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);
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 fdatasync is triggered, let's do in-place-update */
220 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
221 set_inode_flag(inode, FI_NEED_IPU);
222 ret = file_write_and_wait_range(file, start, end);
223 clear_inode_flag(inode, FI_NEED_IPU);
226 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
230 /* if the inode is dirty, let's recover all the time */
231 if (!f2fs_skip_inode_update(inode, datasync)) {
232 f2fs_write_inode(inode, NULL);
237 * if there is no written data, don't waste time to write recovery info.
239 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
240 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
242 /* it may call write_inode just prior to fsync */
243 if (need_inode_page_update(sbi, ino))
246 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
247 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
253 * Both of fdatasync() and fsync() are able to be recovered from
256 down_read(&F2FS_I(inode)->i_sem);
257 cp_reason = need_do_checkpoint(inode);
258 up_read(&F2FS_I(inode)->i_sem);
261 /* all the dirty node pages should be flushed for POR */
262 ret = f2fs_sync_fs(inode->i_sb, 1);
265 * We've secured consistency through sync_fs. Following pino
266 * will be used only for fsynced inodes after checkpoint.
268 try_to_fix_pino(inode);
269 clear_inode_flag(inode, FI_APPEND_WRITE);
270 clear_inode_flag(inode, FI_UPDATE_WRITE);
274 atomic_inc(&sbi->wb_sync_req[NODE]);
275 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
276 atomic_dec(&sbi->wb_sync_req[NODE]);
280 /* if cp_error was enabled, we should avoid infinite loop */
281 if (unlikely(f2fs_cp_error(sbi))) {
286 if (f2fs_need_inode_block_update(sbi, ino)) {
287 f2fs_mark_inode_dirty_sync(inode, true);
288 f2fs_write_inode(inode, NULL);
293 * If it's atomic_write, it's just fine to keep write ordering. So
294 * here we don't need to wait for node write completion, since we use
295 * node chain which serializes node blocks. If one of node writes are
296 * reordered, we can see simply broken chain, resulting in stopping
297 * roll-forward recovery. It means we'll recover all or none node blocks
301 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
306 /* once recovery info is written, don't need to tack this */
307 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
308 clear_inode_flag(inode, FI_APPEND_WRITE);
310 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
311 ret = f2fs_issue_flush(sbi, inode->i_ino);
313 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
314 clear_inode_flag(inode, FI_UPDATE_WRITE);
315 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
317 f2fs_update_time(sbi, REQ_TIME);
319 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
320 f2fs_trace_ios(NULL, 1);
324 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
326 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
328 return f2fs_do_sync_file(file, start, end, datasync, false);
331 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
332 pgoff_t pgofs, int whence)
337 if (whence != SEEK_DATA)
340 /* find first dirty page index */
341 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
350 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
351 pgoff_t dirty, pgoff_t pgofs, int whence)
355 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
356 is_valid_data_blkaddr(sbi, blkaddr))
360 if (blkaddr == NULL_ADDR)
367 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
369 struct inode *inode = file->f_mapping->host;
370 loff_t maxbytes = inode->i_sb->s_maxbytes;
371 struct dnode_of_data dn;
372 pgoff_t pgofs, end_offset, dirty;
373 loff_t data_ofs = offset;
379 isize = i_size_read(inode);
383 /* handle inline data case */
384 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
385 if (whence == SEEK_HOLE)
390 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
392 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
394 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
395 set_new_dnode(&dn, inode, NULL, NULL, 0);
396 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
397 if (err && err != -ENOENT) {
399 } else if (err == -ENOENT) {
400 /* direct node does not exists */
401 if (whence == SEEK_DATA) {
402 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
409 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
411 /* find data/hole in dnode block */
412 for (; dn.ofs_in_node < end_offset;
413 dn.ofs_in_node++, pgofs++,
414 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
417 blkaddr = datablock_addr(dn.inode,
418 dn.node_page, dn.ofs_in_node);
420 if (__is_valid_data_blkaddr(blkaddr) &&
421 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
422 blkaddr, DATA_GENERIC)) {
427 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
436 if (whence == SEEK_DATA)
439 if (whence == SEEK_HOLE && data_ofs > isize)
442 return vfs_setpos(file, data_ofs, maxbytes);
448 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
450 struct inode *inode = file->f_mapping->host;
451 loff_t maxbytes = inode->i_sb->s_maxbytes;
457 return generic_file_llseek_size(file, offset, whence,
458 maxbytes, i_size_read(inode));
463 return f2fs_seek_block(file, offset, whence);
469 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
471 struct inode *inode = file_inode(file);
474 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
477 /* we don't need to use inline_data strictly */
478 err = f2fs_convert_inline_inode(inode);
483 vma->vm_ops = &f2fs_file_vm_ops;
487 static int f2fs_file_open(struct inode *inode, struct file *filp)
489 int err = fscrypt_file_open(inode, filp);
494 filp->f_mode |= FMODE_NOWAIT;
496 return dquot_file_open(inode, filp);
499 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
501 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
502 struct f2fs_node *raw_node;
503 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
507 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
508 base = get_extra_isize(dn->inode);
510 raw_node = F2FS_NODE(dn->node_page);
511 addr = blkaddr_in_node(raw_node) + base + ofs;
513 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
514 block_t blkaddr = le32_to_cpu(*addr);
516 if (blkaddr == NULL_ADDR)
519 dn->data_blkaddr = NULL_ADDR;
520 f2fs_set_data_blkaddr(dn);
522 if (__is_valid_data_blkaddr(blkaddr) &&
523 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
526 f2fs_invalidate_blocks(sbi, blkaddr);
527 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
528 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
535 * once we invalidate valid blkaddr in range [ofs, ofs + count],
536 * we will invalidate all blkaddr in the whole range.
