linux - fs/f2fs/file.c

5911 lines · 142.3 KiB

1	// SPDX-License-Identifier: GPL-2.0
2	/*
3	* fs/f2fs/file.c
4	*
5	* Copyright (c) 2012 Samsung Electronics Co., Ltd.
6	* http://www.samsung.com/
7	*/
8	#include <linux/blk-crypto.h>
9	#include <linux/fs.h>
10	#include <linux/f2fs_fs.h>
11	#include <linux/stat.h>
12	#include <linux/writeback.h>
13	#include <linux/blkdev.h>
14	#include <linux/falloc.h>
15	#include <linux/filelock.h>
16	#include <linux/types.h>
17	#include <linux/compat.h>
18	#include <linux/uaccess.h>
19	#include <linux/mount.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>
28
29	#include "f2fs.h"
30	#include "node.h"
31	#include "segment.h"
32	#include "xattr.h"
33	#include "acl.h"
34	#include "gc.h"
35	#include "iostat.h"
36	#include <trace/events/f2fs.h>
37	#include <uapi/linux/f2fs.h>
38
39	static int fill_zero(struct inode *inode, pgoff_t index,
40	loff_t start, loff_t len);
41
42	static int do_zero_post_eof_page(struct inode *inode, loff_t new_size)
43	{
44	loff_t old_size = i_size_read(inode);
45	unsigned int offset, len;
46	pgoff_t index;
47	int err;
48
49	offset = old_size & (PAGE_SIZE - 1);
50
51	if (!offset)
52	return 0;
53
54	len = min_t(loff_t, PAGE_SIZE - offset, new_size - old_size);
55	index = old_size >> PAGE_SHIFT;
56
57	if (f2fs_has_inline_data(inode)) {
58	/* data post eof should be always zero */
59	if (new_size <= MAX_INLINE_DATA(inode))
60	return 0;
61	err = f2fs_convert_inline_inode(inode);
62	if (err)
63	return err;
64	}
65
66	err = fill_zero(inode, index, offset, len);
67	if (err)
68	return err;
69	return filemap_write_and_wait_range(inode->i_mapping,
70	old_size, old_size + len - 1);
71	}
72
73	static int f2fs_zero_post_eof_page(struct inode *inode,
74	loff_t new_size, bool lock, bool writeback)
75	{
76	loff_t old_size = i_size_read(inode);
77	bool strict =
78	F2FS_OPTION(F2FS_I_SB(inode)).fsync_mode == FSYNC_MODE_STRICT;
79
80	if (old_size >= new_size)
81	return 0;
82
83	if (!strict && mapping_empty(inode->i_mapping))
84	return 0;
85
86	if (lock)
87	filemap_invalidate_lock(inode->i_mapping);
88	/* zero or drop pages only in range of [old_size, new_size] */
89	truncate_inode_pages_range(inode->i_mapping, old_size, new_size - 1);
90	if (lock)
91	filemap_invalidate_unlock(inode->i_mapping);
92
93	if (!writeback \|\| !strict)
94	return 0;
95	/*
96	* In fsync_mode=strict, when we expand an unaligned EOF size, we
97	* should zero post EOF data and writeback the data immediately,
98	* so that it can avoid exposing stale data after metadata flush
99	* and POR.
100	*/
101	return do_zero_post_eof_page(inode, new_size);
102	}
103
104	static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
105	{
106	struct inode *inode = file_inode(vmf->vma->vm_file);
107	vm_flags_t flags = vmf->vma->vm_flags;
108	vm_fault_t ret;
109
110	ret = filemap_fault(vmf);
111	if (ret & VM_FAULT_LOCKED)
112	f2fs_update_iostat(F2FS_I_SB(inode), inode,
113	APP_MAPPED_READ_IO, F2FS_BLKSIZE);
114
115	trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret);
116
117	return ret;
118	}
119
120	static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
121	{
122	struct folio *folio = page_folio(vmf->page);
123	struct inode *inode = file_inode(vmf->vma->vm_file);
124	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
125	struct dnode_of_data dn;
126	bool need_alloc = !f2fs_is_pinned_file(inode);
127	int err = 0;
128	vm_fault_t ret;
129
130	/*
131	* We only support large folio on the read case.
132	* Don't make any dirty pages.
133	*/
134	if (unlikely(IS_IMMUTABLE(inode)) \|\|
135	mapping_large_folio_support(inode->i_mapping)) {
136	f2fs_err(sbi, "Not expected: immutable: %d large_folio: %d",
137	IS_IMMUTABLE(inode),
138	mapping_large_folio_support(inode->i_mapping));
139	return VM_FAULT_SIGBUS;
140	}
141
142	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
143	err = -EIO;
144	goto out;
145	}
146
147	if (unlikely(f2fs_cp_error(sbi))) {
148	err = -EIO;
149	goto out;
150	}
151
152	if (!f2fs_is_checkpoint_ready(sbi)) {
153	err = -ENOSPC;
154	goto out;
155	}
156
157	err = f2fs_convert_inline_inode(inode);
158	if (err)
159	goto out;
160
161	#ifdef CONFIG_F2FS_FS_COMPRESSION
162	if (f2fs_compressed_file(inode)) {
163	int ret = f2fs_is_compressed_cluster(inode, folio->index);
164
165	if (ret < 0) {
166	err = ret;
167	goto out;
168	} else if (ret) {
169	need_alloc = false;
170	}
171	}
172	#endif
173	/* should do out of any locked page */
174	if (need_alloc)
175	f2fs_balance_fs(sbi, true);
176
177	sb_start_pagefault(inode->i_sb);
178
179	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
180
181	err = f2fs_zero_post_eof_page(inode,
182	(folio->index + 1) << PAGE_SHIFT, true, false);
183	if (err)
184	goto out_pagefault;
185
186	file_update_time(vmf->vma->vm_file);
187	filemap_invalidate_lock_shared(inode->i_mapping);
188
189	folio_lock(folio);
190	if (unlikely(folio->mapping != inode->i_mapping \|\|
191	folio_pos(folio) > i_size_read(inode) \|\|
192	!folio_test_uptodate(folio))) {
193	folio_unlock(folio);
194	err = -EFAULT;
195	goto out_sem;
196	}
197
198	set_new_dnode(&dn, inode, NULL, NULL, 0);
199	if (need_alloc) {
200	/* block allocation */
201	err = f2fs_get_block_locked(&dn, folio->index);
202	} else {
203	err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE);
204	f2fs_put_dnode(&dn);
205	if (f2fs_is_pinned_file(inode) &&
206	!__is_valid_data_blkaddr(dn.data_blkaddr))
207	err = -EIO;
208	}
209
210	if (err) {
211	folio_unlock(folio);
212	goto out_sem;
213	}
214
215	f2fs_folio_wait_writeback(folio, DATA, false, true);
216
217	/* wait for GCed page writeback via META_MAPPING */
218	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
219
220	/*
221	* check to see if the page is mapped already (no holes)
222	*/
223	if (folio_test_mappedtodisk(folio))
224	goto out_sem;
225
226	/* page is wholly or partially inside EOF */
227	if (((loff_t)(folio->index + 1) << PAGE_SHIFT) >
228	i_size_read(inode)) {
229	loff_t offset;
230
231	offset = i_size_read(inode) & ~PAGE_MASK;
232	folio_zero_segment(folio, offset, folio_size(folio));
233	}
234	folio_mark_dirty(folio);
235
236	f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
237	f2fs_update_time(sbi, REQ_TIME);
238
239	out_sem:
240	filemap_invalidate_unlock_shared(inode->i_mapping);
241	out_pagefault:
242	sb_end_pagefault(inode->i_sb);
243	out:
244	ret = vmf_fs_error(err);
245
246	trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret);
247	return ret;
248	}
249
250	static const struct vm_operations_struct f2fs_file_vm_ops = {
251	.fault = f2fs_filemap_fault,
252	.map_pages = filemap_map_pages,
253	.page_mkwrite = f2fs_vm_page_mkwrite,
254	};
255
256	static int get_parent_ino(struct inode inode, nid_t pino)
257	{
258	struct dentry *dentry;
259
260	/*
261	* Make sure to get the non-deleted alias. The alias associated with
262	* the open file descriptor being fsync()'ed may be deleted already.
263	*/
264	dentry = d_find_alias(inode);
265	if (!dentry)
266	return 0;
267
268	*pino = d_parent_ino(dentry);
269	dput(dentry);
270	return 1;
271	}
272
273	static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
274	{
275	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
276	enum cp_reason_type cp_reason = CP_NO_NEEDED;
277
278	if (!S_ISREG(inode->i_mode))
279	cp_reason = CP_NON_REGULAR;
280	else if (f2fs_compressed_file(inode))
281	cp_reason = CP_COMPRESSED;
282	else if (inode->i_nlink != 1)
283	cp_reason = CP_HARDLINK;
284	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
285	cp_reason = CP_SB_NEED_CP;
286	else if (file_wrong_pino(inode))
287	cp_reason = CP_WRONG_PINO;
288	else if (!f2fs_space_for_roll_forward(sbi))
289	cp_reason = CP_NO_SPC_ROLL;
290	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
291	cp_reason = CP_NODE_NEED_CP;
292	else if (test_opt(sbi, FASTBOOT))
293	cp_reason = CP_FASTBOOT_MODE;
294	else if (F2FS_OPTION(sbi).active_logs == 2)
295	cp_reason = CP_SPEC_LOG_NUM;
296	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
297	f2fs_need_dentry_mark(sbi, inode->i_ino) &&
298	f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
299	TRANS_DIR_INO))
300	cp_reason = CP_RECOVER_DIR;
301	else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
302	XATTR_DIR_INO))
303	cp_reason = CP_XATTR_DIR;
304
305	return cp_reason;
306	}
307
308	static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
309	{
310	struct folio *i = filemap_get_folio(NODE_MAPPING(sbi), ino);
311	bool ret = false;
312	/* But we need to avoid that there are some inode updates */
313	if ((!IS_ERR(i) && folio_test_dirty(i)) \|\|
314	f2fs_need_inode_block_update(sbi, ino))
315	ret = true;
316	f2fs_folio_put(i, false);
317	return ret;
318	}
319
320	static void try_to_fix_pino(struct inode *inode)
321	{
322	struct f2fs_inode_info *fi = F2FS_I(inode);
323	nid_t pino;
324
325	f2fs_down_write(&fi->i_sem);
326	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
327	get_parent_ino(inode, &pino)) {
328	f2fs_i_pino_write(inode, pino);
329	file_got_pino(inode);
330	}
331	f2fs_up_write(&fi->i_sem);
332	}
333
334	static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
335	int datasync, bool atomic)
336	{
337	struct inode *inode = file->f_mapping->host;
338	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
339	nid_t ino = inode->i_ino;
340	int ret = 0;
341	enum cp_reason_type cp_reason = 0;
342	struct writeback_control wbc = {
343	.sync_mode = WB_SYNC_ALL,
344	.nr_to_write = LONG_MAX,
345	};
346	unsigned int seq_id = 0;
347
348	if (unlikely(f2fs_readonly(inode->i_sb)))
349	return 0;
350
351	trace_f2fs_sync_file_enter(inode);
352
353	if (S_ISDIR(inode->i_mode))
354	goto go_write;
355
356	/* if fdatasync is triggered, let's do in-place-update */
357	if (datasync \|\| get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
358	set_inode_flag(inode, FI_NEED_IPU);
359	ret = file_write_and_wait_range(file, start, end);
360	clear_inode_flag(inode, FI_NEED_IPU);
361
362	if (ret \|\| is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
363	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
364	return ret;
365	}
366
367	/* if the inode is dirty, let's recover all the time */
368	if (!f2fs_skip_inode_update(inode, datasync)) {
369	f2fs_write_inode(inode, NULL);
370	goto go_write;
371	}
372
373	/*
374	* if there is no written data, don't waste time to write recovery info.
375	*/
376	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
377	!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
378
379	/* it may call write_inode just prior to fsync */
380	if (need_inode_page_update(sbi, ino))
381	goto go_write;
382
383	if (is_inode_flag_set(inode, FI_UPDATE_WRITE) \|\|
384	f2fs_exist_written_data(sbi, ino, UPDATE_INO))
385	goto flush_out;
386	goto out;
387	} else {
388	/*
389	* for OPU case, during fsync(), node can be persisted before
390	* data when lower device doesn't support write barrier, result
391	* in data corruption after SPO.
392	* So for strict fsync mode, force to use atomic write semantics
393	* to keep write order in between data/node and last node to
394	* avoid potential data corruption.
395	*/
396	if (F2FS_OPTION(sbi).fsync_mode ==
397	FSYNC_MODE_STRICT && !atomic)
398	atomic = true;
399	}
400	go_write:
401	/*
402	* Both of fdatasync() and fsync() are able to be recovered from
403	* sudden-power-off.
404	*/
405	f2fs_down_read(&F2FS_I(inode)->i_sem);
406	cp_reason = need_do_checkpoint(inode);
407	f2fs_up_read(&F2FS_I(inode)->i_sem);
408
409	if (cp_reason) {
410	/* all the dirty node pages should be flushed for POR */
411	ret = f2fs_sync_fs(inode->i_sb, 1);
412
413	/*
414	* We've secured consistency through sync_fs. Following pino
415	* will be used only for fsynced inodes after checkpoint.
416	*/
417	try_to_fix_pino(inode);
418	clear_inode_flag(inode, FI_APPEND_WRITE);
419	clear_inode_flag(inode, FI_UPDATE_WRITE);
420	goto out;
421	}
422	sync_nodes:
423	atomic_inc(&sbi->wb_sync_req[NODE]);
424	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
425	atomic_dec(&sbi->wb_sync_req[NODE]);
426	if (ret)
427	goto out;
428
429	/* if cp_error was enabled, we should avoid infinite loop */
430	if (unlikely(f2fs_cp_error(sbi))) {
431	ret = -EIO;
432	goto out;
433	}
434
435	if (f2fs_need_inode_block_update(sbi, ino)) {
436	f2fs_mark_inode_dirty_sync(inode, true);
437	f2fs_write_inode(inode, NULL);
438	goto sync_nodes;
439	}
440
441	/*
442	* If it's atomic_write, it's just fine to keep write ordering. So
443	* here we don't need to wait for node write completion, since we use
444	* node chain which serializes node blocks. If one of node writes are
445	* reordered, we can see simply broken chain, resulting in stopping
446	* roll-forward recovery. It means we'll recover all or none node blocks
447	* given fsync mark.
448	*/
449	if (!atomic) {
450	ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
451	if (ret)
452	goto out;
453	}
454
455	/* once recovery info is written, don't need to tack this */
456	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
457	clear_inode_flag(inode, FI_APPEND_WRITE);
458	flush_out:
459	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
460	ret = f2fs_issue_flush(sbi, inode->i_ino);
461	if (!ret) {
462	f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
463	clear_inode_flag(inode, FI_UPDATE_WRITE);
464	f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
465	}
466	f2fs_update_time(sbi, REQ_TIME);
467	out:
468	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
469	return ret;
470	}
471
472	int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
473	{
474	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
475	return -EIO;
476	return f2fs_do_sync_file(file, start, end, datasync, false);
477	}
478
479	static bool __found_offset(struct address_space *mapping,
480	struct dnode_of_data *dn, pgoff_t index, int whence)
481	{
482	block_t blkaddr = f2fs_data_blkaddr(dn);
483	struct inode *inode = mapping->host;
484	bool compressed_cluster = false;
485
486	if (f2fs_compressed_file(inode)) {
487	block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio,
488	ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size));
489
490	compressed_cluster = first_blkaddr == COMPRESS_ADDR;
491	}
492
493	switch (whence) {
494	case SEEK_DATA:
495	if (__is_valid_data_blkaddr(blkaddr))
496	return true;
497	if (blkaddr == NEW_ADDR &&
498	xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
499	return true;
500	if (compressed_cluster)
501	return true;
502	break;
503	case SEEK_HOLE:
504	if (compressed_cluster)
505	return false;
506	if (blkaddr == NULL_ADDR)
507	return true;
508	break;
509	}
510	return false;
511	}
512
513	static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
514	{
515	struct inode *inode = file->f_mapping->host;
516	loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
517	struct dnode_of_data dn;
518	pgoff_t pgofs, end_offset;
519	loff_t data_ofs = offset;
520	loff_t isize;
521	int err = 0;
522
523	inode_lock_shared(inode);
524
525	isize = i_size_read(inode);
526	if (offset >= isize)
527	goto fail;
528
529	/* handle inline data case */
530	if (f2fs_has_inline_data(inode)) {
531	if (whence == SEEK_HOLE) {
532	data_ofs = isize;
533	goto found;
534	} else if (whence == SEEK_DATA) {
535	data_ofs = offset;
536	goto found;
537	}
538	}
539
540	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
541
542	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
543	set_new_dnode(&dn, inode, NULL, NULL, 0);
544	err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
545	if (err && err != -ENOENT) {
546	goto fail;
547	} else if (err == -ENOENT) {
548	/* direct node does not exists */
549	if (whence == SEEK_DATA) {
550	pgofs = f2fs_get_next_page_offset(&dn, pgofs);
551	continue;
552	} else {
553	goto found;
554	}
555	}
556
557	end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
558
559	/* find data/hole in dnode block */
560	for (; dn.ofs_in_node < end_offset;
561	dn.ofs_in_node++, pgofs++,
562	data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
563	block_t blkaddr;
564
565	blkaddr = f2fs_data_blkaddr(&dn);
566
567	if (__is_valid_data_blkaddr(blkaddr) &&
568	!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
569	blkaddr, DATA_GENERIC_ENHANCE)) {
570	f2fs_put_dnode(&dn);
571	goto fail;
572	}
573
574	if (__found_offset(file->f_mapping, &dn,
575	pgofs, whence)) {
576	f2fs_put_dnode(&dn);
577	goto found;
578	}
579	}
580	f2fs_put_dnode(&dn);
581	}
582
583	if (whence == SEEK_DATA)
584	goto fail;
585	found:
586	if (whence == SEEK_HOLE && data_ofs > isize)
587	data_ofs = isize;
588	inode_unlock_shared(inode);
589	return vfs_setpos(file, data_ofs, maxbytes);
590	fail:
591	inode_unlock_shared(inode);
592	return -ENXIO;
593	}
594
595	static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
596	{
597	struct inode *inode = file->f_mapping->host;
598	loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
599
600	switch (whence) {
601	case SEEK_SET:
602	case SEEK_CUR:
603	case SEEK_END:
604	return generic_file_llseek_size(file, offset, whence,
605	maxbytes, i_size_read(inode));
606	case SEEK_DATA:
607	case SEEK_HOLE:
608	if (offset < 0)
609	return -ENXIO;
610	return f2fs_seek_block(file, offset, whence);
611	}
612
613	return -EINVAL;
614	}
615
616	static int f2fs_file_mmap_prepare(struct vm_area_desc *desc)
617	{
618	struct file *file = desc->file;
619	struct inode *inode = file_inode(file);
620
621	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
622	return -EIO;
623
624	if (!f2fs_is_compress_backend_ready(inode))
625	return -EOPNOTSUPP;
626
627	file_accessed(file);
628	desc->vm_ops = &f2fs_file_vm_ops;
629
630	f2fs_down_read(&F2FS_I(inode)->i_sem);
631	set_inode_flag(inode, FI_MMAP_FILE);
632	f2fs_up_read(&F2FS_I(inode)->i_sem);
633
634	return 0;
635	}
636
637	static int finish_preallocate_blocks(struct inode *inode)
638	{
639	int ret = 0;
640	bool opened;
641
642	f2fs_down_read(&F2FS_I(inode)->i_sem);
643	opened = is_inode_flag_set(inode, FI_OPENED_FILE);
644	f2fs_up_read(&F2FS_I(inode)->i_sem);
645	if (opened)
646	return 0;
647
648	inode_lock(inode);
649	if (is_inode_flag_set(inode, FI_OPENED_FILE))
650	goto out_unlock;
651
652	if (!file_should_truncate(inode))
653	goto out_update;
654
655	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
656	filemap_invalidate_lock(inode->i_mapping);
657
658	truncate_setsize(inode, i_size_read(inode));
659	ret = f2fs_truncate(inode);
660
661	filemap_invalidate_unlock(inode->i_mapping);
662	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
663	if (ret)
664	goto out_unlock;
665
666	file_dont_truncate(inode);
667	out_update:
668	f2fs_down_write(&F2FS_I(inode)->i_sem);
669	set_inode_flag(inode, FI_OPENED_FILE);
670	f2fs_up_write(&F2FS_I(inode)->i_sem);
671	out_unlock:
672	inode_unlock(inode);
673	return ret;
674	}
675
676	static int f2fs_file_open(struct inode inode, struct file filp)
677	{
678	int err = fscrypt_file_open(inode, filp);
679
680	if (err)
681	return err;
682
683	if (!f2fs_is_compress_backend_ready(inode))
684	return -EOPNOTSUPP;
685
686	if (mapping_large_folio_support(inode->i_mapping) &&
687	filp->f_mode & FMODE_WRITE)
688	return -EOPNOTSUPP;
689
690	err = fsverity_file_open(inode, filp);
691	if (err)
692	return err;
693
694	filp->f_mode \|= FMODE_NOWAIT;
695	filp->f_mode \|= FMODE_CAN_ODIRECT;
696
697	err = dquot_file_open(inode, filp);
698	if (err)
699	return err;
700
701	err = finish_preallocate_blocks(inode);
702	if (!err)
703	atomic_inc(&F2FS_I(inode)->open_count);
704	return err;
705	}
706
707	void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
708	{
709	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
710	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
711	__le32 *addr;
712	bool compressed_cluster = false;
713	int cluster_index = 0, valid_blocks = 0;
714	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
715	bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
716	block_t blkstart;
717	int blklen = 0;
718
719	addr = get_dnode_addr(dn->inode, dn->node_folio) + ofs;
720	blkstart = le32_to_cpu(*addr);
721
722	/* Assumption: truncation starts with cluster */
723	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
724	block_t blkaddr = le32_to_cpu(*addr);
725
726	if (f2fs_compressed_file(dn->inode) &&
727	!(cluster_index & (cluster_size - 1))) {
728	if (compressed_cluster)
729	f2fs_i_compr_blocks_update(dn->inode,
730	valid_blocks, false);
731	compressed_cluster = (blkaddr == COMPRESS_ADDR);
732	valid_blocks = 0;
733	}
734
735	if (blkaddr == NULL_ADDR)
736	goto next;
737
738	f2fs_set_data_blkaddr(dn, NULL_ADDR);
739
740	if (__is_valid_data_blkaddr(blkaddr)) {
741	if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE))
742	goto next;
743	if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr,
744	DATA_GENERIC_ENHANCE))
745	goto next;
746	if (compressed_cluster)
747	valid_blocks++;
748	}
749
750	if (blkstart + blklen == blkaddr) {
751	blklen++;
752	} else {
753	f2fs_invalidate_blocks(sbi, blkstart, blklen);
754	blkstart = blkaddr;
755	blklen = 1;
756	}
757
758	if (!released \|\| blkaddr != COMPRESS_ADDR)
759	nr_free++;
760
761	continue;
762
763	next:
764	if (blklen)
765	f2fs_invalidate_blocks(sbi, blkstart, blklen);
766
767	blkstart = le32_to_cpu(*(addr + 1));
768	blklen = 0;
769	}
770
771	if (blklen)
772	f2fs_invalidate_blocks(sbi, blkstart, blklen);
773
774	if (compressed_cluster)
775	f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
776
777	if (nr_free) {
778	pgoff_t fofs;
779	/*
780	* once we invalidate valid blkaddr in range [ofs, ofs + count],
781	* we will invalidate all blkaddr in the whole range.
