xref: /openbmc/linux/fs/f2fs/file.c (revision a17922de)
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 vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
37 {
38 	struct inode *inode = file_inode(vmf->vma->vm_file);
39 	vm_fault_t ret;
40 
41 	down_read(&F2FS_I(inode)->i_mmap_sem);
42 	ret = filemap_fault(vmf);
43 	up_read(&F2FS_I(inode)->i_mmap_sem);
44 
45 	return ret;
46 }
47 
48 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 {
50 	struct page *page = vmf->page;
51 	struct inode *inode = file_inode(vmf->vma->vm_file);
52 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 	struct dnode_of_data dn;
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 		loff_t offset;
99 
100 		offset = i_size_read(inode) & ~PAGE_MASK;
101 		zero_user_segment(page, offset, PAGE_SIZE);
102 	}
103 	set_page_dirty(page);
104 	if (!PageUptodate(page))
105 		SetPageUptodate(page);
106 
107 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
108 
109 	trace_f2fs_vm_page_mkwrite(page, DATA);
110 mapped:
111 	/* fill the page */
112 	f2fs_wait_on_page_writeback(page, DATA, false);
113 
114 	/* wait for GCed page writeback via META_MAPPING */
115 	if (f2fs_post_read_required(inode))
116 		f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
117 
118 out_sem:
119 	up_read(&F2FS_I(inode)->i_mmap_sem);
120 out:
121 	sb_end_pagefault(inode->i_sb);
122 	f2fs_update_time(sbi, REQ_TIME);
123 err:
124 	return block_page_mkwrite_return(err);
125 }
126 
127 static const struct vm_operations_struct f2fs_file_vm_ops = {
128 	.fault		= f2fs_filemap_fault,
129 	.map_pages	= filemap_map_pages,
130 	.page_mkwrite	= f2fs_vm_page_mkwrite,
131 };
132 
133 static int get_parent_ino(struct inode *inode, nid_t *pino)
134 {
135 	struct dentry *dentry;
136 
137 	inode = igrab(inode);
138 	dentry = d_find_any_alias(inode);
139 	iput(inode);
140 	if (!dentry)
141 		return 0;
142 
143 	*pino = parent_ino(dentry);
144 	dput(dentry);
145 	return 1;
146 }
147 
148 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
149 {
150 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
151 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
152 
153 	if (!S_ISREG(inode->i_mode))
154 		cp_reason = CP_NON_REGULAR;
155 	else if (inode->i_nlink != 1)
156 		cp_reason = CP_HARDLINK;
157 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
158 		cp_reason = CP_SB_NEED_CP;
159 	else if (file_wrong_pino(inode))
160 		cp_reason = CP_WRONG_PINO;
161 	else if (!f2fs_space_for_roll_forward(sbi))
162 		cp_reason = CP_NO_SPC_ROLL;
163 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
164 		cp_reason = CP_NODE_NEED_CP;
165 	else if (test_opt(sbi, FASTBOOT))
166 		cp_reason = CP_FASTBOOT_MODE;
167 	else if (F2FS_OPTION(sbi).active_logs == 2)
168 		cp_reason = CP_SPEC_LOG_NUM;
169 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
170 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
171 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
172 							TRANS_DIR_INO))
173 		cp_reason = CP_RECOVER_DIR;
174 
175 	return cp_reason;
176 }
177 
178 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
179 {
180 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
181 	bool ret = false;
182 	/* But we need to avoid that there are some inode updates */
183 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
184 		ret = true;
185 	f2fs_put_page(i, 0);
186 	return ret;
187 }
188 
189 static void try_to_fix_pino(struct inode *inode)
190 {
191 	struct f2fs_inode_info *fi = F2FS_I(inode);
192 	nid_t pino;
193 
194 	down_write(&fi->i_sem);
195 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
196 			get_parent_ino(inode, &pino)) {
197 		f2fs_i_pino_write(inode, pino);
198 		file_got_pino(inode);
199 	}
200 	up_write(&fi->i_sem);
201 }
202 
203 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
204 						int datasync, bool atomic)
205 {
206 	struct inode *inode = file->f_mapping->host;
207 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
208 	nid_t ino = inode->i_ino;
209 	int ret = 0;
210 	enum cp_reason_type cp_reason = 0;
211 	struct writeback_control wbc = {
212 		.sync_mode = WB_SYNC_ALL,
213 		.nr_to_write = LONG_MAX,
214 		.for_reclaim = 0,
215 	};
216 
217 	if (unlikely(f2fs_readonly(inode->i_sb)))
218 		return 0;
219 
220 	trace_f2fs_sync_file_enter(inode);
221 
222 	/* if fdatasync is triggered, let's do in-place-update */
223 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
224 		set_inode_flag(inode, FI_NEED_IPU);
225 	ret = file_write_and_wait_range(file, start, end);
226 	clear_inode_flag(inode, FI_NEED_IPU);
227 
228 	if (ret) {
229 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
230 		return ret;
231 	}
232 
233 	/* if the inode is dirty, let's recover all the time */
234 	if (!f2fs_skip_inode_update(inode, datasync)) {
235 		f2fs_write_inode(inode, NULL);
236 		goto go_write;
237 	}
238 
239 	/*
240 	 * if there is no written data, don't waste time to write recovery info.
241 	 */
242 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
243 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
244 
245 		/* it may call write_inode just prior to fsync */
246 		if (need_inode_page_update(sbi, ino))
247 			goto go_write;
248 
249 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
250 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
251 			goto flush_out;
252 		goto out;
253 	}
254 go_write:
255 	/*
256 	 * Both of fdatasync() and fsync() are able to be recovered from
257 	 * sudden-power-off.
258 	 */
259 	down_read(&F2FS_I(inode)->i_sem);
260 	cp_reason = need_do_checkpoint(inode);
261 	up_read(&F2FS_I(inode)->i_sem);
262 
263 	if (cp_reason) {
264 		/* all the dirty node pages should be flushed for POR */
265 		ret = f2fs_sync_fs(inode->i_sb, 1);
266 
267 		/*
268 		 * We've secured consistency through sync_fs. Following pino
269 		 * will be used only for fsynced inodes after checkpoint.
270 		 */
271 		try_to_fix_pino(inode);
272 		clear_inode_flag(inode, FI_APPEND_WRITE);
273 		clear_inode_flag(inode, FI_UPDATE_WRITE);
274 		goto out;
275 	}
276 sync_nodes:
277 	atomic_inc(&sbi->wb_sync_req[NODE]);
278 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic);
279 	atomic_dec(&sbi->wb_sync_req[NODE]);
280 	if (ret)
281 		goto out;
282 
283 	/* if cp_error was enabled, we should avoid infinite loop */
284 	if (unlikely(f2fs_cp_error(sbi))) {
285 		ret = -EIO;
286 		goto out;
287 	}
288 
289 	if (f2fs_need_inode_block_update(sbi, ino)) {
290 		f2fs_mark_inode_dirty_sync(inode, true);
291 		f2fs_write_inode(inode, NULL);
292 		goto sync_nodes;
293 	}
294 
295 	/*
296 	 * If it's atomic_write, it's just fine to keep write ordering. So
297 	 * here we don't need to wait for node write completion, since we use
298 	 * node chain which serializes node blocks. If one of node writes are
299 	 * reordered, we can see simply broken chain, resulting in stopping
300 	 * roll-forward recovery. It means we'll recover all or none node blocks
301 	 * given fsync mark.
302 	 */
303 	if (!atomic) {
304 		ret = f2fs_wait_on_node_pages_writeback(sbi, ino);
305 		if (ret)
306 			goto out;
307 	}
308 
309 	/* once recovery info is written, don't need to tack this */
310 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
311 	clear_inode_flag(inode, FI_APPEND_WRITE);
312 flush_out:
313 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
314 		ret = f2fs_issue_flush(sbi, inode->i_ino);
315 	if (!ret) {
316 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
317 		clear_inode_flag(inode, FI_UPDATE_WRITE);
318 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
319 	}
320 	f2fs_update_time(sbi, REQ_TIME);
321 out:
322 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
323 	f2fs_trace_ios(NULL, 1);
324 	return ret;
325 }
326 
327 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
328 {
329 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
330 		return -EIO;
331 	return f2fs_do_sync_file(file, start, end, datasync, false);
332 }
333 
334 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
335 						pgoff_t pgofs, int whence)
336 {
337 	struct page *page;
338 	int nr_pages;
339 
340 	if (whence != SEEK_DATA)
341 		return 0;
342 
343 	/* find first dirty page index */
344 	nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
345 				      1, &page);
346 	if (!nr_pages)
347 		return ULONG_MAX;
348 	pgofs = page->index;
349 	put_page(page);
350 	return pgofs;
351 }
352 
353 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
354 							int whence)
355 {
356 	switch (whence) {
357 	case SEEK_DATA:
358 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
359 			is_valid_blkaddr(blkaddr))
360 			return true;
361 		break;
362 	case SEEK_HOLE:
363 		if (blkaddr == NULL_ADDR)
364 			return true;
365 		break;
366 	}
367 	return false;
368 }
369 
370 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
371 {
372 	struct inode *inode = file->f_mapping->host;
373 	loff_t maxbytes = inode->i_sb->s_maxbytes;
374 	struct dnode_of_data dn;
375 	pgoff_t pgofs, end_offset, dirty;
376 	loff_t data_ofs = offset;
377 	loff_t isize;
378 	int err = 0;
379 
380 	inode_lock(inode);
381 
382 	isize = i_size_read(inode);
383 	if (offset >= isize)
384 		goto fail;
385 
386 	/* handle inline data case */
387 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
388 		if (whence == SEEK_HOLE)
389 			data_ofs = isize;
390 		goto found;
391 	}
392 
393 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
394 
395 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
396 
397 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
398 		set_new_dnode(&dn, inode, NULL, NULL, 0);
399 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
400 		if (err && err != -ENOENT) {
401 			goto fail;
402 		} else if (err == -ENOENT) {
403 			/* direct node does not exists */
404 			if (whence == SEEK_DATA) {
405 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
406 				continue;
407 			} else {
408 				goto found;
409 			}
410 		}
411 
412 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
413 
414 		/* find data/hole in dnode block */
415 		for (; dn.ofs_in_node < end_offset;
416 				dn.ofs_in_node++, pgofs++,
417 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
418 			block_t blkaddr;
419 
420 			blkaddr = datablock_addr(dn.inode,
421 					dn.node_page, dn.ofs_in_node);
422 
423 			if (__found_offset(blkaddr, dirty, pgofs, whence)) {
424 				f2fs_put_dnode(&dn);
425 				goto found;
426 			}
427 		}
428 		f2fs_put_dnode(&dn);
429 	}
430 
431 	if (whence == SEEK_DATA)
432 		goto fail;
433 found:
434 	if (whence == SEEK_HOLE && data_ofs > isize)
435 		data_ofs = isize;
436 	inode_unlock(inode);
437 	return vfs_setpos(file, data_ofs, maxbytes);
438 fail:
439 	inode_unlock(inode);
440 	return -ENXIO;
441 }
442 
443 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
444 {
445 	struct inode *inode = file->f_mapping->host;
446 	loff_t maxbytes = inode->i_sb->s_maxbytes;
447 
448 	switch (whence) {
449 	case SEEK_SET:
450 	case SEEK_CUR:
451 	case SEEK_END:
452 		return generic_file_llseek_size(file, offset, whence,
453 						maxbytes, i_size_read(inode));
454 	case SEEK_DATA:
455 	case SEEK_HOLE:
456 		if (offset < 0)
457 			return -ENXIO;
458 		return f2fs_seek_block(file, offset, whence);
459 	}
460 
461 	return -EINVAL;
462 }
463 
464 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
465 {
466 	struct inode *inode = file_inode(file);
467 	int err;
468 
469 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
470 		return -EIO;
471 
472 	/* we don't need to use inline_data strictly */
473 	err = f2fs_convert_inline_inode(inode);
474 	if (err)
475 		return err;
476 
477 	file_accessed(file);
478 	vma->vm_ops = &f2fs_file_vm_ops;
479 	return 0;
480 }
481 
482 static int f2fs_file_open(struct inode *inode, struct file *filp)
483 {
484 	int err = fscrypt_file_open(inode, filp);
485 
486 	if (err)
487 		return err;
488 
489 	filp->f_mode |= FMODE_NOWAIT;
490 
491 	return dquot_file_open(inode, filp);
492 }
493 
494 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
495 {
496 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
497 	struct f2fs_node *raw_node;
498 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
499 	__le32 *addr;
500 	int base = 0;
501 
502 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
503 		base = get_extra_isize(dn->inode);
504 
505 	raw_node = F2FS_NODE(dn->node_page);
506 	addr = blkaddr_in_node(raw_node) + base + ofs;
507 
508 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
509 		block_t blkaddr = le32_to_cpu(*addr);
510 
511 		if (blkaddr == NULL_ADDR)
512 			continue;
513 
514 		dn->data_blkaddr = NULL_ADDR;
515 		f2fs_set_data_blkaddr(dn);
516 		f2fs_invalidate_blocks(sbi, blkaddr);
517 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
518 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
519 		nr_free++;
520 	}
521 
522 	if (nr_free) {
523 		pgoff_t fofs;
524 		/*
525 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
526 		 * we will invalidate all blkaddr in the whole range.
