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>
23 #include <linux/nls.h>
24 #include <linux/sched/signal.h>
25 #include <linux/fileattr.h>
26 #include <linux/fadvise.h>
27 #include <linux/iomap.h>
36 #include <trace/events/f2fs.h>
37 #include <uapi/linux/f2fs.h>
39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
41 struct inode *inode = file_inode(vmf->vma->vm_file);
44 ret = filemap_fault(vmf);
46 f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
49 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
56 struct page *page = vmf->page;
57 struct inode *inode = file_inode(vmf->vma->vm_file);
58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
59 struct dnode_of_data dn;
60 bool need_alloc = true;
63 if (unlikely(IS_IMMUTABLE(inode)))
64 return VM_FAULT_SIGBUS;
66 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
67 return VM_FAULT_SIGBUS;
69 if (unlikely(f2fs_cp_error(sbi))) {
74 if (!f2fs_is_checkpoint_ready(sbi)) {
79 err = f2fs_convert_inline_inode(inode);
83 #ifdef CONFIG_F2FS_FS_COMPRESSION
84 if (f2fs_compressed_file(inode)) {
85 int ret = f2fs_is_compressed_cluster(inode, page->index);
95 /* should do out of any locked page */
97 f2fs_balance_fs(sbi, true);
99 sb_start_pagefault(inode->i_sb);
101 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
103 file_update_time(vmf->vma->vm_file);
104 filemap_invalidate_lock_shared(inode->i_mapping);
106 if (unlikely(page->mapping != inode->i_mapping ||
107 page_offset(page) > i_size_read(inode) ||
108 !PageUptodate(page))) {
115 /* block allocation */
116 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
117 set_new_dnode(&dn, inode, NULL, NULL, 0);
118 err = f2fs_get_block(&dn, page->index);
119 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
122 #ifdef CONFIG_F2FS_FS_COMPRESSION
124 set_new_dnode(&dn, inode, NULL, NULL, 0);
125 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
134 f2fs_wait_on_page_writeback(page, DATA, false, true);
136 /* wait for GCed page writeback via META_MAPPING */
137 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
140 * check to see if the page is mapped already (no holes)
142 if (PageMappedToDisk(page))
145 /* page is wholly or partially inside EOF */
146 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
147 i_size_read(inode)) {
150 offset = i_size_read(inode) & ~PAGE_MASK;
151 zero_user_segment(page, offset, PAGE_SIZE);
153 set_page_dirty(page);
154 if (!PageUptodate(page))
155 SetPageUptodate(page);
157 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
158 f2fs_update_time(sbi, REQ_TIME);
160 trace_f2fs_vm_page_mkwrite(page, DATA);
162 filemap_invalidate_unlock_shared(inode->i_mapping);
164 sb_end_pagefault(inode->i_sb);
166 return block_page_mkwrite_return(err);
169 static const struct vm_operations_struct f2fs_file_vm_ops = {
170 .fault = f2fs_filemap_fault,
171 .map_pages = filemap_map_pages,
172 .page_mkwrite = f2fs_vm_page_mkwrite,
175 static int get_parent_ino(struct inode *inode, nid_t *pino)
177 struct dentry *dentry;
180 * Make sure to get the non-deleted alias. The alias associated with
181 * the open file descriptor being fsync()'ed may be deleted already.
183 dentry = d_find_alias(inode);
187 *pino = parent_ino(dentry);
192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
195 enum cp_reason_type cp_reason = CP_NO_NEEDED;
197 if (!S_ISREG(inode->i_mode))
198 cp_reason = CP_NON_REGULAR;
199 else if (f2fs_compressed_file(inode))
200 cp_reason = CP_COMPRESSED;
201 else if (inode->i_nlink != 1)
202 cp_reason = CP_HARDLINK;
203 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
204 cp_reason = CP_SB_NEED_CP;
205 else if (file_wrong_pino(inode))
206 cp_reason = CP_WRONG_PINO;
207 else if (!f2fs_space_for_roll_forward(sbi))
208 cp_reason = CP_NO_SPC_ROLL;
209 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
210 cp_reason = CP_NODE_NEED_CP;
211 else if (test_opt(sbi, FASTBOOT))
212 cp_reason = CP_FASTBOOT_MODE;
213 else if (F2FS_OPTION(sbi).active_logs == 2)
214 cp_reason = CP_SPEC_LOG_NUM;
215 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
216 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
217 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
219 cp_reason = CP_RECOVER_DIR;
224 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
226 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
228 /* But we need to avoid that there are some inode updates */
229 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
235 static void try_to_fix_pino(struct inode *inode)
237 struct f2fs_inode_info *fi = F2FS_I(inode);
240 f2fs_down_write(&fi->i_sem);
241 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
242 get_parent_ino(inode, &pino)) {
243 f2fs_i_pino_write(inode, pino);
244 file_got_pino(inode);
246 f2fs_up_write(&fi->i_sem);
249 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
250 int datasync, bool atomic)
252 struct inode *inode = file->f_mapping->host;
253 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
254 nid_t ino = inode->i_ino;
256 enum cp_reason_type cp_reason = 0;
257 struct writeback_control wbc = {
258 .sync_mode = WB_SYNC_ALL,
259 .nr_to_write = LONG_MAX,
262 unsigned int seq_id = 0;
264 if (unlikely(f2fs_readonly(inode->i_sb)))
267 trace_f2fs_sync_file_enter(inode);
269 if (S_ISDIR(inode->i_mode))
272 /* if fdatasync is triggered, let's do in-place-update */
273 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
274 set_inode_flag(inode, FI_NEED_IPU);
275 ret = file_write_and_wait_range(file, start, end);
276 clear_inode_flag(inode, FI_NEED_IPU);
278 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
279 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
283 /* if the inode is dirty, let's recover all the time */
284 if (!f2fs_skip_inode_update(inode, datasync)) {
285 f2fs_write_inode(inode, NULL);
290 * if there is no written data, don't waste time to write recovery info.
292 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
293 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
295 /* it may call write_inode just prior to fsync */
296 if (need_inode_page_update(sbi, ino))
299 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
300 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
305 * for OPU case, during fsync(), node can be persisted before
306 * data when lower device doesn't support write barrier, result
307 * in data corruption after SPO.
308 * So for strict fsync mode, force to use atomic write sematics
309 * to keep write order in between data/node and last node to
310 * avoid potential data corruption.
312 if (F2FS_OPTION(sbi).fsync_mode ==
313 FSYNC_MODE_STRICT && !atomic)
318 * Both of fdatasync() and fsync() are able to be recovered from
321 f2fs_down_read(&F2FS_I(inode)->i_sem);
322 cp_reason = need_do_checkpoint(inode);
323 f2fs_up_read(&F2FS_I(inode)->i_sem);
326 /* all the dirty node pages should be flushed for POR */
327 ret = f2fs_sync_fs(inode->i_sb, 1);
330 * We've secured consistency through sync_fs. Following pino
331 * will be used only for fsynced inodes after checkpoint.
333 try_to_fix_pino(inode);
334 clear_inode_flag(inode, FI_APPEND_WRITE);
335 clear_inode_flag(inode, FI_UPDATE_WRITE);
339 atomic_inc(&sbi->wb_sync_req[NODE]);
340 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
341 atomic_dec(&sbi->wb_sync_req[NODE]);
345 /* if cp_error was enabled, we should avoid infinite loop */
346 if (unlikely(f2fs_cp_error(sbi))) {
351 if (f2fs_need_inode_block_update(sbi, ino)) {
352 f2fs_mark_inode_dirty_sync(inode, true);
353 f2fs_write_inode(inode, NULL);
358 * If it's atomic_write, it's just fine to keep write ordering. So
359 * here we don't need to wait for node write completion, since we use
360 * node chain which serializes node blocks. If one of node writes are
361 * reordered, we can see simply broken chain, resulting in stopping
362 * roll-forward recovery. It means we'll recover all or none node blocks
366 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
371 /* once recovery info is written, don't need to tack this */
372 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
373 clear_inode_flag(inode, FI_APPEND_WRITE);
375 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
376 ret = f2fs_issue_flush(sbi, inode->i_ino);
378 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
379 clear_inode_flag(inode, FI_UPDATE_WRITE);
380 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
382 f2fs_update_time(sbi, REQ_TIME);
384 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
388 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
390 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
392 return f2fs_do_sync_file(file, start, end, datasync, false);
395 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
396 pgoff_t index, int whence)
400 if (__is_valid_data_blkaddr(blkaddr))
402 if (blkaddr == NEW_ADDR &&
403 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
407 if (blkaddr == NULL_ADDR)
414 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
416 struct inode *inode = file->f_mapping->host;
417 loff_t maxbytes = inode->i_sb->s_maxbytes;
418 struct dnode_of_data dn;
419 pgoff_t pgofs, end_offset;
420 loff_t data_ofs = offset;
426 isize = i_size_read(inode);
430 /* handle inline data case */
431 if (f2fs_has_inline_data(inode)) {
432 if (whence == SEEK_HOLE) {
435 } else if (whence == SEEK_DATA) {
441 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
443 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
444 set_new_dnode(&dn, inode, NULL, NULL, 0);
445 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
446 if (err && err != -ENOENT) {
448 } else if (err == -ENOENT) {
449 /* direct node does not exists */
450 if (whence == SEEK_DATA) {
451 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
458 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
460 /* find data/hole in dnode block */
461 for (; dn.ofs_in_node < end_offset;
462 dn.ofs_in_node++, pgofs++,
463 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
466 blkaddr = f2fs_data_blkaddr(&dn);
468 if (__is_valid_data_blkaddr(blkaddr) &&
469 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
470 blkaddr, DATA_GENERIC_ENHANCE)) {
475 if (__found_offset(file->f_mapping, blkaddr,
484 if (whence == SEEK_DATA)
487 if (whence == SEEK_HOLE && data_ofs > isize)
490 return vfs_setpos(file, data_ofs, maxbytes);
496 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
498 struct inode *inode = file->f_mapping->host;
499 loff_t maxbytes = inode->i_sb->s_maxbytes;
501 if (f2fs_compressed_file(inode))
502 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
508 return generic_file_llseek_size(file, offset, whence,
509 maxbytes, i_size_read(inode));
514 return f2fs_seek_block(file, offset, whence);
520 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
522 struct inode *inode = file_inode(file);
524 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
527 if (!f2fs_is_compress_backend_ready(inode))
531 vma->vm_ops = &f2fs_file_vm_ops;
532 set_inode_flag(inode, FI_MMAP_FILE);
536 static int f2fs_file_open(struct inode *inode, struct file *filp)
538 int err = fscrypt_file_open(inode, filp);
543 if (!f2fs_is_compress_backend_ready(inode))
546 err = fsverity_file_open(inode, filp);
550 filp->f_mode |= FMODE_NOWAIT;
552 return dquot_file_open(inode, filp);
555 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
557 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
558 struct f2fs_node *raw_node;
559 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
562 bool compressed_cluster = false;
563 int cluster_index = 0, valid_blocks = 0;
564 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
565 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
567 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
568 base = get_extra_isize(dn->inode);
570 raw_node = F2FS_NODE(dn->node_page);
571 addr = blkaddr_in_node(raw_node) + base + ofs;
573 /* Assumption: truncateion starts with cluster */
574 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
575 block_t blkaddr = le32_to_cpu(*addr);
577 if (f2fs_compressed_file(dn->inode) &&
578 !(cluster_index & (cluster_size - 1))) {
579 if (compressed_cluster)
580 f2fs_i_compr_blocks_update(dn->inode,
581 valid_blocks, false);
582 compressed_cluster = (blkaddr == COMPRESS_ADDR);
586 if (blkaddr == NULL_ADDR)
589 dn->data_blkaddr = NULL_ADDR;
590 f2fs_set_data_blkaddr(dn);
592 if (__is_valid_data_blkaddr(blkaddr)) {
593 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
594 DATA_GENERIC_ENHANCE))
596 if (compressed_cluster)
600 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
601 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
603 f2fs_invalidate_blocks(sbi, blkaddr);
605 if (!released || blkaddr != COMPRESS_ADDR)
609 if (compressed_cluster)
610 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
615 * once we invalidate valid blkaddr in range [ofs, ofs + count],
616 * we will invalidate all blkaddr in the whole range.
