4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
12 #include <linux/f2fs_fs.h>
13 #include <linux/stat.h>
14 #include <linux/buffer_head.h>
15 #include <linux/writeback.h>
16 #include <linux/blkdev.h>
17 #include <linux/falloc.h>
18 #include <linux/types.h>
19 #include <linux/compat.h>
20 #include <linux/uaccess.h>
21 #include <linux/mount.h>
22 #include <linux/pagevec.h>
23 #include <linux/uio.h>
24 #include <linux/uuid.h>
25 #include <linux/file.h>
34 #include <trace/events/f2fs.h>
36 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
38 struct inode *inode = file_inode(vmf->vma->vm_file);
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 ret = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
48 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
50 struct page *page = vmf->page;
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 struct dnode_of_data dn;
56 if (unlikely(f2fs_cp_error(sbi))) {
61 sb_start_pagefault(inode->i_sb);
63 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
65 /* block allocation */
67 set_new_dnode(&dn, inode, NULL, NULL, 0);
68 err = f2fs_reserve_block(&dn, page->index);
76 f2fs_balance_fs(sbi, dn.node_changed);
78 file_update_time(vmf->vma->vm_file);
79 down_read(&F2FS_I(inode)->i_mmap_sem);
81 if (unlikely(page->mapping != inode->i_mapping ||
82 page_offset(page) > i_size_read(inode) ||
83 !PageUptodate(page))) {
90 * check to see if the page is mapped already (no holes)
92 if (PageMappedToDisk(page))
95 /* page is wholly or partially inside EOF */
96 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
100 offset = i_size_read(inode) & ~PAGE_MASK;
101 zero_user_segment(page, offset, PAGE_SIZE);
103 set_page_dirty(page);
104 if (!PageUptodate(page))
105 SetPageUptodate(page);
107 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
109 trace_f2fs_vm_page_mkwrite(page, DATA);
112 f2fs_wait_on_page_writeback(page, DATA, false);
114 /* wait for GCed page writeback via META_MAPPING */
115 if (f2fs_post_read_required(inode))
116 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
119 up_read(&F2FS_I(inode)->i_mmap_sem);
121 sb_end_pagefault(inode->i_sb);
122 f2fs_update_time(sbi, REQ_TIME);
124 return block_page_mkwrite_return(err);
127 static const struct vm_operations_struct f2fs_file_vm_ops = {
128 .fault = f2fs_filemap_fault,
129 .map_pages = filemap_map_pages,
130 .page_mkwrite = f2fs_vm_page_mkwrite,
133 static int get_parent_ino(struct inode *inode, nid_t *pino)
135 struct dentry *dentry;
137 inode = igrab(inode);
138 dentry = d_find_any_alias(inode);
143 *pino = parent_ino(dentry);
148 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
150 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
151 enum cp_reason_type cp_reason = CP_NO_NEEDED;
153 if (!S_ISREG(inode->i_mode))
154 cp_reason = CP_NON_REGULAR;
155 else if (inode->i_nlink != 1)
156 cp_reason = CP_HARDLINK;
157 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
158 cp_reason = CP_SB_NEED_CP;
159 else if (file_wrong_pino(inode))
160 cp_reason = CP_WRONG_PINO;
161 else if (!f2fs_space_for_roll_forward(sbi))
162 cp_reason = CP_NO_SPC_ROLL;
163 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
164 cp_reason = CP_NODE_NEED_CP;
165 else if (test_opt(sbi, FASTBOOT))
166 cp_reason = CP_FASTBOOT_MODE;
167 else if (F2FS_OPTION(sbi).active_logs == 2)
168 cp_reason = CP_SPEC_LOG_NUM;
169 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
170 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
171 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
173 cp_reason = CP_RECOVER_DIR;
178 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
180 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
182 /* But we need to avoid that there are some inode updates */
183 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
189 static void try_to_fix_pino(struct inode *inode)
191 struct f2fs_inode_info *fi = F2FS_I(inode);
194 down_write(&fi->i_sem);
195 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
196 get_parent_ino(inode, &pino)) {
197 f2fs_i_pino_write(inode, pino);
198 file_got_pino(inode);
200 up_write(&fi->i_sem);
203 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
204 int datasync, bool atomic)
206 struct inode *inode = file->f_mapping->host;
207 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
208 nid_t ino = inode->i_ino;
210 enum cp_reason_type cp_reason = 0;
211 struct writeback_control wbc = {
212 .sync_mode = WB_SYNC_ALL,
213 .nr_to_write = LONG_MAX,
217 if (unlikely(f2fs_readonly(inode->i_sb)))
220 trace_f2fs_sync_file_enter(inode);
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);
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, ino);
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(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
358 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
359 is_valid_blkaddr(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 (__found_offset(blkaddr, dirty, pgofs, whence)) {
431 if (whence == SEEK_DATA)
434 if (whence == SEEK_HOLE && data_ofs > isize)
437 return vfs_setpos(file, data_ofs, maxbytes);
443 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
445 struct inode *inode = file->f_mapping->host;
446 loff_t maxbytes = inode->i_sb->s_maxbytes;
452 return generic_file_llseek_size(file, offset, whence,
453 maxbytes, i_size_read(inode));
458 return f2fs_seek_block(file, offset, whence);
464 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
466 struct inode *inode = file_inode(file);
469 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
472 /* we don't need to use inline_data strictly */
473 err = f2fs_convert_inline_inode(inode);
478 vma->vm_ops = &f2fs_file_vm_ops;
482 static int f2fs_file_open(struct inode *inode, struct file *filp)
484 int err = fscrypt_file_open(inode, filp);
489 filp->f_mode |= FMODE_NOWAIT;
491 return dquot_file_open(inode, filp);
494 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
496 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
497 struct f2fs_node *raw_node;
498 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
502 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
503 base = get_extra_isize(dn->inode);
505 raw_node = F2FS_NODE(dn->node_page);
506 addr = blkaddr_in_node(raw_node) + base + ofs;
508 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
509 block_t blkaddr = le32_to_cpu(*addr);
511 if (blkaddr == NULL_ADDR)
514 dn->data_blkaddr = NULL_ADDR;
515 f2fs_set_data_blkaddr(dn);
516 f2fs_invalidate_blocks(sbi, blkaddr);
517 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
518 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
525 * once we invalidate valid blkaddr in range [ofs, ofs + count],
526 * we will invalidate all blkaddr in the whole range.
