xref: /openbmc/linux/fs/f2fs/file.c (revision 2572f00d)
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/random.h>
24 
25 #include "f2fs.h"
26 #include "node.h"
27 #include "segment.h"
28 #include "xattr.h"
29 #include "acl.h"
30 #include "gc.h"
31 #include "trace.h"
32 #include <trace/events/f2fs.h>
33 
34 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
35 						struct vm_fault *vmf)
36 {
37 	struct page *page = vmf->page;
38 	struct inode *inode = file_inode(vma->vm_file);
39 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
40 	struct dnode_of_data dn;
41 	int err;
42 
43 	f2fs_balance_fs(sbi);
44 
45 	sb_start_pagefault(inode->i_sb);
46 
47 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
48 
49 	/* block allocation */
50 	f2fs_lock_op(sbi);
51 	set_new_dnode(&dn, inode, NULL, NULL, 0);
52 	err = f2fs_reserve_block(&dn, page->index);
53 	if (err) {
54 		f2fs_unlock_op(sbi);
55 		goto out;
56 	}
57 	f2fs_put_dnode(&dn);
58 	f2fs_unlock_op(sbi);
59 
60 	file_update_time(vma->vm_file);
61 	lock_page(page);
62 	if (unlikely(page->mapping != inode->i_mapping ||
63 			page_offset(page) > i_size_read(inode) ||
64 			!PageUptodate(page))) {
65 		unlock_page(page);
66 		err = -EFAULT;
67 		goto out;
68 	}
69 
70 	/*
71 	 * check to see if the page is mapped already (no holes)
72 	 */
73 	if (PageMappedToDisk(page))
74 		goto mapped;
75 
76 	/* page is wholly or partially inside EOF */
77 	if (((loff_t)(page->index + 1) << PAGE_CACHE_SHIFT) >
78 						i_size_read(inode)) {
79 		unsigned offset;
80 		offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
81 		zero_user_segment(page, offset, PAGE_CACHE_SIZE);
82 	}
83 	set_page_dirty(page);
84 	SetPageUptodate(page);
85 
86 	trace_f2fs_vm_page_mkwrite(page, DATA);
87 mapped:
88 	/* fill the page */
89 	f2fs_wait_on_page_writeback(page, DATA);
90 
91 	/* wait for GCed encrypted page writeback */
92 	if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
93 		f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
94 
95 	/* if gced page is attached, don't write to cold segment */
96 	clear_cold_data(page);
97 out:
98 	sb_end_pagefault(inode->i_sb);
99 	return block_page_mkwrite_return(err);
100 }
101 
102 static const struct vm_operations_struct f2fs_file_vm_ops = {
103 	.fault		= filemap_fault,
104 	.map_pages	= filemap_map_pages,
105 	.page_mkwrite	= f2fs_vm_page_mkwrite,
106 };
107 
108 static int get_parent_ino(struct inode *inode, nid_t *pino)
109 {
110 	struct dentry *dentry;
111 
112 	inode = igrab(inode);
113 	dentry = d_find_any_alias(inode);
114 	iput(inode);
115 	if (!dentry)
116 		return 0;
117 
118 	if (update_dent_inode(inode, inode, &dentry->d_name)) {
119 		dput(dentry);
120 		return 0;
121 	}
122 
123 	*pino = parent_ino(dentry);
124 	dput(dentry);
125 	return 1;
126 }
127 
128 static inline bool need_do_checkpoint(struct inode *inode)
129 {
130 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
131 	bool need_cp = false;
132 
133 	if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
134 		need_cp = true;
135 	else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
136 		need_cp = true;
137 	else if (file_wrong_pino(inode))
138 		need_cp = true;
139 	else if (!space_for_roll_forward(sbi))
140 		need_cp = true;
141 	else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
142 		need_cp = true;
143 	else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
144 		need_cp = true;
145 	else if (test_opt(sbi, FASTBOOT))
146 		need_cp = true;
147 	else if (sbi->active_logs == 2)
148 		need_cp = true;
149 
150 	return need_cp;
151 }
152 
153 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
154 {
155 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
156 	bool ret = false;
157 	/* But we need to avoid that there are some inode updates */
158 	if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
159 		ret = true;
160 	f2fs_put_page(i, 0);
161 	return ret;
162 }
163 
164 static void try_to_fix_pino(struct inode *inode)
165 {
166 	struct f2fs_inode_info *fi = F2FS_I(inode);
167 	nid_t pino;
168 
169 	down_write(&fi->i_sem);
170 	fi->xattr_ver = 0;
171 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
172 			get_parent_ino(inode, &pino)) {
173 		fi->i_pino = pino;
174 		file_got_pino(inode);
175 		up_write(&fi->i_sem);
176 
177 		mark_inode_dirty_sync(inode);
178 		f2fs_write_inode(inode, NULL);
179 	} else {
180 		up_write(&fi->i_sem);
181 	}
182 }
183 
184 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
185 {
186 	struct inode *inode = file->f_mapping->host;
187 	struct f2fs_inode_info *fi = F2FS_I(inode);
188 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
189 	nid_t ino = inode->i_ino;
190 	int ret = 0;
191 	bool need_cp = false;
192 	struct writeback_control wbc = {
193 		.sync_mode = WB_SYNC_ALL,
194 		.nr_to_write = LONG_MAX,
195 		.for_reclaim = 0,
196 	};
197 
198 	if (unlikely(f2fs_readonly(inode->i_sb)))
199 		return 0;
200 
201 	trace_f2fs_sync_file_enter(inode);
202 
203 	/* if fdatasync is triggered, let's do in-place-update */
204 	if (get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
205 		set_inode_flag(fi, FI_NEED_IPU);
206 	ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
207 	clear_inode_flag(fi, FI_NEED_IPU);
208 
209 	if (ret) {
210 		trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
211 		return ret;
212 	}
213 
214 	/* if the inode is dirty, let's recover all the time */
215 	if (!datasync) {
216 		f2fs_write_inode(inode, NULL);
217 		goto go_write;
218 	}
219 
220 	/*
221 	 * if there is no written data, don't waste time to write recovery info.
222 	 */
223 	if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
224 			!exist_written_data(sbi, ino, APPEND_INO)) {
225 
226 		/* it may call write_inode just prior to fsync */
227 		if (need_inode_page_update(sbi, ino))
228 			goto go_write;
229 
230 		if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
231 				exist_written_data(sbi, ino, UPDATE_INO))
232 			goto flush_out;
233 		goto out;
234 	}
235 go_write:
236 	/* guarantee free sections for fsync */
237 	f2fs_balance_fs(sbi);
238 
239 	/*
240 	 * Both of fdatasync() and fsync() are able to be recovered from
241 	 * sudden-power-off.
