xref: /linux/fs/f2fs/file.c (revision a1c3be890440a1769ed6f822376a3e3ab0d42994)
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/events/f2fs.h>
33 #include <uapi/linux/f2fs.h>
34 
35 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
36 {
37 	struct inode *inode = file_inode(vmf->vma->vm_file);
38 	vm_fault_t ret;
39 
40 	down_read(&F2FS_I(inode)->i_mmap_sem);
41 	ret = filemap_fault(vmf);
42 	up_read(&F2FS_I(inode)->i_mmap_sem);
43 
44 	if (!ret)
45 		f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
46 							F2FS_BLKSIZE);
47 
48 	trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
49 
50 	return ret;
51 }
52 
53 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
54 {
55 	struct page *page = vmf->page;
56 	struct inode *inode = file_inode(vmf->vma->vm_file);
57 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
58 	struct dnode_of_data dn;
59 	bool need_alloc = true;
60 	int err = 0;
61 
62 	if (unlikely(IS_IMMUTABLE(inode)))
63 		return VM_FAULT_SIGBUS;
64 
65 	if (unlikely(f2fs_cp_error(sbi))) {
66 		err = -EIO;
67 		goto err;
68 	}
69 
70 	if (!f2fs_is_checkpoint_ready(sbi)) {
71 		err = -ENOSPC;
72 		goto err;
73 	}
74 
75 	err = f2fs_convert_inline_inode(inode);
76 	if (err)
77 		goto err;
78 
79 #ifdef CONFIG_F2FS_FS_COMPRESSION
80 	if (f2fs_compressed_file(inode)) {
81 		int ret = f2fs_is_compressed_cluster(inode, page->index);
82 
83 		if (ret < 0) {
84 			err = ret;
85 			goto err;
86 		} else if (ret) {
87 			if (ret < F2FS_I(inode)->i_cluster_size) {
88 				err = -EAGAIN;
89 				goto err;
90 			}
91 			need_alloc = false;
92 		}
93 	}
94 #endif
95 	/* should do out of any locked page */
96 	if (need_alloc)
97 		f2fs_balance_fs(sbi, true);
98 
99 	sb_start_pagefault(inode->i_sb);
100 
101 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
102 
103 	file_update_time(vmf->vma->vm_file);
104 	down_read(&F2FS_I(inode)->i_mmap_sem);
105 	lock_page(page);
106 	if (unlikely(page->mapping != inode->i_mapping ||
107 			page_offset(page) > i_size_read(inode) ||
108 			!PageUptodate(page))) {
109 		unlock_page(page);
110 		err = -EFAULT;
111 		goto out_sem;
112 	}
113 
114 	if (need_alloc) {
115 		/* block allocation */
116 		f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
117 		set_new_dnode(&dn, inode, NULL, NULL, 0);
118 		err = f2fs_get_block(&dn, page->index);
119 		f2fs_put_dnode(&dn);
120 		f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
121 	}
122 
123 #ifdef CONFIG_F2FS_FS_COMPRESSION
124 	if (!need_alloc) {
125 		set_new_dnode(&dn, inode, NULL, NULL, 0);
126 		err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
127 		f2fs_put_dnode(&dn);
128 	}
129 #endif
130 	if (err) {
131 		unlock_page(page);
132 		goto out_sem;
133 	}
134 
135 	f2fs_wait_on_page_writeback(page, DATA, false, true);
136 
137 	/* wait for GCed page writeback via META_MAPPING */
138 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
139 
140 	/*
141 	 * check to see if the page is mapped already (no holes)
142 	 */
143 	if (PageMappedToDisk(page))
144 		goto out_sem;
145 
146 	/* page is wholly or partially inside EOF */
147 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
148 						i_size_read(inode)) {
149 		loff_t offset;
150 
151 		offset = i_size_read(inode) & ~PAGE_MASK;
152 		zero_user_segment(page, offset, PAGE_SIZE);
153 	}
154 	set_page_dirty(page);
155 	if (!PageUptodate(page))
156 		SetPageUptodate(page);
157 
158 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
159 	f2fs_update_time(sbi, REQ_TIME);
160 
161 	trace_f2fs_vm_page_mkwrite(page, DATA);
162 out_sem:
163 	up_read(&F2FS_I(inode)->i_mmap_sem);
164 
165 	sb_end_pagefault(inode->i_sb);
166 err:
167 	return block_page_mkwrite_return(err);
168 }
169 
170 static const struct vm_operations_struct f2fs_file_vm_ops = {
171 	.fault		= f2fs_filemap_fault,
172 	.map_pages	= filemap_map_pages,
173 	.page_mkwrite	= f2fs_vm_page_mkwrite,
174 };
175 
176 static int get_parent_ino(struct inode *inode, nid_t *pino)
177 {
178 	struct dentry *dentry;
179 
180 	/*
181 	 * Make sure to get the non-deleted alias.  The alias associated with
182 	 * the open file descriptor being fsync()'ed may be deleted already.
183 	 */
184 	dentry = d_find_alias(inode);
185 	if (!dentry)
186 		return 0;
187 
188 	*pino = parent_ino(dentry);
189 	dput(dentry);
190 	return 1;
191 }
192 
193 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
194 {
195 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
196 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
197 
198 	if (!S_ISREG(inode->i_mode))
199 		cp_reason = CP_NON_REGULAR;
200 	else if (f2fs_compressed_file(inode))
201 		cp_reason = CP_COMPRESSED;
202 	else if (inode->i_nlink != 1)
203 		cp_reason = CP_HARDLINK;
204 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
205 		cp_reason = CP_SB_NEED_CP;
206 	else if (file_wrong_pino(inode))
207 		cp_reason = CP_WRONG_PINO;
208 	else if (!f2fs_space_for_roll_forward(sbi))
209 		cp_reason = CP_NO_SPC_ROLL;
210 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
211 		cp_reason = CP_NODE_NEED_CP;
212 	else if (test_opt(sbi, FASTBOOT))
213 		cp_reason = CP_FASTBOOT_MODE;
214 	else if (F2FS_OPTION(sbi).active_logs == 2)
215 		cp_reason = CP_SPEC_LOG_NUM;
216 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
217 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
218 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
219 							TRANS_DIR_INO))
220 		cp_reason = CP_RECOVER_DIR;
221 
222 	return cp_reason;
223 }
224 
225 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
226 {
227 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
228 	bool ret = false;
229 	/* But we need to avoid that there are some inode updates */
230 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
231 		ret = true;
232 	f2fs_put_page(i, 0);
233 	return ret;
234 }
235 
236 static void try_to_fix_pino(struct inode *inode)
237 {
238 	struct f2fs_inode_info *fi = F2FS_I(inode);
239 	nid_t pino;
240 
241 	down_write(&fi->i_sem);
242 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
243 			get_parent_ino(inode, &pino)) {
244 		f2fs_i_pino_write(inode, pino);
245 		file_got_pino(inode);
246 	}
247 	up_write(&fi->i_sem);
248 }
249 
250 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
251 						int datasync, bool atomic)
252 {
253 	struct inode *inode = file->f_mapping->host;
254 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
255 	nid_t ino = inode->i_ino;
256 	int ret = 0;
257 	enum cp_reason_type cp_reason = 0;
258 	struct writeback_control wbc = {
259 		.sync_mode = WB_SYNC_ALL,
260 		.nr_to_write = LONG_MAX,
261 		.for_reclaim = 0,
262 	};
263 	unsigned int seq_id = 0;
264 
265 	if (unlikely(f2fs_readonly(inode->i_sb) ||
266 				is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
267 		return 0;
268 
269 	trace_f2fs_sync_file_enter(inode);
270 
271 	if (S_ISDIR(inode->i_mode))
272 		goto go_write;
273 
274 	/* if fdatasync is triggered, let's do in-place-update */
275 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
276 		set_inode_flag(inode, FI_NEED_IPU);
277 	ret = file_write_and_wait_range(file, start, end);
278 	clear_inode_flag(inode, FI_NEED_IPU);
279 
280 	if (ret) {
281 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
282 		return ret;
283 	}
284 
285 	/* if the inode is dirty, let's recover all the time */
286 	if (!f2fs_skip_inode_update(inode, datasync)) {
287 		f2fs_write_inode(inode, NULL);
288 		goto go_write;
289 	}
290 
291 	/*
292 	 * if there is no written data, don't waste time to write recovery info.
293 	 */
294 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
295 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
296 
297 		/* it may call write_inode just prior to fsync */
298 		if (need_inode_page_update(sbi, ino))
299 			goto go_write;
300 
301 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
302 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
303 			goto flush_out;
304 		goto out;
305 	}
306 go_write:
307 	/*
308 	 * Both of fdatasync() and fsync() are able to be recovered from
309 	 * sudden-power-off.
310 	 */
311 	down_read(&F2FS_I(inode)->i_sem);
312 	cp_reason = need_do_checkpoint(inode);
313 	up_read(&F2FS_I(inode)->i_sem);
314 
315 	if (cp_reason) {
316 		/* all the dirty node pages should be flushed for POR */
317 		ret = f2fs_sync_fs(inode->i_sb, 1);
318 
319 		/*
320 		 * We've secured consistency through sync_fs. Following pino
321 		 * will be used only for fsynced inodes after checkpoint.
322 		 */
323 		try_to_fix_pino(inode);
324 		clear_inode_flag(inode, FI_APPEND_WRITE);
325 		clear_inode_flag(inode, FI_UPDATE_WRITE);
326 		goto out;
327 	}
328 sync_nodes:
329 	atomic_inc(&sbi->wb_sync_req[NODE]);
330 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
331 	atomic_dec(&sbi->wb_sync_req[NODE]);
332 	if (ret)
333 		goto out;
334 
335 	/* if cp_error was enabled, we should avoid infinite loop */
336 	if (unlikely(f2fs_cp_error(sbi))) {
337 		ret = -EIO;
338 		goto out;
339 	}
340 
341 	if (f2fs_need_inode_block_update(sbi, ino)) {
342 		f2fs_mark_inode_dirty_sync(inode, true);
343 		f2fs_write_inode(inode, NULL);
344 		goto sync_nodes;
345 	}
346 
347 	/*
348 	 * If it's atomic_write, it's just fine to keep write ordering. So
349 	 * here we don't need to wait for node write completion, since we use
350 	 * node chain which serializes node blocks. If one of node writes are
351 	 * reordered, we can see simply broken chain, resulting in stopping
352 	 * roll-forward recovery. It means we'll recover all or none node blocks
353 	 * given fsync mark.
354 	 */
355 	if (!atomic) {
356 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
357 		if (ret)
358 			goto out;
359 	}
360 
361 	/* once recovery info is written, don't need to tack this */
362 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
363 	clear_inode_flag(inode, FI_APPEND_WRITE);
364 flush_out:
365 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
366 		ret = f2fs_issue_flush(sbi, inode->i_ino);
367 	if (!ret) {
368 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
369 		clear_inode_flag(inode, FI_UPDATE_WRITE);
370 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
371 	}
372 	f2fs_update_time(sbi, REQ_TIME);
373 out:
374 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
375 	return ret;
376 }
377 
378 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
379 {
380 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
381 		return -EIO;
382 	return f2fs_do_sync_file(file, start, end, datasync, false);
383 }
384 
385 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
386 				pgoff_t index, int whence)
387 {
388 	switch (whence) {
389 	case SEEK_DATA:
390 		if (__is_valid_data_blkaddr(blkaddr))
391 			return true;
392 		if (blkaddr == NEW_ADDR &&
393 		    xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
394 			return true;
395 		break;
396 	case SEEK_HOLE:
397 		if (blkaddr == NULL_ADDR)
398 			return true;
399 		break;
400 	}
401 	return false;
402 }
403 
404 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
405 {
406 	struct inode *inode = file->f_mapping->host;
407 	loff_t maxbytes = inode->i_sb->s_maxbytes;
408 	struct dnode_of_data dn;
409 	pgoff_t pgofs, end_offset;
410 	loff_t data_ofs = offset;
411 	loff_t isize;
412 	int err = 0;
413 
414 	inode_lock(inode);
415 
416 	isize = i_size_read(inode);
417 	if (offset >= isize)
418 		goto fail;
419 
420 	/* handle inline data case */
421 	if (f2fs_has_inline_data(inode)) {
422 		if (whence == SEEK_HOLE) {
423 			data_ofs = isize;
424 			goto found;
425 		} else if (whence == SEEK_DATA) {
426 			data_ofs = offset;
427 			goto found;
428 		}
429 	}
430 
431 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
432 
433 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
434 		set_new_dnode(&dn, inode, NULL, NULL, 0);
435 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
436 		if (err && err != -ENOENT) {
437 			goto fail;
438 		} else if (err == -ENOENT) {
439 			/* direct node does not exists */
440 			if (whence == SEEK_DATA) {
441 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
442 				continue;
443 			} else {
444 				goto found;
445 			}
446 		}
447 
448 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
449 
450 		/* find data/hole in dnode block */
451 		for (; dn.ofs_in_node < end_offset;
452 				dn.ofs_in_node++, pgofs++,
453 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
454 			block_t blkaddr;
455 
456 			blkaddr = f2fs_data_blkaddr(&dn);
457 
458 			if (__is_valid_data_blkaddr(blkaddr) &&
459 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
460 					blkaddr, DATA_GENERIC_ENHANCE)) {
461 				f2fs_put_dnode(&dn);
462 				goto fail;
463 			}
464 
465 			if (__found_offset(file->f_mapping, blkaddr,
466 							pgofs, whence)) {
467 				f2fs_put_dnode(&dn);
468 				goto found;
469 			}
470 		}
471 		f2fs_put_dnode(&dn);
472 	}
473 
474 	if (whence == SEEK_DATA)
475 		goto fail;
476 found:
477 	if (whence == SEEK_HOLE && data_ofs > isize)
478 		data_ofs = isize;
479 	inode_unlock(inode);
480 	return vfs_setpos(file, data_ofs, maxbytes);
481 fail:
482 	inode_unlock(inode);
483 	return -ENXIO;
484 }
485 
486 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
487 {
488 	struct inode *inode = file->f_mapping->host;
489 	loff_t maxbytes = inode->i_sb->s_maxbytes;
490 
491 	if (f2fs_compressed_file(inode))
492 		maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
493 
494 	switch (whence) {
495 	case SEEK_SET:
496 	case SEEK_CUR:
497 	case SEEK_END:
498 		return generic_file_llseek_size(file, offset, whence,
499 						maxbytes, i_size_read(inode));
500 	case SEEK_DATA:
501 	case SEEK_HOLE:
502 		if (offset < 0)
503 			return -ENXIO;
504 		return f2fs_seek_block(file, offset, whence);
505 	}
506 
507 	return -EINVAL;
508 }
509 
510 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
511 {
512 	struct inode *inode = file_inode(file);
513 
514 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
515 		return -EIO;
516 
517 	if (!f2fs_is_compress_backend_ready(inode))
518 		return -EOPNOTSUPP;
519 
520 	file_accessed(file);
521 	vma->vm_ops = &f2fs_file_vm_ops;
522 	set_inode_flag(inode, FI_MMAP_FILE);
523 	return 0;
524 }
525 
526 static int f2fs_file_open(struct inode *inode, struct file *filp)
527 {
528 	int err = fscrypt_file_open(inode, filp);
529 
530 	if (err)
531 		return err;
532 
533 	if (!f2fs_is_compress_backend_ready(inode))
534 		return -EOPNOTSUPP;
535 
536 	err = fsverity_file_open(inode, filp);
537 	if (err)
538 		return err;
539 
540 	filp->f_mode |= FMODE_NOWAIT;
541 
542 	return dquot_file_open(inode, filp);
543 }
544 
545 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
546 {
547 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
548 	struct f2fs_node *raw_node;
549 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
550 	__le32 *addr;
551 	int base = 0;
552 	bool compressed_cluster = false;
553 	int cluster_index = 0, valid_blocks = 0;
554 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
555 	bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
556 
557 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
558 		base = get_extra_isize(dn->inode);
559 
560 	raw_node = F2FS_NODE(dn->node_page);
561 	addr = blkaddr_in_node(raw_node) + base + ofs;
562 
563 	/* Assumption: truncateion starts with cluster */
564 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
565 		block_t blkaddr = le32_to_cpu(*addr);
566 
567 		if (f2fs_compressed_file(dn->inode) &&
568 					!(cluster_index & (cluster_size - 1))) {
569 			if (compressed_cluster)
570 				f2fs_i_compr_blocks_update(dn->inode,
571 							valid_blocks, false);
572 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
573 			valid_blocks = 0;
574 		}
575 
576 		if (blkaddr == NULL_ADDR)
577 			continue;
578 
579 		dn->data_blkaddr = NULL_ADDR;
580 		f2fs_set_data_blkaddr(dn);
581 
582 		if (__is_valid_data_blkaddr(blkaddr)) {
583 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
584 					DATA_GENERIC_ENHANCE))
585 				continue;
586 			if (compressed_cluster)
587 				valid_blocks++;
588 		}
589 
590 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
591 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
592 
593 		f2fs_invalidate_blocks(sbi, blkaddr);
594 
595 		if (!released || blkaddr != COMPRESS_ADDR)
596 			nr_free++;
597 	}
598 
599 	if (compressed_cluster)
600 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
601 
602 	if (nr_free) {
603 		pgoff_t fofs;
604 		/*
605 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
606 		 * we will invalidate all blkaddr in the whole range.
