xref: /linux/fs/f2fs/file.c (revision 0c00ed308d0559fc216be0442a3df124e9e13533)
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/blk-crypto.h>
9 #include <linux/fs.h>
10 #include <linux/f2fs_fs.h>
11 #include <linux/stat.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/filelock.h>
16 #include <linux/types.h>
17 #include <linux/compat.h>
18 #include <linux/uaccess.h>
19 #include <linux/mount.h>
20 #include <linux/pagevec.h>
21 #include <linux/uio.h>
22 #include <linux/uuid.h>
23 #include <linux/file.h>
24 #include <linux/nls.h>
25 #include <linux/sched/signal.h>
26 #include <linux/fileattr.h>
27 #include <linux/fadvise.h>
28 #include <linux/iomap.h>
29 
30 #include "f2fs.h"
31 #include "node.h"
32 #include "segment.h"
33 #include "xattr.h"
34 #include "acl.h"
35 #include "gc.h"
36 #include "iostat.h"
37 #include <trace/events/f2fs.h>
38 #include <uapi/linux/f2fs.h>
39 
40 static void f2fs_zero_post_eof_page(struct inode *inode,
41 					loff_t new_size, bool lock)
42 {
43 	loff_t old_size = i_size_read(inode);
44 
45 	if (old_size >= new_size)
46 		return;
47 
48 	if (mapping_empty(inode->i_mapping))
49 		return;
50 
51 	if (lock)
52 		filemap_invalidate_lock(inode->i_mapping);
53 	/* zero or drop pages only in range of [old_size, new_size] */
54 	truncate_inode_pages_range(inode->i_mapping, old_size, new_size);
55 	if (lock)
56 		filemap_invalidate_unlock(inode->i_mapping);
57 }
58 
59 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
60 {
61 	struct inode *inode = file_inode(vmf->vma->vm_file);
62 	vm_flags_t flags = vmf->vma->vm_flags;
63 	vm_fault_t ret;
64 
65 	ret = filemap_fault(vmf);
66 	if (ret & VM_FAULT_LOCKED)
67 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
68 					APP_MAPPED_READ_IO, F2FS_BLKSIZE);
69 
70 	trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret);
71 
72 	return ret;
73 }
74 
75 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
76 {
77 	struct folio *folio = page_folio(vmf->page);
78 	struct inode *inode = file_inode(vmf->vma->vm_file);
79 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
80 	struct dnode_of_data dn;
81 	bool need_alloc = !f2fs_is_pinned_file(inode);
82 	int err = 0;
83 	vm_fault_t ret;
84 
85 	if (unlikely(IS_IMMUTABLE(inode)))
86 		return VM_FAULT_SIGBUS;
87 
88 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
89 		err = -EIO;
90 		goto out;
91 	}
92 
93 	if (unlikely(f2fs_cp_error(sbi))) {
94 		err = -EIO;
95 		goto out;
96 	}
97 
98 	if (!f2fs_is_checkpoint_ready(sbi)) {
99 		err = -ENOSPC;
100 		goto out;
101 	}
102 
103 	err = f2fs_convert_inline_inode(inode);
104 	if (err)
105 		goto out;
106 
107 #ifdef CONFIG_F2FS_FS_COMPRESSION
108 	if (f2fs_compressed_file(inode)) {
109 		int ret = f2fs_is_compressed_cluster(inode, folio->index);
110 
111 		if (ret < 0) {
112 			err = ret;
113 			goto out;
114 		} else if (ret) {
115 			need_alloc = false;
116 		}
117 	}
118 #endif
119 	/* should do out of any locked page */
120 	if (need_alloc)
121 		f2fs_balance_fs(sbi, true);
122 
123 	sb_start_pagefault(inode->i_sb);
124 
125 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
126 
127 	f2fs_zero_post_eof_page(inode, (folio->index + 1) << PAGE_SHIFT, true);
128 
129 	file_update_time(vmf->vma->vm_file);
130 	filemap_invalidate_lock_shared(inode->i_mapping);
131 
132 	folio_lock(folio);
133 	if (unlikely(folio->mapping != inode->i_mapping ||
134 			folio_pos(folio) > i_size_read(inode) ||
135 			!folio_test_uptodate(folio))) {
136 		folio_unlock(folio);
137 		err = -EFAULT;
138 		goto out_sem;
139 	}
140 
141 	set_new_dnode(&dn, inode, NULL, NULL, 0);
142 	if (need_alloc) {
143 		/* block allocation */
144 		err = f2fs_get_block_locked(&dn, folio->index);
145 	} else {
146 		err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE);
147 		f2fs_put_dnode(&dn);
148 		if (f2fs_is_pinned_file(inode) &&
