xref: /openbmc/linux/fs/f2fs/file.c (revision 78beef62)
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 
24 #include "f2fs.h"
25 #include "node.h"
26 #include "segment.h"
27 #include "xattr.h"
28 #include "acl.h"
29 #include "gc.h"
30 #include "trace.h"
31 #include <trace/events/f2fs.h>
32 
33 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
34 {
35 	struct inode *inode = file_inode(vmf->vma->vm_file);
36 	vm_fault_t ret;
37 
38 	down_read(&F2FS_I(inode)->i_mmap_sem);
39 	ret = filemap_fault(vmf);
40 	up_read(&F2FS_I(inode)->i_mmap_sem);
41 
42 	trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
43 
44 	return ret;
45 }
46 
47 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
48 {
49 	struct page *page = vmf->page;
50 	struct inode *inode = file_inode(vmf->vma->vm_file);
51 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
52 	struct dnode_of_data dn = { .node_changed = false };
53 	int err;
54 
55 	if (unlikely(f2fs_cp_error(sbi))) {
56 		err = -EIO;
57 		goto err;
58 	}
59 
60 	sb_start_pagefault(inode->i_sb);
61 
62 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
63 
64 	file_update_time(vmf->vma->vm_file);
65 	down_read(&F2FS_I(inode)->i_mmap_sem);
66 	lock_page(page);
67 	if (unlikely(page->mapping != inode->i_mapping ||
68 			page_offset(page) > i_size_read(inode) ||
69 			!PageUptodate(page))) {
70 		unlock_page(page);
71 		err = -EFAULT;
72 		goto out_sem;
73 	}
74 
75 	/* block allocation */
76 	__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
77 	set_new_dnode(&dn, inode, NULL, NULL, 0);
78 	err = f2fs_get_block(&dn, page->index);
79 	f2fs_put_dnode(&dn);
80 	__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
81 	if (err) {
82 		unlock_page(page);
83 		goto out_sem;
84 	}
85 
86 	/* fill the page */
87 	f2fs_wait_on_page_writeback(page, DATA, false, true);
88 
89 	/* wait for GCed page writeback via META_MAPPING */
90 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
91 
92 	/*
93 	 * check to see if the page is mapped already (no holes)
94 	 */
95 	if (PageMappedToDisk(page))
96 		goto out_sem;
97 
98 	/* page is wholly or partially inside EOF */
99 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
100 						i_size_read(inode)) {
101 		loff_t offset;
102 
103 		offset = i_size_read(inode) & ~PAGE_MASK;
104 		zero_user_segment(page, offset, PAGE_SIZE);
105 	}
106 	set_page_dirty(page);
107 	if (!PageUptodate(page))
108 		SetPageUptodate(page);
109 
110 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
111 	f2fs_update_time(sbi, REQ_TIME);
112 
113 	trace_f2fs_vm_page_mkwrite(page, DATA);
114 out_sem:
115 	up_read(&F2FS_I(inode)->i_mmap_sem);
116 
117 	f2fs_balance_fs(sbi, dn.node_changed);
118 
119 	sb_end_pagefault(inode->i_sb);
120 err:
121 	return block_page_mkwrite_return(err);
122 }
123 
124 static const struct vm_operations_struct f2fs_file_vm_ops = {
125 	.fault		= f2fs_filemap_fault,
126 	.map_pages	= filemap_map_pages,
127 	.page_mkwrite	= f2fs_vm_page_mkwrite,
128 };
129 
130 static int get_parent_ino(struct inode *inode, nid_t *pino)
131 {
132 	struct dentry *dentry;
133 
134 	inode = igrab(inode);
135 	dentry = d_find_any_alias(inode);
136 	iput(inode);
137 	if (!dentry)
138 		return 0;
139 
140 	*pino = parent_ino(dentry);
141 	dput(dentry);
142 	return 1;
143 }
144 
145 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
146 {
147 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
148 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
149 
150 	if (!S_ISREG(inode->i_mode))
151 		cp_reason = CP_NON_REGULAR;
152 	else if (inode->i_nlink != 1)
153 		cp_reason = CP_HARDLINK;
154 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
155 		cp_reason = CP_SB_NEED_CP;
156 	else if (file_wrong_pino(inode))
157 		cp_reason = CP_WRONG_PINO;
158 	else if (!f2fs_space_for_roll_forward(sbi))
159 		cp_reason = CP_NO_SPC_ROLL;
160 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
161 		cp_reason = CP_NODE_NEED_CP;
162 	else if (test_opt(sbi, FASTBOOT))
163 		cp_reason = CP_FASTBOOT_MODE;
164 	else if (F2FS_OPTION(sbi).active_logs == 2)
165 		cp_reason = CP_SPEC_LOG_NUM;
166 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
167 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
168 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
169 							TRANS_DIR_INO))
170 		cp_reason = CP_RECOVER_DIR;
171 
172 	return cp_reason;
173 }
174 
175 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
176 {
177 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
178 	bool ret = false;
179 	/* But we need to avoid that there are some inode updates */
180 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
181 		ret = true;
182 	f2fs_put_page(i, 0);
183 	return ret;
184 }
185 
186 static void try_to_fix_pino(struct inode *inode)
187 {
188 	struct f2fs_inode_info *fi = F2FS_I(inode);
189 	nid_t pino;
190 
191 	down_write(&fi->i_sem);
192 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
193 			get_parent_ino(inode, &pino)) {
194 		f2fs_i_pino_write(inode, pino);
195 		file_got_pino(inode);
196 	}
197 	up_write(&fi->i_sem);
198 }
199 
200 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
201 						int datasync, bool atomic)
202 {
203 	struct inode *inode = file->f_mapping->host;
204 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
205 	nid_t ino = inode->i_ino;
206 	int ret = 0;
207 	enum cp_reason_type cp_reason = 0;
208 	struct writeback_control wbc = {
209 		.sync_mode = WB_SYNC_ALL,
210 		.nr_to_write = LONG_MAX,
211 		.for_reclaim = 0,
212 	};
213 	unsigned int seq_id = 0;
214 
215 	if (unlikely(f2fs_readonly(inode->i_sb) ||
216 				is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
217 		return 0;
218 
219 	trace_f2fs_sync_file_enter(inode);
220 
221 	if (S_ISDIR(inode->i_mode))
222 		goto go_write;
223 
224 	/* if fdatasync is triggered, let's do in-place-update */
225 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
226 		set_inode_flag(inode, FI_NEED_IPU);
227 	ret = file_write_and_wait_range(file, start, end);
228 	clear_inode_flag(inode, FI_NEED_IPU);
229 
230 	if (ret) {
231 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
232 		return ret;
233 	}
234 
235 	/* if the inode is dirty, let's recover all the time */
236 	if (!f2fs_skip_inode_update(inode, datasync)) {
237 		f2fs_write_inode(inode, NULL);
238 		goto go_write;
239 	}
240 
241 	/*
242 	 * if there is no written data, don't waste time to write recovery info.
243 	 */
244 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
245 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
246 
247 		/* it may call write_inode just prior to fsync */
248 		if (need_inode_page_update(sbi, ino))
249 			goto go_write;
250 
251 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
252 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
253 			goto flush_out;
254 		goto out;
255 	}
256 go_write:
257 	/*
258 	 * Both of fdatasync() and fsync() are able to be recovered from
259 	 * sudden-power-off.
260 	 */
261 	down_read(&F2FS_I(inode)->i_sem);
262 	cp_reason = need_do_checkpoint(inode);
263 	up_read(&F2FS_I(inode)->i_sem);
264 
265 	if (cp_reason) {
266 		/* all the dirty node pages should be flushed for POR */
267 		ret = f2fs_sync_fs(inode->i_sb, 1);
268 
269 		/*
270 		 * We've secured consistency through sync_fs. Following pino
271 		 * will be used only for fsynced inodes after checkpoint.
272 		 */
273 		try_to_fix_pino(inode);
274 		clear_inode_flag(inode, FI_APPEND_WRITE);
275 		clear_inode_flag(inode, FI_UPDATE_WRITE);
276 		goto out;
277 	}
278 sync_nodes:
279 	atomic_inc(&sbi->wb_sync_req[NODE]);
280 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
281 	atomic_dec(&sbi->wb_sync_req[NODE]);
282 	if (ret)
283 		goto out;
284 
285 	/* if cp_error was enabled, we should avoid infinite loop */
286 	if (unlikely(f2fs_cp_error(sbi))) {
287 		ret = -EIO;
288 		goto out;
289 	}
290 
291 	if (f2fs_need_inode_block_update(sbi, ino)) {
292 		f2fs_mark_inode_dirty_sync(inode, true);
293 		f2fs_write_inode(inode, NULL);
294 		goto sync_nodes;
295 	}
296 
297 	/*
298 	 * If it's atomic_write, it's just fine to keep write ordering. So
299 	 * here we don't need to wait for node write completion, since we use
300 	 * node chain which serializes node blocks. If one of node writes are
301 	 * reordered, we can see simply broken chain, resulting in stopping
302 	 * roll-forward recovery. It means we'll recover all or none node blocks
303 	 * given fsync mark.
304 	 */
305 	if (!atomic) {
306 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
307 		if (ret)
308 			goto out;
309 	}
310 
311 	/* once recovery info is written, don't need to tack this */
312 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
313 	clear_inode_flag(inode, FI_APPEND_WRITE);
314 flush_out:
315 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
316 		ret = f2fs_issue_flush(sbi, inode->i_ino);
317 	if (!ret) {
318 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
319 		clear_inode_flag(inode, FI_UPDATE_WRITE);
320 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
321 	}
322 	f2fs_update_time(sbi, REQ_TIME);
323 out:
324 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
325 	f2fs_trace_ios(NULL, 1);
326 	return ret;
327 }
328 
329 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
330 {
331 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
332 		return -EIO;
333 	return f2fs_do_sync_file(file, start, end, datasync, false);
334 }
335 
336 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
337 						pgoff_t pgofs, int whence)
338 {
339 	struct page *page;
340 	int nr_pages;
341 
342 	if (whence != SEEK_DATA)
343 		return 0;
344 
345 	/* find first dirty page index */
346 	nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
347 				      1, &page);
348 	if (!nr_pages)
349 		return ULONG_MAX;
350 	pgofs = page->index;
351 	put_page(page);
352 	return pgofs;
353 }
354 
355 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
356 				pgoff_t dirty, pgoff_t pgofs, int whence)
357 {
358 	switch (whence) {
359 	case SEEK_DATA:
360 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
361 			__is_valid_data_blkaddr(blkaddr))
362 			return true;
363 		break;
364 	case SEEK_HOLE:
365 		if (blkaddr == NULL_ADDR)
366 			return true;
367 		break;
368 	}
369 	return false;
370 }
371 
372 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
373 {
374 	struct inode *inode = file->f_mapping->host;
375 	loff_t maxbytes = inode->i_sb->s_maxbytes;
376 	struct dnode_of_data dn;
377 	pgoff_t pgofs, end_offset, dirty;
378 	loff_t data_ofs = offset;
379 	loff_t isize;
380 	int err = 0;
381 
382 	inode_lock(inode);
383 
384 	isize = i_size_read(inode);
385 	if (offset >= isize)
386 		goto fail;
387 
388 	/* handle inline data case */
389 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
390 		if (whence == SEEK_HOLE)
391 			data_ofs = isize;
392 		goto found;
393 	}
394 
395 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
396 
397 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
398 
399 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
400 		set_new_dnode(&dn, inode, NULL, NULL, 0);
401 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
402 		if (err && err != -ENOENT) {
403 			goto fail;
404 		} else if (err == -ENOENT) {
405 			/* direct node does not exists */
406 			if (whence == SEEK_DATA) {
407 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
408 				continue;
409 			} else {
410 				goto found;
411 			}
412 		}
413 
414 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
415 
416 		/* find data/hole in dnode block */
417 		for (; dn.ofs_in_node < end_offset;
418 				dn.ofs_in_node++, pgofs++,
419 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
420 			block_t blkaddr;
421 
422 			blkaddr = datablock_addr(dn.inode,
423 					dn.node_page, dn.ofs_in_node);
424 
425 			if (__is_valid_data_blkaddr(blkaddr) &&
426 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
427 					blkaddr, DATA_GENERIC_ENHANCE)) {
428 				f2fs_put_dnode(&dn);
429 				goto fail;
430 			}
431 
432 			if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
433 							pgofs, whence)) {
434 				f2fs_put_dnode(&dn);
435 				goto found;
436 			}
437 		}
438 		f2fs_put_dnode(&dn);
439 	}
440 
441 	if (whence == SEEK_DATA)
442 		goto fail;
443 found:
444 	if (whence == SEEK_HOLE && data_ofs > isize)
445 		data_ofs = isize;
446 	inode_unlock(inode);
447 	return vfs_setpos(file, data_ofs, maxbytes);
448 fail:
449 	inode_unlock(inode);
450 	return -ENXIO;
451 }
452 
453 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
454 {
455 	struct inode *inode = file->f_mapping->host;
456 	loff_t maxbytes = inode->i_sb->s_maxbytes;
457 
458 	switch (whence) {
459 	case SEEK_SET:
460 	case SEEK_CUR:
461 	case SEEK_END:
462 		return generic_file_llseek_size(file, offset, whence,
463 						maxbytes, i_size_read(inode));
464 	case SEEK_DATA:
465 	case SEEK_HOLE:
466 		if (offset < 0)
467 			return -ENXIO;
468 		return f2fs_seek_block(file, offset, whence);
469 	}
470 
471 	return -EINVAL;
472 }
473 
474 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
475 {
476 	struct inode *inode = file_inode(file);
477 	int err;
478 
479 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
480 		return -EIO;
481 
482 	/* we don't need to use inline_data strictly */
483 	err = f2fs_convert_inline_inode(inode);
484 	if (err)
485 		return err;
486 
487 	file_accessed(file);
488 	vma->vm_ops = &f2fs_file_vm_ops;
489 	return 0;
490 }
491 
492 static int f2fs_file_open(struct inode *inode, struct file *filp)
493 {
494 	int err = fscrypt_file_open(inode, filp);
495 
496 	if (err)
497 		return err;
498 
499 	err = fsverity_file_open(inode, filp);
500 	if (err)
501 		return err;
502 
503 	filp->f_mode |= FMODE_NOWAIT;
504 
505 	return dquot_file_open(inode, filp);
506 }
507 
508 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
509 {
510 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
511 	struct f2fs_node *raw_node;
512 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
513 	__le32 *addr;
514 	int base = 0;
515 
516 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
517 		base = get_extra_isize(dn->inode);
518 
519 	raw_node = F2FS_NODE(dn->node_page);
520 	addr = blkaddr_in_node(raw_node) + base + ofs;
521 
522 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
523 		block_t blkaddr = le32_to_cpu(*addr);
524 
525 		if (blkaddr == NULL_ADDR)
526 			continue;
527 
528 		dn->data_blkaddr = NULL_ADDR;
529 		f2fs_set_data_blkaddr(dn);
530 
531 		if (__is_valid_data_blkaddr(blkaddr) &&
532 			!f2fs_is_valid_blkaddr(sbi, blkaddr,
533 					DATA_GENERIC_ENHANCE))
534 			continue;
535 
536 		f2fs_invalidate_blocks(sbi, blkaddr);
537 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
538 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
539 		nr_free++;
540 	}
541 
542 	if (nr_free) {
543 		pgoff_t fofs;
544 		/*
545 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
546 		 * we will invalidate all blkaddr in the whole range.
