xref: /openbmc/linux/fs/f2fs/file.c (revision 8795a739)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/file.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/buffer_head.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/types.h>
16 #include <linux/compat.h>
17 #include <linux/uaccess.h>
18 #include <linux/mount.h>
19 #include <linux/pagevec.h>
20 #include <linux/uio.h>
21 #include <linux/uuid.h>
22 #include <linux/file.h>
23 #include <linux/nls.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 vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
35 {
36 	struct inode *inode = file_inode(vmf->vma->vm_file);
37 	vm_fault_t ret;
38 
39 	down_read(&F2FS_I(inode)->i_mmap_sem);
40 	ret = filemap_fault(vmf);
41 	up_read(&F2FS_I(inode)->i_mmap_sem);
42 
43 	trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
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 = { .node_changed = false };
54 	int err;
55 
56 	if (unlikely(f2fs_cp_error(sbi))) {
57 		err = -EIO;
58 		goto err;
59 	}
60 
61 	if (!f2fs_is_checkpoint_ready(sbi)) {
62 		err = -ENOSPC;
63 		goto err;
64 	}
65 
66 	sb_start_pagefault(inode->i_sb);
67 
68 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
69 
70 	file_update_time(vmf->vma->vm_file);
71 	down_read(&F2FS_I(inode)->i_mmap_sem);
72 	lock_page(page);
73 	if (unlikely(page->mapping != inode->i_mapping ||
74 			page_offset(page) > i_size_read(inode) ||
75 			!PageUptodate(page))) {
76 		unlock_page(page);
77 		err = -EFAULT;
78 		goto out_sem;
79 	}
80 
81 	/* block allocation */
82 	__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
83 	set_new_dnode(&dn, inode, NULL, NULL, 0);
84 	err = f2fs_get_block(&dn, page->index);
85 	f2fs_put_dnode(&dn);
86 	__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
87 	if (err) {
88 		unlock_page(page);
89 		goto out_sem;
90 	}
91 
92 	/* fill the page */
93 	f2fs_wait_on_page_writeback(page, DATA, false, true);
94 
95 	/* wait for GCed page writeback via META_MAPPING */
96 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
97 
98 	/*
99 	 * check to see if the page is mapped already (no holes)
100 	 */
101 	if (PageMappedToDisk(page))
102 		goto out_sem;
103 
104 	/* page is wholly or partially inside EOF */
105 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
106 						i_size_read(inode)) {
107 		loff_t offset;
108 
109 		offset = i_size_read(inode) & ~PAGE_MASK;
110 		zero_user_segment(page, offset, PAGE_SIZE);
111 	}
112 	set_page_dirty(page);
113 	if (!PageUptodate(page))
114 		SetPageUptodate(page);
115 
116 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
117 	f2fs_update_time(sbi, REQ_TIME);
118 
119 	trace_f2fs_vm_page_mkwrite(page, DATA);
120 out_sem:
121 	up_read(&F2FS_I(inode)->i_mmap_sem);
122 
123 	f2fs_balance_fs(sbi, dn.node_changed);
124 
125 	sb_end_pagefault(inode->i_sb);
126 err:
127 	return block_page_mkwrite_return(err);
128 }
129 
130 static const struct vm_operations_struct f2fs_file_vm_ops = {
131 	.fault		= f2fs_filemap_fault,
132 	.map_pages	= filemap_map_pages,
133 	.page_mkwrite	= f2fs_vm_page_mkwrite,
134 };
135 
136 static int get_parent_ino(struct inode *inode, nid_t *pino)
137 {
138 	struct dentry *dentry;
139 
140 	inode = igrab(inode);
141 	dentry = d_find_any_alias(inode);
142 	iput(inode);
143 	if (!dentry)
144 		return 0;
145 
146 	*pino = parent_ino(dentry);
147 	dput(dentry);
148 	return 1;
149 }
150 
151 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
152 {
153 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
154 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
155 
156 	if (!S_ISREG(inode->i_mode))
157 		cp_reason = CP_NON_REGULAR;
158 	else if (inode->i_nlink != 1)
159 		cp_reason = CP_HARDLINK;
160 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
161 		cp_reason = CP_SB_NEED_CP;
162 	else if (file_wrong_pino(inode))
163 		cp_reason = CP_WRONG_PINO;
164 	else if (!f2fs_space_for_roll_forward(sbi))
165 		cp_reason = CP_NO_SPC_ROLL;
166 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
167 		cp_reason = CP_NODE_NEED_CP;
168 	else if (test_opt(sbi, FASTBOOT))
169 		cp_reason = CP_FASTBOOT_MODE;
170 	else if (F2FS_OPTION(sbi).active_logs == 2)
171 		cp_reason = CP_SPEC_LOG_NUM;
172 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
173 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
174 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
175 							TRANS_DIR_INO))
176 		cp_reason = CP_RECOVER_DIR;
177 
178 	return cp_reason;
179 }
180 
181 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
182 {
183 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
184 	bool ret = false;
185 	/* But we need to avoid that there are some inode updates */
186 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
187 		ret = true;
188 	f2fs_put_page(i, 0);
189 	return ret;
190 }
191 
192 static void try_to_fix_pino(struct inode *inode)
193 {
194 	struct f2fs_inode_info *fi = F2FS_I(inode);
195 	nid_t pino;
196 
197 	down_write(&fi->i_sem);
198 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
199 			get_parent_ino(inode, &pino)) {
200 		f2fs_i_pino_write(inode, pino);
201 		file_got_pino(inode);
202 	}
203 	up_write(&fi->i_sem);
204 }
205 
206 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
207 						int datasync, bool atomic)
208 {
209 	struct inode *inode = file->f_mapping->host;
210 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
211 	nid_t ino = inode->i_ino;
212 	int ret = 0;
213 	enum cp_reason_type cp_reason = 0;
214 	struct writeback_control wbc = {
215 		.sync_mode = WB_SYNC_ALL,
216 		.nr_to_write = LONG_MAX,
217 		.for_reclaim = 0,
218 	};
219 	unsigned int seq_id = 0;
220 
221 	if (unlikely(f2fs_readonly(inode->i_sb) ||
222 				is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
223 		return 0;
224 
225 	trace_f2fs_sync_file_enter(inode);
226 
227 	if (S_ISDIR(inode->i_mode))
228 		goto go_write;
229 
230 	/* if fdatasync is triggered, let's do in-place-update */
231 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
232 		set_inode_flag(inode, FI_NEED_IPU);
233 	ret = file_write_and_wait_range(file, start, end);
234 	clear_inode_flag(inode, FI_NEED_IPU);
235 
236 	if (ret) {
237 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
238 		return ret;
239 	}
240 
241 	/* if the inode is dirty, let's recover all the time */
242 	if (!f2fs_skip_inode_update(inode, datasync)) {
243 		f2fs_write_inode(inode, NULL);
244 		goto go_write;
245 	}
246 
247 	/*
248 	 * if there is no written data, don't waste time to write recovery info.
249 	 */
250 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
251 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
252 
253 		/* it may call write_inode just prior to fsync */
254 		if (need_inode_page_update(sbi, ino))
255 			goto go_write;
256 
257 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
258 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
259 			goto flush_out;
260 		goto out;
261 	}
262 go_write:
263 	/*
264 	 * Both of fdatasync() and fsync() are able to be recovered from
265 	 * sudden-power-off.
266 	 */
267 	down_read(&F2FS_I(inode)->i_sem);
268 	cp_reason = need_do_checkpoint(inode);
269 	up_read(&F2FS_I(inode)->i_sem);
270 
271 	if (cp_reason) {
272 		/* all the dirty node pages should be flushed for POR */
273 		ret = f2fs_sync_fs(inode->i_sb, 1);
274 
275 		/*
276 		 * We've secured consistency through sync_fs. Following pino
277 		 * will be used only for fsynced inodes after checkpoint.
278 		 */
279 		try_to_fix_pino(inode);
280 		clear_inode_flag(inode, FI_APPEND_WRITE);
281 		clear_inode_flag(inode, FI_UPDATE_WRITE);
282 		goto out;
283 	}
284 sync_nodes:
285 	atomic_inc(&sbi->wb_sync_req[NODE]);
286 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
287 	atomic_dec(&sbi->wb_sync_req[NODE]);
288 	if (ret)
289 		goto out;
290 
291 	/* if cp_error was enabled, we should avoid infinite loop */
292 	if (unlikely(f2fs_cp_error(sbi))) {
293 		ret = -EIO;
294 		goto out;
295 	}
296 
297 	if (f2fs_need_inode_block_update(sbi, ino)) {
298 		f2fs_mark_inode_dirty_sync(inode, true);
299 		f2fs_write_inode(inode, NULL);
300 		goto sync_nodes;
301 	}
302 
303 	/*
304 	 * If it's atomic_write, it's just fine to keep write ordering. So
305 	 * here we don't need to wait for node write completion, since we use
306 	 * node chain which serializes node blocks. If one of node writes are
307 	 * reordered, we can see simply broken chain, resulting in stopping
308 	 * roll-forward recovery. It means we'll recover all or none node blocks
309 	 * given fsync mark.
310 	 */
311 	if (!atomic) {
312 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
313 		if (ret)
314 			goto out;
315 	}
316 
317 	/* once recovery info is written, don't need to tack this */
318 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
319 	clear_inode_flag(inode, FI_APPEND_WRITE);
320 flush_out:
321 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
322 		ret = f2fs_issue_flush(sbi, inode->i_ino);
323 	if (!ret) {
324 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
325 		clear_inode_flag(inode, FI_UPDATE_WRITE);
326 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
327 	}
328 	f2fs_update_time(sbi, REQ_TIME);
329 out:
330 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
331 	f2fs_trace_ios(NULL, 1);
332 	return ret;
333 }
334 
335 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
336 {
337 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
338 		return -EIO;
339 	return f2fs_do_sync_file(file, start, end, datasync, false);
340 }
341 
342 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
343 						pgoff_t pgofs, int whence)
344 {
345 	struct page *page;
346 	int nr_pages;
347 
348 	if (whence != SEEK_DATA)
349 		return 0;
350 
351 	/* find first dirty page index */
352 	nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
353 				      1, &page);
354 	if (!nr_pages)
355 		return ULONG_MAX;
356 	pgofs = page->index;
357 	put_page(page);
358 	return pgofs;
359 }
360 
361 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
362 				pgoff_t dirty, pgoff_t pgofs, int whence)
363 {
364 	switch (whence) {
365 	case SEEK_DATA:
366 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
367 			__is_valid_data_blkaddr(blkaddr))
368 			return true;
369 		break;
370 	case SEEK_HOLE:
371 		if (blkaddr == NULL_ADDR)
372 			return true;
373 		break;
374 	}
375 	return false;
376 }
377 
378 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
379 {
380 	struct inode *inode = file->f_mapping->host;
381 	loff_t maxbytes = inode->i_sb->s_maxbytes;
382 	struct dnode_of_data dn;
383 	pgoff_t pgofs, end_offset, dirty;
384 	loff_t data_ofs = offset;
385 	loff_t isize;
386 	int err = 0;
387 
388 	inode_lock(inode);
389 
390 	isize = i_size_read(inode);
391 	if (offset >= isize)
392 		goto fail;
393 
394 	/* handle inline data case */
395 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
396 		if (whence == SEEK_HOLE)
397 			data_ofs = isize;
398 		goto found;
399 	}
400 
401 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
402 
403 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
404 
405 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
406 		set_new_dnode(&dn, inode, NULL, NULL, 0);
407 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
408 		if (err && err != -ENOENT) {
409 			goto fail;
410 		} else if (err == -ENOENT) {
411 			/* direct node does not exists */
412 			if (whence == SEEK_DATA) {
413 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
414 				continue;
415 			} else {
416 				goto found;
417 			}
418 		}
419 
420 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
421 
422 		/* find data/hole in dnode block */
423 		for (; dn.ofs_in_node < end_offset;
424 				dn.ofs_in_node++, pgofs++,
425 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
426 			block_t blkaddr;
427 
428 			blkaddr = datablock_addr(dn.inode,
429 					dn.node_page, dn.ofs_in_node);
430 
431 			if (__is_valid_data_blkaddr(blkaddr) &&
432 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
433 					blkaddr, DATA_GENERIC_ENHANCE)) {
434 				f2fs_put_dnode(&dn);
435 				goto fail;
436 			}
437 
438 			if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
439 							pgofs, whence)) {
440 				f2fs_put_dnode(&dn);
441 				goto found;
442 			}
443 		}
444 		f2fs_put_dnode(&dn);
445 	}
446 
447 	if (whence == SEEK_DATA)
448 		goto fail;
449 found:
450 	if (whence == SEEK_HOLE && data_ofs > isize)
451 		data_ofs = isize;
452 	inode_unlock(inode);
453 	return vfs_setpos(file, data_ofs, maxbytes);
454 fail:
455 	inode_unlock(inode);
456 	return -ENXIO;
457 }
458 
459 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
460 {
461 	struct inode *inode = file->f_mapping->host;
462 	loff_t maxbytes = inode->i_sb->s_maxbytes;
463 
464 	switch (whence) {
465 	case SEEK_SET:
466 	case SEEK_CUR:
467 	case SEEK_END:
468 		return generic_file_llseek_size(file, offset, whence,
469 						maxbytes, i_size_read(inode));
470 	case SEEK_DATA:
471 	case SEEK_HOLE:
472 		if (offset < 0)
473 			return -ENXIO;
474 		return f2fs_seek_block(file, offset, whence);
475 	}
476 
477 	return -EINVAL;
478 }
479 
480 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
481 {
482 	struct inode *inode = file_inode(file);
483 	int err;
484 
485 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
486 		return -EIO;
487 
488 	/* we don't need to use inline_data strictly */
489 	err = f2fs_convert_inline_inode(inode);
490 	if (err)
491 		return err;
492 
493 	file_accessed(file);
494 	vma->vm_ops = &f2fs_file_vm_ops;
495 	return 0;
496 }
497 
498 static int f2fs_file_open(struct inode *inode, struct file *filp)
499 {
500 	int err = fscrypt_file_open(inode, filp);
501 
502 	if (err)
503 		return err;
504 
505 	err = fsverity_file_open(inode, filp);
506 	if (err)
507 		return err;
508 
509 	filp->f_mode |= FMODE_NOWAIT;
510 
511 	return dquot_file_open(inode, filp);
512 }
513 
514 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
515 {
516 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
517 	struct f2fs_node *raw_node;
518 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
519 	__le32 *addr;
520 	int base = 0;
521 
522 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
523 		base = get_extra_isize(dn->inode);
524 
525 	raw_node = F2FS_NODE(dn->node_page);
526 	addr = blkaddr_in_node(raw_node) + base + ofs;
527 
528 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
529 		block_t blkaddr = le32_to_cpu(*addr);
530 
531 		if (blkaddr == NULL_ADDR)
532 			continue;
533 
534 		dn->data_blkaddr = NULL_ADDR;
535 		f2fs_set_data_blkaddr(dn);
536 
537 		if (__is_valid_data_blkaddr(blkaddr) &&
538 			!f2fs_is_valid_blkaddr(sbi, blkaddr,
539 					DATA_GENERIC_ENHANCE))
540 			continue;
541 
542 		f2fs_invalidate_blocks(sbi, blkaddr);
543 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
544 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
545 		nr_free++;
546 	}
547 
548 	if (nr_free) {
549 		pgoff_t fofs;
550 		/*
551 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
552 		 * we will invalidate all blkaddr in the whole range.