149 		    !__is_valid_data_blkaddr(dn.data_blkaddr))
150 			err = -EIO;
151 	}
152 
153 	if (err) {
154 		folio_unlock(folio);
155 		goto out_sem;
156 	}
157 
158 	f2fs_folio_wait_writeback(folio, DATA, false, true);
159 
160 	/* wait for GCed page writeback via META_MAPPING */
161 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
162 
163 	/*
164 	 * check to see if the page is mapped already (no holes)
165 	 */
166 	if (folio_test_mappedtodisk(folio))
167 		goto out_sem;
168 
169 	/* page is wholly or partially inside EOF */
170 	if (((loff_t)(folio->index + 1) << PAGE_SHIFT) >
171 						i_size_read(inode)) {
172 		loff_t offset;
173 
174 		offset = i_size_read(inode) & ~PAGE_MASK;
175 		folio_zero_segment(folio, offset, folio_size(folio));
176 	}
177 	folio_mark_dirty(folio);
178 
179 	f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
180 	f2fs_update_time(sbi, REQ_TIME);
181 
182 out_sem:
183 	filemap_invalidate_unlock_shared(inode->i_mapping);
184 
185 	sb_end_pagefault(inode->i_sb);
186 out:
187 	ret = vmf_fs_error(err);
188 
189 	trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret);
190 	return ret;
191 }
192 
193 static const struct vm_operations_struct f2fs_file_vm_ops = {
194 	.fault		= f2fs_filemap_fault,
195 	.map_pages	= filemap_map_pages,
196 	.page_mkwrite	= f2fs_vm_page_mkwrite,
197 };
198 
199 static int get_parent_ino(struct inode *inode, nid_t *pino)
200 {
201 	struct dentry *dentry;
202 
203 	/*
204 	 * Make sure to get the non-deleted alias.  The alias associated with
205 	 * the open file descriptor being fsync()'ed may be deleted already.
206 	 */
207 	dentry = d_find_alias(inode);
208 	if (!dentry)
209 		return 0;
210 
211 	*pino = d_parent_ino(dentry);
212 	dput(dentry);
213 	return 1;
214 }
215 
216 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
217 {
218 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
219 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
220 
221 	if (!S_ISREG(inode->i_mode))
222 		cp_reason = CP_NON_REGULAR;
223 	else if (f2fs_compressed_file(inode))
224 		cp_reason = CP_COMPRESSED;
225 	else if (inode->i_nlink != 1)
226 		cp_reason = CP_HARDLINK;
227 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
228 		cp_reason = CP_SB_NEED_CP;
229 	else if (file_wrong_pino(inode))
230 		cp_reason = CP_WRONG_PINO;
231 	else if (!f2fs_space_for_roll_forward(sbi))
232 		cp_reason = CP_NO_SPC_ROLL;
233 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
234 		cp_reason = CP_NODE_NEED_CP;
235 	else if (test_opt(sbi, FASTBOOT))
236 		cp_reason = CP_FASTBOOT_MODE;
237 	else if (F2FS_OPTION(sbi).active_logs == 2)
238 		cp_reason = CP_SPEC_LOG_NUM;
239 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
240 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
241 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
242 							TRANS_DIR_INO))
243 		cp_reason = CP_RECOVER_DIR;
244 	else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
245 							XATTR_DIR_INO))
246 		cp_reason = CP_XATTR_DIR;
247 
248 	return cp_reason;
249 }
250 
251 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
252 {
253 	struct folio *i = filemap_get_folio(NODE_MAPPING(sbi), ino);
254 	bool ret = false;
255 	/* But we need to avoid that there are some inode updates */
256 	if ((!IS_ERR(i) && folio_test_dirty(i)) ||
257 	    f2fs_need_inode_block_update(sbi, ino))
258 		ret = true;
259 	f2fs_folio_put(i, false);
260 	return ret;
261 }
262 
263 static void try_to_fix_pino(struct inode *inode)
264 {
265 	struct f2fs_inode_info *fi = F2FS_I(inode);
266 	nid_t pino;
267 
268 	f2fs_down_write(&fi->i_sem);
269 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
270 			get_parent_ino(inode, &pino)) {
271 		f2fs_i_pino_write(inode, pino);
272 		file_got_pino(inode);
273 	}
274 	f2fs_up_write(&fi->i_sem);