547 		 */
548 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
549 							dn->inode) + ofs;
550 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
551 		dec_valid_block_count(sbi, dn->inode, nr_free);
552 	}
553 	dn->ofs_in_node = ofs;
554 
555 	f2fs_update_time(sbi, REQ_TIME);
556 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
557 					 dn->ofs_in_node, nr_free);
558 }
559 
560 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
561 {
562 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
563 }
564 
565 static int truncate_partial_data_page(struct inode *inode, u64 from,
566 								bool cache_only)
567 {
568 	loff_t offset = from & (PAGE_SIZE - 1);
569 	pgoff_t index = from >> PAGE_SHIFT;
570 	struct address_space *mapping = inode->i_mapping;
571 	struct page *page;
572 
573 	if (!offset && !cache_only)
574 		return 0;
575 
576 	if (cache_only) {
577 		page = find_lock_page(mapping, index);
578 		if (page && PageUptodate(page))
579 			goto truncate_out;
580 		f2fs_put_page(page, 1);
581 		return 0;
582 	}
583 
584 	page = f2fs_get_lock_data_page(inode, index, true);
585 	if (IS_ERR(page))
586 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
587 truncate_out:
588 	f2fs_wait_on_page_writeback(page, DATA, true, true);
589 	zero_user(page, offset, PAGE_SIZE - offset);
590 
591 	/* An encrypted inode should have a key and truncate the last page. */
592 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
593 	if (!cache_only)
594 		set_page_dirty(page);
595 	f2fs_put_page(page, 1);
596 	return 0;
597 }
598 
599 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
600 {
601 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
602 	struct dnode_of_data dn;
603 	pgoff_t free_from;
604 	int count = 0, err = 0;
605 	struct page *ipage;
606 	bool truncate_page = false;
607 
608 	trace_f2fs_truncate_blocks_enter(inode, from);
609 
610 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
611 
612 	if (free_from >= sbi->max_file_blocks)
613 		goto free_partial;
614 
615 	if (lock)
616 		f2fs_lock_op(sbi);
617 
618 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
619 	if (IS_ERR(ipage)) {
620 		err = PTR_ERR(ipage);
621 		goto out;
622 	}
623 
624 	if (f2fs_has_inline_data(inode)) {
625 		f2fs_truncate_inline_inode(inode, ipage, from);
626 		f2fs_put_page(ipage, 1);
627 		truncate_page = true;
628 		goto out;
629 	}
630 
631 	set_new_dnode(&dn, inode, ipage, NULL, 0);
632 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
633 	if (err) {
634 		if (err == -ENOENT)
635 			goto free_next;
636 		goto out;
637 	}
638 
639 	count = ADDRS_PER_PAGE(dn.node_page, inode);
640 
641 	count -= dn.ofs_in_node;
642 	f2fs_bug_on(sbi, count < 0);
643 
644 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
645 		f2fs_truncate_data_blocks_range(&dn, count);
646 		free_from += count;
647 	}
648 
649 	f2fs_put_dnode(&dn);
650 free_next:
651 	err = f2fs_truncate_inode_blocks(inode, free_from);
652 out:
653 	if (lock)
654 		f2fs_unlock_op(sbi);
655 free_partial:
656 	/* lastly zero out the first data page */
657 	if (!err)
658 		err = truncate_partial_data_page(inode, from, truncate_page);
659 
660 	trace_f2fs_truncate_blocks_exit(inode, err);
661 	return err;
662 }
663 
664 int f2fs_truncate(struct inode *inode)
665 {
666 	int err;
667 
668 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
669 		return -EIO;
670 
671 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
672 				S_ISLNK(inode->i_mode)))
673 		return 0;
674 
675 	trace_f2fs_truncate(inode);
676 
677 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
678 		f2fs_show_injection_info(FAULT_TRUNCATE);
679 		return -EIO;
680 	}
681 
682 	/* we should check inline_data size */
683 	if (!f2fs_may_inline_data(inode)) {
684 		err = f2fs_convert_inline_inode(inode);
685 		if (err)
686 			return err;
687 	}
688 
689 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
690 	if (err)
691 		return err;
692 
693 	inode->i_mtime = inode->i_ctime = current_time(inode);
694 	f2fs_mark_inode_dirty_sync(inode, false);
695 	return 0;
696 }
697 
698 int f2fs_getattr(const struct path *path, struct kstat *stat,
699 		 u32 request_mask, unsigned int query_flags)
700 {
701 	struct inode *inode = d_inode(path->dentry);
702 	struct f2fs_inode_info *fi = F2FS_I(inode);
703 	struct f2fs_inode *ri;
704 	unsigned int flags;
705 
706 	if (f2fs_has_extra_attr(inode) &&
707 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
708 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
709 		stat->result_mask |= STATX_BTIME;
710 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
711 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
712 	}
713 
714 	flags = fi->i_flags;
715 	if (flags & F2FS_APPEND_FL)
716 		stat->attributes |= STATX_ATTR_APPEND;
717 	if (IS_ENCRYPTED(inode))
718 		stat->attributes |= STATX_ATTR_ENCRYPTED;
719 	if (flags & F2FS_IMMUTABLE_FL)
720 		stat->attributes |= STATX_ATTR_IMMUTABLE;
721 	if (flags & F2FS_NODUMP_FL)
722 		stat->attributes |= STATX_ATTR_NODUMP;
723 
724 	stat->attributes_mask |= (STATX_ATTR_APPEND |
725 				  STATX_ATTR_ENCRYPTED |
726 				  STATX_ATTR_IMMUTABLE |
727 				  STATX_ATTR_NODUMP);
728 
729 	generic_fillattr(inode, stat);
730 
731 	/* we need to show initial sectors used for inline_data/dentries */
732 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
733 					f2fs_has_inline_dentry(inode))
734 		stat->blocks += (stat->size + 511) >> 9;
735 
736 	return 0;
737 }
738 
739 #ifdef CONFIG_F2FS_FS_POSIX_ACL
740 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
741 {
742 	unsigned int ia_valid = attr->ia_valid;
743 
744 	if (ia_valid & ATTR_UID)
745 		inode->i_uid = attr->ia_uid;
746 	if (ia_valid & ATTR_GID)
747 		inode->i_gid = attr->ia_gid;
748 	if (ia_valid & ATTR_ATIME)
749 		inode->i_atime = timespec64_trunc(attr->ia_atime,
750 						  inode->i_sb->s_time_gran);
751 	if (ia_valid & ATTR_MTIME)
752 		inode->i_mtime = timespec64_trunc(attr->ia_mtime,
753 						  inode->i_sb->s_time_gran);
754 	if (ia_valid & ATTR_CTIME)
755 		inode->i_ctime = timespec64_trunc(attr->ia_ctime,
756 						  inode->i_sb->s_time_gran);
757 	if (ia_valid & ATTR_MODE) {
758 		umode_t mode = attr->ia_mode;
759 
760 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
761 			mode &= ~S_ISGID;
762 		set_acl_inode(inode, mode);
763 	}
764 }
765 #else
766 #define __setattr_copy setattr_copy
767 #endif
768 
769 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
770 {
771 	struct inode *inode = d_inode(dentry);
772 	int err;
773 
774 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
775 		return -EIO;
776 
777 	err = setattr_prepare(dentry, attr);
778 	if (err)
779 		return err;
780 
781 	err = fscrypt_prepare_setattr(dentry, attr);
782 	if (err)
783 		return err;
784 
785 	err = fsverity_prepare_setattr(dentry, attr);
786 	if (err)
787 		return err;
788 
789 	if (is_quota_modification(inode, attr)) {
790 		err = dquot_initialize(inode);
791 		if (err)
792 			return err;
793 	}
794 	if ((attr->ia_valid & ATTR_UID &&
795 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
796 		(attr->ia_valid & ATTR_GID &&
797 		!gid_eq(attr->ia_gid, inode->i_gid))) {
798 		f2fs_lock_op(F2FS_I_SB(inode));
799 		err = dquot_transfer(inode, attr);
800 		if (err) {
801 			set_sbi_flag(F2FS_I_SB(inode),
802 					SBI_QUOTA_NEED_REPAIR);
803 			f2fs_unlock_op(F2FS_I_SB(inode));
804 			return err;
805 		}
806 		/*
807 		 * update uid/gid under lock_op(), so that dquot and inode can
808 		 * be updated atomically.
809 		 */
810 		if (attr->ia_valid & ATTR_UID)
811 			inode->i_uid = attr->ia_uid;
812 		if (attr->ia_valid & ATTR_GID)
813 			inode->i_gid = attr->ia_gid;
814 		f2fs_mark_inode_dirty_sync(inode, true);
815 		f2fs_unlock_op(F2FS_I_SB(inode));
816 	}
817 
818 	if (attr->ia_valid & ATTR_SIZE) {
819 		bool to_smaller = (attr->ia_size <= i_size_read(inode));
820 
821 		down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
822 		down_write(&F2FS_I(inode)->i_mmap_sem);
823 
824 		truncate_setsize(inode, attr->ia_size);
825 
826 		if (to_smaller)
827 			err = f2fs_truncate(inode);
828 		/*
829 		 * do not trim all blocks after i_size if target size is
830 		 * larger than i_size.
