xref: /openbmc/linux/fs/f2fs/file.c (revision d5097be55c21c103d2227591708425aab2e9682d)
1 /*
2  * fs/f2fs/file.c
3  *
4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5  *             http://www.samsung.com/
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  */
11 #include <linux/fs.h>
12 #include <linux/f2fs_fs.h>
13 #include <linux/stat.h>
14 #include <linux/buffer_head.h>
15 #include <linux/writeback.h>
16 #include <linux/blkdev.h>
17 #include <linux/falloc.h>
18 #include <linux/types.h>
19 #include <linux/compat.h>
20 #include <linux/uaccess.h>
21 #include <linux/mount.h>
22 #include <linux/pagevec.h>
23 #include <linux/uio.h>
24 #include <linux/uuid.h>
25 #include <linux/file.h>
26 
27 #include "f2fs.h"
28 #include "node.h"
29 #include "segment.h"
30 #include "xattr.h"
31 #include "acl.h"
32 #include "gc.h"
33 #include "trace.h"
34 #include <trace/events/f2fs.h>
35 
36 static int f2fs_filemap_fault(struct vm_fault *vmf)
37 {
38 	struct inode *inode = file_inode(vmf->vma->vm_file);
39 	int err;
40 
41 	down_read(&F2FS_I(inode)->i_mmap_sem);
42 	err = filemap_fault(vmf);
43 	up_read(&F2FS_I(inode)->i_mmap_sem);
44 
45 	return err;
46 }
47 
48 static int f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 {
50 	struct page *page = vmf->page;
51 	struct inode *inode = file_inode(vmf->vma->vm_file);
52 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 	struct dnode_of_data dn;
54 	int err;
55 
56 	if (unlikely(f2fs_cp_error(sbi))) {
57 		err = -EIO;
58 		goto err;
59 	}
60 
61 	sb_start_pagefault(inode->i_sb);
62 
63 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
64 
65 	/* block allocation */
66 	f2fs_lock_op(sbi);
67 	set_new_dnode(&dn, inode, NULL, NULL, 0);
68 	err = f2fs_reserve_block(&dn, page->index);
69 	if (err) {
70 		f2fs_unlock_op(sbi);
71 		goto out;
72 	}
73 	f2fs_put_dnode(&dn);
74 	f2fs_unlock_op(sbi);
75 
76 	f2fs_balance_fs(sbi, dn.node_changed);
77 
78 	file_update_time(vmf->vma->vm_file);
79 	down_read(&F2FS_I(inode)->i_mmap_sem);
80 	lock_page(page);
81 	if (unlikely(page->mapping != inode->i_mapping ||
82 			page_offset(page) > i_size_read(inode) ||
83 			!PageUptodate(page))) {
84 		unlock_page(page);
85 		err = -EFAULT;
86 		goto out_sem;
87 	}
88 
89 	/*
90 	 * check to see if the page is mapped already (no holes)
91 	 */
92 	if (PageMappedToDisk(page))
93 		goto mapped;
94 
95 	/* page is wholly or partially inside EOF */
96 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
97 						i_size_read(inode)) {
98 		unsigned offset;
99 		offset = i_size_read(inode) & ~PAGE_MASK;
100 		zero_user_segment(page, offset, PAGE_SIZE);
101 	}
102 	set_page_dirty(page);
103 	if (!PageUptodate(page))
104 		SetPageUptodate(page);
105 
106 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
107 
108 	trace_f2fs_vm_page_mkwrite(page, DATA);
109 mapped:
110 	/* fill the page */
111 	f2fs_wait_on_page_writeback(page, DATA, false);
112 
113 	/* wait for GCed encrypted page writeback */
114 	if (f2fs_encrypted_file(inode))
115 		f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
116 
117 out_sem:
118 	up_read(&F2FS_I(inode)->i_mmap_sem);
119 out:
120 	sb_end_pagefault(inode->i_sb);
121 	f2fs_update_time(sbi, REQ_TIME);
122 err:
123 	return block_page_mkwrite_return(err);
124 }
125 
126 static const struct vm_operations_struct f2fs_file_vm_ops = {
127 	.fault		= f2fs_filemap_fault,
128 	.map_pages	= filemap_map_pages,
129 	.page_mkwrite	= f2fs_vm_page_mkwrite,
130 };
131 
132 static int get_parent_ino(struct inode *inode, nid_t *pino)
133 {
134 	struct dentry *dentry;
135 
136 	inode = igrab(inode);
137 	dentry = d_find_any_alias(inode);
138 	iput(inode);
139 	if (!dentry)
140 		return 0;
141 
142 	*pino = parent_ino(dentry);
143 	dput(dentry);
144 	return 1;
145 }
146 
147 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
148 {
149 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
150 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
151 
152 	if (!S_ISREG(inode->i_mode))
153 		cp_reason = CP_NON_REGULAR;
154 	else if (inode->i_nlink != 1)
155 		cp_reason = CP_HARDLINK;
156 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
157 		cp_reason = CP_SB_NEED_CP;
158 	else if (file_wrong_pino(inode))
159 		cp_reason = CP_WRONG_PINO;
160 	else if (!space_for_roll_forward(sbi))
161 		cp_reason = CP_NO_SPC_ROLL;
162 	else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
163 		cp_reason = CP_NODE_NEED_CP;
164 	else if (test_opt(sbi, FASTBOOT))
165 		cp_reason = CP_FASTBOOT_MODE;
166 	else if (sbi->active_logs == 2)
167 		cp_reason = CP_SPEC_LOG_NUM;
168 
169 	return cp_reason;
170 }
171 
172 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
173 {
174 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
175 	bool ret = false;
176 	/* But we need to avoid that there are some inode updates */
177 	if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
178 		ret = true;
179 	f2fs_put_page(i, 0);
180 	return ret;
181 }
182 
183 static void try_to_fix_pino(struct inode *inode)
184 {
185 	struct f2fs_inode_info *fi = F2FS_I(inode);
186 	nid_t pino;
187 
188 	down_write(&fi->i_sem);
189 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
190 			get_parent_ino(inode, &pino)) {
191 		f2fs_i_pino_write(inode, pino);
192 		file_got_pino(inode);
193 	}
194 	up_write(&fi->i_sem);
195 }
196 
197 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
198 						int datasync, bool atomic)
199 {
200 	struct inode *inode = file->f_mapping->host;
201 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
202 	nid_t ino = inode->i_ino;
203 	int ret = 0;
204 	enum cp_reason_type cp_reason = 0;
205 	struct writeback_control wbc = {
206 		.sync_mode = WB_SYNC_ALL,
207 		.nr_to_write = LONG_MAX,
208 		.for_reclaim = 0,
209 	};
210 
211 	if (unlikely(f2fs_readonly(inode->i_sb)))
212 		return 0;
213 
214 	trace_f2fs_sync_file_enter(inode);
215 
216 	/* if fdatasync is triggered, let's do in-place-update */
217 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
218 		set_inode_flag(inode, FI_NEED_IPU);
219 	ret = file_write_and_wait_range(file, start, end);
220 	clear_inode_flag(inode, FI_NEED_IPU);
221 
222 	if (ret) {
223 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
224 		return ret;
225 	}
226 
227 	/* if the inode is dirty, let's recover all the time */
228 	if (!f2fs_skip_inode_update(inode, datasync)) {
229 		f2fs_write_inode(inode, NULL);
230 		goto go_write;
231 	}
232 
233 	/*
234 	 * if there is no written data, don't waste time to write recovery info.
235 	 */
236 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
237 			!exist_written_data(sbi, ino, APPEND_INO)) {
238 
239 		/* it may call write_inode just prior to fsync */
240 		if (need_inode_page_update(sbi, ino))
241 			goto go_write;
242 
243 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
244 				exist_written_data(sbi, ino, UPDATE_INO))
245 			goto flush_out;
246 		goto out;
247 	}
248 go_write:
249 	/*
250 	 * Both of fdatasync() and fsync() are able to be recovered from
251 	 * sudden-power-off.
252 	 */
253 	down_read(&F2FS_I(inode)->i_sem);
254 	cp_reason = need_do_checkpoint(inode);
255 	up_read(&F2FS_I(inode)->i_sem);
256 
257 	if (cp_reason) {
258 		/* all the dirty node pages should be flushed for POR */
259 		ret = f2fs_sync_fs(inode->i_sb, 1);
260 
261 		/*
262 		 * We've secured consistency through sync_fs. Following pino
263 		 * will be used only for fsynced inodes after checkpoint.
264 		 */
265 		try_to_fix_pino(inode);
266 		clear_inode_flag(inode, FI_APPEND_WRITE);
267 		clear_inode_flag(inode, FI_UPDATE_WRITE);
268 		goto out;
269 	}
270 sync_nodes:
271 	ret = fsync_node_pages(sbi, inode, &wbc, atomic);
272 	if (ret)
273 		goto out;
274 
275 	/* if cp_error was enabled, we should avoid infinite loop */
276 	if (unlikely(f2fs_cp_error(sbi))) {
277 		ret = -EIO;
278 		goto out;
279 	}
280 
281 	if (need_inode_block_update(sbi, ino)) {
282 		f2fs_mark_inode_dirty_sync(inode, true);
283 		f2fs_write_inode(inode, NULL);
284 		goto sync_nodes;
285 	}
286 
287 	/*
288 	 * If it's atomic_write, it's just fine to keep write ordering. So
289 	 * here we don't need to wait for node write completion, since we use
290 	 * node chain which serializes node blocks. If one of node writes are
291 	 * reordered, we can see simply broken chain, resulting in stopping
292 	 * roll-forward recovery. It means we'll recover all or none node blocks
293 	 * given fsync mark.