275 }
276 
277 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
278 						int datasync, bool atomic)
279 {
280 	struct inode *inode = file->f_mapping->host;
281 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
282 	nid_t ino = inode->i_ino;
283 	int ret = 0;
284 	enum cp_reason_type cp_reason = 0;
285 	struct writeback_control wbc = {
286 		.sync_mode = WB_SYNC_ALL,
287 		.nr_to_write = LONG_MAX,
288 	};
289 	unsigned int seq_id = 0;
290 
291 	if (unlikely(f2fs_readonly(inode->i_sb)))
292 		return 0;
293 
294 	trace_f2fs_sync_file_enter(inode);
295 
296 	if (S_ISDIR(inode->i_mode))
297 		goto go_write;
298 
299 	/* if fdatasync is triggered, let's do in-place-update */
300 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
301 		set_inode_flag(inode, FI_NEED_IPU);
302 	ret = file_write_and_wait_range(file, start, end);
303 	clear_inode_flag(inode, FI_NEED_IPU);
304 
305 	if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
306 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
307 		return ret;
308 	}
309 
310 	/* if the inode is dirty, let's recover all the time */
311 	if (!f2fs_skip_inode_update(inode, datasync)) {
312 		f2fs_write_inode(inode, NULL);
313 		goto go_write;
314 	}
315 
316 	/*
317 	 * if there is no written data, don't waste time to write recovery info.
318 	 */
319 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
320 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
321 
322 		/* it may call write_inode just prior to fsync */
323 		if (need_inode_page_update(sbi, ino))
324 			goto go_write;
325 
326 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
327 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
328 			goto flush_out;
329 		goto out;
330 	} else {
331 		/*
332 		 * for OPU case, during fsync(), node can be persisted before
333 		 * data when lower device doesn't support write barrier, result
334 		 * in data corruption after SPO.
335 		 * So for strict fsync mode, force to use atomic write semantics
336 		 * to keep write order in between data/node and last node to
337 		 * avoid potential data corruption.
338 		 */
339 		if (F2FS_OPTION(sbi).fsync_mode ==
340 				FSYNC_MODE_STRICT && !atomic)
341 			atomic = true;
342 	}
343 go_write:
344 	/*
345 	 * Both of fdatasync() and fsync() are able to be recovered from
346 	 * sudden-power-off.
347 	 */
348 	f2fs_down_read(&F2FS_I(inode)->i_sem);
349 	cp_reason = need_do_checkpoint(inode);
350 	f2fs_up_read(&F2FS_I(inode)->i_sem);
351 
352 	if (cp_reason) {
353 		/* all the dirty node pages should be flushed for POR */
354 		ret = f2fs_sync_fs(inode->i_sb, 1);
355 
356 		/*
357 		 * We've secured consistency through sync_fs. Following pino
358 		 * will be used only for fsynced inodes after checkpoint.
359 		 */
360 		try_to_fix_pino(inode);
361 		clear_inode_flag(inode, FI_APPEND_WRITE);
362 		clear_inode_flag(inode, FI_UPDATE_WRITE);
363 		goto out;
364 	}
365 sync_nodes:
366 	atomic_inc(&sbi->wb_sync_req[NODE]);
367 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
368 	atomic_dec(&sbi->wb_sync_req[NODE]);
369 	if (ret)
370 		goto out;
371 
372 	/* if cp_error was enabled, we should avoid infinite loop */
373 	if (unlikely(f2fs_cp_error(sbi))) {
374 		ret = -EIO;
375 		goto out;
376 	}
377 
378 	if (f2fs_need_inode_block_update(sbi, ino)) {
379 		f2fs_mark_inode_dirty_sync(inode, true);
380 		f2fs_write_inode(inode, NULL);
381 		goto sync_nodes;
382 	}
383 
384 	/*
385 	 * If it's atomic_write, it's just fine to keep write ordering. So
386 	 * here we don't need to wait for node write completion, since we use
387 	 * node chain which serializes node blocks. If one of node writes are
388 	 * reordered, we can see simply broken chain, resulting in stopping
389 	 * roll-forward recovery. It means we'll recover all or none node blocks
390 	 * given fsync mark.