538 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
540 f2fs_update_extent_cache_range(dn, fofs, 0, len);
541 dec_valid_block_count(sbi, dn->inode, nr_free);
543 dn->ofs_in_node = ofs;
545 f2fs_update_time(sbi, REQ_TIME);
546 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
547 dn->ofs_in_node, nr_free);
550 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
552 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
555 static int truncate_partial_data_page(struct inode *inode, u64 from,
558 loff_t offset = from & (PAGE_SIZE - 1);
559 pgoff_t index = from >> PAGE_SHIFT;
560 struct address_space *mapping = inode->i_mapping;
563 if (!offset && !cache_only)
567 page = find_lock_page(mapping, index);
568 if (page && PageUptodate(page))
570 f2fs_put_page(page, 1);
574 page = f2fs_get_lock_data_page(inode, index, true);
576 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
578 f2fs_wait_on_page_writeback(page, DATA, true);
579 zero_user(page, offset, PAGE_SIZE - offset);
581 /* An encrypted inode should have a key and truncate the last page. */
582 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
584 set_page_dirty(page);
585 f2fs_put_page(page, 1);
589 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock,
592 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
593 struct dnode_of_data dn;
595 int count = 0, err = 0;
597 bool truncate_page = false;
598 int flag = buf_write ? F2FS_GET_BLOCK_PRE_AIO : F2FS_GET_BLOCK_PRE_DIO;
600 trace_f2fs_truncate_blocks_enter(inode, from);
602 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
604 if (free_from >= sbi->max_file_blocks)
608 __do_map_lock(sbi, flag, true);
610 ipage = f2fs_get_node_page(sbi, inode->i_ino);
612 err = PTR_ERR(ipage);
616 if (f2fs_has_inline_data(inode)) {
617 f2fs_truncate_inline_inode(inode, ipage, from);
618 f2fs_put_page(ipage, 1);
619 truncate_page = true;
623 set_new_dnode(&dn, inode, ipage, NULL, 0);
624 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
631 count = ADDRS_PER_PAGE(dn.node_page, inode);
633 count -= dn.ofs_in_node;
634 f2fs_bug_on(sbi, count < 0);
636 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
637 f2fs_truncate_data_blocks_range(&dn, count);
643 err = f2fs_truncate_inode_blocks(inode, free_from);
646 __do_map_lock(sbi, flag, false);
648 /* lastly zero out the first data page */
650 err = truncate_partial_data_page(inode, from, truncate_page);
652 trace_f2fs_truncate_blocks_exit(inode, err);
656 int f2fs_truncate(struct inode *inode)
660 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
663 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
664 S_ISLNK(inode->i_mode)))
667 trace_f2fs_truncate(inode);
669 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
670 f2fs_show_injection_info(FAULT_TRUNCATE);
674 /* we should check inline_data size */
675 if (!f2fs_may_inline_data(inode)) {
676 err = f2fs_convert_inline_inode(inode);
681 err = f2fs_truncate_blocks(inode, i_size_read(inode), true, false);
685 inode->i_mtime = inode->i_ctime = current_time(inode);
686 f2fs_mark_inode_dirty_sync(inode, false);
690 int f2fs_getattr(const struct path *path, struct kstat *stat,
691 u32 request_mask, unsigned int query_flags)
693 struct inode *inode = d_inode(path->dentry);
694 struct f2fs_inode_info *fi = F2FS_I(inode);
695 struct f2fs_inode *ri;
698 if (f2fs_has_extra_attr(inode) &&
699 f2fs_sb_has_inode_crtime(inode->i_sb) &&
700 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
701 stat->result_mask |= STATX_BTIME;
702 stat->btime.tv_sec = fi->i_crtime.tv_sec;
703 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
706 flags = fi->i_flags & F2FS_FL_USER_VISIBLE;
707 if (flags & F2FS_APPEND_FL)
708 stat->attributes |= STATX_ATTR_APPEND;
709 if (flags & F2FS_COMPR_FL)
710 stat->attributes |= STATX_ATTR_COMPRESSED;
711 if (f2fs_encrypted_inode(inode))
712 stat->attributes |= STATX_ATTR_ENCRYPTED;
713 if (flags & F2FS_IMMUTABLE_FL)
714 stat->attributes |= STATX_ATTR_IMMUTABLE;
715 if (flags & F2FS_NODUMP_FL)
716 stat->attributes |= STATX_ATTR_NODUMP;
718 stat->attributes_mask |= (STATX_ATTR_APPEND |
719 STATX_ATTR_COMPRESSED |
720 STATX_ATTR_ENCRYPTED |
721 STATX_ATTR_IMMUTABLE |
724 generic_fillattr(inode, stat);
726 /* we need to show initial sectors used for inline_data/dentries */
727 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
728 f2fs_has_inline_dentry(inode))
729 stat->blocks += (stat->size + 511) >> 9;
734 #ifdef CONFIG_F2FS_FS_POSIX_ACL
735 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
737 unsigned int ia_valid = attr->ia_valid;
739 if (ia_valid & ATTR_UID)
740 inode->i_uid = attr->ia_uid;
741 if (ia_valid & ATTR_GID)
742 inode->i_gid = attr->ia_gid;
743 if (ia_valid & ATTR_ATIME)
744 inode->i_atime = timespec64_trunc(attr->ia_atime,
745 inode->i_sb->s_time_gran);
746 if (ia_valid & ATTR_MTIME)
747 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
748 inode->i_sb->s_time_gran);
749 if (ia_valid & ATTR_CTIME)
750 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
751 inode->i_sb->s_time_gran);
752 if (ia_valid & ATTR_MODE) {
753 umode_t mode = attr->ia_mode;
755 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
757 set_acl_inode(inode, mode);
761 #define __setattr_copy setattr_copy
764 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
766 struct inode *inode = d_inode(dentry);
768 bool size_changed = false;
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);
847 __setattr_copy(inode, attr);
849 if (attr->ia_valid & ATTR_MODE) {
850 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
851 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
852 inode->i_mode = F2FS_I(inode)->i_acl_mode;
853 clear_inode_flag(inode, FI_ACL_MODE);
857 /* file size may changed here */
858 f2fs_mark_inode_dirty_sync(inode, size_changed);
860 /* inode change will produce dirty node pages flushed by checkpoint */
861 f2fs_balance_fs(F2FS_I_SB(inode), true);
866 const struct inode_operations f2fs_file_inode_operations = {
867 .getattr = f2fs_getattr,
868 .setattr = f2fs_setattr,
869 .get_acl = f2fs_get_acl,
870 .set_acl = f2fs_set_acl,
871 #ifdef CONFIG_F2FS_FS_XATTR
872 .listxattr = f2fs_listxattr,
874 .fiemap = f2fs_fiemap,
877 static int fill_zero(struct inode *inode, pgoff_t index,
878 loff_t start, loff_t len)
880 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
886 f2fs_balance_fs(sbi, true);
889 page = f2fs_get_new_data_page(inode, NULL, index, false);
893 return PTR_ERR(page);
895 f2fs_wait_on_page_writeback(page, DATA, true);
896 zero_user(page, start, len);
897 set_page_dirty(page);
898 f2fs_put_page(page, 1);
902 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
906 while (pg_start < pg_end) {
907 struct dnode_of_data dn;
908 pgoff_t end_offset, count;
910 set_new_dnode(&dn, inode, NULL, NULL, 0);
911 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
913 if (err == -ENOENT) {
914 pg_start = f2fs_get_next_page_offset(&dn,
921 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
922 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
924 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
926 f2fs_truncate_data_blocks_range(&dn, count);
934 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
936 pgoff_t pg_start, pg_end;
937 loff_t off_start, off_end;
940 ret = f2fs_convert_inline_inode(inode);
944 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