618 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
620 f2fs_update_extent_cache_range(dn, fofs, 0, len);
621 dec_valid_block_count(sbi, dn->inode, nr_free);
623 dn->ofs_in_node = ofs;
625 f2fs_update_time(sbi, REQ_TIME);
626 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
627 dn->ofs_in_node, nr_free);
630 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
632 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
635 static int truncate_partial_data_page(struct inode *inode, u64 from,
638 loff_t offset = from & (PAGE_SIZE - 1);
639 pgoff_t index = from >> PAGE_SHIFT;
640 struct address_space *mapping = inode->i_mapping;
643 if (!offset && !cache_only)
647 page = find_lock_page(mapping, index);
648 if (page && PageUptodate(page))
650 f2fs_put_page(page, 1);
654 page = f2fs_get_lock_data_page(inode, index, true);
656 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
658 f2fs_wait_on_page_writeback(page, DATA, true, true);
659 zero_user(page, offset, PAGE_SIZE - offset);
661 /* An encrypted inode should have a key and truncate the last page. */
662 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
664 set_page_dirty(page);
665 f2fs_put_page(page, 1);
669 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
671 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
672 struct dnode_of_data dn;
674 int count = 0, err = 0;
676 bool truncate_page = false;
678 trace_f2fs_truncate_blocks_enter(inode, from);
680 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
682 if (free_from >= max_file_blocks(inode))
688 ipage = f2fs_get_node_page(sbi, inode->i_ino);
690 err = PTR_ERR(ipage);
694 if (f2fs_has_inline_data(inode)) {
695 f2fs_truncate_inline_inode(inode, ipage, from);
696 f2fs_put_page(ipage, 1);
697 truncate_page = true;
701 set_new_dnode(&dn, inode, ipage, NULL, 0);
702 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
709 count = ADDRS_PER_PAGE(dn.node_page, inode);
711 count -= dn.ofs_in_node;
712 f2fs_bug_on(sbi, count < 0);
714 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
715 f2fs_truncate_data_blocks_range(&dn, count);
721 err = f2fs_truncate_inode_blocks(inode, free_from);
726 /* lastly zero out the first data page */
728 err = truncate_partial_data_page(inode, from, truncate_page);
730 trace_f2fs_truncate_blocks_exit(inode, err);
734 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
736 u64 free_from = from;
739 #ifdef CONFIG_F2FS_FS_COMPRESSION
741 * for compressed file, only support cluster size
742 * aligned truncation.
744 if (f2fs_compressed_file(inode))
745 free_from = round_up(from,
746 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
749 err = f2fs_do_truncate_blocks(inode, free_from, lock);
753 #ifdef CONFIG_F2FS_FS_COMPRESSION
755 * For compressed file, after release compress blocks, don't allow write
756 * direct, but we should allow write direct after truncate to zero.
758 if (f2fs_compressed_file(inode) && !free_from
759 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
760 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
762 if (from != free_from) {
763 err = f2fs_truncate_partial_cluster(inode, from, lock);
772 int f2fs_truncate(struct inode *inode)
776 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
779 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
780 S_ISLNK(inode->i_mode)))
783 trace_f2fs_truncate(inode);
785 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
786 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
790 err = f2fs_dquot_initialize(inode);
794 /* we should check inline_data size */
795 if (!f2fs_may_inline_data(inode)) {
796 err = f2fs_convert_inline_inode(inode);
801 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
805 inode->i_mtime = inode->i_ctime = current_time(inode);
806 f2fs_mark_inode_dirty_sync(inode, false);
810 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
811 struct kstat *stat, u32 request_mask, unsigned int query_flags)
813 struct inode *inode = d_inode(path->dentry);
814 struct f2fs_inode_info *fi = F2FS_I(inode);
815 struct f2fs_inode *ri = NULL;
818 if (f2fs_has_extra_attr(inode) &&
819 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
820 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
821 stat->result_mask |= STATX_BTIME;
822 stat->btime.tv_sec = fi->i_crtime.tv_sec;
823 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
827 if (flags & F2FS_COMPR_FL)
828 stat->attributes |= STATX_ATTR_COMPRESSED;
829 if (flags & F2FS_APPEND_FL)
830 stat->attributes |= STATX_ATTR_APPEND;
831 if (IS_ENCRYPTED(inode))
832 stat->attributes |= STATX_ATTR_ENCRYPTED;
833 if (flags & F2FS_IMMUTABLE_FL)
834 stat->attributes |= STATX_ATTR_IMMUTABLE;
835 if (flags & F2FS_NODUMP_FL)
836 stat->attributes |= STATX_ATTR_NODUMP;
837 if (IS_VERITY(inode))
838 stat->attributes |= STATX_ATTR_VERITY;
840 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
842 STATX_ATTR_ENCRYPTED |
843 STATX_ATTR_IMMUTABLE |
847 generic_fillattr(mnt_userns, inode, stat);
849 /* we need to show initial sectors used for inline_data/dentries */
850 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
851 f2fs_has_inline_dentry(inode))
852 stat->blocks += (stat->size + 511) >> 9;
857 #ifdef CONFIG_F2FS_FS_POSIX_ACL
858 static void __setattr_copy(struct user_namespace *mnt_userns,
859 struct inode *inode, const struct iattr *attr)
861 unsigned int ia_valid = attr->ia_valid;
863 if (ia_valid & ATTR_UID)
864 inode->i_uid = attr->ia_uid;
865 if (ia_valid & ATTR_GID)
866 inode->i_gid = attr->ia_gid;
867 if (ia_valid & ATTR_ATIME)
868 inode->i_atime = attr->ia_atime;
869 if (ia_valid & ATTR_MTIME)
870 inode->i_mtime = attr->ia_mtime;
871 if (ia_valid & ATTR_CTIME)
872 inode->i_ctime = attr->ia_ctime;
873 if (ia_valid & ATTR_MODE) {
874 umode_t mode = attr->ia_mode;
875 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
877 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
879 set_acl_inode(inode, mode);
883 #define __setattr_copy setattr_copy
886 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
889 struct inode *inode = d_inode(dentry);
892 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
895 if (unlikely(IS_IMMUTABLE(inode)))
898 if (unlikely(IS_APPEND(inode) &&
899 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
900 ATTR_GID | ATTR_TIMES_SET))))
903 if ((attr->ia_valid & ATTR_SIZE) &&
904 !f2fs_is_compress_backend_ready(inode))
907 err = setattr_prepare(mnt_userns, dentry, attr);
911 err = fscrypt_prepare_setattr(dentry, attr);
915 err = fsverity_prepare_setattr(dentry, attr);
919 if (is_quota_modification(inode, attr)) {
920 err = f2fs_dquot_initialize(inode);
924 if ((attr->ia_valid & ATTR_UID &&
925 !uid_eq(attr->ia_uid, inode->i_uid)) ||
926 (attr->ia_valid & ATTR_GID &&
927 !gid_eq(attr->ia_gid, inode->i_gid))) {
928 f2fs_lock_op(F2FS_I_SB(inode));
929 err = dquot_transfer(inode, attr);
931 set_sbi_flag(F2FS_I_SB(inode),
932 SBI_QUOTA_NEED_REPAIR);
933 f2fs_unlock_op(F2FS_I_SB(inode));
937 * update uid/gid under lock_op(), so that dquot and inode can
938 * be updated atomically.
940 if (attr->ia_valid & ATTR_UID)
941 inode->i_uid = attr->ia_uid;
942 if (attr->ia_valid & ATTR_GID)
943 inode->i_gid = attr->ia_gid;
944 f2fs_mark_inode_dirty_sync(inode, true);
945 f2fs_unlock_op(F2FS_I_SB(inode));
948 if (attr->ia_valid & ATTR_SIZE) {
949 loff_t old_size = i_size_read(inode);
951 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
953 * should convert inline inode before i_size_write to
954 * keep smaller than inline_data size with inline flag.
956 err = f2fs_convert_inline_inode(inode);
961 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
962 filemap_invalidate_lock(inode->i_mapping);
964 truncate_setsize(inode, attr->ia_size);
966 if (attr->ia_size <= old_size)
967 err = f2fs_truncate(inode);
969 * do not trim all blocks after i_size if target size is
970 * larger than i_size.
