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;
710 flags = fi->i_flags & F2FS_FL_USER_VISIBLE;
711 if (flags & F2FS_APPEND_FL)
712 stat->attributes |= STATX_ATTR_APPEND;
713 if (flags & F2FS_COMPR_FL)
714 stat->attributes |= STATX_ATTR_COMPRESSED;
715 if (IS_ENCRYPTED(inode))
716 stat->attributes |= STATX_ATTR_ENCRYPTED;
717 if (flags & F2FS_IMMUTABLE_FL)
718 stat->attributes |= STATX_ATTR_IMMUTABLE;
719 if (flags & F2FS_NODUMP_FL)
720 stat->attributes |= STATX_ATTR_NODUMP;
722 stat->attributes_mask |= (STATX_ATTR_APPEND |
723 STATX_ATTR_COMPRESSED |
724 STATX_ATTR_ENCRYPTED |
725 STATX_ATTR_IMMUTABLE |
728 generic_fillattr(inode, stat);
730 /* we need to show initial sectors used for inline_data/dentries */
731 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
732 f2fs_has_inline_dentry(inode))
733 stat->blocks += (stat->size + 511) >> 9;
738 #ifdef CONFIG_F2FS_FS_POSIX_ACL
739 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
741 unsigned int ia_valid = attr->ia_valid;
743 if (ia_valid & ATTR_UID)
744 inode->i_uid = attr->ia_uid;
745 if (ia_valid & ATTR_GID)
746 inode->i_gid = attr->ia_gid;
747 if (ia_valid & ATTR_ATIME)
748 inode->i_atime = timespec64_trunc(attr->ia_atime,
749 inode->i_sb->s_time_gran);
750 if (ia_valid & ATTR_MTIME)
751 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
752 inode->i_sb->s_time_gran);
753 if (ia_valid & ATTR_CTIME)
754 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
755 inode->i_sb->s_time_gran);
756 if (ia_valid & ATTR_MODE) {
757 umode_t mode = attr->ia_mode;
759 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
761 set_acl_inode(inode, mode);
765 #define __setattr_copy setattr_copy
768 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
770 struct inode *inode = d_inode(dentry);
773 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
776 err = setattr_prepare(dentry, attr);
780 err = fscrypt_prepare_setattr(dentry, attr);
784 if (is_quota_modification(inode, attr)) {
785 err = dquot_initialize(inode);
789 if ((attr->ia_valid & ATTR_UID &&
790 !uid_eq(attr->ia_uid, inode->i_uid)) ||
791 (attr->ia_valid & ATTR_GID &&
792 !gid_eq(attr->ia_gid, inode->i_gid))) {
793 f2fs_lock_op(F2FS_I_SB(inode));
794 err = dquot_transfer(inode, attr);
796 set_sbi_flag(F2FS_I_SB(inode),
797 SBI_QUOTA_NEED_REPAIR);
798 f2fs_unlock_op(F2FS_I_SB(inode));
802 * update uid/gid under lock_op(), so that dquot and inode can
803 * be updated atomically.
805 if (attr->ia_valid & ATTR_UID)
806 inode->i_uid = attr->ia_uid;
807 if (attr->ia_valid & ATTR_GID)
808 inode->i_gid = attr->ia_gid;
809 f2fs_mark_inode_dirty_sync(inode, true);
810 f2fs_unlock_op(F2FS_I_SB(inode));
813 if (attr->ia_valid & ATTR_SIZE) {
814 bool to_smaller = (attr->ia_size <= i_size_read(inode));
816 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
817 down_write(&F2FS_I(inode)->i_mmap_sem);
819 truncate_setsize(inode, attr->ia_size);
822 err = f2fs_truncate(inode);
824 * do not trim all blocks after i_size if target size is
825 * larger than i_size.
827 up_write(&F2FS_I(inode)->i_mmap_sem);
828 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
834 /* should convert inline inode here */
835 if (!f2fs_may_inline_data(inode)) {
836 err = f2fs_convert_inline_inode(inode);
840 inode->i_mtime = inode->i_ctime = current_time(inode);
843 down_write(&F2FS_I(inode)->i_sem);
844 F2FS_I(inode)->last_disk_size = i_size_read(inode);
845 up_write(&F2FS_I(inode)->i_sem);
848 __setattr_copy(inode, attr);
850 if (attr->ia_valid & ATTR_MODE) {
851 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
852 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
853 inode->i_mode = F2FS_I(inode)->i_acl_mode;
854 clear_inode_flag(inode, FI_ACL_MODE);
858 /* file size may changed here */
859 f2fs_mark_inode_dirty_sync(inode, true);
861 /* inode change will produce dirty node pages flushed by checkpoint */
862 f2fs_balance_fs(F2FS_I_SB(inode), true);
867 const struct inode_operations f2fs_file_inode_operations = {
868 .getattr = f2fs_getattr,
869 .setattr = f2fs_setattr,
870 .get_acl = f2fs_get_acl,
871 .set_acl = f2fs_set_acl,
872 #ifdef CONFIG_F2FS_FS_XATTR
873 .listxattr = f2fs_listxattr,
875 .fiemap = f2fs_fiemap,
878 static int fill_zero(struct inode *inode, pgoff_t index,
879 loff_t start, loff_t len)
881 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
887 f2fs_balance_fs(sbi, true);
890 page = f2fs_get_new_data_page(inode, NULL, index, false);
894 return PTR_ERR(page);
896 f2fs_wait_on_page_writeback(page, DATA, true, true);
897 zero_user(page, start, len);
898 set_page_dirty(page);
899 f2fs_put_page(page, 1);
903 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
907 while (pg_start < pg_end) {
908 struct dnode_of_data dn;
909 pgoff_t end_offset, count;
911 set_new_dnode(&dn, inode, NULL, NULL, 0);
912 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
914 if (err == -ENOENT) {
915 pg_start = f2fs_get_next_page_offset(&dn,
922 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
923 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
925 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
927 f2fs_truncate_data_blocks_range(&dn, count);
935 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
937 pgoff_t pg_start, pg_end;
938 loff_t off_start, off_end;
941 ret = f2fs_convert_inline_inode(inode);
945 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
946 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
948 off_start = offset & (PAGE_SIZE - 1);
949 off_end = (offset + len) & (PAGE_SIZE - 1);
951 if (pg_start == pg_end) {
952 ret = fill_zero(inode, pg_start, off_start,
953 off_end - off_start);
958 ret = fill_zero(inode, pg_start++, off_start,
959 PAGE_SIZE - off_start);
964 ret = fill_zero(inode, pg_end, 0, off_end);
969 if (pg_start < pg_end) {
970 struct address_space *mapping = inode->i_mapping;
971 loff_t blk_start, blk_end;
972 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
974 f2fs_balance_fs(sbi, true);
976 blk_start = (loff_t)pg_start << PAGE_SHIFT;
977 blk_end = (loff_t)pg_end << PAGE_SHIFT;
979 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
980 down_write(&F2FS_I(inode)->i_mmap_sem);
982 truncate_inode_pages_range(mapping, blk_start,
986 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
989 up_write(&F2FS_I(inode)->i_mmap_sem);
990 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
997 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
998 int *do_replace, pgoff_t off, pgoff_t len)