945 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
947 off_start = offset & (PAGE_SIZE - 1);
948 off_end = (offset + len) & (PAGE_SIZE - 1);
950 if (pg_start == pg_end) {
951 ret = fill_zero(inode, pg_start, off_start,
952 off_end - off_start);
957 ret = fill_zero(inode, pg_start++, off_start,
958 PAGE_SIZE - off_start);
963 ret = fill_zero(inode, pg_end, 0, off_end);
968 if (pg_start < pg_end) {
969 struct address_space *mapping = inode->i_mapping;
970 loff_t blk_start, blk_end;
971 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
973 f2fs_balance_fs(sbi, true);
975 blk_start = (loff_t)pg_start << PAGE_SHIFT;
976 blk_end = (loff_t)pg_end << PAGE_SHIFT;
978 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
979 down_write(&F2FS_I(inode)->i_mmap_sem);
981 truncate_inode_pages_range(mapping, blk_start,
985 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
988 up_write(&F2FS_I(inode)->i_mmap_sem);
989 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
996 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
997 int *do_replace, pgoff_t off, pgoff_t len)
999 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1000 struct dnode_of_data dn;
1004 set_new_dnode(&dn, inode, NULL, NULL, 0);
1005 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1006 if (ret && ret != -ENOENT) {
1008 } else if (ret == -ENOENT) {
1009 if (dn.max_level == 0)
1011 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
1017 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1018 dn.ofs_in_node, len);
1019 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1020 *blkaddr = datablock_addr(dn.inode,
1021 dn.node_page, dn.ofs_in_node);
1022 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1024 if (test_opt(sbi, LFS)) {
1025 f2fs_put_dnode(&dn);
1029 /* do not invalidate this block address */
1030 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1034 f2fs_put_dnode(&dn);
1043 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1044 int *do_replace, pgoff_t off, int len)
1046 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1047 struct dnode_of_data dn;
1050 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1051 if (*do_replace == 0)
1054 set_new_dnode(&dn, inode, NULL, NULL, 0);
1055 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1057 dec_valid_block_count(sbi, inode, 1);
1058 f2fs_invalidate_blocks(sbi, *blkaddr);
1060 f2fs_update_data_blkaddr(&dn, *blkaddr);
1062 f2fs_put_dnode(&dn);
1067 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1068 block_t *blkaddr, int *do_replace,
1069 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1071 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1076 if (blkaddr[i] == NULL_ADDR && !full) {
1081 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1082 struct dnode_of_data dn;
1083 struct node_info ni;
1087 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1088 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1092 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1094 f2fs_put_dnode(&dn);
1098 ilen = min((pgoff_t)
1099 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1100 dn.ofs_in_node, len - i);
1102 dn.data_blkaddr = datablock_addr(dn.inode,
1103 dn.node_page, dn.ofs_in_node);
1104 f2fs_truncate_data_blocks_range(&dn, 1);
1106 if (do_replace[i]) {
1107 f2fs_i_blocks_write(src_inode,
1109 f2fs_i_blocks_write(dst_inode,
1111 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1112 blkaddr[i], ni.version, true, false);
1118 new_size = (dst + i) << PAGE_SHIFT;
1119 if (dst_inode->i_size < new_size)
1120 f2fs_i_size_write(dst_inode, new_size);
1121 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1123 f2fs_put_dnode(&dn);
1125 struct page *psrc, *pdst;
1127 psrc = f2fs_get_lock_data_page(src_inode,
1130 return PTR_ERR(psrc);
1131 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1134 f2fs_put_page(psrc, 1);
1135 return PTR_ERR(pdst);
1137 f2fs_copy_page(psrc, pdst);
1138 set_page_dirty(pdst);
1139 f2fs_put_page(pdst, 1);
1140 f2fs_put_page(psrc, 1);
1142 ret = f2fs_truncate_hole(src_inode,
1143 src + i, src + i + 1);
1152 static int __exchange_data_block(struct inode *src_inode,
1153 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1154 pgoff_t len, bool full)
1156 block_t *src_blkaddr;
1162 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1164 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1165 array_size(olen, sizeof(block_t)),
1170 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1171 array_size(olen, sizeof(int)),
1174 kvfree(src_blkaddr);
1178 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1179 do_replace, src, olen);
1183 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1184 do_replace, src, dst, olen, full);
1192 kvfree(src_blkaddr);
1198 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1199 kvfree(src_blkaddr);
1204 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1207 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1208 pgoff_t start = offset >> PAGE_SHIFT;
1209 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1212 f2fs_balance_fs(sbi, true);
1214 /* avoid gc operation during block exchange */
1215 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1216 down_write(&F2FS_I(inode)->i_mmap_sem);
1219 f2fs_drop_extent_tree(inode);
1220 truncate_pagecache(inode, offset);
1221 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1222 f2fs_unlock_op(sbi);
1224 up_write(&F2FS_I(inode)->i_mmap_sem);
1225 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1229 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1234 if (offset + len >= i_size_read(inode))
1237 /* collapse range should be aligned to block size of f2fs. */
1238 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1241 ret = f2fs_convert_inline_inode(inode);
1245 /* write out all dirty pages from offset */
1246 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1250 ret = f2fs_do_collapse(inode, offset, len);
1254 /* write out all moved pages, if possible */
1255 down_write(&F2FS_I(inode)->i_mmap_sem);
1256 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1257 truncate_pagecache(inode, offset);
1259 new_size = i_size_read(inode) - len;
1260 truncate_pagecache(inode, new_size);
1262 ret = f2fs_truncate_blocks(inode, new_size, true, false);
1263 up_write(&F2FS_I(inode)->i_mmap_sem);
1265 f2fs_i_size_write(inode, new_size);
1269 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1272 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1273 pgoff_t index = start;
1274 unsigned int ofs_in_node = dn->ofs_in_node;
1278 for (; index < end; index++, dn->ofs_in_node++) {
1279 if (datablock_addr(dn->inode, dn->node_page,
1280 dn->ofs_in_node) == NULL_ADDR)
1284 dn->ofs_in_node = ofs_in_node;
1285 ret = f2fs_reserve_new_blocks(dn, count);
1289 dn->ofs_in_node = ofs_in_node;
1290 for (index = start; index < end; index++, dn->ofs_in_node++) {
1291 dn->data_blkaddr = datablock_addr(dn->inode,
1292 dn->node_page, dn->ofs_in_node);
1294 * f2fs_reserve_new_blocks will not guarantee entire block
1297 if (dn->data_blkaddr == NULL_ADDR) {
1301 if (dn->data_blkaddr != NEW_ADDR) {
1302 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1303 dn->data_blkaddr = NEW_ADDR;
1304 f2fs_set_data_blkaddr(dn);