294 	 */
295 	if (!atomic) {
296 		ret = wait_on_node_pages_writeback(sbi, ino);
297 		if (ret)
298 			goto out;
299 	}
300 
301 	/* once recovery info is written, don't need to tack this */
302 	remove_ino_entry(sbi, ino, APPEND_INO);
303 	clear_inode_flag(inode, FI_APPEND_WRITE);
304 flush_out:
305 	if (!atomic)
306 		ret = f2fs_issue_flush(sbi, inode->i_ino);
307 	if (!ret) {
308 		remove_ino_entry(sbi, ino, UPDATE_INO);
309 		clear_inode_flag(inode, FI_UPDATE_WRITE);
310 		remove_ino_entry(sbi, ino, FLUSH_INO);
311 	}
312 	f2fs_update_time(sbi, REQ_TIME);
313 out:
314 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
315 	f2fs_trace_ios(NULL, 1);
316 	return ret;
317 }
318 
319 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
320 {
321 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
322 		return -EIO;
323 	return f2fs_do_sync_file(file, start, end, datasync, false);
324 }
325 
326 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
327 						pgoff_t pgofs, int whence)
328 {
329 	struct page *page;
330 	int nr_pages;
331 
332 	if (whence != SEEK_DATA)
333 		return 0;
334 
335 	/* find first dirty page index */
336 	nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
337 				      1, &page);
338 	if (!nr_pages)
339 		return ULONG_MAX;
340 	pgofs = page->index;
341 	put_page(page);
342 	return pgofs;
343 }
344 
345 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
346 							int whence)
347 {
348 	switch (whence) {
349 	case SEEK_DATA:
350 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
351 			(blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
352 			return true;
353 		break;
354 	case SEEK_HOLE:
355 		if (blkaddr == NULL_ADDR)
356 			return true;
357 		break;
358 	}
359 	return false;
360 }
361 
362 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
363 {
364 	struct inode *inode = file->f_mapping->host;
365 	loff_t maxbytes = inode->i_sb->s_maxbytes;
366 	struct dnode_of_data dn;
367 	pgoff_t pgofs, end_offset, dirty;
368 	loff_t data_ofs = offset;
369 	loff_t isize;
370 	int err = 0;
371 
372 	inode_lock(inode);
373 
374 	isize = i_size_read(inode);
375 	if (offset >= isize)
376 		goto fail;
377 
378 	/* handle inline data case */
379 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
380 		if (whence == SEEK_HOLE)
381 			data_ofs = isize;
382 		goto found;
383 	}
384 
385 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
386 
387 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
388 
389 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
390 		set_new_dnode(&dn, inode, NULL, NULL, 0);
391 		err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
392 		if (err && err != -ENOENT) {
393 			goto fail;
394 		} else if (err == -ENOENT) {
395 			/* direct node does not exists */
396 			if (whence == SEEK_DATA) {
397 				pgofs = get_next_page_offset(&dn, pgofs);
398 				continue;
399 			} else {
400 				goto found;
401 			}
402 		}
403 
404 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
405 
406 		/* find data/hole in dnode block */
407 		for (; dn.ofs_in_node < end_offset;
408 				dn.ofs_in_node++, pgofs++,
409 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
410 			block_t blkaddr;
411 			blkaddr = datablock_addr(dn.inode,
412 					dn.node_page, dn.ofs_in_node);
413 
414 			if (__found_offset(blkaddr, dirty, pgofs, whence)) {
415 				f2fs_put_dnode(&dn);
416 				goto found;
417 			}
418 		}
419 		f2fs_put_dnode(&dn);
420 	}
421 
422 	if (whence == SEEK_DATA)
423 		goto fail;
424 found:
425 	if (whence == SEEK_HOLE && data_ofs > isize)
426 		data_ofs = isize;
427 	inode_unlock(inode);
428 	return vfs_setpos(file, data_ofs, maxbytes);
429 fail:
430 	inode_unlock(inode);
431 	return -ENXIO;
432 }
433 
434 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
435 {
436 	struct inode *inode = file->f_mapping->host;
437 	loff_t maxbytes = inode->i_sb->s_maxbytes;
438 
439 	switch (whence) {
440 	case SEEK_SET:
441 	case SEEK_CUR:
442 	case SEEK_END:
443 		return generic_file_llseek_size(file, offset, whence,
444 						maxbytes, i_size_read(inode));
445 	case SEEK_DATA:
446 	case SEEK_HOLE:
447 		if (offset < 0)
448 			return -ENXIO;
449 		return f2fs_seek_block(file, offset, whence);
450 	}
451 
452 	return -EINVAL;
453 }
454 
455 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
456 {
457 	struct inode *inode = file_inode(file);
458 	int err;
459 
460 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
461 		return -EIO;
462 
463 	/* we don't need to use inline_data strictly */
464 	err = f2fs_convert_inline_inode(inode);
465 	if (err)
466 		return err;
467 
468 	file_accessed(file);
469 	vma->vm_ops = &f2fs_file_vm_ops;
470 	return 0;
471 }
472 
473 static int f2fs_file_open(struct inode *inode, struct file *filp)
474 {
475 	int err = fscrypt_file_open(inode, filp);
476 
477 	if (err)
478 		return err;
479 	return dquot_file_open(inode, filp);
480 }
481 
482 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
483 {
484 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
485 	struct f2fs_node *raw_node;
486 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
487 	__le32 *addr;
488 	int base = 0;
489 
490 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
491 		base = get_extra_isize(dn->inode);
492 
493 	raw_node = F2FS_NODE(dn->node_page);
494 	addr = blkaddr_in_node(raw_node) + base + ofs;
495 
496 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
497 		block_t blkaddr = le32_to_cpu(*addr);
498 		if (blkaddr == NULL_ADDR)
499 			continue;
500 
501 		dn->data_blkaddr = NULL_ADDR;
502 		set_data_blkaddr(dn);
503 		invalidate_blocks(sbi, blkaddr);
504 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
505 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
506 		nr_free++;
507 	}
508 
509 	if (nr_free) {
510 		pgoff_t fofs;
511 		/*
512 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
513 		 * we will invalidate all blkaddr in the whole range.
514 		 */
515 		fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
516 							dn->inode) + ofs;
517 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
518 		dec_valid_block_count(sbi, dn->inode, nr_free);
519 	}
520 	dn->ofs_in_node = ofs;
521 
522 	f2fs_update_time(sbi, REQ_TIME);
523 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
524 					 dn->ofs_in_node, nr_free);
525 	return nr_free;
526 }
527 
528 void truncate_data_blocks(struct dnode_of_data *dn)
529 {
530 	truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
531 }
532 
533 static int truncate_partial_data_page(struct inode *inode, u64 from,
534 								bool cache_only)
535 {
536 	unsigned offset = from & (PAGE_SIZE - 1);
537 	pgoff_t index = from >> PAGE_SHIFT;
538 	struct address_space *mapping = inode->i_mapping;
539 	struct page *page;
540 
541 	if (!offset && !cache_only)
542 		return 0;
543 
544 	if (cache_only) {
545 		page = find_lock_page(mapping, index);
546 		if (page && PageUptodate(page))
547 			goto truncate_out;
548 		f2fs_put_page(page, 1);
549 		return 0;
550 	}
551 
552 	page = get_lock_data_page(inode, index, true);
553 	if (IS_ERR(page))
554 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
555 truncate_out:
556 	f2fs_wait_on_page_writeback(page, DATA, true);
557 	zero_user(page, offset, PAGE_SIZE - offset);
558 
559 	/* An encrypted inode should have a key and truncate the last page. */
560 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
561 	if (!cache_only)
562 		set_page_dirty(page);
563 	f2fs_put_page(page, 1);
564 	return 0;
565 }
566 
567 int truncate_blocks(struct inode *inode, u64 from, bool lock)
568 {
569 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
570 	unsigned int blocksize = inode->i_sb->s_blocksize;
571 	struct dnode_of_data dn;
572 	pgoff_t free_from;
573 	int count = 0, err = 0;
574 	struct page *ipage;
575 	bool truncate_page = false;
576 
577 	trace_f2fs_truncate_blocks_enter(inode, from);
578 
579 	free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
580 
581 	if (free_from >= sbi->max_file_blocks)
582 		goto free_partial;
583 
584 	if (lock)
585 		f2fs_lock_op(sbi);
586 
587 	ipage = get_node_page(sbi, inode->i_ino);
588 	if (IS_ERR(ipage)) {
589 		err = PTR_ERR(ipage);
590 		goto out;
591 	}
592 
593 	if (f2fs_has_inline_data(inode)) {
594 		truncate_inline_inode(inode, ipage, from);
595 		f2fs_put_page(ipage, 1);
596 		truncate_page = true;
597 		goto out;
598 	}
599 
600 	set_new_dnode(&dn, inode, ipage, NULL, 0);
601 	err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
602 	if (err) {
603 		if (err == -ENOENT)
604 			goto free_next;
605 		goto out;
606 	}
607 
608 	count = ADDRS_PER_PAGE(dn.node_page, inode);
609 
610 	count -= dn.ofs_in_node;
611 	f2fs_bug_on(sbi, count < 0);
612 
613 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
614 		truncate_data_blocks_range(&dn, count);
615 		free_from += count;
616 	}
617 
618 	f2fs_put_dnode(&dn);
619 free_next:
620 	err = truncate_inode_blocks(inode, free_from);
621 out:
622 	if (lock)
623 		f2fs_unlock_op(sbi);
624 free_partial:
625 	/* lastly zero out the first data page */
626 	if (!err)
627 		err = truncate_partial_data_page(inode, from, truncate_page);
628 
629 	trace_f2fs_truncate_blocks_exit(inode, err);
630 	return err;
631 }
632 
633 int f2fs_truncate(struct inode *inode)
634 {
635 	int err;
636 
637 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
638 		return -EIO;
639 
640 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
641 				S_ISLNK(inode->i_mode)))
642 		return 0;
643 
644 	trace_f2fs_truncate(inode);
645 
646 #ifdef CONFIG_F2FS_FAULT_INJECTION
647 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
648 		f2fs_show_injection_info(FAULT_TRUNCATE);
649 		return -EIO;
650 	}
651 #endif
652 	/* we should check inline_data size */
653 	if (!f2fs_may_inline_data(inode)) {
654 		err = f2fs_convert_inline_inode(inode);
655 		if (err)
656 			return err;
657 	}
658 
659 	err = truncate_blocks(inode, i_size_read(inode), true);
660 	if (err)
661 		return err;
662 
663 	inode->i_mtime = inode->i_ctime = current_time(inode);
664 	f2fs_mark_inode_dirty_sync(inode, false);
665 	return 0;
666 }
667 
668 int f2fs_getattr(const struct path *path, struct kstat *stat,
669 		 u32 request_mask, unsigned int query_flags)
670 {
671 	struct inode *inode = d_inode(path->dentry);
672 	struct f2fs_inode_info *fi = F2FS_I(inode);
673 	unsigned int flags;
674 
675 	flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
676 	if (flags & FS_APPEND_FL)
677 		stat->attributes |= STATX_ATTR_APPEND;
678 	if (flags & FS_COMPR_FL)
679 		stat->attributes |= STATX_ATTR_COMPRESSED;
680 	if (f2fs_encrypted_inode(inode))
681 		stat->attributes |= STATX_ATTR_ENCRYPTED;
682 	if (flags & FS_IMMUTABLE_FL)
683 		stat->attributes |= STATX_ATTR_IMMUTABLE;
684 	if (flags & FS_NODUMP_FL)
685 		stat->attributes |= STATX_ATTR_NODUMP;
686 
687 	stat->attributes_mask |= (STATX_ATTR_APPEND |
688 				  STATX_ATTR_COMPRESSED |
689 				  STATX_ATTR_ENCRYPTED |
690 				  STATX_ATTR_IMMUTABLE |
691 				  STATX_ATTR_NODUMP);
692 
693 	generic_fillattr(inode, stat);
694 
695 	/* we need to show initial sectors used for inline_data/dentries */
696 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
697 					f2fs_has_inline_dentry(inode))
698 		stat->blocks += (stat->size + 511) >> 9;
699 
700 	return 0;
701 }
702 
703 #ifdef CONFIG_F2FS_FS_POSIX_ACL
704 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
705 {
706 	unsigned int ia_valid = attr->ia_valid;
707 
708 	if (ia_valid & ATTR_UID)
709 		inode->i_uid = attr->ia_uid;
710 	if (ia_valid & ATTR_GID)
711 		inode->i_gid = attr->ia_gid;
712 	if (ia_valid & ATTR_ATIME)
713 		inode->i_atime = timespec_trunc(attr->ia_atime,
714 						inode->i_sb->s_time_gran);
715 	if (ia_valid & ATTR_MTIME)
716 		inode->i_mtime = timespec_trunc(attr->ia_mtime,
717 						inode->i_sb->s_time_gran);
718 	if (ia_valid & ATTR_CTIME)
719 		inode->i_ctime = timespec_trunc(attr->ia_ctime,
720 						inode->i_sb->s_time_gran);
721 	if (ia_valid & ATTR_MODE) {
722 		umode_t mode = attr->ia_mode;
723 
724 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
725 			mode &= ~S_ISGID;
726 		set_acl_inode(inode, mode);
727 	}
728 }
729 #else
730 #define __setattr_copy setattr_copy
731 #endif
732 
733 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
734 {
735 	struct inode *inode = d_inode(dentry);
736 	int err;
737 	bool size_changed = false;
738 
739 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
740 		return -EIO;
741 
742 	err = setattr_prepare(dentry, attr);
743 	if (err)
744 		return err;
745 
746 	err = fscrypt_prepare_setattr(dentry, attr);
747 	if (err)
748 		return err;
749 
750 	if (is_quota_modification(inode, attr)) {
751 		err = dquot_initialize(inode);
752 		if (err)
753 			return err;
754 	}
755 	if ((attr->ia_valid & ATTR_UID &&
756 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
757 		(attr->ia_valid & ATTR_GID &&
758 		!gid_eq(attr->ia_gid, inode->i_gid))) {
759 		err = dquot_transfer(inode, attr);
760 		if (err)
761 			return err;
762 	}
763 
764 	if (attr->ia_valid & ATTR_SIZE) {
765 		if (attr->ia_size <= i_size_read(inode)) {
766 			down_write(&F2FS_I(inode)->i_mmap_sem);
767 			truncate_setsize(inode, attr->ia_size);
768 			err = f2fs_truncate(inode);
769 			up_write(&F2FS_I(inode)->i_mmap_sem);
770 			if (err)
771 				return err;
772 		} else {
773 			/*
774 			 * do not trim all blocks after i_size if target size is
775 			 * larger than i_size.