607 		 */
608 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
609 							dn->inode) + ofs;
610 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
611 		dec_valid_block_count(sbi, dn->inode, nr_free);
612 	}
613 	dn->ofs_in_node = ofs;
614 
615 	f2fs_update_time(sbi, REQ_TIME);
616 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
617 					 dn->ofs_in_node, nr_free);
618 }
619 
620 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
621 {
622 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
623 }
624 
625 static int truncate_partial_data_page(struct inode *inode, u64 from,
626 								bool cache_only)
627 {
628 	loff_t offset = from & (PAGE_SIZE - 1);
629 	pgoff_t index = from >> PAGE_SHIFT;
630 	struct address_space *mapping = inode->i_mapping;
631 	struct page *page;
632 
633 	if (!offset && !cache_only)
634 		return 0;
635 
636 	if (cache_only) {
637 		page = find_lock_page(mapping, index);
638 		if (page && PageUptodate(page))
639 			goto truncate_out;
640 		f2fs_put_page(page, 1);
641 		return 0;
642 	}
643 
644 	page = f2fs_get_lock_data_page(inode, index, true);
645 	if (IS_ERR(page))
646 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
647 truncate_out:
648 	f2fs_wait_on_page_writeback(page, DATA, true, true);
649 	zero_user(page, offset, PAGE_SIZE - offset);
650 
651 	/* An encrypted inode should have a key and truncate the last page. */
652 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
653 	if (!cache_only)
654 		set_page_dirty(page);
655 	f2fs_put_page(page, 1);
656 	return 0;
657 }
658 
659 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
660 {
661 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
662 	struct dnode_of_data dn;
663 	pgoff_t free_from;
664 	int count = 0, err = 0;
665 	struct page *ipage;
666 	bool truncate_page = false;
667 
668 	trace_f2fs_truncate_blocks_enter(inode, from);
669 
670 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
671 
672 	if (free_from >= max_file_blocks(inode))
673 		goto free_partial;
674 
675 	if (lock)
676 		f2fs_lock_op(sbi);
677 
678 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
679 	if (IS_ERR(ipage)) {
680 		err = PTR_ERR(ipage);
681 		goto out;
682 	}
683 
684 	if (f2fs_has_inline_data(inode)) {
685 		f2fs_truncate_inline_inode(inode, ipage, from);
686 		f2fs_put_page(ipage, 1);
687 		truncate_page = true;
688 		goto out;
689 	}
690 
691 	set_new_dnode(&dn, inode, ipage, NULL, 0);
692 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
693 	if (err) {
694 		if (err == -ENOENT)
695 			goto free_next;
696 		goto out;
697 	}
698 
699 	count = ADDRS_PER_PAGE(dn.node_page, inode);
700 
701 	count -= dn.ofs_in_node;
702 	f2fs_bug_on(sbi, count < 0);
703 
704 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
705 		f2fs_truncate_data_blocks_range(&dn, count);
706 		free_from += count;
707 	}
708 
709 	f2fs_put_dnode(&dn);
710 free_next:
711 	err = f2fs_truncate_inode_blocks(inode, free_from);
712 out:
713 	if (lock)
714 		f2fs_unlock_op(sbi);
715 free_partial:
716 	/* lastly zero out the first data page */
717 	if (!err)
718 		err = truncate_partial_data_page(inode, from, truncate_page);
719 
720 	trace_f2fs_truncate_blocks_exit(inode, err);
721 	return err;
722 }
723 
724 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
725 {
726 	u64 free_from = from;
727 	int err;
728 
729 #ifdef CONFIG_F2FS_FS_COMPRESSION
730 	/*
731 	 * for compressed file, only support cluster size
732 	 * aligned truncation.
733 	 */
734 	if (f2fs_compressed_file(inode))
735 		free_from = round_up(from,
736 				F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
737 #endif
738 
739 	err = f2fs_do_truncate_blocks(inode, free_from, lock);
740 	if (err)
741 		return err;
742 
743 #ifdef CONFIG_F2FS_FS_COMPRESSION
744 	if (from != free_from) {
745 		err = f2fs_truncate_partial_cluster(inode, from, lock);
746 		if (err)
747 			return err;
748 	}
749 #endif
750 
751 	return 0;
752 }
753 
754 int f2fs_truncate(struct inode *inode)
755 {
756 	int err;
757 
758 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
759 		return -EIO;
760 
761 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
762 				S_ISLNK(inode->i_mode)))
763 		return 0;
764 
765 	trace_f2fs_truncate(inode);
766 
767 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
768 		f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
769 		return -EIO;
770 	}
771 
772 	err = dquot_initialize(inode);
773 	if (err)
774 		return err;
775 
776 	/* we should check inline_data size */
777 	if (!f2fs_may_inline_data(inode)) {
778 		err = f2fs_convert_inline_inode(inode);
779 		if (err)
780 			return err;
781 	}
782 
783 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
784 	if (err)
785 		return err;
786 
787 	inode->i_mtime = inode->i_ctime = current_time(inode);
788 	f2fs_mark_inode_dirty_sync(inode, false);
789 	return 0;
790 }
791 
792 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
793 		 struct kstat *stat, u32 request_mask, unsigned int query_flags)
794 {
795 	struct inode *inode = d_inode(path->dentry);
796 	struct f2fs_inode_info *fi = F2FS_I(inode);
797 	struct f2fs_inode *ri;
798 	unsigned int flags;
799 
800 	if (f2fs_has_extra_attr(inode) &&
801 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
802 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
803 		stat->result_mask |= STATX_BTIME;
804 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
805 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
806 	}
807 
808 	flags = fi->i_flags;
809 	if (flags & F2FS_COMPR_FL)
810 		stat->attributes |= STATX_ATTR_COMPRESSED;
811 	if (flags & F2FS_APPEND_FL)
812 		stat->attributes |= STATX_ATTR_APPEND;
813 	if (IS_ENCRYPTED(inode))
814 		stat->attributes |= STATX_ATTR_ENCRYPTED;
815 	if (flags & F2FS_IMMUTABLE_FL)
816 		stat->attributes |= STATX_ATTR_IMMUTABLE;
817 	if (flags & F2FS_NODUMP_FL)
818 		stat->attributes |= STATX_ATTR_NODUMP;
819 	if (IS_VERITY(inode))
820 		stat->attributes |= STATX_ATTR_VERITY;
821 
822 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
823 				  STATX_ATTR_APPEND |
824 				  STATX_ATTR_ENCRYPTED |
825 				  STATX_ATTR_IMMUTABLE |
826 				  STATX_ATTR_NODUMP |
827 				  STATX_ATTR_VERITY);
828 
829 	generic_fillattr(&init_user_ns, inode, stat);
830 
831 	/* we need to show initial sectors used for inline_data/dentries */
832 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
833 					f2fs_has_inline_dentry(inode))
834 		stat->blocks += (stat->size + 511) >> 9;
835 
836 	return 0;
837 }
838 
839 #ifdef CONFIG_F2FS_FS_POSIX_ACL
840 static void __setattr_copy(struct user_namespace *mnt_userns,
841 			   struct inode *inode, const struct iattr *attr)
842 {
843 	unsigned int ia_valid = attr->ia_valid;
844 
845 	if (ia_valid & ATTR_UID)
846 		inode->i_uid = attr->ia_uid;
847 	if (ia_valid & ATTR_GID)
848 		inode->i_gid = attr->ia_gid;
849 	if (ia_valid & ATTR_ATIME)
850 		inode->i_atime = attr->ia_atime;
851 	if (ia_valid & ATTR_MTIME)
852 		inode->i_mtime = attr->ia_mtime;
853 	if (ia_valid & ATTR_CTIME)
854 		inode->i_ctime = attr->ia_ctime;
855 	if (ia_valid & ATTR_MODE) {
856 		umode_t mode = attr->ia_mode;
857 		kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
858 
859 		if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
860 			mode &= ~S_ISGID;
861 		set_acl_inode(inode, mode);
862 	}
863 }
864 #else
865 #define __setattr_copy setattr_copy
866 #endif
867 
868 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
869 		 struct iattr *attr)
870 {
871 	struct inode *inode = d_inode(dentry);
872 	int err;
873 
874 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
875 		return -EIO;
876 
877 	if (unlikely(IS_IMMUTABLE(inode)))
878 		return -EPERM;
879 
880 	if (unlikely(IS_APPEND(inode) &&
881 			(attr->ia_valid & (ATTR_MODE | ATTR_UID |
882 				  ATTR_GID | ATTR_TIMES_SET))))
883 		return -EPERM;
884 
885 	if ((attr->ia_valid & ATTR_SIZE) &&
886 		!f2fs_is_compress_backend_ready(inode))
887 		return -EOPNOTSUPP;
888 
889 	err = setattr_prepare(&init_user_ns, dentry, attr);
890 	if (err)
891 		return err;
892 
893 	err = fscrypt_prepare_setattr(dentry, attr);
894 	if (err)
895 		return err;
896 
897 	err = fsverity_prepare_setattr(dentry, attr);
898 	if (err)
899 		return err;
900 
901 	if (is_quota_modification(inode, attr)) {
902 		err = dquot_initialize(inode);
903 		if (err)
904 			return err;
905 	}
906 	if ((attr->ia_valid & ATTR_UID &&
907 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
908 		(attr->ia_valid & ATTR_GID &&
909 		!gid_eq(attr->ia_gid, inode->i_gid))) {
910 		f2fs_lock_op(F2FS_I_SB(inode));
911 		err = dquot_transfer(inode, attr);
912 		if (err) {
913 			set_sbi_flag(F2FS_I_SB(inode),
914 					SBI_QUOTA_NEED_REPAIR);
915 			f2fs_unlock_op(F2FS_I_SB(inode));
916 			return err;
917 		}
918 		/*
919 		 * update uid/gid under lock_op(), so that dquot and inode can
920 		 * be updated atomically.
921 		 */
922 		if (attr->ia_valid & ATTR_UID)
923 			inode->i_uid = attr->ia_uid;
924 		if (attr->ia_valid & ATTR_GID)
925 			inode->i_gid = attr->ia_gid;
926 		f2fs_mark_inode_dirty_sync(inode, true);
927 		f2fs_unlock_op(F2FS_I_SB(inode));
928 	}
929 
930 	if (attr->ia_valid & ATTR_SIZE) {
931 		loff_t old_size = i_size_read(inode);
932 
933 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
934 			/*
935 			 * should convert inline inode before i_size_write to
936 			 * keep smaller than inline_data size with inline flag.
937 			 */
938 			err = f2fs_convert_inline_inode(inode);
939 			if (err)
940 				return err;
941 		}
942 
943 		down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
944 		down_write(&F2FS_I(inode)->i_mmap_sem);
945 
946 		truncate_setsize(inode, attr->ia_size);
947 
948 		if (attr->ia_size <= old_size)
949 			err = f2fs_truncate(inode);
950 		/*
951 		 * do not trim all blocks after i_size if target size is
952 		 * larger than i_size.
953 		 */
954 		up_write(&F2FS_I(inode)->i_mmap_sem);
955 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
956 		if (err)
957 			return err;
958 
959 		spin_lock(&F2FS_I(inode)->i_size_lock);
960 		inode->i_mtime = inode->i_ctime = current_time(inode);
961 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
962 		spin_unlock(&F2FS_I(inode)->i_size_lock);
963 	}
964 
965 	__setattr_copy(&init_user_ns, inode, attr);
966 
967 	if (attr->ia_valid & ATTR_MODE) {
968 		err = posix_acl_chmod(&init_user_ns, inode, f2fs_get_inode_mode(inode));
969 
970 		if (is_inode_flag_set(inode, FI_ACL_MODE)) {
971 			if (!err)
972 				inode->i_mode = F2FS_I(inode)->i_acl_mode;
973 			clear_inode_flag(inode, FI_ACL_MODE);
974 		}
975 	}
976 
977 	/* file size may changed here */
978 	f2fs_mark_inode_dirty_sync(inode, true);
979 
980 	/* inode change will produce dirty node pages flushed by checkpoint */
981 	f2fs_balance_fs(F2FS_I_SB(inode), true);
982 
983 	return err;
984 }
985 
986 const struct inode_operations f2fs_file_inode_operations = {
987 	.getattr	= f2fs_getattr,
988 	.setattr	= f2fs_setattr,
989 	.get_acl	= f2fs_get_acl,
990 	.set_acl	= f2fs_set_acl,
991 	.listxattr	= f2fs_listxattr,
992 	.fiemap		= f2fs_fiemap,
993 };
994 
995 static int fill_zero(struct inode *inode, pgoff_t index,
996 					loff_t start, loff_t len)
997 {
998 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
999 	struct page *page;
1000 
1001 	if (!len)
1002 		return 0;
1003 
1004 	f2fs_balance_fs(sbi, true);
1005 
1006 	f2fs_lock_op(sbi);
1007 	page = f2fs_get_new_data_page(inode, NULL, index, false);
1008 	f2fs_unlock_op(sbi);
1009 
1010 	if (IS_ERR(page))
1011 		return PTR_ERR(page);
1012 
1013 	f2fs_wait_on_page_writeback(page, DATA, true, true);
1014 	zero_user(page, start, len);
1015 	set_page_dirty(page);
1016 	f2fs_put_page(page, 1);
1017 	return 0;
1018 }
1019 
1020 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1021 {
1022 	int err;
1023 
1024 	while (pg_start < pg_end) {
1025 		struct dnode_of_data dn;
1026 		pgoff_t end_offset, count;
1027 
1028 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1029 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1030 		if (err) {
1031 			if (err == -ENOENT) {
1032 				pg_start = f2fs_get_next_page_offset(&dn,
1033 								pg_start);
1034 				continue;
1035 			}
1036 			return err;
1037 		}
1038 
1039 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1040 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1041 
1042 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1043 
1044 		f2fs_truncate_data_blocks_range(&dn, count);
1045 		f2fs_put_dnode(&dn);
1046 
1047 		pg_start += count;
1048 	}
1049 	return 0;
1050 }
1051 
1052 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1053 {
1054 	pgoff_t pg_start, pg_end;
1055 	loff_t off_start, off_end;
1056 	int ret;
1057 
1058 	ret = f2fs_convert_inline_inode(inode);
1059 	if (ret)
1060 		return ret;
1061 
1062 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1063 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1064 
1065 	off_start = offset & (PAGE_SIZE - 1);
1066 	off_end = (offset + len) & (PAGE_SIZE - 1);
1067 
1068 	if (pg_start == pg_end) {
1069 		ret = fill_zero(inode, pg_start, off_start,
1070 						off_end - off_start);
1071 		if (ret)
1072 			return ret;
1073 	} else {
1074 		if (off_start) {
1075 			ret = fill_zero(inode, pg_start++, off_start,
1076 						PAGE_SIZE - off_start);
1077 			if (ret)
1078 				return ret;
1079 		}
1080 		if (off_end) {
1081 			ret = fill_zero(inode, pg_end, 0, off_end);
1082 			if (ret)
1083 				return ret;
1084 		}
1085 
1086 		if (pg_start < pg_end) {
1087 			struct address_space *mapping = inode->i_mapping;
1088 			loff_t blk_start, blk_end;
1089 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1090 
1091 			f2fs_balance_fs(sbi, true);
1092 
1093 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1094 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1095 
1096 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1097 			down_write(&F2FS_I(inode)->i_mmap_sem);
1098 
1099 			truncate_inode_pages_range(mapping, blk_start,
1100 					blk_end - 1);
1101 
1102 			f2fs_lock_op(sbi);
1103 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1104 			f2fs_unlock_op(sbi);
1105 
1106 			up_write(&F2FS_I(inode)->i_mmap_sem);
1107 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1108 		}
1109 	}
1110 
1111 	return ret;
1112 }
1113 
1114 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1115 				int *do_replace, pgoff_t off, pgoff_t len)
1116 {
1117 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1118 	struct dnode_of_data dn;
1119 	int ret, done, i;
1120 
1121 next_dnode:
1122 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1123 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1124 	if (ret && ret != -ENOENT) {
1125 		return ret;
1126 	} else if (ret == -ENOENT) {
1127 		if (dn.max_level == 0)
1128 			return -ENOENT;
1129 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1130 						dn.ofs_in_node, len);
1131 		blkaddr += done;
1132 		do_replace += done;
1133 		goto next;
1134 	}
1135 
1136 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1137 							dn.ofs_in_node, len);
1138 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1139 		*blkaddr = f2fs_data_blkaddr(&dn);
1140 
1141 		if (__is_valid_data_blkaddr(*blkaddr) &&
1142 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1143 					DATA_GENERIC_ENHANCE)) {
1144 			f2fs_put_dnode(&dn);
1145 			return -EFSCORRUPTED;
1146 		}
1147 
1148 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1149 
1150 			if (f2fs_lfs_mode(sbi)) {
1151 				f2fs_put_dnode(&dn);
1152 				return -EOPNOTSUPP;
1153 			}
1154 
1155 			/* do not invalidate this block address */
1156 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1157 			*do_replace = 1;
1158 		}
1159 	}
1160 	f2fs_put_dnode(&dn);
1161 next:
1162 	len -= done;
1163 	off += done;
1164 	if (len)
1165 		goto next_dnode;
1166 	return 0;
1167 }
1168 
1169 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1170 				int *do_replace, pgoff_t off, int len)
1171 {
1172 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1173 	struct dnode_of_data dn;
1174 	int ret, i;
1175 
1176 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1177 		if (*do_replace == 0)
1178 			continue;
1179 
1180 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1181 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1182 		if (ret) {
1183 			dec_valid_block_count(sbi, inode, 1);
1184 			f2fs_invalidate_blocks(sbi, *blkaddr);
1185 		} else {
1186 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1187 		}
1188 		f2fs_put_dnode(&dn);
1189 	}
1190 	return 0;
1191 }
1192 
1193 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1194 			block_t *blkaddr, int *do_replace,
1195 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1196 {
1197 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1198 	pgoff_t i = 0;