553 		 */
554 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
555 							dn->inode) + ofs;
556 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
557 		dec_valid_block_count(sbi, dn->inode, nr_free);
558 	}
559 	dn->ofs_in_node = ofs;
560 
561 	f2fs_update_time(sbi, REQ_TIME);
562 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
563 					 dn->ofs_in_node, nr_free);
564 }
565 
566 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
567 {
568 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
569 }
570 
571 static int truncate_partial_data_page(struct inode *inode, u64 from,
572 								bool cache_only)
573 {
574 	loff_t offset = from & (PAGE_SIZE - 1);
575 	pgoff_t index = from >> PAGE_SHIFT;
576 	struct address_space *mapping = inode->i_mapping;
577 	struct page *page;
578 
579 	if (!offset && !cache_only)
580 		return 0;
581 
582 	if (cache_only) {
583 		page = find_lock_page(mapping, index);
584 		if (page && PageUptodate(page))
585 			goto truncate_out;
586 		f2fs_put_page(page, 1);
587 		return 0;
588 	}
589 
590 	page = f2fs_get_lock_data_page(inode, index, true);
591 	if (IS_ERR(page))
592 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
593 truncate_out:
594 	f2fs_wait_on_page_writeback(page, DATA, true, true);
595 	zero_user(page, offset, PAGE_SIZE - offset);
596 
597 	/* An encrypted inode should have a key and truncate the last page. */
598 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
599 	if (!cache_only)
600 		set_page_dirty(page);
601 	f2fs_put_page(page, 1);
602 	return 0;
603 }
604 
605 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
606 {
607 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
608 	struct dnode_of_data dn;
609 	pgoff_t free_from;
610 	int count = 0, err = 0;
611 	struct page *ipage;
612 	bool truncate_page = false;
613 
614 	trace_f2fs_truncate_blocks_enter(inode, from);
615 
616 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
617 
618 	if (free_from >= sbi->max_file_blocks)
619 		goto free_partial;
620 
621 	if (lock)
622 		f2fs_lock_op(sbi);
623 
624 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
625 	if (IS_ERR(ipage)) {
626 		err = PTR_ERR(ipage);
627 		goto out;
628 	}
629 
630 	if (f2fs_has_inline_data(inode)) {
631 		f2fs_truncate_inline_inode(inode, ipage, from);
632 		f2fs_put_page(ipage, 1);
633 		truncate_page = true;
634 		goto out;
635 	}
636 
637 	set_new_dnode(&dn, inode, ipage, NULL, 0);
638 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
639 	if (err) {
640 		if (err == -ENOENT)
641 			goto free_next;
642 		goto out;
643 	}
644 
645 	count = ADDRS_PER_PAGE(dn.node_page, inode);
646 
647 	count -= dn.ofs_in_node;
648 	f2fs_bug_on(sbi, count < 0);
649 
650 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
651 		f2fs_truncate_data_blocks_range(&dn, count);
652 		free_from += count;
653 	}
654 
655 	f2fs_put_dnode(&dn);
656 free_next:
657 	err = f2fs_truncate_inode_blocks(inode, free_from);
658 out:
659 	if (lock)
660 		f2fs_unlock_op(sbi);
661 free_partial:
662 	/* lastly zero out the first data page */
663 	if (!err)
664 		err = truncate_partial_data_page(inode, from, truncate_page);
665 
666 	trace_f2fs_truncate_blocks_exit(inode, err);
667 	return err;
668 }
669 
670 int f2fs_truncate(struct inode *inode)
671 {
672 	int err;
673 
674 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
675 		return -EIO;
676 
677 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
678 				S_ISLNK(inode->i_mode)))
679 		return 0;
680 
681 	trace_f2fs_truncate(inode);
682 
683 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
684 		f2fs_show_injection_info(FAULT_TRUNCATE);
685 		return -EIO;
686 	}
687 
688 	/* we should check inline_data size */
689 	if (!f2fs_may_inline_data(inode)) {
690 		err = f2fs_convert_inline_inode(inode);
691 		if (err)
692 			return err;
693 	}
694 
695 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
696 	if (err)
697 		return err;
698 
699 	inode->i_mtime = inode->i_ctime = current_time(inode);
700 	f2fs_mark_inode_dirty_sync(inode, false);
701 	return 0;
702 }
703 
704 int f2fs_getattr(const struct path *path, struct kstat *stat,
705 		 u32 request_mask, unsigned int query_flags)
706 {
707 	struct inode *inode = d_inode(path->dentry);
708 	struct f2fs_inode_info *fi = F2FS_I(inode);
709 	struct f2fs_inode *ri;
710 	unsigned int flags;
711 
712 	if (f2fs_has_extra_attr(inode) &&
713 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
714 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
715 		stat->result_mask |= STATX_BTIME;
716 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
717 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
718 	}
719 
720 	flags = fi->i_flags;
721 	if (flags & F2FS_APPEND_FL)
722 		stat->attributes |= STATX_ATTR_APPEND;
723 	if (IS_ENCRYPTED(inode))
724 		stat->attributes |= STATX_ATTR_ENCRYPTED;
725 	if (flags & F2FS_IMMUTABLE_FL)
726 		stat->attributes |= STATX_ATTR_IMMUTABLE;
727 	if (flags & F2FS_NODUMP_FL)
728 		stat->attributes |= STATX_ATTR_NODUMP;
729 
730 	stat->attributes_mask |= (STATX_ATTR_APPEND |
731 				  STATX_ATTR_ENCRYPTED |
732 				  STATX_ATTR_IMMUTABLE |
733 				  STATX_ATTR_NODUMP);
734 
735 	generic_fillattr(inode, stat);
736 
737 	/* we need to show initial sectors used for inline_data/dentries */
738 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
739 					f2fs_has_inline_dentry(inode))
740 		stat->blocks += (stat->size + 511) >> 9;
741 
742 	return 0;
743 }
744 
745 #ifdef CONFIG_F2FS_FS_POSIX_ACL
746 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
747 {
748 	unsigned int ia_valid = attr->ia_valid;
749 
750 	if (ia_valid & ATTR_UID)
751 		inode->i_uid = attr->ia_uid;
752 	if (ia_valid & ATTR_GID)
753 		inode->i_gid = attr->ia_gid;
754 	if (ia_valid & ATTR_ATIME) {
755 		inode->i_atime = timestamp_truncate(attr->ia_atime,
756 						  inode);
757 	}
758 	if (ia_valid & ATTR_MTIME) {
759 		inode->i_mtime = timestamp_truncate(attr->ia_mtime,
760 						  inode);
761 	}
762 	if (ia_valid & ATTR_CTIME) {
763 		inode->i_ctime = timestamp_truncate(attr->ia_ctime,
764 						  inode);
765 	}
766 	if (ia_valid & ATTR_MODE) {
767 		umode_t mode = attr->ia_mode;
768 
769 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
770 			mode &= ~S_ISGID;
771 		set_acl_inode(inode, mode);
772 	}
773 }
774 #else
775 #define __setattr_copy setattr_copy
776 #endif
777 
778 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
779 {
780 	struct inode *inode = d_inode(dentry);
781 	int err;
782 
783 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
784 		return -EIO;
785 
786 	err = setattr_prepare(dentry, attr);
787 	if (err)
788 		return err;
789 
790 	err = fscrypt_prepare_setattr(dentry, attr);
791 	if (err)
792 		return err;
793 
794 	err = fsverity_prepare_setattr(dentry, attr);
795 	if (err)
796 		return err;
797 
798 	if (is_quota_modification(inode, attr)) {
799 		err = dquot_initialize(inode);
800 		if (err)
801 			return err;
802 	}
803 	if ((attr->ia_valid & ATTR_UID &&
804 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
805 		(attr->ia_valid & ATTR_GID &&
806 		!gid_eq(attr->ia_gid, inode->i_gid))) {
807 		f2fs_lock_op(F2FS_I_SB(inode));
808 		err = dquot_transfer(inode, attr);
809 		if (err) {
810 			set_sbi_flag(F2FS_I_SB(inode),
811 					SBI_QUOTA_NEED_REPAIR);
812 			f2fs_unlock_op(F2FS_I_SB(inode));
813 			return err;
814 		}
815 		/*
816 		 * update uid/gid under lock_op(), so that dquot and inode can
817 		 * be updated atomically.
818 		 */
819 		if (attr->ia_valid & ATTR_UID)
820 			inode->i_uid = attr->ia_uid;
821 		if (attr->ia_valid & ATTR_GID)
822 			inode->i_gid = attr->ia_gid;
823 		f2fs_mark_inode_dirty_sync(inode, true);
824 		f2fs_unlock_op(F2FS_I_SB(inode));
825 	}
826 
827 	if (attr->ia_valid & ATTR_SIZE) {
828 		loff_t old_size = i_size_read(inode);
829 
830 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
831 			/*
832 			 * should convert inline inode before i_size_write to
833 			 * keep smaller than inline_data size with inline flag.
834 			 */
835 			err = f2fs_convert_inline_inode(inode);
836 			if (err)
837 				return err;
838 		}
839 
840 		down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
841 		down_write(&F2FS_I(inode)->i_mmap_sem);
842 
843 		truncate_setsize(inode, attr->ia_size);
844 
845 		if (attr->ia_size <= old_size)
846 			err = f2fs_truncate(inode);
847 		/*
848 		 * do not trim all blocks after i_size if target size is
849 		 * larger than i_size.
