xref: /openbmc/linux/fs/f2fs/file.c (revision c24c57a4)
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 
25 #include "f2fs.h"
26 #include "node.h"
27 #include "segment.h"
28 #include "xattr.h"
29 #include "acl.h"
30 #include "gc.h"
31 #include "trace.h"
32 #include <trace/events/f2fs.h>
33 
34 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
35 {
36 	struct inode *inode = file_inode(vmf->vma->vm_file);
37 	vm_fault_t ret;
38 
39 	down_read(&F2FS_I(inode)->i_mmap_sem);
40 	ret = filemap_fault(vmf);
41 	up_read(&F2FS_I(inode)->i_mmap_sem);
42 
43 	trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
44 
45 	return ret;
46 }
47 
48 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 {
50 	struct page *page = vmf->page;
51 	struct inode *inode = file_inode(vmf->vma->vm_file);
52 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 	struct dnode_of_data dn = { .node_changed = false };
54 	int err;
55 
56 	if (unlikely(f2fs_cp_error(sbi))) {
57 		err = -EIO;
58 		goto err;
59 	}
60 
61 	if (!f2fs_is_checkpoint_ready(sbi)) {
62 		err = -ENOSPC;
63 		goto err;
64 	}
65 
66 	sb_start_pagefault(inode->i_sb);
67 
68 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
69 
70 	file_update_time(vmf->vma->vm_file);
71 	down_read(&F2FS_I(inode)->i_mmap_sem);
72 	lock_page(page);
73 	if (unlikely(page->mapping != inode->i_mapping ||
74 			page_offset(page) > i_size_read(inode) ||
75 			!PageUptodate(page))) {
76 		unlock_page(page);
77 		err = -EFAULT;
78 		goto out_sem;
79 	}
80 
81 	/* block allocation */
82 	__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
83 	set_new_dnode(&dn, inode, NULL, NULL, 0);
84 	err = f2fs_get_block(&dn, page->index);
85 	f2fs_put_dnode(&dn);
86 	__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
87 	if (err) {
88 		unlock_page(page);
89 		goto out_sem;
90 	}
91 
92 	/* fill the page */
93 	f2fs_wait_on_page_writeback(page, DATA, false, true);
94 
95 	/* wait for GCed page writeback via META_MAPPING */
96 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
97 
98 	/*
99 	 * check to see if the page is mapped already (no holes)
100 	 */
101 	if (PageMappedToDisk(page))
102 		goto out_sem;
103 
104 	/* page is wholly or partially inside EOF */
105 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
106 						i_size_read(inode)) {
107 		loff_t offset;
108 
109 		offset = i_size_read(inode) & ~PAGE_MASK;
110 		zero_user_segment(page, offset, PAGE_SIZE);
111 	}
112 	set_page_dirty(page);
113 	if (!PageUptodate(page))
114 		SetPageUptodate(page);
115 
116 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
117 	f2fs_update_time(sbi, REQ_TIME);
118 
119 	trace_f2fs_vm_page_mkwrite(page, DATA);
120 out_sem:
121 	up_read(&F2FS_I(inode)->i_mmap_sem);
122 
123 	f2fs_balance_fs(sbi, dn.node_changed);
124 
125 	sb_end_pagefault(inode->i_sb);
126 err:
127 	return block_page_mkwrite_return(err);
128 }
129 
130 static const struct vm_operations_struct f2fs_file_vm_ops = {
131 	.fault		= f2fs_filemap_fault,
132 	.map_pages	= filemap_map_pages,
133 	.page_mkwrite	= f2fs_vm_page_mkwrite,
134 };
135 
136 static int get_parent_ino(struct inode *inode, nid_t *pino)
137 {
138 	struct dentry *dentry;
139 
140 	inode = igrab(inode);
141 	dentry = d_find_any_alias(inode);
142 	iput(inode);
143 	if (!dentry)
144 		return 0;
145 
146 	*pino = parent_ino(dentry);
147 	dput(dentry);
148 	return 1;
149 }
150 
151 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
152 {
153 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
154 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
155 
156 	if (!S_ISREG(inode->i_mode))
157 		cp_reason = CP_NON_REGULAR;
158 	else if (inode->i_nlink != 1)
159 		cp_reason = CP_HARDLINK;
160 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
161 		cp_reason = CP_SB_NEED_CP;
162 	else if (file_wrong_pino(inode))
163 		cp_reason = CP_WRONG_PINO;
164 	else if (!f2fs_space_for_roll_forward(sbi))
165 		cp_reason = CP_NO_SPC_ROLL;
166 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
167 		cp_reason = CP_NODE_NEED_CP;
168 	else if (test_opt(sbi, FASTBOOT))
169 		cp_reason = CP_FASTBOOT_MODE;
170 	else if (F2FS_OPTION(sbi).active_logs == 2)
171 		cp_reason = CP_SPEC_LOG_NUM;
172 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
173 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
174 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
175 							TRANS_DIR_INO))
176 		cp_reason = CP_RECOVER_DIR;
177 
178 	return cp_reason;
179 }
180 
181 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
182 {
183 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
184 	bool ret = false;
185 	/* But we need to avoid that there are some inode updates */
186 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
187 		ret = true;
188 	f2fs_put_page(i, 0);
189 	return ret;
190 }
191 
192 static void try_to_fix_pino(struct inode *inode)
193 {
194 	struct f2fs_inode_info *fi = F2FS_I(inode);
195 	nid_t pino;
196 
197 	down_write(&fi->i_sem);
198 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
199 			get_parent_ino(inode, &pino)) {
200 		f2fs_i_pino_write(inode, pino);
201 		file_got_pino(inode);
202 	}
203 	up_write(&fi->i_sem);
204 }
205 
206 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
207 						int datasync, bool atomic)
208 {
209 	struct inode *inode = file->f_mapping->host;
210 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
211 	nid_t ino = inode->i_ino;
212 	int ret = 0;
213 	enum cp_reason_type cp_reason = 0;
214 	struct writeback_control wbc = {
215 		.sync_mode = WB_SYNC_ALL,
216 		.nr_to_write = LONG_MAX,
217 		.for_reclaim = 0,
218 	};
219 	unsigned int seq_id = 0;
220 
221 	if (unlikely(f2fs_readonly(inode->i_sb) ||
222 				is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
223 		return 0;
224 
225 	trace_f2fs_sync_file_enter(inode);
226 
227 	if (S_ISDIR(inode->i_mode))
228 		goto go_write;
229 
230 	/* if fdatasync is triggered, let's do in-place-update */
231 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
232 		set_inode_flag(inode, FI_NEED_IPU);
233 	ret = file_write_and_wait_range(file, start, end);
234 	clear_inode_flag(inode, FI_NEED_IPU);
235 
236 	if (ret) {
237 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
238 		return ret;
239 	}
240 
241 	/* if the inode is dirty, let's recover all the time */
242 	if (!f2fs_skip_inode_update(inode, datasync)) {
243 		f2fs_write_inode(inode, NULL);
244 		goto go_write;
245 	}
246 
247 	/*
248 	 * if there is no written data, don't waste time to write recovery info.
249 	 */
250 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
251 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
252 
253 		/* it may call write_inode just prior to fsync */
254 		if (need_inode_page_update(sbi, ino))
255 			goto go_write;
256 
257 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
258 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
259 			goto flush_out;
260 		goto out;
261 	}
262 go_write:
263 	/*
264 	 * Both of fdatasync() and fsync() are able to be recovered from
265 	 * sudden-power-off.
266 	 */
267 	down_read(&F2FS_I(inode)->i_sem);
268 	cp_reason = need_do_checkpoint(inode);
269 	up_read(&F2FS_I(inode)->i_sem);
270 
271 	if (cp_reason) {
272 		/* all the dirty node pages should be flushed for POR */
273 		ret = f2fs_sync_fs(inode->i_sb, 1);
274 
275 		/*
276 		 * We've secured consistency through sync_fs. Following pino
277 		 * will be used only for fsynced inodes after checkpoint.
278 		 */
279 		try_to_fix_pino(inode);
280 		clear_inode_flag(inode, FI_APPEND_WRITE);
281 		clear_inode_flag(inode, FI_UPDATE_WRITE);
282 		goto out;
283 	}
284 sync_nodes:
285 	atomic_inc(&sbi->wb_sync_req[NODE]);
286 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
287 	atomic_dec(&sbi->wb_sync_req[NODE]);
288 	if (ret)
289 		goto out;
290 
291 	/* if cp_error was enabled, we should avoid infinite loop */
292 	if (unlikely(f2fs_cp_error(sbi))) {
293 		ret = -EIO;
294 		goto out;
295 	}
296 
297 	if (f2fs_need_inode_block_update(sbi, ino)) {
298 		f2fs_mark_inode_dirty_sync(inode, true);
299 		f2fs_write_inode(inode, NULL);
300 		goto sync_nodes;
301 	}
302 
303 	/*
304 	 * If it's atomic_write, it's just fine to keep write ordering. So
305 	 * here we don't need to wait for node write completion, since we use
306 	 * node chain which serializes node blocks. If one of node writes are
307 	 * reordered, we can see simply broken chain, resulting in stopping
308 	 * roll-forward recovery. It means we'll recover all or none node blocks
309 	 * given fsync mark.
310 	 */
311 	if (!atomic) {
312 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
313 		if (ret)
314 			goto out;
315 	}
316 
317 	/* once recovery info is written, don't need to tack this */
318 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
319 	clear_inode_flag(inode, FI_APPEND_WRITE);
320 flush_out:
321 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
322 		ret = f2fs_issue_flush(sbi, inode->i_ino);
323 	if (!ret) {
324 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
325 		clear_inode_flag(inode, FI_UPDATE_WRITE);
326 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
327 	}
328 	f2fs_update_time(sbi, REQ_TIME);
329 out:
330 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
331 	f2fs_trace_ios(NULL, 1);
332 	return ret;
333 }
334 
335 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
336 {
337 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
338 		return -EIO;
339 	return f2fs_do_sync_file(file, start, end, datasync, false);
340 }
341 
342 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
343 						pgoff_t pgofs, int whence)
344 {
345 	struct page *page;
346 	int nr_pages;
347 
348 	if (whence != SEEK_DATA)
349 		return 0;
350 
351 	/* find first dirty page index */
352 	nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
353 				      1, &page);
354 	if (!nr_pages)
355 		return ULONG_MAX;
356 	pgofs = page->index;
357 	put_page(page);
358 	return pgofs;
359 }
360 
361 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
362 				pgoff_t dirty, pgoff_t pgofs, int whence)
363 {
364 	switch (whence) {
365 	case SEEK_DATA:
366 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
367 			__is_valid_data_blkaddr(blkaddr))
368 			return true;
369 		break;
370 	case SEEK_HOLE:
371 		if (blkaddr == NULL_ADDR)
372 			return true;
373 		break;
374 	}
375 	return false;
376 }
377 
378 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
379 {
380 	struct inode *inode = file->f_mapping->host;
381 	loff_t maxbytes = inode->i_sb->s_maxbytes;
382 	struct dnode_of_data dn;
383 	pgoff_t pgofs, end_offset, dirty;
384 	loff_t data_ofs = offset;
385 	loff_t isize;
386 	int err = 0;
387 
388 	inode_lock(inode);
389 
390 	isize = i_size_read(inode);
391 	if (offset >= isize)
392 		goto fail;
393 
394 	/* handle inline data case */
395 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
396 		if (whence == SEEK_HOLE)
397 			data_ofs = isize;
398 		goto found;
399 	}
400 
401 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
402 
403 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
404 
405 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
406 		set_new_dnode(&dn, inode, NULL, NULL, 0);
407 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
408 		if (err && err != -ENOENT) {
409 			goto fail;
410 		} else if (err == -ENOENT) {
411 			/* direct node does not exists */
412 			if (whence == SEEK_DATA) {
413 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
414 				continue;
415 			} else {
416 				goto found;
417 			}
418 		}
419 
420 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
421 
422 		/* find data/hole in dnode block */
423 		for (; dn.ofs_in_node < end_offset;
424 				dn.ofs_in_node++, pgofs++,
425 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
426 			block_t blkaddr;
427 
428 			blkaddr = datablock_addr(dn.inode,
429 					dn.node_page, dn.ofs_in_node);
430 
431 			if (__is_valid_data_blkaddr(blkaddr) &&
432 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
433 					blkaddr, DATA_GENERIC_ENHANCE)) {
434 				f2fs_put_dnode(&dn);
435 				goto fail;
436 			}
437 
438 			if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
439 							pgofs, whence)) {
440 				f2fs_put_dnode(&dn);
441 				goto found;
442 			}
443 		}
444 		f2fs_put_dnode(&dn);
445 	}
446 
447 	if (whence == SEEK_DATA)
448 		goto fail;
449 found:
450 	if (whence == SEEK_HOLE && data_ofs > isize)
451 		data_ofs = isize;
452 	inode_unlock(inode);
453 	return vfs_setpos(file, data_ofs, maxbytes);
454 fail:
455 	inode_unlock(inode);
456 	return -ENXIO;
457 }
458 
459 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
460 {
461 	struct inode *inode = file->f_mapping->host;
462 	loff_t maxbytes = inode->i_sb->s_maxbytes;
463 
464 	switch (whence) {
465 	case SEEK_SET:
466 	case SEEK_CUR:
467 	case SEEK_END:
468 		return generic_file_llseek_size(file, offset, whence,
469 						maxbytes, i_size_read(inode));
470 	case SEEK_DATA:
471 	case SEEK_HOLE:
472 		if (offset < 0)
473 			return -ENXIO;
474 		return f2fs_seek_block(file, offset, whence);
475 	}
476 
477 	return -EINVAL;
478 }
479 
480 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
481 {
482 	struct inode *inode = file_inode(file);
483 	int err;
484 
485 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
486 		return -EIO;
487 
488 	/* we don't need to use inline_data strictly */
489 	err = f2fs_convert_inline_inode(inode);
490 	if (err)
491 		return err;
492 
493 	file_accessed(file);
494 	vma->vm_ops = &f2fs_file_vm_ops;
495 	return 0;
496 }
497 
498 static int f2fs_file_open(struct inode *inode, struct file *filp)
499 {
500 	int err = fscrypt_file_open(inode, filp);
501 
502 	if (err)
503 		return err;
504 
505 	err = fsverity_file_open(inode, filp);
506 	if (err)
507 		return err;
508 
509 	filp->f_mode |= FMODE_NOWAIT;
510 
511 	return dquot_file_open(inode, filp);
512 }
513 
514 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
515 {
516 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
517 	struct f2fs_node *raw_node;
518 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
519 	__le32 *addr;
520 	int base = 0;
521 
522 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
523 		base = get_extra_isize(dn->inode);
524 
525 	raw_node = F2FS_NODE(dn->node_page);
526 	addr = blkaddr_in_node(raw_node) + base + ofs;
527 
528 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
529 		block_t blkaddr = le32_to_cpu(*addr);
530 
531 		if (blkaddr == NULL_ADDR)
532 			continue;
533 
534 		dn->data_blkaddr = NULL_ADDR;
535 		f2fs_set_data_blkaddr(dn);
536 
537 		if (__is_valid_data_blkaddr(blkaddr) &&
538 			!f2fs_is_valid_blkaddr(sbi, blkaddr,
539 					DATA_GENERIC_ENHANCE))
540 			continue;
541 
542 		f2fs_invalidate_blocks(sbi, blkaddr);
543 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
544 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
545 		nr_free++;
546 	}
547 
548 	if (nr_free) {
549 		pgoff_t fofs;
550 		/*
551 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
552 		 * we will invalidate all blkaddr in the whole range.