1199 	int ret;
1200 
1201 	while (i < len) {
1202 		if (blkaddr[i] == NULL_ADDR && !full) {
1203 			i++;
1204 			continue;
1205 		}
1206 
1207 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1208 			struct dnode_of_data dn;
1209 			struct node_info ni;
1210 			size_t new_size;
1211 			pgoff_t ilen;
1212 
1213 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1214 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1215 			if (ret)
1216 				return ret;
1217 
1218 			ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1219 			if (ret) {
1220 				f2fs_put_dnode(&dn);
1221 				return ret;
1222 			}
1223 
1224 			ilen = min((pgoff_t)
1225 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1226 						dn.ofs_in_node, len - i);
1227 			do {
1228 				dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1229 				f2fs_truncate_data_blocks_range(&dn, 1);
1230 
1231 				if (do_replace[i]) {
1232 					f2fs_i_blocks_write(src_inode,
1233 							1, false, false);
1234 					f2fs_i_blocks_write(dst_inode,
1235 							1, true, false);
1236 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1237 					blkaddr[i], ni.version, true, false);
1238 
1239 					do_replace[i] = 0;
1240 				}
1241 				dn.ofs_in_node++;
1242 				i++;
1243 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1244 				if (dst_inode->i_size < new_size)
1245 					f2fs_i_size_write(dst_inode, new_size);
1246 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1247 
1248 			f2fs_put_dnode(&dn);
1249 		} else {
1250 			struct page *psrc, *pdst;
1251 
1252 			psrc = f2fs_get_lock_data_page(src_inode,
1253 							src + i, true);
1254 			if (IS_ERR(psrc))
1255 				return PTR_ERR(psrc);
1256 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1257 								true);
1258 			if (IS_ERR(pdst)) {
1259 				f2fs_put_page(psrc, 1);
1260 				return PTR_ERR(pdst);
1261 			}
1262 			f2fs_copy_page(psrc, pdst);
1263 			set_page_dirty(pdst);
1264 			f2fs_put_page(pdst, 1);
1265 			f2fs_put_page(psrc, 1);
1266 
1267 			ret = f2fs_truncate_hole(src_inode,
1268 						src + i, src + i + 1);
1269 			if (ret)
1270 				return ret;
1271 			i++;
1272 		}
1273 	}
1274 	return 0;
1275 }
1276 
1277 static int __exchange_data_block(struct inode *src_inode,
1278 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1279 			pgoff_t len, bool full)
1280 {
1281 	block_t *src_blkaddr;
1282 	int *do_replace;
1283 	pgoff_t olen;
1284 	int ret;
1285 
1286 	while (len) {
1287 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1288 
1289 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1290 					array_size(olen, sizeof(block_t)),
1291 					GFP_NOFS);
1292 		if (!src_blkaddr)
1293 			return -ENOMEM;
1294 
1295 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1296 					array_size(olen, sizeof(int)),
1297 					GFP_NOFS);
1298 		if (!do_replace) {
1299 			kvfree(src_blkaddr);
1300 			return -ENOMEM;
1301 		}
1302 
1303 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1304 					do_replace, src, olen);
1305 		if (ret)
1306 			goto roll_back;
1307 
1308 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1309 					do_replace, src, dst, olen, full);
1310 		if (ret)
1311 			goto roll_back;
1312 
1313 		src += olen;
1314 		dst += olen;
1315 		len -= olen;
1316 
1317 		kvfree(src_blkaddr);
1318 		kvfree(do_replace);
1319 	}
1320 	return 0;
1321 
1322 roll_back:
1323 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1324 	kvfree(src_blkaddr);
1325 	kvfree(do_replace);
1326 	return ret;
1327 }
1328 
1329 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1330 {
1331 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1332 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1333 	pgoff_t start = offset >> PAGE_SHIFT;
1334 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1335 	int ret;
1336 
1337 	f2fs_balance_fs(sbi, true);
1338 
1339 	/* avoid gc operation during block exchange */
1340 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1341 	down_write(&F2FS_I(inode)->i_mmap_sem);
1342 
1343 	f2fs_lock_op(sbi);
1344 	f2fs_drop_extent_tree(inode);
1345 	truncate_pagecache(inode, offset);
1346 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1347 	f2fs_unlock_op(sbi);
1348 
1349 	up_write(&F2FS_I(inode)->i_mmap_sem);
1350 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1351 	return ret;
1352 }
1353 
1354 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1355 {
1356 	loff_t new_size;
1357 	int ret;
1358 
1359 	if (offset + len >= i_size_read(inode))
1360 		return -EINVAL;
1361 
1362 	/* collapse range should be aligned to block size of f2fs. */
1363 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1364 		return -EINVAL;
1365 
1366 	ret = f2fs_convert_inline_inode(inode);
1367 	if (ret)
1368 		return ret;
1369 
1370 	/* write out all dirty pages from offset */
1371 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1372 	if (ret)
1373 		return ret;
1374 
1375 	ret = f2fs_do_collapse(inode, offset, len);
1376 	if (ret)
1377 		return ret;
1378 
1379 	/* write out all moved pages, if possible */
1380 	down_write(&F2FS_I(inode)->i_mmap_sem);
1381 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1382 	truncate_pagecache(inode, offset);
1383 
1384 	new_size = i_size_read(inode) - len;
1385 	ret = f2fs_truncate_blocks(inode, new_size, true);
1386 	up_write(&F2FS_I(inode)->i_mmap_sem);
1387 	if (!ret)
1388 		f2fs_i_size_write(inode, new_size);
1389 	return ret;
1390 }
1391 
1392 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1393 								pgoff_t end)
1394 {
1395 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1396 	pgoff_t index = start;
1397 	unsigned int ofs_in_node = dn->ofs_in_node;
1398 	blkcnt_t count = 0;
1399 	int ret;
1400 
1401 	for (; index < end; index++, dn->ofs_in_node++) {
1402 		if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1403 			count++;
1404 	}
1405 
1406 	dn->ofs_in_node = ofs_in_node;
1407 	ret = f2fs_reserve_new_blocks(dn, count);
1408 	if (ret)
1409 		return ret;
1410 
1411 	dn->ofs_in_node = ofs_in_node;
1412 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1413 		dn->data_blkaddr = f2fs_data_blkaddr(dn);
1414 		/*
1415 		 * f2fs_reserve_new_blocks will not guarantee entire block
1416 		 * allocation.
1417 		 */
1418 		if (dn->data_blkaddr == NULL_ADDR) {
1419 			ret = -ENOSPC;
1420 			break;
1421 		}
1422 		if (dn->data_blkaddr != NEW_ADDR) {
1423 			f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1424 			dn->data_blkaddr = NEW_ADDR;
1425 			f2fs_set_data_blkaddr(dn);
1426 		}
1427 	}
1428 
1429 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1430 
1431 	return ret;
1432 }
1433 
1434 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1435 								int mode)
1436 {
1437 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1438 	struct address_space *mapping = inode->i_mapping;
1439 	pgoff_t index, pg_start, pg_end;
1440 	loff_t new_size = i_size_read(inode);
1441 	loff_t off_start, off_end;
1442 	int ret = 0;
1443 
1444 	ret = inode_newsize_ok(inode, (len + offset));
1445 	if (ret)
1446 		return ret;
1447 
1448 	ret = f2fs_convert_inline_inode(inode);
1449 	if (ret)
1450 		return ret;
1451 
1452 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1453 	if (ret)
1454 		return ret;
1455 
1456 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1457 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1458 
1459 	off_start = offset & (PAGE_SIZE - 1);
1460 	off_end = (offset + len) & (PAGE_SIZE - 1);
1461 
1462 	if (pg_start == pg_end) {
1463 		ret = fill_zero(inode, pg_start, off_start,
1464 						off_end - off_start);
1465 		if (ret)
1466 			return ret;
1467 
1468 		new_size = max_t(loff_t, new_size, offset + len);
1469 	} else {
1470 		if (off_start) {
1471 			ret = fill_zero(inode, pg_start++, off_start,
1472 						PAGE_SIZE - off_start);
1473 			if (ret)
1474 				return ret;
1475 
1476 			new_size = max_t(loff_t, new_size,
1477 					(loff_t)pg_start << PAGE_SHIFT);
1478 		}
1479 
1480 		for (index = pg_start; index < pg_end;) {
1481 			struct dnode_of_data dn;
1482 			unsigned int end_offset;
1483 			pgoff_t end;
1484 
1485 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1486 			down_write(&F2FS_I(inode)->i_mmap_sem);
1487 
1488 			truncate_pagecache_range(inode,
1489 				(loff_t)index << PAGE_SHIFT,
1490 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1491 
1492 			f2fs_lock_op(sbi);
1493 
1494 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1495 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1496 			if (ret) {
1497 				f2fs_unlock_op(sbi);
1498 				up_write(&F2FS_I(inode)->i_mmap_sem);
1499 				up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1500 				goto out;
1501 			}
1502 
1503 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1504 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1505 
1506 			ret = f2fs_do_zero_range(&dn, index, end);
1507 			f2fs_put_dnode(&dn);
1508 
1509 			f2fs_unlock_op(sbi);
1510 			up_write(&F2FS_I(inode)->i_mmap_sem);
1511 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1512 
1513 			f2fs_balance_fs(sbi, dn.node_changed);
1514 
1515 			if (ret)
1516 				goto out;
1517 
1518 			index = end;
1519 			new_size = max_t(loff_t, new_size,
1520 					(loff_t)index << PAGE_SHIFT);
1521 		}
1522 
1523 		if (off_end) {
1524 			ret = fill_zero(inode, pg_end, 0, off_end);
1525 			if (ret)
1526 				goto out;
1527 
1528 			new_size = max_t(loff_t, new_size, offset + len);
1529 		}
1530 	}
1531 
1532 out:
1533 	if (new_size > i_size_read(inode)) {
1534 		if (mode & FALLOC_FL_KEEP_SIZE)
1535 			file_set_keep_isize(inode);
1536 		else
1537 			f2fs_i_size_write(inode, new_size);
1538 	}
1539 	return ret;
1540 }
1541 
1542 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1543 {
1544 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1545 	pgoff_t nr, pg_start, pg_end, delta, idx;
1546 	loff_t new_size;
1547 	int ret = 0;
1548 
1549 	new_size = i_size_read(inode) + len;
1550 	ret = inode_newsize_ok(inode, new_size);
1551 	if (ret)
1552 		return ret;
1553 
1554 	if (offset >= i_size_read(inode))
1555 		return -EINVAL;
1556 
1557 	/* insert range should be aligned to block size of f2fs. */
1558 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1559 		return -EINVAL;
1560 
1561 	ret = f2fs_convert_inline_inode(inode);
1562 	if (ret)
1563 		return ret;
1564 
1565 	f2fs_balance_fs(sbi, true);
1566 
1567 	down_write(&F2FS_I(inode)->i_mmap_sem);
1568 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1569 	up_write(&F2FS_I(inode)->i_mmap_sem);
1570 	if (ret)
1571 		return ret;
1572 
1573 	/* write out all dirty pages from offset */
1574 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1575 	if (ret)
1576 		return ret;
1577 
1578 	pg_start = offset >> PAGE_SHIFT;
1579 	pg_end = (offset + len) >> PAGE_SHIFT;
1580 	delta = pg_end - pg_start;
1581 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1582 
1583 	/* avoid gc operation during block exchange */
1584 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1585 	down_write(&F2FS_I(inode)->i_mmap_sem);
1586 	truncate_pagecache(inode, offset);
1587 
1588 	while (!ret && idx > pg_start) {
1589 		nr = idx - pg_start;
1590 		if (nr > delta)
1591 			nr = delta;
1592 		idx -= nr;
1593 
1594 		f2fs_lock_op(sbi);
1595 		f2fs_drop_extent_tree(inode);
1596 
1597 		ret = __exchange_data_block(inode, inode, idx,
1598 					idx + delta, nr, false);
1599 		f2fs_unlock_op(sbi);
1600 	}
1601 	up_write(&F2FS_I(inode)->i_mmap_sem);
1602 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1603 
1604 	/* write out all moved pages, if possible */
1605 	down_write(&F2FS_I(inode)->i_mmap_sem);
1606 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1607 	truncate_pagecache(inode, offset);
1608 	up_write(&F2FS_I(inode)->i_mmap_sem);
1609 
1610 	if (!ret)
1611 		f2fs_i_size_write(inode, new_size);
1612 	return ret;
1613 }
1614 
1615 static int expand_inode_data(struct inode *inode, loff_t offset,
1616 					loff_t len, int mode)
1617 {
1618 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1619 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1620 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1621 			.m_may_create = true };
1622 	pgoff_t pg_end;
1623 	loff_t new_size = i_size_read(inode);
1624 	loff_t off_end;
1625 	int err;
1626 
1627 	err = inode_newsize_ok(inode, (len + offset));
1628 	if (err)
1629 		return err;
1630 
1631 	err = f2fs_convert_inline_inode(inode);
1632 	if (err)
1633 		return err;
1634 
1635 	f2fs_balance_fs(sbi, true);
1636 
1637 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1638 	off_end = (offset + len) & (PAGE_SIZE - 1);
1639 
1640 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1641 	map.m_len = pg_end - map.m_lblk;
1642 	if (off_end)
1643 		map.m_len++;
1644 
1645 	if (!map.m_len)
1646 		return 0;
1647 
1648 	if (f2fs_is_pinned_file(inode)) {
1649 		block_t len = (map.m_len >> sbi->log_blocks_per_seg) <<
1650 					sbi->log_blocks_per_seg;
1651 		block_t done = 0;
1652 
1653 		if (map.m_len % sbi->blocks_per_seg)
1654 			len += sbi->blocks_per_seg;
1655 
1656 		map.m_len = sbi->blocks_per_seg;
1657 next_alloc:
1658 		if (has_not_enough_free_secs(sbi, 0,
1659 			GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1660 			down_write(&sbi->gc_lock);
1661 			err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1662 			if (err && err != -ENODATA && err != -EAGAIN)
1663 				goto out_err;
1664 		}
1665 
1666 		down_write(&sbi->pin_sem);
1667 
1668 		f2fs_lock_op(sbi);
1669 		f2fs_allocate_new_segment(sbi, CURSEG_COLD_DATA_PINNED);
1670 		f2fs_unlock_op(sbi);
1671 
1672 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1673 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1674 
1675 		up_write(&sbi->pin_sem);
1676 
1677 		done += map.m_len;
1678 		len -= map.m_len;
1679 		map.m_lblk += map.m_len;
1680 		if (!err && len)
1681 			goto next_alloc;
1682 
1683 		map.m_len = done;
1684 	} else {
1685 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1686 	}
1687 out_err:
1688 	if (err) {
1689 		pgoff_t last_off;
1690 
1691 		if (!map.m_len)
1692 			return err;
1693 
1694 		last_off = map.m_lblk + map.m_len - 1;
1695 
1696 		/* update new size to the failed position */
1697 		new_size = (last_off == pg_end) ? offset + len :
1698 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1699 	} else {
1700 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1701 	}
1702 
1703 	if (new_size > i_size_read(inode)) {
1704 		if (mode & FALLOC_FL_KEEP_SIZE)
1705 			file_set_keep_isize(inode);
1706 		else
1707 			f2fs_i_size_write(inode, new_size);
1708 	}
1709 
1710 	return err;
1711 }
1712 
1713 static long f2fs_fallocate(struct file *file, int mode,
1714 				loff_t offset, loff_t len)
1715 {
1716 	struct inode *inode = file_inode(file);
1717 	long ret = 0;
1718 
1719 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1720 		return -EIO;
1721 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1722 		return -ENOSPC;
1723 	if (!f2fs_is_compress_backend_ready(inode))
1724 		return -EOPNOTSUPP;
1725 
1726 	/* f2fs only support ->fallocate for regular file */
1727 	if (!S_ISREG(inode->i_mode))
1728 		return -EINVAL;
1729 
1730 	if (IS_ENCRYPTED(inode) &&
1731 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1732 		return -EOPNOTSUPP;
1733 
1734 	if (f2fs_compressed_file(inode) &&
1735 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1736 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1737 		return -EOPNOTSUPP;
1738 
1739 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1740 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1741 			FALLOC_FL_INSERT_RANGE))
1742 		return -EOPNOTSUPP;
1743 
1744 	inode_lock(inode);
1745 
1746 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1747 		if (offset >= inode->i_size)
1748 			goto out;
1749 
1750 		ret = punch_hole(inode, offset, len);
1751 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1752 		ret = f2fs_collapse_range(inode, offset, len);
1753 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1754 		ret = f2fs_zero_range(inode, offset, len, mode);
1755 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1756 		ret = f2fs_insert_range(inode, offset, len);
1757 	} else {
1758 		ret = expand_inode_data(inode, offset, len, mode);
1759 	}
1760 
1761 	if (!ret) {
1762 		inode->i_mtime = inode->i_ctime = current_time(inode);
1763 		f2fs_mark_inode_dirty_sync(inode, false);
1764 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1765 	}
1766 
1767 out:
1768 	inode_unlock(inode);
1769 
1770 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1771 	return ret;
1772 }
1773 
1774 static int f2fs_release_file(struct inode *inode, struct file *filp)
1775 {
1776 	/*
1777 	 * f2fs_relase_file is called at every close calls. So we should
1778 	 * not drop any inmemory pages by close called by other process.