1000 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1001 struct dnode_of_data dn;
1005 set_new_dnode(&dn, inode, NULL, NULL, 0);
1006 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1007 if (ret && ret != -ENOENT) {
1009 } else if (ret == -ENOENT) {
1010 if (dn.max_level == 0)
1012 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1019 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1020 dn.ofs_in_node, len);
1021 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1022 *blkaddr = datablock_addr(dn.inode,
1023 dn.node_page, dn.ofs_in_node);
1025 if (__is_valid_data_blkaddr(*blkaddr) &&
1026 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1027 DATA_GENERIC_ENHANCE)) {
1028 f2fs_put_dnode(&dn);
1032 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1034 if (test_opt(sbi, LFS)) {
1035 f2fs_put_dnode(&dn);
1039 /* do not invalidate this block address */
1040 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1044 f2fs_put_dnode(&dn);
1053 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1054 int *do_replace, pgoff_t off, int len)
1056 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1057 struct dnode_of_data dn;
1060 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1061 if (*do_replace == 0)
1064 set_new_dnode(&dn, inode, NULL, NULL, 0);
1065 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1067 dec_valid_block_count(sbi, inode, 1);
1068 f2fs_invalidate_blocks(sbi, *blkaddr);
1070 f2fs_update_data_blkaddr(&dn, *blkaddr);
1072 f2fs_put_dnode(&dn);
1077 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1078 block_t *blkaddr, int *do_replace,
1079 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1081 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1086 if (blkaddr[i] == NULL_ADDR && !full) {
1091 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1092 struct dnode_of_data dn;
1093 struct node_info ni;
1097 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1098 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1102 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1104 f2fs_put_dnode(&dn);
1108 ilen = min((pgoff_t)
1109 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1110 dn.ofs_in_node, len - i);
1112 dn.data_blkaddr = datablock_addr(dn.inode,
1113 dn.node_page, dn.ofs_in_node);
1114 f2fs_truncate_data_blocks_range(&dn, 1);
1116 if (do_replace[i]) {
1117 f2fs_i_blocks_write(src_inode,
1119 f2fs_i_blocks_write(dst_inode,
1121 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1122 blkaddr[i], ni.version, true, false);
1128 new_size = (dst + i) << PAGE_SHIFT;
1129 if (dst_inode->i_size < new_size)
1130 f2fs_i_size_write(dst_inode, new_size);
1131 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1133 f2fs_put_dnode(&dn);
1135 struct page *psrc, *pdst;
1137 psrc = f2fs_get_lock_data_page(src_inode,
1140 return PTR_ERR(psrc);
1141 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1144 f2fs_put_page(psrc, 1);
1145 return PTR_ERR(pdst);
1147 f2fs_copy_page(psrc, pdst);
1148 set_page_dirty(pdst);
1149 f2fs_put_page(pdst, 1);
1150 f2fs_put_page(psrc, 1);
1152 ret = f2fs_truncate_hole(src_inode,
1153 src + i, src + i + 1);
1162 static int __exchange_data_block(struct inode *src_inode,
1163 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1164 pgoff_t len, bool full)
1166 block_t *src_blkaddr;
1172 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1174 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1175 array_size(olen, sizeof(block_t)),
1180 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1181 array_size(olen, sizeof(int)),
1184 kvfree(src_blkaddr);
1188 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1189 do_replace, src, olen);
1193 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1194 do_replace, src, dst, olen, full);
1202 kvfree(src_blkaddr);
1208 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1209 kvfree(src_blkaddr);
1214 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1216 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1217 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1218 pgoff_t start = offset >> PAGE_SHIFT;
1219 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1222 f2fs_balance_fs(sbi, true);
1224 /* avoid gc operation during block exchange */
1225 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1226 down_write(&F2FS_I(inode)->i_mmap_sem);
1229 f2fs_drop_extent_tree(inode);
1230 truncate_pagecache(inode, offset);
1231 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1232 f2fs_unlock_op(sbi);
1234 up_write(&F2FS_I(inode)->i_mmap_sem);
1235 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1239 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1244 if (offset + len >= i_size_read(inode))
1247 /* collapse range should be aligned to block size of f2fs. */
1248 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1251 ret = f2fs_convert_inline_inode(inode);
1255 /* write out all dirty pages from offset */
1256 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1260 ret = f2fs_do_collapse(inode, offset, len);
1264 /* write out all moved pages, if possible */
1265 down_write(&F2FS_I(inode)->i_mmap_sem);
1266 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1267 truncate_pagecache(inode, offset);
1269 new_size = i_size_read(inode) - len;
1270 truncate_pagecache(inode, new_size);
1272 ret = f2fs_truncate_blocks(inode, new_size, true);
1273 up_write(&F2FS_I(inode)->i_mmap_sem);
1275 f2fs_i_size_write(inode, new_size);
1279 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1282 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1283 pgoff_t index = start;
1284 unsigned int ofs_in_node = dn->ofs_in_node;
1288 for (; index < end; index++, dn->ofs_in_node++) {
1289 if (datablock_addr(dn->inode, dn->node_page,
1290 dn->ofs_in_node) == NULL_ADDR)
1294 dn->ofs_in_node = ofs_in_node;
1295 ret = f2fs_reserve_new_blocks(dn, count);
1299 dn->ofs_in_node = ofs_in_node;
1300 for (index = start; index < end; index++, dn->ofs_in_node++) {
1301 dn->data_blkaddr = datablock_addr(dn->inode,
1302 dn->node_page, dn->ofs_in_node);
1304 * f2fs_reserve_new_blocks will not guarantee entire block
1307 if (dn->data_blkaddr == NULL_ADDR) {
1311 if (dn->data_blkaddr != NEW_ADDR) {
1312 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1313 dn->data_blkaddr = NEW_ADDR;
1314 f2fs_set_data_blkaddr(dn);
1318 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1323 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1326 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1327 struct address_space *mapping = inode->i_mapping;