527 		 */
528 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
529 							dn->inode) + ofs;
530 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
531 		dec_valid_block_count(sbi, dn->inode, nr_free);
532 	}
533 	dn->ofs_in_node = ofs;
534 
535 	f2fs_update_time(sbi, REQ_TIME);
536 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
537 					 dn->ofs_in_node, nr_free);
538 }
539 
540 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
541 {
542 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
543 }
544 
545 static int truncate_partial_data_page(struct inode *inode, u64 from,
546 								bool cache_only)
547 {
548 	loff_t offset = from & (PAGE_SIZE - 1);
549 	pgoff_t index = from >> PAGE_SHIFT;
550 	struct address_space *mapping = inode->i_mapping;
551 	struct page *page;
552 
553 	if (!offset && !cache_only)
554 		return 0;
555 
556 	if (cache_only) {
557 		page = find_lock_page(mapping, index);
558 		if (page && PageUptodate(page))
559 			goto truncate_out;
560 		f2fs_put_page(page, 1);
561 		return 0;
562 	}
563 
564 	page = f2fs_get_lock_data_page(inode, index, true);
565 	if (IS_ERR(page))
566 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
567 truncate_out:
568 	f2fs_wait_on_page_writeback(page, DATA, true);
569 	zero_user(page, offset, PAGE_SIZE - offset);
570 
571 	/* An encrypted inode should have a key and truncate the last page. */
572 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
573 	if (!cache_only)
574 		set_page_dirty(page);
575 	f2fs_put_page(page, 1);
576 	return 0;
577 }
578 
579 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
580 {
581 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
582 	struct dnode_of_data dn;
583 	pgoff_t free_from;
584 	int count = 0, err = 0;
585 	struct page *ipage;
586 	bool truncate_page = false;
587 
588 	trace_f2fs_truncate_blocks_enter(inode, from);
589 
590 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
591 
592 	if (free_from >= sbi->max_file_blocks)
593 		goto free_partial;
594 
595 	if (lock)
596 		f2fs_lock_op(sbi);
597 
598 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
599 	if (IS_ERR(ipage)) {
600 		err = PTR_ERR(ipage);
601 		goto out;
602 	}
603 
604 	if (f2fs_has_inline_data(inode)) {
605 		f2fs_truncate_inline_inode(inode, ipage, from);
606 		f2fs_put_page(ipage, 1);
607 		truncate_page = true;
608 		goto out;
609 	}
610 
611 	set_new_dnode(&dn, inode, ipage, NULL, 0);
612 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
613 	if (err) {
614 		if (err == -ENOENT)
615 			goto free_next;
616 		goto out;
617 	}
618 
619 	count = ADDRS_PER_PAGE(dn.node_page, inode);
620 
621 	count -= dn.ofs_in_node;
622 	f2fs_bug_on(sbi, count < 0);
623 
624 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
625 		f2fs_truncate_data_blocks_range(&dn, count);
626 		free_from += count;
627 	}
628 
629 	f2fs_put_dnode(&dn);
630 free_next:
631 	err = f2fs_truncate_inode_blocks(inode, free_from);
632 out:
633 	if (lock)
634 		f2fs_unlock_op(sbi);
635 free_partial:
636 	/* lastly zero out the first data page */
637 	if (!err)
638 		err = truncate_partial_data_page(inode, from, truncate_page);
639 
640 	trace_f2fs_truncate_blocks_exit(inode, err);
641 	return err;
642 }
643 
644 int f2fs_truncate(struct inode *inode)
645 {
646 	int err;
647 
648 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
649 		return -EIO;
650 
651 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
652 				S_ISLNK(inode->i_mode)))
653 		return 0;
654 
655 	trace_f2fs_truncate(inode);
656 
657 #ifdef CONFIG_F2FS_FAULT_INJECTION
658 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
659 		f2fs_show_injection_info(FAULT_TRUNCATE);
660 		return -EIO;
661 	}
662 #endif
663 	/* we should check inline_data size */
664 	if (!f2fs_may_inline_data(inode)) {
665 		err = f2fs_convert_inline_inode(inode);
666 		if (err)
667 			return err;
668 	}
669 
670 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
671 	if (err)
672 		return err;
673 
674 	inode->i_mtime = inode->i_ctime = current_time(inode);
675 	f2fs_mark_inode_dirty_sync(inode, false);
676 	return 0;
677 }
678 
679 int f2fs_getattr(const struct path *path, struct kstat *stat,
680 		 u32 request_mask, unsigned int query_flags)
681 {
682 	struct inode *inode = d_inode(path->dentry);
683 	struct f2fs_inode_info *fi = F2FS_I(inode);
684 	struct f2fs_inode *ri;
685 	unsigned int flags;
686 
687 	if (f2fs_has_extra_attr(inode) &&
688 			f2fs_sb_has_inode_crtime(inode->i_sb) &&
689 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
690 		stat->result_mask |= STATX_BTIME;
691 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
692 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
693 	}
694 
695 	flags = fi->i_flags & F2FS_FL_USER_VISIBLE;
696 	if (flags & F2FS_APPEND_FL)
697 		stat->attributes |= STATX_ATTR_APPEND;
698 	if (flags & F2FS_COMPR_FL)
699 		stat->attributes |= STATX_ATTR_COMPRESSED;
700 	if (f2fs_encrypted_inode(inode))
701 		stat->attributes |= STATX_ATTR_ENCRYPTED;
702 	if (flags & F2FS_IMMUTABLE_FL)
703 		stat->attributes |= STATX_ATTR_IMMUTABLE;
704 	if (flags & F2FS_NODUMP_FL)
705 		stat->attributes |= STATX_ATTR_NODUMP;
706 
707 	stat->attributes_mask |= (STATX_ATTR_APPEND |
708 				  STATX_ATTR_COMPRESSED |
709 				  STATX_ATTR_ENCRYPTED |
710 				  STATX_ATTR_IMMUTABLE |
711 				  STATX_ATTR_NODUMP);
712 
713 	generic_fillattr(inode, stat);
714 
715 	/* we need to show initial sectors used for inline_data/dentries */
716 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
717 					f2fs_has_inline_dentry(inode))
718 		stat->blocks += (stat->size + 511) >> 9;
719 
720 	return 0;
721 }
722 
723 #ifdef CONFIG_F2FS_FS_POSIX_ACL
724 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
725 {
726 	unsigned int ia_valid = attr->ia_valid;
727 
728 	if (ia_valid & ATTR_UID)
729 		inode->i_uid = attr->ia_uid;
730 	if (ia_valid & ATTR_GID)
731 		inode->i_gid = attr->ia_gid;
732 	if (ia_valid & ATTR_ATIME)
733 		inode->i_atime = timespec64_trunc(attr->ia_atime,
734 						  inode->i_sb->s_time_gran);
735 	if (ia_valid & ATTR_MTIME)
736 		inode->i_mtime = timespec64_trunc(attr->ia_mtime,
737 						  inode->i_sb->s_time_gran);
738 	if (ia_valid & ATTR_CTIME)
739 		inode->i_ctime = timespec64_trunc(attr->ia_ctime,
740 						  inode->i_sb->s_time_gran);
741 	if (ia_valid & ATTR_MODE) {
742 		umode_t mode = attr->ia_mode;
743 
744 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
745 			mode &= ~S_ISGID;
746 		set_acl_inode(inode, mode);
747 	}
748 }
749 #else
750 #define __setattr_copy setattr_copy
751 #endif
752 
753 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
754 {
755 	struct inode *inode = d_inode(dentry);
756 	int err;
757 	bool size_changed = false;
758 
759 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
760 		return -EIO;
761 
762 	err = setattr_prepare(dentry, attr);
763 	if (err)
764 		return err;
765 
766 	err = fscrypt_prepare_setattr(dentry, attr);
767 	if (err)
768 		return err;
769 
770 	if (is_quota_modification(inode, attr)) {
771 		err = dquot_initialize(inode);
772 		if (err)
773 			return err;
774 	}
775 	if ((attr->ia_valid & ATTR_UID &&
776 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
777 		(attr->ia_valid & ATTR_GID &&
778 		!gid_eq(attr->ia_gid, inode->i_gid))) {
779 		err = dquot_transfer(inode, attr);
780 		if (err)
781 			return err;
782 	}
783 
784 	if (attr->ia_valid & ATTR_SIZE) {
785 		if (attr->ia_size <= i_size_read(inode)) {
786 			down_write(&F2FS_I(inode)->i_mmap_sem);
787 			truncate_setsize(inode, attr->ia_size);
788 			err = f2fs_truncate(inode);
789 			up_write(&F2FS_I(inode)->i_mmap_sem);
790 			if (err)
791 				return err;
792 		} else {
793 			/*
794 			 * do not trim all blocks after i_size if target size is
795 			 * larger than i_size.
