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 filp->f_mode |= FMODE_NOWAIT;
501 return dquot_file_open(inode, filp);
504 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
506 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
507 struct f2fs_node *raw_node;
508 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
512 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
513 base = get_extra_isize(dn->inode);
515 raw_node = F2FS_NODE(dn->node_page);
516 addr = blkaddr_in_node(raw_node) + base + ofs;
518 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
519 block_t blkaddr = le32_to_cpu(*addr);
521 if (blkaddr == NULL_ADDR)
524 dn->data_blkaddr = NULL_ADDR;
525 f2fs_set_data_blkaddr(dn);
527 if (__is_valid_data_blkaddr(blkaddr) &&
528 !f2fs_is_valid_blkaddr(sbi, blkaddr,
529 DATA_GENERIC_ENHANCE))
532 f2fs_invalidate_blocks(sbi, blkaddr);
533 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
534 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
541 * once we invalidate valid blkaddr in range [ofs, ofs + count],
542 * we will invalidate all blkaddr in the whole range.
544 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
546 f2fs_update_extent_cache_range(dn, fofs, 0, len);
547 dec_valid_block_count(sbi, dn->inode, nr_free);
549 dn->ofs_in_node = ofs;
551 f2fs_update_time(sbi, REQ_TIME);
552 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
553 dn->ofs_in_node, nr_free);
556 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
558 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
561 static int truncate_partial_data_page(struct inode *inode, u64 from,
564 loff_t offset = from & (PAGE_SIZE - 1);
565 pgoff_t index = from >> PAGE_SHIFT;
566 struct address_space *mapping = inode->i_mapping;
569 if (!offset && !cache_only)
573 page = find_lock_page(mapping, index);
574 if (page && PageUptodate(page))
576 f2fs_put_page(page, 1);
580 page = f2fs_get_lock_data_page(inode, index, true);
582 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
584 f2fs_wait_on_page_writeback(page, DATA, true, true);
585 zero_user(page, offset, PAGE_SIZE - offset);
587 /* An encrypted inode should have a key and truncate the last page. */
588 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
590 set_page_dirty(page);
591 f2fs_put_page(page, 1);
595 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
597 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
598 struct dnode_of_data dn;
600 int count = 0, err = 0;
602 bool truncate_page = false;
604 trace_f2fs_truncate_blocks_enter(inode, from);
606 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
608 if (free_from >= sbi->max_file_blocks)
614 ipage = f2fs_get_node_page(sbi, inode->i_ino);
616 err = PTR_ERR(ipage);
620 if (f2fs_has_inline_data(inode)) {
621 f2fs_truncate_inline_inode(inode, ipage, from);
622 f2fs_put_page(ipage, 1);
623 truncate_page = true;
627 set_new_dnode(&dn, inode, ipage, NULL, 0);
628 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
635 count = ADDRS_PER_PAGE(dn.node_page, inode);
637 count -= dn.ofs_in_node;
638 f2fs_bug_on(sbi, count < 0);
640 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
641 f2fs_truncate_data_blocks_range(&dn, count);
647 err = f2fs_truncate_inode_blocks(inode, free_from);
652 /* lastly zero out the first data page */
654 err = truncate_partial_data_page(inode, from, truncate_page);
656 trace_f2fs_truncate_blocks_exit(inode, err);
660 int f2fs_truncate(struct inode *inode)
664 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
667 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
668 S_ISLNK(inode->i_mode)))
671 trace_f2fs_truncate(inode);
673 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
674 f2fs_show_injection_info(FAULT_TRUNCATE);
678 /* we should check inline_data size */
679 if (!f2fs_may_inline_data(inode)) {
680 err = f2fs_convert_inline_inode(inode);
685 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
689 inode->i_mtime = inode->i_ctime = current_time(inode);
690 f2fs_mark_inode_dirty_sync(inode, false);
694 int f2fs_getattr(const struct path *path, struct kstat *stat,
695 u32 request_mask, unsigned int query_flags)
697 struct inode *inode = d_inode(path->dentry);
698 struct f2fs_inode_info *fi = F2FS_I(inode);
699 struct f2fs_inode *ri;
702 if (f2fs_has_extra_attr(inode) &&
703 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
704 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
705 stat->result_mask |= STATX_BTIME;
706 stat->btime.tv_sec = fi->i_crtime.tv_sec;
707 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
711 if (flags & F2FS_APPEND_FL)
712 stat->attributes |= STATX_ATTR_APPEND;
713 if (IS_ENCRYPTED(inode))
714 stat->attributes |= STATX_ATTR_ENCRYPTED;
715 if (flags & F2FS_IMMUTABLE_FL)
716 stat->attributes |= STATX_ATTR_IMMUTABLE;
717 if (flags & F2FS_NODUMP_FL)
718 stat->attributes |= STATX_ATTR_NODUMP;
720 stat->attributes_mask |= (STATX_ATTR_APPEND |
721 STATX_ATTR_ENCRYPTED |
722 STATX_ATTR_IMMUTABLE |
725 generic_fillattr(inode, stat);
727 /* we need to show initial sectors used for inline_data/dentries */
728 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
729 f2fs_has_inline_dentry(inode))
730 stat->blocks += (stat->size + 511) >> 9;
735 #ifdef CONFIG_F2FS_FS_POSIX_ACL
736 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
738 unsigned int ia_valid = attr->ia_valid;
740 if (ia_valid & ATTR_UID)
741 inode->i_uid = attr->ia_uid;
742 if (ia_valid & ATTR_GID)
743 inode->i_gid = attr->ia_gid;
744 if (ia_valid & ATTR_ATIME)
745 inode->i_atime = timespec64_trunc(attr->ia_atime,
746 inode->i_sb->s_time_gran);
747 if (ia_valid & ATTR_MTIME)
748 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
749 inode->i_sb->s_time_gran);
750 if (ia_valid & ATTR_CTIME)
751 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
752 inode->i_sb->s_time_gran);
753 if (ia_valid & ATTR_MODE) {
754 umode_t mode = attr->ia_mode;
756 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
758 set_acl_inode(inode, mode);
762 #define __setattr_copy setattr_copy
765 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
767 struct inode *inode = d_inode(dentry);
770 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
773 err = setattr_prepare(dentry, attr);
777 err = fscrypt_prepare_setattr(dentry, attr);
781 if (is_quota_modification(inode, attr)) {
782 err = dquot_initialize(inode);
786 if ((attr->ia_valid & ATTR_UID &&
787 !uid_eq(attr->ia_uid, inode->i_uid)) ||
788 (attr->ia_valid & ATTR_GID &&
789 !gid_eq(attr->ia_gid, inode->i_gid))) {
790 f2fs_lock_op(F2FS_I_SB(inode));
791 err = dquot_transfer(inode, attr);
793 set_sbi_flag(F2FS_I_SB(inode),
794 SBI_QUOTA_NEED_REPAIR);
795 f2fs_unlock_op(F2FS_I_SB(inode));
799 * update uid/gid under lock_op(), so that dquot and inode can
800 * be updated atomically.
