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>
31 #include <trace/events/f2fs.h>
33 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
35 struct inode *inode = file_inode(vmf->vma->vm_file);
38 down_read(&F2FS_I(inode)->i_mmap_sem);
39 ret = filemap_fault(vmf);
40 up_read(&F2FS_I(inode)->i_mmap_sem);
42 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
47 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 struct page *page = vmf->page;
50 struct inode *inode = file_inode(vmf->vma->vm_file);
51 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
52 struct dnode_of_data dn = { .node_changed = false };
55 if (unlikely(f2fs_cp_error(sbi))) {
60 sb_start_pagefault(inode->i_sb);
62 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
64 file_update_time(vmf->vma->vm_file);
65 down_read(&F2FS_I(inode)->i_mmap_sem);
67 if (unlikely(page->mapping != inode->i_mapping ||
68 page_offset(page) > i_size_read(inode) ||
69 !PageUptodate(page))) {
75 /* block allocation */
76 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
77 set_new_dnode(&dn, inode, NULL, NULL, 0);
78 err = f2fs_get_block(&dn, page->index);
80 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
87 f2fs_wait_on_page_writeback(page, DATA, false, true);
89 /* wait for GCed page writeback via META_MAPPING */
90 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
93 * check to see if the page is mapped already (no holes)
95 if (PageMappedToDisk(page))
98 /* page is wholly or partially inside EOF */
99 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
100 i_size_read(inode)) {
103 offset = i_size_read(inode) & ~PAGE_MASK;
104 zero_user_segment(page, offset, PAGE_SIZE);
106 set_page_dirty(page);
107 if (!PageUptodate(page))
108 SetPageUptodate(page);
110 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
111 f2fs_update_time(sbi, REQ_TIME);
113 trace_f2fs_vm_page_mkwrite(page, DATA);
115 up_read(&F2FS_I(inode)->i_mmap_sem);
117 f2fs_balance_fs(sbi, dn.node_changed);
119 sb_end_pagefault(inode->i_sb);
121 return block_page_mkwrite_return(err);
124 static const struct vm_operations_struct f2fs_file_vm_ops = {
125 .fault = f2fs_filemap_fault,
126 .map_pages = filemap_map_pages,
127 .page_mkwrite = f2fs_vm_page_mkwrite,
130 static int get_parent_ino(struct inode *inode, nid_t *pino)
132 struct dentry *dentry;
134 inode = igrab(inode);
135 dentry = d_find_any_alias(inode);
140 *pino = parent_ino(dentry);
145 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
147 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
148 enum cp_reason_type cp_reason = CP_NO_NEEDED;
150 if (!S_ISREG(inode->i_mode))
151 cp_reason = CP_NON_REGULAR;
152 else if (inode->i_nlink != 1)
153 cp_reason = CP_HARDLINK;
154 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
155 cp_reason = CP_SB_NEED_CP;
156 else if (file_wrong_pino(inode))
157 cp_reason = CP_WRONG_PINO;
158 else if (!f2fs_space_for_roll_forward(sbi))
159 cp_reason = CP_NO_SPC_ROLL;
160 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
161 cp_reason = CP_NODE_NEED_CP;
162 else if (test_opt(sbi, FASTBOOT))
163 cp_reason = CP_FASTBOOT_MODE;
164 else if (F2FS_OPTION(sbi).active_logs == 2)
165 cp_reason = CP_SPEC_LOG_NUM;
166 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
167 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
168 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
170 cp_reason = CP_RECOVER_DIR;
175 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
177 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
179 /* But we need to avoid that there are some inode updates */
180 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
186 static void try_to_fix_pino(struct inode *inode)
188 struct f2fs_inode_info *fi = F2FS_I(inode);
191 down_write(&fi->i_sem);
192 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
193 get_parent_ino(inode, &pino)) {
194 f2fs_i_pino_write(inode, pino);
195 file_got_pino(inode);
197 up_write(&fi->i_sem);
200 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
201 int datasync, bool atomic)
203 struct inode *inode = file->f_mapping->host;
204 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
205 nid_t ino = inode->i_ino;
207 enum cp_reason_type cp_reason = 0;
208 struct writeback_control wbc = {
209 .sync_mode = WB_SYNC_ALL,
210 .nr_to_write = LONG_MAX,
213 unsigned int seq_id = 0;
215 if (unlikely(f2fs_readonly(inode->i_sb) ||
216 is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
219 trace_f2fs_sync_file_enter(inode);
221 if (S_ISDIR(inode->i_mode))
224 /* if fdatasync is triggered, let's do in-place-update */
225 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
226 set_inode_flag(inode, FI_NEED_IPU);
227 ret = file_write_and_wait_range(file, start, end);
228 clear_inode_flag(inode, FI_NEED_IPU);
231 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
235 /* if the inode is dirty, let's recover all the time */
236 if (!f2fs_skip_inode_update(inode, datasync)) {
237 f2fs_write_inode(inode, NULL);
242 * if there is no written data, don't waste time to write recovery info.
244 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
245 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
247 /* it may call write_inode just prior to fsync */
248 if (need_inode_page_update(sbi, ino))
251 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
252 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
258 * Both of fdatasync() and fsync() are able to be recovered from
261 down_read(&F2FS_I(inode)->i_sem);
262 cp_reason = need_do_checkpoint(inode);
263 up_read(&F2FS_I(inode)->i_sem);
266 /* all the dirty node pages should be flushed for POR */
267 ret = f2fs_sync_fs(inode->i_sb, 1);
270 * We've secured consistency through sync_fs. Following pino
271 * will be used only for fsynced inodes after checkpoint.
273 try_to_fix_pino(inode);
274 clear_inode_flag(inode, FI_APPEND_WRITE);
275 clear_inode_flag(inode, FI_UPDATE_WRITE);
279 atomic_inc(&sbi->wb_sync_req[NODE]);
280 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
281 atomic_dec(&sbi->wb_sync_req[NODE]);
285 /* if cp_error was enabled, we should avoid infinite loop */
286 if (unlikely(f2fs_cp_error(sbi))) {
291 if (f2fs_need_inode_block_update(sbi, ino)) {
292 f2fs_mark_inode_dirty_sync(inode, true);
293 f2fs_write_inode(inode, NULL);
298 * If it's atomic_write, it's just fine to keep write ordering. So
299 * here we don't need to wait for node write completion, since we use
300 * node chain which serializes node blocks. If one of node writes are
301 * reordered, we can see simply broken chain, resulting in stopping
302 * roll-forward recovery. It means we'll recover all or none node blocks
306 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
311 /* once recovery info is written, don't need to tack this */
312 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
313 clear_inode_flag(inode, FI_APPEND_WRITE);
315 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
316 ret = f2fs_issue_flush(sbi, inode->i_ino);
318 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
319 clear_inode_flag(inode, FI_UPDATE_WRITE);
320 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
322 f2fs_update_time(sbi, REQ_TIME);
324 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
325 f2fs_trace_ios(NULL, 1);
329 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
331 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
333 return f2fs_do_sync_file(file, start, end, datasync, false);
336 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
337 pgoff_t pgofs, int whence)
342 if (whence != SEEK_DATA)
345 /* find first dirty page index */
346 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
355 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
356 pgoff_t dirty, pgoff_t pgofs, int whence)
360 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
361 __is_valid_data_blkaddr(blkaddr))
365 if (blkaddr == NULL_ADDR)
372 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
374 struct inode *inode = file->f_mapping->host;
375 loff_t maxbytes = inode->i_sb->s_maxbytes;
376 struct dnode_of_data dn;
377 pgoff_t pgofs, end_offset, dirty;
378 loff_t data_ofs = offset;
384 isize = i_size_read(inode);
388 /* handle inline data case */
389 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
390 if (whence == SEEK_HOLE)
395 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
397 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
399 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
400 set_new_dnode(&dn, inode, NULL, NULL, 0);
401 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
402 if (err && err != -ENOENT) {
404 } else if (err == -ENOENT) {
405 /* direct node does not exists */
406 if (whence == SEEK_DATA) {
407 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
414 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
416 /* find data/hole in dnode block */
417 for (; dn.ofs_in_node < end_offset;
418 dn.ofs_in_node++, pgofs++,
419 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
422 blkaddr = datablock_addr(dn.inode,
423 dn.node_page, dn.ofs_in_node);
425 if (__is_valid_data_blkaddr(blkaddr) &&
426 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
427 blkaddr, DATA_GENERIC_ENHANCE)) {
432 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
441 if (whence == SEEK_DATA)
444 if (whence == SEEK_HOLE && data_ofs > isize)
447 return vfs_setpos(file, data_ofs, maxbytes);
453 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
455 struct inode *inode = file->f_mapping->host;
456 loff_t maxbytes = inode->i_sb->s_maxbytes;
462 return generic_file_llseek_size(file, offset, whence,
463 maxbytes, i_size_read(inode));
468 return f2fs_seek_block(file, offset, whence);
474 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
476 struct inode *inode = file_inode(file);
479 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
482 /* we don't need to use inline_data strictly */
483 err = f2fs_convert_inline_inode(inode);
488 vma->vm_ops = &f2fs_file_vm_ops;
492 static int f2fs_file_open(struct inode *inode, struct file *filp)
494 int err = fscrypt_file_open(inode, filp);
499 err = fsverity_file_open(inode, filp);
503 filp->f_mode |= FMODE_NOWAIT;
505 return dquot_file_open(inode, filp);
508 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
510 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
511 struct f2fs_node *raw_node;
512 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
516 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
517 base = get_extra_isize(dn->inode);
519 raw_node = F2FS_NODE(dn->node_page);
520 addr = blkaddr_in_node(raw_node) + base + ofs;
522 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
523 block_t blkaddr = le32_to_cpu(*addr);
525 if (blkaddr == NULL_ADDR)
528 dn->data_blkaddr = NULL_ADDR;
529 f2fs_set_data_blkaddr(dn);
531 if (__is_valid_data_blkaddr(blkaddr) &&
532 !f2fs_is_valid_blkaddr(sbi, blkaddr,
533 DATA_GENERIC_ENHANCE))
536 f2fs_invalidate_blocks(sbi, blkaddr);
537 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
538 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
545 * once we invalidate valid blkaddr in range [ofs, ofs + count],
546 * we will invalidate all blkaddr in the whole range.
