xref: /linux/fs/f2fs/file.c (revision 1448f8acf4cc61197a228bdb7126e7eeb92760fe)
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/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/buffer_head.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/types.h>
16 #include <linux/compat.h>
17 #include <linux/uaccess.h>
18 #include <linux/mount.h>
19 #include <linux/pagevec.h>
20 #include <linux/uio.h>
21 #include <linux/uuid.h>
22 #include <linux/file.h>
23 #include <linux/nls.h>
24 #include <linux/sched/signal.h>
25 
26 #include "f2fs.h"
27 #include "node.h"
28 #include "segment.h"
29 #include "xattr.h"
30 #include "acl.h"
31 #include "gc.h"
32 #include "trace.h"
33 #include <trace/events/f2fs.h>
34 #include <uapi/linux/f2fs.h>
35 
36 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
37 {
38 	struct inode *inode = file_inode(vmf->vma->vm_file);
39 	vm_fault_t ret;
40 
41 	down_read(&F2FS_I(inode)->i_mmap_sem);
42 	ret = filemap_fault(vmf);
43 	up_read(&F2FS_I(inode)->i_mmap_sem);
44 
45 	if (!ret)
46 		f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
47 							F2FS_BLKSIZE);
48 
49 	trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
50 
51 	return ret;
52 }
53 
54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
55 {
56 	struct page *page = vmf->page;
57 	struct inode *inode = file_inode(vmf->vma->vm_file);
58 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
59 	struct dnode_of_data dn;
60 	bool need_alloc = true;
61 	int err = 0;
62 
63 	if (unlikely(f2fs_cp_error(sbi))) {
64 		err = -EIO;
65 		goto err;
66 	}
67 
68 	if (!f2fs_is_checkpoint_ready(sbi)) {
69 		err = -ENOSPC;
70 		goto err;
71 	}
72 
73 #ifdef CONFIG_F2FS_FS_COMPRESSION
74 	if (f2fs_compressed_file(inode)) {
75 		int ret = f2fs_is_compressed_cluster(inode, page->index);
76 
77 		if (ret < 0) {
78 			err = ret;
79 			goto err;
80 		} else if (ret) {
81 			if (ret < F2FS_I(inode)->i_cluster_size) {
82 				err = -EAGAIN;
83 				goto err;
84 			}
85 			need_alloc = false;
86 		}
87 	}
88 #endif
89 	/* should do out of any locked page */
90 	if (need_alloc)
91 		f2fs_balance_fs(sbi, true);
92 
93 	sb_start_pagefault(inode->i_sb);
94 
95 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
96 
97 	file_update_time(vmf->vma->vm_file);
98 	down_read(&F2FS_I(inode)->i_mmap_sem);
99 	lock_page(page);
100 	if (unlikely(page->mapping != inode->i_mapping ||
101 			page_offset(page) > i_size_read(inode) ||
102 			!PageUptodate(page))) {
103 		unlock_page(page);
104 		err = -EFAULT;
105 		goto out_sem;
106 	}
107 
108 	if (need_alloc) {
109 		/* block allocation */
110 		f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
111 		set_new_dnode(&dn, inode, NULL, NULL, 0);
112 		err = f2fs_get_block(&dn, page->index);
113 		f2fs_put_dnode(&dn);
114 		f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
115 	}
116 
117 #ifdef CONFIG_F2FS_FS_COMPRESSION
118 	if (!need_alloc) {
119 		set_new_dnode(&dn, inode, NULL, NULL, 0);
120 		err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
121 		f2fs_put_dnode(&dn);
122 	}
123 #endif
124 	if (err) {
125 		unlock_page(page);
126 		goto out_sem;
127 	}
128 
129 	f2fs_wait_on_page_writeback(page, DATA, false, true);
130 
131 	/* wait for GCed page writeback via META_MAPPING */
132 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
133 
134 	/*
135 	 * check to see if the page is mapped already (no holes)
136 	 */
137 	if (PageMappedToDisk(page))
138 		goto out_sem;
139 
140 	/* page is wholly or partially inside EOF */
141 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
142 						i_size_read(inode)) {
143 		loff_t offset;
144 
145 		offset = i_size_read(inode) & ~PAGE_MASK;
146 		zero_user_segment(page, offset, PAGE_SIZE);
147 	}
148 	set_page_dirty(page);
149 	if (!PageUptodate(page))
150 		SetPageUptodate(page);
151 
152 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
153 	f2fs_update_time(sbi, REQ_TIME);
154 
155 	trace_f2fs_vm_page_mkwrite(page, DATA);
156 out_sem:
157 	up_read(&F2FS_I(inode)->i_mmap_sem);
158 
159 	sb_end_pagefault(inode->i_sb);
160 err:
161 	return block_page_mkwrite_return(err);
162 }
163 
164 static const struct vm_operations_struct f2fs_file_vm_ops = {
165 	.fault		= f2fs_filemap_fault,
166 	.map_pages	= filemap_map_pages,
167 	.page_mkwrite	= f2fs_vm_page_mkwrite,
168 };
169 
170 static int get_parent_ino(struct inode *inode, nid_t *pino)
171 {
172 	struct dentry *dentry;
173 
174 	/*
175 	 * Make sure to get the non-deleted alias.  The alias associated with
176 	 * the open file descriptor being fsync()'ed may be deleted already.
177 	 */
178 	dentry = d_find_alias(inode);
179 	if (!dentry)
180 		return 0;
181 
182 	*pino = parent_ino(dentry);
183 	dput(dentry);
184 	return 1;
185 }
186 
187 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
188 {
189 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
190 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
191 
192 	if (!S_ISREG(inode->i_mode))
193 		cp_reason = CP_NON_REGULAR;
194 	else if (f2fs_compressed_file(inode))
195 		cp_reason = CP_COMPRESSED;
196 	else if (inode->i_nlink != 1)
197 		cp_reason = CP_HARDLINK;
198 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
199 		cp_reason = CP_SB_NEED_CP;
200 	else if (file_wrong_pino(inode))
201 		cp_reason = CP_WRONG_PINO;
202 	else if (!f2fs_space_for_roll_forward(sbi))
203 		cp_reason = CP_NO_SPC_ROLL;
204 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
205 		cp_reason = CP_NODE_NEED_CP;
206 	else if (test_opt(sbi, FASTBOOT))
207 		cp_reason = CP_FASTBOOT_MODE;
208 	else if (F2FS_OPTION(sbi).active_logs == 2)
209 		cp_reason = CP_SPEC_LOG_NUM;
210 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
211 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
212 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
213 							TRANS_DIR_INO))
214 		cp_reason = CP_RECOVER_DIR;
215 
216 	return cp_reason;
217 }
218 
219 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
220 {
221 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
222 	bool ret = false;
223 	/* But we need to avoid that there are some inode updates */
224 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
225 		ret = true;
226 	f2fs_put_page(i, 0);
227 	return ret;
228 }
229 
230 static void try_to_fix_pino(struct inode *inode)
231 {
232 	struct f2fs_inode_info *fi = F2FS_I(inode);
233 	nid_t pino;
234 
235 	down_write(&fi->i_sem);
236 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
237 			get_parent_ino(inode, &pino)) {
238 		f2fs_i_pino_write(inode, pino);
239 		file_got_pino(inode);
240 	}
241 	up_write(&fi->i_sem);
242 }
243 
244 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
245 						int datasync, bool atomic)
246 {
247 	struct inode *inode = file->f_mapping->host;
248 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
249 	nid_t ino = inode->i_ino;
250 	int ret = 0;
251 	enum cp_reason_type cp_reason = 0;
252 	struct writeback_control wbc = {
253 		.sync_mode = WB_SYNC_ALL,
254 		.nr_to_write = LONG_MAX,
255 		.for_reclaim = 0,
256 	};
257 	unsigned int seq_id = 0;
258 
259 	if (unlikely(f2fs_readonly(inode->i_sb) ||
260 				is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
261 		return 0;
262 
263 	trace_f2fs_sync_file_enter(inode);
264 
265 	if (S_ISDIR(inode->i_mode))
266 		goto go_write;
267 
268 	/* if fdatasync is triggered, let's do in-place-update */
269 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
270 		set_inode_flag(inode, FI_NEED_IPU);
271 	ret = file_write_and_wait_range(file, start, end);
272 	clear_inode_flag(inode, FI_NEED_IPU);
273 
274 	if (ret) {
275 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
276 		return ret;
277 	}
278 
279 	/* if the inode is dirty, let's recover all the time */
280 	if (!f2fs_skip_inode_update(inode, datasync)) {
281 		f2fs_write_inode(inode, NULL);
282 		goto go_write;
283 	}
284 
285 	/*
286 	 * if there is no written data, don't waste time to write recovery info.
287 	 */
288 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
289 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
290 
291 		/* it may call write_inode just prior to fsync */
292 		if (need_inode_page_update(sbi, ino))
293 			goto go_write;
294 
295 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
296 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
297 			goto flush_out;
298 		goto out;
299 	}
300 go_write:
301 	/*
302 	 * Both of fdatasync() and fsync() are able to be recovered from
303 	 * sudden-power-off.
304 	 */
305 	down_read(&F2FS_I(inode)->i_sem);
306 	cp_reason = need_do_checkpoint(inode);
307 	up_read(&F2FS_I(inode)->i_sem);
308 
309 	if (cp_reason) {
310 		/* all the dirty node pages should be flushed for POR */
311 		ret = f2fs_sync_fs(inode->i_sb, 1);
312 
313 		/*
314 		 * We've secured consistency through sync_fs. Following pino
315 		 * will be used only for fsynced inodes after checkpoint.
316 		 */
317 		try_to_fix_pino(inode);
318 		clear_inode_flag(inode, FI_APPEND_WRITE);
319 		clear_inode_flag(inode, FI_UPDATE_WRITE);
320 		goto out;
321 	}
322 sync_nodes:
323 	atomic_inc(&sbi->wb_sync_req[NODE]);
324 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
325 	atomic_dec(&sbi->wb_sync_req[NODE]);
326 	if (ret)
327 		goto out;
328 
329 	/* if cp_error was enabled, we should avoid infinite loop */
330 	if (unlikely(f2fs_cp_error(sbi))) {
331 		ret = -EIO;
332 		goto out;
333 	}
334 
335 	if (f2fs_need_inode_block_update(sbi, ino)) {
336 		f2fs_mark_inode_dirty_sync(inode, true);
337 		f2fs_write_inode(inode, NULL);
338 		goto sync_nodes;
339 	}
340 
341 	/*
342 	 * If it's atomic_write, it's just fine to keep write ordering. So
343 	 * here we don't need to wait for node write completion, since we use
344 	 * node chain which serializes node blocks. If one of node writes are
345 	 * reordered, we can see simply broken chain, resulting in stopping
346 	 * roll-forward recovery. It means we'll recover all or none node blocks
347 	 * given fsync mark.
348 	 */
349 	if (!atomic) {
350 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
351 		if (ret)
352 			goto out;
353 	}
354 
355 	/* once recovery info is written, don't need to tack this */
356 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
357 	clear_inode_flag(inode, FI_APPEND_WRITE);
358 flush_out:
359 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
360 		ret = f2fs_issue_flush(sbi, inode->i_ino);
361 	if (!ret) {
362 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
363 		clear_inode_flag(inode, FI_UPDATE_WRITE);
364 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
365 	}
366 	f2fs_update_time(sbi, REQ_TIME);
367 out:
368 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
369 	f2fs_trace_ios(NULL, 1);
370 	return ret;
371 }
372 
373 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
374 {
375 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
376 		return -EIO;
377 	return f2fs_do_sync_file(file, start, end, datasync, false);
378 }
379 
380 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
381 				pgoff_t index, int whence)
382 {
383 	switch (whence) {
384 	case SEEK_DATA:
385 		if (__is_valid_data_blkaddr(blkaddr))
386 			return true;
387 		if (blkaddr == NEW_ADDR &&
388 		    xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
389 			return true;
390 		break;
391 	case SEEK_HOLE:
392 		if (blkaddr == NULL_ADDR)
393 			return true;
394 		break;
395 	}
396 	return false;
397 }
398 
399 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
400 {
401 	struct inode *inode = file->f_mapping->host;
402 	loff_t maxbytes = inode->i_sb->s_maxbytes;
403 	struct dnode_of_data dn;
404 	pgoff_t pgofs, end_offset;
405 	loff_t data_ofs = offset;
406 	loff_t isize;
407 	int err = 0;
408 
409 	inode_lock(inode);
410 
411 	isize = i_size_read(inode);
412 	if (offset >= isize)
413 		goto fail;
414 
415 	/* handle inline data case */
416 	if (f2fs_has_inline_data(inode)) {
417 		if (whence == SEEK_HOLE) {
418 			data_ofs = isize;
419 			goto found;
420 		} else if (whence == SEEK_DATA) {
421 			data_ofs = offset;
422 			goto found;
423 		}
424 	}
425 
426 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
427 
428 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
429 		set_new_dnode(&dn, inode, NULL, NULL, 0);
430 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
431 		if (err && err != -ENOENT) {
432 			goto fail;
433 		} else if (err == -ENOENT) {
434 			/* direct node does not exists */
435 			if (whence == SEEK_DATA) {
436 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
437 				continue;
438 			} else {
439 				goto found;
440 			}
441 		}
442 
443 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
444 
445 		/* find data/hole in dnode block */
446 		for (; dn.ofs_in_node < end_offset;
447 				dn.ofs_in_node++, pgofs++,
448 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
449 			block_t blkaddr;
450 
451 			blkaddr = f2fs_data_blkaddr(&dn);
452 
453 			if (__is_valid_data_blkaddr(blkaddr) &&
454 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
455 					blkaddr, DATA_GENERIC_ENHANCE)) {
456 				f2fs_put_dnode(&dn);
457 				goto fail;
458 			}
459 
460 			if (__found_offset(file->f_mapping, blkaddr,
461 							pgofs, whence)) {
462 				f2fs_put_dnode(&dn);
463 				goto found;
464 			}
465 		}
466 		f2fs_put_dnode(&dn);
467 	}
468 
469 	if (whence == SEEK_DATA)
470 		goto fail;
471 found:
472 	if (whence == SEEK_HOLE && data_ofs > isize)
473 		data_ofs = isize;
474 	inode_unlock(inode);
475 	return vfs_setpos(file, data_ofs, maxbytes);
476 fail:
477 	inode_unlock(inode);
478 	return -ENXIO;
479 }
480 
481 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
482 {
483 	struct inode *inode = file->f_mapping->host;
484 	loff_t maxbytes = inode->i_sb->s_maxbytes;
485 
486 	switch (whence) {
487 	case SEEK_SET:
488 	case SEEK_CUR:
489 	case SEEK_END:
490 		return generic_file_llseek_size(file, offset, whence,
491 						maxbytes, i_size_read(inode));
492 	case SEEK_DATA:
493 	case SEEK_HOLE:
494 		if (offset < 0)
495 			return -ENXIO;
496 		return f2fs_seek_block(file, offset, whence);
497 	}
498 
499 	return -EINVAL;
500 }
501 
502 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
503 {
504 	struct inode *inode = file_inode(file);
505 	int err;
506 
507 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
508 		return -EIO;
509 
510 	if (!f2fs_is_compress_backend_ready(inode))
511 		return -EOPNOTSUPP;
512 
513 	/* we don't need to use inline_data strictly */
514 	err = f2fs_convert_inline_inode(inode);
515 	if (err)
516 		return err;
517 
518 	file_accessed(file);
519 	vma->vm_ops = &f2fs_file_vm_ops;
520 	set_inode_flag(inode, FI_MMAP_FILE);
521 	return 0;
522 }
523 
524 static int f2fs_file_open(struct inode *inode, struct file *filp)
525 {
526 	int err = fscrypt_file_open(inode, filp);
527 
528 	if (err)
529 		return err;
530 
531 	if (!f2fs_is_compress_backend_ready(inode))
532 		return -EOPNOTSUPP;
533 
534 	err = fsverity_file_open(inode, filp);
535 	if (err)
536 		return err;
537 
538 	filp->f_mode |= FMODE_NOWAIT;
539 
540 	return dquot_file_open(inode, filp);
541 }
542 
543 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
544 {
545 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
546 	struct f2fs_node *raw_node;
547 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
548 	__le32 *addr;
549 	int base = 0;
550 	bool compressed_cluster = false;
551 	int cluster_index = 0, valid_blocks = 0;
552 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
553 	bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
554 
555 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
556 		base = get_extra_isize(dn->inode);
557 
558 	raw_node = F2FS_NODE(dn->node_page);
559 	addr = blkaddr_in_node(raw_node) + base + ofs;
560 
561 	/* Assumption: truncateion starts with cluster */
562 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
563 		block_t blkaddr = le32_to_cpu(*addr);
564 
565 		if (f2fs_compressed_file(dn->inode) &&
566 					!(cluster_index & (cluster_size - 1))) {
567 			if (compressed_cluster)
568 				f2fs_i_compr_blocks_update(dn->inode,
569 							valid_blocks, false);
570 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
571 			valid_blocks = 0;
572 		}
573 
574 		if (blkaddr == NULL_ADDR)
575 			continue;
576 
577 		dn->data_blkaddr = NULL_ADDR;
578 		f2fs_set_data_blkaddr(dn);
579 
580 		if (__is_valid_data_blkaddr(blkaddr)) {
581 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
582 					DATA_GENERIC_ENHANCE))
583 				continue;
584 			if (compressed_cluster)
585 				valid_blocks++;
586 		}
587 
588 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
589 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
590 
591 		f2fs_invalidate_blocks(sbi, blkaddr);
592 
593 		if (!released || blkaddr != COMPRESS_ADDR)
594 			nr_free++;
595 	}
596 
597 	if (compressed_cluster)
598 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
599 
600 	if (nr_free) {
601 		pgoff_t fofs;
602 		/*
603 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
604 		 * we will invalidate all blkaddr in the whole range.