242 	 */
243 	down_read(&fi->i_sem);
244 	need_cp = need_do_checkpoint(inode);
245 	up_read(&fi->i_sem);
246 
247 	if (need_cp) {
248 		/* all the dirty node pages should be flushed for POR */
249 		ret = f2fs_sync_fs(inode->i_sb, 1);
250 
251 		/*
252 		 * We've secured consistency through sync_fs. Following pino
253 		 * will be used only for fsynced inodes after checkpoint.
254 		 */
255 		try_to_fix_pino(inode);
256 		clear_inode_flag(fi, FI_APPEND_WRITE);
257 		clear_inode_flag(fi, FI_UPDATE_WRITE);
258 		goto out;
259 	}
260 sync_nodes:
261 	sync_node_pages(sbi, ino, &wbc);
262 
263 	/* if cp_error was enabled, we should avoid infinite loop */
264 	if (unlikely(f2fs_cp_error(sbi)))
265 		goto out;
266 
267 	if (need_inode_block_update(sbi, ino)) {
268 		mark_inode_dirty_sync(inode);
269 		f2fs_write_inode(inode, NULL);
270 		goto sync_nodes;
271 	}
272 
273 	ret = wait_on_node_pages_writeback(sbi, ino);
274 	if (ret)
275 		goto out;
276 
277 	/* once recovery info is written, don't need to tack this */
278 	remove_dirty_inode(sbi, ino, APPEND_INO);
279 	clear_inode_flag(fi, FI_APPEND_WRITE);
280 flush_out:
281 	remove_dirty_inode(sbi, ino, UPDATE_INO);
282 	clear_inode_flag(fi, FI_UPDATE_WRITE);
283 	ret = f2fs_issue_flush(sbi);
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 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
291 						pgoff_t pgofs, int whence)
292 {
293 	struct pagevec pvec;
294 	int nr_pages;
295 
296 	if (whence != SEEK_DATA)
297 		return 0;
298 
299 	/* find first dirty page index */
300 	pagevec_init(&pvec, 0);
301 	nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
302 					PAGECACHE_TAG_DIRTY, 1);
303 	pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
304 	pagevec_release(&pvec);
305 	return pgofs;
306 }
307 
308 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
309 							int whence)
310 {
311 	switch (whence) {
312 	case SEEK_DATA:
313 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
314 			(blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
315 			return true;
316 		break;
317 	case SEEK_HOLE:
318 		if (blkaddr == NULL_ADDR)
319 			return true;
320 		break;
321 	}
322 	return false;
323 }
324 
325 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
326 {
327 	struct inode *inode = file->f_mapping->host;
328 	loff_t maxbytes = inode->i_sb->s_maxbytes;
329 	struct dnode_of_data dn;
330 	pgoff_t pgofs, end_offset, dirty;
331 	loff_t data_ofs = offset;
332 	loff_t isize;
333 	int err = 0;
334 
335 	mutex_lock(&inode->i_mutex);
336 
337 	isize = i_size_read(inode);
338 	if (offset >= isize)
339 		goto fail;
340 
341 	/* handle inline data case */
342 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
343 		if (whence == SEEK_HOLE)
344 			data_ofs = isize;
345 		goto found;
346 	}
347 
348 	pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
349 
350 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
351 
352 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
353 		set_new_dnode(&dn, inode, NULL, NULL, 0);
354 		err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
355 		if (err && err != -ENOENT) {
356 			goto fail;
357 		} else if (err == -ENOENT) {
358 			/* direct node does not exists */
359 			if (whence == SEEK_DATA) {
360 				pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
361 							F2FS_I(inode));
362 				continue;
363 			} else {
364 				goto found;
365 			}
366 		}
367 
368 		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
369 
370 		/* find data/hole in dnode block */
371 		for (; dn.ofs_in_node < end_offset;
372 				dn.ofs_in_node++, pgofs++,
373 				data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
374 			block_t blkaddr;
375 			blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
376 
377 			if (__found_offset(blkaddr, dirty, pgofs, whence)) {
378 				f2fs_put_dnode(&dn);
379 				goto found;
380 			}
381 		}
382 		f2fs_put_dnode(&dn);
383 	}
384 
385 	if (whence == SEEK_DATA)
386 		goto fail;
387 found:
388 	if (whence == SEEK_HOLE && data_ofs > isize)
389 		data_ofs = isize;
390 	mutex_unlock(&inode->i_mutex);
391 	return vfs_setpos(file, data_ofs, maxbytes);
392 fail:
393 	mutex_unlock(&inode->i_mutex);
394 	return -ENXIO;
395 }
396 
397 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
398 {
399 	struct inode *inode = file->f_mapping->host;
400 	loff_t maxbytes = inode->i_sb->s_maxbytes;
401 
402 	switch (whence) {
403 	case SEEK_SET:
404 	case SEEK_CUR:
405 	case SEEK_END:
406 		return generic_file_llseek_size(file, offset, whence,
407 						maxbytes, i_size_read(inode));
408 	case SEEK_DATA:
409 	case SEEK_HOLE:
410 		if (offset < 0)
411 			return -ENXIO;
412 		return f2fs_seek_block(file, offset, whence);
413 	}
414 
415 	return -EINVAL;
416 }
417 
418 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
419 {
420 	struct inode *inode = file_inode(file);
421 
422 	if (f2fs_encrypted_inode(inode)) {
423 		int err = f2fs_get_encryption_info(inode);
424 		if (err)
425 			return 0;
426 	}
427 
428 	/* we don't need to use inline_data strictly */
429 	if (f2fs_has_inline_data(inode)) {
430 		int err = f2fs_convert_inline_inode(inode);
431 		if (err)
432 			return err;
433 	}
434 
435 	file_accessed(file);
436 	vma->vm_ops = &f2fs_file_vm_ops;
437 	return 0;
438 }
439 
440 static int f2fs_file_open(struct inode *inode, struct file *filp)
441 {
442 	int ret = generic_file_open(inode, filp);
443 
444 	if (!ret && f2fs_encrypted_inode(inode)) {
445 		ret = f2fs_get_encryption_info(inode);
446 		if (ret)
447 			ret = -EACCES;
448 	}
449 	return ret;
450 }
451 
452 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
453 {
454 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
455 	struct f2fs_node *raw_node;
456 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
457 	__le32 *addr;
458 
459 	raw_node = F2FS_NODE(dn->node_page);
460 	addr = blkaddr_in_node(raw_node) + ofs;
461 
462 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
463 		block_t blkaddr = le32_to_cpu(*addr);
464 		if (blkaddr == NULL_ADDR)
465 			continue;
466 
467 		dn->data_blkaddr = NULL_ADDR;
468 		set_data_blkaddr(dn);
469 		invalidate_blocks(sbi, blkaddr);
470 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
471 			clear_inode_flag(F2FS_I(dn->inode),
472 						FI_FIRST_BLOCK_WRITTEN);
473 		nr_free++;
474 	}
475 
476 	if (nr_free) {
477 		pgoff_t fofs;
478 		/*
479 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
480 		 * we will invalidate all blkaddr in the whole range.
