xref: /openbmc/linux/fs/f2fs/file.c (revision 305c8388)
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 	if (!ret)
44 		f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
45 							F2FS_BLKSIZE);
46 
47 	trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
48 
49 	return ret;
50 }
51 
52 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
53 {
54 	struct page *page = vmf->page;
55 	struct inode *inode = file_inode(vmf->vma->vm_file);
56 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
57 	struct dnode_of_data dn;
58 	bool need_alloc = true;
59 	int err = 0;
60 
61 	if (unlikely(f2fs_cp_error(sbi))) {
62 		err = -EIO;
63 		goto err;
64 	}
65 
66 	if (!f2fs_is_checkpoint_ready(sbi)) {
67 		err = -ENOSPC;
68 		goto err;
69 	}
70 
71 #ifdef CONFIG_F2FS_FS_COMPRESSION
72 	if (f2fs_compressed_file(inode)) {
73 		int ret = f2fs_is_compressed_cluster(inode, page->index);
74 
75 		if (ret < 0) {
76 			err = ret;
77 			goto err;
78 		} else if (ret) {
79 			if (ret < F2FS_I(inode)->i_cluster_size) {
80 				err = -EAGAIN;
81 				goto err;
82 			}
83 			need_alloc = false;
84 		}
85 	}
86 #endif
87 	/* should do out of any locked page */
88 	if (need_alloc)
89 		f2fs_balance_fs(sbi, true);
90 
91 	sb_start_pagefault(inode->i_sb);
92 
93 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
94 
95 	file_update_time(vmf->vma->vm_file);
96 	down_read(&F2FS_I(inode)->i_mmap_sem);
97 	lock_page(page);
98 	if (unlikely(page->mapping != inode->i_mapping ||
99 			page_offset(page) > i_size_read(inode) ||
100 			!PageUptodate(page))) {
101 		unlock_page(page);
102 		err = -EFAULT;
103 		goto out_sem;
104 	}
105 
106 	if (need_alloc) {
107 		/* block allocation */
108 		__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
109 		set_new_dnode(&dn, inode, NULL, NULL, 0);
110 		err = f2fs_get_block(&dn, page->index);
111 		f2fs_put_dnode(&dn);
112 		__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
113 	}
114 
115 #ifdef CONFIG_F2FS_FS_COMPRESSION
116 	if (!need_alloc) {
117 		set_new_dnode(&dn, inode, NULL, NULL, 0);
118 		err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
119 		f2fs_put_dnode(&dn);
120 	}
121 #endif
122 	if (err) {
123 		unlock_page(page);
124 		goto out_sem;
125 	}
126 
127 	f2fs_wait_on_page_writeback(page, DATA, false, true);
128 
129 	/* wait for GCed page writeback via META_MAPPING */
130 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
131 
132 	/*
133 	 * check to see if the page is mapped already (no holes)
134 	 */
135 	if (PageMappedToDisk(page))
136 		goto out_sem;
137 
138 	/* page is wholly or partially inside EOF */
139 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
140 						i_size_read(inode)) {
141 		loff_t offset;
142 
143 		offset = i_size_read(inode) & ~PAGE_MASK;
144 		zero_user_segment(page, offset, PAGE_SIZE);
145 	}
146 	set_page_dirty(page);
147 	if (!PageUptodate(page))
148 		SetPageUptodate(page);
149 
150 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
151 	f2fs_update_time(sbi, REQ_TIME);
152 
153 	trace_f2fs_vm_page_mkwrite(page, DATA);
154 out_sem:
155 	up_read(&F2FS_I(inode)->i_mmap_sem);
156 
157 	sb_end_pagefault(inode->i_sb);
158 err:
159 	return block_page_mkwrite_return(err);
160 }
161 
162 static const struct vm_operations_struct f2fs_file_vm_ops = {
163 	.fault		= f2fs_filemap_fault,
164 	.map_pages	= filemap_map_pages,
165 	.page_mkwrite	= f2fs_vm_page_mkwrite,
166 };
167 
168 static int get_parent_ino(struct inode *inode, nid_t *pino)
169 {
170 	struct dentry *dentry;
171 
172 	/*
173 	 * Make sure to get the non-deleted alias.  The alias associated with
174 	 * the open file descriptor being fsync()'ed may be deleted already.
175 	 */
176 	dentry = d_find_alias(inode);
177 	if (!dentry)
178 		return 0;
179 
180 	*pino = parent_ino(dentry);
181 	dput(dentry);
182 	return 1;
183 }
184 
185 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
186 {
187 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
188 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
189 
190 	if (!S_ISREG(inode->i_mode))
191 		cp_reason = CP_NON_REGULAR;
192 	else if (f2fs_compressed_file(inode))
193 		cp_reason = CP_COMPRESSED;
194 	else if (inode->i_nlink != 1)
195 		cp_reason = CP_HARDLINK;
196 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
197 		cp_reason = CP_SB_NEED_CP;
198 	else if (file_wrong_pino(inode))
199 		cp_reason = CP_WRONG_PINO;
200 	else if (!f2fs_space_for_roll_forward(sbi))
201 		cp_reason = CP_NO_SPC_ROLL;
202 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
203 		cp_reason = CP_NODE_NEED_CP;
204 	else if (test_opt(sbi, FASTBOOT))
205 		cp_reason = CP_FASTBOOT_MODE;
206 	else if (F2FS_OPTION(sbi).active_logs == 2)
207 		cp_reason = CP_SPEC_LOG_NUM;
208 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
209 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
210 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
211 							TRANS_DIR_INO))
212 		cp_reason = CP_RECOVER_DIR;
213 
214 	return cp_reason;
215 }
216 
217 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
218 {
219 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
220 	bool ret = false;
221 	/* But we need to avoid that there are some inode updates */
222 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
223 		ret = true;
224 	f2fs_put_page(i, 0);
225 	return ret;
226 }
227 
228 static void try_to_fix_pino(struct inode *inode)
229 {
230 	struct f2fs_inode_info *fi = F2FS_I(inode);
231 	nid_t pino;
232 
233 	down_write(&fi->i_sem);
234 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
235 			get_parent_ino(inode, &pino)) {
236 		f2fs_i_pino_write(inode, pino);
237 		file_got_pino(inode);
238 	}
239 	up_write(&fi->i_sem);
240 }
241 
242 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
243 						int datasync, bool atomic)
244 {
245 	struct inode *inode = file->f_mapping->host;
246 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
247 	nid_t ino = inode->i_ino;
248 	int ret = 0;
249 	enum cp_reason_type cp_reason = 0;
250 	struct writeback_control wbc = {
251 		.sync_mode = WB_SYNC_ALL,
252 		.nr_to_write = LONG_MAX,
253 		.for_reclaim = 0,
254 	};
255 	unsigned int seq_id = 0;
256 
257 	if (unlikely(f2fs_readonly(inode->i_sb) ||
258 				is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
259 		return 0;
260 
261 	trace_f2fs_sync_file_enter(inode);
262 
263 	if (S_ISDIR(inode->i_mode))
264 		goto go_write;
265 
266 	/* if fdatasync is triggered, let's do in-place-update */
267 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
268 		set_inode_flag(inode, FI_NEED_IPU);
269 	ret = file_write_and_wait_range(file, start, end);
270 	clear_inode_flag(inode, FI_NEED_IPU);
271 
272 	if (ret) {
273 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
274 		return ret;
275 	}
276 
277 	/* if the inode is dirty, let's recover all the time */
278 	if (!f2fs_skip_inode_update(inode, datasync)) {
279 		f2fs_write_inode(inode, NULL);
280 		goto go_write;
281 	}
282 
283 	/*
284 	 * if there is no written data, don't waste time to write recovery info.
285 	 */
286 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
287 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
288 
289 		/* it may call write_inode just prior to fsync */
290 		if (need_inode_page_update(sbi, ino))
291 			goto go_write;
292 
293 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
294 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
295 			goto flush_out;
296 		goto out;
297 	}
298 go_write:
299 	/*
300 	 * Both of fdatasync() and fsync() are able to be recovered from
301 	 * sudden-power-off.
302 	 */
303 	down_read(&F2FS_I(inode)->i_sem);
304 	cp_reason = need_do_checkpoint(inode);
305 	up_read(&F2FS_I(inode)->i_sem);
306 
307 	if (cp_reason) {
308 		/* all the dirty node pages should be flushed for POR */
309 		ret = f2fs_sync_fs(inode->i_sb, 1);
310 
311 		/*
312 		 * We've secured consistency through sync_fs. Following pino
313 		 * will be used only for fsynced inodes after checkpoint.
314 		 */
315 		try_to_fix_pino(inode);
316 		clear_inode_flag(inode, FI_APPEND_WRITE);
317 		clear_inode_flag(inode, FI_UPDATE_WRITE);
318 		goto out;
319 	}
320 sync_nodes:
321 	atomic_inc(&sbi->wb_sync_req[NODE]);
322 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
323 	atomic_dec(&sbi->wb_sync_req[NODE]);
324 	if (ret)
325 		goto out;
326 
327 	/* if cp_error was enabled, we should avoid infinite loop */
328 	if (unlikely(f2fs_cp_error(sbi))) {
329 		ret = -EIO;
330 		goto out;
331 	}
332 
333 	if (f2fs_need_inode_block_update(sbi, ino)) {
334 		f2fs_mark_inode_dirty_sync(inode, true);
335 		f2fs_write_inode(inode, NULL);
336 		goto sync_nodes;
337 	}
338 
339 	/*
340 	 * If it's atomic_write, it's just fine to keep write ordering. So
341 	 * here we don't need to wait for node write completion, since we use
342 	 * node chain which serializes node blocks. If one of node writes are
343 	 * reordered, we can see simply broken chain, resulting in stopping
344 	 * roll-forward recovery. It means we'll recover all or none node blocks
345 	 * given fsync mark.
346 	 */
347 	if (!atomic) {
348 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
349 		if (ret)
350 			goto out;
351 	}
352 
353 	/* once recovery info is written, don't need to tack this */
354 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
355 	clear_inode_flag(inode, FI_APPEND_WRITE);
356 flush_out:
357 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
358 		ret = f2fs_issue_flush(sbi, inode->i_ino);
359 	if (!ret) {
360 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
361 		clear_inode_flag(inode, FI_UPDATE_WRITE);
362 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
363 	}
364 	f2fs_update_time(sbi, REQ_TIME);
365 out:
366 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
367 	f2fs_trace_ios(NULL, 1);
368 	return ret;
369 }
370 
371 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
372 {
373 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
374 		return -EIO;
375 	return f2fs_do_sync_file(file, start, end, datasync, false);
376 }
377 
378 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
379 						pgoff_t pgofs, int whence)
380 {
381 	struct page *page;
382 	int nr_pages;
383 
384 	if (whence != SEEK_DATA)
385 		return 0;
386 
387 	/* find first dirty page index */
388 	nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
389 				      1, &page);
390 	if (!nr_pages)
391 		return ULONG_MAX;
392 	pgofs = page->index;
393 	put_page(page);
394 	return pgofs;
395 }
396 
397 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
398 				pgoff_t dirty, pgoff_t pgofs, int whence)
399 {
400 	switch (whence) {
401 	case SEEK_DATA:
402 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
403 			__is_valid_data_blkaddr(blkaddr))
404 			return true;
405 		break;
406 	case SEEK_HOLE:
407 		if (blkaddr == NULL_ADDR)
408 			return true;
409 		break;
410 	}
411 	return false;
412 }
413 
414 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
415 {
416 	struct inode *inode = file->f_mapping->host;
417 	loff_t maxbytes = inode->i_sb->s_maxbytes;
418 	struct dnode_of_data dn;
419 	pgoff_t pgofs, end_offset, dirty;
420 	loff_t data_ofs = offset;
421 	loff_t isize;
422 	int err = 0;
423 
424 	inode_lock(inode);
425 
426 	isize = i_size_read(inode);
427 	if (offset >= isize)
428 		goto fail;
429 
430 	/* handle inline data case */
431 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
432 		if (whence == SEEK_HOLE)
433 			data_ofs = isize;
434 		goto found;
435 	}
436 
437 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
438 
439 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
440 
441 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
442 		set_new_dnode(&dn, inode, NULL, NULL, 0);
443 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
444 		if (err && err != -ENOENT) {
445 			goto fail;
446 		} else if (err == -ENOENT) {
447 			/* direct node does not exists */
448 			if (whence == SEEK_DATA) {
449 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
450 				continue;
451 			} else {
452 				goto found;
453 			}
454 		}
455 
456 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
457 
458 		/* find data/hole in dnode block */
459 		for (; dn.ofs_in_node < end_offset;
460 				dn.ofs_in_node++, pgofs++,
461 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
462 			block_t blkaddr;
463 
464 			blkaddr = f2fs_data_blkaddr(&dn);
465 
466 			if (__is_valid_data_blkaddr(blkaddr) &&
467 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
468 					blkaddr, DATA_GENERIC_ENHANCE)) {
469 				f2fs_put_dnode(&dn);
470 				goto fail;
471 			}
472 
473 			if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
474 							pgofs, whence)) {
475 				f2fs_put_dnode(&dn);
476 				goto found;
477 			}
478 		}
479 		f2fs_put_dnode(&dn);
480 	}
481 
482 	if (whence == SEEK_DATA)
483 		goto fail;
484 found:
485 	if (whence == SEEK_HOLE && data_ofs > isize)
486 		data_ofs = isize;
487 	inode_unlock(inode);
488 	return vfs_setpos(file, data_ofs, maxbytes);
489 fail:
490 	inode_unlock(inode);
491 	return -ENXIO;
492 }
493 
494 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
495 {
496 	struct inode *inode = file->f_mapping->host;
497 	loff_t maxbytes = inode->i_sb->s_maxbytes;
498 
499 	switch (whence) {
500 	case SEEK_SET:
501 	case SEEK_CUR:
502 	case SEEK_END:
503 		return generic_file_llseek_size(file, offset, whence,
504 						maxbytes, i_size_read(inode));
505 	case SEEK_DATA:
506 	case SEEK_HOLE:
507 		if (offset < 0)
508 			return -ENXIO;
509 		return f2fs_seek_block(file, offset, whence);
510 	}
511 
512 	return -EINVAL;
513 }
514 
515 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
516 {
517 	struct inode *inode = file_inode(file);
518 	int err;
519 
520 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
521 		return -EIO;
522 
523 	if (!f2fs_is_compress_backend_ready(inode))
524 		return -EOPNOTSUPP;
525 
526 	/* we don't need to use inline_data strictly */
527 	err = f2fs_convert_inline_inode(inode);
528 	if (err)
529 		return err;
530 
531 	file_accessed(file);
532 	vma->vm_ops = &f2fs_file_vm_ops;
533 	set_inode_flag(inode, FI_MMAP_FILE);
534 	return 0;
535 }
536 
537 static int f2fs_file_open(struct inode *inode, struct file *filp)
538 {
539 	int err = fscrypt_file_open(inode, filp);
540 
541 	if (err)
542 		return err;
543 
544 	if (!f2fs_is_compress_backend_ready(inode))
545 		return -EOPNOTSUPP;
546 
547 	err = fsverity_file_open(inode, filp);
548 	if (err)
549 		return err;
550 
551 	filp->f_mode |= FMODE_NOWAIT;
552 
553 	return dquot_file_open(inode, filp);
554 }
555 
556 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
557 {
558 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
559 	struct f2fs_node *raw_node;
560 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
561 	__le32 *addr;
562 	int base = 0;
563 	bool compressed_cluster = false;
564 	int cluster_index = 0, valid_blocks = 0;
565 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
566 	bool released = !F2FS_I(dn->inode)->i_compr_blocks;
567 
568 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
569 		base = get_extra_isize(dn->inode);
570 
571 	raw_node = F2FS_NODE(dn->node_page);
572 	addr = blkaddr_in_node(raw_node) + base + ofs;
573 
574 	/* Assumption: truncateion starts with cluster */
575 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
576 		block_t blkaddr = le32_to_cpu(*addr);
577 
578 		if (f2fs_compressed_file(dn->inode) &&
579 					!(cluster_index & (cluster_size - 1))) {
580 			if (compressed_cluster)
581 				f2fs_i_compr_blocks_update(dn->inode,
582 							valid_blocks, false);
583 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
584 			valid_blocks = 0;
585 		}
586 
587 		if (blkaddr == NULL_ADDR)
588 			continue;
589 
590 		dn->data_blkaddr = NULL_ADDR;
591 		f2fs_set_data_blkaddr(dn);
592 
593 		if (__is_valid_data_blkaddr(blkaddr)) {
594 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
595 					DATA_GENERIC_ENHANCE))
596 				continue;
597 			if (compressed_cluster)
598 				valid_blocks++;
599 		}
600 
601 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
602 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
603 
604 		f2fs_invalidate_blocks(sbi, blkaddr);
605 
606 		if (!released || blkaddr != COMPRESS_ADDR)
607 			nr_free++;
608 	}
609 
610 	if (compressed_cluster)
611 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
612 
613 	if (nr_free) {
614 		pgoff_t fofs;
615 		/*
616 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
617 		 * we will invalidate all blkaddr in the whole range.