850 		 */
851 		up_write(&F2FS_I(inode)->i_mmap_sem);
852 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
853 		if (err)
854 			return err;
855 
856 		down_write(&F2FS_I(inode)->i_sem);
857 		inode->i_mtime = inode->i_ctime = current_time(inode);
858 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
859 		up_write(&F2FS_I(inode)->i_sem);
860 	}
861 
862 	__setattr_copy(inode, attr);
863 
864 	if (attr->ia_valid & ATTR_MODE) {
865 		err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
866 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
867 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
868 			clear_inode_flag(inode, FI_ACL_MODE);
869 		}
870 	}
871 
872 	/* file size may changed here */
873 	f2fs_mark_inode_dirty_sync(inode, true);
874 
875 	/* inode change will produce dirty node pages flushed by checkpoint */
876 	f2fs_balance_fs(F2FS_I_SB(inode), true);
877 
878 	return err;
879 }
880 
881 const struct inode_operations f2fs_file_inode_operations = {
882 	.getattr	= f2fs_getattr,
883 	.setattr	= f2fs_setattr,
884 	.get_acl	= f2fs_get_acl,
885 	.set_acl	= f2fs_set_acl,
886 #ifdef CONFIG_F2FS_FS_XATTR
887 	.listxattr	= f2fs_listxattr,
888 #endif
889 	.fiemap		= f2fs_fiemap,
890 };
891 
892 static int fill_zero(struct inode *inode, pgoff_t index,
893 					loff_t start, loff_t len)
894 {
895 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
896 	struct page *page;
897 
898 	if (!len)
899 		return 0;
900 
901 	f2fs_balance_fs(sbi, true);
902 
903 	f2fs_lock_op(sbi);
904 	page = f2fs_get_new_data_page(inode, NULL, index, false);
905 	f2fs_unlock_op(sbi);
906 
907 	if (IS_ERR(page))
908 		return PTR_ERR(page);
909 
910 	f2fs_wait_on_page_writeback(page, DATA, true, true);
911 	zero_user(page, start, len);
912 	set_page_dirty(page);
913 	f2fs_put_page(page, 1);
914 	return 0;
915 }
916 
917 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
918 {
919 	int err;
920 
921 	while (pg_start < pg_end) {
922 		struct dnode_of_data dn;
923 		pgoff_t end_offset, count;
924 
925 		set_new_dnode(&dn, inode, NULL, NULL, 0);
926 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
927 		if (err) {
928 			if (err == -ENOENT) {
929 				pg_start = f2fs_get_next_page_offset(&dn,
930 								pg_start);
931 				continue;
932 			}
933 			return err;
934 		}
935 
936 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
937 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
938 
939 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
940 
941 		f2fs_truncate_data_blocks_range(&dn, count);
942 		f2fs_put_dnode(&dn);
943 
944 		pg_start += count;
945 	}
946 	return 0;
947 }
948 
949 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
950 {
951 	pgoff_t pg_start, pg_end;
952 	loff_t off_start, off_end;
953 	int ret;
954 
955 	ret = f2fs_convert_inline_inode(inode);
956 	if (ret)
957 		return ret;
958 
959 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
960 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
961 
962 	off_start = offset & (PAGE_SIZE - 1);
963 	off_end = (offset + len) & (PAGE_SIZE - 1);
964 
965 	if (pg_start == pg_end) {
966 		ret = fill_zero(inode, pg_start, off_start,
967 						off_end - off_start);
968 		if (ret)
969 			return ret;
970 	} else {
971 		if (off_start) {
972 			ret = fill_zero(inode, pg_start++, off_start,
973 						PAGE_SIZE - off_start);
974 			if (ret)
975 				return ret;
976 		}
977 		if (off_end) {
978 			ret = fill_zero(inode, pg_end, 0, off_end);
979 			if (ret)
980 				return ret;
981 		}
982 
983 		if (pg_start < pg_end) {
984 			struct address_space *mapping = inode->i_mapping;
985 			loff_t blk_start, blk_end;
986 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
987 
988 			f2fs_balance_fs(sbi, true);
989 
990 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
991 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
992 
993 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
994 			down_write(&F2FS_I(inode)->i_mmap_sem);
995 
996 			truncate_inode_pages_range(mapping, blk_start,
997 					blk_end - 1);
998 
999 			f2fs_lock_op(sbi);
1000 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1001 			f2fs_unlock_op(sbi);
1002 
1003 			up_write(&F2FS_I(inode)->i_mmap_sem);
1004 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1005 		}
1006 	}
1007 
1008 	return ret;
1009 }
1010 
1011 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1012 				int *do_replace, pgoff_t off, pgoff_t len)
1013 {
1014 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1015 	struct dnode_of_data dn;
1016 	int ret, done, i;
1017 
1018 next_dnode:
1019 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1020 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1021 	if (ret && ret != -ENOENT) {
1022 		return ret;
1023 	} else if (ret == -ENOENT) {
1024 		if (dn.max_level == 0)
1025 			return -ENOENT;
1026 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1027 									len);
1028 		blkaddr += done;
1029 		do_replace += done;
1030 		goto next;
1031 	}
1032 
1033 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1034 							dn.ofs_in_node, len);
1035 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1036 		*blkaddr = datablock_addr(dn.inode,
1037 					dn.node_page, dn.ofs_in_node);
1038 
1039 		if (__is_valid_data_blkaddr(*blkaddr) &&
1040 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1041 					DATA_GENERIC_ENHANCE)) {
1042 			f2fs_put_dnode(&dn);
1043 			return -EFSCORRUPTED;
1044 		}
1045 
1046 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1047 
1048 			if (test_opt(sbi, LFS)) {
1049 				f2fs_put_dnode(&dn);
1050 				return -EOPNOTSUPP;
1051 			}
1052 
1053 			/* do not invalidate this block address */
1054 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1055 			*do_replace = 1;
1056 		}
1057 	}
1058 	f2fs_put_dnode(&dn);
1059 next:
1060 	len -= done;
1061 	off += done;
1062 	if (len)
1063 		goto next_dnode;
1064 	return 0;
1065 }
1066 
1067 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1068 				int *do_replace, pgoff_t off, int len)
1069 {
1070 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1071 	struct dnode_of_data dn;
1072 	int ret, i;
1073 
1074 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1075 		if (*do_replace == 0)
1076 			continue;
1077 
1078 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1079 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1080 		if (ret) {
1081 			dec_valid_block_count(sbi, inode, 1);
1082 			f2fs_invalidate_blocks(sbi, *blkaddr);
1083 		} else {
1084 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1085 		}
1086 		f2fs_put_dnode(&dn);
1087 	}
1088 	return 0;
1089 }
1090 
1091 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1092 			block_t *blkaddr, int *do_replace,
1093 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1094 {
1095 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1096 	pgoff_t i = 0;
1097 	int ret;
1098 
1099 	while (i < len) {
1100 		if (blkaddr[i] == NULL_ADDR && !full) {
1101 			i++;
1102 			continue;
1103 		}
1104 
1105 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1106 			struct dnode_of_data dn;
1107 			struct node_info ni;
1108 			size_t new_size;
1109 			pgoff_t ilen;
1110 
1111 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1112 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1113 			if (ret)
1114 				return ret;
1115 
1116 			ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1117 			if (ret) {
1118 				f2fs_put_dnode(&dn);
1119 				return ret;
1120 			}
1121 
1122 			ilen = min((pgoff_t)
1123 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1124 						dn.ofs_in_node, len - i);
1125 			do {
1126 				dn.data_blkaddr = datablock_addr(dn.inode,
1127 						dn.node_page, dn.ofs_in_node);
1128 				f2fs_truncate_data_blocks_range(&dn, 1);
1129 
1130 				if (do_replace[i]) {
1131 					f2fs_i_blocks_write(src_inode,
1132 							1, false, false);
1133 					f2fs_i_blocks_write(dst_inode,
1134 							1, true, false);
1135 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1136 					blkaddr[i], ni.version, true, false);
1137 
1138 					do_replace[i] = 0;
1139 				}
1140 				dn.ofs_in_node++;
1141 				i++;
1142 				new_size = (dst + i) << PAGE_SHIFT;
1143 				if (dst_inode->i_size < new_size)
1144 					f2fs_i_size_write(dst_inode, new_size);
1145 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1146 
1147 			f2fs_put_dnode(&dn);
1148 		} else {
1149 			struct page *psrc, *pdst;
1150 
1151 			psrc = f2fs_get_lock_data_page(src_inode,
1152 							src + i, true);
1153 			if (IS_ERR(psrc))
1154 				return PTR_ERR(psrc);
1155 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1156 								true);
1157 			if (IS_ERR(pdst)) {
1158 				f2fs_put_page(psrc, 1);
1159 				return PTR_ERR(pdst);
1160 			}
1161 			f2fs_copy_page(psrc, pdst);
1162 			set_page_dirty(pdst);
1163 			f2fs_put_page(pdst, 1);
1164 			f2fs_put_page(psrc, 1);
1165 
1166 			ret = f2fs_truncate_hole(src_inode,
1167 						src + i, src + i + 1);
1168 			if (ret)
1169 				return ret;
1170 			i++;
1171 		}
1172 	}
1173 	return 0;
1174 }
1175 
1176 static int __exchange_data_block(struct inode *src_inode,
1177 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1178 			pgoff_t len, bool full)
1179 {
1180 	block_t *src_blkaddr;
1181 	int *do_replace;
1182 	pgoff_t olen;
1183 	int ret;
1184 
1185 	while (len) {
1186 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1187 
1188 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1189 					array_size(olen, sizeof(block_t)),
1190 					GFP_KERNEL);
1191 		if (!src_blkaddr)
1192 			return -ENOMEM;
1193 
1194 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1195 					array_size(olen, sizeof(int)),
1196 					GFP_KERNEL);
1197 		if (!do_replace) {
1198 			kvfree(src_blkaddr);
1199 			return -ENOMEM;
1200 		}
1201 
1202 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1203 					do_replace, src, olen);
1204 		if (ret)
1205 			goto roll_back;
1206 
1207 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1208 					do_replace, src, dst, olen, full);
1209 		if (ret)
1210 			goto roll_back;
1211 
1212 		src += olen;
1213 		dst += olen;
1214 		len -= olen;
1215 
1216 		kvfree(src_blkaddr);
1217 		kvfree(do_replace);
1218 	}
1219 	return 0;
1220 
1221 roll_back:
1222 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1223 	kvfree(src_blkaddr);
1224 	kvfree(do_replace);
1225 	return ret;
1226 }
1227 
1228 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1229 {
1230 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1231 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1232 	pgoff_t start = offset >> PAGE_SHIFT;
1233 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1234 	int ret;
1235 
1236 	f2fs_balance_fs(sbi, true);
1237 
1238 	/* avoid gc operation during block exchange */
1239 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1240 	down_write(&F2FS_I(inode)->i_mmap_sem);
1241 
1242 	f2fs_lock_op(sbi);
1243 	f2fs_drop_extent_tree(inode);
1244 	truncate_pagecache(inode, offset);
1245 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1246 	f2fs_unlock_op(sbi);
1247 
1248 	up_write(&F2FS_I(inode)->i_mmap_sem);
1249 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1250 	return ret;
1251 }
1252 
1253 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1254 {
1255 	loff_t new_size;
1256 	int ret;
1257 
1258 	if (offset + len >= i_size_read(inode))
1259 		return -EINVAL;
1260 
1261 	/* collapse range should be aligned to block size of f2fs. */
1262 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1263 		return -EINVAL;
1264 
1265 	ret = f2fs_convert_inline_inode(inode);
1266 	if (ret)
1267 		return ret;
1268 
1269 	/* write out all dirty pages from offset */
1270 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1271 	if (ret)
1272 		return ret;
1273 
1274 	ret = f2fs_do_collapse(inode, offset, len);
1275 	if (ret)
1276 		return ret;
1277 
1278 	/* write out all moved pages, if possible */
1279 	down_write(&F2FS_I(inode)->i_mmap_sem);
1280 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1281 	truncate_pagecache(inode, offset);
1282 
1283 	new_size = i_size_read(inode) - len;
1284 	truncate_pagecache(inode, new_size);
1285 
1286 	ret = f2fs_truncate_blocks(inode, new_size, true);
1287 	up_write(&F2FS_I(inode)->i_mmap_sem);
1288 	if (!ret)
1289 		f2fs_i_size_write(inode, new_size);
1290 	return ret;
1291 }
1292 
1293 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1294 								pgoff_t end)
1295 {
1296 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1297 	pgoff_t index = start;
1298 	unsigned int ofs_in_node = dn->ofs_in_node;
1299 	blkcnt_t count = 0;
1300 	int ret;
1301 
1302 	for (; index < end; index++, dn->ofs_in_node++) {
1303 		if (datablock_addr(dn->inode, dn->node_page,
1304 					dn->ofs_in_node) == NULL_ADDR)
1305 			count++;
1306 	}
1307 
1308 	dn->ofs_in_node = ofs_in_node;
1309 	ret = f2fs_reserve_new_blocks(dn, count);
1310 	if (ret)
1311 		return ret;
1312 
1313 	dn->ofs_in_node = ofs_in_node;
1314 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1315 		dn->data_blkaddr = datablock_addr(dn->inode,
1316 					dn->node_page, dn->ofs_in_node);
1317 		/*
1318 		 * f2fs_reserve_new_blocks will not guarantee entire block
1319 		 * allocation.