802 if (attr->ia_valid & ATTR_UID)
803 inode->i_uid = attr->ia_uid;
804 if (attr->ia_valid & ATTR_GID)
805 inode->i_gid = attr->ia_gid;
806 f2fs_mark_inode_dirty_sync(inode, true);
807 f2fs_unlock_op(F2FS_I_SB(inode));
810 if (attr->ia_valid & ATTR_SIZE) {
811 bool to_smaller = (attr->ia_size <= i_size_read(inode));
813 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
814 down_write(&F2FS_I(inode)->i_mmap_sem);
816 truncate_setsize(inode, attr->ia_size);
819 err = f2fs_truncate(inode);
821 * do not trim all blocks after i_size if target size is
822 * larger than i_size.
824 up_write(&F2FS_I(inode)->i_mmap_sem);
825 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
831 /* should convert inline inode here */
832 if (!f2fs_may_inline_data(inode)) {
833 err = f2fs_convert_inline_inode(inode);
837 inode->i_mtime = inode->i_ctime = current_time(inode);
840 down_write(&F2FS_I(inode)->i_sem);
841 F2FS_I(inode)->last_disk_size = i_size_read(inode);
842 up_write(&F2FS_I(inode)->i_sem);
845 __setattr_copy(inode, attr);
847 if (attr->ia_valid & ATTR_MODE) {
848 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
849 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
850 inode->i_mode = F2FS_I(inode)->i_acl_mode;
851 clear_inode_flag(inode, FI_ACL_MODE);
855 /* file size may changed here */
856 f2fs_mark_inode_dirty_sync(inode, true);
858 /* inode change will produce dirty node pages flushed by checkpoint */
859 f2fs_balance_fs(F2FS_I_SB(inode), true);
864 const struct inode_operations f2fs_file_inode_operations = {
865 .getattr = f2fs_getattr,
866 .setattr = f2fs_setattr,
867 .get_acl = f2fs_get_acl,
868 .set_acl = f2fs_set_acl,
869 #ifdef CONFIG_F2FS_FS_XATTR
870 .listxattr = f2fs_listxattr,
872 .fiemap = f2fs_fiemap,
875 static int fill_zero(struct inode *inode, pgoff_t index,
876 loff_t start, loff_t len)
878 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
884 f2fs_balance_fs(sbi, true);
887 page = f2fs_get_new_data_page(inode, NULL, index, false);
891 return PTR_ERR(page);
893 f2fs_wait_on_page_writeback(page, DATA, true, true);
894 zero_user(page, start, len);
895 set_page_dirty(page);
896 f2fs_put_page(page, 1);
900 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
904 while (pg_start < pg_end) {
905 struct dnode_of_data dn;
906 pgoff_t end_offset, count;
908 set_new_dnode(&dn, inode, NULL, NULL, 0);
909 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
911 if (err == -ENOENT) {
912 pg_start = f2fs_get_next_page_offset(&dn,
919 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
920 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
922 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
924 f2fs_truncate_data_blocks_range(&dn, count);
932 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
934 pgoff_t pg_start, pg_end;
935 loff_t off_start, off_end;
938 ret = f2fs_convert_inline_inode(inode);
942 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
943 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
945 off_start = offset & (PAGE_SIZE - 1);
946 off_end = (offset + len) & (PAGE_SIZE - 1);
948 if (pg_start == pg_end) {
949 ret = fill_zero(inode, pg_start, off_start,
950 off_end - off_start);
955 ret = fill_zero(inode, pg_start++, off_start,
956 PAGE_SIZE - off_start);
961 ret = fill_zero(inode, pg_end, 0, off_end);
966 if (pg_start < pg_end) {
967 struct address_space *mapping = inode->i_mapping;
968 loff_t blk_start, blk_end;
969 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
971 f2fs_balance_fs(sbi, true);
973 blk_start = (loff_t)pg_start << PAGE_SHIFT;
974 blk_end = (loff_t)pg_end << PAGE_SHIFT;
976 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
977 down_write(&F2FS_I(inode)->i_mmap_sem);
979 truncate_inode_pages_range(mapping, blk_start,
983 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
986 up_write(&F2FS_I(inode)->i_mmap_sem);
987 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
994 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
995 int *do_replace, pgoff_t off, pgoff_t len)
997 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
998 struct dnode_of_data dn;
1002 set_new_dnode(&dn, inode, NULL, NULL, 0);
1003 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1004 if (ret && ret != -ENOENT) {
1006 } else if (ret == -ENOENT) {
1007 if (dn.max_level == 0)
1009 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1016 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1017 dn.ofs_in_node, len);
1018 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1019 *blkaddr = datablock_addr(dn.inode,
1020 dn.node_page, dn.ofs_in_node);
1022 if (__is_valid_data_blkaddr(*blkaddr) &&
1023 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1024 DATA_GENERIC_ENHANCE)) {
1025 f2fs_put_dnode(&dn);
1026 return -EFSCORRUPTED;
1029 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1031 if (test_opt(sbi, LFS)) {
1032 f2fs_put_dnode(&dn);
1036 /* do not invalidate this block address */
1037 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1041 f2fs_put_dnode(&dn);
1050 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1051 int *do_replace, pgoff_t off, int len)
1053 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1054 struct dnode_of_data dn;
1057 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1058 if (*do_replace == 0)
1061 set_new_dnode(&dn, inode, NULL, NULL, 0);
1062 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1064 dec_valid_block_count(sbi, inode, 1);
1065 f2fs_invalidate_blocks(sbi, *blkaddr);
1067 f2fs_update_data_blkaddr(&dn, *blkaddr);
1069 f2fs_put_dnode(&dn);
1074 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1075 block_t *blkaddr, int *do_replace,
1076 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1078 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1083 if (blkaddr[i] == NULL_ADDR && !full) {
1088 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1089 struct dnode_of_data dn;
1090 struct node_info ni;
1094 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1095 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1099 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1101 f2fs_put_dnode(&dn);
1105 ilen = min((pgoff_t)
1106 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1107 dn.ofs_in_node, len - i);
1109 dn.data_blkaddr = datablock_addr(dn.inode,
1110 dn.node_page, dn.ofs_in_node);
1111 f2fs_truncate_data_blocks_range(&dn, 1);
1113 if (do_replace[i]) {
1114 f2fs_i_blocks_write(src_inode,
1116 f2fs_i_blocks_write(dst_inode,
1118 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1119 blkaddr[i], ni.version, true, false);
1125 new_size = (dst + i) << PAGE_SHIFT;
1126 if (dst_inode->i_size < new_size)
1127 f2fs_i_size_write(dst_inode, new_size);
1128 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1130 f2fs_put_dnode(&dn);
1132 struct page *psrc, *pdst;
1134 psrc = f2fs_get_lock_data_page(src_inode,
1137 return PTR_ERR(psrc);
1138 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1141 f2fs_put_page(psrc, 1);
1142 return PTR_ERR(pdst);
1144 f2fs_copy_page(psrc, pdst);
1145 set_page_dirty(pdst);
1146 f2fs_put_page(pdst, 1);
1147 f2fs_put_page(psrc, 1);
1149 ret = f2fs_truncate_hole(src_inode,
1150 src + i, src + i + 1);
1159 static int __exchange_data_block(struct inode *src_inode,
1160 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1161 pgoff_t len, bool full)