548 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
550 f2fs_update_extent_cache_range(dn, fofs, 0, len);
551 dec_valid_block_count(sbi, dn->inode, nr_free);
553 dn->ofs_in_node = ofs;
555 f2fs_update_time(sbi, REQ_TIME);
556 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
557 dn->ofs_in_node, nr_free);
560 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
562 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
565 static int truncate_partial_data_page(struct inode *inode, u64 from,
568 loff_t offset = from & (PAGE_SIZE - 1);
569 pgoff_t index = from >> PAGE_SHIFT;
570 struct address_space *mapping = inode->i_mapping;
573 if (!offset && !cache_only)
577 page = find_lock_page(mapping, index);
578 if (page && PageUptodate(page))
580 f2fs_put_page(page, 1);
584 page = f2fs_get_lock_data_page(inode, index, true);
586 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
588 f2fs_wait_on_page_writeback(page, DATA, true, true);
589 zero_user(page, offset, PAGE_SIZE - offset);
591 /* An encrypted inode should have a key and truncate the last page. */
592 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
594 set_page_dirty(page);
595 f2fs_put_page(page, 1);
599 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
601 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
602 struct dnode_of_data dn;
604 int count = 0, err = 0;
606 bool truncate_page = false;
608 trace_f2fs_truncate_blocks_enter(inode, from);
610 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
612 if (free_from >= sbi->max_file_blocks)
618 ipage = f2fs_get_node_page(sbi, inode->i_ino);
620 err = PTR_ERR(ipage);
624 if (f2fs_has_inline_data(inode)) {
625 f2fs_truncate_inline_inode(inode, ipage, from);
626 f2fs_put_page(ipage, 1);
627 truncate_page = true;
631 set_new_dnode(&dn, inode, ipage, NULL, 0);
632 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
639 count = ADDRS_PER_PAGE(dn.node_page, inode);
641 count -= dn.ofs_in_node;
642 f2fs_bug_on(sbi, count < 0);
644 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
645 f2fs_truncate_data_blocks_range(&dn, count);
651 err = f2fs_truncate_inode_blocks(inode, free_from);
656 /* lastly zero out the first data page */
658 err = truncate_partial_data_page(inode, from, truncate_page);
660 trace_f2fs_truncate_blocks_exit(inode, err);
664 int f2fs_truncate(struct inode *inode)
668 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
671 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
672 S_ISLNK(inode->i_mode)))
675 trace_f2fs_truncate(inode);
677 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
678 f2fs_show_injection_info(FAULT_TRUNCATE);
682 /* we should check inline_data size */
683 if (!f2fs_may_inline_data(inode)) {
684 err = f2fs_convert_inline_inode(inode);
689 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
693 inode->i_mtime = inode->i_ctime = current_time(inode);
694 f2fs_mark_inode_dirty_sync(inode, false);
698 int f2fs_getattr(const struct path *path, struct kstat *stat,
699 u32 request_mask, unsigned int query_flags)
701 struct inode *inode = d_inode(path->dentry);
702 struct f2fs_inode_info *fi = F2FS_I(inode);
703 struct f2fs_inode *ri;
706 if (f2fs_has_extra_attr(inode) &&
707 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
708 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
709 stat->result_mask |= STATX_BTIME;
710 stat->btime.tv_sec = fi->i_crtime.tv_sec;
711 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
715 if (flags & F2FS_APPEND_FL)
716 stat->attributes |= STATX_ATTR_APPEND;
717 if (IS_ENCRYPTED(inode))
718 stat->attributes |= STATX_ATTR_ENCRYPTED;
719 if (flags & F2FS_IMMUTABLE_FL)
720 stat->attributes |= STATX_ATTR_IMMUTABLE;
721 if (flags & F2FS_NODUMP_FL)
722 stat->attributes |= STATX_ATTR_NODUMP;
724 stat->attributes_mask |= (STATX_ATTR_APPEND |
725 STATX_ATTR_ENCRYPTED |
726 STATX_ATTR_IMMUTABLE |
729 generic_fillattr(inode, stat);
731 /* we need to show initial sectors used for inline_data/dentries */
732 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
733 f2fs_has_inline_dentry(inode))
734 stat->blocks += (stat->size + 511) >> 9;
739 #ifdef CONFIG_F2FS_FS_POSIX_ACL
740 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
742 unsigned int ia_valid = attr->ia_valid;
744 if (ia_valid & ATTR_UID)
745 inode->i_uid = attr->ia_uid;
746 if (ia_valid & ATTR_GID)
747 inode->i_gid = attr->ia_gid;
748 if (ia_valid & ATTR_ATIME) {
749 inode->i_atime = timestamp_truncate(attr->ia_atime,
752 if (ia_valid & ATTR_MTIME) {
753 inode->i_mtime = timestamp_truncate(attr->ia_mtime,
756 if (ia_valid & ATTR_CTIME) {
757 inode->i_ctime = timestamp_truncate(attr->ia_ctime,
760 if (ia_valid & ATTR_MODE) {
761 umode_t mode = attr->ia_mode;
763 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
765 set_acl_inode(inode, mode);
769 #define __setattr_copy setattr_copy
772 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
774 struct inode *inode = d_inode(dentry);
777 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
780 err = setattr_prepare(dentry, attr);
784 err = fscrypt_prepare_setattr(dentry, attr);
788 err = fsverity_prepare_setattr(dentry, attr);
792 if (is_quota_modification(inode, attr)) {
793 err = dquot_initialize(inode);
797 if ((attr->ia_valid & ATTR_UID &&
798 !uid_eq(attr->ia_uid, inode->i_uid)) ||
799 (attr->ia_valid & ATTR_GID &&
800 !gid_eq(attr->ia_gid, inode->i_gid))) {
801 f2fs_lock_op(F2FS_I_SB(inode));
802 err = dquot_transfer(inode, attr);
804 set_sbi_flag(F2FS_I_SB(inode),
805 SBI_QUOTA_NEED_REPAIR);
806 f2fs_unlock_op(F2FS_I_SB(inode));
810 * update uid/gid under lock_op(), so that dquot and inode can
811 * be updated atomically.