831 		 */
832 		up_write(&F2FS_I(inode)->i_mmap_sem);
833 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
834 
835 		if (err)
836 			return err;
837 
838 		if (!to_smaller) {
839 			/* should convert inline inode here */
840 			if (!f2fs_may_inline_data(inode)) {
841 				err = f2fs_convert_inline_inode(inode);
842 				if (err)
843 					return err;
844 			}
845 			inode->i_mtime = inode->i_ctime = current_time(inode);
846 		}
847 
848 		down_write(&F2FS_I(inode)->i_sem);
849 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
850 		up_write(&F2FS_I(inode)->i_sem);
851 	}
852 
853 	__setattr_copy(inode, attr);
854 
855 	if (attr->ia_valid & ATTR_MODE) {
856 		err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
857 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
858 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
859 			clear_inode_flag(inode, FI_ACL_MODE);
860 		}
861 	}
862 
863 	/* file size may changed here */
864 	f2fs_mark_inode_dirty_sync(inode, true);
865 
866 	/* inode change will produce dirty node pages flushed by checkpoint */
867 	f2fs_balance_fs(F2FS_I_SB(inode), true);
868 
869 	return err;
870 }
871 
872 const struct inode_operations f2fs_file_inode_operations = {
873 	.getattr	= f2fs_getattr,
874 	.setattr	= f2fs_setattr,
875 	.get_acl	= f2fs_get_acl,
876 	.set_acl	= f2fs_set_acl,
877 #ifdef CONFIG_F2FS_FS_XATTR
878 	.listxattr	= f2fs_listxattr,
879 #endif
880 	.fiemap		= f2fs_fiemap,
881 };
882 
883 static int fill_zero(struct inode *inode, pgoff_t index,
884 					loff_t start, loff_t len)
885 {
886 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
887 	struct page *page;
888 
889 	if (!len)
890 		return 0;
891 
892 	f2fs_balance_fs(sbi, true);
893 
894 	f2fs_lock_op(sbi);
895 	page = f2fs_get_new_data_page(inode, NULL, index, false);
896 	f2fs_unlock_op(sbi);
897 
898 	if (IS_ERR(page))
899 		return PTR_ERR(page);
900 
901 	f2fs_wait_on_page_writeback(page, DATA, true, true);
902 	zero_user(page, start, len);
903 	set_page_dirty(page);
904 	f2fs_put_page(page, 1);
905 	return 0;
906 }
907 
908 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
909 {
910 	int err;
911 
912 	while (pg_start < pg_end) {
913 		struct dnode_of_data dn;
914 		pgoff_t end_offset, count;
915 
916 		set_new_dnode(&dn, inode, NULL, NULL, 0);
917 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
918 		if (err) {
919 			if (err == -ENOENT) {
920 				pg_start = f2fs_get_next_page_offset(&dn,
921 								pg_start);
922 				continue;
923 			}
924 			return err;
925 		}
926 
927 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
928 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
929 
930 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
931 
932 		f2fs_truncate_data_blocks_range(&dn, count);
933 		f2fs_put_dnode(&dn);
934 
935 		pg_start += count;
936 	}
937 	return 0;
938 }
939 
940 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
941 {
942 	pgoff_t pg_start, pg_end;
943 	loff_t off_start, off_end;
944 	int ret;
945 
946 	ret = f2fs_convert_inline_inode(inode);
947 	if (ret)
948 		return ret;
949 
950 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
951 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
952 
953 	off_start = offset & (PAGE_SIZE - 1);
954 	off_end = (offset + len) & (PAGE_SIZE - 1);
955 
956 	if (pg_start == pg_end) {
957 		ret = fill_zero(inode, pg_start, off_start,
958 						off_end - off_start);
959 		if (ret)
960 			return ret;
961 	} else {
962 		if (off_start) {
963 			ret = fill_zero(inode, pg_start++, off_start,
964 						PAGE_SIZE - off_start);
965 			if (ret)
966 				return ret;
967 		}
968 		if (off_end) {
969 			ret = fill_zero(inode, pg_end, 0, off_end);
970 			if (ret)
971 				return ret;
972 		}
973 
974 		if (pg_start < pg_end) {
975 			struct address_space *mapping = inode->i_mapping;
976 			loff_t blk_start, blk_end;
977 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
978 
979 			f2fs_balance_fs(sbi, true);
980 
981 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
982 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
983 
984 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
985 			down_write(&F2FS_I(inode)->i_mmap_sem);
986 
987 			truncate_inode_pages_range(mapping, blk_start,
988 					blk_end - 1);
989 
990 			f2fs_lock_op(sbi);
991 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
992 			f2fs_unlock_op(sbi);
993 
994 			up_write(&F2FS_I(inode)->i_mmap_sem);
995 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
996 		}
997 	}
998 
999 	return ret;
1000 }
1001 
1002 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1003 				int *do_replace, pgoff_t off, pgoff_t len)
1004 {
1005 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1006 	struct dnode_of_data dn;
1007 	int ret, done, i;
1008 
1009 next_dnode:
1010 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1011 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1012 	if (ret && ret != -ENOENT) {
1013 		return ret;
1014 	} else if (ret == -ENOENT) {
1015 		if (dn.max_level == 0)
1016 			return -ENOENT;
1017 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1018 									len);
1019 		blkaddr += done;
1020 		do_replace += done;
1021 		goto next;
1022 	}
1023 
1024 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1025 							dn.ofs_in_node, len);
1026 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1027 		*blkaddr = datablock_addr(dn.inode,
1028 					dn.node_page, dn.ofs_in_node);
1029 
1030 		if (__is_valid_data_blkaddr(*blkaddr) &&
1031 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1032 					DATA_GENERIC_ENHANCE)) {
1033 			f2fs_put_dnode(&dn);
1034 			return -EFSCORRUPTED;
1035 		}
1036 
1037 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1038 
1039 			if (test_opt(sbi, LFS)) {
1040 				f2fs_put_dnode(&dn);
1041 				return -ENOTSUPP;
1042 			}
1043 
1044 			/* do not invalidate this block address */
1045 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1046 			*do_replace = 1;
1047 		}
1048 	}
1049 	f2fs_put_dnode(&dn);
1050 next:
1051 	len -= done;
1052 	off += done;
1053 	if (len)
1054 		goto next_dnode;
1055 	return 0;
1056 }
1057 
1058 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1059 				int *do_replace, pgoff_t off, int len)
1060 {
1061 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1062 	struct dnode_of_data dn;
1063 	int ret, i;
1064 
1065 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1066 		if (*do_replace == 0)
1067 			continue;
1068 
1069 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1070 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1071 		if (ret) {
1072 			dec_valid_block_count(sbi, inode, 1);
1073 			f2fs_invalidate_blocks(sbi, *blkaddr);
1074 		} else {
1075 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1076 		}
1077 		f2fs_put_dnode(&dn);
1078 	}
1079 	return 0;
1080 }
1081 
1082 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1083 			block_t *blkaddr, int *do_replace,
1084 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1085 {
1086 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1087 	pgoff_t i = 0;
1088 	int ret;
1089 
1090 	while (i < len) {
1091 		if (blkaddr[i] == NULL_ADDR && !full) {
1092 			i++;
1093 			continue;
1094 		}
1095 
1096 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1097 			struct dnode_of_data dn;
1098 			struct node_info ni;
1099 			size_t new_size;
1100 			pgoff_t ilen;
1101 
1102 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1103 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1104 			if (ret)
1105 				return ret;
1106 
1107 			ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1108 			if (ret) {
1109 				f2fs_put_dnode(&dn);
1110 				return ret;
1111 			}
1112 
1113 			ilen = min((pgoff_t)
1114 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1115 						dn.ofs_in_node, len - i);
1116 			do {
1117 				dn.data_blkaddr = datablock_addr(dn.inode,
1118 						dn.node_page, dn.ofs_in_node);
1119 				f2fs_truncate_data_blocks_range(&dn, 1);
1120 
1121 				if (do_replace[i]) {
1122 					f2fs_i_blocks_write(src_inode,
1123 							1, false, false);
1124 					f2fs_i_blocks_write(dst_inode,
1125 							1, true, false);
1126 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1127 					blkaddr[i], ni.version, true, false);
1128 
1129 					do_replace[i] = 0;
1130 				}
1131 				dn.ofs_in_node++;
1132 				i++;
1133 				new_size = (dst + i) << PAGE_SHIFT;
1134 				if (dst_inode->i_size < new_size)
1135 					f2fs_i_size_write(dst_inode, new_size);
1136 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1137 
1138 			f2fs_put_dnode(&dn);
1139 		} else {
1140 			struct page *psrc, *pdst;
1141 
1142 			psrc = f2fs_get_lock_data_page(src_inode,
1143 							src + i, true);
1144 			if (IS_ERR(psrc))
1145 				return PTR_ERR(psrc);
1146 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1147 								true);
1148 			if (IS_ERR(pdst)) {
1149 				f2fs_put_page(psrc, 1);
1150 				return PTR_ERR(pdst);
1151 			}
1152 			f2fs_copy_page(psrc, pdst);
1153 			set_page_dirty(pdst);
1154 			f2fs_put_page(pdst, 1);
1155 			f2fs_put_page(psrc, 1);
1156 
1157 			ret = f2fs_truncate_hole(src_inode,
1158 						src + i, src + i + 1);
1159 			if (ret)
1160 				return ret;
1161 			i++;
1162 		}
1163 	}
1164 	return 0;
1165 }
1166 
1167 static int __exchange_data_block(struct inode *src_inode,
1168 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1169 			pgoff_t len, bool full)
1170 {
1171 	block_t *src_blkaddr;
1172 	int *do_replace;
1173 	pgoff_t olen;
1174 	int ret;
1175 
1176 	while (len) {
1177 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1178 
1179 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1180 					array_size(olen, sizeof(block_t)),
1181 					GFP_KERNEL);
1182 		if (!src_blkaddr)
1183 			return -ENOMEM;
1184 
1185 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1186 					array_size(olen, sizeof(int)),
1187 					GFP_KERNEL);
1188 		if (!do_replace) {
1189 			kvfree(src_blkaddr);
1190 			return -ENOMEM;
1191 		}
1192 
1193 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1194 					do_replace, src, olen);
1195 		if (ret)
1196 			goto roll_back;
1197 
1198 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1199 					do_replace, src, dst, olen, full);
1200 		if (ret)
1201 			goto roll_back;
1202 
1203 		src += olen;
1204 		dst += olen;
1205 		len -= olen;
1206 
1207 		kvfree(src_blkaddr);
1208 		kvfree(do_replace);
1209 	}
1210 	return 0;
1211 
1212 roll_back:
1213 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1214 	kvfree(src_blkaddr);
1215 	kvfree(do_replace);
1216 	return ret;
1217 }
1218 
1219 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1220 {
1221 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1222 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1223 	pgoff_t start = offset >> PAGE_SHIFT;
1224 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1225 	int ret;
1226 
1227 	f2fs_balance_fs(sbi, true);
1228 
1229 	/* avoid gc operation during block exchange */
1230 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1231 	down_write(&F2FS_I(inode)->i_mmap_sem);
1232 
1233 	f2fs_lock_op(sbi);
1234 	f2fs_drop_extent_tree(inode);
1235 	truncate_pagecache(inode, offset);
1236 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1237 	f2fs_unlock_op(sbi);
1238 
1239 	up_write(&F2FS_I(inode)->i_mmap_sem);
1240 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1241 	return ret;
1242 }
1243 
1244 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1245 {
1246 	loff_t new_size;
1247 	int ret;
1248 
1249 	if (offset + len >= i_size_read(inode))
1250 		return -EINVAL;
1251 
1252 	/* collapse range should be aligned to block size of f2fs. */
1253 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1254 		return -EINVAL;
1255 
1256 	ret = f2fs_convert_inline_inode(inode);
1257 	if (ret)
1258 		return ret;
1259 
1260 	/* write out all dirty pages from offset */
1261 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1262 	if (ret)
1263 		return ret;
1264 
1265 	ret = f2fs_do_collapse(inode, offset, len);
1266 	if (ret)
1267 		return ret;
1268 
1269 	/* write out all moved pages, if possible */
1270 	down_write(&F2FS_I(inode)->i_mmap_sem);
1271 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1272 	truncate_pagecache(inode, offset);
1273 
1274 	new_size = i_size_read(inode) - len;
1275 	truncate_pagecache(inode, new_size);
1276 
1277 	ret = f2fs_truncate_blocks(inode, new_size, true);
1278 	up_write(&F2FS_I(inode)->i_mmap_sem);
1279 	if (!ret)
1280 		f2fs_i_size_write(inode, new_size);
1281 	return ret;
1282 }
1283 
1284 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1285 								pgoff_t end)
1286 {
1287 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1288 	pgoff_t index = start;
1289 	unsigned int ofs_in_node = dn->ofs_in_node;
1290 	blkcnt_t count = 0;
1291 	int ret;
1292 
1293 	for (; index < end; index++, dn->ofs_in_node++) {
1294 		if (datablock_addr(dn->inode, dn->node_page,
1295 					dn->ofs_in_node) == NULL_ADDR)
1296 			count++;
1297 	}
1298 
1299 	dn->ofs_in_node = ofs_in_node;
1300 	ret = f2fs_reserve_new_blocks(dn, count);
1301 	if (ret)
1302 		return ret;
1303 
1304 	dn->ofs_in_node = ofs_in_node;
1305 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1306 		dn->data_blkaddr = datablock_addr(dn->inode,
1307 					dn->node_page, dn->ofs_in_node);
1308 		/*
1309 		 * f2fs_reserve_new_blocks will not guarantee entire block
1310 		 * allocation.
