xref: /linux/fs/f2fs/file.c (revision 109d59b900e78834c66657dd4748fcedb9a1fe8d)
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 	else if (need_dentry_mark(sbi, inode->i_ino) &&
169 		exist_written_data(sbi, F2FS_I(inode)->i_pino, 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)) || 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 
214 	if (unlikely(f2fs_readonly(inode->i_sb)))
215 		return 0;
216 
217 	trace_f2fs_sync_file_enter(inode);
218 
219 	/* if fdatasync is triggered, let's do in-place-update */
220 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
221 		set_inode_flag(inode, FI_NEED_IPU);
222 	ret = file_write_and_wait_range(file, start, end);
223 	clear_inode_flag(inode, FI_NEED_IPU);
224 
225 	if (ret) {
226 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
227 		return ret;
228 	}
229 
230 	/* if the inode is dirty, let's recover all the time */
231 	if (!f2fs_skip_inode_update(inode, datasync)) {
232 		f2fs_write_inode(inode, NULL);
233 		goto go_write;
234 	}
235 
236 	/*
237 	 * if there is no written data, don't waste time to write recovery info.
238 	 */
239 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
240 			!exist_written_data(sbi, ino, APPEND_INO)) {
241 
242 		/* it may call write_inode just prior to fsync */
243 		if (need_inode_page_update(sbi, ino))
244 			goto go_write;
245 
246 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
247 				exist_written_data(sbi, ino, UPDATE_INO))
248 			goto flush_out;
249 		goto out;
250 	}
251 go_write:
252 	/*
253 	 * Both of fdatasync() and fsync() are able to be recovered from
254 	 * sudden-power-off.
255 	 */
256 	down_read(&F2FS_I(inode)->i_sem);
257 	cp_reason = need_do_checkpoint(inode);
258 	up_read(&F2FS_I(inode)->i_sem);
259 
260 	if (cp_reason) {
261 		/* all the dirty node pages should be flushed for POR */
262 		ret = f2fs_sync_fs(inode->i_sb, 1);
263 
264 		/*
265 		 * We've secured consistency through sync_fs. Following pino
266 		 * will be used only for fsynced inodes after checkpoint.
267 		 */
268 		try_to_fix_pino(inode);
269 		clear_inode_flag(inode, FI_APPEND_WRITE);
270 		clear_inode_flag(inode, FI_UPDATE_WRITE);
271 		goto out;
272 	}
273 sync_nodes:
274 	ret = fsync_node_pages(sbi, inode, &wbc, atomic);
275 	if (ret)
276 		goto out;
277 
278 	/* if cp_error was enabled, we should avoid infinite loop */
279 	if (unlikely(f2fs_cp_error(sbi))) {
280 		ret = -EIO;
281 		goto out;
282 	}
283 
284 	if (need_inode_block_update(sbi, ino)) {
285 		f2fs_mark_inode_dirty_sync(inode, true);
286 		f2fs_write_inode(inode, NULL);
287 		goto sync_nodes;
288 	}
289 
290 	/*
291 	 * If it's atomic_write, it's just fine to keep write ordering. So
292 	 * here we don't need to wait for node write completion, since we use
293 	 * node chain which serializes node blocks. If one of node writes are
294 	 * reordered, we can see simply broken chain, resulting in stopping
295 	 * roll-forward recovery. It means we'll recover all or none node blocks
296 	 * given fsync mark.
297 	 */
298 	if (!atomic) {
299 		ret = wait_on_node_pages_writeback(sbi, ino);
300 		if (ret)
301 			goto out;
302 	}
303 
304 	/* once recovery info is written, don't need to tack this */
305 	remove_ino_entry(sbi, ino, APPEND_INO);
306 	clear_inode_flag(inode, FI_APPEND_WRITE);
307 flush_out:
308 	if (!atomic)
309 		ret = f2fs_issue_flush(sbi, inode->i_ino);
310 	if (!ret) {
311 		remove_ino_entry(sbi, ino, UPDATE_INO);
312 		clear_inode_flag(inode, FI_UPDATE_WRITE);
313 		remove_ino_entry(sbi, ino, FLUSH_INO);
314 	}
315 	f2fs_update_time(sbi, REQ_TIME);
316 out:
317 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
318 	f2fs_trace_ios(NULL, 1);
319 	return ret;
320 }
321 
322 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
323 {
324 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
325 		return -EIO;
326 	return f2fs_do_sync_file(file, start, end, datasync, false);
327 }
328 
329 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
330 						pgoff_t pgofs, int whence)
331 {
332 	struct page *page;
333 	int nr_pages;
334 
335 	if (whence != SEEK_DATA)
336 		return 0;
337 
338 	/* find first dirty page index */
339 	nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
340 				      1, &page);
341 	if (!nr_pages)
342 		return ULONG_MAX;
343 	pgofs = page->index;
344 	put_page(page);
345 	return pgofs;
346 }
347 
348 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
349 							int whence)
350 {
351 	switch (whence) {
352 	case SEEK_DATA:
353 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
354 			(blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
355 			return true;
356 		break;
357 	case SEEK_HOLE:
358 		if (blkaddr == NULL_ADDR)
359 			return true;
360 		break;
361 	}
362 	return false;
363 }
364 
365 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
366 {
367 	struct inode *inode = file->f_mapping->host;
368 	loff_t maxbytes = inode->i_sb->s_maxbytes;
369 	struct dnode_of_data dn;
370 	pgoff_t pgofs, end_offset, dirty;
371 	loff_t data_ofs = offset;
372 	loff_t isize;
373 	int err = 0;
374 
375 	inode_lock(inode);
376 
377 	isize = i_size_read(inode);
378 	if (offset >= isize)
379 		goto fail;
380 
381 	/* handle inline data case */
382 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
383 		if (whence == SEEK_HOLE)
384 			data_ofs = isize;
385 		goto found;
386 	}
387 
388 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
389 
390 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
391 
392 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
393 		set_new_dnode(&dn, inode, NULL, NULL, 0);
394 		err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
395 		if (err && err != -ENOENT) {
396 			goto fail;
397 		} else if (err == -ENOENT) {
398 			/* direct node does not exists */
399 			if (whence == SEEK_DATA) {
400 				pgofs = get_next_page_offset(&dn, pgofs);
401 				continue;
402 			} else {
403 				goto found;
404 			}
405 		}
406 
407 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
408 
409 		/* find data/hole in dnode block */
410 		for (; dn.ofs_in_node < end_offset;
411 				dn.ofs_in_node++, pgofs++,
412 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
413 			block_t blkaddr;
414 			blkaddr = datablock_addr(dn.inode,
415 					dn.node_page, dn.ofs_in_node);
416 
417 			if (__found_offset(blkaddr, dirty, pgofs, whence)) {
418 				f2fs_put_dnode(&dn);
419 				goto found;
420 			}
421 		}
422 		f2fs_put_dnode(&dn);
423 	}
424 
425 	if (whence == SEEK_DATA)
426 		goto fail;
427 found:
428 	if (whence == SEEK_HOLE && data_ofs > isize)
429 		data_ofs = isize;
430 	inode_unlock(inode);
431 	return vfs_setpos(file, data_ofs, maxbytes);
432 fail:
433 	inode_unlock(inode);
434 	return -ENXIO;
435 }
436 
437 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
438 {
439 	struct inode *inode = file->f_mapping->host;
440 	loff_t maxbytes = inode->i_sb->s_maxbytes;
441 
442 	switch (whence) {
443 	case SEEK_SET:
444 	case SEEK_CUR:
445 	case SEEK_END:
446 		return generic_file_llseek_size(file, offset, whence,
447 						maxbytes, i_size_read(inode));
448 	case SEEK_DATA:
449 	case SEEK_HOLE:
450 		if (offset < 0)
451 			return -ENXIO;
452 		return f2fs_seek_block(file, offset, whence);
453 	}
454 
455 	return -EINVAL;
456 }
457 
458 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
459 {
460 	struct inode *inode = file_inode(file);
461 	int err;
462 
463 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
464 		return -EIO;
465 
466 	/* we don't need to use inline_data strictly */
467 	err = f2fs_convert_inline_inode(inode);
468 	if (err)
469 		return err;
470 
471 	file_accessed(file);
472 	vma->vm_ops = &f2fs_file_vm_ops;
473 	return 0;
474 }
475 
476 static int f2fs_file_open(struct inode *inode, struct file *filp)
477 {
478 	int err = fscrypt_file_open(inode, filp);
479 
480 	if (err)
481 		return err;
482 	return dquot_file_open(inode, filp);
483 }
484 
485 void truncate_data_blocks_range(struct dnode_of_data *dn, int count)
486 {
487 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
488 	struct f2fs_node *raw_node;
489 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
490 	__le32 *addr;
491 	int base = 0;
492 
493 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
494 		base = get_extra_isize(dn->inode);
495 
496 	raw_node = F2FS_NODE(dn->node_page);
497 	addr = blkaddr_in_node(raw_node) + base + ofs;
498 
499 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
500 		block_t blkaddr = le32_to_cpu(*addr);
501 		if (blkaddr == NULL_ADDR)
502 			continue;
503 
504 		dn->data_blkaddr = NULL_ADDR;
505 		set_data_blkaddr(dn);
506 		invalidate_blocks(sbi, blkaddr);
507 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
508 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
509 		nr_free++;
510 	}
511 
512 	if (nr_free) {
513 		pgoff_t fofs;
514 		/*
515 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
516 		 * we will invalidate all blkaddr in the whole range.