1163 block_t *src_blkaddr;
1169 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1171 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1172 array_size(olen, sizeof(block_t)),
1177 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1178 array_size(olen, sizeof(int)),
1181 kvfree(src_blkaddr);
1185 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1186 do_replace, src, olen);
1190 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1191 do_replace, src, dst, olen, full);
1199 kvfree(src_blkaddr);
1205 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1206 kvfree(src_blkaddr);
1211 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1213 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1214 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1215 pgoff_t start = offset >> PAGE_SHIFT;
1216 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1219 f2fs_balance_fs(sbi, true);
1221 /* avoid gc operation during block exchange */
1222 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1223 down_write(&F2FS_I(inode)->i_mmap_sem);
1226 f2fs_drop_extent_tree(inode);
1227 truncate_pagecache(inode, offset);
1228 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1229 f2fs_unlock_op(sbi);
1231 up_write(&F2FS_I(inode)->i_mmap_sem);
1232 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1236 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1241 if (offset + len >= i_size_read(inode))
1244 /* collapse range should be aligned to block size of f2fs. */
1245 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1248 ret = f2fs_convert_inline_inode(inode);
1252 /* write out all dirty pages from offset */
1253 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1257 ret = f2fs_do_collapse(inode, offset, len);
1261 /* write out all moved pages, if possible */
1262 down_write(&F2FS_I(inode)->i_mmap_sem);
1263 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1264 truncate_pagecache(inode, offset);
1266 new_size = i_size_read(inode) - len;
1267 truncate_pagecache(inode, new_size);
1269 ret = f2fs_truncate_blocks(inode, new_size, true);
1270 up_write(&F2FS_I(inode)->i_mmap_sem);
1272 f2fs_i_size_write(inode, new_size);
1276 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1279 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1280 pgoff_t index = start;
1281 unsigned int ofs_in_node = dn->ofs_in_node;
1285 for (; index < end; index++, dn->ofs_in_node++) {
1286 if (datablock_addr(dn->inode, dn->node_page,
1287 dn->ofs_in_node) == NULL_ADDR)
1291 dn->ofs_in_node = ofs_in_node;
1292 ret = f2fs_reserve_new_blocks(dn, count);
1296 dn->ofs_in_node = ofs_in_node;
1297 for (index = start; index < end; index++, dn->ofs_in_node++) {
1298 dn->data_blkaddr = datablock_addr(dn->inode,
1299 dn->node_page, dn->ofs_in_node);
1301 * f2fs_reserve_new_blocks will not guarantee entire block
1304 if (dn->data_blkaddr == NULL_ADDR) {
1308 if (dn->data_blkaddr != NEW_ADDR) {
1309 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1310 dn->data_blkaddr = NEW_ADDR;
1311 f2fs_set_data_blkaddr(dn);
1315 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1320 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1323 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1324 struct address_space *mapping = inode->i_mapping;
1325 pgoff_t index, pg_start, pg_end;
1326 loff_t new_size = i_size_read(inode);
1327 loff_t off_start, off_end;
1330 ret = inode_newsize_ok(inode, (len + offset));
1334 ret = f2fs_convert_inline_inode(inode);
1338 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1342 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1343 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1345 off_start = offset & (PAGE_SIZE - 1);
1346 off_end = (offset + len) & (PAGE_SIZE - 1);
1348 if (pg_start == pg_end) {
1349 ret = fill_zero(inode, pg_start, off_start,
1350 off_end - off_start);
1354 new_size = max_t(loff_t, new_size, offset + len);
1357 ret = fill_zero(inode, pg_start++, off_start,
1358 PAGE_SIZE - off_start);
1362 new_size = max_t(loff_t, new_size,
1363 (loff_t)pg_start << PAGE_SHIFT);
1366 for (index = pg_start; index < pg_end;) {
1367 struct dnode_of_data dn;
1368 unsigned int end_offset;
1371 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1372 down_write(&F2FS_I(inode)->i_mmap_sem);
1374 truncate_pagecache_range(inode,
1375 (loff_t)index << PAGE_SHIFT,
1376 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1380 set_new_dnode(&dn, inode, NULL, NULL, 0);
1381 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1383 f2fs_unlock_op(sbi);
1384 up_write(&F2FS_I(inode)->i_mmap_sem);
1385 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1389 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1390 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1392 ret = f2fs_do_zero_range(&dn, index, end);
1393 f2fs_put_dnode(&dn);
1395 f2fs_unlock_op(sbi);
1396 up_write(&F2FS_I(inode)->i_mmap_sem);
1397 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1399 f2fs_balance_fs(sbi, dn.node_changed);
1405 new_size = max_t(loff_t, new_size,
1406 (loff_t)index << PAGE_SHIFT);
1410 ret = fill_zero(inode, pg_end, 0, off_end);
1414 new_size = max_t(loff_t, new_size, offset + len);
1419 if (new_size > i_size_read(inode)) {
1420 if (mode & FALLOC_FL_KEEP_SIZE)
1421 file_set_keep_isize(inode);
1423 f2fs_i_size_write(inode, new_size);
1428 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1430 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1431 pgoff_t nr, pg_start, pg_end, delta, idx;
1435 new_size = i_size_read(inode) + len;
1436 ret = inode_newsize_ok(inode, new_size);
1440 if (offset >= i_size_read(inode))
1443 /* insert range should be aligned to block size of f2fs. */
1444 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1447 ret = f2fs_convert_inline_inode(inode);
1451 f2fs_balance_fs(sbi, true);
1453 down_write(&F2FS_I(inode)->i_mmap_sem);
1454 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1455 up_write(&F2FS_I(inode)->i_mmap_sem);
1459 /* write out all dirty pages from offset */
1460 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1464 pg_start = offset >> PAGE_SHIFT;
1465 pg_end = (offset + len) >> PAGE_SHIFT;
1466 delta = pg_end - pg_start;
1467 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1469 /* avoid gc operation during block exchange */
1470 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1471 down_write(&F2FS_I(inode)->i_mmap_sem);
1472 truncate_pagecache(inode, offset);
1474 while (!ret && idx > pg_start) {
1475 nr = idx - pg_start;
1481 f2fs_drop_extent_tree(inode);
1483 ret = __exchange_data_block(inode, inode, idx,
1484 idx + delta, nr, false);
1485 f2fs_unlock_op(sbi);
1487 up_write(&F2FS_I(inode)->i_mmap_sem);
1488 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1490 /* write out all moved pages, if possible */
1491 down_write(&F2FS_I(inode)->i_mmap_sem);
1492 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1493 truncate_pagecache(inode, offset);
1494 up_write(&F2FS_I(inode)->i_mmap_sem);
1497 f2fs_i_size_write(inode, new_size);
1501 static int expand_inode_data(struct inode *inode, loff_t offset,
1502 loff_t len, int mode)
1504 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1505 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1506 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1507 .m_may_create = true };