1311 		 */
1312 		if (dn->data_blkaddr == NULL_ADDR) {
1313 			ret = -ENOSPC;
1314 			break;
1315 		}
1316 		if (dn->data_blkaddr != NEW_ADDR) {
1317 			f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1318 			dn->data_blkaddr = NEW_ADDR;
1319 			f2fs_set_data_blkaddr(dn);
1320 		}
1321 	}
1322 
1323 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1324 
1325 	return ret;
1326 }
1327 
1328 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1329 								int mode)
1330 {
1331 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1332 	struct address_space *mapping = inode->i_mapping;
1333 	pgoff_t index, pg_start, pg_end;
1334 	loff_t new_size = i_size_read(inode);
1335 	loff_t off_start, off_end;
1336 	int ret = 0;
1337 
1338 	ret = inode_newsize_ok(inode, (len + offset));
1339 	if (ret)
1340 		return ret;
1341 
1342 	ret = f2fs_convert_inline_inode(inode);
1343 	if (ret)
1344 		return ret;
1345 
1346 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1347 	if (ret)
1348 		return ret;
1349 
1350 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1351 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1352 
1353 	off_start = offset & (PAGE_SIZE - 1);
1354 	off_end = (offset + len) & (PAGE_SIZE - 1);
1355 
1356 	if (pg_start == pg_end) {
1357 		ret = fill_zero(inode, pg_start, off_start,
1358 						off_end - off_start);
1359 		if (ret)
1360 			return ret;
1361 
1362 		new_size = max_t(loff_t, new_size, offset + len);
1363 	} else {
1364 		if (off_start) {
1365 			ret = fill_zero(inode, pg_start++, off_start,
1366 						PAGE_SIZE - off_start);
1367 			if (ret)
1368 				return ret;
1369 
1370 			new_size = max_t(loff_t, new_size,
1371 					(loff_t)pg_start << PAGE_SHIFT);
1372 		}
1373 
1374 		for (index = pg_start; index < pg_end;) {
1375 			struct dnode_of_data dn;
1376 			unsigned int end_offset;
1377 			pgoff_t end;
1378 
1379 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1380 			down_write(&F2FS_I(inode)->i_mmap_sem);
1381 
1382 			truncate_pagecache_range(inode,
1383 				(loff_t)index << PAGE_SHIFT,
1384 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1385 
1386 			f2fs_lock_op(sbi);
1387 
1388 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1389 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1390 			if (ret) {
1391 				f2fs_unlock_op(sbi);
1392 				up_write(&F2FS_I(inode)->i_mmap_sem);
1393 				up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1394 				goto out;
1395 			}
1396 
1397 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1398 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1399 
1400 			ret = f2fs_do_zero_range(&dn, index, end);
1401 			f2fs_put_dnode(&dn);
1402 
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 
1407 			f2fs_balance_fs(sbi, dn.node_changed);
1408 
1409 			if (ret)
1410 				goto out;
1411 
1412 			index = end;
1413 			new_size = max_t(loff_t, new_size,
1414 					(loff_t)index << PAGE_SHIFT);
1415 		}
1416 
1417 		if (off_end) {
1418 			ret = fill_zero(inode, pg_end, 0, off_end);
1419 			if (ret)
1420 				goto out;
1421 
1422 			new_size = max_t(loff_t, new_size, offset + len);
1423 		}
1424 	}
1425 
1426 out:
1427 	if (new_size > i_size_read(inode)) {
1428 		if (mode & FALLOC_FL_KEEP_SIZE)
1429 			file_set_keep_isize(inode);
1430 		else
1431 			f2fs_i_size_write(inode, new_size);
1432 	}
1433 	return ret;
1434 }
1435 
1436 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1437 {
1438 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1439 	pgoff_t nr, pg_start, pg_end, delta, idx;
1440 	loff_t new_size;
1441 	int ret = 0;
1442 
1443 	new_size = i_size_read(inode) + len;
1444 	ret = inode_newsize_ok(inode, new_size);
1445 	if (ret)
1446 		return ret;
1447 
1448 	if (offset >= i_size_read(inode))
1449 		return -EINVAL;
1450 
1451 	/* insert range should be aligned to block size of f2fs. */
1452 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1453 		return -EINVAL;
1454 
1455 	ret = f2fs_convert_inline_inode(inode);
1456 	if (ret)
1457 		return ret;
1458 
1459 	f2fs_balance_fs(sbi, true);
1460 
1461 	down_write(&F2FS_I(inode)->i_mmap_sem);
1462 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1463 	up_write(&F2FS_I(inode)->i_mmap_sem);
1464 	if (ret)
1465 		return ret;
1466 
1467 	/* write out all dirty pages from offset */
1468 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1469 	if (ret)
1470 		return ret;
1471 
1472 	pg_start = offset >> PAGE_SHIFT;
1473 	pg_end = (offset + len) >> PAGE_SHIFT;
1474 	delta = pg_end - pg_start;
1475 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1476 
1477 	/* avoid gc operation during block exchange */
1478 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1479 	down_write(&F2FS_I(inode)->i_mmap_sem);
1480 	truncate_pagecache(inode, offset);
1481 
1482 	while (!ret && idx > pg_start) {
1483 		nr = idx - pg_start;
1484 		if (nr > delta)
1485 			nr = delta;
1486 		idx -= nr;
1487 
1488 		f2fs_lock_op(sbi);
1489 		f2fs_drop_extent_tree(inode);
1490 
1491 		ret = __exchange_data_block(inode, inode, idx,
1492 					idx + delta, nr, false);
1493 		f2fs_unlock_op(sbi);
1494 	}
1495 	up_write(&F2FS_I(inode)->i_mmap_sem);
1496 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1497 
1498 	/* write out all moved pages, if possible */
1499 	down_write(&F2FS_I(inode)->i_mmap_sem);
1500 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1501 	truncate_pagecache(inode, offset);
1502 	up_write(&F2FS_I(inode)->i_mmap_sem);
1503 
1504 	if (!ret)
1505 		f2fs_i_size_write(inode, new_size);
1506 	return ret;
1507 }
1508 
1509 static int expand_inode_data(struct inode *inode, loff_t offset,
1510 					loff_t len, int mode)
1511 {
1512 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1513 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1514 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1515 			.m_may_create = true };
1516 	pgoff_t pg_end;
1517 	loff_t new_size = i_size_read(inode);
1518 	loff_t off_end;
1519 	int err;
1520 
1521 	err = inode_newsize_ok(inode, (len + offset));
1522 	if (err)
1523 		return err;
1524 
1525 	err = f2fs_convert_inline_inode(inode);
1526 	if (err)
1527 		return err;
1528 
1529 	f2fs_balance_fs(sbi, true);
1530 
1531 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1532 	off_end = (offset + len) & (PAGE_SIZE - 1);
1533 
1534 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1535 	map.m_len = pg_end - map.m_lblk;
1536 	if (off_end)
1537 		map.m_len++;
1538 
1539 	if (f2fs_is_pinned_file(inode))
1540 		map.m_seg_type = CURSEG_COLD_DATA;
1541 
1542 	err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
1543 						F2FS_GET_BLOCK_PRE_DIO :
1544 						F2FS_GET_BLOCK_PRE_AIO));
1545 	if (err) {
1546 		pgoff_t last_off;
1547 
1548 		if (!map.m_len)
1549 			return err;
1550 
1551 		last_off = map.m_lblk + map.m_len - 1;
1552 
1553 		/* update new size to the failed position */
1554 		new_size = (last_off == pg_end) ? offset + len :
1555 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1556 	} else {
1557 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1558 	}
1559 
1560 	if (new_size > i_size_read(inode)) {
1561 		if (mode & FALLOC_FL_KEEP_SIZE)
1562 			file_set_keep_isize(inode);
1563 		else
1564 			f2fs_i_size_write(inode, new_size);
1565 	}
1566 
1567 	return err;
1568 }
1569 
1570 static long f2fs_fallocate(struct file *file, int mode,
1571 				loff_t offset, loff_t len)
1572 {
1573 	struct inode *inode = file_inode(file);
1574 	long ret = 0;
1575 
1576 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1577 		return -EIO;
1578 
1579 	/* f2fs only support ->fallocate for regular file */
1580 	if (!S_ISREG(inode->i_mode))
1581 		return -EINVAL;
1582 
1583 	if (IS_ENCRYPTED(inode) &&
1584 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1585 		return -EOPNOTSUPP;
1586 
1587 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1588 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1589 			FALLOC_FL_INSERT_RANGE))
1590 		return -EOPNOTSUPP;
1591 
1592 	inode_lock(inode);
1593 
1594 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1595 		if (offset >= inode->i_size)
1596 			goto out;
1597 
1598 		ret = punch_hole(inode, offset, len);
1599 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1600 		ret = f2fs_collapse_range(inode, offset, len);
1601 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1602 		ret = f2fs_zero_range(inode, offset, len, mode);
1603 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1604 		ret = f2fs_insert_range(inode, offset, len);
1605 	} else {
1606 		ret = expand_inode_data(inode, offset, len, mode);
1607 	}
1608 
1609 	if (!ret) {
1610 		inode->i_mtime = inode->i_ctime = current_time(inode);
1611 		f2fs_mark_inode_dirty_sync(inode, false);
1612 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1613 	}
1614 
1615 out:
1616 	inode_unlock(inode);
1617 
1618 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1619 	return ret;
1620 }
1621 
1622 static int f2fs_release_file(struct inode *inode, struct file *filp)
1623 {
1624 	/*
1625 	 * f2fs_relase_file is called at every close calls. So we should
1626 	 * not drop any inmemory pages by close called by other process.
1627 	 */
1628 	if (!(filp->f_mode & FMODE_WRITE) ||
1629 			atomic_read(&inode->i_writecount) != 1)
1630 		return 0;
1631 
1632 	/* some remained atomic pages should discarded */
1633 	if (f2fs_is_atomic_file(inode))
1634 		f2fs_drop_inmem_pages(inode);
1635 	if (f2fs_is_volatile_file(inode)) {
1636 		set_inode_flag(inode, FI_DROP_CACHE);
1637 		filemap_fdatawrite(inode->i_mapping);
1638 		clear_inode_flag(inode, FI_DROP_CACHE);
1639 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1640 		stat_dec_volatile_write(inode);
1641 	}
1642 	return 0;
1643 }
1644 
1645 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1646 {
1647 	struct inode *inode = file_inode(file);
1648 
1649 	/*
1650 	 * If the process doing a transaction is crashed, we should do
1651 	 * roll-back. Otherwise, other reader/write can see corrupted database
1652 	 * until all the writers close its file. Since this should be done
1653 	 * before dropping file lock, it needs to do in ->flush.