1308 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1313 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1316 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1317 struct address_space *mapping = inode->i_mapping;
1318 pgoff_t index, pg_start, pg_end;
1319 loff_t new_size = i_size_read(inode);
1320 loff_t off_start, off_end;
1323 ret = inode_newsize_ok(inode, (len + offset));
1327 ret = f2fs_convert_inline_inode(inode);
1331 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1335 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1336 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1338 off_start = offset & (PAGE_SIZE - 1);
1339 off_end = (offset + len) & (PAGE_SIZE - 1);
1341 if (pg_start == pg_end) {
1342 ret = fill_zero(inode, pg_start, off_start,
1343 off_end - off_start);
1347 new_size = max_t(loff_t, new_size, offset + len);
1350 ret = fill_zero(inode, pg_start++, off_start,
1351 PAGE_SIZE - off_start);
1355 new_size = max_t(loff_t, new_size,
1356 (loff_t)pg_start << PAGE_SHIFT);
1359 for (index = pg_start; index < pg_end;) {
1360 struct dnode_of_data dn;
1361 unsigned int end_offset;
1364 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1365 down_write(&F2FS_I(inode)->i_mmap_sem);
1367 truncate_pagecache_range(inode,
1368 (loff_t)index << PAGE_SHIFT,
1369 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1373 set_new_dnode(&dn, inode, NULL, NULL, 0);
1374 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1376 f2fs_unlock_op(sbi);
1377 up_write(&F2FS_I(inode)->i_mmap_sem);
1378 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1382 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1383 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1385 ret = f2fs_do_zero_range(&dn, index, end);
1386 f2fs_put_dnode(&dn);
1388 f2fs_unlock_op(sbi);
1389 up_write(&F2FS_I(inode)->i_mmap_sem);
1390 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1392 f2fs_balance_fs(sbi, dn.node_changed);
1398 new_size = max_t(loff_t, new_size,
1399 (loff_t)index << PAGE_SHIFT);
1403 ret = fill_zero(inode, pg_end, 0, off_end);
1407 new_size = max_t(loff_t, new_size, offset + len);
1412 if (new_size > i_size_read(inode)) {
1413 if (mode & FALLOC_FL_KEEP_SIZE)
1414 file_set_keep_isize(inode);
1416 f2fs_i_size_write(inode, new_size);
1421 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1423 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1424 pgoff_t nr, pg_start, pg_end, delta, idx;
1428 new_size = i_size_read(inode) + len;
1429 ret = inode_newsize_ok(inode, new_size);
1433 if (offset >= i_size_read(inode))
1436 /* insert range should be aligned to block size of f2fs. */
1437 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1440 ret = f2fs_convert_inline_inode(inode);
1444 f2fs_balance_fs(sbi, true);
1446 down_write(&F2FS_I(inode)->i_mmap_sem);
1447 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true, false);
1448 up_write(&F2FS_I(inode)->i_mmap_sem);
1452 /* write out all dirty pages from offset */
1453 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1457 pg_start = offset >> PAGE_SHIFT;
1458 pg_end = (offset + len) >> PAGE_SHIFT;
1459 delta = pg_end - pg_start;
1460 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1462 /* avoid gc operation during block exchange */
1463 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1464 down_write(&F2FS_I(inode)->i_mmap_sem);
1465 truncate_pagecache(inode, offset);
1467 while (!ret && idx > pg_start) {
1468 nr = idx - pg_start;
1474 f2fs_drop_extent_tree(inode);
1476 ret = __exchange_data_block(inode, inode, idx,
1477 idx + delta, nr, false);
1478 f2fs_unlock_op(sbi);
1480 up_write(&F2FS_I(inode)->i_mmap_sem);
1481 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1483 /* write out all moved pages, if possible */
1484 down_write(&F2FS_I(inode)->i_mmap_sem);
1485 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1486 truncate_pagecache(inode, offset);
1487 up_write(&F2FS_I(inode)->i_mmap_sem);
1490 f2fs_i_size_write(inode, new_size);
1494 static int expand_inode_data(struct inode *inode, loff_t offset,
1495 loff_t len, int mode)
1497 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1498 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1499 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1501 loff_t new_size = i_size_read(inode);
1505 err = inode_newsize_ok(inode, (len + offset));
1509 err = f2fs_convert_inline_inode(inode);
1513 f2fs_balance_fs(sbi, true);
1515 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1516 off_end = (offset + len) & (PAGE_SIZE - 1);
1518 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1519 map.m_len = pg_end - map.m_lblk;
1523 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1530 last_off = map.m_lblk + map.m_len - 1;
1532 /* update new size to the failed position */
1533 new_size = (last_off == pg_end) ? offset + len :
1534 (loff_t)(last_off + 1) << PAGE_SHIFT;
1536 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1539 if (new_size > i_size_read(inode)) {
1540 if (mode & FALLOC_FL_KEEP_SIZE)
1541 file_set_keep_isize(inode);
1543 f2fs_i_size_write(inode, new_size);
1549 static long f2fs_fallocate(struct file *file, int mode,
1550 loff_t offset, loff_t len)
1552 struct inode *inode = file_inode(file);
1555 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1558 /* f2fs only support ->fallocate for regular file */
1559 if (!S_ISREG(inode->i_mode))
1562 if (f2fs_encrypted_inode(inode) &&
1563 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1566 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1567 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1568 FALLOC_FL_INSERT_RANGE))
1573 if (mode & FALLOC_FL_PUNCH_HOLE) {
1574 if (offset >= inode->i_size)
1577 ret = punch_hole(inode, offset, len);
1578 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1579 ret = f2fs_collapse_range(inode, offset, len);
1580 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1581 ret = f2fs_zero_range(inode, offset, len, mode);
1582 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1583 ret = f2fs_insert_range(inode, offset, len);
1585 ret = expand_inode_data(inode, offset, len, mode);
1589 inode->i_mtime = inode->i_ctime = current_time(inode);
1590 f2fs_mark_inode_dirty_sync(inode, false);
1591 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1595 inode_unlock(inode);
1597 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1601 static int f2fs_release_file(struct inode *inode, struct file *filp)
1604 * f2fs_relase_file is called at every close calls. So we should
1605 * not drop any inmemory pages by close called by other process.
1607 if (!(filp->f_mode & FMODE_WRITE) ||
1608 atomic_read(&inode->i_writecount) != 1)
1611 /* some remained atomic pages should discarded */
1612 if (f2fs_is_atomic_file(inode))
1613 f2fs_drop_inmem_pages(inode);
1614 if (f2fs_is_volatile_file(inode)) {
1615 set_inode_flag(inode, FI_DROP_CACHE);
1616 filemap_fdatawrite(inode->i_mapping);
1617 clear_inode_flag(inode, FI_DROP_CACHE);
1618 clear_inode_flag(inode, FI_VOLATILE_FILE);
1619 stat_dec_volatile_write(inode);
1624 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1626 struct inode *inode = file_inode(file);
1629 * If the process doing a transaction is crashed, we should do
1630 * roll-back. Otherwise, other reader/write can see corrupted database
1631 * until all the writers close its file. Since this should be done
1632 * before dropping file lock, it needs to do in ->flush.