1509 loff_t new_size = i_size_read(inode);
1513 err = inode_newsize_ok(inode, (len + offset));
1517 err = f2fs_convert_inline_inode(inode);
1521 f2fs_balance_fs(sbi, true);
1523 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1524 off_end = (offset + len) & (PAGE_SIZE - 1);
1526 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1527 map.m_len = pg_end - map.m_lblk;
1531 if (f2fs_is_pinned_file(inode))
1532 map.m_seg_type = CURSEG_COLD_DATA;
1534 err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
1535 F2FS_GET_BLOCK_PRE_DIO :
1536 F2FS_GET_BLOCK_PRE_AIO));
1543 last_off = map.m_lblk + map.m_len - 1;
1545 /* update new size to the failed position */
1546 new_size = (last_off == pg_end) ? offset + len :
1547 (loff_t)(last_off + 1) << PAGE_SHIFT;
1549 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1552 if (new_size > i_size_read(inode)) {
1553 if (mode & FALLOC_FL_KEEP_SIZE)
1554 file_set_keep_isize(inode);
1556 f2fs_i_size_write(inode, new_size);
1562 static long f2fs_fallocate(struct file *file, int mode,
1563 loff_t offset, loff_t len)
1565 struct inode *inode = file_inode(file);
1568 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1571 /* f2fs only support ->fallocate for regular file */
1572 if (!S_ISREG(inode->i_mode))
1575 if (IS_ENCRYPTED(inode) &&
1576 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1579 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1580 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1581 FALLOC_FL_INSERT_RANGE))
1586 if (mode & FALLOC_FL_PUNCH_HOLE) {
1587 if (offset >= inode->i_size)
1590 ret = punch_hole(inode, offset, len);
1591 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1592 ret = f2fs_collapse_range(inode, offset, len);
1593 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1594 ret = f2fs_zero_range(inode, offset, len, mode);
1595 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1596 ret = f2fs_insert_range(inode, offset, len);
1598 ret = expand_inode_data(inode, offset, len, mode);
1602 inode->i_mtime = inode->i_ctime = current_time(inode);
1603 f2fs_mark_inode_dirty_sync(inode, false);
1604 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1608 inode_unlock(inode);
1610 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1614 static int f2fs_release_file(struct inode *inode, struct file *filp)
1617 * f2fs_relase_file is called at every close calls. So we should
1618 * not drop any inmemory pages by close called by other process.
1620 if (!(filp->f_mode & FMODE_WRITE) ||
1621 atomic_read(&inode->i_writecount) != 1)
1624 /* some remained atomic pages should discarded */
1625 if (f2fs_is_atomic_file(inode))
1626 f2fs_drop_inmem_pages(inode);
1627 if (f2fs_is_volatile_file(inode)) {
1628 set_inode_flag(inode, FI_DROP_CACHE);
1629 filemap_fdatawrite(inode->i_mapping);
1630 clear_inode_flag(inode, FI_DROP_CACHE);
1631 clear_inode_flag(inode, FI_VOLATILE_FILE);
1632 stat_dec_volatile_write(inode);
1637 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1639 struct inode *inode = file_inode(file);
1642 * If the process doing a transaction is crashed, we should do
1643 * roll-back. Otherwise, other reader/write can see corrupted database
1644 * until all the writers close its file. Since this should be done
1645 * before dropping file lock, it needs to do in ->flush.
1647 if (f2fs_is_atomic_file(inode) &&
1648 F2FS_I(inode)->inmem_task == current)
1649 f2fs_drop_inmem_pages(inode);
1653 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1655 struct f2fs_inode_info *fi = F2FS_I(inode);
1657 /* Is it quota file? Do not allow user to mess with it */
1658 if (IS_NOQUOTA(inode))
1661 fi->i_flags = iflags | (fi->i_flags & ~mask);
1663 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1664 set_inode_flag(inode, FI_PROJ_INHERIT);
1666 clear_inode_flag(inode, FI_PROJ_INHERIT);
1668 inode->i_ctime = current_time(inode);
1669 f2fs_set_inode_flags(inode);
1670 f2fs_mark_inode_dirty_sync(inode, true);
1674 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1677 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1678 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1679 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1680 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1683 static const struct {
1686 } f2fs_fsflags_map[] = {
1687 { F2FS_SYNC_FL, FS_SYNC_FL },
1688 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1689 { F2FS_APPEND_FL, FS_APPEND_FL },
1690 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1691 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1692 { F2FS_INDEX_FL, FS_INDEX_FL },
1693 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1694 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1697 #define F2FS_GETTABLE_FS_FL ( \
1705 FS_PROJINHERIT_FL | \
1707 FS_INLINE_DATA_FL | \
1710 #define F2FS_SETTABLE_FS_FL ( \
1719 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1720 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1725 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1726 if (iflags & f2fs_fsflags_map[i].iflag)
1727 fsflags |= f2fs_fsflags_map[i].fsflag;
1732 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1733 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1738 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1739 if (fsflags & f2fs_fsflags_map[i].fsflag)
1740 iflags |= f2fs_fsflags_map[i].iflag;
1745 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1747 struct inode *inode = file_inode(filp);
1748 struct f2fs_inode_info *fi = F2FS_I(inode);
1749 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1751 if (IS_ENCRYPTED(inode))
1752 fsflags |= FS_ENCRYPT_FL;
1753 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1754 fsflags |= FS_INLINE_DATA_FL;
1755 if (is_inode_flag_set(inode, FI_PIN_FILE))
1756 fsflags |= FS_NOCOW_FL;
1758 fsflags &= F2FS_GETTABLE_FS_FL;
1760 return put_user(fsflags, (int __user *)arg);
1763 static int f2fs_ioc_setflags(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, old_fsflags;
1771 if (!inode_owner_or_capable(inode))
1774 if (get_user(fsflags, (int __user *)arg))
1777 if (fsflags & ~F2FS_GETTABLE_FS_FL)
1779 fsflags &= F2FS_SETTABLE_FS_FL;
1781 iflags = f2fs_fsflags_to_iflags(fsflags);
1782 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1785 ret = mnt_want_write_file(filp);
1791 old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1792 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1796 ret = f2fs_setflags_common(inode, iflags,
1797 f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1799 inode_unlock(inode);
1800 mnt_drop_write_file(filp);
1804 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1806 struct inode *inode = file_inode(filp);
1808 return put_user(inode->i_generation, (int __user *)arg);
1811 static int f2fs_ioc_start_atomic_write(struct file *filp)
1813 struct inode *inode = file_inode(filp);
1816 if (!inode_owner_or_capable(inode))
1819 if (!S_ISREG(inode->i_mode))
1822 ret = mnt_want_write_file(filp);
1828 if (f2fs_is_atomic_file(inode)) {
1829 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1834 ret = f2fs_convert_inline_inode(inode);
1838 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1841 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1842 * f2fs_is_atomic_file.