481 		 */
482 		fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
483 						F2FS_I(dn->inode)) + ofs;
484 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
485 		dec_valid_block_count(sbi, dn->inode, nr_free);
486 		set_page_dirty(dn->node_page);
487 		sync_inode_page(dn);
488 	}
489 	dn->ofs_in_node = ofs;
490 
491 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
492 					 dn->ofs_in_node, nr_free);
493 	return nr_free;
494 }
495 
496 void truncate_data_blocks(struct dnode_of_data *dn)
497 {
498 	truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
499 }
500 
501 static int truncate_partial_data_page(struct inode *inode, u64 from,
502 								bool cache_only)
503 {
504 	unsigned offset = from & (PAGE_CACHE_SIZE - 1);
505 	pgoff_t index = from >> PAGE_CACHE_SHIFT;
506 	struct address_space *mapping = inode->i_mapping;
507 	struct page *page;
508 
509 	if (!offset && !cache_only)
510 		return 0;
511 
512 	if (cache_only) {
513 		page = f2fs_grab_cache_page(mapping, index, false);
514 		if (page && PageUptodate(page))
515 			goto truncate_out;
516 		f2fs_put_page(page, 1);
517 		return 0;
518 	}
519 
520 	page = get_lock_data_page(inode, index, true);
521 	if (IS_ERR(page))
522 		return 0;
523 truncate_out:
524 	f2fs_wait_on_page_writeback(page, DATA);
525 	zero_user(page, offset, PAGE_CACHE_SIZE - offset);
526 	if (!cache_only || !f2fs_encrypted_inode(inode) || !S_ISREG(inode->i_mode))
527 		set_page_dirty(page);
528 	f2fs_put_page(page, 1);
529 	return 0;
530 }
531 
532 int truncate_blocks(struct inode *inode, u64 from, bool lock)
533 {
534 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
535 	unsigned int blocksize = inode->i_sb->s_blocksize;
536 	struct dnode_of_data dn;
537 	pgoff_t free_from;
538 	int count = 0, err = 0;
539 	struct page *ipage;
540 	bool truncate_page = false;
541 
542 	trace_f2fs_truncate_blocks_enter(inode, from);
543 
544 	free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
545 
546 	if (lock)
547 		f2fs_lock_op(sbi);
548 
549 	ipage = get_node_page(sbi, inode->i_ino);
550 	if (IS_ERR(ipage)) {
551 		err = PTR_ERR(ipage);
552 		goto out;
553 	}
554 
555 	if (f2fs_has_inline_data(inode)) {
556 		if (truncate_inline_inode(ipage, from))
557 			set_page_dirty(ipage);
558 		f2fs_put_page(ipage, 1);
559 		truncate_page = true;
560 		goto out;
561 	}
562 
563 	set_new_dnode(&dn, inode, ipage, NULL, 0);
564 	err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
565 	if (err) {
566 		if (err == -ENOENT)
567 			goto free_next;
568 		goto out;
569 	}
570 
571 	count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
572 
573 	count -= dn.ofs_in_node;
574 	f2fs_bug_on(sbi, count < 0);
575 
576 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
577 		truncate_data_blocks_range(&dn, count);
578 		free_from += count;
579 	}
580 
581 	f2fs_put_dnode(&dn);
582 free_next:
583 	err = truncate_inode_blocks(inode, free_from);
584 out:
585 	if (lock)
586 		f2fs_unlock_op(sbi);
587 
588 	/* lastly zero out the first data page */
589 	if (!err)
590 		err = truncate_partial_data_page(inode, from, truncate_page);
591 
592 	trace_f2fs_truncate_blocks_exit(inode, err);
593 	return err;
594 }
595 
596 int f2fs_truncate(struct inode *inode, bool lock)
597 {
598 	int err;
599 
600 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
601 				S_ISLNK(inode->i_mode)))
602 		return 0;
603 
604 	trace_f2fs_truncate(inode);
605 
606 	/* we should check inline_data size */
607 	if (f2fs_has_inline_data(inode) && !f2fs_may_inline_data(inode)) {
608 		err = f2fs_convert_inline_inode(inode);
609 		if (err)
610 			return err;
611 	}
612 
613 	err = truncate_blocks(inode, i_size_read(inode), lock);
614 	if (err)
615 		return err;
616 
617 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
618 	mark_inode_dirty(inode);
619 	return 0;
620 }
621 
622 int f2fs_getattr(struct vfsmount *mnt,
623 			 struct dentry *dentry, struct kstat *stat)
624 {
625 	struct inode *inode = d_inode(dentry);
626 	generic_fillattr(inode, stat);
627 	stat->blocks <<= 3;
628 	return 0;
629 }
630 
631 #ifdef CONFIG_F2FS_FS_POSIX_ACL
632 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
633 {
634 	struct f2fs_inode_info *fi = F2FS_I(inode);
635 	unsigned int ia_valid = attr->ia_valid;
636 
637 	if (ia_valid & ATTR_UID)
638 		inode->i_uid = attr->ia_uid;
639 	if (ia_valid & ATTR_GID)
640 		inode->i_gid = attr->ia_gid;
641 	if (ia_valid & ATTR_ATIME)
642 		inode->i_atime = timespec_trunc(attr->ia_atime,
643 						inode->i_sb->s_time_gran);
644 	if (ia_valid & ATTR_MTIME)
645 		inode->i_mtime = timespec_trunc(attr->ia_mtime,
646 						inode->i_sb->s_time_gran);
647 	if (ia_valid & ATTR_CTIME)
648 		inode->i_ctime = timespec_trunc(attr->ia_ctime,
649 						inode->i_sb->s_time_gran);
650 	if (ia_valid & ATTR_MODE) {
651 		umode_t mode = attr->ia_mode;
652 
653 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
654 			mode &= ~S_ISGID;
655 		set_acl_inode(fi, mode);
656 	}
657 }
658 #else
659 #define __setattr_copy setattr_copy
660 #endif
661 
662 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
663 {
664 	struct inode *inode = d_inode(dentry);
665 	struct f2fs_inode_info *fi = F2FS_I(inode);
666 	int err;
667 
668 	err = inode_change_ok(inode, attr);
669 	if (err)
670 		return err;
671 
672 	if (attr->ia_valid & ATTR_SIZE) {
673 		if (f2fs_encrypted_inode(inode) &&
674 				f2fs_get_encryption_info(inode))
675 			return -EACCES;
676 
677 		if (attr->ia_size <= i_size_read(inode)) {
678 			truncate_setsize(inode, attr->ia_size);
679 			err = f2fs_truncate(inode, true);
680 			if (err)
681 				return err;
682 			f2fs_balance_fs(F2FS_I_SB(inode));
683 		} else {
684 			/*
685 			 * do not trim all blocks after i_size if target size is
686 			 * larger than i_size.