1328 pgoff_t index, pg_start, pg_end;
1329 loff_t new_size = i_size_read(inode);
1330 loff_t off_start, off_end;
1333 ret = inode_newsize_ok(inode, (len + offset));
1337 ret = f2fs_convert_inline_inode(inode);
1341 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1345 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1346 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1348 off_start = offset & (PAGE_SIZE - 1);
1349 off_end = (offset + len) & (PAGE_SIZE - 1);
1351 if (pg_start == pg_end) {
1352 ret = fill_zero(inode, pg_start, off_start,
1353 off_end - off_start);
1357 new_size = max_t(loff_t, new_size, offset + len);
1360 ret = fill_zero(inode, pg_start++, off_start,
1361 PAGE_SIZE - off_start);
1365 new_size = max_t(loff_t, new_size,
1366 (loff_t)pg_start << PAGE_SHIFT);
1369 for (index = pg_start; index < pg_end;) {
1370 struct dnode_of_data dn;
1371 unsigned int end_offset;
1374 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1375 down_write(&F2FS_I(inode)->i_mmap_sem);
1377 truncate_pagecache_range(inode,
1378 (loff_t)index << PAGE_SHIFT,
1379 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1383 set_new_dnode(&dn, inode, NULL, NULL, 0);
1384 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1386 f2fs_unlock_op(sbi);
1387 up_write(&F2FS_I(inode)->i_mmap_sem);
1388 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1392 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1393 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1395 ret = f2fs_do_zero_range(&dn, index, end);
1396 f2fs_put_dnode(&dn);
1398 f2fs_unlock_op(sbi);
1399 up_write(&F2FS_I(inode)->i_mmap_sem);
1400 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1402 f2fs_balance_fs(sbi, dn.node_changed);
1408 new_size = max_t(loff_t, new_size,
1409 (loff_t)index << PAGE_SHIFT);
1413 ret = fill_zero(inode, pg_end, 0, off_end);
1417 new_size = max_t(loff_t, new_size, offset + len);
1422 if (new_size > i_size_read(inode)) {
1423 if (mode & FALLOC_FL_KEEP_SIZE)
1424 file_set_keep_isize(inode);
1426 f2fs_i_size_write(inode, new_size);
1431 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1433 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1434 pgoff_t nr, pg_start, pg_end, delta, idx;
1438 new_size = i_size_read(inode) + len;
1439 ret = inode_newsize_ok(inode, new_size);
1443 if (offset >= i_size_read(inode))
1446 /* insert range should be aligned to block size of f2fs. */
1447 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1450 ret = f2fs_convert_inline_inode(inode);
1454 f2fs_balance_fs(sbi, true);
1456 down_write(&F2FS_I(inode)->i_mmap_sem);
1457 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1458 up_write(&F2FS_I(inode)->i_mmap_sem);
1462 /* write out all dirty pages from offset */
1463 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1467 pg_start = offset >> PAGE_SHIFT;
1468 pg_end = (offset + len) >> PAGE_SHIFT;
1469 delta = pg_end - pg_start;
1470 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1472 /* avoid gc operation during block exchange */
1473 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1474 down_write(&F2FS_I(inode)->i_mmap_sem);
1475 truncate_pagecache(inode, offset);
1477 while (!ret && idx > pg_start) {
1478 nr = idx - pg_start;
1484 f2fs_drop_extent_tree(inode);
1486 ret = __exchange_data_block(inode, inode, idx,
1487 idx + delta, nr, false);
1488 f2fs_unlock_op(sbi);
1490 up_write(&F2FS_I(inode)->i_mmap_sem);
1491 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1493 /* write out all moved pages, if possible */
1494 down_write(&F2FS_I(inode)->i_mmap_sem);
1495 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1496 truncate_pagecache(inode, offset);
1497 up_write(&F2FS_I(inode)->i_mmap_sem);
1500 f2fs_i_size_write(inode, new_size);
1504 static int expand_inode_data(struct inode *inode, loff_t offset,
1505 loff_t len, int mode)
1507 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1508 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1509 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1510 .m_may_create = true };
1512 loff_t new_size = i_size_read(inode);
1516 err = inode_newsize_ok(inode, (len + offset));
1520 err = f2fs_convert_inline_inode(inode);
1524 f2fs_balance_fs(sbi, true);
1526 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1527 off_end = (offset + len) & (PAGE_SIZE - 1);
1529 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1530 map.m_len = pg_end - map.m_lblk;
1534 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1541 last_off = map.m_lblk + map.m_len - 1;
1543 /* update new size to the failed position */
1544 new_size = (last_off == pg_end) ? offset + len :
1545 (loff_t)(last_off + 1) << PAGE_SHIFT;
1547 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1550 if (new_size > i_size_read(inode)) {
1551 if (mode & FALLOC_FL_KEEP_SIZE)
1552 file_set_keep_isize(inode);
1554 f2fs_i_size_write(inode, new_size);
1560 static long f2fs_fallocate(struct file *file, int mode,
1561 loff_t offset, loff_t len)
1563 struct inode *inode = file_inode(file);
1566 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1569 /* f2fs only support ->fallocate for regular file */
1570 if (!S_ISREG(inode->i_mode))
1573 if (IS_ENCRYPTED(inode) &&
1574 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1577 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1578 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1579 FALLOC_FL_INSERT_RANGE))
1584 if (mode & FALLOC_FL_PUNCH_HOLE) {
1585 if (offset >= inode->i_size)
1588 ret = punch_hole(inode, offset, len);
1589 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1590 ret = f2fs_collapse_range(inode, offset, len);
1591 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1592 ret = f2fs_zero_range(inode, offset, len, mode);
1593 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1594 ret = f2fs_insert_range(inode, offset, len);
1596 ret = expand_inode_data(inode, offset, len, mode);
1600 inode->i_mtime = inode->i_ctime = current_time(inode);
1601 f2fs_mark_inode_dirty_sync(inode, false);
1602 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1606 inode_unlock(inode);
1608 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1612 static int f2fs_release_file(struct inode *inode, struct file *filp)
1615 * f2fs_relase_file is called at every close calls. So we should
1616 * not drop any inmemory pages by close called by other process.
