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);
45 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
47 struct page *page = vmf->page;
48 struct inode *inode = file_inode(vmf->vma->vm_file);
49 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
50 struct dnode_of_data dn = { .node_changed = false };
53 if (unlikely(f2fs_cp_error(sbi))) {
58 sb_start_pagefault(inode->i_sb);
60 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
62 file_update_time(vmf->vma->vm_file);
63 down_read(&F2FS_I(inode)->i_mmap_sem);
65 if (unlikely(page->mapping != inode->i_mapping ||
66 page_offset(page) > i_size_read(inode) ||
67 !PageUptodate(page))) {
73 /* block allocation */
74 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
75 set_new_dnode(&dn, inode, NULL, NULL, 0);
76 err = f2fs_get_block(&dn, page->index);
78 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
85 f2fs_wait_on_page_writeback(page, DATA, false, true);
87 /* wait for GCed page writeback via META_MAPPING */
88 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
91 * check to see if the page is mapped already (no holes)
93 if (PageMappedToDisk(page))
96 /* page is wholly or partially inside EOF */
97 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
101 offset = i_size_read(inode) & ~PAGE_MASK;
102 zero_user_segment(page, offset, PAGE_SIZE);
104 set_page_dirty(page);
105 if (!PageUptodate(page))
106 SetPageUptodate(page);
108 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
109 f2fs_update_time(sbi, REQ_TIME);
111 trace_f2fs_vm_page_mkwrite(page, DATA);
113 up_read(&F2FS_I(inode)->i_mmap_sem);
115 f2fs_balance_fs(sbi, dn.node_changed);
117 sb_end_pagefault(inode->i_sb);
119 return block_page_mkwrite_return(err);
122 static const struct vm_operations_struct f2fs_file_vm_ops = {
123 .fault = f2fs_filemap_fault,
124 .map_pages = filemap_map_pages,
125 .page_mkwrite = f2fs_vm_page_mkwrite,
128 static int get_parent_ino(struct inode *inode, nid_t *pino)
130 struct dentry *dentry;
132 inode = igrab(inode);
133 dentry = d_find_any_alias(inode);
138 *pino = parent_ino(dentry);
143 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
145 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
146 enum cp_reason_type cp_reason = CP_NO_NEEDED;
148 if (!S_ISREG(inode->i_mode))
149 cp_reason = CP_NON_REGULAR;
150 else if (inode->i_nlink != 1)
151 cp_reason = CP_HARDLINK;
152 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
153 cp_reason = CP_SB_NEED_CP;
154 else if (file_wrong_pino(inode))
155 cp_reason = CP_WRONG_PINO;
156 else if (!f2fs_space_for_roll_forward(sbi))
157 cp_reason = CP_NO_SPC_ROLL;
158 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
159 cp_reason = CP_NODE_NEED_CP;
160 else if (test_opt(sbi, FASTBOOT))
161 cp_reason = CP_FASTBOOT_MODE;
162 else if (F2FS_OPTION(sbi).active_logs == 2)
163 cp_reason = CP_SPEC_LOG_NUM;
164 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
165 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
166 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
168 cp_reason = CP_RECOVER_DIR;
173 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
175 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
177 /* But we need to avoid that there are some inode updates */
178 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
184 static void try_to_fix_pino(struct inode *inode)
186 struct f2fs_inode_info *fi = F2FS_I(inode);
189 down_write(&fi->i_sem);
190 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
191 get_parent_ino(inode, &pino)) {
192 f2fs_i_pino_write(inode, pino);
193 file_got_pino(inode);
195 up_write(&fi->i_sem);
198 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
199 int datasync, bool atomic)
201 struct inode *inode = file->f_mapping->host;
202 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
203 nid_t ino = inode->i_ino;
205 enum cp_reason_type cp_reason = 0;
206 struct writeback_control wbc = {
207 .sync_mode = WB_SYNC_ALL,
208 .nr_to_write = LONG_MAX,
211 unsigned int seq_id = 0;
213 if (unlikely(f2fs_readonly(inode->i_sb) ||
214 is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
217 trace_f2fs_sync_file_enter(inode);
219 if (S_ISDIR(inode->i_mode))
222 /* if fdatasync is triggered, let's do in-place-update */
223 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
224 set_inode_flag(inode, FI_NEED_IPU);
225 ret = file_write_and_wait_range(file, start, end);
226 clear_inode_flag(inode, FI_NEED_IPU);
229 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
233 /* if the inode is dirty, let's recover all the time */
234 if (!f2fs_skip_inode_update(inode, datasync)) {
235 f2fs_write_inode(inode, NULL);
240 * if there is no written data, don't waste time to write recovery info.
242 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
243 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
245 /* it may call write_inode just prior to fsync */
246 if (need_inode_page_update(sbi, ino))
249 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
250 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
256 * Both of fdatasync() and fsync() are able to be recovered from
259 down_read(&F2FS_I(inode)->i_sem);
260 cp_reason = need_do_checkpoint(inode);
261 up_read(&F2FS_I(inode)->i_sem);
264 /* all the dirty node pages should be flushed for POR */
265 ret = f2fs_sync_fs(inode->i_sb, 1);
268 * We've secured consistency through sync_fs. Following pino
269 * will be used only for fsynced inodes after checkpoint.
271 try_to_fix_pino(inode);
272 clear_inode_flag(inode, FI_APPEND_WRITE);
273 clear_inode_flag(inode, FI_UPDATE_WRITE);
277 atomic_inc(&sbi->wb_sync_req[NODE]);
278 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
279 atomic_dec(&sbi->wb_sync_req[NODE]);
283 /* if cp_error was enabled, we should avoid infinite loop */
284 if (unlikely(f2fs_cp_error(sbi))) {
289 if (f2fs_need_inode_block_update(sbi, ino)) {
290 f2fs_mark_inode_dirty_sync(inode, true);
291 f2fs_write_inode(inode, NULL);
296 * If it's atomic_write, it's just fine to keep write ordering. So
297 * here we don't need to wait for node write completion, since we use
298 * node chain which serializes node blocks. If one of node writes are
299 * reordered, we can see simply broken chain, resulting in stopping
300 * roll-forward recovery. It means we'll recover all or none node blocks
304 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
309 /* once recovery info is written, don't need to tack this */
310 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
311 clear_inode_flag(inode, FI_APPEND_WRITE);
313 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
314 ret = f2fs_issue_flush(sbi, inode->i_ino);
316 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
317 clear_inode_flag(inode, FI_UPDATE_WRITE);
318 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
320 f2fs_update_time(sbi, REQ_TIME);
322 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
323 f2fs_trace_ios(NULL, 1);
327 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
329 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
331 return f2fs_do_sync_file(file, start, end, datasync, false);
334 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
335 pgoff_t pgofs, int whence)
340 if (whence != SEEK_DATA)
343 /* find first dirty page index */
344 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
353 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
354 pgoff_t dirty, pgoff_t pgofs, int whence)
358 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
359 is_valid_data_blkaddr(sbi, blkaddr))
363 if (blkaddr == NULL_ADDR)
370 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
372 struct inode *inode = file->f_mapping->host;
373 loff_t maxbytes = inode->i_sb->s_maxbytes;
374 struct dnode_of_data dn;
375 pgoff_t pgofs, end_offset, dirty;
376 loff_t data_ofs = offset;
382 isize = i_size_read(inode);
386 /* handle inline data case */
387 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
388 if (whence == SEEK_HOLE)
393 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
395 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
397 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
398 set_new_dnode(&dn, inode, NULL, NULL, 0);
399 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
400 if (err && err != -ENOENT) {
402 } else if (err == -ENOENT) {
403 /* direct node does not exists */
404 if (whence == SEEK_DATA) {
405 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
412 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
414 /* find data/hole in dnode block */
415 for (; dn.ofs_in_node < end_offset;
416 dn.ofs_in_node++, pgofs++,
417 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
420 blkaddr = datablock_addr(dn.inode,
421 dn.node_page, dn.ofs_in_node);
423 if (__is_valid_data_blkaddr(blkaddr) &&
424 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
425 blkaddr, DATA_GENERIC)) {
430 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
439 if (whence == SEEK_DATA)
442 if (whence == SEEK_HOLE && data_ofs > isize)
445 return vfs_setpos(file, data_ofs, maxbytes);
451 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
453 struct inode *inode = file->f_mapping->host;
454 loff_t maxbytes = inode->i_sb->s_maxbytes;
460 return generic_file_llseek_size(file, offset, whence,
461 maxbytes, i_size_read(inode));
466 return f2fs_seek_block(file, offset, whence);
472 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
474 struct inode *inode = file_inode(file);
477 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
480 /* we don't need to use inline_data strictly */
481 err = f2fs_convert_inline_inode(inode);
486 vma->vm_ops = &f2fs_file_vm_ops;
490 static int f2fs_file_open(struct inode *inode, struct file *filp)
492 int err = fscrypt_file_open(inode, filp);
497 filp->f_mode |= FMODE_NOWAIT;
499 return dquot_file_open(inode, filp);
502 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
504 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
505 struct f2fs_node *raw_node;
506 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
510 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
511 base = get_extra_isize(dn->inode);
513 raw_node = F2FS_NODE(dn->node_page);
514 addr = blkaddr_in_node(raw_node) + base + ofs;
516 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
517 block_t blkaddr = le32_to_cpu(*addr);
519 if (blkaddr == NULL_ADDR)
522 dn->data_blkaddr = NULL_ADDR;
523 f2fs_set_data_blkaddr(dn);
525 if (__is_valid_data_blkaddr(blkaddr) &&
526 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
529 f2fs_invalidate_blocks(sbi, blkaddr);
530 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
531 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
538 * once we invalidate valid blkaddr in range [ofs, ofs + count],
539 * we will invalidate all blkaddr in the whole range.