687 			 */
688 			truncate_setsize(inode, attr->ia_size);
689 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
690 		}
691 	}
692 
693 	__setattr_copy(inode, attr);
694 
695 	if (attr->ia_valid & ATTR_MODE) {
696 		err = posix_acl_chmod(inode, get_inode_mode(inode));
697 		if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
698 			inode->i_mode = fi->i_acl_mode;
699 			clear_inode_flag(fi, FI_ACL_MODE);
700 		}
701 	}
702 
703 	mark_inode_dirty(inode);
704 	return err;
705 }
706 
707 const struct inode_operations f2fs_file_inode_operations = {
708 	.getattr	= f2fs_getattr,
709 	.setattr	= f2fs_setattr,
710 	.get_acl	= f2fs_get_acl,
711 	.set_acl	= f2fs_set_acl,
712 #ifdef CONFIG_F2FS_FS_XATTR
713 	.setxattr	= generic_setxattr,
714 	.getxattr	= generic_getxattr,
715 	.listxattr	= f2fs_listxattr,
716 	.removexattr	= generic_removexattr,
717 #endif
718 	.fiemap		= f2fs_fiemap,
719 };
720 
721 static int fill_zero(struct inode *inode, pgoff_t index,
722 					loff_t start, loff_t len)
723 {
724 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
725 	struct page *page;
726 
727 	if (!len)
728 		return 0;
729 
730 	f2fs_balance_fs(sbi);
731 
732 	f2fs_lock_op(sbi);
733 	page = get_new_data_page(inode, NULL, index, false);
734 	f2fs_unlock_op(sbi);
735 
736 	if (IS_ERR(page))
737 		return PTR_ERR(page);
738 
739 	f2fs_wait_on_page_writeback(page, DATA);
740 	zero_user(page, start, len);
741 	set_page_dirty(page);
742 	f2fs_put_page(page, 1);
743 	return 0;
744 }
745 
746 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
747 {
748 	int err;
749 
750 	while (pg_start < pg_end) {
751 		struct dnode_of_data dn;
752 		pgoff_t end_offset, count;
753 
754 		set_new_dnode(&dn, inode, NULL, NULL, 0);
755 		err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
756 		if (err) {
757 			if (err == -ENOENT) {
758 				pg_start++;
759 				continue;
760 			}
761 			return err;
762 		}
763 
764 		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
765 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
766 
767 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
768 
769 		truncate_data_blocks_range(&dn, count);
770 		f2fs_put_dnode(&dn);
771 
772 		pg_start += count;
773 	}
774 	return 0;
775 }
776 
777 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
778 {
779 	pgoff_t pg_start, pg_end;
780 	loff_t off_start, off_end;
781 	int ret = 0;
782 
783 	if (f2fs_has_inline_data(inode)) {
784 		ret = f2fs_convert_inline_inode(inode);
785 		if (ret)
786 			return ret;
787 	}
788 
789 	pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
790 	pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
791 
792 	off_start = offset & (PAGE_CACHE_SIZE - 1);
793 	off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
794 
795 	if (pg_start == pg_end) {
796 		ret = fill_zero(inode, pg_start, off_start,
797 						off_end - off_start);
798 		if (ret)
799 			return ret;
800 	} else {
801 		if (off_start) {
802 			ret = fill_zero(inode, pg_start++, off_start,
803 						PAGE_CACHE_SIZE - off_start);
804 			if (ret)
805 				return ret;
806 		}
807 		if (off_end) {
808 			ret = fill_zero(inode, pg_end, 0, off_end);
809 			if (ret)
810 				return ret;
811 		}
812 
813 		if (pg_start < pg_end) {
814 			struct address_space *mapping = inode->i_mapping;
815 			loff_t blk_start, blk_end;
816 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
817 
818 			f2fs_balance_fs(sbi);
819 
820 			blk_start = (loff_t)pg_start << PAGE_CACHE_SHIFT;
821 			blk_end = (loff_t)pg_end << PAGE_CACHE_SHIFT;
822 			truncate_inode_pages_range(mapping, blk_start,
823 					blk_end - 1);
824 
825 			f2fs_lock_op(sbi);
826 			ret = truncate_hole(inode, pg_start, pg_end);
827 			f2fs_unlock_op(sbi);
828 		}
829 	}
830 
831 	return ret;
832 }
833 
834 static int __exchange_data_block(struct inode *inode, pgoff_t src,
835 					pgoff_t dst, bool full)
836 {
837 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
838 	struct dnode_of_data dn;
839 	block_t new_addr;
840 	bool do_replace = false;
841 	int ret;
842 
843 	set_new_dnode(&dn, inode, NULL, NULL, 0);
844 	ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
845 	if (ret && ret != -ENOENT) {
846 		return ret;
847 	} else if (ret == -ENOENT) {
848 		new_addr = NULL_ADDR;
849 	} else {
850 		new_addr = dn.data_blkaddr;
851 		if (!is_checkpointed_data(sbi, new_addr)) {
852 			dn.data_blkaddr = NULL_ADDR;
853 			/* do not invalidate this block address */
854 			set_data_blkaddr(&dn);
855 			f2fs_update_extent_cache(&dn);
856 			do_replace = true;
857 		}
858 		f2fs_put_dnode(&dn);
859 	}
860 
861 	if (new_addr == NULL_ADDR)
862 		return full ? truncate_hole(inode, dst, dst + 1) : 0;
863 
864 	if (do_replace) {
865 		struct page *ipage = get_node_page(sbi, inode->i_ino);
866 		struct node_info ni;
867 
868 		if (IS_ERR(ipage)) {
869 			ret = PTR_ERR(ipage);
870 			goto err_out;
871 		}
872 
873 		set_new_dnode(&dn, inode, ipage, NULL, 0);
874 		ret = f2fs_reserve_block(&dn, dst);
875 		if (ret)
876 			goto err_out;
877 
878 		truncate_data_blocks_range(&dn, 1);
879 
880 		get_node_info(sbi, dn.nid, &ni);
881 		f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
882 				ni.version, true);
883 		f2fs_put_dnode(&dn);
884 	} else {
885 		struct page *psrc, *pdst;
886 
887 		psrc = get_lock_data_page(inode, src, true);
888 		if (IS_ERR(psrc))
889 			return PTR_ERR(psrc);
890 		pdst = get_new_data_page(inode, NULL, dst, false);
891 		if (IS_ERR(pdst)) {
892 			f2fs_put_page(psrc, 1);
893 			return PTR_ERR(pdst);
894 		}
895 		f2fs_copy_page(psrc, pdst);
896 		set_page_dirty(pdst);
897 		f2fs_put_page(pdst, 1);
898 		f2fs_put_page(psrc, 1);
899 
900 		return truncate_hole(inode, src, src + 1);
901 	}
902 	return 0;
903 
904 err_out:
905 	if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
906 		dn.data_blkaddr = new_addr;
907 		set_data_blkaddr(&dn);
908 		f2fs_update_extent_cache(&dn);
909 		f2fs_put_dnode(&dn);
910 	}
911 	return ret;
912 }
913 
914 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
915 {
916 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
917 	pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
918 	int ret = 0;
919 
920 	for (; end < nrpages; start++, end++) {
921 		f2fs_balance_fs(sbi);
922 		f2fs_lock_op(sbi);
923 		ret = __exchange_data_block(inode, end, start, true);
924 		f2fs_unlock_op(sbi);
925 		if (ret)
926 			break;
927 	}
928 	return ret;
929 }
930 
931 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
932 {
933 	pgoff_t pg_start, pg_end;
934 	loff_t new_size;
935 	int ret;
936 
937 	if (offset + len >= i_size_read(inode))
938 		return -EINVAL;
939 
940 	/* collapse range should be aligned to block size of f2fs. */
941 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
942 		return -EINVAL;
943 
944 	f2fs_balance_fs(F2FS_I_SB(inode));
945 
946 	if (f2fs_has_inline_data(inode)) {
947 		ret = f2fs_convert_inline_inode(inode);
948 		if (ret)
949 			return ret;
950 	}
951 
952 	pg_start = offset >> PAGE_CACHE_SHIFT;
953 	pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
954 
955 	/* write out all dirty pages from offset */
956 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
957 	if (ret)
958 		return ret;
959 
960 	truncate_pagecache(inode, offset);
961 
962 	ret = f2fs_do_collapse(inode, pg_start, pg_end);
963 	if (ret)
964 		return ret;
965 
966 	/* write out all moved pages, if possible */
967 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
968 	truncate_pagecache(inode, offset);
969 
970 	new_size = i_size_read(inode) - len;
971 	truncate_pagecache(inode, new_size);
972 
973 	ret = truncate_blocks(inode, new_size, true);
974 	if (!ret)
975 		i_size_write(inode, new_size);
976 
977 	return ret;
978 }
979 
980 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
981 								int mode)
982 {
983 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
984 	struct address_space *mapping = inode->i_mapping;
985 	pgoff_t index, pg_start, pg_end;
986 	loff_t new_size = i_size_read(inode);
987 	loff_t off_start, off_end;
988 	int ret = 0;
989 
990 	ret = inode_newsize_ok(inode, (len + offset));
991 	if (ret)
992 		return ret;
993 
994 	f2fs_balance_fs(sbi);
995 
996 	if (f2fs_has_inline_data(inode)) {
997 		ret = f2fs_convert_inline_inode(inode);
998 		if (ret)
999 			return ret;
1000 	}
1001 
1002 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1003 	if (ret)
1004 		return ret;
1005 
1006 	truncate_pagecache_range(inode, offset, offset + len - 1);
1007 
1008 	pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1009 	pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1010 
1011 	off_start = offset & (PAGE_CACHE_SIZE - 1);
1012 	off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1013 
1014 	if (pg_start == pg_end) {
1015 		ret = fill_zero(inode, pg_start, off_start,
1016 						off_end - off_start);
1017 		if (ret)
1018 			return ret;
1019 
1020 		if (offset + len > new_size)
1021 			new_size = offset + len;
1022 		new_size = max_t(loff_t, new_size, offset + len);
1023 	} else {
1024 		if (off_start) {
1025 			ret = fill_zero(inode, pg_start++, off_start,
1026 						PAGE_CACHE_SIZE - off_start);
1027 			if (ret)
1028 				return ret;
1029 
1030 			new_size = max_t(loff_t, new_size,
1031 					(loff_t)pg_start << PAGE_CACHE_SHIFT);
1032 		}
1033 
1034 		for (index = pg_start; index < pg_end; index++) {
1035 			struct dnode_of_data dn;
1036 			struct page *ipage;
1037 
1038 			f2fs_lock_op(sbi);
1039 
1040 			ipage = get_node_page(sbi, inode->i_ino);
1041 			if (IS_ERR(ipage)) {
1042 				ret = PTR_ERR(ipage);
1043 				f2fs_unlock_op(sbi);
1044 				goto out;
1045 			}
1046 
1047 			set_new_dnode(&dn, inode, ipage, NULL, 0);
1048 			ret = f2fs_reserve_block(&dn, index);
1049 			if (ret) {
1050 				f2fs_unlock_op(sbi);
1051 				goto out;
1052 			}
1053 
1054 			if (dn.data_blkaddr != NEW_ADDR) {
1055 				invalidate_blocks(sbi, dn.data_blkaddr);
1056 
1057 				dn.data_blkaddr = NEW_ADDR;
1058 				set_data_blkaddr(&dn);
1059 
1060 				dn.data_blkaddr = NULL_ADDR;
1061 				f2fs_update_extent_cache(&dn);
1062 			}
1063 			f2fs_put_dnode(&dn);
1064 			f2fs_unlock_op(sbi);
1065 
1066 			new_size = max_t(loff_t, new_size,
1067 				(loff_t)(index + 1) << PAGE_CACHE_SHIFT);
1068 		}
1069 
1070 		if (off_end) {
1071 			ret = fill_zero(inode, pg_end, 0, off_end);
1072 			if (ret)
1073 				goto out;
1074 
1075 			new_size = max_t(loff_t, new_size, offset + len);
1076 		}
1077 	}
1078 
1079 out:
1080 	if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1081 		i_size_write(inode, new_size);
1082 		mark_inode_dirty(inode);
1083 		update_inode_page(inode);
1084 	}
1085 
1086 	return ret;
1087 }
1088 
1089 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1090 {
1091 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1092 	pgoff_t pg_start, pg_end, delta, nrpages, idx;
1093 	loff_t new_size;
1094 	int ret = 0;
1095 
1096 	new_size = i_size_read(inode) + len;
1097 	if (new_size > inode->i_sb->s_maxbytes)
1098 		return -EFBIG;
1099 
1100 	if (offset >= i_size_read(inode))
1101 		return -EINVAL;
1102 
1103 	/* insert range should be aligned to block size of f2fs. */
1104 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1105 		return -EINVAL;
1106 
1107 	f2fs_balance_fs(sbi);
1108 
1109 	if (f2fs_has_inline_data(inode)) {
1110 		ret = f2fs_convert_inline_inode(inode);
1111 		if (ret)
1112 			return ret;
1113 	}
1114 
1115 	ret = truncate_blocks(inode, i_size_read(inode), true);
1116 	if (ret)
1117 		return ret;