1654 	 */
1655 	if (f2fs_is_atomic_file(inode) &&
1656 			F2FS_I(inode)->inmem_task == current)
1657 		f2fs_drop_inmem_pages(inode);
1658 	return 0;
1659 }
1660 
1661 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1662 {
1663 	struct f2fs_inode_info *fi = F2FS_I(inode);
1664 
1665 	/* Is it quota file? Do not allow user to mess with it */
1666 	if (IS_NOQUOTA(inode))
1667 		return -EPERM;
1668 
1669 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1670 
1671 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1672 		set_inode_flag(inode, FI_PROJ_INHERIT);
1673 	else
1674 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1675 
1676 	inode->i_ctime = current_time(inode);
1677 	f2fs_set_inode_flags(inode);
1678 	f2fs_mark_inode_dirty_sync(inode, true);
1679 	return 0;
1680 }
1681 
1682 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1683 
1684 /*
1685  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1686  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1687  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1688  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1689  */
1690 
1691 static const struct {
1692 	u32 iflag;
1693 	u32 fsflag;
1694 } f2fs_fsflags_map[] = {
1695 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1696 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1697 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1698 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1699 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1700 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1701 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1702 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1703 };
1704 
1705 #define F2FS_GETTABLE_FS_FL (		\
1706 		FS_SYNC_FL |		\
1707 		FS_IMMUTABLE_FL |	\
1708 		FS_APPEND_FL |		\
1709 		FS_NODUMP_FL |		\
1710 		FS_NOATIME_FL |		\
1711 		FS_INDEX_FL |		\
1712 		FS_DIRSYNC_FL |		\
1713 		FS_PROJINHERIT_FL |	\
1714 		FS_ENCRYPT_FL |		\
1715 		FS_INLINE_DATA_FL |	\
1716 		FS_NOCOW_FL |		\
1717 		FS_VERITY_FL)
1718 
1719 #define F2FS_SETTABLE_FS_FL (		\
1720 		FS_SYNC_FL |		\
1721 		FS_IMMUTABLE_FL |	\
1722 		FS_APPEND_FL |		\
1723 		FS_NODUMP_FL |		\
1724 		FS_NOATIME_FL |		\
1725 		FS_DIRSYNC_FL |		\
1726 		FS_PROJINHERIT_FL)
1727 
1728 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1729 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1730 {
1731 	u32 fsflags = 0;
1732 	int i;
1733 
1734 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1735 		if (iflags & f2fs_fsflags_map[i].iflag)
1736 			fsflags |= f2fs_fsflags_map[i].fsflag;
1737 
1738 	return fsflags;
1739 }
1740 
1741 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1742 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1743 {
1744 	u32 iflags = 0;
1745 	int i;
1746 
1747 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1748 		if (fsflags & f2fs_fsflags_map[i].fsflag)
1749 			iflags |= f2fs_fsflags_map[i].iflag;
1750 
1751 	return iflags;
1752 }
1753 
1754 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1755 {
1756 	struct inode *inode = file_inode(filp);
1757 	struct f2fs_inode_info *fi = F2FS_I(inode);
1758 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1759 
1760 	if (IS_ENCRYPTED(inode))
1761 		fsflags |= FS_ENCRYPT_FL;
1762 	if (IS_VERITY(inode))
1763 		fsflags |= FS_VERITY_FL;
1764 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1765 		fsflags |= FS_INLINE_DATA_FL;
1766 	if (is_inode_flag_set(inode, FI_PIN_FILE))
1767 		fsflags |= FS_NOCOW_FL;
1768 
1769 	fsflags &= F2FS_GETTABLE_FS_FL;
1770 
1771 	return put_user(fsflags, (int __user *)arg);
1772 }
1773 
1774 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1775 {
1776 	struct inode *inode = file_inode(filp);
1777 	struct f2fs_inode_info *fi = F2FS_I(inode);
1778 	u32 fsflags, old_fsflags;
1779 	u32 iflags;
1780 	int ret;
1781 
1782 	if (!inode_owner_or_capable(inode))
1783 		return -EACCES;
1784 
1785 	if (get_user(fsflags, (int __user *)arg))
1786 		return -EFAULT;
1787 
1788 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
1789 		return -EOPNOTSUPP;
1790 	fsflags &= F2FS_SETTABLE_FS_FL;
1791 
1792 	iflags = f2fs_fsflags_to_iflags(fsflags);
1793 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1794 		return -EOPNOTSUPP;
1795 
1796 	ret = mnt_want_write_file(filp);
1797 	if (ret)
1798 		return ret;
1799 
1800 	inode_lock(inode);
1801 
1802 	old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1803 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1804 	if (ret)
1805 		goto out;
1806 
1807 	ret = f2fs_setflags_common(inode, iflags,
1808 			f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1809 out:
1810 	inode_unlock(inode);
1811 	mnt_drop_write_file(filp);
1812 	return ret;
1813 }
1814 
1815 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1816 {
1817 	struct inode *inode = file_inode(filp);
1818 
1819 	return put_user(inode->i_generation, (int __user *)arg);
1820 }
1821 
1822 static int f2fs_ioc_start_atomic_write(struct file *filp)
1823 {
1824 	struct inode *inode = file_inode(filp);
1825 	int ret;
1826 
1827 	if (!inode_owner_or_capable(inode))
1828 		return -EACCES;
1829 
1830 	if (!S_ISREG(inode->i_mode))
1831 		return -EINVAL;
1832 
1833 	ret = mnt_want_write_file(filp);
1834 	if (ret)
1835 		return ret;
1836 
1837 	inode_lock(inode);
1838 
1839 	if (f2fs_is_atomic_file(inode)) {
1840 		if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1841 			ret = -EINVAL;
1842 		goto out;
1843 	}
1844 
1845 	ret = f2fs_convert_inline_inode(inode);
1846 	if (ret)
1847 		goto out;
1848 
1849 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1850 
1851 	/*
1852 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1853 	 * f2fs_is_atomic_file.
1854 	 */
1855 	if (get_dirty_pages(inode))
1856 		f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1857 			  inode->i_ino, get_dirty_pages(inode));
1858 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1859 	if (ret) {
1860 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1861 		goto out;
1862 	}
1863 
1864 	set_inode_flag(inode, FI_ATOMIC_FILE);
1865 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1866 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1867 
1868 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1869 	F2FS_I(inode)->inmem_task = current;
1870 	stat_inc_atomic_write(inode);
1871 	stat_update_max_atomic_write(inode);
1872 out:
1873 	inode_unlock(inode);
1874 	mnt_drop_write_file(filp);
1875 	return ret;
1876 }
1877 
1878 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1879 {
1880 	struct inode *inode = file_inode(filp);
1881 	int ret;
1882 
1883 	if (!inode_owner_or_capable(inode))
1884 		return -EACCES;
1885 
1886 	ret = mnt_want_write_file(filp);
1887 	if (ret)
1888 		return ret;
1889 
1890 	f2fs_balance_fs(F2FS_I_SB(inode), true);
1891 
1892 	inode_lock(inode);
1893 
1894 	if (f2fs_is_volatile_file(inode)) {
1895 		ret = -EINVAL;
1896 		goto err_out;
1897 	}
1898 
1899 	if (f2fs_is_atomic_file(inode)) {
1900 		ret = f2fs_commit_inmem_pages(inode);
1901 		if (ret)
1902 			goto err_out;
1903 
1904 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1905 		if (!ret) {
1906 			clear_inode_flag(inode, FI_ATOMIC_FILE);
1907 			F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1908 			stat_dec_atomic_write(inode);
1909 		}
1910 	} else {
1911 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1912 	}
1913 err_out:
1914 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1915 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1916 		ret = -EINVAL;
1917 	}
1918 	inode_unlock(inode);
1919 	mnt_drop_write_file(filp);
1920 	return ret;
1921 }
1922 
1923 static int f2fs_ioc_start_volatile_write(struct file *filp)
1924 {
1925 	struct inode *inode = file_inode(filp);
1926 	int ret;
1927 
1928 	if (!inode_owner_or_capable(inode))
1929 		return -EACCES;
1930 
1931 	if (!S_ISREG(inode->i_mode))
1932 		return -EINVAL;
1933 
1934 	ret = mnt_want_write_file(filp);
1935 	if (ret)
1936 		return ret;
1937 
1938 	inode_lock(inode);
1939 
1940 	if (f2fs_is_volatile_file(inode))
1941 		goto out;
1942 
1943 	ret = f2fs_convert_inline_inode(inode);
1944 	if (ret)
1945 		goto out;
1946 
1947 	stat_inc_volatile_write(inode);
1948 	stat_update_max_volatile_write(inode);
1949 
1950 	set_inode_flag(inode, FI_VOLATILE_FILE);
1951 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1952 out:
1953 	inode_unlock(inode);
1954 	mnt_drop_write_file(filp);
1955 	return ret;
1956 }
1957 
1958 static int f2fs_ioc_release_volatile_write(struct file *filp)
1959 {
1960 	struct inode *inode = file_inode(filp);
1961 	int ret;
1962 
1963 	if (!inode_owner_or_capable(inode))
1964 		return -EACCES;
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 	if (!f2fs_is_first_block_written(inode)) {
1976 		ret = truncate_partial_data_page(inode, 0, true);
1977 		goto out;
1978 	}
1979 
1980 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1981 out:
1982 	inode_unlock(inode);
1983 	mnt_drop_write_file(filp);
1984 	return ret;
1985 }
1986 
1987 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1988 {
1989 	struct inode *inode = file_inode(filp);
1990 	int ret;
1991 
1992 	if (!inode_owner_or_capable(inode))
1993 		return -EACCES;
1994 
1995 	ret = mnt_want_write_file(filp);
1996 	if (ret)
1997 		return ret;
1998 
1999 	inode_lock(inode);
2000 
2001 	if (f2fs_is_atomic_file(inode))
2002 		f2fs_drop_inmem_pages(inode);
2003 	if (f2fs_is_volatile_file(inode)) {
2004 		clear_inode_flag(inode, FI_VOLATILE_FILE);
2005 		stat_dec_volatile_write(inode);
2006 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2007 	}
2008 
2009 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2010 
2011 	inode_unlock(inode);
2012 
2013 	mnt_drop_write_file(filp);
2014 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2015 	return ret;
2016 }
2017 
2018 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2019 {
2020 	struct inode *inode = file_inode(filp);
2021 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2022 	struct super_block *sb = sbi->sb;
2023 	__u32 in;
2024 	int ret = 0;
2025 
2026 	if (!capable(CAP_SYS_ADMIN))
2027 		return -EPERM;
2028 
2029 	if (get_user(in, (__u32 __user *)arg))
2030 		return -EFAULT;
2031 
2032 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2033 		ret = mnt_want_write_file(filp);
2034 		if (ret)
2035 			return ret;
2036 	}
2037 
2038 	switch (in) {
2039 	case F2FS_GOING_DOWN_FULLSYNC:
2040 		sb = freeze_bdev(sb->s_bdev);
2041 		if (IS_ERR(sb)) {
2042 			ret = PTR_ERR(sb);
2043 			goto out;
2044 		}
2045 		if (sb) {
2046 			f2fs_stop_checkpoint(sbi, false);
2047 			set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2048 			thaw_bdev(sb->s_bdev, sb);
2049 		}
2050 		break;
2051 	case F2FS_GOING_DOWN_METASYNC:
2052 		/* do checkpoint only */
2053 		ret = f2fs_sync_fs(sb, 1);
2054 		if (ret)
2055 			goto out;
2056 		f2fs_stop_checkpoint(sbi, false);
2057 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2058 		break;
2059 	case F2FS_GOING_DOWN_NOSYNC:
2060 		f2fs_stop_checkpoint(sbi, false);
2061 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2062 		break;
2063 	case F2FS_GOING_DOWN_METAFLUSH:
2064 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2065 		f2fs_stop_checkpoint(sbi, false);
2066 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2067 		break;
2068 	case F2FS_GOING_DOWN_NEED_FSCK:
2069 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2070 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2071 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2072 		/* do checkpoint only */
2073 		ret = f2fs_sync_fs(sb, 1);
2074 		goto out;
2075 	default:
2076 		ret = -EINVAL;
2077 		goto out;
2078 	}
2079 
2080 	f2fs_stop_gc_thread(sbi);
2081 	f2fs_stop_discard_thread(sbi);
2082 
2083 	f2fs_drop_discard_cmd(sbi);
2084 	clear_opt(sbi, DISCARD);
2085 
2086 	f2fs_update_time(sbi, REQ_TIME);
2087 out:
2088 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2089 		mnt_drop_write_file(filp);
2090 
2091 	trace_f2fs_shutdown(sbi, in, ret);