1618 if (!(filp->f_mode & FMODE_WRITE) ||
1619 atomic_read(&inode->i_writecount) != 1)
1622 /* some remained atomic pages should discarded */
1623 if (f2fs_is_atomic_file(inode))
1624 f2fs_drop_inmem_pages(inode);
1625 if (f2fs_is_volatile_file(inode)) {
1626 set_inode_flag(inode, FI_DROP_CACHE);
1627 filemap_fdatawrite(inode->i_mapping);
1628 clear_inode_flag(inode, FI_DROP_CACHE);
1629 clear_inode_flag(inode, FI_VOLATILE_FILE);
1630 stat_dec_volatile_write(inode);
1635 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1637 struct inode *inode = file_inode(file);
1640 * If the process doing a transaction is crashed, we should do
1641 * roll-back. Otherwise, other reader/write can see corrupted database
1642 * until all the writers close its file. Since this should be done
1643 * before dropping file lock, it needs to do in ->flush.
1645 if (f2fs_is_atomic_file(inode) &&
1646 F2FS_I(inode)->inmem_task == current)
1647 f2fs_drop_inmem_pages(inode);
1651 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1653 struct inode *inode = file_inode(filp);
1654 struct f2fs_inode_info *fi = F2FS_I(inode);
1655 unsigned int flags = fi->i_flags;
1657 if (IS_ENCRYPTED(inode))
1658 flags |= F2FS_ENCRYPT_FL;
1659 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1660 flags |= F2FS_INLINE_DATA_FL;
1661 if (is_inode_flag_set(inode, FI_PIN_FILE))
1662 flags |= F2FS_NOCOW_FL;
1664 flags &= F2FS_FL_USER_VISIBLE;
1666 return put_user(flags, (int __user *)arg);
1669 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1671 struct f2fs_inode_info *fi = F2FS_I(inode);
1672 unsigned int oldflags;
1674 /* Is it quota file? Do not allow user to mess with it */
1675 if (IS_NOQUOTA(inode))
1678 flags = f2fs_mask_flags(inode->i_mode, flags);
1680 oldflags = fi->i_flags;
1682 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL))
1683 if (!capable(CAP_LINUX_IMMUTABLE))
1686 flags = flags & F2FS_FL_USER_MODIFIABLE;
1687 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE;
1688 fi->i_flags = flags;
1690 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1691 set_inode_flag(inode, FI_PROJ_INHERIT);
1693 clear_inode_flag(inode, FI_PROJ_INHERIT);
1695 inode->i_ctime = current_time(inode);
1696 f2fs_set_inode_flags(inode);
1697 f2fs_mark_inode_dirty_sync(inode, true);
1701 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1703 struct inode *inode = file_inode(filp);
1707 if (!inode_owner_or_capable(inode))
1710 if (get_user(flags, (int __user *)arg))
1713 ret = mnt_want_write_file(filp);
1719 ret = __f2fs_ioc_setflags(inode, flags);
1721 inode_unlock(inode);
1722 mnt_drop_write_file(filp);
1726 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1728 struct inode *inode = file_inode(filp);
1730 return put_user(inode->i_generation, (int __user *)arg);
1733 static int f2fs_ioc_start_atomic_write(struct file *filp)
1735 struct inode *inode = file_inode(filp);
1738 if (!inode_owner_or_capable(inode))
1741 if (!S_ISREG(inode->i_mode))
1744 ret = mnt_want_write_file(filp);
1750 if (f2fs_is_atomic_file(inode)) {
1751 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1756 ret = f2fs_convert_inline_inode(inode);
1760 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1763 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1764 * f2fs_is_atomic_file.
1766 if (get_dirty_pages(inode))
1767 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1768 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1769 inode->i_ino, get_dirty_pages(inode));
1770 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1772 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1776 set_inode_flag(inode, FI_ATOMIC_FILE);
1777 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1778 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1780 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1781 F2FS_I(inode)->inmem_task = current;
1782 stat_inc_atomic_write(inode);
1783 stat_update_max_atomic_write(inode);
1785 inode_unlock(inode);
1786 mnt_drop_write_file(filp);
1790 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1792 struct inode *inode = file_inode(filp);
1795 if (!inode_owner_or_capable(inode))
1798 ret = mnt_want_write_file(filp);
1802 f2fs_balance_fs(F2FS_I_SB(inode), true);
1806 if (f2fs_is_volatile_file(inode)) {
1811 if (f2fs_is_atomic_file(inode)) {
1812 ret = f2fs_commit_inmem_pages(inode);
1816 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1818 clear_inode_flag(inode, FI_ATOMIC_FILE);
1819 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1820 stat_dec_atomic_write(inode);
1823 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1826 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1827 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1830 inode_unlock(inode);
1831 mnt_drop_write_file(filp);
1835 static int f2fs_ioc_start_volatile_write(struct file *filp)
1837 struct inode *inode = file_inode(filp);
1840 if (!inode_owner_or_capable(inode))
1843 if (!S_ISREG(inode->i_mode))
1846 ret = mnt_want_write_file(filp);
1852 if (f2fs_is_volatile_file(inode))
1855 ret = f2fs_convert_inline_inode(inode);
1859 stat_inc_volatile_write(inode);
1860 stat_update_max_volatile_write(inode);
1862 set_inode_flag(inode, FI_VOLATILE_FILE);
1863 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1865 inode_unlock(inode);
1866 mnt_drop_write_file(filp);
1870 static int f2fs_ioc_release_volatile_write(struct file *filp)
1872 struct inode *inode = file_inode(filp);
1875 if (!inode_owner_or_capable(inode))
1878 ret = mnt_want_write_file(filp);
1884 if (!f2fs_is_volatile_file(inode))
1887 if (!f2fs_is_first_block_written(inode)) {
1888 ret = truncate_partial_data_page(inode, 0, true);
1892 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1894 inode_unlock(inode);
1895 mnt_drop_write_file(filp);
1899 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1901 struct inode *inode = file_inode(filp);
1904 if (!inode_owner_or_capable(inode))
1907 ret = mnt_want_write_file(filp);
1913 if (f2fs_is_atomic_file(inode))
1914 f2fs_drop_inmem_pages(inode);
1915 if (f2fs_is_volatile_file(inode)) {
1916 clear_inode_flag(inode, FI_VOLATILE_FILE);
1917 stat_dec_volatile_write(inode);
1918 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1921 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1923 inode_unlock(inode);
1925 mnt_drop_write_file(filp);
1926 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1930 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1932 struct inode *inode = file_inode(filp);