541 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
543 f2fs_update_extent_cache_range(dn, fofs, 0, len);
544 dec_valid_block_count(sbi, dn->inode, nr_free);
546 dn->ofs_in_node = ofs;
548 f2fs_update_time(sbi, REQ_TIME);
549 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
550 dn->ofs_in_node, nr_free);
553 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
555 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
558 static int truncate_partial_data_page(struct inode *inode, u64 from,
561 loff_t offset = from & (PAGE_SIZE - 1);
562 pgoff_t index = from >> PAGE_SHIFT;
563 struct address_space *mapping = inode->i_mapping;
566 if (!offset && !cache_only)
570 page = find_lock_page(mapping, index);
571 if (page && PageUptodate(page))
573 f2fs_put_page(page, 1);
577 page = f2fs_get_lock_data_page(inode, index, true);
579 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
581 f2fs_wait_on_page_writeback(page, DATA, true, true);
582 zero_user(page, offset, PAGE_SIZE - offset);
584 /* An encrypted inode should have a key and truncate the last page. */
585 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
587 set_page_dirty(page);
588 f2fs_put_page(page, 1);
592 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
594 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
595 struct dnode_of_data dn;
597 int count = 0, err = 0;
599 bool truncate_page = false;
601 trace_f2fs_truncate_blocks_enter(inode, from);
603 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
605 if (free_from >= sbi->max_file_blocks)
611 ipage = f2fs_get_node_page(sbi, inode->i_ino);
613 err = PTR_ERR(ipage);
617 if (f2fs_has_inline_data(inode)) {
618 f2fs_truncate_inline_inode(inode, ipage, from);
619 f2fs_put_page(ipage, 1);
620 truncate_page = true;
624 set_new_dnode(&dn, inode, ipage, NULL, 0);
625 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
632 count = ADDRS_PER_PAGE(dn.node_page, inode);
634 count -= dn.ofs_in_node;
635 f2fs_bug_on(sbi, count < 0);
637 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
638 f2fs_truncate_data_blocks_range(&dn, count);
644 err = f2fs_truncate_inode_blocks(inode, free_from);
649 /* lastly zero out the first data page */
651 err = truncate_partial_data_page(inode, from, truncate_page);
653 trace_f2fs_truncate_blocks_exit(inode, err);
657 int f2fs_truncate(struct inode *inode)
661 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
664 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
665 S_ISLNK(inode->i_mode)))
668 trace_f2fs_truncate(inode);
670 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
671 f2fs_show_injection_info(FAULT_TRUNCATE);
675 /* we should check inline_data size */
676 if (!f2fs_may_inline_data(inode)) {
677 err = f2fs_convert_inline_inode(inode);
682 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
686 inode->i_mtime = inode->i_ctime = current_time(inode);
687 f2fs_mark_inode_dirty_sync(inode, false);
691 int f2fs_getattr(const struct path *path, struct kstat *stat,
692 u32 request_mask, unsigned int query_flags)
694 struct inode *inode = d_inode(path->dentry);
695 struct f2fs_inode_info *fi = F2FS_I(inode);
696 struct f2fs_inode *ri;
699 if (f2fs_has_extra_attr(inode) &&
700 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
701 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
702 stat->result_mask |= STATX_BTIME;
703 stat->btime.tv_sec = fi->i_crtime.tv_sec;
704 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
707 flags = fi->i_flags & F2FS_FL_USER_VISIBLE;
708 if (flags & F2FS_APPEND_FL)
709 stat->attributes |= STATX_ATTR_APPEND;
710 if (flags & F2FS_COMPR_FL)
711 stat->attributes |= STATX_ATTR_COMPRESSED;
712 if (IS_ENCRYPTED(inode))
713 stat->attributes |= STATX_ATTR_ENCRYPTED;
714 if (flags & F2FS_IMMUTABLE_FL)
715 stat->attributes |= STATX_ATTR_IMMUTABLE;
716 if (flags & F2FS_NODUMP_FL)
717 stat->attributes |= STATX_ATTR_NODUMP;
719 stat->attributes_mask |= (STATX_ATTR_APPEND |
720 STATX_ATTR_COMPRESSED |
721 STATX_ATTR_ENCRYPTED |
722 STATX_ATTR_IMMUTABLE |
725 generic_fillattr(inode, stat);
727 /* we need to show initial sectors used for inline_data/dentries */
728 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
729 f2fs_has_inline_dentry(inode))
730 stat->blocks += (stat->size + 511) >> 9;
735 #ifdef CONFIG_F2FS_FS_POSIX_ACL
736 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
738 unsigned int ia_valid = attr->ia_valid;
740 if (ia_valid & ATTR_UID)
741 inode->i_uid = attr->ia_uid;
742 if (ia_valid & ATTR_GID)
743 inode->i_gid = attr->ia_gid;
744 if (ia_valid & ATTR_ATIME)
745 inode->i_atime = timespec64_trunc(attr->ia_atime,
746 inode->i_sb->s_time_gran);
747 if (ia_valid & ATTR_MTIME)
748 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
749 inode->i_sb->s_time_gran);
750 if (ia_valid & ATTR_CTIME)
751 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
752 inode->i_sb->s_time_gran);
753 if (ia_valid & ATTR_MODE) {
754 umode_t mode = attr->ia_mode;
756 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
758 set_acl_inode(inode, mode);
762 #define __setattr_copy setattr_copy
765 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
767 struct inode *inode = d_inode(dentry);
770 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
773 err = setattr_prepare(dentry, attr);
777 err = fscrypt_prepare_setattr(dentry, attr);
781 if (is_quota_modification(inode, attr)) {
782 err = dquot_initialize(inode);
786 if ((attr->ia_valid & ATTR_UID &&
787 !uid_eq(attr->ia_uid, inode->i_uid)) ||
788 (attr->ia_valid & ATTR_GID &&
789 !gid_eq(attr->ia_gid, inode->i_gid))) {
790 f2fs_lock_op(F2FS_I_SB(inode));
791 err = dquot_transfer(inode, attr);
793 set_sbi_flag(F2FS_I_SB(inode),
794 SBI_QUOTA_NEED_REPAIR);
795 f2fs_unlock_op(F2FS_I_SB(inode));
799 * update uid/gid under lock_op(), so that dquot and inode can
800 * be updated atomically.