796 			 */
797 			down_write(&F2FS_I(inode)->i_mmap_sem);
798 			truncate_setsize(inode, attr->ia_size);
799 			up_write(&F2FS_I(inode)->i_mmap_sem);
800 
801 			/* should convert inline inode here */
802 			if (!f2fs_may_inline_data(inode)) {
803 				err = f2fs_convert_inline_inode(inode);
804 				if (err)
805 					return err;
806 			}
807 			inode->i_mtime = inode->i_ctime = current_time(inode);
808 		}
809 
810 		down_write(&F2FS_I(inode)->i_sem);
811 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
812 		up_write(&F2FS_I(inode)->i_sem);
813 
814 		size_changed = true;
815 	}
816 
817 	__setattr_copy(inode, attr);
818 
819 	if (attr->ia_valid & ATTR_MODE) {
820 		err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
821 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
822 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
823 			clear_inode_flag(inode, FI_ACL_MODE);
824 		}
825 	}
826 
827 	/* file size may changed here */
828 	f2fs_mark_inode_dirty_sync(inode, size_changed);
829 
830 	/* inode change will produce dirty node pages flushed by checkpoint */
831 	f2fs_balance_fs(F2FS_I_SB(inode), true);
832 
833 	return err;
834 }
835 
836 const struct inode_operations f2fs_file_inode_operations = {
837 	.getattr	= f2fs_getattr,
838 	.setattr	= f2fs_setattr,
839 	.get_acl	= f2fs_get_acl,
840 	.set_acl	= f2fs_set_acl,
841 #ifdef CONFIG_F2FS_FS_XATTR
842 	.listxattr	= f2fs_listxattr,
843 #endif
844 	.fiemap		= f2fs_fiemap,
845 };
846 
847 static int fill_zero(struct inode *inode, pgoff_t index,
848 					loff_t start, loff_t len)
849 {
850 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
851 	struct page *page;
852 
853 	if (!len)
854 		return 0;
855 
856 	f2fs_balance_fs(sbi, true);
857 
858 	f2fs_lock_op(sbi);
859 	page = f2fs_get_new_data_page(inode, NULL, index, false);
860 	f2fs_unlock_op(sbi);
861 
862 	if (IS_ERR(page))
863 		return PTR_ERR(page);
864 
865 	f2fs_wait_on_page_writeback(page, DATA, true);
866 	zero_user(page, start, len);
867 	set_page_dirty(page);
868 	f2fs_put_page(page, 1);
869 	return 0;
870 }
871 
872 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
873 {
874 	int err;
875 
876 	while (pg_start < pg_end) {
877 		struct dnode_of_data dn;
878 		pgoff_t end_offset, count;
879 
880 		set_new_dnode(&dn, inode, NULL, NULL, 0);
881 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
882 		if (err) {
883 			if (err == -ENOENT) {
884 				pg_start = f2fs_get_next_page_offset(&dn,
885 								pg_start);
886 				continue;
887 			}
888 			return err;
889 		}
890 
891 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
892 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
893 
894 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
895 
896 		f2fs_truncate_data_blocks_range(&dn, count);
897 		f2fs_put_dnode(&dn);
898 
899 		pg_start += count;
900 	}
901 	return 0;
902 }
903 
904 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
905 {
906 	pgoff_t pg_start, pg_end;
907 	loff_t off_start, off_end;
908 	int ret;
909 
910 	ret = f2fs_convert_inline_inode(inode);
911 	if (ret)
912 		return ret;
913 
914 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
915 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
916 
917 	off_start = offset & (PAGE_SIZE - 1);
918 	off_end = (offset + len) & (PAGE_SIZE - 1);
919 
920 	if (pg_start == pg_end) {
921 		ret = fill_zero(inode, pg_start, off_start,
922 						off_end - off_start);
923 		if (ret)
924 			return ret;
925 	} else {
926 		if (off_start) {
927 			ret = fill_zero(inode, pg_start++, off_start,
928 						PAGE_SIZE - off_start);
929 			if (ret)
930 				return ret;
931 		}
932 		if (off_end) {
933 			ret = fill_zero(inode, pg_end, 0, off_end);
934 			if (ret)
935 				return ret;
936 		}
937 
938 		if (pg_start < pg_end) {
939 			struct address_space *mapping = inode->i_mapping;
940 			loff_t blk_start, blk_end;
941 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
942 
943 			f2fs_balance_fs(sbi, true);
944 
945 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
946 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
947 			down_write(&F2FS_I(inode)->i_mmap_sem);
948 			truncate_inode_pages_range(mapping, blk_start,
949 					blk_end - 1);
950 
951 			f2fs_lock_op(sbi);
952 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
953 			f2fs_unlock_op(sbi);
954 			up_write(&F2FS_I(inode)->i_mmap_sem);
955 		}
956 	}
957 
958 	return ret;
959 }
960 
961 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
962 				int *do_replace, pgoff_t off, pgoff_t len)
963 {
964 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
965 	struct dnode_of_data dn;
966 	int ret, done, i;
967 
968 next_dnode:
969 	set_new_dnode(&dn, inode, NULL, NULL, 0);
970 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
971 	if (ret && ret != -ENOENT) {
972 		return ret;
973 	} else if (ret == -ENOENT) {
974 		if (dn.max_level == 0)
975 			return -ENOENT;
976 		done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
977 		blkaddr += done;
978 		do_replace += done;
979 		goto next;
980 	}
981 
982 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
983 							dn.ofs_in_node, len);
984 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
985 		*blkaddr = datablock_addr(dn.inode,
986 					dn.node_page, dn.ofs_in_node);
987 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
988 
989 			if (test_opt(sbi, LFS)) {
990 				f2fs_put_dnode(&dn);
991 				return -ENOTSUPP;
992 			}
993 
994 			/* do not invalidate this block address */
995 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
996 			*do_replace = 1;
997 		}
998 	}
999 	f2fs_put_dnode(&dn);
1000 next:
1001 	len -= done;
1002 	off += done;
1003 	if (len)
1004 		goto next_dnode;
1005 	return 0;
1006 }
1007 
1008 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1009 				int *do_replace, pgoff_t off, int len)
1010 {
1011 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1012 	struct dnode_of_data dn;
1013 	int ret, i;
1014 
1015 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1016 		if (*do_replace == 0)
1017 			continue;
1018 
1019 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1020 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1021 		if (ret) {
1022 			dec_valid_block_count(sbi, inode, 1);
1023 			f2fs_invalidate_blocks(sbi, *blkaddr);
1024 		} else {
1025 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1026 		}
1027 		f2fs_put_dnode(&dn);
1028 	}
1029 	return 0;
1030 }
1031 
1032 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1033 			block_t *blkaddr, int *do_replace,
1034 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1035 {
1036 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1037 	pgoff_t i = 0;
1038 	int ret;
1039 
1040 	while (i < len) {
1041 		if (blkaddr[i] == NULL_ADDR && !full) {
1042 			i++;
1043 			continue;
1044 		}
1045 
1046 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1047 			struct dnode_of_data dn;
1048 			struct node_info ni;
1049 			size_t new_size;
1050 			pgoff_t ilen;
1051 
1052 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1053 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1054 			if (ret)
1055 				return ret;
1056 
1057 			f2fs_get_node_info(sbi, dn.nid, &ni);
1058 			ilen = min((pgoff_t)
1059 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1060 						dn.ofs_in_node, len - i);
1061 			do {
1062 				dn.data_blkaddr = datablock_addr(dn.inode,
1063 						dn.node_page, dn.ofs_in_node);
1064 				f2fs_truncate_data_blocks_range(&dn, 1);
1065 
1066 				if (do_replace[i]) {
1067 					f2fs_i_blocks_write(src_inode,
1068 							1, false, false);
1069 					f2fs_i_blocks_write(dst_inode,
1070 							1, true, false);
1071 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1072 					blkaddr[i], ni.version, true, false);
1073 
1074 					do_replace[i] = 0;
1075 				}
1076 				dn.ofs_in_node++;
1077 				i++;
1078 				new_size = (dst + i) << PAGE_SHIFT;
1079 				if (dst_inode->i_size < new_size)
1080 					f2fs_i_size_write(dst_inode, new_size);
1081 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1082 
1083 			f2fs_put_dnode(&dn);
1084 		} else {
1085 			struct page *psrc, *pdst;
1086 
1087 			psrc = f2fs_get_lock_data_page(src_inode,
1088 							src + i, true);
1089 			if (IS_ERR(psrc))
1090 				return PTR_ERR(psrc);
1091 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1092 								true);
1093 			if (IS_ERR(pdst)) {
1094 				f2fs_put_page(psrc, 1);
1095 				return PTR_ERR(pdst);
1096 			}
1097 			f2fs_copy_page(psrc, pdst);
1098 			set_page_dirty(pdst);
1099 			f2fs_put_page(pdst, 1);
1100 			f2fs_put_page(psrc, 1);
1101 
1102 			ret = f2fs_truncate_hole(src_inode,
1103 						src + i, src + i + 1);
1104 			if (ret)
1105 				return ret;
1106 			i++;
1107 		}
1108 	}
1109 	return 0;
1110 }
1111 
1112 static int __exchange_data_block(struct inode *src_inode,
1113 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1114 			pgoff_t len, bool full)
1115 {
1116 	block_t *src_blkaddr;
1117 	int *do_replace;
1118 	pgoff_t olen;
1119 	int ret;
1120 
1121 	while (len) {
1122 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1123 
1124 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1125 					array_size(olen, sizeof(block_t)),
1126 					GFP_KERNEL);
1127 		if (!src_blkaddr)
1128 			return -ENOMEM;
1129 
1130 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1131 					array_size(olen, sizeof(int)),
1132 					GFP_KERNEL);
1133 		if (!do_replace) {
1134 			kvfree(src_blkaddr);
1135 			return -ENOMEM;
1136 		}
1137 
1138 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1139 					do_replace, src, olen);
1140 		if (ret)
1141 			goto roll_back;
1142 
1143 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1144 					do_replace, src, dst, olen, full);
1145 		if (ret)
1146 			goto roll_back;
1147 
1148 		src += olen;
1149 		dst += olen;
1150 		len -= olen;
1151 
1152 		kvfree(src_blkaddr);
1153 		kvfree(do_replace);
1154 	}
1155 	return 0;
1156 
1157 roll_back:
1158 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1159 	kvfree(src_blkaddr);
1160 	kvfree(do_replace);
1161 	return ret;
1162 }
1163 
1164 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1165 {
1166 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1167 	pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1168 	int ret;
1169 
1170 	f2fs_balance_fs(sbi, true);
1171 	f2fs_lock_op(sbi);
1172 
1173 	f2fs_drop_extent_tree(inode);
1174 
1175 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1176 	f2fs_unlock_op(sbi);
1177 	return ret;
1178 }
1179 
1180 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1181 {
1182 	pgoff_t pg_start, pg_end;
1183 	loff_t new_size;
1184 	int ret;
1185 
1186 	if (offset + len >= i_size_read(inode))
1187 		return -EINVAL;
1188 
1189 	/* collapse range should be aligned to block size of f2fs. */
1190 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1191 		return -EINVAL;
1192 
1193 	ret = f2fs_convert_inline_inode(inode);
1194 	if (ret)
1195 		return ret;
1196 
1197 	pg_start = offset >> PAGE_SHIFT;
1198 	pg_end = (offset + len) >> PAGE_SHIFT;
1199 
1200 	/* avoid gc operation during block exchange */
1201 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1202 
1203 	down_write(&F2FS_I(inode)->i_mmap_sem);
1204 	/* write out all dirty pages from offset */
1205 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1206 	if (ret)
1207 		goto out_unlock;
1208 
1209 	truncate_pagecache(inode, offset);
1210 
1211 	ret = f2fs_do_collapse(inode, pg_start, pg_end);
1212 	if (ret)
1213 		goto out_unlock;
1214 
1215 	/* write out all moved pages, if possible */
1216 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1217 	truncate_pagecache(inode, offset);
1218 
1219 	new_size = i_size_read(inode) - len;
1220 	truncate_pagecache(inode, new_size);
1221 
1222 	ret = f2fs_truncate_blocks(inode, new_size, true);
1223 	if (!ret)
1224 		f2fs_i_size_write(inode, new_size);
1225 out_unlock:
1226 	up_write(&F2FS_I(inode)->i_mmap_sem);
1227 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1228 	return ret;
1229 }
1230 
1231 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1232 								pgoff_t end)
1233 {
1234 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1235 	pgoff_t index = start;
1236 	unsigned int ofs_in_node = dn->ofs_in_node;
1237 	blkcnt_t count = 0;
1238 	int ret;
1239 
1240 	for (; index < end; index++, dn->ofs_in_node++) {
1241 		if (datablock_addr(dn->inode, dn->node_page,
1242 					dn->ofs_in_node) == NULL_ADDR)
1243 			count++;
1244 	}
1245 
1246 	dn->ofs_in_node = ofs_in_node;
1247 	ret = f2fs_reserve_new_blocks(dn, count);
1248 	if (ret)
1249 		return ret;
1250 
1251 	dn->ofs_in_node = ofs_in_node;
1252 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1253 		dn->data_blkaddr = datablock_addr(dn->inode,
1254 					dn->node_page, dn->ofs_in_node);
1255 		/*
1256 		 * f2fs_reserve_new_blocks will not guarantee entire block
1257 		 * allocation.