1634 if (f2fs_is_atomic_file(inode) &&
1635 F2FS_I(inode)->inmem_task == current)
1636 f2fs_drop_inmem_pages(inode);
1640 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1642 struct inode *inode = file_inode(filp);
1643 struct f2fs_inode_info *fi = F2FS_I(inode);
1644 unsigned int flags = fi->i_flags;
1646 if (f2fs_encrypted_inode(inode))
1647 flags |= F2FS_ENCRYPT_FL;
1648 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1649 flags |= F2FS_INLINE_DATA_FL;
1651 flags &= F2FS_FL_USER_VISIBLE;
1653 return put_user(flags, (int __user *)arg);
1656 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1658 struct f2fs_inode_info *fi = F2FS_I(inode);
1659 unsigned int oldflags;
1661 /* Is it quota file? Do not allow user to mess with it */
1662 if (IS_NOQUOTA(inode))
1665 flags = f2fs_mask_flags(inode->i_mode, flags);
1667 oldflags = fi->i_flags;
1669 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL))
1670 if (!capable(CAP_LINUX_IMMUTABLE))
1673 flags = flags & F2FS_FL_USER_MODIFIABLE;
1674 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE;
1675 fi->i_flags = flags;
1677 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1678 set_inode_flag(inode, FI_PROJ_INHERIT);
1680 clear_inode_flag(inode, FI_PROJ_INHERIT);
1682 inode->i_ctime = current_time(inode);
1683 f2fs_set_inode_flags(inode);
1684 f2fs_mark_inode_dirty_sync(inode, false);
1688 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1690 struct inode *inode = file_inode(filp);
1694 if (!inode_owner_or_capable(inode))
1697 if (get_user(flags, (int __user *)arg))
1700 ret = mnt_want_write_file(filp);
1706 ret = __f2fs_ioc_setflags(inode, flags);
1708 inode_unlock(inode);
1709 mnt_drop_write_file(filp);
1713 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1715 struct inode *inode = file_inode(filp);
1717 return put_user(inode->i_generation, (int __user *)arg);
1720 static int f2fs_ioc_start_atomic_write(struct file *filp)
1722 struct inode *inode = file_inode(filp);
1725 if (!inode_owner_or_capable(inode))
1728 if (!S_ISREG(inode->i_mode))
1731 ret = mnt_want_write_file(filp);
1737 if (f2fs_is_atomic_file(inode)) {
1738 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1743 ret = f2fs_convert_inline_inode(inode);
1747 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1749 if (!get_dirty_pages(inode))
1752 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1753 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1754 inode->i_ino, get_dirty_pages(inode));
1755 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1757 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1761 set_inode_flag(inode, FI_ATOMIC_FILE);
1762 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1763 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1765 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1766 F2FS_I(inode)->inmem_task = current;
1767 stat_inc_atomic_write(inode);
1768 stat_update_max_atomic_write(inode);
1770 inode_unlock(inode);
1771 mnt_drop_write_file(filp);
1775 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1777 struct inode *inode = file_inode(filp);
1780 if (!inode_owner_or_capable(inode))
1783 ret = mnt_want_write_file(filp);
1787 f2fs_balance_fs(F2FS_I_SB(inode), true);
1791 if (f2fs_is_volatile_file(inode)) {
1796 if (f2fs_is_atomic_file(inode)) {
1797 ret = f2fs_commit_inmem_pages(inode);
1801 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1803 clear_inode_flag(inode, FI_ATOMIC_FILE);
1804 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1805 stat_dec_atomic_write(inode);
1808 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1811 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1812 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1815 inode_unlock(inode);
1816 mnt_drop_write_file(filp);
1820 static int f2fs_ioc_start_volatile_write(struct file *filp)
1822 struct inode *inode = file_inode(filp);
1825 if (!inode_owner_or_capable(inode))
1828 if (!S_ISREG(inode->i_mode))
1831 ret = mnt_want_write_file(filp);
1837 if (f2fs_is_volatile_file(inode))
1840 ret = f2fs_convert_inline_inode(inode);
1844 stat_inc_volatile_write(inode);
1845 stat_update_max_volatile_write(inode);
1847 set_inode_flag(inode, FI_VOLATILE_FILE);
1848 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1850 inode_unlock(inode);
1851 mnt_drop_write_file(filp);
1855 static int f2fs_ioc_release_volatile_write(struct file *filp)
1857 struct inode *inode = file_inode(filp);
1860 if (!inode_owner_or_capable(inode))
1863 ret = mnt_want_write_file(filp);
1869 if (!f2fs_is_volatile_file(inode))
1872 if (!f2fs_is_first_block_written(inode)) {
1873 ret = truncate_partial_data_page(inode, 0, true);
1877 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1879 inode_unlock(inode);
1880 mnt_drop_write_file(filp);
1884 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1886 struct inode *inode = file_inode(filp);
1889 if (!inode_owner_or_capable(inode))
1892 ret = mnt_want_write_file(filp);
1898 if (f2fs_is_atomic_file(inode))
1899 f2fs_drop_inmem_pages(inode);
1900 if (f2fs_is_volatile_file(inode)) {
1901 clear_inode_flag(inode, FI_VOLATILE_FILE);
1902 stat_dec_volatile_write(inode);
1903 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1906 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1908 inode_unlock(inode);
1910 mnt_drop_write_file(filp);
1911 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1915 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1917 struct inode *inode = file_inode(filp);
1918 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1919 struct super_block *sb = sbi->sb;
1923 if (!capable(CAP_SYS_ADMIN))
1926 if (get_user(in, (__u32 __user *)arg))
1929 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1930 ret = mnt_want_write_file(filp);
1936 case F2FS_GOING_DOWN_FULLSYNC:
1937 sb = freeze_bdev(sb->s_bdev);
1943 f2fs_stop_checkpoint(sbi, false);
1944 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1945 thaw_bdev(sb->s_bdev, sb);
1948 case F2FS_GOING_DOWN_METASYNC:
1949 /* do checkpoint only */
1950 ret = f2fs_sync_fs(sb, 1);
1953 f2fs_stop_checkpoint(sbi, false);
1954 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1956 case F2FS_GOING_DOWN_NOSYNC:
1957 f2fs_stop_checkpoint(sbi, false);
1958 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1960 case F2FS_GOING_DOWN_METAFLUSH:
1961 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1962 f2fs_stop_checkpoint(sbi, false);
1963 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1970 f2fs_stop_gc_thread(sbi);
1971 f2fs_stop_discard_thread(sbi);
1973 f2fs_drop_discard_cmd(sbi);
1974 clear_opt(sbi, DISCARD);
1976 f2fs_update_time(sbi, REQ_TIME);
1978 if (in != F2FS_GOING_DOWN_FULLSYNC)
1979 mnt_drop_write_file(filp);
1983 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1985 struct inode *inode = file_inode(filp);
1986 struct super_block *sb = inode->i_sb;
1987 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1988 struct fstrim_range range;
1991 if (!capable(CAP_SYS_ADMIN))
1994 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
1997 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2001 ret = mnt_want_write_file(filp);
2005 range.minlen = max((unsigned int)range.minlen,
2006 q->limits.discard_granularity);
2007 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2008 mnt_drop_write_file(filp);
2012 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2015 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2019 static bool uuid_is_nonzero(__u8 u[16])
2023 for (i = 0; i < 16; i++)
2029 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2031 struct inode *inode = file_inode(filp);
2033 if (!f2fs_sb_has_encrypt(inode->i_sb))
2036 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2038 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2041 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2043 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
2045 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2048 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2050 struct inode *inode = file_inode(filp);
2051 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2054 if (!f2fs_sb_has_encrypt(inode->i_sb))
2057 err = mnt_want_write_file(filp);
2061 down_write(&sbi->sb_lock);
2063 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2066 /* update superblock with uuid */
2067 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2069 err = f2fs_commit_super(sbi, false);
2072 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2076 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2080 up_write(&sbi->sb_lock);
2081 mnt_drop_write_file(filp);
2085 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2087 struct inode *inode = file_inode(filp);
2088 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2092 if (!capable(CAP_SYS_ADMIN))
2095 if (get_user(sync, (__u32 __user *)arg))
2098 if (f2fs_readonly(sbi->sb))
2101 ret = mnt_want_write_file(filp);
2106 if (!mutex_trylock(&sbi->gc_mutex)) {
2111 mutex_lock(&sbi->gc_mutex);
2114 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2116 mnt_drop_write_file(filp);
2120 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2122 struct inode *inode = file_inode(filp);
2123 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2124 struct f2fs_gc_range range;
2128 if (!capable(CAP_SYS_ADMIN))
2131 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2135 if (f2fs_readonly(sbi->sb))
2138 end = range.start + range.len;
2139 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2143 ret = mnt_want_write_file(filp);
2149 if (!mutex_trylock(&sbi->gc_mutex)) {
2154 mutex_lock(&sbi->gc_mutex);
2157 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2158 range.start += sbi->blocks_per_seg;
2159 if (range.start <= end)
2162 mnt_drop_write_file(filp);
2166 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2168 struct inode *inode = file_inode(filp);
2169 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2172 if (!capable(CAP_SYS_ADMIN))
2175 if (f2fs_readonly(sbi->sb))
2178 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2179 f2fs_msg(sbi->sb, KERN_INFO,
2180 "Skipping Checkpoint. Checkpoints currently disabled.");
2184 ret = mnt_want_write_file(filp);
2188 ret = f2fs_sync_fs(sbi->sb, 1);
2190 mnt_drop_write_file(filp);
2194 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2196 struct f2fs_defragment *range)
2198 struct inode *inode = file_inode(filp);
2199 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2200 .m_seg_type = NO_CHECK_TYPE };
2201 struct extent_info ei = {0, 0, 0};
2202 pgoff_t pg_start, pg_end, next_pgofs;
2203 unsigned int blk_per_seg = sbi->blocks_per_seg;
2204 unsigned int total = 0, sec_num;
2205 block_t blk_end = 0;
2206 bool fragmented = false;
2209 /* if in-place-update policy is enabled, don't waste time here */
2210 if (f2fs_should_update_inplace(inode, NULL))
2213 pg_start = range->start >> PAGE_SHIFT;
2214 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2216 f2fs_balance_fs(sbi, true);
2220 /* writeback all dirty pages in the range */
2221 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2222 range->start + range->len - 1);
2227 * lookup mapping info in extent cache, skip defragmenting if physical
2228 * block addresses are continuous.