605 		 */
606 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
607 							dn->inode) + ofs;
608 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
609 		dec_valid_block_count(sbi, dn->inode, nr_free);
610 	}
611 	dn->ofs_in_node = ofs;
612 
613 	f2fs_update_time(sbi, REQ_TIME);
614 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
615 					 dn->ofs_in_node, nr_free);
616 }
617 
618 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
619 {
620 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
621 }
622 
623 static int truncate_partial_data_page(struct inode *inode, u64 from,
624 								bool cache_only)
625 {
626 	loff_t offset = from & (PAGE_SIZE - 1);
627 	pgoff_t index = from >> PAGE_SHIFT;
628 	struct address_space *mapping = inode->i_mapping;
629 	struct page *page;
630 
631 	if (!offset && !cache_only)
632 		return 0;
633 
634 	if (cache_only) {
635 		page = find_lock_page(mapping, index);
636 		if (page && PageUptodate(page))
637 			goto truncate_out;
638 		f2fs_put_page(page, 1);
639 		return 0;
640 	}
641 
642 	page = f2fs_get_lock_data_page(inode, index, true);
643 	if (IS_ERR(page))
644 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
645 truncate_out:
646 	f2fs_wait_on_page_writeback(page, DATA, true, true);
647 	zero_user(page, offset, PAGE_SIZE - offset);
648 
649 	/* An encrypted inode should have a key and truncate the last page. */
650 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
651 	if (!cache_only)
652 		set_page_dirty(page);
653 	f2fs_put_page(page, 1);
654 	return 0;
655 }
656 
657 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
658 {
659 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
660 	struct dnode_of_data dn;
661 	pgoff_t free_from;
662 	int count = 0, err = 0;
663 	struct page *ipage;
664 	bool truncate_page = false;
665 
666 	trace_f2fs_truncate_blocks_enter(inode, from);
667 
668 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
669 
670 	if (free_from >= sbi->max_file_blocks)
671 		goto free_partial;
672 
673 	if (lock)
674 		f2fs_lock_op(sbi);
675 
676 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
677 	if (IS_ERR(ipage)) {
678 		err = PTR_ERR(ipage);
679 		goto out;
680 	}
681 
682 	if (f2fs_has_inline_data(inode)) {
683 		f2fs_truncate_inline_inode(inode, ipage, from);
684 		f2fs_put_page(ipage, 1);
685 		truncate_page = true;
686 		goto out;
687 	}
688 
689 	set_new_dnode(&dn, inode, ipage, NULL, 0);
690 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
691 	if (err) {
692 		if (err == -ENOENT)
693 			goto free_next;
694 		goto out;
695 	}
696 
697 	count = ADDRS_PER_PAGE(dn.node_page, inode);
698 
699 	count -= dn.ofs_in_node;
700 	f2fs_bug_on(sbi, count < 0);
701 
702 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
703 		f2fs_truncate_data_blocks_range(&dn, count);
704 		free_from += count;
705 	}
706 
707 	f2fs_put_dnode(&dn);
708 free_next:
709 	err = f2fs_truncate_inode_blocks(inode, free_from);
710 out:
711 	if (lock)
712 		f2fs_unlock_op(sbi);
713 free_partial:
714 	/* lastly zero out the first data page */
715 	if (!err)
716 		err = truncate_partial_data_page(inode, from, truncate_page);
717 
718 	trace_f2fs_truncate_blocks_exit(inode, err);
719 	return err;
720 }
721 
722 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
723 {
724 	u64 free_from = from;
725 	int err;
726 
727 #ifdef CONFIG_F2FS_FS_COMPRESSION
728 	/*
729 	 * for compressed file, only support cluster size
730 	 * aligned truncation.
731 	 */
732 	if (f2fs_compressed_file(inode))
733 		free_from = round_up(from,
734 				F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
735 #endif
736 
737 	err = f2fs_do_truncate_blocks(inode, free_from, lock);
738 	if (err)
739 		return err;
740 
741 #ifdef CONFIG_F2FS_FS_COMPRESSION
742 	if (from != free_from) {
743 		err = f2fs_truncate_partial_cluster(inode, from, lock);
744 		if (err)
745 			return err;
746 	}
747 #endif
748 
749 	return 0;
750 }
751 
752 int f2fs_truncate(struct inode *inode)
753 {
754 	int err;
755 
756 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
757 		return -EIO;
758 
759 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
760 				S_ISLNK(inode->i_mode)))
761 		return 0;
762 
763 	trace_f2fs_truncate(inode);
764 
765 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
766 		f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
767 		return -EIO;
768 	}
769 
770 	/* we should check inline_data size */
771 	if (!f2fs_may_inline_data(inode)) {
772 		err = f2fs_convert_inline_inode(inode);
773 		if (err)
774 			return err;
775 	}
776 
777 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
778 	if (err)
779 		return err;
780 
781 	inode->i_mtime = inode->i_ctime = current_time(inode);
782 	f2fs_mark_inode_dirty_sync(inode, false);
783 	return 0;
784 }
785 
786 int f2fs_getattr(const struct path *path, struct kstat *stat,
787 		 u32 request_mask, unsigned int query_flags)
788 {
789 	struct inode *inode = d_inode(path->dentry);
790 	struct f2fs_inode_info *fi = F2FS_I(inode);
791 	struct f2fs_inode *ri;
792 	unsigned int flags;
793 
794 	if (f2fs_has_extra_attr(inode) &&
795 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
796 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
797 		stat->result_mask |= STATX_BTIME;
798 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
799 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
800 	}
801 
802 	flags = fi->i_flags;
803 	if (flags & F2FS_COMPR_FL)
804 		stat->attributes |= STATX_ATTR_COMPRESSED;
805 	if (flags & F2FS_APPEND_FL)
806 		stat->attributes |= STATX_ATTR_APPEND;
807 	if (IS_ENCRYPTED(inode))
808 		stat->attributes |= STATX_ATTR_ENCRYPTED;
809 	if (flags & F2FS_IMMUTABLE_FL)
810 		stat->attributes |= STATX_ATTR_IMMUTABLE;
811 	if (flags & F2FS_NODUMP_FL)
812 		stat->attributes |= STATX_ATTR_NODUMP;
813 	if (IS_VERITY(inode))
814 		stat->attributes |= STATX_ATTR_VERITY;
815 
816 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
817 				  STATX_ATTR_APPEND |
818 				  STATX_ATTR_ENCRYPTED |
819 				  STATX_ATTR_IMMUTABLE |
820 				  STATX_ATTR_NODUMP |
821 				  STATX_ATTR_VERITY);
822 
823 	generic_fillattr(inode, stat);
824 
825 	/* we need to show initial sectors used for inline_data/dentries */
826 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
827 					f2fs_has_inline_dentry(inode))
828 		stat->blocks += (stat->size + 511) >> 9;
829 
830 	return 0;
831 }
832 
833 #ifdef CONFIG_F2FS_FS_POSIX_ACL
834 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
835 {
836 	unsigned int ia_valid = attr->ia_valid;
837 
838 	if (ia_valid & ATTR_UID)
839 		inode->i_uid = attr->ia_uid;
840 	if (ia_valid & ATTR_GID)
841 		inode->i_gid = attr->ia_gid;
842 	if (ia_valid & ATTR_ATIME)
843 		inode->i_atime = attr->ia_atime;
844 	if (ia_valid & ATTR_MTIME)
845 		inode->i_mtime = attr->ia_mtime;
846 	if (ia_valid & ATTR_CTIME)
847 		inode->i_ctime = attr->ia_ctime;
848 	if (ia_valid & ATTR_MODE) {
849 		umode_t mode = attr->ia_mode;
850 
851 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
852 			mode &= ~S_ISGID;
853 		set_acl_inode(inode, mode);
854 	}
855 }
856 #else
857 #define __setattr_copy setattr_copy
858 #endif
859 
860 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
861 {
862 	struct inode *inode = d_inode(dentry);
863 	int err;
864 
865 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
866 		return -EIO;
867 
868 	if ((attr->ia_valid & ATTR_SIZE) &&
869 		!f2fs_is_compress_backend_ready(inode))
870 		return -EOPNOTSUPP;
871 
872 	err = setattr_prepare(dentry, attr);
873 	if (err)
874 		return err;
875 
876 	err = fscrypt_prepare_setattr(dentry, attr);
877 	if (err)
878 		return err;
879 
880 	err = fsverity_prepare_setattr(dentry, attr);
881 	if (err)
882 		return err;
883 
884 	if (is_quota_modification(inode, attr)) {
885 		err = dquot_initialize(inode);
886 		if (err)
887 			return err;
888 	}
889 	if ((attr->ia_valid & ATTR_UID &&
890 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
891 		(attr->ia_valid & ATTR_GID &&
892 		!gid_eq(attr->ia_gid, inode->i_gid))) {
893 		f2fs_lock_op(F2FS_I_SB(inode));
894 		err = dquot_transfer(inode, attr);
895 		if (err) {
896 			set_sbi_flag(F2FS_I_SB(inode),
897 					SBI_QUOTA_NEED_REPAIR);
898 			f2fs_unlock_op(F2FS_I_SB(inode));
899 			return err;
900 		}
901 		/*
902 		 * update uid/gid under lock_op(), so that dquot and inode can
903 		 * be updated atomically.
904 		 */
905 		if (attr->ia_valid & ATTR_UID)
906 			inode->i_uid = attr->ia_uid;
907 		if (attr->ia_valid & ATTR_GID)
908 			inode->i_gid = attr->ia_gid;
909 		f2fs_mark_inode_dirty_sync(inode, true);
910 		f2fs_unlock_op(F2FS_I_SB(inode));
911 	}
912 
913 	if (attr->ia_valid & ATTR_SIZE) {
914 		loff_t old_size = i_size_read(inode);
915 
916 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
917 			/*
918 			 * should convert inline inode before i_size_write to
919 			 * keep smaller than inline_data size with inline flag.
920 			 */
921 			err = f2fs_convert_inline_inode(inode);
922 			if (err)
923 				return err;
924 		}
925 
926 		down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
927 		down_write(&F2FS_I(inode)->i_mmap_sem);
928 
929 		truncate_setsize(inode, attr->ia_size);
930 
931 		if (attr->ia_size <= old_size)
932 			err = f2fs_truncate(inode);
933 		/*
934 		 * do not trim all blocks after i_size if target size is
935 		 * larger than i_size.
936 		 */
937 		up_write(&F2FS_I(inode)->i_mmap_sem);
938 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
939 		if (err)
940 			return err;
941 
942 		spin_lock(&F2FS_I(inode)->i_size_lock);
943 		inode->i_mtime = inode->i_ctime = current_time(inode);
944 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
945 		spin_unlock(&F2FS_I(inode)->i_size_lock);
946 	}
947 
948 	__setattr_copy(inode, attr);
949 
950 	if (attr->ia_valid & ATTR_MODE) {
951 		err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
952 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
953 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
954 			clear_inode_flag(inode, FI_ACL_MODE);
955 		}
956 	}
957 
958 	/* file size may changed here */
959 	f2fs_mark_inode_dirty_sync(inode, true);
960 
961 	/* inode change will produce dirty node pages flushed by checkpoint */
962 	f2fs_balance_fs(F2FS_I_SB(inode), true);
963 
964 	return err;
965 }
966 
967 const struct inode_operations f2fs_file_inode_operations = {
968 	.getattr	= f2fs_getattr,
969 	.setattr	= f2fs_setattr,
970 	.get_acl	= f2fs_get_acl,
971 	.set_acl	= f2fs_set_acl,
972 	.listxattr	= f2fs_listxattr,
973 	.fiemap		= f2fs_fiemap,
974 };
975 
976 static int fill_zero(struct inode *inode, pgoff_t index,
977 					loff_t start, loff_t len)
978 {
979 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
980 	struct page *page;
981 
982 	if (!len)
983 		return 0;
984 
985 	f2fs_balance_fs(sbi, true);
986 
987 	f2fs_lock_op(sbi);
988 	page = f2fs_get_new_data_page(inode, NULL, index, false);
989 	f2fs_unlock_op(sbi);
990 
991 	if (IS_ERR(page))
992 		return PTR_ERR(page);
993 
994 	f2fs_wait_on_page_writeback(page, DATA, true, true);
995 	zero_user(page, start, len);
996 	set_page_dirty(page);
997 	f2fs_put_page(page, 1);
998 	return 0;
999 }
1000 
1001 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1002 {
1003 	int err;
1004 
1005 	while (pg_start < pg_end) {
1006 		struct dnode_of_data dn;
1007 		pgoff_t end_offset, count;
1008 
1009 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1010 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1011 		if (err) {
1012 			if (err == -ENOENT) {
1013 				pg_start = f2fs_get_next_page_offset(&dn,
1014 								pg_start);
1015 				continue;
1016 			}
1017 			return err;
1018 		}
1019 
1020 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1021 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1022 
1023 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1024 
1025 		f2fs_truncate_data_blocks_range(&dn, count);
1026 		f2fs_put_dnode(&dn);
1027 
1028 		pg_start += count;
1029 	}
1030 	return 0;
1031 }
1032 
1033 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1034 {
1035 	pgoff_t pg_start, pg_end;
1036 	loff_t off_start, off_end;
1037 	int ret;
1038 
1039 	ret = f2fs_convert_inline_inode(inode);
1040 	if (ret)
1041 		return ret;
1042 
1043 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1044 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1045 
1046 	off_start = offset & (PAGE_SIZE - 1);
1047 	off_end = (offset + len) & (PAGE_SIZE - 1);
1048 
1049 	if (pg_start == pg_end) {
1050 		ret = fill_zero(inode, pg_start, off_start,
1051 						off_end - off_start);
1052 		if (ret)
1053 			return ret;
1054 	} else {
1055 		if (off_start) {
1056 			ret = fill_zero(inode, pg_start++, off_start,
1057 						PAGE_SIZE - off_start);
1058 			if (ret)
1059 				return ret;
1060 		}
1061 		if (off_end) {
1062 			ret = fill_zero(inode, pg_end, 0, off_end);
1063 			if (ret)
1064 				return ret;
1065 		}
1066 
1067 		if (pg_start < pg_end) {
1068 			struct address_space *mapping = inode->i_mapping;
1069 			loff_t blk_start, blk_end;
1070 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1071 
1072 			f2fs_balance_fs(sbi, true);
1073 
1074 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1075 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1076 
1077 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1078 			down_write(&F2FS_I(inode)->i_mmap_sem);
1079 
1080 			truncate_inode_pages_range(mapping, blk_start,
1081 					blk_end - 1);
1082 
1083 			f2fs_lock_op(sbi);
1084 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1085 			f2fs_unlock_op(sbi);
1086 
1087 			up_write(&F2FS_I(inode)->i_mmap_sem);
1088 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1089 		}
1090 	}
1091 
1092 	return ret;
1093 }
1094 
1095 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1096 				int *do_replace, pgoff_t off, pgoff_t len)
1097 {
1098 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1099 	struct dnode_of_data dn;
1100 	int ret, done, i;
1101 
1102 next_dnode:
1103 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1104 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1105 	if (ret && ret != -ENOENT) {
1106 		return ret;
1107 	} else if (ret == -ENOENT) {
1108 		if (dn.max_level == 0)
1109 			return -ENOENT;
1110 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1111 						dn.ofs_in_node, len);
1112 		blkaddr += done;
1113 		do_replace += done;
1114 		goto next;
1115 	}
1116 
1117 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1118 							dn.ofs_in_node, len);
1119 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1120 		*blkaddr = f2fs_data_blkaddr(&dn);
1121 
1122 		if (__is_valid_data_blkaddr(*blkaddr) &&
1123 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1124 					DATA_GENERIC_ENHANCE)) {
1125 			f2fs_put_dnode(&dn);
1126 			return -EFSCORRUPTED;
1127 		}
1128 
1129 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1130 
1131 			if (f2fs_lfs_mode(sbi)) {
1132 				f2fs_put_dnode(&dn);
1133 				return -EOPNOTSUPP;
1134 			}
1135 
1136 			/* do not invalidate this block address */
1137 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1138 			*do_replace = 1;
1139 		}
1140 	}
1141 	f2fs_put_dnode(&dn);
1142 next:
1143 	len -= done;
1144 	off += done;
1145 	if (len)
1146 		goto next_dnode;
1147 	return 0;
1148 }
1149 
1150 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1151 				int *do_replace, pgoff_t off, int len)
1152 {
1153 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1154 	struct dnode_of_data dn;
1155 	int ret, i;
1156 
1157 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1158 		if (*do_replace == 0)
1159 			continue;
1160 
1161 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1162 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1163 		if (ret) {
1164 			dec_valid_block_count(sbi, inode, 1);
1165 			f2fs_invalidate_blocks(sbi, *blkaddr);
1166 		} else {
1167 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1168 		}
1169 		f2fs_put_dnode(&dn);
1170 	}
1171 	return 0;
1172 }
1173 
1174 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1175 			block_t *blkaddr, int *do_replace,
1176 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1177 {
1178 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1179 	pgoff_t i = 0;
1180 	int ret;
1181 
1182 	while (i < len) {
1183 		if (blkaddr[i] == NULL_ADDR && !full) {
1184 			i++;
1185 			continue;
1186 		}
1187 
1188 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1189 			struct dnode_of_data dn;
1190 			struct node_info ni;