802 if (attr->ia_valid & ATTR_UID)
803 inode->i_uid = attr->ia_uid;
804 if (attr->ia_valid & ATTR_GID)
805 inode->i_gid = attr->ia_gid;
806 f2fs_mark_inode_dirty_sync(inode, true);
807 f2fs_unlock_op(F2FS_I_SB(inode));
810 if (attr->ia_valid & ATTR_SIZE) {
811 bool to_smaller = (attr->ia_size <= i_size_read(inode));
813 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
814 down_write(&F2FS_I(inode)->i_mmap_sem);
816 truncate_setsize(inode, attr->ia_size);
819 err = f2fs_truncate(inode);
821 * do not trim all blocks after i_size if target size is
822 * larger than i_size.
824 up_write(&F2FS_I(inode)->i_mmap_sem);
825 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
831 /* should convert inline inode here */
832 if (!f2fs_may_inline_data(inode)) {
833 err = f2fs_convert_inline_inode(inode);
837 inode->i_mtime = inode->i_ctime = current_time(inode);
840 down_write(&F2FS_I(inode)->i_sem);
841 F2FS_I(inode)->last_disk_size = i_size_read(inode);
842 up_write(&F2FS_I(inode)->i_sem);
845 __setattr_copy(inode, attr);
847 if (attr->ia_valid & ATTR_MODE) {
848 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
849 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
850 inode->i_mode = F2FS_I(inode)->i_acl_mode;
851 clear_inode_flag(inode, FI_ACL_MODE);
855 /* file size may changed here */
856 f2fs_mark_inode_dirty_sync(inode, true);
858 /* inode change will produce dirty node pages flushed by checkpoint */
859 f2fs_balance_fs(F2FS_I_SB(inode), true);
864 const struct inode_operations f2fs_file_inode_operations = {
865 .getattr = f2fs_getattr,
866 .setattr = f2fs_setattr,
867 .get_acl = f2fs_get_acl,
868 .set_acl = f2fs_set_acl,
869 #ifdef CONFIG_F2FS_FS_XATTR
870 .listxattr = f2fs_listxattr,
872 .fiemap = f2fs_fiemap,
875 static int fill_zero(struct inode *inode, pgoff_t index,
876 loff_t start, loff_t len)
878 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
884 f2fs_balance_fs(sbi, true);
887 page = f2fs_get_new_data_page(inode, NULL, index, false);
891 return PTR_ERR(page);
893 f2fs_wait_on_page_writeback(page, DATA, true, true);
894 zero_user(page, start, len);
895 set_page_dirty(page);
896 f2fs_put_page(page, 1);
900 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
904 while (pg_start < pg_end) {
905 struct dnode_of_data dn;
906 pgoff_t end_offset, count;
908 set_new_dnode(&dn, inode, NULL, NULL, 0);
909 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
911 if (err == -ENOENT) {
912 pg_start = f2fs_get_next_page_offset(&dn,
919 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
920 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
922 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
924 f2fs_truncate_data_blocks_range(&dn, count);
932 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
934 pgoff_t pg_start, pg_end;
935 loff_t off_start, off_end;
938 ret = f2fs_convert_inline_inode(inode);
942 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
943 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
945 off_start = offset & (PAGE_SIZE - 1);
946 off_end = (offset + len) & (PAGE_SIZE - 1);
948 if (pg_start == pg_end) {
949 ret = fill_zero(inode, pg_start, off_start,
950 off_end - off_start);
955 ret = fill_zero(inode, pg_start++, off_start,
956 PAGE_SIZE - off_start);
961 ret = fill_zero(inode, pg_end, 0, off_end);
966 if (pg_start < pg_end) {
967 struct address_space *mapping = inode->i_mapping;
968 loff_t blk_start, blk_end;
969 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
971 f2fs_balance_fs(sbi, true);
973 blk_start = (loff_t)pg_start << PAGE_SHIFT;
974 blk_end = (loff_t)pg_end << PAGE_SHIFT;
976 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
977 down_write(&F2FS_I(inode)->i_mmap_sem);
979 truncate_inode_pages_range(mapping, blk_start,
983 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
986 up_write(&F2FS_I(inode)->i_mmap_sem);
987 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
994 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
995 int *do_replace, pgoff_t off, pgoff_t len)
997 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
998 struct dnode_of_data dn;
1002 set_new_dnode(&dn, inode, NULL, NULL, 0);
1003 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1004 if (ret && ret != -ENOENT) {
1006 } else if (ret == -ENOENT) {
1007 if (dn.max_level == 0)
1009 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
1015 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1016 dn.ofs_in_node, len);
1017 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1018 *blkaddr = datablock_addr(dn.inode,
1019 dn.node_page, dn.ofs_in_node);
1020 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1022 if (test_opt(sbi, LFS)) {
1023 f2fs_put_dnode(&dn);
1027 /* do not invalidate this block address */
1028 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1032 f2fs_put_dnode(&dn);
1041 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1042 int *do_replace, pgoff_t off, int len)
1044 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1045 struct dnode_of_data dn;
1048 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1049 if (*do_replace == 0)
1052 set_new_dnode(&dn, inode, NULL, NULL, 0);
1053 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1055 dec_valid_block_count(sbi, inode, 1);
1056 f2fs_invalidate_blocks(sbi, *blkaddr);
1058 f2fs_update_data_blkaddr(&dn, *blkaddr);
1060 f2fs_put_dnode(&dn);
1065 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1066 block_t *blkaddr, int *do_replace,
1067 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1069 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1074 if (blkaddr[i] == NULL_ADDR && !full) {
1079 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1080 struct dnode_of_data dn;
1081 struct node_info ni;
1085 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1086 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1090 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1092 f2fs_put_dnode(&dn);
1096 ilen = min((pgoff_t)
1097 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1098 dn.ofs_in_node, len - i);
1100 dn.data_blkaddr = datablock_addr(dn.inode,
1101 dn.node_page, dn.ofs_in_node);
1102 f2fs_truncate_data_blocks_range(&dn, 1);
1104 if (do_replace[i]) {
1105 f2fs_i_blocks_write(src_inode,
1107 f2fs_i_blocks_write(dst_inode,
1109 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1110 blkaddr[i], ni.version, true, false);
1116 new_size = (dst + i) << PAGE_SHIFT;
1117 if (dst_inode->i_size < new_size)
1118 f2fs_i_size_write(dst_inode, new_size);
1119 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1121 f2fs_put_dnode(&dn);
1123 struct page *psrc, *pdst;
1125 psrc = f2fs_get_lock_data_page(src_inode,
1128 return PTR_ERR(psrc);
1129 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1132 f2fs_put_page(psrc, 1);
1133 return PTR_ERR(pdst);
1135 f2fs_copy_page(psrc, pdst);
1136 set_page_dirty(pdst);
1137 f2fs_put_page(pdst, 1);
1138 f2fs_put_page(psrc, 1);
1140 ret = f2fs_truncate_hole(src_inode,
1141 src + i, src + i + 1);
1150 static int __exchange_data_block(struct inode *src_inode,
1151 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1152 pgoff_t len, bool full)
1154 block_t *src_blkaddr;
1160 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1162 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1163 array_size(olen, sizeof(block_t)),
1168 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1169 array_size(olen, sizeof(int)),
1172 kvfree(src_blkaddr);
1176 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1177 do_replace, src, olen);
1181 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1182 do_replace, src, dst, olen, full);
1190 kvfree(src_blkaddr);
1196 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1197 kvfree(src_blkaddr);
1202 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1204 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1205 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1206 pgoff_t start = offset >> PAGE_SHIFT;
1207 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1210 f2fs_balance_fs(sbi, true);
1212 /* avoid gc operation during block exchange */
1213 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1214 down_write(&F2FS_I(inode)->i_mmap_sem);
1217 f2fs_drop_extent_tree(inode);
1218 truncate_pagecache(inode, offset);
1219 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1220 f2fs_unlock_op(sbi);
1222 up_write(&F2FS_I(inode)->i_mmap_sem);