776 			 */
777 			down_write(&F2FS_I(inode)->i_mmap_sem);
778 			truncate_setsize(inode, attr->ia_size);
779 			up_write(&F2FS_I(inode)->i_mmap_sem);
780 
781 			/* should convert inline inode here */
782 			if (!f2fs_may_inline_data(inode)) {
783 				err = f2fs_convert_inline_inode(inode);
784 				if (err)
785 					return err;
786 			}
787 			inode->i_mtime = inode->i_ctime = current_time(inode);
788 		}
789 
790 		down_write(&F2FS_I(inode)->i_sem);
791 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
792 		up_write(&F2FS_I(inode)->i_sem);
793 
794 		size_changed = true;
795 	}
796 
797 	__setattr_copy(inode, attr);
798 
799 	if (attr->ia_valid & ATTR_MODE) {
800 		err = posix_acl_chmod(inode, get_inode_mode(inode));
801 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
802 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
803 			clear_inode_flag(inode, FI_ACL_MODE);
804 		}
805 	}
806 
807 	/* file size may changed here */
808 	f2fs_mark_inode_dirty_sync(inode, size_changed);
809 
810 	/* inode change will produce dirty node pages flushed by checkpoint */
811 	f2fs_balance_fs(F2FS_I_SB(inode), true);
812 
813 	return err;
814 }
815 
816 const struct inode_operations f2fs_file_inode_operations = {
817 	.getattr	= f2fs_getattr,
818 	.setattr	= f2fs_setattr,
819 	.get_acl	= f2fs_get_acl,
820 	.set_acl	= f2fs_set_acl,
821 #ifdef CONFIG_F2FS_FS_XATTR
822 	.listxattr	= f2fs_listxattr,
823 #endif
824 	.fiemap		= f2fs_fiemap,
825 };
826 
827 static int fill_zero(struct inode *inode, pgoff_t index,
828 					loff_t start, loff_t len)
829 {
830 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
831 	struct page *page;
832 
833 	if (!len)
834 		return 0;
835 
836 	f2fs_balance_fs(sbi, true);
837 
838 	f2fs_lock_op(sbi);
839 	page = get_new_data_page(inode, NULL, index, false);
840 	f2fs_unlock_op(sbi);
841 
842 	if (IS_ERR(page))
843 		return PTR_ERR(page);
844 
845 	f2fs_wait_on_page_writeback(page, DATA, true);
846 	zero_user(page, start, len);
847 	set_page_dirty(page);
848 	f2fs_put_page(page, 1);
849 	return 0;
850 }
851 
852 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
853 {
854 	int err;
855 
856 	while (pg_start < pg_end) {
857 		struct dnode_of_data dn;
858 		pgoff_t end_offset, count;
859 
860 		set_new_dnode(&dn, inode, NULL, NULL, 0);
861 		err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
862 		if (err) {
863 			if (err == -ENOENT) {
864 				pg_start = get_next_page_offset(&dn, pg_start);
865 				continue;
866 			}
867 			return err;
868 		}
869 
870 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
871 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
872 
873 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
874 
875 		truncate_data_blocks_range(&dn, count);
876 		f2fs_put_dnode(&dn);
877 
878 		pg_start += count;
879 	}
880 	return 0;
881 }
882 
883 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
884 {
885 	pgoff_t pg_start, pg_end;
886 	loff_t off_start, off_end;
887 	int ret;
888 
889 	ret = f2fs_convert_inline_inode(inode);
890 	if (ret)
891 		return ret;
892 
893 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
894 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
895 
896 	off_start = offset & (PAGE_SIZE - 1);
897 	off_end = (offset + len) & (PAGE_SIZE - 1);
898 
899 	if (pg_start == pg_end) {
900 		ret = fill_zero(inode, pg_start, off_start,
901 						off_end - off_start);
902 		if (ret)
903 			return ret;
904 	} else {
905 		if (off_start) {
906 			ret = fill_zero(inode, pg_start++, off_start,
907 						PAGE_SIZE - off_start);
908 			if (ret)
909 				return ret;
910 		}
911 		if (off_end) {
912 			ret = fill_zero(inode, pg_end, 0, off_end);
913 			if (ret)
914 				return ret;
915 		}
916 
917 		if (pg_start < pg_end) {
918 			struct address_space *mapping = inode->i_mapping;
919 			loff_t blk_start, blk_end;
920 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
921 
922 			f2fs_balance_fs(sbi, true);
923 
924 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
925 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
926 			down_write(&F2FS_I(inode)->i_mmap_sem);
927 			truncate_inode_pages_range(mapping, blk_start,
928 					blk_end - 1);
929 
930 			f2fs_lock_op(sbi);
931 			ret = truncate_hole(inode, pg_start, pg_end);
932 			f2fs_unlock_op(sbi);
933 			up_write(&F2FS_I(inode)->i_mmap_sem);
934 		}
935 	}
936 
937 	return ret;
938 }
939 
940 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
941 				int *do_replace, pgoff_t off, pgoff_t len)
942 {
943 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
944 	struct dnode_of_data dn;
945 	int ret, done, i;
946 
947 next_dnode:
948 	set_new_dnode(&dn, inode, NULL, NULL, 0);
949 	ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
950 	if (ret && ret != -ENOENT) {
951 		return ret;
952 	} else if (ret == -ENOENT) {
953 		if (dn.max_level == 0)
954 			return -ENOENT;
955 		done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
956 		blkaddr += done;
957 		do_replace += done;
958 		goto next;
959 	}
960 
961 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
962 							dn.ofs_in_node, len);
963 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
964 		*blkaddr = datablock_addr(dn.inode,
965 					dn.node_page, dn.ofs_in_node);
966 		if (!is_checkpointed_data(sbi, *blkaddr)) {
967 
968 			if (test_opt(sbi, LFS)) {
969 				f2fs_put_dnode(&dn);
970 				return -ENOTSUPP;
971 			}
972 
973 			/* do not invalidate this block address */
974 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
975 			*do_replace = 1;
976 		}
977 	}
978 	f2fs_put_dnode(&dn);
979 next:
980 	len -= done;
981 	off += done;
982 	if (len)
983 		goto next_dnode;
984 	return 0;
985 }
986 
987 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
988 				int *do_replace, pgoff_t off, int len)
989 {
990 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
991 	struct dnode_of_data dn;
992 	int ret, i;
993 
994 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
995 		if (*do_replace == 0)
996 			continue;
997 
998 		set_new_dnode(&dn, inode, NULL, NULL, 0);
999 		ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1000 		if (ret) {
1001 			dec_valid_block_count(sbi, inode, 1);
1002 			invalidate_blocks(sbi, *blkaddr);
1003 		} else {
1004 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1005 		}
1006 		f2fs_put_dnode(&dn);
1007 	}
1008 	return 0;
1009 }
1010 
1011 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1012 			block_t *blkaddr, int *do_replace,
1013 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1014 {
1015 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1016 	pgoff_t i = 0;
1017 	int ret;
1018 
1019 	while (i < len) {
1020 		if (blkaddr[i] == NULL_ADDR && !full) {
1021 			i++;
1022 			continue;
1023 		}
1024 
1025 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1026 			struct dnode_of_data dn;
1027 			struct node_info ni;
1028 			size_t new_size;
1029 			pgoff_t ilen;
1030 
1031 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1032 			ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1033 			if (ret)
1034 				return ret;
1035 
1036 			get_node_info(sbi, dn.nid, &ni);
1037 			ilen = min((pgoff_t)
1038 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1039 						dn.ofs_in_node, len - i);
1040 			do {
1041 				dn.data_blkaddr = datablock_addr(dn.inode,
1042 						dn.node_page, dn.ofs_in_node);
1043 				truncate_data_blocks_range(&dn, 1);
1044 
1045 				if (do_replace[i]) {
1046 					f2fs_i_blocks_write(src_inode,
1047 							1, false, false);
1048 					f2fs_i_blocks_write(dst_inode,
1049 							1, true, false);
1050 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1051 					blkaddr[i], ni.version, true, false);
1052 
1053 					do_replace[i] = 0;
1054 				}
1055 				dn.ofs_in_node++;
1056 				i++;
1057 				new_size = (dst + i) << PAGE_SHIFT;
1058 				if (dst_inode->i_size < new_size)
1059 					f2fs_i_size_write(dst_inode, new_size);
1060 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1061 
1062 			f2fs_put_dnode(&dn);
1063 		} else {
1064 			struct page *psrc, *pdst;
1065 
1066 			psrc = get_lock_data_page(src_inode, src + i, true);
1067 			if (IS_ERR(psrc))
1068 				return PTR_ERR(psrc);
1069 			pdst = get_new_data_page(dst_inode, NULL, dst + i,
1070 								true);
1071 			if (IS_ERR(pdst)) {
1072 				f2fs_put_page(psrc, 1);
1073 				return PTR_ERR(pdst);
1074 			}
1075 			f2fs_copy_page(psrc, pdst);
1076 			set_page_dirty(pdst);
1077 			f2fs_put_page(pdst, 1);
1078 			f2fs_put_page(psrc, 1);
1079 
1080 			ret = truncate_hole(src_inode, src + i, src + i + 1);
1081 			if (ret)
1082 				return ret;
1083 			i++;
1084 		}
1085 	}
1086 	return 0;
1087 }
1088 
1089 static int __exchange_data_block(struct inode *src_inode,
1090 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1091 			pgoff_t len, bool full)
1092 {
1093 	block_t *src_blkaddr;
1094 	int *do_replace;
1095 	pgoff_t olen;
1096 	int ret;
1097 
1098 	while (len) {
1099 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1100 
1101 		src_blkaddr = kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1102 		if (!src_blkaddr)
1103 			return -ENOMEM;
1104 