1933 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1934 struct super_block *sb = sbi->sb;
1938 if (!capable(CAP_SYS_ADMIN))
1941 if (get_user(in, (__u32 __user *)arg))
1944 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1945 ret = mnt_want_write_file(filp);
1951 case F2FS_GOING_DOWN_FULLSYNC:
1952 sb = freeze_bdev(sb->s_bdev);
1958 f2fs_stop_checkpoint(sbi, false);
1959 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1960 thaw_bdev(sb->s_bdev, sb);
1963 case F2FS_GOING_DOWN_METASYNC:
1964 /* do checkpoint only */
1965 ret = f2fs_sync_fs(sb, 1);
1968 f2fs_stop_checkpoint(sbi, false);
1969 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1971 case F2FS_GOING_DOWN_NOSYNC:
1972 f2fs_stop_checkpoint(sbi, false);
1973 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1975 case F2FS_GOING_DOWN_METAFLUSH:
1976 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1977 f2fs_stop_checkpoint(sbi, false);
1978 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1980 case F2FS_GOING_DOWN_NEED_FSCK:
1981 set_sbi_flag(sbi, SBI_NEED_FSCK);
1982 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
1983 set_sbi_flag(sbi, SBI_IS_DIRTY);
1984 /* do checkpoint only */
1985 ret = f2fs_sync_fs(sb, 1);
1992 f2fs_stop_gc_thread(sbi);
1993 f2fs_stop_discard_thread(sbi);
1995 f2fs_drop_discard_cmd(sbi);
1996 clear_opt(sbi, DISCARD);
1998 f2fs_update_time(sbi, REQ_TIME);
2000 if (in != F2FS_GOING_DOWN_FULLSYNC)
2001 mnt_drop_write_file(filp);
2003 trace_f2fs_shutdown(sbi, in, ret);
2008 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2010 struct inode *inode = file_inode(filp);
2011 struct super_block *sb = inode->i_sb;
2012 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2013 struct fstrim_range range;
2016 if (!capable(CAP_SYS_ADMIN))
2019 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2022 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2026 ret = mnt_want_write_file(filp);
2030 range.minlen = max((unsigned int)range.minlen,
2031 q->limits.discard_granularity);
2032 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2033 mnt_drop_write_file(filp);
2037 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2040 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2044 static bool uuid_is_nonzero(__u8 u[16])
2048 for (i = 0; i < 16; i++)
2054 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2056 struct inode *inode = file_inode(filp);
2058 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2061 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2063 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2066 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2068 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2070 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2073 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2075 struct inode *inode = file_inode(filp);
2076 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2079 if (!f2fs_sb_has_encrypt(sbi))
2082 err = mnt_want_write_file(filp);
2086 down_write(&sbi->sb_lock);
2088 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2091 /* update superblock with uuid */
2092 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2094 err = f2fs_commit_super(sbi, false);
2097 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2101 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2105 up_write(&sbi->sb_lock);
2106 mnt_drop_write_file(filp);
2110 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2112 struct inode *inode = file_inode(filp);
2113 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2117 if (!capable(CAP_SYS_ADMIN))
2120 if (get_user(sync, (__u32 __user *)arg))
2123 if (f2fs_readonly(sbi->sb))
2126 ret = mnt_want_write_file(filp);
2131 if (!mutex_trylock(&sbi->gc_mutex)) {
2136 mutex_lock(&sbi->gc_mutex);
2139 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2141 mnt_drop_write_file(filp);
2145 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2147 struct inode *inode = file_inode(filp);
2148 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2149 struct f2fs_gc_range range;
2153 if (!capable(CAP_SYS_ADMIN))
2156 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2160 if (f2fs_readonly(sbi->sb))
2163 end = range.start + range.len;
2164 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2168 ret = mnt_want_write_file(filp);
2174 if (!mutex_trylock(&sbi->gc_mutex)) {
2179 mutex_lock(&sbi->gc_mutex);
2182 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2183 range.start += BLKS_PER_SEC(sbi);
2184 if (range.start <= end)
2187 mnt_drop_write_file(filp);
2191 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2193 struct inode *inode = file_inode(filp);
2194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2197 if (!capable(CAP_SYS_ADMIN))
2200 if (f2fs_readonly(sbi->sb))
2203 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2204 f2fs_msg(sbi->sb, KERN_INFO,
2205 "Skipping Checkpoint. Checkpoints currently disabled.");
2209 ret = mnt_want_write_file(filp);
2213 ret = f2fs_sync_fs(sbi->sb, 1);
2215 mnt_drop_write_file(filp);
2219 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2221 struct f2fs_defragment *range)
2223 struct inode *inode = file_inode(filp);
2224 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2225 .m_seg_type = NO_CHECK_TYPE ,
2226 .m_may_create = false };
2227 struct extent_info ei = {0, 0, 0};
2228 pgoff_t pg_start, pg_end, next_pgofs;
2229 unsigned int blk_per_seg = sbi->blocks_per_seg;
2230 unsigned int total = 0, sec_num;
2231 block_t blk_end = 0;
2232 bool fragmented = false;
2235 /* if in-place-update policy is enabled, don't waste time here */
2236 if (f2fs_should_update_inplace(inode, NULL))
2239 pg_start = range->start >> PAGE_SHIFT;
2240 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2242 f2fs_balance_fs(sbi, true);
2246 /* writeback all dirty pages in the range */
2247 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2248 range->start + range->len - 1);
2253 * lookup mapping info in extent cache, skip defragmenting if physical
2254 * block addresses are continuous.
2256 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2257 if (ei.fofs + ei.len >= pg_end)
2261 map.m_lblk = pg_start;
2262 map.m_next_pgofs = &next_pgofs;
2265 * lookup mapping info in dnode page cache, skip defragmenting if all
2266 * physical block addresses are continuous even if there are hole(s)
2267 * in logical blocks.
