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 = timespec64_trunc(attr->ia_atime,
750 inode->i_sb->s_time_gran);
751 if (ia_valid & ATTR_MTIME)
752 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
753 inode->i_sb->s_time_gran);
754 if (ia_valid & ATTR_CTIME)
755 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
756 inode->i_sb->s_time_gran);
757 if (ia_valid & ATTR_MODE) {
758 umode_t mode = attr->ia_mode;
760 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
762 set_acl_inode(inode, mode);
766 #define __setattr_copy setattr_copy
769 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
771 struct inode *inode = d_inode(dentry);
774 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
777 err = setattr_prepare(dentry, attr);
781 err = fscrypt_prepare_setattr(dentry, attr);
785 err = fsverity_prepare_setattr(dentry, attr);
789 if (is_quota_modification(inode, attr)) {
790 err = dquot_initialize(inode);
794 if ((attr->ia_valid & ATTR_UID &&
795 !uid_eq(attr->ia_uid, inode->i_uid)) ||
796 (attr->ia_valid & ATTR_GID &&
797 !gid_eq(attr->ia_gid, inode->i_gid))) {
798 f2fs_lock_op(F2FS_I_SB(inode));
799 err = dquot_transfer(inode, attr);
801 set_sbi_flag(F2FS_I_SB(inode),
802 SBI_QUOTA_NEED_REPAIR);
803 f2fs_unlock_op(F2FS_I_SB(inode));
807 * update uid/gid under lock_op(), so that dquot and inode can
808 * be updated atomically.
810 if (attr->ia_valid & ATTR_UID)
811 inode->i_uid = attr->ia_uid;
812 if (attr->ia_valid & ATTR_GID)
813 inode->i_gid = attr->ia_gid;
814 f2fs_mark_inode_dirty_sync(inode, true);
815 f2fs_unlock_op(F2FS_I_SB(inode));
818 if (attr->ia_valid & ATTR_SIZE) {
819 bool to_smaller = (attr->ia_size <= i_size_read(inode));
821 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
822 down_write(&F2FS_I(inode)->i_mmap_sem);
824 truncate_setsize(inode, attr->ia_size);
827 err = f2fs_truncate(inode);
829 * do not trim all blocks after i_size if target size is
830 * larger than i_size.
832 up_write(&F2FS_I(inode)->i_mmap_sem);
833 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
839 /* should convert inline inode here */
840 if (!f2fs_may_inline_data(inode)) {
841 err = f2fs_convert_inline_inode(inode);
845 inode->i_mtime = inode->i_ctime = current_time(inode);
848 down_write(&F2FS_I(inode)->i_sem);
849 F2FS_I(inode)->last_disk_size = i_size_read(inode);
850 up_write(&F2FS_I(inode)->i_sem);
853 __setattr_copy(inode, attr);
855 if (attr->ia_valid & ATTR_MODE) {
856 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
857 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
858 inode->i_mode = F2FS_I(inode)->i_acl_mode;
859 clear_inode_flag(inode, FI_ACL_MODE);
863 /* file size may changed here */
864 f2fs_mark_inode_dirty_sync(inode, true);
866 /* inode change will produce dirty node pages flushed by checkpoint */
867 f2fs_balance_fs(F2FS_I_SB(inode), true);
872 const struct inode_operations f2fs_file_inode_operations = {
873 .getattr = f2fs_getattr,
874 .setattr = f2fs_setattr,
875 .get_acl = f2fs_get_acl,
876 .set_acl = f2fs_set_acl,
877 #ifdef CONFIG_F2FS_FS_XATTR
878 .listxattr = f2fs_listxattr,
880 .fiemap = f2fs_fiemap,
883 static int fill_zero(struct inode *inode, pgoff_t index,
884 loff_t start, loff_t len)
886 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
892 f2fs_balance_fs(sbi, true);
895 page = f2fs_get_new_data_page(inode, NULL, index, false);
899 return PTR_ERR(page);
901 f2fs_wait_on_page_writeback(page, DATA, true, true);
902 zero_user(page, start, len);
903 set_page_dirty(page);
904 f2fs_put_page(page, 1);
908 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
912 while (pg_start < pg_end) {
913 struct dnode_of_data dn;
914 pgoff_t end_offset, count;
916 set_new_dnode(&dn, inode, NULL, NULL, 0);
917 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
919 if (err == -ENOENT) {
920 pg_start = f2fs_get_next_page_offset(&dn,
927 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
928 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
930 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
932 f2fs_truncate_data_blocks_range(&dn, count);
940 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
942 pgoff_t pg_start, pg_end;
943 loff_t off_start, off_end;
946 ret = f2fs_convert_inline_inode(inode);
950 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
951 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
953 off_start = offset & (PAGE_SIZE - 1);
954 off_end = (offset + len) & (PAGE_SIZE - 1);
956 if (pg_start == pg_end) {
957 ret = fill_zero(inode, pg_start, off_start,
958 off_end - off_start);
963 ret = fill_zero(inode, pg_start++, off_start,
964 PAGE_SIZE - off_start);
969 ret = fill_zero(inode, pg_end, 0, off_end);
974 if (pg_start < pg_end) {
975 struct address_space *mapping = inode->i_mapping;
976 loff_t blk_start, blk_end;
977 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
979 f2fs_balance_fs(sbi, true);
981 blk_start = (loff_t)pg_start << PAGE_SHIFT;
982 blk_end = (loff_t)pg_end << PAGE_SHIFT;
984 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
985 down_write(&F2FS_I(inode)->i_mmap_sem);
987 truncate_inode_pages_range(mapping, blk_start,
991 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
994 up_write(&F2FS_I(inode)->i_mmap_sem);
995 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1002 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1003 int *do_replace, pgoff_t off, pgoff_t len)
1005 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1006 struct dnode_of_data dn;
1010 set_new_dnode(&dn, inode, NULL, NULL, 0);
1011 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1012 if (ret && ret != -ENOENT) {
1014 } else if (ret == -ENOENT) {
1015 if (dn.max_level == 0)
1017 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1024 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1025 dn.ofs_in_node, len);
1026 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1027 *blkaddr = datablock_addr(dn.inode,
1028 dn.node_page, dn.ofs_in_node);
1030 if (__is_valid_data_blkaddr(*blkaddr) &&
1031 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1032 DATA_GENERIC_ENHANCE)) {
1033 f2fs_put_dnode(&dn);
1034 return -EFSCORRUPTED;
1037 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1039 if (test_opt(sbi, LFS)) {
1040 f2fs_put_dnode(&dn);
1044 /* do not invalidate this block address */
1045 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1049 f2fs_put_dnode(&dn);
1058 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1059 int *do_replace, pgoff_t off, int len)