1258 		 */
1259 		if (dn->data_blkaddr == NULL_ADDR) {
1260 			ret = -ENOSPC;
1261 			break;
1262 		}
1263 		if (dn->data_blkaddr != NEW_ADDR) {
1264 			f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1265 			dn->data_blkaddr = NEW_ADDR;
1266 			f2fs_set_data_blkaddr(dn);
1267 		}
1268 	}
1269 
1270 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1271 
1272 	return ret;
1273 }
1274 
1275 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1276 								int mode)
1277 {
1278 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1279 	struct address_space *mapping = inode->i_mapping;
1280 	pgoff_t index, pg_start, pg_end;
1281 	loff_t new_size = i_size_read(inode);
1282 	loff_t off_start, off_end;
1283 	int ret = 0;
1284 
1285 	ret = inode_newsize_ok(inode, (len + offset));
1286 	if (ret)
1287 		return ret;
1288 
1289 	ret = f2fs_convert_inline_inode(inode);
1290 	if (ret)
1291 		return ret;
1292 
1293 	down_write(&F2FS_I(inode)->i_mmap_sem);
1294 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1295 	if (ret)
1296 		goto out_sem;
1297 
1298 	truncate_pagecache_range(inode, offset, offset + len - 1);
1299 
1300 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1301 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1302 
1303 	off_start = offset & (PAGE_SIZE - 1);
1304 	off_end = (offset + len) & (PAGE_SIZE - 1);
1305 
1306 	if (pg_start == pg_end) {
1307 		ret = fill_zero(inode, pg_start, off_start,
1308 						off_end - off_start);
1309 		if (ret)
1310 			goto out_sem;
1311 
1312 		new_size = max_t(loff_t, new_size, offset + len);
1313 	} else {
1314 		if (off_start) {
1315 			ret = fill_zero(inode, pg_start++, off_start,
1316 						PAGE_SIZE - off_start);
1317 			if (ret)
1318 				goto out_sem;
1319 
1320 			new_size = max_t(loff_t, new_size,
1321 					(loff_t)pg_start << PAGE_SHIFT);
1322 		}
1323 
1324 		for (index = pg_start; index < pg_end;) {
1325 			struct dnode_of_data dn;
1326 			unsigned int end_offset;
1327 			pgoff_t end;
1328 
1329 			f2fs_lock_op(sbi);
1330 
1331 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1332 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1333 			if (ret) {
1334 				f2fs_unlock_op(sbi);
1335 				goto out;
1336 			}
1337 
1338 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1339 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1340 
1341 			ret = f2fs_do_zero_range(&dn, index, end);
1342 			f2fs_put_dnode(&dn);
1343 			f2fs_unlock_op(sbi);
1344 
1345 			f2fs_balance_fs(sbi, dn.node_changed);
1346 
1347 			if (ret)
1348 				goto out;
1349 
1350 			index = end;
1351 			new_size = max_t(loff_t, new_size,
1352 					(loff_t)index << PAGE_SHIFT);
1353 		}
1354 
1355 		if (off_end) {
1356 			ret = fill_zero(inode, pg_end, 0, off_end);
1357 			if (ret)
1358 				goto out;
1359 
1360 			new_size = max_t(loff_t, new_size, offset + len);
1361 		}
1362 	}
1363 
1364 out:
1365 	if (new_size > i_size_read(inode)) {
1366 		if (mode & FALLOC_FL_KEEP_SIZE)
1367 			file_set_keep_isize(inode);
1368 		else
1369 			f2fs_i_size_write(inode, new_size);
1370 	}
1371 out_sem:
1372 	up_write(&F2FS_I(inode)->i_mmap_sem);
1373 
1374 	return ret;
1375 }
1376 
1377 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1378 {
1379 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1380 	pgoff_t nr, pg_start, pg_end, delta, idx;
1381 	loff_t new_size;
1382 	int ret = 0;
1383 
1384 	new_size = i_size_read(inode) + len;
1385 	ret = inode_newsize_ok(inode, new_size);
1386 	if (ret)
1387 		return ret;
1388 
1389 	if (offset >= i_size_read(inode))
1390 		return -EINVAL;
1391 
1392 	/* insert range should be aligned to block size of f2fs. */
1393 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1394 		return -EINVAL;
1395 
1396 	ret = f2fs_convert_inline_inode(inode);
1397 	if (ret)
1398 		return ret;
1399 
1400 	f2fs_balance_fs(sbi, true);
1401 
1402 	/* avoid gc operation during block exchange */
1403 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1404 
1405 	down_write(&F2FS_I(inode)->i_mmap_sem);
1406 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1407 	if (ret)
1408 		goto out;
1409 
1410 	/* write out all dirty pages from offset */
1411 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1412 	if (ret)
1413 		goto out;
1414 
1415 	truncate_pagecache(inode, offset);
1416 
1417 	pg_start = offset >> PAGE_SHIFT;
1418 	pg_end = (offset + len) >> PAGE_SHIFT;
1419 	delta = pg_end - pg_start;
1420 	idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1421 
1422 	while (!ret && idx > pg_start) {
1423 		nr = idx - pg_start;
1424 		if (nr > delta)
1425 			nr = delta;
1426 		idx -= nr;
1427 
1428 		f2fs_lock_op(sbi);
1429 		f2fs_drop_extent_tree(inode);
1430 
1431 		ret = __exchange_data_block(inode, inode, idx,
1432 					idx + delta, nr, false);
1433 		f2fs_unlock_op(sbi);
1434 	}
1435 
1436 	/* write out all moved pages, if possible */
1437 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1438 	truncate_pagecache(inode, offset);
1439 
1440 	if (!ret)
1441 		f2fs_i_size_write(inode, new_size);
1442 out:
1443 	up_write(&F2FS_I(inode)->i_mmap_sem);
1444 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1445 	return ret;
1446 }
1447 
1448 static int expand_inode_data(struct inode *inode, loff_t offset,
1449 					loff_t len, int mode)
1450 {
1451 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1452 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1453 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1454 	pgoff_t pg_end;
1455 	loff_t new_size = i_size_read(inode);
1456 	loff_t off_end;
1457 	int err;
1458 
1459 	err = inode_newsize_ok(inode, (len + offset));
1460 	if (err)
1461 		return err;
1462 
1463 	err = f2fs_convert_inline_inode(inode);
1464 	if (err)
1465 		return err;
1466 
1467 	f2fs_balance_fs(sbi, true);
1468 
1469 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1470 	off_end = (offset + len) & (PAGE_SIZE - 1);
1471 
1472 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1473 	map.m_len = pg_end - map.m_lblk;
1474 	if (off_end)
1475 		map.m_len++;
1476 
1477 	err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1478 	if (err) {
1479 		pgoff_t last_off;
1480 
1481 		if (!map.m_len)
1482 			return err;
1483 
1484 		last_off = map.m_lblk + map.m_len - 1;
1485 
1486 		/* update new size to the failed position */
1487 		new_size = (last_off == pg_end) ? offset + len :
1488 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1489 	} else {
1490 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1491 	}
1492 
1493 	if (new_size > i_size_read(inode)) {
1494 		if (mode & FALLOC_FL_KEEP_SIZE)
1495 			file_set_keep_isize(inode);
1496 		else
1497 			f2fs_i_size_write(inode, new_size);
1498 	}
1499 
1500 	return err;
1501 }
1502 
1503 static long f2fs_fallocate(struct file *file, int mode,
1504 				loff_t offset, loff_t len)
1505 {
1506 	struct inode *inode = file_inode(file);
1507 	long ret = 0;
1508 
1509 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1510 		return -EIO;
1511 
1512 	/* f2fs only support ->fallocate for regular file */
1513 	if (!S_ISREG(inode->i_mode))
1514 		return -EINVAL;
1515 
1516 	if (f2fs_encrypted_inode(inode) &&
1517 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1518 		return -EOPNOTSUPP;
1519 
1520 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1521 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1522 			FALLOC_FL_INSERT_RANGE))
1523 		return -EOPNOTSUPP;
1524 
1525 	inode_lock(inode);
1526 
1527 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1528 		if (offset >= inode->i_size)
1529 			goto out;
1530 
1531 		ret = punch_hole(inode, offset, len);
1532 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1533 		ret = f2fs_collapse_range(inode, offset, len);
1534 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1535 		ret = f2fs_zero_range(inode, offset, len, mode);
1536 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1537 		ret = f2fs_insert_range(inode, offset, len);
1538 	} else {
1539 		ret = expand_inode_data(inode, offset, len, mode);
1540 	}
1541 
1542 	if (!ret) {
1543 		inode->i_mtime = inode->i_ctime = current_time(inode);
1544 		f2fs_mark_inode_dirty_sync(inode, false);
1545 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1546 	}
1547 
1548 out:
1549 	inode_unlock(inode);
1550 
1551 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1552 	return ret;
1553 }
1554 
1555 static int f2fs_release_file(struct inode *inode, struct file *filp)
1556 {
1557 	/*
1558 	 * f2fs_relase_file is called at every close calls. So we should
1559 	 * not drop any inmemory pages by close called by other process.
1560 	 */
1561 	if (!(filp->f_mode & FMODE_WRITE) ||
1562 			atomic_read(&inode->i_writecount) != 1)
1563 		return 0;
1564 
1565 	/* some remained atomic pages should discarded */
1566 	if (f2fs_is_atomic_file(inode))
1567 		f2fs_drop_inmem_pages(inode);
1568 	if (f2fs_is_volatile_file(inode)) {
1569 		set_inode_flag(inode, FI_DROP_CACHE);
1570 		filemap_fdatawrite(inode->i_mapping);
1571 		clear_inode_flag(inode, FI_DROP_CACHE);
1572 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1573 		stat_dec_volatile_write(inode);
1574 	}
1575 	return 0;
1576 }
1577 
1578 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1579 {
1580 	struct inode *inode = file_inode(file);
1581 
1582 	/*
1583 	 * If the process doing a transaction is crashed, we should do
1584 	 * roll-back. Otherwise, other reader/write can see corrupted database
1585 	 * until all the writers close its file. Since this should be done
1586 	 * before dropping file lock, it needs to do in ->flush.