553 		 */
554 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
555 							dn->inode) + ofs;
556 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
557 		dec_valid_block_count(sbi, dn->inode, nr_free);
558 	}
559 	dn->ofs_in_node = ofs;
560 
561 	f2fs_update_time(sbi, REQ_TIME);
562 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
563 					 dn->ofs_in_node, nr_free);
564 }
565 
566 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
567 {
568 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
569 }
570 
571 static int truncate_partial_data_page(struct inode *inode, u64 from,
572 								bool cache_only)
573 {
574 	loff_t offset = from & (PAGE_SIZE - 1);
575 	pgoff_t index = from >> PAGE_SHIFT;
576 	struct address_space *mapping = inode->i_mapping;
577 	struct page *page;
578 
579 	if (!offset && !cache_only)
580 		return 0;
581 
582 	if (cache_only) {
583 		page = find_lock_page(mapping, index);
584 		if (page && PageUptodate(page))
585 			goto truncate_out;
586 		f2fs_put_page(page, 1);
587 		return 0;
588 	}
589 
590 	page = f2fs_get_lock_data_page(inode, index, true);
591 	if (IS_ERR(page))
592 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
593 truncate_out:
594 	f2fs_wait_on_page_writeback(page, DATA, true, true);
595 	zero_user(page, offset, PAGE_SIZE - offset);
596 
597 	/* An encrypted inode should have a key and truncate the last page. */
598 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
599 	if (!cache_only)
600 		set_page_dirty(page);
601 	f2fs_put_page(page, 1);
602 	return 0;
603 }
604 
605 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
606 {
607 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
608 	struct dnode_of_data dn;
609 	pgoff_t free_from;
610 	int count = 0, err = 0;
611 	struct page *ipage;
612 	bool truncate_page = false;
613 
614 	trace_f2fs_truncate_blocks_enter(inode, from);
615 
616 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
617 
618 	if (free_from >= sbi->max_file_blocks)
619 		goto free_partial;
620 
621 	if (lock)
622 		f2fs_lock_op(sbi);
623 
624 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
625 	if (IS_ERR(ipage)) {
626 		err = PTR_ERR(ipage);
627 		goto out;
628 	}
629 
630 	if (f2fs_has_inline_data(inode)) {
631 		f2fs_truncate_inline_inode(inode, ipage, from);
632 		f2fs_put_page(ipage, 1);
633 		truncate_page = true;
634 		goto out;
635 	}
636 
637 	set_new_dnode(&dn, inode, ipage, NULL, 0);
638 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
639 	if (err) {
640 		if (err == -ENOENT)
641 			goto free_next;
642 		goto out;
643 	}
644 
645 	count = ADDRS_PER_PAGE(dn.node_page, inode);
646 
647 	count -= dn.ofs_in_node;
648 	f2fs_bug_on(sbi, count < 0);
649 
650 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
651 		f2fs_truncate_data_blocks_range(&dn, count);
652 		free_from += count;
653 	}
654 
655 	f2fs_put_dnode(&dn);
656 free_next:
657 	err = f2fs_truncate_inode_blocks(inode, free_from);
658 out:
659 	if (lock)
660 		f2fs_unlock_op(sbi);
661 free_partial:
662 	/* lastly zero out the first data page */
663 	if (!err)
664 		err = truncate_partial_data_page(inode, from, truncate_page);
665 
666 	trace_f2fs_truncate_blocks_exit(inode, err);
667 	return err;
668 }
669 
670 int f2fs_truncate(struct inode *inode)
671 {
672 	int err;
673 
674 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
675 		return -EIO;
676 
677 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
678 				S_ISLNK(inode->i_mode)))
679 		return 0;
680 
681 	trace_f2fs_truncate(inode);
682 
683 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
684 		f2fs_show_injection_info(FAULT_TRUNCATE);
685 		return -EIO;
686 	}
687 
688 	/* we should check inline_data size */
689 	if (!f2fs_may_inline_data(inode)) {
690 		err = f2fs_convert_inline_inode(inode);
691 		if (err)
692 			return err;
693 	}
694 
695 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
696 	if (err)
697 		return err;
698 
699 	inode->i_mtime = inode->i_ctime = current_time(inode);
700 	f2fs_mark_inode_dirty_sync(inode, false);
701 	return 0;
702 }
703 
704 int f2fs_getattr(const struct path *path, struct kstat *stat,
705 		 u32 request_mask, unsigned int query_flags)
706 {
707 	struct inode *inode = d_inode(path->dentry);
708 	struct f2fs_inode_info *fi = F2FS_I(inode);
709 	struct f2fs_inode *ri;
710 	unsigned int flags;
711 
712 	if (f2fs_has_extra_attr(inode) &&
713 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
714 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
715 		stat->result_mask |= STATX_BTIME;
716 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
717 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
718 	}
719 
720 	flags = fi->i_flags;
721 	if (flags & F2FS_APPEND_FL)
722 		stat->attributes |= STATX_ATTR_APPEND;
723 	if (IS_ENCRYPTED(inode))
724 		stat->attributes |= STATX_ATTR_ENCRYPTED;
725 	if (flags & F2FS_IMMUTABLE_FL)
726 		stat->attributes |= STATX_ATTR_IMMUTABLE;
727 	if (flags & F2FS_NODUMP_FL)
728 		stat->attributes |= STATX_ATTR_NODUMP;
729 	if (IS_VERITY(inode))
730 		stat->attributes |= STATX_ATTR_VERITY;
731 
732 	stat->attributes_mask |= (STATX_ATTR_APPEND |
733 				  STATX_ATTR_ENCRYPTED |
734 				  STATX_ATTR_IMMUTABLE |
735 				  STATX_ATTR_NODUMP |
736 				  STATX_ATTR_VERITY);
737 
738 	generic_fillattr(inode, stat);
739 
740 	/* we need to show initial sectors used for inline_data/dentries */
741 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
742 					f2fs_has_inline_dentry(inode))
743 		stat->blocks += (stat->size + 511) >> 9;
744 
745 	return 0;
746 }
747 
748 #ifdef CONFIG_F2FS_FS_POSIX_ACL
749 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
750 {
751 	unsigned int ia_valid = attr->ia_valid;
752 
753 	if (ia_valid & ATTR_UID)
754 		inode->i_uid = attr->ia_uid;
755 	if (ia_valid & ATTR_GID)
756 		inode->i_gid = attr->ia_gid;
757 	if (ia_valid & ATTR_ATIME) {
758 		inode->i_atime = timestamp_truncate(attr->ia_atime,
759 						  inode);
760 	}
761 	if (ia_valid & ATTR_MTIME) {
762 		inode->i_mtime = timestamp_truncate(attr->ia_mtime,
763 						  inode);
764 	}
765 	if (ia_valid & ATTR_CTIME) {
766 		inode->i_ctime = timestamp_truncate(attr->ia_ctime,
767 						  inode);
768 	}
769 	if (ia_valid & ATTR_MODE) {
770 		umode_t mode = attr->ia_mode;
771 
772 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
773 			mode &= ~S_ISGID;
774 		set_acl_inode(inode, mode);
775 	}
776 }
777 #else
778 #define __setattr_copy setattr_copy
779 #endif
780 
781 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
782 {
783 	struct inode *inode = d_inode(dentry);
784 	int err;
785 
786 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
787 		return -EIO;
788 
789 	err = setattr_prepare(dentry, attr);
790 	if (err)
791 		return err;
792 
793 	err = fscrypt_prepare_setattr(dentry, attr);
794 	if (err)
795 		return err;
796 
797 	err = fsverity_prepare_setattr(dentry, attr);
798 	if (err)
799 		return err;
800 
801 	if (is_quota_modification(inode, attr)) {
802 		err = dquot_initialize(inode);
803 		if (err)
804 			return err;
805 	}
806 	if ((attr->ia_valid & ATTR_UID &&
807 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
808 		(attr->ia_valid & ATTR_GID &&
809 		!gid_eq(attr->ia_gid, inode->i_gid))) {
810 		f2fs_lock_op(F2FS_I_SB(inode));
811 		err = dquot_transfer(inode, attr);
812 		if (err) {
813 			set_sbi_flag(F2FS_I_SB(inode),
814 					SBI_QUOTA_NEED_REPAIR);
815 			f2fs_unlock_op(F2FS_I_SB(inode));
816 			return err;
817 		}
818 		/*
819 		 * update uid/gid under lock_op(), so that dquot and inode can
820 		 * be updated atomically.
821 		 */
822 		if (attr->ia_valid & ATTR_UID)
823 			inode->i_uid = attr->ia_uid;
824 		if (attr->ia_valid & ATTR_GID)
825 			inode->i_gid = attr->ia_gid;
826 		f2fs_mark_inode_dirty_sync(inode, true);
827 		f2fs_unlock_op(F2FS_I_SB(inode));
828 	}
829 
830 	if (attr->ia_valid & ATTR_SIZE) {
831 		loff_t old_size = i_size_read(inode);
832 
833 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
834 			/*
835 			 * should convert inline inode before i_size_write to
836 			 * keep smaller than inline_data size with inline flag.
837 			 */
838 			err = f2fs_convert_inline_inode(inode);
839 			if (err)
840 				return err;
841 		}
842 
843 		down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
844 		down_write(&F2FS_I(inode)->i_mmap_sem);
845 
846 		truncate_setsize(inode, attr->ia_size);
847 
848 		if (attr->ia_size <= old_size)
849 			err = f2fs_truncate(inode);
850 		/*
851 		 * do not trim all blocks after i_size if target size is
852 		 * larger than i_size.