517 		 */
518 		fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
519 							dn->inode) + ofs;
520 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
521 		dec_valid_block_count(sbi, dn->inode, nr_free);
522 	}
523 	dn->ofs_in_node = ofs;
524 
525 	f2fs_update_time(sbi, REQ_TIME);
526 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
527 					 dn->ofs_in_node, nr_free);
528 }
529 
530 void truncate_data_blocks(struct dnode_of_data *dn)
531 {
532 	truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
533 }
534 
535 static int truncate_partial_data_page(struct inode *inode, u64 from,
536 								bool cache_only)
537 {
538 	unsigned offset = from & (PAGE_SIZE - 1);
539 	pgoff_t index = from >> PAGE_SHIFT;
540 	struct address_space *mapping = inode->i_mapping;
541 	struct page *page;
542 
543 	if (!offset && !cache_only)
544 		return 0;
545 
546 	if (cache_only) {
547 		page = find_lock_page(mapping, index);
548 		if (page && PageUptodate(page))
549 			goto truncate_out;
550 		f2fs_put_page(page, 1);
551 		return 0;
552 	}
553 
554 	page = get_lock_data_page(inode, index, true);
555 	if (IS_ERR(page))
556 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
557 truncate_out:
558 	f2fs_wait_on_page_writeback(page, DATA, true);
559 	zero_user(page, offset, PAGE_SIZE - offset);
560 
561 	/* An encrypted inode should have a key and truncate the last page. */
562 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
563 	if (!cache_only)
564 		set_page_dirty(page);
565 	f2fs_put_page(page, 1);
566 	return 0;
567 }
568 
569 int truncate_blocks(struct inode *inode, u64 from, bool lock)
570 {
571 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
572 	unsigned int blocksize = inode->i_sb->s_blocksize;
573 	struct dnode_of_data dn;
574 	pgoff_t free_from;
575 	int count = 0, err = 0;
576 	struct page *ipage;
577 	bool truncate_page = false;
578 
579 	trace_f2fs_truncate_blocks_enter(inode, from);
580 
581 	free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
582 
583 	if (free_from >= sbi->max_file_blocks)
584 		goto free_partial;
585 
586 	if (lock)
587 		f2fs_lock_op(sbi);
588 
589 	ipage = get_node_page(sbi, inode->i_ino);
590 	if (IS_ERR(ipage)) {
591 		err = PTR_ERR(ipage);
592 		goto out;
593 	}
594 
595 	if (f2fs_has_inline_data(inode)) {
596 		truncate_inline_inode(inode, ipage, from);
597 		f2fs_put_page(ipage, 1);
598 		truncate_page = true;
599 		goto out;
600 	}
601 
602 	set_new_dnode(&dn, inode, ipage, NULL, 0);
603 	err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
604 	if (err) {
605 		if (err == -ENOENT)
606 			goto free_next;
607 		goto out;
608 	}
609 
610 	count = ADDRS_PER_PAGE(dn.node_page, inode);
611 
612 	count -= dn.ofs_in_node;
613 	f2fs_bug_on(sbi, count < 0);
614 
615 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
616 		truncate_data_blocks_range(&dn, count);
617 		free_from += count;
618 	}
619 
620 	f2fs_put_dnode(&dn);
621 free_next:
622 	err = truncate_inode_blocks(inode, free_from);
623 out:
624 	if (lock)
625 		f2fs_unlock_op(sbi);
626 free_partial:
627 	/* lastly zero out the first data page */
628 	if (!err)
629 		err = truncate_partial_data_page(inode, from, truncate_page);
630 
631 	trace_f2fs_truncate_blocks_exit(inode, err);
632 	return err;
633 }
634 
635 int f2fs_truncate(struct inode *inode)
636 {
637 	int err;
638 
639 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
640 		return -EIO;
641 
642 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
643 				S_ISLNK(inode->i_mode)))
644 		return 0;
645 
646 	trace_f2fs_truncate(inode);
647 
648 #ifdef CONFIG_F2FS_FAULT_INJECTION
649 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
650 		f2fs_show_injection_info(FAULT_TRUNCATE);
651 		return -EIO;
652 	}
653 #endif
654 	/* we should check inline_data size */
655 	if (!f2fs_may_inline_data(inode)) {
656 		err = f2fs_convert_inline_inode(inode);
657 		if (err)
658 			return err;
659 	}
660 
661 	err = truncate_blocks(inode, i_size_read(inode), true);
662 	if (err)
663 		return err;
664 
665 	inode->i_mtime = inode->i_ctime = current_time(inode);
666 	f2fs_mark_inode_dirty_sync(inode, false);
667 	return 0;
668 }
669 
670 int f2fs_getattr(const struct path *path, struct kstat *stat,
671 		 u32 request_mask, unsigned int query_flags)
672 {
673 	struct inode *inode = d_inode(path->dentry);
674 	struct f2fs_inode_info *fi = F2FS_I(inode);
675 	struct f2fs_inode *ri;
676 	unsigned int flags;
677 
678 	if (f2fs_has_extra_attr(inode) &&
679 			f2fs_sb_has_inode_crtime(inode->i_sb) &&
680 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
681 		stat->result_mask |= STATX_BTIME;
682 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
683 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
684 	}
685 
686 	flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
687 	if (flags & FS_APPEND_FL)
688 		stat->attributes |= STATX_ATTR_APPEND;
689 	if (flags & FS_COMPR_FL)
690 		stat->attributes |= STATX_ATTR_COMPRESSED;
691 	if (f2fs_encrypted_inode(inode))
692 		stat->attributes |= STATX_ATTR_ENCRYPTED;
693 	if (flags & FS_IMMUTABLE_FL)
694 		stat->attributes |= STATX_ATTR_IMMUTABLE;
695 	if (flags & FS_NODUMP_FL)
696 		stat->attributes |= STATX_ATTR_NODUMP;
697 
698 	stat->attributes_mask |= (STATX_ATTR_APPEND |
699 				  STATX_ATTR_COMPRESSED |
700 				  STATX_ATTR_ENCRYPTED |
701 				  STATX_ATTR_IMMUTABLE |
702 				  STATX_ATTR_NODUMP);
703 
704 	generic_fillattr(inode, stat);
705 
706 	/* we need to show initial sectors used for inline_data/dentries */
707 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
708 					f2fs_has_inline_dentry(inode))
709 		stat->blocks += (stat->size + 511) >> 9;
710 
711 	return 0;
712 }
713 
714 #ifdef CONFIG_F2FS_FS_POSIX_ACL
715 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
716 {
717 	unsigned int ia_valid = attr->ia_valid;
718 
719 	if (ia_valid & ATTR_UID)
720 		inode->i_uid = attr->ia_uid;
721 	if (ia_valid & ATTR_GID)
722 		inode->i_gid = attr->ia_gid;
723 	if (ia_valid & ATTR_ATIME)
724 		inode->i_atime = timespec_trunc(attr->ia_atime,
725 						inode->i_sb->s_time_gran);
726 	if (ia_valid & ATTR_MTIME)
727 		inode->i_mtime = timespec_trunc(attr->ia_mtime,
728 						inode->i_sb->s_time_gran);
729 	if (ia_valid & ATTR_CTIME)
730 		inode->i_ctime = timespec_trunc(attr->ia_ctime,
731 						inode->i_sb->s_time_gran);
732 	if (ia_valid & ATTR_MODE) {
733 		umode_t mode = attr->ia_mode;
734 
735 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
736 			mode &= ~S_ISGID;
737 		set_acl_inode(inode, mode);
738 	}
739 }
740 #else
741 #define __setattr_copy setattr_copy
742 #endif
743 
744 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
745 {
746 	struct inode *inode = d_inode(dentry);
747 	int err;
748 	bool size_changed = false;
749 
750 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
751 		return -EIO;
752 
753 	err = setattr_prepare(dentry, attr);
754 	if (err)
755 		return err;
756 
757 	err = fscrypt_prepare_setattr(dentry, attr);
758 	if (err)
759 		return err;
760 
761 	if (is_quota_modification(inode, attr)) {
762 		err = dquot_initialize(inode);
763 		if (err)
764 			return err;
765 	}
766 	if ((attr->ia_valid & ATTR_UID &&
767 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
768 		(attr->ia_valid & ATTR_GID &&
769 		!gid_eq(attr->ia_gid, inode->i_gid))) {
770 		err = dquot_transfer(inode, attr);
771 		if (err)
772 			return err;
773 	}
774 
775 	if (attr->ia_valid & ATTR_SIZE) {
776 		if (attr->ia_size <= i_size_read(inode)) {
777 			down_write(&F2FS_I(inode)->i_mmap_sem);
778 			truncate_setsize(inode, attr->ia_size);
779 			err = f2fs_truncate(inode);
780 			up_write(&F2FS_I(inode)->i_mmap_sem);
781 			if (err)
782 				return err;
783 		} else {
784 			/*
785 			 * do not trim all blocks after i_size if target size is
786 			 * larger than i_size.