528 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
530 f2fs_update_extent_cache_range(dn, fofs, 0, len);
531 dec_valid_block_count(sbi, dn->inode, nr_free);
533 dn->ofs_in_node = ofs;
535 f2fs_update_time(sbi, REQ_TIME);
536 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
537 dn->ofs_in_node, nr_free);
540 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
542 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
545 static int truncate_partial_data_page(struct inode *inode, u64 from,
548 loff_t offset = from & (PAGE_SIZE - 1);
549 pgoff_t index = from >> PAGE_SHIFT;
550 struct address_space *mapping = inode->i_mapping;
553 if (!offset && !cache_only)
557 page = find_lock_page(mapping, index);
558 if (page && PageUptodate(page))
560 f2fs_put_page(page, 1);
564 page = f2fs_get_lock_data_page(inode, index, true);
566 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
568 f2fs_wait_on_page_writeback(page, DATA, true);
569 zero_user(page, offset, PAGE_SIZE - offset);
571 /* An encrypted inode should have a key and truncate the last page. */
572 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
574 set_page_dirty(page);
575 f2fs_put_page(page, 1);
579 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
581 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
582 struct dnode_of_data dn;
584 int count = 0, err = 0;
586 bool truncate_page = false;
588 trace_f2fs_truncate_blocks_enter(inode, from);
590 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
592 if (free_from >= sbi->max_file_blocks)
598 ipage = f2fs_get_node_page(sbi, inode->i_ino);
600 err = PTR_ERR(ipage);
604 if (f2fs_has_inline_data(inode)) {
605 f2fs_truncate_inline_inode(inode, ipage, from);
606 f2fs_put_page(ipage, 1);
607 truncate_page = true;
611 set_new_dnode(&dn, inode, ipage, NULL, 0);
612 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
619 count = ADDRS_PER_PAGE(dn.node_page, inode);
621 count -= dn.ofs_in_node;
622 f2fs_bug_on(sbi, count < 0);
624 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
625 f2fs_truncate_data_blocks_range(&dn, count);
631 err = f2fs_truncate_inode_blocks(inode, free_from);
636 /* lastly zero out the first data page */
638 err = truncate_partial_data_page(inode, from, truncate_page);
640 trace_f2fs_truncate_blocks_exit(inode, err);
644 int f2fs_truncate(struct inode *inode)
648 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
651 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
652 S_ISLNK(inode->i_mode)))
655 trace_f2fs_truncate(inode);
657 #ifdef CONFIG_F2FS_FAULT_INJECTION
658 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
659 f2fs_show_injection_info(FAULT_TRUNCATE);
663 /* we should check inline_data size */
664 if (!f2fs_may_inline_data(inode)) {
665 err = f2fs_convert_inline_inode(inode);
670 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
674 inode->i_mtime = inode->i_ctime = current_time(inode);
675 f2fs_mark_inode_dirty_sync(inode, false);
679 int f2fs_getattr(const struct path *path, struct kstat *stat,
680 u32 request_mask, unsigned int query_flags)
682 struct inode *inode = d_inode(path->dentry);
683 struct f2fs_inode_info *fi = F2FS_I(inode);
684 struct f2fs_inode *ri;
687 if (f2fs_has_extra_attr(inode) &&
688 f2fs_sb_has_inode_crtime(inode->i_sb) &&
689 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
690 stat->result_mask |= STATX_BTIME;
691 stat->btime.tv_sec = fi->i_crtime.tv_sec;
692 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
695 flags = fi->i_flags & F2FS_FL_USER_VISIBLE;
696 if (flags & F2FS_APPEND_FL)
697 stat->attributes |= STATX_ATTR_APPEND;
698 if (flags & F2FS_COMPR_FL)
699 stat->attributes |= STATX_ATTR_COMPRESSED;
700 if (f2fs_encrypted_inode(inode))
701 stat->attributes |= STATX_ATTR_ENCRYPTED;
702 if (flags & F2FS_IMMUTABLE_FL)
703 stat->attributes |= STATX_ATTR_IMMUTABLE;
704 if (flags & F2FS_NODUMP_FL)
705 stat->attributes |= STATX_ATTR_NODUMP;
707 stat->attributes_mask |= (STATX_ATTR_APPEND |
708 STATX_ATTR_COMPRESSED |
709 STATX_ATTR_ENCRYPTED |
710 STATX_ATTR_IMMUTABLE |
713 generic_fillattr(inode, stat);
715 /* we need to show initial sectors used for inline_data/dentries */
716 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
717 f2fs_has_inline_dentry(inode))
718 stat->blocks += (stat->size + 511) >> 9;
723 #ifdef CONFIG_F2FS_FS_POSIX_ACL
724 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
726 unsigned int ia_valid = attr->ia_valid;
728 if (ia_valid & ATTR_UID)
729 inode->i_uid = attr->ia_uid;
730 if (ia_valid & ATTR_GID)
731 inode->i_gid = attr->ia_gid;
732 if (ia_valid & ATTR_ATIME)
733 inode->i_atime = timespec64_trunc(attr->ia_atime,
734 inode->i_sb->s_time_gran);
735 if (ia_valid & ATTR_MTIME)
736 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
737 inode->i_sb->s_time_gran);
738 if (ia_valid & ATTR_CTIME)
739 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
740 inode->i_sb->s_time_gran);
741 if (ia_valid & ATTR_MODE) {
742 umode_t mode = attr->ia_mode;
744 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
746 set_acl_inode(inode, mode);
750 #define __setattr_copy setattr_copy
753 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
755 struct inode *inode = d_inode(dentry);
757 bool size_changed = false;
759 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
762 err = setattr_prepare(dentry, attr);
766 err = fscrypt_prepare_setattr(dentry, attr);
770 if (is_quota_modification(inode, attr)) {
771 err = dquot_initialize(inode);
775 if ((attr->ia_valid & ATTR_UID &&
776 !uid_eq(attr->ia_uid, inode->i_uid)) ||
777 (attr->ia_valid & ATTR_GID &&
778 !gid_eq(attr->ia_gid, inode->i_gid))) {
779 err = dquot_transfer(inode, attr);
784 if (attr->ia_valid & ATTR_SIZE) {
785 bool to_smaller = (attr->ia_size <= i_size_read(inode));
787 down_write(&F2FS_I(inode)->i_mmap_sem);
788 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
790 truncate_setsize(inode, attr->ia_size);
793 err = f2fs_truncate(inode);
795 * do not trim all blocks after i_size if target size is
796 * larger than i_size.