1061 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1062 struct dnode_of_data dn;
1065 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1066 if (*do_replace == 0)
1069 set_new_dnode(&dn, inode, NULL, NULL, 0);
1070 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1072 dec_valid_block_count(sbi, inode, 1);
1073 f2fs_invalidate_blocks(sbi, *blkaddr);
1075 f2fs_update_data_blkaddr(&dn, *blkaddr);
1077 f2fs_put_dnode(&dn);
1082 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1083 block_t *blkaddr, int *do_replace,
1084 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1086 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1091 if (blkaddr[i] == NULL_ADDR && !full) {
1096 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1097 struct dnode_of_data dn;
1098 struct node_info ni;
1102 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1103 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1107 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1109 f2fs_put_dnode(&dn);
1113 ilen = min((pgoff_t)
1114 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1115 dn.ofs_in_node, len - i);
1117 dn.data_blkaddr = datablock_addr(dn.inode,
1118 dn.node_page, dn.ofs_in_node);
1119 f2fs_truncate_data_blocks_range(&dn, 1);
1121 if (do_replace[i]) {
1122 f2fs_i_blocks_write(src_inode,
1124 f2fs_i_blocks_write(dst_inode,
1126 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1127 blkaddr[i], ni.version, true, false);
1133 new_size = (dst + i) << PAGE_SHIFT;
1134 if (dst_inode->i_size < new_size)
1135 f2fs_i_size_write(dst_inode, new_size);
1136 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1138 f2fs_put_dnode(&dn);
1140 struct page *psrc, *pdst;
1142 psrc = f2fs_get_lock_data_page(src_inode,
1145 return PTR_ERR(psrc);
1146 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1149 f2fs_put_page(psrc, 1);
1150 return PTR_ERR(pdst);
1152 f2fs_copy_page(psrc, pdst);
1153 set_page_dirty(pdst);
1154 f2fs_put_page(pdst, 1);
1155 f2fs_put_page(psrc, 1);
1157 ret = f2fs_truncate_hole(src_inode,
1158 src + i, src + i + 1);
1167 static int __exchange_data_block(struct inode *src_inode,
1168 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1169 pgoff_t len, bool full)
1171 block_t *src_blkaddr;
1177 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1179 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1180 array_size(olen, sizeof(block_t)),
1185 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1186 array_size(olen, sizeof(int)),
1189 kvfree(src_blkaddr);
1193 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1194 do_replace, src, olen);
1198 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1199 do_replace, src, dst, olen, full);
1207 kvfree(src_blkaddr);
1213 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1214 kvfree(src_blkaddr);
1219 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1221 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1222 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1223 pgoff_t start = offset >> PAGE_SHIFT;
1224 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1227 f2fs_balance_fs(sbi, true);
1229 /* avoid gc operation during block exchange */
1230 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1231 down_write(&F2FS_I(inode)->i_mmap_sem);
1234 f2fs_drop_extent_tree(inode);
1235 truncate_pagecache(inode, offset);
1236 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1237 f2fs_unlock_op(sbi);
1239 up_write(&F2FS_I(inode)->i_mmap_sem);
1240 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1244 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1249 if (offset + len >= i_size_read(inode))
1252 /* collapse range should be aligned to block size of f2fs. */
1253 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1256 ret = f2fs_convert_inline_inode(inode);
1260 /* write out all dirty pages from offset */
1261 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1265 ret = f2fs_do_collapse(inode, offset, len);
1269 /* write out all moved pages, if possible */
1270 down_write(&F2FS_I(inode)->i_mmap_sem);
1271 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1272 truncate_pagecache(inode, offset);
1274 new_size = i_size_read(inode) - len;
1275 truncate_pagecache(inode, new_size);
1277 ret = f2fs_truncate_blocks(inode, new_size, true);
1278 up_write(&F2FS_I(inode)->i_mmap_sem);
1280 f2fs_i_size_write(inode, new_size);
1284 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1287 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1288 pgoff_t index = start;
1289 unsigned int ofs_in_node = dn->ofs_in_node;
1293 for (; index < end; index++, dn->ofs_in_node++) {
1294 if (datablock_addr(dn->inode, dn->node_page,
1295 dn->ofs_in_node) == NULL_ADDR)
1299 dn->ofs_in_node = ofs_in_node;
1300 ret = f2fs_reserve_new_blocks(dn, count);
1304 dn->ofs_in_node = ofs_in_node;
1305 for (index = start; index < end; index++, dn->ofs_in_node++) {
1306 dn->data_blkaddr = datablock_addr(dn->inode,
1307 dn->node_page, dn->ofs_in_node);
1309 * f2fs_reserve_new_blocks will not guarantee entire block
1312 if (dn->data_blkaddr == NULL_ADDR) {
1316 if (dn->data_blkaddr != NEW_ADDR) {
1317 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1318 dn->data_blkaddr = NEW_ADDR;
1319 f2fs_set_data_blkaddr(dn);
1323 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1328 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1331 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1332 struct address_space *mapping = inode->i_mapping;
1333 pgoff_t index, pg_start, pg_end;
1334 loff_t new_size = i_size_read(inode);
1335 loff_t off_start, off_end;
1338 ret = inode_newsize_ok(inode, (len + offset));
1342 ret = f2fs_convert_inline_inode(inode);
1346 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1350 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1351 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1353 off_start = offset & (PAGE_SIZE - 1);
1354 off_end = (offset + len) & (PAGE_SIZE - 1);
1356 if (pg_start == pg_end) {
1357 ret = fill_zero(inode, pg_start, off_start,
1358 off_end - off_start);
1362 new_size = max_t(loff_t, new_size, offset + len);
1365 ret = fill_zero(inode, pg_start++, off_start,
1366 PAGE_SIZE - off_start);
1370 new_size = max_t(loff_t, new_size,
1371 (loff_t)pg_start << PAGE_SHIFT);
1374 for (index = pg_start; index < pg_end;) {
1375 struct dnode_of_data dn;
1376 unsigned int end_offset;