972 filemap_invalidate_unlock(inode->i_mapping);
973 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
977 spin_lock(&F2FS_I(inode)->i_size_lock);
978 inode->i_mtime = inode->i_ctime = current_time(inode);
979 F2FS_I(inode)->last_disk_size = i_size_read(inode);
980 spin_unlock(&F2FS_I(inode)->i_size_lock);
983 __setattr_copy(mnt_userns, inode, attr);
985 if (attr->ia_valid & ATTR_MODE) {
986 err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
988 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
990 inode->i_mode = F2FS_I(inode)->i_acl_mode;
991 clear_inode_flag(inode, FI_ACL_MODE);
995 /* file size may changed here */
996 f2fs_mark_inode_dirty_sync(inode, true);
998 /* inode change will produce dirty node pages flushed by checkpoint */
999 f2fs_balance_fs(F2FS_I_SB(inode), true);
1004 const struct inode_operations f2fs_file_inode_operations = {
1005 .getattr = f2fs_getattr,
1006 .setattr = f2fs_setattr,
1007 .get_acl = f2fs_get_acl,
1008 .set_acl = f2fs_set_acl,
1009 .listxattr = f2fs_listxattr,
1010 .fiemap = f2fs_fiemap,
1011 .fileattr_get = f2fs_fileattr_get,
1012 .fileattr_set = f2fs_fileattr_set,
1015 static int fill_zero(struct inode *inode, pgoff_t index,
1016 loff_t start, loff_t len)
1018 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1024 f2fs_balance_fs(sbi, true);
1027 page = f2fs_get_new_data_page(inode, NULL, index, false);
1028 f2fs_unlock_op(sbi);
1031 return PTR_ERR(page);
1033 f2fs_wait_on_page_writeback(page, DATA, true, true);
1034 zero_user(page, start, len);
1035 set_page_dirty(page);
1036 f2fs_put_page(page, 1);
1040 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1044 while (pg_start < pg_end) {
1045 struct dnode_of_data dn;
1046 pgoff_t end_offset, count;
1048 set_new_dnode(&dn, inode, NULL, NULL, 0);
1049 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1051 if (err == -ENOENT) {
1052 pg_start = f2fs_get_next_page_offset(&dn,
1059 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1060 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1062 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1064 f2fs_truncate_data_blocks_range(&dn, count);
1065 f2fs_put_dnode(&dn);
1072 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1074 pgoff_t pg_start, pg_end;
1075 loff_t off_start, off_end;
1078 ret = f2fs_convert_inline_inode(inode);
1082 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1083 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1085 off_start = offset & (PAGE_SIZE - 1);
1086 off_end = (offset + len) & (PAGE_SIZE - 1);
1088 if (pg_start == pg_end) {
1089 ret = fill_zero(inode, pg_start, off_start,
1090 off_end - off_start);
1095 ret = fill_zero(inode, pg_start++, off_start,
1096 PAGE_SIZE - off_start);
1101 ret = fill_zero(inode, pg_end, 0, off_end);
1106 if (pg_start < pg_end) {
1107 loff_t blk_start, blk_end;
1108 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1110 f2fs_balance_fs(sbi, true);
1112 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1113 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1115 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1116 filemap_invalidate_lock(inode->i_mapping);
1118 truncate_pagecache_range(inode, blk_start, blk_end - 1);
1121 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1122 f2fs_unlock_op(sbi);
1124 filemap_invalidate_unlock(inode->i_mapping);
1125 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1132 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1133 int *do_replace, pgoff_t off, pgoff_t len)
1135 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1136 struct dnode_of_data dn;
1140 set_new_dnode(&dn, inode, NULL, NULL, 0);
1141 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1142 if (ret && ret != -ENOENT) {
1144 } else if (ret == -ENOENT) {
1145 if (dn.max_level == 0)
1147 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1148 dn.ofs_in_node, len);
1154 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1155 dn.ofs_in_node, len);
1156 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1157 *blkaddr = f2fs_data_blkaddr(&dn);
1159 if (__is_valid_data_blkaddr(*blkaddr) &&
1160 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1161 DATA_GENERIC_ENHANCE)) {
1162 f2fs_put_dnode(&dn);
1163 return -EFSCORRUPTED;
1166 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1168 if (f2fs_lfs_mode(sbi)) {
1169 f2fs_put_dnode(&dn);
1173 /* do not invalidate this block address */
1174 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1178 f2fs_put_dnode(&dn);
1187 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1188 int *do_replace, pgoff_t off, int len)
1190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1191 struct dnode_of_data dn;
1194 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1195 if (*do_replace == 0)
1198 set_new_dnode(&dn, inode, NULL, NULL, 0);
1199 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1201 dec_valid_block_count(sbi, inode, 1);
1202 f2fs_invalidate_blocks(sbi, *blkaddr);
1204 f2fs_update_data_blkaddr(&dn, *blkaddr);
1206 f2fs_put_dnode(&dn);
1211 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1212 block_t *blkaddr, int *do_replace,
1213 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1215 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1220 if (blkaddr[i] == NULL_ADDR && !full) {
1225 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1226 struct dnode_of_data dn;
1227 struct node_info ni;
1231 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1232 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1236 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1238 f2fs_put_dnode(&dn);
1242 ilen = min((pgoff_t)
1243 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1244 dn.ofs_in_node, len - i);
1246 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1247 f2fs_truncate_data_blocks_range(&dn, 1);
1249 if (do_replace[i]) {
1250 f2fs_i_blocks_write(src_inode,
1252 f2fs_i_blocks_write(dst_inode,
1254 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1255 blkaddr[i], ni.version, true, false);
1261 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1262 if (dst_inode->i_size < new_size)
1263 f2fs_i_size_write(dst_inode, new_size);
1264 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1266 f2fs_put_dnode(&dn);
1268 struct page *psrc, *pdst;
1270 psrc = f2fs_get_lock_data_page(src_inode,
1273 return PTR_ERR(psrc);
1274 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1277 f2fs_put_page(psrc, 1);
1278 return PTR_ERR(pdst);
1280 f2fs_copy_page(psrc, pdst);
1281 set_page_dirty(pdst);
1282 f2fs_put_page(pdst, 1);
1283 f2fs_put_page(psrc, 1);
1285 ret = f2fs_truncate_hole(src_inode,
1286 src + i, src + i + 1);
1295 static int __exchange_data_block(struct inode *src_inode,
1296 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1297 pgoff_t len, bool full)
1299 block_t *src_blkaddr;
1305 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1307 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1308 array_size(olen, sizeof(block_t)),
1313 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1314 array_size(olen, sizeof(int)),
1317 kvfree(src_blkaddr);
1321 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1322 do_replace, src, olen);
1326 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1327 do_replace, src, dst, olen, full);
1335 kvfree(src_blkaddr);
1341 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1342 kvfree(src_blkaddr);
1347 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1349 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1350 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1351 pgoff_t start = offset >> PAGE_SHIFT;
1352 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1355 f2fs_balance_fs(sbi, true);
1357 /* avoid gc operation during block exchange */
1358 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1359 filemap_invalidate_lock(inode->i_mapping);
1362 f2fs_drop_extent_tree(inode);
1363 truncate_pagecache(inode, offset);
1364 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1365 f2fs_unlock_op(sbi);
1367 filemap_invalidate_unlock(inode->i_mapping);
1368 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1372 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1377 if (offset + len >= i_size_read(inode))
1380 /* collapse range should be aligned to block size of f2fs. */
1381 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1384 ret = f2fs_convert_inline_inode(inode);
1388 /* write out all dirty pages from offset */
1389 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1393 ret = f2fs_do_collapse(inode, offset, len);
1397 /* write out all moved pages, if possible */
1398 filemap_invalidate_lock(inode->i_mapping);
1399 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1400 truncate_pagecache(inode, offset);
1402 new_size = i_size_read(inode) - len;
1403 ret = f2fs_truncate_blocks(inode, new_size, true);
1404 filemap_invalidate_unlock(inode->i_mapping);
1406 f2fs_i_size_write(inode, new_size);
1410 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1413 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1414 pgoff_t index = start;
1415 unsigned int ofs_in_node = dn->ofs_in_node;
1419 for (; index < end; index++, dn->ofs_in_node++) {
1420 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1424 dn->ofs_in_node = ofs_in_node;
1425 ret = f2fs_reserve_new_blocks(dn, count);
1429 dn->ofs_in_node = ofs_in_node;
1430 for (index = start; index < end; index++, dn->ofs_in_node++) {
1431 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1433 * f2fs_reserve_new_blocks will not guarantee entire block
1436 if (dn->data_blkaddr == NULL_ADDR) {
1440 if (dn->data_blkaddr != NEW_ADDR) {
1441 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1442 dn->data_blkaddr = NEW_ADDR;
1443 f2fs_set_data_blkaddr(dn);
1447 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1452 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1455 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1456 struct address_space *mapping = inode->i_mapping;
1457 pgoff_t index, pg_start, pg_end;
1458 loff_t new_size = i_size_read(inode);
1459 loff_t off_start, off_end;
1462 ret = inode_newsize_ok(inode, (len + offset));
1466 ret = f2fs_convert_inline_inode(inode);
1470 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1474 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1475 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1477 off_start = offset & (PAGE_SIZE - 1);
1478 off_end = (offset + len) & (PAGE_SIZE - 1);
1480 if (pg_start == pg_end) {
1481 ret = fill_zero(inode, pg_start, off_start,
1482 off_end - off_start);
1486 new_size = max_t(loff_t, new_size, offset + len);
1489 ret = fill_zero(inode, pg_start++, off_start,
1490 PAGE_SIZE - off_start);
1494 new_size = max_t(loff_t, new_size,
1495 (loff_t)pg_start << PAGE_SHIFT);
1498 for (index = pg_start; index < pg_end;) {
1499 struct dnode_of_data dn;
1500 unsigned int end_offset;
1503 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1504 filemap_invalidate_lock(mapping);
1506 truncate_pagecache_range(inode,
1507 (loff_t)index << PAGE_SHIFT,
1508 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1512 set_new_dnode(&dn, inode, NULL, NULL, 0);
1513 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1515 f2fs_unlock_op(sbi);
1516 filemap_invalidate_unlock(mapping);
1517 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1521 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1522 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1524 ret = f2fs_do_zero_range(&dn, index, end);
1525 f2fs_put_dnode(&dn);
1527 f2fs_unlock_op(sbi);
1528 filemap_invalidate_unlock(mapping);
1529 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1531 f2fs_balance_fs(sbi, dn.node_changed);
1537 new_size = max_t(loff_t, new_size,
1538 (loff_t)index << PAGE_SHIFT);
1542 ret = fill_zero(inode, pg_end, 0, off_end);
1546 new_size = max_t(loff_t, new_size, offset + len);
1551 if (new_size > i_size_read(inode)) {
1552 if (mode & FALLOC_FL_KEEP_SIZE)
1553 file_set_keep_isize(inode);
1555 f2fs_i_size_write(inode, new_size);
1560 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1562 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1563 struct address_space *mapping = inode->i_mapping;
1564 pgoff_t nr, pg_start, pg_end, delta, idx;
1568 new_size = i_size_read(inode) + len;
1569 ret = inode_newsize_ok(inode, new_size);
1573 if (offset >= i_size_read(inode))
1576 /* insert range should be aligned to block size of f2fs. */
1577 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1580 ret = f2fs_convert_inline_inode(inode);
1584 f2fs_balance_fs(sbi, true);
1586 filemap_invalidate_lock(mapping);
1587 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1588 filemap_invalidate_unlock(mapping);
1592 /* write out all dirty pages from offset */
1593 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1597 pg_start = offset >> PAGE_SHIFT;
1598 pg_end = (offset + len) >> PAGE_SHIFT;
1599 delta = pg_end - pg_start;
1600 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1602 /* avoid gc operation during block exchange */
1603 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1604 filemap_invalidate_lock(mapping);
1605 truncate_pagecache(inode, offset);
1607 while (!ret && idx > pg_start) {
1608 nr = idx - pg_start;
1614 f2fs_drop_extent_tree(inode);
1616 ret = __exchange_data_block(inode, inode, idx,
1617 idx + delta, nr, false);
1618 f2fs_unlock_op(sbi);
1620 filemap_invalidate_unlock(mapping);
1621 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1623 /* write out all moved pages, if possible */
1624 filemap_invalidate_lock(mapping);
1625 filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1626 truncate_pagecache(inode, offset);
1627 filemap_invalidate_unlock(mapping);
1630 f2fs_i_size_write(inode, new_size);
1634 static int expand_inode_data(struct inode *inode, loff_t offset,
1635 loff_t len, int mode)
1637 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1638 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1639 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1640 .m_may_create = true };
1641 pgoff_t pg_start, pg_end;
1642 loff_t new_size = i_size_read(inode);
1644 block_t expanded = 0;
1647 err = inode_newsize_ok(inode, (len + offset));
1651 err = f2fs_convert_inline_inode(inode);
1655 f2fs_balance_fs(sbi, true);
1657 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1658 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1659 off_end = (offset + len) & (PAGE_SIZE - 1);
1661 map.m_lblk = pg_start;
1662 map.m_len = pg_end - pg_start;
1669 if (f2fs_is_pinned_file(inode)) {
1670 block_t sec_blks = BLKS_PER_SEC(sbi);
1671 block_t sec_len = roundup(map.m_len, sec_blks);
1673 map.m_len = sec_blks;
1675 if (has_not_enough_free_secs(sbi, 0,
1676 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1677 f2fs_down_write(&sbi->gc_lock);
1678 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1679 if (err && err != -ENODATA && err != -EAGAIN)
1683 f2fs_down_write(&sbi->pin_sem);
1686 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1687 f2fs_unlock_op(sbi);
1689 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1690 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1691 file_dont_truncate(inode);
1693 f2fs_up_write(&sbi->pin_sem);
1695 expanded += map.m_len;
1696 sec_len -= map.m_len;
1697 map.m_lblk += map.m_len;
1698 if (!err && sec_len)
1701 map.m_len = expanded;
1703 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1704 expanded = map.m_len;
1713 last_off = pg_start + expanded - 1;
1715 /* update new size to the failed position */
1716 new_size = (last_off == pg_end) ? offset + len :
1717 (loff_t)(last_off + 1) << PAGE_SHIFT;
1719 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1722 if (new_size > i_size_read(inode)) {
1723 if (mode & FALLOC_FL_KEEP_SIZE)
1724 file_set_keep_isize(inode);
1726 f2fs_i_size_write(inode, new_size);
1732 static long f2fs_fallocate(struct file *file, int mode,
1733 loff_t offset, loff_t len)
1735 struct inode *inode = file_inode(file);
1738 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1740 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1742 if (!f2fs_is_compress_backend_ready(inode))
1745 /* f2fs only support ->fallocate for regular file */
1746 if (!S_ISREG(inode->i_mode))
1749 if (IS_ENCRYPTED(inode) &&
1750 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1754 * Pinned file should not support partial trucation since the block
1755 * can be used by applications.