798 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
799 up_write(&F2FS_I(inode)->i_mmap_sem);
805 /* should convert inline inode here */
806 if (!f2fs_may_inline_data(inode)) {
807 err = f2fs_convert_inline_inode(inode);
811 inode->i_mtime = inode->i_ctime = current_time(inode);
814 down_write(&F2FS_I(inode)->i_sem);
815 F2FS_I(inode)->last_disk_size = i_size_read(inode);
816 up_write(&F2FS_I(inode)->i_sem);
821 __setattr_copy(inode, attr);
823 if (attr->ia_valid & ATTR_MODE) {
824 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
825 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
826 inode->i_mode = F2FS_I(inode)->i_acl_mode;
827 clear_inode_flag(inode, FI_ACL_MODE);
831 /* file size may changed here */
832 f2fs_mark_inode_dirty_sync(inode, size_changed);
834 /* inode change will produce dirty node pages flushed by checkpoint */
835 f2fs_balance_fs(F2FS_I_SB(inode), true);
840 const struct inode_operations f2fs_file_inode_operations = {
841 .getattr = f2fs_getattr,
842 .setattr = f2fs_setattr,
843 .get_acl = f2fs_get_acl,
844 .set_acl = f2fs_set_acl,
845 #ifdef CONFIG_F2FS_FS_XATTR
846 .listxattr = f2fs_listxattr,
848 .fiemap = f2fs_fiemap,
851 static int fill_zero(struct inode *inode, pgoff_t index,
852 loff_t start, loff_t len)
854 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
860 f2fs_balance_fs(sbi, true);
863 page = f2fs_get_new_data_page(inode, NULL, index, false);
867 return PTR_ERR(page);
869 f2fs_wait_on_page_writeback(page, DATA, true);
870 zero_user(page, start, len);
871 set_page_dirty(page);
872 f2fs_put_page(page, 1);
876 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
880 while (pg_start < pg_end) {
881 struct dnode_of_data dn;
882 pgoff_t end_offset, count;
884 set_new_dnode(&dn, inode, NULL, NULL, 0);
885 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
887 if (err == -ENOENT) {
888 pg_start = f2fs_get_next_page_offset(&dn,
895 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
896 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
898 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
900 f2fs_truncate_data_blocks_range(&dn, count);
908 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
910 pgoff_t pg_start, pg_end;
911 loff_t off_start, off_end;
914 ret = f2fs_convert_inline_inode(inode);
918 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
919 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
921 off_start = offset & (PAGE_SIZE - 1);
922 off_end = (offset + len) & (PAGE_SIZE - 1);
924 if (pg_start == pg_end) {
925 ret = fill_zero(inode, pg_start, off_start,
926 off_end - off_start);
931 ret = fill_zero(inode, pg_start++, off_start,
932 PAGE_SIZE - off_start);
937 ret = fill_zero(inode, pg_end, 0, off_end);
942 if (pg_start < pg_end) {
943 struct address_space *mapping = inode->i_mapping;
944 loff_t blk_start, blk_end;
945 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
947 f2fs_balance_fs(sbi, true);
949 blk_start = (loff_t)pg_start << PAGE_SHIFT;
950 blk_end = (loff_t)pg_end << PAGE_SHIFT;
952 down_write(&F2FS_I(inode)->i_mmap_sem);
953 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
955 truncate_inode_pages_range(mapping, blk_start,
959 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
962 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
963 up_write(&F2FS_I(inode)->i_mmap_sem);
970 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
971 int *do_replace, pgoff_t off, pgoff_t len)
973 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
974 struct dnode_of_data dn;
978 set_new_dnode(&dn, inode, NULL, NULL, 0);
979 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
980 if (ret && ret != -ENOENT) {
982 } else if (ret == -ENOENT) {
983 if (dn.max_level == 0)
985 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
991 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
992 dn.ofs_in_node, len);
993 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
994 *blkaddr = datablock_addr(dn.inode,
995 dn.node_page, dn.ofs_in_node);
996 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
998 if (test_opt(sbi, LFS)) {
1003 /* do not invalidate this block address */
1004 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1008 f2fs_put_dnode(&dn);
1017 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1018 int *do_replace, pgoff_t off, int len)
1020 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1021 struct dnode_of_data dn;
1024 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1025 if (*do_replace == 0)
1028 set_new_dnode(&dn, inode, NULL, NULL, 0);
1029 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1031 dec_valid_block_count(sbi, inode, 1);
1032 f2fs_invalidate_blocks(sbi, *blkaddr);
1034 f2fs_update_data_blkaddr(&dn, *blkaddr);
1036 f2fs_put_dnode(&dn);
1041 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1042 block_t *blkaddr, int *do_replace,
1043 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1045 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1050 if (blkaddr[i] == NULL_ADDR && !full) {
1055 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1056 struct dnode_of_data dn;
1057 struct node_info ni;
1061 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1062 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1066 f2fs_get_node_info(sbi, dn.nid, &ni);
1067 ilen = min((pgoff_t)
1068 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1069 dn.ofs_in_node, len - i);
1071 dn.data_blkaddr = datablock_addr(dn.inode,
1072 dn.node_page, dn.ofs_in_node);
1073 f2fs_truncate_data_blocks_range(&dn, 1);
1075 if (do_replace[i]) {
1076 f2fs_i_blocks_write(src_inode,
1078 f2fs_i_blocks_write(dst_inode,
1080 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1081 blkaddr[i], ni.version, true, false);
1087 new_size = (dst + i) << PAGE_SHIFT;
1088 if (dst_inode->i_size < new_size)
1089 f2fs_i_size_write(dst_inode, new_size);
1090 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1092 f2fs_put_dnode(&dn);
1094 struct page *psrc, *pdst;
1096 psrc = f2fs_get_lock_data_page(src_inode,
1099 return PTR_ERR(psrc);
1100 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1103 f2fs_put_page(psrc, 1);
1104 return PTR_ERR(pdst);
1106 f2fs_copy_page(psrc, pdst);
1107 set_page_dirty(pdst);
1108 f2fs_put_page(pdst, 1);
1109 f2fs_put_page(psrc, 1);
1111 ret = f2fs_truncate_hole(src_inode,
1112 src + i, src + i + 1);
1121 static int __exchange_data_block(struct inode *src_inode,
1122 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1123 pgoff_t len, bool full)
1125 block_t *src_blkaddr;
1131 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1133 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1134 array_size(olen, sizeof(block_t)),
1139 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1140 array_size(olen, sizeof(int)),
1143 kvfree(src_blkaddr);
1147 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1148 do_replace, src, olen);
1152 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1153 do_replace, src, dst, olen, full);
1161 kvfree(src_blkaddr);
1167 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1168 kvfree(src_blkaddr);
1173 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1175 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1176 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1179 f2fs_balance_fs(sbi, true);
1182 f2fs_drop_extent_tree(inode);
1184 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1185 f2fs_unlock_op(sbi);
1189 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1191 pgoff_t pg_start, pg_end;
1195 if (offset + len >= i_size_read(inode))
1198 /* collapse range should be aligned to block size of f2fs. */
1199 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1202 ret = f2fs_convert_inline_inode(inode);
1206 pg_start = offset >> PAGE_SHIFT;
1207 pg_end = (offset + len) >> PAGE_SHIFT;
1209 /* avoid gc operation during block exchange */
1210 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1212 down_write(&F2FS_I(inode)->i_mmap_sem);
1213 /* write out all dirty pages from offset */
1214 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1218 truncate_pagecache(inode, offset);
1220 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1224 /* write out all moved pages, if possible */
1225 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1226 truncate_pagecache(inode, offset);
1228 new_size = i_size_read(inode) - len;