1191 			size_t new_size;
1192 			pgoff_t ilen;
1193 
1194 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1195 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1196 			if (ret)
1197 				return ret;
1198 
1199 			ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1200 			if (ret) {
1201 				f2fs_put_dnode(&dn);
1202 				return ret;
1203 			}
1204 
1205 			ilen = min((pgoff_t)
1206 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1207 						dn.ofs_in_node, len - i);
1208 			do {
1209 				dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1210 				f2fs_truncate_data_blocks_range(&dn, 1);
1211 
1212 				if (do_replace[i]) {
1213 					f2fs_i_blocks_write(src_inode,
1214 							1, false, false);
1215 					f2fs_i_blocks_write(dst_inode,
1216 							1, true, false);
1217 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1218 					blkaddr[i], ni.version, true, false);
1219 
1220 					do_replace[i] = 0;
1221 				}
1222 				dn.ofs_in_node++;
1223 				i++;
1224 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1225 				if (dst_inode->i_size < new_size)
1226 					f2fs_i_size_write(dst_inode, new_size);
1227 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1228 
1229 			f2fs_put_dnode(&dn);
1230 		} else {
1231 			struct page *psrc, *pdst;
1232 
1233 			psrc = f2fs_get_lock_data_page(src_inode,
1234 							src + i, true);
1235 			if (IS_ERR(psrc))
1236 				return PTR_ERR(psrc);
1237 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1238 								true);
1239 			if (IS_ERR(pdst)) {
1240 				f2fs_put_page(psrc, 1);
1241 				return PTR_ERR(pdst);
1242 			}
1243 			f2fs_copy_page(psrc, pdst);
1244 			set_page_dirty(pdst);
1245 			f2fs_put_page(pdst, 1);
1246 			f2fs_put_page(psrc, 1);
1247 
1248 			ret = f2fs_truncate_hole(src_inode,
1249 						src + i, src + i + 1);
1250 			if (ret)
1251 				return ret;
1252 			i++;
1253 		}
1254 	}
1255 	return 0;
1256 }
1257 
1258 static int __exchange_data_block(struct inode *src_inode,
1259 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1260 			pgoff_t len, bool full)
1261 {
1262 	block_t *src_blkaddr;
1263 	int *do_replace;
1264 	pgoff_t olen;
1265 	int ret;
1266 
1267 	while (len) {
1268 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1269 
1270 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1271 					array_size(olen, sizeof(block_t)),
1272 					GFP_NOFS);
1273 		if (!src_blkaddr)
1274 			return -ENOMEM;
1275 
1276 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1277 					array_size(olen, sizeof(int)),
1278 					GFP_NOFS);
1279 		if (!do_replace) {
1280 			kvfree(src_blkaddr);
1281 			return -ENOMEM;
1282 		}
1283 
1284 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1285 					do_replace, src, olen);
1286 		if (ret)
1287 			goto roll_back;
1288 
1289 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1290 					do_replace, src, dst, olen, full);
1291 		if (ret)
1292 			goto roll_back;
1293 
1294 		src += olen;
1295 		dst += olen;
1296 		len -= olen;
1297 
1298 		kvfree(src_blkaddr);
1299 		kvfree(do_replace);
1300 	}
1301 	return 0;
1302 
1303 roll_back:
1304 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1305 	kvfree(src_blkaddr);
1306 	kvfree(do_replace);
1307 	return ret;
1308 }
1309 
1310 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1311 {
1312 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1313 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1314 	pgoff_t start = offset >> PAGE_SHIFT;
1315 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1316 	int ret;
1317 
1318 	f2fs_balance_fs(sbi, true);
1319 
1320 	/* avoid gc operation during block exchange */
1321 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1322 	down_write(&F2FS_I(inode)->i_mmap_sem);
1323 
1324 	f2fs_lock_op(sbi);
1325 	f2fs_drop_extent_tree(inode);
1326 	truncate_pagecache(inode, offset);
1327 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1328 	f2fs_unlock_op(sbi);
1329 
1330 	up_write(&F2FS_I(inode)->i_mmap_sem);
1331 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1332 	return ret;
1333 }
1334 
1335 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1336 {
1337 	loff_t new_size;
1338 	int ret;
1339 
1340 	if (offset + len >= i_size_read(inode))
1341 		return -EINVAL;
1342 
1343 	/* collapse range should be aligned to block size of f2fs. */
1344 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1345 		return -EINVAL;
1346 
1347 	ret = f2fs_convert_inline_inode(inode);
1348 	if (ret)
1349 		return ret;
1350 
1351 	/* write out all dirty pages from offset */
1352 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1353 	if (ret)
1354 		return ret;
1355 
1356 	ret = f2fs_do_collapse(inode, offset, len);
1357 	if (ret)
1358 		return ret;
1359 
1360 	/* write out all moved pages, if possible */
1361 	down_write(&F2FS_I(inode)->i_mmap_sem);
1362 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1363 	truncate_pagecache(inode, offset);
1364 
1365 	new_size = i_size_read(inode) - len;
1366 	ret = f2fs_truncate_blocks(inode, new_size, true);
1367 	up_write(&F2FS_I(inode)->i_mmap_sem);
1368 	if (!ret)
1369 		f2fs_i_size_write(inode, new_size);
1370 	return ret;
1371 }
1372 
1373 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1374 								pgoff_t end)
1375 {
1376 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1377 	pgoff_t index = start;
1378 	unsigned int ofs_in_node = dn->ofs_in_node;
1379 	blkcnt_t count = 0;
1380 	int ret;
1381 
1382 	for (; index < end; index++, dn->ofs_in_node++) {
1383 		if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1384 			count++;
1385 	}
1386 
1387 	dn->ofs_in_node = ofs_in_node;
1388 	ret = f2fs_reserve_new_blocks(dn, count);
1389 	if (ret)
1390 		return ret;
1391 
1392 	dn->ofs_in_node = ofs_in_node;
1393 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1394 		dn->data_blkaddr = f2fs_data_blkaddr(dn);
1395 		/*
1396 		 * f2fs_reserve_new_blocks will not guarantee entire block
1397 		 * allocation.
1398 		 */
1399 		if (dn->data_blkaddr == NULL_ADDR) {
1400 			ret = -ENOSPC;
1401 			break;
1402 		}
1403 		if (dn->data_blkaddr != NEW_ADDR) {
1404 			f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1405 			dn->data_blkaddr = NEW_ADDR;
1406 			f2fs_set_data_blkaddr(dn);
1407 		}
1408 	}
1409 
1410 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1411 
1412 	return ret;
1413 }
1414 
1415 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1416 								int mode)
1417 {
1418 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1419 	struct address_space *mapping = inode->i_mapping;
1420 	pgoff_t index, pg_start, pg_end;
1421 	loff_t new_size = i_size_read(inode);
1422 	loff_t off_start, off_end;
1423 	int ret = 0;
1424 
1425 	ret = inode_newsize_ok(inode, (len + offset));
1426 	if (ret)
1427 		return ret;
1428 
1429 	ret = f2fs_convert_inline_inode(inode);
1430 	if (ret)
1431 		return ret;
1432 
1433 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1434 	if (ret)
1435 		return ret;
1436 
1437 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1438 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1439 
1440 	off_start = offset & (PAGE_SIZE - 1);
1441 	off_end = (offset + len) & (PAGE_SIZE - 1);
1442 
1443 	if (pg_start == pg_end) {
1444 		ret = fill_zero(inode, pg_start, off_start,
1445 						off_end - off_start);
1446 		if (ret)
1447 			return ret;
1448 
1449 		new_size = max_t(loff_t, new_size, offset + len);
1450 	} else {
1451 		if (off_start) {
1452 			ret = fill_zero(inode, pg_start++, off_start,
1453 						PAGE_SIZE - off_start);
1454 			if (ret)
1455 				return ret;
1456 
1457 			new_size = max_t(loff_t, new_size,
1458 					(loff_t)pg_start << PAGE_SHIFT);
1459 		}
1460 
1461 		for (index = pg_start; index < pg_end;) {
1462 			struct dnode_of_data dn;
1463 			unsigned int end_offset;
1464 			pgoff_t end;
1465 
1466 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1467 			down_write(&F2FS_I(inode)->i_mmap_sem);
1468 
1469 			truncate_pagecache_range(inode,
1470 				(loff_t)index << PAGE_SHIFT,
1471 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1472 
1473 			f2fs_lock_op(sbi);
1474 
1475 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1476 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1477 			if (ret) {
1478 				f2fs_unlock_op(sbi);
1479 				up_write(&F2FS_I(inode)->i_mmap_sem);
1480 				up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1481 				goto out;
1482 			}
1483 
1484 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1485 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1486 
1487 			ret = f2fs_do_zero_range(&dn, index, end);
1488 			f2fs_put_dnode(&dn);
1489 
1490 			f2fs_unlock_op(sbi);
1491 			up_write(&F2FS_I(inode)->i_mmap_sem);
1492 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1493 
1494 			f2fs_balance_fs(sbi, dn.node_changed);
1495 
1496 			if (ret)
1497 				goto out;
1498 
1499 			index = end;
1500 			new_size = max_t(loff_t, new_size,
1501 					(loff_t)index << PAGE_SHIFT);
1502 		}
1503 
1504 		if (off_end) {
1505 			ret = fill_zero(inode, pg_end, 0, off_end);
1506 			if (ret)
1507 				goto out;
1508 
1509 			new_size = max_t(loff_t, new_size, offset + len);
1510 		}
1511 	}
1512 
1513 out:
1514 	if (new_size > i_size_read(inode)) {
1515 		if (mode & FALLOC_FL_KEEP_SIZE)
1516 			file_set_keep_isize(inode);
1517 		else
1518 			f2fs_i_size_write(inode, new_size);
1519 	}
1520 	return ret;
1521 }
1522 
1523 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1524 {
1525 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1526 	pgoff_t nr, pg_start, pg_end, delta, idx;
1527 	loff_t new_size;
1528 	int ret = 0;
1529 
1530 	new_size = i_size_read(inode) + len;
1531 	ret = inode_newsize_ok(inode, new_size);
1532 	if (ret)
1533 		return ret;
1534 
1535 	if (offset >= i_size_read(inode))
1536 		return -EINVAL;
1537 
1538 	/* insert range should be aligned to block size of f2fs. */
1539 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1540 		return -EINVAL;
1541 
1542 	ret = f2fs_convert_inline_inode(inode);
1543 	if (ret)
1544 		return ret;
1545 
1546 	f2fs_balance_fs(sbi, true);
1547 
1548 	down_write(&F2FS_I(inode)->i_mmap_sem);
1549 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1550 	up_write(&F2FS_I(inode)->i_mmap_sem);
1551 	if (ret)
1552 		return ret;
1553 
1554 	/* write out all dirty pages from offset */
1555 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1556 	if (ret)
1557 		return ret;
1558 
1559 	pg_start = offset >> PAGE_SHIFT;
1560 	pg_end = (offset + len) >> PAGE_SHIFT;
1561 	delta = pg_end - pg_start;
1562 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1563 
1564 	/* avoid gc operation during block exchange */
1565 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1566 	down_write(&F2FS_I(inode)->i_mmap_sem);
1567 	truncate_pagecache(inode, offset);
1568 
1569 	while (!ret && idx > pg_start) {
1570 		nr = idx - pg_start;
1571 		if (nr > delta)
1572 			nr = delta;
1573 		idx -= nr;
1574 
1575 		f2fs_lock_op(sbi);
1576 		f2fs_drop_extent_tree(inode);
1577 
1578 		ret = __exchange_data_block(inode, inode, idx,
1579 					idx + delta, nr, false);
1580 		f2fs_unlock_op(sbi);
1581 	}
1582 	up_write(&F2FS_I(inode)->i_mmap_sem);
1583 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1584 
1585 	/* write out all moved pages, if possible */
1586 	down_write(&F2FS_I(inode)->i_mmap_sem);
1587 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1588 	truncate_pagecache(inode, offset);
1589 	up_write(&F2FS_I(inode)->i_mmap_sem);
1590 
1591 	if (!ret)
1592 		f2fs_i_size_write(inode, new_size);
1593 	return ret;
1594 }
1595 
1596 static int expand_inode_data(struct inode *inode, loff_t offset,
1597 					loff_t len, int mode)
1598 {
1599 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1600 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1601 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1602 			.m_may_create = true };
1603 	pgoff_t pg_end;
1604 	loff_t new_size = i_size_read(inode);
1605 	loff_t off_end;
1606 	int err;
1607 
1608 	err = inode_newsize_ok(inode, (len + offset));
1609 	if (err)
1610 		return err;
1611 
1612 	err = f2fs_convert_inline_inode(inode);
1613 	if (err)
1614 		return err;
1615 
1616 	f2fs_balance_fs(sbi, true);
1617 
1618 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1619 	off_end = (offset + len) & (PAGE_SIZE - 1);
1620 
1621 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1622 	map.m_len = pg_end - map.m_lblk;
1623 	if (off_end)
1624 		map.m_len++;
1625 
1626 	if (!map.m_len)
1627 		return 0;
1628 
1629 	if (f2fs_is_pinned_file(inode)) {
1630 		block_t len = (map.m_len >> sbi->log_blocks_per_seg) <<
1631 					sbi->log_blocks_per_seg;
1632 		block_t done = 0;
1633 
1634 		if (map.m_len % sbi->blocks_per_seg)
1635 			len += sbi->blocks_per_seg;
1636 
1637 		map.m_len = sbi->blocks_per_seg;
1638 next_alloc:
1639 		if (has_not_enough_free_secs(sbi, 0,
1640 			GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1641 			down_write(&sbi->gc_lock);
1642 			err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1643 			if (err && err != -ENODATA && err != -EAGAIN)
1644 				goto out_err;
1645 		}
1646 
1647 		down_write(&sbi->pin_sem);
1648 
1649 		f2fs_lock_op(sbi);
1650 		f2fs_allocate_new_segment(sbi, CURSEG_COLD_DATA_PINNED);
1651 		f2fs_unlock_op(sbi);
1652 
1653 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1654 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1655 
1656 		up_write(&sbi->pin_sem);
1657 
1658 		done += map.m_len;
1659 		len -= map.m_len;
1660 		map.m_lblk += map.m_len;
1661 		if (!err && len)
1662 			goto next_alloc;
1663 
1664 		map.m_len = done;
1665 	} else {
1666 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1667 	}
1668 out_err:
1669 	if (err) {
1670 		pgoff_t last_off;
1671 
1672 		if (!map.m_len)
1673 			return err;
1674 
1675 		last_off = map.m_lblk + map.m_len - 1;
1676 
1677 		/* update new size to the failed position */
1678 		new_size = (last_off == pg_end) ? offset + len :
1679 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1680 	} else {
1681 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1682 	}
1683 
1684 	if (new_size > i_size_read(inode)) {
1685 		if (mode & FALLOC_FL_KEEP_SIZE)
1686 			file_set_keep_isize(inode);
1687 		else
1688 			f2fs_i_size_write(inode, new_size);
1689 	}
1690 
1691 	return err;
1692 }
1693 
1694 static long f2fs_fallocate(struct file *file, int mode,
1695 				loff_t offset, loff_t len)
1696 {
1697 	struct inode *inode = file_inode(file);
1698 	long ret = 0;
1699 
1700 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1701 		return -EIO;
1702 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1703 		return -ENOSPC;
1704 	if (!f2fs_is_compress_backend_ready(inode))
1705 		return -EOPNOTSUPP;
1706 
1707 	/* f2fs only support ->fallocate for regular file */
1708 	if (!S_ISREG(inode->i_mode))
1709 		return -EINVAL;
1710 
1711 	if (IS_ENCRYPTED(inode) &&
1712 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1713 		return -EOPNOTSUPP;
1714 
1715 	if (f2fs_compressed_file(inode) &&
1716 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1717 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1718 		return -EOPNOTSUPP;
1719 
1720 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1721 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1722 			FALLOC_FL_INSERT_RANGE))
1723 		return -EOPNOTSUPP;
1724 
1725 	inode_lock(inode);
1726 
1727 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1728 		if (offset >= inode->i_size)
1729 			goto out;
1730 
1731 		ret = punch_hole(inode, offset, len);
1732 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1733 		ret = f2fs_collapse_range(inode, offset, len);
1734 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1735 		ret = f2fs_zero_range(inode, offset, len, mode);
1736 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1737 		ret = f2fs_insert_range(inode, offset, len);
1738 	} else {
1739 		ret = expand_inode_data(inode, offset, len, mode);
1740 	}
1741 
1742 	if (!ret) {
1743 		inode->i_mtime = inode->i_ctime = current_time(inode);
1744 		f2fs_mark_inode_dirty_sync(inode, false);
1745 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1746 	}
1747 
1748 out:
1749 	inode_unlock(inode);
1750 
1751 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1752 	return ret;
1753 }
1754 
1755 static int f2fs_release_file(struct inode *inode, struct file *filp)
1756 {
1757 	/*
1758 	 * f2fs_relase_file is called at every close calls. So we should
1759 	 * not drop any inmemory pages by close called by other process.