391 	 */
392 	if (!atomic) {
393 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
394 		if (ret)
395 			goto out;
396 	}
397 
398 	/* once recovery info is written, don't need to tack this */
399 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
400 	clear_inode_flag(inode, FI_APPEND_WRITE);
401 flush_out:
402 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
403 		ret = f2fs_issue_flush(sbi, inode->i_ino);
404 	if (!ret) {
405 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
406 		clear_inode_flag(inode, FI_UPDATE_WRITE);
407 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
408 	}
409 	f2fs_update_time(sbi, REQ_TIME);
410 out:
411 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
412 	return ret;
413 }
414 
415 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
416 {
417 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
418 		return -EIO;
419 	return f2fs_do_sync_file(file, start, end, datasync, false);
420 }
421 
422 static bool __found_offset(struct address_space *mapping,
423 		struct dnode_of_data *dn, pgoff_t index, int whence)
424 {
425 	block_t blkaddr = f2fs_data_blkaddr(dn);
426 	struct inode *inode = mapping->host;
427 	bool compressed_cluster = false;
428 
429 	if (f2fs_compressed_file(inode)) {
430 		block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio,
431 		    ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size));
432 
433 		compressed_cluster = first_blkaddr == COMPRESS_ADDR;
434 	}
435 
436 	switch (whence) {
437 	case SEEK_DATA:
438 		if (__is_valid_data_blkaddr(blkaddr))
439 			return true;
440 		if (blkaddr == NEW_ADDR &&
441 		    xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
442 			return true;
443 		if (compressed_cluster)
444 			return true;
445 		break;
446 	case SEEK_HOLE:
447 		if (compressed_cluster)
448 			return false;
449 		if (blkaddr == NULL_ADDR)
450 			return true;
451 		break;
452 	}
453 	return false;
454 }
455 
456 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
457 {
458 	struct inode *inode = file->f_mapping->host;
459 	loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
460 	struct dnode_of_data dn;
461 	pgoff_t pgofs, end_offset;
462 	loff_t data_ofs = offset;
463 	loff_t isize;
464 	int err = 0;
465 
466 	inode_lock_shared(inode);
467 
468 	isize = i_size_read(inode);
469 	if (offset >= isize)
470 		goto fail;
471 
472 	/* handle inline data case */
473 	if (f2fs_has_inline_data(inode)) {
474 		if (whence == SEEK_HOLE) {
475 			data_ofs = isize;
476 			goto found;
477 		} else if (whence == SEEK_DATA) {
478 			data_ofs = offset;
479 			goto found;
480 		}
481 	}
482 
483 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
484 
485 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
486 		set_new_dnode(&dn, inode, NULL, NULL, 0);
487 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
488 		if (err && err != -ENOENT) {
489 			goto fail;
490 		} else if (err == -ENOENT) {
491 			/* direct node does not exists */
492 			if (whence == SEEK_DATA) {
493 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
494 				continue;
495 			} else {
496 				goto found;
497 			}
498 		}
499 
500 		end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
501 
502 		/* find data/hole in dnode block */
503 		for (; dn.ofs_in_node < end_offset;
504 				dn.ofs_in_node++, pgofs++,
505 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
506 			block_t blkaddr;
507 
508 			blkaddr = f2fs_data_blkaddr(&dn);
509 
510 			if (__is_valid_data_blkaddr(blkaddr) &&
511 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
512 					blkaddr, DATA_GENERIC_ENHANCE)) {
513 				f2fs_put_dnode(&dn);
514 				goto fail;
515 			}
516 
517 			if (__found_offset(file->f_mapping, &dn,
518 							pgofs, whence)) {
519 				f2fs_put_dnode(&dn);
520 				goto found;
521 			}
522 		}
523 		f2fs_put_dnode(&dn);
524 	}
525 