787 			 */
788 			down_write(&F2FS_I(inode)->i_mmap_sem);
789 			truncate_setsize(inode, attr->ia_size);
790 			up_write(&F2FS_I(inode)->i_mmap_sem);
791 
792 			/* should convert inline inode here */
793 			if (!f2fs_may_inline_data(inode)) {
794 				err = f2fs_convert_inline_inode(inode);
795 				if (err)
796 					return err;
797 			}
798 			inode->i_mtime = inode->i_ctime = current_time(inode);
799 		}
800 
801 		down_write(&F2FS_I(inode)->i_sem);
802 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
803 		up_write(&F2FS_I(inode)->i_sem);
804 
805 		size_changed = true;
806 	}
807 
808 	__setattr_copy(inode, attr);
809 
810 	if (attr->ia_valid & ATTR_MODE) {
811 		err = posix_acl_chmod(inode, get_inode_mode(inode));
812 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
813 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
814 			clear_inode_flag(inode, FI_ACL_MODE);
815 		}
816 	}
817 
818 	/* file size may changed here */
819 	f2fs_mark_inode_dirty_sync(inode, size_changed);
820 
821 	/* inode change will produce dirty node pages flushed by checkpoint */
822 	f2fs_balance_fs(F2FS_I_SB(inode), true);
823 
824 	return err;
825 }
826 
827 const struct inode_operations f2fs_file_inode_operations = {
828 	.getattr	= f2fs_getattr,
829 	.setattr	= f2fs_setattr,
830 	.get_acl	= f2fs_get_acl,
831 	.set_acl	= f2fs_set_acl,
832 #ifdef CONFIG_F2FS_FS_XATTR
833 	.listxattr	= f2fs_listxattr,
834 #endif
835 	.fiemap		= f2fs_fiemap,
836 };
837 
838 static int fill_zero(struct inode *inode, pgoff_t index,
839 					loff_t start, loff_t len)
840 {
841 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
842 	struct page *page;
843 
844 	if (!len)
845 		return 0;
846 
847 	f2fs_balance_fs(sbi, true);
848 
849 	f2fs_lock_op(sbi);
850 	page = get_new_data_page(inode, NULL, index, false);
851 	f2fs_unlock_op(sbi);
852 
853 	if (IS_ERR(page))
854 		return PTR_ERR(page);
855 
856 	f2fs_wait_on_page_writeback(page, DATA, true);
857 	zero_user(page, start, len);
858 	set_page_dirty(page);
859 	f2fs_put_page(page, 1);
860 	return 0;
861 }
862 
863 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
864 {
865 	int err;
866 
867 	while (pg_start < pg_end) {
868 		struct dnode_of_data dn;
869 		pgoff_t end_offset, count;
870 
871 		set_new_dnode(&dn, inode, NULL, NULL, 0);
872 		err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
873 		if (err) {
874 			if (err == -ENOENT) {
875 				pg_start = get_next_page_offset(&dn, pg_start);
876 				continue;
877 			}
878 			return err;
879 		}
880 
881 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
882 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
883 
884 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
885 
886 		truncate_data_blocks_range(&dn, count);
887 		f2fs_put_dnode(&dn);
888 
889 		pg_start += count;
890 	}
891 	return 0;
892 }
893 
894 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
895 {
896 	pgoff_t pg_start, pg_end;
897 	loff_t off_start, off_end;
898 	int ret;
899 
900 	ret = f2fs_convert_inline_inode(inode);
901 	if (ret)
902 		return ret;
903 
904 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
905 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
906 
907 	off_start = offset & (PAGE_SIZE - 1);
908 	off_end = (offset + len) & (PAGE_SIZE - 1);
909 
910 	if (pg_start == pg_end) {
911 		ret = fill_zero(inode, pg_start, off_start,
912 						off_end - off_start);
913 		if (ret)
914 			return ret;
915 	} else {
916 		if (off_start) {
917 			ret = fill_zero(inode, pg_start++, off_start,
918 						PAGE_SIZE - off_start);
919 			if (ret)
920 				return ret;
921 		}
922 		if (off_end) {
923 			ret = fill_zero(inode, pg_end, 0, off_end);
924 			if (ret)
925 				return ret;
926 		}
927 
928 		if (pg_start < pg_end) {
929 			struct address_space *mapping = inode->i_mapping;
930 			loff_t blk_start, blk_end;
931 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
932 
933 			f2fs_balance_fs(sbi, true);
934 
935 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
936 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
937 			down_write(&F2FS_I(inode)->i_mmap_sem);
938 			truncate_inode_pages_range(mapping, blk_start,
939 					blk_end - 1);
940 
941 			f2fs_lock_op(sbi);
942 			ret = truncate_hole(inode, pg_start, pg_end);
943 			f2fs_unlock_op(sbi);
944 			up_write(&F2FS_I(inode)->i_mmap_sem);
945 		}
946 	}
947 
948 	return ret;
949 }
950 
951 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
952 				int *do_replace, pgoff_t off, pgoff_t len)
953 {
954 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
955 	struct dnode_of_data dn;
956 	int ret, done, i;
957 
958 next_dnode:
959 	set_new_dnode(&dn, inode, NULL, NULL, 0);
960 	ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
961 	if (ret && ret != -ENOENT) {
962 		return ret;
963 	} else if (ret == -ENOENT) {
964 		if (dn.max_level == 0)
965 			return -ENOENT;
966 		done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
967 		blkaddr += done;
968 		do_replace += done;
969 		goto next;
970 	}
971 
972 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
973 							dn.ofs_in_node, len);
974 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
975 		*blkaddr = datablock_addr(dn.inode,
976 					dn.node_page, dn.ofs_in_node);
977 		if (!is_checkpointed_data(sbi, *blkaddr)) {
978 
979 			if (test_opt(sbi, LFS)) {
980 				f2fs_put_dnode(&dn);
981 				return -ENOTSUPP;
982 			}
983 
984 			/* do not invalidate this block address */
985 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
986 			*do_replace = 1;
987 		}
988 	}
989 	f2fs_put_dnode(&dn);
990 next:
991 	len -= done;
992 	off += done;
993 	if (len)
994 		goto next_dnode;
995 	return 0;
996 }
997 
998 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
999 				int *do_replace, pgoff_t off, int len)
1000 {
1001 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1002 	struct dnode_of_data dn;
1003 	int ret, i;
1004 
1005 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1006 		if (*do_replace == 0)
1007 			continue;
1008 
1009 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1010 		ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1011 		if (ret) {
1012 			dec_valid_block_count(sbi, inode, 1);
1013 			invalidate_blocks(sbi, *blkaddr);
1014 		} else {
1015 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1016 		}
1017 		f2fs_put_dnode(&dn);
1018 	}
1019 	return 0;
1020 }
1021 
1022 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1023 			block_t *blkaddr, int *do_replace,
1024 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1025 {
1026 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1027 	pgoff_t i = 0;
1028 	int ret;
1029 
1030 	while (i < len) {
1031 		if (blkaddr[i] == NULL_ADDR && !full) {
1032 			i++;
1033 			continue;
1034 		}
1035 
1036 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1037 			struct dnode_of_data dn;
1038 			struct node_info ni;
1039 			size_t new_size;
1040 			pgoff_t ilen;
1041 
1042 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1043 			ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1044 			if (ret)
1045 				return ret;
1046 
1047 			get_node_info(sbi, dn.nid, &ni);
1048 			ilen = min((pgoff_t)
1049 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1050 						dn.ofs_in_node, len - i);
1051 			do {
1052 				dn.data_blkaddr = datablock_addr(dn.inode,
1053 						dn.node_page, dn.ofs_in_node);
1054 				truncate_data_blocks_range(&dn, 1);
1055 
1056 				if (do_replace[i]) {
1057 					f2fs_i_blocks_write(src_inode,
1058 							1, false, false);
1059 					f2fs_i_blocks_write(dst_inode,
1060 							1, true, false);
1061 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1062 					blkaddr[i], ni.version, true, false);
1063 
1064 					do_replace[i] = 0;
1065 				}
1066 				dn.ofs_in_node++;
1067 				i++;
1068 				new_size = (dst + i) << PAGE_SHIFT;
1069 				if (dst_inode->i_size < new_size)
1070 					f2fs_i_size_write(dst_inode, new_size);
1071 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1072 
1073 			f2fs_put_dnode(&dn);
1074 		} else {
1075 			struct page *psrc, *pdst;
1076 
1077 			psrc = get_lock_data_page(src_inode, src + i, true);
1078 			if (IS_ERR(psrc))
1079 				return PTR_ERR(psrc);
1080 			pdst = get_new_data_page(dst_inode, NULL, dst + i,
1081 								true);
1082 			if (IS_ERR(pdst)) {
1083 				f2fs_put_page(psrc, 1);
1084 				return PTR_ERR(pdst);
1085 			}
1086 			f2fs_copy_page(psrc, pdst);
1087 			set_page_dirty(pdst);
1088 			f2fs_put_page(pdst, 1);
1089 			f2fs_put_page(psrc, 1);
1090 
1091 			ret = truncate_hole(src_inode, src + i, src + i + 1);
1092 			if (ret)
1093 				return ret;
1094 			i++;
1095 		}
1096 	}
1097 	return 0;
1098 }
1099 
1100 static int __exchange_data_block(struct inode *src_inode,
1101 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1102 			pgoff_t len, bool full)
1103 {
1104 	block_t *src_blkaddr;
1105 	int *do_replace;
1106 	pgoff_t olen;
1107 	int ret;
1108 
1109 	while (len) {
1110 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1111 
1112 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1113 					sizeof(block_t) * olen, GFP_KERNEL);
1114 		if (!src_blkaddr)
1115 			return -ENOMEM;
1116 
1117 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1118 					sizeof(int) * olen, GFP_KERNEL);
1119 		if (!do_replace) {
1120 			kvfree(src_blkaddr);
1121 			return -ENOMEM;
1122 		}
1123 
1124 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1125 					do_replace, src, olen);
1126 		if (ret)
1127 			goto roll_back;
1128 
1129 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1130 					do_replace, src, dst, olen, full);
1131 		if (ret)
1132 			goto roll_back;
1133 
1134 		src += olen;
1135 		dst += olen;
1136 		len -= olen;
1137 
1138 		kvfree(src_blkaddr);
1139 		kvfree(do_replace);
1140 	}
1141 	return 0;
1142 
1143 roll_back:
1144 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1145 	kvfree(src_blkaddr);
1146 	kvfree(do_replace);
1147 	return ret;
1148 }
1149 
1150 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1151 {
1152 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1153 	pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1154 	int ret;
1155 
1156 	f2fs_balance_fs(sbi, true);
1157 	f2fs_lock_op(sbi);
1158 
1159 	f2fs_drop_extent_tree(inode);
1160 
1161 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1162 	f2fs_unlock_op(sbi);
1163 	return ret;
1164 }
1165 
1166 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1167 {
1168 	pgoff_t pg_start, pg_end;
1169 	loff_t new_size;
1170 	int ret;
1171 
1172 	if (offset + len >= i_size_read(inode))
1173 		return -EINVAL;
1174 
1175 	/* collapse range should be aligned to block size of f2fs. */
1176 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1177 		return -EINVAL;
1178 
1179 	ret = f2fs_convert_inline_inode(inode);
1180 	if (ret)
1181 		return ret;
1182 
1183 	pg_start = offset >> PAGE_SHIFT;
1184 	pg_end = (offset + len) >> PAGE_SHIFT;
1185 
1186 	/* avoid gc operation during block exchange */
1187 	down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1188 
1189 	down_write(&F2FS_I(inode)->i_mmap_sem);
1190 	/* write out all dirty pages from offset */
1191 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1192 	if (ret)
1193 		goto out_unlock;
1194 
1195 	truncate_pagecache(inode, offset);
1196 
1197 	ret = f2fs_do_collapse(inode, pg_start, pg_end);
1198 	if (ret)
1199 		goto out_unlock;
1200 
1201 	/* write out all moved pages, if possible */
1202 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1203 	truncate_pagecache(inode, offset);
1204 
1205 	new_size = i_size_read(inode) - len;
1206 	truncate_pagecache(inode, new_size);
1207 
1208 	ret = truncate_blocks(inode, new_size, true);
1209 	if (!ret)
1210 		f2fs_i_size_write(inode, new_size);
1211 out_unlock:
1212 	up_write(&F2FS_I(inode)->i_mmap_sem);
1213 	up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1214 	return ret;
1215 }
1216 
1217 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1218 								pgoff_t end)
1219 {
1220 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1221 	pgoff_t index = start;
1222 	unsigned int ofs_in_node = dn->ofs_in_node;
1223 	blkcnt_t count = 0;
1224 	int ret;
1225 
1226 	for (; index < end; index++, dn->ofs_in_node++) {
1227 		if (datablock_addr(dn->inode, dn->node_page,
1228 					dn->ofs_in_node) == NULL_ADDR)
1229 			count++;
1230 	}
1231 
1232 	dn->ofs_in_node = ofs_in_node;
1233 	ret = reserve_new_blocks(dn, count);
1234 	if (ret)
1235 		return ret;
1236 
1237 	dn->ofs_in_node = ofs_in_node;
1238 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1239 		dn->data_blkaddr = datablock_addr(dn->inode,
1240 					dn->node_page, dn->ofs_in_node);
1241 		/*
1242 		 * reserve_new_blocks will not guarantee entire block
1243 		 * allocation.