618 		 */
619 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
620 							dn->inode) + ofs;
621 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
622 		dec_valid_block_count(sbi, dn->inode, nr_free);
623 	}
624 	dn->ofs_in_node = ofs;
625 
626 	f2fs_update_time(sbi, REQ_TIME);
627 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
628 					 dn->ofs_in_node, nr_free);
629 }
630 
631 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
632 {
633 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
634 }
635 
636 static int truncate_partial_data_page(struct inode *inode, u64 from,
637 								bool cache_only)
638 {
639 	loff_t offset = from & (PAGE_SIZE - 1);
640 	pgoff_t index = from >> PAGE_SHIFT;
641 	struct address_space *mapping = inode->i_mapping;
642 	struct page *page;
643 
644 	if (!offset && !cache_only)
645 		return 0;
646 
647 	if (cache_only) {
648 		page = find_lock_page(mapping, index);
649 		if (page && PageUptodate(page))
650 			goto truncate_out;
651 		f2fs_put_page(page, 1);
652 		return 0;
653 	}
654 
655 	page = f2fs_get_lock_data_page(inode, index, true);
656 	if (IS_ERR(page))
657 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
658 truncate_out:
659 	f2fs_wait_on_page_writeback(page, DATA, true, true);
660 	zero_user(page, offset, PAGE_SIZE - offset);
661 
662 	/* An encrypted inode should have a key and truncate the last page. */
663 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
664 	if (!cache_only)
665 		set_page_dirty(page);
666 	f2fs_put_page(page, 1);
667 	return 0;
668 }
669 
670 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
671 {
672 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
673 	struct dnode_of_data dn;
674 	pgoff_t free_from;
675 	int count = 0, err = 0;
676 	struct page *ipage;
677 	bool truncate_page = false;
678 
679 	trace_f2fs_truncate_blocks_enter(inode, from);
680 
681 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
682 
683 	if (free_from >= sbi->max_file_blocks)
684 		goto free_partial;
685 
686 	if (lock)
687 		f2fs_lock_op(sbi);
688 
689 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
690 	if (IS_ERR(ipage)) {
691 		err = PTR_ERR(ipage);
692 		goto out;
693 	}
694 
695 	if (f2fs_has_inline_data(inode)) {
696 		f2fs_truncate_inline_inode(inode, ipage, from);
697 		f2fs_put_page(ipage, 1);
698 		truncate_page = true;
699 		goto out;
700 	}
701 
702 	set_new_dnode(&dn, inode, ipage, NULL, 0);
703 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
704 	if (err) {
705 		if (err == -ENOENT)
706 			goto free_next;
707 		goto out;
708 	}
709 
710 	count = ADDRS_PER_PAGE(dn.node_page, inode);
711 
712 	count -= dn.ofs_in_node;
713 	f2fs_bug_on(sbi, count < 0);
714 
715 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
716 		f2fs_truncate_data_blocks_range(&dn, count);
717 		free_from += count;
718 	}
719 
720 	f2fs_put_dnode(&dn);
721 free_next:
722 	err = f2fs_truncate_inode_blocks(inode, free_from);
723 out:
724 	if (lock)
725 		f2fs_unlock_op(sbi);
726 free_partial:
727 	/* lastly zero out the first data page */
728 	if (!err)
729 		err = truncate_partial_data_page(inode, from, truncate_page);
730 
731 	trace_f2fs_truncate_blocks_exit(inode, err);
732 	return err;
733 }
734 
735 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
736 {
737 	u64 free_from = from;
738 	int err;
739 
740 #ifdef CONFIG_F2FS_FS_COMPRESSION
741 	/*
742 	 * for compressed file, only support cluster size
743 	 * aligned truncation.
744 	 */
745 	if (f2fs_compressed_file(inode))
746 		free_from = round_up(from,
747 				F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
748 #endif
749 
750 	err = f2fs_do_truncate_blocks(inode, free_from, lock);
751 	if (err)
752 		return err;
753 
754 #ifdef CONFIG_F2FS_FS_COMPRESSION
755 	if (from != free_from)
756 		err = f2fs_truncate_partial_cluster(inode, from, lock);
757 #endif
758 
759 	return err;
760 }
761 
762 int f2fs_truncate(struct inode *inode)
763 {
764 	int err;
765 
766 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
767 		return -EIO;
768 
769 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
770 				S_ISLNK(inode->i_mode)))
771 		return 0;
772 
773 	trace_f2fs_truncate(inode);
774 
775 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
776 		f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
777 		return -EIO;
778 	}
779 
780 	/* we should check inline_data size */
781 	if (!f2fs_may_inline_data(inode)) {
782 		err = f2fs_convert_inline_inode(inode);
783 		if (err)
784 			return err;
785 	}
786 
787 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
788 	if (err)
789 		return err;
790 
791 	inode->i_mtime = inode->i_ctime = current_time(inode);
792 	f2fs_mark_inode_dirty_sync(inode, false);
793 	return 0;
794 }
795 
796 int f2fs_getattr(const struct path *path, struct kstat *stat,
797 		 u32 request_mask, unsigned int query_flags)
798 {
799 	struct inode *inode = d_inode(path->dentry);
800 	struct f2fs_inode_info *fi = F2FS_I(inode);
801 	struct f2fs_inode *ri;
802 	unsigned int flags;
803 
804 	if (f2fs_has_extra_attr(inode) &&
805 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
806 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
807 		stat->result_mask |= STATX_BTIME;
808 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
809 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
810 	}
811 
812 	flags = fi->i_flags;
813 	if (flags & F2FS_COMPR_FL)
814 		stat->attributes |= STATX_ATTR_COMPRESSED;
815 	if (flags & F2FS_APPEND_FL)
816 		stat->attributes |= STATX_ATTR_APPEND;
817 	if (IS_ENCRYPTED(inode))
818 		stat->attributes |= STATX_ATTR_ENCRYPTED;
819 	if (flags & F2FS_IMMUTABLE_FL)
820 		stat->attributes |= STATX_ATTR_IMMUTABLE;
821 	if (flags & F2FS_NODUMP_FL)
822 		stat->attributes |= STATX_ATTR_NODUMP;
823 	if (IS_VERITY(inode))
824 		stat->attributes |= STATX_ATTR_VERITY;
825 
826 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
827 				  STATX_ATTR_APPEND |
828 				  STATX_ATTR_ENCRYPTED |
829 				  STATX_ATTR_IMMUTABLE |
830 				  STATX_ATTR_NODUMP |
831 				  STATX_ATTR_VERITY);
832 
833 	generic_fillattr(inode, stat);
834 
835 	/* we need to show initial sectors used for inline_data/dentries */
836 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
837 					f2fs_has_inline_dentry(inode))
838 		stat->blocks += (stat->size + 511) >> 9;
839 
840 	return 0;
841 }
842 
843 #ifdef CONFIG_F2FS_FS_POSIX_ACL
844 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
845 {
846 	unsigned int ia_valid = attr->ia_valid;
847 
848 	if (ia_valid & ATTR_UID)
849 		inode->i_uid = attr->ia_uid;
850 	if (ia_valid & ATTR_GID)
851 		inode->i_gid = attr->ia_gid;
852 	if (ia_valid & ATTR_ATIME)
853 		inode->i_atime = attr->ia_atime;
854 	if (ia_valid & ATTR_MTIME)
855 		inode->i_mtime = attr->ia_mtime;
856 	if (ia_valid & ATTR_CTIME)
857 		inode->i_ctime = attr->ia_ctime;
858 	if (ia_valid & ATTR_MODE) {
859 		umode_t mode = attr->ia_mode;
860 
861 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
862 			mode &= ~S_ISGID;
863 		set_acl_inode(inode, mode);
864 	}
865 }
866 #else
867 #define __setattr_copy setattr_copy
868 #endif
869 
870 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
871 {
872 	struct inode *inode = d_inode(dentry);
873 	int err;
874 
875 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
876 		return -EIO;
877 
878 	if ((attr->ia_valid & ATTR_SIZE) &&
879 		!f2fs_is_compress_backend_ready(inode))
880 		return -EOPNOTSUPP;
881 
882 	err = setattr_prepare(dentry, attr);
883 	if (err)
884 		return err;
885 
886 	err = fscrypt_prepare_setattr(dentry, attr);
887 	if (err)
888 		return err;
889 
890 	err = fsverity_prepare_setattr(dentry, attr);
891 	if (err)
892 		return err;
893 
894 	if (is_quota_modification(inode, attr)) {
895 		err = dquot_initialize(inode);
896 		if (err)
897 			return err;
898 	}
899 	if ((attr->ia_valid & ATTR_UID &&
900 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
901 		(attr->ia_valid & ATTR_GID &&
902 		!gid_eq(attr->ia_gid, inode->i_gid))) {
903 		f2fs_lock_op(F2FS_I_SB(inode));
904 		err = dquot_transfer(inode, attr);
905 		if (err) {
906 			set_sbi_flag(F2FS_I_SB(inode),
907 					SBI_QUOTA_NEED_REPAIR);
908 			f2fs_unlock_op(F2FS_I_SB(inode));
909 			return err;
910 		}
911 		/*
912 		 * update uid/gid under lock_op(), so that dquot and inode can
913 		 * be updated atomically.
914 		 */
915 		if (attr->ia_valid & ATTR_UID)
916 			inode->i_uid = attr->ia_uid;
917 		if (attr->ia_valid & ATTR_GID)
918 			inode->i_gid = attr->ia_gid;
919 		f2fs_mark_inode_dirty_sync(inode, true);
920 		f2fs_unlock_op(F2FS_I_SB(inode));
921 	}
922 
923 	if (attr->ia_valid & ATTR_SIZE) {
924 		loff_t old_size = i_size_read(inode);
925 
926 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
927 			/*
928 			 * should convert inline inode before i_size_write to
929 			 * keep smaller than inline_data size with inline flag.
930 			 */
931 			err = f2fs_convert_inline_inode(inode);
932 			if (err)
933 				return err;
934 		}
935 
936 		down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
937 		down_write(&F2FS_I(inode)->i_mmap_sem);
938 
939 		truncate_setsize(inode, attr->ia_size);
940 
941 		if (attr->ia_size <= old_size)
942 			err = f2fs_truncate(inode);
943 		/*
944 		 * do not trim all blocks after i_size if target size is
945 		 * larger than i_size.