853 		 */
854 		up_write(&F2FS_I(inode)->i_mmap_sem);
855 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
856 		if (err)
857 			return err;
858 
859 		down_write(&F2FS_I(inode)->i_sem);
860 		inode->i_mtime = inode->i_ctime = current_time(inode);
861 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
862 		up_write(&F2FS_I(inode)->i_sem);
863 	}
864 
865 	__setattr_copy(inode, attr);
866 
867 	if (attr->ia_valid & ATTR_MODE) {
868 		err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
869 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
870 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
871 			clear_inode_flag(inode, FI_ACL_MODE);
872 		}
873 	}
874 
875 	/* file size may changed here */
876 	f2fs_mark_inode_dirty_sync(inode, true);
877 
878 	/* inode change will produce dirty node pages flushed by checkpoint */
879 	f2fs_balance_fs(F2FS_I_SB(inode), true);
880 
881 	return err;
882 }
883 
884 const struct inode_operations f2fs_file_inode_operations = {
885 	.getattr	= f2fs_getattr,
886 	.setattr	= f2fs_setattr,
887 	.get_acl	= f2fs_get_acl,
888 	.set_acl	= f2fs_set_acl,
889 #ifdef CONFIG_F2FS_FS_XATTR
890 	.listxattr	= f2fs_listxattr,
891 #endif
892 	.fiemap		= f2fs_fiemap,
893 };
894 
895 static int fill_zero(struct inode *inode, pgoff_t index,
896 					loff_t start, loff_t len)
897 {
898 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
899 	struct page *page;
900 
901 	if (!len)
902 		return 0;
903 
904 	f2fs_balance_fs(sbi, true);
905 
906 	f2fs_lock_op(sbi);
907 	page = f2fs_get_new_data_page(inode, NULL, index, false);
908 	f2fs_unlock_op(sbi);
909 
910 	if (IS_ERR(page))
911 		return PTR_ERR(page);
912 
913 	f2fs_wait_on_page_writeback(page, DATA, true, true);
914 	zero_user(page, start, len);
915 	set_page_dirty(page);
916 	f2fs_put_page(page, 1);
917 	return 0;
918 }
919 
920 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
921 {
922 	int err;
923 
924 	while (pg_start < pg_end) {
925 		struct dnode_of_data dn;
926 		pgoff_t end_offset, count;
927 
928 		set_new_dnode(&dn, inode, NULL, NULL, 0);
929 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
930 		if (err) {
931 			if (err == -ENOENT) {
932 				pg_start = f2fs_get_next_page_offset(&dn,
933 								pg_start);
934 				continue;
935 			}
936 			return err;
937 		}
938 
939 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
940 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
941 
942 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
943 
944 		f2fs_truncate_data_blocks_range(&dn, count);
945 		f2fs_put_dnode(&dn);
946 
947 		pg_start += count;
948 	}
949 	return 0;
950 }
951 
952 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
953 {
954 	pgoff_t pg_start, pg_end;
955 	loff_t off_start, off_end;
956 	int ret;
957 
958 	ret = f2fs_convert_inline_inode(inode);
959 	if (ret)
960 		return ret;
961 
962 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
963 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
964 
965 	off_start = offset & (PAGE_SIZE - 1);
966 	off_end = (offset + len) & (PAGE_SIZE - 1);
967 
968 	if (pg_start == pg_end) {
969 		ret = fill_zero(inode, pg_start, off_start,
970 						off_end - off_start);
971 		if (ret)
972 			return ret;
973 	} else {
974 		if (off_start) {
975 			ret = fill_zero(inode, pg_start++, off_start,
976 						PAGE_SIZE - off_start);
977 			if (ret)
978 				return ret;
979 		}
980 		if (off_end) {
981 			ret = fill_zero(inode, pg_end, 0, off_end);
982 			if (ret)
983 				return ret;
984 		}
985 
986 		if (pg_start < pg_end) {
987 			struct address_space *mapping = inode->i_mapping;
988 			loff_t blk_start, blk_end;
989 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
990 
991 			f2fs_balance_fs(sbi, true);
992 
993 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
994 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
995 
996 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
997 			down_write(&F2FS_I(inode)->i_mmap_sem);
998 
999 			truncate_inode_pages_range(mapping, blk_start,
1000 					blk_end - 1);
1001 
1002 			f2fs_lock_op(sbi);
1003 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1004 			f2fs_unlock_op(sbi);
1005 
1006 			up_write(&F2FS_I(inode)->i_mmap_sem);
1007 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1008 		}
1009 	}
1010 
1011 	return ret;
1012 }
1013 
1014 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1015 				int *do_replace, pgoff_t off, pgoff_t len)
1016 {
1017 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1018 	struct dnode_of_data dn;
1019 	int ret, done, i;
1020 
1021 next_dnode:
1022 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1023 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1024 	if (ret && ret != -ENOENT) {
1025 		return ret;
1026 	} else if (ret == -ENOENT) {
1027 		if (dn.max_level == 0)
1028 			return -ENOENT;
1029 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1030 									len);
1031 		blkaddr += done;
1032 		do_replace += done;
1033 		goto next;
1034 	}
1035 
1036 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1037 							dn.ofs_in_node, len);
1038 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1039 		*blkaddr = datablock_addr(dn.inode,
1040 					dn.node_page, dn.ofs_in_node);
1041 
1042 		if (__is_valid_data_blkaddr(*blkaddr) &&
1043 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1044 					DATA_GENERIC_ENHANCE)) {
1045 			f2fs_put_dnode(&dn);
1046 			return -EFSCORRUPTED;
1047 		}
1048 
1049 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1050 
1051 			if (test_opt(sbi, LFS)) {
1052 				f2fs_put_dnode(&dn);
1053 				return -EOPNOTSUPP;
1054 			}
1055 
1056 			/* do not invalidate this block address */
1057 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1058 			*do_replace = 1;
1059 		}
1060 	}
1061 	f2fs_put_dnode(&dn);
1062 next:
1063 	len -= done;
1064 	off += done;
1065 	if (len)
1066 		goto next_dnode;
1067 	return 0;
1068 }
1069 
1070 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1071 				int *do_replace, pgoff_t off, int len)
1072 {
1073 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1074 	struct dnode_of_data dn;
1075 	int ret, i;
1076 
1077 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1078 		if (*do_replace == 0)
1079 			continue;
1080 
1081 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1082 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1083 		if (ret) {
1084 			dec_valid_block_count(sbi, inode, 1);
1085 			f2fs_invalidate_blocks(sbi, *blkaddr);
1086 		} else {
1087 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1088 		}
1089 		f2fs_put_dnode(&dn);
1090 	}
1091 	return 0;
1092 }
1093 
1094 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1095 			block_t *blkaddr, int *do_replace,
1096 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1097 {
1098 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1099 	pgoff_t i = 0;
1100 	int ret;
1101 
1102 	while (i < len) {
1103 		if (blkaddr[i] == NULL_ADDR && !full) {
1104 			i++;
1105 			continue;
1106 		}
1107 
1108 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1109 			struct dnode_of_data dn;
1110 			struct node_info ni;
1111 			size_t new_size;
1112 			pgoff_t ilen;
1113 
1114 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1115 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1116 			if (ret)
1117 				return ret;
1118 
1119 			ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1120 			if (ret) {
1121 				f2fs_put_dnode(&dn);
1122 				return ret;
1123 			}
1124 
1125 			ilen = min((pgoff_t)
1126 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1127 						dn.ofs_in_node, len - i);
1128 			do {
1129 				dn.data_blkaddr = datablock_addr(dn.inode,
1130 						dn.node_page, dn.ofs_in_node);
1131 				f2fs_truncate_data_blocks_range(&dn, 1);
1132 
1133 				if (do_replace[i]) {
1134 					f2fs_i_blocks_write(src_inode,
1135 							1, false, false);
1136 					f2fs_i_blocks_write(dst_inode,
1137 							1, true, false);
1138 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1139 					blkaddr[i], ni.version, true, false);
1140 
1141 					do_replace[i] = 0;
1142 				}
1143 				dn.ofs_in_node++;
1144 				i++;
1145 				new_size = (dst + i) << PAGE_SHIFT;
1146 				if (dst_inode->i_size < new_size)
1147 					f2fs_i_size_write(dst_inode, new_size);
1148 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1149 
1150 			f2fs_put_dnode(&dn);
1151 		} else {
1152 			struct page *psrc, *pdst;
1153 
1154 			psrc = f2fs_get_lock_data_page(src_inode,
1155 							src + i, true);
1156 			if (IS_ERR(psrc))
1157 				return PTR_ERR(psrc);
1158 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1159 								true);
1160 			if (IS_ERR(pdst)) {
1161 				f2fs_put_page(psrc, 1);
1162 				return PTR_ERR(pdst);
1163 			}
1164 			f2fs_copy_page(psrc, pdst);
1165 			set_page_dirty(pdst);
1166 			f2fs_put_page(pdst, 1);
1167 			f2fs_put_page(psrc, 1);
1168 
1169 			ret = f2fs_truncate_hole(src_inode,
1170 						src + i, src + i + 1);
1171 			if (ret)
1172 				return ret;
1173 			i++;
1174 		}
1175 	}
1176 	return 0;
1177 }
1178 
1179 static int __exchange_data_block(struct inode *src_inode,
1180 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1181 			pgoff_t len, bool full)
1182 {
1183 	block_t *src_blkaddr;
1184 	int *do_replace;
1185 	pgoff_t olen;
1186 	int ret;
1187 
1188 	while (len) {
1189 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1190 
1191 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1192 					array_size(olen, sizeof(block_t)),
1193 					GFP_KERNEL);
1194 		if (!src_blkaddr)
1195 			return -ENOMEM;
1196 
1197 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1198 					array_size(olen, sizeof(int)),
1199 					GFP_KERNEL);
1200 		if (!do_replace) {
1201 			kvfree(src_blkaddr);
1202 			return -ENOMEM;
1203 		}
1204 
1205 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1206 					do_replace, src, olen);
1207 		if (ret)
1208 			goto roll_back;
1209 
1210 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1211 					do_replace, src, dst, olen, full);
1212 		if (ret)
1213 			goto roll_back;
1214 
1215 		src += olen;
1216 		dst += olen;
1217 		len -= olen;
1218 
1219 		kvfree(src_blkaddr);
1220 		kvfree(do_replace);
1221 	}
1222 	return 0;
1223 
1224 roll_back:
1225 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1226 	kvfree(src_blkaddr);
1227 	kvfree(do_replace);
1228 	return ret;
1229 }
1230 
1231 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1232 {
1233 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1234 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1235 	pgoff_t start = offset >> PAGE_SHIFT;
1236 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1237 	int ret;
1238 
1239 	f2fs_balance_fs(sbi, true);
1240 
1241 	/* avoid gc operation during block exchange */
1242 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1243 	down_write(&F2FS_I(inode)->i_mmap_sem);
1244 
1245 	f2fs_lock_op(sbi);
1246 	f2fs_drop_extent_tree(inode);
1247 	truncate_pagecache(inode, offset);
1248 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1249 	f2fs_unlock_op(sbi);
1250 
1251 	up_write(&F2FS_I(inode)->i_mmap_sem);
1252 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1253 	return ret;
1254 }
1255 
1256 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1257 {
1258 	loff_t new_size;
1259 	int ret;
1260 
1261 	if (offset + len >= i_size_read(inode))
1262 		return -EINVAL;
1263 
1264 	/* collapse range should be aligned to block size of f2fs. */
1265 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1266 		return -EINVAL;
1267 
1268 	ret = f2fs_convert_inline_inode(inode);
1269 	if (ret)
1270 		return ret;
1271 
1272 	/* write out all dirty pages from offset */
1273 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1274 	if (ret)
1275 		return ret;
1276 
1277 	ret = f2fs_do_collapse(inode, offset, len);
1278 	if (ret)
1279 		return ret;
1280 
1281 	/* write out all moved pages, if possible */
1282 	down_write(&F2FS_I(inode)->i_mmap_sem);
1283 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1284 	truncate_pagecache(inode, offset);
1285 
1286 	new_size = i_size_read(inode) - len;
1287 	truncate_pagecache(inode, new_size);
1288 
1289 	ret = f2fs_truncate_blocks(inode, new_size, true);
1290 	up_write(&F2FS_I(inode)->i_mmap_sem);
1291 	if (!ret)
1292 		f2fs_i_size_write(inode, new_size);
1293 	return ret;
1294 }
1295 
1296 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1297 								pgoff_t end)
1298 {
1299 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1300 	pgoff_t index = start;
1301 	unsigned int ofs_in_node = dn->ofs_in_node;
1302 	blkcnt_t count = 0;
1303 	int ret;
1304 
1305 	for (; index < end; index++, dn->ofs_in_node++) {
1306 		if (datablock_addr(dn->inode, dn->node_page,
1307 					dn->ofs_in_node) == NULL_ADDR)
1308 			count++;
1309 	}
1310 
1311 	dn->ofs_in_node = ofs_in_node;
1312 	ret = f2fs_reserve_new_blocks(dn, count);
1313 	if (ret)
1314 		return ret;
1315 
1316 	dn->ofs_in_node = ofs_in_node;
1317 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1318 		dn->data_blkaddr = datablock_addr(dn->inode,
1319 					dn->node_page, dn->ofs_in_node);
1320 		/*
1321 		 * f2fs_reserve_new_blocks will not guarantee entire block
1322 		 * allocation.