2269 while (map.m_lblk < pg_end) {
2270 map.m_len = pg_end - map.m_lblk;
2271 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2275 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2276 map.m_lblk = next_pgofs;
2280 if (blk_end && blk_end != map.m_pblk)
2283 /* record total count of block that we're going to move */
2286 blk_end = map.m_pblk + map.m_len;
2288 map.m_lblk += map.m_len;
2294 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2297 * make sure there are enough free section for LFS allocation, this can
2298 * avoid defragment running in SSR mode when free section are allocated
2301 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2306 map.m_lblk = pg_start;
2307 map.m_len = pg_end - pg_start;
2310 while (map.m_lblk < pg_end) {
2315 map.m_len = pg_end - map.m_lblk;
2316 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2320 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2321 map.m_lblk = next_pgofs;
2325 set_inode_flag(inode, FI_DO_DEFRAG);
2328 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2331 page = f2fs_get_lock_data_page(inode, idx, true);
2333 err = PTR_ERR(page);
2337 set_page_dirty(page);
2338 f2fs_put_page(page, 1);
2347 if (idx < pg_end && cnt < blk_per_seg)
2350 clear_inode_flag(inode, FI_DO_DEFRAG);
2352 err = filemap_fdatawrite(inode->i_mapping);
2357 clear_inode_flag(inode, FI_DO_DEFRAG);
2359 inode_unlock(inode);
2361 range->len = (u64)total << PAGE_SHIFT;
2365 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2367 struct inode *inode = file_inode(filp);
2368 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2369 struct f2fs_defragment range;
2372 if (!capable(CAP_SYS_ADMIN))
2375 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2378 if (f2fs_readonly(sbi->sb))
2381 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2385 /* verify alignment of offset & size */
2386 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2389 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2390 sbi->max_file_blocks))
2393 err = mnt_want_write_file(filp);
2397 err = f2fs_defragment_range(sbi, filp, &range);
2398 mnt_drop_write_file(filp);
2400 f2fs_update_time(sbi, REQ_TIME);
2404 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2411 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2412 struct file *file_out, loff_t pos_out, size_t len)
2414 struct inode *src = file_inode(file_in);
2415 struct inode *dst = file_inode(file_out);
2416 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2417 size_t olen = len, dst_max_i_size = 0;
2421 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2422 src->i_sb != dst->i_sb)
2425 if (unlikely(f2fs_readonly(src->i_sb)))
2428 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2431 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2435 if (pos_in == pos_out)
2437 if (pos_out > pos_in && pos_out < pos_in + len)
2444 if (!inode_trylock(dst))
2449 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2452 olen = len = src->i_size - pos_in;
2453 if (pos_in + len == src->i_size)
2454 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2460 dst_osize = dst->i_size;
2461 if (pos_out + olen > dst->i_size)
2462 dst_max_i_size = pos_out + olen;
2464 /* verify the end result is block aligned */
2465 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2466 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2467 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2470 ret = f2fs_convert_inline_inode(src);
2474 ret = f2fs_convert_inline_inode(dst);
2478 /* write out all dirty pages from offset */
2479 ret = filemap_write_and_wait_range(src->i_mapping,
2480 pos_in, pos_in + len);
2484 ret = filemap_write_and_wait_range(dst->i_mapping,
2485 pos_out, pos_out + len);
2489 f2fs_balance_fs(sbi, true);
2491 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2494 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2499 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2500 pos_out >> F2FS_BLKSIZE_BITS,
2501 len >> F2FS_BLKSIZE_BITS, false);
2505 f2fs_i_size_write(dst, dst_max_i_size);
2506 else if (dst_osize != dst->i_size)
2507 f2fs_i_size_write(dst, dst_osize);
2509 f2fs_unlock_op(sbi);
2512 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2514 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2523 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2525 struct f2fs_move_range range;
2529 if (!(filp->f_mode & FMODE_READ) ||
2530 !(filp->f_mode & FMODE_WRITE))
2533 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2537 dst = fdget(range.dst_fd);
2541 if (!(dst.file->f_mode & FMODE_WRITE)) {
2546 err = mnt_want_write_file(filp);
2550 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2551 range.pos_out, range.len);
2553 mnt_drop_write_file(filp);
2557 if (copy_to_user((struct f2fs_move_range __user *)arg,
2558 &range, sizeof(range)))
2565 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2567 struct inode *inode = file_inode(filp);
2568 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2569 struct sit_info *sm = SIT_I(sbi);
2570 unsigned int start_segno = 0, end_segno = 0;
2571 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2572 struct f2fs_flush_device range;
2575 if (!capable(CAP_SYS_ADMIN))
2578 if (f2fs_readonly(sbi->sb))
2581 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2584 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2588 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2589 __is_large_section(sbi)) {
2590 f2fs_msg(sbi->sb, KERN_WARNING,
2591 "Can't flush %u in %d for segs_per_sec %u != 1",
2592 range.dev_num, sbi->s_ndevs,
2597 ret = mnt_want_write_file(filp);
2601 if (range.dev_num != 0)
2602 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2603 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2605 start_segno = sm->last_victim[FLUSH_DEVICE];
2606 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2607 start_segno = dev_start_segno;
2608 end_segno = min(start_segno + range.segments, dev_end_segno);
2610 while (start_segno < end_segno) {
2611 if (!mutex_trylock(&sbi->gc_mutex)) {
2615 sm->last_victim[GC_CB] = end_segno + 1;
2616 sm->last_victim[GC_GREEDY] = end_segno + 1;
2617 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2618 ret = f2fs_gc(sbi, true, true, start_segno);
2626 mnt_drop_write_file(filp);
2630 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2632 struct inode *inode = file_inode(filp);