1223 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1227 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1232 if (offset + len >= i_size_read(inode))
1235 /* collapse range should be aligned to block size of f2fs. */
1236 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1239 ret = f2fs_convert_inline_inode(inode);
1243 /* write out all dirty pages from offset */
1244 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1248 ret = f2fs_do_collapse(inode, offset, len);
1252 /* write out all moved pages, if possible */
1253 down_write(&F2FS_I(inode)->i_mmap_sem);
1254 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1255 truncate_pagecache(inode, offset);
1257 new_size = i_size_read(inode) - len;
1258 truncate_pagecache(inode, new_size);
1260 ret = f2fs_truncate_blocks(inode, new_size, true);
1261 up_write(&F2FS_I(inode)->i_mmap_sem);
1263 f2fs_i_size_write(inode, new_size);
1267 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1270 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1271 pgoff_t index = start;
1272 unsigned int ofs_in_node = dn->ofs_in_node;
1276 for (; index < end; index++, dn->ofs_in_node++) {
1277 if (datablock_addr(dn->inode, dn->node_page,
1278 dn->ofs_in_node) == NULL_ADDR)
1282 dn->ofs_in_node = ofs_in_node;
1283 ret = f2fs_reserve_new_blocks(dn, count);
1287 dn->ofs_in_node = ofs_in_node;
1288 for (index = start; index < end; index++, dn->ofs_in_node++) {
1289 dn->data_blkaddr = datablock_addr(dn->inode,
1290 dn->node_page, dn->ofs_in_node);
1292 * f2fs_reserve_new_blocks will not guarantee entire block
1295 if (dn->data_blkaddr == NULL_ADDR) {
1299 if (dn->data_blkaddr != NEW_ADDR) {
1300 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1301 dn->data_blkaddr = NEW_ADDR;
1302 f2fs_set_data_blkaddr(dn);
1306 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1311 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1314 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1315 struct address_space *mapping = inode->i_mapping;
1316 pgoff_t index, pg_start, pg_end;
1317 loff_t new_size = i_size_read(inode);
1318 loff_t off_start, off_end;
1321 ret = inode_newsize_ok(inode, (len + offset));
1325 ret = f2fs_convert_inline_inode(inode);
1329 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1333 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1334 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1336 off_start = offset & (PAGE_SIZE - 1);
1337 off_end = (offset + len) & (PAGE_SIZE - 1);
1339 if (pg_start == pg_end) {
1340 ret = fill_zero(inode, pg_start, off_start,
1341 off_end - off_start);
1345 new_size = max_t(loff_t, new_size, offset + len);
1348 ret = fill_zero(inode, pg_start++, off_start,
1349 PAGE_SIZE - off_start);
1353 new_size = max_t(loff_t, new_size,
1354 (loff_t)pg_start << PAGE_SHIFT);
1357 for (index = pg_start; index < pg_end;) {
1358 struct dnode_of_data dn;
1359 unsigned int end_offset;
1362 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1363 down_write(&F2FS_I(inode)->i_mmap_sem);
1365 truncate_pagecache_range(inode,
1366 (loff_t)index << PAGE_SHIFT,
1367 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1371 set_new_dnode(&dn, inode, NULL, NULL, 0);
1372 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1374 f2fs_unlock_op(sbi);
1375 up_write(&F2FS_I(inode)->i_mmap_sem);
1376 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1380 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1381 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1383 ret = f2fs_do_zero_range(&dn, index, end);
1384 f2fs_put_dnode(&dn);
1386 f2fs_unlock_op(sbi);
1387 up_write(&F2FS_I(inode)->i_mmap_sem);
1388 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1390 f2fs_balance_fs(sbi, dn.node_changed);
1396 new_size = max_t(loff_t, new_size,
1397 (loff_t)index << PAGE_SHIFT);
1401 ret = fill_zero(inode, pg_end, 0, off_end);
1405 new_size = max_t(loff_t, new_size, offset + len);
1410 if (new_size > i_size_read(inode)) {
1411 if (mode & FALLOC_FL_KEEP_SIZE)
1412 file_set_keep_isize(inode);
1414 f2fs_i_size_write(inode, new_size);
1419 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1421 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1422 pgoff_t nr, pg_start, pg_end, delta, idx;
1426 new_size = i_size_read(inode) + len;
1427 ret = inode_newsize_ok(inode, new_size);
1431 if (offset >= i_size_read(inode))
1434 /* insert range should be aligned to block size of f2fs. */
1435 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1438 ret = f2fs_convert_inline_inode(inode);
1442 f2fs_balance_fs(sbi, true);
1444 down_write(&F2FS_I(inode)->i_mmap_sem);
1445 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1446 up_write(&F2FS_I(inode)->i_mmap_sem);
1450 /* write out all dirty pages from offset */
1451 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1455 pg_start = offset >> PAGE_SHIFT;
1456 pg_end = (offset + len) >> PAGE_SHIFT;
1457 delta = pg_end - pg_start;
1458 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1460 /* avoid gc operation during block exchange */
1461 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1462 down_write(&F2FS_I(inode)->i_mmap_sem);
1463 truncate_pagecache(inode, offset);
1465 while (!ret && idx > pg_start) {
1466 nr = idx - pg_start;
1472 f2fs_drop_extent_tree(inode);
1474 ret = __exchange_data_block(inode, inode, idx,
1475 idx + delta, nr, false);
1476 f2fs_unlock_op(sbi);
1478 up_write(&F2FS_I(inode)->i_mmap_sem);
1479 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1481 /* write out all moved pages, if possible */
1482 down_write(&F2FS_I(inode)->i_mmap_sem);
1483 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1484 truncate_pagecache(inode, offset);
1485 up_write(&F2FS_I(inode)->i_mmap_sem);
1488 f2fs_i_size_write(inode, new_size);
1492 static int expand_inode_data(struct inode *inode, loff_t offset,
1493 loff_t len, int mode)
1495 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1496 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1497 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1498 .m_may_create = true };
1500 loff_t new_size = i_size_read(inode);
1504 err = inode_newsize_ok(inode, (len + offset));
1508 err = f2fs_convert_inline_inode(inode);
1512 f2fs_balance_fs(sbi, true);
1514 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1515 off_end = (offset + len) & (PAGE_SIZE - 1);
1517 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1518 map.m_len = pg_end - map.m_lblk;
1522 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1529 last_off = map.m_lblk + map.m_len - 1;
1531 /* update new size to the failed position */
1532 new_size = (last_off == pg_end) ? offset + len :
1533 (loff_t)(last_off + 1) << PAGE_SHIFT;
1535 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1538 if (new_size > i_size_read(inode)) {
1539 if (mode & FALLOC_FL_KEEP_SIZE)
1540 file_set_keep_isize(inode);
1542 f2fs_i_size_write(inode, new_size);
1548 static long f2fs_fallocate(struct file *file, int mode,
1549 loff_t offset, loff_t len)
1551 struct inode *inode = file_inode(file);
1554 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1557 /* f2fs only support ->fallocate for regular file */
1558 if (!S_ISREG(inode->i_mode))
1561 if (IS_ENCRYPTED(inode) &&
1562 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1565 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1566 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1567 FALLOC_FL_INSERT_RANGE))
1572 if (mode & FALLOC_FL_PUNCH_HOLE) {
1573 if (offset >= inode->i_size)
1576 ret = punch_hole(inode, offset, len);
1577 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1578 ret = f2fs_collapse_range(inode, offset, len);
1579 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1580 ret = f2fs_zero_range(inode, offset, len, mode);
1581 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1582 ret = f2fs_insert_range(inode, offset, len);
1584 ret = expand_inode_data(inode, offset, len, mode);
1588 inode->i_mtime = inode->i_ctime = current_time(inode);
1589 f2fs_mark_inode_dirty_sync(inode, false);
1590 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1594 inode_unlock(inode);
1596 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1600 static int f2fs_release_file(struct inode *inode, struct file *filp)
1603 * f2fs_relase_file is called at every close calls. So we should
1604 * not drop any inmemory pages by close called by other process.