813 if (attr->ia_valid & ATTR_UID)
814 inode->i_uid = attr->ia_uid;
815 if (attr->ia_valid & ATTR_GID)
816 inode->i_gid = attr->ia_gid;
817 f2fs_mark_inode_dirty_sync(inode, true);
818 f2fs_unlock_op(F2FS_I_SB(inode));
821 if (attr->ia_valid & ATTR_SIZE) {
822 bool to_smaller = (attr->ia_size <= i_size_read(inode));
824 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
825 down_write(&F2FS_I(inode)->i_mmap_sem);
827 truncate_setsize(inode, attr->ia_size);
830 err = f2fs_truncate(inode);
832 * do not trim all blocks after i_size if target size is
833 * larger than i_size.
835 up_write(&F2FS_I(inode)->i_mmap_sem);
836 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
842 /* should convert inline inode here */
843 if (!f2fs_may_inline_data(inode)) {
844 err = f2fs_convert_inline_inode(inode);
848 inode->i_mtime = inode->i_ctime = current_time(inode);
851 down_write(&F2FS_I(inode)->i_sem);
852 F2FS_I(inode)->last_disk_size = i_size_read(inode);
853 up_write(&F2FS_I(inode)->i_sem);
856 __setattr_copy(inode, attr);
858 if (attr->ia_valid & ATTR_MODE) {
859 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
860 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
861 inode->i_mode = F2FS_I(inode)->i_acl_mode;
862 clear_inode_flag(inode, FI_ACL_MODE);
866 /* file size may changed here */
867 f2fs_mark_inode_dirty_sync(inode, true);
869 /* inode change will produce dirty node pages flushed by checkpoint */
870 f2fs_balance_fs(F2FS_I_SB(inode), true);
875 const struct inode_operations f2fs_file_inode_operations = {
876 .getattr = f2fs_getattr,
877 .setattr = f2fs_setattr,
878 .get_acl = f2fs_get_acl,
879 .set_acl = f2fs_set_acl,
880 #ifdef CONFIG_F2FS_FS_XATTR
881 .listxattr = f2fs_listxattr,
883 .fiemap = f2fs_fiemap,
886 static int fill_zero(struct inode *inode, pgoff_t index,
887 loff_t start, loff_t len)
889 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
895 f2fs_balance_fs(sbi, true);
898 page = f2fs_get_new_data_page(inode, NULL, index, false);
902 return PTR_ERR(page);
904 f2fs_wait_on_page_writeback(page, DATA, true, true);
905 zero_user(page, start, len);
906 set_page_dirty(page);
907 f2fs_put_page(page, 1);
911 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
915 while (pg_start < pg_end) {
916 struct dnode_of_data dn;
917 pgoff_t end_offset, count;
919 set_new_dnode(&dn, inode, NULL, NULL, 0);
920 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
922 if (err == -ENOENT) {
923 pg_start = f2fs_get_next_page_offset(&dn,
930 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
931 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
933 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
935 f2fs_truncate_data_blocks_range(&dn, count);
943 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
945 pgoff_t pg_start, pg_end;
946 loff_t off_start, off_end;
949 ret = f2fs_convert_inline_inode(inode);
953 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
954 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
956 off_start = offset & (PAGE_SIZE - 1);
957 off_end = (offset + len) & (PAGE_SIZE - 1);
959 if (pg_start == pg_end) {
960 ret = fill_zero(inode, pg_start, off_start,
961 off_end - off_start);
966 ret = fill_zero(inode, pg_start++, off_start,
967 PAGE_SIZE - off_start);
972 ret = fill_zero(inode, pg_end, 0, off_end);
977 if (pg_start < pg_end) {
978 struct address_space *mapping = inode->i_mapping;
979 loff_t blk_start, blk_end;
980 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
982 f2fs_balance_fs(sbi, true);
984 blk_start = (loff_t)pg_start << PAGE_SHIFT;
985 blk_end = (loff_t)pg_end << PAGE_SHIFT;
987 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
988 down_write(&F2FS_I(inode)->i_mmap_sem);
990 truncate_inode_pages_range(mapping, blk_start,
994 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
997 up_write(&F2FS_I(inode)->i_mmap_sem);
998 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1005 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1006 int *do_replace, pgoff_t off, pgoff_t len)
1008 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1009 struct dnode_of_data dn;
1013 set_new_dnode(&dn, inode, NULL, NULL, 0);
1014 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1015 if (ret && ret != -ENOENT) {
1017 } else if (ret == -ENOENT) {
1018 if (dn.max_level == 0)
1020 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1027 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1028 dn.ofs_in_node, len);
1029 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1030 *blkaddr = datablock_addr(dn.inode,
1031 dn.node_page, dn.ofs_in_node);
1033 if (__is_valid_data_blkaddr(*blkaddr) &&
1034 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1035 DATA_GENERIC_ENHANCE)) {
1036 f2fs_put_dnode(&dn);
1037 return -EFSCORRUPTED;
1040 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1042 if (test_opt(sbi, LFS)) {
1043 f2fs_put_dnode(&dn);
1047 /* do not invalidate this block address */
1048 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1052 f2fs_put_dnode(&dn);
1061 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1062 int *do_replace, pgoff_t off, int len)
1064 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1065 struct dnode_of_data dn;
1068 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1069 if (*do_replace == 0)
1072 set_new_dnode(&dn, inode, NULL, NULL, 0);
1073 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1075 dec_valid_block_count(sbi, inode, 1);
1076 f2fs_invalidate_blocks(sbi, *blkaddr);
1078 f2fs_update_data_blkaddr(&dn, *blkaddr);
1080 f2fs_put_dnode(&dn);
1085 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1086 block_t *blkaddr, int *do_replace,
1087 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1089 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1094 if (blkaddr[i] == NULL_ADDR && !full) {
1099 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1100 struct dnode_of_data dn;
1101 struct node_info ni;
1105 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1106 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1110 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1112 f2fs_put_dnode(&dn);
1116 ilen = min((pgoff_t)
1117 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1118 dn.ofs_in_node, len - i);
1120 dn.data_blkaddr = datablock_addr(dn.inode,
1121 dn.node_page, dn.ofs_in_node);
1122 f2fs_truncate_data_blocks_range(&dn, 1);
1124 if (do_replace[i]) {
1125 f2fs_i_blocks_write(src_inode,
1127 f2fs_i_blocks_write(dst_inode,
1129 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1130 blkaddr[i], ni.version, true, false);
1136 new_size = (dst + i) << PAGE_SHIFT;
1137 if (dst_inode->i_size < new_size)
1138 f2fs_i_size_write(dst_inode, new_size);
1139 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1141 f2fs_put_dnode(&dn);
1143 struct page *psrc, *pdst;
1145 psrc = f2fs_get_lock_data_page(src_inode,
1148 return PTR_ERR(psrc);
1149 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1152 f2fs_put_page(psrc, 1);
1153 return PTR_ERR(pdst);
1155 f2fs_copy_page(psrc, pdst);
1156 set_page_dirty(pdst);
1157 f2fs_put_page(pdst, 1);
1158 f2fs_put_page(psrc, 1);
1160 ret = f2fs_truncate_hole(src_inode,
1161 src + i, src + i + 1);
1170 static int __exchange_data_block(struct inode *src_inode,
1171 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1172 pgoff_t len, bool full)
1174 block_t *src_blkaddr;
1180 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1182 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1183 array_size(olen, sizeof(block_t)),
1188 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1189 array_size(olen, sizeof(int)),
1192 kvfree(src_blkaddr);
1196 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1197 do_replace, src, olen);
1201 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1202 do_replace, src, dst, olen, full);
1210 kvfree(src_blkaddr);
1216 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1217 kvfree(src_blkaddr);
1222 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1224 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1225 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1226 pgoff_t start = offset >> PAGE_SHIFT;
1227 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1230 f2fs_balance_fs(sbi, true);
1232 /* avoid gc operation during block exchange */
1233 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1234 down_write(&F2FS_I(inode)->i_mmap_sem);
1237 f2fs_drop_extent_tree(inode);
1238 truncate_pagecache(inode, offset);
1239 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1240 f2fs_unlock_op(sbi);