946 		 */
947 		up_write(&F2FS_I(inode)->i_mmap_sem);
948 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
949 		if (err)
950 			return err;
951 
952 		spin_lock(&F2FS_I(inode)->i_size_lock);
953 		inode->i_mtime = inode->i_ctime = current_time(inode);
954 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
955 		spin_unlock(&F2FS_I(inode)->i_size_lock);
956 	}
957 
958 	__setattr_copy(inode, attr);
959 
960 	if (attr->ia_valid & ATTR_MODE) {
961 		err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
962 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
963 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
964 			clear_inode_flag(inode, FI_ACL_MODE);
965 		}
966 	}
967 
968 	/* file size may changed here */
969 	f2fs_mark_inode_dirty_sync(inode, true);
970 
971 	/* inode change will produce dirty node pages flushed by checkpoint */
972 	f2fs_balance_fs(F2FS_I_SB(inode), true);
973 
974 	return err;
975 }
976 
977 const struct inode_operations f2fs_file_inode_operations = {
978 	.getattr	= f2fs_getattr,
979 	.setattr	= f2fs_setattr,
980 	.get_acl	= f2fs_get_acl,
981 	.set_acl	= f2fs_set_acl,
982 	.listxattr	= f2fs_listxattr,
983 	.fiemap		= f2fs_fiemap,
984 };
985 
986 static int fill_zero(struct inode *inode, pgoff_t index,
987 					loff_t start, loff_t len)
988 {
989 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
990 	struct page *page;
991 
992 	if (!len)
993 		return 0;
994 
995 	f2fs_balance_fs(sbi, true);
996 
997 	f2fs_lock_op(sbi);
998 	page = f2fs_get_new_data_page(inode, NULL, index, false);
999 	f2fs_unlock_op(sbi);
1000 
1001 	if (IS_ERR(page))
1002 		return PTR_ERR(page);
1003 
1004 	f2fs_wait_on_page_writeback(page, DATA, true, true);
1005 	zero_user(page, start, len);
1006 	set_page_dirty(page);
1007 	f2fs_put_page(page, 1);
1008 	return 0;
1009 }
1010 
1011 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1012 {
1013 	int err;
1014 
1015 	while (pg_start < pg_end) {
1016 		struct dnode_of_data dn;
1017 		pgoff_t end_offset, count;
1018 
1019 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1020 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1021 		if (err) {
1022 			if (err == -ENOENT) {
1023 				pg_start = f2fs_get_next_page_offset(&dn,
1024 								pg_start);
1025 				continue;
1026 			}
1027 			return err;
1028 		}
1029 
1030 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1031 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1032 
1033 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1034 
1035 		f2fs_truncate_data_blocks_range(&dn, count);
1036 		f2fs_put_dnode(&dn);
1037 
1038 		pg_start += count;
1039 	}
1040 	return 0;
1041 }
1042 
1043 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1044 {
1045 	pgoff_t pg_start, pg_end;
1046 	loff_t off_start, off_end;
1047 	int ret;
1048 
1049 	ret = f2fs_convert_inline_inode(inode);
1050 	if (ret)
1051 		return ret;
1052 
1053 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1054 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1055 
1056 	off_start = offset & (PAGE_SIZE - 1);
1057 	off_end = (offset + len) & (PAGE_SIZE - 1);
1058 
1059 	if (pg_start == pg_end) {
1060 		ret = fill_zero(inode, pg_start, off_start,
1061 						off_end - off_start);
1062 		if (ret)
1063 			return ret;
1064 	} else {
1065 		if (off_start) {
1066 			ret = fill_zero(inode, pg_start++, off_start,
1067 						PAGE_SIZE - off_start);
1068 			if (ret)
1069 				return ret;
1070 		}
1071 		if (off_end) {
1072 			ret = fill_zero(inode, pg_end, 0, off_end);
1073 			if (ret)
1074 				return ret;
1075 		}
1076 
1077 		if (pg_start < pg_end) {
1078 			struct address_space *mapping = inode->i_mapping;
1079 			loff_t blk_start, blk_end;
1080 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1081 
1082 			f2fs_balance_fs(sbi, true);
1083 
1084 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1085 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1086 
1087 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1088 			down_write(&F2FS_I(inode)->i_mmap_sem);
1089 
1090 			truncate_inode_pages_range(mapping, blk_start,
1091 					blk_end - 1);
1092 
1093 			f2fs_lock_op(sbi);
1094 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1095 			f2fs_unlock_op(sbi);
1096 
1097 			up_write(&F2FS_I(inode)->i_mmap_sem);
1098 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1099 		}
1100 	}
1101 
1102 	return ret;
1103 }
1104 
1105 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1106 				int *do_replace, pgoff_t off, pgoff_t len)
1107 {
1108 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1109 	struct dnode_of_data dn;
1110 	int ret, done, i;
1111 
1112 next_dnode:
1113 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1114 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1115 	if (ret && ret != -ENOENT) {
1116 		return ret;
1117 	} else if (ret == -ENOENT) {
1118 		if (dn.max_level == 0)
1119 			return -ENOENT;
1120 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1121 						dn.ofs_in_node, len);
1122 		blkaddr += done;
1123 		do_replace += done;
1124 		goto next;
1125 	}
1126 
1127 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1128 							dn.ofs_in_node, len);
1129 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1130 		*blkaddr = f2fs_data_blkaddr(&dn);
1131 
1132 		if (__is_valid_data_blkaddr(*blkaddr) &&
1133 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1134 					DATA_GENERIC_ENHANCE)) {
1135 			f2fs_put_dnode(&dn);
1136 			return -EFSCORRUPTED;
1137 		}
1138 
1139 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1140 
1141 			if (f2fs_lfs_mode(sbi)) {
1142 				f2fs_put_dnode(&dn);
1143 				return -EOPNOTSUPP;
1144 			}
1145 
1146 			/* do not invalidate this block address */
1147 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1148 			*do_replace = 1;
1149 		}
1150 	}
1151 	f2fs_put_dnode(&dn);
1152 next:
1153 	len -= done;
1154 	off += done;
1155 	if (len)
1156 		goto next_dnode;
1157 	return 0;
1158 }
1159 
1160 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1161 				int *do_replace, pgoff_t off, int len)
1162 {
1163 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1164 	struct dnode_of_data dn;
1165 	int ret, i;
1166 
1167 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1168 		if (*do_replace == 0)
1169 			continue;
1170 
1171 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1172 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1173 		if (ret) {
1174 			dec_valid_block_count(sbi, inode, 1);
1175 			f2fs_invalidate_blocks(sbi, *blkaddr);
1176 		} else {
1177 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1178 		}
1179 		f2fs_put_dnode(&dn);
1180 	}
1181 	return 0;
1182 }
1183 
1184 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1185 			block_t *blkaddr, int *do_replace,
1186 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1187 {
1188 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1189 	pgoff_t i = 0;
1190 	int ret;
1191 
1192 	while (i < len) {
1193 		if (blkaddr[i] == NULL_ADDR && !full) {
1194 			i++;
1195 			continue;
1196 		}
1197 
1198 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1199 			struct dnode_of_data dn;
1200 			struct node_info ni;
1201 			size_t new_size;
1202 			pgoff_t ilen;
1203 
1204 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1205 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1206 			if (ret)
1207 				return ret;
1208 
1209 			ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1210 			if (ret) {
1211 				f2fs_put_dnode(&dn);
1212 				return ret;
1213 			}
1214 
1215 			ilen = min((pgoff_t)
1216 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1217 						dn.ofs_in_node, len - i);
1218 			do {
1219 				dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1220 				f2fs_truncate_data_blocks_range(&dn, 1);
1221 
1222 				if (do_replace[i]) {
1223 					f2fs_i_blocks_write(src_inode,
1224 							1, false, false);
1225 					f2fs_i_blocks_write(dst_inode,
1226 							1, true, false);
1227 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1228 					blkaddr[i], ni.version, true, false);
1229 
1230 					do_replace[i] = 0;
1231 				}
1232 				dn.ofs_in_node++;
1233 				i++;
1234 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1235 				if (dst_inode->i_size < new_size)
1236 					f2fs_i_size_write(dst_inode, new_size);
1237 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1238 
1239 			f2fs_put_dnode(&dn);
1240 		} else {
1241 			struct page *psrc, *pdst;
1242 
1243 			psrc = f2fs_get_lock_data_page(src_inode,
1244 							src + i, true);
1245 			if (IS_ERR(psrc))
1246 				return PTR_ERR(psrc);
1247 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1248 								true);
1249 			if (IS_ERR(pdst)) {
1250 				f2fs_put_page(psrc, 1);
1251 				return PTR_ERR(pdst);
1252 			}
1253 			f2fs_copy_page(psrc, pdst);
1254 			set_page_dirty(pdst);
1255 			f2fs_put_page(pdst, 1);
1256 			f2fs_put_page(psrc, 1);
1257 
1258 			ret = f2fs_truncate_hole(src_inode,
1259 						src + i, src + i + 1);
1260 			if (ret)
1261 				return ret;
1262 			i++;
1263 		}
1264 	}
1265 	return 0;
1266 }
1267 
1268 static int __exchange_data_block(struct inode *src_inode,
1269 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1270 			pgoff_t len, bool full)
1271 {
1272 	block_t *src_blkaddr;
1273 	int *do_replace;
1274 	pgoff_t olen;
1275 	int ret;
1276 
1277 	while (len) {
1278 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1279 
1280 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1281 					array_size(olen, sizeof(block_t)),
1282 					GFP_NOFS);
1283 		if (!src_blkaddr)
1284 			return -ENOMEM;
1285 
1286 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1287 					array_size(olen, sizeof(int)),
1288 					GFP_NOFS);
1289 		if (!do_replace) {
1290 			kvfree(src_blkaddr);
1291 			return -ENOMEM;
1292 		}
1293 
1294 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1295 					do_replace, src, olen);
1296 		if (ret)
1297 			goto roll_back;
1298 
1299 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1300 					do_replace, src, dst, olen, full);
1301 		if (ret)
1302 			goto roll_back;
1303 
1304 		src += olen;
1305 		dst += olen;
1306 		len -= olen;
1307 
1308 		kvfree(src_blkaddr);
1309 		kvfree(do_replace);
1310 	}
1311 	return 0;
1312 
1313 roll_back:
1314 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1315 	kvfree(src_blkaddr);
1316 	kvfree(do_replace);
1317 	return ret;
1318 }
1319 
1320 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1321 {
1322 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1323 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1324 	pgoff_t start = offset >> PAGE_SHIFT;
1325 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1326 	int ret;
1327 
1328 	f2fs_balance_fs(sbi, true);
1329 
1330 	/* avoid gc operation during block exchange */
1331 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1332 	down_write(&F2FS_I(inode)->i_mmap_sem);
1333 
1334 	f2fs_lock_op(sbi);
1335 	f2fs_drop_extent_tree(inode);
1336 	truncate_pagecache(inode, offset);
1337 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1338 	f2fs_unlock_op(sbi);
1339 
1340 	up_write(&F2FS_I(inode)->i_mmap_sem);
1341 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1342 	return ret;
1343 }
1344 
1345 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1346 {
1347 	loff_t new_size;
1348 	int ret;
1349 
1350 	if (offset + len >= i_size_read(inode))
1351 		return -EINVAL;
1352 
1353 	/* collapse range should be aligned to block size of f2fs. */
1354 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1355 		return -EINVAL;
1356 
1357 	ret = f2fs_convert_inline_inode(inode);
1358 	if (ret)
1359 		return ret;
1360 
1361 	/* write out all dirty pages from offset */
1362 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1363 	if (ret)
1364 		return ret;
1365 
1366 	ret = f2fs_do_collapse(inode, offset, len);
1367 	if (ret)
1368 		return ret;
1369 
1370 	/* write out all moved pages, if possible */
1371 	down_write(&F2FS_I(inode)->i_mmap_sem);
1372 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1373 	truncate_pagecache(inode, offset);
1374 
1375 	new_size = i_size_read(inode) - len;
1376 	truncate_pagecache(inode, new_size);
1377 
1378 	ret = f2fs_truncate_blocks(inode, new_size, true);
1379 	up_write(&F2FS_I(inode)->i_mmap_sem);
1380 	if (!ret)
1381 		f2fs_i_size_write(inode, new_size);
1382 	return ret;
1383 }
1384 
1385 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1386 								pgoff_t end)
1387 {
1388 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1389 	pgoff_t index = start;
1390 	unsigned int ofs_in_node = dn->ofs_in_node;
1391 	blkcnt_t count = 0;
1392 	int ret;
1393 
1394 	for (; index < end; index++, dn->ofs_in_node++) {
1395 		if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1396 			count++;
1397 	}
1398 
1399 	dn->ofs_in_node = ofs_in_node;
1400 	ret = f2fs_reserve_new_blocks(dn, count);
1401 	if (ret)
1402 		return ret;
1403 
1404 	dn->ofs_in_node = ofs_in_node;
1405 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1406 		dn->data_blkaddr = f2fs_data_blkaddr(dn);
1407 		/*
1408 		 * f2fs_reserve_new_blocks will not guarantee entire block
1409 		 * allocation.