1760 	 */
1761 	if (!(filp->f_mode & FMODE_WRITE) ||
1762 			atomic_read(&inode->i_writecount) != 1)
1763 		return 0;
1764 
1765 	/* some remained atomic pages should discarded */
1766 	if (f2fs_is_atomic_file(inode))
1767 		f2fs_drop_inmem_pages(inode);
1768 	if (f2fs_is_volatile_file(inode)) {
1769 		set_inode_flag(inode, FI_DROP_CACHE);
1770 		filemap_fdatawrite(inode->i_mapping);
1771 		clear_inode_flag(inode, FI_DROP_CACHE);
1772 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1773 		stat_dec_volatile_write(inode);
1774 	}
1775 	return 0;
1776 }
1777 
1778 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1779 {
1780 	struct inode *inode = file_inode(file);
1781 
1782 	/*
1783 	 * If the process doing a transaction is crashed, we should do
1784 	 * roll-back. Otherwise, other reader/write can see corrupted database
1785 	 * until all the writers close its file. Since this should be done
1786 	 * before dropping file lock, it needs to do in ->flush.
1787 	 */
1788 	if (f2fs_is_atomic_file(inode) &&
1789 			F2FS_I(inode)->inmem_task == current)
1790 		f2fs_drop_inmem_pages(inode);
1791 	return 0;
1792 }
1793 
1794 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1795 {
1796 	struct f2fs_inode_info *fi = F2FS_I(inode);
1797 	u32 masked_flags = fi->i_flags & mask;
1798 
1799 	f2fs_bug_on(F2FS_I_SB(inode), (iflags & ~mask));
1800 
1801 	/* Is it quota file? Do not allow user to mess with it */
1802 	if (IS_NOQUOTA(inode))
1803 		return -EPERM;
1804 
1805 	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1806 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1807 			return -EOPNOTSUPP;
1808 		if (!f2fs_empty_dir(inode))
1809 			return -ENOTEMPTY;
1810 	}
1811 
1812 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1813 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1814 			return -EOPNOTSUPP;
1815 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1816 			return -EINVAL;
1817 	}
1818 
1819 	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1820 		if (masked_flags & F2FS_COMPR_FL) {
1821 			if (!f2fs_disable_compressed_file(inode))
1822 				return -EINVAL;
1823 		}
1824 		if (iflags & F2FS_NOCOMP_FL)
1825 			return -EINVAL;
1826 		if (iflags & F2FS_COMPR_FL) {
1827 			if (!f2fs_may_compress(inode))
1828 				return -EINVAL;
1829 			if (S_ISREG(inode->i_mode) && inode->i_size)
1830 				return -EINVAL;
1831 
1832 			set_compress_context(inode);
1833 		}
1834 	}
1835 	if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1836 		if (masked_flags & F2FS_COMPR_FL)
1837 			return -EINVAL;
1838 	}
1839 
1840 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1841 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1842 					(fi->i_flags & F2FS_NOCOMP_FL));
1843 
1844 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1845 		set_inode_flag(inode, FI_PROJ_INHERIT);
1846 	else
1847 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1848 
1849 	inode->i_ctime = current_time(inode);
1850 	f2fs_set_inode_flags(inode);
1851 	f2fs_mark_inode_dirty_sync(inode, true);
1852 	return 0;
1853 }
1854 
1855 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1856 
1857 /*
1858  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1859  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1860  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1861  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1862  */
1863 
1864 static const struct {
1865 	u32 iflag;
1866 	u32 fsflag;
1867 } f2fs_fsflags_map[] = {
1868 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
1869 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1870 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1871 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1872 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1873 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1874 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
1875 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1876 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1877 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1878 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
1879 };
1880 
1881 #define F2FS_GETTABLE_FS_FL (		\
1882 		FS_COMPR_FL |		\
1883 		FS_SYNC_FL |		\
1884 		FS_IMMUTABLE_FL |	\
1885 		FS_APPEND_FL |		\
1886 		FS_NODUMP_FL |		\
1887 		FS_NOATIME_FL |		\
1888 		FS_NOCOMP_FL |		\
1889 		FS_INDEX_FL |		\
1890 		FS_DIRSYNC_FL |		\
1891 		FS_PROJINHERIT_FL |	\
1892 		FS_ENCRYPT_FL |		\
1893 		FS_INLINE_DATA_FL |	\
1894 		FS_NOCOW_FL |		\
1895 		FS_VERITY_FL |		\
1896 		FS_CASEFOLD_FL)
1897 
1898 #define F2FS_SETTABLE_FS_FL (		\
1899 		FS_COMPR_FL |		\
1900 		FS_SYNC_FL |		\
1901 		FS_IMMUTABLE_FL |	\
1902 		FS_APPEND_FL |		\
1903 		FS_NODUMP_FL |		\
1904 		FS_NOATIME_FL |		\
1905 		FS_NOCOMP_FL |		\
1906 		FS_DIRSYNC_FL |		\
1907 		FS_PROJINHERIT_FL |	\
1908 		FS_CASEFOLD_FL)
1909 
1910 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1911 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1912 {
1913 	u32 fsflags = 0;
1914 	int i;
1915 
1916 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1917 		if (iflags & f2fs_fsflags_map[i].iflag)
1918 			fsflags |= f2fs_fsflags_map[i].fsflag;
1919 
1920 	return fsflags;
1921 }
1922 
1923 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1924 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1925 {
1926 	u32 iflags = 0;
1927 	int i;
1928 
1929 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1930 		if (fsflags & f2fs_fsflags_map[i].fsflag)
1931 			iflags |= f2fs_fsflags_map[i].iflag;
1932 
1933 	return iflags;
1934 }
1935 
1936 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1937 {
1938 	struct inode *inode = file_inode(filp);
1939 	struct f2fs_inode_info *fi = F2FS_I(inode);
1940 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1941 
1942 	if (IS_ENCRYPTED(inode))
1943 		fsflags |= FS_ENCRYPT_FL;
1944 	if (IS_VERITY(inode))
1945 		fsflags |= FS_VERITY_FL;
1946 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1947 		fsflags |= FS_INLINE_DATA_FL;
1948 	if (is_inode_flag_set(inode, FI_PIN_FILE))
1949 		fsflags |= FS_NOCOW_FL;
1950 
1951 	fsflags &= F2FS_GETTABLE_FS_FL;
1952 
1953 	return put_user(fsflags, (int __user *)arg);
1954 }
1955 
1956 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1957 {
1958 	struct inode *inode = file_inode(filp);
1959 	struct f2fs_inode_info *fi = F2FS_I(inode);
1960 	u32 fsflags, old_fsflags;
1961 	u32 iflags;
1962 	int ret;
1963 
1964 	if (!inode_owner_or_capable(inode))
1965 		return -EACCES;
1966 
1967 	if (get_user(fsflags, (int __user *)arg))
1968 		return -EFAULT;
1969 
1970 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
1971 		return -EOPNOTSUPP;
1972 	fsflags &= F2FS_SETTABLE_FS_FL;
1973 
1974 	iflags = f2fs_fsflags_to_iflags(fsflags);
1975 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1976 		return -EOPNOTSUPP;
1977 
1978 	ret = mnt_want_write_file(filp);
1979 	if (ret)
1980 		return ret;
1981 
1982 	inode_lock(inode);
1983 
1984 	old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1985 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1986 	if (ret)
1987 		goto out;
1988 
1989 	ret = f2fs_setflags_common(inode, iflags,
1990 			f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1991 out:
1992 	inode_unlock(inode);
1993 	mnt_drop_write_file(filp);
1994 	return ret;
1995 }
1996 
1997 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1998 {
1999 	struct inode *inode = file_inode(filp);
2000 
2001 	return put_user(inode->i_generation, (int __user *)arg);
2002 }
2003 
2004 static int f2fs_ioc_start_atomic_write(struct file *filp)
2005 {
2006 	struct inode *inode = file_inode(filp);
2007 	struct f2fs_inode_info *fi = F2FS_I(inode);
2008 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2009 	int ret;
2010 
2011 	if (!inode_owner_or_capable(inode))
2012 		return -EACCES;
2013 
2014 	if (!S_ISREG(inode->i_mode))
2015 		return -EINVAL;
2016 
2017 	if (filp->f_flags & O_DIRECT)
2018 		return -EINVAL;
2019 
2020 	ret = mnt_want_write_file(filp);
2021 	if (ret)
2022 		return ret;
2023 
2024 	inode_lock(inode);
2025 
2026 	f2fs_disable_compressed_file(inode);
2027 
2028 	if (f2fs_is_atomic_file(inode)) {
2029 		if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2030 			ret = -EINVAL;
2031 		goto out;
2032 	}
2033 
2034 	ret = f2fs_convert_inline_inode(inode);
2035 	if (ret)
2036 		goto out;
2037 
2038 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2039 
2040 	/*
2041 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2042 	 * f2fs_is_atomic_file.
2043 	 */
2044 	if (get_dirty_pages(inode))
2045 		f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2046 			  inode->i_ino, get_dirty_pages(inode));
2047 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2048 	if (ret) {
2049 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2050 		goto out;
2051 	}
2052 
2053 	spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2054 	if (list_empty(&fi->inmem_ilist))
2055 		list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2056 	sbi->atomic_files++;
2057 	spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2058 
2059 	/* add inode in inmem_list first and set atomic_file */
2060 	set_inode_flag(inode, FI_ATOMIC_FILE);
2061 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2062 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2063 
2064 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2065 	F2FS_I(inode)->inmem_task = current;
2066 	stat_update_max_atomic_write(inode);
2067 out:
2068 	inode_unlock(inode);
2069 	mnt_drop_write_file(filp);
2070 	return ret;
2071 }
2072 
2073 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2074 {
2075 	struct inode *inode = file_inode(filp);
2076 	int ret;
2077 
2078 	if (!inode_owner_or_capable(inode))
2079 		return -EACCES;
2080 
2081 	ret = mnt_want_write_file(filp);
2082 	if (ret)
2083 		return ret;
2084 
2085 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2086 
2087 	inode_lock(inode);
2088 
2089 	if (f2fs_is_volatile_file(inode)) {
2090 		ret = -EINVAL;
2091 		goto err_out;
2092 	}
2093 
2094 	if (f2fs_is_atomic_file(inode)) {
2095 		ret = f2fs_commit_inmem_pages(inode);
2096 		if (ret)
2097 			goto err_out;
2098 
2099 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2100 		if (!ret)
2101 			f2fs_drop_inmem_pages(inode);
2102 	} else {
2103 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2104 	}
2105 err_out:
2106 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2107 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2108 		ret = -EINVAL;
2109 	}
2110 	inode_unlock(inode);
2111 	mnt_drop_write_file(filp);
2112 	return ret;
2113 }
2114 
2115 static int f2fs_ioc_start_volatile_write(struct file *filp)
2116 {
2117 	struct inode *inode = file_inode(filp);
2118 	int ret;
2119 
2120 	if (!inode_owner_or_capable(inode))
2121 		return -EACCES;
2122 
2123 	if (!S_ISREG(inode->i_mode))
2124 		return -EINVAL;
2125 
2126 	ret = mnt_want_write_file(filp);
2127 	if (ret)
2128 		return ret;
2129 
2130 	inode_lock(inode);
2131 
2132 	if (f2fs_is_volatile_file(inode))
2133 		goto out;
2134 
2135 	ret = f2fs_convert_inline_inode(inode);
2136 	if (ret)
2137 		goto out;
2138 
2139 	stat_inc_volatile_write(inode);
2140 	stat_update_max_volatile_write(inode);
2141 
2142 	set_inode_flag(inode, FI_VOLATILE_FILE);
2143 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2144 out:
2145 	inode_unlock(inode);
2146 	mnt_drop_write_file(filp);
2147 	return ret;
2148 }
2149 
2150 static int f2fs_ioc_release_volatile_write(struct file *filp)
2151 {
2152 	struct inode *inode = file_inode(filp);
2153 	int ret;
2154 
2155 	if (!inode_owner_or_capable(inode))
2156 		return -EACCES;
2157 
2158 	ret = mnt_want_write_file(filp);
2159 	if (ret)
2160 		return ret;
2161 
2162 	inode_lock(inode);
2163 
2164 	if (!f2fs_is_volatile_file(inode))
2165 		goto out;
2166 
2167 	if (!f2fs_is_first_block_written(inode)) {
2168 		ret = truncate_partial_data_page(inode, 0, true);
2169 		goto out;
2170 	}
2171 
2172 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2173 out:
2174 	inode_unlock(inode);
2175 	mnt_drop_write_file(filp);
2176 	return ret;
2177 }
2178 
2179 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2180 {
2181 	struct inode *inode = file_inode(filp);
2182 	int ret;
2183 
2184 	if (!inode_owner_or_capable(inode))
2185 		return -EACCES;
2186 
2187 	ret = mnt_want_write_file(filp);
2188 	if (ret)
2189 		return ret;
2190 
2191 	inode_lock(inode);
2192 
2193 	if (f2fs_is_atomic_file(inode))
2194 		f2fs_drop_inmem_pages(inode);
2195 	if (f2fs_is_volatile_file(inode)) {
2196 		clear_inode_flag(inode, FI_VOLATILE_FILE);
2197 		stat_dec_volatile_write(inode);
2198 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2199 	}
2200 
2201 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2202 
2203 	inode_unlock(inode);
2204 
2205 	mnt_drop_write_file(filp);
2206 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2207 	return ret;
2208 }
2209 
2210 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2211 {
2212 	struct inode *inode = file_inode(filp);
2213 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2214 	struct super_block *sb = sbi->sb;
2215 	__u32 in;
2216 	int ret = 0;
2217 
2218 	if (!capable(CAP_SYS_ADMIN))
2219 		return -EPERM;
2220 
2221 	if (get_user(in, (__u32 __user *)arg))
2222 		return -EFAULT;
2223 
2224 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2225 		ret = mnt_want_write_file(filp);
2226 		if (ret) {
2227 			if (ret == -EROFS) {
2228 				ret = 0;
2229 				f2fs_stop_checkpoint(sbi, false);
2230 				set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2231 				trace_f2fs_shutdown(sbi, in, ret);
2232 			}
2233 			return ret;
2234 		}
2235 	}
2236 
2237 	switch (in) {
2238 	case F2FS_GOING_DOWN_FULLSYNC:
2239 		ret = freeze_bdev(sb->s_bdev);
2240 		if (ret)
2241 			goto out;
2242 		f2fs_stop_checkpoint(sbi, false);
2243 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2244 		thaw_bdev(sb->s_bdev);
2245 		break;
2246 	case F2FS_GOING_DOWN_METASYNC:
2247 		/* do checkpoint only */
2248 		ret = f2fs_sync_fs(sb, 1);
2249 		if (ret)
2250 			goto out;
2251 		f2fs_stop_checkpoint(sbi, false);
2252 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2253 		break;
2254 	case F2FS_GOING_DOWN_NOSYNC:
2255 		f2fs_stop_checkpoint(sbi, false);
2256 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2257 		break;
2258 	case F2FS_GOING_DOWN_METAFLUSH:
2259 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2260 		f2fs_stop_checkpoint(sbi, false);
2261 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2262 		break;
2263 	case F2FS_GOING_DOWN_NEED_FSCK:
2264 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2265 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2266 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2267 		/* do checkpoint only */
2268 		ret = f2fs_sync_fs(sb, 1);
2269 		goto out;
2270 	default:
2271 		ret = -EINVAL;
2272 		goto out;
2273 	}
2274 
2275 	f2fs_stop_gc_thread(sbi);
2276 	f2fs_stop_discard_thread(sbi);
2277 
2278 	f2fs_drop_discard_cmd(sbi);
2279 	clear_opt(sbi, DISCARD);
2280 
2281 	f2fs_update_time(sbi, REQ_TIME);
2282 out:
2283 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2284 		mnt_drop_write_file(filp);
2285 
2286 	trace_f2fs_shutdown(sbi, in, ret);
2287 
2288 	return ret;
2289 }
2290 
2291 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2292 {
2293 	struct inode *inode = file_inode(filp);
2294 	struct super_block *sb = inode->i_sb;
2295 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
2296 	struct fstrim_range range;
2297 	int ret;
2298 
2299 	if (!capable(CAP_SYS_ADMIN))
2300 		return -EPERM;
2301 
2302 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2303 		return -EOPNOTSUPP;
2304 
2305 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2306 				sizeof(range)))
2307 		return -EFAULT;
2308 
2309 	ret = mnt_want_write_file(filp);
2310 	if (ret)
2311 		return ret;
2312 