782	*/
783	fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_folio),
784	dn->inode) + ofs;
785	f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
786	f2fs_update_age_extent_cache_range(dn, fofs, len);
787	dec_valid_block_count(sbi, dn->inode, nr_free);
788	}
789	dn->ofs_in_node = ofs;
790
791	f2fs_update_time(sbi, REQ_TIME);
792	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
793	dn->ofs_in_node, nr_free);
794	}
795
796	static int truncate_partial_data_page(struct inode *inode, u64 from,
797	bool cache_only)
798	{
799	loff_t offset = from & (PAGE_SIZE - 1);
800	pgoff_t index = from >> PAGE_SHIFT;
801	struct address_space *mapping = inode->i_mapping;
802	struct folio *folio;
803
804	if (!offset && !cache_only)
805	return 0;
806
807	if (cache_only) {
808	folio = filemap_lock_folio(mapping, index);
809	if (IS_ERR(folio))
810	return 0;
811	if (folio_test_uptodate(folio))
812	goto truncate_out;
813	f2fs_folio_put(folio, true);
814	return 0;
815	}
816
817	folio = f2fs_get_lock_data_folio(inode, index, true);
818	if (IS_ERR(folio))
819	return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio);
820	truncate_out:
821	f2fs_folio_wait_writeback(folio, DATA, true, true);
822	folio_zero_segment(folio, offset, folio_size(folio));
823
824	/* An encrypted inode should have a key and truncate the last page. */
825	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
826	if (!cache_only)
827	folio_mark_dirty(folio);
828	f2fs_folio_put(folio, true);
829	return 0;
830	}
831
832	int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
833	{
834	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
835	struct dnode_of_data dn;
836	struct f2fs_lock_context lc;
837	pgoff_t free_from;
838	int count = 0, err = 0;
839	struct folio *ifolio;
840	bool truncate_page = false;
841
842	trace_f2fs_truncate_blocks_enter(inode, from);
843
844	if (IS_DEVICE_ALIASING(inode) && from) {
845	err = -EINVAL;
846	goto out_err;
847	}
848
849	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
850
851	if (free_from >= max_file_blocks(inode))
852	goto free_partial;
853
854	if (lock)
855	f2fs_lock_op(sbi, &lc);
856
857	ifolio = f2fs_get_inode_folio(sbi, inode->i_ino);
858	if (IS_ERR(ifolio)) {
859	err = PTR_ERR(ifolio);
860	goto out;
861	}
862
863	if (IS_DEVICE_ALIASING(inode)) {
864	struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
865	struct extent_info ei;
866
867	read_lock(&et->lock);
868	ei = et->largest;
869	read_unlock(&et->lock);
870
871	f2fs_invalidate_blocks(sbi, ei.blk, ei.len);
872
873	dec_valid_block_count(sbi, inode, ei.len);
874	f2fs_update_time(sbi, REQ_TIME);
875
876	f2fs_drop_extent_tree(inode);
877
878	f2fs_folio_put(ifolio, true);
879	goto out;
880	}
881
882	if (f2fs_has_inline_data(inode)) {
883	f2fs_truncate_inline_inode(inode, ifolio, from);
884	f2fs_folio_put(ifolio, true);
885	truncate_page = true;
886	goto out;
887	}
888
889	set_new_dnode(&dn, inode, ifolio, NULL, 0);
890	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
891	if (err) {
892	if (err == -ENOENT)
893	goto free_next;
894	goto out;
895	}
896
897	count = ADDRS_PER_PAGE(dn.node_folio, inode);
898
899	count -= dn.ofs_in_node;
900	f2fs_bug_on(sbi, count < 0);
901
902	if (dn.ofs_in_node \|\| IS_INODE(dn.node_folio)) {
903	f2fs_truncate_data_blocks_range(&dn, count);
904	free_from += count;
905	}
906
907	f2fs_put_dnode(&dn);
908	free_next:
909	err = f2fs_truncate_inode_blocks(inode, free_from);
910	out:
911	if (lock)
912	f2fs_unlock_op(sbi, &lc);
913	free_partial:
914	/* lastly zero out the first data page */
915	if (!err)
916	err = truncate_partial_data_page(inode, from, truncate_page);
917	out_err:
918	trace_f2fs_truncate_blocks_exit(inode, err);
919	return err;
920	}
921
922	int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
923	{
924	u64 free_from = from;
925	int err;
926
927	#ifdef CONFIG_F2FS_FS_COMPRESSION
928	/*
929	* for compressed file, only support cluster size
930	* aligned truncation.
931	*/
932	if (f2fs_compressed_file(inode))
933	free_from = round_up(from,
934	F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
935	#endif
936
937	err = f2fs_do_truncate_blocks(inode, free_from, lock);
938	if (err)
939	return err;
940
941	#ifdef CONFIG_F2FS_FS_COMPRESSION
942	/*
943	* For compressed file, after release compress blocks, don't allow write
944	* direct, but we should allow write direct after truncate to zero.
945	*/
946	if (f2fs_compressed_file(inode) && !free_from
947	&& is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
948	clear_inode_flag(inode, FI_COMPRESS_RELEASED);
949
950	if (from != free_from) {
951	err = f2fs_truncate_partial_cluster(inode, from, lock);
952	if (err)
953	return err;
954	}
955	#endif
956
957	return 0;
958	}
959
960	int f2fs_truncate(struct inode *inode)
961	{
962	int err;
963
964	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
965	return -EIO;
966
967	if (!(S_ISREG(inode->i_mode) \|\| S_ISDIR(inode->i_mode) \|\|
968	S_ISLNK(inode->i_mode)))
969	return 0;
970
971	trace_f2fs_truncate(inode);
972
973	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
974	return -EIO;
975
976	err = f2fs_dquot_initialize(inode);
977	if (err)
978	return err;
979
980	/* we should check inline_data size */
981	if (!f2fs_may_inline_data(inode)) {
982	err = f2fs_convert_inline_inode(inode);
983	if (err) {
984	/*
985	* Always truncate page #0 to avoid page cache
986	* leak in evict() path.
987	*/
988	truncate_inode_pages_range(inode->i_mapping,
989	F2FS_BLK_TO_BYTES(0),
990	F2FS_BLK_END_BYTES(0));
991	return err;
992	}
993	}
994
995	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
996	if (err)
997	return err;
998
999	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1000	f2fs_mark_inode_dirty_sync(inode, false);
1001	return 0;
1002	}
1003
1004	static bool f2fs_force_buffered_io(struct inode *inode, int rw)
1005	{
1006	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1007
1008	if (fsverity_active(inode))
1009	return true;
1010	if (f2fs_compressed_file(inode))
1011	return true;
1012	/*
1013	* only force direct read to use buffered IO, for direct write,
1014	* it expects inline data conversion before committing IO.
1015	*/
1016	if (f2fs_has_inline_data(inode) && rw == READ)
1017	return true;
1018
1019	/* disallow direct IO if any of devices has unaligned blksize */
1020	if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
1021	return true;
1022	/*
1023	* for blkzoned device, fallback direct IO to buffered IO, so
1024	* all IOs can be serialized by log-structured write.
1025	*/
1026	if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) &&
1027	!f2fs_is_pinned_file(inode))
1028	return true;
1029	if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
1030	return true;
1031
1032	return false;
1033	}
1034
1035	int f2fs_getattr(const struct mnt_idmap idmap, const struct path path,
1036	struct kstat *stat, u32 request_mask, unsigned int query_flags)
1037	{
1038	struct inode *inode = d_inode(path->dentry);
1039	struct f2fs_inode_info *fi = F2FS_I(inode);
1040	struct f2fs_inode *ri = NULL;
1041	unsigned int flags;
1042
1043	if (f2fs_has_extra_attr(inode) &&
1044	f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
1045	F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
1046	stat->result_mask \|= STATX_BTIME;
1047	stat->btime.tv_sec = fi->i_crtime.tv_sec;
1048	stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
1049	}
1050
1051	/*
1052	* Return the DIO alignment restrictions if requested.
1053	*
1054	* f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN
1055	* cannot represent that, so in that case we report no DIO support.
1056	*/
1057	if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
1058	unsigned int bsize = i_blocksize(inode);
1059
1060	stat->result_mask \|= STATX_DIOALIGN;
1061	if (!f2fs_force_buffered_io(inode, WRITE)) {
1062	stat->dio_mem_align = bsize;
1063	stat->dio_offset_align = bsize;
1064	}
1065	}
1066
1067	flags = fi->i_flags;
1068	if (flags & F2FS_COMPR_FL)
1069	stat->attributes \|= STATX_ATTR_COMPRESSED;
1070	if (flags & F2FS_APPEND_FL)
1071	stat->attributes \|= STATX_ATTR_APPEND;
1072	if (IS_ENCRYPTED(inode))
1073	stat->attributes \|= STATX_ATTR_ENCRYPTED;
1074	if (flags & F2FS_IMMUTABLE_FL)
1075	stat->attributes \|= STATX_ATTR_IMMUTABLE;
1076	if (flags & F2FS_NODUMP_FL)
1077	stat->attributes \|= STATX_ATTR_NODUMP;
1078
1079	stat->attributes_mask \|= (STATX_ATTR_COMPRESSED \|
1080	STATX_ATTR_APPEND \|
1081	STATX_ATTR_ENCRYPTED \|
1082	STATX_ATTR_IMMUTABLE \|
1083	STATX_ATTR_NODUMP);
1084
1085	generic_fillattr(idmap, request_mask, inode, stat);
1086
1087	/* we need to show initial sectors used for inline_data/dentries */
1088	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) \|\|
1089	f2fs_has_inline_dentry(inode))
1090	stat->blocks += (stat->size + 511) >> 9;
1091
1092	return 0;
1093	}
1094
1095	#ifdef CONFIG_F2FS_FS_POSIX_ACL
1096	static void __setattr_copy(const struct mnt_idmap *idmap,
1097	struct inode inode, const struct iattr attr)
1098	{
1099	unsigned int ia_valid = attr->ia_valid;
1100
1101	i_uid_update(idmap, attr, inode);
1102	i_gid_update(idmap, attr, inode);
1103	if (ia_valid & ATTR_ATIME)
1104	inode_set_atime_to_ts(inode, attr->ia_atime);
1105	if (ia_valid & ATTR_MTIME)
1106	inode_set_mtime_to_ts(inode, attr->ia_mtime);
1107	if (ia_valid & ATTR_CTIME)
1108	inode_set_ctime_to_ts(inode, attr->ia_ctime);
1109	if (ia_valid & ATTR_MODE) {
1110	umode_t mode = attr->ia_mode;
1111
1112	if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode)))
1113	mode &= ~S_ISGID;
1114	set_acl_inode(inode, mode);
1115	}
1116	}
1117	#else
1118	#define __setattr_copy setattr_copy
1119	#endif
1120
1121	int f2fs_setattr(const struct mnt_idmap idmap, struct dentry dentry,
1122	struct iattr *attr)
1123	{
1124	struct inode *inode = d_inode(dentry);
1125	struct f2fs_inode_info *fi = F2FS_I(inode);
1126	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1127	int err;
1128
1129	if (unlikely(f2fs_cp_error(sbi)))
1130	return -EIO;
1131
1132	err = setattr_prepare(idmap, dentry, attr);
1133	if (err)
1134	return err;
1135
1136	err = fscrypt_prepare_setattr(dentry, attr);
1137	if (err)
1138	return err;
1139
1140	if (unlikely(IS_IMMUTABLE(inode)))
1141	return -EPERM;
1142
1143	if (unlikely(IS_APPEND(inode) &&
1144	(attr->ia_valid & (ATTR_MODE \| ATTR_UID \|
1145	ATTR_GID \| ATTR_TIMES_SET))))
1146	return -EPERM;
1147
1148	if ((attr->ia_valid & ATTR_SIZE)) {
1149	if (mapping_large_folio_support(inode->i_mapping))
1150	return -EOPNOTSUPP;
1151	if (IS_DEVICE_ALIASING(inode))
1152	return -EPERM;
1153	if (!f2fs_is_compress_backend_ready(inode))
1154	return -EOPNOTSUPP;
1155	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) &&
1156	!IS_ALIGNED(attr->ia_size,
1157	F2FS_BLK_TO_BYTES(fi->i_cluster_size)))
1158	return -EINVAL;
1159
1160	if (f2fs_is_pinned_file(inode)) {
1161	/*
1162	* It may break section-aligned fallocate recovery
1163	* mechanism, so do not allow larger size truncation.
1164	*/
1165	if (attr->ia_size > i_size_read(inode))
1166	return -EINVAL;
1167	/*
1168	* To prevent scattered pin block generation, we don't
1169	* allow smaller/equal size unaligned truncation for
1170	* pinned file.
1171	*/
1172	else if (!IS_ALIGNED(attr->ia_size,
1173	F2FS_BLK_TO_BYTES(CAP_BLKS_PER_SEC(sbi))))
1174	return -EINVAL;
1175	}
1176	}
1177
1178	if (is_quota_modification(idmap, inode, attr)) {
1179	err = f2fs_dquot_initialize(inode);
1180	if (err)
1181	return err;
1182	}
1183	if (i_uid_needs_update(idmap, attr, inode) \|\|
1184	i_gid_needs_update(idmap, attr, inode)) {
1185	struct f2fs_lock_context lc;
1186
1187	f2fs_lock_op(sbi, &lc);
1188	err = dquot_transfer(idmap, inode, attr);
1189	if (err) {
1190	set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
1191	f2fs_unlock_op(sbi, &lc);
1192	return err;
1193	}
1194	/*
1195	* update uid/gid under lock_op(), so that dquot and inode can
1196	* be updated atomically.
1197	*/
1198	i_uid_update(idmap, attr, inode);
1199	i_gid_update(idmap, attr, inode);
1200	f2fs_mark_inode_dirty_sync(inode, true);
1201	f2fs_unlock_op(sbi, &lc);
1202	}
1203
1204	if (attr->ia_valid & ATTR_SIZE) {
1205	loff_t old_size = i_size_read(inode);
1206
1207	if (attr->ia_size > MAX_INLINE_DATA(inode)) {
1208	/*
1209	* should convert inline inode before i_size_write to
1210	* keep smaller than inline_data size with inline flag.
1211	*/
1212	err = f2fs_convert_inline_inode(inode);
1213	if (err)
1214	return err;
1215	}
1216
1217	/*
1218	* wait for inflight dio, blocks should be removed after
1219	* IO completion.
1220	*/
1221	if (attr->ia_size < old_size)
1222	inode_dio_wait(inode);
1223
1224	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
1225	filemap_invalidate_lock(inode->i_mapping);
1226
1227	if (attr->ia_size > old_size) {
1228	err = f2fs_zero_post_eof_page(inode,
1229	attr->ia_size, false, true);
1230	if (err)
1231	goto err_out;
1232	}
1233	truncate_setsize(inode, attr->ia_size);
1234
1235	if (attr->ia_size <= old_size)
1236	err = f2fs_truncate(inode);
1237	/*
1238	* do not trim all blocks after i_size if target size is
1239	* larger than i_size.
1240	*/
1241	err_out:
1242	filemap_invalidate_unlock(inode->i_mapping);
1243	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
1244	if (err)
1245	return err;
1246
1247	spin_lock(&fi->i_size_lock);
1248	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1249	fi->last_disk_size = i_size_read(inode);
1250	spin_unlock(&fi->i_size_lock);
1251	}
1252
1253	__setattr_copy(idmap, inode, attr);
1254
1255	if (attr->ia_valid & ATTR_MODE) {
1256	err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
1257
1258	if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1259	if (!err)
1260	inode->i_mode = fi->i_acl_mode;
1261	clear_inode_flag(inode, FI_ACL_MODE);
1262	}
1263	}
1264
1265	/* file size may changed here */
1266	f2fs_mark_inode_dirty_sync(inode, true);
1267
1268	/* inode change will produce dirty node pages flushed by checkpoint */
1269	f2fs_balance_fs(sbi, true);
1270
1271	return err;
1272	}
1273
1274	const struct inode_operations f2fs_file_inode_operations = {
1275	.getattr = f2fs_getattr,
1276	.setattr = f2fs_setattr,
1277	.get_inode_acl = f2fs_get_acl,
1278	.set_acl = f2fs_set_acl,
1279	.listxattr = f2fs_listxattr,
1280	.fiemap = f2fs_fiemap,
1281	.fileattr_get = f2fs_fileattr_get,
1282	.fileattr_set = f2fs_fileattr_set,
1283	};
1284
1285	static int fill_zero(struct inode *inode, pgoff_t index,
1286	loff_t start, loff_t len)
1287	{
1288	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1289	struct folio *folio;
1290	struct f2fs_lock_context lc;
1291
1292	if (!len)
1293	return 0;
1294
1295	f2fs_balance_fs(sbi, true);
1296
1297	f2fs_lock_op(sbi, &lc);
1298	folio = f2fs_get_new_data_folio(inode, NULL, index, false);
1299	f2fs_unlock_op(sbi, &lc);
1300
1301	if (IS_ERR(folio))
1302	return PTR_ERR(folio);
1303
1304	f2fs_folio_wait_writeback(folio, DATA, true, true);
1305	folio_zero_range(folio, start, len);
1306	folio_mark_dirty(folio);
1307	f2fs_folio_put(folio, true);
1308	return 0;
1309	}
1310
1311	int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1312	{
1313	int err;
1314
1315	while (pg_start < pg_end) {
1316	struct dnode_of_data dn;
1317	pgoff_t end_offset, count;
1318
1319	set_new_dnode(&dn, inode, NULL, NULL, 0);
1320	err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1321	if (err) {
1322	if (err == -ENOENT) {
1323	pg_start = f2fs_get_next_page_offset(&dn,
1324	pg_start);
1325	continue;
1326	}
1327	return err;
1328	}
1329
1330	end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
1331	count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1332
1333	f2fs_bug_on(F2FS_I_SB(inode), count == 0 \|\| count > end_offset);
1334
1335	f2fs_truncate_data_blocks_range(&dn, count);
1336	f2fs_put_dnode(&dn);
1337
1338	pg_start += count;
1339	}
1340	return 0;
1341	}
1342
1343	static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
1344	{
1345	pgoff_t pg_start, pg_end;
1346	loff_t off_start, off_end;
1347	int ret;
1348
1349	ret = f2fs_convert_inline_inode(inode);
1350	if (ret)
1351	return ret;
1352
1353	ret = f2fs_zero_post_eof_page(inode, offset + len, true, false);
1354	if (ret)
1355	return ret;
1356
1357	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1358	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1359
1360	off_start = offset & (PAGE_SIZE - 1);
1361	off_end = (offset + len) & (PAGE_SIZE - 1);
1362
1363	if (pg_start == pg_end) {
1364	ret = fill_zero(inode, pg_start, off_start,
1365	off_end - off_start);
1366	if (ret)
1367	return ret;
1368	} else {
1369	if (off_start) {
1370	ret = fill_zero(inode, pg_start++, off_start,
1371	PAGE_SIZE - off_start);
1372	if (ret)
1373	return ret;
1374	}
1375	if (off_end) {
1376	ret = fill_zero(inode, pg_end, 0, off_end);
1377	if (ret)
1378	return ret;
1379	}
1380
1381	if (pg_start < pg_end) {
1382	loff_t blk_start, blk_end;
1383	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1384	struct f2fs_lock_context lc;
1385
1386	f2fs_balance_fs(sbi, true);
1387
1388	blk_start = (loff_t)pg_start << PAGE_SHIFT;
1389	blk_end = (loff_t)pg_end << PAGE_SHIFT;
1390
1391	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1392	filemap_invalidate_lock(inode->i_mapping);
1393
1394	truncate_pagecache_range(inode, blk_start, blk_end - 1);
1395
1396	f2fs_lock_op(sbi, &lc);
1397	ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1398	f2fs_unlock_op(sbi, &lc);
1399
1400	filemap_invalidate_unlock(inode->i_mapping);
1401	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1402	}
1403	}
1404
1405	return ret;
1406	}
1407
1408	static int __read_out_blkaddrs(struct inode inode, block_t blkaddr,
1409	int *do_replace, pgoff_t off, pgoff_t len)
1410	{
1411	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1412	struct dnode_of_data dn;
1413	int ret, done, i;
1414
1415	next_dnode:
1416	set_new_dnode(&dn, inode, NULL, NULL, 0);
1417	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1418	if (ret && ret != -ENOENT) {
1419	return ret;
1420	} else if (ret == -ENOENT) {
1421	if (dn.max_level == 0)
1422	return -ENOENT;
1423	done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1424	dn.ofs_in_node, len);
1425	blkaddr += done;
1426	do_replace += done;
1427	goto next;
1428	}
1429
1430	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_folio, inode) -
1431	dn.ofs_in_node, len);
1432	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1433	*blkaddr = f2fs_data_blkaddr(&dn);
1434
1435	if (__is_valid_data_blkaddr(*blkaddr) &&
1436	!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1437	DATA_GENERIC_ENHANCE)) {
1438	f2fs_put_dnode(&dn);
1439	return -EFSCORRUPTED;
1440	}
1441
1442	if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1443
1444	if (f2fs_lfs_mode(sbi)) {
1445	f2fs_put_dnode(&dn);
1446	return -EOPNOTSUPP;
1447	}
1448
1449	/* do not invalidate this block address */
1450	f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1451	*do_replace = 1;
1452	}
1453	}
1454	f2fs_put_dnode(&dn);
1455	next:
1456	len -= done;
1457	off += done;
1458	if (len)
1459	goto next_dnode;
1460	return 0;
1461	}
1462
1463	static int __roll_back_blkaddrs(struct inode inode, block_t blkaddr,
1464	int *do_replace, pgoff_t off, int len)
1465	{
1466	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1467	struct dnode_of_data dn;
1468	int ret, i;
1469
1470	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1471	if (*do_replace == 0)
1472	continue;
1473
1474	set_new_dnode(&dn, inode, NULL, NULL, 0);
1475	ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1476	if (ret) {
1477	dec_valid_block_count(sbi, inode, 1);
1478	f2fs_invalidate_blocks(sbi, *blkaddr, 1);
1479	} else {
1480	f2fs_update_data_blkaddr(&dn, *blkaddr);
1481	}
1482	f2fs_put_dnode(&dn);
1483	}
1484	return 0;
1485	}
1486
1487	static int __clone_blkaddrs(struct inode src_inode, struct inode dst_inode,
1488	block_t blkaddr, int do_replace,
1489	pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1490	{
1491	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1492	pgoff_t i = 0;
1493	int ret;
1494
1495	while (i < len) {
1496	if (blkaddr[i] == NULL_ADDR && !full) {
1497	i++;
1498	continue;
1499	}
1500
1501	if (do_replace[i] \|\| blkaddr[i] == NULL_ADDR) {
1502	struct dnode_of_data dn;
1503	struct node_info ni;
1504	size_t new_size;
1505	pgoff_t ilen;
1506
1507	set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1508	ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1509	if (ret)
1510	return ret;
1511
1512	ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1513	if (ret) {
1514	f2fs_put_dnode(&dn);
1515	return ret;
1516	}
1517
1518	ilen = min((pgoff_t)
1519	ADDRS_PER_PAGE(dn.node_folio, dst_inode) -
1520	dn.ofs_in_node, len - i);
1521	do {
1522	dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1523	f2fs_truncate_data_blocks_range(&dn, 1);
1524
1525	if (do_replace[i]) {
1526	f2fs_i_blocks_write(src_inode,
1527	1, false, false);
1528	f2fs_i_blocks_write(dst_inode,
1529	1, true, false);
1530	f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1531	blkaddr[i], ni.version, true, false);
1532
1533	do_replace[i] = 0;
1534	}
1535	dn.ofs_in_node++;
1536	i++;
1537	new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1538	if (dst_inode->i_size < new_size)
1539	f2fs_i_size_write(dst_inode, new_size);
1540	} while (--ilen && (do_replace[i] \|\| blkaddr[i] == NULL_ADDR));
1541
1542	f2fs_put_dnode(&dn);
1543	} else {
1544	struct folio fsrc, fdst;
1545
1546	fsrc = f2fs_get_lock_data_folio(src_inode,
1547	src + i, true);
1548	if (IS_ERR(fsrc))
1549	return PTR_ERR(fsrc);
1550	fdst = f2fs_get_new_data_folio(dst_inode, NULL, dst + i,
1551	true);
1552	if (IS_ERR(fdst)) {
1553	f2fs_folio_put(fsrc, true);
1554	return PTR_ERR(fdst);
1555	}
1556
1557	f2fs_folio_wait_writeback(fdst, DATA, true, true);
1558
1559	memcpy_folio(fdst, 0, fsrc, 0, PAGE_SIZE);
1560	folio_mark_dirty(fdst);
1561	folio_set_f2fs_gcing(fdst);
1562	f2fs_folio_put(fdst, true);
1563	f2fs_folio_put(fsrc, true);
1564
1565	ret = f2fs_truncate_hole(src_inode,
1566	src + i, src + i + 1);
1567	if (ret)
1568	return ret;
1569	i++;
1570	}
1571	}
1572	return 0;
1573	}
1574
1575	static int __exchange_data_block(struct inode *src_inode,
1576	struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1577	pgoff_t len, bool full)
1578	{
1579	block_t *src_blkaddr;
1580	int *do_replace;
1581	pgoff_t olen;
1582	int ret;
1583
1584	while (len) {
1585	olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1586
1587	src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1588	array_size(olen, sizeof(block_t)),
1589	GFP_NOFS);
1590	if (!src_blkaddr)
1591	return -ENOMEM;
1592
1593	do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1594	array_size(olen, sizeof(int)),
1595	GFP_NOFS);
1596	if (!do_replace) {
1597	kvfree(src_blkaddr);
1598	return -ENOMEM;
1599	}
1600
1601	ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1602	do_replace, src, olen);
1603	if (ret)
1604	goto roll_back;
1605
1606	ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1607	do_replace, src, dst, olen, full);
1608	if (ret)
1609	goto roll_back;
1610
1611	src += olen;
1612	dst += olen;
1613	len -= olen;
1614
1615	kvfree(src_blkaddr);
1616	kvfree(do_replace);
1617	}
1618	return 0;
1619
1620	roll_back:
1621	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1622	kvfree(src_blkaddr);
1623	kvfree(do_replace);
1624	return ret;
1625	}
1626
1627	static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1628	{
1629	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1630	struct f2fs_lock_context lc;
1631	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1632	pgoff_t start = offset >> PAGE_SHIFT;
1633	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1634	int ret;
1635
1636	f2fs_balance_fs(sbi, true);
1637
1638	/* avoid gc operation during block exchange */
1639	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1640	filemap_invalidate_lock(inode->i_mapping);
1641
1642	ret = f2fs_zero_post_eof_page(inode, offset + len, false, false);
1643	if (ret)
1644	goto out_unlock;
1645
1646	f2fs_lock_op(sbi, &lc);
1647	f2fs_drop_extent_tree(inode);
1648	truncate_pagecache(inode, offset);
1649	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1650	f2fs_unlock_op(sbi, &lc);
1651
1652	out_unlock:
1653	filemap_invalidate_unlock(inode->i_mapping);
1654	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1655	return ret;
1656	}
1657
1658	static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1659	{
1660	loff_t new_size;
1661	int ret;
1662
1663	if (offset + len >= i_size_read(inode))
1664	return -EINVAL;
1665
1666	/* collapse range should be aligned to block size of f2fs. */
1667	if (offset & (F2FS_BLKSIZE - 1) \|\| len & (F2FS_BLKSIZE - 1))
1668	return -EINVAL;
1669
1670	ret = f2fs_convert_inline_inode(inode);
1671	if (ret)
1672	return ret;
1673
1674	/* write out all dirty pages from offset */
1675	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1676	if (ret)
1677	return ret;
1678
1679	ret = f2fs_do_collapse(inode, offset, len);
1680	if (ret)
1681	return ret;
1682
1683	/* write out all moved pages, if possible */
1684	filemap_invalidate_lock(inode->i_mapping);
1685	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1686	if (ret)
1687	goto out_unlock;
1688	truncate_pagecache(inode, offset);
1689
1690	new_size = i_size_read(inode) - len;
1691	ret = f2fs_truncate_blocks(inode, new_size, true);
1692	out_unlock:
1693	filemap_invalidate_unlock(inode->i_mapping);
1694	if (!ret)
1695	f2fs_i_size_write(inode, new_size);
1696	return ret;
1697	}
1698
1699	static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1700	pgoff_t end)
1701	{
1702	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1703	pgoff_t index = start;
1704	unsigned int ofs_in_node = dn->ofs_in_node;
1705	blkcnt_t count = 0;
1706	int ret;
1707
1708	for (; index < end; index++, dn->ofs_in_node++) {
1709	if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1710	count++;
1711	}
1712
1713	dn->ofs_in_node = ofs_in_node;
1714	ret = f2fs_reserve_new_blocks(dn, count);
1715	if (ret)
1716	return ret;
1717
1718	dn->ofs_in_node = ofs_in_node;
1719	for (index = start; index < end; index++, dn->ofs_in_node++) {
1720	dn->data_blkaddr = f2fs_data_blkaddr(dn);
1721	/*
1722	* f2fs_reserve_new_blocks will not guarantee entire block
1723	* allocation.