1229 truncate_pagecache(inode, new_size);
1231 ret = f2fs_truncate_blocks(inode, new_size, true);
1233 f2fs_i_size_write(inode, new_size);
1235 up_write(&F2FS_I(inode)->i_mmap_sem);
1236 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1240 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1243 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1244 pgoff_t index = start;
1245 unsigned int ofs_in_node = dn->ofs_in_node;
1249 for (; index < end; index++, dn->ofs_in_node++) {
1250 if (datablock_addr(dn->inode, dn->node_page,
1251 dn->ofs_in_node) == NULL_ADDR)
1255 dn->ofs_in_node = ofs_in_node;
1256 ret = f2fs_reserve_new_blocks(dn, count);
1260 dn->ofs_in_node = ofs_in_node;
1261 for (index = start; index < end; index++, dn->ofs_in_node++) {
1262 dn->data_blkaddr = datablock_addr(dn->inode,
1263 dn->node_page, dn->ofs_in_node);
1265 * f2fs_reserve_new_blocks will not guarantee entire block
1268 if (dn->data_blkaddr == NULL_ADDR) {
1272 if (dn->data_blkaddr != NEW_ADDR) {
1273 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1274 dn->data_blkaddr = NEW_ADDR;
1275 f2fs_set_data_blkaddr(dn);
1279 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1284 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1287 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1288 struct address_space *mapping = inode->i_mapping;
1289 pgoff_t index, pg_start, pg_end;
1290 loff_t new_size = i_size_read(inode);
1291 loff_t off_start, off_end;
1294 ret = inode_newsize_ok(inode, (len + offset));
1298 ret = f2fs_convert_inline_inode(inode);
1302 down_write(&F2FS_I(inode)->i_mmap_sem);
1303 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1307 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1308 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1310 off_start = offset & (PAGE_SIZE - 1);
1311 off_end = (offset + len) & (PAGE_SIZE - 1);
1313 if (pg_start == pg_end) {
1314 ret = fill_zero(inode, pg_start, off_start,
1315 off_end - off_start);
1319 new_size = max_t(loff_t, new_size, offset + len);
1322 ret = fill_zero(inode, pg_start++, off_start,
1323 PAGE_SIZE - off_start);
1327 new_size = max_t(loff_t, new_size,
1328 (loff_t)pg_start << PAGE_SHIFT);
1331 for (index = pg_start; index < pg_end;) {
1332 struct dnode_of_data dn;
1333 unsigned int end_offset;
1336 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1338 truncate_pagecache_range(inode,
1339 (loff_t)index << PAGE_SHIFT,
1340 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1344 set_new_dnode(&dn, inode, NULL, NULL, 0);
1345 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1347 f2fs_unlock_op(sbi);
1348 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1352 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1353 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1355 ret = f2fs_do_zero_range(&dn, index, end);
1356 f2fs_put_dnode(&dn);
1358 f2fs_unlock_op(sbi);
1359 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1361 f2fs_balance_fs(sbi, dn.node_changed);
1367 new_size = max_t(loff_t, new_size,
1368 (loff_t)index << PAGE_SHIFT);
1372 ret = fill_zero(inode, pg_end, 0, off_end);
1376 new_size = max_t(loff_t, new_size, offset + len);
1381 if (new_size > i_size_read(inode)) {
1382 if (mode & FALLOC_FL_KEEP_SIZE)
1383 file_set_keep_isize(inode);
1385 f2fs_i_size_write(inode, new_size);
1388 up_write(&F2FS_I(inode)->i_mmap_sem);
1393 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1395 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1396 pgoff_t nr, pg_start, pg_end, delta, idx;
1400 new_size = i_size_read(inode) + len;
1401 ret = inode_newsize_ok(inode, new_size);
1405 if (offset >= i_size_read(inode))
1408 /* insert range should be aligned to block size of f2fs. */
1409 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1412 ret = f2fs_convert_inline_inode(inode);
1416 f2fs_balance_fs(sbi, true);
1418 /* avoid gc operation during block exchange */
1419 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1421 down_write(&F2FS_I(inode)->i_mmap_sem);
1422 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1426 /* write out all dirty pages from offset */
1427 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1431 truncate_pagecache(inode, offset);
1433 pg_start = offset >> PAGE_SHIFT;
1434 pg_end = (offset + len) >> PAGE_SHIFT;
1435 delta = pg_end - pg_start;
1436 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1438 while (!ret && idx > pg_start) {
1439 nr = idx - pg_start;
1445 f2fs_drop_extent_tree(inode);
1447 ret = __exchange_data_block(inode, inode, idx,
1448 idx + delta, nr, false);
1449 f2fs_unlock_op(sbi);
1452 /* write out all moved pages, if possible */
1453 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1454 truncate_pagecache(inode, offset);
1457 f2fs_i_size_write(inode, new_size);
1459 up_write(&F2FS_I(inode)->i_mmap_sem);
1460 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1464 static int expand_inode_data(struct inode *inode, loff_t offset,
1465 loff_t len, int mode)
1467 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1468 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1469 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1471 loff_t new_size = i_size_read(inode);
1475 err = inode_newsize_ok(inode, (len + offset));
1479 err = f2fs_convert_inline_inode(inode);
1483 f2fs_balance_fs(sbi, true);
1485 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1486 off_end = (offset + len) & (PAGE_SIZE - 1);
1488 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1489 map.m_len = pg_end - map.m_lblk;
1493 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1500 last_off = map.m_lblk + map.m_len - 1;
1502 /* update new size to the failed position */
1503 new_size = (last_off == pg_end) ? offset + len :
1504 (loff_t)(last_off + 1) << PAGE_SHIFT;
1506 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1509 if (new_size > i_size_read(inode)) {
1510 if (mode & FALLOC_FL_KEEP_SIZE)
1511 file_set_keep_isize(inode);
1513 f2fs_i_size_write(inode, new_size);
1519 static long f2fs_fallocate(struct file *file, int mode,
1520 loff_t offset, loff_t len)
1522 struct inode *inode = file_inode(file);
1525 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1528 /* f2fs only support ->fallocate for regular file */
1529 if (!S_ISREG(inode->i_mode))
1532 if (f2fs_encrypted_inode(inode) &&
1533 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1536 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1537 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1538 FALLOC_FL_INSERT_RANGE))
1543 if (mode & FALLOC_FL_PUNCH_HOLE) {
1544 if (offset >= inode->i_size)
1547 ret = punch_hole(inode, offset, len);
1548 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1549 ret = f2fs_collapse_range(inode, offset, len);
1550 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1551 ret = f2fs_zero_range(inode, offset, len, mode);
1552 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1553 ret = f2fs_insert_range(inode, offset, len);
1555 ret = expand_inode_data(inode, offset, len, mode);
1559 inode->i_mtime = inode->i_ctime = current_time(inode);
1560 f2fs_mark_inode_dirty_sync(inode, false);
1561 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1565 inode_unlock(inode);
1567 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1571 static int f2fs_release_file(struct inode *inode, struct file *filp)
1574 * f2fs_relase_file is called at every close calls. So we should
1575 * not drop any inmemory pages by close called by other process.
1577 if (!(filp->f_mode & FMODE_WRITE) ||
1578 atomic_read(&inode->i_writecount) != 1)
1581 /* some remained atomic pages should discarded */
1582 if (f2fs_is_atomic_file(inode))
1583 f2fs_drop_inmem_pages(inode);
1584 if (f2fs_is_volatile_file(inode)) {
1585 set_inode_flag(inode, FI_DROP_CACHE);
1586 filemap_fdatawrite(inode->i_mapping);
1587 clear_inode_flag(inode, FI_DROP_CACHE);
1588 clear_inode_flag(inode, FI_VOLATILE_FILE);
1589 stat_dec_volatile_write(inode);
1594 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1596 struct inode *inode = file_inode(file);
1599 * If the process doing a transaction is crashed, we should do
1600 * roll-back. Otherwise, other reader/write can see corrupted database
1601 * until all the writers close its file. Since this should be done
1602 * before dropping file lock, it needs to do in ->flush.