1244 		 */
1245 		if (dn->data_blkaddr == NULL_ADDR) {
1246 			ret = -ENOSPC;
1247 			break;
1248 		}
1249 		if (dn->data_blkaddr != NEW_ADDR) {
1250 			invalidate_blocks(sbi, dn->data_blkaddr);
1251 			dn->data_blkaddr = NEW_ADDR;
1252 			set_data_blkaddr(dn);
1253 		}
1254 	}
1255 
1256 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1257 
1258 	return ret;
1259 }
1260 
1261 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1262 								int mode)
1263 {
1264 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1265 	struct address_space *mapping = inode->i_mapping;
1266 	pgoff_t index, pg_start, pg_end;
1267 	loff_t new_size = i_size_read(inode);
1268 	loff_t off_start, off_end;
1269 	int ret = 0;
1270 
1271 	ret = inode_newsize_ok(inode, (len + offset));
1272 	if (ret)
1273 		return ret;
1274 
1275 	ret = f2fs_convert_inline_inode(inode);
1276 	if (ret)
1277 		return ret;
1278 
1279 	down_write(&F2FS_I(inode)->i_mmap_sem);
1280 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1281 	if (ret)
1282 		goto out_sem;
1283 
1284 	truncate_pagecache_range(inode, offset, offset + len - 1);
1285 
1286 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1287 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1288 
1289 	off_start = offset & (PAGE_SIZE - 1);
1290 	off_end = (offset + len) & (PAGE_SIZE - 1);
1291 
1292 	if (pg_start == pg_end) {
1293 		ret = fill_zero(inode, pg_start, off_start,
1294 						off_end - off_start);
1295 		if (ret)
1296 			goto out_sem;
1297 
1298 		new_size = max_t(loff_t, new_size, offset + len);
1299 	} else {
1300 		if (off_start) {
1301 			ret = fill_zero(inode, pg_start++, off_start,
1302 						PAGE_SIZE - off_start);
1303 			if (ret)
1304 				goto out_sem;
1305 
1306 			new_size = max_t(loff_t, new_size,
1307 					(loff_t)pg_start << PAGE_SHIFT);
1308 		}
1309 
1310 		for (index = pg_start; index < pg_end;) {
1311 			struct dnode_of_data dn;
1312 			unsigned int end_offset;
1313 			pgoff_t end;
1314 
1315 			f2fs_lock_op(sbi);
1316 
1317 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1318 			ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1319 			if (ret) {
1320 				f2fs_unlock_op(sbi);
1321 				goto out;
1322 			}
1323 
1324 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1325 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1326 
1327 			ret = f2fs_do_zero_range(&dn, index, end);
1328 			f2fs_put_dnode(&dn);
1329 			f2fs_unlock_op(sbi);
1330 
1331 			f2fs_balance_fs(sbi, dn.node_changed);
1332 
1333 			if (ret)
1334 				goto out;
1335 
1336 			index = end;
1337 			new_size = max_t(loff_t, new_size,
1338 					(loff_t)index << PAGE_SHIFT);
1339 		}
1340 
1341 		if (off_end) {
1342 			ret = fill_zero(inode, pg_end, 0, off_end);
1343 			if (ret)
1344 				goto out;
1345 
1346 			new_size = max_t(loff_t, new_size, offset + len);
1347 		}
1348 	}
1349 
1350 out:
1351 	if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1352 		f2fs_i_size_write(inode, new_size);
1353 out_sem:
1354 	up_write(&F2FS_I(inode)->i_mmap_sem);
1355 
1356 	return ret;
1357 }
1358 
1359 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1360 {
1361 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1362 	pgoff_t nr, pg_start, pg_end, delta, idx;
1363 	loff_t new_size;
1364 	int ret = 0;
1365 
1366 	new_size = i_size_read(inode) + len;
1367 	ret = inode_newsize_ok(inode, new_size);
1368 	if (ret)
1369 		return ret;
1370 
1371 	if (offset >= i_size_read(inode))
1372 		return -EINVAL;
1373 
1374 	/* insert range should be aligned to block size of f2fs. */
1375 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1376 		return -EINVAL;
1377 
1378 	ret = f2fs_convert_inline_inode(inode);
1379 	if (ret)
1380 		return ret;
1381 
1382 	f2fs_balance_fs(sbi, true);
1383 
1384 	/* avoid gc operation during block exchange */
1385 	down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1386 
1387 	down_write(&F2FS_I(inode)->i_mmap_sem);
1388 	ret = truncate_blocks(inode, i_size_read(inode), true);
1389 	if (ret)
1390 		goto out;
1391 
1392 	/* write out all dirty pages from offset */
1393 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1394 	if (ret)
1395 		goto out;
1396 
1397 	truncate_pagecache(inode, offset);
1398 
1399 	pg_start = offset >> PAGE_SHIFT;
1400 	pg_end = (offset + len) >> PAGE_SHIFT;
1401 	delta = pg_end - pg_start;
1402 	idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1403 
1404 	while (!ret && idx > pg_start) {
1405 		nr = idx - pg_start;
1406 		if (nr > delta)
1407 			nr = delta;
1408 		idx -= nr;
1409 
1410 		f2fs_lock_op(sbi);
1411 		f2fs_drop_extent_tree(inode);
1412 
1413 		ret = __exchange_data_block(inode, inode, idx,
1414 					idx + delta, nr, false);
1415 		f2fs_unlock_op(sbi);
1416 	}
1417 
1418 	/* write out all moved pages, if possible */
1419 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1420 	truncate_pagecache(inode, offset);
1421 
1422 	if (!ret)
1423 		f2fs_i_size_write(inode, new_size);
1424 out:
1425 	up_write(&F2FS_I(inode)->i_mmap_sem);
1426 	up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1427 	return ret;
1428 }
1429 
1430 static int expand_inode_data(struct inode *inode, loff_t offset,
1431 					loff_t len, int mode)
1432 {
1433 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1434 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1435 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1436 	pgoff_t pg_end;
1437 	loff_t new_size = i_size_read(inode);
1438 	loff_t off_end;
1439 	int err;
1440 
1441 	err = inode_newsize_ok(inode, (len + offset));
1442 	if (err)
1443 		return err;
1444 
1445 	err = f2fs_convert_inline_inode(inode);
1446 	if (err)
1447 		return err;
1448 
1449 	f2fs_balance_fs(sbi, true);
1450 
1451 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1452 	off_end = (offset + len) & (PAGE_SIZE - 1);
1453 
1454 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1455 	map.m_len = pg_end - map.m_lblk;
1456 	if (off_end)
1457 		map.m_len++;
1458 
1459 	err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1460 	if (err) {
1461 		pgoff_t last_off;
1462 
1463 		if (!map.m_len)
1464 			return err;
1465 
1466 		last_off = map.m_lblk + map.m_len - 1;
1467 
1468 		/* update new size to the failed position */
1469 		new_size = (last_off == pg_end) ? offset + len:
1470 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1471 	} else {
1472 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1473 	}
1474 
1475 	if (new_size > i_size_read(inode)) {
1476 		if (mode & FALLOC_FL_KEEP_SIZE)
1477 			file_set_keep_isize(inode);
1478 		else
1479 			f2fs_i_size_write(inode, new_size);
1480 	}
1481 
1482 	return err;
1483 }
1484 
1485 static long f2fs_fallocate(struct file *file, int mode,
1486 				loff_t offset, loff_t len)
1487 {
1488 	struct inode *inode = file_inode(file);
1489 	long ret = 0;
1490 
1491 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1492 		return -EIO;
1493 
1494 	/* f2fs only support ->fallocate for regular file */
1495 	if (!S_ISREG(inode->i_mode))
1496 		return -EINVAL;
1497 
1498 	if (f2fs_encrypted_inode(inode) &&
1499 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1500 		return -EOPNOTSUPP;
1501 
1502 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1503 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1504 			FALLOC_FL_INSERT_RANGE))
1505 		return -EOPNOTSUPP;
1506 
1507 	inode_lock(inode);
1508 
1509 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1510 		if (offset >= inode->i_size)
1511 			goto out;
1512 
1513 		ret = punch_hole(inode, offset, len);
1514 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1515 		ret = f2fs_collapse_range(inode, offset, len);
1516 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1517 		ret = f2fs_zero_range(inode, offset, len, mode);
1518 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1519 		ret = f2fs_insert_range(inode, offset, len);
1520 	} else {
1521 		ret = expand_inode_data(inode, offset, len, mode);
1522 	}
1523 
1524 	if (!ret) {
1525 		inode->i_mtime = inode->i_ctime = current_time(inode);
1526 		f2fs_mark_inode_dirty_sync(inode, false);
1527 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1528 	}
1529 
1530 out:
1531 	inode_unlock(inode);
1532 
1533 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1534 	return ret;
1535 }
1536 
1537 static int f2fs_release_file(struct inode *inode, struct file *filp)
1538 {
1539 	/*
1540 	 * f2fs_relase_file is called at every close calls. So we should
1541 	 * not drop any inmemory pages by close called by other process.
1542 	 */
1543 	if (!(filp->f_mode & FMODE_WRITE) ||
1544 			atomic_read(&inode->i_writecount) != 1)
1545 		return 0;
1546 
1547 	/* some remained atomic pages should discarded */
1548 	if (f2fs_is_atomic_file(inode))
1549 		drop_inmem_pages(inode);
1550 	if (f2fs_is_volatile_file(inode)) {
1551 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1552 		stat_dec_volatile_write(inode);
1553 		set_inode_flag(inode, FI_DROP_CACHE);
1554 		filemap_fdatawrite(inode->i_mapping);
1555 		clear_inode_flag(inode, FI_DROP_CACHE);
1556 	}
1557 	return 0;
1558 }
1559 
1560 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1561 {
1562 	struct inode *inode = file_inode(file);
1563 
1564 	/*
1565 	 * If the process doing a transaction is crashed, we should do
1566 	 * roll-back. Otherwise, other reader/write can see corrupted database
1567 	 * until all the writers close its file. Since this should be done
1568 	 * before dropping file lock, it needs to do in ->flush.