2633 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2635 /* Must validate to set it with SQLite behavior in Android. */
2636 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2638 return put_user(sb_feature, (u32 __user *)arg);
2642 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2644 struct dquot *transfer_to[MAXQUOTAS] = {};
2645 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2646 struct super_block *sb = sbi->sb;
2649 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2650 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2651 err = __dquot_transfer(inode, transfer_to);
2653 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2654 dqput(transfer_to[PRJQUOTA]);
2659 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2661 struct inode *inode = file_inode(filp);
2662 struct f2fs_inode_info *fi = F2FS_I(inode);
2663 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2668 if (!f2fs_sb_has_project_quota(sbi)) {
2669 if (projid != F2FS_DEF_PROJID)
2675 if (!f2fs_has_extra_attr(inode))
2678 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2680 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2684 /* Is it quota file? Do not allow user to mess with it */
2685 if (IS_NOQUOTA(inode))
2688 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2690 return PTR_ERR(ipage);
2692 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2695 f2fs_put_page(ipage, 1);
2698 f2fs_put_page(ipage, 1);
2700 err = dquot_initialize(inode);
2705 err = f2fs_transfer_project_quota(inode, kprojid);
2709 F2FS_I(inode)->i_projid = kprojid;
2710 inode->i_ctime = current_time(inode);
2711 f2fs_mark_inode_dirty_sync(inode, true);
2713 f2fs_unlock_op(sbi);
2717 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2722 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2724 if (projid != F2FS_DEF_PROJID)
2730 /* Transfer internal flags to xflags */
2731 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2735 if (iflags & F2FS_SYNC_FL)
2736 xflags |= FS_XFLAG_SYNC;
2737 if (iflags & F2FS_IMMUTABLE_FL)
2738 xflags |= FS_XFLAG_IMMUTABLE;
2739 if (iflags & F2FS_APPEND_FL)
2740 xflags |= FS_XFLAG_APPEND;
2741 if (iflags & F2FS_NODUMP_FL)
2742 xflags |= FS_XFLAG_NODUMP;
2743 if (iflags & F2FS_NOATIME_FL)
2744 xflags |= FS_XFLAG_NOATIME;
2745 if (iflags & F2FS_PROJINHERIT_FL)
2746 xflags |= FS_XFLAG_PROJINHERIT;
2750 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2751 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2752 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2754 /* Transfer xflags flags to internal */
2755 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2757 unsigned long iflags = 0;
2759 if (xflags & FS_XFLAG_SYNC)
2760 iflags |= F2FS_SYNC_FL;
2761 if (xflags & FS_XFLAG_IMMUTABLE)
2762 iflags |= F2FS_IMMUTABLE_FL;
2763 if (xflags & FS_XFLAG_APPEND)
2764 iflags |= F2FS_APPEND_FL;
2765 if (xflags & FS_XFLAG_NODUMP)
2766 iflags |= F2FS_NODUMP_FL;
2767 if (xflags & FS_XFLAG_NOATIME)
2768 iflags |= F2FS_NOATIME_FL;
2769 if (xflags & FS_XFLAG_PROJINHERIT)
2770 iflags |= F2FS_PROJINHERIT_FL;
2775 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2777 struct inode *inode = file_inode(filp);
2778 struct f2fs_inode_info *fi = F2FS_I(inode);
2781 memset(&fa, 0, sizeof(struct fsxattr));
2782 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2783 F2FS_FL_USER_VISIBLE);
2785 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2786 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2789 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2794 static int f2fs_ioctl_check_project(struct inode *inode, struct fsxattr *fa)
2797 * Project Quota ID state is only allowed to change from within the init
2798 * namespace. Enforce that restriction only if we are trying to change
2799 * the quota ID state. Everything else is allowed in user namespaces.
2801 if (current_user_ns() == &init_user_ns)
2804 if (__kprojid_val(F2FS_I(inode)->i_projid) != fa->fsx_projid)
2807 if (F2FS_I(inode)->i_flags & F2FS_PROJINHERIT_FL) {
2808 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT))
2811 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
2818 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2820 struct inode *inode = file_inode(filp);
2821 struct f2fs_inode_info *fi = F2FS_I(inode);
2826 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2829 /* Make sure caller has proper permission */
2830 if (!inode_owner_or_capable(inode))
2833 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2836 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2837 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2840 err = mnt_want_write_file(filp);
2845 err = f2fs_ioctl_check_project(inode, &fa);
2848 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2849 (flags & F2FS_FL_XFLAG_VISIBLE);
2850 err = __f2fs_ioc_setflags(inode, flags);
2854 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2856 inode_unlock(inode);
2857 mnt_drop_write_file(filp);
2861 int f2fs_pin_file_control(struct inode *inode, bool inc)
2863 struct f2fs_inode_info *fi = F2FS_I(inode);
2864 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2866 /* Use i_gc_failures for normal file as a risk signal. */
2868 f2fs_i_gc_failures_write(inode,
2869 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2871 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2872 f2fs_msg(sbi->sb, KERN_WARNING,
2873 "%s: Enable GC = ino %lx after %x GC trials",
2874 __func__, inode->i_ino,
2875 fi->i_gc_failures[GC_FAILURE_PIN]);
2876 clear_inode_flag(inode, FI_PIN_FILE);
2882 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2884 struct inode *inode = file_inode(filp);
2888 if (!capable(CAP_SYS_ADMIN))
2891 if (get_user(pin, (__u32 __user *)arg))
2894 if (!S_ISREG(inode->i_mode))
2897 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2900 ret = mnt_want_write_file(filp);
2906 if (f2fs_should_update_outplace(inode, NULL)) {
2912 clear_inode_flag(inode, FI_PIN_FILE);
2913 f2fs_i_gc_failures_write(inode, 0);
2917 if (f2fs_pin_file_control(inode, false)) {
2921 ret = f2fs_convert_inline_inode(inode);
2925 set_inode_flag(inode, FI_PIN_FILE);
2926 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2928 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2930 inode_unlock(inode);
2931 mnt_drop_write_file(filp);
2935 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2937 struct inode *inode = file_inode(filp);
2940 if (is_inode_flag_set(inode, FI_PIN_FILE))