1105 		do_replace = kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1106 		if (!do_replace) {
1107 			kvfree(src_blkaddr);
1108 			return -ENOMEM;
1109 		}
1110 
1111 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1112 					do_replace, src, olen);
1113 		if (ret)
1114 			goto roll_back;
1115 
1116 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1117 					do_replace, src, dst, olen, full);
1118 		if (ret)
1119 			goto roll_back;
1120 
1121 		src += olen;
1122 		dst += olen;
1123 		len -= olen;
1124 
1125 		kvfree(src_blkaddr);
1126 		kvfree(do_replace);
1127 	}
1128 	return 0;
1129 
1130 roll_back:
1131 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1132 	kvfree(src_blkaddr);
1133 	kvfree(do_replace);
1134 	return ret;
1135 }
1136 
1137 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1138 {
1139 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1140 	pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1141 	int ret;
1142 
1143 	f2fs_balance_fs(sbi, true);
1144 	f2fs_lock_op(sbi);
1145 
1146 	f2fs_drop_extent_tree(inode);
1147 
1148 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1149 	f2fs_unlock_op(sbi);
1150 	return ret;
1151 }
1152 
1153 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1154 {
1155 	pgoff_t pg_start, pg_end;
1156 	loff_t new_size;
1157 	int ret;
1158 
1159 	if (offset + len >= i_size_read(inode))
1160 		return -EINVAL;
1161 
1162 	/* collapse range should be aligned to block size of f2fs. */
1163 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1164 		return -EINVAL;
1165 
1166 	ret = f2fs_convert_inline_inode(inode);
1167 	if (ret)
1168 		return ret;
1169 
1170 	pg_start = offset >> PAGE_SHIFT;
1171 	pg_end = (offset + len) >> PAGE_SHIFT;
1172 
1173 	/* avoid gc operation during block exchange */
1174 	down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1175 
1176 	down_write(&F2FS_I(inode)->i_mmap_sem);
1177 	/* write out all dirty pages from offset */
1178 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1179 	if (ret)
1180 		goto out_unlock;
1181 
1182 	truncate_pagecache(inode, offset);
1183 
1184 	ret = f2fs_do_collapse(inode, pg_start, pg_end);
1185 	if (ret)
1186 		goto out_unlock;
1187 
1188 	/* write out all moved pages, if possible */
1189 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1190 	truncate_pagecache(inode, offset);
1191 
1192 	new_size = i_size_read(inode) - len;
1193 	truncate_pagecache(inode, new_size);
1194 
1195 	ret = truncate_blocks(inode, new_size, true);
1196 	if (!ret)
1197 		f2fs_i_size_write(inode, new_size);
1198 out_unlock:
1199 	up_write(&F2FS_I(inode)->i_mmap_sem);
1200 	up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1201 	return ret;
1202 }
1203 
1204 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1205 								pgoff_t end)
1206 {
1207 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1208 	pgoff_t index = start;
1209 	unsigned int ofs_in_node = dn->ofs_in_node;
1210 	blkcnt_t count = 0;
1211 	int ret;
1212 
1213 	for (; index < end; index++, dn->ofs_in_node++) {
1214 		if (datablock_addr(dn->inode, dn->node_page,
1215 					dn->ofs_in_node) == NULL_ADDR)
1216 			count++;
1217 	}
1218 
1219 	dn->ofs_in_node = ofs_in_node;
1220 	ret = reserve_new_blocks(dn, count);
1221 	if (ret)
1222 		return ret;
1223 
1224 	dn->ofs_in_node = ofs_in_node;
1225 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1226 		dn->data_blkaddr = datablock_addr(dn->inode,
1227 					dn->node_page, dn->ofs_in_node);
1228 		/*
1229 		 * reserve_new_blocks will not guarantee entire block
1230 		 * allocation.
1231 		 */
1232 		if (dn->data_blkaddr == NULL_ADDR) {
1233 			ret = -ENOSPC;
1234 			break;
1235 		}
1236 		if (dn->data_blkaddr != NEW_ADDR) {
1237 			invalidate_blocks(sbi, dn->data_blkaddr);
1238 			dn->data_blkaddr = NEW_ADDR;
1239 			set_data_blkaddr(dn);
1240 		}
1241 	}
1242 
1243 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1244 
1245 	return ret;
1246 }
1247 
1248 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1249 								int mode)
1250 {
1251 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1252 	struct address_space *mapping = inode->i_mapping;
1253 	pgoff_t index, pg_start, pg_end;
1254 	loff_t new_size = i_size_read(inode);
1255 	loff_t off_start, off_end;
1256 	int ret = 0;
1257 
1258 	ret = inode_newsize_ok(inode, (len + offset));
1259 	if (ret)
1260 		return ret;
1261 
1262 	ret = f2fs_convert_inline_inode(inode);
1263 	if (ret)
1264 		return ret;
1265 
1266 	down_write(&F2FS_I(inode)->i_mmap_sem);
1267 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1268 	if (ret)
1269 		goto out_sem;
1270 
1271 	truncate_pagecache_range(inode, offset, offset + len - 1);
1272 
1273 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1274 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1275 
1276 	off_start = offset & (PAGE_SIZE - 1);
1277 	off_end = (offset + len) & (PAGE_SIZE - 1);
1278 
1279 	if (pg_start == pg_end) {
1280 		ret = fill_zero(inode, pg_start, off_start,
1281 						off_end - off_start);
1282 		if (ret)
1283 			goto out_sem;
1284 
1285 		new_size = max_t(loff_t, new_size, offset + len);
1286 	} else {
1287 		if (off_start) {
1288 			ret = fill_zero(inode, pg_start++, off_start,
1289 						PAGE_SIZE - off_start);
1290 			if (ret)
1291 				goto out_sem;
1292 
1293 			new_size = max_t(loff_t, new_size,
1294 					(loff_t)pg_start << PAGE_SHIFT);
1295 		}
1296 
1297 		for (index = pg_start; index < pg_end;) {
1298 			struct dnode_of_data dn;
1299 			unsigned int end_offset;
1300 			pgoff_t end;
1301 
1302 			f2fs_lock_op(sbi);
1303 
1304 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1305 			ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1306 			if (ret) {
1307 				f2fs_unlock_op(sbi);
1308 				goto out;
1309 			}
1310 
1311 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1312 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1313 
1314 			ret = f2fs_do_zero_range(&dn, index, end);
1315 			f2fs_put_dnode(&dn);
1316 			f2fs_unlock_op(sbi);
1317 
1318 			f2fs_balance_fs(sbi, dn.node_changed);
1319 
1320 			if (ret)
1321 				goto out;
1322 
1323 			index = end;
1324 			new_size = max_t(loff_t, new_size,
1325 					(loff_t)index << PAGE_SHIFT);
1326 		}
1327 
1328 		if (off_end) {
1329 			ret = fill_zero(inode, pg_end, 0, off_end);
1330 			if (ret)
1331 				goto out;
1332 
1333 			new_size = max_t(loff_t, new_size, offset + len);
1334 		}
1335 	}
1336 
1337 out:
1338 	if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1339 		f2fs_i_size_write(inode, new_size);
1340 out_sem:
1341 	up_write(&F2FS_I(inode)->i_mmap_sem);
1342 
1343 	return ret;
1344 }
1345 
1346 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1347 {
1348 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1349 	pgoff_t nr, pg_start, pg_end, delta, idx;
1350 	loff_t new_size;
1351 	int ret = 0;
1352 
1353 	new_size = i_size_read(inode) + len;
1354 	ret = inode_newsize_ok(inode, new_size);
1355 	if (ret)
1356 		return ret;
1357 
1358 	if (offset >= i_size_read(inode))
1359 		return -EINVAL;
1360 
1361 	/* insert range should be aligned to block size of f2fs. */
1362 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1363 		return -EINVAL;
1364 
1365 	ret = f2fs_convert_inline_inode(inode);
1366 	if (ret)
1367 		return ret;
1368 
1369 	f2fs_balance_fs(sbi, true);
1370 
1371 	/* avoid gc operation during block exchange */
1372 	down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1373 
1374 	down_write(&F2FS_I(inode)->i_mmap_sem);
1375 	ret = truncate_blocks(inode, i_size_read(inode), true);
1376 	if (ret)
1377 		goto out;
1378 
1379 	/* write out all dirty pages from offset */
1380 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1381 	if (ret)
1382 		goto out;
1383 
1384 	truncate_pagecache(inode, offset);
1385 
1386 	pg_start = offset >> PAGE_SHIFT;
1387 	pg_end = (offset + len) >> PAGE_SHIFT;
1388 	delta = pg_end - pg_start;
1389 	idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1390 
1391 	while (!ret && idx > pg_start) {
1392 		nr = idx - pg_start;
1393 		if (nr > delta)
1394 			nr = delta;
1395 		idx -= nr;
1396 
1397 		f2fs_lock_op(sbi);
1398 		f2fs_drop_extent_tree(inode);
1399 
1400 		ret = __exchange_data_block(inode, inode, idx,
1401 					idx + delta, nr, false);
1402 		f2fs_unlock_op(sbi);
1403 	}
1404 
1405 	/* write out all moved pages, if possible */
1406 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1407 	truncate_pagecache(inode, offset);
1408 
1409 	if (!ret)
1410 		f2fs_i_size_write(inode, new_size);
1411 out:
1412 	up_write(&F2FS_I(inode)->i_mmap_sem);
1413 	up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1414 	return ret;
1415 }
1416 
1417 static int expand_inode_data(struct inode *inode, loff_t offset,
1418 					loff_t len, int mode)
1419 {
1420 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1421 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1422 					.m_seg_type = NO_CHECK_TYPE };
1423 	pgoff_t pg_end;
1424 	loff_t new_size = i_size_read(inode);
1425 	loff_t off_end;
1426 	int err;
1427 
1428 	err = inode_newsize_ok(inode, (len + offset));
1429 	if (err)
1430 		return err;
1431 
1432 	err = f2fs_convert_inline_inode(inode);
1433 	if (err)
1434 		return err;
1435 