1410 		 */
1411 		if (dn->data_blkaddr == NULL_ADDR) {
1412 			ret = -ENOSPC;
1413 			break;
1414 		}
1415 		if (dn->data_blkaddr != NEW_ADDR) {
1416 			f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1417 			dn->data_blkaddr = NEW_ADDR;
1418 			f2fs_set_data_blkaddr(dn);
1419 		}
1420 	}
1421 
1422 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1423 
1424 	return ret;
1425 }
1426 
1427 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1428 								int mode)
1429 {
1430 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1431 	struct address_space *mapping = inode->i_mapping;
1432 	pgoff_t index, pg_start, pg_end;
1433 	loff_t new_size = i_size_read(inode);
1434 	loff_t off_start, off_end;
1435 	int ret = 0;
1436 
1437 	ret = inode_newsize_ok(inode, (len + offset));
1438 	if (ret)
1439 		return ret;
1440 
1441 	ret = f2fs_convert_inline_inode(inode);
1442 	if (ret)
1443 		return ret;
1444 
1445 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1446 	if (ret)
1447 		return ret;
1448 
1449 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1450 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1451 
1452 	off_start = offset & (PAGE_SIZE - 1);
1453 	off_end = (offset + len) & (PAGE_SIZE - 1);
1454 
1455 	if (pg_start == pg_end) {
1456 		ret = fill_zero(inode, pg_start, off_start,
1457 						off_end - off_start);
1458 		if (ret)
1459 			return ret;
1460 
1461 		new_size = max_t(loff_t, new_size, offset + len);
1462 	} else {
1463 		if (off_start) {
1464 			ret = fill_zero(inode, pg_start++, off_start,
1465 						PAGE_SIZE - off_start);
1466 			if (ret)
1467 				return ret;
1468 
1469 			new_size = max_t(loff_t, new_size,
1470 					(loff_t)pg_start << PAGE_SHIFT);
1471 		}
1472 
1473 		for (index = pg_start; index < pg_end;) {
1474 			struct dnode_of_data dn;
1475 			unsigned int end_offset;
1476 			pgoff_t end;
1477 
1478 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1479 			down_write(&F2FS_I(inode)->i_mmap_sem);
1480 
1481 			truncate_pagecache_range(inode,
1482 				(loff_t)index << PAGE_SHIFT,
1483 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1484 
1485 			f2fs_lock_op(sbi);
1486 
1487 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1488 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1489 			if (ret) {
1490 				f2fs_unlock_op(sbi);
1491 				up_write(&F2FS_I(inode)->i_mmap_sem);
1492 				up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1493 				goto out;
1494 			}
1495 
1496 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1497 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1498 
1499 			ret = f2fs_do_zero_range(&dn, index, end);
1500 			f2fs_put_dnode(&dn);
1501 
1502 			f2fs_unlock_op(sbi);
1503 			up_write(&F2FS_I(inode)->i_mmap_sem);
1504 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1505 
1506 			f2fs_balance_fs(sbi, dn.node_changed);
1507 
1508 			if (ret)
1509 				goto out;
1510 
1511 			index = end;
1512 			new_size = max_t(loff_t, new_size,
1513 					(loff_t)index << PAGE_SHIFT);
1514 		}
1515 
1516 		if (off_end) {
1517 			ret = fill_zero(inode, pg_end, 0, off_end);
1518 			if (ret)
1519 				goto out;
1520 
1521 			new_size = max_t(loff_t, new_size, offset + len);
1522 		}
1523 	}
1524 
1525 out:
1526 	if (new_size > i_size_read(inode)) {
1527 		if (mode & FALLOC_FL_KEEP_SIZE)
1528 			file_set_keep_isize(inode);
1529 		else
1530 			f2fs_i_size_write(inode, new_size);
1531 	}
1532 	return ret;
1533 }
1534 
1535 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1536 {
1537 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1538 	pgoff_t nr, pg_start, pg_end, delta, idx;
1539 	loff_t new_size;
1540 	int ret = 0;
1541 
1542 	new_size = i_size_read(inode) + len;
1543 	ret = inode_newsize_ok(inode, new_size);
1544 	if (ret)
1545 		return ret;
1546 
1547 	if (offset >= i_size_read(inode))
1548 		return -EINVAL;
1549 
1550 	/* insert range should be aligned to block size of f2fs. */
1551 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1552 		return -EINVAL;
1553 
1554 	ret = f2fs_convert_inline_inode(inode);
1555 	if (ret)
1556 		return ret;
1557 
1558 	f2fs_balance_fs(sbi, true);
1559 
1560 	down_write(&F2FS_I(inode)->i_mmap_sem);
1561 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1562 	up_write(&F2FS_I(inode)->i_mmap_sem);
1563 	if (ret)
1564 		return ret;
1565 
1566 	/* write out all dirty pages from offset */
1567 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1568 	if (ret)
1569 		return ret;
1570 
1571 	pg_start = offset >> PAGE_SHIFT;
1572 	pg_end = (offset + len) >> PAGE_SHIFT;
1573 	delta = pg_end - pg_start;
1574 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1575 
1576 	/* avoid gc operation during block exchange */
1577 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1578 	down_write(&F2FS_I(inode)->i_mmap_sem);
1579 	truncate_pagecache(inode, offset);
1580 
1581 	while (!ret && idx > pg_start) {
1582 		nr = idx - pg_start;
1583 		if (nr > delta)
1584 			nr = delta;
1585 		idx -= nr;
1586 
1587 		f2fs_lock_op(sbi);
1588 		f2fs_drop_extent_tree(inode);
1589 
1590 		ret = __exchange_data_block(inode, inode, idx,
1591 					idx + delta, nr, false);
1592 		f2fs_unlock_op(sbi);
1593 	}
1594 	up_write(&F2FS_I(inode)->i_mmap_sem);
1595 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1596 
1597 	/* write out all moved pages, if possible */
1598 	down_write(&F2FS_I(inode)->i_mmap_sem);
1599 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1600 	truncate_pagecache(inode, offset);
1601 	up_write(&F2FS_I(inode)->i_mmap_sem);
1602 
1603 	if (!ret)
1604 		f2fs_i_size_write(inode, new_size);
1605 	return ret;
1606 }
1607 
1608 static int expand_inode_data(struct inode *inode, loff_t offset,
1609 					loff_t len, int mode)
1610 {
1611 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1612 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1613 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1614 			.m_may_create = true };
1615 	pgoff_t pg_end;
1616 	loff_t new_size = i_size_read(inode);
1617 	loff_t off_end;
1618 	int err;
1619 
1620 	err = inode_newsize_ok(inode, (len + offset));
1621 	if (err)
1622 		return err;
1623 
1624 	err = f2fs_convert_inline_inode(inode);
1625 	if (err)
1626 		return err;
1627 
1628 	f2fs_balance_fs(sbi, true);
1629 
1630 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1631 	off_end = (offset + len) & (PAGE_SIZE - 1);
1632 
1633 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1634 	map.m_len = pg_end - map.m_lblk;
1635 	if (off_end)
1636 		map.m_len++;
1637 
1638 	if (!map.m_len)
1639 		return 0;
1640 
1641 	if (f2fs_is_pinned_file(inode)) {
1642 		block_t len = (map.m_len >> sbi->log_blocks_per_seg) <<
1643 					sbi->log_blocks_per_seg;
1644 		block_t done = 0;
1645 
1646 		if (map.m_len % sbi->blocks_per_seg)
1647 			len += sbi->blocks_per_seg;
1648 
1649 		map.m_len = sbi->blocks_per_seg;
1650 next_alloc:
1651 		if (has_not_enough_free_secs(sbi, 0,
1652 			GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1653 			down_write(&sbi->gc_lock);
1654 			err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1655 			if (err && err != -ENODATA && err != -EAGAIN)
1656 				goto out_err;
1657 		}
1658 
1659 		down_write(&sbi->pin_sem);
1660 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1661 
1662 		f2fs_lock_op(sbi);
1663 		f2fs_allocate_new_segments(sbi, CURSEG_COLD_DATA);
1664 		f2fs_unlock_op(sbi);
1665 
1666 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1667 		up_write(&sbi->pin_sem);
1668 
1669 		done += map.m_len;
1670 		len -= map.m_len;
1671 		map.m_lblk += map.m_len;
1672 		if (!err && len)
1673 			goto next_alloc;
1674 
1675 		map.m_len = done;
1676 	} else {
1677 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1678 	}
1679 out_err:
1680 	if (err) {
1681 		pgoff_t last_off;
1682 
1683 		if (!map.m_len)
1684 			return err;
1685 
1686 		last_off = map.m_lblk + map.m_len - 1;
1687 
1688 		/* update new size to the failed position */
1689 		new_size = (last_off == pg_end) ? offset + len :
1690 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1691 	} else {
1692 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1693 	}
1694 
1695 	if (new_size > i_size_read(inode)) {
1696 		if (mode & FALLOC_FL_KEEP_SIZE)
1697 			file_set_keep_isize(inode);
1698 		else
1699 			f2fs_i_size_write(inode, new_size);
1700 	}
1701 
1702 	return err;
1703 }
1704 
1705 static long f2fs_fallocate(struct file *file, int mode,
1706 				loff_t offset, loff_t len)
1707 {
1708 	struct inode *inode = file_inode(file);
1709 	long ret = 0;
1710 
1711 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1712 		return -EIO;
1713 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1714 		return -ENOSPC;
1715 	if (!f2fs_is_compress_backend_ready(inode))
1716 		return -EOPNOTSUPP;
1717 
1718 	/* f2fs only support ->fallocate for regular file */
1719 	if (!S_ISREG(inode->i_mode))
1720 		return -EINVAL;
1721 
1722 	if (IS_ENCRYPTED(inode) &&
1723 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1724 		return -EOPNOTSUPP;
1725 
1726 	if (f2fs_compressed_file(inode) &&
1727 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1728 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1729 		return -EOPNOTSUPP;
1730 
1731 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1732 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1733 			FALLOC_FL_INSERT_RANGE))
1734 		return -EOPNOTSUPP;
1735 
1736 	inode_lock(inode);
1737 
1738 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1739 		if (offset >= inode->i_size)
1740 			goto out;
1741 
1742 		ret = punch_hole(inode, offset, len);
1743 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1744 		ret = f2fs_collapse_range(inode, offset, len);
1745 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1746 		ret = f2fs_zero_range(inode, offset, len, mode);
1747 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1748 		ret = f2fs_insert_range(inode, offset, len);
1749 	} else {
1750 		ret = expand_inode_data(inode, offset, len, mode);
1751 	}
1752 
1753 	if (!ret) {
1754 		inode->i_mtime = inode->i_ctime = current_time(inode);
1755 		f2fs_mark_inode_dirty_sync(inode, false);
1756 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1757 	}
1758 
1759 out:
1760 	inode_unlock(inode);
1761 
1762 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1763 	return ret;
1764 }
1765 
1766 static int f2fs_release_file(struct inode *inode, struct file *filp)
1767 {
1768 	/*
1769 	 * f2fs_relase_file is called at every close calls. So we should
1770 	 * not drop any inmemory pages by close called by other process.
1771 	 */
1772 	if (!(filp->f_mode & FMODE_WRITE) ||
1773 			atomic_read(&inode->i_writecount) != 1)
1774 		return 0;
1775 
1776 	/* some remained atomic pages should discarded */
1777 	if (f2fs_is_atomic_file(inode))
1778 		f2fs_drop_inmem_pages(inode);
1779 	if (f2fs_is_volatile_file(inode)) {
1780 		set_inode_flag(inode, FI_DROP_CACHE);
1781 		filemap_fdatawrite(inode->i_mapping);
1782 		clear_inode_flag(inode, FI_DROP_CACHE);
1783 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1784 		stat_dec_volatile_write(inode);
1785 	}
1786 	return 0;
1787 }
1788 
1789 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1790 {
1791 	struct inode *inode = file_inode(file);
1792 
1793 	/*
1794 	 * If the process doing a transaction is crashed, we should do
1795 	 * roll-back. Otherwise, other reader/write can see corrupted database
1796 	 * until all the writers close its file. Since this should be done
1797 	 * before dropping file lock, it needs to do in ->flush.
1798 	 */
1799 	if (f2fs_is_atomic_file(inode) &&
1800 			F2FS_I(inode)->inmem_task == current)
1801 		f2fs_drop_inmem_pages(inode);
1802 	return 0;
1803 }
1804 
1805 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1806 {
1807 	struct f2fs_inode_info *fi = F2FS_I(inode);
1808 	u32 masked_flags = fi->i_flags & mask;
1809 
1810 	f2fs_bug_on(F2FS_I_SB(inode), (iflags & ~mask));
1811 
1812 	/* Is it quota file? Do not allow user to mess with it */
1813 	if (IS_NOQUOTA(inode))
1814 		return -EPERM;
1815 
1816 	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1817 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1818 			return -EOPNOTSUPP;
1819 		if (!f2fs_empty_dir(inode))
1820 			return -ENOTEMPTY;
1821 	}
1822 
1823 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1824 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1825 			return -EOPNOTSUPP;
1826 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1827 			return -EINVAL;
1828 	}
1829 
1830 	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1831 		if (masked_flags & F2FS_COMPR_FL) {
1832 			if (f2fs_disable_compressed_file(inode))
1833 				return -EINVAL;
1834 		}
1835 		if (iflags & F2FS_NOCOMP_FL)
1836 			return -EINVAL;
1837 		if (iflags & F2FS_COMPR_FL) {
1838 			if (!f2fs_may_compress(inode))
1839 				return -EINVAL;
1840 
1841 			set_compress_context(inode);
1842 		}
1843 	}
1844 	if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1845 		if (masked_flags & F2FS_COMPR_FL)
1846 			return -EINVAL;
1847 	}
1848 
1849 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1850 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1851 					(fi->i_flags & F2FS_NOCOMP_FL));
1852 
1853 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1854 		set_inode_flag(inode, FI_PROJ_INHERIT);
1855 	else
1856 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1857 
1858 	inode->i_ctime = current_time(inode);
1859 	f2fs_set_inode_flags(inode);
1860 	f2fs_mark_inode_dirty_sync(inode, true);
1861 	return 0;
1862 }
1863 
1864 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1865 
1866 /*
1867  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1868  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1869  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1870  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1871  */
1872 
1873 static const struct {
1874 	u32 iflag;
1875 	u32 fsflag;
1876 } f2fs_fsflags_map[] = {
1877 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
1878 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1879 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1880 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1881 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1882 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1883 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
1884 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1885 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1886 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1887 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
1888 };
1889 
1890 #define F2FS_GETTABLE_FS_FL (		\
1891 		FS_COMPR_FL |		\
1892 		FS_SYNC_FL |		\
1893 		FS_IMMUTABLE_FL |	\
1894 		FS_APPEND_FL |		\
1895 		FS_NODUMP_FL |		\
1896 		FS_NOATIME_FL |		\
1897 		FS_NOCOMP_FL |		\
1898 		FS_INDEX_FL |		\
1899 		FS_DIRSYNC_FL |		\
1900 		FS_PROJINHERIT_FL |	\
1901 		FS_ENCRYPT_FL |		\
1902 		FS_INLINE_DATA_FL |	\
1903 		FS_NOCOW_FL |		\
1904 		FS_VERITY_FL |		\
1905 		FS_CASEFOLD_FL)
1906 
1907 #define F2FS_SETTABLE_FS_FL (		\
1908 		FS_COMPR_FL |		\
1909 		FS_SYNC_FL |		\
1910 		FS_IMMUTABLE_FL |	\
1911 		FS_APPEND_FL |		\
1912 		FS_NODUMP_FL |		\
1913 		FS_NOATIME_FL |		\
1914 		FS_NOCOMP_FL |		\
1915 		FS_DIRSYNC_FL |		\
1916 		FS_PROJINHERIT_FL |	\
1917 		FS_CASEFOLD_FL)
1918 
1919 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1920 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1921 {
1922 	u32 fsflags = 0;
1923 	int i;
1924 
1925 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1926 		if (iflags & f2fs_fsflags_map[i].iflag)
1927 			fsflags |= f2fs_fsflags_map[i].fsflag;
1928 
1929 	return fsflags;
1930 }
1931 
1932 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1933 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1934 {
1935 	u32 iflags = 0;
1936 	int i;
1937 
1938 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1939 		if (fsflags & f2fs_fsflags_map[i].fsflag)
1940 			iflags |= f2fs_fsflags_map[i].iflag;
1941 
1942 	return iflags;
1943 }
1944 
1945 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1946 {
1947 	struct inode *inode = file_inode(filp);
1948 	struct f2fs_inode_info *fi = F2FS_I(inode);
1949 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1950 
1951 	if (IS_ENCRYPTED(inode))
1952 		fsflags |= FS_ENCRYPT_FL;
1953 	if (IS_VERITY(inode))
1954 		fsflags |= FS_VERITY_FL;
1955 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1956 		fsflags |= FS_INLINE_DATA_FL;
1957 	if (is_inode_flag_set(inode, FI_PIN_FILE))
1958 		fsflags |= FS_NOCOW_FL;
1959 
1960 	fsflags &= F2FS_GETTABLE_FS_FL;
1961 
1962 	return put_user(fsflags, (int __user *)arg);
1963 }
1964 
1965 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1966 {
1967 	struct inode *inode = file_inode(filp);
1968 	struct f2fs_inode_info *fi = F2FS_I(inode);
1969 	u32 fsflags, old_fsflags;
1970 	u32 iflags;
1971 	int ret;
1972 
1973 	if (!inode_owner_or_capable(inode))
1974 		return -EACCES;
1975 
1976 	if (get_user(fsflags, (int __user *)arg))
1977 		return -EFAULT;
1978 
1979 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
1980 		return -EOPNOTSUPP;
1981 	fsflags &= F2FS_SETTABLE_FS_FL;
1982 
1983 	iflags = f2fs_fsflags_to_iflags(fsflags);
1984 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1985 		return -EOPNOTSUPP;
1986 
1987 	ret = mnt_want_write_file(filp);
1988 	if (ret)
1989 		return ret;
1990 
1991 	inode_lock(inode);
1992 
1993 	old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1994 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1995 	if (ret)
1996 		goto out;
1997 
1998 	ret = f2fs_setflags_common(inode, iflags,
1999 			f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
2000 out:
2001 	inode_unlock(inode);
2002 	mnt_drop_write_file(filp);
2003 	return ret;
2004 }
2005 
2006 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2007 {
2008 	struct inode *inode = file_inode(filp);
2009 
2010 	return put_user(inode->i_generation, (int __user *)arg);
2011 }
2012 
2013 static int f2fs_ioc_start_atomic_write(struct file *filp)
2014 {
2015 	struct inode *inode = file_inode(filp);
2016 	struct f2fs_inode_info *fi = F2FS_I(inode);
2017 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2018 	int ret;
2019 
2020 	if (!inode_owner_or_capable(inode))
2021 		return -EACCES;
2022 
2023 	if (!S_ISREG(inode->i_mode))
2024 		return -EINVAL;
2025 
2026 	if (filp->f_flags & O_DIRECT)
2027 		return -EINVAL;
2028 
2029 	ret = mnt_want_write_file(filp);
2030 	if (ret)
2031 		return ret;
2032 
2033 	inode_lock(inode);
2034 
2035 	f2fs_disable_compressed_file(inode);
2036 
2037 	if (f2fs_is_atomic_file(inode)) {
2038 		if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2039 			ret = -EINVAL;
2040 		goto out;
2041 	}
2042 
2043 	ret = f2fs_convert_inline_inode(inode);
2044 	if (ret)
2045 		goto out;
2046 
2047 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2048 
2049 	/*
2050 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2051 	 * f2fs_is_atomic_file.