1320 		 */
1321 		if (dn->data_blkaddr == NULL_ADDR) {
1322 			ret = -ENOSPC;
1323 			break;
1324 		}
1325 		if (dn->data_blkaddr != NEW_ADDR) {
1326 			f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1327 			dn->data_blkaddr = NEW_ADDR;
1328 			f2fs_set_data_blkaddr(dn);
1329 		}
1330 	}
1331 
1332 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1333 
1334 	return ret;
1335 }
1336 
1337 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1338 								int mode)
1339 {
1340 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1341 	struct address_space *mapping = inode->i_mapping;
1342 	pgoff_t index, pg_start, pg_end;
1343 	loff_t new_size = i_size_read(inode);
1344 	loff_t off_start, off_end;
1345 	int ret = 0;
1346 
1347 	ret = inode_newsize_ok(inode, (len + offset));
1348 	if (ret)
1349 		return ret;
1350 
1351 	ret = f2fs_convert_inline_inode(inode);
1352 	if (ret)
1353 		return ret;
1354 
1355 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1356 	if (ret)
1357 		return ret;
1358 
1359 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1360 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1361 
1362 	off_start = offset & (PAGE_SIZE - 1);
1363 	off_end = (offset + len) & (PAGE_SIZE - 1);
1364 
1365 	if (pg_start == pg_end) {
1366 		ret = fill_zero(inode, pg_start, off_start,
1367 						off_end - off_start);
1368 		if (ret)
1369 			return ret;
1370 
1371 		new_size = max_t(loff_t, new_size, offset + len);
1372 	} else {
1373 		if (off_start) {
1374 			ret = fill_zero(inode, pg_start++, off_start,
1375 						PAGE_SIZE - off_start);
1376 			if (ret)
1377 				return ret;
1378 
1379 			new_size = max_t(loff_t, new_size,
1380 					(loff_t)pg_start << PAGE_SHIFT);
1381 		}
1382 
1383 		for (index = pg_start; index < pg_end;) {
1384 			struct dnode_of_data dn;
1385 			unsigned int end_offset;
1386 			pgoff_t end;
1387 
1388 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1389 			down_write(&F2FS_I(inode)->i_mmap_sem);
1390 
1391 			truncate_pagecache_range(inode,
1392 				(loff_t)index << PAGE_SHIFT,
1393 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1394 
1395 			f2fs_lock_op(sbi);
1396 
1397 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1398 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1399 			if (ret) {
1400 				f2fs_unlock_op(sbi);
1401 				up_write(&F2FS_I(inode)->i_mmap_sem);
1402 				up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1403 				goto out;
1404 			}
1405 
1406 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1407 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1408 
1409 			ret = f2fs_do_zero_range(&dn, index, end);
1410 			f2fs_put_dnode(&dn);
1411 
1412 			f2fs_unlock_op(sbi);
1413 			up_write(&F2FS_I(inode)->i_mmap_sem);
1414 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1415 
1416 			f2fs_balance_fs(sbi, dn.node_changed);
1417 
1418 			if (ret)
1419 				goto out;
1420 
1421 			index = end;
1422 			new_size = max_t(loff_t, new_size,
1423 					(loff_t)index << PAGE_SHIFT);
1424 		}
1425 
1426 		if (off_end) {
1427 			ret = fill_zero(inode, pg_end, 0, off_end);
1428 			if (ret)
1429 				goto out;
1430 
1431 			new_size = max_t(loff_t, new_size, offset + len);
1432 		}
1433 	}
1434 
1435 out:
1436 	if (new_size > i_size_read(inode)) {
1437 		if (mode & FALLOC_FL_KEEP_SIZE)
1438 			file_set_keep_isize(inode);
1439 		else
1440 			f2fs_i_size_write(inode, new_size);
1441 	}
1442 	return ret;
1443 }
1444 
1445 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1446 {
1447 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1448 	pgoff_t nr, pg_start, pg_end, delta, idx;
1449 	loff_t new_size;
1450 	int ret = 0;
1451 
1452 	new_size = i_size_read(inode) + len;
1453 	ret = inode_newsize_ok(inode, new_size);
1454 	if (ret)
1455 		return ret;
1456 
1457 	if (offset >= i_size_read(inode))
1458 		return -EINVAL;
1459 
1460 	/* insert range should be aligned to block size of f2fs. */
1461 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1462 		return -EINVAL;
1463 
1464 	ret = f2fs_convert_inline_inode(inode);
1465 	if (ret)
1466 		return ret;
1467 
1468 	f2fs_balance_fs(sbi, true);
1469 
1470 	down_write(&F2FS_I(inode)->i_mmap_sem);
1471 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1472 	up_write(&F2FS_I(inode)->i_mmap_sem);
1473 	if (ret)
1474 		return ret;
1475 
1476 	/* write out all dirty pages from offset */
1477 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1478 	if (ret)
1479 		return ret;
1480 
1481 	pg_start = offset >> PAGE_SHIFT;
1482 	pg_end = (offset + len) >> PAGE_SHIFT;
1483 	delta = pg_end - pg_start;
1484 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1485 
1486 	/* avoid gc operation during block exchange */
1487 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1488 	down_write(&F2FS_I(inode)->i_mmap_sem);
1489 	truncate_pagecache(inode, offset);
1490 
1491 	while (!ret && idx > pg_start) {
1492 		nr = idx - pg_start;
1493 		if (nr > delta)
1494 			nr = delta;
1495 		idx -= nr;
1496 
1497 		f2fs_lock_op(sbi);
1498 		f2fs_drop_extent_tree(inode);
1499 
1500 		ret = __exchange_data_block(inode, inode, idx,
1501 					idx + delta, nr, false);
1502 		f2fs_unlock_op(sbi);
1503 	}
1504 	up_write(&F2FS_I(inode)->i_mmap_sem);
1505 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1506 
1507 	/* write out all moved pages, if possible */
1508 	down_write(&F2FS_I(inode)->i_mmap_sem);
1509 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1510 	truncate_pagecache(inode, offset);
1511 	up_write(&F2FS_I(inode)->i_mmap_sem);
1512 
1513 	if (!ret)
1514 		f2fs_i_size_write(inode, new_size);
1515 	return ret;
1516 }
1517 
1518 static int expand_inode_data(struct inode *inode, loff_t offset,
1519 					loff_t len, int mode)
1520 {
1521 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1522 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1523 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1524 			.m_may_create = true };
1525 	pgoff_t pg_end;
1526 	loff_t new_size = i_size_read(inode);
1527 	loff_t off_end;
1528 	int err;
1529 
1530 	err = inode_newsize_ok(inode, (len + offset));
1531 	if (err)
1532 		return err;
1533 
1534 	err = f2fs_convert_inline_inode(inode);
1535 	if (err)
1536 		return err;
1537 
1538 	f2fs_balance_fs(sbi, true);
1539 
1540 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1541 	off_end = (offset + len) & (PAGE_SIZE - 1);
1542 
1543 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1544 	map.m_len = pg_end - map.m_lblk;
1545 	if (off_end)
1546 		map.m_len++;
1547 
1548 	if (f2fs_is_pinned_file(inode))
1549 		map.m_seg_type = CURSEG_COLD_DATA;
1550 
1551 	err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
1552 						F2FS_GET_BLOCK_PRE_DIO :
1553 						F2FS_GET_BLOCK_PRE_AIO));
1554 	if (err) {
1555 		pgoff_t last_off;
1556 
1557 		if (!map.m_len)
1558 			return err;
1559 
1560 		last_off = map.m_lblk + map.m_len - 1;
1561 
1562 		/* update new size to the failed position */
1563 		new_size = (last_off == pg_end) ? offset + len :
1564 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1565 	} else {
1566 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1567 	}
1568 
1569 	if (new_size > i_size_read(inode)) {
1570 		if (mode & FALLOC_FL_KEEP_SIZE)
1571 			file_set_keep_isize(inode);
1572 		else
1573 			f2fs_i_size_write(inode, new_size);
1574 	}
1575 
1576 	return err;
1577 }
1578 
1579 static long f2fs_fallocate(struct file *file, int mode,
1580 				loff_t offset, loff_t len)
1581 {
1582 	struct inode *inode = file_inode(file);
1583 	long ret = 0;
1584 
1585 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1586 		return -EIO;
1587 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1588 		return -ENOSPC;
1589 
1590 	/* f2fs only support ->fallocate for regular file */
1591 	if (!S_ISREG(inode->i_mode))
1592 		return -EINVAL;
1593 
1594 	if (IS_ENCRYPTED(inode) &&
1595 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1596 		return -EOPNOTSUPP;
1597 
1598 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1599 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1600 			FALLOC_FL_INSERT_RANGE))
1601 		return -EOPNOTSUPP;
1602 
1603 	inode_lock(inode);
1604 
1605 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1606 		if (offset >= inode->i_size)
1607 			goto out;
1608 
1609 		ret = punch_hole(inode, offset, len);
1610 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1611 		ret = f2fs_collapse_range(inode, offset, len);
1612 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1613 		ret = f2fs_zero_range(inode, offset, len, mode);
1614 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1615 		ret = f2fs_insert_range(inode, offset, len);
1616 	} else {
1617 		ret = expand_inode_data(inode, offset, len, mode);
1618 	}
1619 
1620 	if (!ret) {
1621 		inode->i_mtime = inode->i_ctime = current_time(inode);
1622 		f2fs_mark_inode_dirty_sync(inode, false);
1623 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1624 	}
1625 
1626 out:
1627 	inode_unlock(inode);
1628 
1629 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1630 	return ret;
1631 }
1632 
1633 static int f2fs_release_file(struct inode *inode, struct file *filp)
1634 {
1635 	/*
1636 	 * f2fs_relase_file is called at every close calls. So we should
1637 	 * not drop any inmemory pages by close called by other process.
1638 	 */
1639 	if (!(filp->f_mode & FMODE_WRITE) ||
1640 			atomic_read(&inode->i_writecount) != 1)
1641 		return 0;
1642 
1643 	/* some remained atomic pages should discarded */
1644 	if (f2fs_is_atomic_file(inode))
1645 		f2fs_drop_inmem_pages(inode);
1646 	if (f2fs_is_volatile_file(inode)) {
1647 		set_inode_flag(inode, FI_DROP_CACHE);
1648 		filemap_fdatawrite(inode->i_mapping);
1649 		clear_inode_flag(inode, FI_DROP_CACHE);
1650 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1651 		stat_dec_volatile_write(inode);
1652 	}
1653 	return 0;
1654 }
1655 
1656 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1657 {
1658 	struct inode *inode = file_inode(file);
1659 
1660 	/*
1661 	 * If the process doing a transaction is crashed, we should do
1662 	 * roll-back. Otherwise, other reader/write can see corrupted database
1663 	 * until all the writers close its file. Since this should be done
1664 	 * before dropping file lock, it needs to do in ->flush.
