xref: /openbmc/linux/fs/f2fs/file.c (revision bc5aa3a0)
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/uuid.h>
24 #include <linux/file.h>
25 
26 #include "f2fs.h"
27 #include "node.h"
28 #include "segment.h"
29 #include "xattr.h"
30 #include "acl.h"
31 #include "gc.h"
32 #include "trace.h"
33 #include <trace/events/f2fs.h>
34 
35 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
36 						struct vm_fault *vmf)
37 {
38 	struct page *page = vmf->page;
39 	struct inode *inode = file_inode(vma->vm_file);
40 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
41 	struct dnode_of_data dn;
42 	int err;
43 
44 	sb_start_pagefault(inode->i_sb);
45 
46 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
47 
48 	/* block allocation */
49 	f2fs_lock_op(sbi);
50 	set_new_dnode(&dn, inode, NULL, NULL, 0);
51 	err = f2fs_reserve_block(&dn, page->index);
52 	if (err) {
53 		f2fs_unlock_op(sbi);
54 		goto out;
55 	}
56 	f2fs_put_dnode(&dn);
57 	f2fs_unlock_op(sbi);
58 
59 	f2fs_balance_fs(sbi, dn.node_changed);
60 
61 	file_update_time(vma->vm_file);
62 	lock_page(page);
63 	if (unlikely(page->mapping != inode->i_mapping ||
64 			page_offset(page) > i_size_read(inode) ||
65 			!PageUptodate(page))) {
66 		unlock_page(page);
67 		err = -EFAULT;
68 		goto out;
69 	}
70 
71 	/*
72 	 * check to see if the page is mapped already (no holes)
73 	 */
74 	if (PageMappedToDisk(page))
75 		goto mapped;
76 
77 	/* page is wholly or partially inside EOF */
78 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
79 						i_size_read(inode)) {
80 		unsigned offset;
81 		offset = i_size_read(inode) & ~PAGE_MASK;
82 		zero_user_segment(page, offset, PAGE_SIZE);
83 	}
84 	set_page_dirty(page);
85 	if (!PageUptodate(page))
86 		SetPageUptodate(page);
87 
88 	trace_f2fs_vm_page_mkwrite(page, DATA);
89 mapped:
90 	/* fill the page */
91 	f2fs_wait_on_page_writeback(page, DATA, false);
92 
93 	/* wait for GCed encrypted page writeback */
94 	if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
95 		f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
96 
97 	/* if gced page is attached, don't write to cold segment */
98 	clear_cold_data(page);
99 out:
100 	sb_end_pagefault(inode->i_sb);
101 	f2fs_update_time(sbi, REQ_TIME);
102 	return block_page_mkwrite_return(err);
103 }
104 
105 static const struct vm_operations_struct f2fs_file_vm_ops = {
106 	.fault		= filemap_fault,
107 	.map_pages	= filemap_map_pages,
108 	.page_mkwrite	= f2fs_vm_page_mkwrite,
109 };
110 
111 static int get_parent_ino(struct inode *inode, nid_t *pino)
112 {
113 	struct dentry *dentry;
114 
115 	inode = igrab(inode);
116 	dentry = d_find_any_alias(inode);
117 	iput(inode);
118 	if (!dentry)
119 		return 0;
120 
121 	if (update_dent_inode(inode, inode, &dentry->d_name)) {
122 		dput(dentry);
123 		return 0;
124 	}
125 
126 	*pino = parent_ino(dentry);
127 	dput(dentry);
128 	return 1;
129 }
130 
131 static inline bool need_do_checkpoint(struct inode *inode)
132 {
133 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
134 	bool need_cp = false;
135 
136 	if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
137 		need_cp = true;
138 	else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
139 		need_cp = true;
140 	else if (file_wrong_pino(inode))
141 		need_cp = true;
142 	else if (!space_for_roll_forward(sbi))
143 		need_cp = true;
144 	else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
145 		need_cp = true;
146 	else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
147 		need_cp = true;
148 	else if (test_opt(sbi, FASTBOOT))
149 		need_cp = true;
150 	else if (sbi->active_logs == 2)
151 		need_cp = true;
152 
153 	return need_cp;
154 }
155 
156 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
157 {
158 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
159 	bool ret = false;
160 	/* But we need to avoid that there are some inode updates */
161 	if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
162 		ret = true;
163 	f2fs_put_page(i, 0);
164 	return ret;
165 }
166 
167 static void try_to_fix_pino(struct inode *inode)
168 {
169 	struct f2fs_inode_info *fi = F2FS_I(inode);
170 	nid_t pino;
171 
172 	down_write(&fi->i_sem);
173 	fi->xattr_ver = 0;
174 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
175 			get_parent_ino(inode, &pino)) {
176 		f2fs_i_pino_write(inode, pino);
177 		file_got_pino(inode);
178 	}
179 	up_write(&fi->i_sem);
180 }
181 
182 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
183 						int datasync, bool atomic)
184 {
185 	struct inode *inode = file->f_mapping->host;
186 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
187 	nid_t ino = inode->i_ino;
188 	int ret = 0;
189 	bool need_cp = false;
190 	struct writeback_control wbc = {
191 		.sync_mode = WB_SYNC_ALL,
192 		.nr_to_write = LONG_MAX,
193 		.for_reclaim = 0,
194 	};
195 
196 	if (unlikely(f2fs_readonly(inode->i_sb)))
197 		return 0;
198 
199 	trace_f2fs_sync_file_enter(inode);
200 
201 	/* if fdatasync is triggered, let's do in-place-update */
202 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
203 		set_inode_flag(inode, FI_NEED_IPU);
204 	ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
205 	clear_inode_flag(inode, FI_NEED_IPU);
206 
207 	if (ret) {
208 		trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
209 		return ret;
210 	}
211 
212 	/* if the inode is dirty, let's recover all the time */
213 	if (!datasync && !f2fs_skip_inode_update(inode)) {
214 		f2fs_write_inode(inode, NULL);
215 		goto go_write;
216 	}
217 
218 	/*
219 	 * if there is no written data, don't waste time to write recovery info.
220 	 */
221 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
222 			!exist_written_data(sbi, ino, APPEND_INO)) {
223 
224 		/* it may call write_inode just prior to fsync */
225 		if (need_inode_page_update(sbi, ino))
226 			goto go_write;
227 
228 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
229 				exist_written_data(sbi, ino, UPDATE_INO))
230 			goto flush_out;
231 		goto out;
232 	}
233 go_write:
234 	/*
235 	 * Both of fdatasync() and fsync() are able to be recovered from
236 	 * sudden-power-off.
237 	 */
238 	down_read(&F2FS_I(inode)->i_sem);
239 	need_cp = need_do_checkpoint(inode);
240 	up_read(&F2FS_I(inode)->i_sem);
241 
242 	if (need_cp) {
243 		/* all the dirty node pages should be flushed for POR */
244 		ret = f2fs_sync_fs(inode->i_sb, 1);
245 
246 		/*
247 		 * We've secured consistency through sync_fs. Following pino
248 		 * will be used only for fsynced inodes after checkpoint.