1242 up_write(&F2FS_I(inode)->i_mmap_sem);
1243 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1247 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1252 if (offset + len >= i_size_read(inode))
1255 /* collapse range should be aligned to block size of f2fs. */
1256 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1259 ret = f2fs_convert_inline_inode(inode);
1263 /* write out all dirty pages from offset */
1264 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1268 ret = f2fs_do_collapse(inode, offset, len);
1272 /* write out all moved pages, if possible */
1273 down_write(&F2FS_I(inode)->i_mmap_sem);
1274 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1275 truncate_pagecache(inode, offset);
1277 new_size = i_size_read(inode) - len;
1278 truncate_pagecache(inode, new_size);
1280 ret = f2fs_truncate_blocks(inode, new_size, true);
1281 up_write(&F2FS_I(inode)->i_mmap_sem);
1283 f2fs_i_size_write(inode, new_size);
1287 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1290 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1291 pgoff_t index = start;
1292 unsigned int ofs_in_node = dn->ofs_in_node;
1296 for (; index < end; index++, dn->ofs_in_node++) {
1297 if (datablock_addr(dn->inode, dn->node_page,
1298 dn->ofs_in_node) == NULL_ADDR)
1302 dn->ofs_in_node = ofs_in_node;
1303 ret = f2fs_reserve_new_blocks(dn, count);
1307 dn->ofs_in_node = ofs_in_node;
1308 for (index = start; index < end; index++, dn->ofs_in_node++) {
1309 dn->data_blkaddr = datablock_addr(dn->inode,
1310 dn->node_page, dn->ofs_in_node);
1312 * f2fs_reserve_new_blocks will not guarantee entire block
1315 if (dn->data_blkaddr == NULL_ADDR) {
1319 if (dn->data_blkaddr != NEW_ADDR) {
1320 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1321 dn->data_blkaddr = NEW_ADDR;
1322 f2fs_set_data_blkaddr(dn);
1326 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1331 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1334 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1335 struct address_space *mapping = inode->i_mapping;
1336 pgoff_t index, pg_start, pg_end;
1337 loff_t new_size = i_size_read(inode);
1338 loff_t off_start, off_end;
1341 ret = inode_newsize_ok(inode, (len + offset));
1345 ret = f2fs_convert_inline_inode(inode);
1349 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1353 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1354 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1356 off_start = offset & (PAGE_SIZE - 1);
1357 off_end = (offset + len) & (PAGE_SIZE - 1);
1359 if (pg_start == pg_end) {
1360 ret = fill_zero(inode, pg_start, off_start,
1361 off_end - off_start);
1365 new_size = max_t(loff_t, new_size, offset + len);
1368 ret = fill_zero(inode, pg_start++, off_start,
1369 PAGE_SIZE - off_start);
1373 new_size = max_t(loff_t, new_size,
1374 (loff_t)pg_start << PAGE_SHIFT);
1377 for (index = pg_start; index < pg_end;) {
1378 struct dnode_of_data dn;
1379 unsigned int end_offset;
1382 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1383 down_write(&F2FS_I(inode)->i_mmap_sem);
1385 truncate_pagecache_range(inode,
1386 (loff_t)index << PAGE_SHIFT,
1387 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1391 set_new_dnode(&dn, inode, NULL, NULL, 0);
1392 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1394 f2fs_unlock_op(sbi);
1395 up_write(&F2FS_I(inode)->i_mmap_sem);
1396 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1400 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1401 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1403 ret = f2fs_do_zero_range(&dn, index, end);
1404 f2fs_put_dnode(&dn);
1406 f2fs_unlock_op(sbi);
1407 up_write(&F2FS_I(inode)->i_mmap_sem);
1408 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1410 f2fs_balance_fs(sbi, dn.node_changed);
1416 new_size = max_t(loff_t, new_size,
1417 (loff_t)index << PAGE_SHIFT);
1421 ret = fill_zero(inode, pg_end, 0, off_end);
1425 new_size = max_t(loff_t, new_size, offset + len);
1430 if (new_size > i_size_read(inode)) {
1431 if (mode & FALLOC_FL_KEEP_SIZE)
1432 file_set_keep_isize(inode);
1434 f2fs_i_size_write(inode, new_size);
1439 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1441 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1442 pgoff_t nr, pg_start, pg_end, delta, idx;
1446 new_size = i_size_read(inode) + len;
1447 ret = inode_newsize_ok(inode, new_size);
1451 if (offset >= i_size_read(inode))
1454 /* insert range should be aligned to block size of f2fs. */
1455 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1458 ret = f2fs_convert_inline_inode(inode);
1462 f2fs_balance_fs(sbi, true);
1464 down_write(&F2FS_I(inode)->i_mmap_sem);
1465 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1466 up_write(&F2FS_I(inode)->i_mmap_sem);
1470 /* write out all dirty pages from offset */
1471 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1475 pg_start = offset >> PAGE_SHIFT;
1476 pg_end = (offset + len) >> PAGE_SHIFT;
1477 delta = pg_end - pg_start;
1478 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1480 /* avoid gc operation during block exchange */
1481 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1482 down_write(&F2FS_I(inode)->i_mmap_sem);
1483 truncate_pagecache(inode, offset);
1485 while (!ret && idx > pg_start) {
1486 nr = idx - pg_start;
1492 f2fs_drop_extent_tree(inode);
1494 ret = __exchange_data_block(inode, inode, idx,
1495 idx + delta, nr, false);
1496 f2fs_unlock_op(sbi);
1498 up_write(&F2FS_I(inode)->i_mmap_sem);
1499 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1501 /* write out all moved pages, if possible */
1502 down_write(&F2FS_I(inode)->i_mmap_sem);
1503 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1504 truncate_pagecache(inode, offset);
1505 up_write(&F2FS_I(inode)->i_mmap_sem);
1508 f2fs_i_size_write(inode, new_size);
1512 static int expand_inode_data(struct inode *inode, loff_t offset,
1513 loff_t len, int mode)
1515 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1516 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1517 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1518 .m_may_create = true };
1520 loff_t new_size = i_size_read(inode);
1524 err = inode_newsize_ok(inode, (len + offset));
1528 err = f2fs_convert_inline_inode(inode);
1532 f2fs_balance_fs(sbi, true);
1534 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1535 off_end = (offset + len) & (PAGE_SIZE - 1);
1537 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1538 map.m_len = pg_end - map.m_lblk;
1542 if (f2fs_is_pinned_file(inode))
1543 map.m_seg_type = CURSEG_COLD_DATA;
1545 err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
1546 F2FS_GET_BLOCK_PRE_DIO :
1547 F2FS_GET_BLOCK_PRE_AIO));
1554 last_off = map.m_lblk + map.m_len - 1;
1556 /* update new size to the failed position */
1557 new_size = (last_off == pg_end) ? offset + len :
1558 (loff_t)(last_off + 1) << PAGE_SHIFT;
1560 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1563 if (new_size > i_size_read(inode)) {
1564 if (mode & FALLOC_FL_KEEP_SIZE)
1565 file_set_keep_isize(inode);
1567 f2fs_i_size_write(inode, new_size);
1573 static long f2fs_fallocate(struct file *file, int mode,
1574 loff_t offset, loff_t len)
1576 struct inode *inode = file_inode(file);
1579 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1582 /* f2fs only support ->fallocate for regular file */
1583 if (!S_ISREG(inode->i_mode))
1586 if (IS_ENCRYPTED(inode) &&
1587 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1590 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1591 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1592 FALLOC_FL_INSERT_RANGE))
1597 if (mode & FALLOC_FL_PUNCH_HOLE) {
1598 if (offset >= inode->i_size)
1601 ret = punch_hole(inode, offset, len);
1602 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1603 ret = f2fs_collapse_range(inode, offset, len);
1604 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1605 ret = f2fs_zero_range(inode, offset, len, mode);
1606 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1607 ret = f2fs_insert_range(inode, offset, len);
1609 ret = expand_inode_data(inode, offset, len, mode);
1613 inode->i_mtime = inode->i_ctime = current_time(inode);
1614 f2fs_mark_inode_dirty_sync(inode, false);
1615 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1619 inode_unlock(inode);
1621 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1625 static int f2fs_release_file(struct inode *inode, struct file *filp)
1628 * f2fs_relase_file is called at every close calls. So we should
1629 * not drop any inmemory pages by close called by other process.