1587 	 */
1588 	if (f2fs_is_atomic_file(inode) &&
1589 			F2FS_I(inode)->inmem_task == current)
1590 		f2fs_drop_inmem_pages(inode);
1591 	return 0;
1592 }
1593 
1594 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1595 {
1596 	struct inode *inode = file_inode(filp);
1597 	struct f2fs_inode_info *fi = F2FS_I(inode);
1598 	unsigned int flags = fi->i_flags;
1599 
1600 	if (file_is_encrypt(inode))
1601 		flags |= F2FS_ENCRYPT_FL;
1602 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1603 		flags |= F2FS_INLINE_DATA_FL;
1604 
1605 	flags &= F2FS_FL_USER_VISIBLE;
1606 
1607 	return put_user(flags, (int __user *)arg);
1608 }
1609 
1610 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1611 {
1612 	struct f2fs_inode_info *fi = F2FS_I(inode);
1613 	unsigned int oldflags;
1614 
1615 	/* Is it quota file? Do not allow user to mess with it */
1616 	if (IS_NOQUOTA(inode))
1617 		return -EPERM;
1618 
1619 	flags = f2fs_mask_flags(inode->i_mode, flags);
1620 
1621 	oldflags = fi->i_flags;
1622 
1623 	if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL))
1624 		if (!capable(CAP_LINUX_IMMUTABLE))
1625 			return -EPERM;
1626 
1627 	flags = flags & F2FS_FL_USER_MODIFIABLE;
1628 	flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE;
1629 	fi->i_flags = flags;
1630 
1631 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1632 		set_inode_flag(inode, FI_PROJ_INHERIT);
1633 	else
1634 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1635 
1636 	inode->i_ctime = current_time(inode);
1637 	f2fs_set_inode_flags(inode);
1638 	f2fs_mark_inode_dirty_sync(inode, false);
1639 	return 0;
1640 }
1641 
1642 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1643 {
1644 	struct inode *inode = file_inode(filp);
1645 	unsigned int flags;
1646 	int ret;
1647 
1648 	if (!inode_owner_or_capable(inode))
1649 		return -EACCES;
1650 
1651 	if (get_user(flags, (int __user *)arg))
1652 		return -EFAULT;
1653 
1654 	ret = mnt_want_write_file(filp);
1655 	if (ret)
1656 		return ret;
1657 
1658 	inode_lock(inode);
1659 
1660 	ret = __f2fs_ioc_setflags(inode, flags);
1661 
1662 	inode_unlock(inode);
1663 	mnt_drop_write_file(filp);
1664 	return ret;
1665 }
1666 
1667 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1668 {
1669 	struct inode *inode = file_inode(filp);
1670 
1671 	return put_user(inode->i_generation, (int __user *)arg);
1672 }
1673 
1674 static int f2fs_ioc_start_atomic_write(struct file *filp)
1675 {
1676 	struct inode *inode = file_inode(filp);
1677 	int ret;
1678 
1679 	if (!inode_owner_or_capable(inode))
1680 		return -EACCES;
1681 
1682 	if (!S_ISREG(inode->i_mode))
1683 		return -EINVAL;
1684 
1685 	ret = mnt_want_write_file(filp);
1686 	if (ret)
1687 		return ret;
1688 
1689 	inode_lock(inode);
1690 
1691 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1692 
1693 	if (f2fs_is_atomic_file(inode))
1694 		goto out;
1695 
1696 	ret = f2fs_convert_inline_inode(inode);
1697 	if (ret)
1698 		goto out;
1699 
1700 	if (!get_dirty_pages(inode))
1701 		goto skip_flush;
1702 
1703 	f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1704 		"Unexpected flush for atomic writes: ino=%lu, npages=%u",
1705 					inode->i_ino, get_dirty_pages(inode));
1706 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1707 	if (ret)
1708 		goto out;
1709 skip_flush:
1710 	set_inode_flag(inode, FI_ATOMIC_FILE);
1711 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1712 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1713 
1714 	F2FS_I(inode)->inmem_task = current;
1715 	stat_inc_atomic_write(inode);
1716 	stat_update_max_atomic_write(inode);
1717 out:
1718 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1719 	inode_unlock(inode);
1720 	mnt_drop_write_file(filp);
1721 	return ret;
1722 }
1723 
1724 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1725 {
1726 	struct inode *inode = file_inode(filp);
1727 	int ret;
1728 
1729 	if (!inode_owner_or_capable(inode))
1730 		return -EACCES;
1731 
1732 	ret = mnt_want_write_file(filp);
1733 	if (ret)
1734 		return ret;
1735 
1736 	inode_lock(inode);
1737 
1738 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1739 
1740 	if (f2fs_is_volatile_file(inode)) {
1741 		ret = -EINVAL;
1742 		goto err_out;
1743 	}
1744 
1745 	if (f2fs_is_atomic_file(inode)) {
1746 		ret = f2fs_commit_inmem_pages(inode);
1747 		if (ret)
1748 			goto err_out;
1749 
1750 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1751 		if (!ret) {
1752 			clear_inode_flag(inode, FI_ATOMIC_FILE);
1753 			F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1754 			stat_dec_atomic_write(inode);
1755 		}
1756 	} else {
1757 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1758 	}
1759 err_out:
1760 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1761 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1762 		ret = -EINVAL;
1763 	}
1764 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1765 	inode_unlock(inode);
1766 	mnt_drop_write_file(filp);
1767 	return ret;
1768 }
1769 
1770 static int f2fs_ioc_start_volatile_write(struct file *filp)
1771 {
1772 	struct inode *inode = file_inode(filp);
1773 	int ret;
1774 
1775 	if (!inode_owner_or_capable(inode))
1776 		return -EACCES;
1777 
1778 	if (!S_ISREG(inode->i_mode))
1779 		return -EINVAL;
1780 
1781 	ret = mnt_want_write_file(filp);
1782 	if (ret)
1783 		return ret;
1784 
1785 	inode_lock(inode);
1786 
1787 	if (f2fs_is_volatile_file(inode))
1788 		goto out;
1789 
1790 	ret = f2fs_convert_inline_inode(inode);
1791 	if (ret)
1792 		goto out;
1793 
1794 	stat_inc_volatile_write(inode);
1795 	stat_update_max_volatile_write(inode);
1796 
1797 	set_inode_flag(inode, FI_VOLATILE_FILE);
1798 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1799 out:
1800 	inode_unlock(inode);
1801 	mnt_drop_write_file(filp);
1802 	return ret;
1803 }
1804 
1805 static int f2fs_ioc_release_volatile_write(struct file *filp)
1806 {
1807 	struct inode *inode = file_inode(filp);
1808 	int ret;
1809 
1810 	if (!inode_owner_or_capable(inode))
1811 		return -EACCES;
1812 
1813 	ret = mnt_want_write_file(filp);
1814 	if (ret)
1815 		return ret;
1816 
1817 	inode_lock(inode);
1818 
1819 	if (!f2fs_is_volatile_file(inode))
1820 		goto out;
1821 
1822 	if (!f2fs_is_first_block_written(inode)) {
1823 		ret = truncate_partial_data_page(inode, 0, true);
1824 		goto out;
1825 	}
1826 
1827 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1828 out:
1829 	inode_unlock(inode);
1830 	mnt_drop_write_file(filp);
1831 	return ret;
1832 }
1833 
1834 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1835 {
1836 	struct inode *inode = file_inode(filp);
1837 	int ret;
1838 
1839 	if (!inode_owner_or_capable(inode))
1840 		return -EACCES;
1841 
1842 	ret = mnt_want_write_file(filp);
1843 	if (ret)
1844 		return ret;
1845 
1846 	inode_lock(inode);
1847 
1848 	if (f2fs_is_atomic_file(inode))
1849 		f2fs_drop_inmem_pages(inode);
1850 	if (f2fs_is_volatile_file(inode)) {
1851 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1852 		stat_dec_volatile_write(inode);
1853 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1854 	}
1855 
1856 	inode_unlock(inode);
1857 
1858 	mnt_drop_write_file(filp);
1859 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1860 	return ret;
1861 }
1862 
1863 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1864 {
1865 	struct inode *inode = file_inode(filp);
1866 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1867 	struct super_block *sb = sbi->sb;
1868 	__u32 in;
1869 	int ret;
1870 
1871 	if (!capable(CAP_SYS_ADMIN))
1872 		return -EPERM;
1873 
1874 	if (get_user(in, (__u32 __user *)arg))
1875 		return -EFAULT;
1876 
1877 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
1878 		ret = mnt_want_write_file(filp);
1879 		if (ret)
1880 			return ret;
1881 	}
1882 
1883 	switch (in) {
1884 	case F2FS_GOING_DOWN_FULLSYNC:
1885 		sb = freeze_bdev(sb->s_bdev);
1886 		if (IS_ERR(sb)) {
1887 			ret = PTR_ERR(sb);
1888 			goto out;
1889 		}
1890 		if (sb) {
1891 			f2fs_stop_checkpoint(sbi, false);
1892 			thaw_bdev(sb->s_bdev, sb);
1893 		}
1894 		break;
1895 	case F2FS_GOING_DOWN_METASYNC:
1896 		/* do checkpoint only */
1897 		ret = f2fs_sync_fs(sb, 1);
1898 		if (ret)
1899 			goto out;
1900 		f2fs_stop_checkpoint(sbi, false);
1901 		break;
1902 	case F2FS_GOING_DOWN_NOSYNC:
1903 		f2fs_stop_checkpoint(sbi, false);
1904 		break;
1905 	case F2FS_GOING_DOWN_METAFLUSH:
1906 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1907 		f2fs_stop_checkpoint(sbi, false);
1908 		break;
1909 	default:
1910 		ret = -EINVAL;
1911 		goto out;
1912 	}
1913 
1914 	f2fs_stop_gc_thread(sbi);
1915 	f2fs_stop_discard_thread(sbi);
1916 
1917 	f2fs_drop_discard_cmd(sbi);
1918 	clear_opt(sbi, DISCARD);
1919 
1920 	f2fs_update_time(sbi, REQ_TIME);
1921 out:
1922 	if (in != F2FS_GOING_DOWN_FULLSYNC)
1923 		mnt_drop_write_file(filp);
1924 	return ret;
1925 }
1926 
1927 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1928 {
1929 	struct inode *inode = file_inode(filp);
1930 	struct super_block *sb = inode->i_sb;
1931 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
1932 	struct fstrim_range range;
1933 	int ret;
1934 
1935 	if (!capable(CAP_SYS_ADMIN))
1936 		return -EPERM;
1937 
1938 	if (!blk_queue_discard(q))
1939 		return -EOPNOTSUPP;
1940 
1941 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1942 				sizeof(range)))
1943 		return -EFAULT;
1944 
1945 	ret = mnt_want_write_file(filp);
1946 	if (ret)
1947 		return ret;
1948 
1949 	range.minlen = max((unsigned int)range.minlen,
1950 				q->limits.discard_granularity);
1951 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1952 	mnt_drop_write_file(filp);
1953 	if (ret < 0)
1954 		return ret;
1955 
1956 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
1957 				sizeof(range)))
1958 		return -EFAULT;
1959 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1960 	return 0;
1961 }
1962 
1963 static bool uuid_is_nonzero(__u8 u[16])
1964 {
1965 	int i;
1966 
1967 	for (i = 0; i < 16; i++)
1968 		if (u[i])
1969 			return true;
1970 	return false;
1971 }
1972 
1973 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1974 {