1436 	f2fs_balance_fs(sbi, true);
1437 
1438 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1439 	off_end = (offset + len) & (PAGE_SIZE - 1);
1440 
1441 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1442 	map.m_len = pg_end - map.m_lblk;
1443 	if (off_end)
1444 		map.m_len++;
1445 
1446 	err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1447 	if (err) {
1448 		pgoff_t last_off;
1449 
1450 		if (!map.m_len)
1451 			return err;
1452 
1453 		last_off = map.m_lblk + map.m_len - 1;
1454 
1455 		/* update new size to the failed position */
1456 		new_size = (last_off == pg_end) ? offset + len:
1457 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1458 	} else {
1459 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1460 	}
1461 
1462 	if (new_size > i_size_read(inode)) {
1463 		if (mode & FALLOC_FL_KEEP_SIZE)
1464 			file_set_keep_isize(inode);
1465 		else
1466 			f2fs_i_size_write(inode, new_size);
1467 	}
1468 
1469 	return err;
1470 }
1471 
1472 static long f2fs_fallocate(struct file *file, int mode,
1473 				loff_t offset, loff_t len)
1474 {
1475 	struct inode *inode = file_inode(file);
1476 	long ret = 0;
1477 
1478 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1479 		return -EIO;
1480 
1481 	/* f2fs only support ->fallocate for regular file */
1482 	if (!S_ISREG(inode->i_mode))
1483 		return -EINVAL;
1484 
1485 	if (f2fs_encrypted_inode(inode) &&
1486 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1487 		return -EOPNOTSUPP;
1488 
1489 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1490 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1491 			FALLOC_FL_INSERT_RANGE))
1492 		return -EOPNOTSUPP;
1493 
1494 	inode_lock(inode);
1495 
1496 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1497 		if (offset >= inode->i_size)
1498 			goto out;
1499 
1500 		ret = punch_hole(inode, offset, len);
1501 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1502 		ret = f2fs_collapse_range(inode, offset, len);
1503 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1504 		ret = f2fs_zero_range(inode, offset, len, mode);
1505 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1506 		ret = f2fs_insert_range(inode, offset, len);
1507 	} else {
1508 		ret = expand_inode_data(inode, offset, len, mode);
1509 	}
1510 
1511 	if (!ret) {
1512 		inode->i_mtime = inode->i_ctime = current_time(inode);
1513 		f2fs_mark_inode_dirty_sync(inode, false);
1514 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1515 	}
1516 
1517 out:
1518 	inode_unlock(inode);
1519 
1520 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1521 	return ret;
1522 }
1523 
1524 static int f2fs_release_file(struct inode *inode, struct file *filp)
1525 {
1526 	/*
1527 	 * f2fs_relase_file is called at every close calls. So we should
1528 	 * not drop any inmemory pages by close called by other process.
1529 	 */
1530 	if (!(filp->f_mode & FMODE_WRITE) ||
1531 			atomic_read(&inode->i_writecount) != 1)
1532 		return 0;
1533 
1534 	/* some remained atomic pages should discarded */
1535 	if (f2fs_is_atomic_file(inode))
1536 		drop_inmem_pages(inode);
1537 	if (f2fs_is_volatile_file(inode)) {
1538 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1539 		stat_dec_volatile_write(inode);
1540 		set_inode_flag(inode, FI_DROP_CACHE);
1541 		filemap_fdatawrite(inode->i_mapping);
1542 		clear_inode_flag(inode, FI_DROP_CACHE);
1543 	}
1544 	return 0;
1545 }
1546 
1547 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1548 {
1549 	struct inode *inode = file_inode(file);
1550 
1551 	/*
1552 	 * If the process doing a transaction is crashed, we should do
1553 	 * roll-back. Otherwise, other reader/write can see corrupted database
1554 	 * until all the writers close its file. Since this should be done
1555 	 * before dropping file lock, it needs to do in ->flush.
1556 	 */
1557 	if (f2fs_is_atomic_file(inode) &&
1558 			F2FS_I(inode)->inmem_task == current)
1559 		drop_inmem_pages(inode);
1560 	return 0;
1561 }
1562 
1563 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1564 {
1565 	struct inode *inode = file_inode(filp);
1566 	struct f2fs_inode_info *fi = F2FS_I(inode);
1567 	unsigned int flags = fi->i_flags &
1568 			(FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1569 	return put_user(flags, (int __user *)arg);
1570 }
1571 
1572 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1573 {
1574 	struct f2fs_inode_info *fi = F2FS_I(inode);
1575 	unsigned int oldflags;
1576 
1577 	/* Is it quota file? Do not allow user to mess with it */
1578 	if (IS_NOQUOTA(inode))
1579 		return -EPERM;
1580 
1581 	flags = f2fs_mask_flags(inode->i_mode, flags);
1582 
1583 	oldflags = fi->i_flags;
1584 
1585 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1586 		if (!capable(CAP_LINUX_IMMUTABLE))
1587 			return -EPERM;
1588 
1589 	flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1590 	flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1591 	fi->i_flags = flags;
1592 
1593 	if (fi->i_flags & FS_PROJINHERIT_FL)
1594 		set_inode_flag(inode, FI_PROJ_INHERIT);
1595 	else
1596 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1597 
1598 	inode->i_ctime = current_time(inode);
1599 	f2fs_set_inode_flags(inode);
1600 	f2fs_mark_inode_dirty_sync(inode, false);
1601 	return 0;
1602 }
1603 
1604 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1605 {
1606 	struct inode *inode = file_inode(filp);
1607 	unsigned int flags;
1608 	int ret;
1609 
1610 	if (!inode_owner_or_capable(inode))
1611 		return -EACCES;
1612 
1613 	if (get_user(flags, (int __user *)arg))
1614 		return -EFAULT;
1615 
1616 	ret = mnt_want_write_file(filp);
1617 	if (ret)
1618 		return ret;
1619 
1620 	inode_lock(inode);
1621 
1622 	ret = __f2fs_ioc_setflags(inode, flags);
1623 
1624 	inode_unlock(inode);
1625 	mnt_drop_write_file(filp);
1626 	return ret;
1627 }
1628 
1629 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1630 {
1631 	struct inode *inode = file_inode(filp);
1632 
1633 	return put_user(inode->i_generation, (int __user *)arg);
1634 }
1635 
1636 static int f2fs_ioc_start_atomic_write(struct file *filp)
1637 {
1638 	struct inode *inode = file_inode(filp);
1639 	int ret;
1640 
1641 	if (!inode_owner_or_capable(inode))
1642 		return -EACCES;
1643 
1644 	if (!S_ISREG(inode->i_mode))
1645 		return -EINVAL;
1646 
1647 	ret = mnt_want_write_file(filp);
1648 	if (ret)
1649 		return ret;
1650 
1651 	inode_lock(inode);
1652 
1653 	if (f2fs_is_atomic_file(inode))
1654 		goto out;
1655 
1656 	ret = f2fs_convert_inline_inode(inode);
1657 	if (ret)
1658 		goto out;
1659 
1660 	set_inode_flag(inode, FI_ATOMIC_FILE);
1661 	set_inode_flag(inode, FI_HOT_DATA);
1662 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1663 
1664 	if (!get_dirty_pages(inode))
1665 		goto inc_stat;
1666 
1667 	f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1668 		"Unexpected flush for atomic writes: ino=%lu, npages=%u",
1669 					inode->i_ino, get_dirty_pages(inode));
1670 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1671 	if (ret) {
1672 		clear_inode_flag(inode, FI_ATOMIC_FILE);
1673 		clear_inode_flag(inode, FI_HOT_DATA);
1674 		goto out;
1675 	}
1676 
1677 inc_stat:
1678 	F2FS_I(inode)->inmem_task = current;
1679 	stat_inc_atomic_write(inode);
1680 	stat_update_max_atomic_write(inode);
1681 out:
1682 	inode_unlock(inode);
1683 	mnt_drop_write_file(filp);
1684 	return ret;
1685 }
1686 
1687 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1688 {
1689 	struct inode *inode = file_inode(filp);
1690 	int ret;
1691 
1692 	if (!inode_owner_or_capable(inode))
1693 		return -EACCES;
1694 
1695 	ret = mnt_want_write_file(filp);
1696 	if (ret)
1697 		return ret;
1698 
1699 	inode_lock(inode);
1700 
1701 	if (f2fs_is_volatile_file(inode))
1702 		goto err_out;
1703 
1704 	if (f2fs_is_atomic_file(inode)) {
1705 		ret = commit_inmem_pages(inode);
1706 		if (ret)
1707 			goto err_out;
1708 
1709 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1710 		if (!ret) {
1711 			clear_inode_flag(inode, FI_ATOMIC_FILE);
1712 			clear_inode_flag(inode, FI_HOT_DATA);
1713 			stat_dec_atomic_write(inode);
1714 		}
1715 	} else {
1716 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1717 	}
1718 err_out:
1719 	inode_unlock(inode);
1720 	mnt_drop_write_file(filp);
1721 	return ret;
1722 }
1723 
1724 static int f2fs_ioc_start_volatile_write(struct file *filp)
1725 {
1726 	struct inode *inode = file_inode(filp);
1727 	int ret;
1728 
1729 	if (!inode_owner_or_capable(inode))
1730 		return -EACCES;
1731 
1732 	if (!S_ISREG(inode->i_mode))
1733 		return -EINVAL;
1734 
1735 	ret = mnt_want_write_file(filp);
1736 	if (ret)
1737 		return ret;
1738 
1739 	inode_lock(inode);
1740 
1741 	if (f2fs_is_volatile_file(inode))
1742 		goto out;
1743 
1744 	ret = f2fs_convert_inline_inode(inode);
1745 	if (ret)
1746 		goto out;
1747 
1748 	stat_inc_volatile_write(inode);
1749 	stat_update_max_volatile_write(inode);
1750 
1751 	set_inode_flag(inode, FI_VOLATILE_FILE);
1752 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1753 out:
1754 	inode_unlock(inode);
1755 	mnt_drop_write_file(filp);
1756 	return ret;
1757 }
1758 
1759 static int f2fs_ioc_release_volatile_write(struct file *filp)
1760 {