1724	*/
1725	if (dn->data_blkaddr == NULL_ADDR) {
1726	ret = -ENOSPC;
1727	break;
1728	}
1729
1730	if (dn->data_blkaddr == NEW_ADDR)
1731	continue;
1732
1733	if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1734	DATA_GENERIC_ENHANCE)) {
1735	ret = -EFSCORRUPTED;
1736	break;
1737	}
1738
1739	f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1);
1740	f2fs_set_data_blkaddr(dn, NEW_ADDR);
1741	}
1742
1743	if (index > start) {
1744	f2fs_update_read_extent_cache_range(dn, start, 0,
1745	index - start);
1746	f2fs_update_age_extent_cache_range(dn, start, index - start);
1747	}
1748
1749	return ret;
1750	}
1751
1752	static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1753	int mode)
1754	{
1755	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1756	struct address_space *mapping = inode->i_mapping;
1757	pgoff_t index, pg_start, pg_end;
1758	loff_t new_size = i_size_read(inode);
1759	loff_t off_start, off_end;
1760	int ret = 0;
1761
1762	ret = inode_newsize_ok(inode, (len + offset));
1763	if (ret)
1764	return ret;
1765
1766	ret = f2fs_convert_inline_inode(inode);
1767	if (ret)
1768	return ret;
1769
1770	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1771	if (ret)
1772	return ret;
1773
1774	ret = f2fs_zero_post_eof_page(inode, offset + len, true, false);
1775	if (ret)
1776	return ret;
1777
1778	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1779	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1780
1781	off_start = offset & (PAGE_SIZE - 1);
1782	off_end = (offset + len) & (PAGE_SIZE - 1);
1783
1784	if (pg_start == pg_end) {
1785	ret = fill_zero(inode, pg_start, off_start,
1786	off_end - off_start);
1787	if (ret)
1788	return ret;
1789
1790	new_size = max_t(loff_t, new_size, offset + len);
1791	} else {
1792	if (off_start) {
1793	ret = fill_zero(inode, pg_start++, off_start,
1794	PAGE_SIZE - off_start);
1795	if (ret)
1796	return ret;
1797
1798	new_size = max_t(loff_t, new_size,
1799	(loff_t)pg_start << PAGE_SHIFT);
1800	}
1801
1802	for (index = pg_start; index < pg_end;) {
1803	struct dnode_of_data dn;
1804	struct f2fs_lock_context lc;
1805	unsigned int end_offset;
1806	pgoff_t end;
1807
1808	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1809	filemap_invalidate_lock(mapping);
1810
1811	truncate_pagecache_range(inode,
1812	(loff_t)index << PAGE_SHIFT,
1813	((loff_t)pg_end << PAGE_SHIFT) - 1);
1814
1815	f2fs_lock_op(sbi, &lc);
1816
1817	set_new_dnode(&dn, inode, NULL, NULL, 0);
1818	ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1819	if (ret) {
1820	f2fs_unlock_op(sbi, &lc);
1821	filemap_invalidate_unlock(mapping);
1822	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1823	goto out;
1824	}
1825
1826	end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
1827	end = min(pg_end, end_offset - dn.ofs_in_node + index);
1828
1829	ret = f2fs_do_zero_range(&dn, index, end);
1830	f2fs_put_dnode(&dn);
1831
1832	f2fs_unlock_op(sbi, &lc);
1833	filemap_invalidate_unlock(mapping);
1834	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1835
1836	f2fs_balance_fs(sbi, dn.node_changed);
1837
1838	if (ret)
1839	goto out;
1840
1841	index = end;
1842	new_size = max_t(loff_t, new_size,
1843	(loff_t)index << PAGE_SHIFT);
1844	}
1845
1846	if (off_end) {
1847	ret = fill_zero(inode, pg_end, 0, off_end);
1848	if (ret)
1849	goto out;
1850
1851	new_size = max_t(loff_t, new_size, offset + len);
1852	}
1853	}
1854
1855	out:
1856	if (new_size > i_size_read(inode)) {
1857	if (mode & FALLOC_FL_KEEP_SIZE)
1858	file_set_keep_isize(inode);
1859	else
1860	f2fs_i_size_write(inode, new_size);
1861	}
1862	return ret;
1863	}
1864
1865	static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1866	{
1867	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1868	struct address_space *mapping = inode->i_mapping;
1869	pgoff_t nr, pg_start, pg_end, delta, idx;
1870	loff_t new_size;
1871	int ret = 0;
1872
1873	new_size = i_size_read(inode) + len;
1874	ret = inode_newsize_ok(inode, new_size);
1875	if (ret)
1876	return ret;
1877
1878	if (offset >= i_size_read(inode))
1879	return -EINVAL;
1880
1881	/* insert range should be aligned to block size of f2fs. */
1882	if (offset & (F2FS_BLKSIZE - 1) \|\| len & (F2FS_BLKSIZE - 1))
1883	return -EINVAL;
1884
1885	ret = f2fs_convert_inline_inode(inode);
1886	if (ret)
1887	return ret;
1888
1889	f2fs_balance_fs(sbi, true);
1890
1891	filemap_invalidate_lock(mapping);
1892	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1893	filemap_invalidate_unlock(mapping);
1894	if (ret)
1895	return ret;
1896
1897	/* write out all dirty pages from offset */
1898	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1899	if (ret)
1900	return ret;
1901
1902	pg_start = offset >> PAGE_SHIFT;
1903	pg_end = (offset + len) >> PAGE_SHIFT;
1904	delta = pg_end - pg_start;
1905	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1906
1907	/* avoid gc operation during block exchange */
1908	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1909	filemap_invalidate_lock(mapping);
1910
1911	ret = f2fs_zero_post_eof_page(inode, offset + len, false, false);
1912	if (ret)
1913	goto out_unlock;
1914	truncate_pagecache(inode, offset);
1915
1916	while (!ret && idx > pg_start) {
1917	struct f2fs_lock_context lc;
1918
1919	nr = idx - pg_start;
1920	if (nr > delta)
1921	nr = delta;
1922	idx -= nr;
1923
1924	f2fs_lock_op(sbi, &lc);
1925	f2fs_drop_extent_tree(inode);
1926
1927	ret = __exchange_data_block(inode, inode, idx,
1928	idx + delta, nr, false);
1929	f2fs_unlock_op(sbi, &lc);
1930	}
1931	out_unlock:
1932	filemap_invalidate_unlock(mapping);
1933	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1934	if (ret)
1935	return ret;
1936
1937	/* write out all moved pages, if possible */
1938	filemap_invalidate_lock(mapping);
1939	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1940	truncate_pagecache(inode, offset);
1941	filemap_invalidate_unlock(mapping);
1942
1943	if (!ret)
1944	f2fs_i_size_write(inode, new_size);
1945	return ret;
1946	}
1947
1948	static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
1949	loff_t len, int mode)
1950	{
1951	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1952	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1953	.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1954	.m_may_create = true };
1955	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1956	.init_gc_type = FG_GC,
1957	.should_migrate_blocks = false,
1958	.err_gc_skipped = true,
1959	.nr_free_secs = 0 };
1960	pgoff_t pg_start, pg_end;
1961	loff_t new_size;
1962	loff_t off_end;
1963	block_t expanded = 0;
1964	int err;
1965
1966	err = inode_newsize_ok(inode, (len + offset));
1967	if (err)
1968	return err;
1969
1970	err = f2fs_convert_inline_inode(inode);
1971	if (err)
1972	return err;
1973
1974	err = f2fs_zero_post_eof_page(inode, offset + len, true, true);
1975	if (err)
1976	return err;
1977
1978	f2fs_balance_fs(sbi, true);
1979
1980	pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1981	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1982	off_end = (offset + len) & (PAGE_SIZE - 1);
1983
1984	map.m_lblk = pg_start;
1985	map.m_len = pg_end - pg_start;
1986	if (off_end)
1987	map.m_len++;
1988
1989	if (!map.m_len)
1990	return 0;
1991
1992	if (f2fs_is_pinned_file(inode)) {
1993	block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1994	block_t sec_len;
1995
1996	if (map.m_lblk % sec_blks) {
1997	pg_start = rounddown(map.m_lblk, sec_blks);
1998	map.m_lblk = pg_start;
1999	map.m_len = pg_end - pg_start;
2000	if (off_end)
2001	map.m_len++;
2002	}
2003	sec_len = roundup(map.m_len, sec_blks);
2004	map.m_len = sec_blks;
2005	next_alloc:
2006	f2fs_down_write(&sbi->pin_sem);
2007
2008	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2009	if (has_not_enough_free_secs(sbi, 0, 0)) {
2010	f2fs_up_write(&sbi->pin_sem);
2011	err = -ENOSPC;
2012	f2fs_warn_ratelimited(sbi,
2013	"ino:%llu, start:%lu, end:%lu, need to trigger GC to "
2014	"reclaim enough free segment when checkpoint is enabled",
2015	inode->i_ino, pg_start, pg_end);
2016	goto out_err;
2017	}
2018	}
2019
2020	if (has_not_enough_free_secs(sbi, 0,
2021	sbi->reserved_pin_section)) {
2022	f2fs_down_write_trace(&sbi->gc_lock, &gc_control.lc);
2023	stat_inc_gc_call_count(sbi, FOREGROUND);
2024	err = f2fs_gc(sbi, &gc_control);
2025	if (err && err != -ENODATA) {
2026	f2fs_up_write(&sbi->pin_sem);
2027	goto out_err;
2028	}
2029	}
2030
2031	err = f2fs_allocate_pinning_section(sbi);
2032	if (err) {
2033	f2fs_up_write(&sbi->pin_sem);
2034	goto out_err;
2035	}
2036
2037	map.m_seg_type = CURSEG_COLD_DATA_PINNED;
2038	err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
2039	file_dont_truncate(inode);
2040
2041	f2fs_up_write(&sbi->pin_sem);
2042
2043	expanded += map.m_len;
2044	sec_len -= map.m_len;
2045	map.m_lblk += map.m_len;
2046	if (!err && sec_len)
2047	goto next_alloc;
2048
2049	map.m_len = expanded;
2050	} else {
2051	err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
2052	expanded = map.m_len;
2053	}
2054	out_err:
2055	if (err) {
2056	pgoff_t last_off;
2057
2058	if (!expanded)
2059	return err;
2060
2061	last_off = pg_start + expanded - 1;
2062
2063	/* update new size to the failed position */
2064	new_size = (last_off == pg_end) ? offset + len :
2065	(loff_t)(last_off + 1) << PAGE_SHIFT;
2066	} else {
2067	new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
2068	}
2069
2070	if (new_size > i_size_read(inode)) {
2071	if (mode & FALLOC_FL_KEEP_SIZE)
2072	file_set_keep_isize(inode);
2073	else
2074	f2fs_i_size_write(inode, new_size);
2075	}
2076
2077	return err;
2078	}
2079
2080	static long f2fs_fallocate(struct file *file, int mode,
2081	loff_t offset, loff_t len)
2082	{
2083	struct inode *inode = file_inode(file);
2084	long ret = 0;
2085
2086	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2087	return -EIO;
2088	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
2089	return -ENOSPC;
2090	if (!f2fs_is_compress_backend_ready(inode) \|\| IS_DEVICE_ALIASING(inode))
2091	return -EOPNOTSUPP;
2092
2093	/* f2fs only support ->fallocate for regular file */
2094	if (!S_ISREG(inode->i_mode))
2095	return -EINVAL;
2096
2097	if (IS_ENCRYPTED(inode) &&
2098	(mode & (FALLOC_FL_COLLAPSE_RANGE \| FALLOC_FL_INSERT_RANGE)))
2099	return -EOPNOTSUPP;
2100
2101	if (mode & ~(FALLOC_FL_KEEP_SIZE \| FALLOC_FL_PUNCH_HOLE \|
2102	FALLOC_FL_COLLAPSE_RANGE \| FALLOC_FL_ZERO_RANGE \|
2103	FALLOC_FL_INSERT_RANGE))
2104	return -EOPNOTSUPP;
2105
2106	inode_lock(inode);
2107
2108	/*
2109	* Pinned file should not support partial truncation since the block
2110	* can be used by applications.
2111	*/
2112	if ((f2fs_compressed_file(inode) \|\| f2fs_is_pinned_file(inode)) &&
2113	(mode & (FALLOC_FL_PUNCH_HOLE \| FALLOC_FL_COLLAPSE_RANGE \|
2114	FALLOC_FL_ZERO_RANGE \| FALLOC_FL_INSERT_RANGE))) {
2115	ret = -EOPNOTSUPP;
2116	goto out;
2117	}
2118
2119	ret = file_modified(file);
2120	if (ret)
2121	goto out;
2122
2123	/*
2124	* wait for inflight dio, blocks should be removed after IO
2125	* completion.
2126	*/
2127	inode_dio_wait(inode);
2128
2129	if (mode & FALLOC_FL_PUNCH_HOLE) {
2130	if (offset >= inode->i_size)
2131	goto out;
2132
2133	ret = f2fs_punch_hole(inode, offset, len);
2134	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
2135	ret = f2fs_collapse_range(inode, offset, len);
2136	} else if (mode & FALLOC_FL_ZERO_RANGE) {
2137	ret = f2fs_zero_range(inode, offset, len, mode);
2138	} else if (mode & FALLOC_FL_INSERT_RANGE) {
2139	ret = f2fs_insert_range(inode, offset, len);
2140	} else {
2141	ret = f2fs_expand_inode_data(inode, offset, len, mode);
2142	}
2143
2144	if (!ret) {
2145	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2146	f2fs_mark_inode_dirty_sync(inode, false);
2147	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2148	}
2149
2150	out:
2151	inode_unlock(inode);
2152
2153	trace_f2fs_fallocate(inode, mode, offset, len, ret);
2154	return ret;
2155	}
2156
2157	static int f2fs_release_file(struct inode inode, struct file filp)
2158	{
2159	if (atomic_dec_and_test(&F2FS_I(inode)->open_count))
2160	f2fs_remove_donate_inode(inode);
2161
2162	/*
2163	* f2fs_release_file is called at every close calls. So we should
2164	* not drop any inmemory pages by close called by other process.
2165	*/
2166	if (!(filp->f_mode & FMODE_WRITE) \|\|
2167	atomic_read(&inode->i_writecount) != 1)
2168	return 0;
2169
2170	inode_lock(inode);
2171	f2fs_abort_atomic_write(inode, true);
2172	inode_unlock(inode);
2173
2174	return 0;
2175	}
2176
2177	static int f2fs_file_flush(struct file *file, fl_owner_t id)
2178	{
2179	struct inode *inode = file_inode(file);
2180
2181	/*
2182	* If the process doing a transaction is crashed, we should do
2183	* roll-back. Otherwise, other reader/write can see corrupted database
2184	* until all the writers close its file. Since this should be done
2185	* before dropping file lock, it needs to do in ->flush.
2186	*/
2187	if (F2FS_I(inode)->atomic_write_task == current &&
2188	(current->flags & PF_EXITING)) {
2189	inode_lock(inode);
2190	f2fs_abort_atomic_write(inode, true);
2191	inode_unlock(inode);
2192	}
2193
2194	return 0;
2195	}
2196
2197	static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
2198	{
2199	struct f2fs_inode_info *fi = F2FS_I(inode);
2200	u32 masked_flags = fi->i_flags & mask;
2201
2202	/* mask can be shrunk by flags_valid selector */
2203	iflags &= mask;
2204
2205	/* Is it quota file? Do not allow user to mess with it */
2206	if (IS_NOQUOTA(inode))
2207	return -EPERM;
2208
2209	if (IS_DEVICE_ALIASING(inode))
2210	return -EPERM;
2211
2212	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
2213	if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
2214	return -EOPNOTSUPP;
2215	if (!f2fs_empty_dir(inode))
2216	return -ENOTEMPTY;
2217	}
2218
2219	if (iflags & (F2FS_COMPR_FL \| F2FS_NOCOMP_FL)) {
2220	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
2221	return -EOPNOTSUPP;
2222	if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
2223	return -EINVAL;
2224	}
2225
2226	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
2227	if (masked_flags & F2FS_COMPR_FL) {
2228	if (!f2fs_disable_compressed_file(inode))
2229	return -EINVAL;
2230	} else {
2231	/* try to convert inline_data to support compression */
2232	int err = f2fs_convert_inline_inode(inode);
2233	if (err)
2234	return err;
2235
2236	f2fs_down_write(&fi->i_sem);
2237	if (!f2fs_may_compress(inode) \|\|
2238	atomic_read(&fi->writeback) \|\|
2239	(S_ISREG(inode->i_mode) &&
2240	F2FS_HAS_BLOCKS(inode))) {
2241	f2fs_up_write(&fi->i_sem);
2242	return -EINVAL;
2243	}
2244	err = set_compress_context(inode);
2245	f2fs_up_write(&fi->i_sem);
2246
2247	if (err)
2248	return err;
2249	}
2250	}
2251
2252	fi->i_flags = iflags \| (fi->i_flags & ~mask);
2253	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
2254	(fi->i_flags & F2FS_NOCOMP_FL));
2255
2256	if (fi->i_flags & F2FS_PROJINHERIT_FL)
2257	set_inode_flag(inode, FI_PROJ_INHERIT);
2258	else
2259	clear_inode_flag(inode, FI_PROJ_INHERIT);
2260
2261	inode_set_ctime_current(inode);
2262	f2fs_set_inode_flags(inode);
2263	f2fs_mark_inode_dirty_sync(inode, true);
2264	return 0;
2265	}
2266
2267	/* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
2268
2269	/*
2270	* To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
2271	* for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
2272	* F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
2273	* its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
2274	*
2275	* Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
2276	* FS_IOC_FSSETXATTR is done by the VFS.