1604 if (f2fs_is_atomic_file(inode) &&
1605 F2FS_I(inode)->inmem_task == current)
1606 f2fs_drop_inmem_pages(inode);
1610 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1612 struct inode *inode = file_inode(filp);
1613 struct f2fs_inode_info *fi = F2FS_I(inode);
1614 unsigned int flags = fi->i_flags;
1616 if (file_is_encrypt(inode))
1617 flags |= F2FS_ENCRYPT_FL;
1618 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1619 flags |= F2FS_INLINE_DATA_FL;
1621 flags &= F2FS_FL_USER_VISIBLE;
1623 return put_user(flags, (int __user *)arg);
1626 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1628 struct f2fs_inode_info *fi = F2FS_I(inode);
1629 unsigned int oldflags;
1631 /* Is it quota file? Do not allow user to mess with it */
1632 if (IS_NOQUOTA(inode))
1635 flags = f2fs_mask_flags(inode->i_mode, flags);
1637 oldflags = fi->i_flags;
1639 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL))
1640 if (!capable(CAP_LINUX_IMMUTABLE))
1643 flags = flags & F2FS_FL_USER_MODIFIABLE;
1644 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE;
1645 fi->i_flags = flags;
1647 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1648 set_inode_flag(inode, FI_PROJ_INHERIT);
1650 clear_inode_flag(inode, FI_PROJ_INHERIT);
1652 inode->i_ctime = current_time(inode);
1653 f2fs_set_inode_flags(inode);
1654 f2fs_mark_inode_dirty_sync(inode, false);
1658 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1660 struct inode *inode = file_inode(filp);
1664 if (!inode_owner_or_capable(inode))
1667 if (get_user(flags, (int __user *)arg))
1670 ret = mnt_want_write_file(filp);
1676 ret = __f2fs_ioc_setflags(inode, flags);
1678 inode_unlock(inode);
1679 mnt_drop_write_file(filp);
1683 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1685 struct inode *inode = file_inode(filp);
1687 return put_user(inode->i_generation, (int __user *)arg);
1690 static int f2fs_ioc_start_atomic_write(struct file *filp)
1692 struct inode *inode = file_inode(filp);
1695 if (!inode_owner_or_capable(inode))
1698 if (!S_ISREG(inode->i_mode))
1701 ret = mnt_want_write_file(filp);
1707 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1709 if (f2fs_is_atomic_file(inode))
1712 ret = f2fs_convert_inline_inode(inode);
1716 if (!get_dirty_pages(inode))
1719 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1720 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1721 inode->i_ino, get_dirty_pages(inode));
1722 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1726 set_inode_flag(inode, FI_ATOMIC_FILE);
1727 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1728 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1730 F2FS_I(inode)->inmem_task = current;
1731 stat_inc_atomic_write(inode);
1732 stat_update_max_atomic_write(inode);
1734 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1735 inode_unlock(inode);
1736 mnt_drop_write_file(filp);
1740 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1742 struct inode *inode = file_inode(filp);
1745 if (!inode_owner_or_capable(inode))
1748 ret = mnt_want_write_file(filp);
1754 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1756 if (f2fs_is_volatile_file(inode)) {
1761 if (f2fs_is_atomic_file(inode)) {
1762 ret = f2fs_commit_inmem_pages(inode);
1766 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1768 clear_inode_flag(inode, FI_ATOMIC_FILE);
1769 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1770 stat_dec_atomic_write(inode);
1773 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1776 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1777 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1780 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1781 inode_unlock(inode);
1782 mnt_drop_write_file(filp);
1786 static int f2fs_ioc_start_volatile_write(struct file *filp)
1788 struct inode *inode = file_inode(filp);
1791 if (!inode_owner_or_capable(inode))
1794 if (!S_ISREG(inode->i_mode))
1797 ret = mnt_want_write_file(filp);
1803 if (f2fs_is_volatile_file(inode))
1806 ret = f2fs_convert_inline_inode(inode);
1810 stat_inc_volatile_write(inode);
1811 stat_update_max_volatile_write(inode);
1813 set_inode_flag(inode, FI_VOLATILE_FILE);
1814 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1816 inode_unlock(inode);
1817 mnt_drop_write_file(filp);
1821 static int f2fs_ioc_release_volatile_write(struct file *filp)
1823 struct inode *inode = file_inode(filp);
1826 if (!inode_owner_or_capable(inode))
1829 ret = mnt_want_write_file(filp);
1835 if (!f2fs_is_volatile_file(inode))
1838 if (!f2fs_is_first_block_written(inode)) {
1839 ret = truncate_partial_data_page(inode, 0, true);
1843 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1845 inode_unlock(inode);
1846 mnt_drop_write_file(filp);
1850 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1852 struct inode *inode = file_inode(filp);
1855 if (!inode_owner_or_capable(inode))
1858 ret = mnt_want_write_file(filp);
1864 if (f2fs_is_atomic_file(inode))
1865 f2fs_drop_inmem_pages(inode);
1866 if (f2fs_is_volatile_file(inode)) {
1867 clear_inode_flag(inode, FI_VOLATILE_FILE);
1868 stat_dec_volatile_write(inode);
1869 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1872 inode_unlock(inode);
1874 mnt_drop_write_file(filp);
1875 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1879 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1881 struct inode *inode = file_inode(filp);
1882 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1883 struct super_block *sb = sbi->sb;
1887 if (!capable(CAP_SYS_ADMIN))
1890 if (get_user(in, (__u32 __user *)arg))
1893 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1894 ret = mnt_want_write_file(filp);
1900 case F2FS_GOING_DOWN_FULLSYNC:
1901 sb = freeze_bdev(sb->s_bdev);
1907 f2fs_stop_checkpoint(sbi, false);
1908 thaw_bdev(sb->s_bdev, sb);
1911 case F2FS_GOING_DOWN_METASYNC:
1912 /* do checkpoint only */
1913 ret = f2fs_sync_fs(sb, 1);
1916 f2fs_stop_checkpoint(sbi, false);
1918 case F2FS_GOING_DOWN_NOSYNC:
1919 f2fs_stop_checkpoint(sbi, false);
1921 case F2FS_GOING_DOWN_METAFLUSH:
1922 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1923 f2fs_stop_checkpoint(sbi, false);
1930 f2fs_stop_gc_thread(sbi);
1931 f2fs_stop_discard_thread(sbi);
1933 f2fs_drop_discard_cmd(sbi);
1934 clear_opt(sbi, DISCARD);
1936 f2fs_update_time(sbi, REQ_TIME);
1938 if (in != F2FS_GOING_DOWN_FULLSYNC)
1939 mnt_drop_write_file(filp);
1943 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1945 struct inode *inode = file_inode(filp);
1946 struct super_block *sb = inode->i_sb;
1947 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1948 struct fstrim_range range;
1951 if (!capable(CAP_SYS_ADMIN))
1954 if (!blk_queue_discard(q))
1957 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1961 ret = mnt_want_write_file(filp);
1965 range.minlen = max((unsigned int)range.minlen,
1966 q->limits.discard_granularity);
1967 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1968 mnt_drop_write_file(filp);
1972 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1975 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1979 static bool uuid_is_nonzero(__u8 u[16])
1983 for (i = 0; i < 16; i++)
1989 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1991 struct inode *inode = file_inode(filp);
1993 if (!f2fs_sb_has_encrypt(inode->i_sb))