2313 	range.minlen = max((unsigned int)range.minlen,
2314 				q->limits.discard_granularity);
2315 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2316 	mnt_drop_write_file(filp);
2317 	if (ret < 0)
2318 		return ret;
2319 
2320 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2321 				sizeof(range)))
2322 		return -EFAULT;
2323 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2324 	return 0;
2325 }
2326 
2327 static bool uuid_is_nonzero(__u8 u[16])
2328 {
2329 	int i;
2330 
2331 	for (i = 0; i < 16; i++)
2332 		if (u[i])
2333 			return true;
2334 	return false;
2335 }
2336 
2337 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2338 {
2339 	struct inode *inode = file_inode(filp);
2340 
2341 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2342 		return -EOPNOTSUPP;
2343 
2344 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2345 
2346 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2347 }
2348 
2349 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2350 {
2351 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2352 		return -EOPNOTSUPP;
2353 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2354 }
2355 
2356 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2357 {
2358 	struct inode *inode = file_inode(filp);
2359 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2360 	int err;
2361 
2362 	if (!f2fs_sb_has_encrypt(sbi))
2363 		return -EOPNOTSUPP;
2364 
2365 	err = mnt_want_write_file(filp);
2366 	if (err)
2367 		return err;
2368 
2369 	down_write(&sbi->sb_lock);
2370 
2371 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2372 		goto got_it;
2373 
2374 	/* update superblock with uuid */
2375 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2376 
2377 	err = f2fs_commit_super(sbi, false);
2378 	if (err) {
2379 		/* undo new data */
2380 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2381 		goto out_err;
2382 	}
2383 got_it:
2384 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2385 									16))
2386 		err = -EFAULT;
2387 out_err:
2388 	up_write(&sbi->sb_lock);
2389 	mnt_drop_write_file(filp);
2390 	return err;
2391 }
2392 
2393 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2394 					     unsigned long arg)
2395 {
2396 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2397 		return -EOPNOTSUPP;
2398 
2399 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2400 }
2401 
2402 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2403 {
2404 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2405 		return -EOPNOTSUPP;
2406 
2407 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2408 }
2409 
2410 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2411 {
2412 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2413 		return -EOPNOTSUPP;
2414 
2415 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2416 }
2417 
2418 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2419 						    unsigned long arg)
2420 {
2421 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2422 		return -EOPNOTSUPP;
2423 
2424 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2425 }
2426 
2427 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2428 					      unsigned long arg)
2429 {
2430 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2431 		return -EOPNOTSUPP;
2432 
2433 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2434 }
2435 
2436 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2437 {
2438 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2439 		return -EOPNOTSUPP;
2440 
2441 	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2442 }
2443 
2444 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2445 {
2446 	struct inode *inode = file_inode(filp);
2447 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2448 	__u32 sync;
2449 	int ret;
2450 
2451 	if (!capable(CAP_SYS_ADMIN))
2452 		return -EPERM;
2453 
2454 	if (get_user(sync, (__u32 __user *)arg))
2455 		return -EFAULT;
2456 
2457 	if (f2fs_readonly(sbi->sb))
2458 		return -EROFS;
2459 
2460 	ret = mnt_want_write_file(filp);
2461 	if (ret)
2462 		return ret;
2463 
2464 	if (!sync) {
2465 		if (!down_write_trylock(&sbi->gc_lock)) {
2466 			ret = -EBUSY;
2467 			goto out;
2468 		}
2469 	} else {
2470 		down_write(&sbi->gc_lock);
2471 	}
2472 
2473 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2474 out:
2475 	mnt_drop_write_file(filp);
2476 	return ret;
2477 }
2478 
2479 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2480 {
2481 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2482 	u64 end;
2483 	int ret;
2484 
2485 	if (!capable(CAP_SYS_ADMIN))
2486 		return -EPERM;
2487 	if (f2fs_readonly(sbi->sb))
2488 		return -EROFS;
2489 
2490 	end = range->start + range->len;
2491 	if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2492 					end >= MAX_BLKADDR(sbi))
2493 		return -EINVAL;
2494 
2495 	ret = mnt_want_write_file(filp);
2496 	if (ret)
2497 		return ret;
2498 
2499 do_more:
2500 	if (!range->sync) {
2501 		if (!down_write_trylock(&sbi->gc_lock)) {
2502 			ret = -EBUSY;
2503 			goto out;
2504 		}
2505 	} else {
2506 		down_write(&sbi->gc_lock);
2507 	}
2508 
2509 	ret = f2fs_gc(sbi, range->sync, true, GET_SEGNO(sbi, range->start));
2510 	if (ret) {
2511 		if (ret == -EBUSY)
2512 			ret = -EAGAIN;
2513 		goto out;
2514 	}
2515 	range->start += BLKS_PER_SEC(sbi);
2516 	if (range->start <= end)
2517 		goto do_more;
2518 out:
2519 	mnt_drop_write_file(filp);
2520 	return ret;
2521 }
2522 
2523 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2524 {
2525 	struct f2fs_gc_range range;
2526 
2527 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2528 							sizeof(range)))
2529 		return -EFAULT;
2530 	return __f2fs_ioc_gc_range(filp, &range);
2531 }
2532 
2533 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2534 {
2535 	struct inode *inode = file_inode(filp);
2536 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2537 	int ret;
2538 
2539 	if (!capable(CAP_SYS_ADMIN))
2540 		return -EPERM;
2541 
2542 	if (f2fs_readonly(sbi->sb))
2543 		return -EROFS;
2544 
2545 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2546 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2547 		return -EINVAL;
2548 	}
2549 
2550 	ret = mnt_want_write_file(filp);
2551 	if (ret)
2552 		return ret;
2553 
2554 	ret = f2fs_sync_fs(sbi->sb, 1);
2555 
2556 	mnt_drop_write_file(filp);
2557 	return ret;
2558 }
2559 
2560 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2561 					struct file *filp,
2562 					struct f2fs_defragment *range)
2563 {
2564 	struct inode *inode = file_inode(filp);
2565 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2566 					.m_seg_type = NO_CHECK_TYPE ,
2567 					.m_may_create = false };
2568 	struct extent_info ei = {0, 0, 0};
2569 	pgoff_t pg_start, pg_end, next_pgofs;
2570 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2571 	unsigned int total = 0, sec_num;
2572 	block_t blk_end = 0;
2573 	bool fragmented = false;
2574 	int err;
2575 
2576 	/* if in-place-update policy is enabled, don't waste time here */
2577 	if (f2fs_should_update_inplace(inode, NULL))
2578 		return -EINVAL;
2579 
2580 	pg_start = range->start >> PAGE_SHIFT;
2581 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2582 
2583 	f2fs_balance_fs(sbi, true);
2584 
2585 	inode_lock(inode);
2586 
2587 	/* writeback all dirty pages in the range */
2588 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2589 						range->start + range->len - 1);
2590 	if (err)
2591 		goto out;
2592 
2593 	/*
2594 	 * lookup mapping info in extent cache, skip defragmenting if physical
2595 	 * block addresses are continuous.
2596 	 */
2597 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2598 		if (ei.fofs + ei.len >= pg_end)
2599 			goto out;
2600 	}
2601 
2602 	map.m_lblk = pg_start;
2603 	map.m_next_pgofs = &next_pgofs;
2604 
2605 	/*
2606 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2607 	 * physical block addresses are continuous even if there are hole(s)
2608 	 * in logical blocks.
2609 	 */
2610 	while (map.m_lblk < pg_end) {
2611 		map.m_len = pg_end - map.m_lblk;
2612 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2613 		if (err)
2614 			goto out;
2615 
2616 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2617 			map.m_lblk = next_pgofs;
2618 			continue;
2619 		}
2620 
2621 		if (blk_end && blk_end != map.m_pblk)
2622 			fragmented = true;
2623 
2624 		/* record total count of block that we're going to move */
2625 		total += map.m_len;
2626 
2627 		blk_end = map.m_pblk + map.m_len;
2628 
2629 		map.m_lblk += map.m_len;
2630 	}
2631 
2632 	if (!fragmented) {
2633 		total = 0;
2634 		goto out;
2635 	}
2636 
2637 	sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2638 
2639 	/*
2640 	 * make sure there are enough free section for LFS allocation, this can
2641 	 * avoid defragment running in SSR mode when free section are allocated
2642 	 * intensively
2643 	 */
2644 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2645 		err = -EAGAIN;
2646 		goto out;
2647 	}
2648 
2649 	map.m_lblk = pg_start;
2650 	map.m_len = pg_end - pg_start;
2651 	total = 0;
2652 
2653 	while (map.m_lblk < pg_end) {
2654 		pgoff_t idx;
2655 		int cnt = 0;
2656 
2657 do_map:
2658 		map.m_len = pg_end - map.m_lblk;
2659 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2660 		if (err)
2661 			goto clear_out;
2662 
2663 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2664 			map.m_lblk = next_pgofs;
2665 			goto check;
2666 		}
2667 
2668 		set_inode_flag(inode, FI_DO_DEFRAG);
2669 
2670 		idx = map.m_lblk;
2671 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2672 			struct page *page;
2673 
2674 			page = f2fs_get_lock_data_page(inode, idx, true);
2675 			if (IS_ERR(page)) {
2676 				err = PTR_ERR(page);
2677 				goto clear_out;
2678 			}
2679 
2680 			set_page_dirty(page);
2681 			f2fs_put_page(page, 1);
2682 
2683 			idx++;
2684 			cnt++;
2685 			total++;
2686 		}
2687 
2688 		map.m_lblk = idx;
2689 check:
2690 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
2691 			goto do_map;
2692 
2693 		clear_inode_flag(inode, FI_DO_DEFRAG);
2694 
2695 		err = filemap_fdatawrite(inode->i_mapping);
2696 		if (err)
2697 			goto out;
2698 	}
2699 clear_out:
2700 	clear_inode_flag(inode, FI_DO_DEFRAG);
2701 out:
2702 	inode_unlock(inode);
2703 	if (!err)
2704 		range->len = (u64)total << PAGE_SHIFT;
2705 	return err;
2706 }
2707 
2708 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2709 {
2710 	struct inode *inode = file_inode(filp);
2711 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2712 	struct f2fs_defragment range;
2713 	int err;
2714 
2715 	if (!capable(CAP_SYS_ADMIN))
2716 		return -EPERM;
2717 
2718 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2719 		return -EINVAL;
2720 
2721 	if (f2fs_readonly(sbi->sb))
2722 		return -EROFS;
2723 
2724 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2725 							sizeof(range)))
2726 		return -EFAULT;
2727 
2728 	/* verify alignment of offset & size */
2729 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2730 		return -EINVAL;
2731 
2732 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2733 					sbi->max_file_blocks))
2734 		return -EINVAL;
2735 
2736 	err = mnt_want_write_file(filp);
2737 	if (err)
2738 		return err;
2739 
2740 	err = f2fs_defragment_range(sbi, filp, &range);
2741 	mnt_drop_write_file(filp);
2742 
2743 	f2fs_update_time(sbi, REQ_TIME);
2744 	if (err < 0)
2745 		return err;
2746 
2747 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2748 							sizeof(range)))
2749 		return -EFAULT;
2750 
2751 	return 0;
2752 }
2753 
2754 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2755 			struct file *file_out, loff_t pos_out, size_t len)
2756 {
2757 	struct inode *src = file_inode(file_in);
2758 	struct inode *dst = file_inode(file_out);
2759 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2760 	size_t olen = len, dst_max_i_size = 0;
2761 	size_t dst_osize;
2762 	int ret;
2763 
2764 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2765 				src->i_sb != dst->i_sb)
2766 		return -EXDEV;
2767 
2768 	if (unlikely(f2fs_readonly(src->i_sb)))
2769 		return -EROFS;
2770 
2771 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2772 		return -EINVAL;
2773 
2774 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2775 		return -EOPNOTSUPP;
2776 
2777 	if (pos_out < 0 || pos_in < 0)
2778 		return -EINVAL;
2779 
2780 	if (src == dst) {
2781 		if (pos_in == pos_out)
2782 			return 0;
2783 		if (pos_out > pos_in && pos_out < pos_in + len)
2784 			return -EINVAL;
2785 	}
2786 
2787 	inode_lock(src);
2788 	if (src != dst) {
2789 		ret = -EBUSY;
2790 		if (!inode_trylock(dst))
2791 			goto out;
2792 	}
2793 
2794 	ret = -EINVAL;
2795 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2796 		goto out_unlock;
2797 	if (len == 0)
2798 		olen = len = src->i_size - pos_in;
2799 	if (pos_in + len == src->i_size)
2800 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2801 	if (len == 0) {
2802 		ret = 0;
2803 		goto out_unlock;
2804 	}
2805 
2806 	dst_osize = dst->i_size;
2807 	if (pos_out + olen > dst->i_size)
2808 		dst_max_i_size = pos_out + olen;
2809 
2810 	/* verify the end result is block aligned */
2811 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2812 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2813 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2814 		goto out_unlock;
2815 
2816 	ret = f2fs_convert_inline_inode(src);
2817 	if (ret)
2818 		goto out_unlock;
2819 
2820 	ret = f2fs_convert_inline_inode(dst);
2821 	if (ret)
2822 		goto out_unlock;
2823 
2824 	/* write out all dirty pages from offset */
2825 	ret = filemap_write_and_wait_range(src->i_mapping,
2826 					pos_in, pos_in + len);
2827 	if (ret)
2828 		goto out_unlock;
2829 
2830 	ret = filemap_write_and_wait_range(dst->i_mapping,
2831 					pos_out, pos_out + len);
2832 	if (ret)
2833 		goto out_unlock;
2834 
2835 	f2fs_balance_fs(sbi, true);
2836 
2837 	down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2838 	if (src != dst) {
2839 		ret = -EBUSY;
2840 		if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2841 			goto out_src;
2842 	}
2843 
2844 	f2fs_lock_op(sbi);
2845 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2846 				pos_out >> F2FS_BLKSIZE_BITS,
2847 				len >> F2FS_BLKSIZE_BITS, false);
2848 
2849 	if (!ret) {
2850 		if (dst_max_i_size)
2851 			f2fs_i_size_write(dst, dst_max_i_size);
2852 		else if (dst_osize != dst->i_size)
2853 			f2fs_i_size_write(dst, dst_osize);
2854 	}
2855 	f2fs_unlock_op(sbi);
2856 
2857 	if (src != dst)
2858 		up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2859 out_src:
2860 	up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2861 out_unlock:
2862 	if (src != dst)
2863 		inode_unlock(dst);
2864 out:
2865 	inode_unlock(src);
2866 	return ret;
2867 }
2868 
2869 static int __f2fs_ioc_move_range(struct file *filp,
2870 				struct f2fs_move_range *range)
2871 {
2872 	struct fd dst;
2873 	int err;
2874 
2875 	if (!(filp->f_mode & FMODE_READ) ||
2876 			!(filp->f_mode & FMODE_WRITE))
2877 		return -EBADF;
2878 
2879 	dst = fdget(range->dst_fd);
2880 	if (!dst.file)
2881 		return -EBADF;
2882 
2883 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2884 		err = -EBADF;
2885 		goto err_out;
2886 	}
2887 
2888 	err = mnt_want_write_file(filp);
2889 	if (err)
2890 		goto err_out;
2891 
2892 	err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2893 					range->pos_out, range->len);
2894 
2895 	mnt_drop_write_file(filp);
2896 err_out:
2897 	fdput(dst);
2898 	return err;
2899 }
2900 
2901 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2902 {
2903 	struct f2fs_move_range range;
2904 
2905 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2906 							sizeof(range)))
2907 		return -EFAULT;
2908 	return __f2fs_ioc_move_range(filp, &range);
2909 }
2910 
2911 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2912 {
2913 	struct inode *inode = file_inode(filp);