249 		 */
250 		try_to_fix_pino(inode);
251 		clear_inode_flag(inode, FI_APPEND_WRITE);
252 		clear_inode_flag(inode, FI_UPDATE_WRITE);
253 		goto out;
254 	}
255 sync_nodes:
256 	ret = fsync_node_pages(sbi, inode, &wbc, atomic);
257 	if (ret)
258 		goto out;
259 
260 	/* if cp_error was enabled, we should avoid infinite loop */
261 	if (unlikely(f2fs_cp_error(sbi))) {
262 		ret = -EIO;
263 		goto out;
264 	}
265 
266 	if (need_inode_block_update(sbi, ino)) {
267 		f2fs_mark_inode_dirty_sync(inode);
268 		f2fs_write_inode(inode, NULL);
269 		goto sync_nodes;
270 	}
271 
272 	ret = wait_on_node_pages_writeback(sbi, ino);
273 	if (ret)
274 		goto out;
275 
276 	/* once recovery info is written, don't need to tack this */
277 	remove_ino_entry(sbi, ino, APPEND_INO);
278 	clear_inode_flag(inode, FI_APPEND_WRITE);
279 flush_out:
280 	remove_ino_entry(sbi, ino, UPDATE_INO);
281 	clear_inode_flag(inode, FI_UPDATE_WRITE);
282 	ret = f2fs_issue_flush(sbi);
283 	f2fs_update_time(sbi, REQ_TIME);
284 out:
285 	trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
286 	f2fs_trace_ios(NULL, 1);
287 	return ret;
288 }
289 
290 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
291 {
292 	return f2fs_do_sync_file(file, start, end, datasync, false);
293 }
294 
295 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
296 						pgoff_t pgofs, int whence)
297 {
298 	struct pagevec pvec;
299 	int nr_pages;
300 
301 	if (whence != SEEK_DATA)
302 		return 0;
303 
304 	/* find first dirty page index */
305 	pagevec_init(&pvec, 0);
306 	nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
307 					PAGECACHE_TAG_DIRTY, 1);
308 	pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
309 	pagevec_release(&pvec);
310 	return pgofs;
311 }
312 
313 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
314 							int whence)
315 {
316 	switch (whence) {
317 	case SEEK_DATA:
318 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
319 			(blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
320 			return true;
321 		break;
322 	case SEEK_HOLE:
323 		if (blkaddr == NULL_ADDR)
324 			return true;
325 		break;
326 	}
327 	return false;
328 }
329 
330 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
331 {
332 	struct inode *inode = file->f_mapping->host;
333 	loff_t maxbytes = inode->i_sb->s_maxbytes;
334 	struct dnode_of_data dn;
335 	pgoff_t pgofs, end_offset, dirty;
336 	loff_t data_ofs = offset;
337 	loff_t isize;
338 	int err = 0;
339 
340 	inode_lock(inode);
341 
342 	isize = i_size_read(inode);
343 	if (offset >= isize)
344 		goto fail;
345 
346 	/* handle inline data case */
347 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
348 		if (whence == SEEK_HOLE)
349 			data_ofs = isize;
350 		goto found;
351 	}
352 
353 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
354 
355 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
356 
357 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
358 		set_new_dnode(&dn, inode, NULL, NULL, 0);
359 		err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
360 		if (err && err != -ENOENT) {
361 			goto fail;
362 		} else if (err == -ENOENT) {
363 			/* direct node does not exists */
364 			if (whence == SEEK_DATA) {
365 				pgofs = get_next_page_offset(&dn, pgofs);
366 				continue;
367 			} else {
368 				goto found;
369 			}
370 		}
371 
372 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
373 
374 		/* find data/hole in dnode block */
375 		for (; dn.ofs_in_node < end_offset;
376 				dn.ofs_in_node++, pgofs++,
377 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
378 			block_t blkaddr;
379 			blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
380 
381 			if (__found_offset(blkaddr, dirty, pgofs, whence)) {
382 				f2fs_put_dnode(&dn);
383 				goto found;
384 			}
385 		}
386 		f2fs_put_dnode(&dn);
387 	}
388 
389 	if (whence == SEEK_DATA)
390 		goto fail;
391 found:
392 	if (whence == SEEK_HOLE && data_ofs > isize)
393 		data_ofs = isize;
394 	inode_unlock(inode);
395 	return vfs_setpos(file, data_ofs, maxbytes);
396 fail:
397 	inode_unlock(inode);
398 	return -ENXIO;
399 }
400 
401 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
402 {
403 	struct inode *inode = file->f_mapping->host;
404 	loff_t maxbytes = inode->i_sb->s_maxbytes;
405 
406 	switch (whence) {
407 	case SEEK_SET:
408 	case SEEK_CUR:
409 	case SEEK_END:
410 		return generic_file_llseek_size(file, offset, whence,
411 						maxbytes, i_size_read(inode));
412 	case SEEK_DATA:
413 	case SEEK_HOLE:
414 		if (offset < 0)
415 			return -ENXIO;
416 		return f2fs_seek_block(file, offset, whence);
417 	}
418 
419 	return -EINVAL;
420 }
421 
422 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
423 {
424 	struct inode *inode = file_inode(file);
425 	int err;
426 
427 	if (f2fs_encrypted_inode(inode)) {
428 		err = fscrypt_get_encryption_info(inode);
429 		if (err)
430 			return 0;
431 		if (!f2fs_encrypted_inode(inode))
432 			return -ENOKEY;
433 	}
434 
435 	/* we don't need to use inline_data strictly */
436 	err = f2fs_convert_inline_inode(inode);
437 	if (err)
438 		return err;
439 
440 	file_accessed(file);
441 	vma->vm_ops = &f2fs_file_vm_ops;
442 	return 0;
443 }
444 
445 static int f2fs_file_open(struct inode *inode, struct file *filp)
446 {
447 	int ret = generic_file_open(inode, filp);
448 	struct dentry *dir;
449 
450 	if (!ret && f2fs_encrypted_inode(inode)) {
451 		ret = fscrypt_get_encryption_info(inode);
452 		if (ret)
453 			return -EACCES;
454 		if (!fscrypt_has_encryption_key(inode))
455 			return -ENOKEY;
456 	}
457 	dir = dget_parent(file_dentry(filp));
458 	if (f2fs_encrypted_inode(d_inode(dir)) &&
459 			!fscrypt_has_permitted_context(d_inode(dir), inode)) {
460 		dput(dir);
461 		return -EPERM;
462 	}
463 	dput(dir);
464 	return ret;
465 }
466 
467 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
468 {
469 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
470 	struct f2fs_node *raw_node;
471 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
472 	__le32 *addr;
473 
474 	raw_node = F2FS_NODE(dn->node_page);
475 	addr = blkaddr_in_node(raw_node) + ofs;
476 
477 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
478 		block_t blkaddr = le32_to_cpu(*addr);
479 		if (blkaddr == NULL_ADDR)
480 			continue;
481 
482 		dn->data_blkaddr = NULL_ADDR;
483 		set_data_blkaddr(dn);
484 		invalidate_blocks(sbi, blkaddr);
485 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
486 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
487 		nr_free++;
488 	}
489 
490 	if (nr_free) {
491 		pgoff_t fofs;
492 		/*
493 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
494 		 * we will invalidate all blkaddr in the whole range.
495 		 */
496 		fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
497 							dn->inode) + ofs;
498 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
499 		dec_valid_block_count(sbi, dn->inode, nr_free);
500 	}
501 	dn->ofs_in_node = ofs;
502 
503 	f2fs_update_time(sbi, REQ_TIME);
504 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
505 					 dn->ofs_in_node, nr_free);
506 	return nr_free;
507 }
508 
509 void truncate_data_blocks(struct dnode_of_data *dn)
510 {
511 	truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
512 }
513 
514 static int truncate_partial_data_page(struct inode *inode, u64 from,
515 								bool cache_only)
516 {
517 	unsigned offset = from & (PAGE_SIZE - 1);
518 	pgoff_t index = from >> PAGE_SHIFT;
519 	struct address_space *mapping = inode->i_mapping;
520 	struct page *page;
521 
522 	if (!offset && !cache_only)
523 		return 0;
524 
525 	if (cache_only) {
526 		page = f2fs_grab_cache_page(mapping, index, false);
527 		if (page && PageUptodate(page))
528 			goto truncate_out;
529 		f2fs_put_page(page, 1);
530 		return 0;
531 	}
532 
533 	page = get_lock_data_page(inode, index, true);
534 	if (IS_ERR(page))
535 		return 0;
536 truncate_out:
537 	f2fs_wait_on_page_writeback(page, DATA, true);
538 	zero_user(page, offset, PAGE_SIZE - offset);
539 	if (!cache_only || !f2fs_encrypted_inode(inode) ||