2092 
2093 	return ret;
2094 }
2095 
2096 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2097 {
2098 	struct inode *inode = file_inode(filp);
2099 	struct super_block *sb = inode->i_sb;
2100 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
2101 	struct fstrim_range range;
2102 	int ret;
2103 
2104 	if (!capable(CAP_SYS_ADMIN))
2105 		return -EPERM;
2106 
2107 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2108 		return -EOPNOTSUPP;
2109 
2110 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2111 				sizeof(range)))
2112 		return -EFAULT;
2113 
2114 	ret = mnt_want_write_file(filp);
2115 	if (ret)
2116 		return ret;
2117 
2118 	range.minlen = max((unsigned int)range.minlen,
2119 				q->limits.discard_granularity);
2120 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2121 	mnt_drop_write_file(filp);
2122 	if (ret < 0)
2123 		return ret;
2124 
2125 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2126 				sizeof(range)))
2127 		return -EFAULT;
2128 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2129 	return 0;
2130 }
2131 
2132 static bool uuid_is_nonzero(__u8 u[16])
2133 {
2134 	int i;
2135 
2136 	for (i = 0; i < 16; i++)
2137 		if (u[i])
2138 			return true;
2139 	return false;
2140 }
2141 
2142 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2143 {
2144 	struct inode *inode = file_inode(filp);
2145 
2146 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2147 		return -EOPNOTSUPP;
2148 
2149 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2150 
2151 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2152 }
2153 
2154 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2155 {
2156 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2157 		return -EOPNOTSUPP;
2158 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2159 }
2160 
2161 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2162 {
2163 	struct inode *inode = file_inode(filp);
2164 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2165 	int err;
2166 
2167 	if (!f2fs_sb_has_encrypt(sbi))
2168 		return -EOPNOTSUPP;
2169 
2170 	err = mnt_want_write_file(filp);
2171 	if (err)
2172 		return err;
2173 
2174 	down_write(&sbi->sb_lock);
2175 
2176 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2177 		goto got_it;
2178 
2179 	/* update superblock with uuid */
2180 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2181 
2182 	err = f2fs_commit_super(sbi, false);
2183 	if (err) {
2184 		/* undo new data */
2185 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2186 		goto out_err;
2187 	}
2188 got_it:
2189 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2190 									16))
2191 		err = -EFAULT;
2192 out_err:
2193 	up_write(&sbi->sb_lock);
2194 	mnt_drop_write_file(filp);
2195 	return err;
2196 }
2197 
2198 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2199 					     unsigned long arg)
2200 {
2201 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2202 		return -EOPNOTSUPP;
2203 
2204 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2205 }
2206 
2207 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2208 {
2209 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2210 		return -EOPNOTSUPP;
2211 
2212 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2213 }
2214 
2215 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2216 {
2217 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2218 		return -EOPNOTSUPP;
2219 
2220 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2221 }
2222 
2223 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2224 						    unsigned long arg)
2225 {
2226 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2227 		return -EOPNOTSUPP;
2228 
2229 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2230 }
2231 
2232 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2233 					      unsigned long arg)
2234 {
2235 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2236 		return -EOPNOTSUPP;
2237 
2238 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2239 }
2240 
2241 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2242 {
2243 	struct inode *inode = file_inode(filp);
2244 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2245 	__u32 sync;
2246 	int ret;
2247 
2248 	if (!capable(CAP_SYS_ADMIN))
2249 		return -EPERM;
2250 
2251 	if (get_user(sync, (__u32 __user *)arg))
2252 		return -EFAULT;
2253 
2254 	if (f2fs_readonly(sbi->sb))
2255 		return -EROFS;
2256 
2257 	ret = mnt_want_write_file(filp);
2258 	if (ret)
2259 		return ret;
2260 
2261 	if (!sync) {
2262 		if (!mutex_trylock(&sbi->gc_mutex)) {
2263 			ret = -EBUSY;
2264 			goto out;
2265 		}
2266 	} else {
2267 		mutex_lock(&sbi->gc_mutex);
2268 	}
2269 
2270 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2271 out:
2272 	mnt_drop_write_file(filp);
2273 	return ret;
2274 }
2275 
2276 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2277 {
2278 	struct inode *inode = file_inode(filp);
2279 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2280 	struct f2fs_gc_range range;
2281 	u64 end;
2282 	int ret;
2283 
2284 	if (!capable(CAP_SYS_ADMIN))
2285 		return -EPERM;
2286 
2287 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2288 							sizeof(range)))
2289 		return -EFAULT;
2290 
2291 	if (f2fs_readonly(sbi->sb))
2292 		return -EROFS;
2293 
2294 	end = range.start + range.len;
2295 	if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2296 		return -EINVAL;
2297 	}
2298 
2299 	ret = mnt_want_write_file(filp);
2300 	if (ret)
2301 		return ret;
2302 
2303 do_more:
2304 	if (!range.sync) {
2305 		if (!mutex_trylock(&sbi->gc_mutex)) {
2306 			ret = -EBUSY;
2307 			goto out;
2308 		}
2309 	} else {
2310 		mutex_lock(&sbi->gc_mutex);
2311 	}
2312 
2313 	ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2314 	range.start += BLKS_PER_SEC(sbi);
2315 	if (range.start <= end)
2316 		goto do_more;
2317 out:
2318 	mnt_drop_write_file(filp);
2319 	return ret;
2320 }
2321 
2322 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2323 {
2324 	struct inode *inode = file_inode(filp);
2325 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2326 	int ret;
2327 
2328 	if (!capable(CAP_SYS_ADMIN))
2329 		return -EPERM;
2330 
2331 	if (f2fs_readonly(sbi->sb))
2332 		return -EROFS;
2333 
2334 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2335 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2336 		return -EINVAL;
2337 	}
2338 
2339 	ret = mnt_want_write_file(filp);
2340 	if (ret)
2341 		return ret;
2342 
2343 	ret = f2fs_sync_fs(sbi->sb, 1);
2344 
2345 	mnt_drop_write_file(filp);
2346 	return ret;
2347 }
2348 
2349 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2350 					struct file *filp,
2351 					struct f2fs_defragment *range)
2352 {
2353 	struct inode *inode = file_inode(filp);
2354 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2355 					.m_seg_type = NO_CHECK_TYPE ,
2356 					.m_may_create = false };
2357 	struct extent_info ei = {0, 0, 0};
2358 	pgoff_t pg_start, pg_end, next_pgofs;
2359 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2360 	unsigned int total = 0, sec_num;
2361 	block_t blk_end = 0;
2362 	bool fragmented = false;
2363 	int err;
2364 
2365 	/* if in-place-update policy is enabled, don't waste time here */
2366 	if (f2fs_should_update_inplace(inode, NULL))
2367 		return -EINVAL;
2368 
2369 	pg_start = range->start >> PAGE_SHIFT;
2370 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2371 
2372 	f2fs_balance_fs(sbi, true);
2373 
2374 	inode_lock(inode);
2375 
2376 	/* writeback all dirty pages in the range */
2377 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2378 						range->start + range->len - 1);
2379 	if (err)
2380 		goto out;
2381 
2382 	/*
2383 	 * lookup mapping info in extent cache, skip defragmenting if physical
2384 	 * block addresses are continuous.
2385 	 */
2386 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2387 		if (ei.fofs + ei.len >= pg_end)
2388 			goto out;
2389 	}
2390 
2391 	map.m_lblk = pg_start;
2392 	map.m_next_pgofs = &next_pgofs;
2393 
2394 	/*
2395 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2396 	 * physical block addresses are continuous even if there are hole(s)
2397 	 * in logical blocks.
2398 	 */
2399 	while (map.m_lblk < pg_end) {
2400 		map.m_len = pg_end - map.m_lblk;
2401 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2402 		if (err)
2403 			goto out;
2404 
2405 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2406 			map.m_lblk = next_pgofs;
2407 			continue;
2408 		}
2409 
2410 		if (blk_end && blk_end != map.m_pblk)
2411 			fragmented = true;
2412 
2413 		/* record total count of block that we're going to move */
2414 		total += map.m_len;
2415 
2416 		blk_end = map.m_pblk + map.m_len;
2417 
2418 		map.m_lblk += map.m_len;
2419 	}
2420 
2421 	if (!fragmented)
2422 		goto out;
2423 
2424 	sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2425 
2426 	/*
2427 	 * make sure there are enough free section for LFS allocation, this can
2428 	 * avoid defragment running in SSR mode when free section are allocated
2429 	 * intensively
2430 	 */
2431 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2432 		err = -EAGAIN;
2433 		goto out;
2434 	}
2435 
2436 	map.m_lblk = pg_start;
2437 	map.m_len = pg_end - pg_start;
2438 	total = 0;
2439 
2440 	while (map.m_lblk < pg_end) {
2441 		pgoff_t idx;
2442 		int cnt = 0;
2443 
2444 do_map:
2445 		map.m_len = pg_end - map.m_lblk;
2446 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2447 		if (err)
2448 			goto clear_out;
2449 
2450 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2451 			map.m_lblk = next_pgofs;
2452 			continue;
2453 		}
2454 
2455 		set_inode_flag(inode, FI_DO_DEFRAG);
2456 
2457 		idx = map.m_lblk;
2458 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2459 			struct page *page;
2460 
2461 			page = f2fs_get_lock_data_page(inode, idx, true);
2462 			if (IS_ERR(page)) {
2463 				err = PTR_ERR(page);
2464 				goto clear_out;
2465 			}
2466 
2467 			set_page_dirty(page);
2468 			f2fs_put_page(page, 1);
2469 
2470 			idx++;
2471 			cnt++;
2472 			total++;
2473 		}
2474 
2475 		map.m_lblk = idx;
2476 
2477 		if (idx < pg_end && cnt < blk_per_seg)
2478 			goto do_map;
2479 
2480 		clear_inode_flag(inode, FI_DO_DEFRAG);
2481 
2482 		err = filemap_fdatawrite(inode->i_mapping);
2483 		if (err)
2484 			goto out;
2485 	}
2486 clear_out:
2487 	clear_inode_flag(inode, FI_DO_DEFRAG);
2488 out:
2489 	inode_unlock(inode);
2490 	if (!err)
2491 		range->len = (u64)total << PAGE_SHIFT;
2492 	return err;
2493 }
2494 
2495 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2496 {
2497 	struct inode *inode = file_inode(filp);
2498 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2499 	struct f2fs_defragment range;
2500 	int err;
2501 
2502 	if (!capable(CAP_SYS_ADMIN))
2503 		return -EPERM;
2504 
2505 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2506 		return -EINVAL;
2507 
2508 	if (f2fs_readonly(sbi->sb))
2509 		return -EROFS;
2510 
2511 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2512 							sizeof(range)))
2513 		return -EFAULT;
2514 
2515 	/* verify alignment of offset & size */
2516 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2517 		return -EINVAL;
2518 
2519 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2520 					sbi->max_file_blocks))
2521 		return -EINVAL;
2522 
2523 	err = mnt_want_write_file(filp);
2524 	if (err)
2525 		return err;
2526 
2527 	err = f2fs_defragment_range(sbi, filp, &range);
2528 	mnt_drop_write_file(filp);
2529 
2530 	f2fs_update_time(sbi, REQ_TIME);
2531 	if (err < 0)
2532 		return err;
2533 
2534 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2535 							sizeof(range)))
2536 		return -EFAULT;
2537 
2538 	return 0;
2539 }
2540 
2541 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2542 			struct file *file_out, loff_t pos_out, size_t len)
2543 {
2544 	struct inode *src = file_inode(file_in);
2545 	struct inode *dst = file_inode(file_out);
2546 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2547 	size_t olen = len, dst_max_i_size = 0;