1996 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1998 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2001 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2003 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
2005 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2008 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2010 struct inode *inode = file_inode(filp);
2011 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2014 if (!f2fs_sb_has_encrypt(inode->i_sb))
2017 err = mnt_want_write_file(filp);
2021 down_write(&sbi->sb_lock);
2023 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2026 /* update superblock with uuid */
2027 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2029 err = f2fs_commit_super(sbi, false);
2032 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2036 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2040 up_write(&sbi->sb_lock);
2041 mnt_drop_write_file(filp);
2045 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2047 struct inode *inode = file_inode(filp);
2048 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2052 if (!capable(CAP_SYS_ADMIN))
2055 if (get_user(sync, (__u32 __user *)arg))
2058 if (f2fs_readonly(sbi->sb))
2061 ret = mnt_want_write_file(filp);
2066 if (!mutex_trylock(&sbi->gc_mutex)) {
2071 mutex_lock(&sbi->gc_mutex);
2074 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2076 mnt_drop_write_file(filp);
2080 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2082 struct inode *inode = file_inode(filp);
2083 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2084 struct f2fs_gc_range range;
2088 if (!capable(CAP_SYS_ADMIN))
2091 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2095 if (f2fs_readonly(sbi->sb))
2098 end = range.start + range.len;
2099 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2103 ret = mnt_want_write_file(filp);
2109 if (!mutex_trylock(&sbi->gc_mutex)) {
2114 mutex_lock(&sbi->gc_mutex);
2117 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2118 range.start += sbi->blocks_per_seg;
2119 if (range.start <= end)
2122 mnt_drop_write_file(filp);
2126 static int f2fs_ioc_f2fs_write_checkpoint(struct file *filp, unsigned long arg)
2128 struct inode *inode = file_inode(filp);
2129 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2132 if (!capable(CAP_SYS_ADMIN))
2135 if (f2fs_readonly(sbi->sb))
2138 ret = mnt_want_write_file(filp);
2142 ret = f2fs_sync_fs(sbi->sb, 1);
2144 mnt_drop_write_file(filp);
2148 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2150 struct f2fs_defragment *range)
2152 struct inode *inode = file_inode(filp);
2153 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2154 .m_seg_type = NO_CHECK_TYPE };
2155 struct extent_info ei = {0, 0, 0};
2156 pgoff_t pg_start, pg_end, next_pgofs;
2157 unsigned int blk_per_seg = sbi->blocks_per_seg;
2158 unsigned int total = 0, sec_num;
2159 block_t blk_end = 0;
2160 bool fragmented = false;
2163 /* if in-place-update policy is enabled, don't waste time here */
2164 if (f2fs_should_update_inplace(inode, NULL))
2167 pg_start = range->start >> PAGE_SHIFT;
2168 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2170 f2fs_balance_fs(sbi, true);
2174 /* writeback all dirty pages in the range */
2175 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2176 range->start + range->len - 1);
2181 * lookup mapping info in extent cache, skip defragmenting if physical
2182 * block addresses are continuous.
2184 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2185 if (ei.fofs + ei.len >= pg_end)
2189 map.m_lblk = pg_start;
2190 map.m_next_pgofs = &next_pgofs;
2193 * lookup mapping info in dnode page cache, skip defragmenting if all
2194 * physical block addresses are continuous even if there are hole(s)
2195 * in logical blocks.
2197 while (map.m_lblk < pg_end) {
2198 map.m_len = pg_end - map.m_lblk;
2199 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2203 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2204 map.m_lblk = next_pgofs;
2208 if (blk_end && blk_end != map.m_pblk)
2211 /* record total count of block that we're going to move */
2214 blk_end = map.m_pblk + map.m_len;
2216 map.m_lblk += map.m_len;
2222 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2225 * make sure there are enough free section for LFS allocation, this can
2226 * avoid defragment running in SSR mode when free section are allocated
2229 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2234 map.m_lblk = pg_start;
2235 map.m_len = pg_end - pg_start;
2238 while (map.m_lblk < pg_end) {
2243 map.m_len = pg_end - map.m_lblk;
2244 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2248 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2249 map.m_lblk = next_pgofs;
2253 set_inode_flag(inode, FI_DO_DEFRAG);
2256 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2259 page = f2fs_get_lock_data_page(inode, idx, true);
2261 err = PTR_ERR(page);
2265 set_page_dirty(page);
2266 f2fs_put_page(page, 1);
2275 if (idx < pg_end && cnt < blk_per_seg)
2278 clear_inode_flag(inode, FI_DO_DEFRAG);
2280 err = filemap_fdatawrite(inode->i_mapping);
2285 clear_inode_flag(inode, FI_DO_DEFRAG);
2287 inode_unlock(inode);
2289 range->len = (u64)total << PAGE_SHIFT;
2293 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2295 struct inode *inode = file_inode(filp);
2296 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2297 struct f2fs_defragment range;
2300 if (!capable(CAP_SYS_ADMIN))
2303 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2306 if (f2fs_readonly(sbi->sb))
2309 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2313 /* verify alignment of offset & size */
2314 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2317 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2318 sbi->max_file_blocks))
2321 err = mnt_want_write_file(filp);
2325 err = f2fs_defragment_range(sbi, filp, &range);
2326 mnt_drop_write_file(filp);
2328 f2fs_update_time(sbi, REQ_TIME);
2332 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2339 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2340 struct file *file_out, loff_t pos_out, size_t len)
2342 struct inode *src = file_inode(file_in);
2343 struct inode *dst = file_inode(file_out);
2344 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2345 size_t olen = len, dst_max_i_size = 0;
2349 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2350 src->i_sb != dst->i_sb)
2353 if (unlikely(f2fs_readonly(src->i_sb)))
2356 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2359 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2363 if (pos_in == pos_out)
2365 if (pos_out > pos_in && pos_out < pos_in + len)
2370 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2373 if (!inode_trylock(dst))
2375 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) {
2382 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2385 olen = len = src->i_size - pos_in;
2386 if (pos_in + len == src->i_size)
2387 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2393 dst_osize = dst->i_size;
2394 if (pos_out + olen > dst->i_size)
2395 dst_max_i_size = pos_out + olen;
2397 /* verify the end result is block aligned */
2398 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2399 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2400 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2403 ret = f2fs_convert_inline_inode(src);