1323 		 */
1324 		if (dn->data_blkaddr == NULL_ADDR) {
1325 			ret = -ENOSPC;
1326 			break;
1327 		}
1328 		if (dn->data_blkaddr != NEW_ADDR) {
1329 			f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1330 			dn->data_blkaddr = NEW_ADDR;
1331 			f2fs_set_data_blkaddr(dn);
1332 		}
1333 	}
1334 
1335 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1336 
1337 	return ret;
1338 }
1339 
1340 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1341 								int mode)
1342 {
1343 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1344 	struct address_space *mapping = inode->i_mapping;
1345 	pgoff_t index, pg_start, pg_end;
1346 	loff_t new_size = i_size_read(inode);
1347 	loff_t off_start, off_end;
1348 	int ret = 0;
1349 
1350 	ret = inode_newsize_ok(inode, (len + offset));
1351 	if (ret)
1352 		return ret;
1353 
1354 	ret = f2fs_convert_inline_inode(inode);
1355 	if (ret)
1356 		return ret;
1357 
1358 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1359 	if (ret)
1360 		return ret;
1361 
1362 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1363 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1364 
1365 	off_start = offset & (PAGE_SIZE - 1);
1366 	off_end = (offset + len) & (PAGE_SIZE - 1);
1367 
1368 	if (pg_start == pg_end) {
1369 		ret = fill_zero(inode, pg_start, off_start,
1370 						off_end - off_start);
1371 		if (ret)
1372 			return ret;
1373 
1374 		new_size = max_t(loff_t, new_size, offset + len);
1375 	} else {
1376 		if (off_start) {
1377 			ret = fill_zero(inode, pg_start++, off_start,
1378 						PAGE_SIZE - off_start);
1379 			if (ret)
1380 				return ret;
1381 
1382 			new_size = max_t(loff_t, new_size,
1383 					(loff_t)pg_start << PAGE_SHIFT);
1384 		}
1385 
1386 		for (index = pg_start; index < pg_end;) {
1387 			struct dnode_of_data dn;
1388 			unsigned int end_offset;
1389 			pgoff_t end;
1390 
1391 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1392 			down_write(&F2FS_I(inode)->i_mmap_sem);
1393 
1394 			truncate_pagecache_range(inode,
1395 				(loff_t)index << PAGE_SHIFT,
1396 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1397 
1398 			f2fs_lock_op(sbi);
1399 
1400 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1401 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1402 			if (ret) {
1403 				f2fs_unlock_op(sbi);
1404 				up_write(&F2FS_I(inode)->i_mmap_sem);
1405 				up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1406 				goto out;
1407 			}
1408 
1409 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1410 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1411 
1412 			ret = f2fs_do_zero_range(&dn, index, end);
1413 			f2fs_put_dnode(&dn);
1414 
1415 			f2fs_unlock_op(sbi);
1416 			up_write(&F2FS_I(inode)->i_mmap_sem);
1417 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1418 
1419 			f2fs_balance_fs(sbi, dn.node_changed);
1420 
1421 			if (ret)
1422 				goto out;
1423 
1424 			index = end;
1425 			new_size = max_t(loff_t, new_size,
1426 					(loff_t)index << PAGE_SHIFT);
1427 		}
1428 
1429 		if (off_end) {
1430 			ret = fill_zero(inode, pg_end, 0, off_end);
1431 			if (ret)
1432 				goto out;
1433 
1434 			new_size = max_t(loff_t, new_size, offset + len);
1435 		}
1436 	}
1437 
1438 out:
1439 	if (new_size > i_size_read(inode)) {
1440 		if (mode & FALLOC_FL_KEEP_SIZE)
1441 			file_set_keep_isize(inode);
1442 		else
1443 			f2fs_i_size_write(inode, new_size);
1444 	}
1445 	return ret;
1446 }
1447 
1448 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1449 {
1450 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1451 	pgoff_t nr, pg_start, pg_end, delta, idx;
1452 	loff_t new_size;
1453 	int ret = 0;
1454 
1455 	new_size = i_size_read(inode) + len;
1456 	ret = inode_newsize_ok(inode, new_size);
1457 	if (ret)
1458 		return ret;
1459 
1460 	if (offset >= i_size_read(inode))
1461 		return -EINVAL;
1462 
1463 	/* insert range should be aligned to block size of f2fs. */
1464 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1465 		return -EINVAL;
1466 
1467 	ret = f2fs_convert_inline_inode(inode);
1468 	if (ret)
1469 		return ret;
1470 
1471 	f2fs_balance_fs(sbi, true);
1472 
1473 	down_write(&F2FS_I(inode)->i_mmap_sem);
1474 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1475 	up_write(&F2FS_I(inode)->i_mmap_sem);
1476 	if (ret)
1477 		return ret;
1478 
1479 	/* write out all dirty pages from offset */
1480 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1481 	if (ret)
1482 		return ret;
1483 
1484 	pg_start = offset >> PAGE_SHIFT;
1485 	pg_end = (offset + len) >> PAGE_SHIFT;
1486 	delta = pg_end - pg_start;
1487 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1488 
1489 	/* avoid gc operation during block exchange */
1490 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1491 	down_write(&F2FS_I(inode)->i_mmap_sem);
1492 	truncate_pagecache(inode, offset);
1493 
1494 	while (!ret && idx > pg_start) {
1495 		nr = idx - pg_start;
1496 		if (nr > delta)
1497 			nr = delta;
1498 		idx -= nr;
1499 
1500 		f2fs_lock_op(sbi);
1501 		f2fs_drop_extent_tree(inode);
1502 
1503 		ret = __exchange_data_block(inode, inode, idx,
1504 					idx + delta, nr, false);
1505 		f2fs_unlock_op(sbi);
1506 	}
1507 	up_write(&F2FS_I(inode)->i_mmap_sem);
1508 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1509 
1510 	/* write out all moved pages, if possible */
1511 	down_write(&F2FS_I(inode)->i_mmap_sem);
1512 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1513 	truncate_pagecache(inode, offset);
1514 	up_write(&F2FS_I(inode)->i_mmap_sem);
1515 
1516 	if (!ret)
1517 		f2fs_i_size_write(inode, new_size);
1518 	return ret;
1519 }
1520 
1521 static int expand_inode_data(struct inode *inode, loff_t offset,
1522 					loff_t len, int mode)
1523 {
1524 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1525 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1526 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1527 			.m_may_create = true };
1528 	pgoff_t pg_end;
1529 	loff_t new_size = i_size_read(inode);
1530 	loff_t off_end;
1531 	int err;
1532 
1533 	err = inode_newsize_ok(inode, (len + offset));
1534 	if (err)
1535 		return err;
1536 
1537 	err = f2fs_convert_inline_inode(inode);
1538 	if (err)
1539 		return err;
1540 
1541 	f2fs_balance_fs(sbi, true);
1542 
1543 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1544 	off_end = (offset + len) & (PAGE_SIZE - 1);
1545 
1546 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1547 	map.m_len = pg_end - map.m_lblk;
1548 	if (off_end)
1549 		map.m_len++;
1550 
1551 	if (f2fs_is_pinned_file(inode))
1552 		map.m_seg_type = CURSEG_COLD_DATA;
1553 
1554 	err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
1555 						F2FS_GET_BLOCK_PRE_DIO :
1556 						F2FS_GET_BLOCK_PRE_AIO));
1557 	if (err) {
1558 		pgoff_t last_off;
1559 
1560 		if (!map.m_len)
1561 			return err;
1562 
1563 		last_off = map.m_lblk + map.m_len - 1;
1564 
1565 		/* update new size to the failed position */
1566 		new_size = (last_off == pg_end) ? offset + len :
1567 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1568 	} else {
1569 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1570 	}
1571 
1572 	if (new_size > i_size_read(inode)) {
1573 		if (mode & FALLOC_FL_KEEP_SIZE)
1574 			file_set_keep_isize(inode);
1575 		else
1576 			f2fs_i_size_write(inode, new_size);
1577 	}
1578 
1579 	return err;
1580 }
1581 
1582 static long f2fs_fallocate(struct file *file, int mode,
1583 				loff_t offset, loff_t len)
1584 {
1585 	struct inode *inode = file_inode(file);
1586 	long ret = 0;
1587 
1588 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1589 		return -EIO;
1590 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1591 		return -ENOSPC;
1592 
1593 	/* f2fs only support ->fallocate for regular file */
1594 	if (!S_ISREG(inode->i_mode))
1595 		return -EINVAL;
1596 
1597 	if (IS_ENCRYPTED(inode) &&
1598 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1599 		return -EOPNOTSUPP;
1600 
1601 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1602 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1603 			FALLOC_FL_INSERT_RANGE))
1604 		return -EOPNOTSUPP;
1605 
1606 	inode_lock(inode);
1607 
1608 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1609 		if (offset >= inode->i_size)
1610 			goto out;
1611 
1612 		ret = punch_hole(inode, offset, len);
1613 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1614 		ret = f2fs_collapse_range(inode, offset, len);
1615 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1616 		ret = f2fs_zero_range(inode, offset, len, mode);
1617 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1618 		ret = f2fs_insert_range(inode, offset, len);
1619 	} else {
1620 		ret = expand_inode_data(inode, offset, len, mode);
1621 	}
1622 
1623 	if (!ret) {
1624 		inode->i_mtime = inode->i_ctime = current_time(inode);
1625 		f2fs_mark_inode_dirty_sync(inode, false);
1626 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1627 	}
1628 
1629 out:
1630 	inode_unlock(inode);
1631 
1632 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1633 	return ret;
1634 }
1635 
1636 static int f2fs_release_file(struct inode *inode, struct file *filp)
1637 {
1638 	/*
1639 	 * f2fs_relase_file is called at every close calls. So we should
1640 	 * not drop any inmemory pages by close called by other process.
1641 	 */
1642 	if (!(filp->f_mode & FMODE_WRITE) ||
1643 			atomic_read(&inode->i_writecount) != 1)
1644 		return 0;
1645 
1646 	/* some remained atomic pages should discarded */
1647 	if (f2fs_is_atomic_file(inode))
1648 		f2fs_drop_inmem_pages(inode);
1649 	if (f2fs_is_volatile_file(inode)) {
1650 		set_inode_flag(inode, FI_DROP_CACHE);
1651 		filemap_fdatawrite(inode->i_mapping);
1652 		clear_inode_flag(inode, FI_DROP_CACHE);
1653 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1654 		stat_dec_volatile_write(inode);
1655 	}
1656 	return 0;
1657 }
1658 
1659 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1660 {
1661 	struct inode *inode = file_inode(file);
1662 
1663 	/*
1664 	 * If the process doing a transaction is crashed, we should do
1665 	 * roll-back. Otherwise, other reader/write can see corrupted database
1666 	 * until all the writers close its file. Since this should be done
1667 	 * before dropping file lock, it needs to do in ->flush.