2277	*/
2278
2279	static const struct {
2280	u32 iflag;
2281	u32 fsflag;
2282	} f2fs_fsflags_map[] = {
2283	{ F2FS_COMPR_FL, FS_COMPR_FL },
2284	{ F2FS_SYNC_FL, FS_SYNC_FL },
2285	{ F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
2286	{ F2FS_APPEND_FL, FS_APPEND_FL },
2287	{ F2FS_NODUMP_FL, FS_NODUMP_FL },
2288	{ F2FS_NOATIME_FL, FS_NOATIME_FL },
2289	{ F2FS_NOCOMP_FL, FS_NOCOMP_FL },
2290	{ F2FS_INDEX_FL, FS_INDEX_FL },
2291	{ F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
2292	{ F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
2293	{ F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
2294	};
2295
2296	#define F2FS_GETTABLE_FS_FL ( \
2297	FS_COMPR_FL \| \
2298	FS_SYNC_FL \| \
2299	FS_IMMUTABLE_FL \| \
2300	FS_APPEND_FL \| \
2301	FS_NODUMP_FL \| \
2302	FS_NOATIME_FL \| \
2303	FS_NOCOMP_FL \| \
2304	FS_INDEX_FL \| \
2305	FS_DIRSYNC_FL \| \
2306	FS_PROJINHERIT_FL \| \
2307	FS_ENCRYPT_FL \| \
2308	FS_INLINE_DATA_FL \| \
2309	FS_NOCOW_FL \| \
2310	FS_VERITY_FL \| \
2311	FS_CASEFOLD_FL)
2312
2313	#define F2FS_SETTABLE_FS_FL ( \
2314	FS_COMPR_FL \| \
2315	FS_SYNC_FL \| \
2316	FS_IMMUTABLE_FL \| \
2317	FS_APPEND_FL \| \
2318	FS_NODUMP_FL \| \
2319	FS_NOATIME_FL \| \
2320	FS_NOCOMP_FL \| \
2321	FS_DIRSYNC_FL \| \
2322	FS_PROJINHERIT_FL \| \
2323	FS_CASEFOLD_FL)
2324
2325	/* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2326	static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2327	{
2328	u32 fsflags = 0;
2329	int i;
2330
2331	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2332	if (iflags & f2fs_fsflags_map[i].iflag)
2333	fsflags \|= f2fs_fsflags_map[i].fsflag;
2334
2335	return fsflags;
2336	}
2337
2338	/* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2339	static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2340	{
2341	u32 iflags = 0;
2342	int i;
2343
2344	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2345	if (fsflags & f2fs_fsflags_map[i].fsflag)
2346	iflags \|= f2fs_fsflags_map[i].iflag;
2347
2348	return iflags;
2349	}
2350
2351	static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2352	{
2353	struct inode *inode = file_inode(filp);
2354
2355	return put_user(inode->i_generation, (int __user *)arg);
2356	}
2357
2358	static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2359	{
2360	struct inode *inode = file_inode(filp);
2361	const struct mnt_idmap *idmap = file_mnt_idmap(filp);
2362	struct f2fs_inode_info *fi = F2FS_I(inode);
2363	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2364	loff_t isize;
2365	int ret;
2366
2367	if (!(filp->f_mode & FMODE_WRITE))
2368	return -EBADF;
2369
2370	if (!inode_owner_or_capable(idmap, inode))
2371	return -EACCES;
2372
2373	if (!S_ISREG(inode->i_mode))
2374	return -EINVAL;
2375
2376	if (filp->f_flags & O_DIRECT)
2377	return -EINVAL;
2378
2379	ret = mnt_want_write_file(filp);
2380	if (ret)
2381	return ret;
2382
2383	inode_lock(inode);
2384
2385	if (!f2fs_disable_compressed_file(inode) \|\|
2386	f2fs_is_pinned_file(inode)) {
2387	ret = -EINVAL;
2388	goto out;
2389	}
2390
2391	if (f2fs_is_atomic_file(inode))
2392	goto out;
2393
2394	ret = f2fs_convert_inline_inode(inode);
2395	if (ret)
2396	goto out;
2397
2398	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2399	f2fs_down_write(&fi->i_gc_rwsem[READ]);
2400
2401	/*
2402	* Should wait end_io to count F2FS_WB_CP_DATA correctly by
2403	* f2fs_is_atomic_file.
2404	*/
2405	if (get_dirty_pages(inode))
2406	f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%llu, npages=%u",
2407	inode->i_ino, get_dirty_pages(inode));
2408	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2409	if (ret)
2410	goto out_unlock;
2411
2412	/* Check if the inode already has a COW inode */
2413	if (fi->cow_inode == NULL) {
2414	/* Create a COW inode for atomic write */
2415	struct dentry *dentry = file_dentry(filp);
2416	struct inode *dir = d_inode(dentry->d_parent);
2417
2418	ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode);
2419	if (ret)
2420	goto out_unlock;
2421
2422	set_inode_flag(fi->cow_inode, FI_COW_FILE);
2423	clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2424
2425	/* Set the COW inode's atomic_inode to the atomic inode */
2426	F2FS_I(fi->cow_inode)->atomic_inode = inode;
2427	} else {
2428	/* Reuse the already created COW inode */
2429	f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode));
2430
2431	invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1);
2432
2433	ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2434	if (ret)
2435	goto out_unlock;
2436	}
2437
2438	f2fs_write_inode(inode, NULL);
2439
2440	stat_inc_atomic_inode(inode);
2441
2442	set_inode_flag(inode, FI_ATOMIC_FILE);
2443
2444	isize = i_size_read(inode);
2445	fi->original_i_size = isize;
2446	if (truncate) {
2447	set_inode_flag(inode, FI_ATOMIC_REPLACE);
2448	truncate_inode_pages_final(inode->i_mapping);
2449	f2fs_i_size_write(inode, 0);
2450	isize = 0;
2451	}
2452	f2fs_i_size_write(fi->cow_inode, isize);
2453
2454	out_unlock:
2455	f2fs_up_write(&fi->i_gc_rwsem[READ]);
2456	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2457	if (ret)
2458	goto out;
2459
2460	f2fs_update_time(sbi, REQ_TIME);
2461	fi->atomic_write_task = current;
2462	stat_update_max_atomic_write(inode);
2463	fi->atomic_write_cnt = 0;
2464	out:
2465	inode_unlock(inode);
2466	mnt_drop_write_file(filp);
2467	return ret;
2468	}
2469
2470	static int f2fs_ioc_commit_atomic_write(struct file *filp)
2471	{
2472	struct inode *inode = file_inode(filp);
2473	const struct mnt_idmap *idmap = file_mnt_idmap(filp);
2474	int ret;
2475
2476	if (!(filp->f_mode & FMODE_WRITE))
2477	return -EBADF;
2478
2479	if (!inode_owner_or_capable(idmap, inode))
2480	return -EACCES;
2481
2482	ret = mnt_want_write_file(filp);
2483	if (ret)
2484	return ret;
2485
2486	f2fs_balance_fs(F2FS_I_SB(inode), true);
2487
2488	inode_lock(inode);
2489
2490	if (f2fs_is_atomic_file(inode)) {
2491	ret = f2fs_commit_atomic_write(inode);
2492	if (!ret)
2493	ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2494
2495	f2fs_abort_atomic_write(inode, ret);
2496	} else {
2497	ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2498	}
2499
2500	inode_unlock(inode);
2501	mnt_drop_write_file(filp);
2502	return ret;
2503	}
2504
2505	static int f2fs_ioc_abort_atomic_write(struct file *filp)
2506	{
2507	struct inode *inode = file_inode(filp);
2508	const struct mnt_idmap *idmap = file_mnt_idmap(filp);
2509	int ret;
2510
2511	if (!(filp->f_mode & FMODE_WRITE))
2512	return -EBADF;
2513
2514	if (!inode_owner_or_capable(idmap, inode))
2515	return -EACCES;
2516
2517	ret = mnt_want_write_file(filp);
2518	if (ret)
2519	return ret;
2520
2521	inode_lock(inode);
2522
2523	f2fs_abort_atomic_write(inode, true);
2524
2525	inode_unlock(inode);
2526
2527	mnt_drop_write_file(filp);
2528	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2529	return ret;
2530	}
2531
2532	int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
2533	bool readonly, bool need_lock)
2534	{
2535	struct super_block *sb = sbi->sb;
2536	int ret = 0;
2537
2538	switch (flag) {
2539	case F2FS_GOING_DOWN_FULLSYNC:
2540	ret = bdev_freeze(sb->s_bdev);
2541	if (ret)
2542	goto out;
2543	f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2544	bdev_thaw(sb->s_bdev);
2545	break;
2546	case F2FS_GOING_DOWN_METASYNC:
2547	/* do checkpoint only */
2548	ret = f2fs_sync_fs(sb, 1);
2549	if (ret) {
2550	if (ret == -EIO)
2551	ret = 0;
2552	goto out;
2553	}
2554	f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2555	break;
2556	case F2FS_GOING_DOWN_NOSYNC:
2557	f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2558	break;
2559	case F2FS_GOING_DOWN_METAFLUSH:
2560	f2fs_sync_meta_pages(sbi, LONG_MAX, FS_META_IO);
2561	f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2562	break;
2563	case F2FS_GOING_DOWN_NEED_FSCK:
2564	set_sbi_flag(sbi, SBI_NEED_FSCK);
2565	set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2566	set_sbi_flag(sbi, SBI_IS_DIRTY);
2567	/* do checkpoint only */
2568	ret = f2fs_sync_fs(sb, 1);
2569	if (ret == -EIO)
2570	ret = 0;
2571	goto out;
2572	default:
2573	ret = -EINVAL;
2574	goto out;
2575	}
2576
2577	if (readonly)
2578	goto out;
2579
2580	/*
2581	* grab sb->s_umount to avoid racing w/ remount() and other shutdown
2582	* paths.
2583	*/
2584	if (need_lock)
2585	down_write(&sbi->sb->s_umount);
2586
2587	f2fs_stop_gc_thread(sbi);
2588	f2fs_stop_discard_thread(sbi);
2589
2590	f2fs_drop_discard_cmd(sbi);
2591	clear_opt(sbi, DISCARD);
2592
2593	if (need_lock)
2594	up_write(&sbi->sb->s_umount);
2595
2596	f2fs_update_time(sbi, REQ_TIME);
2597	out:
2598
2599	trace_f2fs_shutdown(sbi, flag, ret);
2600
2601	return ret;
2602	}
2603
2604	static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2605	{
2606	struct inode *inode = file_inode(filp);
2607	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2608	__u32 in;
2609	int ret;
2610	bool need_drop = false, readonly = false;
2611
2612	if (!capable(CAP_SYS_ADMIN))
2613	return -EPERM;
2614
2615	if (get_user(in, (__u32 __user *)arg))
2616	return -EFAULT;
2617
2618	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2619	ret = mnt_want_write_file(filp);
2620	if (ret) {
2621	if (ret != -EROFS)
2622	return ret;
2623
2624	/* fallback to nosync shutdown for readonly fs */
2625	in = F2FS_GOING_DOWN_NOSYNC;
2626	readonly = true;
2627	} else {
2628	need_drop = true;
2629	}
2630	}
2631
2632	ret = f2fs_do_shutdown(sbi, in, readonly, true);
2633
2634	if (need_drop)
2635	mnt_drop_write_file(filp);
2636
2637	return ret;
2638	}
2639
2640	static int f2fs_keep_noreuse_range(struct inode *inode,
2641	loff_t offset, loff_t len)
2642	{
2643	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2644	u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
2645	u64 start, end;
2646	int ret = 0;
2647
2648	if (!S_ISREG(inode->i_mode))
2649	return 0;
2650
2651	if (offset >= max_bytes \|\| len > max_bytes \|\|
2652	(offset + len) > max_bytes)
2653	return 0;
2654
2655	start = offset >> PAGE_SHIFT;
2656	end = DIV_ROUND_UP(offset + len, PAGE_SIZE);
2657
2658	inode_lock(inode);
2659	if (f2fs_is_atomic_file(inode)) {
2660	inode_unlock(inode);
2661	return 0;
2662	}
2663
2664	spin_lock(&sbi->inode_lock[DONATE_INODE]);
2665	/* let's remove the range, if len = 0 */
2666	if (!len) {
2667	if (!list_empty(&F2FS_I(inode)->gdonate_list)) {
2668	list_del_init(&F2FS_I(inode)->gdonate_list);
2669	sbi->donate_files--;
2670	if (is_inode_flag_set(inode, FI_DONATE_FINISHED))
2671	ret = -EALREADY;
2672	else
2673	set_inode_flag(inode, FI_DONATE_FINISHED);
2674	} else
2675	ret = -ENOENT;
2676	} else {
2677	if (list_empty(&F2FS_I(inode)->gdonate_list)) {
2678	list_add_tail(&F2FS_I(inode)->gdonate_list,
2679	&sbi->inode_list[DONATE_INODE]);
2680	sbi->donate_files++;
2681	} else {
2682	list_move_tail(&F2FS_I(inode)->gdonate_list,
2683	&sbi->inode_list[DONATE_INODE]);
2684	}
2685	F2FS_I(inode)->donate_start = start;
2686	F2FS_I(inode)->donate_end = end - 1;
2687	clear_inode_flag(inode, FI_DONATE_FINISHED);
2688	}
2689	spin_unlock(&sbi->inode_lock[DONATE_INODE]);
2690	inode_unlock(inode);
2691
2692	return ret;
2693	}
2694
2695	static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2696	{
2697	struct inode *inode = file_inode(filp);
2698	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2699	struct fstrim_range range;
2700	int ret;
2701
2702	if (!capable(CAP_SYS_ADMIN))
2703	return -EPERM;
2704
2705	if (!f2fs_hw_support_discard(sbi))
2706	return -EOPNOTSUPP;
2707
2708	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2709	sizeof(range)))
2710	return -EFAULT;
2711
2712	ret = mnt_want_write_file(filp);
2713	if (ret)
2714	return ret;
2715
2716	range.minlen = max_t(unsigned int, range.minlen,
2717	f2fs_hw_discard_granularity(sbi));
2718	ret = f2fs_trim_fs(sbi, &range);
2719	mnt_drop_write_file(filp);
2720	if (ret < 0)
2721	return ret;
2722
2723	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2724	sizeof(range)))
2725	return -EFAULT;
2726	f2fs_update_time(sbi, REQ_TIME);
2727	return 0;
2728	}
2729
2730	static bool uuid_is_nonzero(__u8 u[16])
2731	{
2732	int i;
2733
2734	for (i = 0; i < 16; i++)
2735	if (u[i])
2736	return true;
2737	return false;
2738	}
2739
2740	static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2741	{
2742	struct inode *inode = file_inode(filp);
2743	int ret;
2744
2745	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2746	return -EOPNOTSUPP;
2747
2748	ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2749	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2750	return ret;
2751	}
2752
2753	static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2754	{
2755	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2756	return -EOPNOTSUPP;
2757	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2758	}
2759
2760	static int f2fs_ioc_get_dev_alias_status(struct file *filp, unsigned long arg)
2761	{
2762	struct inode *inode = file_inode(filp);
2763
2764	if (!IS_DEVICE_ALIASING(inode))
2765	return -EINVAL;
2766
2767	return put_user(F2FS_HAS_BLOCKS(inode) ? F2FS_DEV_ALIAS_STATUS_RESERVED :
2768	F2FS_DEV_ALIAS_STATUS_RELEASED, (u32 __user *)arg);
2769	}
2770
2771	static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2772	{
2773	struct inode *inode = file_inode(filp);
2774	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2775	u8 encrypt_pw_salt[16];
2776	int err;
2777
2778	if (!f2fs_sb_has_encrypt(sbi))
2779	return -EOPNOTSUPP;
2780
2781	err = mnt_want_write_file(filp);
2782	if (err)
2783	return err;
2784
2785	f2fs_down_write(&sbi->sb_lock);
2786
2787	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2788	goto got_it;
2789
2790	/* update superblock with uuid */
2791	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2792
2793	err = f2fs_commit_super(sbi, false);
2794	if (err) {
2795	/* undo new data */
2796	memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2797	goto out_err;
2798	}
2799	got_it:
2800	memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2801	out_err:
2802	f2fs_up_write(&sbi->sb_lock);
2803	mnt_drop_write_file(filp);
2804
2805	if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2806	err = -EFAULT;
2807
2808	return err;
2809	}
2810
2811	static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2812	unsigned long arg)
2813	{
2814	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2815	return -EOPNOTSUPP;
2816
2817	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2818	}
2819
2820	static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2821	{
2822	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2823	return -EOPNOTSUPP;
2824
2825	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2826	}
2827
2828	static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2829	{
2830	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2831	return -EOPNOTSUPP;
2832
2833	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2834	}
2835
2836	static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2837	unsigned long arg)
2838	{
2839	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2840	return -EOPNOTSUPP;
2841
2842	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2843	}
2844
2845	static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2846	unsigned long arg)
2847	{
2848	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2849	return -EOPNOTSUPP;
2850
2851	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2852	}
2853
2854	static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2855	{
2856	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2857	return -EOPNOTSUPP;
2858
2859	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2860	}
2861
2862	static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2863	{
2864	struct inode *inode = file_inode(filp);
2865	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2866	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2867	.no_bg_gc = false,
2868	.should_migrate_blocks = false,
2869	.nr_free_secs = 0 };
2870	__u32 sync;
2871	int ret;
2872
2873	if (!capable(CAP_SYS_ADMIN))
2874	return -EPERM;
2875
2876	if (get_user(sync, (__u32 __user *)arg))
2877	return -EFAULT;
2878
2879	if (f2fs_readonly(sbi->sb))
2880	return -EROFS;
2881
2882	ret = mnt_want_write_file(filp);
2883	if (ret)
2884	return ret;
2885
2886	if (!sync) {
2887	if (!f2fs_down_write_trylock_trace(&sbi->gc_lock,
2888	&gc_control.lc)) {
2889	ret = -EBUSY;
2890	goto out;
2891	}
2892	} else {
2893	f2fs_down_write_trace(&sbi->gc_lock, &gc_control.lc);
2894	}
2895
2896	gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2897	gc_control.err_gc_skipped = sync;
2898	stat_inc_gc_call_count(sbi, FOREGROUND);
2899	ret = f2fs_gc(sbi, &gc_control);
2900	out:
2901	mnt_drop_write_file(filp);
2902	return ret;
2903	}
2904
2905	static int __f2fs_ioc_gc_range(struct file filp, struct f2fs_gc_range range)
2906	{
2907	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2908	struct f2fs_gc_control gc_control = {
2909	.init_gc_type = range->sync ? FG_GC : BG_GC,
2910	.no_bg_gc = false,
2911	.should_migrate_blocks = false,
2912	.err_gc_skipped = range->sync,
2913	.nr_free_secs = 0 };
2914	u64 end;
2915	int ret;
2916
2917	if (!capable(CAP_SYS_ADMIN))
2918	return -EPERM;
2919	if (f2fs_readonly(sbi->sb))
2920	return -EROFS;
2921
2922	end = range->start + range->len;
2923	if (end < range->start \|\| range->start < MAIN_BLKADDR(sbi) \|\|
2924	end >= MAX_BLKADDR(sbi))
2925	return -EINVAL;
2926
2927	ret = mnt_want_write_file(filp);
2928	if (ret)
2929	return ret;
2930
2931	do_more:
2932	if (!range->sync) {
2933	if (!f2fs_down_write_trylock_trace(&sbi->gc_lock, &gc_control.lc)) {
2934	ret = -EBUSY;
2935	goto out;
2936	}
2937	} else {
2938	f2fs_down_write_trace(&sbi->gc_lock, &gc_control.lc);
2939	}
2940
2941	gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2942	stat_inc_gc_call_count(sbi, FOREGROUND);
2943	ret = f2fs_gc(sbi, &gc_control);
2944	if (ret) {
2945	if (ret == -EBUSY)
2946	ret = -EAGAIN;
2947	goto out;
2948	}
2949	range->start += CAP_BLKS_PER_SEC(sbi);
2950	if (range->start <= end)
2951	goto do_more;
2952	out:
2953	mnt_drop_write_file(filp);
2954	return ret;
2955	}
2956
2957	static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2958	{
2959	struct f2fs_gc_range range;
2960
2961	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2962	sizeof(range)))
2963	return -EFAULT;
2964	return __f2fs_ioc_gc_range(filp, &range);
2965	}
2966
2967	static int f2fs_ioc_write_checkpoint(struct file *filp)
2968	{
2969	struct inode *inode = file_inode(filp);
2970	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2971	int ret;
2972
2973	if (!capable(CAP_SYS_ADMIN))
2974	return -EPERM;
2975
2976	if (f2fs_readonly(sbi->sb))
2977	return -EROFS;
2978
2979	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2980	f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2981	return -EINVAL;
2982	}
2983
2984	ret = mnt_want_write_file(filp);
2985	if (ret)
2986	return ret;
2987
2988	ret = f2fs_sync_fs(sbi->sb, 1);
2989
2990	mnt_drop_write_file(filp);
2991	return ret;
2992	}
2993
2994	static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2995	struct file *filp,
2996	struct f2fs_defragment *range)
2997	{
2998	struct inode *inode = file_inode(filp);
2999	struct f2fs_map_blocks map = { .m_next_extent = NULL,
3000	.m_seg_type = NO_CHECK_TYPE,
3001	.m_may_create = false };
3002	struct extent_info ei = {};
3003	pgoff_t pg_start, pg_end, next_pgofs;
3004	unsigned int total = 0, sec_num;
3005	block_t blk_end = 0;
3006	bool fragmented = false;
3007	int err;
3008
3009	f2fs_balance_fs(sbi, true);
3010
3011	inode_lock(inode);
3012	pg_start = range->start >> PAGE_SHIFT;
3013	pg_end = min_t(pgoff_t,
3014	(range->start + range->len) >> PAGE_SHIFT,
3015	DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));
3016
3017	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) \|\|
3018	f2fs_is_atomic_file(inode)) {
3019	err = -EINVAL;
3020	goto unlock_out;
3021	}
3022
3023	/* if in-place-update policy is enabled, don't waste time here */
3024	set_inode_flag(inode, FI_OPU_WRITE);
3025	if (f2fs_should_update_inplace(inode, NULL)) {
3026	err = -EINVAL;
3027	goto out;
3028	}
3029
3030	/* writeback all dirty pages in the range */
3031	err = filemap_write_and_wait_range(inode->i_mapping,
3032	pg_start << PAGE_SHIFT,
3033	(pg_end << PAGE_SHIFT) - 1);
3034	if (err)
3035	goto out;
3036
3037	/*
3038	* lookup mapping info in extent cache, skip defragmenting if physical
3039	* block addresses are continuous.
3040	*/
3041	if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
3042	if ((pgoff_t)ei.fofs + ei.len >= pg_end)
3043	goto out;
3044	}
3045
3046	map.m_lblk = pg_start;
3047	map.m_next_pgofs = &next_pgofs;
3048
3049	/*
3050	* lookup mapping info in dnode page cache, skip defragmenting if all
3051	* physical block addresses are continuous even if there are hole(s)
3052	* in logical blocks.