1757 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1758 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1759 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1762 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1763 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1764 FALLOC_FL_INSERT_RANGE))
1769 if (mode & FALLOC_FL_PUNCH_HOLE) {
1770 if (offset >= inode->i_size)
1773 ret = punch_hole(inode, offset, len);
1774 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1775 ret = f2fs_collapse_range(inode, offset, len);
1776 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1777 ret = f2fs_zero_range(inode, offset, len, mode);
1778 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1779 ret = f2fs_insert_range(inode, offset, len);
1781 ret = expand_inode_data(inode, offset, len, mode);
1785 inode->i_mtime = inode->i_ctime = current_time(inode);
1786 f2fs_mark_inode_dirty_sync(inode, false);
1787 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1791 inode_unlock(inode);
1793 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1797 static int f2fs_release_file(struct inode *inode, struct file *filp)
1800 * f2fs_relase_file is called at every close calls. So we should
1801 * not drop any inmemory pages by close called by other process.
1803 if (!(filp->f_mode & FMODE_WRITE) ||
1804 atomic_read(&inode->i_writecount) != 1)
1807 /* some remained atomic pages should discarded */
1808 if (f2fs_is_atomic_file(inode))
1809 f2fs_drop_inmem_pages(inode);
1810 if (f2fs_is_volatile_file(inode)) {
1811 set_inode_flag(inode, FI_DROP_CACHE);
1812 filemap_fdatawrite(inode->i_mapping);
1813 clear_inode_flag(inode, FI_DROP_CACHE);
1814 clear_inode_flag(inode, FI_VOLATILE_FILE);
1815 stat_dec_volatile_write(inode);
1820 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1822 struct inode *inode = file_inode(file);
1825 * If the process doing a transaction is crashed, we should do
1826 * roll-back. Otherwise, other reader/write can see corrupted database
1827 * until all the writers close its file. Since this should be done
1828 * before dropping file lock, it needs to do in ->flush.
1830 if (f2fs_is_atomic_file(inode) &&
1831 F2FS_I(inode)->inmem_task == current)
1832 f2fs_drop_inmem_pages(inode);
1836 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1838 struct f2fs_inode_info *fi = F2FS_I(inode);
1839 u32 masked_flags = fi->i_flags & mask;
1841 /* mask can be shrunk by flags_valid selector */
1844 /* Is it quota file? Do not allow user to mess with it */
1845 if (IS_NOQUOTA(inode))
1848 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1849 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1851 if (!f2fs_empty_dir(inode))
1855 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1856 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1858 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1862 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1863 if (masked_flags & F2FS_COMPR_FL) {
1864 if (!f2fs_disable_compressed_file(inode))
1867 if (iflags & F2FS_NOCOMP_FL)
1869 if (iflags & F2FS_COMPR_FL) {
1870 if (!f2fs_may_compress(inode))
1872 if (S_ISREG(inode->i_mode) && inode->i_size)
1875 set_compress_context(inode);
1878 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1879 if (masked_flags & F2FS_COMPR_FL)
1883 fi->i_flags = iflags | (fi->i_flags & ~mask);
1884 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1885 (fi->i_flags & F2FS_NOCOMP_FL));
1887 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1888 set_inode_flag(inode, FI_PROJ_INHERIT);
1890 clear_inode_flag(inode, FI_PROJ_INHERIT);
1892 inode->i_ctime = current_time(inode);
1893 f2fs_set_inode_flags(inode);
1894 f2fs_mark_inode_dirty_sync(inode, true);
1898 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1901 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1902 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1903 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1904 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1906 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1907 * FS_IOC_FSSETXATTR is done by the VFS.
1910 static const struct {
1913 } f2fs_fsflags_map[] = {
1914 { F2FS_COMPR_FL, FS_COMPR_FL },
1915 { F2FS_SYNC_FL, FS_SYNC_FL },
1916 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1917 { F2FS_APPEND_FL, FS_APPEND_FL },
1918 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1919 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1920 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
1921 { F2FS_INDEX_FL, FS_INDEX_FL },
1922 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1923 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1924 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1927 #define F2FS_GETTABLE_FS_FL ( \
1937 FS_PROJINHERIT_FL | \
1939 FS_INLINE_DATA_FL | \
1944 #define F2FS_SETTABLE_FS_FL ( \
1953 FS_PROJINHERIT_FL | \
1956 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1957 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1962 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1963 if (iflags & f2fs_fsflags_map[i].iflag)
1964 fsflags |= f2fs_fsflags_map[i].fsflag;
1969 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1970 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1975 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1976 if (fsflags & f2fs_fsflags_map[i].fsflag)
1977 iflags |= f2fs_fsflags_map[i].iflag;
1982 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1984 struct inode *inode = file_inode(filp);
1986 return put_user(inode->i_generation, (int __user *)arg);
1989 static int f2fs_ioc_start_atomic_write(struct file *filp)
1991 struct inode *inode = file_inode(filp);
1992 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
1993 struct f2fs_inode_info *fi = F2FS_I(inode);
1994 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1997 if (!inode_owner_or_capable(mnt_userns, inode))
2000 if (!S_ISREG(inode->i_mode))
2003 if (filp->f_flags & O_DIRECT)
2006 ret = mnt_want_write_file(filp);
2012 if (!f2fs_disable_compressed_file(inode)) {
2017 if (f2fs_is_atomic_file(inode)) {
2018 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2023 ret = f2fs_convert_inline_inode(inode);
2027 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2030 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2031 * f2fs_is_atomic_file.
2033 if (get_dirty_pages(inode))
2034 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2035 inode->i_ino, get_dirty_pages(inode));
2036 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2038 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2042 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2043 if (list_empty(&fi->inmem_ilist))
2044 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2045 sbi->atomic_files++;
2046 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2048 /* add inode in inmem_list first and set atomic_file */
2049 set_inode_flag(inode, FI_ATOMIC_FILE);
2050 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2051 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2053 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2054 F2FS_I(inode)->inmem_task = current;
2055 stat_update_max_atomic_write(inode);
2057 inode_unlock(inode);
2058 mnt_drop_write_file(filp);
2062 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2064 struct inode *inode = file_inode(filp);
2065 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2068 if (!inode_owner_or_capable(mnt_userns, inode))
2071 ret = mnt_want_write_file(filp);
2075 f2fs_balance_fs(F2FS_I_SB(inode), true);
2079 if (f2fs_is_volatile_file(inode)) {
2084 if (f2fs_is_atomic_file(inode)) {
2085 ret = f2fs_commit_inmem_pages(inode);
2089 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2091 f2fs_drop_inmem_pages(inode);
2093 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2096 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2097 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2100 inode_unlock(inode);
2101 mnt_drop_write_file(filp);
2105 static int f2fs_ioc_start_volatile_write(struct file *filp)
2107 struct inode *inode = file_inode(filp);
2108 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2111 if (!inode_owner_or_capable(mnt_userns, inode))
2114 if (!S_ISREG(inode->i_mode))
2117 ret = mnt_want_write_file(filp);
2123 if (f2fs_is_volatile_file(inode))
2126 ret = f2fs_convert_inline_inode(inode);
2130 stat_inc_volatile_write(inode);
2131 stat_update_max_volatile_write(inode);
2133 set_inode_flag(inode, FI_VOLATILE_FILE);
2134 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2136 inode_unlock(inode);
2137 mnt_drop_write_file(filp);
2141 static int f2fs_ioc_release_volatile_write(struct file *filp)
2143 struct inode *inode = file_inode(filp);
2144 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2147 if (!inode_owner_or_capable(mnt_userns, inode))
2150 ret = mnt_want_write_file(filp);
2156 if (!f2fs_is_volatile_file(inode))
2159 if (!f2fs_is_first_block_written(inode)) {
2160 ret = truncate_partial_data_page(inode, 0, true);
2164 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2166 inode_unlock(inode);
2167 mnt_drop_write_file(filp);
2171 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2173 struct inode *inode = file_inode(filp);
2174 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2177 if (!inode_owner_or_capable(mnt_userns, inode))
2180 ret = mnt_want_write_file(filp);
2186 if (f2fs_is_atomic_file(inode))
2187 f2fs_drop_inmem_pages(inode);
2188 if (f2fs_is_volatile_file(inode)) {
2189 clear_inode_flag(inode, FI_VOLATILE_FILE);
2190 stat_dec_volatile_write(inode);
2191 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2194 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2196 inode_unlock(inode);
2198 mnt_drop_write_file(filp);
2199 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2203 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2205 struct inode *inode = file_inode(filp);
2206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2207 struct super_block *sb = sbi->sb;
2211 if (!capable(CAP_SYS_ADMIN))
2214 if (get_user(in, (__u32 __user *)arg))
2217 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2218 ret = mnt_want_write_file(filp);
2220 if (ret == -EROFS) {
2222 f2fs_stop_checkpoint(sbi, false);
2223 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2224 trace_f2fs_shutdown(sbi, in, ret);
2231 case F2FS_GOING_DOWN_FULLSYNC:
2232 ret = freeze_bdev(sb->s_bdev);
2235 f2fs_stop_checkpoint(sbi, false);
2236 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2237 thaw_bdev(sb->s_bdev);
2239 case F2FS_GOING_DOWN_METASYNC:
2240 /* do checkpoint only */
2241 ret = f2fs_sync_fs(sb, 1);
2244 f2fs_stop_checkpoint(sbi, false);
2245 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2247 case F2FS_GOING_DOWN_NOSYNC:
2248 f2fs_stop_checkpoint(sbi, false);
2249 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2251 case F2FS_GOING_DOWN_METAFLUSH:
2252 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2253 f2fs_stop_checkpoint(sbi, false);
2254 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2256 case F2FS_GOING_DOWN_NEED_FSCK:
2257 set_sbi_flag(sbi, SBI_NEED_FSCK);
2258 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2259 set_sbi_flag(sbi, SBI_IS_DIRTY);
2260 /* do checkpoint only */
2261 ret = f2fs_sync_fs(sb, 1);
2268 f2fs_stop_gc_thread(sbi);
2269 f2fs_stop_discard_thread(sbi);
2271 f2fs_drop_discard_cmd(sbi);
2272 clear_opt(sbi, DISCARD);
2274 f2fs_update_time(sbi, REQ_TIME);
2276 if (in != F2FS_GOING_DOWN_FULLSYNC)
2277 mnt_drop_write_file(filp);
2279 trace_f2fs_shutdown(sbi, in, ret);
2284 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2286 struct inode *inode = file_inode(filp);
2287 struct super_block *sb = inode->i_sb;
2288 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2289 struct fstrim_range range;
2292 if (!capable(CAP_SYS_ADMIN))
2295 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2298 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2302 ret = mnt_want_write_file(filp);
2306 range.minlen = max((unsigned int)range.minlen,
2307 q->limits.discard_granularity);
2308 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2309 mnt_drop_write_file(filp);
2313 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2316 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2320 static bool uuid_is_nonzero(__u8 u[16])
2324 for (i = 0; i < 16; i++)
2330 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2332 struct inode *inode = file_inode(filp);
2334 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2337 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2339 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2342 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2344 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2346 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2349 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2351 struct inode *inode = file_inode(filp);
2352 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2355 if (!f2fs_sb_has_encrypt(sbi))
2358 err = mnt_want_write_file(filp);
2362 f2fs_down_write(&sbi->sb_lock);
2364 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2367 /* update superblock with uuid */
2368 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2370 err = f2fs_commit_super(sbi, false);
2373 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2377 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2381 f2fs_up_write(&sbi->sb_lock);
2382 mnt_drop_write_file(filp);
2386 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2389 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2392 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2395 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2397 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2400 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2403 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2405 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2408 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2411 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2414 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2417 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2420 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2423 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2426 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2429 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2431 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2434 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2437 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2439 struct inode *inode = file_inode(filp);
2440 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2444 if (!capable(CAP_SYS_ADMIN))
2447 if (get_user(sync, (__u32 __user *)arg))
2450 if (f2fs_readonly(sbi->sb))
2453 ret = mnt_want_write_file(filp);
2458 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2463 f2fs_down_write(&sbi->gc_lock);
2466 ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO);
2468 mnt_drop_write_file(filp);
2472 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2474 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2478 if (!capable(CAP_SYS_ADMIN))
2480 if (f2fs_readonly(sbi->sb))
2483 end = range->start + range->len;
2484 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2485 end >= MAX_BLKADDR(sbi))
2488 ret = mnt_want_write_file(filp);
2494 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2499 f2fs_down_write(&sbi->gc_lock);
2502 ret = f2fs_gc(sbi, range->sync, true, false,
2503 GET_SEGNO(sbi, range->start));
2509 range->start += BLKS_PER_SEC(sbi);
2510 if (range->start <= end)
2513 mnt_drop_write_file(filp);
2517 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2519 struct f2fs_gc_range range;
2521 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2524 return __f2fs_ioc_gc_range(filp, &range);
2527 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2529 struct inode *inode = file_inode(filp);
2530 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2533 if (!capable(CAP_SYS_ADMIN))
2536 if (f2fs_readonly(sbi->sb))
2539 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2540 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2544 ret = mnt_want_write_file(filp);
2548 ret = f2fs_sync_fs(sbi->sb, 1);
2550 mnt_drop_write_file(filp);
2554 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2556 struct f2fs_defragment *range)
2558 struct inode *inode = file_inode(filp);
2559 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2560 .m_seg_type = NO_CHECK_TYPE,
2561 .m_may_create = false };
2562 struct extent_info ei = {0, 0, 0};
2563 pgoff_t pg_start, pg_end, next_pgofs;
2564 unsigned int blk_per_seg = sbi->blocks_per_seg;
2565 unsigned int total = 0, sec_num;
2566 block_t blk_end = 0;
2567 bool fragmented = false;
2570 pg_start = range->start >> PAGE_SHIFT;
2571 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2573 f2fs_balance_fs(sbi, true);
2577 /* if in-place-update policy is enabled, don't waste time here */
2578 set_inode_flag(inode, FI_OPU_WRITE);
2579 if (f2fs_should_update_inplace(inode, NULL)) {
2584 /* writeback all dirty pages in the range */
2585 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2586 range->start + range->len - 1);
2591 * lookup mapping info in extent cache, skip defragmenting if physical
2592 * block addresses are continuous.