2407 ret = f2fs_convert_inline_inode(dst);
2411 /* write out all dirty pages from offset */
2412 ret = filemap_write_and_wait_range(src->i_mapping,
2413 pos_in, pos_in + len);
2417 ret = filemap_write_and_wait_range(dst->i_mapping,
2418 pos_out, pos_out + len);
2422 f2fs_balance_fs(sbi, true);
2424 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2425 pos_out >> F2FS_BLKSIZE_BITS,
2426 len >> F2FS_BLKSIZE_BITS, false);
2430 f2fs_i_size_write(dst, dst_max_i_size);
2431 else if (dst_osize != dst->i_size)
2432 f2fs_i_size_write(dst, dst_osize);
2434 f2fs_unlock_op(sbi);
2437 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2441 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2446 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2448 struct f2fs_move_range range;
2452 if (!(filp->f_mode & FMODE_READ) ||
2453 !(filp->f_mode & FMODE_WRITE))
2456 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2460 dst = fdget(range.dst_fd);
2464 if (!(dst.file->f_mode & FMODE_WRITE)) {
2469 err = mnt_want_write_file(filp);
2473 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2474 range.pos_out, range.len);
2476 mnt_drop_write_file(filp);
2480 if (copy_to_user((struct f2fs_move_range __user *)arg,
2481 &range, sizeof(range)))
2488 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2490 struct inode *inode = file_inode(filp);
2491 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2492 struct sit_info *sm = SIT_I(sbi);
2493 unsigned int start_segno = 0, end_segno = 0;
2494 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2495 struct f2fs_flush_device range;
2498 if (!capable(CAP_SYS_ADMIN))
2501 if (f2fs_readonly(sbi->sb))
2504 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2508 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2509 sbi->segs_per_sec != 1) {
2510 f2fs_msg(sbi->sb, KERN_WARNING,
2511 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2512 range.dev_num, sbi->s_ndevs,
2517 ret = mnt_want_write_file(filp);
2521 if (range.dev_num != 0)
2522 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2523 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2525 start_segno = sm->last_victim[FLUSH_DEVICE];
2526 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2527 start_segno = dev_start_segno;
2528 end_segno = min(start_segno + range.segments, dev_end_segno);
2530 while (start_segno < end_segno) {
2531 if (!mutex_trylock(&sbi->gc_mutex)) {
2535 sm->last_victim[GC_CB] = end_segno + 1;
2536 sm->last_victim[GC_GREEDY] = end_segno + 1;
2537 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2538 ret = f2fs_gc(sbi, true, true, start_segno);
2546 mnt_drop_write_file(filp);
2550 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2552 struct inode *inode = file_inode(filp);
2553 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2555 /* Must validate to set it with SQLite behavior in Android. */
2556 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2558 return put_user(sb_feature, (u32 __user *)arg);
2562 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2564 struct inode *inode = file_inode(filp);
2565 struct f2fs_inode_info *fi = F2FS_I(inode);
2566 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2567 struct super_block *sb = sbi->sb;
2568 struct dquot *transfer_to[MAXQUOTAS] = {};
2573 if (!f2fs_sb_has_project_quota(sb)) {
2574 if (projid != F2FS_DEF_PROJID)
2580 if (!f2fs_has_extra_attr(inode))
2583 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2585 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2588 err = mnt_want_write_file(filp);
2595 /* Is it quota file? Do not allow user to mess with it */
2596 if (IS_NOQUOTA(inode))
2599 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2600 if (IS_ERR(ipage)) {
2601 err = PTR_ERR(ipage);
2605 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2608 f2fs_put_page(ipage, 1);
2611 f2fs_put_page(ipage, 1);
2613 err = dquot_initialize(inode);
2617 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2618 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2619 err = __dquot_transfer(inode, transfer_to);
2620 dqput(transfer_to[PRJQUOTA]);
2625 F2FS_I(inode)->i_projid = kprojid;
2626 inode->i_ctime = current_time(inode);
2628 f2fs_mark_inode_dirty_sync(inode, true);
2630 inode_unlock(inode);
2631 mnt_drop_write_file(filp);
2635 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2637 if (projid != F2FS_DEF_PROJID)
2643 /* Transfer internal flags to xflags */
2644 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2648 if (iflags & F2FS_SYNC_FL)
2649 xflags |= FS_XFLAG_SYNC;
2650 if (iflags & F2FS_IMMUTABLE_FL)
2651 xflags |= FS_XFLAG_IMMUTABLE;
2652 if (iflags & F2FS_APPEND_FL)
2653 xflags |= FS_XFLAG_APPEND;
2654 if (iflags & F2FS_NODUMP_FL)
2655 xflags |= FS_XFLAG_NODUMP;
2656 if (iflags & F2FS_NOATIME_FL)
2657 xflags |= FS_XFLAG_NOATIME;
2658 if (iflags & F2FS_PROJINHERIT_FL)
2659 xflags |= FS_XFLAG_PROJINHERIT;
2663 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2664 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2665 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2667 /* Transfer xflags flags to internal */
2668 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2670 unsigned long iflags = 0;
2672 if (xflags & FS_XFLAG_SYNC)
2673 iflags |= F2FS_SYNC_FL;
2674 if (xflags & FS_XFLAG_IMMUTABLE)
2675 iflags |= F2FS_IMMUTABLE_FL;
2676 if (xflags & FS_XFLAG_APPEND)
2677 iflags |= F2FS_APPEND_FL;
2678 if (xflags & FS_XFLAG_NODUMP)
2679 iflags |= F2FS_NODUMP_FL;
2680 if (xflags & FS_XFLAG_NOATIME)
2681 iflags |= F2FS_NOATIME_FL;
2682 if (xflags & FS_XFLAG_PROJINHERIT)
2683 iflags |= F2FS_PROJINHERIT_FL;
2688 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2690 struct inode *inode = file_inode(filp);
2691 struct f2fs_inode_info *fi = F2FS_I(inode);
2694 memset(&fa, 0, sizeof(struct fsxattr));
2695 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2696 F2FS_FL_USER_VISIBLE);
2698 if (f2fs_sb_has_project_quota(inode->i_sb))
2699 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2702 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2707 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2709 struct inode *inode = file_inode(filp);
2710 struct f2fs_inode_info *fi = F2FS_I(inode);
2715 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2718 /* Make sure caller has proper permission */
2719 if (!inode_owner_or_capable(inode))
2722 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2725 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2726 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2729 err = mnt_want_write_file(filp);
2734 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2735 (flags & F2FS_FL_XFLAG_VISIBLE);
2736 err = __f2fs_ioc_setflags(inode, flags);
2737 inode_unlock(inode);
2738 mnt_drop_write_file(filp);
2742 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2749 int f2fs_pin_file_control(struct inode *inode, bool inc)
2751 struct f2fs_inode_info *fi = F2FS_I(inode);
2752 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2754 /* Use i_gc_failures for normal file as a risk signal. */
2756 f2fs_i_gc_failures_write(inode,
2757 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2759 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2760 f2fs_msg(sbi->sb, KERN_WARNING,
2761 "%s: Enable GC = ino %lx after %x GC trials\n",
2762 __func__, inode->i_ino,