2230 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2231 if (ei.fofs + ei.len >= pg_end)
2235 map.m_lblk = pg_start;
2236 map.m_next_pgofs = &next_pgofs;
2239 * lookup mapping info in dnode page cache, skip defragmenting if all
2240 * physical block addresses are continuous even if there are hole(s)
2241 * in logical blocks.
2243 while (map.m_lblk < pg_end) {
2244 map.m_len = pg_end - map.m_lblk;
2245 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2249 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2250 map.m_lblk = next_pgofs;
2254 if (blk_end && blk_end != map.m_pblk)
2257 /* record total count of block that we're going to move */
2260 blk_end = map.m_pblk + map.m_len;
2262 map.m_lblk += map.m_len;
2268 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2271 * make sure there are enough free section for LFS allocation, this can
2272 * avoid defragment running in SSR mode when free section are allocated
2275 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2280 map.m_lblk = pg_start;
2281 map.m_len = pg_end - pg_start;
2284 while (map.m_lblk < pg_end) {
2289 map.m_len = pg_end - map.m_lblk;
2290 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2294 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2295 map.m_lblk = next_pgofs;
2299 set_inode_flag(inode, FI_DO_DEFRAG);
2302 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2305 page = f2fs_get_lock_data_page(inode, idx, true);
2307 err = PTR_ERR(page);
2311 set_page_dirty(page);
2312 f2fs_put_page(page, 1);
2321 if (idx < pg_end && cnt < blk_per_seg)
2324 clear_inode_flag(inode, FI_DO_DEFRAG);
2326 err = filemap_fdatawrite(inode->i_mapping);
2331 clear_inode_flag(inode, FI_DO_DEFRAG);
2333 inode_unlock(inode);
2335 range->len = (u64)total << PAGE_SHIFT;
2339 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2341 struct inode *inode = file_inode(filp);
2342 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2343 struct f2fs_defragment range;
2346 if (!capable(CAP_SYS_ADMIN))
2349 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2352 if (f2fs_readonly(sbi->sb))
2355 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2359 /* verify alignment of offset & size */
2360 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2363 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2364 sbi->max_file_blocks))
2367 err = mnt_want_write_file(filp);
2371 err = f2fs_defragment_range(sbi, filp, &range);
2372 mnt_drop_write_file(filp);
2374 f2fs_update_time(sbi, REQ_TIME);
2378 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2385 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2386 struct file *file_out, loff_t pos_out, size_t len)
2388 struct inode *src = file_inode(file_in);
2389 struct inode *dst = file_inode(file_out);
2390 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2391 size_t olen = len, dst_max_i_size = 0;
2395 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2396 src->i_sb != dst->i_sb)
2399 if (unlikely(f2fs_readonly(src->i_sb)))
2402 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2405 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2409 if (pos_in == pos_out)
2411 if (pos_out > pos_in && pos_out < pos_in + len)
2418 if (!inode_trylock(dst))
2423 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2426 olen = len = src->i_size - pos_in;
2427 if (pos_in + len == src->i_size)
2428 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2434 dst_osize = dst->i_size;
2435 if (pos_out + olen > dst->i_size)
2436 dst_max_i_size = pos_out + olen;
2438 /* verify the end result is block aligned */
2439 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2440 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2441 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2444 ret = f2fs_convert_inline_inode(src);
2448 ret = f2fs_convert_inline_inode(dst);
2452 /* write out all dirty pages from offset */
2453 ret = filemap_write_and_wait_range(src->i_mapping,
2454 pos_in, pos_in + len);
2458 ret = filemap_write_and_wait_range(dst->i_mapping,
2459 pos_out, pos_out + len);
2463 f2fs_balance_fs(sbi, true);
2465 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2468 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2473 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2474 pos_out >> F2FS_BLKSIZE_BITS,
2475 len >> F2FS_BLKSIZE_BITS, false);
2479 f2fs_i_size_write(dst, dst_max_i_size);
2480 else if (dst_osize != dst->i_size)
2481 f2fs_i_size_write(dst, dst_osize);
2483 f2fs_unlock_op(sbi);
2486 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2488 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2497 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2499 struct f2fs_move_range range;
2503 if (!(filp->f_mode & FMODE_READ) ||
2504 !(filp->f_mode & FMODE_WRITE))
2507 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2511 dst = fdget(range.dst_fd);
2515 if (!(dst.file->f_mode & FMODE_WRITE)) {
2520 err = mnt_want_write_file(filp);
2524 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2525 range.pos_out, range.len);
2527 mnt_drop_write_file(filp);
2531 if (copy_to_user((struct f2fs_move_range __user *)arg,
2532 &range, sizeof(range)))
2539 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2541 struct inode *inode = file_inode(filp);
2542 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2543 struct sit_info *sm = SIT_I(sbi);
2544 unsigned int start_segno = 0, end_segno = 0;
2545 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2546 struct f2fs_flush_device range;
2549 if (!capable(CAP_SYS_ADMIN))
2552 if (f2fs_readonly(sbi->sb))
2555 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2558 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2562 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2563 sbi->segs_per_sec != 1) {
2564 f2fs_msg(sbi->sb, KERN_WARNING,
2565 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2566 range.dev_num, sbi->s_ndevs,
2571 ret = mnt_want_write_file(filp);
2575 if (range.dev_num != 0)
2576 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2577 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2579 start_segno = sm->last_victim[FLUSH_DEVICE];
2580 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2581 start_segno = dev_start_segno;
2582 end_segno = min(start_segno + range.segments, dev_end_segno);
2584 while (start_segno < end_segno) {
2585 if (!mutex_trylock(&sbi->gc_mutex)) {
2589 sm->last_victim[GC_CB] = end_segno + 1;
2590 sm->last_victim[GC_GREEDY] = end_segno + 1;