540 					!S_ISREG(inode->i_mode))
541 		set_page_dirty(page);
542 	f2fs_put_page(page, 1);
543 	return 0;
544 }
545 
546 int truncate_blocks(struct inode *inode, u64 from, bool lock)
547 {
548 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
549 	unsigned int blocksize = inode->i_sb->s_blocksize;
550 	struct dnode_of_data dn;
551 	pgoff_t free_from;
552 	int count = 0, err = 0;
553 	struct page *ipage;
554 	bool truncate_page = false;
555 
556 	trace_f2fs_truncate_blocks_enter(inode, from);
557 
558 	free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
559 
560 	if (free_from >= sbi->max_file_blocks)
561 		goto free_partial;
562 
563 	if (lock)
564 		f2fs_lock_op(sbi);
565 
566 	ipage = get_node_page(sbi, inode->i_ino);
567 	if (IS_ERR(ipage)) {
568 		err = PTR_ERR(ipage);
569 		goto out;
570 	}
571 
572 	if (f2fs_has_inline_data(inode)) {
573 		if (truncate_inline_inode(ipage, from))
574 			set_page_dirty(ipage);
575 		f2fs_put_page(ipage, 1);
576 		truncate_page = true;
577 		goto out;
578 	}
579 
580 	set_new_dnode(&dn, inode, ipage, NULL, 0);
581 	err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
582 	if (err) {
583 		if (err == -ENOENT)
584 			goto free_next;
585 		goto out;
586 	}
587 
588 	count = ADDRS_PER_PAGE(dn.node_page, inode);
589 
590 	count -= dn.ofs_in_node;
591 	f2fs_bug_on(sbi, count < 0);
592 
593 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
594 		truncate_data_blocks_range(&dn, count);
595 		free_from += count;
596 	}
597 
598 	f2fs_put_dnode(&dn);
599 free_next:
600 	err = truncate_inode_blocks(inode, free_from);
601 out:
602 	if (lock)
603 		f2fs_unlock_op(sbi);
604 free_partial:
605 	/* lastly zero out the first data page */
606 	if (!err)
607 		err = truncate_partial_data_page(inode, from, truncate_page);
608 
609 	trace_f2fs_truncate_blocks_exit(inode, err);
610 	return err;
611 }
612 
613 int f2fs_truncate(struct inode *inode)
614 {
615 	int err;
616 
617 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
618 				S_ISLNK(inode->i_mode)))
619 		return 0;
620 
621 	trace_f2fs_truncate(inode);
622 
623 	/* we should check inline_data size */
624 	if (!f2fs_may_inline_data(inode)) {
625 		err = f2fs_convert_inline_inode(inode);
626 		if (err)
627 			return err;
628 	}
629 
630 	err = truncate_blocks(inode, i_size_read(inode), true);
631 	if (err)
632 		return err;
633 
634 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
635 	f2fs_mark_inode_dirty_sync(inode);
636 	return 0;
637 }
638 
639 int f2fs_getattr(struct vfsmount *mnt,
640 			 struct dentry *dentry, struct kstat *stat)
641 {
642 	struct inode *inode = d_inode(dentry);
643 	generic_fillattr(inode, stat);
644 	stat->blocks <<= 3;
645 	return 0;
646 }
647 
648 #ifdef CONFIG_F2FS_FS_POSIX_ACL
649 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
650 {
651 	unsigned int ia_valid = attr->ia_valid;
652 
653 	if (ia_valid & ATTR_UID)
654 		inode->i_uid = attr->ia_uid;
655 	if (ia_valid & ATTR_GID)
656 		inode->i_gid = attr->ia_gid;
657 	if (ia_valid & ATTR_ATIME)
658 		inode->i_atime = timespec_trunc(attr->ia_atime,
659 						inode->i_sb->s_time_gran);
660 	if (ia_valid & ATTR_MTIME)
661 		inode->i_mtime = timespec_trunc(attr->ia_mtime,
662 						inode->i_sb->s_time_gran);
663 	if (ia_valid & ATTR_CTIME)
664 		inode->i_ctime = timespec_trunc(attr->ia_ctime,
665 						inode->i_sb->s_time_gran);
666 	if (ia_valid & ATTR_MODE) {
667 		umode_t mode = attr->ia_mode;
668 
669 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
670 			mode &= ~S_ISGID;
671 		set_acl_inode(inode, mode);
672 	}
673 }
674 #else
675 #define __setattr_copy setattr_copy
676 #endif
677 
678 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
679 {
680 	struct inode *inode = d_inode(dentry);
681 	int err;
682 
683 	err = inode_change_ok(inode, attr);
684 	if (err)
685 		return err;
686 
687 	if (attr->ia_valid & ATTR_SIZE) {
688 		if (f2fs_encrypted_inode(inode) &&
689 				fscrypt_get_encryption_info(inode))
690 			return -EACCES;
691 
692 		if (attr->ia_size <= i_size_read(inode)) {
693 			truncate_setsize(inode, attr->ia_size);
694 			err = f2fs_truncate(inode);
695 			if (err)
696 				return err;
697 			f2fs_balance_fs(F2FS_I_SB(inode), true);
698 		} else {
699 			/*
700 			 * do not trim all blocks after i_size if target size is
701 			 * larger than i_size.
702 			 */
703 			truncate_setsize(inode, attr->ia_size);
704 
705 			/* should convert inline inode here */
706 			if (!f2fs_may_inline_data(inode)) {
707 				err = f2fs_convert_inline_inode(inode);
708 				if (err)
709 					return err;
710 			}
711 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
712 		}
713 	}
714 
715 	__setattr_copy(inode, attr);
716 
717 	if (attr->ia_valid & ATTR_MODE) {
718 		err = posix_acl_chmod(inode, get_inode_mode(inode));
719 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
720 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
721 			clear_inode_flag(inode, FI_ACL_MODE);
722 		}
723 	}
724 
725 	f2fs_mark_inode_dirty_sync(inode);
726 	return err;
727 }
728 
729 const struct inode_operations f2fs_file_inode_operations = {
730 	.getattr	= f2fs_getattr,
731 	.setattr	= f2fs_setattr,
732 	.get_acl	= f2fs_get_acl,
733 	.set_acl	= f2fs_set_acl,
734 #ifdef CONFIG_F2FS_FS_XATTR
735 	.setxattr	= generic_setxattr,
736 	.getxattr	= generic_getxattr,
737 	.listxattr	= f2fs_listxattr,
738 	.removexattr	= generic_removexattr,
739 #endif
740 	.fiemap		= f2fs_fiemap,
741 };
742 
743 static int fill_zero(struct inode *inode, pgoff_t index,
744 					loff_t start, loff_t len)
745 {
746 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
747 	struct page *page;
748 
749 	if (!len)
750 		return 0;
751 
752 	f2fs_balance_fs(sbi, true);
753 
754 	f2fs_lock_op(sbi);
755 	page = get_new_data_page(inode, NULL, index, false);
756 	f2fs_unlock_op(sbi);
757 
758 	if (IS_ERR(page))
759 		return PTR_ERR(page);
760 
761 	f2fs_wait_on_page_writeback(page, DATA, true);
762 	zero_user(page, start, len);
763 	set_page_dirty(page);
764 	f2fs_put_page(page, 1);
765 	return 0;
766 }
767 
768 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
769 {
770 	int err;
771 
772 	while (pg_start < pg_end) {
773 		struct dnode_of_data dn;
774 		pgoff_t end_offset, count;
775 
776 		set_new_dnode(&dn, inode, NULL, NULL, 0);
777 		err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
778 		if (err) {
779 			if (err == -ENOENT) {
780 				pg_start++;
781 				continue;
782 			}
783 			return err;
784 		}
785 
786 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
787 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
788 
789 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
790 
791 		truncate_data_blocks_range(&dn, count);
792 		f2fs_put_dnode(&dn);
793 
794 		pg_start += count;
795 	}
796 	return 0;
797 }
798 
799 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
800 {
801 	pgoff_t pg_start, pg_end;
802 	loff_t off_start, off_end;
803 	int ret;
804 
805 	ret = f2fs_convert_inline_inode(inode);
806 	if (ret)
807 		return ret;
808 
809 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
810 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
811 
812 	off_start = offset & (PAGE_SIZE - 1);
813 	off_end = (offset + len) & (PAGE_SIZE - 1);
814 
815 	if (pg_start == pg_end) {
816 		ret = fill_zero(inode, pg_start, off_start,
817 						off_end - off_start);
818 		if (ret)
819 			return ret;
820 	} else {
821 		if (off_start) {
822 			ret = fill_zero(inode, pg_start++, off_start,
823 						PAGE_SIZE - off_start);
824 			if (ret)
825 				return ret;
826 		}
827 		if (off_end) {
828 			ret = fill_zero(inode, pg_end, 0, off_end);
829 			if (ret)
830 				return ret;
831 		}
832 
833 		if (pg_start < pg_end) {
834 			struct address_space *mapping = inode->i_mapping;
835 			loff_t blk_start, blk_end;
836 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
837 
838 			f2fs_balance_fs(sbi, true);
839 
840 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
841 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
842 			truncate_inode_pages_range(mapping, blk_start,
