
3320 case F2FS_IOC_RESIZE_FS:
3321 case F2FS_IOC_GET_VOLUME_NAME:
3322 case F2FS_IOC_SET_VOLUME_NAME:
3325 return -ENOIOCTLCMD;
3327 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3331 const struct file_operations f2fs_file_operations = {
3332 .llseek = f2fs_llseek,
3333 .read_iter = generic_file_read_iter,
3334 .write_iter = f2fs_file_write_iter,
3335 .open = f2fs_file_open,
3336 .release = f2fs_release_file,
3337 .mmap = f2fs_file_mmap,
3338 .flush = f2fs_file_flush,
3339 .fsync = f2fs_sync_file,
3340 .fallocate = f2fs_fallocate,
3341 .unlocked_ioctl = f2fs_ioctl,
3342 #ifdef CONFIG_COMPAT
3343 .compat_ioctl = f2fs_compat_ioctl,
3345 .splice_read = generic_file_splice_read,
3346 .splice_write = iter_file_splice_write,