1779 	 */
1780 	if (!(filp->f_mode & FMODE_WRITE) ||
1781 			atomic_read(&inode->i_writecount) != 1)
1782 		return 0;
1783 
1784 	/* some remained atomic pages should discarded */
1785 	if (f2fs_is_atomic_file(inode))
1786 		f2fs_drop_inmem_pages(inode);
1787 	if (f2fs_is_volatile_file(inode)) {
1788 		set_inode_flag(inode, FI_DROP_CACHE);
1789 		filemap_fdatawrite(inode->i_mapping);
1790 		clear_inode_flag(inode, FI_DROP_CACHE);
1791 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1792 		stat_dec_volatile_write(inode);
1793 	}
1794 	return 0;
1795 }
1796 
1797 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1798 {
1799 	struct inode *inode = file_inode(file);
1800 
1801 	/*
1802 	 * If the process doing a transaction is crashed, we should do
1803 	 * roll-back. Otherwise, other reader/write can see corrupted database
1804 	 * until all the writers close its file. Since this should be done
1805 	 * before dropping file lock, it needs to do in ->flush.
1806 	 */
1807 	if (f2fs_is_atomic_file(inode) &&
1808 			F2FS_I(inode)->inmem_task == current)
1809 		f2fs_drop_inmem_pages(inode);
1810 	return 0;
1811 }
1812 
1813 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1814 {
1815 	struct f2fs_inode_info *fi = F2FS_I(inode);
1816 	u32 masked_flags = fi->i_flags & mask;
1817 
1818 	f2fs_bug_on(F2FS_I_SB(inode), (iflags & ~mask));
1819 
1820 	/* Is it quota file? Do not allow user to mess with it */
1821 	if (IS_NOQUOTA(inode))
1822 		return -EPERM;
1823 
1824 	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1825 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1826 			return -EOPNOTSUPP;
1827 		if (!f2fs_empty_dir(inode))
1828 			return -ENOTEMPTY;
1829 	}
1830 
1831 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1832 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1833 			return -EOPNOTSUPP;
1834 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1835 			return -EINVAL;
1836 	}
1837 
1838 	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1839 		if (masked_flags & F2FS_COMPR_FL) {
1840 			if (!f2fs_disable_compressed_file(inode))
1841 				return -EINVAL;
1842 		}
1843 		if (iflags & F2FS_NOCOMP_FL)
1844 			return -EINVAL;
1845 		if (iflags & F2FS_COMPR_FL) {
1846 			if (!f2fs_may_compress(inode))
1847 				return -EINVAL;
1848 			if (S_ISREG(inode->i_mode) && inode->i_size)
1849 				return -EINVAL;
1850 
1851 			set_compress_context(inode);
1852 		}
1853 	}
1854 	if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1855 		if (masked_flags & F2FS_COMPR_FL)
1856 			return -EINVAL;
1857 	}
1858 
1859 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1860 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1861 					(fi->i_flags & F2FS_NOCOMP_FL));
1862 
1863 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1864 		set_inode_flag(inode, FI_PROJ_INHERIT);
1865 	else
1866 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1867 
1868 	inode->i_ctime = current_time(inode);
1869 	f2fs_set_inode_flags(inode);
1870 	f2fs_mark_inode_dirty_sync(inode, true);
1871 	return 0;
1872 }
1873 
1874 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1875 
1876 /*
1877  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1878  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1879  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1880  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1881  */
1882 
1883 static const struct {
1884 	u32 iflag;
1885 	u32 fsflag;
1886 } f2fs_fsflags_map[] = {
1887 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
1888 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1889 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1890 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1891 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1892 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1893 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
1894 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1895 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1896 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1897 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
1898 };
1899 
1900 #define F2FS_GETTABLE_FS_FL (		\
1901 		FS_COMPR_FL |		\
1902 		FS_SYNC_FL |		\
1903 		FS_IMMUTABLE_FL |	\
1904 		FS_APPEND_FL |		\
1905 		FS_NODUMP_FL |		\
1906 		FS_NOATIME_FL |		\
1907 		FS_NOCOMP_FL |		\
1908 		FS_INDEX_FL |		\
1909 		FS_DIRSYNC_FL |		\
1910 		FS_PROJINHERIT_FL |	\
1911 		FS_ENCRYPT_FL |		\
1912 		FS_INLINE_DATA_FL |	\
1913 		FS_NOCOW_FL |		\
1914 		FS_VERITY_FL |		\
1915 		FS_CASEFOLD_FL)
1916 
1917 #define F2FS_SETTABLE_FS_FL (		\
1918 		FS_COMPR_FL |		\
1919 		FS_SYNC_FL |		\
1920 		FS_IMMUTABLE_FL |	\
1921 		FS_APPEND_FL |		\
1922 		FS_NODUMP_FL |		\
1923 		FS_NOATIME_FL |		\
1924 		FS_NOCOMP_FL |		\
1925 		FS_DIRSYNC_FL |		\
1926 		FS_PROJINHERIT_FL |	\
1927 		FS_CASEFOLD_FL)
1928 
1929 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1930 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1931 {
1932 	u32 fsflags = 0;
1933 	int i;
1934 
1935 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1936 		if (iflags & f2fs_fsflags_map[i].iflag)
1937 			fsflags |= f2fs_fsflags_map[i].fsflag;
1938 
1939 	return fsflags;
1940 }
1941 
1942 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1943 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1944 {
1945 	u32 iflags = 0;
1946 	int i;
1947 
1948 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1949 		if (fsflags & f2fs_fsflags_map[i].fsflag)
1950 			iflags |= f2fs_fsflags_map[i].iflag;
1951 
1952 	return iflags;
1953 }
1954 
1955 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1956 {
1957 	struct inode *inode = file_inode(filp);
1958 	struct f2fs_inode_info *fi = F2FS_I(inode);
1959 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1960 
1961 	if (IS_ENCRYPTED(inode))
1962 		fsflags |= FS_ENCRYPT_FL;
1963 	if (IS_VERITY(inode))
1964 		fsflags |= FS_VERITY_FL;
1965 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1966 		fsflags |= FS_INLINE_DATA_FL;
1967 	if (is_inode_flag_set(inode, FI_PIN_FILE))
1968 		fsflags |= FS_NOCOW_FL;
1969 
1970 	fsflags &= F2FS_GETTABLE_FS_FL;
1971 
1972 	return put_user(fsflags, (int __user *)arg);
1973 }
1974 
1975 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1976 {
1977 	struct inode *inode = file_inode(filp);
1978 	struct f2fs_inode_info *fi = F2FS_I(inode);
1979 	u32 fsflags, old_fsflags;
1980 	u32 iflags;
1981 	int ret;
1982 
1983 	if (!inode_owner_or_capable(&init_user_ns, inode))
1984 		return -EACCES;
1985 
1986 	if (get_user(fsflags, (int __user *)arg))
1987 		return -EFAULT;
1988 
1989 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
1990 		return -EOPNOTSUPP;
1991 	fsflags &= F2FS_SETTABLE_FS_FL;
1992 
1993 	iflags = f2fs_fsflags_to_iflags(fsflags);
1994 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1995 		return -EOPNOTSUPP;
1996 
1997 	ret = mnt_want_write_file(filp);
1998 	if (ret)
1999 		return ret;
2000 
2001 	inode_lock(inode);
2002 
2003 	old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
2004 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
2005 	if (ret)
2006 		goto out;
2007 
2008 	ret = f2fs_setflags_common(inode, iflags,
2009 			f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
2010 out:
2011 	inode_unlock(inode);
2012 	mnt_drop_write_file(filp);
2013 	return ret;
2014 }
2015 
2016 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2017 {
2018 	struct inode *inode = file_inode(filp);
2019 
2020 	return put_user(inode->i_generation, (int __user *)arg);
2021 }
2022 
2023 static int f2fs_ioc_start_atomic_write(struct file *filp)
2024 {
2025 	struct inode *inode = file_inode(filp);
2026 	struct f2fs_inode_info *fi = F2FS_I(inode);
2027 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2028 	int ret;
2029 
2030 	if (!inode_owner_or_capable(&init_user_ns, inode))
2031 		return -EACCES;
2032 
2033 	if (!S_ISREG(inode->i_mode))
2034 		return -EINVAL;
2035 
2036 	if (filp->f_flags & O_DIRECT)
2037 		return -EINVAL;
2038 
2039 	ret = mnt_want_write_file(filp);
2040 	if (ret)
2041 		return ret;
2042 
2043 	inode_lock(inode);
2044 
2045 	f2fs_disable_compressed_file(inode);
2046 
2047 	if (f2fs_is_atomic_file(inode)) {
2048 		if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2049 			ret = -EINVAL;
2050 		goto out;
2051 	}
2052 
2053 	ret = f2fs_convert_inline_inode(inode);
2054 	if (ret)
2055 		goto out;
2056 
2057 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2058 
2059 	/*
2060 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2061 	 * f2fs_is_atomic_file.
2062 	 */
2063 	if (get_dirty_pages(inode))
2064 		f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2065 			  inode->i_ino, get_dirty_pages(inode));
2066 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2067 	if (ret) {
2068 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2069 		goto out;
2070 	}
2071 
2072 	spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2073 	if (list_empty(&fi->inmem_ilist))
2074 		list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2075 	sbi->atomic_files++;
2076 	spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2077 
2078 	/* add inode in inmem_list first and set atomic_file */
2079 	set_inode_flag(inode, FI_ATOMIC_FILE);
2080 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2081 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2082 
2083 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2084 	F2FS_I(inode)->inmem_task = current;
2085 	stat_update_max_atomic_write(inode);
2086 out:
2087 	inode_unlock(inode);
2088 	mnt_drop_write_file(filp);
2089 	return ret;
2090 }
2091 
2092 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2093 {
2094 	struct inode *inode = file_inode(filp);
2095 	int ret;
2096 
2097 	if (!inode_owner_or_capable(&init_user_ns, inode))
2098 		return -EACCES;
2099 
2100 	ret = mnt_want_write_file(filp);
2101 	if (ret)
2102 		return ret;
2103 
2104 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2105 
2106 	inode_lock(inode);
2107 
2108 	if (f2fs_is_volatile_file(inode)) {
2109 		ret = -EINVAL;
2110 		goto err_out;
2111 	}
2112 
2113 	if (f2fs_is_atomic_file(inode)) {
2114 		ret = f2fs_commit_inmem_pages(inode);
2115 		if (ret)
2116 			goto err_out;
2117 
2118 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2119 		if (!ret)
2120 			f2fs_drop_inmem_pages(inode);
2121 	} else {
2122 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2123 	}
2124 err_out:
2125 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2126 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2127 		ret = -EINVAL;
2128 	}
2129 	inode_unlock(inode);
2130 	mnt_drop_write_file(filp);
2131 	return ret;
2132 }
2133 
2134 static int f2fs_ioc_start_volatile_write(struct file *filp)
2135 {
2136 	struct inode *inode = file_inode(filp);
2137 	int ret;
2138 
2139 	if (!inode_owner_or_capable(&init_user_ns, inode))
2140 		return -EACCES;
2141 
2142 	if (!S_ISREG(inode->i_mode))
2143 		return -EINVAL;
2144 
2145 	ret = mnt_want_write_file(filp);
2146 	if (ret)
2147 		return ret;
2148 
2149 	inode_lock(inode);
2150 
2151 	if (f2fs_is_volatile_file(inode))
2152 		goto out;
2153 
2154 	ret = f2fs_convert_inline_inode(inode);
2155 	if (ret)
2156 		goto out;
2157 
2158 	stat_inc_volatile_write(inode);
2159 	stat_update_max_volatile_write(inode);
2160 
2161 	set_inode_flag(inode, FI_VOLATILE_FILE);
2162 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2163 out:
2164 	inode_unlock(inode);
2165 	mnt_drop_write_file(filp);
2166 	return ret;
2167 }
2168 
2169 static int f2fs_ioc_release_volatile_write(struct file *filp)
2170 {
2171 	struct inode *inode = file_inode(filp);
2172 	int ret;
2173 
2174 	if (!inode_owner_or_capable(&init_user_ns, inode))
2175 		return -EACCES;
2176 
2177 	ret = mnt_want_write_file(filp);
2178 	if (ret)
2179 		return ret;
2180 
2181 	inode_lock(inode);
2182 
2183 	if (!f2fs_is_volatile_file(inode))
2184 		goto out;
2185 
2186 	if (!f2fs_is_first_block_written(inode)) {
2187 		ret = truncate_partial_data_page(inode, 0, true);
2188 		goto out;
2189 	}
2190 
2191 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2192 out:
2193 	inode_unlock(inode);
2194 	mnt_drop_write_file(filp);
2195 	return ret;
2196 }
2197 
2198 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2199 {
2200 	struct inode *inode = file_inode(filp);
2201 	int ret;
2202 
2203 	if (!inode_owner_or_capable(&init_user_ns, inode))
2204 		return -EACCES;
2205 
2206 	ret = mnt_want_write_file(filp);
2207 	if (ret)
2208 		return ret;
2209 
2210 	inode_lock(inode);
2211 
2212 	if (f2fs_is_atomic_file(inode))
2213 		f2fs_drop_inmem_pages(inode);
2214 	if (f2fs_is_volatile_file(inode)) {
2215 		clear_inode_flag(inode, FI_VOLATILE_FILE);
2216 		stat_dec_volatile_write(inode);
2217 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2218 	}
2219 
2220 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2221 
2222 	inode_unlock(inode);
2223 
2224 	mnt_drop_write_file(filp);
2225 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2226 	return ret;
2227 }
2228 
2229 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2230 {
2231 	struct inode *inode = file_inode(filp);
2232 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2233 	struct super_block *sb = sbi->sb;
2234 	__u32 in;
2235 	int ret = 0;
2236 
2237 	if (!capable(CAP_SYS_ADMIN))
2238 		return -EPERM;
2239 
2240 	if (get_user(in, (__u32 __user *)arg))
2241 		return -EFAULT;
2242 
2243 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2244 		ret = mnt_want_write_file(filp);
2245 		if (ret) {
2246 			if (ret == -EROFS) {
2247 				ret = 0;
2248 				f2fs_stop_checkpoint(sbi, false);
2249 				set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2250 				trace_f2fs_shutdown(sbi, in, ret);
2251 			}
2252 			return ret;
2253 		}
2254 	}
2255 
2256 	switch (in) {
2257 	case F2FS_GOING_DOWN_FULLSYNC:
2258 		ret = freeze_bdev(sb->s_bdev);
2259 		if (ret)
2260 			goto out;
2261 		f2fs_stop_checkpoint(sbi, false);
2262 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2263 		thaw_bdev(sb->s_bdev);
2264 		break;
2265 	case F2FS_GOING_DOWN_METASYNC:
2266 		/* do checkpoint only */
2267 		ret = f2fs_sync_fs(sb, 1);
2268 		if (ret)
2269 			goto out;
2270 		f2fs_stop_checkpoint(sbi, false);
2271 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2272 		break;
2273 	case F2FS_GOING_DOWN_NOSYNC:
2274 		f2fs_stop_checkpoint(sbi, false);
2275 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2276 		break;
2277 	case F2FS_GOING_DOWN_METAFLUSH:
2278 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2279 		f2fs_stop_checkpoint(sbi, false);
2280 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2281 		break;
2282 	case F2FS_GOING_DOWN_NEED_FSCK:
2283 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2284 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2285 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2286 		/* do checkpoint only */
2287 		ret = f2fs_sync_fs(sb, 1);
2288 		goto out;
2289 	default:
2290 		ret = -EINVAL;
2291 		goto out;
2292 	}
2293 
2294 	f2fs_stop_gc_thread(sbi);
2295 	f2fs_stop_discard_thread(sbi);
2296 
2297 	f2fs_drop_discard_cmd(sbi);
2298 	clear_opt(sbi, DISCARD);
2299 
2300 	f2fs_update_time(sbi, REQ_TIME);
2301 out:
2302 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2303 		mnt_drop_write_file(filp);
2304 
2305 	trace_f2fs_shutdown(sbi, in, ret);
2306 
2307 	return ret;
2308 }
2309 
2310 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2311 {
2312 	struct inode *inode = file_inode(filp);
2313 	struct super_block *sb = inode->i_sb;
2314 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
2315 	struct fstrim_range range;
2316 	int ret;
2317 
2318 	if (!capable(CAP_SYS_ADMIN))
2319 		return -EPERM;
2320 
2321 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2322 		return -EOPNOTSUPP;
2323 
2324 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2325 				sizeof(range)))
2326 		return -EFAULT;
2327 
2328 	ret = mnt_want_write_file(filp);
2329 	if (ret)
2330 		return ret;
2331 
2332 	range.minlen = max((unsigned int)range.minlen,