1844 if (get_dirty_pages(inode))
1845 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1846 inode->i_ino, get_dirty_pages(inode));
1847 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1849 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1853 set_inode_flag(inode, FI_ATOMIC_FILE);
1854 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1855 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1857 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1858 F2FS_I(inode)->inmem_task = current;
1859 stat_inc_atomic_write(inode);
1860 stat_update_max_atomic_write(inode);
1862 inode_unlock(inode);
1863 mnt_drop_write_file(filp);
1867 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1869 struct inode *inode = file_inode(filp);
1872 if (!inode_owner_or_capable(inode))
1875 ret = mnt_want_write_file(filp);
1879 f2fs_balance_fs(F2FS_I_SB(inode), true);
1883 if (f2fs_is_volatile_file(inode)) {
1888 if (f2fs_is_atomic_file(inode)) {
1889 ret = f2fs_commit_inmem_pages(inode);
1893 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1895 clear_inode_flag(inode, FI_ATOMIC_FILE);
1896 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1897 stat_dec_atomic_write(inode);
1900 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1903 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1904 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1907 inode_unlock(inode);
1908 mnt_drop_write_file(filp);
1912 static int f2fs_ioc_start_volatile_write(struct file *filp)
1914 struct inode *inode = file_inode(filp);
1917 if (!inode_owner_or_capable(inode))
1920 if (!S_ISREG(inode->i_mode))
1923 ret = mnt_want_write_file(filp);
1929 if (f2fs_is_volatile_file(inode))
1932 ret = f2fs_convert_inline_inode(inode);
1936 stat_inc_volatile_write(inode);
1937 stat_update_max_volatile_write(inode);
1939 set_inode_flag(inode, FI_VOLATILE_FILE);
1940 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1942 inode_unlock(inode);
1943 mnt_drop_write_file(filp);
1947 static int f2fs_ioc_release_volatile_write(struct file *filp)
1949 struct inode *inode = file_inode(filp);
1952 if (!inode_owner_or_capable(inode))
1955 ret = mnt_want_write_file(filp);
1961 if (!f2fs_is_volatile_file(inode))
1964 if (!f2fs_is_first_block_written(inode)) {
1965 ret = truncate_partial_data_page(inode, 0, true);
1969 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1971 inode_unlock(inode);
1972 mnt_drop_write_file(filp);
1976 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1978 struct inode *inode = file_inode(filp);
1981 if (!inode_owner_or_capable(inode))
1984 ret = mnt_want_write_file(filp);
1990 if (f2fs_is_atomic_file(inode))
1991 f2fs_drop_inmem_pages(inode);
1992 if (f2fs_is_volatile_file(inode)) {
1993 clear_inode_flag(inode, FI_VOLATILE_FILE);
1994 stat_dec_volatile_write(inode);
1995 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1998 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2000 inode_unlock(inode);
2002 mnt_drop_write_file(filp);
2003 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2007 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2009 struct inode *inode = file_inode(filp);
2010 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2011 struct super_block *sb = sbi->sb;
2015 if (!capable(CAP_SYS_ADMIN))
2018 if (get_user(in, (__u32 __user *)arg))
2021 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2022 ret = mnt_want_write_file(filp);
2028 case F2FS_GOING_DOWN_FULLSYNC:
2029 sb = freeze_bdev(sb->s_bdev);
2035 f2fs_stop_checkpoint(sbi, false);
2036 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2037 thaw_bdev(sb->s_bdev, sb);
2040 case F2FS_GOING_DOWN_METASYNC:
2041 /* do checkpoint only */
2042 ret = f2fs_sync_fs(sb, 1);
2045 f2fs_stop_checkpoint(sbi, false);
2046 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2048 case F2FS_GOING_DOWN_NOSYNC:
2049 f2fs_stop_checkpoint(sbi, false);
2050 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2052 case F2FS_GOING_DOWN_METAFLUSH:
2053 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2054 f2fs_stop_checkpoint(sbi, false);
2055 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2057 case F2FS_GOING_DOWN_NEED_FSCK:
2058 set_sbi_flag(sbi, SBI_NEED_FSCK);
2059 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2060 set_sbi_flag(sbi, SBI_IS_DIRTY);
2061 /* do checkpoint only */
2062 ret = f2fs_sync_fs(sb, 1);
2069 f2fs_stop_gc_thread(sbi);
2070 f2fs_stop_discard_thread(sbi);
2072 f2fs_drop_discard_cmd(sbi);
2073 clear_opt(sbi, DISCARD);
2075 f2fs_update_time(sbi, REQ_TIME);
2077 if (in != F2FS_GOING_DOWN_FULLSYNC)
2078 mnt_drop_write_file(filp);
2080 trace_f2fs_shutdown(sbi, in, ret);
2085 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2087 struct inode *inode = file_inode(filp);
2088 struct super_block *sb = inode->i_sb;
2089 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2090 struct fstrim_range range;
2093 if (!capable(CAP_SYS_ADMIN))
2096 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2099 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2103 ret = mnt_want_write_file(filp);
2107 range.minlen = max((unsigned int)range.minlen,
2108 q->limits.discard_granularity);
2109 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2110 mnt_drop_write_file(filp);
2114 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2117 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2121 static bool uuid_is_nonzero(__u8 u[16])
2125 for (i = 0; i < 16; i++)
2131 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2133 struct inode *inode = file_inode(filp);
2135 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2138 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2140 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2143 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2145 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2147 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2150 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2152 struct inode *inode = file_inode(filp);
2153 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2156 if (!f2fs_sb_has_encrypt(sbi))
2159 err = mnt_want_write_file(filp);
2163 down_write(&sbi->sb_lock);
2165 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2168 /* update superblock with uuid */
2169 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2171 err = f2fs_commit_super(sbi, false);
2174 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2178 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2182 up_write(&sbi->sb_lock);
2183 mnt_drop_write_file(filp);
2187 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2189 struct inode *inode = file_inode(filp);
2190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2194 if (!capable(CAP_SYS_ADMIN))
2197 if (get_user(sync, (__u32 __user *)arg))
2200 if (f2fs_readonly(sbi->sb))
2203 ret = mnt_want_write_file(filp);
2208 if (!mutex_trylock(&sbi->gc_mutex)) {
2213 mutex_lock(&sbi->gc_mutex);
2216 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2218 mnt_drop_write_file(filp);
2222 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2224 struct inode *inode = file_inode(filp);