1118 
1119 	/* write out all dirty pages from offset */
1120 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1121 	if (ret)
1122 		return ret;
1123 
1124 	truncate_pagecache(inode, offset);
1125 
1126 	pg_start = offset >> PAGE_CACHE_SHIFT;
1127 	pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
1128 	delta = pg_end - pg_start;
1129 	nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1130 
1131 	for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1132 		f2fs_lock_op(sbi);
1133 		ret = __exchange_data_block(inode, idx, idx + delta, false);
1134 		f2fs_unlock_op(sbi);
1135 		if (ret)
1136 			break;
1137 	}
1138 
1139 	/* write out all moved pages, if possible */
1140 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1141 	truncate_pagecache(inode, offset);
1142 
1143 	if (!ret)
1144 		i_size_write(inode, new_size);
1145 	return ret;
1146 }
1147 
1148 static int expand_inode_data(struct inode *inode, loff_t offset,
1149 					loff_t len, int mode)
1150 {
1151 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1152 	pgoff_t index, pg_start, pg_end;
1153 	loff_t new_size = i_size_read(inode);
1154 	loff_t off_start, off_end;
1155 	int ret = 0;
1156 
1157 	f2fs_balance_fs(sbi);
1158 
1159 	ret = inode_newsize_ok(inode, (len + offset));
1160 	if (ret)
1161 		return ret;
1162 
1163 	if (f2fs_has_inline_data(inode)) {
1164 		ret = f2fs_convert_inline_inode(inode);
1165 		if (ret)
1166 			return ret;
1167 	}
1168 
1169 	pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1170 	pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1171 
1172 	off_start = offset & (PAGE_CACHE_SIZE - 1);
1173 	off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1174 
1175 	f2fs_lock_op(sbi);
1176 
1177 	for (index = pg_start; index <= pg_end; index++) {
1178 		struct dnode_of_data dn;
1179 
1180 		if (index == pg_end && !off_end)
1181 			goto noalloc;
1182 
1183 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1184 		ret = f2fs_reserve_block(&dn, index);
1185 		if (ret)
1186 			break;
1187 noalloc:
1188 		if (pg_start == pg_end)
1189 			new_size = offset + len;
1190 		else if (index == pg_start && off_start)
1191 			new_size = (loff_t)(index + 1) << PAGE_CACHE_SHIFT;
1192 		else if (index == pg_end)
1193 			new_size = ((loff_t)index << PAGE_CACHE_SHIFT) +
1194 								off_end;
1195 		else
1196 			new_size += PAGE_CACHE_SIZE;
1197 	}
1198 
1199 	if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1200 		i_size_read(inode) < new_size) {
1201 		i_size_write(inode, new_size);
1202 		mark_inode_dirty(inode);
1203 		update_inode_page(inode);
1204 	}
1205 	f2fs_unlock_op(sbi);
1206 
1207 	return ret;
1208 }
1209 
1210 static long f2fs_fallocate(struct file *file, int mode,
1211 				loff_t offset, loff_t len)
1212 {
1213 	struct inode *inode = file_inode(file);
1214 	long ret = 0;
1215 
1216 	/* f2fs only support ->fallocate for regular file */
1217 	if (!S_ISREG(inode->i_mode))
1218 		return -EINVAL;
1219 
1220 	if (f2fs_encrypted_inode(inode) &&
1221 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1222 		return -EOPNOTSUPP;
1223 
1224 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1225 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1226 			FALLOC_FL_INSERT_RANGE))
1227 		return -EOPNOTSUPP;
1228 
1229 	mutex_lock(&inode->i_mutex);
1230 
1231 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1232 		if (offset >= inode->i_size)
1233 			goto out;
1234 
1235 		ret = punch_hole(inode, offset, len);
1236 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1237 		ret = f2fs_collapse_range(inode, offset, len);
1238 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1239 		ret = f2fs_zero_range(inode, offset, len, mode);
1240 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1241 		ret = f2fs_insert_range(inode, offset, len);
1242 	} else {
1243 		ret = expand_inode_data(inode, offset, len, mode);
1244 	}
1245 
1246 	if (!ret) {
1247 		inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1248 		mark_inode_dirty(inode);
1249 	}
1250 
1251 out:
1252 	mutex_unlock(&inode->i_mutex);
1253 
1254 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1255 	return ret;
1256 }
1257 
1258 static int f2fs_release_file(struct inode *inode, struct file *filp)
1259 {
1260 	/* some remained atomic pages should discarded */
1261 	if (f2fs_is_atomic_file(inode))
1262 		commit_inmem_pages(inode, true);
1263 	if (f2fs_is_volatile_file(inode)) {
1264 		set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1265 		filemap_fdatawrite(inode->i_mapping);
1266 		clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1267 	}
1268 	return 0;
1269 }
1270 
1271 #define F2FS_REG_FLMASK		(~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1272 #define F2FS_OTHER_FLMASK	(FS_NODUMP_FL | FS_NOATIME_FL)
1273 
1274 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1275 {
1276 	if (S_ISDIR(mode))
1277 		return flags;
1278 	else if (S_ISREG(mode))
1279 		return flags & F2FS_REG_FLMASK;
1280 	else
1281 		return flags & F2FS_OTHER_FLMASK;
1282 }
1283 
1284 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1285 {
1286 	struct inode *inode = file_inode(filp);
1287 	struct f2fs_inode_info *fi = F2FS_I(inode);
1288 	unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1289 	return put_user(flags, (int __user *)arg);
1290 }
1291 
1292 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1293 {
1294 	struct inode *inode = file_inode(filp);
1295 	struct f2fs_inode_info *fi = F2FS_I(inode);
1296 	unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1297 	unsigned int oldflags;
1298 	int ret;
1299 
1300 	ret = mnt_want_write_file(filp);
1301 	if (ret)
1302 		return ret;
1303 
1304 	if (!inode_owner_or_capable(inode)) {
1305 		ret = -EACCES;