2594 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2595 if (ei.fofs + ei.len >= pg_end)
2599 map.m_lblk = pg_start;
2600 map.m_next_pgofs = &next_pgofs;
2603 * lookup mapping info in dnode page cache, skip defragmenting if all
2604 * physical block addresses are continuous even if there are hole(s)
2605 * in logical blocks.
2607 while (map.m_lblk < pg_end) {
2608 map.m_len = pg_end - map.m_lblk;
2609 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2613 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2614 map.m_lblk = next_pgofs;
2618 if (blk_end && blk_end != map.m_pblk)
2621 /* record total count of block that we're going to move */
2624 blk_end = map.m_pblk + map.m_len;
2626 map.m_lblk += map.m_len;
2634 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2637 * make sure there are enough free section for LFS allocation, this can
2638 * avoid defragment running in SSR mode when free section are allocated
2641 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2646 map.m_lblk = pg_start;
2647 map.m_len = pg_end - pg_start;
2650 while (map.m_lblk < pg_end) {
2655 map.m_len = pg_end - map.m_lblk;
2656 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2660 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2661 map.m_lblk = next_pgofs;
2665 set_inode_flag(inode, FI_SKIP_WRITES);
2668 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2671 page = f2fs_get_lock_data_page(inode, idx, true);
2673 err = PTR_ERR(page);
2677 set_page_dirty(page);
2678 f2fs_put_page(page, 1);
2687 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2690 clear_inode_flag(inode, FI_SKIP_WRITES);
2692 err = filemap_fdatawrite(inode->i_mapping);
2697 clear_inode_flag(inode, FI_SKIP_WRITES);
2699 clear_inode_flag(inode, FI_OPU_WRITE);
2700 inode_unlock(inode);
2702 range->len = (u64)total << PAGE_SHIFT;
2706 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2708 struct inode *inode = file_inode(filp);
2709 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2710 struct f2fs_defragment range;
2713 if (!capable(CAP_SYS_ADMIN))
2716 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2719 if (f2fs_readonly(sbi->sb))
2722 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2726 /* verify alignment of offset & size */
2727 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2730 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2731 max_file_blocks(inode)))
2734 err = mnt_want_write_file(filp);
2738 err = f2fs_defragment_range(sbi, filp, &range);
2739 mnt_drop_write_file(filp);
2741 f2fs_update_time(sbi, REQ_TIME);
2745 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2752 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2753 struct file *file_out, loff_t pos_out, size_t len)
2755 struct inode *src = file_inode(file_in);
2756 struct inode *dst = file_inode(file_out);
2757 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2758 size_t olen = len, dst_max_i_size = 0;
2762 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2763 src->i_sb != dst->i_sb)
2766 if (unlikely(f2fs_readonly(src->i_sb)))
2769 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2772 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2775 if (pos_out < 0 || pos_in < 0)
2779 if (pos_in == pos_out)
2781 if (pos_out > pos_in && pos_out < pos_in + len)
2788 if (!inode_trylock(dst))
2793 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2796 olen = len = src->i_size - pos_in;
2797 if (pos_in + len == src->i_size)
2798 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2804 dst_osize = dst->i_size;
2805 if (pos_out + olen > dst->i_size)
2806 dst_max_i_size = pos_out + olen;
2808 /* verify the end result is block aligned */
2809 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2810 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2811 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2814 ret = f2fs_convert_inline_inode(src);
2818 ret = f2fs_convert_inline_inode(dst);
2822 /* write out all dirty pages from offset */
2823 ret = filemap_write_and_wait_range(src->i_mapping,
2824 pos_in, pos_in + len);
2828 ret = filemap_write_and_wait_range(dst->i_mapping,
2829 pos_out, pos_out + len);
2833 f2fs_balance_fs(sbi, true);
2835 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2838 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2843 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2844 pos_out >> F2FS_BLKSIZE_BITS,
2845 len >> F2FS_BLKSIZE_BITS, false);
2849 f2fs_i_size_write(dst, dst_max_i_size);
2850 else if (dst_osize != dst->i_size)
2851 f2fs_i_size_write(dst, dst_osize);
2853 f2fs_unlock_op(sbi);
2856 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2858 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2867 static int __f2fs_ioc_move_range(struct file *filp,
2868 struct f2fs_move_range *range)
2873 if (!(filp->f_mode & FMODE_READ) ||
2874 !(filp->f_mode & FMODE_WRITE))
2877 dst = fdget(range->dst_fd);
2881 if (!(dst.file->f_mode & FMODE_WRITE)) {
2886 err = mnt_want_write_file(filp);
2890 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2891 range->pos_out, range->len);
2893 mnt_drop_write_file(filp);
2899 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2901 struct f2fs_move_range range;
2903 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2906 return __f2fs_ioc_move_range(filp, &range);
2909 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2911 struct inode *inode = file_inode(filp);
2912 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2913 struct sit_info *sm = SIT_I(sbi);
2914 unsigned int start_segno = 0, end_segno = 0;
2915 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2916 struct f2fs_flush_device range;
2919 if (!capable(CAP_SYS_ADMIN))
2922 if (f2fs_readonly(sbi->sb))
2925 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2928 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2932 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2933 __is_large_section(sbi)) {
2934 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2935 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2939 ret = mnt_want_write_file(filp);
2943 if (range.dev_num != 0)
2944 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2945 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2947 start_segno = sm->last_victim[FLUSH_DEVICE];
2948 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2949 start_segno = dev_start_segno;
2950 end_segno = min(start_segno + range.segments, dev_end_segno);
2952 while (start_segno < end_segno) {
2953 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2957 sm->last_victim[GC_CB] = end_segno + 1;
2958 sm->last_victim[GC_GREEDY] = end_segno + 1;
2959 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2960 ret = f2fs_gc(sbi, true, true, true, start_segno);
2968 mnt_drop_write_file(filp);
2972 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2974 struct inode *inode = file_inode(filp);
2975 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2977 /* Must validate to set it with SQLite behavior in Android. */
2978 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2980 return put_user(sb_feature, (u32 __user *)arg);
2984 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2986 struct dquot *transfer_to[MAXQUOTAS] = {};
2987 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2988 struct super_block *sb = sbi->sb;
2991 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2992 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2993 err = __dquot_transfer(inode, transfer_to);
2995 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2996 dqput(transfer_to[PRJQUOTA]);
3001 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3003 struct f2fs_inode_info *fi = F2FS_I(inode);
3004 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3005 struct f2fs_inode *ri = NULL;
3009 if (!f2fs_sb_has_project_quota(sbi)) {
3010 if (projid != F2FS_DEF_PROJID)
3016 if (!f2fs_has_extra_attr(inode))
3019 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3021 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
3025 /* Is it quota file? Do not allow user to mess with it */
3026 if (IS_NOQUOTA(inode))
3029 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3032 err = f2fs_dquot_initialize(inode);
3037 err = f2fs_transfer_project_quota(inode, kprojid);
3041 F2FS_I(inode)->i_projid = kprojid;
3042 inode->i_ctime = current_time(inode);
3043 f2fs_mark_inode_dirty_sync(inode, true);
3045 f2fs_unlock_op(sbi);
3049 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3054 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3056 if (projid != F2FS_DEF_PROJID)
3062 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3064 struct inode *inode = d_inode(dentry);
3065 struct f2fs_inode_info *fi = F2FS_I(inode);
3066 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3068 if (IS_ENCRYPTED(inode))
3069 fsflags |= FS_ENCRYPT_FL;
3070 if (IS_VERITY(inode))
3071 fsflags |= FS_VERITY_FL;
3072 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3073 fsflags |= FS_INLINE_DATA_FL;
3074 if (is_inode_flag_set(inode, FI_PIN_FILE))
3075 fsflags |= FS_NOCOW_FL;
3077 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3079 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3080 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3085 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3086 struct dentry *dentry, struct fileattr *fa)
3088 struct inode *inode = d_inode(dentry);
3089 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3093 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3095 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3097 if (fsflags & ~F2FS_GETTABLE_FS_FL)
3099 fsflags &= F2FS_SETTABLE_FS_FL;
3100 if (!fa->flags_valid)
3101 mask &= FS_COMMON_FL;
3103 iflags = f2fs_fsflags_to_iflags(fsflags);
3104 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3107 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3109 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3114 int f2fs_pin_file_control(struct inode *inode, bool inc)
3116 struct f2fs_inode_info *fi = F2FS_I(inode);
3117 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3119 /* Use i_gc_failures for normal file as a risk signal. */
3121 f2fs_i_gc_failures_write(inode,
3122 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3124 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3125 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3126 __func__, inode->i_ino,
3127 fi->i_gc_failures[GC_FAILURE_PIN]);
3128 clear_inode_flag(inode, FI_PIN_FILE);
3134 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3136 struct inode *inode = file_inode(filp);
3140 if (get_user(pin, (__u32 __user *)arg))
3143 if (!S_ISREG(inode->i_mode))
3146 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3149 ret = mnt_want_write_file(filp);
3156 clear_inode_flag(inode, FI_PIN_FILE);
3157 f2fs_i_gc_failures_write(inode, 0);
3161 if (f2fs_should_update_outplace(inode, NULL)) {
3166 if (f2fs_pin_file_control(inode, false)) {
3171 ret = f2fs_convert_inline_inode(inode);
3175 if (!f2fs_disable_compressed_file(inode)) {
3180 set_inode_flag(inode, FI_PIN_FILE);
3181 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3183 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3185 inode_unlock(inode);
3186 mnt_drop_write_file(filp);
3190 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3192 struct inode *inode = file_inode(filp);
3195 if (is_inode_flag_set(inode, FI_PIN_FILE))
3196 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3197 return put_user(pin, (u32 __user *)arg);