1631 if (!(filp->f_mode & FMODE_WRITE) ||
1632 atomic_read(&inode->i_writecount) != 1)
1635 /* some remained atomic pages should discarded */
1636 if (f2fs_is_atomic_file(inode))
1637 f2fs_drop_inmem_pages(inode);
1638 if (f2fs_is_volatile_file(inode)) {
1639 set_inode_flag(inode, FI_DROP_CACHE);
1640 filemap_fdatawrite(inode->i_mapping);
1641 clear_inode_flag(inode, FI_DROP_CACHE);
1642 clear_inode_flag(inode, FI_VOLATILE_FILE);
1643 stat_dec_volatile_write(inode);
1648 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1650 struct inode *inode = file_inode(file);
1653 * If the process doing a transaction is crashed, we should do
1654 * roll-back. Otherwise, other reader/write can see corrupted database
1655 * until all the writers close its file. Since this should be done
1656 * before dropping file lock, it needs to do in ->flush.
1658 if (f2fs_is_atomic_file(inode) &&
1659 F2FS_I(inode)->inmem_task == current)
1660 f2fs_drop_inmem_pages(inode);
1664 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1666 struct f2fs_inode_info *fi = F2FS_I(inode);
1668 /* Is it quota file? Do not allow user to mess with it */
1669 if (IS_NOQUOTA(inode))
1672 fi->i_flags = iflags | (fi->i_flags & ~mask);
1674 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1675 set_inode_flag(inode, FI_PROJ_INHERIT);
1677 clear_inode_flag(inode, FI_PROJ_INHERIT);
1679 inode->i_ctime = current_time(inode);
1680 f2fs_set_inode_flags(inode);
1681 f2fs_mark_inode_dirty_sync(inode, true);
1685 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1688 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1689 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1690 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1691 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1694 static const struct {
1697 } f2fs_fsflags_map[] = {
1698 { F2FS_SYNC_FL, FS_SYNC_FL },
1699 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1700 { F2FS_APPEND_FL, FS_APPEND_FL },
1701 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1702 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1703 { F2FS_INDEX_FL, FS_INDEX_FL },
1704 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1705 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1708 #define F2FS_GETTABLE_FS_FL ( \
1716 FS_PROJINHERIT_FL | \
1718 FS_INLINE_DATA_FL | \
1722 #define F2FS_SETTABLE_FS_FL ( \
1731 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1732 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1737 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1738 if (iflags & f2fs_fsflags_map[i].iflag)
1739 fsflags |= f2fs_fsflags_map[i].fsflag;
1744 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1745 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1750 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1751 if (fsflags & f2fs_fsflags_map[i].fsflag)
1752 iflags |= f2fs_fsflags_map[i].iflag;
1757 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1759 struct inode *inode = file_inode(filp);
1760 struct f2fs_inode_info *fi = F2FS_I(inode);
1761 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1763 if (IS_ENCRYPTED(inode))
1764 fsflags |= FS_ENCRYPT_FL;
1765 if (IS_VERITY(inode))
1766 fsflags |= FS_VERITY_FL;
1767 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1768 fsflags |= FS_INLINE_DATA_FL;
1769 if (is_inode_flag_set(inode, FI_PIN_FILE))
1770 fsflags |= FS_NOCOW_FL;
1772 fsflags &= F2FS_GETTABLE_FS_FL;
1774 return put_user(fsflags, (int __user *)arg);
1777 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1779 struct inode *inode = file_inode(filp);
1780 struct f2fs_inode_info *fi = F2FS_I(inode);
1781 u32 fsflags, old_fsflags;
1785 if (!inode_owner_or_capable(inode))
1788 if (get_user(fsflags, (int __user *)arg))
1791 if (fsflags & ~F2FS_GETTABLE_FS_FL)
1793 fsflags &= F2FS_SETTABLE_FS_FL;
1795 iflags = f2fs_fsflags_to_iflags(fsflags);
1796 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1799 ret = mnt_want_write_file(filp);
1805 old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1806 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1810 ret = f2fs_setflags_common(inode, iflags,
1811 f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1813 inode_unlock(inode);
1814 mnt_drop_write_file(filp);
1818 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1820 struct inode *inode = file_inode(filp);
1822 return put_user(inode->i_generation, (int __user *)arg);
1825 static int f2fs_ioc_start_atomic_write(struct file *filp)
1827 struct inode *inode = file_inode(filp);
1830 if (!inode_owner_or_capable(inode))
1833 if (!S_ISREG(inode->i_mode))
1836 ret = mnt_want_write_file(filp);
1842 if (f2fs_is_atomic_file(inode)) {
1843 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1848 ret = f2fs_convert_inline_inode(inode);
1852 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1855 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1856 * f2fs_is_atomic_file.
1858 if (get_dirty_pages(inode))
1859 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1860 inode->i_ino, get_dirty_pages(inode));
1861 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1863 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1867 set_inode_flag(inode, FI_ATOMIC_FILE);
1868 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1869 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1871 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1872 F2FS_I(inode)->inmem_task = current;
1873 stat_inc_atomic_write(inode);
1874 stat_update_max_atomic_write(inode);
1876 inode_unlock(inode);
1877 mnt_drop_write_file(filp);
1881 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1883 struct inode *inode = file_inode(filp);
1886 if (!inode_owner_or_capable(inode))
1889 ret = mnt_want_write_file(filp);
1893 f2fs_balance_fs(F2FS_I_SB(inode), true);
1897 if (f2fs_is_volatile_file(inode)) {
1902 if (f2fs_is_atomic_file(inode)) {
1903 ret = f2fs_commit_inmem_pages(inode);
1907 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1909 clear_inode_flag(inode, FI_ATOMIC_FILE);
1910 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1911 stat_dec_atomic_write(inode);
1914 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1917 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1918 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1921 inode_unlock(inode);
1922 mnt_drop_write_file(filp);
1926 static int f2fs_ioc_start_volatile_write(struct file *filp)
1928 struct inode *inode = file_inode(filp);
1931 if (!inode_owner_or_capable(inode))
1934 if (!S_ISREG(inode->i_mode))
1937 ret = mnt_want_write_file(filp);
1943 if (f2fs_is_volatile_file(inode))
1946 ret = f2fs_convert_inline_inode(inode);
1950 stat_inc_volatile_write(inode);
1951 stat_update_max_volatile_write(inode);
1953 set_inode_flag(inode, FI_VOLATILE_FILE);
1954 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1956 inode_unlock(inode);
1957 mnt_drop_write_file(filp);
1961 static int f2fs_ioc_release_volatile_write(struct file *filp)
1963 struct inode *inode = file_inode(filp);
1966 if (!inode_owner_or_capable(inode))
1969 ret = mnt_want_write_file(filp);
1975 if (!f2fs_is_volatile_file(inode))
1978 if (!f2fs_is_first_block_written(inode)) {
1979 ret = truncate_partial_data_page(inode, 0, true);
1983 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1985 inode_unlock(inode);
1986 mnt_drop_write_file(filp);
1990 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1992 struct inode *inode = file_inode(filp);
1995 if (!inode_owner_or_capable(inode))
1998 ret = mnt_want_write_file(filp);
2004 if (f2fs_is_atomic_file(inode))
2005 f2fs_drop_inmem_pages(inode);
2006 if (f2fs_is_volatile_file(inode)) {
2007 clear_inode_flag(inode, FI_VOLATILE_FILE);
2008 stat_dec_volatile_write(inode);
2009 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2012 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2014 inode_unlock(inode);
2016 mnt_drop_write_file(filp);
2017 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2021 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2023 struct inode *inode = file_inode(filp);
2024 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2025 struct super_block *sb = sbi->sb;
2029 if (!capable(CAP_SYS_ADMIN))
2032 if (get_user(in, (__u32 __user *)arg))
2035 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2036 ret = mnt_want_write_file(filp);
2042 case F2FS_GOING_DOWN_FULLSYNC:
2043 sb = freeze_bdev(sb->s_bdev);
2049 f2fs_stop_checkpoint(sbi, false);
2050 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2051 thaw_bdev(sb->s_bdev, sb);
2054 case F2FS_GOING_DOWN_METASYNC:
2055 /* do checkpoint only */
2056 ret = f2fs_sync_fs(sb, 1);
2059 f2fs_stop_checkpoint(sbi, false);
2060 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2062 case F2FS_GOING_DOWN_NOSYNC:
2063 f2fs_stop_checkpoint(sbi, false);
2064 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2066 case F2FS_GOING_DOWN_METAFLUSH:
2067 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2068 f2fs_stop_checkpoint(sbi, false);