1761 	struct inode *inode = file_inode(filp);
1762 	int ret;
1763 
1764 	if (!inode_owner_or_capable(inode))
1765 		return -EACCES;
1766 
1767 	ret = mnt_want_write_file(filp);
1768 	if (ret)
1769 		return ret;
1770 
1771 	inode_lock(inode);
1772 
1773 	if (!f2fs_is_volatile_file(inode))
1774 		goto out;
1775 
1776 	if (!f2fs_is_first_block_written(inode)) {
1777 		ret = truncate_partial_data_page(inode, 0, true);
1778 		goto out;
1779 	}
1780 
1781 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1782 out:
1783 	inode_unlock(inode);
1784 	mnt_drop_write_file(filp);
1785 	return ret;
1786 }
1787 
1788 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1789 {
1790 	struct inode *inode = file_inode(filp);
1791 	int ret;
1792 
1793 	if (!inode_owner_or_capable(inode))
1794 		return -EACCES;
1795 
1796 	ret = mnt_want_write_file(filp);
1797 	if (ret)
1798 		return ret;
1799 
1800 	inode_lock(inode);
1801 
1802 	if (f2fs_is_atomic_file(inode))
1803 		drop_inmem_pages(inode);
1804 	if (f2fs_is_volatile_file(inode)) {
1805 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1806 		stat_dec_volatile_write(inode);
1807 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1808 	}
1809 
1810 	inode_unlock(inode);
1811 
1812 	mnt_drop_write_file(filp);
1813 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1814 	return ret;
1815 }
1816 
1817 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1818 {
1819 	struct inode *inode = file_inode(filp);
1820 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1821 	struct super_block *sb = sbi->sb;
1822 	__u32 in;
1823 	int ret;
1824 
1825 	if (!capable(CAP_SYS_ADMIN))
1826 		return -EPERM;
1827 
1828 	if (get_user(in, (__u32 __user *)arg))
1829 		return -EFAULT;
1830 
1831 	ret = mnt_want_write_file(filp);
1832 	if (ret)
1833 		return ret;
1834 
1835 	switch (in) {
1836 	case F2FS_GOING_DOWN_FULLSYNC:
1837 		sb = freeze_bdev(sb->s_bdev);
1838 		if (sb && !IS_ERR(sb)) {
1839 			f2fs_stop_checkpoint(sbi, false);
1840 			thaw_bdev(sb->s_bdev, sb);
1841 		}
1842 		break;
1843 	case F2FS_GOING_DOWN_METASYNC:
1844 		/* do checkpoint only */
1845 		f2fs_sync_fs(sb, 1);
1846 		f2fs_stop_checkpoint(sbi, false);
1847 		break;
1848 	case F2FS_GOING_DOWN_NOSYNC:
1849 		f2fs_stop_checkpoint(sbi, false);
1850 		break;
1851 	case F2FS_GOING_DOWN_METAFLUSH:
1852 		sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1853 		f2fs_stop_checkpoint(sbi, false);
1854 		break;
1855 	default:
1856 		ret = -EINVAL;
1857 		goto out;
1858 	}
1859 	f2fs_update_time(sbi, REQ_TIME);
1860 out:
1861 	mnt_drop_write_file(filp);
1862 	return ret;
1863 }
1864 
1865 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1866 {
1867 	struct inode *inode = file_inode(filp);
1868 	struct super_block *sb = inode->i_sb;
1869 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
1870 	struct fstrim_range range;
1871 	int ret;
1872 
1873 	if (!capable(CAP_SYS_ADMIN))
1874 		return -EPERM;
1875 
1876 	if (!blk_queue_discard(q))
1877 		return -EOPNOTSUPP;
1878 
1879 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1880 				sizeof(range)))
1881 		return -EFAULT;
1882 
1883 	ret = mnt_want_write_file(filp);
1884 	if (ret)
1885 		return ret;
1886 
1887 	range.minlen = max((unsigned int)range.minlen,
1888 				q->limits.discard_granularity);
1889 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1890 	mnt_drop_write_file(filp);
1891 	if (ret < 0)
1892 		return ret;
1893 
1894 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
1895 				sizeof(range)))
1896 		return -EFAULT;
1897 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1898 	return 0;
1899 }
1900 
1901 static bool uuid_is_nonzero(__u8 u[16])
1902 {
1903 	int i;
1904 
1905 	for (i = 0; i < 16; i++)
1906 		if (u[i])
1907 			return true;
1908 	return false;
1909 }
1910 
1911 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1912 {
1913 	struct inode *inode = file_inode(filp);
1914 
1915 	if (!f2fs_sb_has_crypto(inode->i_sb))
1916 		return -EOPNOTSUPP;
1917 
1918 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1919 
1920 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1921 }
1922 
1923 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1924 {
1925 	if (!f2fs_sb_has_crypto(file_inode(filp)->i_sb))
1926 		return -EOPNOTSUPP;
1927 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1928 }
1929 
1930 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1931 {
1932 	struct inode *inode = file_inode(filp);
1933 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1934 	int err;
1935 
1936 	if (!f2fs_sb_has_crypto(inode->i_sb))
1937 		return -EOPNOTSUPP;
1938 
1939 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1940 		goto got_it;
1941 
1942 	err = mnt_want_write_file(filp);
1943 	if (err)
1944 		return err;
1945 
1946 	/* update superblock with uuid */
1947 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1948 
1949 	err = f2fs_commit_super(sbi, false);
1950 	if (err) {
1951 		/* undo new data */
1952 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1953 		mnt_drop_write_file(filp);
1954 		return err;
1955 	}
1956 	mnt_drop_write_file(filp);
1957 got_it:
1958 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1959 									16))
1960 		return -EFAULT;
1961 	return 0;
1962 }
1963 
1964 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1965 {
1966 	struct inode *inode = file_inode(filp);
1967 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1968 	__u32 sync;
1969 	int ret;
1970 
1971 	if (!capable(CAP_SYS_ADMIN))
1972 		return -EPERM;
1973 
1974 	if (get_user(sync, (__u32 __user *)arg))
1975 		return -EFAULT;
1976 
1977 	if (f2fs_readonly(sbi->sb))
1978 		return -EROFS;
1979 
1980 	ret = mnt_want_write_file(filp);
1981 	if (ret)
1982 		return ret;
1983 
1984 	if (!sync) {
1985 		if (!mutex_trylock(&sbi->gc_mutex)) {
1986 			ret = -EBUSY;
1987 			goto out;
1988 		}
1989 	} else {
1990 		mutex_lock(&sbi->gc_mutex);
1991 	}
1992 
1993 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
1994 out:
1995 	mnt_drop_write_file(filp);
1996 	return ret;
1997 }
1998 
1999 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2000 {
2001 	struct inode *inode = file_inode(filp);
2002 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2003 	struct f2fs_gc_range range;
2004 	u64 end;
2005 	int ret;
2006 
2007 	if (!capable(CAP_SYS_ADMIN))
2008 		return -EPERM;
2009 
2010 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2011 							sizeof(range)))
2012 		return -EFAULT;
2013 
2014 	if (f2fs_readonly(sbi->sb))
2015 		return -EROFS;
2016 
2017 	ret = mnt_want_write_file(filp);
2018 	if (ret)
2019 		return ret;
2020 
2021 	end = range.start + range.len;
2022 	if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi))
2023 		return -EINVAL;
2024 do_more:
2025 	if (!range.sync) {
2026 		if (!mutex_trylock(&sbi->gc_mutex)) {
2027 			ret = -EBUSY;
2028 			goto out;
2029 		}
2030 	} else {
2031 		mutex_lock(&sbi->gc_mutex);
2032 	}
2033 
2034 	ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2035 	range.start += sbi->blocks_per_seg;
2036 	if (range.start <= end)
2037 		goto do_more;
2038 out:
2039 	mnt_drop_write_file(filp);
2040 	return ret;
2041 }
2042 
2043 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2044 {
2045 	struct inode *inode = file_inode(filp);
2046 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2047 	int ret;
2048 
2049 	if (!capable(CAP_SYS_ADMIN))
2050 		return -EPERM;
2051 
2052 	if (f2fs_readonly(sbi->sb))
2053 		return -EROFS;
2054 
2055 	ret = mnt_want_write_file(filp);
2056 	if (ret)
2057 		return ret;
2058 
2059 	ret = f2fs_sync_fs(sbi->sb, 1);
2060 
2061 	mnt_drop_write_file(filp);
2062 	return ret;
2063 }
2064 
2065 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2066 					struct file *filp,
2067 					struct f2fs_defragment *range)
2068 {
2069 	struct inode *inode = file_inode(filp);
2070 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
2071 					.m_seg_type = NO_CHECK_TYPE };
2072 	struct extent_info ei = {0,0,0};
2073 	pgoff_t pg_start, pg_end;
2074 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2075 	unsigned int total = 0, sec_num;
2076 	block_t blk_end = 0;
2077 	bool fragmented = false;
2078 	int err;
2079 
2080 	/* if in-place-update policy is enabled, don't waste time here */
2081 	if (need_inplace_update_policy(inode, NULL))
2082 		return -EINVAL;
2083 
2084 	pg_start = range->start >> PAGE_SHIFT;
2085 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2086 
2087 	f2fs_balance_fs(sbi, true);
2088 
2089 	inode_lock(inode);
2090 
2091 	/* writeback all dirty pages in the range */
2092 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2093 						range->start + range->len - 1);
2094 	if (err)
2095 		goto out;
2096 
2097 	/*
2098 	 * lookup mapping info in extent cache, skip defragmenting if physical
2099 	 * block addresses are continuous.
2100 	 */
2101 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2102 		if (ei.fofs + ei.len >= pg_end)
2103 			goto out;
2104 	}
2105 
2106 	map.m_lblk = pg_start;
2107 
2108 	/*
2109 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2110 	 * physical block addresses are continuous even if there are hole(s)
2111 	 * in logical blocks.