2591 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2592 ret = f2fs_gc(sbi, true, true, start_segno);
2600 mnt_drop_write_file(filp);
2604 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2606 struct inode *inode = file_inode(filp);
2607 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2609 /* Must validate to set it with SQLite behavior in Android. */
2610 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2612 return put_user(sb_feature, (u32 __user *)arg);
2616 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2618 struct dquot *transfer_to[MAXQUOTAS] = {};
2619 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2620 struct super_block *sb = sbi->sb;
2623 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2624 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2625 err = __dquot_transfer(inode, transfer_to);
2627 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2628 dqput(transfer_to[PRJQUOTA]);
2633 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2635 struct inode *inode = file_inode(filp);
2636 struct f2fs_inode_info *fi = F2FS_I(inode);
2637 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2638 struct super_block *sb = sbi->sb;
2643 if (!f2fs_sb_has_project_quota(sb)) {
2644 if (projid != F2FS_DEF_PROJID)
2650 if (!f2fs_has_extra_attr(inode))
2653 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2655 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2659 /* Is it quota file? Do not allow user to mess with it */
2660 if (IS_NOQUOTA(inode))
2663 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2665 return PTR_ERR(ipage);
2667 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2670 f2fs_put_page(ipage, 1);
2673 f2fs_put_page(ipage, 1);
2675 err = dquot_initialize(inode);
2680 err = f2fs_transfer_project_quota(inode, kprojid);
2684 F2FS_I(inode)->i_projid = kprojid;
2685 inode->i_ctime = current_time(inode);
2686 f2fs_mark_inode_dirty_sync(inode, true);
2688 f2fs_unlock_op(sbi);
2692 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2697 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2699 if (projid != F2FS_DEF_PROJID)
2705 /* Transfer internal flags to xflags */
2706 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2710 if (iflags & F2FS_SYNC_FL)
2711 xflags |= FS_XFLAG_SYNC;
2712 if (iflags & F2FS_IMMUTABLE_FL)
2713 xflags |= FS_XFLAG_IMMUTABLE;
2714 if (iflags & F2FS_APPEND_FL)
2715 xflags |= FS_XFLAG_APPEND;
2716 if (iflags & F2FS_NODUMP_FL)
2717 xflags |= FS_XFLAG_NODUMP;
2718 if (iflags & F2FS_NOATIME_FL)
2719 xflags |= FS_XFLAG_NOATIME;
2720 if (iflags & F2FS_PROJINHERIT_FL)
2721 xflags |= FS_XFLAG_PROJINHERIT;
2725 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2726 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2727 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2729 /* Transfer xflags flags to internal */
2730 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2732 unsigned long iflags = 0;
2734 if (xflags & FS_XFLAG_SYNC)
2735 iflags |= F2FS_SYNC_FL;
2736 if (xflags & FS_XFLAG_IMMUTABLE)
2737 iflags |= F2FS_IMMUTABLE_FL;
2738 if (xflags & FS_XFLAG_APPEND)
2739 iflags |= F2FS_APPEND_FL;
2740 if (xflags & FS_XFLAG_NODUMP)
2741 iflags |= F2FS_NODUMP_FL;
2742 if (xflags & FS_XFLAG_NOATIME)
2743 iflags |= F2FS_NOATIME_FL;
2744 if (xflags & FS_XFLAG_PROJINHERIT)
2745 iflags |= F2FS_PROJINHERIT_FL;
2750 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2752 struct inode *inode = file_inode(filp);
2753 struct f2fs_inode_info *fi = F2FS_I(inode);
2756 memset(&fa, 0, sizeof(struct fsxattr));
2757 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2758 F2FS_FL_USER_VISIBLE);
2760 if (f2fs_sb_has_project_quota(inode->i_sb))
2761 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2764 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2769 static int f2fs_ioctl_check_project(struct inode *inode, struct fsxattr *fa)
2772 * Project Quota ID state is only allowed to change from within the init
2773 * namespace. Enforce that restriction only if we are trying to change
2774 * the quota ID state. Everything else is allowed in user namespaces.
2776 if (current_user_ns() == &init_user_ns)
2779 if (__kprojid_val(F2FS_I(inode)->i_projid) != fa->fsx_projid)
2782 if (F2FS_I(inode)->i_flags & F2FS_PROJINHERIT_FL) {
2783 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT))
2786 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
2793 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2795 struct inode *inode = file_inode(filp);
2796 struct f2fs_inode_info *fi = F2FS_I(inode);
2801 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2804 /* Make sure caller has proper permission */
2805 if (!inode_owner_or_capable(inode))
2808 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2811 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2812 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2815 err = mnt_want_write_file(filp);
2820 err = f2fs_ioctl_check_project(inode, &fa);
2823 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2824 (flags & F2FS_FL_XFLAG_VISIBLE);
2825 err = __f2fs_ioc_setflags(inode, flags);
2829 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2831 inode_unlock(inode);
2832 mnt_drop_write_file(filp);
2836 int f2fs_pin_file_control(struct inode *inode, bool inc)
2838 struct f2fs_inode_info *fi = F2FS_I(inode);
2839 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2841 /* Use i_gc_failures for normal file as a risk signal. */
2843 f2fs_i_gc_failures_write(inode,
2844 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2846 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2847 f2fs_msg(sbi->sb, KERN_WARNING,
2848 "%s: Enable GC = ino %lx after %x GC trials\n",
2849 __func__, inode->i_ino,
2850 fi->i_gc_failures[GC_FAILURE_PIN]);
2851 clear_inode_flag(inode, FI_PIN_FILE);
2857 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2859 struct inode *inode = file_inode(filp);
2863 if (!inode_owner_or_capable(inode))
2866 if (get_user(pin, (__u32 __user *)arg))
2869 if (!S_ISREG(inode->i_mode))
2872 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2875 ret = mnt_want_write_file(filp);
2881 if (f2fs_should_update_outplace(inode, NULL)) {
2887 clear_inode_flag(inode, FI_PIN_FILE);
2888 f2fs_i_gc_failures_write(inode, 0);
2892 if (f2fs_pin_file_control(inode, false)) {
2896 ret = f2fs_convert_inline_inode(inode);
2900 set_inode_flag(inode, FI_PIN_FILE);
2901 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2903 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2905 inode_unlock(inode);