2914 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2915 	struct sit_info *sm = SIT_I(sbi);
2916 	unsigned int start_segno = 0, end_segno = 0;
2917 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2918 	struct f2fs_flush_device range;
2919 	int ret;
2920 
2921 	if (!capable(CAP_SYS_ADMIN))
2922 		return -EPERM;
2923 
2924 	if (f2fs_readonly(sbi->sb))
2925 		return -EROFS;
2926 
2927 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2928 		return -EINVAL;
2929 
2930 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2931 							sizeof(range)))
2932 		return -EFAULT;
2933 
2934 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2935 			__is_large_section(sbi)) {
2936 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2937 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2938 		return -EINVAL;
2939 	}
2940 
2941 	ret = mnt_want_write_file(filp);
2942 	if (ret)
2943 		return ret;
2944 
2945 	if (range.dev_num != 0)
2946 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2947 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2948 
2949 	start_segno = sm->last_victim[FLUSH_DEVICE];
2950 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2951 		start_segno = dev_start_segno;
2952 	end_segno = min(start_segno + range.segments, dev_end_segno);
2953 
2954 	while (start_segno < end_segno) {
2955 		if (!down_write_trylock(&sbi->gc_lock)) {
2956 			ret = -EBUSY;
2957 			goto out;
2958 		}
2959 		sm->last_victim[GC_CB] = end_segno + 1;
2960 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2961 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2962 		ret = f2fs_gc(sbi, true, true, start_segno);
2963 		if (ret == -EAGAIN)
2964 			ret = 0;
2965 		else if (ret < 0)
2966 			break;
2967 		start_segno++;
2968 	}
2969 out:
2970 	mnt_drop_write_file(filp);
2971 	return ret;
2972 }
2973 
2974 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2975 {
2976 	struct inode *inode = file_inode(filp);
2977 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2978 
2979 	/* Must validate to set it with SQLite behavior in Android. */
2980 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2981 
2982 	return put_user(sb_feature, (u32 __user *)arg);
2983 }
2984 
2985 #ifdef CONFIG_QUOTA
2986 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2987 {
2988 	struct dquot *transfer_to[MAXQUOTAS] = {};
2989 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2990 	struct super_block *sb = sbi->sb;
2991 	int err = 0;
2992 
2993 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2994 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2995 		err = __dquot_transfer(inode, transfer_to);
2996 		if (err)
2997 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2998 		dqput(transfer_to[PRJQUOTA]);
2999 	}
3000 	return err;
3001 }
3002 
3003 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3004 {
3005 	struct inode *inode = file_inode(filp);
3006 	struct f2fs_inode_info *fi = F2FS_I(inode);
3007 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3008 	struct page *ipage;
3009 	kprojid_t kprojid;
3010 	int err;
3011 
3012 	if (!f2fs_sb_has_project_quota(sbi)) {
3013 		if (projid != F2FS_DEF_PROJID)
3014 			return -EOPNOTSUPP;
3015 		else
3016 			return 0;
3017 	}
3018 
3019 	if (!f2fs_has_extra_attr(inode))
3020 		return -EOPNOTSUPP;
3021 
3022 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3023 
3024 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
3025 		return 0;
3026 
3027 	err = -EPERM;
3028 	/* Is it quota file? Do not allow user to mess with it */
3029 	if (IS_NOQUOTA(inode))
3030 		return err;
3031 
3032 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
3033 	if (IS_ERR(ipage))
3034 		return PTR_ERR(ipage);
3035 
3036 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
3037 								i_projid)) {
3038 		err = -EOVERFLOW;
3039 		f2fs_put_page(ipage, 1);
3040 		return err;
3041 	}
3042 	f2fs_put_page(ipage, 1);
3043 
3044 	err = dquot_initialize(inode);
3045 	if (err)
3046 		return err;
3047 
3048 	f2fs_lock_op(sbi);
3049 	err = f2fs_transfer_project_quota(inode, kprojid);
3050 	if (err)
3051 		goto out_unlock;
3052 
3053 	F2FS_I(inode)->i_projid = kprojid;
3054 	inode->i_ctime = current_time(inode);
3055 	f2fs_mark_inode_dirty_sync(inode, true);
3056 out_unlock:
3057 	f2fs_unlock_op(sbi);
3058 	return err;
3059 }
3060 #else
3061 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3062 {
3063 	return 0;
3064 }
3065 
3066 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3067 {
3068 	if (projid != F2FS_DEF_PROJID)
3069 		return -EOPNOTSUPP;
3070 	return 0;
3071 }
3072 #endif
3073 
3074 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
3075 
3076 /*
3077  * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
3078  * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
3079  * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
3080  */
3081 
3082 static const struct {
3083 	u32 iflag;
3084 	u32 xflag;
3085 } f2fs_xflags_map[] = {
3086 	{ F2FS_SYNC_FL,		FS_XFLAG_SYNC },
3087 	{ F2FS_IMMUTABLE_FL,	FS_XFLAG_IMMUTABLE },
3088 	{ F2FS_APPEND_FL,	FS_XFLAG_APPEND },
3089 	{ F2FS_NODUMP_FL,	FS_XFLAG_NODUMP },
3090 	{ F2FS_NOATIME_FL,	FS_XFLAG_NOATIME },
3091 	{ F2FS_PROJINHERIT_FL,	FS_XFLAG_PROJINHERIT },
3092 };
3093 
3094 #define F2FS_SUPPORTED_XFLAGS (		\
3095 		FS_XFLAG_SYNC |		\
3096 		FS_XFLAG_IMMUTABLE |	\
3097 		FS_XFLAG_APPEND |	\
3098 		FS_XFLAG_NODUMP |	\
3099 		FS_XFLAG_NOATIME |	\
3100 		FS_XFLAG_PROJINHERIT)
3101 
3102 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
3103 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
3104 {
3105 	u32 xflags = 0;
3106 	int i;
3107 
3108 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3109 		if (iflags & f2fs_xflags_map[i].iflag)
3110 			xflags |= f2fs_xflags_map[i].xflag;
3111 
3112 	return xflags;
3113 }
3114 
3115 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
3116 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
3117 {
3118 	u32 iflags = 0;
3119 	int i;
3120 
3121 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3122 		if (xflags & f2fs_xflags_map[i].xflag)
3123 			iflags |= f2fs_xflags_map[i].iflag;
3124 
3125 	return iflags;
3126 }
3127 
3128 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
3129 {
3130 	struct f2fs_inode_info *fi = F2FS_I(inode);
3131 
3132 	simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
3133 
3134 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3135 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3136 }
3137 
3138 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
3139 {
3140 	struct inode *inode = file_inode(filp);
3141 	struct fsxattr fa;
3142 
3143 	f2fs_fill_fsxattr(inode, &fa);
3144 
3145 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
3146 		return -EFAULT;
3147 	return 0;
3148 }
3149 
3150 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
3151 {
3152 	struct inode *inode = file_inode(filp);
3153 	struct fsxattr fa, old_fa;
3154 	u32 iflags;
3155 	int err;
3156 
3157 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
3158 		return -EFAULT;
3159 
3160 	/* Make sure caller has proper permission */
3161 	if (!inode_owner_or_capable(inode))
3162 		return -EACCES;
3163 
3164 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
3165 		return -EOPNOTSUPP;
3166 
3167 	iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
3168 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3169 		return -EOPNOTSUPP;
3170 
3171 	err = mnt_want_write_file(filp);
3172 	if (err)
3173 		return err;
3174 
3175 	inode_lock(inode);
3176 
3177 	f2fs_fill_fsxattr(inode, &old_fa);
3178 	err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
3179 	if (err)
3180 		goto out;
3181 
3182 	err = f2fs_setflags_common(inode, iflags,
3183 			f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
3184 	if (err)
3185 		goto out;
3186 
3187 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3188 out:
3189 	inode_unlock(inode);
3190 	mnt_drop_write_file(filp);
3191 	return err;
3192 }
3193 
3194 int f2fs_pin_file_control(struct inode *inode, bool inc)
3195 {
3196 	struct f2fs_inode_info *fi = F2FS_I(inode);
3197 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3198 
3199 	/* Use i_gc_failures for normal file as a risk signal. */
3200 	if (inc)
3201 		f2fs_i_gc_failures_write(inode,
3202 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3203 
3204 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3205 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3206 			  __func__, inode->i_ino,
3207 			  fi->i_gc_failures[GC_FAILURE_PIN]);
3208 		clear_inode_flag(inode, FI_PIN_FILE);
3209 		return -EAGAIN;
3210 	}
3211 	return 0;
3212 }
3213 
3214 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3215 {
3216 	struct inode *inode = file_inode(filp);
3217 	__u32 pin;
3218 	int ret = 0;
3219 
3220 	if (get_user(pin, (__u32 __user *)arg))
3221 		return -EFAULT;
3222 
3223 	if (!S_ISREG(inode->i_mode))
3224 		return -EINVAL;
3225 
3226 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3227 		return -EROFS;
3228 
3229 	ret = mnt_want_write_file(filp);
3230 	if (ret)
3231 		return ret;
3232 
3233 	inode_lock(inode);
3234 
3235 	if (f2fs_should_update_outplace(inode, NULL)) {
3236 		ret = -EINVAL;
3237 		goto out;
3238 	}
3239 
3240 	if (!pin) {
3241 		clear_inode_flag(inode, FI_PIN_FILE);
3242 		f2fs_i_gc_failures_write(inode, 0);
3243 		goto done;
3244 	}
3245 
3246 	if (f2fs_pin_file_control(inode, false)) {
3247 		ret = -EAGAIN;
3248 		goto out;
3249 	}
3250 
3251 	ret = f2fs_convert_inline_inode(inode);
3252 	if (ret)
3253 		goto out;
3254 
3255 	if (!f2fs_disable_compressed_file(inode)) {
3256 		ret = -EOPNOTSUPP;
3257 		goto out;
3258 	}
3259 
3260 	set_inode_flag(inode, FI_PIN_FILE);
3261 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3262 done:
3263 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3264 out:
3265 	inode_unlock(inode);
3266 	mnt_drop_write_file(filp);
3267 	return ret;
3268 }
3269 
3270 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3271 {
3272 	struct inode *inode = file_inode(filp);
3273 	__u32 pin = 0;
3274 
3275 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3276 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3277 	return put_user(pin, (u32 __user *)arg);
3278 }
3279 
3280 int f2fs_precache_extents(struct inode *inode)
3281 {
3282 	struct f2fs_inode_info *fi = F2FS_I(inode);
3283 	struct f2fs_map_blocks map;
3284 	pgoff_t m_next_extent;
3285 	loff_t end;
3286 	int err;
3287 
3288 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3289 		return -EOPNOTSUPP;
3290 
3291 	map.m_lblk = 0;
3292 	map.m_next_pgofs = NULL;
3293 	map.m_next_extent = &m_next_extent;
3294 	map.m_seg_type = NO_CHECK_TYPE;
3295 	map.m_may_create = false;
3296 	end = F2FS_I_SB(inode)->max_file_blocks;
3297 
3298 	while (map.m_lblk < end) {
3299 		map.m_len = end - map.m_lblk;
3300 
3301 		down_write(&fi->i_gc_rwsem[WRITE]);
3302 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3303 		up_write(&fi->i_gc_rwsem[WRITE]);
3304 		if (err)
3305 			return err;
3306 
3307 		map.m_lblk = m_next_extent;
3308 	}
3309 
3310 	return err;
3311 }
3312 
3313 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3314 {
3315 	return f2fs_precache_extents(file_inode(filp));
3316 }
3317 
3318 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3319 {
3320 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3321 	__u64 block_count;
3322 
3323 	if (!capable(CAP_SYS_ADMIN))
3324 		return -EPERM;
3325 
3326 	if (f2fs_readonly(sbi->sb))
3327 		return -EROFS;
3328 
3329 	if (copy_from_user(&block_count, (void __user *)arg,
3330 			   sizeof(block_count)))
3331 		return -EFAULT;
3332 
3333 	return f2fs_resize_fs(sbi, block_count);
3334 }
3335 
3336 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3337 {
3338 	struct inode *inode = file_inode(filp);
3339 
3340 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3341 
3342 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3343 		f2fs_warn(F2FS_I_SB(inode),
3344 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3345 			  inode->i_ino);
3346 		return -EOPNOTSUPP;
3347 	}
3348 
3349 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3350 }
3351 
3352 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3353 {
3354 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3355 		return -EOPNOTSUPP;
3356 
3357 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3358 }
3359 
3360 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3361 {
3362 	struct inode *inode = file_inode(filp);
3363 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3364 	char *vbuf;
3365 	int count;
3366 	int err = 0;
3367 
3368 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3369 	if (!vbuf)
3370 		return -ENOMEM;
3371 
3372 	down_read(&sbi->sb_lock);
3373 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3374 			ARRAY_SIZE(sbi->raw_super->volume_name),
3375 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3376 	up_read(&sbi->sb_lock);
3377 
3378 	if (copy_to_user((char __user *)arg, vbuf,
3379 				min(FSLABEL_MAX, count)))
3380 		err = -EFAULT;
3381 
3382 	kfree(vbuf);
3383 	return err;
3384 }
3385 
3386 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3387 {
3388 	struct inode *inode = file_inode(filp);
3389 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3390 	char *vbuf;
3391 	int err = 0;
3392 
3393 	if (!capable(CAP_SYS_ADMIN))
3394 		return -EPERM;
3395 
3396 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3397 	if (IS_ERR(vbuf))
3398 		return PTR_ERR(vbuf);
3399 
3400 	err = mnt_want_write_file(filp);
3401 	if (err)
3402 		goto out;
3403 
3404 	down_write(&sbi->sb_lock);
3405 
3406 	memset(sbi->raw_super->volume_name, 0,
3407 			sizeof(sbi->raw_super->volume_name));
3408 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3409 			sbi->raw_super->volume_name,
3410 			ARRAY_SIZE(sbi->raw_super->volume_name));
3411 
3412 	err = f2fs_commit_super(sbi, false);
3413 
3414 	up_write(&sbi->sb_lock);
3415 
3416 	mnt_drop_write_file(filp);
3417 out:
3418 	kfree(vbuf);
3419 	return err;
3420 }
3421 
3422 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3423 {
3424 	struct inode *inode = file_inode(filp);
3425 	__u64 blocks;
3426 
3427 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3428 		return -EOPNOTSUPP;
3429 
3430 	if (!f2fs_compressed_file(inode))
3431 		return -EINVAL;
3432 
3433 	blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3434 	return put_user(blocks, (u64 __user *)arg);
3435 }
3436 
3437 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3438 {
3439 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3440 	unsigned int released_blocks = 0;
3441 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3442 	block_t blkaddr;
3443 	int i;
3444 
3445 	for (i = 0; i < count; i++) {
3446 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3447 						dn->ofs_in_node + i);
3448 
3449 		if (!__is_valid_data_blkaddr(blkaddr))
3450 			continue;
3451 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3452 					DATA_GENERIC_ENHANCE)))
3453 			return -EFSCORRUPTED;
3454 	}
3455 
3456 	while (count) {
3457 		int compr_blocks = 0;
3458 
3459 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3460 			blkaddr = f2fs_data_blkaddr(dn);
3461 
3462 			if (i == 0) {
3463 				if (blkaddr == COMPRESS_ADDR)
3464 					continue;
3465 				dn->ofs_in_node += cluster_size;
3466 				goto next;
3467 			}
3468 
3469 			if (__is_valid_data_blkaddr(blkaddr))
3470 				compr_blocks++;
3471 
3472 			if (blkaddr != NEW_ADDR)
3473 				continue;
3474 
3475 			dn->data_blkaddr = NULL_ADDR;
3476 			f2fs_set_data_blkaddr(dn);
3477 		}
3478 
3479 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3480 		dec_valid_block_count(sbi, dn->inode,
3481 					cluster_size - compr_blocks);
3482 
3483 		released_blocks += cluster_size - compr_blocks;
3484 next:
3485 		count -= cluster_size;
3486 	}
3487 
3488 	return released_blocks;
3489 }
3490 