1379 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1380 down_write(&F2FS_I(inode)->i_mmap_sem);
1382 truncate_pagecache_range(inode,
1383 (loff_t)index << PAGE_SHIFT,
1384 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1388 set_new_dnode(&dn, inode, NULL, NULL, 0);
1389 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1391 f2fs_unlock_op(sbi);
1392 up_write(&F2FS_I(inode)->i_mmap_sem);
1393 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1397 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1398 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1400 ret = f2fs_do_zero_range(&dn, index, end);
1401 f2fs_put_dnode(&dn);
1403 f2fs_unlock_op(sbi);
1404 up_write(&F2FS_I(inode)->i_mmap_sem);
1405 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1407 f2fs_balance_fs(sbi, dn.node_changed);
1413 new_size = max_t(loff_t, new_size,
1414 (loff_t)index << PAGE_SHIFT);
1418 ret = fill_zero(inode, pg_end, 0, off_end);
1422 new_size = max_t(loff_t, new_size, offset + len);
1427 if (new_size > i_size_read(inode)) {
1428 if (mode & FALLOC_FL_KEEP_SIZE)
1429 file_set_keep_isize(inode);
1431 f2fs_i_size_write(inode, new_size);
1436 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1438 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1439 pgoff_t nr, pg_start, pg_end, delta, idx;
1443 new_size = i_size_read(inode) + len;
1444 ret = inode_newsize_ok(inode, new_size);
1448 if (offset >= i_size_read(inode))
1451 /* insert range should be aligned to block size of f2fs. */
1452 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1455 ret = f2fs_convert_inline_inode(inode);
1459 f2fs_balance_fs(sbi, true);
1461 down_write(&F2FS_I(inode)->i_mmap_sem);
1462 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1463 up_write(&F2FS_I(inode)->i_mmap_sem);
1467 /* write out all dirty pages from offset */
1468 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1472 pg_start = offset >> PAGE_SHIFT;
1473 pg_end = (offset + len) >> PAGE_SHIFT;
1474 delta = pg_end - pg_start;
1475 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1477 /* avoid gc operation during block exchange */
1478 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1479 down_write(&F2FS_I(inode)->i_mmap_sem);
1480 truncate_pagecache(inode, offset);
1482 while (!ret && idx > pg_start) {
1483 nr = idx - pg_start;
1489 f2fs_drop_extent_tree(inode);
1491 ret = __exchange_data_block(inode, inode, idx,
1492 idx + delta, nr, false);
1493 f2fs_unlock_op(sbi);
1495 up_write(&F2FS_I(inode)->i_mmap_sem);
1496 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1498 /* write out all moved pages, if possible */
1499 down_write(&F2FS_I(inode)->i_mmap_sem);
1500 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1501 truncate_pagecache(inode, offset);
1502 up_write(&F2FS_I(inode)->i_mmap_sem);
1505 f2fs_i_size_write(inode, new_size);
1509 static int expand_inode_data(struct inode *inode, loff_t offset,
1510 loff_t len, int mode)
1512 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1513 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1514 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1515 .m_may_create = true };
1517 loff_t new_size = i_size_read(inode);
1521 err = inode_newsize_ok(inode, (len + offset));
1525 err = f2fs_convert_inline_inode(inode);
1529 f2fs_balance_fs(sbi, true);
1531 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1532 off_end = (offset + len) & (PAGE_SIZE - 1);
1534 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1535 map.m_len = pg_end - map.m_lblk;
1539 if (f2fs_is_pinned_file(inode))
1540 map.m_seg_type = CURSEG_COLD_DATA;
1542 err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
1543 F2FS_GET_BLOCK_PRE_DIO :
1544 F2FS_GET_BLOCK_PRE_AIO));
1551 last_off = map.m_lblk + map.m_len - 1;
1553 /* update new size to the failed position */
1554 new_size = (last_off == pg_end) ? offset + len :
1555 (loff_t)(last_off + 1) << PAGE_SHIFT;
1557 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1560 if (new_size > i_size_read(inode)) {
1561 if (mode & FALLOC_FL_KEEP_SIZE)
1562 file_set_keep_isize(inode);
1564 f2fs_i_size_write(inode, new_size);
1570 static long f2fs_fallocate(struct file *file, int mode,
1571 loff_t offset, loff_t len)
1573 struct inode *inode = file_inode(file);
1576 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1579 /* f2fs only support ->fallocate for regular file */
1580 if (!S_ISREG(inode->i_mode))
1583 if (IS_ENCRYPTED(inode) &&
1584 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1587 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1588 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1589 FALLOC_FL_INSERT_RANGE))
1594 if (mode & FALLOC_FL_PUNCH_HOLE) {
1595 if (offset >= inode->i_size)
1598 ret = punch_hole(inode, offset, len);
1599 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1600 ret = f2fs_collapse_range(inode, offset, len);
1601 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1602 ret = f2fs_zero_range(inode, offset, len, mode);
1603 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1604 ret = f2fs_insert_range(inode, offset, len);
1606 ret = expand_inode_data(inode, offset, len, mode);
1610 inode->i_mtime = inode->i_ctime = current_time(inode);
1611 f2fs_mark_inode_dirty_sync(inode, false);
1612 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1616 inode_unlock(inode);
1618 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1622 static int f2fs_release_file(struct inode *inode, struct file *filp)
1625 * f2fs_relase_file is called at every close calls. So we should
1626 * not drop any inmemory pages by close called by other process.
1628 if (!(filp->f_mode & FMODE_WRITE) ||
1629 atomic_read(&inode->i_writecount) != 1)
1632 /* some remained atomic pages should discarded */
1633 if (f2fs_is_atomic_file(inode))
1634 f2fs_drop_inmem_pages(inode);
1635 if (f2fs_is_volatile_file(inode)) {
1636 set_inode_flag(inode, FI_DROP_CACHE);
1637 filemap_fdatawrite(inode->i_mapping);
1638 clear_inode_flag(inode, FI_DROP_CACHE);
1639 clear_inode_flag(inode, FI_VOLATILE_FILE);
1640 stat_dec_volatile_write(inode);
1645 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1647 struct inode *inode = file_inode(file);
1650 * If the process doing a transaction is crashed, we should do
1651 * roll-back. Otherwise, other reader/write can see corrupted database
1652 * until all the writers close its file. Since this should be done
1653 * before dropping file lock, it needs to do in ->flush.