2333 				q->limits.discard_granularity);
2334 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2335 	mnt_drop_write_file(filp);
2336 	if (ret < 0)
2337 		return ret;
2338 
2339 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2340 				sizeof(range)))
2341 		return -EFAULT;
2342 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2343 	return 0;
2344 }
2345 
2346 static bool uuid_is_nonzero(__u8 u[16])
2347 {
2348 	int i;
2349 
2350 	for (i = 0; i < 16; i++)
2351 		if (u[i])
2352 			return true;
2353 	return false;
2354 }
2355 
2356 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2357 {
2358 	struct inode *inode = file_inode(filp);
2359 
2360 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2361 		return -EOPNOTSUPP;
2362 
2363 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2364 
2365 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2366 }
2367 
2368 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2369 {
2370 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2371 		return -EOPNOTSUPP;
2372 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2373 }
2374 
2375 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2376 {
2377 	struct inode *inode = file_inode(filp);
2378 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2379 	int err;
2380 
2381 	if (!f2fs_sb_has_encrypt(sbi))
2382 		return -EOPNOTSUPP;
2383 
2384 	err = mnt_want_write_file(filp);
2385 	if (err)
2386 		return err;
2387 
2388 	down_write(&sbi->sb_lock);
2389 
2390 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2391 		goto got_it;
2392 
2393 	/* update superblock with uuid */
2394 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2395 
2396 	err = f2fs_commit_super(sbi, false);
2397 	if (err) {
2398 		/* undo new data */
2399 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2400 		goto out_err;
2401 	}
2402 got_it:
2403 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2404 									16))
2405 		err = -EFAULT;
2406 out_err:
2407 	up_write(&sbi->sb_lock);
2408 	mnt_drop_write_file(filp);
2409 	return err;
2410 }
2411 
2412 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2413 					     unsigned long arg)
2414 {
2415 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2416 		return -EOPNOTSUPP;
2417 
2418 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2419 }
2420 
2421 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2422 {
2423 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2424 		return -EOPNOTSUPP;
2425 
2426 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2427 }
2428 
2429 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2430 {
2431 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2432 		return -EOPNOTSUPP;
2433 
2434 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2435 }
2436 
2437 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2438 						    unsigned long arg)
2439 {
2440 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2441 		return -EOPNOTSUPP;
2442 
2443 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2444 }
2445 
2446 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2447 					      unsigned long arg)
2448 {
2449 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2450 		return -EOPNOTSUPP;
2451 
2452 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2453 }
2454 
2455 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2456 {
2457 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2458 		return -EOPNOTSUPP;
2459 
2460 	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2461 }
2462 
2463 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2464 {
2465 	struct inode *inode = file_inode(filp);
2466 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2467 	__u32 sync;
2468 	int ret;
2469 
2470 	if (!capable(CAP_SYS_ADMIN))
2471 		return -EPERM;
2472 
2473 	if (get_user(sync, (__u32 __user *)arg))
2474 		return -EFAULT;
2475 
2476 	if (f2fs_readonly(sbi->sb))
2477 		return -EROFS;
2478 
2479 	ret = mnt_want_write_file(filp);
2480 	if (ret)
2481 		return ret;
2482 
2483 	if (!sync) {
2484 		if (!down_write_trylock(&sbi->gc_lock)) {
2485 			ret = -EBUSY;
2486 			goto out;
2487 		}
2488 	} else {
2489 		down_write(&sbi->gc_lock);
2490 	}
2491 
2492 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2493 out:
2494 	mnt_drop_write_file(filp);
2495 	return ret;
2496 }
2497 
2498 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2499 {
2500 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2501 	u64 end;
2502 	int ret;
2503 
2504 	if (!capable(CAP_SYS_ADMIN))
2505 		return -EPERM;
2506 	if (f2fs_readonly(sbi->sb))
2507 		return -EROFS;
2508 
2509 	end = range->start + range->len;
2510 	if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2511 					end >= MAX_BLKADDR(sbi))
2512 		return -EINVAL;
2513 
2514 	ret = mnt_want_write_file(filp);
2515 	if (ret)
2516 		return ret;
2517 
2518 do_more:
2519 	if (!range->sync) {
2520 		if (!down_write_trylock(&sbi->gc_lock)) {
2521 			ret = -EBUSY;
2522 			goto out;
2523 		}
2524 	} else {
2525 		down_write(&sbi->gc_lock);
2526 	}
2527 
2528 	ret = f2fs_gc(sbi, range->sync, true, GET_SEGNO(sbi, range->start));
2529 	if (ret) {
2530 		if (ret == -EBUSY)
2531 			ret = -EAGAIN;
2532 		goto out;
2533 	}
2534 	range->start += BLKS_PER_SEC(sbi);
2535 	if (range->start <= end)
2536 		goto do_more;
2537 out:
2538 	mnt_drop_write_file(filp);
2539 	return ret;
2540 }
2541 
2542 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2543 {
2544 	struct f2fs_gc_range range;
2545 
2546 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2547 							sizeof(range)))
2548 		return -EFAULT;
2549 	return __f2fs_ioc_gc_range(filp, &range);
2550 }
2551 
2552 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2553 {
2554 	struct inode *inode = file_inode(filp);
2555 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2556 	int ret;
2557 
2558 	if (!capable(CAP_SYS_ADMIN))
2559 		return -EPERM;
2560 
2561 	if (f2fs_readonly(sbi->sb))
2562 		return -EROFS;
2563 
2564 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2565 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2566 		return -EINVAL;
2567 	}
2568 
2569 	ret = mnt_want_write_file(filp);
2570 	if (ret)
2571 		return ret;
2572 
2573 	ret = f2fs_sync_fs(sbi->sb, 1);
2574 
2575 	mnt_drop_write_file(filp);
2576 	return ret;
2577 }
2578 
2579 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2580 					struct file *filp,
2581 					struct f2fs_defragment *range)
2582 {
2583 	struct inode *inode = file_inode(filp);
2584 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2585 					.m_seg_type = NO_CHECK_TYPE ,
2586 					.m_may_create = false };
2587 	struct extent_info ei = {0, 0, 0};
2588 	pgoff_t pg_start, pg_end, next_pgofs;
2589 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2590 	unsigned int total = 0, sec_num;
2591 	block_t blk_end = 0;
2592 	bool fragmented = false;
2593 	int err;
2594 
2595 	/* if in-place-update policy is enabled, don't waste time here */
2596 	if (f2fs_should_update_inplace(inode, NULL))
2597 		return -EINVAL;
2598 
2599 	pg_start = range->start >> PAGE_SHIFT;
2600 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2601 
2602 	f2fs_balance_fs(sbi, true);
2603 
2604 	inode_lock(inode);
2605 
2606 	/* writeback all dirty pages in the range */
2607 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2608 						range->start + range->len - 1);
2609 	if (err)
2610 		goto out;
2611 
2612 	/*
2613 	 * lookup mapping info in extent cache, skip defragmenting if physical
2614 	 * block addresses are continuous.
2615 	 */
2616 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2617 		if (ei.fofs + ei.len >= pg_end)
2618 			goto out;
2619 	}
2620 
2621 	map.m_lblk = pg_start;
2622 	map.m_next_pgofs = &next_pgofs;
2623 
2624 	/*
2625 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2626 	 * physical block addresses are continuous even if there are hole(s)
2627 	 * in logical blocks.
2628 	 */
2629 	while (map.m_lblk < pg_end) {
2630 		map.m_len = pg_end - map.m_lblk;
2631 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2632 		if (err)
2633 			goto out;
2634 
2635 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2636 			map.m_lblk = next_pgofs;
2637 			continue;
2638 		}
2639 
2640 		if (blk_end && blk_end != map.m_pblk)
2641 			fragmented = true;
2642 
2643 		/* record total count of block that we're going to move */
2644 		total += map.m_len;
2645 
2646 		blk_end = map.m_pblk + map.m_len;
2647 
2648 		map.m_lblk += map.m_len;
2649 	}
2650 
2651 	if (!fragmented) {
2652 		total = 0;
2653 		goto out;
2654 	}
2655 
2656 	sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2657 
2658 	/*
2659 	 * make sure there are enough free section for LFS allocation, this can
2660 	 * avoid defragment running in SSR mode when free section are allocated
2661 	 * intensively
2662 	 */
2663 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2664 		err = -EAGAIN;
2665 		goto out;
2666 	}
2667 
2668 	map.m_lblk = pg_start;
2669 	map.m_len = pg_end - pg_start;
2670 	total = 0;
2671 
2672 	while (map.m_lblk < pg_end) {
2673 		pgoff_t idx;
2674 		int cnt = 0;
2675 
2676 do_map:
2677 		map.m_len = pg_end - map.m_lblk;
2678 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2679 		if (err)
2680 			goto clear_out;
2681 
2682 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2683 			map.m_lblk = next_pgofs;
2684 			goto check;
2685 		}
2686 
2687 		set_inode_flag(inode, FI_DO_DEFRAG);
2688 
2689 		idx = map.m_lblk;
2690 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2691 			struct page *page;
2692 
2693 			page = f2fs_get_lock_data_page(inode, idx, true);
2694 			if (IS_ERR(page)) {
2695 				err = PTR_ERR(page);
2696 				goto clear_out;
2697 			}
2698 
2699 			set_page_dirty(page);
2700 			f2fs_put_page(page, 1);
2701 
2702 			idx++;
2703 			cnt++;
2704 			total++;
2705 		}
2706 
2707 		map.m_lblk = idx;
2708 check:
2709 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
2710 			goto do_map;
2711 
2712 		clear_inode_flag(inode, FI_DO_DEFRAG);
2713 
2714 		err = filemap_fdatawrite(inode->i_mapping);
2715 		if (err)
2716 			goto out;
2717 	}
2718 clear_out:
2719 	clear_inode_flag(inode, FI_DO_DEFRAG);
2720 out:
2721 	inode_unlock(inode);
2722 	if (!err)
2723 		range->len = (u64)total << PAGE_SHIFT;
2724 	return err;
2725 }
2726 
2727 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2728 {
2729 	struct inode *inode = file_inode(filp);
2730 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2731 	struct f2fs_defragment range;
2732 	int err;
2733 
2734 	if (!capable(CAP_SYS_ADMIN))
2735 		return -EPERM;
2736 
2737 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2738 		return -EINVAL;
2739 
2740 	if (f2fs_readonly(sbi->sb))
2741 		return -EROFS;
2742 
2743 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2744 							sizeof(range)))
2745 		return -EFAULT;
2746 
2747 	/* verify alignment of offset & size */
2748 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2749 		return -EINVAL;
2750 
2751 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2752 					max_file_blocks(inode)))
2753 		return -EINVAL;
2754 
2755 	err = mnt_want_write_file(filp);
2756 	if (err)
2757 		return err;
2758 
2759 	err = f2fs_defragment_range(sbi, filp, &range);
2760 	mnt_drop_write_file(filp);
2761 
2762 	f2fs_update_time(sbi, REQ_TIME);
2763 	if (err < 0)
2764 		return err;
2765 
2766 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2767 							sizeof(range)))
2768 		return -EFAULT;
2769 
2770 	return 0;
2771 }
2772 
2773 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2774 			struct file *file_out, loff_t pos_out, size_t len)
2775 {
2776 	struct inode *src = file_inode(file_in);
2777 	struct inode *dst = file_inode(file_out);
2778 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2779 	size_t olen = len, dst_max_i_size = 0;
2780 	size_t dst_osize;
2781 	int ret;
2782 
2783 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2784 				src->i_sb != dst->i_sb)
2785 		return -EXDEV;
2786 
2787 	if (unlikely(f2fs_readonly(src->i_sb)))
2788 		return -EROFS;
2789 
2790 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2791 		return -EINVAL;
2792 
2793 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2794 		return -EOPNOTSUPP;
2795 
2796 	if (pos_out < 0 || pos_in < 0)
2797 		return -EINVAL;
2798 
2799 	if (src == dst) {
2800 		if (pos_in == pos_out)
2801 			return 0;
2802 		if (pos_out > pos_in && pos_out < pos_in + len)
2803 			return -EINVAL;
2804 	}
2805 
2806 	inode_lock(src);
2807 	if (src != dst) {
2808 		ret = -EBUSY;
2809 		if (!inode_trylock(dst))
2810 			goto out;
2811 	}
2812 
2813 	ret = -EINVAL;
2814 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2815 		goto out_unlock;
2816 	if (len == 0)
2817 		olen = len = src->i_size - pos_in;
2818 	if (pos_in + len == src->i_size)
2819 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2820 	if (len == 0) {
2821 		ret = 0;
2822 		goto out_unlock;
2823 	}
2824 
2825 	dst_osize = dst->i_size;
2826 	if (pos_out + olen > dst->i_size)
2827 		dst_max_i_size = pos_out + olen;
2828 
2829 	/* verify the end result is block aligned */
2830 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2831 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2832 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2833 		goto out_unlock;
2834 
2835 	ret = f2fs_convert_inline_inode(src);
2836 	if (ret)
2837 		goto out_unlock;
2838 
2839 	ret = f2fs_convert_inline_inode(dst);
2840 	if (ret)
2841 		goto out_unlock;
2842 
2843 	/* write out all dirty pages from offset */
2844 	ret = filemap_write_and_wait_range(src->i_mapping,
2845 					pos_in, pos_in + len);
2846 	if (ret)
2847 		goto out_unlock;
2848 
2849 	ret = filemap_write_and_wait_range(dst->i_mapping,
2850 					pos_out, pos_out + len);
2851 	if (ret)
2852 		goto out_unlock;
2853 
2854 	f2fs_balance_fs(sbi, true);
2855 
2856 	down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2857 	if (src != dst) {
2858 		ret = -EBUSY;
2859 		if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2860 			goto out_src;
2861 	}
2862 
2863 	f2fs_lock_op(sbi);
2864 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2865 				pos_out >> F2FS_BLKSIZE_BITS,
2866 				len >> F2FS_BLKSIZE_BITS, false);
2867 
2868 	if (!ret) {
2869 		if (dst_max_i_size)
2870 			f2fs_i_size_write(dst, dst_max_i_size);
2871 		else if (dst_osize != dst->i_size)
2872 			f2fs_i_size_write(dst, dst_osize);
2873 	}
2874 	f2fs_unlock_op(sbi);
2875 
2876 	if (src != dst)
2877 		up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2878 out_src:
2879 	up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2880 out_unlock:
2881 	if (src != dst)
2882 		inode_unlock(dst);
2883 out:
2884 	inode_unlock(src);
2885 	return ret;
2886 }
2887 
2888 static int __f2fs_ioc_move_range(struct file *filp,
2889 				struct f2fs_move_range *range)
2890 {
2891 	struct fd dst;
2892 	int err;
2893 
2894 	if (!(filp->f_mode & FMODE_READ) ||
2895 			!(filp->f_mode & FMODE_WRITE))
2896 		return -EBADF;
2897 
2898 	dst = fdget(range->dst_fd);
2899 	if (!dst.file)
2900 		return -EBADF;
2901 
2902 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2903 		err = -EBADF;
2904 		goto err_out;
2905 	}
2906 
2907 	err = mnt_want_write_file(filp);
2908 	if (err)
2909 		goto err_out;
2910 
2911 	err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2912 					range->pos_out, range->len);
2913 
2914 	mnt_drop_write_file(filp);
2915 err_out:
2916 	fdput(dst);
2917 	return err;
2918 }
2919 
2920 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2921 {
2922 	struct f2fs_move_range range;
2923 
2924 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2925 							sizeof(range)))
2926 		return -EFAULT;
2927 	return __f2fs_ioc_move_range(filp, &range);
2928 }
2929 
2930 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2931 {
2932 	struct inode *inode = file_inode(filp);
2933 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2934 	struct sit_info *sm = SIT_I(sbi);
2935 	unsigned int start_segno = 0, end_segno = 0;