2548 	size_t dst_osize;
2549 	int ret;
2550 
2551 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2552 				src->i_sb != dst->i_sb)
2553 		return -EXDEV;
2554 
2555 	if (unlikely(f2fs_readonly(src->i_sb)))
2556 		return -EROFS;
2557 
2558 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2559 		return -EINVAL;
2560 
2561 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2562 		return -EOPNOTSUPP;
2563 
2564 	if (src == dst) {
2565 		if (pos_in == pos_out)
2566 			return 0;
2567 		if (pos_out > pos_in && pos_out < pos_in + len)
2568 			return -EINVAL;
2569 	}
2570 
2571 	inode_lock(src);
2572 	if (src != dst) {
2573 		ret = -EBUSY;
2574 		if (!inode_trylock(dst))
2575 			goto out;
2576 	}
2577 
2578 	ret = -EINVAL;
2579 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2580 		goto out_unlock;
2581 	if (len == 0)
2582 		olen = len = src->i_size - pos_in;
2583 	if (pos_in + len == src->i_size)
2584 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2585 	if (len == 0) {
2586 		ret = 0;
2587 		goto out_unlock;
2588 	}
2589 
2590 	dst_osize = dst->i_size;
2591 	if (pos_out + olen > dst->i_size)
2592 		dst_max_i_size = pos_out + olen;
2593 
2594 	/* verify the end result is block aligned */
2595 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2596 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2597 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2598 		goto out_unlock;
2599 
2600 	ret = f2fs_convert_inline_inode(src);
2601 	if (ret)
2602 		goto out_unlock;
2603 
2604 	ret = f2fs_convert_inline_inode(dst);
2605 	if (ret)
2606 		goto out_unlock;
2607 
2608 	/* write out all dirty pages from offset */
2609 	ret = filemap_write_and_wait_range(src->i_mapping,
2610 					pos_in, pos_in + len);
2611 	if (ret)
2612 		goto out_unlock;
2613 
2614 	ret = filemap_write_and_wait_range(dst->i_mapping,
2615 					pos_out, pos_out + len);
2616 	if (ret)
2617 		goto out_unlock;
2618 
2619 	f2fs_balance_fs(sbi, true);
2620 
2621 	down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2622 	if (src != dst) {
2623 		ret = -EBUSY;
2624 		if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2625 			goto out_src;
2626 	}
2627 
2628 	f2fs_lock_op(sbi);
2629 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2630 				pos_out >> F2FS_BLKSIZE_BITS,
2631 				len >> F2FS_BLKSIZE_BITS, false);
2632 
2633 	if (!ret) {
2634 		if (dst_max_i_size)
2635 			f2fs_i_size_write(dst, dst_max_i_size);
2636 		else if (dst_osize != dst->i_size)
2637 			f2fs_i_size_write(dst, dst_osize);
2638 	}
2639 	f2fs_unlock_op(sbi);
2640 
2641 	if (src != dst)
2642 		up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2643 out_src:
2644 	up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2645 out_unlock:
2646 	if (src != dst)
2647 		inode_unlock(dst);
2648 out:
2649 	inode_unlock(src);
2650 	return ret;
2651 }
2652 
2653 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2654 {
2655 	struct f2fs_move_range range;
2656 	struct fd dst;
2657 	int err;
2658 
2659 	if (!(filp->f_mode & FMODE_READ) ||
2660 			!(filp->f_mode & FMODE_WRITE))
2661 		return -EBADF;
2662 
2663 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2664 							sizeof(range)))
2665 		return -EFAULT;
2666 
2667 	dst = fdget(range.dst_fd);
2668 	if (!dst.file)
2669 		return -EBADF;
2670 
2671 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2672 		err = -EBADF;
2673 		goto err_out;
2674 	}
2675 
2676 	err = mnt_want_write_file(filp);
2677 	if (err)
2678 		goto err_out;
2679 
2680 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2681 					range.pos_out, range.len);
2682 
2683 	mnt_drop_write_file(filp);
2684 	if (err)
2685 		goto err_out;
2686 
2687 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2688 						&range, sizeof(range)))
2689 		err = -EFAULT;
2690 err_out:
2691 	fdput(dst);
2692 	return err;
2693 }
2694 
2695 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2696 {
2697 	struct inode *inode = file_inode(filp);
2698 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2699 	struct sit_info *sm = SIT_I(sbi);
2700 	unsigned int start_segno = 0, end_segno = 0;
2701 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2702 	struct f2fs_flush_device range;
2703 	int ret;
2704 
2705 	if (!capable(CAP_SYS_ADMIN))
2706 		return -EPERM;
2707 
2708 	if (f2fs_readonly(sbi->sb))
2709 		return -EROFS;
2710 
2711 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2712 		return -EINVAL;
2713 
2714 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2715 							sizeof(range)))
2716 		return -EFAULT;
2717 
2718 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2719 			__is_large_section(sbi)) {
2720 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2721 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2722 		return -EINVAL;
2723 	}
2724 
2725 	ret = mnt_want_write_file(filp);
2726 	if (ret)
2727 		return ret;
2728 
2729 	if (range.dev_num != 0)
2730 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2731 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2732 
2733 	start_segno = sm->last_victim[FLUSH_DEVICE];
2734 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2735 		start_segno = dev_start_segno;
2736 	end_segno = min(start_segno + range.segments, dev_end_segno);
2737 
2738 	while (start_segno < end_segno) {
2739 		if (!mutex_trylock(&sbi->gc_mutex)) {
2740 			ret = -EBUSY;
2741 			goto out;
2742 		}
2743 		sm->last_victim[GC_CB] = end_segno + 1;
2744 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2745 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2746 		ret = f2fs_gc(sbi, true, true, start_segno);
2747 		if (ret == -EAGAIN)
2748 			ret = 0;
2749 		else if (ret < 0)
2750 			break;
2751 		start_segno++;
2752 	}
2753 out:
2754 	mnt_drop_write_file(filp);
2755 	return ret;
2756 }
2757 
2758 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2759 {
2760 	struct inode *inode = file_inode(filp);
2761 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2762 
2763 	/* Must validate to set it with SQLite behavior in Android. */
2764 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2765 
2766 	return put_user(sb_feature, (u32 __user *)arg);
2767 }
2768 
2769 #ifdef CONFIG_QUOTA
2770 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2771 {
2772 	struct dquot *transfer_to[MAXQUOTAS] = {};
2773 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2774 	struct super_block *sb = sbi->sb;
2775 	int err = 0;
2776 
2777 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2778 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2779 		err = __dquot_transfer(inode, transfer_to);
2780 		if (err)
2781 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2782 		dqput(transfer_to[PRJQUOTA]);
2783 	}
2784 	return err;
2785 }
2786 
2787 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2788 {
2789 	struct inode *inode = file_inode(filp);
2790 	struct f2fs_inode_info *fi = F2FS_I(inode);
2791 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2792 	struct page *ipage;
2793 	kprojid_t kprojid;
2794 	int err;
2795 
2796 	if (!f2fs_sb_has_project_quota(sbi)) {
2797 		if (projid != F2FS_DEF_PROJID)
2798 			return -EOPNOTSUPP;
2799 		else
2800 			return 0;
2801 	}
2802 
2803 	if (!f2fs_has_extra_attr(inode))
2804 		return -EOPNOTSUPP;
2805 
2806 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2807 
2808 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2809 		return 0;
2810 
2811 	err = -EPERM;
2812 	/* Is it quota file? Do not allow user to mess with it */
2813 	if (IS_NOQUOTA(inode))
2814 		return err;
2815 
2816 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
2817 	if (IS_ERR(ipage))
2818 		return PTR_ERR(ipage);
2819 
2820 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2821 								i_projid)) {
2822 		err = -EOVERFLOW;
2823 		f2fs_put_page(ipage, 1);
2824 		return err;
2825 	}
2826 	f2fs_put_page(ipage, 1);
2827 
2828 	err = dquot_initialize(inode);
2829 	if (err)
2830 		return err;
2831 
2832 	f2fs_lock_op(sbi);
2833 	err = f2fs_transfer_project_quota(inode, kprojid);
2834 	if (err)
2835 		goto out_unlock;
2836 
2837 	F2FS_I(inode)->i_projid = kprojid;
2838 	inode->i_ctime = current_time(inode);
2839 	f2fs_mark_inode_dirty_sync(inode, true);
2840 out_unlock:
2841 	f2fs_unlock_op(sbi);
2842 	return err;
2843 }
2844 #else
2845 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2846 {
2847 	return 0;
2848 }
2849 
2850 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2851 {
2852 	if (projid != F2FS_DEF_PROJID)
2853 		return -EOPNOTSUPP;
2854 	return 0;
2855 }
2856 #endif
2857 
2858 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
2859 
2860 /*
2861  * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
2862  * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
2863  * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
2864  */
2865 
2866 static const struct {
2867 	u32 iflag;
2868 	u32 xflag;
2869 } f2fs_xflags_map[] = {
2870 	{ F2FS_SYNC_FL,		FS_XFLAG_SYNC },
2871 	{ F2FS_IMMUTABLE_FL,	FS_XFLAG_IMMUTABLE },
2872 	{ F2FS_APPEND_FL,	FS_XFLAG_APPEND },
2873 	{ F2FS_NODUMP_FL,	FS_XFLAG_NODUMP },
2874 	{ F2FS_NOATIME_FL,	FS_XFLAG_NOATIME },
2875 	{ F2FS_PROJINHERIT_FL,	FS_XFLAG_PROJINHERIT },
2876 };
2877 
2878 #define F2FS_SUPPORTED_XFLAGS (		\
2879 		FS_XFLAG_SYNC |		\
2880 		FS_XFLAG_IMMUTABLE |	\
2881 		FS_XFLAG_APPEND |	\
2882 		FS_XFLAG_NODUMP |	\
2883 		FS_XFLAG_NOATIME |	\
2884 		FS_XFLAG_PROJINHERIT)
2885 
2886 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
2887 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
2888 {
2889 	u32 xflags = 0;
2890 	int i;
2891 
2892 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2893 		if (iflags & f2fs_xflags_map[i].iflag)
2894 			xflags |= f2fs_xflags_map[i].xflag;
2895 
2896 	return xflags;
2897 }
2898 
2899 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
2900 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
2901 {
2902 	u32 iflags = 0;
2903 	int i;
2904 
2905 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2906 		if (xflags & f2fs_xflags_map[i].xflag)
2907 			iflags |= f2fs_xflags_map[i].iflag;
2908 
2909 	return iflags;
2910 }
2911 
2912 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
2913 {
2914 	struct f2fs_inode_info *fi = F2FS_I(inode);
2915 
2916 	simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
2917 
2918 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2919 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2920 }
2921 
2922 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2923 {
2924 	struct inode *inode = file_inode(filp);
2925 	struct fsxattr fa;
2926 
2927 	f2fs_fill_fsxattr(inode, &fa);
2928 
2929 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2930 		return -EFAULT;
2931 	return 0;
2932 }
2933 
2934 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2935 {
2936 	struct inode *inode = file_inode(filp);
2937 	struct fsxattr fa, old_fa;
2938 	u32 iflags;
2939 	int err;
2940 
2941 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2942 		return -EFAULT;
2943 
2944 	/* Make sure caller has proper permission */
2945 	if (!inode_owner_or_capable(inode))
2946 		return -EACCES;
2947 
2948 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
2949 		return -EOPNOTSUPP;
2950 
2951 	iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2952 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2953 		return -EOPNOTSUPP;
2954 
2955 	err = mnt_want_write_file(filp);
2956 	if (err)
2957 		return err;
2958 
2959 	inode_lock(inode);
2960 
2961 	f2fs_fill_fsxattr(inode, &old_fa);
2962 	err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
2963 	if (err)
2964 		goto out;
2965 
2966 	err = f2fs_setflags_common(inode, iflags,
2967 			f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
2968 	if (err)
2969 		goto out;
2970 
2971 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2972 out:
2973 	inode_unlock(inode);
2974 	mnt_drop_write_file(filp);
2975 	return err;
2976 }
2977 
2978 int f2fs_pin_file_control(struct inode *inode, bool inc)