2052 	 */
2053 	if (get_dirty_pages(inode))
2054 		f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2055 			  inode->i_ino, get_dirty_pages(inode));
2056 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2057 	if (ret) {
2058 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2059 		goto out;
2060 	}
2061 
2062 	spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2063 	if (list_empty(&fi->inmem_ilist))
2064 		list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2065 	sbi->atomic_files++;
2066 	spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2067 
2068 	/* add inode in inmem_list first and set atomic_file */
2069 	set_inode_flag(inode, FI_ATOMIC_FILE);
2070 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2071 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2072 
2073 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2074 	F2FS_I(inode)->inmem_task = current;
2075 	stat_update_max_atomic_write(inode);
2076 out:
2077 	inode_unlock(inode);
2078 	mnt_drop_write_file(filp);
2079 	return ret;
2080 }
2081 
2082 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2083 {
2084 	struct inode *inode = file_inode(filp);
2085 	int ret;
2086 
2087 	if (!inode_owner_or_capable(inode))
2088 		return -EACCES;
2089 
2090 	ret = mnt_want_write_file(filp);
2091 	if (ret)
2092 		return ret;
2093 
2094 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2095 
2096 	inode_lock(inode);
2097 
2098 	if (f2fs_is_volatile_file(inode)) {
2099 		ret = -EINVAL;
2100 		goto err_out;
2101 	}
2102 
2103 	if (f2fs_is_atomic_file(inode)) {
2104 		ret = f2fs_commit_inmem_pages(inode);
2105 		if (ret)
2106 			goto err_out;
2107 
2108 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2109 		if (!ret)
2110 			f2fs_drop_inmem_pages(inode);
2111 	} else {
2112 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2113 	}
2114 err_out:
2115 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2116 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2117 		ret = -EINVAL;
2118 	}
2119 	inode_unlock(inode);
2120 	mnt_drop_write_file(filp);
2121 	return ret;
2122 }
2123 
2124 static int f2fs_ioc_start_volatile_write(struct file *filp)
2125 {
2126 	struct inode *inode = file_inode(filp);
2127 	int ret;
2128 
2129 	if (!inode_owner_or_capable(inode))
2130 		return -EACCES;
2131 
2132 	if (!S_ISREG(inode->i_mode))
2133 		return -EINVAL;
2134 
2135 	ret = mnt_want_write_file(filp);
2136 	if (ret)
2137 		return ret;
2138 
2139 	inode_lock(inode);
2140 
2141 	if (f2fs_is_volatile_file(inode))
2142 		goto out;
2143 
2144 	ret = f2fs_convert_inline_inode(inode);
2145 	if (ret)
2146 		goto out;
2147 
2148 	stat_inc_volatile_write(inode);
2149 	stat_update_max_volatile_write(inode);
2150 
2151 	set_inode_flag(inode, FI_VOLATILE_FILE);
2152 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2153 out:
2154 	inode_unlock(inode);
2155 	mnt_drop_write_file(filp);
2156 	return ret;
2157 }
2158 
2159 static int f2fs_ioc_release_volatile_write(struct file *filp)
2160 {
2161 	struct inode *inode = file_inode(filp);
2162 	int ret;
2163 
2164 	if (!inode_owner_or_capable(inode))
2165 		return -EACCES;
2166 
2167 	ret = mnt_want_write_file(filp);
2168 	if (ret)
2169 		return ret;
2170 
2171 	inode_lock(inode);
2172 
2173 	if (!f2fs_is_volatile_file(inode))
2174 		goto out;
2175 
2176 	if (!f2fs_is_first_block_written(inode)) {
2177 		ret = truncate_partial_data_page(inode, 0, true);
2178 		goto out;
2179 	}
2180 
2181 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2182 out:
2183 	inode_unlock(inode);
2184 	mnt_drop_write_file(filp);
2185 	return ret;
2186 }
2187 
2188 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2189 {
2190 	struct inode *inode = file_inode(filp);
2191 	int ret;
2192 
2193 	if (!inode_owner_or_capable(inode))
2194 		return -EACCES;
2195 
2196 	ret = mnt_want_write_file(filp);
2197 	if (ret)
2198 		return ret;
2199 
2200 	inode_lock(inode);
2201 
2202 	if (f2fs_is_atomic_file(inode))
2203 		f2fs_drop_inmem_pages(inode);
2204 	if (f2fs_is_volatile_file(inode)) {
2205 		clear_inode_flag(inode, FI_VOLATILE_FILE);
2206 		stat_dec_volatile_write(inode);
2207 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2208 	}
2209 
2210 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2211 
2212 	inode_unlock(inode);
2213 
2214 	mnt_drop_write_file(filp);
2215 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2216 	return ret;
2217 }
2218 
2219 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2220 {
2221 	struct inode *inode = file_inode(filp);
2222 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2223 	struct super_block *sb = sbi->sb;
2224 	__u32 in;
2225 	int ret = 0;
2226 
2227 	if (!capable(CAP_SYS_ADMIN))
2228 		return -EPERM;
2229 
2230 	if (get_user(in, (__u32 __user *)arg))
2231 		return -EFAULT;
2232 
2233 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2234 		ret = mnt_want_write_file(filp);
2235 		if (ret) {
2236 			if (ret == -EROFS) {
2237 				ret = 0;
2238 				f2fs_stop_checkpoint(sbi, false);
2239 				set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2240 				trace_f2fs_shutdown(sbi, in, ret);
2241 			}
2242 			return ret;
2243 		}
2244 	}
2245 
2246 	switch (in) {
2247 	case F2FS_GOING_DOWN_FULLSYNC:
2248 		sb = freeze_bdev(sb->s_bdev);
2249 		if (IS_ERR(sb)) {
2250 			ret = PTR_ERR(sb);
2251 			goto out;
2252 		}
2253 		if (sb) {
2254 			f2fs_stop_checkpoint(sbi, false);
2255 			set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2256 			thaw_bdev(sb->s_bdev, sb);
2257 		}
2258 		break;
2259 	case F2FS_GOING_DOWN_METASYNC:
2260 		/* do checkpoint only */
2261 		ret = f2fs_sync_fs(sb, 1);
2262 		if (ret)
2263 			goto out;
2264 		f2fs_stop_checkpoint(sbi, false);
2265 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2266 		break;
2267 	case F2FS_GOING_DOWN_NOSYNC:
2268 		f2fs_stop_checkpoint(sbi, false);
2269 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2270 		break;
2271 	case F2FS_GOING_DOWN_METAFLUSH:
2272 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2273 		f2fs_stop_checkpoint(sbi, false);
2274 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2275 		break;
2276 	case F2FS_GOING_DOWN_NEED_FSCK:
2277 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2278 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2279 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2280 		/* do checkpoint only */
2281 		ret = f2fs_sync_fs(sb, 1);
2282 		goto out;
2283 	default:
2284 		ret = -EINVAL;
2285 		goto out;
2286 	}
2287 
2288 	f2fs_stop_gc_thread(sbi);
2289 	f2fs_stop_discard_thread(sbi);
2290 
2291 	f2fs_drop_discard_cmd(sbi);
2292 	clear_opt(sbi, DISCARD);
2293 
2294 	f2fs_update_time(sbi, REQ_TIME);
2295 out:
2296 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2297 		mnt_drop_write_file(filp);
2298 
2299 	trace_f2fs_shutdown(sbi, in, ret);
2300 
2301 	return ret;
2302 }
2303 
2304 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2305 {
2306 	struct inode *inode = file_inode(filp);
2307 	struct super_block *sb = inode->i_sb;
2308 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
2309 	struct fstrim_range range;
2310 	int ret;
2311 
2312 	if (!capable(CAP_SYS_ADMIN))
2313 		return -EPERM;
2314 
2315 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2316 		return -EOPNOTSUPP;
2317 
2318 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2319 				sizeof(range)))
2320 		return -EFAULT;
2321 
2322 	ret = mnt_want_write_file(filp);
2323 	if (ret)
2324 		return ret;
2325 
2326 	range.minlen = max((unsigned int)range.minlen,
2327 				q->limits.discard_granularity);
2328 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2329 	mnt_drop_write_file(filp);
2330 	if (ret < 0)
2331 		return ret;
2332 
2333 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2334 				sizeof(range)))
2335 		return -EFAULT;
2336 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2337 	return 0;
2338 }
2339 
2340 static bool uuid_is_nonzero(__u8 u[16])
2341 {
2342 	int i;
2343 
2344 	for (i = 0; i < 16; i++)
2345 		if (u[i])
2346 			return true;
2347 	return false;
2348 }
2349 
2350 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2351 {
2352 	struct inode *inode = file_inode(filp);
2353 
2354 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2355 		return -EOPNOTSUPP;
2356 
2357 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2358 
2359 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2360 }
2361 
2362 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2363 {
2364 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2365 		return -EOPNOTSUPP;
2366 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2367 }
2368 
2369 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2370 {
2371 	struct inode *inode = file_inode(filp);
2372 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2373 	int err;
2374 
2375 	if (!f2fs_sb_has_encrypt(sbi))
2376 		return -EOPNOTSUPP;
2377 
2378 	err = mnt_want_write_file(filp);
2379 	if (err)
2380 		return err;
2381 
2382 	down_write(&sbi->sb_lock);
2383 
2384 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2385 		goto got_it;
2386 
2387 	/* update superblock with uuid */
2388 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2389 
2390 	err = f2fs_commit_super(sbi, false);
2391 	if (err) {
2392 		/* undo new data */
2393 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2394 		goto out_err;
2395 	}
2396 got_it:
2397 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2398 									16))
2399 		err = -EFAULT;
2400 out_err:
2401 	up_write(&sbi->sb_lock);
2402 	mnt_drop_write_file(filp);
2403 	return err;
2404 }
2405 
2406 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2407 					     unsigned long arg)
2408 {
2409 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2410 		return -EOPNOTSUPP;
2411 
2412 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2413 }
2414 
2415 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2416 {
2417 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2418 		return -EOPNOTSUPP;
2419 
2420 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2421 }
2422 
2423 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2424 {
2425 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2426 		return -EOPNOTSUPP;
2427 
2428 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2429 }
2430 
2431 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2432 						    unsigned long arg)
2433 {
2434 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2435 		return -EOPNOTSUPP;
2436 
2437 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2438 }
2439 
2440 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2441 					      unsigned long arg)
2442 {
2443 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2444 		return -EOPNOTSUPP;
2445 
2446 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2447 }
2448 
2449 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2450 {
2451 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2452 		return -EOPNOTSUPP;
2453 
2454 	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2455 }
2456 
2457 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2458 {
2459 	struct inode *inode = file_inode(filp);
2460 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2461 	__u32 sync;
2462 	int ret;
2463 
2464 	if (!capable(CAP_SYS_ADMIN))
2465 		return -EPERM;
2466 
2467 	if (get_user(sync, (__u32 __user *)arg))
2468 		return -EFAULT;
2469 
2470 	if (f2fs_readonly(sbi->sb))
2471 		return -EROFS;
2472 
2473 	ret = mnt_want_write_file(filp);
2474 	if (ret)
2475 		return ret;
2476 
2477 	if (!sync) {
2478 		if (!down_write_trylock(&sbi->gc_lock)) {
2479 			ret = -EBUSY;
2480 			goto out;
2481 		}
2482 	} else {
2483 		down_write(&sbi->gc_lock);
2484 	}
2485 
2486 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2487 out:
2488 	mnt_drop_write_file(filp);
2489 	return ret;
2490 }
2491 
2492 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2493 {
2494 	struct inode *inode = file_inode(filp);
2495 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2496 	struct f2fs_gc_range range;
2497 	u64 end;
2498 	int ret;
2499 
2500 	if (!capable(CAP_SYS_ADMIN))
2501 		return -EPERM;
2502 
2503 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2504 							sizeof(range)))
2505 		return -EFAULT;
2506 
2507 	if (f2fs_readonly(sbi->sb))
2508 		return -EROFS;
2509 
2510 	end = range.start + range.len;
2511 	if (end < range.start || range.start < MAIN_BLKADDR(sbi) ||
2512 					end >= MAX_BLKADDR(sbi))
2513 		return -EINVAL;
2514 
2515 	ret = mnt_want_write_file(filp);
2516 	if (ret)
2517 		return ret;
2518 
2519 do_more:
2520 	if (!range.sync) {
2521 		if (!down_write_trylock(&sbi->gc_lock)) {
2522 			ret = -EBUSY;
2523 			goto out;
2524 		}
2525 	} else {
2526 		down_write(&sbi->gc_lock);
2527 	}
2528 
2529 	ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2530 	range.start += BLKS_PER_SEC(sbi);
2531 	if (range.start <= end)
2532 		goto do_more;
2533 out:
2534 	mnt_drop_write_file(filp);
2535 	return ret;
2536 }
2537 
2538 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2539 {
2540 	struct inode *inode = file_inode(filp);
2541 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2542 	int ret;
2543 
2544 	if (!capable(CAP_SYS_ADMIN))
2545 		return -EPERM;
2546 
2547 	if (f2fs_readonly(sbi->sb))
2548 		return -EROFS;
2549 
2550 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2551 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2552 		return -EINVAL;
2553 	}
2554 
2555 	ret = mnt_want_write_file(filp);
2556 	if (ret)
2557 		return ret;
2558 
2559 	ret = f2fs_sync_fs(sbi->sb, 1);
2560 
2561 	mnt_drop_write_file(filp);
2562 	return ret;
2563 }
2564 
2565 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2566 					struct file *filp,
2567 					struct f2fs_defragment *range)
2568 {
2569 	struct inode *inode = file_inode(filp);
2570 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2571 					.m_seg_type = NO_CHECK_TYPE ,
2572 					.m_may_create = false };
2573 	struct extent_info ei = {0, 0, 0};
2574 	pgoff_t pg_start, pg_end, next_pgofs;
2575 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2576 	unsigned int total = 0, sec_num;
2577 	block_t blk_end = 0;
2578 	bool fragmented = false;
2579 	int err;
2580 
2581 	/* if in-place-update policy is enabled, don't waste time here */
2582 	if (f2fs_should_update_inplace(inode, NULL))
2583 		return -EINVAL;
2584 
2585 	pg_start = range->start >> PAGE_SHIFT;
2586 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2587 
2588 	f2fs_balance_fs(sbi, true);
2589 
2590 	inode_lock(inode);
2591 
2592 	/* writeback all dirty pages in the range */
2593 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2594 						range->start + range->len - 1);
2595 	if (err)
2596 		goto out;
2597 
2598 	/*
2599 	 * lookup mapping info in extent cache, skip defragmenting if physical
2600 	 * block addresses are continuous.