3053	*/
3054	while (map.m_lblk < pg_end) {
3055	map.m_len = pg_end - map.m_lblk;
3056	err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
3057	if (err)
3058	goto out;
3059
3060	if (!(map.m_flags & F2FS_MAP_FLAGS)) {
3061	map.m_lblk = next_pgofs;
3062	continue;
3063	}
3064
3065	if (blk_end && blk_end != map.m_pblk)
3066	fragmented = true;
3067
3068	/* record total count of block that we're going to move */
3069	total += map.m_len;
3070
3071	blk_end = map.m_pblk + map.m_len;
3072
3073	map.m_lblk += map.m_len;
3074	}
3075
3076	if (!fragmented) {
3077	total = 0;
3078	goto out;
3079	}
3080
3081	sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
3082
3083	/*
3084	* make sure there are enough free section for LFS allocation, this can
3085	* avoid defragment running in SSR mode when free section are allocated
3086	* intensively
3087	*/
3088	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
3089	err = -EAGAIN;
3090	goto out;
3091	}
3092
3093	map.m_lblk = pg_start;
3094	map.m_len = pg_end - pg_start;
3095	total = 0;
3096
3097	while (map.m_lblk < pg_end) {
3098	pgoff_t idx;
3099	int cnt = 0;
3100
3101	do_map:
3102	map.m_len = pg_end - map.m_lblk;
3103	err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
3104	if (err)
3105	goto clear_out;
3106
3107	if (!(map.m_flags & F2FS_MAP_FLAGS)) {
3108	map.m_lblk = next_pgofs;
3109	goto check;
3110	}
3111
3112	set_inode_flag(inode, FI_SKIP_WRITES);
3113
3114	idx = map.m_lblk;
3115	while (idx < map.m_lblk + map.m_len &&
3116	cnt < BLKS_PER_SEG(sbi)) {
3117	struct folio *folio;
3118
3119	folio = f2fs_get_lock_data_folio(inode, idx, true);
3120	if (IS_ERR(folio)) {
3121	err = PTR_ERR(folio);
3122	goto clear_out;
3123	}
3124
3125	f2fs_folio_wait_writeback(folio, DATA, true, true);
3126
3127	folio_mark_dirty(folio);
3128	folio_set_f2fs_gcing(folio);
3129	f2fs_folio_put(folio, true);
3130
3131	idx++;
3132	cnt++;
3133	total++;
3134	}
3135
3136	map.m_lblk = idx;
3137	check:
3138	if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi))
3139	goto do_map;
3140
3141	clear_inode_flag(inode, FI_SKIP_WRITES);
3142
3143	err = filemap_fdatawrite(inode->i_mapping);
3144	if (err)
3145	goto out;
3146	}
3147	clear_out:
3148	clear_inode_flag(inode, FI_SKIP_WRITES);
3149	out:
3150	clear_inode_flag(inode, FI_OPU_WRITE);
3151	unlock_out:
3152	inode_unlock(inode);
3153	if (!err) {
3154	range->len = (u64)total << PAGE_SHIFT;
3155	stat_inc_defrag_blk_count(sbi, total);
3156	}
3157	return err;
3158	}
3159
3160	static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
3161	{
3162	struct inode *inode = file_inode(filp);
3163	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3164	struct f2fs_defragment range;
3165	int err;
3166
3167	if (!capable(CAP_SYS_ADMIN))
3168	return -EPERM;
3169
3170	if (!S_ISREG(inode->i_mode))
3171	return -EINVAL;
3172
3173	if (f2fs_readonly(sbi->sb))
3174	return -EROFS;
3175
3176	if (IS_DEVICE_ALIASING(inode))
3177	return -EOPNOTSUPP;
3178
3179	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
3180	sizeof(range)))
3181	return -EFAULT;
3182
3183	/* verify alignment of offset & size */
3184	if (range.start & (F2FS_BLKSIZE - 1) \|\| range.len & (F2FS_BLKSIZE - 1))
3185	return -EINVAL;
3186
3187	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
3188	max_file_blocks(inode)))
3189	return -EINVAL;
3190
3191	err = mnt_want_write_file(filp);
3192	if (err)
3193	return err;
3194
3195	err = f2fs_defragment_range(sbi, filp, &range);
3196	mnt_drop_write_file(filp);
3197
3198	if (range.len)
3199	f2fs_update_time(sbi, REQ_TIME);
3200	if (err < 0)
3201	return err;
3202
3203	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
3204	sizeof(range)))
3205	return -EFAULT;
3206
3207	return 0;
3208	}
3209
3210	static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
3211	struct file *file_out, loff_t pos_out, size_t len)
3212	{
3213	struct inode *src = file_inode(file_in);
3214	struct inode *dst = file_inode(file_out);
3215	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
3216	struct f2fs_lock_context lc;
3217	size_t olen = len;
3218	loff_t dst_max_i_size = 0;
3219	loff_t dst_osize, dst_end;
3220	int ret;
3221
3222	if (file_in->f_path.mnt != file_out->f_path.mnt \|\|
3223	src->i_sb != dst->i_sb)
3224	return -EXDEV;
3225
3226	if (unlikely(f2fs_readonly(src->i_sb)))
3227	return -EROFS;
3228
3229	if (!S_ISREG(src->i_mode) \|\| !S_ISREG(dst->i_mode))
3230	return -EINVAL;
3231
3232	if (IS_ENCRYPTED(src) \|\| IS_ENCRYPTED(dst) \|\|
3233	IS_DEVICE_ALIASING(src) \|\| IS_DEVICE_ALIASING(dst))
3234	return -EOPNOTSUPP;
3235
3236	if (pos_out < 0 \|\| pos_in < 0)
3237	return -EINVAL;
3238
3239	if (src == dst) {
3240	if (pos_in == pos_out)
3241	return 0;
3242	}
3243
3244	inode_lock(src);
3245	if (src != dst) {
3246	ret = -EBUSY;
3247	if (!inode_trylock(dst))
3248	goto out;
3249	}
3250
3251	if (f2fs_compressed_file(src) \|\| f2fs_compressed_file(dst) \|\|
3252	f2fs_is_pinned_file(src) \|\| f2fs_is_pinned_file(dst)) {
3253	ret = -EOPNOTSUPP;
3254	goto out_unlock;
3255	}
3256
3257	if (f2fs_is_atomic_file(src) \|\| f2fs_is_atomic_file(dst)) {
3258	ret = -EINVAL;
3259	goto out_unlock;
3260	}
3261
3262	ret = -EINVAL;
3263	if (pos_in + len > src->i_size \|\| pos_in + len < pos_in)
3264	goto out_unlock;
3265	if (len == 0)
3266	olen = len = src->i_size - pos_in;
3267	if (src == dst && pos_out > pos_in && pos_out < pos_in + len)
3268	goto out_unlock;
3269	if (pos_in + len == src->i_size)
3270	len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
3271	if (len == 0) {
3272	ret = 0;
3273	goto out_unlock;
3274	}
3275
3276	dst_osize = dst->i_size;
3277	if (olen > LLONG_MAX - pos_out)
3278	goto out_unlock;
3279	dst_end = pos_out + olen;
3280	if (dst_end > dst->i_size) {
3281	ret = inode_newsize_ok(dst, dst_end);
3282	if (ret)
3283	goto out_unlock;
3284	dst_max_i_size = dst_end;
3285	}
3286
3287	/* verify the end result is block aligned */
3288	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) \|\|
3289	!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) \|\|
3290	!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
3291	goto out_unlock;
3292
3293	ret = f2fs_convert_inline_inode(src);
3294	if (ret)
3295	goto out_unlock;
3296
3297	ret = f2fs_convert_inline_inode(dst);
3298	if (ret)
3299	goto out_unlock;
3300
3301	/* write out all dirty pages from offset */
3302	ret = filemap_write_and_wait_range(src->i_mapping,
3303	pos_in, pos_in + len);
3304	if (ret)
3305	goto out_unlock;
3306
3307	ret = filemap_write_and_wait_range(dst->i_mapping,
3308	pos_out, pos_out + len);
3309	if (ret)
3310	goto out_unlock;
3311
3312	f2fs_balance_fs(sbi, true);
3313
3314	f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3315	if (src != dst) {
3316	ret = -EBUSY;
3317	if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
3318	goto out_src;
3319	}
3320
3321	f2fs_lock_op(sbi, &lc);
3322	ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in),
3323	F2FS_BYTES_TO_BLK(pos_out),
3324	F2FS_BYTES_TO_BLK(len), false);
3325
3326	if (!ret) {
3327	if (dst_max_i_size)
3328	f2fs_i_size_write(dst, dst_max_i_size);
3329	else if (dst_osize != dst->i_size)
3330	f2fs_i_size_write(dst, dst_osize);
3331	}
3332	f2fs_unlock_op(sbi, &lc);
3333
3334	if (src != dst)
3335	f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
3336	out_src:
3337	f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3338	if (ret)
3339	goto out_unlock;
3340
3341	inode_set_mtime_to_ts(src, inode_set_ctime_current(src));
3342	f2fs_mark_inode_dirty_sync(src, false);
3343	if (src != dst) {
3344	inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst));
3345	f2fs_mark_inode_dirty_sync(dst, false);
3346	}
3347	f2fs_update_time(sbi, REQ_TIME);
3348
3349	out_unlock:
3350	if (src != dst)
3351	inode_unlock(dst);
3352	out:
3353	inode_unlock(src);
3354	return ret;
3355	}
3356
3357	static int __f2fs_ioc_move_range(struct file *filp,
3358	struct f2fs_move_range *range)
3359	{
3360	int err;
3361
3362	if (!(filp->f_mode & FMODE_READ) \|\|
3363	!(filp->f_mode & FMODE_WRITE))
3364	return -EBADF;
3365
3366	CLASS(fd, dst)(range->dst_fd);
3367	if (fd_empty(dst))
3368	return -EBADF;
3369
3370	if (!(fd_file(dst)->f_mode & FMODE_WRITE))
3371	return -EBADF;
3372
3373	err = mnt_want_write_file(filp);
3374	if (err)
3375	return err;
3376
3377	err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst),
3378	range->pos_out, range->len);
3379
3380	mnt_drop_write_file(filp);
3381	return err;
3382	}
3383
3384	static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
3385	{
3386	struct f2fs_move_range range;
3387
3388	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
3389	sizeof(range)))
3390	return -EFAULT;
3391	return __f2fs_ioc_move_range(filp, &range);
3392	}
3393
3394	static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
3395	{
3396	struct inode *inode = file_inode(filp);
3397	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3398	struct sit_info *sm = SIT_I(sbi);
3399	unsigned int start_segno = 0, end_segno = 0;
3400	unsigned int dev_start_segno = 0, dev_end_segno = 0;
3401	struct f2fs_flush_device range;
3402	struct f2fs_gc_control gc_control = {
3403	.init_gc_type = FG_GC,
3404	.should_migrate_blocks = true,
3405	.err_gc_skipped = true,
3406	.nr_free_secs = 0 };
3407	int ret;
3408
3409	if (!capable(CAP_SYS_ADMIN))
3410	return -EPERM;
3411
3412	if (f2fs_readonly(sbi->sb))
3413	return -EROFS;
3414
3415	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3416	return -EINVAL;
3417
3418	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
3419	sizeof(range)))
3420	return -EFAULT;
3421
3422	if (!f2fs_is_multi_device(sbi) \|\| sbi->s_ndevs - 1 <= range.dev_num \|\|
3423	__is_large_section(sbi)) {
3424	f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1",
3425	range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi));
3426	return -EINVAL;
3427	}
3428
3429	ret = mnt_want_write_file(filp);
3430	if (ret)
3431	return ret;
3432
3433	if (range.dev_num != 0)
3434	dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
3435	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
3436
3437	start_segno = sm->last_victim[FLUSH_DEVICE];
3438	if (start_segno < dev_start_segno \|\| start_segno >= dev_end_segno)
3439	start_segno = dev_start_segno;
3440	end_segno = min(start_segno + range.segments, dev_end_segno);
3441
3442	while (start_segno < end_segno) {
3443	if (!f2fs_down_write_trylock_trace(&sbi->gc_lock, &gc_control.lc)) {
3444	ret = -EBUSY;
3445	goto out;
3446	}
3447	sm->last_victim[GC_CB] = end_segno + 1;
3448	sm->last_victim[GC_GREEDY] = end_segno + 1;
3449	sm->last_victim[ALLOC_NEXT] = end_segno + 1;
3450
3451	gc_control.victim_segno = start_segno;
3452	stat_inc_gc_call_count(sbi, FOREGROUND);
3453	ret = f2fs_gc(sbi, &gc_control);
3454	if (ret == -EAGAIN)
3455	ret = 0;
3456	else if (ret < 0)
3457	break;
3458	start_segno++;
3459	}
3460	out:
3461	mnt_drop_write_file(filp);
3462	return ret;
3463	}
3464
3465	static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3466	{
3467	struct inode *inode = file_inode(filp);
3468	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3469
3470	/* Must validate to set it with SQLite behavior in Android. */
3471	sb_feature \|= F2FS_FEATURE_ATOMIC_WRITE;
3472
3473	return put_user(sb_feature, (u32 __user *)arg);
3474	}
3475
3476	#ifdef CONFIG_QUOTA
3477	int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3478	{
3479	struct dquot *transfer_to[MAXQUOTAS] = {};
3480	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3481	struct super_block *sb = sbi->sb;
3482	int err;
3483
3484	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3485	if (IS_ERR(transfer_to[PRJQUOTA]))
3486	return PTR_ERR(transfer_to[PRJQUOTA]);
3487
3488	err = __dquot_transfer(inode, transfer_to);
3489	if (err)
3490	set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3491	dqput(transfer_to[PRJQUOTA]);
3492	return err;
3493	}
3494
3495	static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3496	{
3497	struct f2fs_inode_info *fi = F2FS_I(inode);
3498	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3499	struct f2fs_inode *ri = NULL;
3500	struct f2fs_lock_context lc;
3501	kprojid_t kprojid;
3502	int err;
3503
3504	if (!f2fs_sb_has_project_quota(sbi)) {
3505	if (projid != F2FS_DEF_PROJID)
3506	return -EOPNOTSUPP;
3507	else
3508	return 0;
3509	}
3510
3511	if (!f2fs_has_extra_attr(inode))
3512	return -EOPNOTSUPP;
3513
3514	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3515
3516	if (projid_eq(kprojid, fi->i_projid))
3517	return 0;
3518
3519	err = -EPERM;
3520	/* Is it quota file? Do not allow user to mess with it */
3521	if (IS_NOQUOTA(inode))
3522	return err;
3523
3524	if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3525	return -EOVERFLOW;
3526
3527	err = f2fs_dquot_initialize(inode);
3528	if (err)
3529	return err;
3530
3531	f2fs_lock_op(sbi, &lc);
3532	err = f2fs_transfer_project_quota(inode, kprojid);
3533	if (err)
3534	goto out_unlock;
3535
3536	fi->i_projid = kprojid;
3537	inode_set_ctime_current(inode);
3538	f2fs_mark_inode_dirty_sync(inode, true);
3539	out_unlock:
3540	f2fs_unlock_op(sbi, &lc);
3541	return err;
3542	}
3543	#else
3544	int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3545	{
3546	return 0;
3547	}
3548
3549	static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3550	{
3551	if (projid != F2FS_DEF_PROJID)
3552	return -EOPNOTSUPP;
3553	return 0;
3554	}
3555	#endif
3556
3557	int f2fs_fileattr_get(struct dentry dentry, struct file_kattr fa)
3558	{
3559	struct inode *inode = d_inode(dentry);
3560	struct f2fs_inode_info *fi = F2FS_I(inode);
3561	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3562
3563	if (IS_ENCRYPTED(inode))
3564	fsflags \|= FS_ENCRYPT_FL;
3565	if (IS_VERITY(inode))
3566	fsflags \|= FS_VERITY_FL;
3567	if (f2fs_has_inline_data(inode) \|\| f2fs_has_inline_dentry(inode))
3568	fsflags \|= FS_INLINE_DATA_FL;
3569	if (is_inode_flag_set(inode, FI_PIN_FILE))
3570	fsflags \|= FS_NOCOW_FL;
3571
3572	fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3573
3574	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3575	fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3576
3577	return 0;
3578	}
3579
3580	int f2fs_fileattr_set(const struct mnt_idmap *idmap,
3581	struct dentry dentry, struct file_kattr fa)
3582	{
3583	struct inode *inode = d_inode(dentry);
3584	u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3585	u32 iflags;
3586	int err;
3587
3588	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3589	return -EIO;
3590	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3591	return -ENOSPC;
3592	if (fsflags & ~F2FS_GETTABLE_FS_FL)
3593	return -EOPNOTSUPP;
3594	fsflags &= F2FS_SETTABLE_FS_FL;
3595	if (!fa->flags_valid)
3596	mask &= FS_COMMON_FL;
3597
3598	iflags = f2fs_fsflags_to_iflags(fsflags);
3599	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3600	return -EOPNOTSUPP;
3601
3602	err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3603	if (!err)
3604	err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3605
3606	return err;
3607	}
3608
3609	int f2fs_pin_file_control(struct inode *inode, bool inc)
3610	{
3611	struct f2fs_inode_info *fi = F2FS_I(inode);
3612	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3613
3614	if (IS_DEVICE_ALIASING(inode))
3615	return -EINVAL;
3616
3617	if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) {
3618	f2fs_warn(sbi, "%s: Enable GC = ino %llx after %x GC trials",
3619	__func__, inode->i_ino, fi->i_gc_failures);
3620	clear_inode_flag(inode, FI_PIN_FILE);
3621	return -EAGAIN;
3622	}
3623
3624	/* Use i_gc_failures for normal file as a risk signal. */
3625	if (inc)
3626	f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
3627
3628	return 0;
3629	}
3630
3631	static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3632	{
3633	struct inode *inode = file_inode(filp);
3634	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3635	__u32 pin;
3636	int ret = 0;
3637
3638	if (get_user(pin, (__u32 __user *)arg))
3639	return -EFAULT;
3640
3641	if (!S_ISREG(inode->i_mode))
3642	return -EINVAL;
3643
3644	if (f2fs_readonly(sbi->sb))
3645	return -EROFS;
3646
3647	if (!pin && IS_DEVICE_ALIASING(inode))
3648	return -EOPNOTSUPP;
3649
3650	ret = mnt_want_write_file(filp);
3651	if (ret)
3652	return ret;
3653
3654	inode_lock(inode);
3655
3656	if (f2fs_is_atomic_file(inode)) {
3657	ret = -EINVAL;
3658	goto out;
3659	}
3660
3661	if (!pin) {
3662	clear_inode_flag(inode, FI_PIN_FILE);
3663	f2fs_i_gc_failures_write(inode, 0);
3664	goto done;
3665	} else if (f2fs_is_pinned_file(inode)) {
3666	goto done;
3667	}
3668
3669	if (F2FS_HAS_BLOCKS(inode)) {
3670	ret = -EFBIG;
3671	goto out;
3672	}
3673
3674	/* Let's allow file pinning on zoned device. */
3675	if (!f2fs_sb_has_blkzoned(sbi) &&
3676	f2fs_should_update_outplace(inode, NULL)) {
3677	ret = -EINVAL;
3678	goto out;
3679	}
3680
3681	if (f2fs_pin_file_control(inode, false)) {
3682	ret = -EAGAIN;
3683	goto out;
3684	}
3685
3686	ret = f2fs_convert_inline_inode(inode);
3687	if (ret)
3688	goto out;
3689
3690	if (!f2fs_disable_compressed_file(inode)) {
3691	ret = -EOPNOTSUPP;
3692	goto out;
3693	}
3694
3695	set_inode_flag(inode, FI_PIN_FILE);
3696	ret = F2FS_I(inode)->i_gc_failures;
3697	done:
3698	f2fs_update_time(sbi, REQ_TIME);
3699	out:
3700	inode_unlock(inode);
3701	mnt_drop_write_file(filp);
3702	return ret;
3703	}
3704
3705	static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3706	{
3707	struct inode *inode = file_inode(filp);
3708	__u32 pin = 0;
3709
3710	if (is_inode_flag_set(inode, FI_PIN_FILE))
3711	pin = F2FS_I(inode)->i_gc_failures;
3712	return put_user(pin, (u32 __user *)arg);
3713	}
3714
3715	static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg)
3716	{
3717	return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0,
3718	(u32 __user *)arg);
3719	}
3720
3721	static bool f2fs_get_dev_alias_extent(struct f2fs_sb_info *sbi,
3722	struct dentry *dentry,
3723	struct extent_info *ei)
3724	{
3725	int i;
3726
3727	for (i = 1; i < sbi->s_ndevs; i++) {
3728	char *name = strrchr(FDEV(i).path, '/');
3729
3730	name = name ? name + 1 : FDEV(i).path;
3731	if (strcmp(name, dentry->d_name.name))
3732	continue;
3733
3734	ei->blk = FDEV(i).start_blk;
3735	ei->len = FDEV(i).total_segments << sbi->log_blocks_per_seg;
3736	ei->fofs = 0;
3737	return true;
3738	}
3739	return false;
3740	}
3741
3742	static int f2fs_ioc_reserve_dev_alias(struct file *filp)
3743	{
3744	struct inode *inode = file_inode(filp);
3745	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3746	struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
3747	struct extent_info ei;
3748	struct cp_control cpc = { CP_SYNC, 0, 0, 0 };
3749	struct f2fs_lock_context lc, glc;
3750	blkcnt_t count;
3751	unsigned int start, end;
3752	int type, err;
3753
3754	if (!capable(CAP_SYS_ADMIN))
3755	return -EPERM;
3756
3757	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3758	return -EINVAL;
3759
3760	err = mnt_want_write_file(filp);
3761	if (err)
3762	return err;
3763
3764	inode_lock(inode);
3765
3766	if (!IS_DEVICE_ALIASING(inode)) {
3767	err = -EINVAL;
3768	goto out_inode_unlock;
3769	}
3770
3771	if (F2FS_HAS_BLOCKS(inode)) {
3772	err = 0;
3773	goto out_inode_unlock;
3774	}
3775
3776	if (!f2fs_get_dev_alias_extent(sbi, filp->f_path.dentry, &ei)) {
3777	f2fs_warn(sbi, "device alias file (%s, ino=%llu) has no matching device",
3778	filp->f_path.dentry->d_name.name,
3779	(unsigned long long)inode->i_ino);
3780	set_sbi_flag(sbi, SBI_NEED_FSCK);
3781	f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
3782	err = -EFSCORRUPTED;
3783	goto out_inode_unlock;
3784	}
3785
3786	spin_lock(&sbi->stat_lock);
3787	if (sbi->total_valid_block_count + ei.len >
3788	get_available_block_count(sbi, inode, true)) {
3789	spin_unlock(&sbi->stat_lock);
3790	err = -ENOSPC;
3791	goto out_inode_unlock;
3792	}
3793	sbi->alias_reserved_blocks += ei.len;
3794	spin_unlock(&sbi->stat_lock);
3795
3796	spin_lock(&FREE_I(sbi)->segmap_lock);
3797	FDEV(f2fs_target_device_index(sbi, ei.blk)).is_reserving = true;
3798	spin_unlock(&FREE_I(sbi)->segmap_lock);
3799
3800	start = GET_SEGNO(sbi, ei.blk);
3801	end = GET_SEGNO(sbi, ei.blk + ei.len - 1);
3802
3803	/* Acquire gc_lock for victim reset, curseg resize, and range GC */
3804	f2fs_down_write_trace(&sbi->gc_lock, &glc);
3805
3806	/* Reset the victim information to prevent GC from targeting the range */
3807	f2fs_reset_gc_victim_resource(sbi, start, end);
3808
3809	/* Move out cursegs from the target range */
3810	for (type = CURSEG_HOT_DATA; type < NR_CURSEG_PERSIST_TYPE; type++) {