1655 if (f2fs_is_atomic_file(inode) &&
1656 F2FS_I(inode)->inmem_task == current)
1657 f2fs_drop_inmem_pages(inode);
1661 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1663 struct f2fs_inode_info *fi = F2FS_I(inode);
1665 /* Is it quota file? Do not allow user to mess with it */
1666 if (IS_NOQUOTA(inode))
1669 fi->i_flags = iflags | (fi->i_flags & ~mask);
1671 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1672 set_inode_flag(inode, FI_PROJ_INHERIT);
1674 clear_inode_flag(inode, FI_PROJ_INHERIT);
1676 inode->i_ctime = current_time(inode);
1677 f2fs_set_inode_flags(inode);
1678 f2fs_mark_inode_dirty_sync(inode, true);
1682 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1685 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1686 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1687 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1688 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1691 static const struct {
1694 } f2fs_fsflags_map[] = {
1695 { F2FS_SYNC_FL, FS_SYNC_FL },
1696 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1697 { F2FS_APPEND_FL, FS_APPEND_FL },
1698 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1699 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1700 { F2FS_INDEX_FL, FS_INDEX_FL },
1701 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1702 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1705 #define F2FS_GETTABLE_FS_FL ( \
1713 FS_PROJINHERIT_FL | \
1715 FS_INLINE_DATA_FL | \
1719 #define F2FS_SETTABLE_FS_FL ( \
1728 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1729 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1734 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1735 if (iflags & f2fs_fsflags_map[i].iflag)
1736 fsflags |= f2fs_fsflags_map[i].fsflag;
1741 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1742 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1747 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1748 if (fsflags & f2fs_fsflags_map[i].fsflag)
1749 iflags |= f2fs_fsflags_map[i].iflag;
1754 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1756 struct inode *inode = file_inode(filp);
1757 struct f2fs_inode_info *fi = F2FS_I(inode);
1758 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1760 if (IS_ENCRYPTED(inode))
1761 fsflags |= FS_ENCRYPT_FL;
1762 if (IS_VERITY(inode))
1763 fsflags |= FS_VERITY_FL;
1764 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1765 fsflags |= FS_INLINE_DATA_FL;
1766 if (is_inode_flag_set(inode, FI_PIN_FILE))
1767 fsflags |= FS_NOCOW_FL;
1769 fsflags &= F2FS_GETTABLE_FS_FL;
1771 return put_user(fsflags, (int __user *)arg);
1774 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1776 struct inode *inode = file_inode(filp);
1777 struct f2fs_inode_info *fi = F2FS_I(inode);
1778 u32 fsflags, old_fsflags;
1782 if (!inode_owner_or_capable(inode))
1785 if (get_user(fsflags, (int __user *)arg))
1788 if (fsflags & ~F2FS_GETTABLE_FS_FL)
1790 fsflags &= F2FS_SETTABLE_FS_FL;
1792 iflags = f2fs_fsflags_to_iflags(fsflags);
1793 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1796 ret = mnt_want_write_file(filp);
1802 old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1803 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1807 ret = f2fs_setflags_common(inode, iflags,
1808 f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1810 inode_unlock(inode);
1811 mnt_drop_write_file(filp);
1815 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1817 struct inode *inode = file_inode(filp);
1819 return put_user(inode->i_generation, (int __user *)arg);
1822 static int f2fs_ioc_start_atomic_write(struct file *filp)
1824 struct inode *inode = file_inode(filp);
1827 if (!inode_owner_or_capable(inode))
1830 if (!S_ISREG(inode->i_mode))
1833 ret = mnt_want_write_file(filp);
1839 if (f2fs_is_atomic_file(inode)) {
1840 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1845 ret = f2fs_convert_inline_inode(inode);
1849 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1852 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1853 * f2fs_is_atomic_file.
1855 if (get_dirty_pages(inode))
1856 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1857 inode->i_ino, get_dirty_pages(inode));
1858 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1860 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1864 set_inode_flag(inode, FI_ATOMIC_FILE);
1865 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1866 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1868 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1869 F2FS_I(inode)->inmem_task = current;
1870 stat_inc_atomic_write(inode);
1871 stat_update_max_atomic_write(inode);
1873 inode_unlock(inode);
1874 mnt_drop_write_file(filp);
1878 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1880 struct inode *inode = file_inode(filp);
1883 if (!inode_owner_or_capable(inode))
1886 ret = mnt_want_write_file(filp);
1890 f2fs_balance_fs(F2FS_I_SB(inode), true);
1894 if (f2fs_is_volatile_file(inode)) {
1899 if (f2fs_is_atomic_file(inode)) {
1900 ret = f2fs_commit_inmem_pages(inode);
1904 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1906 clear_inode_flag(inode, FI_ATOMIC_FILE);
1907 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1908 stat_dec_atomic_write(inode);
1911 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1914 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1915 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1918 inode_unlock(inode);
1919 mnt_drop_write_file(filp);
1923 static int f2fs_ioc_start_volatile_write(struct file *filp)
1925 struct inode *inode = file_inode(filp);
1928 if (!inode_owner_or_capable(inode))
1931 if (!S_ISREG(inode->i_mode))
1934 ret = mnt_want_write_file(filp);
1940 if (f2fs_is_volatile_file(inode))
1943 ret = f2fs_convert_inline_inode(inode);
1947 stat_inc_volatile_write(inode);
1948 stat_update_max_volatile_write(inode);
1950 set_inode_flag(inode, FI_VOLATILE_FILE);
1951 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1953 inode_unlock(inode);
1954 mnt_drop_write_file(filp);
1958 static int f2fs_ioc_release_volatile_write(struct file *filp)
1960 struct inode *inode = file_inode(filp);
1963 if (!inode_owner_or_capable(inode))
1966 ret = mnt_want_write_file(filp);
1972 if (!f2fs_is_volatile_file(inode))
1975 if (!f2fs_is_first_block_written(inode)) {
1976 ret = truncate_partial_data_page(inode, 0, true);
1980 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1982 inode_unlock(inode);
1983 mnt_drop_write_file(filp);
1987 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1989 struct inode *inode = file_inode(filp);
1992 if (!inode_owner_or_capable(inode))
1995 ret = mnt_want_write_file(filp);
2001 if (f2fs_is_atomic_file(inode))
2002 f2fs_drop_inmem_pages(inode);
2003 if (f2fs_is_volatile_file(inode)) {
2004 clear_inode_flag(inode, FI_VOLATILE_FILE);
2005 stat_dec_volatile_write(inode);
2006 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2009 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2011 inode_unlock(inode);
2013 mnt_drop_write_file(filp);
2014 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2018 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2020 struct inode *inode = file_inode(filp);