1306 		goto out;
1307 	}
1308 
1309 	if (get_user(flags, (int __user *)arg)) {
1310 		ret = -EFAULT;
1311 		goto out;
1312 	}
1313 
1314 	flags = f2fs_mask_flags(inode->i_mode, flags);
1315 
1316 	mutex_lock(&inode->i_mutex);
1317 
1318 	oldflags = fi->i_flags;
1319 
1320 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1321 		if (!capable(CAP_LINUX_IMMUTABLE)) {
1322 			mutex_unlock(&inode->i_mutex);
1323 			ret = -EPERM;
1324 			goto out;
1325 		}
1326 	}
1327 
1328 	flags = flags & FS_FL_USER_MODIFIABLE;
1329 	flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1330 	fi->i_flags = flags;
1331 	mutex_unlock(&inode->i_mutex);
1332 
1333 	f2fs_set_inode_flags(inode);
1334 	inode->i_ctime = CURRENT_TIME;
1335 	mark_inode_dirty(inode);
1336 out:
1337 	mnt_drop_write_file(filp);
1338 	return ret;
1339 }
1340 
1341 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1342 {
1343 	struct inode *inode = file_inode(filp);
1344 
1345 	return put_user(inode->i_generation, (int __user *)arg);
1346 }
1347 
1348 static int f2fs_ioc_start_atomic_write(struct file *filp)
1349 {
1350 	struct inode *inode = file_inode(filp);
1351 	int ret;
1352 
1353 	if (!inode_owner_or_capable(inode))
1354 		return -EACCES;
1355 
1356 	f2fs_balance_fs(F2FS_I_SB(inode));
1357 
1358 	if (f2fs_is_atomic_file(inode))
1359 		return 0;
1360 
1361 	ret = f2fs_convert_inline_inode(inode);
1362 	if (ret)
1363 		return ret;
1364 
1365 	set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1366 	return 0;
1367 }
1368 
1369 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1370 {
1371 	struct inode *inode = file_inode(filp);
1372 	int ret;
1373 
1374 	if (!inode_owner_or_capable(inode))
1375 		return -EACCES;
1376 
1377 	if (f2fs_is_volatile_file(inode))
1378 		return 0;
1379 
1380 	ret = mnt_want_write_file(filp);
1381 	if (ret)
1382 		return ret;
1383 
1384 	if (f2fs_is_atomic_file(inode)) {
1385 		clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1386 		ret = commit_inmem_pages(inode, false);
1387 		if (ret)
1388 			goto err_out;
1389 	}
1390 
1391 	ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1392 err_out:
1393 	mnt_drop_write_file(filp);
1394 	return ret;
1395 }
1396 
1397 static int f2fs_ioc_start_volatile_write(struct file *filp)
1398 {
1399 	struct inode *inode = file_inode(filp);
1400 	int ret;
1401 
1402 	if (!inode_owner_or_capable(inode))
1403 		return -EACCES;
1404 
1405 	if (f2fs_is_volatile_file(inode))
1406 		return 0;
1407 
1408 	ret = f2fs_convert_inline_inode(inode);
1409 	if (ret)
1410 		return ret;
1411 
1412 	set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1413 	return 0;
1414 }
1415 
1416 static int f2fs_ioc_release_volatile_write(struct file *filp)
1417 {
1418 	struct inode *inode = file_inode(filp);
1419 
1420 	if (!inode_owner_or_capable(inode))
1421 		return -EACCES;
1422 
1423 	if (!f2fs_is_volatile_file(inode))
1424 		return 0;
1425 
1426 	if (!f2fs_is_first_block_written(inode))
1427 		return truncate_partial_data_page(inode, 0, true);
1428 
1429 	return punch_hole(inode, 0, F2FS_BLKSIZE);
1430 }
1431 
1432 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1433 {
1434 	struct inode *inode = file_inode(filp);
1435 	int ret;
1436 
1437 	if (!inode_owner_or_capable(inode))
1438 		return -EACCES;
1439 
1440 	ret = mnt_want_write_file(filp);
1441 	if (ret)
1442 		return ret;
1443 
1444 	f2fs_balance_fs(F2FS_I_SB(inode));
1445 
1446 	clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1447 	clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1448 	commit_inmem_pages(inode, true);
1449 
1450 	mnt_drop_write_file(filp);
1451 	return ret;
1452 }
1453 
1454 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1455 {
1456 	struct inode *inode = file_inode(filp);
1457 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1458 	struct super_block *sb = sbi->sb;
1459 	__u32 in;
1460 
1461 	if (!capable(CAP_SYS_ADMIN))
1462 		return -EPERM;
1463 
1464 	if (get_user(in, (__u32 __user *)arg))
1465 		return -EFAULT;
1466 
1467 	switch (in) {
1468 	case F2FS_GOING_DOWN_FULLSYNC:
1469 		sb = freeze_bdev(sb->s_bdev);
1470 		if (sb && !IS_ERR(sb)) {
1471 			f2fs_stop_checkpoint(sbi);
1472 			thaw_bdev(sb->s_bdev, sb);
1473 		}
1474 		break;
1475 	case F2FS_GOING_DOWN_METASYNC:
1476 		/* do checkpoint only */
1477 		f2fs_sync_fs(sb, 1);
1478 		f2fs_stop_checkpoint(sbi);
1479 		break;
1480 	case F2FS_GOING_DOWN_NOSYNC:
1481 		f2fs_stop_checkpoint(sbi);
1482 		break;
1483 	case F2FS_GOING_DOWN_METAFLUSH:
1484 		sync_meta_pages(sbi, META, LONG_MAX);
1485 		f2fs_stop_checkpoint(sbi);
1486 		break;
1487 	default:
1488 		return -EINVAL;
1489 	}
1490 	return 0;
1491 }
1492 
1493 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1494 {
1495 	struct inode *inode = file_inode(filp);
1496 	struct super_block *sb = inode->i_sb;
1497 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
1498 	struct fstrim_range range;
1499 	int ret;
1500 
1501 	if (!capable(CAP_SYS_ADMIN))
1502 		return -EPERM;
1503 
1504 	if (!blk_queue_discard(q))
1505 		return -EOPNOTSUPP;
1506 
1507 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1508 				sizeof(range)))
1509 		return -EFAULT;
1510 
1511 	range.minlen = max((unsigned int)range.minlen,
1512 				q->limits.discard_granularity);
1513 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1514 	if (ret < 0)
1515 		return ret;
1516 
1517 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
1518 				sizeof(range)))
1519 		return -EFAULT;
1520 	return 0;
1521 }
1522 
1523 static bool uuid_is_nonzero(__u8 u[16])
1524 {
1525 	int i;
1526 
1527 	for (i = 0; i < 16; i++)
1528 		if (u[i])
1529 			return true;
1530 	return false;
1531 }
1532 
1533 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1534 {