1665 	 */
1666 	if (f2fs_is_atomic_file(inode) &&
1667 			F2FS_I(inode)->inmem_task == current)
1668 		f2fs_drop_inmem_pages(inode);
1669 	return 0;
1670 }
1671 
1672 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1673 {
1674 	struct f2fs_inode_info *fi = F2FS_I(inode);
1675 
1676 	/* Is it quota file? Do not allow user to mess with it */
1677 	if (IS_NOQUOTA(inode))
1678 		return -EPERM;
1679 
1680 	if ((iflags ^ fi->i_flags) & F2FS_CASEFOLD_FL) {
1681 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1682 			return -EOPNOTSUPP;
1683 		if (!f2fs_empty_dir(inode))
1684 			return -ENOTEMPTY;
1685 	}
1686 
1687 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1688 
1689 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1690 		set_inode_flag(inode, FI_PROJ_INHERIT);
1691 	else
1692 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1693 
1694 	inode->i_ctime = current_time(inode);
1695 	f2fs_set_inode_flags(inode);
1696 	f2fs_mark_inode_dirty_sync(inode, true);
1697 	return 0;
1698 }
1699 
1700 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1701 
1702 /*
1703  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1704  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1705  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1706  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1707  */
1708 
1709 static const struct {
1710 	u32 iflag;
1711 	u32 fsflag;
1712 } f2fs_fsflags_map[] = {
1713 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1714 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1715 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1716 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1717 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1718 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1719 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1720 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1721 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
1722 };
1723 
1724 #define F2FS_GETTABLE_FS_FL (		\
1725 		FS_SYNC_FL |		\
1726 		FS_IMMUTABLE_FL |	\
1727 		FS_APPEND_FL |		\
1728 		FS_NODUMP_FL |		\
1729 		FS_NOATIME_FL |		\
1730 		FS_INDEX_FL |		\
1731 		FS_DIRSYNC_FL |		\
1732 		FS_PROJINHERIT_FL |	\
1733 		FS_ENCRYPT_FL |		\
1734 		FS_INLINE_DATA_FL |	\
1735 		FS_NOCOW_FL |		\
1736 		FS_VERITY_FL |		\
1737 		FS_CASEFOLD_FL)
1738 
1739 #define F2FS_SETTABLE_FS_FL (		\
1740 		FS_SYNC_FL |		\
1741 		FS_IMMUTABLE_FL |	\
1742 		FS_APPEND_FL |		\
1743 		FS_NODUMP_FL |		\
1744 		FS_NOATIME_FL |		\
1745 		FS_DIRSYNC_FL |		\
1746 		FS_PROJINHERIT_FL |	\
1747 		FS_CASEFOLD_FL)
1748 
1749 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1750 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1751 {
1752 	u32 fsflags = 0;
1753 	int i;
1754 
1755 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1756 		if (iflags & f2fs_fsflags_map[i].iflag)
1757 			fsflags |= f2fs_fsflags_map[i].fsflag;
1758 
1759 	return fsflags;
1760 }
1761 
1762 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1763 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1764 {
1765 	u32 iflags = 0;
1766 	int i;
1767 
1768 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1769 		if (fsflags & f2fs_fsflags_map[i].fsflag)
1770 			iflags |= f2fs_fsflags_map[i].iflag;
1771 
1772 	return iflags;
1773 }
1774 
1775 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1776 {
1777 	struct inode *inode = file_inode(filp);
1778 	struct f2fs_inode_info *fi = F2FS_I(inode);
1779 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1780 
1781 	if (IS_ENCRYPTED(inode))
1782 		fsflags |= FS_ENCRYPT_FL;
1783 	if (IS_VERITY(inode))
1784 		fsflags |= FS_VERITY_FL;
1785 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1786 		fsflags |= FS_INLINE_DATA_FL;
1787 	if (is_inode_flag_set(inode, FI_PIN_FILE))
1788 		fsflags |= FS_NOCOW_FL;
1789 
1790 	fsflags &= F2FS_GETTABLE_FS_FL;
1791 
1792 	return put_user(fsflags, (int __user *)arg);
1793 }
1794 
1795 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1796 {
1797 	struct inode *inode = file_inode(filp);
1798 	struct f2fs_inode_info *fi = F2FS_I(inode);
1799 	u32 fsflags, old_fsflags;
1800 	u32 iflags;
1801 	int ret;
1802 
1803 	if (!inode_owner_or_capable(inode))
1804 		return -EACCES;
1805 
1806 	if (get_user(fsflags, (int __user *)arg))
1807 		return -EFAULT;
1808 
1809 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
1810 		return -EOPNOTSUPP;
1811 	fsflags &= F2FS_SETTABLE_FS_FL;
1812 
1813 	iflags = f2fs_fsflags_to_iflags(fsflags);
1814 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1815 		return -EOPNOTSUPP;
1816 
1817 	ret = mnt_want_write_file(filp);
1818 	if (ret)
1819 		return ret;
1820 
1821 	inode_lock(inode);
1822 
1823 	old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1824 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1825 	if (ret)
1826 		goto out;
1827 
1828 	ret = f2fs_setflags_common(inode, iflags,
1829 			f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1830 out:
1831 	inode_unlock(inode);
1832 	mnt_drop_write_file(filp);
1833 	return ret;
1834 }
1835 
1836 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1837 {
1838 	struct inode *inode = file_inode(filp);
1839 
1840 	return put_user(inode->i_generation, (int __user *)arg);
1841 }
1842 
1843 static int f2fs_ioc_start_atomic_write(struct file *filp)
1844 {
1845 	struct inode *inode = file_inode(filp);
1846 	struct f2fs_inode_info *fi = F2FS_I(inode);
1847 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1848 	int ret;
1849 
1850 	if (!inode_owner_or_capable(inode))
1851 		return -EACCES;
1852 
1853 	if (!S_ISREG(inode->i_mode))
1854 		return -EINVAL;
1855 
1856 	if (filp->f_flags & O_DIRECT)
1857 		return -EINVAL;
1858 
1859 	ret = mnt_want_write_file(filp);
1860 	if (ret)
1861 		return ret;
1862 
1863 	inode_lock(inode);
1864 
1865 	if (f2fs_is_atomic_file(inode)) {
1866 		if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1867 			ret = -EINVAL;
1868 		goto out;
1869 	}
1870 
1871 	ret = f2fs_convert_inline_inode(inode);
1872 	if (ret)
1873 		goto out;
1874 
1875 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1876 
1877 	/*
1878 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1879 	 * f2fs_is_atomic_file.
1880 	 */
1881 	if (get_dirty_pages(inode))
1882 		f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1883 			  inode->i_ino, get_dirty_pages(inode));
1884 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1885 	if (ret) {
1886 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1887 		goto out;
1888 	}
1889 
1890 	spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
1891 	if (list_empty(&fi->inmem_ilist))
1892 		list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
1893 	spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
1894 
1895 	/* add inode in inmem_list first and set atomic_file */
1896 	set_inode_flag(inode, FI_ATOMIC_FILE);
1897 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1898 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1899 
1900 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1901 	F2FS_I(inode)->inmem_task = current;
1902 	stat_inc_atomic_write(inode);
1903 	stat_update_max_atomic_write(inode);
1904 out:
1905 	inode_unlock(inode);
1906 	mnt_drop_write_file(filp);
1907 	return ret;
1908 }
1909 
1910 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1911 {
1912 	struct inode *inode = file_inode(filp);
1913 	int ret;
1914 
1915 	if (!inode_owner_or_capable(inode))
1916 		return -EACCES;
1917 
1918 	ret = mnt_want_write_file(filp);
1919 	if (ret)
1920 		return ret;
1921 
1922 	f2fs_balance_fs(F2FS_I_SB(inode), true);
1923 
1924 	inode_lock(inode);
1925 
1926 	if (f2fs_is_volatile_file(inode)) {
1927 		ret = -EINVAL;
1928 		goto err_out;
1929 	}
1930 
1931 	if (f2fs_is_atomic_file(inode)) {
1932 		ret = f2fs_commit_inmem_pages(inode);
1933 		if (ret)
1934 			goto err_out;
1935 
1936 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1937 		if (!ret)
1938 			f2fs_drop_inmem_pages(inode);
1939 	} else {
1940 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1941 	}
1942 err_out:
1943 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1944 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1945 		ret = -EINVAL;
1946 	}
1947 	inode_unlock(inode);
1948 	mnt_drop_write_file(filp);
1949 	return ret;
1950 }
1951 
1952 static int f2fs_ioc_start_volatile_write(struct file *filp)
1953 {
1954 	struct inode *inode = file_inode(filp);
1955 	int ret;
1956 
1957 	if (!inode_owner_or_capable(inode))
1958 		return -EACCES;
1959 
1960 	if (!S_ISREG(inode->i_mode))
1961 		return -EINVAL;
1962 
1963 	ret = mnt_want_write_file(filp);
1964 	if (ret)
1965 		return ret;
1966 
1967 	inode_lock(inode);
1968 
1969 	if (f2fs_is_volatile_file(inode))
1970 		goto out;
1971 
1972 	ret = f2fs_convert_inline_inode(inode);
1973 	if (ret)
1974 		goto out;
1975 
1976 	stat_inc_volatile_write(inode);
1977 	stat_update_max_volatile_write(inode);
1978 
1979 	set_inode_flag(inode, FI_VOLATILE_FILE);
1980 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1981 out:
1982 	inode_unlock(inode);
1983 	mnt_drop_write_file(filp);
1984 	return ret;
1985 }
1986 
1987 static int f2fs_ioc_release_volatile_write(struct file *filp)
1988 {
1989 	struct inode *inode = file_inode(filp);
1990 	int ret;
1991 
1992 	if (!inode_owner_or_capable(inode))
1993 		return -EACCES;
1994 
1995 	ret = mnt_want_write_file(filp);
1996 	if (ret)
1997 		return ret;
1998 
1999 	inode_lock(inode);
2000 
2001 	if (!f2fs_is_volatile_file(inode))
2002 		goto out;
2003 
2004 	if (!f2fs_is_first_block_written(inode)) {
2005 		ret = truncate_partial_data_page(inode, 0, true);
2006 		goto out;
2007 	}
2008 
2009 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2010 out:
2011 	inode_unlock(inode);
2012 	mnt_drop_write_file(filp);
2013 	return ret;
2014 }
2015 
2016 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2017 {
2018 	struct inode *inode = file_inode(filp);
2019 	int ret;
2020 
2021 	if (!inode_owner_or_capable(inode))
2022 		return -EACCES;
2023 
2024 	ret = mnt_want_write_file(filp);
2025 	if (ret)
2026 		return ret;
2027 
2028 	inode_lock(inode);
2029 
2030 	if (f2fs_is_atomic_file(inode))
2031 		f2fs_drop_inmem_pages(inode);
2032 	if (f2fs_is_volatile_file(inode)) {
2033 		clear_inode_flag(inode, FI_VOLATILE_FILE);
2034 		stat_dec_volatile_write(inode);
2035 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2036 	}
2037 
2038 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2039 
2040 	inode_unlock(inode);
2041 
2042 	mnt_drop_write_file(filp);
2043 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2044 	return ret;
2045 }
2046 
2047 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2048 {
2049 	struct inode *inode = file_inode(filp);
2050 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2051 	struct super_block *sb = sbi->sb;
2052 	__u32 in;
2053 	int ret = 0;
2054 
2055 	if (!capable(CAP_SYS_ADMIN))
2056 		return -EPERM;
2057 
2058 	if (get_user(in, (__u32 __user *)arg))
2059 		return -EFAULT;
2060 
2061 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2062 		ret = mnt_want_write_file(filp);
2063 		if (ret)
2064 			return ret;
2065 	}
2066 
2067 	switch (in) {
2068 	case F2FS_GOING_DOWN_FULLSYNC:
2069 		sb = freeze_bdev(sb->s_bdev);
2070 		if (IS_ERR(sb)) {
2071 			ret = PTR_ERR(sb);
2072 			goto out;
2073 		}
2074 		if (sb) {
2075 			f2fs_stop_checkpoint(sbi, false);
2076 			set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2077 			thaw_bdev(sb->s_bdev, sb);
2078 		}
2079 		break;
2080 	case F2FS_GOING_DOWN_METASYNC:
2081 		/* do checkpoint only */
2082 		ret = f2fs_sync_fs(sb, 1);
2083 		if (ret)
2084 			goto out;
2085 		f2fs_stop_checkpoint(sbi, false);
2086 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2087 		break;
2088 	case F2FS_GOING_DOWN_NOSYNC:
2089 		f2fs_stop_checkpoint(sbi, false);
2090 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2091 		break;
2092 	case F2FS_GOING_DOWN_METAFLUSH:
2093 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2094 		f2fs_stop_checkpoint(sbi, false);
2095 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2096 		break;
2097 	case F2FS_GOING_DOWN_NEED_FSCK:
2098 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2099 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2100 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2101 		/* do checkpoint only */
2102 		ret = f2fs_sync_fs(sb, 1);
2103 		goto out;
2104 	default:
2105 		ret = -EINVAL;
2106 		goto out;
2107 	}
2108 
2109 	f2fs_stop_gc_thread(sbi);
2110 	f2fs_stop_discard_thread(sbi);
2111 
2112 	f2fs_drop_discard_cmd(sbi);
2113 	clear_opt(sbi, DISCARD);
2114 
2115 	f2fs_update_time(sbi, REQ_TIME);
2116 out:
2117 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2118 		mnt_drop_write_file(filp);
2119 
2120 	trace_f2fs_shutdown(sbi, in, ret);
2121 
2122 	return ret;
2123 }
2124 
2125 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2126 {
2127 	struct inode *inode = file_inode(filp);
2128 	struct super_block *sb = inode->i_sb;
2129 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
2130 	struct fstrim_range range;
2131 	int ret;
2132 
2133 	if (!capable(CAP_SYS_ADMIN))
2134 		return -EPERM;
2135 
2136 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2137 		return -EOPNOTSUPP;
2138 
2139 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2140 				sizeof(range)))
2141 		return -EFAULT;
2142 
2143 	ret = mnt_want_write_file(filp);
2144 	if (ret)
2145 		return ret;
2146 
2147 	range.minlen = max((unsigned int)range.minlen,
2148 				q->limits.discard_granularity);
2149 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2150 	mnt_drop_write_file(filp);
2151 	if (ret < 0)
2152 		return ret;
2153 
2154 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2155 				sizeof(range)))
2156 		return -EFAULT;
2157 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2158 	return 0;
2159 }
2160 
2161 static bool uuid_is_nonzero(__u8 u[16])
2162 {
2163 	int i;
2164 
2165 	for (i = 0; i < 16; i++)
2166 		if (u[i])
2167 			return true;
2168 	return false;
2169 }
2170 
2171 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2172 {
2173 	struct inode *inode = file_inode(filp);
2174 
2175 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2176 		return -EOPNOTSUPP;
2177 
2178 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2179 
2180 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2181 }
2182 
2183 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2184 {
2185 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2186 		return -EOPNOTSUPP;
2187 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2188 }
2189 
2190 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2191 {
2192 	struct inode *inode = file_inode(filp);
2193 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2194 	int err;
2195 
2196 	if (!f2fs_sb_has_encrypt(sbi))
2197 		return -EOPNOTSUPP;
2198 
2199 	err = mnt_want_write_file(filp);
2200 	if (err)
2201 		return err;
2202 
2203 	down_write(&sbi->sb_lock);
2204 
2205 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2206 		goto got_it;
2207 
2208 	/* update superblock with uuid */
2209 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2210 
2211 	err = f2fs_commit_super(sbi, false);
2212 	if (err) {
2213 		/* undo new data */
2214 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2215 		goto out_err;
2216 	}
2217 got_it:
2218 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2219 									16))
2220 		err = -EFAULT;
2221 out_err:
2222 	up_write(&sbi->sb_lock);
2223 	mnt_drop_write_file(filp);
2224 	return err;
2225 }
2226 
2227 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2228 					     unsigned long arg)
2229 {
2230 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2231 		return -EOPNOTSUPP;
2232 
2233 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2234 }
2235 
2236 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2237 {
2238 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2239 		return -EOPNOTSUPP;
2240 
2241 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2242 }
2243 
2244 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2245 {
2246 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2247 		return -EOPNOTSUPP;
2248 
2249 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2250 }
2251 
2252 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2253 						    unsigned long arg)
2254 {
2255 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2256 		return -EOPNOTSUPP;
2257 
2258 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2259 }
2260 
2261 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2262 					      unsigned long arg)
2263 {
2264 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2265 		return -EOPNOTSUPP;
2266 
2267 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2268 }
2269 
2270 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2271 {
2272 	struct inode *inode = file_inode(filp);
2273 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2274 	__u32 sync;
2275 	int ret;
2276 
2277 	if (!capable(CAP_SYS_ADMIN))
2278 		return -EPERM;
2279 
2280 	if (get_user(sync, (__u32 __user *)arg))
2281 		return -EFAULT;
2282 
2283 	if (f2fs_readonly(sbi->sb))
2284 		return -EROFS;
2285 
2286 	ret = mnt_want_write_file(filp);
2287 	if (ret)
2288 		return ret;
2289 
2290 	if (!sync) {
2291 		if (!mutex_trylock(&sbi->gc_mutex)) {
2292 			ret = -EBUSY;
2293 			goto out;
2294 		}
2295 	} else {
2296 		mutex_lock(&sbi->gc_mutex);
2297 	}
2298 
2299 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2300 out:
2301 	mnt_drop_write_file(filp);
2302 	return ret;
2303 }
2304 
2305 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2306 {
2307 	struct inode *inode = file_inode(filp);
2308 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2309 	struct f2fs_gc_range range;
2310 	u64 end;
2311 	int ret;
2312 
2313 	if (!capable(CAP_SYS_ADMIN))
2314 		return -EPERM;
2315 
2316 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2317 							sizeof(range)))
2318 		return -EFAULT;
2319 
2320 	if (f2fs_readonly(sbi->sb))
2321 		return -EROFS;
2322 
2323 	end = range.start + range.len;
2324 	if (end < range.start || range.start < MAIN_BLKADDR(sbi) ||
2325 					end >= MAX_BLKADDR(sbi))
2326 		return -EINVAL;
2327 
2328 	ret = mnt_want_write_file(filp);
2329 	if (ret)
2330 		return ret;
2331 
2332 do_more:
2333 	if (!range.sync) {
2334 		if (!mutex_trylock(&sbi->gc_mutex)) {
2335 			ret = -EBUSY;
2336 			goto out;
2337 		}
2338 	} else {
2339 		mutex_lock(&sbi->gc_mutex);
2340 	}
2341 
2342 	ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2343 	range.start += BLKS_PER_SEC(sbi);
2344 	if (range.start <= end)
2345 		goto do_more;
2346 out:
2347 	mnt_drop_write_file(filp);
2348 	return ret;
2349 }
2350 
2351 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2352 {
2353 	struct inode *inode = file_inode(filp);
2354 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2355 	int ret;
2356 
2357 	if (!capable(CAP_SYS_ADMIN))
2358 		return -EPERM;
2359 
2360 	if (f2fs_readonly(sbi->sb))
2361 		return -EROFS;
2362 
2363 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2364 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2365 		return -EINVAL;
2366 	}
2367 
2368 	ret = mnt_want_write_file(filp);
2369 	if (ret)
2370 		return ret;
2371 
2372 	ret = f2fs_sync_fs(sbi->sb, 1);
2373 
2374 	mnt_drop_write_file(filp);
2375 	return ret;
2376 }
2377 
2378 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2379 					struct file *filp,
2380 					struct f2fs_defragment *range)
2381 {
2382 	struct inode *inode = file_inode(filp);
2383 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2384 					.m_seg_type = NO_CHECK_TYPE ,
2385 					.m_may_create = false };
2386 	struct extent_info ei = {0, 0, 0};
2387 	pgoff_t pg_start, pg_end, next_pgofs;
2388 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2389 	unsigned int total = 0, sec_num;
2390 	block_t blk_end = 0;
2391 	bool fragmented = false;
2392 	int err;
2393 
2394 	/* if in-place-update policy is enabled, don't waste time here */
2395 	if (f2fs_should_update_inplace(inode, NULL))
2396 		return -EINVAL;
2397 
2398 	pg_start = range->start >> PAGE_SHIFT;
2399 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2400 
2401 	f2fs_balance_fs(sbi, true);
2402 
2403 	inode_lock(inode);
2404 
2405 	/* writeback all dirty pages in the range */
2406 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2407 						range->start + range->len - 1);
2408 	if (err)
2409 		goto out;
2410 
2411 	/*
2412 	 * lookup mapping info in extent cache, skip defragmenting if physical
2413 	 * block addresses are continuous.
2414 	 */
2415 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2416 		if (ei.fofs + ei.len >= pg_end)
2417 			goto out;
2418 	}
2419 
2420 	map.m_lblk = pg_start;
2421 	map.m_next_pgofs = &next_pgofs;
2422 
2423 	/*
2424 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2425 	 * physical block addresses are continuous even if there are hole(s)
2426 	 * in logical blocks.
2427 	 */
2428 	while (map.m_lblk < pg_end) {
2429 		map.m_len = pg_end - map.m_lblk;
2430 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2431 		if (err)
2432 			goto out;
2433 
2434 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2435 			map.m_lblk = next_pgofs;
2436 			continue;
2437 		}
2438 
2439 		if (blk_end && blk_end != map.m_pblk)
2440 			fragmented = true;
2441 
2442 		/* record total count of block that we're going to move */
2443 		total += map.m_len;
2444 
2445 		blk_end = map.m_pblk + map.m_len;
2446 
2447 		map.m_lblk += map.m_len;
2448 	}
2449 
2450 	if (!fragmented) {
2451 		total = 0;
2452 		goto out;
2453 	}
2454 
2455 	sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2456 
2457 	/*
2458 	 * make sure there are enough free section for LFS allocation, this can
2459 	 * avoid defragment running in SSR mode when free section are allocated
2460 	 * intensively
2461 	 */
2462 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2463 		err = -EAGAIN;
2464 		goto out;
2465 	}
2466 
2467 	map.m_lblk = pg_start;
2468 	map.m_len = pg_end - pg_start;
2469 	total = 0;
2470 
2471 	while (map.m_lblk < pg_end) {
2472 		pgoff_t idx;
2473 		int cnt = 0;
2474 
2475 do_map:
2476 		map.m_len = pg_end - map.m_lblk;
2477 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2478 		if (err)
2479 			goto clear_out;
2480 
2481 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2482 			map.m_lblk = next_pgofs;
2483 			goto check;
2484 		}
2485 
2486 		set_inode_flag(inode, FI_DO_DEFRAG);
2487 
2488 		idx = map.m_lblk;
2489 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2490 			struct page *page;
2491 
2492 			page = f2fs_get_lock_data_page(inode, idx, true);
2493 			if (IS_ERR(page)) {
2494 				err = PTR_ERR(page);
2495 				goto clear_out;
2496 			}
2497 
2498 			set_page_dirty(page);
2499 			f2fs_put_page(page, 1);
2500 
2501 			idx++;
2502 			cnt++;
2503 			total++;
2504 		}
2505 
2506 		map.m_lblk = idx;
2507 check:
2508 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
2509 			goto do_map;
2510 
2511 		clear_inode_flag(inode, FI_DO_DEFRAG);
2512 
2513 		err = filemap_fdatawrite(inode->i_mapping);
2514 		if (err)
2515 			goto out;
2516 	}
2517 clear_out:
2518 	clear_inode_flag(inode, FI_DO_DEFRAG);
2519 out:
2520 	inode_unlock(inode);
2521 	if (!err)
2522 		range->len = (u64)total << PAGE_SHIFT;
2523 	return err;
2524 }
2525 
2526 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2527 {
2528 	struct inode *inode = file_inode(filp);
2529 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2530 	struct f2fs_defragment range;
2531 	int err;
2532 
2533 	if (!capable(CAP_SYS_ADMIN))
2534 		return -EPERM;
2535 
2536 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2537 		return -EINVAL;
2538 
2539 	if (f2fs_readonly(sbi->sb))
2540 		return -EROFS;
2541 
2542 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2543 							sizeof(range)))
2544 		return -EFAULT;
2545 
2546 	/* verify alignment of offset & size */
2547 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2548 		return -EINVAL;
2549 
2550 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2551 					sbi->max_file_blocks))
2552 		return -EINVAL;
2553 
2554 	err = mnt_want_write_file(filp);
2555 	if (err)
2556 		return err;
2557 
2558 	err = f2fs_defragment_range(sbi, filp, &range);
2559 	mnt_drop_write_file(filp);
2560 
2561 	f2fs_update_time(sbi, REQ_TIME);
2562 	if (err < 0)
2563 		return err;
2564 
2565 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2566 							sizeof(range)))
2567 		return -EFAULT;
2568 
2569 	return 0;
2570 }
2571 
2572 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2573 			struct file *file_out, loff_t pos_out, size_t len)
2574 {
2575 	struct inode *src = file_inode(file_in);
2576 	struct inode *dst = file_inode(file_out);
2577 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2578 	size_t olen = len, dst_max_i_size = 0;
2579 	size_t dst_osize;
2580 	int ret;
2581 
2582 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2583 				src->i_sb != dst->i_sb)
2584 		return -EXDEV;
2585 
2586 	if (unlikely(f2fs_readonly(src->i_sb)))
2587 		return -EROFS;
2588 
2589 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2590 		return -EINVAL;
2591 
2592 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2593 		return -EOPNOTSUPP;
2594 
2595 	if (src == dst) {
2596 		if (pos_in == pos_out)