2112 	 */
2113 	while (map.m_lblk < pg_end) {
2114 		map.m_len = pg_end - map.m_lblk;
2115 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2116 		if (err)
2117 			goto out;
2118 
2119 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2120 			map.m_lblk++;
2121 			continue;
2122 		}
2123 
2124 		if (blk_end && blk_end != map.m_pblk) {
2125 			fragmented = true;
2126 			break;
2127 		}
2128 		blk_end = map.m_pblk + map.m_len;
2129 
2130 		map.m_lblk += map.m_len;
2131 	}
2132 
2133 	if (!fragmented)
2134 		goto out;
2135 
2136 	map.m_lblk = pg_start;
2137 	map.m_len = pg_end - pg_start;
2138 
2139 	sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2140 
2141 	/*
2142 	 * make sure there are enough free section for LFS allocation, this can
2143 	 * avoid defragment running in SSR mode when free section are allocated
2144 	 * intensively
2145 	 */
2146 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2147 		err = -EAGAIN;
2148 		goto out;
2149 	}
2150 
2151 	while (map.m_lblk < pg_end) {
2152 		pgoff_t idx;
2153 		int cnt = 0;
2154 
2155 do_map:
2156 		map.m_len = pg_end - map.m_lblk;
2157 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2158 		if (err)
2159 			goto clear_out;
2160 
2161 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2162 			map.m_lblk++;
2163 			continue;
2164 		}
2165 
2166 		set_inode_flag(inode, FI_DO_DEFRAG);
2167 
2168 		idx = map.m_lblk;
2169 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2170 			struct page *page;
2171 
2172 			page = get_lock_data_page(inode, idx, true);
2173 			if (IS_ERR(page)) {
2174 				err = PTR_ERR(page);
2175 				goto clear_out;
2176 			}
2177 
2178 			set_page_dirty(page);
2179 			f2fs_put_page(page, 1);
2180 
2181 			idx++;
2182 			cnt++;
2183 			total++;
2184 		}
2185 
2186 		map.m_lblk = idx;
2187 
2188 		if (idx < pg_end && cnt < blk_per_seg)
2189 			goto do_map;
2190 
2191 		clear_inode_flag(inode, FI_DO_DEFRAG);
2192 
2193 		err = filemap_fdatawrite(inode->i_mapping);
2194 		if (err)
2195 			goto out;
2196 	}
2197 clear_out:
2198 	clear_inode_flag(inode, FI_DO_DEFRAG);
2199 out:
2200 	inode_unlock(inode);
2201 	if (!err)
2202 		range->len = (u64)total << PAGE_SHIFT;
2203 	return err;
2204 }
2205 
2206 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2207 {
2208 	struct inode *inode = file_inode(filp);
2209 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2210 	struct f2fs_defragment range;
2211 	int err;
2212 
2213 	if (!capable(CAP_SYS_ADMIN))
2214 		return -EPERM;
2215 
2216 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2217 		return -EINVAL;
2218 
2219 	if (f2fs_readonly(sbi->sb))
2220 		return -EROFS;
2221 
2222 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2223 							sizeof(range)))
2224 		return -EFAULT;
2225 
2226 	/* verify alignment of offset & size */
2227 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2228 		return -EINVAL;
2229 
2230 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2231 					sbi->max_file_blocks))
2232 		return -EINVAL;
2233 
2234 	err = mnt_want_write_file(filp);
2235 	if (err)
2236 		return err;
2237 
2238 	err = f2fs_defragment_range(sbi, filp, &range);
2239 	mnt_drop_write_file(filp);
2240 
2241 	f2fs_update_time(sbi, REQ_TIME);
2242 	if (err < 0)
2243 		return err;
2244 
2245 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2246 							sizeof(range)))
2247 		return -EFAULT;
2248 
2249 	return 0;
2250 }
2251 
2252 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2253 			struct file *file_out, loff_t pos_out, size_t len)
2254 {
2255 	struct inode *src = file_inode(file_in);
2256 	struct inode *dst = file_inode(file_out);
2257 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2258 	size_t olen = len, dst_max_i_size = 0;
2259 	size_t dst_osize;
2260 	int ret;
2261 
2262 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2263 				src->i_sb != dst->i_sb)
2264 		return -EXDEV;
2265 
2266 	if (unlikely(f2fs_readonly(src->i_sb)))
2267 		return -EROFS;
2268 
2269 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2270 		return -EINVAL;
2271 
2272 	if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2273 		return -EOPNOTSUPP;
2274 
2275 	if (src == dst) {
2276 		if (pos_in == pos_out)
2277 			return 0;
2278 		if (pos_out > pos_in && pos_out < pos_in + len)
2279 			return -EINVAL;
2280 	}
2281 
2282 	inode_lock(src);
2283 	down_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2284 	if (src != dst) {
2285 		ret = -EBUSY;
2286 		if (!inode_trylock(dst))
2287 			goto out;
2288 		if (!down_write_trylock(&F2FS_I(dst)->dio_rwsem[WRITE])) {
2289 			inode_unlock(dst);
2290 			goto out;
2291 		}
2292 	}
2293 
2294 	ret = -EINVAL;
2295 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2296 		goto out_unlock;
2297 	if (len == 0)
2298 		olen = len = src->i_size - pos_in;
2299 	if (pos_in + len == src->i_size)
2300 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2301 	if (len == 0) {
2302 		ret = 0;
2303 		goto out_unlock;
2304 	}
2305 
2306 	dst_osize = dst->i_size;
2307 	if (pos_out + olen > dst->i_size)
2308 		dst_max_i_size = pos_out + olen;
2309 
2310 	/* verify the end result is block aligned */
2311 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2312 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2313 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2314 		goto out_unlock;
2315 
2316 	ret = f2fs_convert_inline_inode(src);
2317 	if (ret)
2318 		goto out_unlock;
2319 
2320 	ret = f2fs_convert_inline_inode(dst);
2321 	if (ret)
2322 		goto out_unlock;
2323 
2324 	/* write out all dirty pages from offset */
2325 	ret = filemap_write_and_wait_range(src->i_mapping,
2326 					pos_in, pos_in + len);
2327 	if (ret)
2328 		goto out_unlock;
2329 
2330 	ret = filemap_write_and_wait_range(dst->i_mapping,
2331 					pos_out, pos_out + len);
2332 	if (ret)
2333 		goto out_unlock;
2334 
2335 	f2fs_balance_fs(sbi, true);
2336 	f2fs_lock_op(sbi);
2337 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2338 				pos_out >> F2FS_BLKSIZE_BITS,
2339 				len >> F2FS_BLKSIZE_BITS, false);
2340 
2341 	if (!ret) {
2342 		if (dst_max_i_size)
2343 			f2fs_i_size_write(dst, dst_max_i_size);
2344 		else if (dst_osize != dst->i_size)
2345 			f2fs_i_size_write(dst, dst_osize);
2346 	}
2347 	f2fs_unlock_op(sbi);
2348 out_unlock:
2349 	if (src != dst) {
2350 		up_write(&F2FS_I(dst)->dio_rwsem[WRITE]);
2351 		inode_unlock(dst);
2352 	}
2353 out:
2354 	up_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2355 	inode_unlock(src);
2356 	return ret;
2357 }
2358 
2359 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2360 {
2361 	struct f2fs_move_range range;
2362 	struct fd dst;
2363 	int err;
2364 
2365 	if (!(filp->f_mode & FMODE_READ) ||
2366 			!(filp->f_mode & FMODE_WRITE))
2367 		return -EBADF;
2368 
2369 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2370 							sizeof(range)))
2371 		return -EFAULT;
2372 
2373 	dst = fdget(range.dst_fd);
2374 	if (!dst.file)
2375 		return -EBADF;
2376 
2377 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2378 		err = -EBADF;
2379 		goto err_out;
2380 	}
2381 
2382 	err = mnt_want_write_file(filp);
2383 	if (err)
2384 		goto err_out;
2385 
2386 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2387 					range.pos_out, range.len);
2388 
2389 	mnt_drop_write_file(filp);
2390 	if (err)
2391 		goto err_out;
2392 
2393 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2394 						&range, sizeof(range)))
2395 		err = -EFAULT;
2396 err_out:
2397 	fdput(dst);
2398 	return err;
2399 }
2400 
2401 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2402 {
2403 	struct inode *inode = file_inode(filp);
2404 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2405 	struct sit_info *sm = SIT_I(sbi);
2406 	unsigned int start_segno = 0, end_segno = 0;
2407 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2408 	struct f2fs_flush_device range;
2409 	int ret;
2410 
2411 	if (!capable(CAP_SYS_ADMIN))
2412 		return -EPERM;
2413 
2414 	if (f2fs_readonly(sbi->sb))
2415 		return -EROFS;
2416 
2417 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2418 							sizeof(range)))
2419 		return -EFAULT;
2420 
2421 	if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2422 			sbi->segs_per_sec != 1) {
2423 		f2fs_msg(sbi->sb, KERN_WARNING,
2424 			"Can't flush %u in %d for segs_per_sec %u != 1\n",
2425 				range.dev_num, sbi->s_ndevs,
2426 				sbi->segs_per_sec);
2427 		return -EINVAL;
2428 	}
2429 
2430 	ret = mnt_want_write_file(filp);
2431 	if (ret)
2432 		return ret;
2433 
2434 	if (range.dev_num != 0)
2435 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2436 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2437 
2438 	start_segno = sm->last_victim[FLUSH_DEVICE];
2439 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2440 		start_segno = dev_start_segno;
2441 	end_segno = min(start_segno + range.segments, dev_end_segno);
2442 
2443 	while (start_segno < end_segno) {
2444 		if (!mutex_trylock(&sbi->gc_mutex)) {
2445 			ret = -EBUSY;
2446 			goto out;
2447 		}
2448 		sm->last_victim[GC_CB] = end_segno + 1;
2449 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2450 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2451 		ret = f2fs_gc(sbi, true, true, start_segno);
2452 		if (ret == -EAGAIN)
2453 			ret = 0;
2454 		else if (ret < 0)
2455 			break;
2456 		start_segno++;
2457 	}
2458 out:
2459 	mnt_drop_write_file(filp);
2460 	return ret;
2461 }
2462 
2463 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2464 {
2465 	struct inode *inode = file_inode(filp);
2466 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2467 
2468 	/* Must validate to set it with SQLite behavior in Android. */
2469 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2470 
2471 	return put_user(sb_feature, (u32 __user *)arg);
2472 }
2473 
2474 #ifdef CONFIG_QUOTA
2475 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2476 {
2477 	struct inode *inode = file_inode(filp);
2478 	struct f2fs_inode_info *fi = F2FS_I(inode);
2479 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2480 	struct super_block *sb = sbi->sb;
2481 	struct dquot *transfer_to[MAXQUOTAS] = {};
2482 	struct page *ipage;
2483 	kprojid_t kprojid;
2484 	int err;
2485 
2486 	if (!f2fs_sb_has_project_quota(sb)) {