2021 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2022 struct super_block *sb = sbi->sb;
2026 if (!capable(CAP_SYS_ADMIN))
2029 if (get_user(in, (__u32 __user *)arg))
2032 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2033 ret = mnt_want_write_file(filp);
2039 case F2FS_GOING_DOWN_FULLSYNC:
2040 sb = freeze_bdev(sb->s_bdev);
2046 f2fs_stop_checkpoint(sbi, false);
2047 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2048 thaw_bdev(sb->s_bdev, sb);
2051 case F2FS_GOING_DOWN_METASYNC:
2052 /* do checkpoint only */
2053 ret = f2fs_sync_fs(sb, 1);
2056 f2fs_stop_checkpoint(sbi, false);
2057 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2059 case F2FS_GOING_DOWN_NOSYNC:
2060 f2fs_stop_checkpoint(sbi, false);
2061 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2063 case F2FS_GOING_DOWN_METAFLUSH:
2064 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2065 f2fs_stop_checkpoint(sbi, false);
2066 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2068 case F2FS_GOING_DOWN_NEED_FSCK:
2069 set_sbi_flag(sbi, SBI_NEED_FSCK);
2070 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2071 set_sbi_flag(sbi, SBI_IS_DIRTY);
2072 /* do checkpoint only */
2073 ret = f2fs_sync_fs(sb, 1);
2080 f2fs_stop_gc_thread(sbi);
2081 f2fs_stop_discard_thread(sbi);
2083 f2fs_drop_discard_cmd(sbi);
2084 clear_opt(sbi, DISCARD);
2086 f2fs_update_time(sbi, REQ_TIME);
2088 if (in != F2FS_GOING_DOWN_FULLSYNC)
2089 mnt_drop_write_file(filp);
2091 trace_f2fs_shutdown(sbi, in, ret);
2096 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2098 struct inode *inode = file_inode(filp);
2099 struct super_block *sb = inode->i_sb;
2100 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2101 struct fstrim_range range;
2104 if (!capable(CAP_SYS_ADMIN))
2107 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2110 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2114 ret = mnt_want_write_file(filp);
2118 range.minlen = max((unsigned int)range.minlen,
2119 q->limits.discard_granularity);
2120 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2121 mnt_drop_write_file(filp);
2125 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2128 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2132 static bool uuid_is_nonzero(__u8 u[16])
2136 for (i = 0; i < 16; i++)
2142 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2144 struct inode *inode = file_inode(filp);
2146 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2149 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2151 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2154 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2156 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2158 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2161 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2163 struct inode *inode = file_inode(filp);
2164 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2167 if (!f2fs_sb_has_encrypt(sbi))
2170 err = mnt_want_write_file(filp);
2174 down_write(&sbi->sb_lock);
2176 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2179 /* update superblock with uuid */
2180 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2182 err = f2fs_commit_super(sbi, false);
2185 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2189 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2193 up_write(&sbi->sb_lock);
2194 mnt_drop_write_file(filp);
2198 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2201 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2204 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2207 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2209 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2212 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2215 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2217 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2220 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2223 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2226 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2229 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2232 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2235 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2238 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2241 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2243 struct inode *inode = file_inode(filp);
2244 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2248 if (!capable(CAP_SYS_ADMIN))
2251 if (get_user(sync, (__u32 __user *)arg))
2254 if (f2fs_readonly(sbi->sb))
2257 ret = mnt_want_write_file(filp);
2262 if (!mutex_trylock(&sbi->gc_mutex)) {
2267 mutex_lock(&sbi->gc_mutex);
2270 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2272 mnt_drop_write_file(filp);
2276 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2278 struct inode *inode = file_inode(filp);
2279 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2280 struct f2fs_gc_range range;
2284 if (!capable(CAP_SYS_ADMIN))
2287 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2291 if (f2fs_readonly(sbi->sb))
2294 end = range.start + range.len;
2295 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2299 ret = mnt_want_write_file(filp);
2305 if (!mutex_trylock(&sbi->gc_mutex)) {
2310 mutex_lock(&sbi->gc_mutex);
2313 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2314 range.start += BLKS_PER_SEC(sbi);
2315 if (range.start <= end)
2318 mnt_drop_write_file(filp);
2322 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2324 struct inode *inode = file_inode(filp);
2325 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2328 if (!capable(CAP_SYS_ADMIN))
2331 if (f2fs_readonly(sbi->sb))
2334 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2335 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2339 ret = mnt_want_write_file(filp);
2343 ret = f2fs_sync_fs(sbi->sb, 1);
2345 mnt_drop_write_file(filp);
2349 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2351 struct f2fs_defragment *range)
2353 struct inode *inode = file_inode(filp);
2354 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2355 .m_seg_type = NO_CHECK_TYPE ,
2356 .m_may_create = false };
2357 struct extent_info ei = {0, 0, 0};
2358 pgoff_t pg_start, pg_end, next_pgofs;
2359 unsigned int blk_per_seg = sbi->blocks_per_seg;
2360 unsigned int total = 0, sec_num;
2361 block_t blk_end = 0;
2362 bool fragmented = false;
2365 /* if in-place-update policy is enabled, don't waste time here */
2366 if (f2fs_should_update_inplace(inode, NULL))
2369 pg_start = range->start >> PAGE_SHIFT;
2370 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2372 f2fs_balance_fs(sbi, true);
2376 /* writeback all dirty pages in the range */
2377 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2378 range->start + range->len - 1);
2383 * lookup mapping info in extent cache, skip defragmenting if physical
2384 * block addresses are continuous.
2386 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2387 if (ei.fofs + ei.len >= pg_end)
2391 map.m_lblk = pg_start;
2392 map.m_next_pgofs = &next_pgofs;
2395 * lookup mapping info in dnode page cache, skip defragmenting if all
2396 * physical block addresses are continuous even if there are hole(s)
2397 * in logical blocks.