2601 	 */
2602 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2603 		if (ei.fofs + ei.len >= pg_end)
2604 			goto out;
2605 	}
2606 
2607 	map.m_lblk = pg_start;
2608 	map.m_next_pgofs = &next_pgofs;
2609 
2610 	/*
2611 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2612 	 * physical block addresses are continuous even if there are hole(s)
2613 	 * in logical blocks.
2614 	 */
2615 	while (map.m_lblk < pg_end) {
2616 		map.m_len = pg_end - map.m_lblk;
2617 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2618 		if (err)
2619 			goto out;
2620 
2621 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2622 			map.m_lblk = next_pgofs;
2623 			continue;
2624 		}
2625 
2626 		if (blk_end && blk_end != map.m_pblk)
2627 			fragmented = true;
2628 
2629 		/* record total count of block that we're going to move */
2630 		total += map.m_len;
2631 
2632 		blk_end = map.m_pblk + map.m_len;
2633 
2634 		map.m_lblk += map.m_len;
2635 	}
2636 
2637 	if (!fragmented) {
2638 		total = 0;
2639 		goto out;
2640 	}
2641 
2642 	sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2643 
2644 	/*
2645 	 * make sure there are enough free section for LFS allocation, this can
2646 	 * avoid defragment running in SSR mode when free section are allocated
2647 	 * intensively
2648 	 */
2649 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2650 		err = -EAGAIN;
2651 		goto out;
2652 	}
2653 
2654 	map.m_lblk = pg_start;
2655 	map.m_len = pg_end - pg_start;
2656 	total = 0;
2657 
2658 	while (map.m_lblk < pg_end) {
2659 		pgoff_t idx;
2660 		int cnt = 0;
2661 
2662 do_map:
2663 		map.m_len = pg_end - map.m_lblk;
2664 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2665 		if (err)
2666 			goto clear_out;
2667 
2668 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2669 			map.m_lblk = next_pgofs;
2670 			goto check;
2671 		}
2672 
2673 		set_inode_flag(inode, FI_DO_DEFRAG);
2674 
2675 		idx = map.m_lblk;
2676 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2677 			struct page *page;
2678 
2679 			page = f2fs_get_lock_data_page(inode, idx, true);
2680 			if (IS_ERR(page)) {
2681 				err = PTR_ERR(page);
2682 				goto clear_out;
2683 			}
2684 
2685 			set_page_dirty(page);
2686 			f2fs_put_page(page, 1);
2687 
2688 			idx++;
2689 			cnt++;
2690 			total++;
2691 		}
2692 
2693 		map.m_lblk = idx;
2694 check:
2695 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
2696 			goto do_map;
2697 
2698 		clear_inode_flag(inode, FI_DO_DEFRAG);
2699 
2700 		err = filemap_fdatawrite(inode->i_mapping);
2701 		if (err)
2702 			goto out;
2703 	}
2704 clear_out:
2705 	clear_inode_flag(inode, FI_DO_DEFRAG);
2706 out:
2707 	inode_unlock(inode);
2708 	if (!err)
2709 		range->len = (u64)total << PAGE_SHIFT;
2710 	return err;
2711 }
2712 
2713 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2714 {
2715 	struct inode *inode = file_inode(filp);
2716 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2717 	struct f2fs_defragment range;
2718 	int err;
2719 
2720 	if (!capable(CAP_SYS_ADMIN))
2721 		return -EPERM;
2722 
2723 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2724 		return -EINVAL;
2725 
2726 	if (f2fs_readonly(sbi->sb))
2727 		return -EROFS;
2728 
2729 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2730 							sizeof(range)))
2731 		return -EFAULT;
2732 
2733 	/* verify alignment of offset & size */
2734 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2735 		return -EINVAL;
2736 
2737 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2738 					sbi->max_file_blocks))
2739 		return -EINVAL;
2740 
2741 	err = mnt_want_write_file(filp);
2742 	if (err)
2743 		return err;
2744 
2745 	err = f2fs_defragment_range(sbi, filp, &range);
2746 	mnt_drop_write_file(filp);
2747 
2748 	f2fs_update_time(sbi, REQ_TIME);
2749 	if (err < 0)
2750 		return err;
2751 
2752 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2753 							sizeof(range)))
2754 		return -EFAULT;
2755 
2756 	return 0;
2757 }
2758 
2759 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2760 			struct file *file_out, loff_t pos_out, size_t len)
2761 {
2762 	struct inode *src = file_inode(file_in);
2763 	struct inode *dst = file_inode(file_out);
2764 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2765 	size_t olen = len, dst_max_i_size = 0;
2766 	size_t dst_osize;
2767 	int ret;
2768 
2769 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2770 				src->i_sb != dst->i_sb)
2771 		return -EXDEV;
2772 
2773 	if (unlikely(f2fs_readonly(src->i_sb)))
2774 		return -EROFS;
2775 
2776 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2777 		return -EINVAL;
2778 
2779 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2780 		return -EOPNOTSUPP;
2781 
2782 	if (src == dst) {
2783 		if (pos_in == pos_out)
2784 			return 0;
2785 		if (pos_out > pos_in && pos_out < pos_in + len)
2786 			return -EINVAL;
2787 	}
2788 
2789 	inode_lock(src);
2790 	if (src != dst) {
2791 		ret = -EBUSY;
2792 		if (!inode_trylock(dst))
2793 			goto out;
2794 	}
2795 
2796 	ret = -EINVAL;
2797 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2798 		goto out_unlock;
2799 	if (len == 0)
2800 		olen = len = src->i_size - pos_in;
2801 	if (pos_in + len == src->i_size)
2802 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2803 	if (len == 0) {
2804 		ret = 0;
2805 		goto out_unlock;
2806 	}
2807 
2808 	dst_osize = dst->i_size;
2809 	if (pos_out + olen > dst->i_size)
2810 		dst_max_i_size = pos_out + olen;
2811 
2812 	/* verify the end result is block aligned */
2813 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2814 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2815 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2816 		goto out_unlock;
2817 
2818 	ret = f2fs_convert_inline_inode(src);
2819 	if (ret)
2820 		goto out_unlock;
2821 
2822 	ret = f2fs_convert_inline_inode(dst);
2823 	if (ret)
2824 		goto out_unlock;
2825 
2826 	/* write out all dirty pages from offset */
2827 	ret = filemap_write_and_wait_range(src->i_mapping,
2828 					pos_in, pos_in + len);
2829 	if (ret)
2830 		goto out_unlock;
2831 
2832 	ret = filemap_write_and_wait_range(dst->i_mapping,
2833 					pos_out, pos_out + len);
2834 	if (ret)
2835 		goto out_unlock;
2836 
2837 	f2fs_balance_fs(sbi, true);
2838 
2839 	down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2840 	if (src != dst) {
2841 		ret = -EBUSY;
2842 		if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2843 			goto out_src;
2844 	}
2845 
2846 	f2fs_lock_op(sbi);
2847 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2848 				pos_out >> F2FS_BLKSIZE_BITS,
2849 				len >> F2FS_BLKSIZE_BITS, false);
2850 
2851 	if (!ret) {
2852 		if (dst_max_i_size)
2853 			f2fs_i_size_write(dst, dst_max_i_size);
2854 		else if (dst_osize != dst->i_size)
2855 			f2fs_i_size_write(dst, dst_osize);
2856 	}
2857 	f2fs_unlock_op(sbi);
2858 
2859 	if (src != dst)
2860 		up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2861 out_src:
2862 	up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2863 out_unlock:
2864 	if (src != dst)
2865 		inode_unlock(dst);
2866 out:
2867 	inode_unlock(src);
2868 	return ret;
2869 }
2870 
2871 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2872 {
2873 	struct f2fs_move_range range;
2874 	struct fd dst;
2875 	int err;
2876 
2877 	if (!(filp->f_mode & FMODE_READ) ||
2878 			!(filp->f_mode & FMODE_WRITE))
2879 		return -EBADF;
2880 
2881 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2882 							sizeof(range)))
2883 		return -EFAULT;
2884 
2885 	dst = fdget(range.dst_fd);
2886 	if (!dst.file)
2887 		return -EBADF;
2888 
2889 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2890 		err = -EBADF;
2891 		goto err_out;
2892 	}
2893 
2894 	err = mnt_want_write_file(filp);
2895 	if (err)
2896 		goto err_out;
2897 
2898 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2899 					range.pos_out, range.len);
2900 
2901 	mnt_drop_write_file(filp);
2902 	if (err)
2903 		goto err_out;
2904 
2905 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2906 						&range, sizeof(range)))
2907 		err = -EFAULT;
2908 err_out:
2909 	fdput(dst);
2910 	return err;
2911 }
2912 
2913 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2914 {
2915 	struct inode *inode = file_inode(filp);
2916 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2917 	struct sit_info *sm = SIT_I(sbi);
2918 	unsigned int start_segno = 0, end_segno = 0;
2919 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2920 	struct f2fs_flush_device range;
2921 	int ret;
2922 
2923 	if (!capable(CAP_SYS_ADMIN))
2924 		return -EPERM;
2925 
2926 	if (f2fs_readonly(sbi->sb))
2927 		return -EROFS;
2928 
2929 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2930 		return -EINVAL;
2931 
2932 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2933 							sizeof(range)))
2934 		return -EFAULT;
2935 
2936 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2937 			__is_large_section(sbi)) {
2938 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2939 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2940 		return -EINVAL;
2941 	}
2942 
2943 	ret = mnt_want_write_file(filp);
2944 	if (ret)
2945 		return ret;
2946 
2947 	if (range.dev_num != 0)
2948 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2949 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2950 
2951 	start_segno = sm->last_victim[FLUSH_DEVICE];
2952 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2953 		start_segno = dev_start_segno;
2954 	end_segno = min(start_segno + range.segments, dev_end_segno);
2955 
2956 	while (start_segno < end_segno) {
2957 		if (!down_write_trylock(&sbi->gc_lock)) {
2958 			ret = -EBUSY;
2959 			goto out;
2960 		}
2961 		sm->last_victim[GC_CB] = end_segno + 1;
2962 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2963 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2964 		ret = f2fs_gc(sbi, true, true, start_segno);
2965 		if (ret == -EAGAIN)
2966 			ret = 0;
2967 		else if (ret < 0)
2968 			break;
2969 		start_segno++;
2970 	}
2971 out:
2972 	mnt_drop_write_file(filp);
2973 	return ret;
2974 }
2975 
2976 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2977 {
2978 	struct inode *inode = file_inode(filp);
2979 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2980 
2981 	/* Must validate to set it with SQLite behavior in Android. */
2982 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2983 
2984 	return put_user(sb_feature, (u32 __user *)arg);
2985 }
2986 
2987 #ifdef CONFIG_QUOTA
2988 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2989 {
2990 	struct dquot *transfer_to[MAXQUOTAS] = {};
2991 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2992 	struct super_block *sb = sbi->sb;
2993 	int err = 0;
2994 
2995 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2996 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2997 		err = __dquot_transfer(inode, transfer_to);
2998 		if (err)
2999 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3000 		dqput(transfer_to[PRJQUOTA]);
3001 	}
3002 	return err;
3003 }
3004 
3005 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3006 {
3007 	struct inode *inode = file_inode(filp);
3008 	struct f2fs_inode_info *fi = F2FS_I(inode);
3009 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3010 	struct page *ipage;
3011 	kprojid_t kprojid;
3012 	int err;
3013 
3014 	if (!f2fs_sb_has_project_quota(sbi)) {
3015 		if (projid != F2FS_DEF_PROJID)
3016 			return -EOPNOTSUPP;
3017 		else
3018 			return 0;
3019 	}
3020 
3021 	if (!f2fs_has_extra_attr(inode))
3022 		return -EOPNOTSUPP;
3023 
3024 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3025 
3026 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
3027 		return 0;
3028 
3029 	err = -EPERM;
3030 	/* Is it quota file? Do not allow user to mess with it */
3031 	if (IS_NOQUOTA(inode))
3032 		return err;
3033 
3034 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
3035 	if (IS_ERR(ipage))
3036 		return PTR_ERR(ipage);
3037 
3038 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
3039 								i_projid)) {
3040 		err = -EOVERFLOW;
3041 		f2fs_put_page(ipage, 1);
3042 		return err;
3043 	}
3044 	f2fs_put_page(ipage, 1);
3045 
3046 	err = dquot_initialize(inode);
3047 	if (err)
3048 		return err;
3049 
3050 	f2fs_lock_op(sbi);
3051 	err = f2fs_transfer_project_quota(inode, kprojid);
3052 	if (err)
3053 		goto out_unlock;
3054 
3055 	F2FS_I(inode)->i_projid = kprojid;
3056 	inode->i_ctime = current_time(inode);
3057 	f2fs_mark_inode_dirty_sync(inode, true);
3058 out_unlock:
3059 	f2fs_unlock_op(sbi);
3060 	return err;
3061 }
3062 #else
3063 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3064 {
3065 	return 0;
3066 }
3067 
3068 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3069 {
3070 	if (projid != F2FS_DEF_PROJID)
3071 		return -EOPNOTSUPP;
3072 	return 0;
3073 }
3074 #endif
3075 
3076 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
3077 
3078 /*
3079  * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
3080  * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
3081  * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
3082  */
3083 
3084 static const struct {
3085 	u32 iflag;
3086 	u32 xflag;
3087 } f2fs_xflags_map[] = {
3088 	{ F2FS_SYNC_FL,		FS_XFLAG_SYNC },
3089 	{ F2FS_IMMUTABLE_FL,	FS_XFLAG_IMMUTABLE },
3090 	{ F2FS_APPEND_FL,	FS_XFLAG_APPEND },
3091 	{ F2FS_NODUMP_FL,	FS_XFLAG_NODUMP },
3092 	{ F2FS_NOATIME_FL,	FS_XFLAG_NOATIME },
3093 	{ F2FS_PROJINHERIT_FL,	FS_XFLAG_PROJINHERIT },
3094 };
3095 
3096 #define F2FS_SUPPORTED_XFLAGS (		\
3097 		FS_XFLAG_SYNC |		\
3098 		FS_XFLAG_IMMUTABLE |	\
3099 		FS_XFLAG_APPEND |	\
3100 		FS_XFLAG_NODUMP |	\
3101 		FS_XFLAG_NOATIME |	\
3102 		FS_XFLAG_PROJINHERIT)
3103 
3104 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