3200 int f2fs_precache_extents(struct inode *inode)
3202 struct f2fs_inode_info *fi = F2FS_I(inode);
3203 struct f2fs_map_blocks map;
3204 pgoff_t m_next_extent;
3208 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3212 map.m_next_pgofs = NULL;
3213 map.m_next_extent = &m_next_extent;
3214 map.m_seg_type = NO_CHECK_TYPE;
3215 map.m_may_create = false;
3216 end = max_file_blocks(inode);
3218 while (map.m_lblk < end) {
3219 map.m_len = end - map.m_lblk;
3221 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3222 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3223 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3227 map.m_lblk = m_next_extent;
3233 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3235 return f2fs_precache_extents(file_inode(filp));
3238 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3240 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3243 if (!capable(CAP_SYS_ADMIN))
3246 if (f2fs_readonly(sbi->sb))
3249 if (copy_from_user(&block_count, (void __user *)arg,
3250 sizeof(block_count)))
3253 return f2fs_resize_fs(sbi, block_count);
3256 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3258 struct inode *inode = file_inode(filp);
3260 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3262 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3263 f2fs_warn(F2FS_I_SB(inode),
3264 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3269 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3272 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3274 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3277 return fsverity_ioctl_measure(filp, (void __user *)arg);
3280 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3282 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3285 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3288 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3290 struct inode *inode = file_inode(filp);
3291 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3296 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3300 f2fs_down_read(&sbi->sb_lock);
3301 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3302 ARRAY_SIZE(sbi->raw_super->volume_name),
3303 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3304 f2fs_up_read(&sbi->sb_lock);
3306 if (copy_to_user((char __user *)arg, vbuf,
3307 min(FSLABEL_MAX, count)))
3314 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3316 struct inode *inode = file_inode(filp);
3317 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3321 if (!capable(CAP_SYS_ADMIN))
3324 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3326 return PTR_ERR(vbuf);
3328 err = mnt_want_write_file(filp);
3332 f2fs_down_write(&sbi->sb_lock);
3334 memset(sbi->raw_super->volume_name, 0,
3335 sizeof(sbi->raw_super->volume_name));
3336 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3337 sbi->raw_super->volume_name,
3338 ARRAY_SIZE(sbi->raw_super->volume_name));
3340 err = f2fs_commit_super(sbi, false);
3342 f2fs_up_write(&sbi->sb_lock);
3344 mnt_drop_write_file(filp);
3350 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3352 struct inode *inode = file_inode(filp);
3355 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3358 if (!f2fs_compressed_file(inode))
3361 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3362 return put_user(blocks, (u64 __user *)arg);
3365 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3367 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3368 unsigned int released_blocks = 0;
3369 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3373 for (i = 0; i < count; i++) {
3374 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3375 dn->ofs_in_node + i);
3377 if (!__is_valid_data_blkaddr(blkaddr))
3379 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3380 DATA_GENERIC_ENHANCE)))
3381 return -EFSCORRUPTED;
3385 int compr_blocks = 0;
3387 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3388 blkaddr = f2fs_data_blkaddr(dn);
3391 if (blkaddr == COMPRESS_ADDR)
3393 dn->ofs_in_node += cluster_size;
3397 if (__is_valid_data_blkaddr(blkaddr))
3400 if (blkaddr != NEW_ADDR)
3403 dn->data_blkaddr = NULL_ADDR;
3404 f2fs_set_data_blkaddr(dn);
3407 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3408 dec_valid_block_count(sbi, dn->inode,
3409 cluster_size - compr_blocks);
3411 released_blocks += cluster_size - compr_blocks;
3413 count -= cluster_size;
3416 return released_blocks;
3419 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3421 struct inode *inode = file_inode(filp);
3422 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3423 pgoff_t page_idx = 0, last_idx;
3424 unsigned int released_blocks = 0;
3428 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3431 if (!f2fs_compressed_file(inode))
3434 if (f2fs_readonly(sbi->sb))
3437 ret = mnt_want_write_file(filp);
3441 f2fs_balance_fs(F2FS_I_SB(inode), true);
3445 writecount = atomic_read(&inode->i_writecount);
3446 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3447 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3452 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3457 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3461 set_inode_flag(inode, FI_COMPRESS_RELEASED);
3462 inode->i_ctime = current_time(inode);
3463 f2fs_mark_inode_dirty_sync(inode, true);
3465 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3468 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3469 filemap_invalidate_lock(inode->i_mapping);
3471 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3473 while (page_idx < last_idx) {
3474 struct dnode_of_data dn;
3475 pgoff_t end_offset, count;
3477 set_new_dnode(&dn, inode, NULL, NULL, 0);
3478 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3480 if (ret == -ENOENT) {
3481 page_idx = f2fs_get_next_page_offset(&dn,
3489 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3490 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3491 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3493 ret = release_compress_blocks(&dn, count);
3495 f2fs_put_dnode(&dn);
3501 released_blocks += ret;
3504 filemap_invalidate_unlock(inode->i_mapping);
3505 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3507 inode_unlock(inode);
3509 mnt_drop_write_file(filp);
3512 ret = put_user(released_blocks, (u64 __user *)arg);
3513 } else if (released_blocks &&
3514 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3515 set_sbi_flag(sbi, SBI_NEED_FSCK);
3516 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3517 "iblocks=%llu, released=%u, compr_blocks=%u, "
3519 __func__, inode->i_ino, inode->i_blocks,
3521 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3527 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3529 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3530 unsigned int reserved_blocks = 0;
3531 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3535 for (i = 0; i < count; i++) {
3536 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3537 dn->ofs_in_node + i);
3539 if (!__is_valid_data_blkaddr(blkaddr))
3541 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3542 DATA_GENERIC_ENHANCE)))
3543 return -EFSCORRUPTED;
3547 int compr_blocks = 0;
3551 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3552 blkaddr = f2fs_data_blkaddr(dn);
3555 if (blkaddr == COMPRESS_ADDR)
3557 dn->ofs_in_node += cluster_size;
3561 if (__is_valid_data_blkaddr(blkaddr)) {
3566 dn->data_blkaddr = NEW_ADDR;
3567 f2fs_set_data_blkaddr(dn);
3570 reserved = cluster_size - compr_blocks;
3571 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3575 if (reserved != cluster_size - compr_blocks)
3578 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3580 reserved_blocks += reserved;
3582 count -= cluster_size;
3585 return reserved_blocks;
3588 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3590 struct inode *inode = file_inode(filp);
3591 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3592 pgoff_t page_idx = 0, last_idx;
3593 unsigned int reserved_blocks = 0;
3596 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3599 if (!f2fs_compressed_file(inode))
3602 if (f2fs_readonly(sbi->sb))
3605 ret = mnt_want_write_file(filp);
3609 if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3612 f2fs_balance_fs(F2FS_I_SB(inode), true);
3616 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3621 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3622 filemap_invalidate_lock(inode->i_mapping);
3624 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3626 while (page_idx < last_idx) {
3627 struct dnode_of_data dn;
3628 pgoff_t end_offset, count;
3630 set_new_dnode(&dn, inode, NULL, NULL, 0);
3631 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3633 if (ret == -ENOENT) {
3634 page_idx = f2fs_get_next_page_offset(&dn,
3642 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3643 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3644 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3646 ret = reserve_compress_blocks(&dn, count);
3648 f2fs_put_dnode(&dn);
3654 reserved_blocks += ret;
3657 filemap_invalidate_unlock(inode->i_mapping);
3658 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3661 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3662 inode->i_ctime = current_time(inode);
3663 f2fs_mark_inode_dirty_sync(inode, true);
3666 inode_unlock(inode);
3668 mnt_drop_write_file(filp);
3671 ret = put_user(reserved_blocks, (u64 __user *)arg);
3672 } else if (reserved_blocks &&
3673 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3674 set_sbi_flag(sbi, SBI_NEED_FSCK);
3675 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3676 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3678 __func__, inode->i_ino, inode->i_blocks,
3680 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3686 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3687 pgoff_t off, block_t block, block_t len, u32 flags)
3689 struct request_queue *q = bdev_get_queue(bdev);
3690 sector_t sector = SECTOR_FROM_BLOCK(block);
3691 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3697 if (flags & F2FS_TRIM_FILE_DISCARD)
3698 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3699 blk_queue_secure_erase(q) ?
3700 BLKDEV_DISCARD_SECURE : 0);
3702 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3703 if (IS_ENCRYPTED(inode))
3704 ret = fscrypt_zeroout_range(inode, off, block, len);
3706 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3713 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3715 struct inode *inode = file_inode(filp);
3716 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3717 struct address_space *mapping = inode->i_mapping;
3718 struct block_device *prev_bdev = NULL;
3719 struct f2fs_sectrim_range range;
3720 pgoff_t index, pg_end, prev_index = 0;
3721 block_t prev_block = 0, len = 0;
3723 bool to_end = false;
3726 if (!(filp->f_mode & FMODE_WRITE))
3729 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3733 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3734 !S_ISREG(inode->i_mode))
3737 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3738 !f2fs_hw_support_discard(sbi)) ||
3739 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3740 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3743 file_start_write(filp);
3746 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3747 range.start >= inode->i_size) {
3755 if (inode->i_size - range.start > range.len) {
3756 end_addr = range.start + range.len;
3758 end_addr = range.len == (u64)-1 ?