3491 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3492 {
3493 	struct inode *inode = file_inode(filp);
3494 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3495 	pgoff_t page_idx = 0, last_idx;
3496 	unsigned int released_blocks = 0;
3497 	int ret;
3498 	int writecount;
3499 
3500 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3501 		return -EOPNOTSUPP;
3502 
3503 	if (!f2fs_compressed_file(inode))
3504 		return -EINVAL;
3505 
3506 	if (f2fs_readonly(sbi->sb))
3507 		return -EROFS;
3508 
3509 	ret = mnt_want_write_file(filp);
3510 	if (ret)
3511 		return ret;
3512 
3513 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3514 
3515 	inode_lock(inode);
3516 
3517 	writecount = atomic_read(&inode->i_writecount);
3518 	if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3519 			(!(filp->f_mode & FMODE_WRITE) && writecount)) {
3520 		ret = -EBUSY;
3521 		goto out;
3522 	}
3523 
3524 	if (IS_IMMUTABLE(inode)) {
3525 		ret = -EINVAL;
3526 		goto out;
3527 	}
3528 
3529 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3530 	if (ret)
3531 		goto out;
3532 
3533 	F2FS_I(inode)->i_flags |= F2FS_IMMUTABLE_FL;
3534 	f2fs_set_inode_flags(inode);
3535 	inode->i_ctime = current_time(inode);
3536 	f2fs_mark_inode_dirty_sync(inode, true);
3537 
3538 	if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3539 		goto out;
3540 
3541 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3542 	down_write(&F2FS_I(inode)->i_mmap_sem);
3543 
3544 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3545 
3546 	while (page_idx < last_idx) {
3547 		struct dnode_of_data dn;
3548 		pgoff_t end_offset, count;
3549 
3550 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3551 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3552 		if (ret) {
3553 			if (ret == -ENOENT) {
3554 				page_idx = f2fs_get_next_page_offset(&dn,
3555 								page_idx);
3556 				ret = 0;
3557 				continue;
3558 			}
3559 			break;
3560 		}
3561 
3562 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3563 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3564 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3565 
3566 		ret = release_compress_blocks(&dn, count);
3567 
3568 		f2fs_put_dnode(&dn);
3569 
3570 		if (ret < 0)
3571 			break;
3572 
3573 		page_idx += count;
3574 		released_blocks += ret;
3575 	}
3576 
3577 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3578 	up_write(&F2FS_I(inode)->i_mmap_sem);
3579 out:
3580 	inode_unlock(inode);
3581 
3582 	mnt_drop_write_file(filp);
3583 
3584 	if (ret >= 0) {
3585 		ret = put_user(released_blocks, (u64 __user *)arg);
3586 	} else if (released_blocks &&
3587 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3588 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3589 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3590 			"iblocks=%llu, released=%u, compr_blocks=%u, "
3591 			"run fsck to fix.",
3592 			__func__, inode->i_ino, inode->i_blocks,
3593 			released_blocks,
3594 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
3595 	}
3596 
3597 	return ret;
3598 }
3599 
3600 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3601 {
3602 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3603 	unsigned int reserved_blocks = 0;
3604 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3605 	block_t blkaddr;
3606 	int i;
3607 
3608 	for (i = 0; i < count; i++) {
3609 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3610 						dn->ofs_in_node + i);
3611 
3612 		if (!__is_valid_data_blkaddr(blkaddr))
3613 			continue;
3614 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3615 					DATA_GENERIC_ENHANCE)))
3616 			return -EFSCORRUPTED;
3617 	}
3618 
3619 	while (count) {
3620 		int compr_blocks = 0;
3621 		blkcnt_t reserved;
3622 		int ret;
3623 
3624 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3625 			blkaddr = f2fs_data_blkaddr(dn);
3626 
3627 			if (i == 0) {
3628 				if (blkaddr == COMPRESS_ADDR)
3629 					continue;
3630 				dn->ofs_in_node += cluster_size;
3631 				goto next;
3632 			}
3633 
3634 			if (__is_valid_data_blkaddr(blkaddr)) {
3635 				compr_blocks++;
3636 				continue;
3637 			}
3638 
3639 			dn->data_blkaddr = NEW_ADDR;
3640 			f2fs_set_data_blkaddr(dn);
3641 		}
3642 
3643 		reserved = cluster_size - compr_blocks;
3644 		ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3645 		if (ret)
3646 			return ret;
3647 
3648 		if (reserved != cluster_size - compr_blocks)
3649 			return -ENOSPC;
3650 
3651 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3652 
3653 		reserved_blocks += reserved;
3654 next:
3655 		count -= cluster_size;
3656 	}
3657 
3658 	return reserved_blocks;
3659 }
3660 
3661 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3662 {
3663 	struct inode *inode = file_inode(filp);
3664 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3665 	pgoff_t page_idx = 0, last_idx;
3666 	unsigned int reserved_blocks = 0;
3667 	int ret;
3668 
3669 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3670 		return -EOPNOTSUPP;
3671 
3672 	if (!f2fs_compressed_file(inode))
3673 		return -EINVAL;
3674 
3675 	if (f2fs_readonly(sbi->sb))
3676 		return -EROFS;
3677 
3678 	ret = mnt_want_write_file(filp);
3679 	if (ret)
3680 		return ret;
3681 
3682 	if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3683 		goto out;
3684 
3685 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3686 
3687 	inode_lock(inode);
3688 
3689 	if (!IS_IMMUTABLE(inode)) {
3690 		ret = -EINVAL;
3691 		goto unlock_inode;
3692 	}
3693 
3694 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3695 	down_write(&F2FS_I(inode)->i_mmap_sem);
3696 
3697 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3698 
3699 	while (page_idx < last_idx) {
3700 		struct dnode_of_data dn;
3701 		pgoff_t end_offset, count;
3702 
3703 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3704 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3705 		if (ret) {
3706 			if (ret == -ENOENT) {
3707 				page_idx = f2fs_get_next_page_offset(&dn,
3708 								page_idx);
3709 				ret = 0;
3710 				continue;
3711 			}
3712 			break;
3713 		}
3714 
3715 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3716 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3717 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3718 
3719 		ret = reserve_compress_blocks(&dn, count);
3720 
3721 		f2fs_put_dnode(&dn);
3722 
3723 		if (ret < 0)
3724 			break;
3725 
3726 		page_idx += count;
3727 		reserved_blocks += ret;
3728 	}
3729 
3730 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3731 	up_write(&F2FS_I(inode)->i_mmap_sem);
3732 
3733 	if (ret >= 0) {
3734 		F2FS_I(inode)->i_flags &= ~F2FS_IMMUTABLE_FL;
3735 		f2fs_set_inode_flags(inode);
3736 		inode->i_ctime = current_time(inode);
3737 		f2fs_mark_inode_dirty_sync(inode, true);
3738 	}
3739 unlock_inode:
3740 	inode_unlock(inode);
3741 out:
3742 	mnt_drop_write_file(filp);
3743 
3744 	if (ret >= 0) {
3745 		ret = put_user(reserved_blocks, (u64 __user *)arg);
3746 	} else if (reserved_blocks &&
3747 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3748 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3749 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3750 			"iblocks=%llu, reserved=%u, compr_blocks=%u, "
3751 			"run fsck to fix.",
3752 			__func__, inode->i_ino, inode->i_blocks,
3753 			reserved_blocks,
3754 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
3755 	}
3756 
3757 	return ret;
3758 }
3759 
3760 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3761 		pgoff_t off, block_t block, block_t len, u32 flags)
3762 {
3763 	struct request_queue *q = bdev_get_queue(bdev);
3764 	sector_t sector = SECTOR_FROM_BLOCK(block);
3765 	sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3766 	int ret = 0;
3767 
3768 	if (!q)
3769 		return -ENXIO;
3770 
3771 	if (flags & F2FS_TRIM_FILE_DISCARD)
3772 		ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3773 						blk_queue_secure_erase(q) ?
3774 						BLKDEV_DISCARD_SECURE : 0);
3775 
3776 	if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3777 		if (IS_ENCRYPTED(inode))
3778 			ret = fscrypt_zeroout_range(inode, off, block, len);
3779 		else
3780 			ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3781 					GFP_NOFS, 0);
3782 	}
3783 
3784 	return ret;
3785 }
3786 
3787 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3788 {
3789 	struct inode *inode = file_inode(filp);
3790 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3791 	struct address_space *mapping = inode->i_mapping;
3792 	struct block_device *prev_bdev = NULL;
3793 	struct f2fs_sectrim_range range;
3794 	pgoff_t index, pg_end, prev_index = 0;
3795 	block_t prev_block = 0, len = 0;
3796 	loff_t end_addr;
3797 	bool to_end = false;
3798 	int ret = 0;
3799 
3800 	if (!(filp->f_mode & FMODE_WRITE))
3801 		return -EBADF;
3802 
3803 	if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3804 				sizeof(range)))
3805 		return -EFAULT;
3806 
3807 	if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3808 			!S_ISREG(inode->i_mode))
3809 		return -EINVAL;
3810 
3811 	if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3812 			!f2fs_hw_support_discard(sbi)) ||
3813 			((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3814 			 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3815 		return -EOPNOTSUPP;
3816 
3817 	file_start_write(filp);
3818 	inode_lock(inode);
3819 
3820 	if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3821 			range.start >= inode->i_size) {
3822 		ret = -EINVAL;
3823 		goto err;
3824 	}
3825 
3826 	if (range.len == 0)
3827 		goto err;
3828 
3829 	if (inode->i_size - range.start > range.len) {
3830 		end_addr = range.start + range.len;
3831 	} else {
3832 		end_addr = range.len == (u64)-1 ?
3833 			sbi->sb->s_maxbytes : inode->i_size;
3834 		to_end = true;
3835 	}
3836 
3837 	if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3838 			(!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3839 		ret = -EINVAL;
3840 		goto err;
3841 	}
3842 
3843 	index = F2FS_BYTES_TO_BLK(range.start);
3844 	pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3845 
3846 	ret = f2fs_convert_inline_inode(inode);
3847 	if (ret)
3848 		goto err;
3849 
3850 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3851 	down_write(&F2FS_I(inode)->i_mmap_sem);
3852 
3853 	ret = filemap_write_and_wait_range(mapping, range.start,
3854 			to_end ? LLONG_MAX : end_addr - 1);
3855 	if (ret)
3856 		goto out;
3857 
3858 	truncate_inode_pages_range(mapping, range.start,
3859 			to_end ? -1 : end_addr - 1);
3860 
3861 	while (index < pg_end) {
3862 		struct dnode_of_data dn;
3863 		pgoff_t end_offset, count;
3864 		int i;
3865 
3866 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3867 		ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3868 		if (ret) {
3869 			if (ret == -ENOENT) {
3870 				index = f2fs_get_next_page_offset(&dn, index);
3871 				continue;
3872 			}
3873 			goto out;
3874 		}
3875 
3876 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3877 		count = min(end_offset - dn.ofs_in_node, pg_end - index);
3878 		for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3879 			struct block_device *cur_bdev;
3880 			block_t blkaddr = f2fs_data_blkaddr(&dn);
3881 
3882 			if (!__is_valid_data_blkaddr(blkaddr))
3883 				continue;
3884 
3885 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3886 						DATA_GENERIC_ENHANCE)) {
3887 				ret = -EFSCORRUPTED;
3888 				f2fs_put_dnode(&dn);
3889 				goto out;
3890 			}
3891 
3892 			cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3893 			if (f2fs_is_multi_device(sbi)) {
3894 				int di = f2fs_target_device_index(sbi, blkaddr);
3895 
3896 				blkaddr -= FDEV(di).start_blk;
3897 			}
3898 
3899 			if (len) {
3900 				if (prev_bdev == cur_bdev &&
3901 						index == prev_index + len &&
3902 						blkaddr == prev_block + len) {
3903 					len++;
3904 				} else {
3905 					ret = f2fs_secure_erase(prev_bdev,
3906 						inode, prev_index, prev_block,
3907 						len, range.flags);
3908 					if (ret) {
3909 						f2fs_put_dnode(&dn);
3910 						goto out;
3911 					}
3912 
3913 					len = 0;
3914 				}
3915 			}
3916 
3917 			if (!len) {
3918 				prev_bdev = cur_bdev;
3919 				prev_index = index;
3920 				prev_block = blkaddr;
3921 				len = 1;
3922 			}
3923 		}
3924 
3925 		f2fs_put_dnode(&dn);
3926 
3927 		if (fatal_signal_pending(current)) {
3928 			ret = -EINTR;
3929 			goto out;
3930 		}
3931 		cond_resched();
3932 	}
3933 
3934 	if (len)
3935 		ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3936 				prev_block, len, range.flags);
3937 out:
3938 	up_write(&F2FS_I(inode)->i_mmap_sem);
3939 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3940 err:
3941 	inode_unlock(inode);
3942 	file_end_write(filp);
3943 
3944 	return ret;
3945 }
3946 
3947 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3948 {
3949 	struct inode *inode = file_inode(filp);
3950 	struct f2fs_comp_option option;
3951 
3952 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3953 		return -EOPNOTSUPP;
3954 
3955 	inode_lock_shared(inode);
3956 
3957 	if (!f2fs_compressed_file(inode)) {
3958 		inode_unlock_shared(inode);
3959 		return -ENODATA;
3960 	}
3961 
3962 	option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3963 	option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3964 
3965 	inode_unlock_shared(inode);
3966 
3967 	if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3968 				sizeof(option)))
3969 		return -EFAULT;
3970 
3971 	return 0;
3972 }
3973 
3974 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3975 {
3976 	struct inode *inode = file_inode(filp);
3977 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3978 	struct f2fs_comp_option option;
3979 	int ret = 0;
3980 
3981 	if (!f2fs_sb_has_compression(sbi))
3982 		return -EOPNOTSUPP;
3983 
3984 	if (!(filp->f_mode & FMODE_WRITE))
3985 		return -EBADF;
3986 
3987 	if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3988 				sizeof(option)))
3989 		return -EFAULT;
3990 
3991 	if (!f2fs_compressed_file(inode) ||
3992 			option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3993 			option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3994 			option.algorithm >= COMPRESS_MAX)
3995 		return -EINVAL;
3996 
3997 	file_start_write(filp);
3998 	inode_lock(inode);
3999 
4000 	if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4001 		ret = -EBUSY;
4002 		goto out;
4003 	}
4004 
4005 	if (inode->i_size != 0) {
4006 		ret = -EFBIG;
4007 		goto out;
4008 	}
4009 
4010 	F2FS_I(inode)->i_compress_algorithm = option.algorithm;
4011 	F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
4012 	F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
4013 	f2fs_mark_inode_dirty_sync(inode, true);
4014 
4015 	if (!f2fs_is_compress_backend_ready(inode))
4016 		f2fs_warn(sbi, "compression algorithm is successfully set, "
4017 			"but current kernel doesn't support this algorithm.");
4018 out:
4019 	inode_unlock(inode);
4020 	file_end_write(filp);
4021 
4022 	return ret;
4023 }
4024 
4025 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4026 {
4027 	DEFINE_READAHEAD(ractl, NULL, inode->i_mapping, page_idx);
4028 	struct address_space *mapping = inode->i_mapping;
4029 	struct page *page;
4030 	pgoff_t redirty_idx = page_idx;
4031 	int i, page_len = 0, ret = 0;
4032 
4033 	page_cache_ra_unbounded(&ractl, len, 0);
4034 
4035 	for (i = 0; i < len; i++, page_idx++) {
4036 		page = read_cache_page(mapping, page_idx, NULL, NULL);
4037 		if (IS_ERR(page)) {
4038 			ret = PTR_ERR(page);
4039 			break;
4040 		}
4041 		page_len++;
4042 	}
4043 
4044 	for (i = 0; i < page_len; i++, redirty_idx++) {
4045 		page = find_lock_page(mapping, redirty_idx);
4046 		if (!page)
4047 			ret = -ENOENT;
4048 		set_page_dirty(page);
4049 		f2fs_put_page(page, 1);
4050 		f2fs_put_page(page, 0);
4051 	}
4052 
4053 	return ret;
4054 }
4055 
4056 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
4057 {
4058 	struct inode *inode = file_inode(filp);