3105 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
3106 {
3107 	u32 xflags = 0;
3108 	int i;
3109 
3110 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3111 		if (iflags & f2fs_xflags_map[i].iflag)
3112 			xflags |= f2fs_xflags_map[i].xflag;
3113 
3114 	return xflags;
3115 }
3116 
3117 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
3118 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
3119 {
3120 	u32 iflags = 0;
3121 	int i;
3122 
3123 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3124 		if (xflags & f2fs_xflags_map[i].xflag)
3125 			iflags |= f2fs_xflags_map[i].iflag;
3126 
3127 	return iflags;
3128 }
3129 
3130 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
3131 {
3132 	struct f2fs_inode_info *fi = F2FS_I(inode);
3133 
3134 	simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
3135 
3136 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3137 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3138 }
3139 
3140 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
3141 {
3142 	struct inode *inode = file_inode(filp);
3143 	struct fsxattr fa;
3144 
3145 	f2fs_fill_fsxattr(inode, &fa);
3146 
3147 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
3148 		return -EFAULT;
3149 	return 0;
3150 }
3151 
3152 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
3153 {
3154 	struct inode *inode = file_inode(filp);
3155 	struct fsxattr fa, old_fa;
3156 	u32 iflags;
3157 	int err;
3158 
3159 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
3160 		return -EFAULT;
3161 
3162 	/* Make sure caller has proper permission */
3163 	if (!inode_owner_or_capable(inode))
3164 		return -EACCES;
3165 
3166 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
3167 		return -EOPNOTSUPP;
3168 
3169 	iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
3170 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3171 		return -EOPNOTSUPP;
3172 
3173 	err = mnt_want_write_file(filp);
3174 	if (err)
3175 		return err;
3176 
3177 	inode_lock(inode);
3178 
3179 	f2fs_fill_fsxattr(inode, &old_fa);
3180 	err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
3181 	if (err)
3182 		goto out;
3183 
3184 	err = f2fs_setflags_common(inode, iflags,
3185 			f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
3186 	if (err)
3187 		goto out;
3188 
3189 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3190 out:
3191 	inode_unlock(inode);
3192 	mnt_drop_write_file(filp);
3193 	return err;
3194 }
3195 
3196 int f2fs_pin_file_control(struct inode *inode, bool inc)
3197 {
3198 	struct f2fs_inode_info *fi = F2FS_I(inode);
3199 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3200 
3201 	/* Use i_gc_failures for normal file as a risk signal. */
3202 	if (inc)
3203 		f2fs_i_gc_failures_write(inode,
3204 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3205 
3206 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3207 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3208 			  __func__, inode->i_ino,
3209 			  fi->i_gc_failures[GC_FAILURE_PIN]);
3210 		clear_inode_flag(inode, FI_PIN_FILE);
3211 		return -EAGAIN;
3212 	}
3213 	return 0;
3214 }
3215 
3216 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3217 {
3218 	struct inode *inode = file_inode(filp);
3219 	__u32 pin;
3220 	int ret = 0;
3221 
3222 	if (get_user(pin, (__u32 __user *)arg))
3223 		return -EFAULT;
3224 
3225 	if (!S_ISREG(inode->i_mode))
3226 		return -EINVAL;
3227 
3228 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3229 		return -EROFS;
3230 
3231 	ret = mnt_want_write_file(filp);
3232 	if (ret)
3233 		return ret;
3234 
3235 	inode_lock(inode);
3236 
3237 	if (f2fs_should_update_outplace(inode, NULL)) {
3238 		ret = -EINVAL;
3239 		goto out;
3240 	}
3241 
3242 	if (!pin) {
3243 		clear_inode_flag(inode, FI_PIN_FILE);
3244 		f2fs_i_gc_failures_write(inode, 0);
3245 		goto done;
3246 	}
3247 
3248 	if (f2fs_pin_file_control(inode, false)) {
3249 		ret = -EAGAIN;
3250 		goto out;
3251 	}
3252 
3253 	ret = f2fs_convert_inline_inode(inode);
3254 	if (ret)
3255 		goto out;
3256 
3257 	if (f2fs_disable_compressed_file(inode)) {
3258 		ret = -EOPNOTSUPP;
3259 		goto out;
3260 	}
3261 
3262 	set_inode_flag(inode, FI_PIN_FILE);
3263 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3264 done:
3265 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3266 out:
3267 	inode_unlock(inode);
3268 	mnt_drop_write_file(filp);
3269 	return ret;
3270 }
3271 
3272 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3273 {
3274 	struct inode *inode = file_inode(filp);
3275 	__u32 pin = 0;
3276 
3277 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3278 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3279 	return put_user(pin, (u32 __user *)arg);
3280 }
3281 
3282 int f2fs_precache_extents(struct inode *inode)
3283 {
3284 	struct f2fs_inode_info *fi = F2FS_I(inode);
3285 	struct f2fs_map_blocks map;
3286 	pgoff_t m_next_extent;
3287 	loff_t end;
3288 	int err;
3289 
3290 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3291 		return -EOPNOTSUPP;
3292 
3293 	map.m_lblk = 0;
3294 	map.m_next_pgofs = NULL;
3295 	map.m_next_extent = &m_next_extent;
3296 	map.m_seg_type = NO_CHECK_TYPE;
3297 	map.m_may_create = false;
3298 	end = F2FS_I_SB(inode)->max_file_blocks;
3299 
3300 	while (map.m_lblk < end) {
3301 		map.m_len = end - map.m_lblk;
3302 
3303 		down_write(&fi->i_gc_rwsem[WRITE]);
3304 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3305 		up_write(&fi->i_gc_rwsem[WRITE]);
3306 		if (err)
3307 			return err;
3308 
3309 		map.m_lblk = m_next_extent;
3310 	}
3311 
3312 	return err;
3313 }
3314 
3315 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3316 {
3317 	return f2fs_precache_extents(file_inode(filp));
3318 }
3319 
3320 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3321 {
3322 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3323 	__u64 block_count;
3324 
3325 	if (!capable(CAP_SYS_ADMIN))
3326 		return -EPERM;
3327 
3328 	if (f2fs_readonly(sbi->sb))
3329 		return -EROFS;
3330 
3331 	if (copy_from_user(&block_count, (void __user *)arg,
3332 			   sizeof(block_count)))
3333 		return -EFAULT;
3334 
3335 	return f2fs_resize_fs(sbi, block_count);
3336 }
3337 
3338 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3339 {
3340 	struct inode *inode = file_inode(filp);
3341 
3342 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3343 
3344 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3345 		f2fs_warn(F2FS_I_SB(inode),
3346 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3347 			  inode->i_ino);
3348 		return -EOPNOTSUPP;
3349 	}
3350 
3351 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3352 }
3353 
3354 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3355 {
3356 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3357 		return -EOPNOTSUPP;
3358 
3359 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3360 }
3361 
3362 static int f2fs_get_volume_name(struct file *filp, unsigned long arg)
3363 {
3364 	struct inode *inode = file_inode(filp);
3365 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3366 	char *vbuf;
3367 	int count;
3368 	int err = 0;
3369 
3370 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3371 	if (!vbuf)
3372 		return -ENOMEM;
3373 
3374 	down_read(&sbi->sb_lock);
3375 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3376 			ARRAY_SIZE(sbi->raw_super->volume_name),
3377 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3378 	up_read(&sbi->sb_lock);
3379 
3380 	if (copy_to_user((char __user *)arg, vbuf,
3381 				min(FSLABEL_MAX, count)))
3382 		err = -EFAULT;
3383 
3384 	kvfree(vbuf);
3385 	return err;
3386 }
3387 
3388 static int f2fs_set_volume_name(struct file *filp, unsigned long arg)
3389 {
3390 	struct inode *inode = file_inode(filp);
3391 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3392 	char *vbuf;
3393 	int err = 0;
3394 
3395 	if (!capable(CAP_SYS_ADMIN))
3396 		return -EPERM;
3397 
3398 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3399 	if (IS_ERR(vbuf))
3400 		return PTR_ERR(vbuf);
3401 
3402 	err = mnt_want_write_file(filp);
3403 	if (err)
3404 		goto out;
3405 
3406 	down_write(&sbi->sb_lock);
3407 
3408 	memset(sbi->raw_super->volume_name, 0,
3409 			sizeof(sbi->raw_super->volume_name));
3410 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3411 			sbi->raw_super->volume_name,
3412 			ARRAY_SIZE(sbi->raw_super->volume_name));
3413 
3414 	err = f2fs_commit_super(sbi, false);
3415 
3416 	up_write(&sbi->sb_lock);
3417 
3418 	mnt_drop_write_file(filp);
3419 out:
3420 	kfree(vbuf);
3421 	return err;
3422 }
3423 
3424 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3425 {
3426 	struct inode *inode = file_inode(filp);
3427 	__u64 blocks;
3428 
3429 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3430 		return -EOPNOTSUPP;
3431 
3432 	if (!f2fs_compressed_file(inode))
3433 		return -EINVAL;
3434 
3435 	blocks = F2FS_I(inode)->i_compr_blocks;
3436 	return put_user(blocks, (u64 __user *)arg);
3437 }
3438 
3439 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3440 {
3441 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3442 	unsigned int released_blocks = 0;
3443 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3444 	block_t blkaddr;
3445 	int i;
3446 
3447 	for (i = 0; i < count; i++) {
3448 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3449 						dn->ofs_in_node + i);
3450 
3451 		if (!__is_valid_data_blkaddr(blkaddr))
3452 			continue;
3453 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3454 					DATA_GENERIC_ENHANCE)))
3455 			return -EFSCORRUPTED;
3456 	}
3457 
3458 	while (count) {
3459 		int compr_blocks = 0;
3460 
3461 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3462 			blkaddr = f2fs_data_blkaddr(dn);
3463 
3464 			if (i == 0) {
3465 				if (blkaddr == COMPRESS_ADDR)
3466 					continue;
3467 				dn->ofs_in_node += cluster_size;
3468 				goto next;
3469 			}
3470 
3471 			if (__is_valid_data_blkaddr(blkaddr))
3472 				compr_blocks++;
3473 
3474 			if (blkaddr != NEW_ADDR)
3475 				continue;
3476 
3477 			dn->data_blkaddr = NULL_ADDR;
3478 			f2fs_set_data_blkaddr(dn);
3479 		}
3480 
3481 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3482 		dec_valid_block_count(sbi, dn->inode,
3483 					cluster_size - compr_blocks);
3484 
3485 		released_blocks += cluster_size - compr_blocks;
3486 next:
3487 		count -= cluster_size;
3488 	}
3489 
3490 	return released_blocks;
3491 }
3492 
3493 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3494 {
3495 	struct inode *inode = file_inode(filp);
3496 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3497 	pgoff_t page_idx = 0, last_idx;
3498 	unsigned int released_blocks = 0;
3499 	int ret;
3500 	int writecount;
3501 
3502 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3503 		return -EOPNOTSUPP;
3504 
3505 	if (!f2fs_compressed_file(inode))
3506 		return -EINVAL;
3507 
3508 	if (f2fs_readonly(sbi->sb))
3509 		return -EROFS;
3510 
3511 	ret = mnt_want_write_file(filp);
3512 	if (ret)
3513 		return ret;
3514 
3515 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3516 
3517 	inode_lock(inode);
3518 
3519 	writecount = atomic_read(&inode->i_writecount);
3520 	if ((filp->f_mode & FMODE_WRITE && writecount != 1) || writecount) {
3521 		ret = -EBUSY;
3522 		goto out;
3523 	}
3524 
3525 	if (IS_IMMUTABLE(inode)) {
3526 		ret = -EINVAL;
3527 		goto out;
3528 	}
3529 
3530 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3531 	if (ret)
3532 		goto out;
3533 
3534 	if (!F2FS_I(inode)->i_compr_blocks)
3535 		goto out;
3536 
3537 	F2FS_I(inode)->i_flags |= F2FS_IMMUTABLE_FL;
3538 	f2fs_set_inode_flags(inode);
3539 	inode->i_ctime = current_time(inode);
3540 	f2fs_mark_inode_dirty_sync(inode, true);
3541 
3542 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3543 	down_write(&F2FS_I(inode)->i_mmap_sem);
3544 
3545 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3546 
3547 	while (page_idx < last_idx) {
3548 		struct dnode_of_data dn;
3549 		pgoff_t end_offset, count;
3550 
3551 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3552 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3553 		if (ret) {
3554 			if (ret == -ENOENT) {
3555 				page_idx = f2fs_get_next_page_offset(&dn,
3556 								page_idx);
3557 				ret = 0;
3558 				continue;
3559 			}
3560 			break;
3561 		}
3562 
3563 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3564 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3565 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3566 
3567 		ret = release_compress_blocks(&dn, count);
3568 
3569 		f2fs_put_dnode(&dn);
3570 
3571 		if (ret < 0)
3572 			break;
3573 
3574 		page_idx += count;
3575 		released_blocks += ret;
3576 	}
3577 
3578 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3579 	up_write(&F2FS_I(inode)->i_mmap_sem);
3580 out:
3581 	inode_unlock(inode);
3582 
3583 	mnt_drop_write_file(filp);
3584 
3585 	if (ret >= 0) {
3586 		ret = put_user(released_blocks, (u64 __user *)arg);
3587 	} else if (released_blocks && F2FS_I(inode)->i_compr_blocks) {
3588 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3589 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3590 			"iblocks=%llu, released=%u, compr_blocks=%llu, "
3591 			"run fsck to fix.",
3592 			__func__, inode->i_ino, inode->i_blocks,
3593 			released_blocks,
3594 			F2FS_I(inode)->i_compr_blocks);
3595 	}
3596 
3597 	return ret;
3598 }
3599 
3600 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3601 {
3602 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3603 	unsigned int reserved_blocks = 0;
3604 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3605 	block_t blkaddr;
3606 	int i;
3607 
3608 	for (i = 0; i < count; i++) {
3609 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3610 						dn->ofs_in_node + i);