2936 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2937 	struct f2fs_flush_device range;
2938 	int ret;
2939 
2940 	if (!capable(CAP_SYS_ADMIN))
2941 		return -EPERM;
2942 
2943 	if (f2fs_readonly(sbi->sb))
2944 		return -EROFS;
2945 
2946 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2947 		return -EINVAL;
2948 
2949 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2950 							sizeof(range)))
2951 		return -EFAULT;
2952 
2953 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2954 			__is_large_section(sbi)) {
2955 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2956 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2957 		return -EINVAL;
2958 	}
2959 
2960 	ret = mnt_want_write_file(filp);
2961 	if (ret)
2962 		return ret;
2963 
2964 	if (range.dev_num != 0)
2965 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2966 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2967 
2968 	start_segno = sm->last_victim[FLUSH_DEVICE];
2969 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2970 		start_segno = dev_start_segno;
2971 	end_segno = min(start_segno + range.segments, dev_end_segno);
2972 
2973 	while (start_segno < end_segno) {
2974 		if (!down_write_trylock(&sbi->gc_lock)) {
2975 			ret = -EBUSY;
2976 			goto out;
2977 		}
2978 		sm->last_victim[GC_CB] = end_segno + 1;
2979 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2980 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2981 		ret = f2fs_gc(sbi, true, true, start_segno);
2982 		if (ret == -EAGAIN)
2983 			ret = 0;
2984 		else if (ret < 0)
2985 			break;
2986 		start_segno++;
2987 	}
2988 out:
2989 	mnt_drop_write_file(filp);
2990 	return ret;
2991 }
2992 
2993 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2994 {
2995 	struct inode *inode = file_inode(filp);
2996 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2997 
2998 	/* Must validate to set it with SQLite behavior in Android. */
2999 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3000 
3001 	return put_user(sb_feature, (u32 __user *)arg);
3002 }
3003 
3004 #ifdef CONFIG_QUOTA
3005 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3006 {
3007 	struct dquot *transfer_to[MAXQUOTAS] = {};
3008 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3009 	struct super_block *sb = sbi->sb;
3010 	int err = 0;
3011 
3012 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3013 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
3014 		err = __dquot_transfer(inode, transfer_to);
3015 		if (err)
3016 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3017 		dqput(transfer_to[PRJQUOTA]);
3018 	}
3019 	return err;
3020 }
3021 
3022 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3023 {
3024 	struct inode *inode = file_inode(filp);
3025 	struct f2fs_inode_info *fi = F2FS_I(inode);
3026 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3027 	struct page *ipage;
3028 	kprojid_t kprojid;
3029 	int err;
3030 
3031 	if (!f2fs_sb_has_project_quota(sbi)) {
3032 		if (projid != F2FS_DEF_PROJID)
3033 			return -EOPNOTSUPP;
3034 		else
3035 			return 0;
3036 	}
3037 
3038 	if (!f2fs_has_extra_attr(inode))
3039 		return -EOPNOTSUPP;
3040 
3041 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3042 
3043 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
3044 		return 0;
3045 
3046 	err = -EPERM;
3047 	/* Is it quota file? Do not allow user to mess with it */
3048 	if (IS_NOQUOTA(inode))
3049 		return err;
3050 
3051 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
3052 	if (IS_ERR(ipage))
3053 		return PTR_ERR(ipage);
3054 
3055 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
3056 								i_projid)) {
3057 		err = -EOVERFLOW;
3058 		f2fs_put_page(ipage, 1);
3059 		return err;
3060 	}
3061 	f2fs_put_page(ipage, 1);
3062 
3063 	err = dquot_initialize(inode);
3064 	if (err)
3065 		return err;
3066 
3067 	f2fs_lock_op(sbi);
3068 	err = f2fs_transfer_project_quota(inode, kprojid);
3069 	if (err)
3070 		goto out_unlock;
3071 
3072 	F2FS_I(inode)->i_projid = kprojid;
3073 	inode->i_ctime = current_time(inode);
3074 	f2fs_mark_inode_dirty_sync(inode, true);
3075 out_unlock:
3076 	f2fs_unlock_op(sbi);
3077 	return err;
3078 }
3079 #else
3080 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3081 {
3082 	return 0;
3083 }
3084 
3085 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3086 {
3087 	if (projid != F2FS_DEF_PROJID)
3088 		return -EOPNOTSUPP;
3089 	return 0;
3090 }
3091 #endif
3092 
3093 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
3094 
3095 /*
3096  * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
3097  * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
3098  * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
3099  */
3100 
3101 static const struct {
3102 	u32 iflag;
3103 	u32 xflag;
3104 } f2fs_xflags_map[] = {
3105 	{ F2FS_SYNC_FL,		FS_XFLAG_SYNC },
3106 	{ F2FS_IMMUTABLE_FL,	FS_XFLAG_IMMUTABLE },
3107 	{ F2FS_APPEND_FL,	FS_XFLAG_APPEND },
3108 	{ F2FS_NODUMP_FL,	FS_XFLAG_NODUMP },
3109 	{ F2FS_NOATIME_FL,	FS_XFLAG_NOATIME },
3110 	{ F2FS_PROJINHERIT_FL,	FS_XFLAG_PROJINHERIT },
3111 };
3112 
3113 #define F2FS_SUPPORTED_XFLAGS (		\
3114 		FS_XFLAG_SYNC |		\
3115 		FS_XFLAG_IMMUTABLE |	\
3116 		FS_XFLAG_APPEND |	\
3117 		FS_XFLAG_NODUMP |	\
3118 		FS_XFLAG_NOATIME |	\
3119 		FS_XFLAG_PROJINHERIT)
3120 
3121 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
3122 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
3123 {
3124 	u32 xflags = 0;
3125 	int i;
3126 
3127 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3128 		if (iflags & f2fs_xflags_map[i].iflag)
3129 			xflags |= f2fs_xflags_map[i].xflag;
3130 
3131 	return xflags;
3132 }
3133 
3134 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
3135 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
3136 {
3137 	u32 iflags = 0;
3138 	int i;
3139 
3140 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3141 		if (xflags & f2fs_xflags_map[i].xflag)
3142 			iflags |= f2fs_xflags_map[i].iflag;
3143 
3144 	return iflags;
3145 }
3146 
3147 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
3148 {
3149 	struct f2fs_inode_info *fi = F2FS_I(inode);
3150 
3151 	simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
3152 
3153 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3154 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3155 }
3156 
3157 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
3158 {
3159 	struct inode *inode = file_inode(filp);
3160 	struct fsxattr fa;
3161 
3162 	f2fs_fill_fsxattr(inode, &fa);
3163 
3164 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
3165 		return -EFAULT;
3166 	return 0;
3167 }
3168 
3169 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
3170 {
3171 	struct inode *inode = file_inode(filp);
3172 	struct fsxattr fa, old_fa;
3173 	u32 iflags;
3174 	int err;
3175 
3176 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
3177 		return -EFAULT;
3178 
3179 	/* Make sure caller has proper permission */
3180 	if (!inode_owner_or_capable(&init_user_ns, inode))
3181 		return -EACCES;
3182 
3183 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
3184 		return -EOPNOTSUPP;
3185 
3186 	iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
3187 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3188 		return -EOPNOTSUPP;
3189 
3190 	err = mnt_want_write_file(filp);
3191 	if (err)
3192 		return err;
3193 
3194 	inode_lock(inode);
3195 
3196 	f2fs_fill_fsxattr(inode, &old_fa);
3197 	err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
3198 	if (err)
3199 		goto out;
3200 
3201 	err = f2fs_setflags_common(inode, iflags,
3202 			f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
3203 	if (err)
3204 		goto out;
3205 
3206 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3207 out:
3208 	inode_unlock(inode);
3209 	mnt_drop_write_file(filp);
3210 	return err;
3211 }
3212 
3213 int f2fs_pin_file_control(struct inode *inode, bool inc)
3214 {
3215 	struct f2fs_inode_info *fi = F2FS_I(inode);
3216 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3217 
3218 	/* Use i_gc_failures for normal file as a risk signal. */
3219 	if (inc)
3220 		f2fs_i_gc_failures_write(inode,
3221 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3222 
3223 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3224 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3225 			  __func__, inode->i_ino,
3226 			  fi->i_gc_failures[GC_FAILURE_PIN]);
3227 		clear_inode_flag(inode, FI_PIN_FILE);
3228 		return -EAGAIN;
3229 	}
3230 	return 0;
3231 }
3232 
3233 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3234 {
3235 	struct inode *inode = file_inode(filp);
3236 	__u32 pin;
3237 	int ret = 0;
3238 
3239 	if (get_user(pin, (__u32 __user *)arg))
3240 		return -EFAULT;
3241 
3242 	if (!S_ISREG(inode->i_mode))
3243 		return -EINVAL;
3244 
3245 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3246 		return -EROFS;
3247 
3248 	ret = mnt_want_write_file(filp);
3249 	if (ret)
3250 		return ret;
3251 
3252 	inode_lock(inode);
3253 
3254 	if (f2fs_should_update_outplace(inode, NULL)) {
3255 		ret = -EINVAL;
3256 		goto out;
3257 	}
3258 
3259 	if (!pin) {
3260 		clear_inode_flag(inode, FI_PIN_FILE);
3261 		f2fs_i_gc_failures_write(inode, 0);
3262 		goto done;
3263 	}
3264 
3265 	if (f2fs_pin_file_control(inode, false)) {
3266 		ret = -EAGAIN;
3267 		goto out;
3268 	}
3269 
3270 	ret = f2fs_convert_inline_inode(inode);
3271 	if (ret)
3272 		goto out;
3273 
3274 	if (!f2fs_disable_compressed_file(inode)) {
3275 		ret = -EOPNOTSUPP;
3276 		goto out;
3277 	}
3278 
3279 	set_inode_flag(inode, FI_PIN_FILE);
3280 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3281 done:
3282 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3283 out:
3284 	inode_unlock(inode);
3285 	mnt_drop_write_file(filp);
3286 	return ret;
3287 }
3288 
3289 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3290 {
3291 	struct inode *inode = file_inode(filp);
3292 	__u32 pin = 0;
3293 
3294 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3295 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3296 	return put_user(pin, (u32 __user *)arg);
3297 }
3298 
3299 int f2fs_precache_extents(struct inode *inode)
3300 {
3301 	struct f2fs_inode_info *fi = F2FS_I(inode);
3302 	struct f2fs_map_blocks map;
3303 	pgoff_t m_next_extent;
3304 	loff_t end;
3305 	int err;
3306 
3307 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3308 		return -EOPNOTSUPP;
3309 
3310 	map.m_lblk = 0;
3311 	map.m_next_pgofs = NULL;
3312 	map.m_next_extent = &m_next_extent;
3313 	map.m_seg_type = NO_CHECK_TYPE;
3314 	map.m_may_create = false;
3315 	end = max_file_blocks(inode);
3316 
3317 	while (map.m_lblk < end) {
3318 		map.m_len = end - map.m_lblk;
3319 
3320 		down_write(&fi->i_gc_rwsem[WRITE]);
3321 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3322 		up_write(&fi->i_gc_rwsem[WRITE]);
3323 		if (err)
3324 			return err;
3325 
3326 		map.m_lblk = m_next_extent;
3327 	}
3328 
3329 	return err;
3330 }
3331 
3332 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3333 {
3334 	return f2fs_precache_extents(file_inode(filp));
3335 }
3336 
3337 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3338 {
3339 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3340 	__u64 block_count;
3341 
3342 	if (!capable(CAP_SYS_ADMIN))
3343 		return -EPERM;
3344 
3345 	if (f2fs_readonly(sbi->sb))
3346 		return -EROFS;
3347 
3348 	if (copy_from_user(&block_count, (void __user *)arg,
3349 			   sizeof(block_count)))
3350 		return -EFAULT;
3351 
3352 	return f2fs_resize_fs(sbi, block_count);
3353 }
3354 
3355 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3356 {
3357 	struct inode *inode = file_inode(filp);
3358 
3359 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3360 
3361 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3362 		f2fs_warn(F2FS_I_SB(inode),
3363 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3364 			  inode->i_ino);
3365 		return -EOPNOTSUPP;
3366 	}
3367 
3368 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3369 }
3370 
3371 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3372 {
3373 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3374 		return -EOPNOTSUPP;
3375 
3376 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3377 }
3378 
3379 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3380 {
3381 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3382 		return -EOPNOTSUPP;
3383 
3384 	return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3385 }
3386 
3387 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3388 {
3389 	struct inode *inode = file_inode(filp);
3390 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3391 	char *vbuf;
3392 	int count;
3393 	int err = 0;
3394 
3395 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3396 	if (!vbuf)
3397 		return -ENOMEM;
3398 
3399 	down_read(&sbi->sb_lock);
3400 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3401 			ARRAY_SIZE(sbi->raw_super->volume_name),
3402 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3403 	up_read(&sbi->sb_lock);
3404 
3405 	if (copy_to_user((char __user *)arg, vbuf,
3406 				min(FSLABEL_MAX, count)))
3407 		err = -EFAULT;
3408 
3409 	kfree(vbuf);
3410 	return err;
3411 }
3412 
3413 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3414 {
3415 	struct inode *inode = file_inode(filp);
3416 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3417 	char *vbuf;
3418 	int err = 0;
3419 
3420 	if (!capable(CAP_SYS_ADMIN))
3421 		return -EPERM;
3422 
3423 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3424 	if (IS_ERR(vbuf))
3425 		return PTR_ERR(vbuf);
3426 
3427 	err = mnt_want_write_file(filp);
3428 	if (err)
3429 		goto out;
3430 
3431 	down_write(&sbi->sb_lock);
3432 
3433 	memset(sbi->raw_super->volume_name, 0,
3434 			sizeof(sbi->raw_super->volume_name));
3435 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3436 			sbi->raw_super->volume_name,
3437 			ARRAY_SIZE(sbi->raw_super->volume_name));
3438 
3439 	err = f2fs_commit_super(sbi, false);
3440 
3441 	up_write(&sbi->sb_lock);
3442 
3443 	mnt_drop_write_file(filp);
3444 out:
3445 	kfree(vbuf);
3446 	return err;
3447 }
3448 
3449 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3450 {
3451 	struct inode *inode = file_inode(filp);
3452 	__u64 blocks;
3453 
3454 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3455 		return -EOPNOTSUPP;
3456 
3457 	if (!f2fs_compressed_file(inode))
3458 		return -EINVAL;
3459 
3460 	blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3461 	return put_user(blocks, (u64 __user *)arg);
3462 }
3463 
3464 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3465 {
3466 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3467 	unsigned int released_blocks = 0;
3468 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3469 	block_t blkaddr;
3470 	int i;
3471 
3472 	for (i = 0; i < count; i++) {
3473 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3474 						dn->ofs_in_node + i);
3475 
3476 		if (!__is_valid_data_blkaddr(blkaddr))
3477 			continue;
3478 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3479 					DATA_GENERIC_ENHANCE)))
3480 			return -EFSCORRUPTED;
3481 	}
3482 
3483 	while (count) {
3484 		int compr_blocks = 0;
3485 
3486 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3487 			blkaddr = f2fs_data_blkaddr(dn);
3488 
3489 			if (i == 0) {
3490 				if (blkaddr == COMPRESS_ADDR)
3491 					continue;
3492 				dn->ofs_in_node += cluster_size;
3493 				goto next;
3494 			}
3495 
3496 			if (__is_valid_data_blkaddr(blkaddr))
3497 				compr_blocks++;
3498 
3499 			if (blkaddr != NEW_ADDR)
3500 				continue;
3501 
3502 			dn->data_blkaddr = NULL_ADDR;
3503 			f2fs_set_data_blkaddr(dn);
3504 		}
3505 
3506 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3507 		dec_valid_block_count(sbi, dn->inode,
3508 					cluster_size - compr_blocks);
3509 
3510 		released_blocks += cluster_size - compr_blocks;
3511 next:
3512 		count -= cluster_size;
3513 	}
3514 
3515 	return released_blocks;
3516 }
3517 
3518 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3519 {