1606 if (!(filp->f_mode & FMODE_WRITE) ||
1607 atomic_read(&inode->i_writecount) != 1)
1610 /* some remained atomic pages should discarded */
1611 if (f2fs_is_atomic_file(inode))
1612 f2fs_drop_inmem_pages(inode);
1613 if (f2fs_is_volatile_file(inode)) {
1614 set_inode_flag(inode, FI_DROP_CACHE);
1615 filemap_fdatawrite(inode->i_mapping);
1616 clear_inode_flag(inode, FI_DROP_CACHE);
1617 clear_inode_flag(inode, FI_VOLATILE_FILE);
1618 stat_dec_volatile_write(inode);
1623 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1625 struct inode *inode = file_inode(file);
1628 * If the process doing a transaction is crashed, we should do
1629 * roll-back. Otherwise, other reader/write can see corrupted database
1630 * until all the writers close its file. Since this should be done
1631 * before dropping file lock, it needs to do in ->flush.
1633 if (f2fs_is_atomic_file(inode) &&
1634 F2FS_I(inode)->inmem_task == current)
1635 f2fs_drop_inmem_pages(inode);
1639 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1641 struct inode *inode = file_inode(filp);
1642 struct f2fs_inode_info *fi = F2FS_I(inode);
1643 unsigned int flags = fi->i_flags;
1645 if (IS_ENCRYPTED(inode))
1646 flags |= F2FS_ENCRYPT_FL;
1647 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1648 flags |= F2FS_INLINE_DATA_FL;
1649 if (is_inode_flag_set(inode, FI_PIN_FILE))
1650 flags |= F2FS_NOCOW_FL;
1652 flags &= F2FS_FL_USER_VISIBLE;
1654 return put_user(flags, (int __user *)arg);
1657 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1659 struct f2fs_inode_info *fi = F2FS_I(inode);
1660 unsigned int oldflags;
1662 /* Is it quota file? Do not allow user to mess with it */
1663 if (IS_NOQUOTA(inode))
1666 flags = f2fs_mask_flags(inode->i_mode, flags);
1668 oldflags = fi->i_flags;
1670 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL))
1671 if (!capable(CAP_LINUX_IMMUTABLE))
1674 flags = flags & F2FS_FL_USER_MODIFIABLE;
1675 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE;
1676 fi->i_flags = flags;
1678 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1679 set_inode_flag(inode, FI_PROJ_INHERIT);
1681 clear_inode_flag(inode, FI_PROJ_INHERIT);
1683 inode->i_ctime = current_time(inode);
1684 f2fs_set_inode_flags(inode);
1685 f2fs_mark_inode_dirty_sync(inode, true);
1689 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1691 struct inode *inode = file_inode(filp);
1695 if (!inode_owner_or_capable(inode))
1698 if (get_user(flags, (int __user *)arg))
1701 ret = mnt_want_write_file(filp);
1707 ret = __f2fs_ioc_setflags(inode, flags);
1709 inode_unlock(inode);
1710 mnt_drop_write_file(filp);
1714 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1716 struct inode *inode = file_inode(filp);
1718 return put_user(inode->i_generation, (int __user *)arg);
1721 static int f2fs_ioc_start_atomic_write(struct file *filp)
1723 struct inode *inode = file_inode(filp);
1726 if (!inode_owner_or_capable(inode))
1729 if (!S_ISREG(inode->i_mode))
1732 ret = mnt_want_write_file(filp);
1738 if (f2fs_is_atomic_file(inode)) {
1739 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1744 ret = f2fs_convert_inline_inode(inode);
1748 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1751 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1752 * f2fs_is_atomic_file.
1754 if (get_dirty_pages(inode))
1755 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1756 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1757 inode->i_ino, get_dirty_pages(inode));
1758 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1760 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1764 set_inode_flag(inode, FI_ATOMIC_FILE);
1765 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1766 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1768 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1769 F2FS_I(inode)->inmem_task = current;
1770 stat_inc_atomic_write(inode);
1771 stat_update_max_atomic_write(inode);
1773 inode_unlock(inode);
1774 mnt_drop_write_file(filp);
1778 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1780 struct inode *inode = file_inode(filp);
1783 if (!inode_owner_or_capable(inode))
1786 ret = mnt_want_write_file(filp);
1790 f2fs_balance_fs(F2FS_I_SB(inode), true);
1794 if (f2fs_is_volatile_file(inode)) {
1799 if (f2fs_is_atomic_file(inode)) {
1800 ret = f2fs_commit_inmem_pages(inode);
1804 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1806 clear_inode_flag(inode, FI_ATOMIC_FILE);
1807 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1808 stat_dec_atomic_write(inode);
1811 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1814 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1815 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1818 inode_unlock(inode);
1819 mnt_drop_write_file(filp);
1823 static int f2fs_ioc_start_volatile_write(struct file *filp)
1825 struct inode *inode = file_inode(filp);
1828 if (!inode_owner_or_capable(inode))
1831 if (!S_ISREG(inode->i_mode))
1834 ret = mnt_want_write_file(filp);
1840 if (f2fs_is_volatile_file(inode))
1843 ret = f2fs_convert_inline_inode(inode);
1847 stat_inc_volatile_write(inode);
1848 stat_update_max_volatile_write(inode);
1850 set_inode_flag(inode, FI_VOLATILE_FILE);
1851 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1853 inode_unlock(inode);
1854 mnt_drop_write_file(filp);
1858 static int f2fs_ioc_release_volatile_write(struct file *filp)
1860 struct inode *inode = file_inode(filp);
1863 if (!inode_owner_or_capable(inode))
1866 ret = mnt_want_write_file(filp);
1872 if (!f2fs_is_volatile_file(inode))
1875 if (!f2fs_is_first_block_written(inode)) {
1876 ret = truncate_partial_data_page(inode, 0, true);
1880 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1882 inode_unlock(inode);
1883 mnt_drop_write_file(filp);
1887 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1889 struct inode *inode = file_inode(filp);
1892 if (!inode_owner_or_capable(inode))
1895 ret = mnt_want_write_file(filp);
1901 if (f2fs_is_atomic_file(inode))
1902 f2fs_drop_inmem_pages(inode);
1903 if (f2fs_is_volatile_file(inode)) {
1904 clear_inode_flag(inode, FI_VOLATILE_FILE);
1905 stat_dec_volatile_write(inode);
1906 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1909 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1911 inode_unlock(inode);
1913 mnt_drop_write_file(filp);
1914 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1918 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1920 struct inode *inode = file_inode(filp);
1921 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1922 struct super_block *sb = sbi->sb;
1926 if (!capable(CAP_SYS_ADMIN))
1929 if (get_user(in, (__u32 __user *)arg))
1932 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1933 ret = mnt_want_write_file(filp);
1939 case F2FS_GOING_DOWN_FULLSYNC:
1940 sb = freeze_bdev(sb->s_bdev);
1946 f2fs_stop_checkpoint(sbi, false);
1947 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1948 thaw_bdev(sb->s_bdev, sb);
1951 case F2FS_GOING_DOWN_METASYNC:
1952 /* do checkpoint only */
1953 ret = f2fs_sync_fs(sb, 1);
1956 f2fs_stop_checkpoint(sbi, false);
1957 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1959 case F2FS_GOING_DOWN_NOSYNC:
1960 f2fs_stop_checkpoint(sbi, false);
1961 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1963 case F2FS_GOING_DOWN_METAFLUSH:
1964 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1965 f2fs_stop_checkpoint(sbi, false);
1966 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1968 case F2FS_GOING_DOWN_NEED_FSCK:
1969 set_sbi_flag(sbi, SBI_NEED_FSCK);
1970 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
1971 set_sbi_flag(sbi, SBI_IS_DIRTY);
1972 /* do checkpoint only */