2906 mnt_drop_write_file(filp);
2910 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2912 struct inode *inode = file_inode(filp);
2915 if (is_inode_flag_set(inode, FI_PIN_FILE))
2916 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2917 return put_user(pin, (u32 __user *)arg);
2920 int f2fs_precache_extents(struct inode *inode)
2922 struct f2fs_inode_info *fi = F2FS_I(inode);
2923 struct f2fs_map_blocks map;
2924 pgoff_t m_next_extent;
2928 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2932 map.m_next_pgofs = NULL;
2933 map.m_next_extent = &m_next_extent;
2934 map.m_seg_type = NO_CHECK_TYPE;
2935 end = F2FS_I_SB(inode)->max_file_blocks;
2937 while (map.m_lblk < end) {
2938 map.m_len = end - map.m_lblk;
2940 down_write(&fi->i_gc_rwsem[WRITE]);
2941 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2942 up_write(&fi->i_gc_rwsem[WRITE]);
2946 map.m_lblk = m_next_extent;
2952 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2954 return f2fs_precache_extents(file_inode(filp));
2957 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2959 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2963 case F2FS_IOC_GETFLAGS:
2964 return f2fs_ioc_getflags(filp, arg);
2965 case F2FS_IOC_SETFLAGS:
2966 return f2fs_ioc_setflags(filp, arg);
2967 case F2FS_IOC_GETVERSION:
2968 return f2fs_ioc_getversion(filp, arg);
2969 case F2FS_IOC_START_ATOMIC_WRITE:
2970 return f2fs_ioc_start_atomic_write(filp);
2971 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2972 return f2fs_ioc_commit_atomic_write(filp);
2973 case F2FS_IOC_START_VOLATILE_WRITE:
2974 return f2fs_ioc_start_volatile_write(filp);
2975 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2976 return f2fs_ioc_release_volatile_write(filp);
2977 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2978 return f2fs_ioc_abort_volatile_write(filp);
2979 case F2FS_IOC_SHUTDOWN:
2980 return f2fs_ioc_shutdown(filp, arg);
2982 return f2fs_ioc_fitrim(filp, arg);
2983 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2984 return f2fs_ioc_set_encryption_policy(filp, arg);
2985 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2986 return f2fs_ioc_get_encryption_policy(filp, arg);
2987 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2988 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2989 case F2FS_IOC_GARBAGE_COLLECT:
2990 return f2fs_ioc_gc(filp, arg);
2991 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2992 return f2fs_ioc_gc_range(filp, arg);
2993 case F2FS_IOC_WRITE_CHECKPOINT:
2994 return f2fs_ioc_write_checkpoint(filp, arg);
2995 case F2FS_IOC_DEFRAGMENT:
2996 return f2fs_ioc_defragment(filp, arg);
2997 case F2FS_IOC_MOVE_RANGE:
2998 return f2fs_ioc_move_range(filp, arg);
2999 case F2FS_IOC_FLUSH_DEVICE:
3000 return f2fs_ioc_flush_device(filp, arg);
3001 case F2FS_IOC_GET_FEATURES:
3002 return f2fs_ioc_get_features(filp, arg);
3003 case F2FS_IOC_FSGETXATTR:
3004 return f2fs_ioc_fsgetxattr(filp, arg);
3005 case F2FS_IOC_FSSETXATTR:
3006 return f2fs_ioc_fssetxattr(filp, arg);
3007 case F2FS_IOC_GET_PIN_FILE:
3008 return f2fs_ioc_get_pin_file(filp, arg);
3009 case F2FS_IOC_SET_PIN_FILE:
3010 return f2fs_ioc_set_pin_file(filp, arg);
3011 case F2FS_IOC_PRECACHE_EXTENTS:
3012 return f2fs_ioc_precache_extents(filp, arg);
3018 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3020 struct file *file = iocb->ki_filp;
3021 struct inode *inode = file_inode(file);
3024 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3027 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
3030 if (!inode_trylock(inode)) {
3031 if (iocb->ki_flags & IOCB_NOWAIT)
3036 ret = generic_write_checks(iocb, from);
3038 bool preallocated = false;
3039 size_t target_size = 0;
3042 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3043 set_inode_flag(inode, FI_NO_PREALLOC);
3045 if ((iocb->ki_flags & IOCB_NOWAIT) &&
3046 (iocb->ki_flags & IOCB_DIRECT)) {
3047 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3048 iov_iter_count(from)) ||
3049 f2fs_has_inline_data(inode) ||
3050 f2fs_force_buffered_io(inode,
3052 clear_inode_flag(inode,
3054 inode_unlock(inode);
3059 preallocated = true;
3060 target_size = iocb->ki_pos + iov_iter_count(from);
3062 err = f2fs_preallocate_blocks(iocb, from);
3064 clear_inode_flag(inode, FI_NO_PREALLOC);
3065 inode_unlock(inode);
3069 ret = __generic_file_write_iter(iocb, from);
3070 clear_inode_flag(inode, FI_NO_PREALLOC);
3072 /* if we couldn't write data, we should deallocate blocks. */
3073 if (preallocated && i_size_read(inode) < target_size)
3074 f2fs_truncate(inode);
3077 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3079 inode_unlock(inode);
3082 ret = generic_write_sync(iocb, ret);
3086 #ifdef CONFIG_COMPAT
3087 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3090 case F2FS_IOC32_GETFLAGS:
3091 cmd = F2FS_IOC_GETFLAGS;
3093 case F2FS_IOC32_SETFLAGS:
3094 cmd = F2FS_IOC_SETFLAGS;
3096 case F2FS_IOC32_GETVERSION:
3097 cmd = F2FS_IOC_GETVERSION;
3099 case F2FS_IOC_START_ATOMIC_WRITE:
3100 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3101 case F2FS_IOC_START_VOLATILE_WRITE:
3102 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3103 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3104 case F2FS_IOC_SHUTDOWN:
3105 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3106 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3107 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3108 case F2FS_IOC_GARBAGE_COLLECT:
3109 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3110 case F2FS_IOC_WRITE_CHECKPOINT:
3111 case F2FS_IOC_DEFRAGMENT:
3112 case F2FS_IOC_MOVE_RANGE:
3113 case F2FS_IOC_FLUSH_DEVICE:
3114 case F2FS_IOC_GET_FEATURES:
3115 case F2FS_IOC_FSGETXATTR:
3116 case F2FS_IOC_FSSETXATTR:
3117 case F2FS_IOC_GET_PIN_FILE:
3118 case F2FS_IOC_SET_PIN_FILE:
3119 case F2FS_IOC_PRECACHE_EXTENTS:
3122 return -ENOIOCTLCMD;
3124 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3128 const struct file_operations f2fs_file_operations = {
3129 .llseek = f2fs_llseek,
3130 .read_iter = generic_file_read_iter,
3131 .write_iter = f2fs_file_write_iter,
3132 .open = f2fs_file_open,
3133 .release = f2fs_release_file,
3134 .mmap = f2fs_file_mmap,
3135 .flush = f2fs_file_flush,
3136 .fsync = f2fs_sync_file,
3137 .fallocate = f2fs_fallocate,
3138 .unlocked_ioctl = f2fs_ioctl,
3139 #ifdef CONFIG_COMPAT
3140 .compat_ioctl = f2fs_compat_ioctl,
3142 .splice_read = generic_file_splice_read,
3143 .splice_write = iter_file_splice_write,