2069 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2071 case F2FS_GOING_DOWN_NEED_FSCK:
2072 set_sbi_flag(sbi, SBI_NEED_FSCK);
2073 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2074 set_sbi_flag(sbi, SBI_IS_DIRTY);
2075 /* do checkpoint only */
2076 ret = f2fs_sync_fs(sb, 1);
2083 f2fs_stop_gc_thread(sbi);
2084 f2fs_stop_discard_thread(sbi);
2086 f2fs_drop_discard_cmd(sbi);
2087 clear_opt(sbi, DISCARD);
2089 f2fs_update_time(sbi, REQ_TIME);
2091 if (in != F2FS_GOING_DOWN_FULLSYNC)
2092 mnt_drop_write_file(filp);
2094 trace_f2fs_shutdown(sbi, in, ret);
2099 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2101 struct inode *inode = file_inode(filp);
2102 struct super_block *sb = inode->i_sb;
2103 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2104 struct fstrim_range range;
2107 if (!capable(CAP_SYS_ADMIN))
2110 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2113 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2117 ret = mnt_want_write_file(filp);
2121 range.minlen = max((unsigned int)range.minlen,
2122 q->limits.discard_granularity);
2123 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2124 mnt_drop_write_file(filp);
2128 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2131 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2135 static bool uuid_is_nonzero(__u8 u[16])
2139 for (i = 0; i < 16; i++)
2145 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2147 struct inode *inode = file_inode(filp);
2149 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2152 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2154 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2157 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2159 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2161 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2164 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2166 struct inode *inode = file_inode(filp);
2167 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2170 if (!f2fs_sb_has_encrypt(sbi))
2173 err = mnt_want_write_file(filp);
2177 down_write(&sbi->sb_lock);
2179 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2182 /* update superblock with uuid */
2183 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2185 err = f2fs_commit_super(sbi, false);
2188 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2192 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2196 up_write(&sbi->sb_lock);
2197 mnt_drop_write_file(filp);
2201 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2204 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2207 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2210 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2212 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2215 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2218 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2220 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2223 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2226 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2229 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2232 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2235 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2238 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2241 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2244 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2246 struct inode *inode = file_inode(filp);
2247 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2251 if (!capable(CAP_SYS_ADMIN))
2254 if (get_user(sync, (__u32 __user *)arg))
2257 if (f2fs_readonly(sbi->sb))
2260 ret = mnt_want_write_file(filp);
2265 if (!mutex_trylock(&sbi->gc_mutex)) {
2270 mutex_lock(&sbi->gc_mutex);
2273 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2275 mnt_drop_write_file(filp);
2279 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2281 struct inode *inode = file_inode(filp);
2282 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2283 struct f2fs_gc_range range;
2287 if (!capable(CAP_SYS_ADMIN))
2290 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2294 if (f2fs_readonly(sbi->sb))
2297 end = range.start + range.len;
2298 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2302 ret = mnt_want_write_file(filp);
2308 if (!mutex_trylock(&sbi->gc_mutex)) {
2313 mutex_lock(&sbi->gc_mutex);
2316 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2317 range.start += BLKS_PER_SEC(sbi);
2318 if (range.start <= end)
2321 mnt_drop_write_file(filp);
2325 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2327 struct inode *inode = file_inode(filp);
2328 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2331 if (!capable(CAP_SYS_ADMIN))
2334 if (f2fs_readonly(sbi->sb))
2337 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2338 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2342 ret = mnt_want_write_file(filp);
2346 ret = f2fs_sync_fs(sbi->sb, 1);
2348 mnt_drop_write_file(filp);
2352 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2354 struct f2fs_defragment *range)
2356 struct inode *inode = file_inode(filp);
2357 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2358 .m_seg_type = NO_CHECK_TYPE ,
2359 .m_may_create = false };
2360 struct extent_info ei = {0, 0, 0};
2361 pgoff_t pg_start, pg_end, next_pgofs;
2362 unsigned int blk_per_seg = sbi->blocks_per_seg;
2363 unsigned int total = 0, sec_num;
2364 block_t blk_end = 0;
2365 bool fragmented = false;
2368 /* if in-place-update policy is enabled, don't waste time here */
2369 if (f2fs_should_update_inplace(inode, NULL))
2372 pg_start = range->start >> PAGE_SHIFT;
2373 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2375 f2fs_balance_fs(sbi, true);
2379 /* writeback all dirty pages in the range */
2380 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2381 range->start + range->len - 1);
2386 * lookup mapping info in extent cache, skip defragmenting if physical
2387 * block addresses are continuous.
2389 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2390 if (ei.fofs + ei.len >= pg_end)
2394 map.m_lblk = pg_start;
2395 map.m_next_pgofs = &next_pgofs;
2398 * lookup mapping info in dnode page cache, skip defragmenting if all
2399 * physical block addresses are continuous even if there are hole(s)
2400 * in logical blocks.
2402 while (map.m_lblk < pg_end) {
2403 map.m_len = pg_end - map.m_lblk;
2404 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2408 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2409 map.m_lblk = next_pgofs;
2413 if (blk_end && blk_end != map.m_pblk)
2416 /* record total count of block that we're going to move */
2419 blk_end = map.m_pblk + map.m_len;
2421 map.m_lblk += map.m_len;
2427 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2430 * make sure there are enough free section for LFS allocation, this can
2431 * avoid defragment running in SSR mode when free section are allocated
2434 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2439 map.m_lblk = pg_start;
2440 map.m_len = pg_end - pg_start;
2443 while (map.m_lblk < pg_end) {
2448 map.m_len = pg_end - map.m_lblk;
2449 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2453 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2454 map.m_lblk = next_pgofs;
2458 set_inode_flag(inode, FI_DO_DEFRAG);
2461 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2464 page = f2fs_get_lock_data_page(inode, idx, true);
2466 err = PTR_ERR(page);
2470 set_page_dirty(page);
2471 f2fs_put_page(page, 1);
2480 if (idx < pg_end && cnt < blk_per_seg)
2483 clear_inode_flag(inode, FI_DO_DEFRAG);
2485 err = filemap_fdatawrite(inode->i_mapping);
2490 clear_inode_flag(inode, FI_DO_DEFRAG);
2492 inode_unlock(inode);
2494 range->len = (u64)total << PAGE_SHIFT;
2498 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2500 struct inode *inode = file_inode(filp);
2501 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2502 struct f2fs_defragment range;
2505 if (!capable(CAP_SYS_ADMIN))
2508 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2511 if (f2fs_readonly(sbi->sb))
2514 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2518 /* verify alignment of offset & size */
2519 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2522 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2523 sbi->max_file_blocks))
2526 err = mnt_want_write_file(filp);
2530 err = f2fs_defragment_range(sbi, filp, &range);
2531 mnt_drop_write_file(filp);
2533 f2fs_update_time(sbi, REQ_TIME);
2537 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2544 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2545 struct file *file_out, loff_t pos_out, size_t len)
2547 struct inode *src = file_inode(file_in);
2548 struct inode *dst = file_inode(file_out);
2549 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2550 size_t olen = len, dst_max_i_size = 0;
2554 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2555 src->i_sb != dst->i_sb)
2558 if (unlikely(f2fs_readonly(src->i_sb)))