3811	err = f2fs_allocate_segment_for_resize(sbi, type, start, end);
3812	if (err)
3813	goto out_gc_unlock;
3814	}
3815
3816	f2fs_lock_op(sbi, &lc);
3817
3818	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
3819	err = -EINVAL;
3820	f2fs_unlock_op(sbi, &lc);
3821	goto out_gc_unlock;
3822	}
3823
3824	/* do GC to move out valid blocks in the range all at once! */
3825	err = f2fs_gc_range(sbi, start, end, false, 0, false);
3826	if (err) {
3827	f2fs_unlock_op(sbi, &lc);
3828	goto out_gc_unlock;
3829	}
3830
3831	count = ei.len;
3832	err = inc_valid_block_count(sbi, inode, &count, false, true);
3833	if (err) {
3834	f2fs_unlock_op(sbi, &lc);
3835	goto out_gc_unlock;
3836	}
3837
3838	write_lock(&et->lock);
3839	et->largest = ei;
3840	write_unlock(&et->lock);
3841	clear_inode_flag(inode, FI_NO_EXTENT);
3842
3843	f2fs_reserve_device_alias(sbi, ei.blk, ei.len);
3844
3845	i_size_write(inode, (loff_t)ei.len << sbi->log_blocksize);
3846	f2fs_update_inode_page(inode);
3847
3848	spin_lock(&FREE_I(sbi)->segmap_lock);
3849	FDEV(f2fs_target_device_index(sbi, ei.blk)).is_reserving = false;
3850	spin_unlock(&FREE_I(sbi)->segmap_lock);
3851
3852	f2fs_unlock_op(sbi, &lc);
3853	f2fs_up_write_trace(&sbi->gc_lock, &glc);
3854
3855	inode_unlock(inode);
3856	mnt_drop_write_file(filp);
3857
3858	return f2fs_write_checkpoint(sbi, &cpc);
3859
3860	out_gc_unlock:
3861	spin_lock(&sbi->stat_lock);
3862	sbi->alias_reserved_blocks -= ei.len;
3863	spin_unlock(&sbi->stat_lock);
3864
3865	spin_lock(&FREE_I(sbi)->segmap_lock);
3866	FDEV(f2fs_target_device_index(sbi, ei.blk)).is_reserving = false;
3867	spin_unlock(&FREE_I(sbi)->segmap_lock);
3868	f2fs_up_write_trace(&sbi->gc_lock, &glc);
3869
3870	out_inode_unlock:
3871	inode_unlock(inode);
3872	mnt_drop_write_file(filp);
3873	return err;
3874	}
3875
3876	static int f2fs_ioc_release_dev_alias(struct file *filp)
3877	{
3878	struct inode *inode = file_inode(filp);
3879	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3880	struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
3881	struct extent_info ei = {0, };
3882	struct cp_control cpc = { CP_SYNC, 0, 0, 0 };
3883	struct f2fs_lock_context lc, glc;
3884	int err;
3885
3886	if (!capable(CAP_SYS_ADMIN))
3887	return -EPERM;
3888
3889	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3890	return -EINVAL;
3891
3892	err = mnt_want_write_file(filp);
3893	if (err)
3894	return err;
3895
3896	inode_lock(inode);
3897
3898	if (!IS_DEVICE_ALIASING(inode)) {
3899	err = -EINVAL;
3900	goto out_inode_unlock;
3901	}
3902
3903	if (!F2FS_HAS_BLOCKS(inode)) {
3904	err = 0;
3905	goto out_inode_unlock;
3906	}
3907
3908	err = filemap_write_and_wait(inode->i_mapping);
3909	if (err)
3910	goto out_inode_unlock;
3911
3912	read_lock(&et->lock);
3913	ei = et->largest;
3914	read_unlock(&et->lock);
3915
3916	f2fs_down_write_trace(&sbi->gc_lock, &glc);
3917	f2fs_lock_op(sbi, &lc);
3918
3919	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
3920	err = -EINVAL;
3921	f2fs_unlock_op(sbi, &lc);
3922	f2fs_up_write_trace(&sbi->gc_lock, &glc);
3923	goto out_inode_unlock;
3924	}
3925
3926	filemap_invalidate_lock(inode->i_mapping);
3927	truncate_setsize(inode, 0);
3928
3929	err = f2fs_truncate_blocks(inode, 0, false);
3930	if (err)
3931	i_size_write(inode, (loff_t)ei.len << sbi->log_blocksize);
3932	filemap_invalidate_unlock(inode->i_mapping);
3933
3934	if (err) {
3935	f2fs_unlock_op(sbi, &lc);
3936	f2fs_up_write_trace(&sbi->gc_lock, &glc);
3937	goto out_inode_unlock;
3938	}
3939
3940	f2fs_update_inode_page(inode);
3941
3942	f2fs_unlock_op(sbi, &lc);
3943	f2fs_up_write_trace(&sbi->gc_lock, &glc);
3944
3945	inode_unlock(inode);
3946	mnt_drop_write_file(filp);
3947
3948	return f2fs_write_checkpoint(sbi, &cpc);
3949
3950	out_inode_unlock:
3951	inode_unlock(inode);
3952	mnt_drop_write_file(filp);
3953	return err;
3954	}
3955
3956	static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg)
3957	{
3958	struct inode *inode = file_inode(filp);
3959	__u32 level;
3960
3961	if (get_user(level, (__u32 __user *)arg))
3962	return -EFAULT;
3963
3964	if (!S_ISREG(inode->i_mode) \|\| level >= F2FS_IOPRIO_MAX)
3965	return -EINVAL;
3966
3967	inode_lock(inode);
3968	F2FS_I(inode)->ioprio_hint = level;
3969	inode_unlock(inode);
3970	return 0;
3971	}
3972
3973	int f2fs_precache_extents(struct inode *inode)
3974	{
3975	struct f2fs_inode_info *fi = F2FS_I(inode);
3976	struct f2fs_map_blocks map;
3977	pgoff_t m_next_extent;
3978	loff_t end;
3979	int err;
3980
3981	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3982	return -EOPNOTSUPP;
3983
3984	map.m_lblk = 0;
3985	map.m_pblk = 0;
3986	map.m_next_pgofs = NULL;
3987	map.m_next_extent = &m_next_extent;
3988	map.m_seg_type = NO_CHECK_TYPE;
3989	map.m_may_create = false;
3990	end = F2FS_BLK_ALIGN(i_size_read(inode));
3991
3992	while (map.m_lblk < end) {
3993	map.m_len = end - map.m_lblk;
3994
3995	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3996	err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
3997	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3998	if (err \|\| !map.m_len)
3999	return err;
4000
4001	map.m_lblk = m_next_extent;
4002	}
4003
4004	return 0;
4005	}
4006
4007	static int f2fs_ioc_precache_extents(struct file *filp)
4008	{
4009	return f2fs_precache_extents(file_inode(filp));
4010	}
4011
4012	static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
4013	{
4014	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
4015	__u64 block_count;
4016
4017	if (!capable(CAP_SYS_ADMIN))
4018	return -EPERM;
4019
4020	if (f2fs_readonly(sbi->sb))
4021	return -EROFS;
4022
4023	if (copy_from_user(&block_count, (void __user *)arg,
4024	sizeof(block_count)))
4025	return -EFAULT;
4026
4027	return f2fs_resize_fs(filp, block_count);
4028	}
4029
4030	static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
4031	{
4032	struct inode *inode = file_inode(filp);
4033
4034	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
4035
4036	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
4037	f2fs_warn(F2FS_I_SB(inode),
4038	"Can't enable fs-verity on inode %llu: the verity feature is not enabled on this filesystem",
4039	inode->i_ino);
4040	return -EOPNOTSUPP;
4041	}
4042
4043	return fsverity_ioctl_enable(filp, (const void __user *)arg);
4044	}
4045
4046	static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
4047	{
4048	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
4049	return -EOPNOTSUPP;
4050
4051	return fsverity_ioctl_measure(filp, (void __user *)arg);
4052	}
4053
4054	static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
4055	{
4056	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
4057	return -EOPNOTSUPP;
4058
4059	return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
4060	}
4061
4062	static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
4063	{
4064	struct inode *inode = file_inode(filp);
4065	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4066	char *vbuf;
4067	int count;
4068	int err = 0;
4069
4070	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
4071	if (!vbuf)
4072	return -ENOMEM;
4073
4074	f2fs_down_read(&sbi->sb_lock);
4075	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
4076	ARRAY_SIZE(sbi->raw_super->volume_name),
4077	UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
4078	f2fs_up_read(&sbi->sb_lock);
4079
4080	if (copy_to_user((char __user *)arg, vbuf,
4081	min(FSLABEL_MAX, count)))
4082	err = -EFAULT;
4083
4084	kfree(vbuf);
4085	return err;
4086	}
4087
4088	static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
4089	{
4090	struct inode *inode = file_inode(filp);
4091	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4092	char *vbuf;
4093	int err = 0;
4094
4095	if (!capable(CAP_SYS_ADMIN))
4096	return -EPERM;
4097
4098	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
4099	if (IS_ERR(vbuf))
4100	return PTR_ERR(vbuf);
4101
4102	err = mnt_want_write_file(filp);
4103	if (err)
4104	goto out;
4105
4106	f2fs_down_write(&sbi->sb_lock);
4107
4108	memset(sbi->raw_super->volume_name, 0,
4109	sizeof(sbi->raw_super->volume_name));
4110	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
4111	sbi->raw_super->volume_name,
4112	ARRAY_SIZE(sbi->raw_super->volume_name));
4113
4114	err = f2fs_commit_super(sbi, false);
4115
4116	f2fs_up_write(&sbi->sb_lock);
4117
4118	mnt_drop_write_file(filp);
4119	out:
4120	kfree(vbuf);
4121	return err;
4122	}
4123
4124	static int f2fs_get_compress_blocks(struct inode inode, __u64 blocks)
4125	{
4126	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4127	return -EOPNOTSUPP;
4128
4129	if (!f2fs_compressed_file(inode))
4130	return -EINVAL;
4131
4132	*blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
4133
4134	return 0;
4135	}
4136
4137	static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg)
4138	{
4139	struct inode *inode = file_inode(filp);
4140	__u64 blocks;
4141	int ret;
4142
4143	ret = f2fs_get_compress_blocks(inode, &blocks);
4144	if (ret < 0)
4145	return ret;
4146
4147	return put_user(blocks, (u64 __user *)arg);
4148	}
4149
4150	static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
4151	{
4152	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
4153	unsigned int released_blocks = 0;
4154	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
4155	block_t blkaddr;
4156	int i;
4157
4158	for (i = 0; i < count; i++) {
4159	blkaddr = data_blkaddr(dn->inode, dn->node_folio,
4160	dn->ofs_in_node + i);
4161
4162	if (!__is_valid_data_blkaddr(blkaddr))
4163	continue;
4164	if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
4165	DATA_GENERIC_ENHANCE)))
4166	return -EFSCORRUPTED;
4167	}
4168
4169	while (count) {
4170	int compr_blocks = 0;
4171
4172	for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
4173	blkaddr = f2fs_data_blkaddr(dn);
4174
4175	if (i == 0) {
4176	if (blkaddr == COMPRESS_ADDR)
4177	continue;
4178	dn->ofs_in_node += cluster_size;
4179	goto next;
4180	}
4181
4182	if (__is_valid_data_blkaddr(blkaddr))
4183	compr_blocks++;
4184
4185	if (blkaddr != NEW_ADDR)
4186	continue;
4187
4188	f2fs_set_data_blkaddr(dn, NULL_ADDR);
4189	}
4190
4191	f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
4192	dec_valid_block_count(sbi, dn->inode,
4193	cluster_size - compr_blocks);
4194
4195	released_blocks += cluster_size - compr_blocks;
4196	next:
4197	count -= cluster_size;
4198	}
4199
4200	return released_blocks;
4201	}
4202
4203	static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
4204	{
4205	struct inode *inode = file_inode(filp);
4206	struct f2fs_inode_info *fi = F2FS_I(inode);
4207	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4208	struct f2fs_lock_context lc;
4209	pgoff_t page_idx = 0, last_idx;
4210	unsigned int released_blocks = 0;
4211	int ret;
4212	int writecount;
4213
4214	if (!f2fs_sb_has_compression(sbi))
4215	return -EOPNOTSUPP;
4216
4217	if (f2fs_readonly(sbi->sb))
4218	return -EROFS;
4219
4220	ret = mnt_want_write_file(filp);
4221	if (ret)
4222	return ret;
4223
4224	f2fs_balance_fs(sbi, true);
4225
4226	inode_lock(inode);
4227
4228	writecount = atomic_read(&inode->i_writecount);
4229	if ((filp->f_mode & FMODE_WRITE && writecount != 1) \|\|
4230	(!(filp->f_mode & FMODE_WRITE) && writecount)) {
4231	ret = -EBUSY;
4232	goto out;
4233	}
4234
4235	if (!f2fs_compressed_file(inode) \|\|
4236	is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4237	ret = -EINVAL;
4238	goto out;
4239	}
4240
4241	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4242	if (ret)
4243	goto out;
4244
4245	if (!atomic_read(&fi->i_compr_blocks)) {
4246	ret = -EPERM;
4247	goto out;
4248	}
4249
4250	set_inode_flag(inode, FI_COMPRESS_RELEASED);
4251	inode_set_ctime_current(inode);
4252	f2fs_mark_inode_dirty_sync(inode, true);
4253
4254	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
4255	filemap_invalidate_lock(inode->i_mapping);
4256
4257	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4258
4259	while (page_idx < last_idx) {
4260	struct dnode_of_data dn;
4261	pgoff_t end_offset, count;
4262
4263	f2fs_lock_op(sbi, &lc);
4264
4265	set_new_dnode(&dn, inode, NULL, NULL, 0);
4266	ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
4267	if (ret) {
4268	f2fs_unlock_op(sbi, &lc);
4269	if (ret == -ENOENT) {
4270	page_idx = f2fs_get_next_page_offset(&dn,
4271	page_idx);
4272	ret = 0;
4273	continue;
4274	}
4275	break;
4276	}
4277
4278	end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
4279	count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
4280	count = round_up(count, fi->i_cluster_size);
4281
4282	ret = release_compress_blocks(&dn, count);
4283
4284	f2fs_put_dnode(&dn);
4285
4286	f2fs_unlock_op(sbi, &lc);
4287
4288	if (ret < 0)
4289	break;
4290
4291	page_idx += count;
4292	released_blocks += ret;
4293	}
4294
4295	filemap_invalidate_unlock(inode->i_mapping);
4296	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
4297	out:
4298	if (released_blocks)
4299	f2fs_update_time(sbi, REQ_TIME);
4300	inode_unlock(inode);
4301
4302	mnt_drop_write_file(filp);
4303
4304	if (ret >= 0) {
4305	ret = put_user(released_blocks, (u64 __user *)arg);
4306	} else if (released_blocks &&
4307	atomic_read(&fi->i_compr_blocks)) {
4308	set_sbi_flag(sbi, SBI_NEED_FSCK);
4309	f2fs_warn(sbi, "%s: partial blocks were released i_ino=%llx "
4310	"iblocks=%llu, released=%u, compr_blocks=%u, "
4311	"run fsck to fix.",
4312	__func__, inode->i_ino, inode->i_blocks,
4313	released_blocks,
4314	atomic_read(&fi->i_compr_blocks));
4315	}
4316
4317	return ret;
4318	}
4319
4320	static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count,
4321	unsigned int *reserved_blocks)
4322	{
4323	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
4324	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
4325	block_t blkaddr;
4326	int i;
4327
4328	for (i = 0; i < count; i++) {
4329	blkaddr = data_blkaddr(dn->inode, dn->node_folio,
4330	dn->ofs_in_node + i);
4331
4332	if (!__is_valid_data_blkaddr(blkaddr))
4333	continue;
4334	if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
4335	DATA_GENERIC_ENHANCE)))
4336	return -EFSCORRUPTED;
4337	}
4338
4339	while (count) {
4340	int compr_blocks = 0;
4341	blkcnt_t reserved = 0;
4342	blkcnt_t to_reserved;
4343	int ret;
4344
4345	for (i = 0; i < cluster_size; i++) {
4346	blkaddr = data_blkaddr(dn->inode, dn->node_folio,
4347	dn->ofs_in_node + i);
4348
4349	if (i == 0) {
4350	if (blkaddr != COMPRESS_ADDR) {
4351	dn->ofs_in_node += cluster_size;
4352	goto next;
4353	}
4354	continue;
4355	}
4356
4357	/*
4358	* compressed cluster was not released due to it
4359	* fails in release_compress_blocks(), so NEW_ADDR
4360	* is a possible case.
4361	*/
4362	if (blkaddr == NEW_ADDR) {
4363	reserved++;
4364	continue;
4365	}
4366	if (__is_valid_data_blkaddr(blkaddr)) {
4367	compr_blocks++;
4368	continue;
4369	}
4370	}
4371
4372	to_reserved = cluster_size - compr_blocks - reserved;
4373
4374	/* for the case all blocks in cluster were reserved */
4375	if (reserved && to_reserved == 1) {
4376	dn->ofs_in_node += cluster_size;
4377	goto next;
4378	}
4379
4380	ret = inc_valid_block_count(sbi, dn->inode,
4381	&to_reserved, false, false);
4382	if (unlikely(ret))
4383	return ret;
4384
4385	for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
4386	if (f2fs_data_blkaddr(dn) == NULL_ADDR)
4387	f2fs_set_data_blkaddr(dn, NEW_ADDR);
4388	}
4389
4390	f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
4391
4392	*reserved_blocks += to_reserved;
4393	next:
4394	count -= cluster_size;
4395	}
4396
4397	return 0;
4398	}
4399
4400	static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
4401	{
4402	struct inode *inode = file_inode(filp);
4403	struct f2fs_inode_info *fi = F2FS_I(inode);
4404	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4405	pgoff_t page_idx = 0, last_idx;
4406	unsigned int reserved_blocks = 0;
4407	int ret;
4408
4409	if (!f2fs_sb_has_compression(sbi))
4410	return -EOPNOTSUPP;
4411
4412	if (f2fs_readonly(sbi->sb))
4413	return -EROFS;
4414
4415	ret = mnt_want_write_file(filp);
4416	if (ret)
4417	return ret;
4418
4419	f2fs_balance_fs(sbi, true);
4420
4421	inode_lock(inode);
4422
4423	if (!f2fs_compressed_file(inode) \|\|
4424	!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4425	ret = -EINVAL;
4426	goto unlock_inode;
4427	}
4428
4429	if (atomic_read(&fi->i_compr_blocks))
4430	goto unlock_inode;
4431
4432	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
4433	filemap_invalidate_lock(inode->i_mapping);
4434
4435	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4436
4437	while (page_idx < last_idx) {
4438	struct dnode_of_data dn;
4439	struct f2fs_lock_context lc;
4440	pgoff_t end_offset, count;
4441
4442	f2fs_lock_op(sbi, &lc);
4443
4444	set_new_dnode(&dn, inode, NULL, NULL, 0);
4445	ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
4446	if (ret) {
4447	f2fs_unlock_op(sbi, &lc);
4448	if (ret == -ENOENT) {
4449	page_idx = f2fs_get_next_page_offset(&dn,
4450	page_idx);
4451	ret = 0;
4452	continue;
4453	}
4454	break;
4455	}
4456
4457	end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
4458	count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
4459	count = round_up(count, fi->i_cluster_size);
4460
4461	ret = reserve_compress_blocks(&dn, count, &reserved_blocks);
4462
4463	f2fs_put_dnode(&dn);
4464
4465	f2fs_unlock_op(sbi, &lc);
4466
4467	if (ret < 0)
4468	break;
4469
4470	page_idx += count;
4471	}
4472
4473	filemap_invalidate_unlock(inode->i_mapping);
4474	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
4475
4476	if (!ret) {
4477	clear_inode_flag(inode, FI_COMPRESS_RELEASED);
4478	inode_set_ctime_current(inode);
4479	f2fs_mark_inode_dirty_sync(inode, true);
4480	}
4481	unlock_inode:
4482	if (reserved_blocks)
4483	f2fs_update_time(sbi, REQ_TIME);
4484	inode_unlock(inode);
4485	mnt_drop_write_file(filp);
4486
4487	if (!ret) {
4488	ret = put_user(reserved_blocks, (u64 __user *)arg);
4489	} else if (reserved_blocks &&
4490	atomic_read(&fi->i_compr_blocks)) {
4491	set_sbi_flag(sbi, SBI_NEED_FSCK);
4492	f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%llx "
4493	"iblocks=%llu, reserved=%u, compr_blocks=%u, "
4494	"run fsck to fix.",
4495	__func__, inode->i_ino, inode->i_blocks,
4496	reserved_blocks,
4497	atomic_read(&fi->i_compr_blocks));
4498	}
4499
4500	return ret;
4501	}
4502
4503	static int f2fs_secure_erase(struct block_device bdev, struct inode inode,
4504	pgoff_t off, block_t block, block_t len, u32 flags)
4505	{
4506	sector_t sector = SECTOR_FROM_BLOCK(block);
4507	sector_t nr_sects = SECTOR_FROM_BLOCK(len);
4508	int ret = 0;
4509
4510	if (flags & F2FS_TRIM_FILE_DISCARD) {
4511	if (bdev_max_secure_erase_sectors(bdev))
4512	ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
4513	GFP_NOFS);
4514	else
4515	ret = blkdev_issue_discard(bdev, sector, nr_sects,
4516	GFP_NOFS);
4517	}
4518
4519	if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
4520	if (IS_ENCRYPTED(inode))
4521	ret = fscrypt_zeroout_range(inode,
4522	(loff_t)off << inode->i_blkbits, sector,
4523	(u64)len << inode->i_blkbits);
4524	else
4525	ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
4526	GFP_NOFS, 0);
4527	}
4528
4529	return ret;
4530	}
4531
4532	static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
4533	{
4534	struct inode *inode = file_inode(filp);
4535	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4536	struct address_space *mapping = inode->i_mapping;
4537	struct block_device *prev_bdev = NULL;
4538	struct f2fs_sectrim_range range;
4539	pgoff_t index, pg_end, prev_index = 0;
4540	block_t prev_block = 0, len = 0;
4541	loff_t end_addr;
4542	bool to_end = false;
4543	int ret = 0;
4544
4545	if (!(filp->f_mode & FMODE_WRITE))
4546	return -EBADF;
4547
4548	if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
4549	sizeof(range)))
4550	return -EFAULT;
4551
4552	if (range.flags == 0 \|\| (range.flags & ~F2FS_TRIM_FILE_MASK) \|\|
4553	!S_ISREG(inode->i_mode))
4554	return -EINVAL;
4555
4556	if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
4557	!f2fs_hw_support_discard(sbi)) \|\|
4558	((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
4559	IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
4560	return -EOPNOTSUPP;
4561
4562	ret = mnt_want_write_file(filp);
4563	if (ret)
4564	return ret;
4565	inode_lock(inode);
4566
4567	if (f2fs_is_atomic_file(inode) \|\| f2fs_compressed_file(inode) \|\|
4568	range.start >= inode->i_size) {
4569	ret = -EINVAL;
4570	goto err;
4571	}
4572
4573	if (range.len == 0)
4574	goto err;
4575
4576	if (inode->i_size - range.start > range.len) {
4577	end_addr = range.start + range.len;
4578	} else {
4579	end_addr = range.len == (u64)-1 ?