3520 	struct inode *inode = file_inode(filp);
3521 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3522 	pgoff_t page_idx = 0, last_idx;
3523 	unsigned int released_blocks = 0;
3524 	int ret;
3525 	int writecount;
3526 
3527 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3528 		return -EOPNOTSUPP;
3529 
3530 	if (!f2fs_compressed_file(inode))
3531 		return -EINVAL;
3532 
3533 	if (f2fs_readonly(sbi->sb))
3534 		return -EROFS;
3535 
3536 	ret = mnt_want_write_file(filp);
3537 	if (ret)
3538 		return ret;
3539 
3540 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3541 
3542 	inode_lock(inode);
3543 
3544 	writecount = atomic_read(&inode->i_writecount);
3545 	if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3546 			(!(filp->f_mode & FMODE_WRITE) && writecount)) {
3547 		ret = -EBUSY;
3548 		goto out;
3549 	}
3550 
3551 	if (IS_IMMUTABLE(inode)) {
3552 		ret = -EINVAL;
3553 		goto out;
3554 	}
3555 
3556 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3557 	if (ret)
3558 		goto out;
3559 
3560 	F2FS_I(inode)->i_flags |= F2FS_IMMUTABLE_FL;
3561 	f2fs_set_inode_flags(inode);
3562 	inode->i_ctime = current_time(inode);
3563 	f2fs_mark_inode_dirty_sync(inode, true);
3564 
3565 	if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3566 		goto out;
3567 
3568 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3569 	down_write(&F2FS_I(inode)->i_mmap_sem);
3570 
3571 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3572 
3573 	while (page_idx < last_idx) {
3574 		struct dnode_of_data dn;
3575 		pgoff_t end_offset, count;
3576 
3577 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3578 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3579 		if (ret) {
3580 			if (ret == -ENOENT) {
3581 				page_idx = f2fs_get_next_page_offset(&dn,
3582 								page_idx);
3583 				ret = 0;
3584 				continue;
3585 			}
3586 			break;
3587 		}
3588 
3589 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3590 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3591 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3592 
3593 		ret = release_compress_blocks(&dn, count);
3594 
3595 		f2fs_put_dnode(&dn);
3596 
3597 		if (ret < 0)
3598 			break;
3599 
3600 		page_idx += count;
3601 		released_blocks += ret;
3602 	}
3603 
3604 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3605 	up_write(&F2FS_I(inode)->i_mmap_sem);
3606 out:
3607 	inode_unlock(inode);
3608 
3609 	mnt_drop_write_file(filp);
3610 
3611 	if (ret >= 0) {
3612 		ret = put_user(released_blocks, (u64 __user *)arg);
3613 	} else if (released_blocks &&
3614 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3615 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3616 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3617 			"iblocks=%llu, released=%u, compr_blocks=%u, "
3618 			"run fsck to fix.",
3619 			__func__, inode->i_ino, inode->i_blocks,
3620 			released_blocks,
3621 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
3622 	}
3623 
3624 	return ret;
3625 }
3626 
3627 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3628 {
3629 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3630 	unsigned int reserved_blocks = 0;
3631 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3632 	block_t blkaddr;
3633 	int i;
3634 
3635 	for (i = 0; i < count; i++) {
3636 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3637 						dn->ofs_in_node + i);
3638 
3639 		if (!__is_valid_data_blkaddr(blkaddr))
3640 			continue;
3641 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3642 					DATA_GENERIC_ENHANCE)))
3643 			return -EFSCORRUPTED;
3644 	}
3645 
3646 	while (count) {
3647 		int compr_blocks = 0;
3648 		blkcnt_t reserved;
3649 		int ret;
3650 
3651 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3652 			blkaddr = f2fs_data_blkaddr(dn);
3653 
3654 			if (i == 0) {
3655 				if (blkaddr == COMPRESS_ADDR)
3656 					continue;
3657 				dn->ofs_in_node += cluster_size;
3658 				goto next;
3659 			}
3660 
3661 			if (__is_valid_data_blkaddr(blkaddr)) {
3662 				compr_blocks++;
3663 				continue;
3664 			}
3665 
3666 			dn->data_blkaddr = NEW_ADDR;
3667 			f2fs_set_data_blkaddr(dn);
3668 		}
3669 
3670 		reserved = cluster_size - compr_blocks;
3671 		ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3672 		if (ret)
3673 			return ret;
3674 
3675 		if (reserved != cluster_size - compr_blocks)
3676 			return -ENOSPC;
3677 
3678 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3679 
3680 		reserved_blocks += reserved;
3681 next:
3682 		count -= cluster_size;
3683 	}
3684 
3685 	return reserved_blocks;
3686 }
3687 
3688 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3689 {
3690 	struct inode *inode = file_inode(filp);
3691 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3692 	pgoff_t page_idx = 0, last_idx;
3693 	unsigned int reserved_blocks = 0;
3694 	int ret;
3695 
3696 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3697 		return -EOPNOTSUPP;
3698 
3699 	if (!f2fs_compressed_file(inode))
3700 		return -EINVAL;
3701 
3702 	if (f2fs_readonly(sbi->sb))
3703 		return -EROFS;
3704 
3705 	ret = mnt_want_write_file(filp);
3706 	if (ret)
3707 		return ret;
3708 
3709 	if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3710 		goto out;
3711 
3712 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3713 
3714 	inode_lock(inode);
3715 
3716 	if (!IS_IMMUTABLE(inode)) {
3717 		ret = -EINVAL;
3718 		goto unlock_inode;
3719 	}
3720 
3721 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3722 	down_write(&F2FS_I(inode)->i_mmap_sem);
3723 
3724 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3725 
3726 	while (page_idx < last_idx) {
3727 		struct dnode_of_data dn;
3728 		pgoff_t end_offset, count;
3729 
3730 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3731 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3732 		if (ret) {
3733 			if (ret == -ENOENT) {
3734 				page_idx = f2fs_get_next_page_offset(&dn,
3735 								page_idx);
3736 				ret = 0;
3737 				continue;
3738 			}
3739 			break;
3740 		}
3741 
3742 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3743 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3744 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3745 
3746 		ret = reserve_compress_blocks(&dn, count);
3747 
3748 		f2fs_put_dnode(&dn);
3749 
3750 		if (ret < 0)
3751 			break;
3752 
3753 		page_idx += count;
3754 		reserved_blocks += ret;
3755 	}
3756 
3757 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3758 	up_write(&F2FS_I(inode)->i_mmap_sem);
3759 
3760 	if (ret >= 0) {
3761 		F2FS_I(inode)->i_flags &= ~F2FS_IMMUTABLE_FL;
3762 		f2fs_set_inode_flags(inode);
3763 		inode->i_ctime = current_time(inode);
3764 		f2fs_mark_inode_dirty_sync(inode, true);
3765 	}
3766 unlock_inode:
3767 	inode_unlock(inode);
3768 out:
3769 	mnt_drop_write_file(filp);
3770 
3771 	if (ret >= 0) {
3772 		ret = put_user(reserved_blocks, (u64 __user *)arg);
3773 	} else if (reserved_blocks &&
3774 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3775 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3776 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3777 			"iblocks=%llu, reserved=%u, compr_blocks=%u, "
3778 			"run fsck to fix.",
3779 			__func__, inode->i_ino, inode->i_blocks,
3780 			reserved_blocks,
3781 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
3782 	}
3783 
3784 	return ret;
3785 }
3786 
3787 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3788 		pgoff_t off, block_t block, block_t len, u32 flags)
3789 {
3790 	struct request_queue *q = bdev_get_queue(bdev);
3791 	sector_t sector = SECTOR_FROM_BLOCK(block);
3792 	sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3793 	int ret = 0;
3794 
3795 	if (!q)
3796 		return -ENXIO;
3797 
3798 	if (flags & F2FS_TRIM_FILE_DISCARD)
3799 		ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3800 						blk_queue_secure_erase(q) ?
3801 						BLKDEV_DISCARD_SECURE : 0);
3802 
3803 	if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3804 		if (IS_ENCRYPTED(inode))
3805 			ret = fscrypt_zeroout_range(inode, off, block, len);
3806 		else
3807 			ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3808 					GFP_NOFS, 0);
3809 	}
3810 
3811 	return ret;
3812 }
3813 
3814 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3815 {
3816 	struct inode *inode = file_inode(filp);
3817 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3818 	struct address_space *mapping = inode->i_mapping;
3819 	struct block_device *prev_bdev = NULL;
3820 	struct f2fs_sectrim_range range;
3821 	pgoff_t index, pg_end, prev_index = 0;
3822 	block_t prev_block = 0, len = 0;
3823 	loff_t end_addr;
3824 	bool to_end = false;
3825 	int ret = 0;
3826 
3827 	if (!(filp->f_mode & FMODE_WRITE))
3828 		return -EBADF;
3829 
3830 	if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3831 				sizeof(range)))
3832 		return -EFAULT;
3833 
3834 	if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3835 			!S_ISREG(inode->i_mode))
3836 		return -EINVAL;
3837 
3838 	if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3839 			!f2fs_hw_support_discard(sbi)) ||
3840 			((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3841 			 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3842 		return -EOPNOTSUPP;
3843 
3844 	file_start_write(filp);
3845 	inode_lock(inode);
3846 
3847 	if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3848 			range.start >= inode->i_size) {
3849 		ret = -EINVAL;
3850 		goto err;
3851 	}
3852 
3853 	if (range.len == 0)
3854 		goto err;
3855 
3856 	if (inode->i_size - range.start > range.len) {
3857 		end_addr = range.start + range.len;
3858 	} else {
3859 		end_addr = range.len == (u64)-1 ?
3860 			sbi->sb->s_maxbytes : inode->i_size;
3861 		to_end = true;
3862 	}
3863 
3864 	if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3865 			(!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3866 		ret = -EINVAL;
3867 		goto err;
3868 	}
3869 
3870 	index = F2FS_BYTES_TO_BLK(range.start);
3871 	pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3872 
3873 	ret = f2fs_convert_inline_inode(inode);
3874 	if (ret)
3875 		goto err;
3876 
3877 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3878 	down_write(&F2FS_I(inode)->i_mmap_sem);
3879 
3880 	ret = filemap_write_and_wait_range(mapping, range.start,
3881 			to_end ? LLONG_MAX : end_addr - 1);
3882 	if (ret)
3883 		goto out;
3884 
3885 	truncate_inode_pages_range(mapping, range.start,
3886 			to_end ? -1 : end_addr - 1);
3887 
3888 	while (index < pg_end) {
3889 		struct dnode_of_data dn;
3890 		pgoff_t end_offset, count;
3891 		int i;
3892 
3893 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3894 		ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3895 		if (ret) {
3896 			if (ret == -ENOENT) {
3897 				index = f2fs_get_next_page_offset(&dn, index);
3898 				continue;
3899 			}
3900 			goto out;
3901 		}
3902 
3903 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3904 		count = min(end_offset - dn.ofs_in_node, pg_end - index);
3905 		for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3906 			struct block_device *cur_bdev;
3907 			block_t blkaddr = f2fs_data_blkaddr(&dn);
3908 
3909 			if (!__is_valid_data_blkaddr(blkaddr))
3910 				continue;
3911 
3912 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3913 						DATA_GENERIC_ENHANCE)) {
3914 				ret = -EFSCORRUPTED;
3915 				f2fs_put_dnode(&dn);
3916 				goto out;
3917 			}
3918 
3919 			cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3920 			if (f2fs_is_multi_device(sbi)) {
3921 				int di = f2fs_target_device_index(sbi, blkaddr);
3922 
3923 				blkaddr -= FDEV(di).start_blk;
3924 			}
3925 
3926 			if (len) {
3927 				if (prev_bdev == cur_bdev &&
3928 						index == prev_index + len &&
3929 						blkaddr == prev_block + len) {
3930 					len++;
3931 				} else {
3932 					ret = f2fs_secure_erase(prev_bdev,
3933 						inode, prev_index, prev_block,
3934 						len, range.flags);
3935 					if (ret) {
3936 						f2fs_put_dnode(&dn);
3937 						goto out;
3938 					}
3939 
3940 					len = 0;
3941 				}
3942 			}
3943 
3944 			if (!len) {
3945 				prev_bdev = cur_bdev;
3946 				prev_index = index;
3947 				prev_block = blkaddr;
3948 				len = 1;
3949 			}
3950 		}
3951 
3952 		f2fs_put_dnode(&dn);
3953 
3954 		if (fatal_signal_pending(current)) {
3955 			ret = -EINTR;
3956 			goto out;
3957 		}
3958 		cond_resched();
3959 	}
3960 
3961 	if (len)
3962 		ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3963 				prev_block, len, range.flags);
3964 out:
3965 	up_write(&F2FS_I(inode)->i_mmap_sem);
3966 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3967 err:
3968 	inode_unlock(inode);
3969 	file_end_write(filp);
3970 
3971 	return ret;
3972 }
3973 
3974 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3975 {
3976 	struct inode *inode = file_inode(filp);
3977 	struct f2fs_comp_option option;
3978 
3979 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3980 		return -EOPNOTSUPP;
3981 
3982 	inode_lock_shared(inode);
3983 
3984 	if (!f2fs_compressed_file(inode)) {
3985 		inode_unlock_shared(inode);
3986 		return -ENODATA;
3987 	}
3988 
3989 	option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3990 	option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3991 
3992 	inode_unlock_shared(inode);
3993 
3994 	if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3995 				sizeof(option)))
3996 		return -EFAULT;
3997 
3998 	return 0;
3999 }
4000 
4001 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4002 {
4003 	struct inode *inode = file_inode(filp);
4004 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4005 	struct f2fs_comp_option option;
4006 	int ret = 0;
4007 
4008 	if (!f2fs_sb_has_compression(sbi))
4009 		return -EOPNOTSUPP;
4010 
4011 	if (!(filp->f_mode & FMODE_WRITE))
4012 		return -EBADF;
4013 
4014 	if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4015 				sizeof(option)))
4016 		return -EFAULT;
4017 
4018 	if (!f2fs_compressed_file(inode) ||
4019 			option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
4020 			option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
4021 			option.algorithm >= COMPRESS_MAX)
4022 		return -EINVAL;
4023 
4024 	file_start_write(filp);
4025 	inode_lock(inode);
4026 
4027 	if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4028 		ret = -EBUSY;
4029 		goto out;
4030 	}
4031 
4032 	if (inode->i_size != 0) {
4033 		ret = -EFBIG;
4034 		goto out;
4035 	}
4036 
4037 	F2FS_I(inode)->i_compress_algorithm = option.algorithm;
4038 	F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
4039 	F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
4040 	f2fs_mark_inode_dirty_sync(inode, true);
4041 
4042 	if (!f2fs_is_compress_backend_ready(inode))
4043 		f2fs_warn(sbi, "compression algorithm is successfully set, "
4044 			"but current kernel doesn't support this algorithm.");
4045 out:
4046 	inode_unlock(inode);
4047 	file_end_write(filp);
4048 
4049 	return ret;
4050 }
4051 
4052 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4053 {
4054 	DEFINE_READAHEAD(ractl, NULL, inode->i_mapping, page_idx);
4055 	struct address_space *mapping = inode->i_mapping;
4056 	struct page *page;
4057 	pgoff_t redirty_idx = page_idx;
4058 	int i, page_len = 0, ret = 0;
4059 
4060 	page_cache_ra_unbounded(&ractl, len, 0);
4061 
4062 	for (i = 0; i < len; i++, page_idx++) {
4063 		page = read_cache_page(mapping, page_idx, NULL, NULL);
4064 		if (IS_ERR(page)) {
4065 			ret = PTR_ERR(page);
4066 			break;
4067 		}
4068 		page_len++;
4069 	}
4070 
4071 	for (i = 0; i < page_len; i++, redirty_idx++) {
4072 		page = find_lock_page(mapping, redirty_idx);
4073 		if (!page) {
4074 			ret = -ENOMEM;
4075 			break;
4076 		}
4077 		set_page_dirty(page);
4078 		f2fs_put_page(page, 1);
4079 		f2fs_put_page(page, 0);
4080 	}
4081 
4082 	return ret;
4083 }
4084 
4085 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
4086 {
4087 	struct inode *inode = file_inode(filp);
4088 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4089 	struct f2fs_inode_info *fi = F2FS_I(inode);