2979 {
2980 	struct f2fs_inode_info *fi = F2FS_I(inode);
2981 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2982 
2983 	/* Use i_gc_failures for normal file as a risk signal. */
2984 	if (inc)
2985 		f2fs_i_gc_failures_write(inode,
2986 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2987 
2988 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2989 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
2990 			  __func__, inode->i_ino,
2991 			  fi->i_gc_failures[GC_FAILURE_PIN]);
2992 		clear_inode_flag(inode, FI_PIN_FILE);
2993 		return -EAGAIN;
2994 	}
2995 	return 0;
2996 }
2997 
2998 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2999 {
3000 	struct inode *inode = file_inode(filp);
3001 	__u32 pin;
3002 	int ret = 0;
3003 
3004 	if (get_user(pin, (__u32 __user *)arg))
3005 		return -EFAULT;
3006 
3007 	if (!S_ISREG(inode->i_mode))
3008 		return -EINVAL;
3009 
3010 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3011 		return -EROFS;
3012 
3013 	ret = mnt_want_write_file(filp);
3014 	if (ret)
3015 		return ret;
3016 
3017 	inode_lock(inode);
3018 
3019 	if (f2fs_should_update_outplace(inode, NULL)) {
3020 		ret = -EINVAL;
3021 		goto out;
3022 	}
3023 
3024 	if (!pin) {
3025 		clear_inode_flag(inode, FI_PIN_FILE);
3026 		f2fs_i_gc_failures_write(inode, 0);
3027 		goto done;
3028 	}
3029 
3030 	if (f2fs_pin_file_control(inode, false)) {
3031 		ret = -EAGAIN;
3032 		goto out;
3033 	}
3034 	ret = f2fs_convert_inline_inode(inode);
3035 	if (ret)
3036 		goto out;
3037 
3038 	set_inode_flag(inode, FI_PIN_FILE);
3039 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3040 done:
3041 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3042 out:
3043 	inode_unlock(inode);
3044 	mnt_drop_write_file(filp);
3045 	return ret;
3046 }
3047 
3048 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3049 {
3050 	struct inode *inode = file_inode(filp);
3051 	__u32 pin = 0;
3052 
3053 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3054 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3055 	return put_user(pin, (u32 __user *)arg);
3056 }
3057 
3058 int f2fs_precache_extents(struct inode *inode)
3059 {
3060 	struct f2fs_inode_info *fi = F2FS_I(inode);
3061 	struct f2fs_map_blocks map;
3062 	pgoff_t m_next_extent;
3063 	loff_t end;
3064 	int err;
3065 
3066 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3067 		return -EOPNOTSUPP;
3068 
3069 	map.m_lblk = 0;
3070 	map.m_next_pgofs = NULL;
3071 	map.m_next_extent = &m_next_extent;
3072 	map.m_seg_type = NO_CHECK_TYPE;
3073 	map.m_may_create = false;
3074 	end = F2FS_I_SB(inode)->max_file_blocks;
3075 
3076 	while (map.m_lblk < end) {
3077 		map.m_len = end - map.m_lblk;
3078 
3079 		down_write(&fi->i_gc_rwsem[WRITE]);
3080 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3081 		up_write(&fi->i_gc_rwsem[WRITE]);
3082 		if (err)
3083 			return err;
3084 
3085 		map.m_lblk = m_next_extent;
3086 	}
3087 
3088 	return err;
3089 }
3090 
3091 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3092 {
3093 	return f2fs_precache_extents(file_inode(filp));
3094 }
3095 
3096 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3097 {
3098 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3099 	__u64 block_count;
3100 	int ret;
3101 
3102 	if (!capable(CAP_SYS_ADMIN))
3103 		return -EPERM;
3104 
3105 	if (f2fs_readonly(sbi->sb))
3106 		return -EROFS;
3107 
3108 	if (copy_from_user(&block_count, (void __user *)arg,
3109 			   sizeof(block_count)))
3110 		return -EFAULT;
3111 
3112 	ret = f2fs_resize_fs(sbi, block_count);
3113 
3114 	return ret;
3115 }
3116 
3117 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3118 {
3119 	struct inode *inode = file_inode(filp);
3120 
3121 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3122 
3123 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3124 		f2fs_warn(F2FS_I_SB(inode),
3125 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3126 			  inode->i_ino);
3127 		return -EOPNOTSUPP;
3128 	}
3129 
3130 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3131 }
3132 
3133 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3134 {
3135 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3136 		return -EOPNOTSUPP;
3137 
3138 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3139 }
3140 
3141 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3142 {
3143 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3144 		return -EIO;
3145 
3146 	switch (cmd) {
3147 	case F2FS_IOC_GETFLAGS:
3148 		return f2fs_ioc_getflags(filp, arg);
3149 	case F2FS_IOC_SETFLAGS:
3150 		return f2fs_ioc_setflags(filp, arg);
3151 	case F2FS_IOC_GETVERSION:
3152 		return f2fs_ioc_getversion(filp, arg);
3153 	case F2FS_IOC_START_ATOMIC_WRITE:
3154 		return f2fs_ioc_start_atomic_write(filp);
3155 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3156 		return f2fs_ioc_commit_atomic_write(filp);
3157 	case F2FS_IOC_START_VOLATILE_WRITE:
3158 		return f2fs_ioc_start_volatile_write(filp);
3159 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3160 		return f2fs_ioc_release_volatile_write(filp);
3161 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3162 		return f2fs_ioc_abort_volatile_write(filp);
3163 	case F2FS_IOC_SHUTDOWN:
3164 		return f2fs_ioc_shutdown(filp, arg);
3165 	case FITRIM:
3166 		return f2fs_ioc_fitrim(filp, arg);
3167 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3168 		return f2fs_ioc_set_encryption_policy(filp, arg);
3169 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3170 		return f2fs_ioc_get_encryption_policy(filp, arg);
3171 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3172 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3173 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3174 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3175 	case FS_IOC_ADD_ENCRYPTION_KEY:
3176 		return f2fs_ioc_add_encryption_key(filp, arg);
3177 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3178 		return f2fs_ioc_remove_encryption_key(filp, arg);
3179 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3180 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3181 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3182 		return f2fs_ioc_get_encryption_key_status(filp, arg);
3183 	case F2FS_IOC_GARBAGE_COLLECT:
3184 		return f2fs_ioc_gc(filp, arg);
3185 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3186 		return f2fs_ioc_gc_range(filp, arg);
3187 	case F2FS_IOC_WRITE_CHECKPOINT:
3188 		return f2fs_ioc_write_checkpoint(filp, arg);
3189 	case F2FS_IOC_DEFRAGMENT:
3190 		return f2fs_ioc_defragment(filp, arg);
3191 	case F2FS_IOC_MOVE_RANGE:
3192 		return f2fs_ioc_move_range(filp, arg);
3193 	case F2FS_IOC_FLUSH_DEVICE:
3194 		return f2fs_ioc_flush_device(filp, arg);
3195 	case F2FS_IOC_GET_FEATURES:
3196 		return f2fs_ioc_get_features(filp, arg);
3197 	case F2FS_IOC_FSGETXATTR:
3198 		return f2fs_ioc_fsgetxattr(filp, arg);
3199 	case F2FS_IOC_FSSETXATTR:
3200 		return f2fs_ioc_fssetxattr(filp, arg);
3201 	case F2FS_IOC_GET_PIN_FILE:
3202 		return f2fs_ioc_get_pin_file(filp, arg);
3203 	case F2FS_IOC_SET_PIN_FILE:
3204 		return f2fs_ioc_set_pin_file(filp, arg);
3205 	case F2FS_IOC_PRECACHE_EXTENTS:
3206 		return f2fs_ioc_precache_extents(filp, arg);
3207 	case F2FS_IOC_RESIZE_FS:
3208 		return f2fs_ioc_resize_fs(filp, arg);
3209 	case FS_IOC_ENABLE_VERITY:
3210 		return f2fs_ioc_enable_verity(filp, arg);
3211 	case FS_IOC_MEASURE_VERITY:
3212 		return f2fs_ioc_measure_verity(filp, arg);
3213 	default:
3214 		return -ENOTTY;
3215 	}
3216 }
3217 
3218 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3219 {
3220 	struct file *file = iocb->ki_filp;
3221 	struct inode *inode = file_inode(file);
3222 	ssize_t ret;
3223 
3224 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3225 		ret = -EIO;
3226 		goto out;
3227 	}
3228 
3229 	if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT)) {
3230 		ret = -EINVAL;
3231 		goto out;
3232 	}
3233 
3234 	if (!inode_trylock(inode)) {
3235 		if (iocb->ki_flags & IOCB_NOWAIT) {
3236 			ret = -EAGAIN;
3237 			goto out;
3238 		}
3239 		inode_lock(inode);
3240 	}
3241 
3242 	ret = generic_write_checks(iocb, from);
3243 	if (ret > 0) {
3244 		bool preallocated = false;
3245 		size_t target_size = 0;
3246 		int err;
3247 
3248 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3249 			set_inode_flag(inode, FI_NO_PREALLOC);
3250 
3251 		if ((iocb->ki_flags & IOCB_NOWAIT)) {
3252 			if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3253 						iov_iter_count(from)) ||
3254 				f2fs_has_inline_data(inode) ||
3255 				f2fs_force_buffered_io(inode, iocb, from)) {
3256 				clear_inode_flag(inode, FI_NO_PREALLOC);
3257 				inode_unlock(inode);
3258 				ret = -EAGAIN;
3259 				goto out;
3260 			}
3261 		} else {
3262 			preallocated = true;
3263 			target_size = iocb->ki_pos + iov_iter_count(from);
3264 
3265 			err = f2fs_preallocate_blocks(iocb, from);
3266 			if (err) {
3267 				clear_inode_flag(inode, FI_NO_PREALLOC);
3268 				inode_unlock(inode);
3269 				ret = err;
3270 				goto out;
3271 			}
3272 		}
3273 		ret = __generic_file_write_iter(iocb, from);
3274 		clear_inode_flag(inode, FI_NO_PREALLOC);
3275 
3276 		/* if we couldn't write data, we should deallocate blocks. */
3277 		if (preallocated && i_size_read(inode) < target_size)
3278 			f2fs_truncate(inode);
3279 
3280 		if (ret > 0)
3281 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3282 	}
3283 	inode_unlock(inode);
3284 out:
3285 	trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3286 					iov_iter_count(from), ret);
3287 	if (ret > 0)
3288 		ret = generic_write_sync(iocb, ret);
3289 	return ret;
3290 }
3291 
3292 #ifdef CONFIG_COMPAT
3293 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3294 {
3295 	switch (cmd) {
3296 	case F2FS_IOC32_GETFLAGS:
3297 		cmd = F2FS_IOC_GETFLAGS;
3298 		break;
3299 	case F2FS_IOC32_SETFLAGS:
3300 		cmd = F2FS_IOC_SETFLAGS;
3301 		break;
3302 	case F2FS_IOC32_GETVERSION:
3303 		cmd = F2FS_IOC_GETVERSION;
3304 		break;
3305 	case F2FS_IOC_START_ATOMIC_WRITE:
3306 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3307 	case F2FS_IOC_START_VOLATILE_WRITE:
3308 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3309 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3310 	case F2FS_IOC_SHUTDOWN:
3311 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3312 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3313 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3314 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3315 	case FS_IOC_ADD_ENCRYPTION_KEY:
3316 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3317 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3318 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3319 	case F2FS_IOC_GARBAGE_COLLECT:
3320 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3321 	case F2FS_IOC_WRITE_CHECKPOINT:
3322 	case F2FS_IOC_DEFRAGMENT:
3323 	case F2FS_IOC_MOVE_RANGE:
3324 	case F2FS_IOC_FLUSH_DEVICE:
3325 	case F2FS_IOC_GET_FEATURES:
3326 	case F2FS_IOC_FSGETXATTR:
3327 	case F2FS_IOC_FSSETXATTR:
3328 	case F2FS_IOC_GET_PIN_FILE:
3329 	case F2FS_IOC_SET_PIN_FILE:
3330 	case F2FS_IOC_PRECACHE_EXTENTS:
3331 	case F2FS_IOC_RESIZE_FS:
3332 	case FS_IOC_ENABLE_VERITY:
3333 	case FS_IOC_MEASURE_VERITY:
3334 		break;
3335 	default:
3336 		return -ENOIOCTLCMD;
3337 	}
3338 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3339 }
3340 #endif
3341 
3342 const struct file_operations f2fs_file_operations = {
3343 	.llseek		= f2fs_llseek,
3344 	.read_iter	= generic_file_read_iter,
3345 	.write_iter	= f2fs_file_write_iter,
3346 	.open		= f2fs_file_open,
3347 	.release	= f2fs_release_file,
3348 	.mmap		= f2fs_file_mmap,
3349 	.flush		= f2fs_file_flush,
3350 	.fsync		= f2fs_sync_file,
3351 	.fallocate	= f2fs_fallocate,
3352 	.unlocked_ioctl	= f2fs_ioctl,
3353 #ifdef CONFIG_COMPAT
3354 	.compat_ioctl	= f2fs_compat_ioctl,
3355 #endif
3356 	.splice_read	= generic_file_splice_read,
3357 	.splice_write	= iter_file_splice_write,
3358 };
3359