526 	if (whence == SEEK_DATA)
527 		goto fail;
528 found:
529 	if (whence == SEEK_HOLE && data_ofs > isize)
530 		data_ofs = isize;
531 	inode_unlock_shared(inode);
532 	return vfs_setpos(file, data_ofs, maxbytes);
533 fail:
534 	inode_unlock_shared(inode);
535 	return -ENXIO;
536 }
537 
538 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
539 {
540 	struct inode *inode = file->f_mapping->host;
541 	loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
542 
543 	switch (whence) {
544 	case SEEK_SET:
545 	case SEEK_CUR:
546 	case SEEK_END:
547 		return generic_file_llseek_size(file, offset, whence,
548 						maxbytes, i_size_read(inode));
549 	case SEEK_DATA:
550 	case SEEK_HOLE:
551 		if (offset < 0)
552 			return -ENXIO;
553 		return f2fs_seek_block(file, offset, whence);
554 	}
555 
556 	return -EINVAL;
557 }
558 
559 static int f2fs_file_mmap_prepare(struct vm_area_desc *desc)
560 {
561 	struct file *file = desc->file;
562 	struct inode *inode = file_inode(file);
563 
564 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
565 		return -EIO;
566 
567 	if (!f2fs_is_compress_backend_ready(inode))
568 		return -EOPNOTSUPP;
569 
570 	file_accessed(file);
571 	desc->vm_ops = &f2fs_file_vm_ops;
572 
573 	f2fs_down_read(&F2FS_I(inode)->i_sem);
574 	set_inode_flag(inode, FI_MMAP_FILE);
575 	f2fs_up_read(&F2FS_I(inode)->i_sem);
576 
577 	return 0;
578 }
579 
580 static int finish_preallocate_blocks(struct inode *inode)
581 {
582 	int ret = 0;
583 	bool opened;
584 
585 	f2fs_down_read(&F2FS_I(inode)->i_sem);
586 	opened = is_inode_flag_set(inode, FI_OPENED_FILE);
587 	f2fs_up_read(&F2FS_I(inode)->i_sem);
588 	if (opened)
589 		return 0;
590 
591 	inode_lock(inode);
592 	if (is_inode_flag_set(inode, FI_OPENED_FILE))
593 		goto out_unlock;
594 
595 	if (!file_should_truncate(inode))
596 		goto out_update;
597 
598 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
599 	filemap_invalidate_lock(inode->i_mapping);
600 
601 	truncate_setsize(inode, i_size_read(inode));
602 	ret = f2fs_truncate(inode);
603 
604 	filemap_invalidate_unlock(inode->i_mapping);
605 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
606 	if (ret)
607 		goto out_unlock;
608 
609 	file_dont_truncate(inode);
610 out_update:
611 	f2fs_down_write(&F2FS_I(inode)->i_sem);
612 	set_inode_flag(inode, FI_OPENED_FILE);
613 	f2fs_up_write(&F2FS_I(inode)->i_sem);
614 out_unlock:
615 	inode_unlock(inode);
616 	return ret;
617 }
618 
619 static int f2fs_file_open(struct inode *inode, struct file *filp)
620 {
621 	int err = fscrypt_file_open(inode, filp);
622 
623 	if (err)
624 		return err;
625 
626 	if (!f2fs_is_compress_backend_ready(inode))
627 		return -EOPNOTSUPP;
628 
629 	err = fsverity_file_open(inode, filp);
630 	if (err)
631 		return err;
632 
633 	filp->f_mode |= FMODE_NOWAIT;
634 	filp->f_mode |= FMODE_CAN_ODIRECT;
635 
636 	err = dquot_file_open(inode, filp);
637 	if (err)
638 		return err;
639 
640 	err = finish_preallocate_blocks(inode);
641 	if (!err)
642 		atomic_inc(&F2FS_I(inode)->open_count);
643 	return err;
644 }
645 
646 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
647 {
648 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
649 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
650 	__le32 *addr;
651 	bool compressed_cluster = false;
652 	int cluster_index = 0, valid_blocks = 0;
653 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
654 	bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
655 	block_t blkstart;
656 	int blklen = 0;
657 
658 	addr = get_dnode_addr(dn->inode, dn->node_folio) + ofs;
659 	blkstart = le32_to_cpu(*addr);
660 
661 	/* Assumption: truncation starts with cluster */
662 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
663 		block_t blkaddr = le32_to_cpu(*addr);
664 
665 		if (f2fs_compressed_file(dn->inode) &&
666 					!(cluster_index & (cluster_size - 1))) {