843 					blk_end - 1);
844 
845 			f2fs_lock_op(sbi);
846 			ret = truncate_hole(inode, pg_start, pg_end);
847 			f2fs_unlock_op(sbi);
848 		}
849 	}
850 
851 	return ret;
852 }
853 
854 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
855 				int *do_replace, pgoff_t off, pgoff_t len)
856 {
857 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
858 	struct dnode_of_data dn;
859 	int ret, done, i;
860 
861 next_dnode:
862 	set_new_dnode(&dn, inode, NULL, NULL, 0);
863 	ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
864 	if (ret && ret != -ENOENT) {
865 		return ret;
866 	} else if (ret == -ENOENT) {
867 		if (dn.max_level == 0)
868 			return -ENOENT;
869 		done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
870 		blkaddr += done;
871 		do_replace += done;
872 		goto next;
873 	}
874 
875 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
876 							dn.ofs_in_node, len);
877 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
878 		*blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
879 		if (!is_checkpointed_data(sbi, *blkaddr)) {
880 
881 			if (test_opt(sbi, LFS)) {
882 				f2fs_put_dnode(&dn);
883 				return -ENOTSUPP;
884 			}
885 
886 			/* do not invalidate this block address */
887 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
888 			*do_replace = 1;
889 		}
890 	}
891 	f2fs_put_dnode(&dn);
892 next:
893 	len -= done;
894 	off += done;
895 	if (len)
896 		goto next_dnode;
897 	return 0;
898 }
899 
900 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
901 				int *do_replace, pgoff_t off, int len)
902 {
903 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
904 	struct dnode_of_data dn;
905 	int ret, i;
906 
907 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
908 		if (*do_replace == 0)
909 			continue;
910 
911 		set_new_dnode(&dn, inode, NULL, NULL, 0);
912 		ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
913 		if (ret) {
914 			dec_valid_block_count(sbi, inode, 1);
915 			invalidate_blocks(sbi, *blkaddr);
916 		} else {
917 			f2fs_update_data_blkaddr(&dn, *blkaddr);
918 		}
919 		f2fs_put_dnode(&dn);
920 	}
921 	return 0;
922 }
923 
924 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
925 			block_t *blkaddr, int *do_replace,
926 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
927 {
928 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
929 	pgoff_t i = 0;
930 	int ret;
931 
932 	while (i < len) {
933 		if (blkaddr[i] == NULL_ADDR && !full) {
934 			i++;
935 			continue;
936 		}
937 
938 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
939 			struct dnode_of_data dn;
940 			struct node_info ni;
941 			size_t new_size;
942 			pgoff_t ilen;
943 
944 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
945 			ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
946 			if (ret)
947 				return ret;
948 
949 			get_node_info(sbi, dn.nid, &ni);
950 			ilen = min((pgoff_t)
951 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
952 						dn.ofs_in_node, len - i);
953 			do {
954 				dn.data_blkaddr = datablock_addr(dn.node_page,
955 								dn.ofs_in_node);
956 				truncate_data_blocks_range(&dn, 1);
957 
958 				if (do_replace[i]) {
959 					f2fs_i_blocks_write(src_inode,
960 								1, false);
961 					f2fs_i_blocks_write(dst_inode,
962 								1, true);
963 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
964 					blkaddr[i], ni.version, true, false);
965 
966 					do_replace[i] = 0;
967 				}
968 				dn.ofs_in_node++;
969 				i++;
970 				new_size = (dst + i) << PAGE_SHIFT;
971 				if (dst_inode->i_size < new_size)
972 					f2fs_i_size_write(dst_inode, new_size);
973 			} while ((do_replace[i] || blkaddr[i] == NULL_ADDR) && --ilen);
974 
975 			f2fs_put_dnode(&dn);
976 		} else {
977 			struct page *psrc, *pdst;
978 
979 			psrc = get_lock_data_page(src_inode, src + i, true);
980 			if (IS_ERR(psrc))
981 				return PTR_ERR(psrc);
982 			pdst = get_new_data_page(dst_inode, NULL, dst + i,
983 								true);
984 			if (IS_ERR(pdst)) {
985 				f2fs_put_page(psrc, 1);
986 				return PTR_ERR(pdst);
987 			}
988 			f2fs_copy_page(psrc, pdst);
989 			set_page_dirty(pdst);
990 			f2fs_put_page(pdst, 1);
991 			f2fs_put_page(psrc, 1);
992 
993 			ret = truncate_hole(src_inode, src + i, src + i + 1);
994 			if (ret)
995 				return ret;
996 			i++;
997 		}
998 	}
999 	return 0;
1000 }
1001 
1002 static int __exchange_data_block(struct inode *src_inode,
1003 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1004 			pgoff_t len, bool full)
1005 {
1006 	block_t *src_blkaddr;
1007 	int *do_replace;
1008 	pgoff_t olen;
1009 	int ret;
1010 
1011 	while (len) {
1012 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1013 
1014 		src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1015 		if (!src_blkaddr)
1016 			return -ENOMEM;
1017 
1018 		do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1019 		if (!do_replace) {
1020 			kvfree(src_blkaddr);
1021 			return -ENOMEM;
1022 		}
1023 
1024 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1025 					do_replace, src, olen);
1026 		if (ret)
1027 			goto roll_back;
1028 
1029 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1030 					do_replace, src, dst, olen, full);
1031 		if (ret)
1032 			goto roll_back;
1033 
1034 		src += olen;
1035 		dst += olen;
1036 		len -= olen;
1037 
1038 		kvfree(src_blkaddr);
1039 		kvfree(do_replace);
1040 	}
1041 	return 0;
1042 
1043 roll_back:
1044 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1045 	kvfree(src_blkaddr);
1046 	kvfree(do_replace);
1047 	return ret;
1048 }
1049 
1050 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1051 {
1052 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1053 	pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1054 	int ret;
1055 
1056 	f2fs_balance_fs(sbi, true);
1057 	f2fs_lock_op(sbi);
1058 
1059 	f2fs_drop_extent_tree(inode);
1060 
1061 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1062 	f2fs_unlock_op(sbi);
1063 	return ret;
1064 }
1065 
1066 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1067 {
1068 	pgoff_t pg_start, pg_end;
1069 	loff_t new_size;
1070 	int ret;
1071 
1072 	if (offset + len >= i_size_read(inode))
1073 		return -EINVAL;
1074 
1075 	/* collapse range should be aligned to block size of f2fs. */
1076 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1077 		return -EINVAL;
1078 
1079 	ret = f2fs_convert_inline_inode(inode);
1080 	if (ret)
1081 		return ret;
1082 
1083 	pg_start = offset >> PAGE_SHIFT;
1084 	pg_end = (offset + len) >> PAGE_SHIFT;
1085 
1086 	/* write out all dirty pages from offset */
1087 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1088 	if (ret)
1089 		return ret;
1090 
1091 	truncate_pagecache(inode, offset);
1092 
1093 	ret = f2fs_do_collapse(inode, pg_start, pg_end);
1094 	if (ret)
1095 		return ret;
1096 
1097 	/* write out all moved pages, if possible */
1098 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1099 	truncate_pagecache(inode, offset);
1100 
1101 	new_size = i_size_read(inode) - len;
1102 	truncate_pagecache(inode, new_size);
1103 
1104 	ret = truncate_blocks(inode, new_size, true);
1105 	if (!ret)
1106 		f2fs_i_size_write(inode, new_size);
1107 
1108 	return ret;
1109 }
1110 
1111 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1112 								pgoff_t end)
1113 {
1114 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1115 	pgoff_t index = start;
1116 	unsigned int ofs_in_node = dn->ofs_in_node;
1117 	blkcnt_t count = 0;
1118 	int ret;
1119 
1120 	for (; index < end; index++, dn->ofs_in_node++) {
1121 		if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1122 			count++;
1123 	}
1124 
1125 	dn->ofs_in_node = ofs_in_node;
1126 	ret = reserve_new_blocks(dn, count);
1127 	if (ret)
1128 		return ret;
1129 
1130 	dn->ofs_in_node = ofs_in_node;
1131 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1132 		dn->data_blkaddr =
1133 				datablock_addr(dn->node_page, dn->ofs_in_node);
1134 		/*
1135 		 * reserve_new_blocks will not guarantee entire block
1136 		 * allocation.