1569 	 */
1570 	if (f2fs_is_atomic_file(inode) &&
1571 			F2FS_I(inode)->inmem_task == current)
1572 		drop_inmem_pages(inode);
1573 	return 0;
1574 }
1575 
1576 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1577 {
1578 	struct inode *inode = file_inode(filp);
1579 	struct f2fs_inode_info *fi = F2FS_I(inode);
1580 	unsigned int flags = fi->i_flags &
1581 			(FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1582 	return put_user(flags, (int __user *)arg);
1583 }
1584 
1585 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1586 {
1587 	struct f2fs_inode_info *fi = F2FS_I(inode);
1588 	unsigned int oldflags;
1589 
1590 	/* Is it quota file? Do not allow user to mess with it */
1591 	if (IS_NOQUOTA(inode))
1592 		return -EPERM;
1593 
1594 	flags = f2fs_mask_flags(inode->i_mode, flags);
1595 
1596 	oldflags = fi->i_flags;
1597 
1598 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1599 		if (!capable(CAP_LINUX_IMMUTABLE))
1600 			return -EPERM;
1601 
1602 	flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1603 	flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1604 	fi->i_flags = flags;
1605 
1606 	if (fi->i_flags & FS_PROJINHERIT_FL)
1607 		set_inode_flag(inode, FI_PROJ_INHERIT);
1608 	else
1609 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1610 
1611 	inode->i_ctime = current_time(inode);
1612 	f2fs_set_inode_flags(inode);
1613 	f2fs_mark_inode_dirty_sync(inode, false);
1614 	return 0;
1615 }
1616 
1617 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1618 {
1619 	struct inode *inode = file_inode(filp);
1620 	unsigned int flags;
1621 	int ret;
1622 
1623 	if (!inode_owner_or_capable(inode))
1624 		return -EACCES;
1625 
1626 	if (get_user(flags, (int __user *)arg))
1627 		return -EFAULT;
1628 
1629 	ret = mnt_want_write_file(filp);
1630 	if (ret)
1631 		return ret;
1632 
1633 	inode_lock(inode);
1634 
1635 	ret = __f2fs_ioc_setflags(inode, flags);
1636 
1637 	inode_unlock(inode);
1638 	mnt_drop_write_file(filp);
1639 	return ret;
1640 }
1641 
1642 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1643 {
1644 	struct inode *inode = file_inode(filp);
1645 
1646 	return put_user(inode->i_generation, (int __user *)arg);
1647 }
1648 
1649 static int f2fs_ioc_start_atomic_write(struct file *filp)
1650 {
1651 	struct inode *inode = file_inode(filp);
1652 	int ret;
1653 
1654 	if (!inode_owner_or_capable(inode))
1655 		return -EACCES;
1656 
1657 	if (!S_ISREG(inode->i_mode))
1658 		return -EINVAL;
1659 
1660 	ret = mnt_want_write_file(filp);
1661 	if (ret)
1662 		return ret;
1663 
1664 	inode_lock(inode);
1665 
1666 	if (f2fs_is_atomic_file(inode))
1667 		goto out;
1668 
1669 	ret = f2fs_convert_inline_inode(inode);
1670 	if (ret)
1671 		goto out;
1672 
1673 	set_inode_flag(inode, FI_ATOMIC_FILE);
1674 	set_inode_flag(inode, FI_HOT_DATA);
1675 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1676 
1677 	if (!get_dirty_pages(inode))
1678 		goto inc_stat;
1679 
1680 	f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1681 		"Unexpected flush for atomic writes: ino=%lu, npages=%u",
1682 					inode->i_ino, get_dirty_pages(inode));
1683 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1684 	if (ret) {
1685 		clear_inode_flag(inode, FI_ATOMIC_FILE);
1686 		clear_inode_flag(inode, FI_HOT_DATA);
1687 		goto out;
1688 	}
1689 
1690 inc_stat:
1691 	F2FS_I(inode)->inmem_task = current;
1692 	stat_inc_atomic_write(inode);
1693 	stat_update_max_atomic_write(inode);
1694 out:
1695 	inode_unlock(inode);
1696 	mnt_drop_write_file(filp);
1697 	return ret;
1698 }
1699 
1700 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1701 {
1702 	struct inode *inode = file_inode(filp);
1703 	int ret;
1704 
1705 	if (!inode_owner_or_capable(inode))
1706 		return -EACCES;
1707 
1708 	ret = mnt_want_write_file(filp);
1709 	if (ret)
1710 		return ret;
1711 
1712 	inode_lock(inode);
1713 
1714 	if (f2fs_is_volatile_file(inode))
1715 		goto err_out;
1716 
1717 	if (f2fs_is_atomic_file(inode)) {
1718 		ret = commit_inmem_pages(inode);
1719 		if (ret)
1720 			goto err_out;
1721 
1722 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1723 		if (!ret) {
1724 			clear_inode_flag(inode, FI_ATOMIC_FILE);
1725 			clear_inode_flag(inode, FI_HOT_DATA);
1726 			stat_dec_atomic_write(inode);
1727 		}
1728 	} else {
1729 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1730 	}
1731 err_out:
1732 	inode_unlock(inode);
1733 	mnt_drop_write_file(filp);
1734 	return ret;
1735 }
1736 
1737 static int f2fs_ioc_start_volatile_write(struct file *filp)
1738 {
1739 	struct inode *inode = file_inode(filp);
1740 	int ret;
1741 
1742 	if (!inode_owner_or_capable(inode))
1743 		return -EACCES;
1744 
1745 	if (!S_ISREG(inode->i_mode))
1746 		return -EINVAL;
1747 
1748 	ret = mnt_want_write_file(filp);
1749 	if (ret)
1750 		return ret;
1751 
1752 	inode_lock(inode);
1753 
1754 	if (f2fs_is_volatile_file(inode))
1755 		goto out;
1756 
1757 	ret = f2fs_convert_inline_inode(inode);
1758 	if (ret)
1759 		goto out;
1760 
1761 	stat_inc_volatile_write(inode);
1762 	stat_update_max_volatile_write(inode);
1763 
1764 	set_inode_flag(inode, FI_VOLATILE_FILE);
1765 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1766 out:
1767 	inode_unlock(inode);
1768 	mnt_drop_write_file(filp);
1769 	return ret;
1770 }
1771 
1772 static int f2fs_ioc_release_volatile_write(struct file *filp)
1773 {
1774 	struct inode *inode = file_inode(filp);
1775 	int ret;
1776 
1777 	if (!inode_owner_or_capable(inode))
1778 		return -EACCES;
1779 
1780 	ret = mnt_want_write_file(filp);
1781 	if (ret)
1782 		return ret;
1783 
1784 	inode_lock(inode);
1785 
1786 	if (!f2fs_is_volatile_file(inode))
1787 		goto out;
1788 
1789 	if (!f2fs_is_first_block_written(inode)) {
1790 		ret = truncate_partial_data_page(inode, 0, true);
1791 		goto out;
1792 	}
1793 
1794 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1795 out:
1796 	inode_unlock(inode);
1797 	mnt_drop_write_file(filp);
1798 	return ret;
1799 }
1800 
1801 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1802 {
1803 	struct inode *inode = file_inode(filp);
1804 	int ret;
1805 
1806 	if (!inode_owner_or_capable(inode))
1807 		return -EACCES;
1808 
1809 	ret = mnt_want_write_file(filp);
1810 	if (ret)
1811 		return ret;
1812 
1813 	inode_lock(inode);
1814 
1815 	if (f2fs_is_atomic_file(inode))
1816 		drop_inmem_pages(inode);
1817 	if (f2fs_is_volatile_file(inode)) {
1818 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1819 		stat_dec_volatile_write(inode);
1820 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1821 	}
1822 
1823 	inode_unlock(inode);
1824 
1825 	mnt_drop_write_file(filp);
1826 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1827 	return ret;
1828 }
1829 
1830 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1831 {
1832 	struct inode *inode = file_inode(filp);
1833 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1834 	struct super_block *sb = sbi->sb;
1835 	__u32 in;
1836 	int ret;
1837 
1838 	if (!capable(CAP_SYS_ADMIN))
1839 		return -EPERM;
1840 
1841 	if (get_user(in, (__u32 __user *)arg))
1842 		return -EFAULT;
1843 
1844 	ret = mnt_want_write_file(filp);
1845 	if (ret)
1846 		return ret;
1847 
1848 	switch (in) {
1849 	case F2FS_GOING_DOWN_FULLSYNC:
1850 		sb = freeze_bdev(sb->s_bdev);
1851 		if (IS_ERR(sb)) {
1852 			ret = PTR_ERR(sb);
1853 			goto out;
1854 		}
1855 		if (sb) {
1856 			f2fs_stop_checkpoint(sbi, false);
1857 			thaw_bdev(sb->s_bdev, sb);
1858 		}
1859 		break;
1860 	case F2FS_GOING_DOWN_METASYNC:
1861 		/* do checkpoint only */
1862 		ret = f2fs_sync_fs(sb, 1);
1863 		if (ret)
1864 			goto out;
1865 		f2fs_stop_checkpoint(sbi, false);
1866 		break;
1867 	case F2FS_GOING_DOWN_NOSYNC:
1868 		f2fs_stop_checkpoint(sbi, false);
1869 		break;
1870 	case F2FS_GOING_DOWN_METAFLUSH:
1871 		sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1872 		f2fs_stop_checkpoint(sbi, false);
1873 		break;
1874 	default:
1875 		ret = -EINVAL;
1876 		goto out;
1877 	}
1878 
1879 	stop_gc_thread(sbi);
1880 	stop_discard_thread(sbi);
1881 
1882 	drop_discard_cmd(sbi);
1883 	clear_opt(sbi, DISCARD);
1884 
1885 	f2fs_update_time(sbi, REQ_TIME);
1886 out:
1887 	mnt_drop_write_file(filp);
1888 	return ret;
1889 }
1890 
1891 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1892 {
1893 	struct inode *inode = file_inode(filp);
1894 	struct super_block *sb = inode->i_sb;
1895 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
1896 	struct fstrim_range range;
1897 	int ret;
1898 
1899 	if (!capable(CAP_SYS_ADMIN))
1900 		return -EPERM;
1901 
1902 	if (!blk_queue_discard(q))
1903 		return -EOPNOTSUPP;
1904 
1905 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1906 				sizeof(range)))
1907 		return -EFAULT;
1908 
1909 	ret = mnt_want_write_file(filp);
1910 	if (ret)
1911 		return ret;
1912 
1913 	range.minlen = max((unsigned int)range.minlen,
1914 				q->limits.discard_granularity);
1915 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1916 	mnt_drop_write_file(filp);
1917 	if (ret < 0)
1918 		return ret;
1919 
1920 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
1921 				sizeof(range)))
1922 		return -EFAULT;
1923 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1924 	return 0;
1925 }
1926 
1927 static bool uuid_is_nonzero(__u8 u[16])
1928 {
1929 	int i;
1930 
1931 	for (i = 0; i < 16; i++)
1932 		if (u[i])
1933 			return true;
1934 	return false;
1935 }
1936 
1937 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1938 {
1939 	struct inode *inode = file_inode(filp);
1940 
1941 	if (!f2fs_sb_has_crypto(inode->i_sb))
1942 		return -EOPNOTSUPP;
1943 
1944 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1945 
1946 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1947 }