3611 
3612 		if (!__is_valid_data_blkaddr(blkaddr))
3613 			continue;
3614 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3615 					DATA_GENERIC_ENHANCE)))
3616 			return -EFSCORRUPTED;
3617 	}
3618 
3619 	while (count) {
3620 		int compr_blocks = 0;
3621 		blkcnt_t reserved;
3622 		int ret;
3623 
3624 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3625 			blkaddr = f2fs_data_blkaddr(dn);
3626 
3627 			if (i == 0) {
3628 				if (blkaddr == COMPRESS_ADDR)
3629 					continue;
3630 				dn->ofs_in_node += cluster_size;
3631 				goto next;
3632 			}
3633 
3634 			if (__is_valid_data_blkaddr(blkaddr)) {
3635 				compr_blocks++;
3636 				continue;
3637 			}
3638 
3639 			dn->data_blkaddr = NEW_ADDR;
3640 			f2fs_set_data_blkaddr(dn);
3641 		}
3642 
3643 		reserved = cluster_size - compr_blocks;
3644 		ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3645 		if (ret)
3646 			return ret;
3647 
3648 		if (reserved != cluster_size - compr_blocks)
3649 			return -ENOSPC;
3650 
3651 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3652 
3653 		reserved_blocks += reserved;
3654 next:
3655 		count -= cluster_size;
3656 	}
3657 
3658 	return reserved_blocks;
3659 }
3660 
3661 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3662 {
3663 	struct inode *inode = file_inode(filp);
3664 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3665 	pgoff_t page_idx = 0, last_idx;
3666 	unsigned int reserved_blocks = 0;
3667 	int ret;
3668 
3669 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3670 		return -EOPNOTSUPP;
3671 
3672 	if (!f2fs_compressed_file(inode))
3673 		return -EINVAL;
3674 
3675 	if (f2fs_readonly(sbi->sb))
3676 		return -EROFS;
3677 
3678 	ret = mnt_want_write_file(filp);
3679 	if (ret)
3680 		return ret;
3681 
3682 	if (F2FS_I(inode)->i_compr_blocks)
3683 		goto out;
3684 
3685 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3686 
3687 	inode_lock(inode);
3688 
3689 	if (!IS_IMMUTABLE(inode)) {
3690 		ret = -EINVAL;
3691 		goto unlock_inode;
3692 	}
3693 
3694 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3695 	down_write(&F2FS_I(inode)->i_mmap_sem);
3696 
3697 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3698 
3699 	while (page_idx < last_idx) {
3700 		struct dnode_of_data dn;
3701 		pgoff_t end_offset, count;
3702 
3703 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3704 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3705 		if (ret) {
3706 			if (ret == -ENOENT) {
3707 				page_idx = f2fs_get_next_page_offset(&dn,
3708 								page_idx);
3709 				ret = 0;
3710 				continue;
3711 			}
3712 			break;
3713 		}
3714 
3715 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3716 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3717 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3718 
3719 		ret = reserve_compress_blocks(&dn, count);
3720 
3721 		f2fs_put_dnode(&dn);
3722 
3723 		if (ret < 0)
3724 			break;
3725 
3726 		page_idx += count;
3727 		reserved_blocks += ret;
3728 	}
3729 
3730 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3731 	up_write(&F2FS_I(inode)->i_mmap_sem);
3732 
3733 	if (ret >= 0) {
3734 		F2FS_I(inode)->i_flags &= ~F2FS_IMMUTABLE_FL;
3735 		f2fs_set_inode_flags(inode);
3736 		inode->i_ctime = current_time(inode);
3737 		f2fs_mark_inode_dirty_sync(inode, true);
3738 	}
3739 unlock_inode:
3740 	inode_unlock(inode);
3741 out:
3742 	mnt_drop_write_file(filp);
3743 
3744 	if (ret >= 0) {
3745 		ret = put_user(reserved_blocks, (u64 __user *)arg);
3746 	} else if (reserved_blocks && F2FS_I(inode)->i_compr_blocks) {
3747 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3748 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3749 			"iblocks=%llu, reserved=%u, compr_blocks=%llu, "
3750 			"run fsck to fix.",
3751 			__func__, inode->i_ino, inode->i_blocks,
3752 			reserved_blocks,
3753 			F2FS_I(inode)->i_compr_blocks);
3754 	}
3755 
3756 	return ret;
3757 }
3758 
3759 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3760 {
3761 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3762 		return -EIO;
3763 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
3764 		return -ENOSPC;
3765 
3766 	switch (cmd) {
3767 	case F2FS_IOC_GETFLAGS:
3768 		return f2fs_ioc_getflags(filp, arg);
3769 	case F2FS_IOC_SETFLAGS:
3770 		return f2fs_ioc_setflags(filp, arg);
3771 	case F2FS_IOC_GETVERSION:
3772 		return f2fs_ioc_getversion(filp, arg);
3773 	case F2FS_IOC_START_ATOMIC_WRITE:
3774 		return f2fs_ioc_start_atomic_write(filp);
3775 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3776 		return f2fs_ioc_commit_atomic_write(filp);
3777 	case F2FS_IOC_START_VOLATILE_WRITE:
3778 		return f2fs_ioc_start_volatile_write(filp);
3779 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3780 		return f2fs_ioc_release_volatile_write(filp);
3781 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3782 		return f2fs_ioc_abort_volatile_write(filp);
3783 	case F2FS_IOC_SHUTDOWN:
3784 		return f2fs_ioc_shutdown(filp, arg);
3785 	case FITRIM:
3786 		return f2fs_ioc_fitrim(filp, arg);
3787 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3788 		return f2fs_ioc_set_encryption_policy(filp, arg);
3789 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3790 		return f2fs_ioc_get_encryption_policy(filp, arg);
3791 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3792 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3793 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3794 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3795 	case FS_IOC_ADD_ENCRYPTION_KEY:
3796 		return f2fs_ioc_add_encryption_key(filp, arg);
3797 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3798 		return f2fs_ioc_remove_encryption_key(filp, arg);
3799 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3800 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3801 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3802 		return f2fs_ioc_get_encryption_key_status(filp, arg);
3803 	case FS_IOC_GET_ENCRYPTION_NONCE:
3804 		return f2fs_ioc_get_encryption_nonce(filp, arg);
3805 	case F2FS_IOC_GARBAGE_COLLECT:
3806 		return f2fs_ioc_gc(filp, arg);
3807 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3808 		return f2fs_ioc_gc_range(filp, arg);
3809 	case F2FS_IOC_WRITE_CHECKPOINT:
3810 		return f2fs_ioc_write_checkpoint(filp, arg);
3811 	case F2FS_IOC_DEFRAGMENT:
3812 		return f2fs_ioc_defragment(filp, arg);
3813 	case F2FS_IOC_MOVE_RANGE:
3814 		return f2fs_ioc_move_range(filp, arg);
3815 	case F2FS_IOC_FLUSH_DEVICE:
3816 		return f2fs_ioc_flush_device(filp, arg);
3817 	case F2FS_IOC_GET_FEATURES:
3818 		return f2fs_ioc_get_features(filp, arg);
3819 	case F2FS_IOC_FSGETXATTR:
3820 		return f2fs_ioc_fsgetxattr(filp, arg);
3821 	case F2FS_IOC_FSSETXATTR:
3822 		return f2fs_ioc_fssetxattr(filp, arg);
3823 	case F2FS_IOC_GET_PIN_FILE:
3824 		return f2fs_ioc_get_pin_file(filp, arg);
3825 	case F2FS_IOC_SET_PIN_FILE:
3826 		return f2fs_ioc_set_pin_file(filp, arg);
3827 	case F2FS_IOC_PRECACHE_EXTENTS:
3828 		return f2fs_ioc_precache_extents(filp, arg);
3829 	case F2FS_IOC_RESIZE_FS:
3830 		return f2fs_ioc_resize_fs(filp, arg);
3831 	case FS_IOC_ENABLE_VERITY:
3832 		return f2fs_ioc_enable_verity(filp, arg);
3833 	case FS_IOC_MEASURE_VERITY:
3834 		return f2fs_ioc_measure_verity(filp, arg);
3835 	case F2FS_IOC_GET_VOLUME_NAME:
3836 		return f2fs_get_volume_name(filp, arg);
3837 	case F2FS_IOC_SET_VOLUME_NAME:
3838 		return f2fs_set_volume_name(filp, arg);
3839 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
3840 		return f2fs_get_compress_blocks(filp, arg);
3841 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
3842 		return f2fs_release_compress_blocks(filp, arg);
3843 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
3844 		return f2fs_reserve_compress_blocks(filp, arg);
3845 	default:
3846 		return -ENOTTY;
3847 	}
3848 }
3849 
3850 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
3851 {
3852 	struct file *file = iocb->ki_filp;
3853 	struct inode *inode = file_inode(file);
3854 	int ret;
3855 
3856 	if (!f2fs_is_compress_backend_ready(inode))
3857 		return -EOPNOTSUPP;
3858 
3859 	ret = generic_file_read_iter(iocb, iter);
3860 
3861 	if (ret > 0)
3862 		f2fs_update_iostat(F2FS_I_SB(inode), APP_READ_IO, ret);
3863 
3864 	return ret;
3865 }
3866 
3867 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3868 {
3869 	struct file *file = iocb->ki_filp;
3870 	struct inode *inode = file_inode(file);
3871 	ssize_t ret;
3872 
3873 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3874 		ret = -EIO;
3875 		goto out;
3876 	}
3877 
3878 	if (!f2fs_is_compress_backend_ready(inode)) {
3879 		ret = -EOPNOTSUPP;
3880 		goto out;
3881 	}
3882 
3883 	if (iocb->ki_flags & IOCB_NOWAIT) {
3884 		if (!inode_trylock(inode)) {
3885 			ret = -EAGAIN;
3886 			goto out;
3887 		}
3888 	} else {
3889 		inode_lock(inode);
3890 	}
3891 
3892 	ret = generic_write_checks(iocb, from);
3893 	if (ret > 0) {
3894 		bool preallocated = false;
3895 		size_t target_size = 0;
3896 		int err;
3897 
3898 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3899 			set_inode_flag(inode, FI_NO_PREALLOC);
3900 
3901 		if ((iocb->ki_flags & IOCB_NOWAIT)) {
3902 			if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3903 						iov_iter_count(from)) ||
3904 				f2fs_has_inline_data(inode) ||
3905 				f2fs_force_buffered_io(inode, iocb, from)) {
3906 				clear_inode_flag(inode, FI_NO_PREALLOC);
3907 				inode_unlock(inode);
3908 				ret = -EAGAIN;
3909 				goto out;
3910 			}
3911 			goto write;
3912 		}
3913 
3914 		if (is_inode_flag_set(inode, FI_NO_PREALLOC))
3915 			goto write;
3916 
3917 		if (iocb->ki_flags & IOCB_DIRECT) {
3918 			/*
3919 			 * Convert inline data for Direct I/O before entering
3920 			 * f2fs_direct_IO().
3921 			 */
3922 			err = f2fs_convert_inline_inode(inode);
3923 			if (err)
3924 				goto out_err;
3925 			/*
3926 			 * If force_buffere_io() is true, we have to allocate
3927 			 * blocks all the time, since f2fs_direct_IO will fall
3928 			 * back to buffered IO.
3929 			 */
3930 			if (!f2fs_force_buffered_io(inode, iocb, from) &&
3931 					allow_outplace_dio(inode, iocb, from))
3932 				goto write;
3933 		}
3934 		preallocated = true;
3935 		target_size = iocb->ki_pos + iov_iter_count(from);
3936 
3937 		err = f2fs_preallocate_blocks(iocb, from);
3938 		if (err) {
3939 out_err:
3940 			clear_inode_flag(inode, FI_NO_PREALLOC);
3941 			inode_unlock(inode);
3942 			ret = err;
3943 			goto out;
3944 		}
3945 write:
3946 		ret = __generic_file_write_iter(iocb, from);
3947 		clear_inode_flag(inode, FI_NO_PREALLOC);
3948 
3949 		/* if we couldn't write data, we should deallocate blocks. */
3950 		if (preallocated && i_size_read(inode) < target_size)
3951 			f2fs_truncate(inode);
3952 
3953 		if (ret > 0)
3954 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3955 	}
3956 	inode_unlock(inode);
3957 out:
3958 	trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3959 					iov_iter_count(from), ret);
3960 	if (ret > 0)
3961 		ret = generic_write_sync(iocb, ret);
3962 	return ret;
3963 }
3964 
3965 #ifdef CONFIG_COMPAT
3966 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3967 {
3968 	switch (cmd) {
3969 	case F2FS_IOC32_GETFLAGS:
3970 		cmd = F2FS_IOC_GETFLAGS;
3971 		break;
3972 	case F2FS_IOC32_SETFLAGS:
3973 		cmd = F2FS_IOC_SETFLAGS;
3974 		break;
3975 	case F2FS_IOC32_GETVERSION:
3976 		cmd = F2FS_IOC_GETVERSION;
3977 		break;
3978 	case F2FS_IOC_START_ATOMIC_WRITE:
3979 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3980 	case F2FS_IOC_START_VOLATILE_WRITE:
3981 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3982 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3983 	case F2FS_IOC_SHUTDOWN:
3984 	case FITRIM:
3985 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3986 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3987 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3988 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3989 	case FS_IOC_ADD_ENCRYPTION_KEY:
3990 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3991 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3992 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3993 	case FS_IOC_GET_ENCRYPTION_NONCE:
3994 	case F2FS_IOC_GARBAGE_COLLECT:
3995 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3996 	case F2FS_IOC_WRITE_CHECKPOINT:
3997 	case F2FS_IOC_DEFRAGMENT:
3998 	case F2FS_IOC_MOVE_RANGE:
3999 	case F2FS_IOC_FLUSH_DEVICE:
4000 	case F2FS_IOC_GET_FEATURES:
4001 	case F2FS_IOC_FSGETXATTR:
4002 	case F2FS_IOC_FSSETXATTR:
4003 	case F2FS_IOC_GET_PIN_FILE:
4004 	case F2FS_IOC_SET_PIN_FILE:
4005 	case F2FS_IOC_PRECACHE_EXTENTS:
4006 	case F2FS_IOC_RESIZE_FS:
4007 	case FS_IOC_ENABLE_VERITY:
4008 	case FS_IOC_MEASURE_VERITY:
4009 	case F2FS_IOC_GET_VOLUME_NAME:
4010 	case F2FS_IOC_SET_VOLUME_NAME:
4011 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4012 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4013 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4014 		break;
4015 	default:
4016 		return -ENOIOCTLCMD;
4017 	}
4018 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4019 }
4020 #endif
4021 
4022 const struct file_operations f2fs_file_operations = {
4023 	.llseek		= f2fs_llseek,
4024 	.read_iter	= f2fs_file_read_iter,
4025 	.write_iter	= f2fs_file_write_iter,
4026 	.open		= f2fs_file_open,
4027 	.release	= f2fs_release_file,
4028 	.mmap		= f2fs_file_mmap,
4029 	.flush		= f2fs_file_flush,
4030 	.fsync		= f2fs_sync_file,
4031 	.fallocate	= f2fs_fallocate,
4032 	.unlocked_ioctl	= f2fs_ioctl,
4033 #ifdef CONFIG_COMPAT
4034 	.compat_ioctl	= f2fs_compat_ioctl,
4035 #endif
4036 	.splice_read	= generic_file_splice_read,
4037 	.splice_write	= iter_file_splice_write,
4038 };
4039