2763 fi->i_gc_failures[GC_FAILURE_PIN]);
2764 clear_inode_flag(inode, FI_PIN_FILE);
2770 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2772 struct inode *inode = file_inode(filp);
2776 if (!inode_owner_or_capable(inode))
2779 if (get_user(pin, (__u32 __user *)arg))
2782 if (!S_ISREG(inode->i_mode))
2785 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2788 ret = mnt_want_write_file(filp);
2794 if (f2fs_should_update_outplace(inode, NULL)) {
2800 clear_inode_flag(inode, FI_PIN_FILE);
2801 F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN] = 1;
2805 if (f2fs_pin_file_control(inode, false)) {
2809 ret = f2fs_convert_inline_inode(inode);
2813 set_inode_flag(inode, FI_PIN_FILE);
2814 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2816 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2818 inode_unlock(inode);
2819 mnt_drop_write_file(filp);
2823 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2825 struct inode *inode = file_inode(filp);
2828 if (is_inode_flag_set(inode, FI_PIN_FILE))
2829 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2830 return put_user(pin, (u32 __user *)arg);
2833 int f2fs_precache_extents(struct inode *inode)
2835 struct f2fs_inode_info *fi = F2FS_I(inode);
2836 struct f2fs_map_blocks map;
2837 pgoff_t m_next_extent;
2841 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2845 map.m_next_pgofs = NULL;
2846 map.m_next_extent = &m_next_extent;
2847 map.m_seg_type = NO_CHECK_TYPE;
2848 end = F2FS_I_SB(inode)->max_file_blocks;
2850 while (map.m_lblk < end) {
2851 map.m_len = end - map.m_lblk;
2853 down_write(&fi->i_gc_rwsem[WRITE]);
2854 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2855 up_write(&fi->i_gc_rwsem[WRITE]);
2859 map.m_lblk = m_next_extent;
2865 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2867 return f2fs_precache_extents(file_inode(filp));
2870 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2872 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2876 case F2FS_IOC_GETFLAGS:
2877 return f2fs_ioc_getflags(filp, arg);
2878 case F2FS_IOC_SETFLAGS:
2879 return f2fs_ioc_setflags(filp, arg);
2880 case F2FS_IOC_GETVERSION:
2881 return f2fs_ioc_getversion(filp, arg);
2882 case F2FS_IOC_START_ATOMIC_WRITE:
2883 return f2fs_ioc_start_atomic_write(filp);
2884 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2885 return f2fs_ioc_commit_atomic_write(filp);
2886 case F2FS_IOC_START_VOLATILE_WRITE:
2887 return f2fs_ioc_start_volatile_write(filp);
2888 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2889 return f2fs_ioc_release_volatile_write(filp);
2890 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2891 return f2fs_ioc_abort_volatile_write(filp);
2892 case F2FS_IOC_SHUTDOWN:
2893 return f2fs_ioc_shutdown(filp, arg);
2895 return f2fs_ioc_fitrim(filp, arg);
2896 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2897 return f2fs_ioc_set_encryption_policy(filp, arg);
2898 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2899 return f2fs_ioc_get_encryption_policy(filp, arg);
2900 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2901 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2902 case F2FS_IOC_GARBAGE_COLLECT:
2903 return f2fs_ioc_gc(filp, arg);
2904 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2905 return f2fs_ioc_gc_range(filp, arg);
2906 case F2FS_IOC_WRITE_CHECKPOINT:
2907 return f2fs_ioc_f2fs_write_checkpoint(filp, arg);
2908 case F2FS_IOC_DEFRAGMENT:
2909 return f2fs_ioc_defragment(filp, arg);
2910 case F2FS_IOC_MOVE_RANGE:
2911 return f2fs_ioc_move_range(filp, arg);
2912 case F2FS_IOC_FLUSH_DEVICE:
2913 return f2fs_ioc_flush_device(filp, arg);
2914 case F2FS_IOC_GET_FEATURES:
2915 return f2fs_ioc_get_features(filp, arg);
2916 case F2FS_IOC_FSGETXATTR:
2917 return f2fs_ioc_fsgetxattr(filp, arg);
2918 case F2FS_IOC_FSSETXATTR:
2919 return f2fs_ioc_fssetxattr(filp, arg);
2920 case F2FS_IOC_GET_PIN_FILE:
2921 return f2fs_ioc_get_pin_file(filp, arg);
2922 case F2FS_IOC_SET_PIN_FILE:
2923 return f2fs_ioc_set_pin_file(filp, arg);
2924 case F2FS_IOC_PRECACHE_EXTENTS:
2925 return f2fs_ioc_precache_extents(filp, arg);
2931 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2933 struct file *file = iocb->ki_filp;
2934 struct inode *inode = file_inode(file);
2937 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2940 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
2943 if (!inode_trylock(inode)) {
2944 if (iocb->ki_flags & IOCB_NOWAIT)
2949 ret = generic_write_checks(iocb, from);
2951 bool preallocated = false;
2952 size_t target_size = 0;
2955 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2956 set_inode_flag(inode, FI_NO_PREALLOC);
2958 if ((iocb->ki_flags & IOCB_NOWAIT) &&
2959 (iocb->ki_flags & IOCB_DIRECT)) {
2960 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
2961 iov_iter_count(from)) ||
2962 f2fs_has_inline_data(inode) ||
2963 f2fs_force_buffered_io(inode, WRITE)) {
2964 clear_inode_flag(inode,
2966 inode_unlock(inode);
2971 preallocated = true;
2972 target_size = iocb->ki_pos + iov_iter_count(from);
2974 err = f2fs_preallocate_blocks(iocb, from);
2976 clear_inode_flag(inode, FI_NO_PREALLOC);
2977 inode_unlock(inode);
2981 ret = __generic_file_write_iter(iocb, from);
2982 clear_inode_flag(inode, FI_NO_PREALLOC);
2984 /* if we couldn't write data, we should deallocate blocks. */
2985 if (preallocated && i_size_read(inode) < target_size)
2986 f2fs_truncate(inode);
2989 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2991 inode_unlock(inode);
2994 ret = generic_write_sync(iocb, ret);
2998 #ifdef CONFIG_COMPAT
2999 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3002 case F2FS_IOC32_GETFLAGS:
3003 cmd = F2FS_IOC_GETFLAGS;
3005 case F2FS_IOC32_SETFLAGS:
3006 cmd = F2FS_IOC_SETFLAGS;
3008 case F2FS_IOC32_GETVERSION:
3009 cmd = F2FS_IOC_GETVERSION;
3011 case F2FS_IOC_START_ATOMIC_WRITE:
3012 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3013 case F2FS_IOC_START_VOLATILE_WRITE:
3014 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3015 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3016 case F2FS_IOC_SHUTDOWN:
3017 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3018 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3019 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3020 case F2FS_IOC_GARBAGE_COLLECT:
3021 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3022 case F2FS_IOC_WRITE_CHECKPOINT:
3023 case F2FS_IOC_DEFRAGMENT:
3024 case F2FS_IOC_MOVE_RANGE:
3025 case F2FS_IOC_FLUSH_DEVICE:
3026 case F2FS_IOC_GET_FEATURES:
3027 case F2FS_IOC_FSGETXATTR:
3028 case F2FS_IOC_FSSETXATTR:
3029 case F2FS_IOC_GET_PIN_FILE:
3030 case F2FS_IOC_SET_PIN_FILE:
3031 case F2FS_IOC_PRECACHE_EXTENTS:
3034 return -ENOIOCTLCMD;
3036 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3040 const struct file_operations f2fs_file_operations = {
3041 .llseek = f2fs_llseek,
3042 .read_iter = generic_file_read_iter,
3043 .write_iter = f2fs_file_write_iter,
3044 .open = f2fs_file_open,
3045 .release = f2fs_release_file,
3046 .mmap = f2fs_file_mmap,
3047 .flush = f2fs_file_flush,
3048 .fsync = f2fs_sync_file,
3049 .fallocate = f2fs_fallocate,
3050 .unlocked_ioctl = f2fs_ioctl,
3051 #ifdef CONFIG_COMPAT
3052 .compat_ioctl = f2fs_compat_ioctl,
3054 .splice_read = generic_file_splice_read,
3055 .splice_write = iter_file_splice_write,