1948 
1949 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1950 {
1951 	if (!f2fs_sb_has_crypto(file_inode(filp)->i_sb))
1952 		return -EOPNOTSUPP;
1953 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1954 }
1955 
1956 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1957 {
1958 	struct inode *inode = file_inode(filp);
1959 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1960 	int err;
1961 
1962 	if (!f2fs_sb_has_crypto(inode->i_sb))
1963 		return -EOPNOTSUPP;
1964 
1965 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1966 		goto got_it;
1967 
1968 	err = mnt_want_write_file(filp);
1969 	if (err)
1970 		return err;
1971 
1972 	/* update superblock with uuid */
1973 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1974 
1975 	err = f2fs_commit_super(sbi, false);
1976 	if (err) {
1977 		/* undo new data */
1978 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1979 		mnt_drop_write_file(filp);
1980 		return err;
1981 	}
1982 	mnt_drop_write_file(filp);
1983 got_it:
1984 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1985 									16))
1986 		return -EFAULT;
1987 	return 0;
1988 }
1989 
1990 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1991 {
1992 	struct inode *inode = file_inode(filp);
1993 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1994 	__u32 sync;
1995 	int ret;
1996 
1997 	if (!capable(CAP_SYS_ADMIN))
1998 		return -EPERM;
1999 
2000 	if (get_user(sync, (__u32 __user *)arg))
2001 		return -EFAULT;
2002 
2003 	if (f2fs_readonly(sbi->sb))
2004 		return -EROFS;
2005 
2006 	ret = mnt_want_write_file(filp);
2007 	if (ret)
2008 		return ret;
2009 
2010 	if (!sync) {
2011 		if (!mutex_trylock(&sbi->gc_mutex)) {
2012 			ret = -EBUSY;
2013 			goto out;
2014 		}
2015 	} else {
2016 		mutex_lock(&sbi->gc_mutex);
2017 	}
2018 
2019 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2020 out:
2021 	mnt_drop_write_file(filp);
2022 	return ret;
2023 }
2024 
2025 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2026 {
2027 	struct inode *inode = file_inode(filp);
2028 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2029 	struct f2fs_gc_range range;
2030 	u64 end;
2031 	int ret;
2032 
2033 	if (!capable(CAP_SYS_ADMIN))
2034 		return -EPERM;
2035 
2036 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2037 							sizeof(range)))
2038 		return -EFAULT;
2039 
2040 	if (f2fs_readonly(sbi->sb))
2041 		return -EROFS;
2042 
2043 	ret = mnt_want_write_file(filp);
2044 	if (ret)
2045 		return ret;
2046 
2047 	end = range.start + range.len;
2048 	if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi))
2049 		return -EINVAL;
2050 do_more:
2051 	if (!range.sync) {
2052 		if (!mutex_trylock(&sbi->gc_mutex)) {
2053 			ret = -EBUSY;
2054 			goto out;
2055 		}
2056 	} else {
2057 		mutex_lock(&sbi->gc_mutex);
2058 	}
2059 
2060 	ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2061 	range.start += sbi->blocks_per_seg;
2062 	if (range.start <= end)
2063 		goto do_more;
2064 out:
2065 	mnt_drop_write_file(filp);
2066 	return ret;
2067 }
2068 
2069 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2070 {
2071 	struct inode *inode = file_inode(filp);
2072 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2073 	int ret;
2074 
2075 	if (!capable(CAP_SYS_ADMIN))
2076 		return -EPERM;
2077 
2078 	if (f2fs_readonly(sbi->sb))
2079 		return -EROFS;
2080 
2081 	ret = mnt_want_write_file(filp);
2082 	if (ret)
2083 		return ret;
2084 
2085 	ret = f2fs_sync_fs(sbi->sb, 1);
2086 
2087 	mnt_drop_write_file(filp);
2088 	return ret;
2089 }
2090 
2091 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2092 					struct file *filp,
2093 					struct f2fs_defragment *range)
2094 {
2095 	struct inode *inode = file_inode(filp);
2096 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2097 					.m_seg_type = NO_CHECK_TYPE };
2098 	struct extent_info ei = {0,0,0};
2099 	pgoff_t pg_start, pg_end, next_pgofs;
2100 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2101 	unsigned int total = 0, sec_num;
2102 	block_t blk_end = 0;
2103 	bool fragmented = false;
2104 	int err;
2105 
2106 	/* if in-place-update policy is enabled, don't waste time here */
2107 	if (should_update_inplace(inode, NULL))
2108 		return -EINVAL;
2109 
2110 	pg_start = range->start >> PAGE_SHIFT;
2111 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2112 
2113 	f2fs_balance_fs(sbi, true);
2114 
2115 	inode_lock(inode);
2116 
2117 	/* writeback all dirty pages in the range */
2118 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2119 						range->start + range->len - 1);
2120 	if (err)
2121 		goto out;
2122 
2123 	/*
2124 	 * lookup mapping info in extent cache, skip defragmenting if physical
2125 	 * block addresses are continuous.
2126 	 */
2127 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2128 		if (ei.fofs + ei.len >= pg_end)
2129 			goto out;
2130 	}
2131 
2132 	map.m_lblk = pg_start;
2133 	map.m_next_pgofs = &next_pgofs;
2134 
2135 	/*
2136 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2137 	 * physical block addresses are continuous even if there are hole(s)
2138 	 * in logical blocks.
2139 	 */
2140 	while (map.m_lblk < pg_end) {
2141 		map.m_len = pg_end - map.m_lblk;
2142 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2143 		if (err)
2144 			goto out;
2145 
2146 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2147 			map.m_lblk = next_pgofs;
2148 			continue;
2149 		}
2150 
2151 		if (blk_end && blk_end != map.m_pblk)
2152 			fragmented = true;
2153 
2154 		/* record total count of block that we're going to move */
2155 		total += map.m_len;
2156 
2157 		blk_end = map.m_pblk + map.m_len;
2158 
2159 		map.m_lblk += map.m_len;
2160 	}
2161 
2162 	if (!fragmented)
2163 		goto out;
2164 
2165 	sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2166 
2167 	/*
2168 	 * make sure there are enough free section for LFS allocation, this can
2169 	 * avoid defragment running in SSR mode when free section are allocated
2170 	 * intensively
2171 	 */
2172 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2173 		err = -EAGAIN;
2174 		goto out;
2175 	}
2176 
2177 	map.m_lblk = pg_start;
2178 	map.m_len = pg_end - pg_start;
2179 	total = 0;
2180 
2181 	while (map.m_lblk < pg_end) {
2182 		pgoff_t idx;
2183 		int cnt = 0;
2184 
2185 do_map:
2186 		map.m_len = pg_end - map.m_lblk;
2187 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2188 		if (err)
2189 			goto clear_out;
2190 
2191 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2192 			map.m_lblk = next_pgofs;
2193 			continue;
2194 		}
2195 
2196 		set_inode_flag(inode, FI_DO_DEFRAG);
2197 
2198 		idx = map.m_lblk;
2199 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2200 			struct page *page;
2201 
2202 			page = get_lock_data_page(inode, idx, true);
2203 			if (IS_ERR(page)) {
2204 				err = PTR_ERR(page);
2205 				goto clear_out;
2206 			}
2207 
2208 			set_page_dirty(page);
2209 			f2fs_put_page(page, 1);
2210 
2211 			idx++;
2212 			cnt++;
2213 			total++;
2214 		}
2215 
2216 		map.m_lblk = idx;
2217 
2218 		if (idx < pg_end && cnt < blk_per_seg)
2219 			goto do_map;
2220 
2221 		clear_inode_flag(inode, FI_DO_DEFRAG);
2222 
2223 		err = filemap_fdatawrite(inode->i_mapping);
2224 		if (err)
2225 			goto out;
2226 	}
2227 clear_out:
2228 	clear_inode_flag(inode, FI_DO_DEFRAG);
2229 out:
2230 	inode_unlock(inode);
2231 	if (!err)
2232 		range->len = (u64)total << PAGE_SHIFT;
2233 	return err;
2234 }
2235 
2236 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2237 {
2238 	struct inode *inode = file_inode(filp);
2239 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2240 	struct f2fs_defragment range;
2241 	int err;
2242 
2243 	if (!capable(CAP_SYS_ADMIN))
2244 		return -EPERM;
2245 
2246 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2247 		return -EINVAL;
2248 
2249 	if (f2fs_readonly(sbi->sb))
2250 		return -EROFS;
2251 
2252 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2253 							sizeof(range)))
2254 		return -EFAULT;
2255 
2256 	/* verify alignment of offset & size */
2257 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2258 		return -EINVAL;
2259 
2260 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2261 					sbi->max_file_blocks))
2262 		return -EINVAL;
2263 
2264 	err = mnt_want_write_file(filp);
2265 	if (err)
2266 		return err;
2267 
2268 	err = f2fs_defragment_range(sbi, filp, &range);
2269 	mnt_drop_write_file(filp);
2270 
2271 	f2fs_update_time(sbi, REQ_TIME);
2272 	if (err < 0)
2273 		return err;
2274 
2275 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2276 							sizeof(range)))
2277 		return -EFAULT;
2278 
2279 	return 0;
2280 }
2281 
2282 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2283 			struct file *file_out, loff_t pos_out, size_t len)
2284 {
2285 	struct inode *src = file_inode(file_in);
2286 	struct inode *dst = file_inode(file_out);
2287 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2288 	size_t olen = len, dst_max_i_size = 0;
2289 	size_t dst_osize;
2290 	int ret;
2291 
2292 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2293 				src->i_sb != dst->i_sb)
2294 		return -EXDEV;
2295 
2296 	if (unlikely(f2fs_readonly(src->i_sb)))
2297 		return -EROFS;
2298 
2299 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2300 		return -EINVAL;
2301 
2302 	if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2303 		return -EOPNOTSUPP;
2304 
2305 	if (src == dst) {
2306 		if (pos_in == pos_out)
2307 			return 0;
2308 		if (pos_out > pos_in && pos_out < pos_in + len)
2309 			return -EINVAL;
2310 	}
2311 
2312 	inode_lock(src);
2313 	down_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2314 	if (src != dst) {
2315 		ret = -EBUSY;
2316 		if (!inode_trylock(dst))
2317 			goto out;