2487 		if (projid != F2FS_DEF_PROJID)
2488 			return -EOPNOTSUPP;
2489 		else
2490 			return 0;
2491 	}
2492 
2493 	if (!f2fs_has_extra_attr(inode))
2494 		return -EOPNOTSUPP;
2495 
2496 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2497 
2498 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2499 		return 0;
2500 
2501 	err = mnt_want_write_file(filp);
2502 	if (err)
2503 		return err;
2504 
2505 	err = -EPERM;
2506 	inode_lock(inode);
2507 
2508 	/* Is it quota file? Do not allow user to mess with it */
2509 	if (IS_NOQUOTA(inode))
2510 		goto out_unlock;
2511 
2512 	ipage = get_node_page(sbi, inode->i_ino);
2513 	if (IS_ERR(ipage)) {
2514 		err = PTR_ERR(ipage);
2515 		goto out_unlock;
2516 	}
2517 
2518 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2519 								i_projid)) {
2520 		err = -EOVERFLOW;
2521 		f2fs_put_page(ipage, 1);
2522 		goto out_unlock;
2523 	}
2524 	f2fs_put_page(ipage, 1);
2525 
2526 	dquot_initialize(inode);
2527 
2528 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2529 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2530 		err = __dquot_transfer(inode, transfer_to);
2531 		dqput(transfer_to[PRJQUOTA]);
2532 		if (err)
2533 			goto out_dirty;
2534 	}
2535 
2536 	F2FS_I(inode)->i_projid = kprojid;
2537 	inode->i_ctime = current_time(inode);
2538 out_dirty:
2539 	f2fs_mark_inode_dirty_sync(inode, true);
2540 out_unlock:
2541 	inode_unlock(inode);
2542 	mnt_drop_write_file(filp);
2543 	return err;
2544 }
2545 #else
2546 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2547 {
2548 	if (projid != F2FS_DEF_PROJID)
2549 		return -EOPNOTSUPP;
2550 	return 0;
2551 }
2552 #endif
2553 
2554 /* Transfer internal flags to xflags */
2555 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2556 {
2557 	__u32 xflags = 0;
2558 
2559 	if (iflags & FS_SYNC_FL)
2560 		xflags |= FS_XFLAG_SYNC;
2561 	if (iflags & FS_IMMUTABLE_FL)
2562 		xflags |= FS_XFLAG_IMMUTABLE;
2563 	if (iflags & FS_APPEND_FL)
2564 		xflags |= FS_XFLAG_APPEND;
2565 	if (iflags & FS_NODUMP_FL)
2566 		xflags |= FS_XFLAG_NODUMP;
2567 	if (iflags & FS_NOATIME_FL)
2568 		xflags |= FS_XFLAG_NOATIME;
2569 	if (iflags & FS_PROJINHERIT_FL)
2570 		xflags |= FS_XFLAG_PROJINHERIT;
2571 	return xflags;
2572 }
2573 
2574 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2575 				  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2576 				  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2577 
2578 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2579 #define F2FS_FL_XFLAG_VISIBLE		(FS_SYNC_FL | \
2580 					 FS_IMMUTABLE_FL | \
2581 					 FS_APPEND_FL | \
2582 					 FS_NODUMP_FL | \
2583 					 FS_NOATIME_FL | \
2584 					 FS_PROJINHERIT_FL)
2585 
2586 /* Transfer xflags flags to internal */
2587 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2588 {
2589 	unsigned long iflags = 0;
2590 
2591 	if (xflags & FS_XFLAG_SYNC)
2592 		iflags |= FS_SYNC_FL;
2593 	if (xflags & FS_XFLAG_IMMUTABLE)
2594 		iflags |= FS_IMMUTABLE_FL;
2595 	if (xflags & FS_XFLAG_APPEND)
2596 		iflags |= FS_APPEND_FL;
2597 	if (xflags & FS_XFLAG_NODUMP)
2598 		iflags |= FS_NODUMP_FL;
2599 	if (xflags & FS_XFLAG_NOATIME)
2600 		iflags |= FS_NOATIME_FL;
2601 	if (xflags & FS_XFLAG_PROJINHERIT)
2602 		iflags |= FS_PROJINHERIT_FL;
2603 
2604 	return iflags;
2605 }
2606 
2607 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2608 {
2609 	struct inode *inode = file_inode(filp);
2610 	struct f2fs_inode_info *fi = F2FS_I(inode);
2611 	struct fsxattr fa;
2612 
2613 	memset(&fa, 0, sizeof(struct fsxattr));
2614 	fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2615 				(FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2616 
2617 	if (f2fs_sb_has_project_quota(inode->i_sb))
2618 		fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2619 							fi->i_projid);
2620 
2621 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2622 		return -EFAULT;
2623 	return 0;
2624 }
2625 
2626 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2627 {
2628 	struct inode *inode = file_inode(filp);
2629 	struct f2fs_inode_info *fi = F2FS_I(inode);
2630 	struct fsxattr fa;
2631 	unsigned int flags;
2632 	int err;
2633 
2634 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2635 		return -EFAULT;
2636 
2637 	/* Make sure caller has proper permission */
2638 	if (!inode_owner_or_capable(inode))
2639 		return -EACCES;
2640 
2641 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2642 		return -EOPNOTSUPP;
2643 
2644 	flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2645 	if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2646 		return -EOPNOTSUPP;
2647 
2648 	err = mnt_want_write_file(filp);
2649 	if (err)
2650 		return err;
2651 
2652 	inode_lock(inode);
2653 	flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2654 				(flags & F2FS_FL_XFLAG_VISIBLE);
2655 	err = __f2fs_ioc_setflags(inode, flags);
2656 	inode_unlock(inode);
2657 	mnt_drop_write_file(filp);
2658 	if (err)
2659 		return err;
2660 
2661 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2662 	if (err)
2663 		return err;
2664 
2665 	return 0;
2666 }
2667 
2668 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2669 {
2670 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2671 		return -EIO;
2672 
2673 	switch (cmd) {
2674 	case F2FS_IOC_GETFLAGS:
2675 		return f2fs_ioc_getflags(filp, arg);
2676 	case F2FS_IOC_SETFLAGS:
2677 		return f2fs_ioc_setflags(filp, arg);
2678 	case F2FS_IOC_GETVERSION:
2679 		return f2fs_ioc_getversion(filp, arg);
2680 	case F2FS_IOC_START_ATOMIC_WRITE:
2681 		return f2fs_ioc_start_atomic_write(filp);
2682 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2683 		return f2fs_ioc_commit_atomic_write(filp);
2684 	case F2FS_IOC_START_VOLATILE_WRITE:
2685 		return f2fs_ioc_start_volatile_write(filp);
2686 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2687 		return f2fs_ioc_release_volatile_write(filp);
2688 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
2689 		return f2fs_ioc_abort_volatile_write(filp);
2690 	case F2FS_IOC_SHUTDOWN:
2691 		return f2fs_ioc_shutdown(filp, arg);
2692 	case FITRIM:
2693 		return f2fs_ioc_fitrim(filp, arg);
2694 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
2695 		return f2fs_ioc_set_encryption_policy(filp, arg);
2696 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
2697 		return f2fs_ioc_get_encryption_policy(filp, arg);
2698 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2699 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2700 	case F2FS_IOC_GARBAGE_COLLECT:
2701 		return f2fs_ioc_gc(filp, arg);
2702 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2703 		return f2fs_ioc_gc_range(filp, arg);
2704 	case F2FS_IOC_WRITE_CHECKPOINT:
2705 		return f2fs_ioc_write_checkpoint(filp, arg);
2706 	case F2FS_IOC_DEFRAGMENT:
2707 		return f2fs_ioc_defragment(filp, arg);
2708 	case F2FS_IOC_MOVE_RANGE:
2709 		return f2fs_ioc_move_range(filp, arg);
2710 	case F2FS_IOC_FLUSH_DEVICE:
2711 		return f2fs_ioc_flush_device(filp, arg);
2712 	case F2FS_IOC_GET_FEATURES:
2713 		return f2fs_ioc_get_features(filp, arg);
2714 	case F2FS_IOC_FSGETXATTR:
2715 		return f2fs_ioc_fsgetxattr(filp, arg);
2716 	case F2FS_IOC_FSSETXATTR:
2717 		return f2fs_ioc_fssetxattr(filp, arg);
2718 	default:
2719 		return -ENOTTY;
2720 	}
2721 }
2722 
2723 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2724 {
2725 	struct file *file = iocb->ki_filp;
2726 	struct inode *inode = file_inode(file);
2727 	struct blk_plug plug;
2728 	ssize_t ret;
2729 
2730 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2731 		return -EIO;
2732 
2733 	inode_lock(inode);
2734 	ret = generic_write_checks(iocb, from);
2735 	if (ret > 0) {
2736 		int err;
2737 
2738 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2739 			set_inode_flag(inode, FI_NO_PREALLOC);
2740 
2741 		err = f2fs_preallocate_blocks(iocb, from);
2742 		if (err) {
2743 			clear_inode_flag(inode, FI_NO_PREALLOC);
2744 			inode_unlock(inode);
2745 			return err;
2746 		}
2747 		blk_start_plug(&plug);
2748 		ret = __generic_file_write_iter(iocb, from);
2749 		blk_finish_plug(&plug);
2750 		clear_inode_flag(inode, FI_NO_PREALLOC);
2751 
2752 		if (ret > 0)
2753 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2754 	}
2755 	inode_unlock(inode);
2756 
2757 	if (ret > 0)
2758 		ret = generic_write_sync(iocb, ret);
2759 	return ret;
2760 }
2761 
2762 #ifdef CONFIG_COMPAT
2763 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2764 {
2765 	switch (cmd) {
2766 	case F2FS_IOC32_GETFLAGS:
2767 		cmd = F2FS_IOC_GETFLAGS;
2768 		break;
2769 	case F2FS_IOC32_SETFLAGS:
2770 		cmd = F2FS_IOC_SETFLAGS;
2771 		break;
2772 	case F2FS_IOC32_GETVERSION:
2773 		cmd = F2FS_IOC_GETVERSION;
2774 		break;
2775 	case F2FS_IOC_START_ATOMIC_WRITE:
2776 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2777 	case F2FS_IOC_START_VOLATILE_WRITE:
2778 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2779 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
2780 	case F2FS_IOC_SHUTDOWN:
2781 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
2782 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2783 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
2784 	case F2FS_IOC_GARBAGE_COLLECT:
2785 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2786 	case F2FS_IOC_WRITE_CHECKPOINT:
2787 	case F2FS_IOC_DEFRAGMENT:
2788 	case F2FS_IOC_MOVE_RANGE:
2789 	case F2FS_IOC_FLUSH_DEVICE:
2790 	case F2FS_IOC_GET_FEATURES:
2791 	case F2FS_IOC_FSGETXATTR:
2792 	case F2FS_IOC_FSSETXATTR:
2793 		break;
2794 	default:
2795 		return -ENOIOCTLCMD;
2796 	}
2797 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2798 }
2799 #endif
2800 
2801 const struct file_operations f2fs_file_operations = {
2802 	.llseek		= f2fs_llseek,
2803 	.read_iter	= generic_file_read_iter,
2804 	.write_iter	= f2fs_file_write_iter,
2805 	.open		= f2fs_file_open,
2806 	.release	= f2fs_release_file,
2807 	.mmap		= f2fs_file_mmap,
2808 	.flush		= f2fs_file_flush,
2809 	.fsync		= f2fs_sync_file,
2810 	.fallocate	= f2fs_fallocate,
2811 	.unlocked_ioctl	= f2fs_ioctl,
2812 #ifdef CONFIG_COMPAT
2813 	.compat_ioctl	= f2fs_compat_ioctl,
2814 #endif
2815 	.splice_read	= generic_file_splice_read,
2816 	.splice_write	= iter_file_splice_write,
2817 };
2818