1973 ret = f2fs_sync_fs(sb, 1);
1980 f2fs_stop_gc_thread(sbi);
1981 f2fs_stop_discard_thread(sbi);
1983 f2fs_drop_discard_cmd(sbi);
1984 clear_opt(sbi, DISCARD);
1986 f2fs_update_time(sbi, REQ_TIME);
1988 if (in != F2FS_GOING_DOWN_FULLSYNC)
1989 mnt_drop_write_file(filp);
1991 trace_f2fs_shutdown(sbi, in, ret);
1996 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1998 struct inode *inode = file_inode(filp);
1999 struct super_block *sb = inode->i_sb;
2000 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2001 struct fstrim_range range;
2004 if (!capable(CAP_SYS_ADMIN))
2007 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2010 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2014 ret = mnt_want_write_file(filp);
2018 range.minlen = max((unsigned int)range.minlen,
2019 q->limits.discard_granularity);
2020 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2021 mnt_drop_write_file(filp);
2025 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2028 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2032 static bool uuid_is_nonzero(__u8 u[16])
2036 for (i = 0; i < 16; i++)
2042 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2044 struct inode *inode = file_inode(filp);
2046 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2049 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2051 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2054 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2056 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2058 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2061 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2063 struct inode *inode = file_inode(filp);
2064 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2067 if (!f2fs_sb_has_encrypt(sbi))
2070 err = mnt_want_write_file(filp);
2074 down_write(&sbi->sb_lock);
2076 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2079 /* update superblock with uuid */
2080 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2082 err = f2fs_commit_super(sbi, false);
2085 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2089 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2093 up_write(&sbi->sb_lock);
2094 mnt_drop_write_file(filp);
2098 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2100 struct inode *inode = file_inode(filp);
2101 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2105 if (!capable(CAP_SYS_ADMIN))
2108 if (get_user(sync, (__u32 __user *)arg))
2111 if (f2fs_readonly(sbi->sb))
2114 ret = mnt_want_write_file(filp);
2119 if (!mutex_trylock(&sbi->gc_mutex)) {
2124 mutex_lock(&sbi->gc_mutex);
2127 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2129 mnt_drop_write_file(filp);
2133 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2135 struct inode *inode = file_inode(filp);
2136 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2137 struct f2fs_gc_range range;
2141 if (!capable(CAP_SYS_ADMIN))
2144 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2148 if (f2fs_readonly(sbi->sb))
2151 end = range.start + range.len;
2152 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2156 ret = mnt_want_write_file(filp);
2162 if (!mutex_trylock(&sbi->gc_mutex)) {
2167 mutex_lock(&sbi->gc_mutex);
2170 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2171 range.start += BLKS_PER_SEC(sbi);
2172 if (range.start <= end)
2175 mnt_drop_write_file(filp);
2179 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2181 struct inode *inode = file_inode(filp);
2182 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2185 if (!capable(CAP_SYS_ADMIN))
2188 if (f2fs_readonly(sbi->sb))
2191 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2192 f2fs_msg(sbi->sb, KERN_INFO,
2193 "Skipping Checkpoint. Checkpoints currently disabled.");
2197 ret = mnt_want_write_file(filp);
2201 ret = f2fs_sync_fs(sbi->sb, 1);
2203 mnt_drop_write_file(filp);
2207 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2209 struct f2fs_defragment *range)
2211 struct inode *inode = file_inode(filp);
2212 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2213 .m_seg_type = NO_CHECK_TYPE ,
2214 .m_may_create = false };
2215 struct extent_info ei = {0, 0, 0};
2216 pgoff_t pg_start, pg_end, next_pgofs;
2217 unsigned int blk_per_seg = sbi->blocks_per_seg;
2218 unsigned int total = 0, sec_num;
2219 block_t blk_end = 0;
2220 bool fragmented = false;
2223 /* if in-place-update policy is enabled, don't waste time here */
2224 if (f2fs_should_update_inplace(inode, NULL))
2227 pg_start = range->start >> PAGE_SHIFT;
2228 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2230 f2fs_balance_fs(sbi, true);
2234 /* writeback all dirty pages in the range */
2235 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2236 range->start + range->len - 1);
2241 * lookup mapping info in extent cache, skip defragmenting if physical
2242 * block addresses are continuous.
2244 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2245 if (ei.fofs + ei.len >= pg_end)
2249 map.m_lblk = pg_start;
2250 map.m_next_pgofs = &next_pgofs;
2253 * lookup mapping info in dnode page cache, skip defragmenting if all
2254 * physical block addresses are continuous even if there are hole(s)
2255 * in logical blocks.
2257 while (map.m_lblk < pg_end) {
2258 map.m_len = pg_end - map.m_lblk;
2259 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2263 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2264 map.m_lblk = next_pgofs;
2268 if (blk_end && blk_end != map.m_pblk)
2271 /* record total count of block that we're going to move */
2274 blk_end = map.m_pblk + map.m_len;
2276 map.m_lblk += map.m_len;
2282 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2285 * make sure there are enough free section for LFS allocation, this can
2286 * avoid defragment running in SSR mode when free section are allocated
2289 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2294 map.m_lblk = pg_start;
2295 map.m_len = pg_end - pg_start;
2298 while (map.m_lblk < pg_end) {
2303 map.m_len = pg_end - map.m_lblk;
2304 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2308 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2309 map.m_lblk = next_pgofs;
2313 set_inode_flag(inode, FI_DO_DEFRAG);
2316 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2319 page = f2fs_get_lock_data_page(inode, idx, true);
2321 err = PTR_ERR(page);
2325 set_page_dirty(page);
2326 f2fs_put_page(page, 1);
2335 if (idx < pg_end && cnt < blk_per_seg)
2338 clear_inode_flag(inode, FI_DO_DEFRAG);
2340 err = filemap_fdatawrite(inode->i_mapping);
2345 clear_inode_flag(inode, FI_DO_DEFRAG);
2347 inode_unlock(inode);
2349 range->len = (u64)total << PAGE_SHIFT;
2353 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2355 struct inode *inode = file_inode(filp);
2356 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2357 struct f2fs_defragment range;
2360 if (!capable(CAP_SYS_ADMIN))
2363 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2366 if (f2fs_readonly(sbi->sb))
2369 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2373 /* verify alignment of offset & size */
2374 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2377 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2378 sbi->max_file_blocks))
2381 err = mnt_want_write_file(filp);
2385 err = f2fs_defragment_range(sbi, filp, &range);
2386 mnt_drop_write_file(filp);
2388 f2fs_update_time(sbi, REQ_TIME);
2392 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2399 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2400 struct file *file_out, loff_t pos_out, size_t len)
2402 struct inode *src = file_inode(file_in);
2403 struct inode *dst = file_inode(file_out);
2404 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2405 size_t olen = len, dst_max_i_size = 0;