4090 	pgoff_t page_idx = 0, last_idx;
4091 	unsigned int blk_per_seg = sbi->blocks_per_seg;
4092 	int cluster_size = F2FS_I(inode)->i_cluster_size;
4093 	int count, ret;
4094 
4095 	if (!f2fs_sb_has_compression(sbi) ||
4096 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4097 		return -EOPNOTSUPP;
4098 
4099 	if (!(filp->f_mode & FMODE_WRITE))
4100 		return -EBADF;
4101 
4102 	if (!f2fs_compressed_file(inode))
4103 		return -EINVAL;
4104 
4105 	f2fs_balance_fs(F2FS_I_SB(inode), true);
4106 
4107 	file_start_write(filp);
4108 	inode_lock(inode);
4109 
4110 	if (!f2fs_is_compress_backend_ready(inode)) {
4111 		ret = -EOPNOTSUPP;
4112 		goto out;
4113 	}
4114 
4115 	if (f2fs_is_mmap_file(inode)) {
4116 		ret = -EBUSY;
4117 		goto out;
4118 	}
4119 
4120 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4121 	if (ret)
4122 		goto out;
4123 
4124 	if (!atomic_read(&fi->i_compr_blocks))
4125 		goto out;
4126 
4127 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4128 
4129 	count = last_idx - page_idx;
4130 	while (count) {
4131 		int len = min(cluster_size, count);
4132 
4133 		ret = redirty_blocks(inode, page_idx, len);
4134 		if (ret < 0)
4135 			break;
4136 
4137 		if (get_dirty_pages(inode) >= blk_per_seg)
4138 			filemap_fdatawrite(inode->i_mapping);
4139 
4140 		count -= len;
4141 		page_idx += len;
4142 	}
4143 
4144 	if (!ret)
4145 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4146 							LLONG_MAX);
4147 
4148 	if (ret)
4149 		f2fs_warn(sbi, "%s: The file might be partially decompressed "
4150 				"(errno=%d). Please delete the file.\n",
4151 				__func__, ret);
4152 out:
4153 	inode_unlock(inode);
4154 	file_end_write(filp);
4155 
4156 	return ret;
4157 }
4158 
4159 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4160 {
4161 	struct inode *inode = file_inode(filp);
4162 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4163 	pgoff_t page_idx = 0, last_idx;
4164 	unsigned int blk_per_seg = sbi->blocks_per_seg;
4165 	int cluster_size = F2FS_I(inode)->i_cluster_size;
4166 	int count, ret;
4167 
4168 	if (!f2fs_sb_has_compression(sbi) ||
4169 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4170 		return -EOPNOTSUPP;
4171 
4172 	if (!(filp->f_mode & FMODE_WRITE))
4173 		return -EBADF;
4174 
4175 	if (!f2fs_compressed_file(inode))
4176 		return -EINVAL;
4177 
4178 	f2fs_balance_fs(F2FS_I_SB(inode), true);
4179 
4180 	file_start_write(filp);
4181 	inode_lock(inode);
4182 
4183 	if (!f2fs_is_compress_backend_ready(inode)) {
4184 		ret = -EOPNOTSUPP;
4185 		goto out;
4186 	}
4187 
4188 	if (f2fs_is_mmap_file(inode)) {
4189 		ret = -EBUSY;
4190 		goto out;
4191 	}
4192 
4193 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4194 	if (ret)
4195 		goto out;
4196 
4197 	set_inode_flag(inode, FI_ENABLE_COMPRESS);
4198 
4199 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4200 
4201 	count = last_idx - page_idx;
4202 	while (count) {
4203 		int len = min(cluster_size, count);
4204 
4205 		ret = redirty_blocks(inode, page_idx, len);
4206 		if (ret < 0)
4207 			break;
4208 
4209 		if (get_dirty_pages(inode) >= blk_per_seg)
4210 			filemap_fdatawrite(inode->i_mapping);
4211 
4212 		count -= len;
4213 		page_idx += len;
4214 	}
4215 
4216 	if (!ret)
4217 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4218 							LLONG_MAX);
4219 
4220 	clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4221 
4222 	if (ret)
4223 		f2fs_warn(sbi, "%s: The file might be partially compressed "
4224 				"(errno=%d). Please delete the file.\n",
4225 				__func__, ret);
4226 out:
4227 	inode_unlock(inode);
4228 	file_end_write(filp);
4229 
4230 	return ret;
4231 }
4232 
4233 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4234 {
4235 	switch (cmd) {
4236 	case FS_IOC_GETFLAGS:
4237 		return f2fs_ioc_getflags(filp, arg);
4238 	case FS_IOC_SETFLAGS:
4239 		return f2fs_ioc_setflags(filp, arg);
4240 	case FS_IOC_GETVERSION:
4241 		return f2fs_ioc_getversion(filp, arg);
4242 	case F2FS_IOC_START_ATOMIC_WRITE:
4243 		return f2fs_ioc_start_atomic_write(filp);
4244 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4245 		return f2fs_ioc_commit_atomic_write(filp);
4246 	case F2FS_IOC_START_VOLATILE_WRITE:
4247 		return f2fs_ioc_start_volatile_write(filp);
4248 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4249 		return f2fs_ioc_release_volatile_write(filp);
4250 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
4251 		return f2fs_ioc_abort_volatile_write(filp);
4252 	case F2FS_IOC_SHUTDOWN:
4253 		return f2fs_ioc_shutdown(filp, arg);
4254 	case FITRIM:
4255 		return f2fs_ioc_fitrim(filp, arg);
4256 	case FS_IOC_SET_ENCRYPTION_POLICY:
4257 		return f2fs_ioc_set_encryption_policy(filp, arg);
4258 	case FS_IOC_GET_ENCRYPTION_POLICY:
4259 		return f2fs_ioc_get_encryption_policy(filp, arg);
4260 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4261 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4262 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4263 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4264 	case FS_IOC_ADD_ENCRYPTION_KEY:
4265 		return f2fs_ioc_add_encryption_key(filp, arg);
4266 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4267 		return f2fs_ioc_remove_encryption_key(filp, arg);
4268 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4269 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4270 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4271 		return f2fs_ioc_get_encryption_key_status(filp, arg);
4272 	case FS_IOC_GET_ENCRYPTION_NONCE:
4273 		return f2fs_ioc_get_encryption_nonce(filp, arg);
4274 	case F2FS_IOC_GARBAGE_COLLECT:
4275 		return f2fs_ioc_gc(filp, arg);
4276 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4277 		return f2fs_ioc_gc_range(filp, arg);
4278 	case F2FS_IOC_WRITE_CHECKPOINT:
4279 		return f2fs_ioc_write_checkpoint(filp, arg);
4280 	case F2FS_IOC_DEFRAGMENT:
4281 		return f2fs_ioc_defragment(filp, arg);
4282 	case F2FS_IOC_MOVE_RANGE:
4283 		return f2fs_ioc_move_range(filp, arg);
4284 	case F2FS_IOC_FLUSH_DEVICE:
4285 		return f2fs_ioc_flush_device(filp, arg);
4286 	case F2FS_IOC_GET_FEATURES:
4287 		return f2fs_ioc_get_features(filp, arg);
4288 	case FS_IOC_FSGETXATTR:
4289 		return f2fs_ioc_fsgetxattr(filp, arg);
4290 	case FS_IOC_FSSETXATTR:
4291 		return f2fs_ioc_fssetxattr(filp, arg);
4292 	case F2FS_IOC_GET_PIN_FILE:
4293 		return f2fs_ioc_get_pin_file(filp, arg);
4294 	case F2FS_IOC_SET_PIN_FILE:
4295 		return f2fs_ioc_set_pin_file(filp, arg);
4296 	case F2FS_IOC_PRECACHE_EXTENTS:
4297 		return f2fs_ioc_precache_extents(filp, arg);
4298 	case F2FS_IOC_RESIZE_FS:
4299 		return f2fs_ioc_resize_fs(filp, arg);
4300 	case FS_IOC_ENABLE_VERITY:
4301 		return f2fs_ioc_enable_verity(filp, arg);
4302 	case FS_IOC_MEASURE_VERITY:
4303 		return f2fs_ioc_measure_verity(filp, arg);
4304 	case FS_IOC_READ_VERITY_METADATA:
4305 		return f2fs_ioc_read_verity_metadata(filp, arg);
4306 	case FS_IOC_GETFSLABEL:
4307 		return f2fs_ioc_getfslabel(filp, arg);
4308 	case FS_IOC_SETFSLABEL:
4309 		return f2fs_ioc_setfslabel(filp, arg);
4310 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4311 		return f2fs_get_compress_blocks(filp, arg);
4312 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4313 		return f2fs_release_compress_blocks(filp, arg);
4314 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4315 		return f2fs_reserve_compress_blocks(filp, arg);
4316 	case F2FS_IOC_SEC_TRIM_FILE:
4317 		return f2fs_sec_trim_file(filp, arg);
4318 	case F2FS_IOC_GET_COMPRESS_OPTION:
4319 		return f2fs_ioc_get_compress_option(filp, arg);
4320 	case F2FS_IOC_SET_COMPRESS_OPTION:
4321 		return f2fs_ioc_set_compress_option(filp, arg);
4322 	case F2FS_IOC_DECOMPRESS_FILE:
4323 		return f2fs_ioc_decompress_file(filp, arg);
4324 	case F2FS_IOC_COMPRESS_FILE:
4325 		return f2fs_ioc_compress_file(filp, arg);
4326 	default:
4327 		return -ENOTTY;
4328 	}
4329 }
4330 
4331 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4332 {
4333 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4334 		return -EIO;
4335 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4336 		return -ENOSPC;
4337 
4338 	return __f2fs_ioctl(filp, cmd, arg);
4339 }
4340 
4341 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
4342 {
4343 	struct file *file = iocb->ki_filp;
4344 	struct inode *inode = file_inode(file);
4345 	int ret;
4346 
4347 	if (!f2fs_is_compress_backend_ready(inode))
4348 		return -EOPNOTSUPP;
4349 
4350 	ret = generic_file_read_iter(iocb, iter);
4351 
4352 	if (ret > 0)
4353 		f2fs_update_iostat(F2FS_I_SB(inode), APP_READ_IO, ret);
4354 
4355 	return ret;
4356 }
4357 
4358 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4359 {
4360 	struct file *file = iocb->ki_filp;
4361 	struct inode *inode = file_inode(file);
4362 	ssize_t ret;
4363 
4364 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4365 		ret = -EIO;
4366 		goto out;
4367 	}
4368 
4369 	if (!f2fs_is_compress_backend_ready(inode)) {
4370 		ret = -EOPNOTSUPP;
4371 		goto out;
4372 	}
4373 
4374 	if (iocb->ki_flags & IOCB_NOWAIT) {
4375 		if (!inode_trylock(inode)) {
4376 			ret = -EAGAIN;
4377 			goto out;
4378 		}
4379 	} else {
4380 		inode_lock(inode);
4381 	}
4382 
4383 	if (unlikely(IS_IMMUTABLE(inode))) {
4384 		ret = -EPERM;
4385 		goto unlock;
4386 	}
4387 
4388 	ret = generic_write_checks(iocb, from);
4389 	if (ret > 0) {
4390 		bool preallocated = false;
4391 		size_t target_size = 0;
4392 		int err;
4393 
4394 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
4395 			set_inode_flag(inode, FI_NO_PREALLOC);
4396 
4397 		if ((iocb->ki_flags & IOCB_NOWAIT)) {
4398 			if (!f2fs_overwrite_io(inode, iocb->ki_pos,
4399 						iov_iter_count(from)) ||
4400 				f2fs_has_inline_data(inode) ||
4401 				f2fs_force_buffered_io(inode, iocb, from)) {
4402 				clear_inode_flag(inode, FI_NO_PREALLOC);
4403 				inode_unlock(inode);
4404 				ret = -EAGAIN;
4405 				goto out;
4406 			}
4407 			goto write;
4408 		}
4409 
4410 		if (is_inode_flag_set(inode, FI_NO_PREALLOC))
4411 			goto write;
4412 
4413 		if (iocb->ki_flags & IOCB_DIRECT) {
4414 			/*
4415 			 * Convert inline data for Direct I/O before entering
4416 			 * f2fs_direct_IO().
4417 			 */
4418 			err = f2fs_convert_inline_inode(inode);
4419 			if (err)
4420 				goto out_err;
4421 			/*
4422 			 * If force_buffere_io() is true, we have to allocate
4423 			 * blocks all the time, since f2fs_direct_IO will fall
4424 			 * back to buffered IO.
4425 			 */
4426 			if (!f2fs_force_buffered_io(inode, iocb, from) &&
4427 					allow_outplace_dio(inode, iocb, from))
4428 				goto write;
4429 		}
4430 		preallocated = true;
4431 		target_size = iocb->ki_pos + iov_iter_count(from);
4432 
4433 		err = f2fs_preallocate_blocks(iocb, from);
4434 		if (err) {
4435 out_err:
4436 			clear_inode_flag(inode, FI_NO_PREALLOC);
4437 			inode_unlock(inode);
4438 			ret = err;
4439 			goto out;
4440 		}
4441 write:
4442 		ret = __generic_file_write_iter(iocb, from);
4443 		clear_inode_flag(inode, FI_NO_PREALLOC);
4444 
4445 		/* if we couldn't write data, we should deallocate blocks. */
4446 		if (preallocated && i_size_read(inode) < target_size)
4447 			f2fs_truncate(inode);
4448 
4449 		if (ret > 0)
4450 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
4451 	}
4452 unlock:
4453 	inode_unlock(inode);
4454 out:
4455 	trace_f2fs_file_write_iter(inode, iocb->ki_pos,
4456 					iov_iter_count(from), ret);
4457 	if (ret > 0)
4458 		ret = generic_write_sync(iocb, ret);
4459 	return ret;
4460 }
4461 
4462 #ifdef CONFIG_COMPAT
4463 struct compat_f2fs_gc_range {
4464 	u32 sync;
4465 	compat_u64 start;
4466 	compat_u64 len;
4467 };
4468 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE	_IOW(F2FS_IOCTL_MAGIC, 11,\
4469 						struct compat_f2fs_gc_range)
4470 
4471 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4472 {
4473 	struct compat_f2fs_gc_range __user *urange;
4474 	struct f2fs_gc_range range;
4475 	int err;
4476 
4477 	urange = compat_ptr(arg);
4478 	err = get_user(range.sync, &urange->sync);
4479 	err |= get_user(range.start, &urange->start);
4480 	err |= get_user(range.len, &urange->len);
4481 	if (err)
4482 		return -EFAULT;
4483 
4484 	return __f2fs_ioc_gc_range(file, &range);
4485 }
4486 
4487 struct compat_f2fs_move_range {
4488 	u32 dst_fd;
4489 	compat_u64 pos_in;
4490 	compat_u64 pos_out;
4491 	compat_u64 len;
4492 };
4493 #define F2FS_IOC32_MOVE_RANGE		_IOWR(F2FS_IOCTL_MAGIC, 9,	\
4494 					struct compat_f2fs_move_range)
4495 
4496 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4497 {
4498 	struct compat_f2fs_move_range __user *urange;
4499 	struct f2fs_move_range range;
4500 	int err;
4501 
4502 	urange = compat_ptr(arg);
4503 	err = get_user(range.dst_fd, &urange->dst_fd);
4504 	err |= get_user(range.pos_in, &urange->pos_in);
4505 	err |= get_user(range.pos_out, &urange->pos_out);
4506 	err |= get_user(range.len, &urange->len);
4507 	if (err)
4508 		return -EFAULT;
4509 
4510 	return __f2fs_ioc_move_range(file, &range);
4511 }
4512 
4513 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4514 {
4515 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4516 		return -EIO;
4517 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4518 		return -ENOSPC;
4519 
4520 	switch (cmd) {
4521 	case FS_IOC32_GETFLAGS:
4522 		cmd = FS_IOC_GETFLAGS;
4523 		break;
4524 	case FS_IOC32_SETFLAGS:
4525 		cmd = FS_IOC_SETFLAGS;
4526 		break;
4527 	case FS_IOC32_GETVERSION:
4528 		cmd = FS_IOC_GETVERSION;
4529 		break;
4530 	case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4531 		return f2fs_compat_ioc_gc_range(file, arg);
4532 	case F2FS_IOC32_MOVE_RANGE:
4533 		return f2fs_compat_ioc_move_range(file, arg);
4534 	case F2FS_IOC_START_ATOMIC_WRITE:
4535 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4536 	case F2FS_IOC_START_VOLATILE_WRITE:
4537 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4538 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
4539 	case F2FS_IOC_SHUTDOWN:
4540 	case FITRIM:
4541 	case FS_IOC_SET_ENCRYPTION_POLICY:
4542 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4543 	case FS_IOC_GET_ENCRYPTION_POLICY:
4544 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4545 	case FS_IOC_ADD_ENCRYPTION_KEY:
4546 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4547 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4548 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4549 	case FS_IOC_GET_ENCRYPTION_NONCE:
4550 	case F2FS_IOC_GARBAGE_COLLECT:
4551 	case F2FS_IOC_WRITE_CHECKPOINT:
4552 	case F2FS_IOC_DEFRAGMENT:
4553 	case F2FS_IOC_FLUSH_DEVICE:
4554 	case F2FS_IOC_GET_FEATURES:
4555 	case FS_IOC_FSGETXATTR:
4556 	case FS_IOC_FSSETXATTR:
4557 	case F2FS_IOC_GET_PIN_FILE:
4558 	case F2FS_IOC_SET_PIN_FILE:
4559 	case F2FS_IOC_PRECACHE_EXTENTS:
4560 	case F2FS_IOC_RESIZE_FS:
4561 	case FS_IOC_ENABLE_VERITY:
4562 	case FS_IOC_MEASURE_VERITY:
4563 	case FS_IOC_READ_VERITY_METADATA:
4564 	case FS_IOC_GETFSLABEL:
4565 	case FS_IOC_SETFSLABEL:
4566 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4567 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4568 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4569 	case F2FS_IOC_SEC_TRIM_FILE:
4570 	case F2FS_IOC_GET_COMPRESS_OPTION:
4571 	case F2FS_IOC_SET_COMPRESS_OPTION:
4572 	case F2FS_IOC_DECOMPRESS_FILE:
4573 	case F2FS_IOC_COMPRESS_FILE:
4574 		break;
4575 	default:
4576 		return -ENOIOCTLCMD;
4577 	}
4578 	return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4579 }
4580 #endif
4581 
4582 const struct file_operations f2fs_file_operations = {
4583 	.llseek		= f2fs_llseek,
4584 	.read_iter	= f2fs_file_read_iter,
4585 	.write_iter	= f2fs_file_write_iter,
4586 	.open		= f2fs_file_open,
4587 	.release	= f2fs_release_file,
4588 	.mmap		= f2fs_file_mmap,
4589 	.flush		= f2fs_file_flush,
4590 	.fsync		= f2fs_sync_file,
4591 	.fallocate	= f2fs_fallocate,
4592 	.unlocked_ioctl	= f2fs_ioctl,
4593 #ifdef CONFIG_COMPAT
4594 	.compat_ioctl	= f2fs_compat_ioctl,
4595 #endif
4596 	.splice_read	= generic_file_splice_read,
4597 	.splice_write	= iter_file_splice_write,
4598 };
4599