2225 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2226 struct f2fs_gc_range range;
2230 if (!capable(CAP_SYS_ADMIN))
2233 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2237 if (f2fs_readonly(sbi->sb))
2240 end = range.start + range.len;
2241 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2245 ret = mnt_want_write_file(filp);
2251 if (!mutex_trylock(&sbi->gc_mutex)) {
2256 mutex_lock(&sbi->gc_mutex);
2259 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2260 range.start += BLKS_PER_SEC(sbi);
2261 if (range.start <= end)
2264 mnt_drop_write_file(filp);
2268 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2270 struct inode *inode = file_inode(filp);
2271 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2274 if (!capable(CAP_SYS_ADMIN))
2277 if (f2fs_readonly(sbi->sb))
2280 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2281 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2285 ret = mnt_want_write_file(filp);
2289 ret = f2fs_sync_fs(sbi->sb, 1);
2291 mnt_drop_write_file(filp);
2295 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2297 struct f2fs_defragment *range)
2299 struct inode *inode = file_inode(filp);
2300 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2301 .m_seg_type = NO_CHECK_TYPE ,
2302 .m_may_create = false };
2303 struct extent_info ei = {0, 0, 0};
2304 pgoff_t pg_start, pg_end, next_pgofs;
2305 unsigned int blk_per_seg = sbi->blocks_per_seg;
2306 unsigned int total = 0, sec_num;
2307 block_t blk_end = 0;
2308 bool fragmented = false;
2311 /* if in-place-update policy is enabled, don't waste time here */
2312 if (f2fs_should_update_inplace(inode, NULL))
2315 pg_start = range->start >> PAGE_SHIFT;
2316 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2318 f2fs_balance_fs(sbi, true);
2322 /* writeback all dirty pages in the range */
2323 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2324 range->start + range->len - 1);
2329 * lookup mapping info in extent cache, skip defragmenting if physical
2330 * block addresses are continuous.
2332 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2333 if (ei.fofs + ei.len >= pg_end)
2337 map.m_lblk = pg_start;
2338 map.m_next_pgofs = &next_pgofs;
2341 * lookup mapping info in dnode page cache, skip defragmenting if all
2342 * physical block addresses are continuous even if there are hole(s)
2343 * in logical blocks.
2345 while (map.m_lblk < pg_end) {
2346 map.m_len = pg_end - map.m_lblk;
2347 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2351 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2352 map.m_lblk = next_pgofs;
2356 if (blk_end && blk_end != map.m_pblk)
2359 /* record total count of block that we're going to move */
2362 blk_end = map.m_pblk + map.m_len;
2364 map.m_lblk += map.m_len;
2370 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2373 * make sure there are enough free section for LFS allocation, this can
2374 * avoid defragment running in SSR mode when free section are allocated
2377 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2382 map.m_lblk = pg_start;
2383 map.m_len = pg_end - pg_start;
2386 while (map.m_lblk < pg_end) {
2391 map.m_len = pg_end - map.m_lblk;
2392 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2396 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2397 map.m_lblk = next_pgofs;
2401 set_inode_flag(inode, FI_DO_DEFRAG);
2404 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2407 page = f2fs_get_lock_data_page(inode, idx, true);
2409 err = PTR_ERR(page);
2413 set_page_dirty(page);
2414 f2fs_put_page(page, 1);
2423 if (idx < pg_end && cnt < blk_per_seg)
2426 clear_inode_flag(inode, FI_DO_DEFRAG);
2428 err = filemap_fdatawrite(inode->i_mapping);
2433 clear_inode_flag(inode, FI_DO_DEFRAG);
2435 inode_unlock(inode);
2437 range->len = (u64)total << PAGE_SHIFT;
2441 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2443 struct inode *inode = file_inode(filp);
2444 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2445 struct f2fs_defragment range;
2448 if (!capable(CAP_SYS_ADMIN))
2451 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2454 if (f2fs_readonly(sbi->sb))
2457 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2461 /* verify alignment of offset & size */
2462 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2465 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2466 sbi->max_file_blocks))
2469 err = mnt_want_write_file(filp);
2473 err = f2fs_defragment_range(sbi, filp, &range);
2474 mnt_drop_write_file(filp);
2476 f2fs_update_time(sbi, REQ_TIME);
2480 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2487 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2488 struct file *file_out, loff_t pos_out, size_t len)
2490 struct inode *src = file_inode(file_in);
2491 struct inode *dst = file_inode(file_out);
2492 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2493 size_t olen = len, dst_max_i_size = 0;
2497 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2498 src->i_sb != dst->i_sb)
2501 if (unlikely(f2fs_readonly(src->i_sb)))
2504 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2507 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2511 if (pos_in == pos_out)
2513 if (pos_out > pos_in && pos_out < pos_in + len)
2520 if (!inode_trylock(dst))
2525 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2528 olen = len = src->i_size - pos_in;
2529 if (pos_in + len == src->i_size)
2530 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2536 dst_osize = dst->i_size;
2537 if (pos_out + olen > dst->i_size)
2538 dst_max_i_size = pos_out + olen;
2540 /* verify the end result is block aligned */
2541 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2542 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2543 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2546 ret = f2fs_convert_inline_inode(src);
2550 ret = f2fs_convert_inline_inode(dst);
2554 /* write out all dirty pages from offset */
2555 ret = filemap_write_and_wait_range(src->i_mapping,
2556 pos_in, pos_in + len);
2560 ret = filemap_write_and_wait_range(dst->i_mapping,
2561 pos_out, pos_out + len);
2565 f2fs_balance_fs(sbi, true);
2567 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2570 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2575 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2576 pos_out >> F2FS_BLKSIZE_BITS,
2577 len >> F2FS_BLKSIZE_BITS, false);
2581 f2fs_i_size_write(dst, dst_max_i_size);
2582 else if (dst_osize != dst->i_size)
2583 f2fs_i_size_write(dst, dst_osize);
2585 f2fs_unlock_op(sbi);
2588 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2590 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2599 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2601 struct f2fs_move_range range;
2605 if (!(filp->f_mode & FMODE_READ) ||
2606 !(filp->f_mode & FMODE_WRITE))
2609 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2613 dst = fdget(range.dst_fd);
2617 if (!(dst.file->f_mode & FMODE_WRITE)) {
2622 err = mnt_want_write_file(filp);
2626 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2627 range.pos_out, range.len);
2629 mnt_drop_write_file(filp);