1975 	struct inode *inode = file_inode(filp);
1976 
1977 	if (!f2fs_sb_has_encrypt(inode->i_sb))
1978 		return -EOPNOTSUPP;
1979 
1980 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1981 
1982 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1983 }
1984 
1985 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1986 {
1987 	if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
1988 		return -EOPNOTSUPP;
1989 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1990 }
1991 
1992 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1993 {
1994 	struct inode *inode = file_inode(filp);
1995 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1996 	int err;
1997 
1998 	if (!f2fs_sb_has_encrypt(inode->i_sb))
1999 		return -EOPNOTSUPP;
2000 
2001 	err = mnt_want_write_file(filp);
2002 	if (err)
2003 		return err;
2004 
2005 	down_write(&sbi->sb_lock);
2006 
2007 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2008 		goto got_it;
2009 
2010 	/* update superblock with uuid */
2011 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2012 
2013 	err = f2fs_commit_super(sbi, false);
2014 	if (err) {
2015 		/* undo new data */
2016 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2017 		goto out_err;
2018 	}
2019 got_it:
2020 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2021 									16))
2022 		err = -EFAULT;
2023 out_err:
2024 	up_write(&sbi->sb_lock);
2025 	mnt_drop_write_file(filp);
2026 	return err;
2027 }
2028 
2029 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2030 {
2031 	struct inode *inode = file_inode(filp);
2032 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2033 	__u32 sync;
2034 	int ret;
2035 
2036 	if (!capable(CAP_SYS_ADMIN))
2037 		return -EPERM;
2038 
2039 	if (get_user(sync, (__u32 __user *)arg))
2040 		return -EFAULT;
2041 
2042 	if (f2fs_readonly(sbi->sb))
2043 		return -EROFS;
2044 
2045 	ret = mnt_want_write_file(filp);
2046 	if (ret)
2047 		return ret;
2048 
2049 	if (!sync) {
2050 		if (!mutex_trylock(&sbi->gc_mutex)) {
2051 			ret = -EBUSY;
2052 			goto out;
2053 		}
2054 	} else {
2055 		mutex_lock(&sbi->gc_mutex);
2056 	}
2057 
2058 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2059 out:
2060 	mnt_drop_write_file(filp);
2061 	return ret;
2062 }
2063 
2064 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2065 {
2066 	struct inode *inode = file_inode(filp);
2067 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2068 	struct f2fs_gc_range range;
2069 	u64 end;
2070 	int ret;
2071 
2072 	if (!capable(CAP_SYS_ADMIN))
2073 		return -EPERM;
2074 
2075 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2076 							sizeof(range)))
2077 		return -EFAULT;
2078 
2079 	if (f2fs_readonly(sbi->sb))
2080 		return -EROFS;
2081 
2082 	end = range.start + range.len;
2083 	if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2084 		return -EINVAL;
2085 	}
2086 
2087 	ret = mnt_want_write_file(filp);
2088 	if (ret)
2089 		return ret;
2090 
2091 do_more:
2092 	if (!range.sync) {
2093 		if (!mutex_trylock(&sbi->gc_mutex)) {
2094 			ret = -EBUSY;
2095 			goto out;
2096 		}
2097 	} else {
2098 		mutex_lock(&sbi->gc_mutex);
2099 	}
2100 
2101 	ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2102 	range.start += sbi->blocks_per_seg;
2103 	if (range.start <= end)
2104 		goto do_more;
2105 out:
2106 	mnt_drop_write_file(filp);
2107 	return ret;
2108 }
2109 
2110 static int f2fs_ioc_f2fs_write_checkpoint(struct file *filp, unsigned long arg)
2111 {
2112 	struct inode *inode = file_inode(filp);
2113 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2114 	int ret;
2115 
2116 	if (!capable(CAP_SYS_ADMIN))
2117 		return -EPERM;
2118 
2119 	if (f2fs_readonly(sbi->sb))
2120 		return -EROFS;
2121 
2122 	ret = mnt_want_write_file(filp);
2123 	if (ret)
2124 		return ret;
2125 
2126 	ret = f2fs_sync_fs(sbi->sb, 1);
2127 
2128 	mnt_drop_write_file(filp);
2129 	return ret;
2130 }
2131 
2132 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2133 					struct file *filp,
2134 					struct f2fs_defragment *range)
2135 {
2136 	struct inode *inode = file_inode(filp);
2137 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2138 					.m_seg_type = NO_CHECK_TYPE };
2139 	struct extent_info ei = {0, 0, 0};
2140 	pgoff_t pg_start, pg_end, next_pgofs;
2141 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2142 	unsigned int total = 0, sec_num;
2143 	block_t blk_end = 0;
2144 	bool fragmented = false;
2145 	int err;
2146 
2147 	/* if in-place-update policy is enabled, don't waste time here */
2148 	if (f2fs_should_update_inplace(inode, NULL))
2149 		return -EINVAL;
2150 
2151 	pg_start = range->start >> PAGE_SHIFT;
2152 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2153 
2154 	f2fs_balance_fs(sbi, true);
2155 
2156 	inode_lock(inode);
2157 
2158 	/* writeback all dirty pages in the range */
2159 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2160 						range->start + range->len - 1);
2161 	if (err)
2162 		goto out;
2163 
2164 	/*
2165 	 * lookup mapping info in extent cache, skip defragmenting if physical
2166 	 * block addresses are continuous.
2167 	 */
2168 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2169 		if (ei.fofs + ei.len >= pg_end)
2170 			goto out;
2171 	}
2172 
2173 	map.m_lblk = pg_start;
2174 	map.m_next_pgofs = &next_pgofs;
2175 
2176 	/*
2177 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2178 	 * physical block addresses are continuous even if there are hole(s)
2179 	 * in logical blocks.
2180 	 */
2181 	while (map.m_lblk < pg_end) {
2182 		map.m_len = pg_end - map.m_lblk;
2183 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2184 		if (err)
2185 			goto out;
2186 
2187 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2188 			map.m_lblk = next_pgofs;
2189 			continue;
2190 		}
2191 
2192 		if (blk_end && blk_end != map.m_pblk)
2193 			fragmented = true;
2194 
2195 		/* record total count of block that we're going to move */
2196 		total += map.m_len;
2197 
2198 		blk_end = map.m_pblk + map.m_len;
2199 
2200 		map.m_lblk += map.m_len;
2201 	}
2202 
2203 	if (!fragmented)
2204 		goto out;
2205 
2206 	sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2207 
2208 	/*
2209 	 * make sure there are enough free section for LFS allocation, this can
2210 	 * avoid defragment running in SSR mode when free section are allocated
2211 	 * intensively
2212 	 */
2213 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2214 		err = -EAGAIN;
2215 		goto out;
2216 	}
2217 
2218 	map.m_lblk = pg_start;
2219 	map.m_len = pg_end - pg_start;
2220 	total = 0;
2221 
2222 	while (map.m_lblk < pg_end) {
2223 		pgoff_t idx;
2224 		int cnt = 0;
2225 
2226 do_map:
2227 		map.m_len = pg_end - map.m_lblk;
2228 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2229 		if (err)
2230 			goto clear_out;
2231 
2232 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2233 			map.m_lblk = next_pgofs;
2234 			continue;
2235 		}
2236 
2237 		set_inode_flag(inode, FI_DO_DEFRAG);
2238 
2239 		idx = map.m_lblk;
2240 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2241 			struct page *page;
2242 
2243 			page = f2fs_get_lock_data_page(inode, idx, true);
2244 			if (IS_ERR(page)) {
2245 				err = PTR_ERR(page);
2246 				goto clear_out;
2247 			}
2248 
2249 			set_page_dirty(page);
2250 			f2fs_put_page(page, 1);
2251 
2252 			idx++;
2253 			cnt++;
2254 			total++;
2255 		}
2256 
2257 		map.m_lblk = idx;
2258 
2259 		if (idx < pg_end && cnt < blk_per_seg)
2260 			goto do_map;
2261 
2262 		clear_inode_flag(inode, FI_DO_DEFRAG);
2263 
2264 		err = filemap_fdatawrite(inode->i_mapping);
2265 		if (err)
2266 			goto out;
2267 	}
2268 clear_out:
2269 	clear_inode_flag(inode, FI_DO_DEFRAG);
2270 out:
2271 	inode_unlock(inode);
2272 	if (!err)
2273 		range->len = (u64)total << PAGE_SHIFT;
2274 	return err;
2275 }
2276 
2277 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2278 {
2279 	struct inode *inode = file_inode(filp);
2280 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2281 	struct f2fs_defragment range;
2282 	int err;
2283 
2284 	if (!capable(CAP_SYS_ADMIN))
2285 		return -EPERM;
2286 
2287 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2288 		return -EINVAL;
2289 
2290 	if (f2fs_readonly(sbi->sb))
2291 		return -EROFS;
2292 
2293 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2294 							sizeof(range)))
2295 		return -EFAULT;
2296 
2297 	/* verify alignment of offset & size */
2298 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2299 		return -EINVAL;
2300 
2301 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2302 					sbi->max_file_blocks))
2303 		return -EINVAL;
2304 
2305 	err = mnt_want_write_file(filp);
2306 	if (err)
2307 		return err;
2308 
2309 	err = f2fs_defragment_range(sbi, filp, &range);
2310 	mnt_drop_write_file(filp);
2311 
2312 	f2fs_update_time(sbi, REQ_TIME);
2313 	if (err < 0)
2314 		return err;
2315 
2316 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2317 							sizeof(range)))
2318 		return -EFAULT;
2319 
2320 	return 0;
2321 }
2322 
2323 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2324 			struct file *file_out, loff_t pos_out, size_t len)
2325 {
2326 	struct inode *src = file_inode(file_in);
2327 	struct inode *dst = file_inode(file_out);
2328 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2329 	size_t olen = len, dst_max_i_size = 0;
2330 	size_t dst_osize;
2331 	int ret;
2332 
2333 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2334 				src->i_sb != dst->i_sb)
2335 		return -EXDEV;
2336 
2337 	if (unlikely(f2fs_readonly(src->i_sb)))
2338 		return -EROFS;
2339 
2340 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2341 		return -EINVAL;
2342 
2343 	if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2344 		return -EOPNOTSUPP;
2345 
2346 	if (src == dst) {
2347 		if (pos_in == pos_out)
2348 			return 0;
2349 		if (pos_out > pos_in && pos_out < pos_in + len)
2350 			return -EINVAL;
2351 	}
2352 
2353 	inode_lock(src);
2354 	down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2355 	if (src != dst) {
2356 		ret = -EBUSY;
2357 		if (!inode_trylock(dst))
2358 			goto out;