4580	sbi->sb->s_maxbytes : inode->i_size;
4581	to_end = true;
4582	}
4583
4584	if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) \|\|
4585	(!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
4586	ret = -EINVAL;
4587	goto err;
4588	}
4589
4590	index = F2FS_BYTES_TO_BLK(range.start);
4591	pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
4592
4593	ret = f2fs_convert_inline_inode(inode);
4594	if (ret)
4595	goto err;
4596
4597	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4598	filemap_invalidate_lock(mapping);
4599
4600	ret = filemap_write_and_wait_range(mapping, range.start,
4601	to_end ? LLONG_MAX : end_addr - 1);
4602	if (ret)
4603	goto out;
4604
4605	truncate_inode_pages_range(mapping, range.start,
4606	to_end ? -1 : end_addr - 1);
4607
4608	while (index < pg_end) {
4609	struct dnode_of_data dn;
4610	pgoff_t end_offset, count;
4611	int i;
4612
4613	set_new_dnode(&dn, inode, NULL, NULL, 0);
4614	ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4615	if (ret) {
4616	if (ret == -ENOENT) {
4617	index = f2fs_get_next_page_offset(&dn, index);
4618	continue;
4619	}
4620	goto out;
4621	}
4622
4623	end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
4624	count = min(end_offset - dn.ofs_in_node, pg_end - index);
4625	for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
4626	struct block_device *cur_bdev;
4627	block_t blkaddr = f2fs_data_blkaddr(&dn);
4628
4629	if (!__is_valid_data_blkaddr(blkaddr))
4630	continue;
4631
4632	if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
4633	DATA_GENERIC_ENHANCE)) {
4634	ret = -EFSCORRUPTED;
4635	f2fs_put_dnode(&dn);
4636	goto out;
4637	}
4638
4639	cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
4640	if (f2fs_is_multi_device(sbi)) {
4641	int di = f2fs_target_device_index(sbi, blkaddr);
4642
4643	blkaddr -= FDEV(di).start_blk;
4644	}
4645
4646	if (len) {
4647	if (prev_bdev == cur_bdev &&
4648	index == prev_index + len &&
4649	blkaddr == prev_block + len) {
4650	len++;
4651	} else {
4652	ret = f2fs_secure_erase(prev_bdev,
4653	inode, prev_index, prev_block,
4654	len, range.flags);
4655	if (ret) {
4656	f2fs_put_dnode(&dn);
4657	goto out;
4658	}
4659
4660	len = 0;
4661	}
4662	}
4663
4664	if (!len) {
4665	prev_bdev = cur_bdev;
4666	prev_index = index;
4667	prev_block = blkaddr;
4668	len = 1;
4669	}
4670	}
4671
4672	f2fs_put_dnode(&dn);
4673
4674	if (fatal_signal_pending(current)) {
4675	ret = -EINTR;
4676	goto out;
4677	}
4678	cond_resched();
4679	}
4680
4681	if (len)
4682	ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
4683	prev_block, len, range.flags);
4684	f2fs_update_time(sbi, REQ_TIME);
4685	out:
4686	filemap_invalidate_unlock(mapping);
4687	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4688	err:
4689	inode_unlock(inode);
4690	mnt_drop_write_file(filp);
4691
4692	return ret;
4693	}
4694
4695	static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
4696	{
4697	struct inode *inode = file_inode(filp);
4698	struct f2fs_comp_option option;
4699
4700	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4701	return -EOPNOTSUPP;
4702
4703	inode_lock_shared(inode);
4704
4705	if (!f2fs_compressed_file(inode)) {
4706	inode_unlock_shared(inode);
4707	return -ENODATA;
4708	}
4709
4710	option.algorithm = F2FS_I(inode)->i_compress_algorithm;
4711	option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
4712
4713	inode_unlock_shared(inode);
4714
4715	if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
4716	sizeof(option)))
4717	return -EFAULT;
4718
4719	return 0;
4720	}
4721
4722	static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4723	{
4724	struct inode *inode = file_inode(filp);
4725	struct f2fs_inode_info *fi = F2FS_I(inode);
4726	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4727	struct f2fs_comp_option option;
4728	int ret = 0;
4729
4730	if (!f2fs_sb_has_compression(sbi))
4731	return -EOPNOTSUPP;
4732
4733	if (!(filp->f_mode & FMODE_WRITE))
4734	return -EBADF;
4735
4736	if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4737	sizeof(option)))
4738	return -EFAULT;
4739
4740	if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE \|\|
4741	option.log_cluster_size > MAX_COMPRESS_LOG_SIZE \|\|
4742	option.algorithm >= COMPRESS_MAX)
4743	return -EINVAL;
4744
4745	ret = mnt_want_write_file(filp);
4746	if (ret)
4747	return ret;
4748	inode_lock(inode);
4749
4750	f2fs_down_write(&F2FS_I(inode)->i_sem);
4751	if (!f2fs_compressed_file(inode)) {
4752	ret = -EINVAL;
4753	goto out;
4754	}
4755
4756	if (f2fs_is_mmap_file(inode) \|\| get_dirty_pages(inode)) {
4757	ret = -EBUSY;
4758	goto out;
4759	}
4760
4761	if (F2FS_HAS_BLOCKS(inode)) {
4762	ret = -EFBIG;
4763	goto out;
4764	}
4765
4766	fi->i_compress_algorithm = option.algorithm;
4767	fi->i_log_cluster_size = option.log_cluster_size;
4768	fi->i_cluster_size = BIT(option.log_cluster_size);
4769	/* Set default level */
4770	if (fi->i_compress_algorithm == COMPRESS_ZSTD)
4771	fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
4772	else
4773	fi->i_compress_level = 0;
4774	/* Adjust mount option level */
4775	if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm &&
4776	F2FS_OPTION(sbi).compress_level)
4777	fi->i_compress_level = F2FS_OPTION(sbi).compress_level;
4778	f2fs_mark_inode_dirty_sync(inode, true);
4779
4780	if (!f2fs_is_compress_backend_ready(inode))
4781	f2fs_warn(sbi, "compression algorithm is successfully set, "
4782	"but current kernel doesn't support this algorithm.");
4783	out:
4784	f2fs_up_write(&fi->i_sem);
4785	inode_unlock(inode);
4786	mnt_drop_write_file(filp);
4787
4788	return ret;
4789	}
4790
4791	static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4792	{
4793	DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
4794	struct address_space *mapping = inode->i_mapping;
4795	struct folio *folio;
4796	pgoff_t redirty_idx = page_idx;
4797	int page_len = 0, ret = 0;
4798
4799	filemap_invalidate_lock_shared(mapping);
4800	page_cache_ra_unbounded(&ractl, len, 0);
4801	filemap_invalidate_unlock_shared(mapping);
4802
4803	do {
4804	folio = read_cache_folio(mapping, page_idx, NULL, NULL);
4805	if (IS_ERR(folio)) {
4806	ret = PTR_ERR(folio);
4807	break;
4808	}
4809	page_len += folio_nr_pages(folio) - (page_idx - folio->index);
4810	page_idx = folio_next_index(folio);
4811	} while (page_len < len);
4812
4813	while (redirty_idx < page_idx) {
4814	folio = filemap_lock_folio(mapping, redirty_idx);
4815
4816	/* It will never fail, when folio has pinned above */
4817	f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(folio));
4818
4819	f2fs_folio_wait_writeback(folio, DATA, true, true);
4820
4821	folio_mark_dirty(folio);
4822	folio_set_f2fs_gcing(folio);
4823	redirty_idx = folio_next_index(folio);
4824	folio_unlock(folio);
4825	folio_put_refs(folio, 2);
4826	}
4827
4828	return ret;
4829	}
4830
4831	static int f2fs_ioc_decompress_file(struct file *filp)
4832	{
4833	struct inode *inode = file_inode(filp);
4834	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4835	struct f2fs_inode_info *fi = F2FS_I(inode);
4836	pgoff_t page_idx = 0, last_idx, cluster_idx;
4837	int ret;
4838
4839	if (!f2fs_sb_has_compression(sbi) \|\|
4840	F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4841	return -EOPNOTSUPP;
4842
4843	if (!(filp->f_mode & FMODE_WRITE))
4844	return -EBADF;
4845
4846	f2fs_balance_fs(sbi, true);
4847
4848	ret = mnt_want_write_file(filp);
4849	if (ret)
4850	return ret;
4851	inode_lock(inode);
4852
4853	if (!f2fs_is_compress_backend_ready(inode)) {
4854	ret = -EOPNOTSUPP;
4855	goto out;
4856	}
4857
4858	if (!f2fs_compressed_file(inode) \|\|
4859	is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4860	ret = -EINVAL;
4861	goto out;
4862	}
4863
4864	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4865	if (ret)
4866	goto out;
4867
4868	if (!atomic_read(&fi->i_compr_blocks))
4869	goto out;
4870
4871	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4872	last_idx >>= fi->i_log_cluster_size;
4873
4874	for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4875	page_idx = cluster_idx << fi->i_log_cluster_size;
4876
4877	if (!f2fs_is_compressed_cluster(inode, page_idx))
4878	continue;
4879
4880	ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4881	if (ret < 0)
4882	break;
4883
4884	if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4885	ret = filemap_fdatawrite(inode->i_mapping);
4886	if (ret < 0)
4887	break;
4888	}
4889
4890	cond_resched();
4891	if (fatal_signal_pending(current)) {
4892	ret = -EINTR;
4893	break;
4894	}
4895	}
4896
4897	if (!ret)
4898	ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4899	LLONG_MAX);
4900
4901	if (ret)
4902	f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4903	__func__, ret);
4904	f2fs_update_time(sbi, REQ_TIME);
4905	out:
4906	inode_unlock(inode);
4907	mnt_drop_write_file(filp);
4908
4909	return ret;
4910	}
4911
4912	static int f2fs_ioc_compress_file(struct file *filp)
4913	{
4914	struct inode *inode = file_inode(filp);
4915	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4916	struct f2fs_inode_info *fi = F2FS_I(inode);
4917	pgoff_t page_idx = 0, last_idx, cluster_idx;
4918	int ret;
4919
4920	if (!f2fs_sb_has_compression(sbi) \|\|
4921	F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4922	return -EOPNOTSUPP;
4923
4924	if (!(filp->f_mode & FMODE_WRITE))
4925	return -EBADF;
4926
4927	f2fs_balance_fs(sbi, true);
4928
4929	ret = mnt_want_write_file(filp);
4930	if (ret)
4931	return ret;
4932	inode_lock(inode);
4933
4934	if (!f2fs_is_compress_backend_ready(inode)) {
4935	ret = -EOPNOTSUPP;
4936	goto out;
4937	}
4938
4939	if (!f2fs_compressed_file(inode) \|\|
4940	is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4941	ret = -EINVAL;
4942	goto out;
4943	}
4944
4945	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4946	if (ret)
4947	goto out;
4948
4949	set_inode_flag(inode, FI_ENABLE_COMPRESS);
4950
4951	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4952	last_idx >>= fi->i_log_cluster_size;
4953
4954	for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4955	page_idx = cluster_idx << fi->i_log_cluster_size;
4956
4957	if (f2fs_is_sparse_cluster(inode, page_idx))
4958	continue;
4959
4960	ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4961	if (ret < 0)
4962	break;
4963
4964	if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4965	ret = filemap_fdatawrite(inode->i_mapping);
4966	if (ret < 0)
4967	break;
4968	}
4969
4970	cond_resched();
4971	if (fatal_signal_pending(current)) {
4972	ret = -EINTR;
4973	break;
4974	}
4975	}
4976
4977	if (!ret)
4978	ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4979	LLONG_MAX);
4980
4981	clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4982
4983	if (ret)
4984	f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4985	__func__, ret);
4986	f2fs_update_time(sbi, REQ_TIME);
4987	out:
4988	inode_unlock(inode);
4989	mnt_drop_write_file(filp);
4990
4991	return ret;
4992	}
4993
4994	static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4995	{
4996	switch (cmd) {
4997	case FS_IOC_GETVERSION:
4998	return f2fs_ioc_getversion(filp, arg);
4999	case F2FS_IOC_START_ATOMIC_WRITE:
5000	return f2fs_ioc_start_atomic_write(filp, false);
5001	case F2FS_IOC_START_ATOMIC_REPLACE:
5002	return f2fs_ioc_start_atomic_write(filp, true);
5003	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
5004	return f2fs_ioc_commit_atomic_write(filp);
5005	case F2FS_IOC_ABORT_ATOMIC_WRITE:
5006	return f2fs_ioc_abort_atomic_write(filp);
5007	case F2FS_IOC_START_VOLATILE_WRITE:
5008	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
5009	return -EOPNOTSUPP;
5010	case F2FS_IOC_SHUTDOWN:
5011	return f2fs_ioc_shutdown(filp, arg);
5012	case FITRIM:
5013	return f2fs_ioc_fitrim(filp, arg);
5014	case FS_IOC_SET_ENCRYPTION_POLICY:
5015	return f2fs_ioc_set_encryption_policy(filp, arg);
5016	case FS_IOC_GET_ENCRYPTION_POLICY:
5017	return f2fs_ioc_get_encryption_policy(filp, arg);
5018	case FS_IOC_GET_ENCRYPTION_PWSALT:
5019	return f2fs_ioc_get_encryption_pwsalt(filp, arg);
5020	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
5021	return f2fs_ioc_get_encryption_policy_ex(filp, arg);
5022	case FS_IOC_ADD_ENCRYPTION_KEY:
5023	return f2fs_ioc_add_encryption_key(filp, arg);
5024	case FS_IOC_REMOVE_ENCRYPTION_KEY:
5025	return f2fs_ioc_remove_encryption_key(filp, arg);
5026	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
5027	return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
5028	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
5029	return f2fs_ioc_get_encryption_key_status(filp, arg);
5030	case FS_IOC_GET_ENCRYPTION_NONCE:
5031	return f2fs_ioc_get_encryption_nonce(filp, arg);
5032	case F2FS_IOC_GARBAGE_COLLECT:
5033	return f2fs_ioc_gc(filp, arg);
5034	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
5035	return f2fs_ioc_gc_range(filp, arg);
5036	case F2FS_IOC_WRITE_CHECKPOINT:
5037	return f2fs_ioc_write_checkpoint(filp);
5038	case F2FS_IOC_DEFRAGMENT:
5039	return f2fs_ioc_defragment(filp, arg);
5040	case F2FS_IOC_MOVE_RANGE:
5041	return f2fs_ioc_move_range(filp, arg);
5042	case F2FS_IOC_FLUSH_DEVICE:
5043	return f2fs_ioc_flush_device(filp, arg);
5044	case F2FS_IOC_GET_FEATURES:
5045	return f2fs_ioc_get_features(filp, arg);
5046	case F2FS_IOC_GET_PIN_FILE:
5047	return f2fs_ioc_get_pin_file(filp, arg);
5048	case F2FS_IOC_SET_PIN_FILE:
5049	return f2fs_ioc_set_pin_file(filp, arg);
5050	case F2FS_IOC_PRECACHE_EXTENTS:
5051	return f2fs_ioc_precache_extents(filp);
5052	case F2FS_IOC_RESIZE_FS:
5053	return f2fs_ioc_resize_fs(filp, arg);
5054	case FS_IOC_ENABLE_VERITY:
5055	return f2fs_ioc_enable_verity(filp, arg);
5056	case FS_IOC_MEASURE_VERITY:
5057	return f2fs_ioc_measure_verity(filp, arg);
5058	case FS_IOC_READ_VERITY_METADATA:
5059	return f2fs_ioc_read_verity_metadata(filp, arg);
5060	case FS_IOC_GETFSLABEL:
5061	return f2fs_ioc_getfslabel(filp, arg);
5062	case FS_IOC_SETFSLABEL:
5063	return f2fs_ioc_setfslabel(filp, arg);
5064	case F2FS_IOC_GET_COMPRESS_BLOCKS:
5065	return f2fs_ioc_get_compress_blocks(filp, arg);
5066	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
5067	return f2fs_release_compress_blocks(filp, arg);
5068	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
5069	return f2fs_reserve_compress_blocks(filp, arg);
5070	case F2FS_IOC_SEC_TRIM_FILE:
5071	return f2fs_sec_trim_file(filp, arg);
5072	case F2FS_IOC_GET_COMPRESS_OPTION:
5073	return f2fs_ioc_get_compress_option(filp, arg);
5074	case F2FS_IOC_SET_COMPRESS_OPTION:
5075	return f2fs_ioc_set_compress_option(filp, arg);
5076	case F2FS_IOC_DECOMPRESS_FILE:
5077	return f2fs_ioc_decompress_file(filp);
5078	case F2FS_IOC_COMPRESS_FILE:
5079	return f2fs_ioc_compress_file(filp);
5080	case F2FS_IOC_GET_DEV_ALIAS_FILE:
5081	return f2fs_ioc_get_dev_alias_file(filp, arg);
5082	case F2FS_IOC_GET_DEV_ALIAS_STATUS:
5083	return f2fs_ioc_get_dev_alias_status(filp, arg);
5084	case F2FS_IOC_IO_PRIO:
5085	return f2fs_ioc_io_prio(filp, arg);
5086	case F2FS_IOC_RESERVE_DEV_ALIAS:
5087	return f2fs_ioc_reserve_dev_alias(filp);
5088	case F2FS_IOC_RELEASE_DEV_ALIAS:
5089	return f2fs_ioc_release_dev_alias(filp);
5090	default:
5091	return -ENOTTY;
5092	}
5093	}
5094
5095	long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
5096	{
5097	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
5098	return -EIO;
5099	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
5100	return -ENOSPC;
5101
5102	return __f2fs_ioctl(filp, cmd, arg);
5103	}
5104
5105	/*
5106	* Return %true if the given read or write request should use direct I/O, or
5107	* %false if it should use buffered I/O.
5108	*/
5109	static bool f2fs_should_use_dio(struct inode inode, struct kiocb iocb,
5110	struct iov_iter *iter)
5111	{
5112	unsigned int align;
5113
5114	if (!(iocb->ki_flags & IOCB_DIRECT))
5115	return false;
5116
5117	if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
5118	return false;
5119
5120	/*
5121	* Direct I/O not aligned to the disk's logical_block_size will be
5122	* attempted, but will fail with -EINVAL.
5123	*
5124	* f2fs additionally requires that direct I/O be aligned to the
5125	* filesystem block size, which is often a stricter requirement.
5126	* However, f2fs traditionally falls back to buffered I/O on requests
5127	* that are logical_block_size-aligned but not fs-block aligned.
5128	*
5129	* The below logic implements this behavior.
5130	*/
5131	align = iocb->ki_pos \| iov_iter_alignment(iter);
5132	if (!IS_ALIGNED(align, i_blocksize(inode)) &&
5133	IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
5134	return false;
5135
5136	return true;
5137	}
5138
5139	#ifdef CONFIG_F2FS_IOSTAT
5140	static void f2fs_dio_end_bio(struct bio *bio)
5141	{
5142	struct bio_iostat_ctx *iostat_ctx = bio->bi_private;
5143	void *orig_bi_private = iostat_ctx->post_read_ctx;
5144
5145	iostat_update_and_unbind_ctx(bio);
5146	bio->bi_private = orig_bi_private;
5147	iomap_dio_bio_end_io(bio);
5148	}
5149
5150	static void f2fs_dio_iostat_start(struct f2fs_sb_info sbi, struct bio bio)
5151	{
5152	void *bi_private = bio->bi_private;
5153
5154	if (!sbi->iostat_enable)
5155	return;
5156
5157	iostat_alloc_and_bind_ctx(sbi, bio, bi_private);
5158	iostat_update_submit_ctx(bio, DATA);
5159	bio->bi_end_io = f2fs_dio_end_bio;
5160	}
5161	#else
5162	static inline void f2fs_dio_iostat_start(struct f2fs_sb_info *sbi,
5163	struct bio *bio) {}
5164	#endif
5165
5166	static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
5167	unsigned int flags)
5168	{
5169	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
5170
5171	dec_page_count(sbi, F2FS_DIO_READ);
5172	if (error)
5173	return error;
5174	f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
5175	return 0;
5176	}
5177
5178	static void f2fs_dio_read_submit_io(const struct iomap_iter *iter,
5179	struct bio *bio, loff_t file_offset)
5180	{
5181	struct f2fs_sb_info *sbi = F2FS_I_SB(iter->inode);
5182
5183	f2fs_dio_iostat_start(sbi, bio);
5184	blk_crypto_submit_bio(bio);
5185	}
5186
5187	static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
5188	.end_io = f2fs_dio_read_end_io,
5189	.submit_io = f2fs_dio_read_submit_io,
5190	};
5191
5192	static ssize_t f2fs_dio_read_iter(struct kiocb iocb, struct iov_iter to)
5193	{
5194	struct file *file = iocb->ki_filp;
5195	struct inode *inode = file_inode(file);
5196	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5197	struct f2fs_inode_info *fi = F2FS_I(inode);
5198	const loff_t pos = iocb->ki_pos;
5199	const size_t count = iov_iter_count(to);
5200	struct iomap_dio *dio;
5201	ssize_t ret;
5202
5203	if (count == 0)
5204	return 0; /* skip atime update */
5205
5206	trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
5207
5208	if (iocb->ki_flags & IOCB_NOWAIT) {
5209	if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
5210	ret = -EAGAIN;
5211	goto out;
5212	}
5213	} else {
5214	f2fs_down_read(&fi->i_gc_rwsem[READ]);
5215	}
5216
5217	/* dio is not compatible w/ atomic file */
5218	if (f2fs_is_atomic_file(inode)) {
5219	f2fs_up_read(&fi->i_gc_rwsem[READ]);
5220	ret = -EOPNOTSUPP;
5221	goto out;
5222	}
5223
5224	/*
5225	* We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
5226	* the higher-level function iomap_dio_rw() in order to ensure that the
5227	* F2FS_DIO_READ counter will be decremented correctly in all cases.