2597 			return 0;
2598 		if (pos_out > pos_in && pos_out < pos_in + len)
2599 			return -EINVAL;
2600 	}
2601 
2602 	inode_lock(src);
2603 	if (src != dst) {
2604 		ret = -EBUSY;
2605 		if (!inode_trylock(dst))
2606 			goto out;
2607 	}
2608 
2609 	ret = -EINVAL;
2610 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2611 		goto out_unlock;
2612 	if (len == 0)
2613 		olen = len = src->i_size - pos_in;
2614 	if (pos_in + len == src->i_size)
2615 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2616 	if (len == 0) {
2617 		ret = 0;
2618 		goto out_unlock;
2619 	}
2620 
2621 	dst_osize = dst->i_size;
2622 	if (pos_out + olen > dst->i_size)
2623 		dst_max_i_size = pos_out + olen;
2624 
2625 	/* verify the end result is block aligned */
2626 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2627 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2628 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2629 		goto out_unlock;
2630 
2631 	ret = f2fs_convert_inline_inode(src);
2632 	if (ret)
2633 		goto out_unlock;
2634 
2635 	ret = f2fs_convert_inline_inode(dst);
2636 	if (ret)
2637 		goto out_unlock;
2638 
2639 	/* write out all dirty pages from offset */
2640 	ret = filemap_write_and_wait_range(src->i_mapping,
2641 					pos_in, pos_in + len);
2642 	if (ret)
2643 		goto out_unlock;
2644 
2645 	ret = filemap_write_and_wait_range(dst->i_mapping,
2646 					pos_out, pos_out + len);
2647 	if (ret)
2648 		goto out_unlock;
2649 
2650 	f2fs_balance_fs(sbi, true);
2651 
2652 	down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2653 	if (src != dst) {
2654 		ret = -EBUSY;
2655 		if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2656 			goto out_src;
2657 	}
2658 
2659 	f2fs_lock_op(sbi);
2660 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2661 				pos_out >> F2FS_BLKSIZE_BITS,
2662 				len >> F2FS_BLKSIZE_BITS, false);
2663 
2664 	if (!ret) {
2665 		if (dst_max_i_size)
2666 			f2fs_i_size_write(dst, dst_max_i_size);
2667 		else if (dst_osize != dst->i_size)
2668 			f2fs_i_size_write(dst, dst_osize);
2669 	}
2670 	f2fs_unlock_op(sbi);
2671 
2672 	if (src != dst)
2673 		up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2674 out_src:
2675 	up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2676 out_unlock:
2677 	if (src != dst)
2678 		inode_unlock(dst);
2679 out:
2680 	inode_unlock(src);
2681 	return ret;
2682 }
2683 
2684 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2685 {
2686 	struct f2fs_move_range range;
2687 	struct fd dst;
2688 	int err;
2689 
2690 	if (!(filp->f_mode & FMODE_READ) ||
2691 			!(filp->f_mode & FMODE_WRITE))
2692 		return -EBADF;
2693 
2694 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2695 							sizeof(range)))
2696 		return -EFAULT;
2697 
2698 	dst = fdget(range.dst_fd);
2699 	if (!dst.file)
2700 		return -EBADF;
2701 
2702 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2703 		err = -EBADF;
2704 		goto err_out;
2705 	}
2706 
2707 	err = mnt_want_write_file(filp);
2708 	if (err)
2709 		goto err_out;
2710 
2711 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2712 					range.pos_out, range.len);
2713 
2714 	mnt_drop_write_file(filp);
2715 	if (err)
2716 		goto err_out;
2717 
2718 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2719 						&range, sizeof(range)))
2720 		err = -EFAULT;
2721 err_out:
2722 	fdput(dst);
2723 	return err;
2724 }
2725 
2726 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2727 {
2728 	struct inode *inode = file_inode(filp);
2729 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2730 	struct sit_info *sm = SIT_I(sbi);
2731 	unsigned int start_segno = 0, end_segno = 0;
2732 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2733 	struct f2fs_flush_device range;
2734 	int ret;
2735 
2736 	if (!capable(CAP_SYS_ADMIN))
2737 		return -EPERM;
2738 
2739 	if (f2fs_readonly(sbi->sb))
2740 		return -EROFS;
2741 
2742 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2743 		return -EINVAL;
2744 
2745 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2746 							sizeof(range)))
2747 		return -EFAULT;
2748 
2749 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2750 			__is_large_section(sbi)) {
2751 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2752 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2753 		return -EINVAL;
2754 	}
2755 
2756 	ret = mnt_want_write_file(filp);
2757 	if (ret)
2758 		return ret;
2759 
2760 	if (range.dev_num != 0)
2761 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2762 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2763 
2764 	start_segno = sm->last_victim[FLUSH_DEVICE];
2765 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2766 		start_segno = dev_start_segno;
2767 	end_segno = min(start_segno + range.segments, dev_end_segno);
2768 
2769 	while (start_segno < end_segno) {
2770 		if (!mutex_trylock(&sbi->gc_mutex)) {
2771 			ret = -EBUSY;
2772 			goto out;
2773 		}
2774 		sm->last_victim[GC_CB] = end_segno + 1;
2775 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2776 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2777 		ret = f2fs_gc(sbi, true, true, start_segno);
2778 		if (ret == -EAGAIN)
2779 			ret = 0;
2780 		else if (ret < 0)
2781 			break;
2782 		start_segno++;
2783 	}
2784 out:
2785 	mnt_drop_write_file(filp);
2786 	return ret;
2787 }
2788 
2789 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2790 {
2791 	struct inode *inode = file_inode(filp);
2792 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2793 
2794 	/* Must validate to set it with SQLite behavior in Android. */
2795 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2796 
2797 	return put_user(sb_feature, (u32 __user *)arg);
2798 }
2799 
2800 #ifdef CONFIG_QUOTA
2801 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2802 {
2803 	struct dquot *transfer_to[MAXQUOTAS] = {};
2804 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2805 	struct super_block *sb = sbi->sb;
2806 	int err = 0;
2807 
2808 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2809 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2810 		err = __dquot_transfer(inode, transfer_to);
2811 		if (err)
2812 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2813 		dqput(transfer_to[PRJQUOTA]);
2814 	}
2815 	return err;
2816 }
2817 
2818 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2819 {
2820 	struct inode *inode = file_inode(filp);
2821 	struct f2fs_inode_info *fi = F2FS_I(inode);
2822 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2823 	struct page *ipage;
2824 	kprojid_t kprojid;
2825 	int err;
2826 
2827 	if (!f2fs_sb_has_project_quota(sbi)) {
2828 		if (projid != F2FS_DEF_PROJID)
2829 			return -EOPNOTSUPP;
2830 		else
2831 			return 0;
2832 	}
2833 
2834 	if (!f2fs_has_extra_attr(inode))
2835 		return -EOPNOTSUPP;
2836 
2837 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2838 
2839 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2840 		return 0;
2841 
2842 	err = -EPERM;
2843 	/* Is it quota file? Do not allow user to mess with it */
2844 	if (IS_NOQUOTA(inode))
2845 		return err;
2846 
2847 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
2848 	if (IS_ERR(ipage))
2849 		return PTR_ERR(ipage);
2850 
2851 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2852 								i_projid)) {
2853 		err = -EOVERFLOW;
2854 		f2fs_put_page(ipage, 1);
2855 		return err;
2856 	}
2857 	f2fs_put_page(ipage, 1);
2858 
2859 	err = dquot_initialize(inode);
2860 	if (err)
2861 		return err;
2862 
2863 	f2fs_lock_op(sbi);
2864 	err = f2fs_transfer_project_quota(inode, kprojid);
2865 	if (err)
2866 		goto out_unlock;
2867 
2868 	F2FS_I(inode)->i_projid = kprojid;
2869 	inode->i_ctime = current_time(inode);
2870 	f2fs_mark_inode_dirty_sync(inode, true);
2871 out_unlock:
2872 	f2fs_unlock_op(sbi);
2873 	return err;
2874 }
2875 #else
2876 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2877 {
2878 	return 0;
2879 }
2880 
2881 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2882 {
2883 	if (projid != F2FS_DEF_PROJID)
2884 		return -EOPNOTSUPP;
2885 	return 0;
2886 }
2887 #endif
2888 
2889 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
2890 
2891 /*
2892  * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
2893  * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
2894  * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
2895  */
2896 
2897 static const struct {
2898 	u32 iflag;
2899 	u32 xflag;
2900 } f2fs_xflags_map[] = {
2901 	{ F2FS_SYNC_FL,		FS_XFLAG_SYNC },
2902 	{ F2FS_IMMUTABLE_FL,	FS_XFLAG_IMMUTABLE },
2903 	{ F2FS_APPEND_FL,	FS_XFLAG_APPEND },
2904 	{ F2FS_NODUMP_FL,	FS_XFLAG_NODUMP },
2905 	{ F2FS_NOATIME_FL,	FS_XFLAG_NOATIME },
2906 	{ F2FS_PROJINHERIT_FL,	FS_XFLAG_PROJINHERIT },
2907 };
2908 
2909 #define F2FS_SUPPORTED_XFLAGS (		\
2910 		FS_XFLAG_SYNC |		\
2911 		FS_XFLAG_IMMUTABLE |	\
2912 		FS_XFLAG_APPEND |	\
2913 		FS_XFLAG_NODUMP |	\
2914 		FS_XFLAG_NOATIME |	\
2915 		FS_XFLAG_PROJINHERIT)
2916 
2917 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
2918 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
2919 {
2920 	u32 xflags = 0;
2921 	int i;
2922 
2923 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2924 		if (iflags & f2fs_xflags_map[i].iflag)
2925 			xflags |= f2fs_xflags_map[i].xflag;
2926 
2927 	return xflags;
2928 }
2929 
2930 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
2931 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
2932 {
2933 	u32 iflags = 0;
2934 	int i;
2935 
2936 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2937 		if (xflags & f2fs_xflags_map[i].xflag)
2938 			iflags |= f2fs_xflags_map[i].iflag;
2939 
2940 	return iflags;
2941 }
2942 
2943 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
2944 {
2945 	struct f2fs_inode_info *fi = F2FS_I(inode);
2946 
2947 	simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
2948 
2949 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2950 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2951 }
2952 
2953 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2954 {
2955 	struct inode *inode = file_inode(filp);
2956 	struct fsxattr fa;
2957 
2958 	f2fs_fill_fsxattr(inode, &fa);
2959 
2960 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2961 		return -EFAULT;
2962 	return 0;
2963 }
2964 
2965 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2966 {
2967 	struct inode *inode = file_inode(filp);
2968 	struct fsxattr fa, old_fa;
2969 	u32 iflags;
2970 	int err;
2971 
2972 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2973 		return -EFAULT;
2974 
2975 	/* Make sure caller has proper permission */
2976 	if (!inode_owner_or_capable(inode))
2977 		return -EACCES;
2978 
2979 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
2980 		return -EOPNOTSUPP;
2981 
2982 	iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2983 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2984 		return -EOPNOTSUPP;
2985 
2986 	err = mnt_want_write_file(filp);
2987 	if (err)
2988 		return err;
2989 
2990 	inode_lock(inode);
2991 
2992 	f2fs_fill_fsxattr(inode, &old_fa);
2993 	err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
2994 	if (err)
2995 		goto out;
2996 
2997 	err = f2fs_setflags_common(inode, iflags,
2998 			f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
2999 	if (err)
3000 		goto out;
3001 
3002 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3003 out:
3004 	inode_unlock(inode);
3005 	mnt_drop_write_file(filp);
3006 	return err;
3007 }
3008 
3009 int f2fs_pin_file_control(struct inode *inode, bool inc)
3010 {
3011 	struct f2fs_inode_info *fi = F2FS_I(inode);
3012 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3013 
3014 	/* Use i_gc_failures for normal file as a risk signal. */
3015 	if (inc)
3016 		f2fs_i_gc_failures_write(inode,
3017 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3018 
3019 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3020 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3021 			  __func__, inode->i_ino,
3022 			  fi->i_gc_failures[GC_FAILURE_PIN]);
3023 		clear_inode_flag(inode, FI_PIN_FILE);
3024 		return -EAGAIN;
3025 	}
3026 	return 0;
3027 }
3028 
3029 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3030 {
3031 	struct inode *inode = file_inode(filp);
3032 	__u32 pin;
3033 	int ret = 0;
3034 
3035 	if (get_user(pin, (__u32 __user *)arg))
3036 		return -EFAULT;
3037 
3038 	if (!S_ISREG(inode->i_mode))
3039 		return -EINVAL;
3040 
3041 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3042 		return -EROFS;
3043 
3044 	ret = mnt_want_write_file(filp);
3045 	if (ret)
3046 		return ret;
3047 
3048 	inode_lock(inode);
3049 
3050 	if (f2fs_should_update_outplace(inode, NULL)) {
3051 		ret = -EINVAL;
3052 		goto out;
3053 	}