2409 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2410 src->i_sb != dst->i_sb)
2413 if (unlikely(f2fs_readonly(src->i_sb)))
2416 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2419 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2423 if (pos_in == pos_out)
2425 if (pos_out > pos_in && pos_out < pos_in + len)
2432 if (!inode_trylock(dst))
2437 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2440 olen = len = src->i_size - pos_in;
2441 if (pos_in + len == src->i_size)
2442 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2448 dst_osize = dst->i_size;
2449 if (pos_out + olen > dst->i_size)
2450 dst_max_i_size = pos_out + olen;
2452 /* verify the end result is block aligned */
2453 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2454 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2455 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2458 ret = f2fs_convert_inline_inode(src);
2462 ret = f2fs_convert_inline_inode(dst);
2466 /* write out all dirty pages from offset */
2467 ret = filemap_write_and_wait_range(src->i_mapping,
2468 pos_in, pos_in + len);
2472 ret = filemap_write_and_wait_range(dst->i_mapping,
2473 pos_out, pos_out + len);
2477 f2fs_balance_fs(sbi, true);
2479 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2482 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2487 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2488 pos_out >> F2FS_BLKSIZE_BITS,
2489 len >> F2FS_BLKSIZE_BITS, false);
2493 f2fs_i_size_write(dst, dst_max_i_size);
2494 else if (dst_osize != dst->i_size)
2495 f2fs_i_size_write(dst, dst_osize);
2497 f2fs_unlock_op(sbi);
2500 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2502 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2511 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2513 struct f2fs_move_range range;
2517 if (!(filp->f_mode & FMODE_READ) ||
2518 !(filp->f_mode & FMODE_WRITE))
2521 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2525 dst = fdget(range.dst_fd);
2529 if (!(dst.file->f_mode & FMODE_WRITE)) {
2534 err = mnt_want_write_file(filp);
2538 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2539 range.pos_out, range.len);
2541 mnt_drop_write_file(filp);
2545 if (copy_to_user((struct f2fs_move_range __user *)arg,
2546 &range, sizeof(range)))
2553 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2555 struct inode *inode = file_inode(filp);
2556 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2557 struct sit_info *sm = SIT_I(sbi);
2558 unsigned int start_segno = 0, end_segno = 0;
2559 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2560 struct f2fs_flush_device range;
2563 if (!capable(CAP_SYS_ADMIN))
2566 if (f2fs_readonly(sbi->sb))
2569 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2572 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2576 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2577 __is_large_section(sbi)) {
2578 f2fs_msg(sbi->sb, KERN_WARNING,
2579 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2580 range.dev_num, sbi->s_ndevs,
2585 ret = mnt_want_write_file(filp);
2589 if (range.dev_num != 0)
2590 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2591 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2593 start_segno = sm->last_victim[FLUSH_DEVICE];
2594 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2595 start_segno = dev_start_segno;
2596 end_segno = min(start_segno + range.segments, dev_end_segno);
2598 while (start_segno < end_segno) {
2599 if (!mutex_trylock(&sbi->gc_mutex)) {
2603 sm->last_victim[GC_CB] = end_segno + 1;
2604 sm->last_victim[GC_GREEDY] = end_segno + 1;
2605 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2606 ret = f2fs_gc(sbi, true, true, start_segno);
2614 mnt_drop_write_file(filp);
2618 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2620 struct inode *inode = file_inode(filp);
2621 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2623 /* Must validate to set it with SQLite behavior in Android. */
2624 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2626 return put_user(sb_feature, (u32 __user *)arg);
2630 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2632 struct dquot *transfer_to[MAXQUOTAS] = {};
2633 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2634 struct super_block *sb = sbi->sb;
2637 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2638 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2639 err = __dquot_transfer(inode, transfer_to);
2641 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2642 dqput(transfer_to[PRJQUOTA]);
2647 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2649 struct inode *inode = file_inode(filp);
2650 struct f2fs_inode_info *fi = F2FS_I(inode);
2651 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2656 if (!f2fs_sb_has_project_quota(sbi)) {
2657 if (projid != F2FS_DEF_PROJID)
2663 if (!f2fs_has_extra_attr(inode))
2666 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2668 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2672 /* Is it quota file? Do not allow user to mess with it */
2673 if (IS_NOQUOTA(inode))
2676 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2678 return PTR_ERR(ipage);
2680 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2683 f2fs_put_page(ipage, 1);
2686 f2fs_put_page(ipage, 1);
2688 err = dquot_initialize(inode);
2693 err = f2fs_transfer_project_quota(inode, kprojid);
2697 F2FS_I(inode)->i_projid = kprojid;
2698 inode->i_ctime = current_time(inode);
2699 f2fs_mark_inode_dirty_sync(inode, true);
2701 f2fs_unlock_op(sbi);
2705 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2710 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2712 if (projid != F2FS_DEF_PROJID)
2718 /* Transfer internal flags to xflags */
2719 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2723 if (iflags & F2FS_SYNC_FL)
2724 xflags |= FS_XFLAG_SYNC;
2725 if (iflags & F2FS_IMMUTABLE_FL)
2726 xflags |= FS_XFLAG_IMMUTABLE;
2727 if (iflags & F2FS_APPEND_FL)
2728 xflags |= FS_XFLAG_APPEND;
2729 if (iflags & F2FS_NODUMP_FL)
2730 xflags |= FS_XFLAG_NODUMP;
2731 if (iflags & F2FS_NOATIME_FL)
2732 xflags |= FS_XFLAG_NOATIME;
2733 if (iflags & F2FS_PROJINHERIT_FL)
2734 xflags |= FS_XFLAG_PROJINHERIT;
2738 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2739 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2740 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2742 /* Transfer xflags flags to internal */
2743 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2745 unsigned long iflags = 0;
2747 if (xflags & FS_XFLAG_SYNC)
2748 iflags |= F2FS_SYNC_FL;
2749 if (xflags & FS_XFLAG_IMMUTABLE)
2750 iflags |= F2FS_IMMUTABLE_FL;
2751 if (xflags & FS_XFLAG_APPEND)
2752 iflags |= F2FS_APPEND_FL;
2753 if (xflags & FS_XFLAG_NODUMP)
2754 iflags |= F2FS_NODUMP_FL;
2755 if (xflags & FS_XFLAG_NOATIME)
2756 iflags |= F2FS_NOATIME_FL;
2757 if (xflags & FS_XFLAG_PROJINHERIT)
2758 iflags |= F2FS_PROJINHERIT_FL;
2763 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2765 struct inode *inode = file_inode(filp);
2766 struct f2fs_inode_info *fi = F2FS_I(inode);
2769 memset(&fa, 0, sizeof(struct fsxattr));
2770 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2771 F2FS_FL_USER_VISIBLE);
2773 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2774 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2777 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2782 static int f2fs_ioctl_check_project(struct inode *inode, struct fsxattr *fa)
2785 * Project Quota ID state is only allowed to change from within the init
2786 * namespace. Enforce that restriction only if we are trying to change
2787 * the quota ID state. Everything else is allowed in user namespaces.