2399 while (map.m_lblk < pg_end) {
2400 map.m_len = pg_end - map.m_lblk;
2401 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2405 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2406 map.m_lblk = next_pgofs;
2410 if (blk_end && blk_end != map.m_pblk)
2413 /* record total count of block that we're going to move */
2416 blk_end = map.m_pblk + map.m_len;
2418 map.m_lblk += map.m_len;
2424 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2427 * make sure there are enough free section for LFS allocation, this can
2428 * avoid defragment running in SSR mode when free section are allocated
2431 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2436 map.m_lblk = pg_start;
2437 map.m_len = pg_end - pg_start;
2440 while (map.m_lblk < pg_end) {
2445 map.m_len = pg_end - map.m_lblk;
2446 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2450 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2451 map.m_lblk = next_pgofs;
2455 set_inode_flag(inode, FI_DO_DEFRAG);
2458 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2461 page = f2fs_get_lock_data_page(inode, idx, true);
2463 err = PTR_ERR(page);
2467 set_page_dirty(page);
2468 f2fs_put_page(page, 1);
2477 if (idx < pg_end && cnt < blk_per_seg)
2480 clear_inode_flag(inode, FI_DO_DEFRAG);
2482 err = filemap_fdatawrite(inode->i_mapping);
2487 clear_inode_flag(inode, FI_DO_DEFRAG);
2489 inode_unlock(inode);
2491 range->len = (u64)total << PAGE_SHIFT;
2495 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2497 struct inode *inode = file_inode(filp);
2498 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2499 struct f2fs_defragment range;
2502 if (!capable(CAP_SYS_ADMIN))
2505 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2508 if (f2fs_readonly(sbi->sb))
2511 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2515 /* verify alignment of offset & size */
2516 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2519 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2520 sbi->max_file_blocks))
2523 err = mnt_want_write_file(filp);
2527 err = f2fs_defragment_range(sbi, filp, &range);
2528 mnt_drop_write_file(filp);
2530 f2fs_update_time(sbi, REQ_TIME);
2534 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2541 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2542 struct file *file_out, loff_t pos_out, size_t len)
2544 struct inode *src = file_inode(file_in);
2545 struct inode *dst = file_inode(file_out);
2546 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2547 size_t olen = len, dst_max_i_size = 0;
2551 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2552 src->i_sb != dst->i_sb)
2555 if (unlikely(f2fs_readonly(src->i_sb)))
2558 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2561 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2565 if (pos_in == pos_out)
2567 if (pos_out > pos_in && pos_out < pos_in + len)
2574 if (!inode_trylock(dst))
2579 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2582 olen = len = src->i_size - pos_in;
2583 if (pos_in + len == src->i_size)
2584 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2590 dst_osize = dst->i_size;
2591 if (pos_out + olen > dst->i_size)
2592 dst_max_i_size = pos_out + olen;
2594 /* verify the end result is block aligned */
2595 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2596 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2597 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2600 ret = f2fs_convert_inline_inode(src);
2604 ret = f2fs_convert_inline_inode(dst);
2608 /* write out all dirty pages from offset */
2609 ret = filemap_write_and_wait_range(src->i_mapping,
2610 pos_in, pos_in + len);
2614 ret = filemap_write_and_wait_range(dst->i_mapping,
2615 pos_out, pos_out + len);
2619 f2fs_balance_fs(sbi, true);
2621 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2624 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2629 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2630 pos_out >> F2FS_BLKSIZE_BITS,
2631 len >> F2FS_BLKSIZE_BITS, false);
2635 f2fs_i_size_write(dst, dst_max_i_size);
2636 else if (dst_osize != dst->i_size)
2637 f2fs_i_size_write(dst, dst_osize);
2639 f2fs_unlock_op(sbi);
2642 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2644 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2653 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2655 struct f2fs_move_range range;
2659 if (!(filp->f_mode & FMODE_READ) ||
2660 !(filp->f_mode & FMODE_WRITE))
2663 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2667 dst = fdget(range.dst_fd);
2671 if (!(dst.file->f_mode & FMODE_WRITE)) {
2676 err = mnt_want_write_file(filp);
2680 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2681 range.pos_out, range.len);
2683 mnt_drop_write_file(filp);
2687 if (copy_to_user((struct f2fs_move_range __user *)arg,
2688 &range, sizeof(range)))
2695 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2697 struct inode *inode = file_inode(filp);
2698 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2699 struct sit_info *sm = SIT_I(sbi);
2700 unsigned int start_segno = 0, end_segno = 0;
2701 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2702 struct f2fs_flush_device range;
2705 if (!capable(CAP_SYS_ADMIN))
2708 if (f2fs_readonly(sbi->sb))
2711 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2714 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2718 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2719 __is_large_section(sbi)) {
2720 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2721 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2725 ret = mnt_want_write_file(filp);
2729 if (range.dev_num != 0)
2730 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2731 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2733 start_segno = sm->last_victim[FLUSH_DEVICE];
2734 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2735 start_segno = dev_start_segno;
2736 end_segno = min(start_segno + range.segments, dev_end_segno);
2738 while (start_segno < end_segno) {
2739 if (!mutex_trylock(&sbi->gc_mutex)) {
2743 sm->last_victim[GC_CB] = end_segno + 1;
2744 sm->last_victim[GC_GREEDY] = end_segno + 1;
2745 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2746 ret = f2fs_gc(sbi, true, true, start_segno);
2754 mnt_drop_write_file(filp);
2758 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2760 struct inode *inode = file_inode(filp);
2761 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2763 /* Must validate to set it with SQLite behavior in Android. */
2764 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2766 return put_user(sb_feature, (u32 __user *)arg);
2770 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2772 struct dquot *transfer_to[MAXQUOTAS] = {};
2773 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2774 struct super_block *sb = sbi->sb;