2359 		if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) {
2360 			inode_unlock(dst);
2361 			goto out;
2362 		}
2363 	}
2364 
2365 	ret = -EINVAL;
2366 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2367 		goto out_unlock;
2368 	if (len == 0)
2369 		olen = len = src->i_size - pos_in;
2370 	if (pos_in + len == src->i_size)
2371 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2372 	if (len == 0) {
2373 		ret = 0;
2374 		goto out_unlock;
2375 	}
2376 
2377 	dst_osize = dst->i_size;
2378 	if (pos_out + olen > dst->i_size)
2379 		dst_max_i_size = pos_out + olen;
2380 
2381 	/* verify the end result is block aligned */
2382 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2383 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2384 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2385 		goto out_unlock;
2386 
2387 	ret = f2fs_convert_inline_inode(src);
2388 	if (ret)
2389 		goto out_unlock;
2390 
2391 	ret = f2fs_convert_inline_inode(dst);
2392 	if (ret)
2393 		goto out_unlock;
2394 
2395 	/* write out all dirty pages from offset */
2396 	ret = filemap_write_and_wait_range(src->i_mapping,
2397 					pos_in, pos_in + len);
2398 	if (ret)
2399 		goto out_unlock;
2400 
2401 	ret = filemap_write_and_wait_range(dst->i_mapping,
2402 					pos_out, pos_out + len);
2403 	if (ret)
2404 		goto out_unlock;
2405 
2406 	f2fs_balance_fs(sbi, true);
2407 	f2fs_lock_op(sbi);
2408 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2409 				pos_out >> F2FS_BLKSIZE_BITS,
2410 				len >> F2FS_BLKSIZE_BITS, false);
2411 
2412 	if (!ret) {
2413 		if (dst_max_i_size)
2414 			f2fs_i_size_write(dst, dst_max_i_size);
2415 		else if (dst_osize != dst->i_size)
2416 			f2fs_i_size_write(dst, dst_osize);
2417 	}
2418 	f2fs_unlock_op(sbi);
2419 out_unlock:
2420 	if (src != dst) {
2421 		up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2422 		inode_unlock(dst);
2423 	}
2424 out:
2425 	up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2426 	inode_unlock(src);
2427 	return ret;
2428 }
2429 
2430 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2431 {
2432 	struct f2fs_move_range range;
2433 	struct fd dst;
2434 	int err;
2435 
2436 	if (!(filp->f_mode & FMODE_READ) ||
2437 			!(filp->f_mode & FMODE_WRITE))
2438 		return -EBADF;
2439 
2440 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2441 							sizeof(range)))
2442 		return -EFAULT;
2443 
2444 	dst = fdget(range.dst_fd);
2445 	if (!dst.file)
2446 		return -EBADF;
2447 
2448 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2449 		err = -EBADF;
2450 		goto err_out;
2451 	}
2452 
2453 	err = mnt_want_write_file(filp);
2454 	if (err)
2455 		goto err_out;
2456 
2457 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2458 					range.pos_out, range.len);
2459 
2460 	mnt_drop_write_file(filp);
2461 	if (err)
2462 		goto err_out;
2463 
2464 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2465 						&range, sizeof(range)))
2466 		err = -EFAULT;
2467 err_out:
2468 	fdput(dst);
2469 	return err;
2470 }
2471 
2472 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2473 {
2474 	struct inode *inode = file_inode(filp);
2475 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2476 	struct sit_info *sm = SIT_I(sbi);
2477 	unsigned int start_segno = 0, end_segno = 0;
2478 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2479 	struct f2fs_flush_device range;
2480 	int ret;
2481 
2482 	if (!capable(CAP_SYS_ADMIN))
2483 		return -EPERM;
2484 
2485 	if (f2fs_readonly(sbi->sb))
2486 		return -EROFS;
2487 
2488 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2489 							sizeof(range)))
2490 		return -EFAULT;
2491 
2492 	if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2493 			sbi->segs_per_sec != 1) {
2494 		f2fs_msg(sbi->sb, KERN_WARNING,
2495 			"Can't flush %u in %d for segs_per_sec %u != 1\n",
2496 				range.dev_num, sbi->s_ndevs,
2497 				sbi->segs_per_sec);
2498 		return -EINVAL;
2499 	}
2500 
2501 	ret = mnt_want_write_file(filp);
2502 	if (ret)
2503 		return ret;
2504 
2505 	if (range.dev_num != 0)
2506 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2507 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2508 
2509 	start_segno = sm->last_victim[FLUSH_DEVICE];
2510 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2511 		start_segno = dev_start_segno;
2512 	end_segno = min(start_segno + range.segments, dev_end_segno);
2513 
2514 	while (start_segno < end_segno) {
2515 		if (!mutex_trylock(&sbi->gc_mutex)) {
2516 			ret = -EBUSY;
2517 			goto out;
2518 		}
2519 		sm->last_victim[GC_CB] = end_segno + 1;
2520 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2521 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2522 		ret = f2fs_gc(sbi, true, true, start_segno);
2523 		if (ret == -EAGAIN)
2524 			ret = 0;
2525 		else if (ret < 0)
2526 			break;
2527 		start_segno++;
2528 	}
2529 out:
2530 	mnt_drop_write_file(filp);
2531 	return ret;
2532 }
2533 
2534 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2535 {
2536 	struct inode *inode = file_inode(filp);
2537 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2538 
2539 	/* Must validate to set it with SQLite behavior in Android. */
2540 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2541 
2542 	return put_user(sb_feature, (u32 __user *)arg);
2543 }
2544 
2545 #ifdef CONFIG_QUOTA
2546 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2547 {
2548 	struct inode *inode = file_inode(filp);
2549 	struct f2fs_inode_info *fi = F2FS_I(inode);
2550 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2551 	struct super_block *sb = sbi->sb;
2552 	struct dquot *transfer_to[MAXQUOTAS] = {};
2553 	struct page *ipage;
2554 	kprojid_t kprojid;
2555 	int err;
2556 
2557 	if (!f2fs_sb_has_project_quota(sb)) {
2558 		if (projid != F2FS_DEF_PROJID)
2559 			return -EOPNOTSUPP;
2560 		else
2561 			return 0;
2562 	}
2563 
2564 	if (!f2fs_has_extra_attr(inode))
2565 		return -EOPNOTSUPP;
2566 
2567 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2568 
2569 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2570 		return 0;
2571 
2572 	err = mnt_want_write_file(filp);
2573 	if (err)
2574 		return err;
2575 
2576 	err = -EPERM;
2577 	inode_lock(inode);
2578 
2579 	/* Is it quota file? Do not allow user to mess with it */
2580 	if (IS_NOQUOTA(inode))
2581 		goto out_unlock;
2582 
2583 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
2584 	if (IS_ERR(ipage)) {
2585 		err = PTR_ERR(ipage);
2586 		goto out_unlock;
2587 	}
2588 
2589 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2590 								i_projid)) {
2591 		err = -EOVERFLOW;
2592 		f2fs_put_page(ipage, 1);
2593 		goto out_unlock;
2594 	}
2595 	f2fs_put_page(ipage, 1);
2596 
2597 	err = dquot_initialize(inode);
2598 	if (err)
2599 		goto out_unlock;
2600 
2601 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2602 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2603 		err = __dquot_transfer(inode, transfer_to);
2604 		dqput(transfer_to[PRJQUOTA]);
2605 		if (err)
2606 			goto out_dirty;
2607 	}
2608 
2609 	F2FS_I(inode)->i_projid = kprojid;
2610 	inode->i_ctime = current_time(inode);
2611 out_dirty:
2612 	f2fs_mark_inode_dirty_sync(inode, true);
2613 out_unlock:
2614 	inode_unlock(inode);
2615 	mnt_drop_write_file(filp);
2616 	return err;
2617 }
2618 #else
2619 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2620 {
2621 	if (projid != F2FS_DEF_PROJID)
2622 		return -EOPNOTSUPP;
2623 	return 0;
2624 }
2625 #endif
2626 
2627 /* Transfer internal flags to xflags */
2628 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2629 {
2630 	__u32 xflags = 0;
2631 
2632 	if (iflags & F2FS_SYNC_FL)
2633 		xflags |= FS_XFLAG_SYNC;
2634 	if (iflags & F2FS_IMMUTABLE_FL)
2635 		xflags |= FS_XFLAG_IMMUTABLE;
2636 	if (iflags & F2FS_APPEND_FL)
2637 		xflags |= FS_XFLAG_APPEND;
2638 	if (iflags & F2FS_NODUMP_FL)
2639 		xflags |= FS_XFLAG_NODUMP;
2640 	if (iflags & F2FS_NOATIME_FL)
2641 		xflags |= FS_XFLAG_NOATIME;
2642 	if (iflags & F2FS_PROJINHERIT_FL)
2643 		xflags |= FS_XFLAG_PROJINHERIT;
2644 	return xflags;
2645 }
2646 
2647 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2648 				  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2649 				  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2650 
2651 /* Transfer xflags flags to internal */
2652 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2653 {
2654 	unsigned long iflags = 0;
2655 
2656 	if (xflags & FS_XFLAG_SYNC)
2657 		iflags |= F2FS_SYNC_FL;
2658 	if (xflags & FS_XFLAG_IMMUTABLE)
2659 		iflags |= F2FS_IMMUTABLE_FL;
2660 	if (xflags & FS_XFLAG_APPEND)
2661 		iflags |= F2FS_APPEND_FL;
2662 	if (xflags & FS_XFLAG_NODUMP)
2663 		iflags |= F2FS_NODUMP_FL;
2664 	if (xflags & FS_XFLAG_NOATIME)
2665 		iflags |= F2FS_NOATIME_FL;
2666 	if (xflags & FS_XFLAG_PROJINHERIT)
2667 		iflags |= F2FS_PROJINHERIT_FL;
2668 
2669 	return iflags;
2670 }
2671 
2672 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2673 {
2674 	struct inode *inode = file_inode(filp);
2675 	struct f2fs_inode_info *fi = F2FS_I(inode);
2676 	struct fsxattr fa;
2677 
2678 	memset(&fa, 0, sizeof(struct fsxattr));
2679 	fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2680 				F2FS_FL_USER_VISIBLE);
2681 
2682 	if (f2fs_sb_has_project_quota(inode->i_sb))
2683 		fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2684 							fi->i_projid);
2685 
2686 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2687 		return -EFAULT;
2688 	return 0;
2689 }
2690 
2691 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2692 {
2693 	struct inode *inode = file_inode(filp);
2694 	struct f2fs_inode_info *fi = F2FS_I(inode);
2695 	struct fsxattr fa;
2696 	unsigned int flags;
2697 	int err;
2698 
2699 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2700 		return -EFAULT;
2701 
2702 	/* Make sure caller has proper permission */
2703 	if (!inode_owner_or_capable(inode))
2704 		return -EACCES;
2705 
2706 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2707 		return -EOPNOTSUPP;
2708 
2709 	flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2710 	if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2711 		return -EOPNOTSUPP;