1668 	 */
1669 	if (f2fs_is_atomic_file(inode) &&
1670 			F2FS_I(inode)->inmem_task == current)
1671 		f2fs_drop_inmem_pages(inode);
1672 	return 0;
1673 }
1674 
1675 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1676 {
1677 	struct f2fs_inode_info *fi = F2FS_I(inode);
1678 
1679 	/* Is it quota file? Do not allow user to mess with it */
1680 	if (IS_NOQUOTA(inode))
1681 		return -EPERM;
1682 
1683 	if ((iflags ^ fi->i_flags) & F2FS_CASEFOLD_FL) {
1684 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1685 			return -EOPNOTSUPP;
1686 		if (!f2fs_empty_dir(inode))
1687 			return -ENOTEMPTY;
1688 	}
1689 
1690 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1691 
1692 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1693 		set_inode_flag(inode, FI_PROJ_INHERIT);
1694 	else
1695 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1696 
1697 	inode->i_ctime = current_time(inode);
1698 	f2fs_set_inode_flags(inode);
1699 	f2fs_mark_inode_dirty_sync(inode, true);
1700 	return 0;
1701 }
1702 
1703 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1704 
1705 /*
1706  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1707  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1708  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1709  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1710  */
1711 
1712 static const struct {
1713 	u32 iflag;
1714 	u32 fsflag;
1715 } f2fs_fsflags_map[] = {
1716 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1717 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1718 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1719 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1720 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1721 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1722 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1723 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1724 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
1725 };
1726 
1727 #define F2FS_GETTABLE_FS_FL (		\
1728 		FS_SYNC_FL |		\
1729 		FS_IMMUTABLE_FL |	\
1730 		FS_APPEND_FL |		\
1731 		FS_NODUMP_FL |		\
1732 		FS_NOATIME_FL |		\
1733 		FS_INDEX_FL |		\
1734 		FS_DIRSYNC_FL |		\
1735 		FS_PROJINHERIT_FL |	\
1736 		FS_ENCRYPT_FL |		\
1737 		FS_INLINE_DATA_FL |	\
1738 		FS_NOCOW_FL |		\
1739 		FS_VERITY_FL |		\
1740 		FS_CASEFOLD_FL)
1741 
1742 #define F2FS_SETTABLE_FS_FL (		\
1743 		FS_SYNC_FL |		\
1744 		FS_IMMUTABLE_FL |	\
1745 		FS_APPEND_FL |		\
1746 		FS_NODUMP_FL |		\
1747 		FS_NOATIME_FL |		\
1748 		FS_DIRSYNC_FL |		\
1749 		FS_PROJINHERIT_FL |	\
1750 		FS_CASEFOLD_FL)
1751 
1752 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1753 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1754 {
1755 	u32 fsflags = 0;
1756 	int i;
1757 
1758 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1759 		if (iflags & f2fs_fsflags_map[i].iflag)
1760 			fsflags |= f2fs_fsflags_map[i].fsflag;
1761 
1762 	return fsflags;
1763 }
1764 
1765 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1766 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1767 {
1768 	u32 iflags = 0;
1769 	int i;
1770 
1771 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1772 		if (fsflags & f2fs_fsflags_map[i].fsflag)
1773 			iflags |= f2fs_fsflags_map[i].iflag;
1774 
1775 	return iflags;
1776 }
1777 
1778 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1779 {
1780 	struct inode *inode = file_inode(filp);
1781 	struct f2fs_inode_info *fi = F2FS_I(inode);
1782 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1783 
1784 	if (IS_ENCRYPTED(inode))
1785 		fsflags |= FS_ENCRYPT_FL;
1786 	if (IS_VERITY(inode))
1787 		fsflags |= FS_VERITY_FL;
1788 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1789 		fsflags |= FS_INLINE_DATA_FL;
1790 	if (is_inode_flag_set(inode, FI_PIN_FILE))
1791 		fsflags |= FS_NOCOW_FL;
1792 
1793 	fsflags &= F2FS_GETTABLE_FS_FL;
1794 
1795 	return put_user(fsflags, (int __user *)arg);
1796 }
1797 
1798 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1799 {
1800 	struct inode *inode = file_inode(filp);
1801 	struct f2fs_inode_info *fi = F2FS_I(inode);
1802 	u32 fsflags, old_fsflags;
1803 	u32 iflags;
1804 	int ret;
1805 
1806 	if (!inode_owner_or_capable(inode))
1807 		return -EACCES;
1808 
1809 	if (get_user(fsflags, (int __user *)arg))
1810 		return -EFAULT;
1811 
1812 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
1813 		return -EOPNOTSUPP;
1814 	fsflags &= F2FS_SETTABLE_FS_FL;
1815 
1816 	iflags = f2fs_fsflags_to_iflags(fsflags);
1817 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1818 		return -EOPNOTSUPP;
1819 
1820 	ret = mnt_want_write_file(filp);
1821 	if (ret)
1822 		return ret;
1823 
1824 	inode_lock(inode);
1825 
1826 	old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1827 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1828 	if (ret)
1829 		goto out;
1830 
1831 	ret = f2fs_setflags_common(inode, iflags,
1832 			f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1833 out:
1834 	inode_unlock(inode);
1835 	mnt_drop_write_file(filp);
1836 	return ret;
1837 }
1838 
1839 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1840 {
1841 	struct inode *inode = file_inode(filp);
1842 
1843 	return put_user(inode->i_generation, (int __user *)arg);
1844 }
1845 
1846 static int f2fs_ioc_start_atomic_write(struct file *filp)
1847 {
1848 	struct inode *inode = file_inode(filp);
1849 	struct f2fs_inode_info *fi = F2FS_I(inode);
1850 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1851 	int ret;
1852 
1853 	if (!inode_owner_or_capable(inode))
1854 		return -EACCES;
1855 
1856 	if (!S_ISREG(inode->i_mode))
1857 		return -EINVAL;
1858 
1859 	if (filp->f_flags & O_DIRECT)
1860 		return -EINVAL;
1861 
1862 	ret = mnt_want_write_file(filp);
1863 	if (ret)
1864 		return ret;
1865 
1866 	inode_lock(inode);
1867 
1868 	if (f2fs_is_atomic_file(inode)) {
1869 		if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1870 			ret = -EINVAL;
1871 		goto out;
1872 	}
1873 
1874 	ret = f2fs_convert_inline_inode(inode);
1875 	if (ret)
1876 		goto out;
1877 
1878 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1879 
1880 	/*
1881 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1882 	 * f2fs_is_atomic_file.
1883 	 */
1884 	if (get_dirty_pages(inode))
1885 		f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1886 			  inode->i_ino, get_dirty_pages(inode));
1887 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1888 	if (ret) {
1889 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1890 		goto out;
1891 	}
1892 
1893 	spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
1894 	if (list_empty(&fi->inmem_ilist))
1895 		list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
1896 	spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
1897 
1898 	/* add inode in inmem_list first and set atomic_file */
1899 	set_inode_flag(inode, FI_ATOMIC_FILE);
1900 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1901 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1902 
1903 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1904 	F2FS_I(inode)->inmem_task = current;
1905 	stat_inc_atomic_write(inode);
1906 	stat_update_max_atomic_write(inode);
1907 out:
1908 	inode_unlock(inode);
1909 	mnt_drop_write_file(filp);
1910 	return ret;
1911 }
1912 
1913 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1914 {
1915 	struct inode *inode = file_inode(filp);
1916 	int ret;
1917 
1918 	if (!inode_owner_or_capable(inode))
1919 		return -EACCES;
1920 
1921 	ret = mnt_want_write_file(filp);
1922 	if (ret)
1923 		return ret;
1924 
1925 	f2fs_balance_fs(F2FS_I_SB(inode), true);
1926 
1927 	inode_lock(inode);
1928 
1929 	if (f2fs_is_volatile_file(inode)) {
1930 		ret = -EINVAL;
1931 		goto err_out;
1932 	}
1933 
1934 	if (f2fs_is_atomic_file(inode)) {
1935 		ret = f2fs_commit_inmem_pages(inode);
1936 		if (ret)
1937 			goto err_out;
1938 
1939 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1940 		if (!ret)
1941 			f2fs_drop_inmem_pages(inode);
1942 	} else {
1943 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1944 	}
1945 err_out:
1946 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1947 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1948 		ret = -EINVAL;
1949 	}
1950 	inode_unlock(inode);
1951 	mnt_drop_write_file(filp);
1952 	return ret;
1953 }
1954 
1955 static int f2fs_ioc_start_volatile_write(struct file *filp)
1956 {
1957 	struct inode *inode = file_inode(filp);
1958 	int ret;
1959 
1960 	if (!inode_owner_or_capable(inode))
1961 		return -EACCES;
1962 
1963 	if (!S_ISREG(inode->i_mode))
1964 		return -EINVAL;
1965 
1966 	ret = mnt_want_write_file(filp);
1967 	if (ret)
1968 		return ret;
1969 
1970 	inode_lock(inode);
1971 
1972 	if (f2fs_is_volatile_file(inode))
1973 		goto out;
1974 
1975 	ret = f2fs_convert_inline_inode(inode);
1976 	if (ret)
1977 		goto out;
1978 
1979 	stat_inc_volatile_write(inode);
1980 	stat_update_max_volatile_write(inode);
1981 
1982 	set_inode_flag(inode, FI_VOLATILE_FILE);
1983 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1984 out:
1985 	inode_unlock(inode);
1986 	mnt_drop_write_file(filp);
1987 	return ret;
1988 }
1989 
1990 static int f2fs_ioc_release_volatile_write(struct file *filp)
1991 {
1992 	struct inode *inode = file_inode(filp);
1993 	int ret;
1994 
1995 	if (!inode_owner_or_capable(inode))
1996 		return -EACCES;
1997 
1998 	ret = mnt_want_write_file(filp);
1999 	if (ret)
2000 		return ret;
2001 
2002 	inode_lock(inode);
2003 
2004 	if (!f2fs_is_volatile_file(inode))
2005 		goto out;
2006 
2007 	if (!f2fs_is_first_block_written(inode)) {
2008 		ret = truncate_partial_data_page(inode, 0, true);
2009 		goto out;
2010 	}
2011 
2012 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2013 out:
2014 	inode_unlock(inode);
2015 	mnt_drop_write_file(filp);
2016 	return ret;
2017 }
2018 
2019 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2020 {
2021 	struct inode *inode = file_inode(filp);
2022 	int ret;
2023 
2024 	if (!inode_owner_or_capable(inode))
2025 		return -EACCES;
2026 
2027 	ret = mnt_want_write_file(filp);
2028 	if (ret)
2029 		return ret;
2030 
2031 	inode_lock(inode);
2032 
2033 	if (f2fs_is_atomic_file(inode))
2034 		f2fs_drop_inmem_pages(inode);
2035 	if (f2fs_is_volatile_file(inode)) {
2036 		clear_inode_flag(inode, FI_VOLATILE_FILE);
2037 		stat_dec_volatile_write(inode);
2038 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2039 	}
2040 
2041 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2042 
2043 	inode_unlock(inode);
2044 
2045 	mnt_drop_write_file(filp);
2046 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2047 	return ret;
2048 }
2049 
2050 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2051 {
2052 	struct inode *inode = file_inode(filp);
2053 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2054 	struct super_block *sb = sbi->sb;
2055 	__u32 in;
2056 	int ret = 0;
2057 
2058 	if (!capable(CAP_SYS_ADMIN))
2059 		return -EPERM;
2060 
2061 	if (get_user(in, (__u32 __user *)arg))
2062 		return -EFAULT;
2063 
2064 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2065 		ret = mnt_want_write_file(filp);
2066 		if (ret)
2067 			return ret;
2068 	}
2069 
2070 	switch (in) {
2071 	case F2FS_GOING_DOWN_FULLSYNC:
2072 		sb = freeze_bdev(sb->s_bdev);
2073 		if (IS_ERR(sb)) {
2074 			ret = PTR_ERR(sb);
2075 			goto out;
2076 		}
2077 		if (sb) {
2078 			f2fs_stop_checkpoint(sbi, false);
2079 			set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2080 			thaw_bdev(sb->s_bdev, sb);
2081 		}
2082 		break;
2083 	case F2FS_GOING_DOWN_METASYNC:
2084 		/* do checkpoint only */
2085 		ret = f2fs_sync_fs(sb, 1);
2086 		if (ret)
2087 			goto out;
2088 		f2fs_stop_checkpoint(sbi, false);
2089 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2090 		break;
2091 	case F2FS_GOING_DOWN_NOSYNC:
2092 		f2fs_stop_checkpoint(sbi, false);
2093 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2094 		break;
2095 	case F2FS_GOING_DOWN_METAFLUSH:
2096 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2097 		f2fs_stop_checkpoint(sbi, false);
2098 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2099 		break;
2100 	case F2FS_GOING_DOWN_NEED_FSCK:
2101 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2102 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2103 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2104 		/* do checkpoint only */
2105 		ret = f2fs_sync_fs(sb, 1);
2106 		goto out;
2107 	default:
2108 		ret = -EINVAL;
2109 		goto out;
2110 	}
2111 
2112 	f2fs_stop_gc_thread(sbi);
2113 	f2fs_stop_discard_thread(sbi);
2114 
2115 	f2fs_drop_discard_cmd(sbi);
2116 	clear_opt(sbi, DISCARD);
2117 
2118 	f2fs_update_time(sbi, REQ_TIME);
2119 out:
2120 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2121 		mnt_drop_write_file(filp);
2122 
2123 	trace_f2fs_shutdown(sbi, in, ret);
2124 
2125 	return ret;
2126 }
2127 
2128 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2129 {
2130 	struct inode *inode = file_inode(filp);
2131 	struct super_block *sb = inode->i_sb;
2132 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
2133 	struct fstrim_range range;
2134 	int ret;
2135 
2136 	if (!capable(CAP_SYS_ADMIN))
2137 		return -EPERM;
2138 
2139 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2140 		return -EOPNOTSUPP;
2141 
2142 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2143 				sizeof(range)))
2144 		return -EFAULT;
2145 
2146 	ret = mnt_want_write_file(filp);
2147 	if (ret)
2148 		return ret;
2149 
2150 	range.minlen = max((unsigned int)range.minlen,
2151 				q->limits.discard_granularity);
2152 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2153 	mnt_drop_write_file(filp);
2154 	if (ret < 0)
2155 		return ret;
2156 
2157 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2158 				sizeof(range)))
2159 		return -EFAULT;
2160 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2161 	return 0;
2162 }
2163 
2164 static bool uuid_is_nonzero(__u8 u[16])
2165 {
2166 	int i;
2167 
2168 	for (i = 0; i < 16; i++)
2169 		if (u[i])
2170 			return true;
2171 	return false;
2172 }
2173 
2174 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2175 {
2176 	struct inode *inode = file_inode(filp);
2177 
2178 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2179 		return -EOPNOTSUPP;
2180 
2181 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2182 
2183 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2184 }
2185 
2186 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2187 {
2188 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2189 		return -EOPNOTSUPP;
2190 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2191 }
2192 
2193 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2194 {
2195 	struct inode *inode = file_inode(filp);
2196 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2197 	int err;
2198 
2199 	if (!f2fs_sb_has_encrypt(sbi))
2200 		return -EOPNOTSUPP;
2201 
2202 	err = mnt_want_write_file(filp);
2203 	if (err)
2204 		return err;
2205 
2206 	down_write(&sbi->sb_lock);
2207 
2208 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2209 		goto got_it;
2210 
2211 	/* update superblock with uuid */
2212 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2213 
2214 	err = f2fs_commit_super(sbi, false);
2215 	if (err) {
2216 		/* undo new data */
2217 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2218 		goto out_err;
2219 	}
2220 got_it:
2221 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2222 									16))
2223 		err = -EFAULT;
2224 out_err:
2225 	up_write(&sbi->sb_lock);
2226 	mnt_drop_write_file(filp);
2227 	return err;
2228 }
2229 
2230 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2231 					     unsigned long arg)
2232 {
2233 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2234 		return -EOPNOTSUPP;
2235 
2236 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2237 }
2238 
2239 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2240 {
2241 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2242 		return -EOPNOTSUPP;
2243 
2244 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2245 }
2246 
2247 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2248 {
2249 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2250 		return -EOPNOTSUPP;
2251 
2252 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2253 }
2254 
2255 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2256 						    unsigned long arg)
2257 {
2258 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2259 		return -EOPNOTSUPP;
2260 
2261 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2262 }
2263 
2264 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2265 					      unsigned long arg)
2266 {
2267 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2268 		return -EOPNOTSUPP;
2269 
2270 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2271 }
2272 
2273 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2274 {
2275 	struct inode *inode = file_inode(filp);
2276 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2277 	__u32 sync;
2278 	int ret;
2279 
2280 	if (!capable(CAP_SYS_ADMIN))
2281 		return -EPERM;
2282 
2283 	if (get_user(sync, (__u32 __user *)arg))
2284 		return -EFAULT;
2285 
2286 	if (f2fs_readonly(sbi->sb))
2287 		return -EROFS;
2288 
2289 	ret = mnt_want_write_file(filp);
2290 	if (ret)
2291 		return ret;
2292 
2293 	if (!sync) {
2294 		if (!mutex_trylock(&sbi->gc_mutex)) {
2295 			ret = -EBUSY;
2296 			goto out;
2297 		}
2298 	} else {
2299 		mutex_lock(&sbi->gc_mutex);
2300 	}
2301 
2302 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2303 out:
2304 	mnt_drop_write_file(filp);
2305 	return ret;
2306 }
2307 
2308 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2309 {
2310 	struct inode *inode = file_inode(filp);
2311 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2312 	struct f2fs_gc_range range;
2313 	u64 end;
2314 	int ret;
2315 
2316 	if (!capable(CAP_SYS_ADMIN))
2317 		return -EPERM;
2318 
2319 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2320 							sizeof(range)))
2321 		return -EFAULT;
2322 
2323 	if (f2fs_readonly(sbi->sb))
2324 		return -EROFS;
2325 
2326 	end = range.start + range.len;
2327 	if (end < range.start || range.start < MAIN_BLKADDR(sbi) ||
2328 					end >= MAX_BLKADDR(sbi))
2329 		return -EINVAL;
2330 
2331 	ret = mnt_want_write_file(filp);
2332 	if (ret)
2333 		return ret;
2334 
2335 do_more:
2336 	if (!range.sync) {
2337 		if (!mutex_trylock(&sbi->gc_mutex)) {
2338 			ret = -EBUSY;
2339 			goto out;
2340 		}
2341 	} else {
2342 		mutex_lock(&sbi->gc_mutex);
2343 	}
2344 
2345 	ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2346 	range.start += BLKS_PER_SEC(sbi);
2347 	if (range.start <= end)
2348 		goto do_more;
2349 out:
2350 	mnt_drop_write_file(filp);
2351 	return ret;
2352 }
2353 
2354 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2355 {
2356 	struct inode *inode = file_inode(filp);
2357 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2358 	int ret;
2359 
2360 	if (!capable(CAP_SYS_ADMIN))
2361 		return -EPERM;
2362 
2363 	if (f2fs_readonly(sbi->sb))
2364 		return -EROFS;
2365 
2366 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2367 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2368 		return -EINVAL;
2369 	}
2370 
2371 	ret = mnt_want_write_file(filp);
2372 	if (ret)
2373 		return ret;
2374 
2375 	ret = f2fs_sync_fs(sbi->sb, 1);
2376 
2377 	mnt_drop_write_file(filp);
2378 	return ret;
2379 }
2380 
2381 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2382 					struct file *filp,
2383 					struct f2fs_defragment *range)
2384 {
2385 	struct inode *inode = file_inode(filp);
2386 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2387 					.m_seg_type = NO_CHECK_TYPE ,
2388 					.m_may_create = false };
2389 	struct extent_info ei = {0, 0, 0};
2390 	pgoff_t pg_start, pg_end, next_pgofs;
2391 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2392 	unsigned int total = 0, sec_num;
2393 	block_t blk_end = 0;
2394 	bool fragmented = false;
2395 	int err;
2396 
2397 	/* if in-place-update policy is enabled, don't waste time here */
2398 	if (f2fs_should_update_inplace(inode, NULL))
2399 		return -EINVAL;
2400 
2401 	pg_start = range->start >> PAGE_SHIFT;
2402 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2403 
2404 	f2fs_balance_fs(sbi, true);
2405 
2406 	inode_lock(inode);
2407 
2408 	/* writeback all dirty pages in the range */
2409 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2410 						range->start + range->len - 1);
2411 	if (err)
2412 		goto out;
2413 
2414 	/*
2415 	 * lookup mapping info in extent cache, skip defragmenting if physical
2416 	 * block addresses are continuous.
2417 	 */
2418 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2419 		if (ei.fofs + ei.len >= pg_end)
2420 			goto out;
2421 	}
2422 
2423 	map.m_lblk = pg_start;
2424 	map.m_next_pgofs = &next_pgofs;
2425 
2426 	/*
2427 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2428 	 * physical block addresses are continuous even if there are hole(s)
2429 	 * in logical blocks.
2430 	 */
2431 	while (map.m_lblk < pg_end) {
2432 		map.m_len = pg_end - map.m_lblk;
2433 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2434 		if (err)
2435 			goto out;
2436 
2437 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2438 			map.m_lblk = next_pgofs;
2439 			continue;
2440 		}
2441 
2442 		if (blk_end && blk_end != map.m_pblk)
2443 			fragmented = true;
2444 
2445 		/* record total count of block that we're going to move */
2446 		total += map.m_len;
2447 
2448 		blk_end = map.m_pblk + map.m_len;
2449 
2450 		map.m_lblk += map.m_len;
2451 	}
2452 
2453 	if (!fragmented) {
2454 		total = 0;
2455 		goto out;
2456 	}
2457 
2458 	sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2459 
2460 	/*
2461 	 * make sure there are enough free section for LFS allocation, this can
2462 	 * avoid defragment running in SSR mode when free section are allocated
2463 	 * intensively
2464 	 */
2465 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2466 		err = -EAGAIN;
2467 		goto out;
2468 	}
2469 
2470 	map.m_lblk = pg_start;
2471 	map.m_len = pg_end - pg_start;
2472 	total = 0;
2473 
2474 	while (map.m_lblk < pg_end) {
2475 		pgoff_t idx;
2476 		int cnt = 0;
2477 
2478 do_map:
2479 		map.m_len = pg_end - map.m_lblk;
2480 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2481 		if (err)
2482 			goto clear_out;
2483 
2484 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2485 			map.m_lblk = next_pgofs;
2486 			goto check;
2487 		}
2488 
2489 		set_inode_flag(inode, FI_DO_DEFRAG);
2490 
2491 		idx = map.m_lblk;
2492 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2493 			struct page *page;
2494 
2495 			page = f2fs_get_lock_data_page(inode, idx, true);
2496 			if (IS_ERR(page)) {
2497 				err = PTR_ERR(page);
2498 				goto clear_out;
2499 			}
2500 
2501 			set_page_dirty(page);
2502 			f2fs_put_page(page, 1);
2503 
2504 			idx++;
2505 			cnt++;
2506 			total++;
2507 		}
2508 
2509 		map.m_lblk = idx;
2510 check:
2511 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
2512 			goto do_map;
2513 
2514 		clear_inode_flag(inode, FI_DO_DEFRAG);
2515 
2516 		err = filemap_fdatawrite(inode->i_mapping);
2517 		if (err)
2518 			goto out;
2519 	}
2520 clear_out:
2521 	clear_inode_flag(inode, FI_DO_DEFRAG);
2522 out:
2523 	inode_unlock(inode);
2524 	if (!err)
2525 		range->len = (u64)total << PAGE_SHIFT;
2526 	return err;
2527 }
2528 
2529 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2530 {
2531 	struct inode *inode = file_inode(filp);
2532 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2533 	struct f2fs_defragment range;
2534 	int err;
2535 
2536 	if (!capable(CAP_SYS_ADMIN))
2537 		return -EPERM;
2538 
2539 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2540 		return -EINVAL;
2541 
2542 	if (f2fs_readonly(sbi->sb))
2543 		return -EROFS;
2544 
2545 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2546 							sizeof(range)))
2547 		return -EFAULT;
2548 
2549 	/* verify alignment of offset & size */
2550 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2551 		return -EINVAL;
2552 
2553 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2554 					sbi->max_file_blocks))
2555 		return -EINVAL;
2556 
2557 	err = mnt_want_write_file(filp);
2558 	if (err)
2559 		return err;
2560 
2561 	err = f2fs_defragment_range(sbi, filp, &range);
2562 	mnt_drop_write_file(filp);
2563 
2564 	f2fs_update_time(sbi, REQ_TIME);
2565 	if (err < 0)
2566 		return err;
2567 
2568 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2569 							sizeof(range)))
2570 		return -EFAULT;
2571 
2572 	return 0;
2573 }
2574 
2575 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2576 			struct file *file_out, loff_t pos_out, size_t len)
2577 {
2578 	struct inode *src = file_inode(file_in);
2579 	struct inode *dst = file_inode(file_out);
2580 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2581 	size_t olen = len, dst_max_i_size = 0;
2582 	size_t dst_osize;
2583 	int ret;
2584 
2585 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2586 				src->i_sb != dst->i_sb)
2587 		return -EXDEV;
2588 
2589 	if (unlikely(f2fs_readonly(src->i_sb)))
2590 		return -EROFS;
2591 
2592 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2593 		return -EINVAL;
2594 
2595 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2596 		return -EOPNOTSUPP;
2597 
2598 	if (src == dst) {
2599 		if (pos_in == pos_out)