2777 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2778 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2779 err = __dquot_transfer(inode, transfer_to);
2781 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2782 dqput(transfer_to[PRJQUOTA]);
2787 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2789 struct inode *inode = file_inode(filp);
2790 struct f2fs_inode_info *fi = F2FS_I(inode);
2791 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2796 if (!f2fs_sb_has_project_quota(sbi)) {
2797 if (projid != F2FS_DEF_PROJID)
2803 if (!f2fs_has_extra_attr(inode))
2806 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2808 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2812 /* Is it quota file? Do not allow user to mess with it */
2813 if (IS_NOQUOTA(inode))
2816 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2818 return PTR_ERR(ipage);
2820 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2823 f2fs_put_page(ipage, 1);
2826 f2fs_put_page(ipage, 1);
2828 err = dquot_initialize(inode);
2833 err = f2fs_transfer_project_quota(inode, kprojid);
2837 F2FS_I(inode)->i_projid = kprojid;
2838 inode->i_ctime = current_time(inode);
2839 f2fs_mark_inode_dirty_sync(inode, true);
2841 f2fs_unlock_op(sbi);
2845 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2850 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2852 if (projid != F2FS_DEF_PROJID)
2858 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
2861 * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
2862 * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
2863 * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
2866 static const struct {
2869 } f2fs_xflags_map[] = {
2870 { F2FS_SYNC_FL, FS_XFLAG_SYNC },
2871 { F2FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE },
2872 { F2FS_APPEND_FL, FS_XFLAG_APPEND },
2873 { F2FS_NODUMP_FL, FS_XFLAG_NODUMP },
2874 { F2FS_NOATIME_FL, FS_XFLAG_NOATIME },
2875 { F2FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT },
2878 #define F2FS_SUPPORTED_XFLAGS ( \
2880 FS_XFLAG_IMMUTABLE | \
2883 FS_XFLAG_NOATIME | \
2884 FS_XFLAG_PROJINHERIT)
2886 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
2887 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
2892 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2893 if (iflags & f2fs_xflags_map[i].iflag)
2894 xflags |= f2fs_xflags_map[i].xflag;
2899 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
2900 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
2905 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2906 if (xflags & f2fs_xflags_map[i].xflag)
2907 iflags |= f2fs_xflags_map[i].iflag;
2912 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
2914 struct f2fs_inode_info *fi = F2FS_I(inode);
2916 simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
2918 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2919 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2922 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2924 struct inode *inode = file_inode(filp);
2927 f2fs_fill_fsxattr(inode, &fa);
2929 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2934 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2936 struct inode *inode = file_inode(filp);
2937 struct fsxattr fa, old_fa;
2941 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2944 /* Make sure caller has proper permission */
2945 if (!inode_owner_or_capable(inode))
2948 if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
2951 iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2952 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2955 err = mnt_want_write_file(filp);
2961 f2fs_fill_fsxattr(inode, &old_fa);
2962 err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
2966 err = f2fs_setflags_common(inode, iflags,
2967 f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
2971 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2973 inode_unlock(inode);
2974 mnt_drop_write_file(filp);
2978 int f2fs_pin_file_control(struct inode *inode, bool inc)
2980 struct f2fs_inode_info *fi = F2FS_I(inode);
2981 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2983 /* Use i_gc_failures for normal file as a risk signal. */
2985 f2fs_i_gc_failures_write(inode,
2986 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2988 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2989 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
2990 __func__, inode->i_ino,
2991 fi->i_gc_failures[GC_FAILURE_PIN]);
2992 clear_inode_flag(inode, FI_PIN_FILE);
2998 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3000 struct inode *inode = file_inode(filp);
3004 if (get_user(pin, (__u32 __user *)arg))
3007 if (!S_ISREG(inode->i_mode))
3010 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3013 ret = mnt_want_write_file(filp);
3019 if (f2fs_should_update_outplace(inode, NULL)) {
3025 clear_inode_flag(inode, FI_PIN_FILE);
3026 f2fs_i_gc_failures_write(inode, 0);
3030 if (f2fs_pin_file_control(inode, false)) {
3034 ret = f2fs_convert_inline_inode(inode);
3038 set_inode_flag(inode, FI_PIN_FILE);
3039 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3041 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3043 inode_unlock(inode);
3044 mnt_drop_write_file(filp);
3048 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3050 struct inode *inode = file_inode(filp);
3053 if (is_inode_flag_set(inode, FI_PIN_FILE))
3054 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3055 return put_user(pin, (u32 __user *)arg);
3058 int f2fs_precache_extents(struct inode *inode)
3060 struct f2fs_inode_info *fi = F2FS_I(inode);
3061 struct f2fs_map_blocks map;
3062 pgoff_t m_next_extent;
3066 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3070 map.m_next_pgofs = NULL;
3071 map.m_next_extent = &m_next_extent;
3072 map.m_seg_type = NO_CHECK_TYPE;
3073 map.m_may_create = false;
3074 end = F2FS_I_SB(inode)->max_file_blocks;
3076 while (map.m_lblk < end) {
3077 map.m_len = end - map.m_lblk;
3079 down_write(&fi->i_gc_rwsem[WRITE]);
3080 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3081 up_write(&fi->i_gc_rwsem[WRITE]);
3085 map.m_lblk = m_next_extent;
3091 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3093 return f2fs_precache_extents(file_inode(filp));
3096 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3098 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3102 if (!capable(CAP_SYS_ADMIN))
3105 if (f2fs_readonly(sbi->sb))
3108 if (copy_from_user(&block_count, (void __user *)arg,
3109 sizeof(block_count)))
3112 ret = f2fs_resize_fs(sbi, block_count);
3117 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3119 struct inode *inode = file_inode(filp);
3121 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3123 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3124 f2fs_warn(F2FS_I_SB(inode),