5228	*/
5229	inc_page_count(sbi, F2FS_DIO_READ);
5230	dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
5231	&f2fs_iomap_dio_read_ops, 0, NULL, 0);
5232	if (IS_ERR_OR_NULL(dio)) {
5233	ret = PTR_ERR_OR_ZERO(dio);
5234	if (ret != -EIOCBQUEUED)
5235	dec_page_count(sbi, F2FS_DIO_READ);
5236	} else {
5237	ret = iomap_dio_complete(dio);
5238	}
5239
5240	f2fs_up_read(&fi->i_gc_rwsem[READ]);
5241
5242	file_accessed(file);
5243	out:
5244	trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
5245	return ret;
5246	}
5247
5248	static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count,
5249	int rw)
5250	{
5251	struct inode *inode = file_inode(file);
5252	char buf, path;
5253
5254	buf = f2fs_getname(F2FS_I_SB(inode));
5255	if (!buf)
5256	return;
5257	path = dentry_path_raw(file_dentry(file), buf, PATH_MAX);
5258	if (IS_ERR(path))
5259	goto free_buf;
5260	if (rw == WRITE)
5261	trace_f2fs_datawrite_start(inode, pos, count,
5262	current->pid, path, current->comm);
5263	else
5264	trace_f2fs_dataread_start(inode, pos, count,
5265	current->pid, path, current->comm);
5266	free_buf:
5267	f2fs_putname(buf);
5268	}
5269
5270	static ssize_t f2fs_file_read_iter(struct kiocb iocb, struct iov_iter to)
5271	{
5272	struct inode *inode = file_inode(iocb->ki_filp);
5273	const loff_t pos = iocb->ki_pos;
5274	ssize_t ret;
5275	bool dio;
5276
5277	if (!f2fs_is_compress_backend_ready(inode))
5278	return -EOPNOTSUPP;
5279
5280	if (trace_f2fs_dataread_start_enabled())
5281	f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
5282	iov_iter_count(to), READ);
5283
5284	dio = f2fs_should_use_dio(inode, iocb, to);
5285
5286	/* In LFS mode, if there is inflight dio, wait for its completion */
5287	if (f2fs_lfs_mode(F2FS_I_SB(inode)) &&
5288	get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE) &&
5289	(!f2fs_is_pinned_file(inode) \|\| !dio))
5290	inode_dio_wait(inode);
5291
5292	if (dio) {
5293	ret = f2fs_dio_read_iter(iocb, to);
5294	} else {
5295	ret = filemap_read(iocb, to, 0);
5296	if (ret > 0)
5297	f2fs_update_iostat(F2FS_I_SB(inode), inode,
5298	APP_BUFFERED_READ_IO, ret);
5299	}
5300	trace_f2fs_dataread_end(inode, pos, ret);
5301	return ret;
5302	}
5303
5304	static ssize_t f2fs_file_splice_read(struct file in, loff_t ppos,
5305	struct pipe_inode_info *pipe,
5306	size_t len, unsigned int flags)
5307	{
5308	struct inode *inode = file_inode(in);
5309	const loff_t pos = *ppos;
5310	ssize_t ret;
5311
5312	if (!f2fs_is_compress_backend_ready(inode))
5313	return -EOPNOTSUPP;
5314
5315	if (trace_f2fs_dataread_start_enabled())
5316	f2fs_trace_rw_file_path(in, pos, len, READ);
5317
5318	ret = filemap_splice_read(in, ppos, pipe, len, flags);
5319	if (ret > 0)
5320	f2fs_update_iostat(F2FS_I_SB(inode), inode,
5321	APP_BUFFERED_READ_IO, ret);
5322
5323	trace_f2fs_dataread_end(inode, pos, ret);
5324	return ret;
5325	}
5326
5327	static ssize_t f2fs_write_checks(struct kiocb iocb, struct iov_iter from)
5328	{
5329	struct file *file = iocb->ki_filp;
5330	struct inode *inode = file_inode(file);
5331	ssize_t count;
5332	int err;
5333
5334	if (IS_IMMUTABLE(inode))
5335	return -EPERM;
5336
5337	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
5338	return -EPERM;
5339
5340	count = generic_write_checks(iocb, from);
5341	if (count <= 0)
5342	return count;
5343
5344	err = file_modified(file);
5345	if (err)
5346	return err;
5347
5348	err = f2fs_zero_post_eof_page(inode,
5349	iocb->ki_pos + iov_iter_count(from), true, true);
5350	if (err)
5351	return err;
5352	return count;
5353	}
5354
5355	/*
5356	* Preallocate blocks for a write request, if it is possible and helpful to do
5357	* so. Returns a positive number if blocks may have been preallocated, 0 if no
5358	* blocks were preallocated, or a negative errno value if something went
5359	* seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if all the
5360	* requested blocks (not just some of them) have been allocated.
5361	*/
5362	static int f2fs_preallocate_blocks(struct kiocb iocb, struct iov_iter iter,
5363	bool dio)
5364	{
5365	struct inode *inode = file_inode(iocb->ki_filp);
5366	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5367	const loff_t pos = iocb->ki_pos;
5368	const size_t count = iov_iter_count(iter);
5369	struct f2fs_map_blocks map = {};
5370	int flag;
5371	int ret;
5372
5373	/* If it will be an out-of-place direct write, don't bother. */
5374	if (dio && f2fs_lfs_mode(sbi))
5375	return 0;
5376	/*
5377	* Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
5378	* buffered IO, if DIO meets any holes.
5379	*/
5380	if (dio && i_size_read(inode) &&
5381	(F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
5382	return 0;
5383
5384	/* No-wait I/O can't allocate blocks. */
5385	if (iocb->ki_flags & IOCB_NOWAIT)
5386	return 0;
5387
5388	/* If it will be a short write, don't bother. */
5389	if (fault_in_iov_iter_readable(iter, count))
5390	return 0;
5391
5392	if (f2fs_has_inline_data(inode)) {
5393	/* If the data will fit inline, don't bother. */
5394	if (pos + count <= MAX_INLINE_DATA(inode))
5395	return 0;
5396	ret = f2fs_convert_inline_inode(inode);
5397	if (ret)
5398	return ret;
5399	}
5400
5401	/* Do not preallocate blocks that will be written partially in 4KB. */
5402	map.m_lblk = F2FS_BLK_ALIGN(pos);
5403	map.m_len = F2FS_BYTES_TO_BLK(pos + count);
5404	if (map.m_len > map.m_lblk)
5405	map.m_len -= map.m_lblk;
5406	else
5407	return 0;
5408
5409	if (!IS_DEVICE_ALIASING(inode))
5410	map.m_may_create = true;
5411	if (dio) {
5412	map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi,
5413	inode->i_write_hint);
5414	flag = F2FS_GET_BLOCK_PRE_DIO;
5415	} else {
5416	map.m_seg_type = NO_CHECK_TYPE;
5417	flag = F2FS_GET_BLOCK_PRE_AIO;
5418	}
5419
5420	ret = f2fs_map_blocks(inode, &map, flag);
5421	/* -ENOSPC\|-EDQUOT are fine to report the number of allocated blocks. */
5422	if (ret < 0 && !((ret == -ENOSPC \|\| ret == -EDQUOT) && map.m_len > 0))
5423	return ret;
5424	if (ret == 0)
5425	set_inode_flag(inode, FI_PREALLOCATED_ALL);
5426	return map.m_len;
5427	}
5428
5429	static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
5430	struct iov_iter *from)
5431	{
5432	struct file *file = iocb->ki_filp;
5433	struct inode *inode = file_inode(file);
5434	ssize_t ret;
5435
5436	if (iocb->ki_flags & IOCB_NOWAIT)
5437	return -EOPNOTSUPP;
5438
5439	ret = generic_perform_write(iocb, from);
5440
5441	if (ret > 0) {
5442	f2fs_update_iostat(F2FS_I_SB(inode), inode,
5443	APP_BUFFERED_IO, ret);
5444	}
5445	return ret;
5446	}
5447
5448	static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
5449	unsigned int flags)
5450	{
5451	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
5452
5453	dec_page_count(sbi, F2FS_DIO_WRITE);
5454	if (error)
5455	return error;
5456	f2fs_update_time(sbi, REQ_TIME);
5457	f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
5458	return 0;
5459	}
5460
5461	static bool f2fs_valid_write_stream(struct f2fs_sb_info *sbi, u8 write_stream)
5462	{
5463	int i;
5464
5465	if (!write_stream)
5466	return true;
5467	if (!f2fs_is_multi_device(sbi))
5468	return write_stream <= bdev_max_write_streams(sbi->sb->s_bdev);
5469
5470	for (i = 0; i < sbi->s_ndevs; i++)
5471	if (write_stream > bdev_max_write_streams(FDEV(i).bdev))
5472	return false;
5473	return true;
5474	}
5475
5476	static void f2fs_dio_write_submit_io(const struct iomap_iter *iter,
5477	struct bio *bio, loff_t file_offset)
5478	{
5479	struct inode *inode = iter->inode;
5480	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5481	struct kiocb *iocb = iter->private;
5482	enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint);
5483	enum temp_type temp = f2fs_get_segment_temp(sbi, type);
5484
5485	bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp);
5486	bio->bi_write_stream =
5487	iocb->ki_write_stream ? iocb->ki_write_stream :
5488	f2fs_io_type_to_write_stream(bio->bi_bdev, DATA, temp);
5489	f2fs_dio_iostat_start(sbi, bio);
5490	blk_crypto_submit_bio(bio);
5491	}
5492
5493	static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
5494	.end_io = f2fs_dio_write_end_io,
5495	.submit_io = f2fs_dio_write_submit_io,
5496	};
5497
5498	static void f2fs_flush_buffered_write(struct address_space *mapping,
5499	loff_t start_pos, loff_t end_pos)
5500	{
5501	int ret;
5502
5503	ret = filemap_write_and_wait_range(mapping, start_pos, end_pos);
5504	if (ret < 0)
5505	return;
5506	invalidate_mapping_pages(mapping,
5507	start_pos >> PAGE_SHIFT,
5508	end_pos >> PAGE_SHIFT);
5509	}
5510
5511	static ssize_t f2fs_dio_write_iter(struct kiocb iocb, struct iov_iter from,
5512	bool *may_need_sync)
5513	{
5514	struct file *file = iocb->ki_filp;
5515	struct inode *inode = file_inode(file);
5516	struct f2fs_inode_info *fi = F2FS_I(inode);
5517	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5518	const bool do_opu = f2fs_lfs_mode(sbi);
5519	const loff_t pos = iocb->ki_pos;
5520	const ssize_t count = iov_iter_count(from);
5521	unsigned int dio_flags;
5522	struct iomap_dio *dio;
5523	ssize_t ret;
5524
5525	trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
5526
5527	if (!f2fs_valid_write_stream(sbi, iocb->ki_write_stream)) {
5528	ret = -EINVAL;
5529	goto out;
5530	}
5531
5532	if (iocb->ki_flags & IOCB_NOWAIT) {
5533	/* f2fs_convert_inline_inode() and block allocation can block */
5534	if (f2fs_has_inline_data(inode) \|\|
5535	!f2fs_overwrite_io(inode, pos, count)) {
5536	ret = -EAGAIN;
5537	goto out;
5538	}
5539
5540	if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
5541	ret = -EAGAIN;
5542	goto out;
5543	}
5544	if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
5545	f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
5546	ret = -EAGAIN;
5547	goto out;
5548	}
5549	} else {
5550	ret = f2fs_convert_inline_inode(inode);
5551	if (ret)
5552	goto out;
5553
5554	f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
5555	if (do_opu)
5556	f2fs_down_read(&fi->i_gc_rwsem[READ]);
5557	}
5558
5559	/*
5560	* We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
5561	* the higher-level function iomap_dio_rw() in order to ensure that the
5562	* F2FS_DIO_WRITE counter will be decremented correctly in all cases.
5563	*/
5564	inc_page_count(sbi, F2FS_DIO_WRITE);
5565	dio_flags = 0;
5566	if (pos + count > inode->i_size)
5567	dio_flags \|= IOMAP_DIO_FORCE_WAIT;
5568	dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
5569	&f2fs_iomap_dio_write_ops, dio_flags, iocb, 0);
5570	if (IS_ERR_OR_NULL(dio)) {
5571	ret = PTR_ERR_OR_ZERO(dio);
5572	if (ret == -ENOTBLK)
5573	ret = 0;
5574	if (ret != -EIOCBQUEUED)
5575	dec_page_count(sbi, F2FS_DIO_WRITE);
5576	} else {
5577	ret = iomap_dio_complete(dio);
5578	}
5579
5580	if (do_opu)
5581	f2fs_up_read(&fi->i_gc_rwsem[READ]);
5582	f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
5583
5584	if (ret < 0)
5585	goto out;
5586	if (pos + ret > inode->i_size)
5587	f2fs_i_size_write(inode, pos + ret);
5588	if (!do_opu)
5589	set_inode_flag(inode, FI_UPDATE_WRITE);
5590
5591	if (iov_iter_count(from)) {
5592	ssize_t ret2;
5593	loff_t bufio_start_pos = iocb->ki_pos;
5594
5595	/*
5596	* The direct write was partial, so we need to fall back to a
5597	* buffered write for the remainder.
5598	*/
5599
5600	ret2 = f2fs_buffered_write_iter(iocb, from);
5601	if (iov_iter_count(from))
5602	f2fs_write_failed(inode, iocb->ki_pos);
5603	if (ret2 < 0)
5604	goto out;
5605
5606	/*
5607	* Ensure that the pagecache pages are written to disk and
5608	* invalidated to preserve the expected O_DIRECT semantics.
5609	*/
5610	if (ret2 > 0) {
5611	loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
5612
5613	ret += ret2;
5614
5615	f2fs_flush_buffered_write(file->f_mapping,
5616	bufio_start_pos,
5617	bufio_end_pos);
5618	}
5619	} else {
5620	/* iomap_dio_rw() already handled the generic_write_sync(). */
5621	*may_need_sync = false;
5622	}
5623	out:
5624	trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
5625	return ret;
5626	}
5627
5628	static ssize_t f2fs_file_write_iter(struct kiocb iocb, struct iov_iter from)
5629	{
5630	struct inode *inode = file_inode(iocb->ki_filp);
5631	const loff_t orig_pos = iocb->ki_pos;
5632	const size_t orig_count = iov_iter_count(from);
5633	loff_t target_size;
5634	bool dio;
5635	bool may_need_sync = true;
5636	int preallocated;
5637	ssize_t ret;
5638	loff_t bufio_start_pos;
5639
5640	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
5641	ret = -EIO;
5642	goto out;
5643	}
5644
5645	if (!f2fs_is_compress_backend_ready(inode)) {
5646	ret = -EOPNOTSUPP;
5647	goto out;
5648	}
5649
5650	if (iocb->ki_flags & IOCB_NOWAIT) {
5651	if (!inode_trylock(inode)) {
5652	ret = -EAGAIN;
5653	goto out;
5654	}
5655	} else {
5656	inode_lock(inode);
5657	}
5658
5659	ret = f2fs_write_checks(iocb, from);
5660	if (ret <= 0)
5661	goto out_unlock;
5662
5663	if (f2fs_is_pinned_file(inode) &&
5664	!f2fs_overwrite_io(inode, iocb->ki_pos, iov_iter_count(from))) {
5665	ret = -EIO;
5666	goto out_unlock;
5667	}
5668
5669	bufio_start_pos = iocb->ki_pos;
5670
5671	/* Determine whether we will do a direct write or a buffered write. */
5672	dio = f2fs_should_use_dio(inode, iocb, from);
5673
5674	/* dio is not compatible w/ atomic write */
5675	if (dio && f2fs_is_atomic_file(inode)) {
5676	ret = -EOPNOTSUPP;
5677	goto out_unlock;
5678	}
5679
5680	/* Possibly preallocate the blocks for the write. */
5681	target_size = iocb->ki_pos + iov_iter_count(from);
5682	preallocated = f2fs_preallocate_blocks(iocb, from, dio);
5683	if (preallocated < 0) {
5684	ret = preallocated;
5685	} else {
5686	if (trace_f2fs_datawrite_start_enabled())
5687	f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
5688	orig_count, WRITE);
5689
5690	/* Do the actual write. */
5691	ret = dio ?
5692	f2fs_dio_write_iter(iocb, from, &may_need_sync) :
5693	f2fs_buffered_write_iter(iocb, from);
5694
5695	trace_f2fs_datawrite_end(inode, orig_pos, ret);
5696	}
5697
5698	/* Don't leave any preallocated blocks around past i_size. */
5699	if (preallocated && i_size_read(inode) < target_size) {
5700	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5701	filemap_invalidate_lock(inode->i_mapping);
5702	if (!f2fs_truncate(inode))
5703	file_dont_truncate(inode);
5704	filemap_invalidate_unlock(inode->i_mapping);
5705	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5706	} else {
5707	file_dont_truncate(inode);
5708	}
5709
5710	clear_inode_flag(inode, FI_PREALLOCATED_ALL);
5711	out_unlock:
5712	inode_unlock(inode);
5713	out:
5714	trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
5715
5716	if (ret > 0 && may_need_sync)
5717	ret = generic_write_sync(iocb, ret);
5718
5719	/* If buffered IO was forced, flush and drop the data from
5720	* the page cache to preserve O_DIRECT semantics
5721	*/
5722	if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
5723	f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
5724	bufio_start_pos,
5725	bufio_start_pos + ret - 1);
5726
5727	return ret;
5728	}
5729
5730	static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
5731	int advice)
5732	{
5733	struct address_space *mapping;
5734	struct backing_dev_info *bdi;
5735	struct inode *inode = file_inode(filp);
5736	int err;
5737
5738	trace_f2fs_fadvise(inode, offset, len, advice);
5739
5740	if (advice == POSIX_FADV_SEQUENTIAL) {
5741	if (S_ISFIFO(inode->i_mode))
5742	return -ESPIPE;
5743
5744	mapping = filp->f_mapping;
5745	if (!mapping \|\| len < 0)
5746	return -EINVAL;
5747
5748	bdi = inode_to_bdi(mapping->host);
5749	filp->f_ra.ra_pages = bdi->ra_pages *
5750	F2FS_I_SB(inode)->seq_file_ra_mul;
5751	spin_lock(&filp->f_lock);
5752	filp->f_mode &= ~FMODE_RANDOM;
5753	spin_unlock(&filp->f_lock);
5754	return 0;
5755	} else if (advice == POSIX_FADV_WILLNEED && offset == 0) {
5756	/* Load extent cache at the first readahead. */
5757	f2fs_precache_extents(inode);
5758	}
5759
5760	err = generic_fadvise(filp, offset, len, advice);
5761	if (err)
5762	return err;
5763
5764	if (advice == POSIX_FADV_DONTNEED &&
5765	(test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
5766	f2fs_compressed_file(inode)))
5767	f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
5768	else if (advice == POSIX_FADV_NOREUSE)
5769	err = f2fs_keep_noreuse_range(inode, offset, len);
5770	return err;
5771	}
5772
5773	#ifdef CONFIG_COMPAT
5774	struct compat_f2fs_gc_range {
5775	u32 sync;
5776	compat_u64 start;
5777	compat_u64 len;
5778	};
5779	#define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
5780	struct compat_f2fs_gc_range)
5781
5782	static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
5783	{
5784	struct compat_f2fs_gc_range __user *urange;
5785	struct f2fs_gc_range range;
5786	int err;
5787
5788	urange = compat_ptr(arg);
5789	err = get_user(range.sync, &urange->sync);
5790	err \|= get_user(range.start, &urange->start);
5791	err \|= get_user(range.len, &urange->len);
5792	if (err)
5793	return -EFAULT;
5794
5795	return __f2fs_ioc_gc_range(file, &range);
5796	}
5797
5798	struct compat_f2fs_move_range {
5799	u32 dst_fd;
5800	compat_u64 pos_in;
5801	compat_u64 pos_out;
5802	compat_u64 len;
5803	};
5804	#define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
5805	struct compat_f2fs_move_range)
5806
5807	static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
5808	{
5809	struct compat_f2fs_move_range __user *urange;
5810	struct f2fs_move_range range;
5811	int err;
5812
5813	urange = compat_ptr(arg);
5814	err = get_user(range.dst_fd, &urange->dst_fd);
5815	err \|= get_user(range.pos_in, &urange->pos_in);
5816	err \|= get_user(range.pos_out, &urange->pos_out);
5817	err \|= get_user(range.len, &urange->len);
5818	if (err)
5819	return -EFAULT;
5820
5821	return __f2fs_ioc_move_range(file, &range);
5822	}
5823
5824	long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5825	{
5826	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
5827	return -EIO;
5828	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
5829	return -ENOSPC;
5830
5831	switch (cmd) {
5832	case FS_IOC32_GETVERSION:
5833	cmd = FS_IOC_GETVERSION;
5834	break;
5835	case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
5836	return f2fs_compat_ioc_gc_range(file, arg);
5837	case F2FS_IOC32_MOVE_RANGE:
5838	return f2fs_compat_ioc_move_range(file, arg);
5839	case F2FS_IOC_START_ATOMIC_WRITE:
5840	case F2FS_IOC_START_ATOMIC_REPLACE:
5841	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
5842	case F2FS_IOC_START_VOLATILE_WRITE:
5843	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
5844	case F2FS_IOC_ABORT_ATOMIC_WRITE:
5845	case F2FS_IOC_SHUTDOWN:
5846	case FITRIM:
5847	case FS_IOC_SET_ENCRYPTION_POLICY:
5848	case FS_IOC_GET_ENCRYPTION_PWSALT:
5849	case FS_IOC_GET_ENCRYPTION_POLICY:
5850	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
5851	case FS_IOC_ADD_ENCRYPTION_KEY:
5852	case FS_IOC_REMOVE_ENCRYPTION_KEY:
5853	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
5854	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
5855	case FS_IOC_GET_ENCRYPTION_NONCE:
5856	case F2FS_IOC_GARBAGE_COLLECT:
5857	case F2FS_IOC_WRITE_CHECKPOINT:
5858	case F2FS_IOC_DEFRAGMENT:
5859	case F2FS_IOC_FLUSH_DEVICE:
5860	case F2FS_IOC_GET_FEATURES:
5861	case F2FS_IOC_GET_PIN_FILE:
5862	case F2FS_IOC_SET_PIN_FILE:
5863	case F2FS_IOC_PRECACHE_EXTENTS:
5864	case F2FS_IOC_RESIZE_FS:
5865	case FS_IOC_ENABLE_VERITY:
5866	case FS_IOC_MEASURE_VERITY:
5867	case FS_IOC_READ_VERITY_METADATA:
5868	case FS_IOC_GETFSLABEL:
5869	case FS_IOC_SETFSLABEL:
5870	case F2FS_IOC_GET_COMPRESS_BLOCKS:
5871	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
5872	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
5873	case F2FS_IOC_SEC_TRIM_FILE:
5874	case F2FS_IOC_GET_COMPRESS_OPTION:
5875	case F2FS_IOC_SET_COMPRESS_OPTION:
5876	case F2FS_IOC_DECOMPRESS_FILE:
5877	case F2FS_IOC_COMPRESS_FILE:
5878	case F2FS_IOC_GET_DEV_ALIAS_FILE:
5879	case F2FS_IOC_GET_DEV_ALIAS_STATUS:
5880	case F2FS_IOC_IO_PRIO:
5881	case F2FS_IOC_RESERVE_DEV_ALIAS:
5882	case F2FS_IOC_RELEASE_DEV_ALIAS:
5883	break;
5884	default:
5885	return -ENOIOCTLCMD;
5886	}
5887	return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5888	}
5889	#endif
5890
5891	const struct file_operations f2fs_file_operations = {
5892	.llseek = f2fs_llseek,
5893	.read_iter = f2fs_file_read_iter,
5894	.write_iter = f2fs_file_write_iter,
5895	.iopoll = iocb_bio_iopoll,
5896	.open = f2fs_file_open,
5897	.release = f2fs_release_file,
5898	.mmap_prepare = f2fs_file_mmap_prepare,
5899	.flush = f2fs_file_flush,
5900	.fsync = f2fs_sync_file,
5901	.fallocate = f2fs_fallocate,
5902	.unlocked_ioctl = f2fs_ioctl,
5903	#ifdef CONFIG_COMPAT
5904	.compat_ioctl = f2fs_compat_ioctl,
5905	#endif
5906	.splice_read = f2fs_file_splice_read,
5907	.splice_write = iter_file_splice_write,
5908	.fadvise = f2fs_file_fadvise,
5909	.fop_flags = FOP_BUFFER_RASYNC,
5910	.setlease = generic_setlease,
5911	};