2561 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2564 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2568 if (pos_in == pos_out)
2570 if (pos_out > pos_in && pos_out < pos_in + len)
2577 if (!inode_trylock(dst))
2582 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2585 olen = len = src->i_size - pos_in;
2586 if (pos_in + len == src->i_size)
2587 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2593 dst_osize = dst->i_size;
2594 if (pos_out + olen > dst->i_size)
2595 dst_max_i_size = pos_out + olen;
2597 /* verify the end result is block aligned */
2598 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2599 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2600 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2603 ret = f2fs_convert_inline_inode(src);
2607 ret = f2fs_convert_inline_inode(dst);
2611 /* write out all dirty pages from offset */
2612 ret = filemap_write_and_wait_range(src->i_mapping,
2613 pos_in, pos_in + len);
2617 ret = filemap_write_and_wait_range(dst->i_mapping,
2618 pos_out, pos_out + len);
2622 f2fs_balance_fs(sbi, true);
2624 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2627 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2632 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2633 pos_out >> F2FS_BLKSIZE_BITS,
2634 len >> F2FS_BLKSIZE_BITS, false);
2638 f2fs_i_size_write(dst, dst_max_i_size);
2639 else if (dst_osize != dst->i_size)
2640 f2fs_i_size_write(dst, dst_osize);
2642 f2fs_unlock_op(sbi);
2645 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2647 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2656 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2658 struct f2fs_move_range range;
2662 if (!(filp->f_mode & FMODE_READ) ||
2663 !(filp->f_mode & FMODE_WRITE))
2666 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2670 dst = fdget(range.dst_fd);
2674 if (!(dst.file->f_mode & FMODE_WRITE)) {
2679 err = mnt_want_write_file(filp);
2683 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2684 range.pos_out, range.len);
2686 mnt_drop_write_file(filp);
2690 if (copy_to_user((struct f2fs_move_range __user *)arg,
2691 &range, sizeof(range)))
2698 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2700 struct inode *inode = file_inode(filp);
2701 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2702 struct sit_info *sm = SIT_I(sbi);
2703 unsigned int start_segno = 0, end_segno = 0;
2704 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2705 struct f2fs_flush_device range;
2708 if (!capable(CAP_SYS_ADMIN))
2711 if (f2fs_readonly(sbi->sb))
2714 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2717 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2721 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2722 __is_large_section(sbi)) {
2723 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2724 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2728 ret = mnt_want_write_file(filp);
2732 if (range.dev_num != 0)
2733 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2734 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2736 start_segno = sm->last_victim[FLUSH_DEVICE];
2737 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2738 start_segno = dev_start_segno;
2739 end_segno = min(start_segno + range.segments, dev_end_segno);
2741 while (start_segno < end_segno) {
2742 if (!mutex_trylock(&sbi->gc_mutex)) {
2746 sm->last_victim[GC_CB] = end_segno + 1;
2747 sm->last_victim[GC_GREEDY] = end_segno + 1;
2748 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2749 ret = f2fs_gc(sbi, true, true, start_segno);
2757 mnt_drop_write_file(filp);
2761 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2763 struct inode *inode = file_inode(filp);
2764 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2766 /* Must validate to set it with SQLite behavior in Android. */
2767 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2769 return put_user(sb_feature, (u32 __user *)arg);
2773 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2775 struct dquot *transfer_to[MAXQUOTAS] = {};
2776 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2777 struct super_block *sb = sbi->sb;
2780 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2781 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2782 err = __dquot_transfer(inode, transfer_to);
2784 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2785 dqput(transfer_to[PRJQUOTA]);
2790 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2792 struct inode *inode = file_inode(filp);
2793 struct f2fs_inode_info *fi = F2FS_I(inode);
2794 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2799 if (!f2fs_sb_has_project_quota(sbi)) {
2800 if (projid != F2FS_DEF_PROJID)
2806 if (!f2fs_has_extra_attr(inode))
2809 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2811 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2815 /* Is it quota file? Do not allow user to mess with it */
2816 if (IS_NOQUOTA(inode))
2819 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2821 return PTR_ERR(ipage);
2823 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2826 f2fs_put_page(ipage, 1);
2829 f2fs_put_page(ipage, 1);
2831 err = dquot_initialize(inode);
2836 err = f2fs_transfer_project_quota(inode, kprojid);
2840 F2FS_I(inode)->i_projid = kprojid;
2841 inode->i_ctime = current_time(inode);
2842 f2fs_mark_inode_dirty_sync(inode, true);
2844 f2fs_unlock_op(sbi);
2848 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2853 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2855 if (projid != F2FS_DEF_PROJID)
2861 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
2864 * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
2865 * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
2866 * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
2869 static const struct {
2872 } f2fs_xflags_map[] = {
2873 { F2FS_SYNC_FL, FS_XFLAG_SYNC },
2874 { F2FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE },
2875 { F2FS_APPEND_FL, FS_XFLAG_APPEND },
2876 { F2FS_NODUMP_FL, FS_XFLAG_NODUMP },
2877 { F2FS_NOATIME_FL, FS_XFLAG_NOATIME },
2878 { F2FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT },
2881 #define F2FS_SUPPORTED_XFLAGS ( \
2883 FS_XFLAG_IMMUTABLE | \
2886 FS_XFLAG_NOATIME | \
2887 FS_XFLAG_PROJINHERIT)
2889 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
2890 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
2895 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2896 if (iflags & f2fs_xflags_map[i].iflag)
2897 xflags |= f2fs_xflags_map[i].xflag;
2902 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
2903 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
2908 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2909 if (xflags & f2fs_xflags_map[i].xflag)
2910 iflags |= f2fs_xflags_map[i].iflag;
2915 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
2917 struct f2fs_inode_info *fi = F2FS_I(inode);
2919 simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
2921 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2922 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2925 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2927 struct inode *inode = file_inode(filp);
2930 f2fs_fill_fsxattr(inode, &fa);
2932 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2937 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2939 struct inode *inode = file_inode(filp);
2940 struct fsxattr fa, old_fa;
2944 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2947 /* Make sure caller has proper permission */
2948 if (!inode_owner_or_capable(inode))
2951 if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
2954 iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2955 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2958 err = mnt_want_write_file(filp);
2964 f2fs_fill_fsxattr(inode, &old_fa);
2965 err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
2969 err = f2fs_setflags_common(inode, iflags,
2970 f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
2974 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2976 inode_unlock(inode);
2977 mnt_drop_write_file(filp);
2981 int f2fs_pin_file_control(struct inode *inode, bool inc)
2983 struct f2fs_inode_info *fi = F2FS_I(inode);
2984 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2986 /* Use i_gc_failures for normal file as a risk signal. */
2988 f2fs_i_gc_failures_write(inode,
2989 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2991 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2992 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
2993 __func__, inode->i_ino,
2994 fi->i_gc_failures[GC_FAILURE_PIN]);
2995 clear_inode_flag(inode, FI_PIN_FILE);
3001 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3003 struct inode *inode = file_inode(filp);
3007 if (get_user(pin, (__u32 __user *)arg))
3010 if (!S_ISREG(inode->i_mode))
3013 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3016 ret = mnt_want_write_file(filp);
3022 if (f2fs_should_update_outplace(inode, NULL)) {