2633 if (copy_to_user((struct f2fs_move_range __user *)arg,
2634 &range, sizeof(range)))
2641 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2643 struct inode *inode = file_inode(filp);
2644 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2645 struct sit_info *sm = SIT_I(sbi);
2646 unsigned int start_segno = 0, end_segno = 0;
2647 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2648 struct f2fs_flush_device range;
2651 if (!capable(CAP_SYS_ADMIN))
2654 if (f2fs_readonly(sbi->sb))
2657 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2660 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2664 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2665 __is_large_section(sbi)) {
2666 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2667 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2671 ret = mnt_want_write_file(filp);
2675 if (range.dev_num != 0)
2676 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2677 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2679 start_segno = sm->last_victim[FLUSH_DEVICE];
2680 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2681 start_segno = dev_start_segno;
2682 end_segno = min(start_segno + range.segments, dev_end_segno);
2684 while (start_segno < end_segno) {
2685 if (!mutex_trylock(&sbi->gc_mutex)) {
2689 sm->last_victim[GC_CB] = end_segno + 1;
2690 sm->last_victim[GC_GREEDY] = end_segno + 1;
2691 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2692 ret = f2fs_gc(sbi, true, true, start_segno);
2700 mnt_drop_write_file(filp);
2704 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2706 struct inode *inode = file_inode(filp);
2707 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2709 /* Must validate to set it with SQLite behavior in Android. */
2710 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2712 return put_user(sb_feature, (u32 __user *)arg);
2716 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2718 struct dquot *transfer_to[MAXQUOTAS] = {};
2719 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2720 struct super_block *sb = sbi->sb;
2723 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2724 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2725 err = __dquot_transfer(inode, transfer_to);
2727 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2728 dqput(transfer_to[PRJQUOTA]);
2733 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2735 struct inode *inode = file_inode(filp);
2736 struct f2fs_inode_info *fi = F2FS_I(inode);
2737 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2742 if (!f2fs_sb_has_project_quota(sbi)) {
2743 if (projid != F2FS_DEF_PROJID)
2749 if (!f2fs_has_extra_attr(inode))
2752 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2754 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2758 /* Is it quota file? Do not allow user to mess with it */
2759 if (IS_NOQUOTA(inode))
2762 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2764 return PTR_ERR(ipage);
2766 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2769 f2fs_put_page(ipage, 1);
2772 f2fs_put_page(ipage, 1);
2774 err = dquot_initialize(inode);
2779 err = f2fs_transfer_project_quota(inode, kprojid);
2783 F2FS_I(inode)->i_projid = kprojid;
2784 inode->i_ctime = current_time(inode);
2785 f2fs_mark_inode_dirty_sync(inode, true);
2787 f2fs_unlock_op(sbi);
2791 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2796 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2798 if (projid != F2FS_DEF_PROJID)
2804 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
2807 * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
2808 * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
2809 * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
2812 static const struct {
2815 } f2fs_xflags_map[] = {
2816 { F2FS_SYNC_FL, FS_XFLAG_SYNC },
2817 { F2FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE },
2818 { F2FS_APPEND_FL, FS_XFLAG_APPEND },
2819 { F2FS_NODUMP_FL, FS_XFLAG_NODUMP },
2820 { F2FS_NOATIME_FL, FS_XFLAG_NOATIME },
2821 { F2FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT },
2824 #define F2FS_SUPPORTED_XFLAGS ( \
2826 FS_XFLAG_IMMUTABLE | \
2829 FS_XFLAG_NOATIME | \
2830 FS_XFLAG_PROJINHERIT)
2832 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
2833 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
2838 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2839 if (iflags & f2fs_xflags_map[i].iflag)
2840 xflags |= f2fs_xflags_map[i].xflag;
2845 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
2846 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
2851 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2852 if (xflags & f2fs_xflags_map[i].xflag)
2853 iflags |= f2fs_xflags_map[i].iflag;
2858 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
2860 struct f2fs_inode_info *fi = F2FS_I(inode);
2862 simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
2864 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2865 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2868 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2870 struct inode *inode = file_inode(filp);
2873 f2fs_fill_fsxattr(inode, &fa);
2875 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2880 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2882 struct inode *inode = file_inode(filp);
2883 struct fsxattr fa, old_fa;
2887 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2890 /* Make sure caller has proper permission */
2891 if (!inode_owner_or_capable(inode))
2894 if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
2897 iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2898 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2901 err = mnt_want_write_file(filp);
2907 f2fs_fill_fsxattr(inode, &old_fa);
2908 err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
2912 err = f2fs_setflags_common(inode, iflags,
2913 f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
2917 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2919 inode_unlock(inode);
2920 mnt_drop_write_file(filp);
2924 int f2fs_pin_file_control(struct inode *inode, bool inc)
2926 struct f2fs_inode_info *fi = F2FS_I(inode);
2927 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2929 /* Use i_gc_failures for normal file as a risk signal. */
2931 f2fs_i_gc_failures_write(inode,
2932 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2934 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2935 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
2936 __func__, inode->i_ino,
2937 fi->i_gc_failures[GC_FAILURE_PIN]);
2938 clear_inode_flag(inode, FI_PIN_FILE);
2944 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2946 struct inode *inode = file_inode(filp);
2950 if (get_user(pin, (__u32 __user *)arg))
2953 if (!S_ISREG(inode->i_mode))
2956 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2959 ret = mnt_want_write_file(filp);
2965 if (f2fs_should_update_outplace(inode, NULL)) {
2971 clear_inode_flag(inode, FI_PIN_FILE);
2972 f2fs_i_gc_failures_write(inode, 0);
2976 if (f2fs_pin_file_control(inode, false)) {
2980 ret = f2fs_convert_inline_inode(inode);
2984 set_inode_flag(inode, FI_PIN_FILE);