3125 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3130 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3133 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3135 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3138 return fsverity_ioctl_measure(filp, (void __user *)arg);
3141 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3143 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3147 case F2FS_IOC_GETFLAGS:
3148 return f2fs_ioc_getflags(filp, arg);
3149 case F2FS_IOC_SETFLAGS:
3150 return f2fs_ioc_setflags(filp, arg);
3151 case F2FS_IOC_GETVERSION:
3152 return f2fs_ioc_getversion(filp, arg);
3153 case F2FS_IOC_START_ATOMIC_WRITE:
3154 return f2fs_ioc_start_atomic_write(filp);
3155 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3156 return f2fs_ioc_commit_atomic_write(filp);
3157 case F2FS_IOC_START_VOLATILE_WRITE:
3158 return f2fs_ioc_start_volatile_write(filp);
3159 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3160 return f2fs_ioc_release_volatile_write(filp);
3161 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3162 return f2fs_ioc_abort_volatile_write(filp);
3163 case F2FS_IOC_SHUTDOWN:
3164 return f2fs_ioc_shutdown(filp, arg);
3166 return f2fs_ioc_fitrim(filp, arg);
3167 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3168 return f2fs_ioc_set_encryption_policy(filp, arg);
3169 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3170 return f2fs_ioc_get_encryption_policy(filp, arg);
3171 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3172 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3173 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3174 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3175 case FS_IOC_ADD_ENCRYPTION_KEY:
3176 return f2fs_ioc_add_encryption_key(filp, arg);
3177 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3178 return f2fs_ioc_remove_encryption_key(filp, arg);
3179 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3180 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3181 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3182 return f2fs_ioc_get_encryption_key_status(filp, arg);
3183 case F2FS_IOC_GARBAGE_COLLECT:
3184 return f2fs_ioc_gc(filp, arg);
3185 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3186 return f2fs_ioc_gc_range(filp, arg);
3187 case F2FS_IOC_WRITE_CHECKPOINT:
3188 return f2fs_ioc_write_checkpoint(filp, arg);
3189 case F2FS_IOC_DEFRAGMENT:
3190 return f2fs_ioc_defragment(filp, arg);
3191 case F2FS_IOC_MOVE_RANGE:
3192 return f2fs_ioc_move_range(filp, arg);
3193 case F2FS_IOC_FLUSH_DEVICE:
3194 return f2fs_ioc_flush_device(filp, arg);
3195 case F2FS_IOC_GET_FEATURES:
3196 return f2fs_ioc_get_features(filp, arg);
3197 case F2FS_IOC_FSGETXATTR:
3198 return f2fs_ioc_fsgetxattr(filp, arg);
3199 case F2FS_IOC_FSSETXATTR:
3200 return f2fs_ioc_fssetxattr(filp, arg);
3201 case F2FS_IOC_GET_PIN_FILE:
3202 return f2fs_ioc_get_pin_file(filp, arg);
3203 case F2FS_IOC_SET_PIN_FILE:
3204 return f2fs_ioc_set_pin_file(filp, arg);
3205 case F2FS_IOC_PRECACHE_EXTENTS:
3206 return f2fs_ioc_precache_extents(filp, arg);
3207 case F2FS_IOC_RESIZE_FS:
3208 return f2fs_ioc_resize_fs(filp, arg);
3209 case FS_IOC_ENABLE_VERITY:
3210 return f2fs_ioc_enable_verity(filp, arg);
3211 case FS_IOC_MEASURE_VERITY:
3212 return f2fs_ioc_measure_verity(filp, arg);
3218 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3220 struct file *file = iocb->ki_filp;
3221 struct inode *inode = file_inode(file);
3224 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3229 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT)) {
3234 if (!inode_trylock(inode)) {
3235 if (iocb->ki_flags & IOCB_NOWAIT) {
3242 ret = generic_write_checks(iocb, from);
3244 bool preallocated = false;
3245 size_t target_size = 0;
3248 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3249 set_inode_flag(inode, FI_NO_PREALLOC);
3251 if ((iocb->ki_flags & IOCB_NOWAIT)) {
3252 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3253 iov_iter_count(from)) ||
3254 f2fs_has_inline_data(inode) ||
3255 f2fs_force_buffered_io(inode, iocb, from)) {
3256 clear_inode_flag(inode, FI_NO_PREALLOC);
3257 inode_unlock(inode);
3262 preallocated = true;
3263 target_size = iocb->ki_pos + iov_iter_count(from);
3265 err = f2fs_preallocate_blocks(iocb, from);
3267 clear_inode_flag(inode, FI_NO_PREALLOC);
3268 inode_unlock(inode);
3273 ret = __generic_file_write_iter(iocb, from);
3274 clear_inode_flag(inode, FI_NO_PREALLOC);
3276 /* if we couldn't write data, we should deallocate blocks. */
3277 if (preallocated && i_size_read(inode) < target_size)
3278 f2fs_truncate(inode);
3281 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3283 inode_unlock(inode);
3285 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3286 iov_iter_count(from), ret);
3288 ret = generic_write_sync(iocb, ret);
3292 #ifdef CONFIG_COMPAT
3293 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3296 case F2FS_IOC32_GETFLAGS:
3297 cmd = F2FS_IOC_GETFLAGS;
3299 case F2FS_IOC32_SETFLAGS:
3300 cmd = F2FS_IOC_SETFLAGS;
3302 case F2FS_IOC32_GETVERSION:
3303 cmd = F2FS_IOC_GETVERSION;
3305 case F2FS_IOC_START_ATOMIC_WRITE:
3306 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3307 case F2FS_IOC_START_VOLATILE_WRITE:
3308 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3309 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3310 case F2FS_IOC_SHUTDOWN:
3311 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3312 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3313 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3314 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3315 case FS_IOC_ADD_ENCRYPTION_KEY:
3316 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3317 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3318 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3319 case F2FS_IOC_GARBAGE_COLLECT:
3320 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3321 case F2FS_IOC_WRITE_CHECKPOINT:
3322 case F2FS_IOC_DEFRAGMENT:
3323 case F2FS_IOC_MOVE_RANGE:
3324 case F2FS_IOC_FLUSH_DEVICE:
3325 case F2FS_IOC_GET_FEATURES:
3326 case F2FS_IOC_FSGETXATTR:
3327 case F2FS_IOC_FSSETXATTR:
3328 case F2FS_IOC_GET_PIN_FILE:
3329 case F2FS_IOC_SET_PIN_FILE:
3330 case F2FS_IOC_PRECACHE_EXTENTS:
3331 case F2FS_IOC_RESIZE_FS:
3332 case FS_IOC_ENABLE_VERITY:
3333 case FS_IOC_MEASURE_VERITY:
3336 return -ENOIOCTLCMD;
3338 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3342 const struct file_operations f2fs_file_operations = {
3343 .llseek = f2fs_llseek,
3344 .read_iter = generic_file_read_iter,
3345 .write_iter = f2fs_file_write_iter,
3346 .open = f2fs_file_open,
3347 .release = f2fs_release_file,
3348 .mmap = f2fs_file_mmap,
3349 .flush = f2fs_file_flush,
3350 .fsync = f2fs_sync_file,
3351 .fallocate = f2fs_fallocate,
3352 .unlocked_ioctl = f2fs_ioctl,
3353 #ifdef CONFIG_COMPAT
3354 .compat_ioctl = f2fs_compat_ioctl,
3356 .splice_read = generic_file_splice_read,
3357 .splice_write = iter_file_splice_write,