2941 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2942 return put_user(pin, (u32 __user *)arg);
2945 int f2fs_precache_extents(struct inode *inode)
2947 struct f2fs_inode_info *fi = F2FS_I(inode);
2948 struct f2fs_map_blocks map;
2949 pgoff_t m_next_extent;
2953 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2957 map.m_next_pgofs = NULL;
2958 map.m_next_extent = &m_next_extent;
2959 map.m_seg_type = NO_CHECK_TYPE;
2960 map.m_may_create = false;
2961 end = F2FS_I_SB(inode)->max_file_blocks;
2963 while (map.m_lblk < end) {
2964 map.m_len = end - map.m_lblk;
2966 down_write(&fi->i_gc_rwsem[WRITE]);
2967 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2968 up_write(&fi->i_gc_rwsem[WRITE]);
2972 map.m_lblk = m_next_extent;
2978 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2980 return f2fs_precache_extents(file_inode(filp));
2983 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2985 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2989 case F2FS_IOC_GETFLAGS:
2990 return f2fs_ioc_getflags(filp, arg);
2991 case F2FS_IOC_SETFLAGS:
2992 return f2fs_ioc_setflags(filp, arg);
2993 case F2FS_IOC_GETVERSION:
2994 return f2fs_ioc_getversion(filp, arg);
2995 case F2FS_IOC_START_ATOMIC_WRITE:
2996 return f2fs_ioc_start_atomic_write(filp);
2997 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2998 return f2fs_ioc_commit_atomic_write(filp);
2999 case F2FS_IOC_START_VOLATILE_WRITE:
3000 return f2fs_ioc_start_volatile_write(filp);
3001 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3002 return f2fs_ioc_release_volatile_write(filp);
3003 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3004 return f2fs_ioc_abort_volatile_write(filp);
3005 case F2FS_IOC_SHUTDOWN:
3006 return f2fs_ioc_shutdown(filp, arg);
3008 return f2fs_ioc_fitrim(filp, arg);
3009 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3010 return f2fs_ioc_set_encryption_policy(filp, arg);
3011 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3012 return f2fs_ioc_get_encryption_policy(filp, arg);
3013 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3014 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3015 case F2FS_IOC_GARBAGE_COLLECT:
3016 return f2fs_ioc_gc(filp, arg);
3017 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3018 return f2fs_ioc_gc_range(filp, arg);
3019 case F2FS_IOC_WRITE_CHECKPOINT:
3020 return f2fs_ioc_write_checkpoint(filp, arg);
3021 case F2FS_IOC_DEFRAGMENT:
3022 return f2fs_ioc_defragment(filp, arg);
3023 case F2FS_IOC_MOVE_RANGE:
3024 return f2fs_ioc_move_range(filp, arg);
3025 case F2FS_IOC_FLUSH_DEVICE:
3026 return f2fs_ioc_flush_device(filp, arg);
3027 case F2FS_IOC_GET_FEATURES:
3028 return f2fs_ioc_get_features(filp, arg);
3029 case F2FS_IOC_FSGETXATTR:
3030 return f2fs_ioc_fsgetxattr(filp, arg);
3031 case F2FS_IOC_FSSETXATTR:
3032 return f2fs_ioc_fssetxattr(filp, arg);
3033 case F2FS_IOC_GET_PIN_FILE:
3034 return f2fs_ioc_get_pin_file(filp, arg);
3035 case F2FS_IOC_SET_PIN_FILE:
3036 return f2fs_ioc_set_pin_file(filp, arg);
3037 case F2FS_IOC_PRECACHE_EXTENTS:
3038 return f2fs_ioc_precache_extents(filp, arg);
3044 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3046 struct file *file = iocb->ki_filp;
3047 struct inode *inode = file_inode(file);
3050 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3055 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT)) {
3060 if (!inode_trylock(inode)) {
3061 if (iocb->ki_flags & IOCB_NOWAIT) {
3068 ret = generic_write_checks(iocb, from);
3070 bool preallocated = false;
3071 size_t target_size = 0;
3074 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3075 set_inode_flag(inode, FI_NO_PREALLOC);
3077 if ((iocb->ki_flags & IOCB_NOWAIT)) {
3078 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3079 iov_iter_count(from)) ||
3080 f2fs_has_inline_data(inode) ||
3081 f2fs_force_buffered_io(inode, iocb, from)) {
3082 clear_inode_flag(inode, FI_NO_PREALLOC);
3083 inode_unlock(inode);
3088 preallocated = true;
3089 target_size = iocb->ki_pos + iov_iter_count(from);
3091 err = f2fs_preallocate_blocks(iocb, from);
3093 clear_inode_flag(inode, FI_NO_PREALLOC);
3094 inode_unlock(inode);
3099 ret = __generic_file_write_iter(iocb, from);
3100 clear_inode_flag(inode, FI_NO_PREALLOC);
3102 /* if we couldn't write data, we should deallocate blocks. */
3103 if (preallocated && i_size_read(inode) < target_size)
3104 f2fs_truncate(inode);
3107 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3109 inode_unlock(inode);
3111 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3112 iov_iter_count(from), ret);
3114 ret = generic_write_sync(iocb, ret);
3118 #ifdef CONFIG_COMPAT
3119 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3122 case F2FS_IOC32_GETFLAGS:
3123 cmd = F2FS_IOC_GETFLAGS;
3125 case F2FS_IOC32_SETFLAGS:
3126 cmd = F2FS_IOC_SETFLAGS;
3128 case F2FS_IOC32_GETVERSION:
3129 cmd = F2FS_IOC_GETVERSION;
3131 case F2FS_IOC_START_ATOMIC_WRITE:
3132 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3133 case F2FS_IOC_START_VOLATILE_WRITE:
3134 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3135 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3136 case F2FS_IOC_SHUTDOWN:
3137 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3138 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3139 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3140 case F2FS_IOC_GARBAGE_COLLECT:
3141 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3142 case F2FS_IOC_WRITE_CHECKPOINT:
3143 case F2FS_IOC_DEFRAGMENT:
3144 case F2FS_IOC_MOVE_RANGE:
3145 case F2FS_IOC_FLUSH_DEVICE:
3146 case F2FS_IOC_GET_FEATURES:
3147 case F2FS_IOC_FSGETXATTR:
3148 case F2FS_IOC_FSSETXATTR:
3149 case F2FS_IOC_GET_PIN_FILE:
3150 case F2FS_IOC_SET_PIN_FILE:
3151 case F2FS_IOC_PRECACHE_EXTENTS:
3154 return -ENOIOCTLCMD;
3156 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3160 const struct file_operations f2fs_file_operations = {
3161 .llseek = f2fs_llseek,
3162 .read_iter = generic_file_read_iter,
3163 .write_iter = f2fs_file_write_iter,
3164 .open = f2fs_file_open,
3165 .release = f2fs_release_file,
3166 .mmap = f2fs_file_mmap,
3167 .flush = f2fs_file_flush,
3168 .fsync = f2fs_sync_file,
3169 .fallocate = f2fs_fallocate,
3170 .unlocked_ioctl = f2fs_ioctl,
3171 #ifdef CONFIG_COMPAT
3172 .compat_ioctl = f2fs_compat_ioctl,
3174 .splice_read = generic_file_splice_read,
3175 .splice_write = iter_file_splice_write,