4059 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4060 	struct f2fs_inode_info *fi = F2FS_I(inode);
4061 	pgoff_t page_idx = 0, last_idx;
4062 	unsigned int blk_per_seg = sbi->blocks_per_seg;
4063 	int cluster_size = F2FS_I(inode)->i_cluster_size;
4064 	int count, ret;
4065 
4066 	if (!f2fs_sb_has_compression(sbi) ||
4067 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4068 		return -EOPNOTSUPP;
4069 
4070 	if (!(filp->f_mode & FMODE_WRITE))
4071 		return -EBADF;
4072 
4073 	if (!f2fs_compressed_file(inode))
4074 		return -EINVAL;
4075 
4076 	f2fs_balance_fs(F2FS_I_SB(inode), true);
4077 
4078 	file_start_write(filp);
4079 	inode_lock(inode);
4080 
4081 	if (!f2fs_is_compress_backend_ready(inode)) {
4082 		ret = -EOPNOTSUPP;
4083 		goto out;
4084 	}
4085 
4086 	if (f2fs_is_mmap_file(inode)) {
4087 		ret = -EBUSY;
4088 		goto out;
4089 	}
4090 
4091 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4092 	if (ret)
4093 		goto out;
4094 
4095 	if (!atomic_read(&fi->i_compr_blocks))
4096 		goto out;
4097 
4098 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4099 
4100 	count = last_idx - page_idx;
4101 	while (count) {
4102 		int len = min(cluster_size, count);
4103 
4104 		ret = redirty_blocks(inode, page_idx, len);
4105 		if (ret < 0)
4106 			break;
4107 
4108 		if (get_dirty_pages(inode) >= blk_per_seg)
4109 			filemap_fdatawrite(inode->i_mapping);
4110 
4111 		count -= len;
4112 		page_idx += len;
4113 	}
4114 
4115 	if (!ret)
4116 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4117 							LLONG_MAX);
4118 
4119 	if (ret)
4120 		f2fs_warn(sbi, "%s: The file might be partially decompressed "
4121 				"(errno=%d). Please delete the file.\n",
4122 				__func__, ret);
4123 out:
4124 	inode_unlock(inode);
4125 	file_end_write(filp);
4126 
4127 	return ret;
4128 }
4129 
4130 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4131 {
4132 	struct inode *inode = file_inode(filp);
4133 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4134 	pgoff_t page_idx = 0, last_idx;
4135 	unsigned int blk_per_seg = sbi->blocks_per_seg;
4136 	int cluster_size = F2FS_I(inode)->i_cluster_size;
4137 	int count, ret;
4138 
4139 	if (!f2fs_sb_has_compression(sbi) ||
4140 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4141 		return -EOPNOTSUPP;
4142 
4143 	if (!(filp->f_mode & FMODE_WRITE))
4144 		return -EBADF;
4145 
4146 	if (!f2fs_compressed_file(inode))
4147 		return -EINVAL;
4148 
4149 	f2fs_balance_fs(F2FS_I_SB(inode), true);
4150 
4151 	file_start_write(filp);
4152 	inode_lock(inode);
4153 
4154 	if (!f2fs_is_compress_backend_ready(inode)) {
4155 		ret = -EOPNOTSUPP;
4156 		goto out;
4157 	}
4158 
4159 	if (f2fs_is_mmap_file(inode)) {
4160 		ret = -EBUSY;
4161 		goto out;
4162 	}
4163 
4164 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4165 	if (ret)
4166 		goto out;
4167 
4168 	set_inode_flag(inode, FI_ENABLE_COMPRESS);
4169 
4170 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4171 
4172 	count = last_idx - page_idx;
4173 	while (count) {
4174 		int len = min(cluster_size, count);
4175 
4176 		ret = redirty_blocks(inode, page_idx, len);
4177 		if (ret < 0)
4178 			break;
4179 
4180 		if (get_dirty_pages(inode) >= blk_per_seg)
4181 			filemap_fdatawrite(inode->i_mapping);
4182 
4183 		count -= len;
4184 		page_idx += len;
4185 	}
4186 
4187 	if (!ret)
4188 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4189 							LLONG_MAX);
4190 
4191 	clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4192 
4193 	if (ret)
4194 		f2fs_warn(sbi, "%s: The file might be partially compressed "
4195 				"(errno=%d). Please delete the file.\n",
4196 				__func__, ret);
4197 out:
4198 	inode_unlock(inode);
4199 	file_end_write(filp);
4200 
4201 	return ret;
4202 }
4203 
4204 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4205 {
4206 	switch (cmd) {
4207 	case FS_IOC_GETFLAGS:
4208 		return f2fs_ioc_getflags(filp, arg);
4209 	case FS_IOC_SETFLAGS:
4210 		return f2fs_ioc_setflags(filp, arg);
4211 	case FS_IOC_GETVERSION:
4212 		return f2fs_ioc_getversion(filp, arg);
4213 	case F2FS_IOC_START_ATOMIC_WRITE:
4214 		return f2fs_ioc_start_atomic_write(filp);
4215 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4216 		return f2fs_ioc_commit_atomic_write(filp);
4217 	case F2FS_IOC_START_VOLATILE_WRITE:
4218 		return f2fs_ioc_start_volatile_write(filp);
4219 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4220 		return f2fs_ioc_release_volatile_write(filp);
4221 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
4222 		return f2fs_ioc_abort_volatile_write(filp);
4223 	case F2FS_IOC_SHUTDOWN:
4224 		return f2fs_ioc_shutdown(filp, arg);
4225 	case FITRIM:
4226 		return f2fs_ioc_fitrim(filp, arg);
4227 	case FS_IOC_SET_ENCRYPTION_POLICY:
4228 		return f2fs_ioc_set_encryption_policy(filp, arg);
4229 	case FS_IOC_GET_ENCRYPTION_POLICY:
4230 		return f2fs_ioc_get_encryption_policy(filp, arg);
4231 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4232 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4233 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4234 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4235 	case FS_IOC_ADD_ENCRYPTION_KEY:
4236 		return f2fs_ioc_add_encryption_key(filp, arg);
4237 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4238 		return f2fs_ioc_remove_encryption_key(filp, arg);
4239 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4240 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4241 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4242 		return f2fs_ioc_get_encryption_key_status(filp, arg);
4243 	case FS_IOC_GET_ENCRYPTION_NONCE:
4244 		return f2fs_ioc_get_encryption_nonce(filp, arg);
4245 	case F2FS_IOC_GARBAGE_COLLECT:
4246 		return f2fs_ioc_gc(filp, arg);
4247 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4248 		return f2fs_ioc_gc_range(filp, arg);
4249 	case F2FS_IOC_WRITE_CHECKPOINT:
4250 		return f2fs_ioc_write_checkpoint(filp, arg);
4251 	case F2FS_IOC_DEFRAGMENT:
4252 		return f2fs_ioc_defragment(filp, arg);
4253 	case F2FS_IOC_MOVE_RANGE:
4254 		return f2fs_ioc_move_range(filp, arg);
4255 	case F2FS_IOC_FLUSH_DEVICE:
4256 		return f2fs_ioc_flush_device(filp, arg);
4257 	case F2FS_IOC_GET_FEATURES:
4258 		return f2fs_ioc_get_features(filp, arg);
4259 	case FS_IOC_FSGETXATTR:
4260 		return f2fs_ioc_fsgetxattr(filp, arg);
4261 	case FS_IOC_FSSETXATTR:
4262 		return f2fs_ioc_fssetxattr(filp, arg);
4263 	case F2FS_IOC_GET_PIN_FILE:
4264 		return f2fs_ioc_get_pin_file(filp, arg);
4265 	case F2FS_IOC_SET_PIN_FILE:
4266 		return f2fs_ioc_set_pin_file(filp, arg);
4267 	case F2FS_IOC_PRECACHE_EXTENTS:
4268 		return f2fs_ioc_precache_extents(filp, arg);
4269 	case F2FS_IOC_RESIZE_FS:
4270 		return f2fs_ioc_resize_fs(filp, arg);
4271 	case FS_IOC_ENABLE_VERITY:
4272 		return f2fs_ioc_enable_verity(filp, arg);
4273 	case FS_IOC_MEASURE_VERITY:
4274 		return f2fs_ioc_measure_verity(filp, arg);
4275 	case FS_IOC_GETFSLABEL:
4276 		return f2fs_ioc_getfslabel(filp, arg);
4277 	case FS_IOC_SETFSLABEL:
4278 		return f2fs_ioc_setfslabel(filp, arg);
4279 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4280 		return f2fs_get_compress_blocks(filp, arg);
4281 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4282 		return f2fs_release_compress_blocks(filp, arg);
4283 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4284 		return f2fs_reserve_compress_blocks(filp, arg);
4285 	case F2FS_IOC_SEC_TRIM_FILE:
4286 		return f2fs_sec_trim_file(filp, arg);
4287 	case F2FS_IOC_GET_COMPRESS_OPTION:
4288 		return f2fs_ioc_get_compress_option(filp, arg);
4289 	case F2FS_IOC_SET_COMPRESS_OPTION:
4290 		return f2fs_ioc_set_compress_option(filp, arg);
4291 	case F2FS_IOC_DECOMPRESS_FILE:
4292 		return f2fs_ioc_decompress_file(filp, arg);
4293 	case F2FS_IOC_COMPRESS_FILE:
4294 		return f2fs_ioc_compress_file(filp, arg);
4295 	default:
4296 		return -ENOTTY;
4297 	}
4298 }
4299 
4300 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4301 {
4302 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4303 		return -EIO;
4304 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4305 		return -ENOSPC;
4306 
4307 	return __f2fs_ioctl(filp, cmd, arg);
4308 }
4309 
4310 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
4311 {
4312 	struct file *file = iocb->ki_filp;
4313 	struct inode *inode = file_inode(file);
4314 	int ret;
4315 
4316 	if (!f2fs_is_compress_backend_ready(inode))
4317 		return -EOPNOTSUPP;
4318 
4319 	ret = generic_file_read_iter(iocb, iter);
4320 
4321 	if (ret > 0)
4322 		f2fs_update_iostat(F2FS_I_SB(inode), APP_READ_IO, ret);
4323 
4324 	return ret;
4325 }
4326 
4327 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4328 {
4329 	struct file *file = iocb->ki_filp;
4330 	struct inode *inode = file_inode(file);
4331 	ssize_t ret;
4332 
4333 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4334 		ret = -EIO;
4335 		goto out;
4336 	}
4337 
4338 	if (!f2fs_is_compress_backend_ready(inode)) {
4339 		ret = -EOPNOTSUPP;
4340 		goto out;
4341 	}
4342 
4343 	if (iocb->ki_flags & IOCB_NOWAIT) {
4344 		if (!inode_trylock(inode)) {
4345 			ret = -EAGAIN;
4346 			goto out;
4347 		}
4348 	} else {
4349 		inode_lock(inode);
4350 	}
4351 
4352 	ret = generic_write_checks(iocb, from);
4353 	if (ret > 0) {
4354 		bool preallocated = false;
4355 		size_t target_size = 0;
4356 		int err;
4357 
4358 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
4359 			set_inode_flag(inode, FI_NO_PREALLOC);
4360 
4361 		if ((iocb->ki_flags & IOCB_NOWAIT)) {
4362 			if (!f2fs_overwrite_io(inode, iocb->ki_pos,
4363 						iov_iter_count(from)) ||
4364 				f2fs_has_inline_data(inode) ||
4365 				f2fs_force_buffered_io(inode, iocb, from)) {
4366 				clear_inode_flag(inode, FI_NO_PREALLOC);
4367 				inode_unlock(inode);
4368 				ret = -EAGAIN;
4369 				goto out;
4370 			}
4371 			goto write;
4372 		}
4373 
4374 		if (is_inode_flag_set(inode, FI_NO_PREALLOC))
4375 			goto write;
4376 
4377 		if (iocb->ki_flags & IOCB_DIRECT) {
4378 			/*
4379 			 * Convert inline data for Direct I/O before entering
4380 			 * f2fs_direct_IO().
4381 			 */
4382 			err = f2fs_convert_inline_inode(inode);
4383 			if (err)
4384 				goto out_err;
4385 			/*
4386 			 * If force_buffere_io() is true, we have to allocate
4387 			 * blocks all the time, since f2fs_direct_IO will fall
4388 			 * back to buffered IO.
4389 			 */
4390 			if (!f2fs_force_buffered_io(inode, iocb, from) &&
4391 					allow_outplace_dio(inode, iocb, from))
4392 				goto write;
4393 		}
4394 		preallocated = true;
4395 		target_size = iocb->ki_pos + iov_iter_count(from);
4396 
4397 		err = f2fs_preallocate_blocks(iocb, from);
4398 		if (err) {
4399 out_err:
4400 			clear_inode_flag(inode, FI_NO_PREALLOC);
4401 			inode_unlock(inode);
4402 			ret = err;
4403 			goto out;
4404 		}
4405 write:
4406 		ret = __generic_file_write_iter(iocb, from);
4407 		clear_inode_flag(inode, FI_NO_PREALLOC);
4408 
4409 		/* if we couldn't write data, we should deallocate blocks. */
4410 		if (preallocated && i_size_read(inode) < target_size)
4411 			f2fs_truncate(inode);
4412 
4413 		if (ret > 0)
4414 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
4415 	}
4416 	inode_unlock(inode);
4417 out:
4418 	trace_f2fs_file_write_iter(inode, iocb->ki_pos,
4419 					iov_iter_count(from), ret);
4420 	if (ret > 0)
4421 		ret = generic_write_sync(iocb, ret);
4422 	return ret;
4423 }
4424 
4425 #ifdef CONFIG_COMPAT
4426 struct compat_f2fs_gc_range {
4427 	u32 sync;
4428 	compat_u64 start;
4429 	compat_u64 len;
4430 };
4431 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE	_IOW(F2FS_IOCTL_MAGIC, 11,\
4432 						struct compat_f2fs_gc_range)
4433 
4434 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4435 {
4436 	struct compat_f2fs_gc_range __user *urange;
4437 	struct f2fs_gc_range range;
4438 	int err;
4439 
4440 	urange = compat_ptr(arg);
4441 	err = get_user(range.sync, &urange->sync);
4442 	err |= get_user(range.start, &urange->start);
4443 	err |= get_user(range.len, &urange->len);
4444 	if (err)
4445 		return -EFAULT;
4446 
4447 	return __f2fs_ioc_gc_range(file, &range);
4448 }
4449 
4450 struct compat_f2fs_move_range {
4451 	u32 dst_fd;
4452 	compat_u64 pos_in;
4453 	compat_u64 pos_out;
4454 	compat_u64 len;
4455 };
4456 #define F2FS_IOC32_MOVE_RANGE		_IOWR(F2FS_IOCTL_MAGIC, 9,	\
4457 					struct compat_f2fs_move_range)
4458 
4459 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4460 {
4461 	struct compat_f2fs_move_range __user *urange;
4462 	struct f2fs_move_range range;
4463 	int err;
4464 
4465 	urange = compat_ptr(arg);
4466 	err = get_user(range.dst_fd, &urange->dst_fd);
4467 	err |= get_user(range.pos_in, &urange->pos_in);
4468 	err |= get_user(range.pos_out, &urange->pos_out);
4469 	err |= get_user(range.len, &urange->len);
4470 	if (err)
4471 		return -EFAULT;
4472 
4473 	return __f2fs_ioc_move_range(file, &range);
4474 }
4475 
4476 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4477 {
4478 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4479 		return -EIO;
4480 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4481 		return -ENOSPC;
4482 
4483 	switch (cmd) {
4484 	case FS_IOC32_GETFLAGS:
4485 		cmd = FS_IOC_GETFLAGS;
4486 		break;
4487 	case FS_IOC32_SETFLAGS:
4488 		cmd = FS_IOC_SETFLAGS;
4489 		break;
4490 	case FS_IOC32_GETVERSION:
4491 		cmd = FS_IOC_GETVERSION;
4492 		break;
4493 	case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4494 		return f2fs_compat_ioc_gc_range(file, arg);
4495 	case F2FS_IOC32_MOVE_RANGE:
4496 		return f2fs_compat_ioc_move_range(file, arg);
4497 	case F2FS_IOC_START_ATOMIC_WRITE:
4498 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4499 	case F2FS_IOC_START_VOLATILE_WRITE:
4500 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4501 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
4502 	case F2FS_IOC_SHUTDOWN:
4503 	case FITRIM:
4504 	case FS_IOC_SET_ENCRYPTION_POLICY:
4505 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4506 	case FS_IOC_GET_ENCRYPTION_POLICY:
4507 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4508 	case FS_IOC_ADD_ENCRYPTION_KEY:
4509 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4510 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4511 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4512 	case FS_IOC_GET_ENCRYPTION_NONCE:
4513 	case F2FS_IOC_GARBAGE_COLLECT:
4514 	case F2FS_IOC_WRITE_CHECKPOINT:
4515 	case F2FS_IOC_DEFRAGMENT:
4516 	case F2FS_IOC_FLUSH_DEVICE:
4517 	case F2FS_IOC_GET_FEATURES:
4518 	case FS_IOC_FSGETXATTR:
4519 	case FS_IOC_FSSETXATTR:
4520 	case F2FS_IOC_GET_PIN_FILE:
4521 	case F2FS_IOC_SET_PIN_FILE:
4522 	case F2FS_IOC_PRECACHE_EXTENTS:
4523 	case F2FS_IOC_RESIZE_FS:
4524 	case FS_IOC_ENABLE_VERITY:
4525 	case FS_IOC_MEASURE_VERITY:
4526 	case FS_IOC_GETFSLABEL:
4527 	case FS_IOC_SETFSLABEL:
4528 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4529 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4530 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4531 	case F2FS_IOC_SEC_TRIM_FILE:
4532 	case F2FS_IOC_GET_COMPRESS_OPTION:
4533 	case F2FS_IOC_SET_COMPRESS_OPTION:
4534 	case F2FS_IOC_DECOMPRESS_FILE:
4535 	case F2FS_IOC_COMPRESS_FILE:
4536 		break;
4537 	default:
4538 		return -ENOIOCTLCMD;
4539 	}
4540 	return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4541 }
4542 #endif
4543 
4544 const struct file_operations f2fs_file_operations = {
4545 	.llseek		= f2fs_llseek,
4546 	.read_iter	= f2fs_file_read_iter,
4547 	.write_iter	= f2fs_file_write_iter,
4548 	.open		= f2fs_file_open,
4549 	.release	= f2fs_release_file,
4550 	.mmap		= f2fs_file_mmap,
4551 	.flush		= f2fs_file_flush,
4552 	.fsync		= f2fs_sync_file,
4553 	.fallocate	= f2fs_fallocate,
4554 	.unlocked_ioctl	= f2fs_ioctl,
4555 #ifdef CONFIG_COMPAT
4556 	.compat_ioctl	= f2fs_compat_ioctl,
4557 #endif
4558 	.splice_read	= generic_file_splice_read,
4559 	.splice_write	= iter_file_splice_write,
4560 };
4561