667 			if (compressed_cluster)
668 				f2fs_i_compr_blocks_update(dn->inode,
669 							valid_blocks, false);
670 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
671 			valid_blocks = 0;
672 		}
673 
674 		if (blkaddr == NULL_ADDR)
675 			goto next;
676 
677 		f2fs_set_data_blkaddr(dn, NULL_ADDR);
678 
679 		if (__is_valid_data_blkaddr(blkaddr)) {
680 			if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE))
681 				goto next;
682 			if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr,
683 						DATA_GENERIC_ENHANCE))
684 				goto next;
685 			if (compressed_cluster)
686 				valid_blocks++;
687 		}
688 
689 		if (blkstart + blklen == blkaddr) {
690 			blklen++;
691 		} else {
692 			f2fs_invalidate_blocks(sbi, blkstart, blklen);
693 			blkstart = blkaddr;
694 			blklen = 1;
695 		}
696 
697 		if (!released || blkaddr != COMPRESS_ADDR)
698 			nr_free++;
699 
700 		continue;
701 
702 next:
703 		if (blklen)
704 			f2fs_invalidate_blocks(sbi, blkstart, blklen);
705 
706 		blkstart = le32_to_cpu(*(addr + 1));
707 		blklen = 0;
708 	}
709 
710 	if (blklen)
711 		f2fs_invalidate_blocks(sbi, blkstart, blklen);
712 
713 	if (compressed_cluster)
714 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
715 
716 	if (nr_free) {
717 		pgoff_t fofs;
718 		/*
719 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
720 		 * we will invalidate all blkaddr in the whole range.
721 		 */
722 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_folio),
723 							dn->inode) + ofs;
724 		f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
725 		f2fs_update_age_extent_cache_range(dn, fofs, len);
726 		dec_valid_block_count(sbi, dn->inode, nr_free);
727 	}
728 	dn->ofs_in_node = ofs;
729 
730 	f2fs_update_time(sbi, REQ_TIME);
731 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
732 					 dn->ofs_in_node, nr_free);
733 }
734 
735 static int truncate_partial_data_page(struct inode *inode, u64 from,
736 								bool cache_only)
737 {
738 	loff_t offset = from & (PAGE_SIZE - 1);
739 	pgoff_t index = from >> PAGE_SHIFT;
740 	struct address_space *mapping = inode->i_mapping;
741 	struct folio *folio;
742 
743 	if (!offset && !cache_only)
744 		return 0;
745 
746 	if (cache_only) {
747 		folio = filemap_lock_folio(mapping, index);
748 		if (IS_ERR(folio))
749 		       return 0;
750 		if (folio_test_uptodate(folio))
751 			goto truncate_out;
752 		f2fs_folio_put(folio, true);
753 		return 0;
754 	}
755 
756 	folio = f2fs_get_lock_data_folio(inode, index, true);
757 	if (IS_ERR(folio))
758 		return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio);
759 truncate_out:
760 	f2fs_folio_wait_writeback(folio, DATA, true, true);
761 	folio_zero_segment(folio, offset, folio_size(folio));
762 
763 	/* An encrypted inode should have a key and truncate the last page. */
764 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
765 	if (!cache_only)
766 		folio_mark_dirty(folio);
767 	f2fs_folio_put(folio, true);
768 	return 0;
769 }
770 
771 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
772 {
773 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
774 	struct dnode_of_data dn;
775 	pgoff_t free_from;
776 	int count = 0, err = 0;
777 	struct folio *ifolio;
778 	bool truncate_page = false;
779 
780 	trace_f2fs_truncate_blocks_enter(inode, from);
781 
782 	if (IS_DEVICE_ALIASING(inode) && from) {
783 		err = -EINVAL;
784 		goto out_err;
785 	}
786 
787 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
788 
789 	if (free_from >= max_file_blocks(inode))
790 		goto free_partial;
791 
792 	if (lock)
793 		f2fs_lock_op(sbi);
794 
795 	ifolio = f2fs_get_inode_folio(sbi, inode->i_ino);
796 	if (IS_ERR(ifolio)) {
797 		err = PTR_ERR(ifolio);
798 		goto out;
799 	}
800 
801 	if (IS_DEVICE_ALIASING(inode)) {
802 		struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
803 		struct extent_info ei = et->largest;