3028 clear_inode_flag(inode, FI_PIN_FILE);
3029 f2fs_i_gc_failures_write(inode, 0);
3033 if (f2fs_pin_file_control(inode, false)) {
3037 ret = f2fs_convert_inline_inode(inode);
3041 set_inode_flag(inode, FI_PIN_FILE);
3042 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3044 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3046 inode_unlock(inode);
3047 mnt_drop_write_file(filp);
3051 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3053 struct inode *inode = file_inode(filp);
3056 if (is_inode_flag_set(inode, FI_PIN_FILE))
3057 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3058 return put_user(pin, (u32 __user *)arg);
3061 int f2fs_precache_extents(struct inode *inode)
3063 struct f2fs_inode_info *fi = F2FS_I(inode);
3064 struct f2fs_map_blocks map;
3065 pgoff_t m_next_extent;
3069 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3073 map.m_next_pgofs = NULL;
3074 map.m_next_extent = &m_next_extent;
3075 map.m_seg_type = NO_CHECK_TYPE;
3076 map.m_may_create = false;
3077 end = F2FS_I_SB(inode)->max_file_blocks;
3079 while (map.m_lblk < end) {
3080 map.m_len = end - map.m_lblk;
3082 down_write(&fi->i_gc_rwsem[WRITE]);
3083 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3084 up_write(&fi->i_gc_rwsem[WRITE]);
3088 map.m_lblk = m_next_extent;
3094 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3096 return f2fs_precache_extents(file_inode(filp));
3099 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3101 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3105 if (!capable(CAP_SYS_ADMIN))
3108 if (f2fs_readonly(sbi->sb))
3111 if (copy_from_user(&block_count, (void __user *)arg,
3112 sizeof(block_count)))
3115 ret = f2fs_resize_fs(sbi, block_count);
3120 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3122 struct inode *inode = file_inode(filp);
3124 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3126 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3127 f2fs_warn(F2FS_I_SB(inode),
3128 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3133 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3136 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3138 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3141 return fsverity_ioctl_measure(filp, (void __user *)arg);
3144 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3146 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3150 case F2FS_IOC_GETFLAGS:
3151 return f2fs_ioc_getflags(filp, arg);
3152 case F2FS_IOC_SETFLAGS:
3153 return f2fs_ioc_setflags(filp, arg);
3154 case F2FS_IOC_GETVERSION:
3155 return f2fs_ioc_getversion(filp, arg);
3156 case F2FS_IOC_START_ATOMIC_WRITE:
3157 return f2fs_ioc_start_atomic_write(filp);
3158 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3159 return f2fs_ioc_commit_atomic_write(filp);
3160 case F2FS_IOC_START_VOLATILE_WRITE:
3161 return f2fs_ioc_start_volatile_write(filp);
3162 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3163 return f2fs_ioc_release_volatile_write(filp);
3164 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3165 return f2fs_ioc_abort_volatile_write(filp);
3166 case F2FS_IOC_SHUTDOWN:
3167 return f2fs_ioc_shutdown(filp, arg);
3169 return f2fs_ioc_fitrim(filp, arg);
3170 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3171 return f2fs_ioc_set_encryption_policy(filp, arg);
3172 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3173 return f2fs_ioc_get_encryption_policy(filp, arg);
3174 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3175 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3176 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3177 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3178 case FS_IOC_ADD_ENCRYPTION_KEY:
3179 return f2fs_ioc_add_encryption_key(filp, arg);
3180 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3181 return f2fs_ioc_remove_encryption_key(filp, arg);
3182 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3183 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3184 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3185 return f2fs_ioc_get_encryption_key_status(filp, arg);
3186 case F2FS_IOC_GARBAGE_COLLECT:
3187 return f2fs_ioc_gc(filp, arg);
3188 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3189 return f2fs_ioc_gc_range(filp, arg);
3190 case F2FS_IOC_WRITE_CHECKPOINT:
3191 return f2fs_ioc_write_checkpoint(filp, arg);
3192 case F2FS_IOC_DEFRAGMENT:
3193 return f2fs_ioc_defragment(filp, arg);
3194 case F2FS_IOC_MOVE_RANGE:
3195 return f2fs_ioc_move_range(filp, arg);
3196 case F2FS_IOC_FLUSH_DEVICE:
3197 return f2fs_ioc_flush_device(filp, arg);
3198 case F2FS_IOC_GET_FEATURES:
3199 return f2fs_ioc_get_features(filp, arg);
3200 case F2FS_IOC_FSGETXATTR:
3201 return f2fs_ioc_fsgetxattr(filp, arg);
3202 case F2FS_IOC_FSSETXATTR:
3203 return f2fs_ioc_fssetxattr(filp, arg);
3204 case F2FS_IOC_GET_PIN_FILE:
3205 return f2fs_ioc_get_pin_file(filp, arg);
3206 case F2FS_IOC_SET_PIN_FILE:
3207 return f2fs_ioc_set_pin_file(filp, arg);
3208 case F2FS_IOC_PRECACHE_EXTENTS:
3209 return f2fs_ioc_precache_extents(filp, arg);
3210 case F2FS_IOC_RESIZE_FS:
3211 return f2fs_ioc_resize_fs(filp, arg);
3212 case FS_IOC_ENABLE_VERITY:
3213 return f2fs_ioc_enable_verity(filp, arg);
3214 case FS_IOC_MEASURE_VERITY:
3215 return f2fs_ioc_measure_verity(filp, arg);
3221 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3223 struct file *file = iocb->ki_filp;
3224 struct inode *inode = file_inode(file);
3227 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3232 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT)) {
3237 if (!inode_trylock(inode)) {
3238 if (iocb->ki_flags & IOCB_NOWAIT) {
3245 ret = generic_write_checks(iocb, from);
3247 bool preallocated = false;
3248 size_t target_size = 0;
3251 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3252 set_inode_flag(inode, FI_NO_PREALLOC);
3254 if ((iocb->ki_flags & IOCB_NOWAIT)) {
3255 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3256 iov_iter_count(from)) ||
3257 f2fs_has_inline_data(inode) ||
3258 f2fs_force_buffered_io(inode, iocb, from)) {
3259 clear_inode_flag(inode, FI_NO_PREALLOC);
3260 inode_unlock(inode);
3265 preallocated = true;
3266 target_size = iocb->ki_pos + iov_iter_count(from);
3268 err = f2fs_preallocate_blocks(iocb, from);
3270 clear_inode_flag(inode, FI_NO_PREALLOC);
3271 inode_unlock(inode);
3276 ret = __generic_file_write_iter(iocb, from);
3277 clear_inode_flag(inode, FI_NO_PREALLOC);
3279 /* if we couldn't write data, we should deallocate blocks. */
3280 if (preallocated && i_size_read(inode) < target_size)
3281 f2fs_truncate(inode);
3284 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3286 inode_unlock(inode);
3288 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3289 iov_iter_count(from), ret);
3291 ret = generic_write_sync(iocb, ret);
3295 #ifdef CONFIG_COMPAT
3296 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3299 case F2FS_IOC32_GETFLAGS:
3300 cmd = F2FS_IOC_GETFLAGS;
3302 case F2FS_IOC32_SETFLAGS:
3303 cmd = F2FS_IOC_SETFLAGS;
3305 case F2FS_IOC32_GETVERSION:
3306 cmd = F2FS_IOC_GETVERSION;
3308 case F2FS_IOC_START_ATOMIC_WRITE:
3309 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3310 case F2FS_IOC_START_VOLATILE_WRITE:
3311 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3312 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3313 case F2FS_IOC_SHUTDOWN:
3314 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3315 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3316 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3317 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3318 case FS_IOC_ADD_ENCRYPTION_KEY:
3319 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3320 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3321 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3322 case F2FS_IOC_GARBAGE_COLLECT:
3323 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3324 case F2FS_IOC_WRITE_CHECKPOINT:
3325 case F2FS_IOC_DEFRAGMENT:
3326 case F2FS_IOC_MOVE_RANGE:
3327 case F2FS_IOC_FLUSH_DEVICE:
3328 case F2FS_IOC_GET_FEATURES:
3329 case F2FS_IOC_FSGETXATTR:
3330 case F2FS_IOC_FSSETXATTR:
3331 case F2FS_IOC_GET_PIN_FILE:
3332 case F2FS_IOC_SET_PIN_FILE:
3333 case F2FS_IOC_PRECACHE_EXTENTS:
3334 case F2FS_IOC_RESIZE_FS:
3335 case FS_IOC_ENABLE_VERITY:
3336 case FS_IOC_MEASURE_VERITY:
3339 return -ENOIOCTLCMD;
3341 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3345 const struct file_operations f2fs_file_operations = {
3346 .llseek = f2fs_llseek,
3347 .read_iter = generic_file_read_iter,
3348 .write_iter = f2fs_file_write_iter,
3349 .open = f2fs_file_open,
3350 .release = f2fs_release_file,
3351 .mmap = f2fs_file_mmap,
3352 .flush = f2fs_file_flush,
3353 .fsync = f2fs_sync_file,
3354 .fallocate = f2fs_fallocate,
3355 .unlocked_ioctl = f2fs_ioctl,
3356 #ifdef CONFIG_COMPAT
3357 .compat_ioctl = f2fs_compat_ioctl,
3359 .splice_read = generic_file_splice_read,
3360 .splice_write = iter_file_splice_write,