1137 		 */
1138 		if (dn->data_blkaddr == NULL_ADDR) {
1139 			ret = -ENOSPC;
1140 			break;
1141 		}
1142 		if (dn->data_blkaddr != NEW_ADDR) {
1143 			invalidate_blocks(sbi, dn->data_blkaddr);
1144 			dn->data_blkaddr = NEW_ADDR;
1145 			set_data_blkaddr(dn);
1146 		}
1147 	}
1148 
1149 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1150 
1151 	return ret;
1152 }
1153 
1154 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1155 								int mode)
1156 {
1157 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1158 	struct address_space *mapping = inode->i_mapping;
1159 	pgoff_t index, pg_start, pg_end;
1160 	loff_t new_size = i_size_read(inode);
1161 	loff_t off_start, off_end;
1162 	int ret = 0;
1163 
1164 	ret = inode_newsize_ok(inode, (len + offset));
1165 	if (ret)
1166 		return ret;
1167 
1168 	ret = f2fs_convert_inline_inode(inode);
1169 	if (ret)
1170 		return ret;
1171 
1172 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1173 	if (ret)
1174 		return ret;
1175 
1176 	truncate_pagecache_range(inode, offset, offset + len - 1);
1177 
1178 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1179 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1180 
1181 	off_start = offset & (PAGE_SIZE - 1);
1182 	off_end = (offset + len) & (PAGE_SIZE - 1);
1183 
1184 	if (pg_start == pg_end) {
1185 		ret = fill_zero(inode, pg_start, off_start,
1186 						off_end - off_start);
1187 		if (ret)
1188 			return ret;
1189 
1190 		if (offset + len > new_size)
1191 			new_size = offset + len;
1192 		new_size = max_t(loff_t, new_size, offset + len);
1193 	} else {
1194 		if (off_start) {
1195 			ret = fill_zero(inode, pg_start++, off_start,
1196 						PAGE_SIZE - off_start);
1197 			if (ret)
1198 				return ret;
1199 
1200 			new_size = max_t(loff_t, new_size,
1201 					(loff_t)pg_start << PAGE_SHIFT);
1202 		}
1203 
1204 		for (index = pg_start; index < pg_end;) {
1205 			struct dnode_of_data dn;
1206 			unsigned int end_offset;
1207 			pgoff_t end;
1208 
1209 			f2fs_lock_op(sbi);
1210 
1211 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1212 			ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1213 			if (ret) {
1214 				f2fs_unlock_op(sbi);
1215 				goto out;
1216 			}
1217 
1218 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1219 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1220 
1221 			ret = f2fs_do_zero_range(&dn, index, end);
1222 			f2fs_put_dnode(&dn);
1223 			f2fs_unlock_op(sbi);
1224 			if (ret)
1225 				goto out;
1226 
1227 			index = end;
1228 			new_size = max_t(loff_t, new_size,
1229 					(loff_t)index << PAGE_SHIFT);
1230 		}
1231 
1232 		if (off_end) {
1233 			ret = fill_zero(inode, pg_end, 0, off_end);
1234 			if (ret)
1235 				goto out;
1236 
1237 			new_size = max_t(loff_t, new_size, offset + len);
1238 		}
1239 	}
1240 
1241 out:
1242 	if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1243 		f2fs_i_size_write(inode, new_size);
1244 
1245 	return ret;
1246 }
1247 
1248 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1249 {
1250 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1251 	pgoff_t nr, pg_start, pg_end, delta, idx;
1252 	loff_t new_size;
1253 	int ret = 0;
1254 
1255 	new_size = i_size_read(inode) + len;
1256 	if (new_size > inode->i_sb->s_maxbytes)
1257 		return -EFBIG;
1258 
1259 	if (offset >= i_size_read(inode))
1260 		return -EINVAL;
1261 
1262 	/* insert range should be aligned to block size of f2fs. */
1263 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1264 		return -EINVAL;
1265 
1266 	ret = f2fs_convert_inline_inode(inode);
1267 	if (ret)
1268 		return ret;
1269 
1270 	f2fs_balance_fs(sbi, true);
1271 
1272 	ret = truncate_blocks(inode, i_size_read(inode), true);
1273 	if (ret)
1274 		return ret;
1275 
1276 	/* write out all dirty pages from offset */
1277 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1278 	if (ret)
1279 		return ret;
1280 
1281 	truncate_pagecache(inode, offset);
1282 
1283 	pg_start = offset >> PAGE_SHIFT;
1284 	pg_end = (offset + len) >> PAGE_SHIFT;
1285 	delta = pg_end - pg_start;
1286 	idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1287 
1288 	while (!ret && idx > pg_start) {
1289 		nr = idx - pg_start;
1290 		if (nr > delta)
1291 			nr = delta;
1292 		idx -= nr;
1293 
1294 		f2fs_lock_op(sbi);
1295 		f2fs_drop_extent_tree(inode);
1296 
1297 		ret = __exchange_data_block(inode, inode, idx,
1298 					idx + delta, nr, false);
1299 		f2fs_unlock_op(sbi);
1300 	}
1301 
1302 	/* write out all moved pages, if possible */
1303 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1304 	truncate_pagecache(inode, offset);
1305 
1306 	if (!ret)
1307 		f2fs_i_size_write(inode, new_size);
1308 	return ret;
1309 }
1310 
1311 static int expand_inode_data(struct inode *inode, loff_t offset,
1312 					loff_t len, int mode)
1313 {
1314 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1315 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1316 	pgoff_t pg_end;
1317 	loff_t new_size = i_size_read(inode);
1318 	loff_t off_end;
1319 	int ret;
1320 
1321 	ret = inode_newsize_ok(inode, (len + offset));
1322 	if (ret)
1323 		return ret;
1324 
1325 	ret = f2fs_convert_inline_inode(inode);
1326 	if (ret)
1327 		return ret;
1328 
1329 	f2fs_balance_fs(sbi, true);
1330 
1331 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1332 	off_end = (offset + len) & (PAGE_SIZE - 1);
1333 
1334 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1335 	map.m_len = pg_end - map.m_lblk;
1336 	if (off_end)
1337 		map.m_len++;
1338 
1339 	ret = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1340 	if (ret) {
1341 		pgoff_t last_off;
1342 
1343 		if (!map.m_len)
1344 			return ret;
1345 
1346 		last_off = map.m_lblk + map.m_len - 1;
1347 
1348 		/* update new size to the failed position */
1349 		new_size = (last_off == pg_end) ? offset + len:
1350 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1351 	} else {
1352 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1353 	}
1354 
1355 	if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1356 		f2fs_i_size_write(inode, new_size);
1357 
1358 	return ret;
1359 }
1360 
1361 static long f2fs_fallocate(struct file *file, int mode,
1362 				loff_t offset, loff_t len)
1363 {
1364 	struct inode *inode = file_inode(file);
1365 	long ret = 0;
1366 
1367 	/* f2fs only support ->fallocate for regular file */
1368 	if (!S_ISREG(inode->i_mode))
1369 		return -EINVAL;
1370 
1371 	if (f2fs_encrypted_inode(inode) &&
1372 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1373 		return -EOPNOTSUPP;
1374 
1375 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1376 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1377 			FALLOC_FL_INSERT_RANGE))
1378 		return -EOPNOTSUPP;
1379 
1380 	inode_lock(inode);
1381 
1382 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1383 		if (offset >= inode->i_size)
1384 			goto out;
1385 
1386 		ret = punch_hole(inode, offset, len);
1387 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1388 		ret = f2fs_collapse_range(inode, offset, len);
1389 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1390 		ret = f2fs_zero_range(inode, offset, len, mode);
1391 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1392 		ret = f2fs_insert_range(inode, offset, len);
1393 	} else {
1394 		ret = expand_inode_data(inode, offset, len, mode);
1395 	}
1396 
1397 	if (!ret) {
1398 		inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1399 		f2fs_mark_inode_dirty_sync(inode);
1400 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1401 	}
1402 
1403 out:
1404 	inode_unlock(inode);
1405 
1406 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1407 	return ret;
1408 }
1409 
1410 static int f2fs_release_file(struct inode *inode, struct file *filp)
1411 {
1412 	/*
1413 	 * f2fs_relase_file is called at every close calls. So we should
1414 	 * not drop any inmemory pages by close called by other process.