804 
805 		f2fs_invalidate_blocks(sbi, ei.blk, ei.len);
806 
807 		dec_valid_block_count(sbi, inode, ei.len);
808 		f2fs_update_time(sbi, REQ_TIME);
809 
810 		f2fs_folio_put(ifolio, true);
811 		goto out;
812 	}
813 
814 	if (f2fs_has_inline_data(inode)) {
815 		f2fs_truncate_inline_inode(inode, ifolio, from);
816 		f2fs_folio_put(ifolio, true);
817 		truncate_page = true;
818 		goto out;
819 	}
820 
821 	set_new_dnode(&dn, inode, ifolio, NULL, 0);
822 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
823 	if (err) {
824 		if (err == -ENOENT)
825 			goto free_next;
826 		goto out;
827 	}
828 
829 	count = ADDRS_PER_PAGE(dn.node_folio, inode);
830 
831 	count -= dn.ofs_in_node;
832 	f2fs_bug_on(sbi, count < 0);
833 
834 	if (dn.ofs_in_node || IS_INODE(dn.node_folio)) {
835 		f2fs_truncate_data_blocks_range(&dn, count);
836 		free_from += count;
837 	}
838 
839 	f2fs_put_dnode(&dn);
840 free_next:
841 	err = f2fs_truncate_inode_blocks(inode, free_from);
842 out:
843 	if (lock)
844 		f2fs_unlock_op(sbi);
845 free_partial:
846 	/* lastly zero out the first data page */
847 	if (!err)
848 		err = truncate_partial_data_page(inode, from, truncate_page);
849 out_err:
850 	trace_f2fs_truncate_blocks_exit(inode, err);
851 	return err;
852 }
853 
854 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
855 {
856 	u64 free_from = from;
857 	int err;
858 
859 #ifdef CONFIG_F2FS_FS_COMPRESSION
860 	/*
861 	 * for compressed file, only support cluster size
862 	 * aligned truncation.
863 	 */
864 	if (f2fs_compressed_file(inode))
865 		free_from = round_up(from,
866 				F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
867 #endif
868 
869 	err = f2fs_do_truncate_blocks(inode, free_from, lock);
870 	if (err)
871 		return err;
872 
873 #ifdef CONFIG_F2FS_FS_COMPRESSION
874 	/*
875 	 * For compressed file, after release compress blocks, don't allow write
876 	 * direct, but we should allow write direct after truncate to zero.
877 	 */
878 	if (f2fs_compressed_file(inode) && !free_from
879 			&& is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
880 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
881 
882 	if (from != free_from) {
883 		err = f2fs_truncate_partial_cluster(inode, from, lock);
884 		if (err)
885 			return err;
886 	}
887 #endif
888 
889 	return 0;
890 }
891 
892 int f2fs_truncate(struct inode *inode)
893 {
894 	int err;
895 
896 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
897 		return -EIO;
898 
899 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
900 				S_ISLNK(inode->i_mode)))
901 		return 0;
902 
903 	trace_f2fs_truncate(inode);
904 
905 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
906 		return -EIO;
907 
908 	err = f2fs_dquot_initialize(inode);
909 	if (err)
910 		return err;
911 
912 	/* we should check inline_data size */
913 	if (!f2fs_may_inline_data(inode)) {
914 		err = f2fs_convert_inline_inode(inode);
915 		if (err) {
916 			/*
917 			 * Always truncate page #0 to avoid page cache
918 			 * leak in evict() path.
919 			 */
920 			truncate_inode_pages_range(inode->i_mapping,
921 					F2FS_BLK_TO_BYTES(0),
922 					F2FS_BLK_END_BYTES(0));
923 			return err;
924 		}
925 	}
926 
927 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
928 	if (err)
929 		return err;
930 
931 	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
932 	f2fs_mark_inode_dirty_sync(inode, false);
933 	return 0;
934 }
935 
936 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
937 {
938 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
939 
940 	if (!fscrypt_dio_supported(inode))
941 		return true;
942 	if (fsverity_active(inode))
943 		return true;
944 	if (f2fs_compressed_file(inode))
945 		return true;
946 	/*
947 	 * only force direct read to use buffered IO, for direct write,
948 	 * it expects inline data conversion before committing IO.