1535 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1536 	struct f2fs_encryption_policy policy;
1537 	struct inode *inode = file_inode(filp);
1538 
1539 	if (copy_from_user(&policy, (struct f2fs_encryption_policy __user *)arg,
1540 				sizeof(policy)))
1541 		return -EFAULT;
1542 
1543 	return f2fs_process_policy(&policy, inode);
1544 #else
1545 	return -EOPNOTSUPP;
1546 #endif
1547 }
1548 
1549 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1550 {
1551 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1552 	struct f2fs_encryption_policy policy;
1553 	struct inode *inode = file_inode(filp);
1554 	int err;
1555 
1556 	err = f2fs_get_policy(inode, &policy);
1557 	if (err)
1558 		return err;
1559 
1560 	if (copy_to_user((struct f2fs_encryption_policy __user *)arg, &policy,
1561 							sizeof(policy)))
1562 		return -EFAULT;
1563 	return 0;
1564 #else
1565 	return -EOPNOTSUPP;
1566 #endif
1567 }
1568 
1569 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1570 {
1571 	struct inode *inode = file_inode(filp);
1572 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1573 	int err;
1574 
1575 	if (!f2fs_sb_has_crypto(inode->i_sb))
1576 		return -EOPNOTSUPP;
1577 
1578 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1579 		goto got_it;
1580 
1581 	err = mnt_want_write_file(filp);
1582 	if (err)
1583 		return err;
1584 
1585 	/* update superblock with uuid */
1586 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1587 
1588 	err = f2fs_commit_super(sbi, false);
1589 
1590 	mnt_drop_write_file(filp);
1591 	if (err) {
1592 		/* undo new data */
1593 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1594 		return err;
1595 	}
1596 got_it:
1597 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1598 									16))
1599 		return -EFAULT;
1600 	return 0;
1601 }
1602 
1603 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1604 {
1605 	struct inode *inode = file_inode(filp);
1606 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1607 	__u32 sync;
1608 
1609 	if (!capable(CAP_SYS_ADMIN))
1610 		return -EPERM;
1611 
1612 	if (get_user(sync, (__u32 __user *)arg))
1613 		return -EFAULT;
1614 
1615 	if (f2fs_readonly(sbi->sb))
1616 		return -EROFS;
1617 
1618 	if (!sync) {
1619 		if (!mutex_trylock(&sbi->gc_mutex))
1620 			return -EBUSY;
1621 	} else {
1622 		mutex_lock(&sbi->gc_mutex);
1623 	}
1624 
1625 	return f2fs_gc(sbi, sync);
1626 }
1627 
1628 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1629 {
1630 	struct inode *inode = file_inode(filp);
1631 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1632 	struct cp_control cpc;
1633 
1634 	if (!capable(CAP_SYS_ADMIN))
1635 		return -EPERM;
1636 
1637 	if (f2fs_readonly(sbi->sb))
1638 		return -EROFS;
1639 
1640 	cpc.reason = __get_cp_reason(sbi);
1641 
1642 	mutex_lock(&sbi->gc_mutex);
1643 	write_checkpoint(sbi, &cpc);
1644 	mutex_unlock(&sbi->gc_mutex);
1645 
1646 	return 0;
1647 }
1648 
1649 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1650 {
1651 	switch (cmd) {
1652 	case F2FS_IOC_GETFLAGS:
1653 		return f2fs_ioc_getflags(filp, arg);
1654 	case F2FS_IOC_SETFLAGS:
1655 		return f2fs_ioc_setflags(filp, arg);
1656 	case F2FS_IOC_GETVERSION:
1657 		return f2fs_ioc_getversion(filp, arg);
1658 	case F2FS_IOC_START_ATOMIC_WRITE:
1659 		return f2fs_ioc_start_atomic_write(filp);
1660 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1661 		return f2fs_ioc_commit_atomic_write(filp);
1662 	case F2FS_IOC_START_VOLATILE_WRITE:
1663 		return f2fs_ioc_start_volatile_write(filp);
1664 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1665 		return f2fs_ioc_release_volatile_write(filp);
1666 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
1667 		return f2fs_ioc_abort_volatile_write(filp);
1668 	case F2FS_IOC_SHUTDOWN:
1669 		return f2fs_ioc_shutdown(filp, arg);
1670 	case FITRIM:
1671 		return f2fs_ioc_fitrim(filp, arg);
1672 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
1673 		return f2fs_ioc_set_encryption_policy(filp, arg);
1674 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
1675 		return f2fs_ioc_get_encryption_policy(filp, arg);
1676 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1677 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1678 	case F2FS_IOC_GARBAGE_COLLECT:
1679 		return f2fs_ioc_gc(filp, arg);
1680 	case F2FS_IOC_WRITE_CHECKPOINT:
1681 		return f2fs_ioc_write_checkpoint(filp, arg);
1682 	default:
1683 		return -ENOTTY;
1684 	}
1685 }
1686 
1687 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1688 {
1689 	struct inode *inode = file_inode(iocb->ki_filp);
1690 
1691 	if (f2fs_encrypted_inode(inode) &&
1692 				!f2fs_has_encryption_key(inode) &&
1693 				f2fs_get_encryption_info(inode))
1694 		return -EACCES;
1695 
1696 	return generic_file_write_iter(iocb, from);
1697 }
1698 
1699 #ifdef CONFIG_COMPAT
1700 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1701 {
1702 	switch (cmd) {
1703 	case F2FS_IOC32_GETFLAGS:
1704 		cmd = F2FS_IOC_GETFLAGS;
1705 		break;
1706 	case F2FS_IOC32_SETFLAGS:
1707 		cmd = F2FS_IOC_SETFLAGS;
1708 		break;
1709 	default:
1710 		return -ENOIOCTLCMD;
1711 	}
1712 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1713 }
1714 #endif
1715 
1716 const struct file_operations f2fs_file_operations = {
1717 	.llseek		= f2fs_llseek,
1718 	.read_iter	= generic_file_read_iter,
1719 	.write_iter	= f2fs_file_write_iter,
1720 	.open		= f2fs_file_open,
1721 	.release	= f2fs_release_file,
1722 	.mmap		= f2fs_file_mmap,
1723 	.fsync		= f2fs_sync_file,
1724 	.fallocate	= f2fs_fallocate,
1725 	.unlocked_ioctl	= f2fs_ioctl,
1726 #ifdef CONFIG_COMPAT
1727 	.compat_ioctl	= f2fs_compat_ioctl,
1728 #endif
1729 	.splice_read	= generic_file_splice_read,
1730 	.splice_write	= iter_file_splice_write,
1731 };
1732