2789 if (current_user_ns() == &init_user_ns)
2792 if (__kprojid_val(F2FS_I(inode)->i_projid) != fa->fsx_projid)
2795 if (F2FS_I(inode)->i_flags & F2FS_PROJINHERIT_FL) {
2796 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT))
2799 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
2806 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2808 struct inode *inode = file_inode(filp);
2809 struct f2fs_inode_info *fi = F2FS_I(inode);
2814 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2817 /* Make sure caller has proper permission */
2818 if (!inode_owner_or_capable(inode))
2821 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2824 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2825 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2828 err = mnt_want_write_file(filp);
2833 err = f2fs_ioctl_check_project(inode, &fa);
2836 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2837 (flags & F2FS_FL_XFLAG_VISIBLE);
2838 err = __f2fs_ioc_setflags(inode, flags);
2842 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2844 inode_unlock(inode);
2845 mnt_drop_write_file(filp);
2849 int f2fs_pin_file_control(struct inode *inode, bool inc)
2851 struct f2fs_inode_info *fi = F2FS_I(inode);
2852 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2854 /* Use i_gc_failures for normal file as a risk signal. */
2856 f2fs_i_gc_failures_write(inode,
2857 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2859 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2860 f2fs_msg(sbi->sb, KERN_WARNING,
2861 "%s: Enable GC = ino %lx after %x GC trials\n",
2862 __func__, inode->i_ino,
2863 fi->i_gc_failures[GC_FAILURE_PIN]);
2864 clear_inode_flag(inode, FI_PIN_FILE);
2870 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2872 struct inode *inode = file_inode(filp);
2876 if (!capable(CAP_SYS_ADMIN))
2879 if (get_user(pin, (__u32 __user *)arg))
2882 if (!S_ISREG(inode->i_mode))
2885 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2888 ret = mnt_want_write_file(filp);
2894 if (f2fs_should_update_outplace(inode, NULL)) {
2900 clear_inode_flag(inode, FI_PIN_FILE);
2901 f2fs_i_gc_failures_write(inode, 0);
2905 if (f2fs_pin_file_control(inode, false)) {
2909 ret = f2fs_convert_inline_inode(inode);
2913 set_inode_flag(inode, FI_PIN_FILE);
2914 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2916 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2918 inode_unlock(inode);
2919 mnt_drop_write_file(filp);
2923 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2925 struct inode *inode = file_inode(filp);
2928 if (is_inode_flag_set(inode, FI_PIN_FILE))
2929 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2930 return put_user(pin, (u32 __user *)arg);
2933 int f2fs_precache_extents(struct inode *inode)
2935 struct f2fs_inode_info *fi = F2FS_I(inode);
2936 struct f2fs_map_blocks map;
2937 pgoff_t m_next_extent;
2941 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2945 map.m_next_pgofs = NULL;
2946 map.m_next_extent = &m_next_extent;
2947 map.m_seg_type = NO_CHECK_TYPE;
2948 map.m_may_create = false;
2949 end = F2FS_I_SB(inode)->max_file_blocks;
2951 while (map.m_lblk < end) {
2952 map.m_len = end - map.m_lblk;
2954 down_write(&fi->i_gc_rwsem[WRITE]);
2955 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2956 up_write(&fi->i_gc_rwsem[WRITE]);
2960 map.m_lblk = m_next_extent;
2966 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2968 return f2fs_precache_extents(file_inode(filp));
2971 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2973 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2977 case F2FS_IOC_GETFLAGS:
2978 return f2fs_ioc_getflags(filp, arg);
2979 case F2FS_IOC_SETFLAGS:
2980 return f2fs_ioc_setflags(filp, arg);
2981 case F2FS_IOC_GETVERSION:
2982 return f2fs_ioc_getversion(filp, arg);
2983 case F2FS_IOC_START_ATOMIC_WRITE:
2984 return f2fs_ioc_start_atomic_write(filp);
2985 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2986 return f2fs_ioc_commit_atomic_write(filp);
2987 case F2FS_IOC_START_VOLATILE_WRITE:
2988 return f2fs_ioc_start_volatile_write(filp);
2989 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2990 return f2fs_ioc_release_volatile_write(filp);
2991 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2992 return f2fs_ioc_abort_volatile_write(filp);
2993 case F2FS_IOC_SHUTDOWN:
2994 return f2fs_ioc_shutdown(filp, arg);
2996 return f2fs_ioc_fitrim(filp, arg);
2997 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2998 return f2fs_ioc_set_encryption_policy(filp, arg);
2999 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3000 return f2fs_ioc_get_encryption_policy(filp, arg);
3001 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3002 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3003 case F2FS_IOC_GARBAGE_COLLECT:
3004 return f2fs_ioc_gc(filp, arg);
3005 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3006 return f2fs_ioc_gc_range(filp, arg);
3007 case F2FS_IOC_WRITE_CHECKPOINT:
3008 return f2fs_ioc_write_checkpoint(filp, arg);
3009 case F2FS_IOC_DEFRAGMENT:
3010 return f2fs_ioc_defragment(filp, arg);
3011 case F2FS_IOC_MOVE_RANGE:
3012 return f2fs_ioc_move_range(filp, arg);
3013 case F2FS_IOC_FLUSH_DEVICE:
3014 return f2fs_ioc_flush_device(filp, arg);
3015 case F2FS_IOC_GET_FEATURES:
3016 return f2fs_ioc_get_features(filp, arg);
3017 case F2FS_IOC_FSGETXATTR:
3018 return f2fs_ioc_fsgetxattr(filp, arg);
3019 case F2FS_IOC_FSSETXATTR:
3020 return f2fs_ioc_fssetxattr(filp, arg);
3021 case F2FS_IOC_GET_PIN_FILE:
3022 return f2fs_ioc_get_pin_file(filp, arg);
3023 case F2FS_IOC_SET_PIN_FILE:
3024 return f2fs_ioc_set_pin_file(filp, arg);
3025 case F2FS_IOC_PRECACHE_EXTENTS:
3026 return f2fs_ioc_precache_extents(filp, arg);
3032 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3034 struct file *file = iocb->ki_filp;
3035 struct inode *inode = file_inode(file);
3038 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3041 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
3044 if (!inode_trylock(inode)) {
3045 if (iocb->ki_flags & IOCB_NOWAIT)
3050 ret = generic_write_checks(iocb, from);
3052 bool preallocated = false;
3053 size_t target_size = 0;
3056 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3057 set_inode_flag(inode, FI_NO_PREALLOC);
3059 if ((iocb->ki_flags & IOCB_NOWAIT) &&
3060 (iocb->ki_flags & IOCB_DIRECT)) {
3061 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3062 iov_iter_count(from)) ||
3063 f2fs_has_inline_data(inode) ||
3064 f2fs_force_buffered_io(inode,
3066 clear_inode_flag(inode,
3068 inode_unlock(inode);
3073 preallocated = true;
3074 target_size = iocb->ki_pos + iov_iter_count(from);
3076 err = f2fs_preallocate_blocks(iocb, from);
3078 clear_inode_flag(inode, FI_NO_PREALLOC);
3079 inode_unlock(inode);
3083 ret = __generic_file_write_iter(iocb, from);
3084 clear_inode_flag(inode, FI_NO_PREALLOC);
3086 /* if we couldn't write data, we should deallocate blocks. */
3087 if (preallocated && i_size_read(inode) < target_size)
3088 f2fs_truncate(inode);
3091 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3093 inode_unlock(inode);
3096 ret = generic_write_sync(iocb, ret);
3100 #ifdef CONFIG_COMPAT
3101 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3104 case F2FS_IOC32_GETFLAGS:
3105 cmd = F2FS_IOC_GETFLAGS;
3107 case F2FS_IOC32_SETFLAGS:
3108 cmd = F2FS_IOC_SETFLAGS;
3110 case F2FS_IOC32_GETVERSION:
3111 cmd = F2FS_IOC_GETVERSION;
3113 case F2FS_IOC_START_ATOMIC_WRITE:
3114 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3115 case F2FS_IOC_START_VOLATILE_WRITE:
3116 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3117 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3118 case F2FS_IOC_SHUTDOWN:
3119 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3120 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3121 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3122 case F2FS_IOC_GARBAGE_COLLECT:
3123 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3124 case F2FS_IOC_WRITE_CHECKPOINT:
3125 case F2FS_IOC_DEFRAGMENT:
3126 case F2FS_IOC_MOVE_RANGE:
3127 case F2FS_IOC_FLUSH_DEVICE:
3128 case F2FS_IOC_GET_FEATURES:
3129 case F2FS_IOC_FSGETXATTR:
3130 case F2FS_IOC_FSSETXATTR:
3131 case F2FS_IOC_GET_PIN_FILE:
3132 case F2FS_IOC_SET_PIN_FILE:
3133 case F2FS_IOC_PRECACHE_EXTENTS:
3136 return -ENOIOCTLCMD;
3138 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3142 const struct file_operations f2fs_file_operations = {
3143 .llseek = f2fs_llseek,
3144 .read_iter = generic_file_read_iter,
3145 .write_iter = f2fs_file_write_iter,
3146 .open = f2fs_file_open,
3147 .release = f2fs_release_file,
3148 .mmap = f2fs_file_mmap,
3149 .flush = f2fs_file_flush,
3150 .fsync = f2fs_sync_file,
3151 .fallocate = f2fs_fallocate,
3152 .unlocked_ioctl = f2fs_ioctl,
3153 #ifdef CONFIG_COMPAT
3154 .compat_ioctl = f2fs_compat_ioctl,
3156 .splice_read = generic_file_splice_read,
3157 .splice_write = iter_file_splice_write,