1415 	 */
1416 	if (!(filp->f_mode & FMODE_WRITE) ||
1417 			atomic_read(&inode->i_writecount) != 1)
1418 		return 0;
1419 
1420 	/* some remained atomic pages should discarded */
1421 	if (f2fs_is_atomic_file(inode))
1422 		drop_inmem_pages(inode);
1423 	if (f2fs_is_volatile_file(inode)) {
1424 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1425 		set_inode_flag(inode, FI_DROP_CACHE);
1426 		filemap_fdatawrite(inode->i_mapping);
1427 		clear_inode_flag(inode, FI_DROP_CACHE);
1428 	}
1429 	return 0;
1430 }
1431 
1432 #define F2FS_REG_FLMASK		(~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1433 #define F2FS_OTHER_FLMASK	(FS_NODUMP_FL | FS_NOATIME_FL)
1434 
1435 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1436 {
1437 	if (S_ISDIR(mode))
1438 		return flags;
1439 	else if (S_ISREG(mode))
1440 		return flags & F2FS_REG_FLMASK;
1441 	else
1442 		return flags & F2FS_OTHER_FLMASK;
1443 }
1444 
1445 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1446 {
1447 	struct inode *inode = file_inode(filp);
1448 	struct f2fs_inode_info *fi = F2FS_I(inode);
1449 	unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1450 	return put_user(flags, (int __user *)arg);
1451 }
1452 
1453 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1454 {
1455 	struct inode *inode = file_inode(filp);
1456 	struct f2fs_inode_info *fi = F2FS_I(inode);
1457 	unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1458 	unsigned int oldflags;
1459 	int ret;
1460 
1461 	if (!inode_owner_or_capable(inode))
1462 		return -EACCES;
1463 
1464 	if (get_user(flags, (int __user *)arg))
1465 		return -EFAULT;
1466 
1467 	ret = mnt_want_write_file(filp);
1468 	if (ret)
1469 		return ret;
1470 
1471 	flags = f2fs_mask_flags(inode->i_mode, flags);
1472 
1473 	inode_lock(inode);
1474 
1475 	oldflags = fi->i_flags;
1476 
1477 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1478 		if (!capable(CAP_LINUX_IMMUTABLE)) {
1479 			inode_unlock(inode);
1480 			ret = -EPERM;
1481 			goto out;
1482 		}
1483 	}
1484 
1485 	flags = flags & FS_FL_USER_MODIFIABLE;
1486 	flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1487 	fi->i_flags = flags;
1488 	inode_unlock(inode);
1489 
1490 	inode->i_ctime = CURRENT_TIME;
1491 	f2fs_set_inode_flags(inode);
1492 out:
1493 	mnt_drop_write_file(filp);
1494 	return ret;
1495 }
1496 
1497 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1498 {
1499 	struct inode *inode = file_inode(filp);
1500 
1501 	return put_user(inode->i_generation, (int __user *)arg);
1502 }
1503 
1504 static int f2fs_ioc_start_atomic_write(struct file *filp)
1505 {
1506 	struct inode *inode = file_inode(filp);
1507 	int ret;
1508 
1509 	if (!inode_owner_or_capable(inode))
1510 		return -EACCES;
1511 
1512 	ret = mnt_want_write_file(filp);
1513 	if (ret)
1514 		return ret;
1515 
1516 	inode_lock(inode);
1517 
1518 	if (f2fs_is_atomic_file(inode))
1519 		goto out;
1520 
1521 	ret = f2fs_convert_inline_inode(inode);
1522 	if (ret)
1523 		goto out;
1524 
1525 	set_inode_flag(inode, FI_ATOMIC_FILE);
1526 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1527 
1528 	if (!get_dirty_pages(inode))
1529 		goto out;
1530 
1531 	f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1532 		"Unexpected flush for atomic writes: ino=%lu, npages=%lld",
1533 					inode->i_ino, get_dirty_pages(inode));
1534 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1535 	if (ret)
1536 		clear_inode_flag(inode, FI_ATOMIC_FILE);
1537 out:
1538 	inode_unlock(inode);
1539 	mnt_drop_write_file(filp);
1540 	return ret;
1541 }
1542 
1543 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1544 {
1545 	struct inode *inode = file_inode(filp);
1546 	int ret;
1547 
1548 	if (!inode_owner_or_capable(inode))
1549 		return -EACCES;
1550 
1551 	ret = mnt_want_write_file(filp);
1552 	if (ret)
1553 		return ret;
1554 
1555 	inode_lock(inode);
1556 
1557 	if (f2fs_is_volatile_file(inode))
1558 		goto err_out;
1559 
1560 	if (f2fs_is_atomic_file(inode)) {
1561 		clear_inode_flag(inode, FI_ATOMIC_FILE);
1562 		ret = commit_inmem_pages(inode);
1563 		if (ret) {
1564 			set_inode_flag(inode, FI_ATOMIC_FILE);
1565 			goto err_out;
1566 		}
1567 	}
1568 
1569 	ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1570 err_out:
1571 	inode_unlock(inode);
1572 	mnt_drop_write_file(filp);
1573 	return ret;
1574 }
1575 
1576 static int f2fs_ioc_start_volatile_write(struct file *filp)
1577 {
1578 	struct inode *inode = file_inode(filp);
1579 	int ret;
1580 
1581 	if (!inode_owner_or_capable(inode))
1582 		return -EACCES;
1583 
1584 	ret = mnt_want_write_file(filp);
1585 	if (ret)
1586 		return ret;
1587 
1588 	inode_lock(inode);
1589 
1590 	if (f2fs_is_volatile_file(inode))
1591 		goto out;
1592 
1593 	ret = f2fs_convert_inline_inode(inode);
1594 	if (ret)
1595 		goto out;
1596 
1597 	set_inode_flag(inode, FI_VOLATILE_FILE);
1598 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1599 out:
1600 	inode_unlock(inode);
1601 	mnt_drop_write_file(filp);
1602 	return ret;
1603 }
1604 
1605 static int f2fs_ioc_release_volatile_write(struct file *filp)
1606 {
1607 	struct inode *inode = file_inode(filp);
1608 	int ret;
1609 
1610 	if (!inode_owner_or_capable(inode))
1611 		return -EACCES;
1612 
1613 	ret = mnt_want_write_file(filp);
1614 	if (ret)
1615 		return ret;
1616 
1617 	inode_lock(inode);
1618 
1619 	if (!f2fs_is_volatile_file(inode))
1620 		goto out;
1621 
1622 	if (!f2fs_is_first_block_written(inode)) {
1623 		ret = truncate_partial_data_page(inode, 0, true);
1624 		goto out;
1625 	}
1626 
1627 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1628 out:
1629 	inode_unlock(inode);
1630 	mnt_drop_write_file(filp);
1631 	return ret;
1632 }
1633 
1634 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1635 {
1636 	struct inode *inode = file_inode(filp);
1637 	int ret;
1638 
1639 	if (!inode_owner_or_capable(inode))
1640 		return -EACCES;
1641 
1642 	ret = mnt_want_write_file(filp);
1643 	if (ret)
1644 		return ret;
1645 
1646 	inode_lock(inode);
1647 
1648 	if (f2fs_is_atomic_file(inode))
1649 		drop_inmem_pages(inode);
1650 	if (f2fs_is_volatile_file(inode)) {
1651 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1652 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1653 	}
1654 
1655 	inode_unlock(inode);
1656 
1657 	mnt_drop_write_file(filp);
1658 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1659 	return ret;
1660 }
1661 
1662 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1663 {
1664 	struct inode *inode = file_inode(filp);
1665 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1666 	struct super_block *sb = sbi->sb;
1667 	__u32 in;
1668 	int ret;
1669 
1670 	if (!capable(CAP_SYS_ADMIN))
1671 		return -EPERM;
1672 
1673 	if (get_user(in, (__u32 __user *)arg))
1674 		return -EFAULT;
1675 
1676 	ret = mnt_want_write_file(filp);
1677 	if (ret)
1678 		return ret;
1679 
1680 	switch (in) {
1681 	case F2FS_GOING_DOWN_FULLSYNC:
1682 		sb = freeze_bdev(sb->s_bdev);
1683 		if (sb && !IS_ERR(sb)) {
1684 			f2fs_stop_checkpoint(sbi, false);
1685 			thaw_bdev(sb->s_bdev, sb);
1686 		}
1687 		break;
1688 	case F2FS_GOING_DOWN_METASYNC:
1689 		/* do checkpoint only */
1690 		f2fs_sync_fs(sb, 1);
1691 		f2fs_stop_checkpoint(sbi, false);
1692 		break;
1693 	case F2FS_GOING_DOWN_NOSYNC:
1694 		f2fs_stop_checkpoint(sbi, false);
1695 		break;
1696 	case F2FS_GOING_DOWN_METAFLUSH:
1697 		sync_meta_pages(sbi, META, LONG_MAX);
1698 		f2fs_stop_checkpoint(sbi, false);
1699 		break;
1700 	default:
1701 		ret = -EINVAL;
1702 		goto out;
1703 	}
1704 	f2fs_update_time(sbi, REQ_TIME);
1705 out:
1706 	mnt_drop_write_file(filp);
1707 	return ret;
1708 }
1709 
1710 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1711 {
1712 	struct inode *inode = file_inode(filp);
1713 	struct super_block *sb = inode->i_sb;
1714 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
1715 	struct fstrim_range range;
1716 	int ret;
1717 
1718 	if (!capable(CAP_SYS_ADMIN))
1719 		return -EPERM;
1720 
1721 	if (!blk_queue_discard(q))
1722 		return -EOPNOTSUPP;
1723 
1724 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1725 				sizeof(range)))
1726 		return -EFAULT;