2985 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2987 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2989 inode_unlock(inode);
2990 mnt_drop_write_file(filp);
2994 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2996 struct inode *inode = file_inode(filp);
2999 if (is_inode_flag_set(inode, FI_PIN_FILE))
3000 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3001 return put_user(pin, (u32 __user *)arg);
3004 int f2fs_precache_extents(struct inode *inode)
3006 struct f2fs_inode_info *fi = F2FS_I(inode);
3007 struct f2fs_map_blocks map;
3008 pgoff_t m_next_extent;
3012 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3016 map.m_next_pgofs = NULL;
3017 map.m_next_extent = &m_next_extent;
3018 map.m_seg_type = NO_CHECK_TYPE;
3019 map.m_may_create = false;
3020 end = F2FS_I_SB(inode)->max_file_blocks;
3022 while (map.m_lblk < end) {
3023 map.m_len = end - map.m_lblk;
3025 down_write(&fi->i_gc_rwsem[WRITE]);
3026 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3027 up_write(&fi->i_gc_rwsem[WRITE]);
3031 map.m_lblk = m_next_extent;
3037 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3039 return f2fs_precache_extents(file_inode(filp));
3042 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3044 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3048 if (!capable(CAP_SYS_ADMIN))
3051 if (f2fs_readonly(sbi->sb))
3054 if (copy_from_user(&block_count, (void __user *)arg,
3055 sizeof(block_count)))
3058 ret = f2fs_resize_fs(sbi, block_count);
3063 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3065 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3069 case F2FS_IOC_GETFLAGS:
3070 return f2fs_ioc_getflags(filp, arg);
3071 case F2FS_IOC_SETFLAGS:
3072 return f2fs_ioc_setflags(filp, arg);
3073 case F2FS_IOC_GETVERSION:
3074 return f2fs_ioc_getversion(filp, arg);
3075 case F2FS_IOC_START_ATOMIC_WRITE:
3076 return f2fs_ioc_start_atomic_write(filp);
3077 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3078 return f2fs_ioc_commit_atomic_write(filp);
3079 case F2FS_IOC_START_VOLATILE_WRITE:
3080 return f2fs_ioc_start_volatile_write(filp);
3081 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3082 return f2fs_ioc_release_volatile_write(filp);
3083 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3084 return f2fs_ioc_abort_volatile_write(filp);
3085 case F2FS_IOC_SHUTDOWN:
3086 return f2fs_ioc_shutdown(filp, arg);
3088 return f2fs_ioc_fitrim(filp, arg);
3089 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3090 return f2fs_ioc_set_encryption_policy(filp, arg);
3091 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3092 return f2fs_ioc_get_encryption_policy(filp, arg);
3093 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3094 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3095 case F2FS_IOC_GARBAGE_COLLECT:
3096 return f2fs_ioc_gc(filp, arg);
3097 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3098 return f2fs_ioc_gc_range(filp, arg);
3099 case F2FS_IOC_WRITE_CHECKPOINT:
3100 return f2fs_ioc_write_checkpoint(filp, arg);
3101 case F2FS_IOC_DEFRAGMENT:
3102 return f2fs_ioc_defragment(filp, arg);
3103 case F2FS_IOC_MOVE_RANGE:
3104 return f2fs_ioc_move_range(filp, arg);
3105 case F2FS_IOC_FLUSH_DEVICE:
3106 return f2fs_ioc_flush_device(filp, arg);
3107 case F2FS_IOC_GET_FEATURES:
3108 return f2fs_ioc_get_features(filp, arg);
3109 case F2FS_IOC_FSGETXATTR:
3110 return f2fs_ioc_fsgetxattr(filp, arg);
3111 case F2FS_IOC_FSSETXATTR:
3112 return f2fs_ioc_fssetxattr(filp, arg);
3113 case F2FS_IOC_GET_PIN_FILE:
3114 return f2fs_ioc_get_pin_file(filp, arg);
3115 case F2FS_IOC_SET_PIN_FILE:
3116 return f2fs_ioc_set_pin_file(filp, arg);
3117 case F2FS_IOC_PRECACHE_EXTENTS:
3118 return f2fs_ioc_precache_extents(filp, arg);
3119 case F2FS_IOC_RESIZE_FS:
3120 return f2fs_ioc_resize_fs(filp, arg);
3126 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3128 struct file *file = iocb->ki_filp;
3129 struct inode *inode = file_inode(file);
3132 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3137 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT)) {
3142 if (!inode_trylock(inode)) {
3143 if (iocb->ki_flags & IOCB_NOWAIT) {
3150 ret = generic_write_checks(iocb, from);
3152 bool preallocated = false;
3153 size_t target_size = 0;
3156 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3157 set_inode_flag(inode, FI_NO_PREALLOC);
3159 if ((iocb->ki_flags & IOCB_NOWAIT)) {
3160 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3161 iov_iter_count(from)) ||
3162 f2fs_has_inline_data(inode) ||
3163 f2fs_force_buffered_io(inode, iocb, from)) {
3164 clear_inode_flag(inode, FI_NO_PREALLOC);
3165 inode_unlock(inode);
3170 preallocated = true;
3171 target_size = iocb->ki_pos + iov_iter_count(from);
3173 err = f2fs_preallocate_blocks(iocb, from);
3175 clear_inode_flag(inode, FI_NO_PREALLOC);
3176 inode_unlock(inode);
3181 ret = __generic_file_write_iter(iocb, from);
3182 clear_inode_flag(inode, FI_NO_PREALLOC);
3184 /* if we couldn't write data, we should deallocate blocks. */
3185 if (preallocated && i_size_read(inode) < target_size)
3186 f2fs_truncate(inode);
3189 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3191 inode_unlock(inode);
3193 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3194 iov_iter_count(from), ret);
3196 ret = generic_write_sync(iocb, ret);
3200 #ifdef CONFIG_COMPAT
3201 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3204 case F2FS_IOC32_GETFLAGS:
3205 cmd = F2FS_IOC_GETFLAGS;
3207 case F2FS_IOC32_SETFLAGS:
3208 cmd = F2FS_IOC_SETFLAGS;
3210 case F2FS_IOC32_GETVERSION:
3211 cmd = F2FS_IOC_GETVERSION;
3213 case F2FS_IOC_START_ATOMIC_WRITE:
3214 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3215 case F2FS_IOC_START_VOLATILE_WRITE:
3216 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3217 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3218 case F2FS_IOC_SHUTDOWN:
3219 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3220 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3221 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3222 case F2FS_IOC_GARBAGE_COLLECT:
3223 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3224 case F2FS_IOC_WRITE_CHECKPOINT:
3225 case F2FS_IOC_DEFRAGMENT:
3226 case F2FS_IOC_MOVE_RANGE:
3227 case F2FS_IOC_FLUSH_DEVICE:
3228 case F2FS_IOC_GET_FEATURES:
3229 case F2FS_IOC_FSGETXATTR:
3230 case F2FS_IOC_FSSETXATTR:
3231 case F2FS_IOC_GET_PIN_FILE:
3232 case F2FS_IOC_SET_PIN_FILE:
3233 case F2FS_IOC_PRECACHE_EXTENTS:
3234 case F2FS_IOC_RESIZE_FS:
3237 return -ENOIOCTLCMD;
3239 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3243 const struct file_operations f2fs_file_operations = {
3244 .llseek = f2fs_llseek,
3245 .read_iter = generic_file_read_iter,
3246 .write_iter = f2fs_file_write_iter,
3247 .open = f2fs_file_open,
3248 .release = f2fs_release_file,
3249 .mmap = f2fs_file_mmap,
3250 .flush = f2fs_file_flush,
3251 .fsync = f2fs_sync_file,
3252 .fallocate = f2fs_fallocate,
3253 .unlocked_ioctl = f2fs_ioctl,
3254 #ifdef CONFIG_COMPAT
3255 .compat_ioctl = f2fs_compat_ioctl,
3257 .splice_read = generic_file_splice_read,
3258 .splice_write = iter_file_splice_write,