2600 			return 0;
2601 		if (pos_out > pos_in && pos_out < pos_in + len)
2602 			return -EINVAL;
2603 	}
2604 
2605 	inode_lock(src);
2606 	if (src != dst) {
2607 		ret = -EBUSY;
2608 		if (!inode_trylock(dst))
2609 			goto out;
2610 	}
2611 
2612 	ret = -EINVAL;
2613 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2614 		goto out_unlock;
2615 	if (len == 0)
2616 		olen = len = src->i_size - pos_in;
2617 	if (pos_in + len == src->i_size)
2618 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2619 	if (len == 0) {
2620 		ret = 0;
2621 		goto out_unlock;
2622 	}
2623 
2624 	dst_osize = dst->i_size;
2625 	if (pos_out + olen > dst->i_size)
2626 		dst_max_i_size = pos_out + olen;
2627 
2628 	/* verify the end result is block aligned */
2629 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2630 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2631 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2632 		goto out_unlock;
2633 
2634 	ret = f2fs_convert_inline_inode(src);
2635 	if (ret)
2636 		goto out_unlock;
2637 
2638 	ret = f2fs_convert_inline_inode(dst);
2639 	if (ret)
2640 		goto out_unlock;
2641 
2642 	/* write out all dirty pages from offset */
2643 	ret = filemap_write_and_wait_range(src->i_mapping,
2644 					pos_in, pos_in + len);
2645 	if (ret)
2646 		goto out_unlock;
2647 
2648 	ret = filemap_write_and_wait_range(dst->i_mapping,
2649 					pos_out, pos_out + len);
2650 	if (ret)
2651 		goto out_unlock;
2652 
2653 	f2fs_balance_fs(sbi, true);
2654 
2655 	down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2656 	if (src != dst) {
2657 		ret = -EBUSY;
2658 		if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2659 			goto out_src;
2660 	}
2661 
2662 	f2fs_lock_op(sbi);
2663 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2664 				pos_out >> F2FS_BLKSIZE_BITS,
2665 				len >> F2FS_BLKSIZE_BITS, false);
2666 
2667 	if (!ret) {
2668 		if (dst_max_i_size)
2669 			f2fs_i_size_write(dst, dst_max_i_size);
2670 		else if (dst_osize != dst->i_size)
2671 			f2fs_i_size_write(dst, dst_osize);
2672 	}
2673 	f2fs_unlock_op(sbi);
2674 
2675 	if (src != dst)
2676 		up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2677 out_src:
2678 	up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2679 out_unlock:
2680 	if (src != dst)
2681 		inode_unlock(dst);
2682 out:
2683 	inode_unlock(src);
2684 	return ret;
2685 }
2686 
2687 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2688 {
2689 	struct f2fs_move_range range;
2690 	struct fd dst;
2691 	int err;
2692 
2693 	if (!(filp->f_mode & FMODE_READ) ||
2694 			!(filp->f_mode & FMODE_WRITE))
2695 		return -EBADF;
2696 
2697 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2698 							sizeof(range)))
2699 		return -EFAULT;
2700 
2701 	dst = fdget(range.dst_fd);
2702 	if (!dst.file)
2703 		return -EBADF;
2704 
2705 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2706 		err = -EBADF;
2707 		goto err_out;
2708 	}
2709 
2710 	err = mnt_want_write_file(filp);
2711 	if (err)
2712 		goto err_out;
2713 
2714 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2715 					range.pos_out, range.len);
2716 
2717 	mnt_drop_write_file(filp);
2718 	if (err)
2719 		goto err_out;
2720 
2721 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2722 						&range, sizeof(range)))
2723 		err = -EFAULT;
2724 err_out:
2725 	fdput(dst);
2726 	return err;
2727 }
2728 
2729 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2730 {
2731 	struct inode *inode = file_inode(filp);
2732 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2733 	struct sit_info *sm = SIT_I(sbi);
2734 	unsigned int start_segno = 0, end_segno = 0;
2735 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2736 	struct f2fs_flush_device range;
2737 	int ret;
2738 
2739 	if (!capable(CAP_SYS_ADMIN))
2740 		return -EPERM;
2741 
2742 	if (f2fs_readonly(sbi->sb))
2743 		return -EROFS;
2744 
2745 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2746 		return -EINVAL;
2747 
2748 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2749 							sizeof(range)))
2750 		return -EFAULT;
2751 
2752 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2753 			__is_large_section(sbi)) {
2754 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2755 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2756 		return -EINVAL;
2757 	}
2758 
2759 	ret = mnt_want_write_file(filp);
2760 	if (ret)
2761 		return ret;
2762 
2763 	if (range.dev_num != 0)
2764 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2765 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2766 
2767 	start_segno = sm->last_victim[FLUSH_DEVICE];
2768 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2769 		start_segno = dev_start_segno;
2770 	end_segno = min(start_segno + range.segments, dev_end_segno);
2771 
2772 	while (start_segno < end_segno) {
2773 		if (!mutex_trylock(&sbi->gc_mutex)) {
2774 			ret = -EBUSY;
2775 			goto out;
2776 		}
2777 		sm->last_victim[GC_CB] = end_segno + 1;
2778 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2779 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2780 		ret = f2fs_gc(sbi, true, true, start_segno);
2781 		if (ret == -EAGAIN)
2782 			ret = 0;
2783 		else if (ret < 0)
2784 			break;
2785 		start_segno++;
2786 	}
2787 out:
2788 	mnt_drop_write_file(filp);
2789 	return ret;
2790 }
2791 
2792 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2793 {
2794 	struct inode *inode = file_inode(filp);
2795 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2796 
2797 	/* Must validate to set it with SQLite behavior in Android. */
2798 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2799 
2800 	return put_user(sb_feature, (u32 __user *)arg);
2801 }
2802 
2803 #ifdef CONFIG_QUOTA
2804 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2805 {
2806 	struct dquot *transfer_to[MAXQUOTAS] = {};
2807 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2808 	struct super_block *sb = sbi->sb;
2809 	int err = 0;
2810 
2811 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2812 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2813 		err = __dquot_transfer(inode, transfer_to);
2814 		if (err)
2815 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2816 		dqput(transfer_to[PRJQUOTA]);
2817 	}
2818 	return err;
2819 }
2820 
2821 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2822 {
2823 	struct inode *inode = file_inode(filp);
2824 	struct f2fs_inode_info *fi = F2FS_I(inode);
2825 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2826 	struct page *ipage;
2827 	kprojid_t kprojid;
2828 	int err;
2829 
2830 	if (!f2fs_sb_has_project_quota(sbi)) {
2831 		if (projid != F2FS_DEF_PROJID)
2832 			return -EOPNOTSUPP;
2833 		else
2834 			return 0;
2835 	}
2836 
2837 	if (!f2fs_has_extra_attr(inode))
2838 		return -EOPNOTSUPP;
2839 
2840 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2841 
2842 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2843 		return 0;
2844 
2845 	err = -EPERM;
2846 	/* Is it quota file? Do not allow user to mess with it */
2847 	if (IS_NOQUOTA(inode))
2848 		return err;
2849 
2850 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
2851 	if (IS_ERR(ipage))
2852 		return PTR_ERR(ipage);
2853 
2854 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2855 								i_projid)) {
2856 		err = -EOVERFLOW;
2857 		f2fs_put_page(ipage, 1);
2858 		return err;
2859 	}
2860 	f2fs_put_page(ipage, 1);
2861 
2862 	err = dquot_initialize(inode);
2863 	if (err)
2864 		return err;
2865 
2866 	f2fs_lock_op(sbi);
2867 	err = f2fs_transfer_project_quota(inode, kprojid);
2868 	if (err)
2869 		goto out_unlock;
2870 
2871 	F2FS_I(inode)->i_projid = kprojid;
2872 	inode->i_ctime = current_time(inode);
2873 	f2fs_mark_inode_dirty_sync(inode, true);
2874 out_unlock:
2875 	f2fs_unlock_op(sbi);
2876 	return err;
2877 }
2878 #else
2879 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2880 {
2881 	return 0;
2882 }
2883 
2884 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2885 {
2886 	if (projid != F2FS_DEF_PROJID)
2887 		return -EOPNOTSUPP;
2888 	return 0;
2889 }
2890 #endif
2891 
2892 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
2893 
2894 /*
2895  * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
2896  * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
2897  * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
2898  */
2899 
2900 static const struct {
2901 	u32 iflag;
2902 	u32 xflag;
2903 } f2fs_xflags_map[] = {
2904 	{ F2FS_SYNC_FL,		FS_XFLAG_SYNC },
2905 	{ F2FS_IMMUTABLE_FL,	FS_XFLAG_IMMUTABLE },
2906 	{ F2FS_APPEND_FL,	FS_XFLAG_APPEND },
2907 	{ F2FS_NODUMP_FL,	FS_XFLAG_NODUMP },
2908 	{ F2FS_NOATIME_FL,	FS_XFLAG_NOATIME },
2909 	{ F2FS_PROJINHERIT_FL,	FS_XFLAG_PROJINHERIT },
2910 };
2911 
2912 #define F2FS_SUPPORTED_XFLAGS (		\
2913 		FS_XFLAG_SYNC |		\
2914 		FS_XFLAG_IMMUTABLE |	\
2915 		FS_XFLAG_APPEND |	\
2916 		FS_XFLAG_NODUMP |	\
2917 		FS_XFLAG_NOATIME |	\
2918 		FS_XFLAG_PROJINHERIT)
2919 
2920 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
2921 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
2922 {
2923 	u32 xflags = 0;
2924 	int i;
2925 
2926 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2927 		if (iflags & f2fs_xflags_map[i].iflag)
2928 			xflags |= f2fs_xflags_map[i].xflag;
2929 
2930 	return xflags;
2931 }
2932 
2933 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
2934 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
2935 {
2936 	u32 iflags = 0;
2937 	int i;
2938 
2939 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2940 		if (xflags & f2fs_xflags_map[i].xflag)
2941 			iflags |= f2fs_xflags_map[i].iflag;
2942 
2943 	return iflags;
2944 }
2945 
2946 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
2947 {
2948 	struct f2fs_inode_info *fi = F2FS_I(inode);
2949 
2950 	simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
2951 
2952 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2953 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2954 }
2955 
2956 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2957 {
2958 	struct inode *inode = file_inode(filp);
2959 	struct fsxattr fa;
2960 
2961 	f2fs_fill_fsxattr(inode, &fa);
2962 
2963 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2964 		return -EFAULT;
2965 	return 0;
2966 }
2967 
2968 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2969 {
2970 	struct inode *inode = file_inode(filp);
2971 	struct fsxattr fa, old_fa;
2972 	u32 iflags;
2973 	int err;
2974 
2975 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2976 		return -EFAULT;
2977 
2978 	/* Make sure caller has proper permission */
2979 	if (!inode_owner_or_capable(inode))
2980 		return -EACCES;
2981 
2982 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
2983 		return -EOPNOTSUPP;
2984 
2985 	iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2986 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2987 		return -EOPNOTSUPP;
2988 
2989 	err = mnt_want_write_file(filp);
2990 	if (err)
2991 		return err;
2992 
2993 	inode_lock(inode);
2994 
2995 	f2fs_fill_fsxattr(inode, &old_fa);
2996 	err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
2997 	if (err)
2998 		goto out;
2999 
3000 	err = f2fs_setflags_common(inode, iflags,
3001 			f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
3002 	if (err)
3003 		goto out;
3004 
3005 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3006 out:
3007 	inode_unlock(inode);
3008 	mnt_drop_write_file(filp);
3009 	return err;
3010 }
3011 
3012 int f2fs_pin_file_control(struct inode *inode, bool inc)
3013 {
3014 	struct f2fs_inode_info *fi = F2FS_I(inode);
3015 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3016 
3017 	/* Use i_gc_failures for normal file as a risk signal. */
3018 	if (inc)
3019 		f2fs_i_gc_failures_write(inode,
3020 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3021 
3022 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3023 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3024 			  __func__, inode->i_ino,
3025 			  fi->i_gc_failures[GC_FAILURE_PIN]);
3026 		clear_inode_flag(inode, FI_PIN_FILE);
3027 		return -EAGAIN;
3028 	}
3029 	return 0;
3030 }
3031 
3032 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3033 {
3034 	struct inode *inode = file_inode(filp);
3035 	__u32 pin;
3036 	int ret = 0;
3037 
3038 	if (get_user(pin, (__u32 __user *)arg))
3039 		return -EFAULT;
3040 
3041 	if (!S_ISREG(inode->i_mode))
3042 		return -EINVAL;
3043 
3044 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3045 		return -EROFS;
3046 
3047 	ret = mnt_want_write_file(filp);
3048 	if (ret)
3049 		return ret;
3050 
3051 	inode_lock(inode);
3052 
3053 	if (f2fs_should_update_outplace(inode, NULL)) {
3054 		ret = -EINVAL;
3055 		goto out;
3056 	}