2712 
2713 	err = mnt_want_write_file(filp);
2714 	if (err)
2715 		return err;
2716 
2717 	inode_lock(inode);
2718 	flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2719 				(flags & F2FS_FL_XFLAG_VISIBLE);
2720 	err = __f2fs_ioc_setflags(inode, flags);
2721 	inode_unlock(inode);
2722 	mnt_drop_write_file(filp);
2723 	if (err)
2724 		return err;
2725 
2726 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2727 	if (err)
2728 		return err;
2729 
2730 	return 0;
2731 }
2732 
2733 int f2fs_pin_file_control(struct inode *inode, bool inc)
2734 {
2735 	struct f2fs_inode_info *fi = F2FS_I(inode);
2736 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2737 
2738 	/* Use i_gc_failures for normal file as a risk signal. */
2739 	if (inc)
2740 		f2fs_i_gc_failures_write(inode,
2741 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2742 
2743 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2744 		f2fs_msg(sbi->sb, KERN_WARNING,
2745 			"%s: Enable GC = ino %lx after %x GC trials\n",
2746 			__func__, inode->i_ino,
2747 			fi->i_gc_failures[GC_FAILURE_PIN]);
2748 		clear_inode_flag(inode, FI_PIN_FILE);
2749 		return -EAGAIN;
2750 	}
2751 	return 0;
2752 }
2753 
2754 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2755 {
2756 	struct inode *inode = file_inode(filp);
2757 	__u32 pin;
2758 	int ret = 0;
2759 
2760 	if (!inode_owner_or_capable(inode))
2761 		return -EACCES;
2762 
2763 	if (get_user(pin, (__u32 __user *)arg))
2764 		return -EFAULT;
2765 
2766 	if (!S_ISREG(inode->i_mode))
2767 		return -EINVAL;
2768 
2769 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2770 		return -EROFS;
2771 
2772 	ret = mnt_want_write_file(filp);
2773 	if (ret)
2774 		return ret;
2775 
2776 	inode_lock(inode);
2777 
2778 	if (f2fs_should_update_outplace(inode, NULL)) {
2779 		ret = -EINVAL;
2780 		goto out;
2781 	}
2782 
2783 	if (!pin) {
2784 		clear_inode_flag(inode, FI_PIN_FILE);
2785 		F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN] = 1;
2786 		goto done;
2787 	}
2788 
2789 	if (f2fs_pin_file_control(inode, false)) {
2790 		ret = -EAGAIN;
2791 		goto out;
2792 	}
2793 	ret = f2fs_convert_inline_inode(inode);
2794 	if (ret)
2795 		goto out;
2796 
2797 	set_inode_flag(inode, FI_PIN_FILE);
2798 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2799 done:
2800 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2801 out:
2802 	inode_unlock(inode);
2803 	mnt_drop_write_file(filp);
2804 	return ret;
2805 }
2806 
2807 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2808 {
2809 	struct inode *inode = file_inode(filp);
2810 	__u32 pin = 0;
2811 
2812 	if (is_inode_flag_set(inode, FI_PIN_FILE))
2813 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2814 	return put_user(pin, (u32 __user *)arg);
2815 }
2816 
2817 int f2fs_precache_extents(struct inode *inode)
2818 {
2819 	struct f2fs_inode_info *fi = F2FS_I(inode);
2820 	struct f2fs_map_blocks map;
2821 	pgoff_t m_next_extent;
2822 	loff_t end;
2823 	int err;
2824 
2825 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
2826 		return -EOPNOTSUPP;
2827 
2828 	map.m_lblk = 0;
2829 	map.m_next_pgofs = NULL;
2830 	map.m_next_extent = &m_next_extent;
2831 	map.m_seg_type = NO_CHECK_TYPE;
2832 	end = F2FS_I_SB(inode)->max_file_blocks;
2833 
2834 	while (map.m_lblk < end) {
2835 		map.m_len = end - map.m_lblk;
2836 
2837 		down_write(&fi->i_gc_rwsem[WRITE]);
2838 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2839 		up_write(&fi->i_gc_rwsem[WRITE]);
2840 		if (err)
2841 			return err;
2842 
2843 		map.m_lblk = m_next_extent;
2844 	}
2845 
2846 	return err;
2847 }
2848 
2849 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2850 {
2851 	return f2fs_precache_extents(file_inode(filp));
2852 }
2853 
2854 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2855 {
2856 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2857 		return -EIO;
2858 
2859 	switch (cmd) {
2860 	case F2FS_IOC_GETFLAGS:
2861 		return f2fs_ioc_getflags(filp, arg);
2862 	case F2FS_IOC_SETFLAGS:
2863 		return f2fs_ioc_setflags(filp, arg);
2864 	case F2FS_IOC_GETVERSION:
2865 		return f2fs_ioc_getversion(filp, arg);
2866 	case F2FS_IOC_START_ATOMIC_WRITE:
2867 		return f2fs_ioc_start_atomic_write(filp);
2868 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2869 		return f2fs_ioc_commit_atomic_write(filp);
2870 	case F2FS_IOC_START_VOLATILE_WRITE:
2871 		return f2fs_ioc_start_volatile_write(filp);
2872 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2873 		return f2fs_ioc_release_volatile_write(filp);
2874 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
2875 		return f2fs_ioc_abort_volatile_write(filp);
2876 	case F2FS_IOC_SHUTDOWN:
2877 		return f2fs_ioc_shutdown(filp, arg);
2878 	case FITRIM:
2879 		return f2fs_ioc_fitrim(filp, arg);
2880 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
2881 		return f2fs_ioc_set_encryption_policy(filp, arg);
2882 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
2883 		return f2fs_ioc_get_encryption_policy(filp, arg);
2884 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2885 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2886 	case F2FS_IOC_GARBAGE_COLLECT:
2887 		return f2fs_ioc_gc(filp, arg);
2888 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2889 		return f2fs_ioc_gc_range(filp, arg);
2890 	case F2FS_IOC_WRITE_CHECKPOINT:
2891 		return f2fs_ioc_f2fs_write_checkpoint(filp, arg);
2892 	case F2FS_IOC_DEFRAGMENT:
2893 		return f2fs_ioc_defragment(filp, arg);
2894 	case F2FS_IOC_MOVE_RANGE:
2895 		return f2fs_ioc_move_range(filp, arg);
2896 	case F2FS_IOC_FLUSH_DEVICE:
2897 		return f2fs_ioc_flush_device(filp, arg);
2898 	case F2FS_IOC_GET_FEATURES:
2899 		return f2fs_ioc_get_features(filp, arg);
2900 	case F2FS_IOC_FSGETXATTR:
2901 		return f2fs_ioc_fsgetxattr(filp, arg);
2902 	case F2FS_IOC_FSSETXATTR:
2903 		return f2fs_ioc_fssetxattr(filp, arg);
2904 	case F2FS_IOC_GET_PIN_FILE:
2905 		return f2fs_ioc_get_pin_file(filp, arg);
2906 	case F2FS_IOC_SET_PIN_FILE:
2907 		return f2fs_ioc_set_pin_file(filp, arg);
2908 	case F2FS_IOC_PRECACHE_EXTENTS:
2909 		return f2fs_ioc_precache_extents(filp, arg);
2910 	default:
2911 		return -ENOTTY;
2912 	}
2913 }
2914 
2915 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2916 {
2917 	struct file *file = iocb->ki_filp;
2918 	struct inode *inode = file_inode(file);
2919 	ssize_t ret;
2920 
2921 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2922 		return -EIO;
2923 
2924 	if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
2925 		return -EINVAL;
2926 
2927 	if (!inode_trylock(inode)) {
2928 		if (iocb->ki_flags & IOCB_NOWAIT)
2929 			return -EAGAIN;
2930 		inode_lock(inode);
2931 	}
2932 
2933 	ret = generic_write_checks(iocb, from);
2934 	if (ret > 0) {
2935 		bool preallocated = false;
2936 		size_t target_size = 0;
2937 		int err;
2938 
2939 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2940 			set_inode_flag(inode, FI_NO_PREALLOC);
2941 
2942 		if ((iocb->ki_flags & IOCB_NOWAIT) &&
2943 			(iocb->ki_flags & IOCB_DIRECT)) {
2944 				if (!f2fs_overwrite_io(inode, iocb->ki_pos,
2945 						iov_iter_count(from)) ||
2946 					f2fs_has_inline_data(inode) ||
2947 					f2fs_force_buffered_io(inode, WRITE)) {
2948 						clear_inode_flag(inode,
2949 								FI_NO_PREALLOC);
2950 						inode_unlock(inode);
2951 						return -EAGAIN;
2952 				}
2953 
2954 		} else {
2955 			preallocated = true;
2956 			target_size = iocb->ki_pos + iov_iter_count(from);
2957 
2958 			err = f2fs_preallocate_blocks(iocb, from);
2959 			if (err) {
2960 				clear_inode_flag(inode, FI_NO_PREALLOC);
2961 				inode_unlock(inode);
2962 				return err;
2963 			}
2964 		}
2965 		ret = __generic_file_write_iter(iocb, from);
2966 		clear_inode_flag(inode, FI_NO_PREALLOC);
2967 
2968 		/* if we couldn't write data, we should deallocate blocks. */
2969 		if (preallocated && i_size_read(inode) < target_size)
2970 			f2fs_truncate(inode);
2971 
2972 		if (ret > 0)
2973 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2974 	}
2975 	inode_unlock(inode);
2976 
2977 	if (ret > 0)
2978 		ret = generic_write_sync(iocb, ret);
2979 	return ret;
2980 }
2981 
2982 #ifdef CONFIG_COMPAT
2983 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2984 {
2985 	switch (cmd) {
2986 	case F2FS_IOC32_GETFLAGS:
2987 		cmd = F2FS_IOC_GETFLAGS;
2988 		break;
2989 	case F2FS_IOC32_SETFLAGS:
2990 		cmd = F2FS_IOC_SETFLAGS;
2991 		break;
2992 	case F2FS_IOC32_GETVERSION:
2993 		cmd = F2FS_IOC_GETVERSION;
2994 		break;
2995 	case F2FS_IOC_START_ATOMIC_WRITE:
2996 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2997 	case F2FS_IOC_START_VOLATILE_WRITE:
2998 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2999 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3000 	case F2FS_IOC_SHUTDOWN:
3001 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3002 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3003 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3004 	case F2FS_IOC_GARBAGE_COLLECT:
3005 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3006 	case F2FS_IOC_WRITE_CHECKPOINT:
3007 	case F2FS_IOC_DEFRAGMENT:
3008 	case F2FS_IOC_MOVE_RANGE:
3009 	case F2FS_IOC_FLUSH_DEVICE:
3010 	case F2FS_IOC_GET_FEATURES:
3011 	case F2FS_IOC_FSGETXATTR:
3012 	case F2FS_IOC_FSSETXATTR:
3013 	case F2FS_IOC_GET_PIN_FILE:
3014 	case F2FS_IOC_SET_PIN_FILE:
3015 	case F2FS_IOC_PRECACHE_EXTENTS:
3016 		break;
3017 	default:
3018 		return -ENOIOCTLCMD;
3019 	}
3020 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3021 }
3022 #endif
3023 
3024 const struct file_operations f2fs_file_operations = {
3025 	.llseek		= f2fs_llseek,
3026 	.read_iter	= generic_file_read_iter,
3027 	.write_iter	= f2fs_file_write_iter,
3028 	.open		= f2fs_file_open,
3029 	.release	= f2fs_release_file,
3030 	.mmap		= f2fs_file_mmap,
3031 	.flush		= f2fs_file_flush,
3032 	.fsync		= f2fs_sync_file,
3033 	.fallocate	= f2fs_fallocate,
3034 	.unlocked_ioctl	= f2fs_ioctl,
3035 #ifdef CONFIG_COMPAT
3036 	.compat_ioctl	= f2fs_compat_ioctl,
3037 #endif
3038 	.splice_read	= generic_file_splice_read,
3039 	.splice_write	= iter_file_splice_write,
3040 };
3041