2318 		if (!down_write_trylock(&F2FS_I(dst)->dio_rwsem[WRITE])) {
2319 			inode_unlock(dst);
2320 			goto out;
2321 		}
2322 	}
2323 
2324 	ret = -EINVAL;
2325 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2326 		goto out_unlock;
2327 	if (len == 0)
2328 		olen = len = src->i_size - pos_in;
2329 	if (pos_in + len == src->i_size)
2330 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2331 	if (len == 0) {
2332 		ret = 0;
2333 		goto out_unlock;
2334 	}
2335 
2336 	dst_osize = dst->i_size;
2337 	if (pos_out + olen > dst->i_size)
2338 		dst_max_i_size = pos_out + olen;
2339 
2340 	/* verify the end result is block aligned */
2341 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2342 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2343 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2344 		goto out_unlock;
2345 
2346 	ret = f2fs_convert_inline_inode(src);
2347 	if (ret)
2348 		goto out_unlock;
2349 
2350 	ret = f2fs_convert_inline_inode(dst);
2351 	if (ret)
2352 		goto out_unlock;
2353 
2354 	/* write out all dirty pages from offset */
2355 	ret = filemap_write_and_wait_range(src->i_mapping,
2356 					pos_in, pos_in + len);
2357 	if (ret)
2358 		goto out_unlock;
2359 
2360 	ret = filemap_write_and_wait_range(dst->i_mapping,
2361 					pos_out, pos_out + len);
2362 	if (ret)
2363 		goto out_unlock;
2364 
2365 	f2fs_balance_fs(sbi, true);
2366 	f2fs_lock_op(sbi);
2367 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2368 				pos_out >> F2FS_BLKSIZE_BITS,
2369 				len >> F2FS_BLKSIZE_BITS, false);
2370 
2371 	if (!ret) {
2372 		if (dst_max_i_size)
2373 			f2fs_i_size_write(dst, dst_max_i_size);
2374 		else if (dst_osize != dst->i_size)
2375 			f2fs_i_size_write(dst, dst_osize);
2376 	}
2377 	f2fs_unlock_op(sbi);
2378 out_unlock:
2379 	if (src != dst) {
2380 		up_write(&F2FS_I(dst)->dio_rwsem[WRITE]);
2381 		inode_unlock(dst);
2382 	}
2383 out:
2384 	up_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2385 	inode_unlock(src);
2386 	return ret;
2387 }
2388 
2389 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2390 {
2391 	struct f2fs_move_range range;
2392 	struct fd dst;
2393 	int err;
2394 
2395 	if (!(filp->f_mode & FMODE_READ) ||
2396 			!(filp->f_mode & FMODE_WRITE))
2397 		return -EBADF;
2398 
2399 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2400 							sizeof(range)))
2401 		return -EFAULT;
2402 
2403 	dst = fdget(range.dst_fd);
2404 	if (!dst.file)
2405 		return -EBADF;
2406 
2407 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2408 		err = -EBADF;
2409 		goto err_out;
2410 	}
2411 
2412 	err = mnt_want_write_file(filp);
2413 	if (err)
2414 		goto err_out;
2415 
2416 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2417 					range.pos_out, range.len);
2418 
2419 	mnt_drop_write_file(filp);
2420 	if (err)
2421 		goto err_out;
2422 
2423 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2424 						&range, sizeof(range)))
2425 		err = -EFAULT;
2426 err_out:
2427 	fdput(dst);
2428 	return err;
2429 }
2430 
2431 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2432 {
2433 	struct inode *inode = file_inode(filp);
2434 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2435 	struct sit_info *sm = SIT_I(sbi);
2436 	unsigned int start_segno = 0, end_segno = 0;
2437 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2438 	struct f2fs_flush_device range;
2439 	int ret;
2440 
2441 	if (!capable(CAP_SYS_ADMIN))
2442 		return -EPERM;
2443 
2444 	if (f2fs_readonly(sbi->sb))
2445 		return -EROFS;
2446 
2447 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2448 							sizeof(range)))
2449 		return -EFAULT;
2450 
2451 	if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2452 			sbi->segs_per_sec != 1) {
2453 		f2fs_msg(sbi->sb, KERN_WARNING,
2454 			"Can't flush %u in %d for segs_per_sec %u != 1\n",
2455 				range.dev_num, sbi->s_ndevs,
2456 				sbi->segs_per_sec);
2457 		return -EINVAL;
2458 	}
2459 
2460 	ret = mnt_want_write_file(filp);
2461 	if (ret)
2462 		return ret;
2463 
2464 	if (range.dev_num != 0)
2465 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2466 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2467 
2468 	start_segno = sm->last_victim[FLUSH_DEVICE];
2469 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2470 		start_segno = dev_start_segno;
2471 	end_segno = min(start_segno + range.segments, dev_end_segno);
2472 
2473 	while (start_segno < end_segno) {
2474 		if (!mutex_trylock(&sbi->gc_mutex)) {
2475 			ret = -EBUSY;
2476 			goto out;
2477 		}
2478 		sm->last_victim[GC_CB] = end_segno + 1;
2479 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2480 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2481 		ret = f2fs_gc(sbi, true, true, start_segno);
2482 		if (ret == -EAGAIN)
2483 			ret = 0;
2484 		else if (ret < 0)
2485 			break;
2486 		start_segno++;
2487 	}
2488 out:
2489 	mnt_drop_write_file(filp);
2490 	return ret;
2491 }
2492 
2493 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2494 {
2495 	struct inode *inode = file_inode(filp);
2496 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2497 
2498 	/* Must validate to set it with SQLite behavior in Android. */
2499 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2500 
2501 	return put_user(sb_feature, (u32 __user *)arg);
2502 }
2503 
2504 #ifdef CONFIG_QUOTA
2505 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2506 {
2507 	struct inode *inode = file_inode(filp);
2508 	struct f2fs_inode_info *fi = F2FS_I(inode);
2509 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2510 	struct super_block *sb = sbi->sb;
2511 	struct dquot *transfer_to[MAXQUOTAS] = {};
2512 	struct page *ipage;
2513 	kprojid_t kprojid;
2514 	int err;
2515 
2516 	if (!f2fs_sb_has_project_quota(sb)) {
2517 		if (projid != F2FS_DEF_PROJID)
2518 			return -EOPNOTSUPP;
2519 		else
2520 			return 0;
2521 	}
2522 
2523 	if (!f2fs_has_extra_attr(inode))
2524 		return -EOPNOTSUPP;
2525 
2526 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2527 
2528 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2529 		return 0;
2530 
2531 	err = mnt_want_write_file(filp);
2532 	if (err)
2533 		return err;
2534 
2535 	err = -EPERM;
2536 	inode_lock(inode);
2537 
2538 	/* Is it quota file? Do not allow user to mess with it */
2539 	if (IS_NOQUOTA(inode))
2540 		goto out_unlock;
2541 
2542 	ipage = get_node_page(sbi, inode->i_ino);
2543 	if (IS_ERR(ipage)) {
2544 		err = PTR_ERR(ipage);
2545 		goto out_unlock;
2546 	}
2547 
2548 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2549 								i_projid)) {
2550 		err = -EOVERFLOW;
2551 		f2fs_put_page(ipage, 1);
2552 		goto out_unlock;
2553 	}
2554 	f2fs_put_page(ipage, 1);
2555 
2556 	dquot_initialize(inode);
2557 
2558 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2559 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2560 		err = __dquot_transfer(inode, transfer_to);
2561 		dqput(transfer_to[PRJQUOTA]);
2562 		if (err)
2563 			goto out_dirty;
2564 	}
2565 
2566 	F2FS_I(inode)->i_projid = kprojid;
2567 	inode->i_ctime = current_time(inode);
2568 out_dirty:
2569 	f2fs_mark_inode_dirty_sync(inode, true);
2570 out_unlock:
2571 	inode_unlock(inode);
2572 	mnt_drop_write_file(filp);
2573 	return err;
2574 }
2575 #else
2576 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2577 {
2578 	if (projid != F2FS_DEF_PROJID)
2579 		return -EOPNOTSUPP;
2580 	return 0;
2581 }
2582 #endif
2583 
2584 /* Transfer internal flags to xflags */
2585 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2586 {
2587 	__u32 xflags = 0;
2588 
2589 	if (iflags & FS_SYNC_FL)
2590 		xflags |= FS_XFLAG_SYNC;
2591 	if (iflags & FS_IMMUTABLE_FL)
2592 		xflags |= FS_XFLAG_IMMUTABLE;
2593 	if (iflags & FS_APPEND_FL)
2594 		xflags |= FS_XFLAG_APPEND;
2595 	if (iflags & FS_NODUMP_FL)
2596 		xflags |= FS_XFLAG_NODUMP;
2597 	if (iflags & FS_NOATIME_FL)
2598 		xflags |= FS_XFLAG_NOATIME;
2599 	if (iflags & FS_PROJINHERIT_FL)
2600 		xflags |= FS_XFLAG_PROJINHERIT;
2601 	return xflags;
2602 }
2603 
2604 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2605 				  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2606 				  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2607 
2608 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2609 #define F2FS_FL_XFLAG_VISIBLE		(FS_SYNC_FL | \
2610 					 FS_IMMUTABLE_FL | \
2611 					 FS_APPEND_FL | \
2612 					 FS_NODUMP_FL | \
2613 					 FS_NOATIME_FL | \
2614 					 FS_PROJINHERIT_FL)
2615 
2616 /* Transfer xflags flags to internal */
2617 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2618 {
2619 	unsigned long iflags = 0;
2620 
2621 	if (xflags & FS_XFLAG_SYNC)
2622 		iflags |= FS_SYNC_FL;
2623 	if (xflags & FS_XFLAG_IMMUTABLE)
2624 		iflags |= FS_IMMUTABLE_FL;
2625 	if (xflags & FS_XFLAG_APPEND)
2626 		iflags |= FS_APPEND_FL;
2627 	if (xflags & FS_XFLAG_NODUMP)
2628 		iflags |= FS_NODUMP_FL;
2629 	if (xflags & FS_XFLAG_NOATIME)
2630 		iflags |= FS_NOATIME_FL;
2631 	if (xflags & FS_XFLAG_PROJINHERIT)
2632 		iflags |= FS_PROJINHERIT_FL;
2633 
2634 	return iflags;
2635 }
2636 
2637 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2638 {
2639 	struct inode *inode = file_inode(filp);
2640 	struct f2fs_inode_info *fi = F2FS_I(inode);
2641 	struct fsxattr fa;
2642 
2643 	memset(&fa, 0, sizeof(struct fsxattr));
2644 	fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2645 				(FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2646 
2647 	if (f2fs_sb_has_project_quota(inode->i_sb))
2648 		fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2649 							fi->i_projid);
2650 
2651 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2652 		return -EFAULT;
2653 	return 0;
2654 }
2655 
2656 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2657 {