3759 sbi->sb->s_maxbytes : inode->i_size;
3763 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3764 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3769 index = F2FS_BYTES_TO_BLK(range.start);
3770 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3772 ret = f2fs_convert_inline_inode(inode);
3776 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3777 filemap_invalidate_lock(mapping);
3779 ret = filemap_write_and_wait_range(mapping, range.start,
3780 to_end ? LLONG_MAX : end_addr - 1);
3784 truncate_inode_pages_range(mapping, range.start,
3785 to_end ? -1 : end_addr - 1);
3787 while (index < pg_end) {
3788 struct dnode_of_data dn;
3789 pgoff_t end_offset, count;
3792 set_new_dnode(&dn, inode, NULL, NULL, 0);
3793 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3795 if (ret == -ENOENT) {
3796 index = f2fs_get_next_page_offset(&dn, index);
3802 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3803 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3804 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3805 struct block_device *cur_bdev;
3806 block_t blkaddr = f2fs_data_blkaddr(&dn);
3808 if (!__is_valid_data_blkaddr(blkaddr))
3811 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3812 DATA_GENERIC_ENHANCE)) {
3813 ret = -EFSCORRUPTED;
3814 f2fs_put_dnode(&dn);
3818 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3819 if (f2fs_is_multi_device(sbi)) {
3820 int di = f2fs_target_device_index(sbi, blkaddr);
3822 blkaddr -= FDEV(di).start_blk;
3826 if (prev_bdev == cur_bdev &&
3827 index == prev_index + len &&
3828 blkaddr == prev_block + len) {
3831 ret = f2fs_secure_erase(prev_bdev,
3832 inode, prev_index, prev_block,
3835 f2fs_put_dnode(&dn);
3844 prev_bdev = cur_bdev;
3846 prev_block = blkaddr;
3851 f2fs_put_dnode(&dn);
3853 if (fatal_signal_pending(current)) {
3861 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3862 prev_block, len, range.flags);
3864 filemap_invalidate_unlock(mapping);
3865 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3867 inode_unlock(inode);
3868 file_end_write(filp);
3873 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3875 struct inode *inode = file_inode(filp);
3876 struct f2fs_comp_option option;
3878 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3881 inode_lock_shared(inode);
3883 if (!f2fs_compressed_file(inode)) {
3884 inode_unlock_shared(inode);
3888 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3889 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3891 inode_unlock_shared(inode);
3893 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3900 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3902 struct inode *inode = file_inode(filp);
3903 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3904 struct f2fs_comp_option option;
3907 if (!f2fs_sb_has_compression(sbi))
3910 if (!(filp->f_mode & FMODE_WRITE))
3913 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3917 if (!f2fs_compressed_file(inode) ||
3918 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3919 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3920 option.algorithm >= COMPRESS_MAX)
3923 file_start_write(filp);
3926 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3931 if (inode->i_size != 0) {
3936 F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3937 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3938 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3939 f2fs_mark_inode_dirty_sync(inode, true);
3941 if (!f2fs_is_compress_backend_ready(inode))
3942 f2fs_warn(sbi, "compression algorithm is successfully set, "
3943 "but current kernel doesn't support this algorithm.");
3945 inode_unlock(inode);
3946 file_end_write(filp);
3951 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3953 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3954 struct address_space *mapping = inode->i_mapping;
3956 pgoff_t redirty_idx = page_idx;
3957 int i, page_len = 0, ret = 0;
3959 page_cache_ra_unbounded(&ractl, len, 0);
3961 for (i = 0; i < len; i++, page_idx++) {
3962 page = read_cache_page(mapping, page_idx, NULL, NULL);
3964 ret = PTR_ERR(page);
3970 for (i = 0; i < page_len; i++, redirty_idx++) {
3971 page = find_lock_page(mapping, redirty_idx);
3976 set_page_dirty(page);
3977 f2fs_put_page(page, 1);
3978 f2fs_put_page(page, 0);
3984 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3986 struct inode *inode = file_inode(filp);
3987 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3988 struct f2fs_inode_info *fi = F2FS_I(inode);
3989 pgoff_t page_idx = 0, last_idx;
3990 unsigned int blk_per_seg = sbi->blocks_per_seg;
3991 int cluster_size = F2FS_I(inode)->i_cluster_size;
3994 if (!f2fs_sb_has_compression(sbi) ||
3995 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3998 if (!(filp->f_mode & FMODE_WRITE))
4001 if (!f2fs_compressed_file(inode))
4004 f2fs_balance_fs(F2FS_I_SB(inode), true);
4006 file_start_write(filp);
4009 if (!f2fs_is_compress_backend_ready(inode)) {
4014 if (f2fs_is_mmap_file(inode)) {
4019 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4023 if (!atomic_read(&fi->i_compr_blocks))
4026 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4028 count = last_idx - page_idx;
4030 int len = min(cluster_size, count);
4032 ret = redirty_blocks(inode, page_idx, len);
4036 if (get_dirty_pages(inode) >= blk_per_seg)
4037 filemap_fdatawrite(inode->i_mapping);
4044 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4048 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4051 inode_unlock(inode);
4052 file_end_write(filp);
4057 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4059 struct inode *inode = file_inode(filp);
4060 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4061 pgoff_t page_idx = 0, last_idx;
4062 unsigned int blk_per_seg = sbi->blocks_per_seg;
4063 int cluster_size = F2FS_I(inode)->i_cluster_size;
4066 if (!f2fs_sb_has_compression(sbi) ||
4067 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4070 if (!(filp->f_mode & FMODE_WRITE))
4073 if (!f2fs_compressed_file(inode))
4076 f2fs_balance_fs(F2FS_I_SB(inode), true);
4078 file_start_write(filp);
4081 if (!f2fs_is_compress_backend_ready(inode)) {
4086 if (f2fs_is_mmap_file(inode)) {
4091 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4095 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4097 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4099 count = last_idx - page_idx;
4101 int len = min(cluster_size, count);
4103 ret = redirty_blocks(inode, page_idx, len);
4107 if (get_dirty_pages(inode) >= blk_per_seg)
4108 filemap_fdatawrite(inode->i_mapping);
4115 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4118 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4121 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4124 inode_unlock(inode);
4125 file_end_write(filp);
4130 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4133 case FS_IOC_GETVERSION:
4134 return f2fs_ioc_getversion(filp, arg);
4135 case F2FS_IOC_START_ATOMIC_WRITE:
4136 return f2fs_ioc_start_atomic_write(filp);
4137 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4138 return f2fs_ioc_commit_atomic_write(filp);
4139 case F2FS_IOC_START_VOLATILE_WRITE:
4140 return f2fs_ioc_start_volatile_write(filp);
4141 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4142 return f2fs_ioc_release_volatile_write(filp);
4143 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4144 return f2fs_ioc_abort_volatile_write(filp);
4145 case F2FS_IOC_SHUTDOWN:
4146 return f2fs_ioc_shutdown(filp, arg);
4148 return f2fs_ioc_fitrim(filp, arg);
4149 case FS_IOC_SET_ENCRYPTION_POLICY:
4150 return f2fs_ioc_set_encryption_policy(filp, arg);
4151 case FS_IOC_GET_ENCRYPTION_POLICY:
4152 return f2fs_ioc_get_encryption_policy(filp, arg);
4153 case FS_IOC_GET_ENCRYPTION_PWSALT:
4154 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4155 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4156 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4157 case FS_IOC_ADD_ENCRYPTION_KEY:
4158 return f2fs_ioc_add_encryption_key(filp, arg);
4159 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4160 return f2fs_ioc_remove_encryption_key(filp, arg);
4161 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4162 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4163 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4164 return f2fs_ioc_get_encryption_key_status(filp, arg);
4165 case FS_IOC_GET_ENCRYPTION_NONCE:
4166 return f2fs_ioc_get_encryption_nonce(filp, arg);
4167 case F2FS_IOC_GARBAGE_COLLECT:
4168 return f2fs_ioc_gc(filp, arg);
4169 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4170 return f2fs_ioc_gc_range(filp, arg);
4171 case F2FS_IOC_WRITE_CHECKPOINT:
4172 return f2fs_ioc_write_checkpoint(filp, arg);
4173 case F2FS_IOC_DEFRAGMENT:
4174 return f2fs_ioc_defragment(filp, arg);
4175 case F2FS_IOC_MOVE_RANGE:
4176 return f2fs_ioc_move_range(filp, arg);
4177 case F2FS_IOC_FLUSH_DEVICE:
4178 return f2fs_ioc_flush_device(filp, arg);
4179 case F2FS_IOC_GET_FEATURES:
4180 return f2fs_ioc_get_features(filp, arg);
4181 case F2FS_IOC_GET_PIN_FILE:
4182 return f2fs_ioc_get_pin_file(filp, arg);
4183 case F2FS_IOC_SET_PIN_FILE:
4184 return f2fs_ioc_set_pin_file(filp, arg);
4185 case F2FS_IOC_PRECACHE_EXTENTS:
4186 return f2fs_ioc_precache_extents(filp, arg);
4187 case F2FS_IOC_RESIZE_FS:
4188 return f2fs_ioc_resize_fs(filp, arg);
4189 case FS_IOC_ENABLE_VERITY:
4190 return f2fs_ioc_enable_verity(filp, arg);
4191 case FS_IOC_MEASURE_VERITY:
4192 return f2fs_ioc_measure_verity(filp, arg);
4193 case FS_IOC_READ_VERITY_METADATA:
4194 return f2fs_ioc_read_verity_metadata(filp, arg);
4195 case FS_IOC_GETFSLABEL:
4196 return f2fs_ioc_getfslabel(filp, arg);
4197 case FS_IOC_SETFSLABEL:
4198 return f2fs_ioc_setfslabel(filp, arg);
4199 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4200 return f2fs_get_compress_blocks(filp, arg);
4201 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4202 return f2fs_release_compress_blocks(filp, arg);
4203 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4204 return f2fs_reserve_compress_blocks(filp, arg);
4205 case F2FS_IOC_SEC_TRIM_FILE:
4206 return f2fs_sec_trim_file(filp, arg);
4207 case F2FS_IOC_GET_COMPRESS_OPTION:
4208 return f2fs_ioc_get_compress_option(filp, arg);
4209 case F2FS_IOC_SET_COMPRESS_OPTION:
4210 return f2fs_ioc_set_compress_option(filp, arg);
4211 case F2FS_IOC_DECOMPRESS_FILE:
4212 return f2fs_ioc_decompress_file(filp, arg);
4213 case F2FS_IOC_COMPRESS_FILE:
4214 return f2fs_ioc_compress_file(filp, arg);
4220 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4222 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4224 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4227 return __f2fs_ioctl(filp, cmd, arg);
4231 * Return %true if the given read or write request should use direct I/O, or
4232 * %false if it should use buffered I/O.
4234 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4235 struct iov_iter *iter)
4239 if (!(iocb->ki_flags & IOCB_DIRECT))
4242 if (f2fs_force_buffered_io(inode, iocb, iter))
4246 * Direct I/O not aligned to the disk's logical_block_size will be
4247 * attempted, but will fail with -EINVAL.
4249 * f2fs additionally requires that direct I/O be aligned to the
4250 * filesystem block size, which is often a stricter requirement.
4251 * However, f2fs traditionally falls back to buffered I/O on requests
4252 * that are logical_block_size-aligned but not fs-block aligned.
4254 * The below logic implements this behavior.
4256 align = iocb->ki_pos | iov_iter_alignment(iter);
4257 if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4258 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4264 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4267 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4269 dec_page_count(sbi, F2FS_DIO_READ);
4272 f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size);
4276 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4277 .end_io = f2fs_dio_read_end_io,
4280 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4282 struct file *file = iocb->ki_filp;
4283 struct inode *inode = file_inode(file);
4284 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4285 struct f2fs_inode_info *fi = F2FS_I(inode);
4286 const loff_t pos = iocb->ki_pos;
4287 const size_t count = iov_iter_count(to);
4288 struct iomap_dio *dio;
4292 return 0; /* skip atime update */
4294 trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4296 if (iocb->ki_flags & IOCB_NOWAIT) {
4297 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4302 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4306 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4307 * the higher-level function iomap_dio_rw() in order to ensure that the
4308 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4310 inc_page_count(sbi, F2FS_DIO_READ);
4311 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4312 &f2fs_iomap_dio_read_ops, 0, 0);
4313 if (IS_ERR_OR_NULL(dio)) {
4314 ret = PTR_ERR_OR_ZERO(dio);
4315 if (ret != -EIOCBQUEUED)
4316 dec_page_count(sbi, F2FS_DIO_READ);
4318 ret = iomap_dio_complete(dio);
4321 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4323 file_accessed(file);
4325 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4329 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4331 struct inode *inode = file_inode(iocb->ki_filp);
4334 if (!f2fs_is_compress_backend_ready(inode))
4337 if (f2fs_should_use_dio(inode, iocb, to))
4338 return f2fs_dio_read_iter(iocb, to);
4340 ret = filemap_read(iocb, to, 0);
4342 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret);
4346 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4348 struct file *file = iocb->ki_filp;
4349 struct inode *inode = file_inode(file);
4353 if (IS_IMMUTABLE(inode))
4356 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4359 count = generic_write_checks(iocb, from);
4363 err = file_modified(file);
4370 * Preallocate blocks for a write request, if it is possible and helpful to do
4371 * so. Returns a positive number if blocks may have been preallocated, 0 if no
4372 * blocks were preallocated, or a negative errno value if something went
4373 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4374 * requested blocks (not just some of them) have been allocated.