1727 
1728 	ret = mnt_want_write_file(filp);
1729 	if (ret)
1730 		return ret;
1731 
1732 	range.minlen = max((unsigned int)range.minlen,
1733 				q->limits.discard_granularity);
1734 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1735 	mnt_drop_write_file(filp);
1736 	if (ret < 0)
1737 		return ret;
1738 
1739 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
1740 				sizeof(range)))
1741 		return -EFAULT;
1742 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1743 	return 0;
1744 }
1745 
1746 static bool uuid_is_nonzero(__u8 u[16])
1747 {
1748 	int i;
1749 
1750 	for (i = 0; i < 16; i++)
1751 		if (u[i])
1752 			return true;
1753 	return false;
1754 }
1755 
1756 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1757 {
1758 	struct fscrypt_policy policy;
1759 	struct inode *inode = file_inode(filp);
1760 
1761 	if (copy_from_user(&policy, (struct fscrypt_policy __user *)arg,
1762 							sizeof(policy)))
1763 		return -EFAULT;
1764 
1765 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1766 
1767 	return fscrypt_process_policy(filp, &policy);
1768 }
1769 
1770 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1771 {
1772 	struct fscrypt_policy policy;
1773 	struct inode *inode = file_inode(filp);
1774 	int err;
1775 
1776 	err = fscrypt_get_policy(inode, &policy);
1777 	if (err)
1778 		return err;
1779 
1780 	if (copy_to_user((struct fscrypt_policy __user *)arg, &policy, sizeof(policy)))
1781 		return -EFAULT;
1782 	return 0;
1783 }
1784 
1785 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1786 {
1787 	struct inode *inode = file_inode(filp);
1788 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1789 	int err;
1790 
1791 	if (!f2fs_sb_has_crypto(inode->i_sb))
1792 		return -EOPNOTSUPP;
1793 
1794 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1795 		goto got_it;
1796 
1797 	err = mnt_want_write_file(filp);
1798 	if (err)
1799 		return err;
1800 
1801 	/* update superblock with uuid */
1802 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1803 
1804 	err = f2fs_commit_super(sbi, false);
1805 	if (err) {
1806 		/* undo new data */
1807 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1808 		mnt_drop_write_file(filp);
1809 		return err;
1810 	}
1811 	mnt_drop_write_file(filp);
1812 got_it:
1813 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1814 									16))
1815 		return -EFAULT;
1816 	return 0;
1817 }
1818 
1819 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1820 {
1821 	struct inode *inode = file_inode(filp);
1822 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1823 	__u32 sync;
1824 	int ret;
1825 
1826 	if (!capable(CAP_SYS_ADMIN))
1827 		return -EPERM;
1828 
1829 	if (get_user(sync, (__u32 __user *)arg))
1830 		return -EFAULT;
1831 
1832 	if (f2fs_readonly(sbi->sb))
1833 		return -EROFS;
1834 
1835 	ret = mnt_want_write_file(filp);
1836 	if (ret)
1837 		return ret;
1838 
1839 	if (!sync) {
1840 		if (!mutex_trylock(&sbi->gc_mutex)) {
1841 			ret = -EBUSY;
1842 			goto out;
1843 		}
1844 	} else {
1845 		mutex_lock(&sbi->gc_mutex);
1846 	}
1847 
1848 	ret = f2fs_gc(sbi, sync);
1849 out:
1850 	mnt_drop_write_file(filp);
1851 	return ret;
1852 }
1853 
1854 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1855 {
1856 	struct inode *inode = file_inode(filp);
1857 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1858 	int ret;
1859 
1860 	if (!capable(CAP_SYS_ADMIN))
1861 		return -EPERM;
1862 
1863 	if (f2fs_readonly(sbi->sb))
1864 		return -EROFS;
1865 
1866 	ret = mnt_want_write_file(filp);
1867 	if (ret)
1868 		return ret;
1869 
1870 	ret = f2fs_sync_fs(sbi->sb, 1);
1871 
1872 	mnt_drop_write_file(filp);
1873 	return ret;
1874 }
1875 
1876 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1877 					struct file *filp,
1878 					struct f2fs_defragment *range)
1879 {
1880 	struct inode *inode = file_inode(filp);
1881 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1882 	struct extent_info ei;
1883 	pgoff_t pg_start, pg_end;
1884 	unsigned int blk_per_seg = sbi->blocks_per_seg;
1885 	unsigned int total = 0, sec_num;
1886 	unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1887 	block_t blk_end = 0;
1888 	bool fragmented = false;
1889 	int err;
1890 
1891 	/* if in-place-update policy is enabled, don't waste time here */
1892 	if (need_inplace_update(inode))
1893 		return -EINVAL;
1894 
1895 	pg_start = range->start >> PAGE_SHIFT;
1896 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
1897 
1898 	f2fs_balance_fs(sbi, true);
1899 
1900 	inode_lock(inode);
1901 
1902 	/* writeback all dirty pages in the range */
1903 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1904 						range->start + range->len - 1);
1905 	if (err)
1906 		goto out;
1907 
1908 	/*
1909 	 * lookup mapping info in extent cache, skip defragmenting if physical
1910 	 * block addresses are continuous.
1911 	 */
1912 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1913 		if (ei.fofs + ei.len >= pg_end)
1914 			goto out;
1915 	}
1916 
1917 	map.m_lblk = pg_start;
1918 
1919 	/*
1920 	 * lookup mapping info in dnode page cache, skip defragmenting if all
1921 	 * physical block addresses are continuous even if there are hole(s)
1922 	 * in logical blocks.
1923 	 */
1924 	while (map.m_lblk < pg_end) {
1925 		map.m_len = pg_end - map.m_lblk;
1926 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1927 		if (err)
1928 			goto out;
1929 
1930 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1931 			map.m_lblk++;
1932 			continue;
1933 		}
1934 
1935 		if (blk_end && blk_end != map.m_pblk) {
1936 			fragmented = true;
1937 			break;
1938 		}
1939 		blk_end = map.m_pblk + map.m_len;
1940 
1941 		map.m_lblk += map.m_len;
1942 	}
1943 
1944 	if (!fragmented)
1945 		goto out;
1946 
1947 	map.m_lblk = pg_start;
1948 	map.m_len = pg_end - pg_start;
1949 
1950 	sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1951 
1952 	/*
1953 	 * make sure there are enough free section for LFS allocation, this can
1954 	 * avoid defragment running in SSR mode when free section are allocated
1955 	 * intensively
1956 	 */
1957 	if (has_not_enough_free_secs(sbi, sec_num)) {
1958 		err = -EAGAIN;
1959 		goto out;
1960 	}
1961 
1962 	while (map.m_lblk < pg_end) {
1963 		pgoff_t idx;
1964 		int cnt = 0;
1965 
1966 do_map:
1967 		map.m_len = pg_end - map.m_lblk;
1968 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1969 		if (err)
1970 			goto clear_out;
1971 
1972 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1973 			map.m_lblk++;
1974 			continue;
1975 		}
1976 
1977 		set_inode_flag(inode, FI_DO_DEFRAG);
1978 
1979 		idx = map.m_lblk;
1980 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1981 			struct page *page;
1982 
1983 			page = get_lock_data_page(inode, idx, true);
1984 			if (IS_ERR(page)) {
1985 				err = PTR_ERR(page);
1986 				goto clear_out;
1987 			}
1988 
1989 			set_page_dirty(page);
1990 			f2fs_put_page(page, 1);
1991 
1992 			idx++;
1993 			cnt++;
1994 			total++;
1995 		}
1996 
1997 		map.m_lblk = idx;
1998 
1999 		if (idx < pg_end && cnt < blk_per_seg)
2000 			goto do_map;
2001 
2002 		clear_inode_flag(inode, FI_DO_DEFRAG);
2003 
2004 		err = filemap_fdatawrite(inode->i_mapping);
2005 		if (err)
2006 			goto out;
2007 	}
2008 clear_out:
2009 	clear_inode_flag(inode, FI_DO_DEFRAG);
2010 out:
2011 	inode_unlock(inode);
2012 	if (!err)
2013 		range->len = (u64)total << PAGE_SHIFT;
2014 	return err;
2015 }
2016 
2017 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2018 {
2019 	struct inode *inode = file_inode(filp);
2020 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2021 	struct f2fs_defragment range;
2022 	int err;
2023 
2024 	if (!capable(CAP_SYS_ADMIN))
2025 		return -EPERM;
2026 
2027 	if (!S_ISREG(inode->i_mode))
2028 		return -EINVAL;
2029 
2030 	err = mnt_want_write_file(filp);
2031 	if (err)
2032 		return err;
2033 
2034 	if (f2fs_readonly(sbi->sb)) {
2035 		err = -EROFS;
2036 		goto out;
2037 	}
2038 
2039 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2040 							sizeof(range))) {
2041 		err = -EFAULT;
2042 		goto out;
2043 	}
2044 
2045 	/* verify alignment of offset & size */
2046 	if (range.start & (F2FS_BLKSIZE - 1) ||