2658 	struct inode *inode = file_inode(filp);
2659 	struct f2fs_inode_info *fi = F2FS_I(inode);
2660 	struct fsxattr fa;
2661 	unsigned int flags;
2662 	int err;
2663 
2664 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2665 		return -EFAULT;
2666 
2667 	/* Make sure caller has proper permission */
2668 	if (!inode_owner_or_capable(inode))
2669 		return -EACCES;
2670 
2671 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2672 		return -EOPNOTSUPP;
2673 
2674 	flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2675 	if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2676 		return -EOPNOTSUPP;
2677 
2678 	err = mnt_want_write_file(filp);
2679 	if (err)
2680 		return err;
2681 
2682 	inode_lock(inode);
2683 	flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2684 				(flags & F2FS_FL_XFLAG_VISIBLE);
2685 	err = __f2fs_ioc_setflags(inode, flags);
2686 	inode_unlock(inode);
2687 	mnt_drop_write_file(filp);
2688 	if (err)
2689 		return err;
2690 
2691 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2692 	if (err)
2693 		return err;
2694 
2695 	return 0;
2696 }
2697 
2698 int f2fs_pin_file_control(struct inode *inode, bool inc)
2699 {
2700 	struct f2fs_inode_info *fi = F2FS_I(inode);
2701 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2702 
2703 	/* Use i_gc_failures for normal file as a risk signal. */
2704 	if (inc)
2705 		f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
2706 
2707 	if (fi->i_gc_failures > sbi->gc_pin_file_threshold) {
2708 		f2fs_msg(sbi->sb, KERN_WARNING,
2709 			"%s: Enable GC = ino %lx after %x GC trials\n",
2710 			__func__, inode->i_ino, fi->i_gc_failures);
2711 		clear_inode_flag(inode, FI_PIN_FILE);
2712 		return -EAGAIN;
2713 	}
2714 	return 0;
2715 }
2716 
2717 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2718 {
2719 	struct inode *inode = file_inode(filp);
2720 	__u32 pin;
2721 	int ret = 0;
2722 
2723 	if (!inode_owner_or_capable(inode))
2724 		return -EACCES;
2725 
2726 	if (get_user(pin, (__u32 __user *)arg))
2727 		return -EFAULT;
2728 
2729 	if (!S_ISREG(inode->i_mode))
2730 		return -EINVAL;
2731 
2732 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2733 		return -EROFS;
2734 
2735 	ret = mnt_want_write_file(filp);
2736 	if (ret)
2737 		return ret;
2738 
2739 	inode_lock(inode);
2740 
2741 	if (should_update_outplace(inode, NULL)) {
2742 		ret = -EINVAL;
2743 		goto out;
2744 	}
2745 
2746 	if (!pin) {
2747 		clear_inode_flag(inode, FI_PIN_FILE);
2748 		F2FS_I(inode)->i_gc_failures = 1;
2749 		goto done;
2750 	}
2751 
2752 	if (f2fs_pin_file_control(inode, false)) {
2753 		ret = -EAGAIN;
2754 		goto out;
2755 	}
2756 	ret = f2fs_convert_inline_inode(inode);
2757 	if (ret)
2758 		goto out;
2759 
2760 	set_inode_flag(inode, FI_PIN_FILE);
2761 	ret = F2FS_I(inode)->i_gc_failures;
2762 done:
2763 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2764 out:
2765 	inode_unlock(inode);
2766 	mnt_drop_write_file(filp);
2767 	return ret;
2768 }
2769 
2770 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2771 {
2772 	struct inode *inode = file_inode(filp);
2773 	__u32 pin = 0;
2774 
2775 	if (is_inode_flag_set(inode, FI_PIN_FILE))
2776 		pin = F2FS_I(inode)->i_gc_failures;
2777 	return put_user(pin, (u32 __user *)arg);
2778 }
2779 
2780 int f2fs_precache_extents(struct inode *inode)
2781 {
2782 	struct f2fs_inode_info *fi = F2FS_I(inode);
2783 	struct f2fs_map_blocks map;
2784 	pgoff_t m_next_extent;
2785 	loff_t end;
2786 	int err;
2787 
2788 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
2789 		return -EOPNOTSUPP;
2790 
2791 	map.m_lblk = 0;
2792 	map.m_next_pgofs = NULL;
2793 	map.m_next_extent = &m_next_extent;
2794 	map.m_seg_type = NO_CHECK_TYPE;
2795 	end = F2FS_I_SB(inode)->max_file_blocks;
2796 
2797 	while (map.m_lblk < end) {
2798 		map.m_len = end - map.m_lblk;
2799 
2800 		down_write(&fi->dio_rwsem[WRITE]);
2801 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2802 		up_write(&fi->dio_rwsem[WRITE]);
2803 		if (err)
2804 			return err;
2805 
2806 		map.m_lblk = m_next_extent;
2807 	}
2808 
2809 	return err;
2810 }
2811 
2812 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2813 {
2814 	return f2fs_precache_extents(file_inode(filp));
2815 }
2816 
2817 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2818 {
2819 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2820 		return -EIO;
2821 
2822 	switch (cmd) {
2823 	case F2FS_IOC_GETFLAGS:
2824 		return f2fs_ioc_getflags(filp, arg);
2825 	case F2FS_IOC_SETFLAGS:
2826 		return f2fs_ioc_setflags(filp, arg);
2827 	case F2FS_IOC_GETVERSION:
2828 		return f2fs_ioc_getversion(filp, arg);
2829 	case F2FS_IOC_START_ATOMIC_WRITE:
2830 		return f2fs_ioc_start_atomic_write(filp);
2831 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2832 		return f2fs_ioc_commit_atomic_write(filp);
2833 	case F2FS_IOC_START_VOLATILE_WRITE:
2834 		return f2fs_ioc_start_volatile_write(filp);
2835 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2836 		return f2fs_ioc_release_volatile_write(filp);
2837 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
2838 		return f2fs_ioc_abort_volatile_write(filp);
2839 	case F2FS_IOC_SHUTDOWN:
2840 		return f2fs_ioc_shutdown(filp, arg);
2841 	case FITRIM:
2842 		return f2fs_ioc_fitrim(filp, arg);
2843 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
2844 		return f2fs_ioc_set_encryption_policy(filp, arg);
2845 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
2846 		return f2fs_ioc_get_encryption_policy(filp, arg);
2847 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2848 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2849 	case F2FS_IOC_GARBAGE_COLLECT:
2850 		return f2fs_ioc_gc(filp, arg);
2851 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2852 		return f2fs_ioc_gc_range(filp, arg);
2853 	case F2FS_IOC_WRITE_CHECKPOINT:
2854 		return f2fs_ioc_write_checkpoint(filp, arg);
2855 	case F2FS_IOC_DEFRAGMENT:
2856 		return f2fs_ioc_defragment(filp, arg);
2857 	case F2FS_IOC_MOVE_RANGE:
2858 		return f2fs_ioc_move_range(filp, arg);
2859 	case F2FS_IOC_FLUSH_DEVICE:
2860 		return f2fs_ioc_flush_device(filp, arg);
2861 	case F2FS_IOC_GET_FEATURES:
2862 		return f2fs_ioc_get_features(filp, arg);
2863 	case F2FS_IOC_FSGETXATTR:
2864 		return f2fs_ioc_fsgetxattr(filp, arg);
2865 	case F2FS_IOC_FSSETXATTR:
2866 		return f2fs_ioc_fssetxattr(filp, arg);
2867 	case F2FS_IOC_GET_PIN_FILE:
2868 		return f2fs_ioc_get_pin_file(filp, arg);
2869 	case F2FS_IOC_SET_PIN_FILE:
2870 		return f2fs_ioc_set_pin_file(filp, arg);
2871 	case F2FS_IOC_PRECACHE_EXTENTS:
2872 		return f2fs_ioc_precache_extents(filp, arg);
2873 	default:
2874 		return -ENOTTY;
2875 	}
2876 }
2877 
2878 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2879 {
2880 	struct file *file = iocb->ki_filp;
2881 	struct inode *inode = file_inode(file);
2882 	struct blk_plug plug;
2883 	ssize_t ret;
2884 
2885 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2886 		return -EIO;
2887 
2888 	inode_lock(inode);
2889 	ret = generic_write_checks(iocb, from);
2890 	if (ret > 0) {
2891 		int err;
2892 
2893 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2894 			set_inode_flag(inode, FI_NO_PREALLOC);
2895 
2896 		err = f2fs_preallocate_blocks(iocb, from);
2897 		if (err) {
2898 			clear_inode_flag(inode, FI_NO_PREALLOC);
2899 			inode_unlock(inode);
2900 			return err;
2901 		}
2902 		blk_start_plug(&plug);
2903 		ret = __generic_file_write_iter(iocb, from);
2904 		blk_finish_plug(&plug);
2905 		clear_inode_flag(inode, FI_NO_PREALLOC);
2906 
2907 		if (ret > 0)
2908 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2909 	}
2910 	inode_unlock(inode);
2911 
2912 	if (ret > 0)
2913 		ret = generic_write_sync(iocb, ret);
2914 	return ret;
2915 }
2916 
2917 #ifdef CONFIG_COMPAT
2918 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2919 {
2920 	switch (cmd) {
2921 	case F2FS_IOC32_GETFLAGS:
2922 		cmd = F2FS_IOC_GETFLAGS;
2923 		break;
2924 	case F2FS_IOC32_SETFLAGS:
2925 		cmd = F2FS_IOC_SETFLAGS;
2926 		break;
2927 	case F2FS_IOC32_GETVERSION:
2928 		cmd = F2FS_IOC_GETVERSION;
2929 		break;
2930 	case F2FS_IOC_START_ATOMIC_WRITE:
2931 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2932 	case F2FS_IOC_START_VOLATILE_WRITE:
2933 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2934 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
2935 	case F2FS_IOC_SHUTDOWN:
2936 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
2937 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2938 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
2939 	case F2FS_IOC_GARBAGE_COLLECT:
2940 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2941 	case F2FS_IOC_WRITE_CHECKPOINT:
2942 	case F2FS_IOC_DEFRAGMENT:
2943 	case F2FS_IOC_MOVE_RANGE:
2944 	case F2FS_IOC_FLUSH_DEVICE:
2945 	case F2FS_IOC_GET_FEATURES:
2946 	case F2FS_IOC_FSGETXATTR:
2947 	case F2FS_IOC_FSSETXATTR:
2948 	case F2FS_IOC_GET_PIN_FILE:
2949 	case F2FS_IOC_SET_PIN_FILE:
2950 	case F2FS_IOC_PRECACHE_EXTENTS:
2951 		break;
2952 	default:
2953 		return -ENOIOCTLCMD;
2954 	}
2955 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2956 }
2957 #endif
2958 
2959 const struct file_operations f2fs_file_operations = {
2960 	.llseek		= f2fs_llseek,
2961 	.read_iter	= generic_file_read_iter,
2962 	.write_iter	= f2fs_file_write_iter,
2963 	.open		= f2fs_file_open,
2964 	.release	= f2fs_release_file,
2965 	.mmap		= f2fs_file_mmap,
2966 	.flush		= f2fs_file_flush,
2967 	.fsync		= f2fs_sync_file,
2968 	.fallocate	= f2fs_fallocate,
2969 	.unlocked_ioctl	= f2fs_ioctl,
2970 #ifdef CONFIG_COMPAT
2971 	.compat_ioctl	= f2fs_compat_ioctl,
2972 #endif
2973 	.splice_read	= generic_file_splice_read,
2974 	.splice_write	= iter_file_splice_write,
2975 };
2976