4376 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4379 struct inode *inode = file_inode(iocb->ki_filp);
4380 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4381 const loff_t pos = iocb->ki_pos;
4382 const size_t count = iov_iter_count(iter);
4383 struct f2fs_map_blocks map = {};
4387 /* If it will be an out-of-place direct write, don't bother. */
4388 if (dio && f2fs_lfs_mode(sbi))
4391 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4392 * buffered IO, if DIO meets any holes.
4394 if (dio && i_size_read(inode) &&
4395 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4398 /* No-wait I/O can't allocate blocks. */
4399 if (iocb->ki_flags & IOCB_NOWAIT)
4402 /* If it will be a short write, don't bother. */
4403 if (fault_in_iov_iter_readable(iter, count))
4406 if (f2fs_has_inline_data(inode)) {
4407 /* If the data will fit inline, don't bother. */
4408 if (pos + count <= MAX_INLINE_DATA(inode))
4410 ret = f2fs_convert_inline_inode(inode);
4415 /* Do not preallocate blocks that will be written partially in 4KB. */
4416 map.m_lblk = F2FS_BLK_ALIGN(pos);
4417 map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4418 if (map.m_len > map.m_lblk)
4419 map.m_len -= map.m_lblk;
4422 map.m_may_create = true;
4424 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4425 flag = F2FS_GET_BLOCK_PRE_DIO;
4427 map.m_seg_type = NO_CHECK_TYPE;
4428 flag = F2FS_GET_BLOCK_PRE_AIO;
4431 ret = f2fs_map_blocks(inode, &map, 1, flag);
4432 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4433 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4436 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4440 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4441 struct iov_iter *from)
4443 struct file *file = iocb->ki_filp;
4444 struct inode *inode = file_inode(file);
4447 if (iocb->ki_flags & IOCB_NOWAIT)
4450 current->backing_dev_info = inode_to_bdi(inode);
4451 ret = generic_perform_write(file, from, iocb->ki_pos);
4452 current->backing_dev_info = NULL;
4455 iocb->ki_pos += ret;
4456 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret);
4461 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4464 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4466 dec_page_count(sbi, F2FS_DIO_WRITE);
4469 f2fs_update_iostat(sbi, APP_DIRECT_IO, size);
4473 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4474 .end_io = f2fs_dio_write_end_io,
4477 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4478 bool *may_need_sync)
4480 struct file *file = iocb->ki_filp;
4481 struct inode *inode = file_inode(file);
4482 struct f2fs_inode_info *fi = F2FS_I(inode);
4483 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4484 const bool do_opu = f2fs_lfs_mode(sbi);
4485 const loff_t pos = iocb->ki_pos;
4486 const ssize_t count = iov_iter_count(from);
4487 unsigned int dio_flags;
4488 struct iomap_dio *dio;
4491 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4493 if (iocb->ki_flags & IOCB_NOWAIT) {
4494 /* f2fs_convert_inline_inode() and block allocation can block */
4495 if (f2fs_has_inline_data(inode) ||
4496 !f2fs_overwrite_io(inode, pos, count)) {
4501 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4505 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4506 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4511 ret = f2fs_convert_inline_inode(inode);
4515 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4517 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4521 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4522 * the higher-level function iomap_dio_rw() in order to ensure that the
4523 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4525 inc_page_count(sbi, F2FS_DIO_WRITE);
4527 if (pos + count > inode->i_size)
4528 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4529 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4530 &f2fs_iomap_dio_write_ops, dio_flags, 0);
4531 if (IS_ERR_OR_NULL(dio)) {
4532 ret = PTR_ERR_OR_ZERO(dio);
4533 if (ret == -ENOTBLK)
4535 if (ret != -EIOCBQUEUED)
4536 dec_page_count(sbi, F2FS_DIO_WRITE);
4538 ret = iomap_dio_complete(dio);
4542 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4543 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4547 if (pos + ret > inode->i_size)
4548 f2fs_i_size_write(inode, pos + ret);
4550 set_inode_flag(inode, FI_UPDATE_WRITE);
4552 if (iov_iter_count(from)) {
4554 loff_t bufio_start_pos = iocb->ki_pos;
4557 * The direct write was partial, so we need to fall back to a
4558 * buffered write for the remainder.
4561 ret2 = f2fs_buffered_write_iter(iocb, from);
4562 if (iov_iter_count(from))
4563 f2fs_write_failed(inode, iocb->ki_pos);
4568 * Ensure that the pagecache pages are written to disk and
4569 * invalidated to preserve the expected O_DIRECT semantics.
4572 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4576 ret2 = filemap_write_and_wait_range(file->f_mapping,
4581 invalidate_mapping_pages(file->f_mapping,
4582 bufio_start_pos >> PAGE_SHIFT,
4583 bufio_end_pos >> PAGE_SHIFT);
4586 /* iomap_dio_rw() already handled the generic_write_sync(). */
4587 *may_need_sync = false;
4590 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4594 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4596 struct inode *inode = file_inode(iocb->ki_filp);
4597 const loff_t orig_pos = iocb->ki_pos;
4598 const size_t orig_count = iov_iter_count(from);
4601 bool may_need_sync = true;
4605 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4610 if (!f2fs_is_compress_backend_ready(inode)) {
4615 if (iocb->ki_flags & IOCB_NOWAIT) {
4616 if (!inode_trylock(inode)) {
4624 ret = f2fs_write_checks(iocb, from);
4628 /* Determine whether we will do a direct write or a buffered write. */
4629 dio = f2fs_should_use_dio(inode, iocb, from);
4631 /* Possibly preallocate the blocks for the write. */
4632 target_size = iocb->ki_pos + iov_iter_count(from);
4633 preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4634 if (preallocated < 0)
4637 /* Do the actual write. */
4639 f2fs_dio_write_iter(iocb, from, &may_need_sync):
4640 f2fs_buffered_write_iter(iocb, from);
4642 /* Don't leave any preallocated blocks around past i_size. */
4643 if (preallocated && i_size_read(inode) < target_size) {
4644 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4645 filemap_invalidate_lock(inode->i_mapping);
4646 if (!f2fs_truncate(inode))
4647 file_dont_truncate(inode);
4648 filemap_invalidate_unlock(inode->i_mapping);
4649 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4651 file_dont_truncate(inode);
4654 clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4656 inode_unlock(inode);
4658 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4659 if (ret > 0 && may_need_sync)
4660 ret = generic_write_sync(iocb, ret);
4664 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4667 struct address_space *mapping;
4668 struct backing_dev_info *bdi;
4669 struct inode *inode = file_inode(filp);
4672 if (advice == POSIX_FADV_SEQUENTIAL) {
4673 if (S_ISFIFO(inode->i_mode))
4676 mapping = filp->f_mapping;
4677 if (!mapping || len < 0)
4680 bdi = inode_to_bdi(mapping->host);
4681 filp->f_ra.ra_pages = bdi->ra_pages *
4682 F2FS_I_SB(inode)->seq_file_ra_mul;
4683 spin_lock(&filp->f_lock);
4684 filp->f_mode &= ~FMODE_RANDOM;
4685 spin_unlock(&filp->f_lock);
4689 err = generic_fadvise(filp, offset, len, advice);
4690 if (!err && advice == POSIX_FADV_DONTNEED &&
4691 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4692 f2fs_compressed_file(inode))
4693 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4698 #ifdef CONFIG_COMPAT
4699 struct compat_f2fs_gc_range {
4704 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
4705 struct compat_f2fs_gc_range)
4707 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4709 struct compat_f2fs_gc_range __user *urange;
4710 struct f2fs_gc_range range;
4713 urange = compat_ptr(arg);
4714 err = get_user(range.sync, &urange->sync);
4715 err |= get_user(range.start, &urange->start);
4716 err |= get_user(range.len, &urange->len);
4720 return __f2fs_ioc_gc_range(file, &range);
4723 struct compat_f2fs_move_range {
4729 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
4730 struct compat_f2fs_move_range)
4732 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4734 struct compat_f2fs_move_range __user *urange;
4735 struct f2fs_move_range range;
4738 urange = compat_ptr(arg);
4739 err = get_user(range.dst_fd, &urange->dst_fd);
4740 err |= get_user(range.pos_in, &urange->pos_in);
4741 err |= get_user(range.pos_out, &urange->pos_out);
4742 err |= get_user(range.len, &urange->len);
4746 return __f2fs_ioc_move_range(file, &range);
4749 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4751 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4753 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4757 case FS_IOC32_GETVERSION:
4758 cmd = FS_IOC_GETVERSION;
4760 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4761 return f2fs_compat_ioc_gc_range(file, arg);
4762 case F2FS_IOC32_MOVE_RANGE:
4763 return f2fs_compat_ioc_move_range(file, arg);
4764 case F2FS_IOC_START_ATOMIC_WRITE:
4765 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4766 case F2FS_IOC_START_VOLATILE_WRITE:
4767 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4768 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4769 case F2FS_IOC_SHUTDOWN:
4771 case FS_IOC_SET_ENCRYPTION_POLICY:
4772 case FS_IOC_GET_ENCRYPTION_PWSALT:
4773 case FS_IOC_GET_ENCRYPTION_POLICY:
4774 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4775 case FS_IOC_ADD_ENCRYPTION_KEY:
4776 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4777 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4778 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4779 case FS_IOC_GET_ENCRYPTION_NONCE:
4780 case F2FS_IOC_GARBAGE_COLLECT:
4781 case F2FS_IOC_WRITE_CHECKPOINT:
4782 case F2FS_IOC_DEFRAGMENT:
4783 case F2FS_IOC_FLUSH_DEVICE:
4784 case F2FS_IOC_GET_FEATURES:
4785 case F2FS_IOC_GET_PIN_FILE:
4786 case F2FS_IOC_SET_PIN_FILE:
4787 case F2FS_IOC_PRECACHE_EXTENTS:
4788 case F2FS_IOC_RESIZE_FS:
4789 case FS_IOC_ENABLE_VERITY:
4790 case FS_IOC_MEASURE_VERITY:
4791 case FS_IOC_READ_VERITY_METADATA:
4792 case FS_IOC_GETFSLABEL:
4793 case FS_IOC_SETFSLABEL:
4794 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4795 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4796 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4797 case F2FS_IOC_SEC_TRIM_FILE:
4798 case F2FS_IOC_GET_COMPRESS_OPTION:
4799 case F2FS_IOC_SET_COMPRESS_OPTION:
4800 case F2FS_IOC_DECOMPRESS_FILE:
4801 case F2FS_IOC_COMPRESS_FILE:
4804 return -ENOIOCTLCMD;
4806 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4810 const struct file_operations f2fs_file_operations = {
4811 .llseek = f2fs_llseek,
4812 .read_iter = f2fs_file_read_iter,
4813 .write_iter = f2fs_file_write_iter,
4814 .open = f2fs_file_open,
4815 .release = f2fs_release_file,
4816 .mmap = f2fs_file_mmap,
4817 .flush = f2fs_file_flush,
4818 .fsync = f2fs_sync_file,
4819 .fallocate = f2fs_fallocate,
4820 .unlocked_ioctl = f2fs_ioctl,
4821 #ifdef CONFIG_COMPAT
4822 .compat_ioctl = f2fs_compat_ioctl,
4824 .splice_read = generic_file_splice_read,
4825 .splice_write = iter_file_splice_write,
4826 .fadvise = f2fs_file_fadvise,