2047 		range.len & (F2FS_BLKSIZE - 1)) {
2048 		err = -EINVAL;
2049 		goto out;
2050 	}
2051 
2052 	err = f2fs_defragment_range(sbi, filp, &range);
2053 	f2fs_update_time(sbi, REQ_TIME);
2054 	if (err < 0)
2055 		goto out;
2056 
2057 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2058 							sizeof(range)))
2059 		err = -EFAULT;
2060 out:
2061 	mnt_drop_write_file(filp);
2062 	return err;
2063 }
2064 
2065 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2066 			struct file *file_out, loff_t pos_out, size_t len)
2067 {
2068 	struct inode *src = file_inode(file_in);
2069 	struct inode *dst = file_inode(file_out);
2070 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2071 	size_t olen = len, dst_max_i_size = 0;
2072 	size_t dst_osize;
2073 	int ret;
2074 
2075 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2076 				src->i_sb != dst->i_sb)
2077 		return -EXDEV;
2078 
2079 	if (unlikely(f2fs_readonly(src->i_sb)))
2080 		return -EROFS;
2081 
2082 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2083 		return -EINVAL;
2084 
2085 	if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2086 		return -EOPNOTSUPP;
2087 
2088 	inode_lock(src);
2089 	if (src != dst) {
2090 		if (!inode_trylock(dst)) {
2091 			ret = -EBUSY;
2092 			goto out;
2093 		}
2094 	}
2095 
2096 	ret = -EINVAL;
2097 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2098 		goto out_unlock;
2099 	if (len == 0)
2100 		olen = len = src->i_size - pos_in;
2101 	if (pos_in + len == src->i_size)
2102 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2103 	if (len == 0) {
2104 		ret = 0;
2105 		goto out_unlock;
2106 	}
2107 
2108 	dst_osize = dst->i_size;
2109 	if (pos_out + olen > dst->i_size)
2110 		dst_max_i_size = pos_out + olen;
2111 
2112 	/* verify the end result is block aligned */
2113 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2114 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2115 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2116 		goto out_unlock;
2117 
2118 	ret = f2fs_convert_inline_inode(src);
2119 	if (ret)
2120 		goto out_unlock;
2121 
2122 	ret = f2fs_convert_inline_inode(dst);
2123 	if (ret)
2124 		goto out_unlock;
2125 
2126 	/* write out all dirty pages from offset */
2127 	ret = filemap_write_and_wait_range(src->i_mapping,
2128 					pos_in, pos_in + len);
2129 	if (ret)
2130 		goto out_unlock;
2131 
2132 	ret = filemap_write_and_wait_range(dst->i_mapping,
2133 					pos_out, pos_out + len);
2134 	if (ret)
2135 		goto out_unlock;
2136 
2137 	f2fs_balance_fs(sbi, true);
2138 	f2fs_lock_op(sbi);
2139 	ret = __exchange_data_block(src, dst, pos_in,
2140 				pos_out, len >> F2FS_BLKSIZE_BITS, false);
2141 
2142 	if (!ret) {
2143 		if (dst_max_i_size)
2144 			f2fs_i_size_write(dst, dst_max_i_size);
2145 		else if (dst_osize != dst->i_size)
2146 			f2fs_i_size_write(dst, dst_osize);
2147 	}
2148 	f2fs_unlock_op(sbi);
2149 out_unlock:
2150 	if (src != dst)
2151 		inode_unlock(dst);
2152 out:
2153 	inode_unlock(src);
2154 	return ret;
2155 }
2156 
2157 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2158 {
2159 	struct f2fs_move_range range;
2160 	struct fd dst;
2161 	int err;
2162 
2163 	if (!(filp->f_mode & FMODE_READ) ||
2164 			!(filp->f_mode & FMODE_WRITE))
2165 		return -EBADF;
2166 
2167 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2168 							sizeof(range)))
2169 		return -EFAULT;
2170 
2171 	dst = fdget(range.dst_fd);
2172 	if (!dst.file)
2173 		return -EBADF;
2174 
2175 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2176 		err = -EBADF;
2177 		goto err_out;
2178 	}
2179 
2180 	err = mnt_want_write_file(filp);
2181 	if (err)
2182 		goto err_out;
2183 
2184 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2185 					range.pos_out, range.len);
2186 
2187 	mnt_drop_write_file(filp);
2188 
2189 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2190 						&range, sizeof(range)))
2191 		err = -EFAULT;
2192 err_out:
2193 	fdput(dst);
2194 	return err;
2195 }
2196 
2197 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2198 {
2199 	switch (cmd) {
2200 	case F2FS_IOC_GETFLAGS:
2201 		return f2fs_ioc_getflags(filp, arg);
2202 	case F2FS_IOC_SETFLAGS:
2203 		return f2fs_ioc_setflags(filp, arg);
2204 	case F2FS_IOC_GETVERSION:
2205 		return f2fs_ioc_getversion(filp, arg);
2206 	case F2FS_IOC_START_ATOMIC_WRITE:
2207 		return f2fs_ioc_start_atomic_write(filp);
2208 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2209 		return f2fs_ioc_commit_atomic_write(filp);
2210 	case F2FS_IOC_START_VOLATILE_WRITE:
2211 		return f2fs_ioc_start_volatile_write(filp);
2212 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2213 		return f2fs_ioc_release_volatile_write(filp);
2214 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
2215 		return f2fs_ioc_abort_volatile_write(filp);
2216 	case F2FS_IOC_SHUTDOWN:
2217 		return f2fs_ioc_shutdown(filp, arg);
2218 	case FITRIM:
2219 		return f2fs_ioc_fitrim(filp, arg);
2220 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
2221 		return f2fs_ioc_set_encryption_policy(filp, arg);
2222 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
2223 		return f2fs_ioc_get_encryption_policy(filp, arg);
2224 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2225 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2226 	case F2FS_IOC_GARBAGE_COLLECT:
2227 		return f2fs_ioc_gc(filp, arg);
2228 	case F2FS_IOC_WRITE_CHECKPOINT:
2229 		return f2fs_ioc_write_checkpoint(filp, arg);
2230 	case F2FS_IOC_DEFRAGMENT:
2231 		return f2fs_ioc_defragment(filp, arg);
2232 	case F2FS_IOC_MOVE_RANGE:
2233 		return f2fs_ioc_move_range(filp, arg);
2234 	default:
2235 		return -ENOTTY;
2236 	}
2237 }
2238 
2239 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2240 {
2241 	struct file *file = iocb->ki_filp;
2242 	struct inode *inode = file_inode(file);
2243 	struct blk_plug plug;
2244 	ssize_t ret;
2245 
2246 	if (f2fs_encrypted_inode(inode) &&
2247 				!fscrypt_has_encryption_key(inode) &&
2248 				fscrypt_get_encryption_info(inode))
2249 		return -EACCES;
2250 
2251 	inode_lock(inode);
2252 	ret = generic_write_checks(iocb, from);
2253 	if (ret > 0) {
2254 		ret = f2fs_preallocate_blocks(iocb, from);
2255 		if (!ret) {
2256 			blk_start_plug(&plug);
2257 			ret = __generic_file_write_iter(iocb, from);
2258 			blk_finish_plug(&plug);
2259 		}
2260 	}
2261 	inode_unlock(inode);
2262 
2263 	if (ret > 0)
2264 		ret = generic_write_sync(iocb, ret);
2265 	return ret;
2266 }
2267 
2268 #ifdef CONFIG_COMPAT
2269 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2270 {
2271 	switch (cmd) {
2272 	case F2FS_IOC32_GETFLAGS:
2273 		cmd = F2FS_IOC_GETFLAGS;
2274 		break;
2275 	case F2FS_IOC32_SETFLAGS:
2276 		cmd = F2FS_IOC_SETFLAGS;
2277 		break;
2278 	case F2FS_IOC32_GETVERSION:
2279 		cmd = F2FS_IOC_GETVERSION;
2280 		break;
2281 	case F2FS_IOC_START_ATOMIC_WRITE:
2282 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2283 	case F2FS_IOC_START_VOLATILE_WRITE:
2284 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2285 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
2286 	case F2FS_IOC_SHUTDOWN:
2287 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
2288 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2289 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
2290 	case F2FS_IOC_GARBAGE_COLLECT:
2291 	case F2FS_IOC_WRITE_CHECKPOINT:
2292 	case F2FS_IOC_DEFRAGMENT:
2293 		break;
2294 	case F2FS_IOC_MOVE_RANGE:
2295 		break;
2296 	default:
2297 		return -ENOIOCTLCMD;
2298 	}
2299 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2300 }
2301 #endif
2302 
2303 const struct file_operations f2fs_file_operations = {
2304 	.llseek		= f2fs_llseek,
2305 	.read_iter	= generic_file_read_iter,
2306 	.write_iter	= f2fs_file_write_iter,
2307 	.open		= f2fs_file_open,
2308 	.release	= f2fs_release_file,
2309 	.mmap		= f2fs_file_mmap,
2310 	.fsync		= f2fs_sync_file,
2311 	.fallocate	= f2fs_fallocate,
2312 	.unlocked_ioctl	= f2fs_ioctl,
2313 #ifdef CONFIG_COMPAT
2314 	.compat_ioctl	= f2fs_compat_ioctl,
2315 #endif
2316 	.splice_read	= generic_file_splice_read,
2317 	.splice_write	= iter_file_splice_write,
2318 };
2319