3057 
3058 	if (!pin) {
3059 		clear_inode_flag(inode, FI_PIN_FILE);
3060 		f2fs_i_gc_failures_write(inode, 0);
3061 		goto done;
3062 	}
3063 
3064 	if (f2fs_pin_file_control(inode, false)) {
3065 		ret = -EAGAIN;
3066 		goto out;
3067 	}
3068 	ret = f2fs_convert_inline_inode(inode);
3069 	if (ret)
3070 		goto out;
3071 
3072 	set_inode_flag(inode, FI_PIN_FILE);
3073 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3074 done:
3075 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3076 out:
3077 	inode_unlock(inode);
3078 	mnt_drop_write_file(filp);
3079 	return ret;
3080 }
3081 
3082 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3083 {
3084 	struct inode *inode = file_inode(filp);
3085 	__u32 pin = 0;
3086 
3087 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3088 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3089 	return put_user(pin, (u32 __user *)arg);
3090 }
3091 
3092 int f2fs_precache_extents(struct inode *inode)
3093 {
3094 	struct f2fs_inode_info *fi = F2FS_I(inode);
3095 	struct f2fs_map_blocks map;
3096 	pgoff_t m_next_extent;
3097 	loff_t end;
3098 	int err;
3099 
3100 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3101 		return -EOPNOTSUPP;
3102 
3103 	map.m_lblk = 0;
3104 	map.m_next_pgofs = NULL;
3105 	map.m_next_extent = &m_next_extent;
3106 	map.m_seg_type = NO_CHECK_TYPE;
3107 	map.m_may_create = false;
3108 	end = F2FS_I_SB(inode)->max_file_blocks;
3109 
3110 	while (map.m_lblk < end) {
3111 		map.m_len = end - map.m_lblk;
3112 
3113 		down_write(&fi->i_gc_rwsem[WRITE]);
3114 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3115 		up_write(&fi->i_gc_rwsem[WRITE]);
3116 		if (err)
3117 			return err;
3118 
3119 		map.m_lblk = m_next_extent;
3120 	}
3121 
3122 	return err;
3123 }
3124 
3125 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3126 {
3127 	return f2fs_precache_extents(file_inode(filp));
3128 }
3129 
3130 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3131 {
3132 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3133 	__u64 block_count;
3134 	int ret;
3135 
3136 	if (!capable(CAP_SYS_ADMIN))
3137 		return -EPERM;
3138 
3139 	if (f2fs_readonly(sbi->sb))
3140 		return -EROFS;
3141 
3142 	if (copy_from_user(&block_count, (void __user *)arg,
3143 			   sizeof(block_count)))
3144 		return -EFAULT;
3145 
3146 	ret = f2fs_resize_fs(sbi, block_count);
3147 
3148 	return ret;
3149 }
3150 
3151 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3152 {
3153 	struct inode *inode = file_inode(filp);
3154 
3155 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3156 
3157 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3158 		f2fs_warn(F2FS_I_SB(inode),
3159 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3160 			  inode->i_ino);
3161 		return -EOPNOTSUPP;
3162 	}
3163 
3164 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3165 }
3166 
3167 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3168 {
3169 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3170 		return -EOPNOTSUPP;
3171 
3172 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3173 }
3174 
3175 static int f2fs_get_volume_name(struct file *filp, unsigned long arg)
3176 {
3177 	struct inode *inode = file_inode(filp);
3178 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3179 	char *vbuf;
3180 	int count;
3181 	int err = 0;
3182 
3183 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3184 	if (!vbuf)
3185 		return -ENOMEM;
3186 
3187 	down_read(&sbi->sb_lock);
3188 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3189 			ARRAY_SIZE(sbi->raw_super->volume_name),
3190 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3191 	up_read(&sbi->sb_lock);
3192 
3193 	if (copy_to_user((char __user *)arg, vbuf,
3194 				min(FSLABEL_MAX, count)))
3195 		err = -EFAULT;
3196 
3197 	kvfree(vbuf);
3198 	return err;
3199 }
3200 
3201 static int f2fs_set_volume_name(struct file *filp, unsigned long arg)
3202 {
3203 	struct inode *inode = file_inode(filp);
3204 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3205 	char *vbuf;
3206 	int err = 0;
3207 
3208 	if (!capable(CAP_SYS_ADMIN))
3209 		return -EPERM;
3210 
3211 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3212 	if (IS_ERR(vbuf))
3213 		return PTR_ERR(vbuf);
3214 
3215 	err = mnt_want_write_file(filp);
3216 	if (err)
3217 		goto out;
3218 
3219 	down_write(&sbi->sb_lock);
3220 
3221 	memset(sbi->raw_super->volume_name, 0,
3222 			sizeof(sbi->raw_super->volume_name));
3223 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3224 			sbi->raw_super->volume_name,
3225 			ARRAY_SIZE(sbi->raw_super->volume_name));
3226 
3227 	err = f2fs_commit_super(sbi, false);
3228 
3229 	up_write(&sbi->sb_lock);
3230 
3231 	mnt_drop_write_file(filp);
3232 out:
3233 	kfree(vbuf);
3234 	return err;
3235 }
3236 
3237 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3238 {
3239 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3240 		return -EIO;
3241 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
3242 		return -ENOSPC;
3243 
3244 	switch (cmd) {
3245 	case F2FS_IOC_GETFLAGS:
3246 		return f2fs_ioc_getflags(filp, arg);
3247 	case F2FS_IOC_SETFLAGS:
3248 		return f2fs_ioc_setflags(filp, arg);
3249 	case F2FS_IOC_GETVERSION:
3250 		return f2fs_ioc_getversion(filp, arg);
3251 	case F2FS_IOC_START_ATOMIC_WRITE:
3252 		return f2fs_ioc_start_atomic_write(filp);
3253 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3254 		return f2fs_ioc_commit_atomic_write(filp);
3255 	case F2FS_IOC_START_VOLATILE_WRITE:
3256 		return f2fs_ioc_start_volatile_write(filp);
3257 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3258 		return f2fs_ioc_release_volatile_write(filp);
3259 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3260 		return f2fs_ioc_abort_volatile_write(filp);
3261 	case F2FS_IOC_SHUTDOWN:
3262 		return f2fs_ioc_shutdown(filp, arg);
3263 	case FITRIM:
3264 		return f2fs_ioc_fitrim(filp, arg);
3265 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3266 		return f2fs_ioc_set_encryption_policy(filp, arg);
3267 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3268 		return f2fs_ioc_get_encryption_policy(filp, arg);
3269 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3270 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3271 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3272 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3273 	case FS_IOC_ADD_ENCRYPTION_KEY:
3274 		return f2fs_ioc_add_encryption_key(filp, arg);
3275 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3276 		return f2fs_ioc_remove_encryption_key(filp, arg);
3277 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3278 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3279 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3280 		return f2fs_ioc_get_encryption_key_status(filp, arg);
3281 	case F2FS_IOC_GARBAGE_COLLECT:
3282 		return f2fs_ioc_gc(filp, arg);
3283 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3284 		return f2fs_ioc_gc_range(filp, arg);
3285 	case F2FS_IOC_WRITE_CHECKPOINT:
3286 		return f2fs_ioc_write_checkpoint(filp, arg);
3287 	case F2FS_IOC_DEFRAGMENT:
3288 		return f2fs_ioc_defragment(filp, arg);
3289 	case F2FS_IOC_MOVE_RANGE:
3290 		return f2fs_ioc_move_range(filp, arg);
3291 	case F2FS_IOC_FLUSH_DEVICE:
3292 		return f2fs_ioc_flush_device(filp, arg);
3293 	case F2FS_IOC_GET_FEATURES:
3294 		return f2fs_ioc_get_features(filp, arg);
3295 	case F2FS_IOC_FSGETXATTR:
3296 		return f2fs_ioc_fsgetxattr(filp, arg);
3297 	case F2FS_IOC_FSSETXATTR:
3298 		return f2fs_ioc_fssetxattr(filp, arg);
3299 	case F2FS_IOC_GET_PIN_FILE:
3300 		return f2fs_ioc_get_pin_file(filp, arg);
3301 	case F2FS_IOC_SET_PIN_FILE:
3302 		return f2fs_ioc_set_pin_file(filp, arg);
3303 	case F2FS_IOC_PRECACHE_EXTENTS:
3304 		return f2fs_ioc_precache_extents(filp, arg);
3305 	case F2FS_IOC_RESIZE_FS:
3306 		return f2fs_ioc_resize_fs(filp, arg);
3307 	case FS_IOC_ENABLE_VERITY:
3308 		return f2fs_ioc_enable_verity(filp, arg);
3309 	case FS_IOC_MEASURE_VERITY:
3310 		return f2fs_ioc_measure_verity(filp, arg);
3311 	case F2FS_IOC_GET_VOLUME_NAME:
3312 		return f2fs_get_volume_name(filp, arg);
3313 	case F2FS_IOC_SET_VOLUME_NAME:
3314 		return f2fs_set_volume_name(filp, arg);
3315 	default:
3316 		return -ENOTTY;
3317 	}
3318 }
3319 
3320 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3321 {
3322 	struct file *file = iocb->ki_filp;
3323 	struct inode *inode = file_inode(file);
3324 	ssize_t ret;
3325 
3326 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3327 		ret = -EIO;
3328 		goto out;
3329 	}
3330 
3331 	if (iocb->ki_flags & IOCB_NOWAIT) {
3332 		if (!inode_trylock(inode)) {
3333 			ret = -EAGAIN;
3334 			goto out;
3335 		}
3336 	} else {
3337 		inode_lock(inode);
3338 	}
3339 
3340 	ret = generic_write_checks(iocb, from);
3341 	if (ret > 0) {
3342 		bool preallocated = false;
3343 		size_t target_size = 0;
3344 		int err;
3345 
3346 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3347 			set_inode_flag(inode, FI_NO_PREALLOC);
3348 
3349 		if ((iocb->ki_flags & IOCB_NOWAIT)) {
3350 			if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3351 						iov_iter_count(from)) ||
3352 				f2fs_has_inline_data(inode) ||
3353 				f2fs_force_buffered_io(inode, iocb, from)) {
3354 				clear_inode_flag(inode, FI_NO_PREALLOC);
3355 				inode_unlock(inode);
3356 				ret = -EAGAIN;
3357 				goto out;
3358 			}
3359 		} else {
3360 			preallocated = true;
3361 			target_size = iocb->ki_pos + iov_iter_count(from);
3362 
3363 			err = f2fs_preallocate_blocks(iocb, from);
3364 			if (err) {
3365 				clear_inode_flag(inode, FI_NO_PREALLOC);
3366 				inode_unlock(inode);
3367 				ret = err;
3368 				goto out;
3369 			}
3370 		}
3371 		ret = __generic_file_write_iter(iocb, from);
3372 		clear_inode_flag(inode, FI_NO_PREALLOC);
3373 
3374 		/* if we couldn't write data, we should deallocate blocks. */
3375 		if (preallocated && i_size_read(inode) < target_size)
3376 			f2fs_truncate(inode);
3377 
3378 		if (ret > 0)
3379 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3380 	}
3381 	inode_unlock(inode);
3382 out:
3383 	trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3384 					iov_iter_count(from), ret);
3385 	if (ret > 0)
3386 		ret = generic_write_sync(iocb, ret);
3387 	return ret;
3388 }
3389 
3390 #ifdef CONFIG_COMPAT
3391 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3392 {
3393 	switch (cmd) {
3394 	case F2FS_IOC32_GETFLAGS:
3395 		cmd = F2FS_IOC_GETFLAGS;
3396 		break;
3397 	case F2FS_IOC32_SETFLAGS:
3398 		cmd = F2FS_IOC_SETFLAGS;
3399 		break;
3400 	case F2FS_IOC32_GETVERSION:
3401 		cmd = F2FS_IOC_GETVERSION;
3402 		break;
3403 	case F2FS_IOC_START_ATOMIC_WRITE:
3404 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3405 	case F2FS_IOC_START_VOLATILE_WRITE:
3406 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3407 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3408 	case F2FS_IOC_SHUTDOWN:
3409 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3410 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3411 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3412 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3413 	case FS_IOC_ADD_ENCRYPTION_KEY:
3414 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3415 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3416 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3417 	case F2FS_IOC_GARBAGE_COLLECT:
3418 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3419 	case F2FS_IOC_WRITE_CHECKPOINT:
3420 	case F2FS_IOC_DEFRAGMENT:
3421 	case F2FS_IOC_MOVE_RANGE:
3422 	case F2FS_IOC_FLUSH_DEVICE:
3423 	case F2FS_IOC_GET_FEATURES:
3424 	case F2FS_IOC_FSGETXATTR:
3425 	case F2FS_IOC_FSSETXATTR:
3426 	case F2FS_IOC_GET_PIN_FILE:
3427 	case F2FS_IOC_SET_PIN_FILE:
3428 	case F2FS_IOC_PRECACHE_EXTENTS:
3429 	case F2FS_IOC_RESIZE_FS:
3430 	case FS_IOC_ENABLE_VERITY:
3431 	case FS_IOC_MEASURE_VERITY:
3432 	case F2FS_IOC_GET_VOLUME_NAME:
3433 	case F2FS_IOC_SET_VOLUME_NAME:
3434 		break;
3435 	default:
3436 		return -ENOIOCTLCMD;
3437 	}
3438 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3439 }
3440 #endif
3441 
3442 const struct file_operations f2fs_file_operations = {
3443 	.llseek		= f2fs_llseek,
3444 	.read_iter	= generic_file_read_iter,
3445 	.write_iter	= f2fs_file_write_iter,
3446 	.open		= f2fs_file_open,
3447 	.release	= f2fs_release_file,
3448 	.mmap		= f2fs_file_mmap,
3449 	.flush		= f2fs_file_flush,
3450 	.fsync		= f2fs_sync_file,
3451 	.fallocate	= f2fs_fallocate,
3452 	.unlocked_ioctl	= f2fs_ioctl,
3453 #ifdef CONFIG_COMPAT
3454 	.compat_ioctl	= f2fs_compat_ioctl,
3455 #endif
3456 	.splice_read	= generic_file_splice_read,
3457 	.splice_write	= iter_file_splice_write,
3458 };
3459