3054 
3055 	if (!pin) {
3056 		clear_inode_flag(inode, FI_PIN_FILE);
3057 		f2fs_i_gc_failures_write(inode, 0);
3058 		goto done;
3059 	}
3060 
3061 	if (f2fs_pin_file_control(inode, false)) {
3062 		ret = -EAGAIN;
3063 		goto out;
3064 	}
3065 	ret = f2fs_convert_inline_inode(inode);
3066 	if (ret)
3067 		goto out;
3068 
3069 	set_inode_flag(inode, FI_PIN_FILE);
3070 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3071 done:
3072 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3073 out:
3074 	inode_unlock(inode);
3075 	mnt_drop_write_file(filp);
3076 	return ret;
3077 }
3078 
3079 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3080 {
3081 	struct inode *inode = file_inode(filp);
3082 	__u32 pin = 0;
3083 
3084 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3085 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3086 	return put_user(pin, (u32 __user *)arg);
3087 }
3088 
3089 int f2fs_precache_extents(struct inode *inode)
3090 {
3091 	struct f2fs_inode_info *fi = F2FS_I(inode);
3092 	struct f2fs_map_blocks map;
3093 	pgoff_t m_next_extent;
3094 	loff_t end;
3095 	int err;
3096 
3097 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3098 		return -EOPNOTSUPP;
3099 
3100 	map.m_lblk = 0;
3101 	map.m_next_pgofs = NULL;
3102 	map.m_next_extent = &m_next_extent;
3103 	map.m_seg_type = NO_CHECK_TYPE;
3104 	map.m_may_create = false;
3105 	end = F2FS_I_SB(inode)->max_file_blocks;
3106 
3107 	while (map.m_lblk < end) {
3108 		map.m_len = end - map.m_lblk;
3109 
3110 		down_write(&fi->i_gc_rwsem[WRITE]);
3111 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3112 		up_write(&fi->i_gc_rwsem[WRITE]);
3113 		if (err)
3114 			return err;
3115 
3116 		map.m_lblk = m_next_extent;
3117 	}
3118 
3119 	return err;
3120 }
3121 
3122 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3123 {
3124 	return f2fs_precache_extents(file_inode(filp));
3125 }
3126 
3127 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3128 {
3129 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3130 	__u64 block_count;
3131 	int ret;
3132 
3133 	if (!capable(CAP_SYS_ADMIN))
3134 		return -EPERM;
3135 
3136 	if (f2fs_readonly(sbi->sb))
3137 		return -EROFS;
3138 
3139 	if (copy_from_user(&block_count, (void __user *)arg,
3140 			   sizeof(block_count)))
3141 		return -EFAULT;
3142 
3143 	ret = f2fs_resize_fs(sbi, block_count);
3144 
3145 	return ret;
3146 }
3147 
3148 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3149 {
3150 	struct inode *inode = file_inode(filp);
3151 
3152 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3153 
3154 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3155 		f2fs_warn(F2FS_I_SB(inode),
3156 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3157 			  inode->i_ino);
3158 		return -EOPNOTSUPP;
3159 	}
3160 
3161 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3162 }
3163 
3164 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3165 {
3166 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3167 		return -EOPNOTSUPP;
3168 
3169 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3170 }
3171 
3172 static int f2fs_get_volume_name(struct file *filp, unsigned long arg)
3173 {
3174 	struct inode *inode = file_inode(filp);
3175 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3176 	char *vbuf;
3177 	int count;
3178 	int err = 0;
3179 
3180 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3181 	if (!vbuf)
3182 		return -ENOMEM;
3183 
3184 	down_read(&sbi->sb_lock);
3185 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3186 			ARRAY_SIZE(sbi->raw_super->volume_name),
3187 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3188 	up_read(&sbi->sb_lock);
3189 
3190 	if (copy_to_user((char __user *)arg, vbuf,
3191 				min(FSLABEL_MAX, count)))
3192 		err = -EFAULT;
3193 
3194 	kvfree(vbuf);
3195 	return err;
3196 }
3197 
3198 static int f2fs_set_volume_name(struct file *filp, unsigned long arg)
3199 {
3200 	struct inode *inode = file_inode(filp);
3201 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3202 	char *vbuf;
3203 	int err = 0;
3204 
3205 	if (!capable(CAP_SYS_ADMIN))
3206 		return -EPERM;
3207 
3208 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3209 	if (IS_ERR(vbuf))
3210 		return PTR_ERR(vbuf);
3211 
3212 	err = mnt_want_write_file(filp);
3213 	if (err)
3214 		goto out;
3215 
3216 	down_write(&sbi->sb_lock);
3217 
3218 	memset(sbi->raw_super->volume_name, 0,
3219 			sizeof(sbi->raw_super->volume_name));
3220 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3221 			sbi->raw_super->volume_name,
3222 			ARRAY_SIZE(sbi->raw_super->volume_name));
3223 
3224 	err = f2fs_commit_super(sbi, false);
3225 
3226 	up_write(&sbi->sb_lock);
3227 
3228 	mnt_drop_write_file(filp);
3229 out:
3230 	kfree(vbuf);
3231 	return err;
3232 }
3233 
3234 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3235 {
3236 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3237 		return -EIO;
3238 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
3239 		return -ENOSPC;
3240 
3241 	switch (cmd) {
3242 	case F2FS_IOC_GETFLAGS:
3243 		return f2fs_ioc_getflags(filp, arg);
3244 	case F2FS_IOC_SETFLAGS:
3245 		return f2fs_ioc_setflags(filp, arg);
3246 	case F2FS_IOC_GETVERSION:
3247 		return f2fs_ioc_getversion(filp, arg);
3248 	case F2FS_IOC_START_ATOMIC_WRITE:
3249 		return f2fs_ioc_start_atomic_write(filp);
3250 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3251 		return f2fs_ioc_commit_atomic_write(filp);
3252 	case F2FS_IOC_START_VOLATILE_WRITE:
3253 		return f2fs_ioc_start_volatile_write(filp);
3254 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3255 		return f2fs_ioc_release_volatile_write(filp);
3256 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3257 		return f2fs_ioc_abort_volatile_write(filp);
3258 	case F2FS_IOC_SHUTDOWN:
3259 		return f2fs_ioc_shutdown(filp, arg);
3260 	case FITRIM:
3261 		return f2fs_ioc_fitrim(filp, arg);
3262 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3263 		return f2fs_ioc_set_encryption_policy(filp, arg);
3264 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3265 		return f2fs_ioc_get_encryption_policy(filp, arg);
3266 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3267 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3268 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3269 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3270 	case FS_IOC_ADD_ENCRYPTION_KEY:
3271 		return f2fs_ioc_add_encryption_key(filp, arg);
3272 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3273 		return f2fs_ioc_remove_encryption_key(filp, arg);
3274 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3275 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3276 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3277 		return f2fs_ioc_get_encryption_key_status(filp, arg);
3278 	case F2FS_IOC_GARBAGE_COLLECT:
3279 		return f2fs_ioc_gc(filp, arg);
3280 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3281 		return f2fs_ioc_gc_range(filp, arg);
3282 	case F2FS_IOC_WRITE_CHECKPOINT:
3283 		return f2fs_ioc_write_checkpoint(filp, arg);
3284 	case F2FS_IOC_DEFRAGMENT:
3285 		return f2fs_ioc_defragment(filp, arg);
3286 	case F2FS_IOC_MOVE_RANGE:
3287 		return f2fs_ioc_move_range(filp, arg);
3288 	case F2FS_IOC_FLUSH_DEVICE:
3289 		return f2fs_ioc_flush_device(filp, arg);
3290 	case F2FS_IOC_GET_FEATURES:
3291 		return f2fs_ioc_get_features(filp, arg);
3292 	case F2FS_IOC_FSGETXATTR:
3293 		return f2fs_ioc_fsgetxattr(filp, arg);
3294 	case F2FS_IOC_FSSETXATTR:
3295 		return f2fs_ioc_fssetxattr(filp, arg);
3296 	case F2FS_IOC_GET_PIN_FILE:
3297 		return f2fs_ioc_get_pin_file(filp, arg);
3298 	case F2FS_IOC_SET_PIN_FILE:
3299 		return f2fs_ioc_set_pin_file(filp, arg);
3300 	case F2FS_IOC_PRECACHE_EXTENTS:
3301 		return f2fs_ioc_precache_extents(filp, arg);
3302 	case F2FS_IOC_RESIZE_FS:
3303 		return f2fs_ioc_resize_fs(filp, arg);
3304 	case FS_IOC_ENABLE_VERITY:
3305 		return f2fs_ioc_enable_verity(filp, arg);
3306 	case FS_IOC_MEASURE_VERITY:
3307 		return f2fs_ioc_measure_verity(filp, arg);
3308 	case F2FS_IOC_GET_VOLUME_NAME:
3309 		return f2fs_get_volume_name(filp, arg);
3310 	case F2FS_IOC_SET_VOLUME_NAME:
3311 		return f2fs_set_volume_name(filp, arg);
3312 	default:
3313 		return -ENOTTY;
3314 	}
3315 }
3316 
3317 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3318 {
3319 	struct file *file = iocb->ki_filp;
3320 	struct inode *inode = file_inode(file);
3321 	ssize_t ret;
3322 
3323 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3324 		ret = -EIO;
3325 		goto out;
3326 	}
3327 
3328 	if (iocb->ki_flags & IOCB_NOWAIT) {
3329 		if (!inode_trylock(inode)) {
3330 			ret = -EAGAIN;
3331 			goto out;
3332 		}
3333 	} else {
3334 		inode_lock(inode);
3335 	}
3336 
3337 	ret = generic_write_checks(iocb, from);
3338 	if (ret > 0) {
3339 		bool preallocated = false;
3340 		size_t target_size = 0;
3341 		int err;
3342 
3343 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3344 			set_inode_flag(inode, FI_NO_PREALLOC);
3345 
3346 		if ((iocb->ki_flags & IOCB_NOWAIT)) {
3347 			if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3348 						iov_iter_count(from)) ||
3349 				f2fs_has_inline_data(inode) ||
3350 				f2fs_force_buffered_io(inode, iocb, from)) {
3351 				clear_inode_flag(inode, FI_NO_PREALLOC);
3352 				inode_unlock(inode);
3353 				ret = -EAGAIN;
3354 				goto out;
3355 			}
3356 		} else {
3357 			preallocated = true;
3358 			target_size = iocb->ki_pos + iov_iter_count(from);
3359 
3360 			err = f2fs_preallocate_blocks(iocb, from);
3361 			if (err) {
3362 				clear_inode_flag(inode, FI_NO_PREALLOC);
3363 				inode_unlock(inode);
3364 				ret = err;
3365 				goto out;
3366 			}
3367 		}
3368 		ret = __generic_file_write_iter(iocb, from);
3369 		clear_inode_flag(inode, FI_NO_PREALLOC);
3370 
3371 		/* if we couldn't write data, we should deallocate blocks. */
3372 		if (preallocated && i_size_read(inode) < target_size)
3373 			f2fs_truncate(inode);
3374 
3375 		if (ret > 0)
3376 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3377 	}
3378 	inode_unlock(inode);
3379 out:
3380 	trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3381 					iov_iter_count(from), ret);
3382 	if (ret > 0)
3383 		ret = generic_write_sync(iocb, ret);
3384 	return ret;
3385 }
3386 
3387 #ifdef CONFIG_COMPAT
3388 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3389 {
3390 	switch (cmd) {
3391 	case F2FS_IOC32_GETFLAGS:
3392 		cmd = F2FS_IOC_GETFLAGS;
3393 		break;
3394 	case F2FS_IOC32_SETFLAGS:
3395 		cmd = F2FS_IOC_SETFLAGS;
3396 		break;
3397 	case F2FS_IOC32_GETVERSION:
3398 		cmd = F2FS_IOC_GETVERSION;
3399 		break;
3400 	case F2FS_IOC_START_ATOMIC_WRITE:
3401 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3402 	case F2FS_IOC_START_VOLATILE_WRITE:
3403 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3404 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3405 	case F2FS_IOC_SHUTDOWN:
3406 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3407 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3408 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3409 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3410 	case FS_IOC_ADD_ENCRYPTION_KEY:
3411 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3412 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3413 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3414 	case F2FS_IOC_GARBAGE_COLLECT:
3415 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3416 	case F2FS_IOC_WRITE_CHECKPOINT:
3417 	case F2FS_IOC_DEFRAGMENT:
3418 	case F2FS_IOC_MOVE_RANGE:
3419 	case F2FS_IOC_FLUSH_DEVICE:
3420 	case F2FS_IOC_GET_FEATURES:
3421 	case F2FS_IOC_FSGETXATTR:
3422 	case F2FS_IOC_FSSETXATTR:
3423 	case F2FS_IOC_GET_PIN_FILE:
3424 	case F2FS_IOC_SET_PIN_FILE:
3425 	case F2FS_IOC_PRECACHE_EXTENTS:
3426 	case F2FS_IOC_RESIZE_FS:
3427 	case FS_IOC_ENABLE_VERITY:
3428 	case FS_IOC_MEASURE_VERITY:
3429 	case F2FS_IOC_GET_VOLUME_NAME:
3430 	case F2FS_IOC_SET_VOLUME_NAME:
3431 		break;
3432 	default:
3433 		return -ENOIOCTLCMD;
3434 	}
3435 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3436 }
3437 #endif
3438 
3439 const struct file_operations f2fs_file_operations = {
3440 	.llseek		= f2fs_llseek,
3441 	.read_iter	= generic_file_read_iter,
3442 	.write_iter	= f2fs_file_write_iter,
3443 	.open		= f2fs_file_open,
3444 	.release	= f2fs_release_file,
3445 	.mmap		= f2fs_file_mmap,
3446 	.flush		= f2fs_file_flush,
3447 	.fsync		= f2fs_sync_file,
3448 	.fallocate	= f2fs_fallocate,
3449 	.unlocked_ioctl	= f2fs_ioctl,
3450 #ifdef CONFIG_COMPAT
3451 	.compat_ioctl	= f2fs_compat_ioctl,
3452 #endif
3453 	.splice_read	= generic_file_splice_read,
3454 	.splice_write	= iter_file_splice_write,
3455 };
3456