949 	 */
950 	if (f2fs_has_inline_data(inode) && rw == READ)
951 		return true;
952 
953 	/* disallow direct IO if any of devices has unaligned blksize */
954 	if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
955 		return true;
956 	/*
957 	 * for blkzoned device, fallback direct IO to buffered IO, so
958 	 * all IOs can be serialized by log-structured write.
959 	 */
960 	if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) &&
961 	    !f2fs_is_pinned_file(inode))
962 		return true;
963 	if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
964 		return true;
965 
966 	return false;
967 }
968 
969 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path,
970 		 struct kstat *stat, u32 request_mask, unsigned int query_flags)
971 {
972 	struct inode *inode = d_inode(path->dentry);
973 	struct f2fs_inode_info *fi = F2FS_I(inode);
974 	struct f2fs_inode *ri = NULL;
975 	unsigned int flags;
976 
977 	if (f2fs_has_extra_attr(inode) &&
978 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
979 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
980 		stat->result_mask |= STATX_BTIME;
981 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
982 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
983 	}
984 
985 	/*
986 	 * Return the DIO alignment restrictions if requested.  We only return
987 	 * this information when requested, since on encrypted files it might
988 	 * take a fair bit of work to get if the file wasn't opened recently.
989 	 *
990 	 * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
991 	 * cannot represent that, so in that case we report no DIO support.
992 	 */
993 	if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
994 		unsigned int bsize = i_blocksize(inode);
995 
996 		stat->result_mask |= STATX_DIOALIGN;
997 		if (!f2fs_force_buffered_io(inode, WRITE)) {
998 			stat->dio_mem_align = bsize;
999 			stat->dio_offset_align = bsize;
1000 		}
1001 	}
1002 
1003 	flags = fi->i_flags;
1004 	if (flags & F2FS_COMPR_FL)
1005 		stat->attributes |= STATX_ATTR_COMPRESSED;
1006 	if (flags & F2FS_APPEND_FL)
1007 		stat->attributes |= STATX_ATTR_APPEND;
1008 	if (IS_ENCRYPTED(inode))
1009 		stat->attributes |= STATX_ATTR_ENCRYPTED;
1010 	if (flags & F2FS_IMMUTABLE_FL)
1011 		stat->attributes |= STATX_ATTR_IMMUTABLE;
1012 	if (flags & F2FS_NODUMP_FL)
1013 		stat->attributes |= STATX_ATTR_NODUMP;
1014 	if (IS_VERITY(inode))
1015 		stat->attributes |= STATX_ATTR_VERITY;
1016 
1017 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
1018 				  STATX_ATTR_APPEND |
1019 				  STATX_ATTR_ENCRYPTED |
1020 				  STATX_ATTR_IMMUTABLE |
1021 				  STATX_ATTR_NODUMP |
1022 				  STATX_ATTR_VERITY);
1023 
1024 	generic_fillattr(idmap, request_mask, inode, stat);
1025 
1026 	/* we need to show initial sectors used for inline_data/dentries */
1027 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
1028 					f2fs_has_inline_dentry(inode))
1029 		stat->blocks += (stat->size + 511) >> 9;
1030 
1031 	return 0;
1032 }
1033 
1034 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1035 static void __setattr_copy(struct mnt_idmap *idmap,
1036 			   struct inode *inode, const struct iattr *attr)
1037 {
1038 	unsigned int ia_valid = attr->ia_valid;
1039 
1040 	i_uid_update(idmap, attr, inode);
1041 	i_gid_update(idmap, attr, inode);
1042 	if (ia_valid & ATTR_ATIME)
1043 		inode_set_atime_to_ts(inode, attr->ia_atime);
1044 	if (ia_valid & ATTR_MTIME)
1045 		inode_set_mtime_to_ts(inode, attr->ia_mtime);
1046 	if (ia_valid & ATTR_CTIME)
1047 		inode_set_ctime_to_ts(inode, attr->ia_ctime);
1048 	if (ia_valid & ATTR_MODE) {
1049 		umode_t mode = attr->ia_mode;
1050 
1051 		if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode)))
1052 			mode &= ~S_ISGID;
1053 		set_acl_inode(inode, mode);
1054 	}
1055 }
1056 #else
1057 #define __setattr_copy setattr_copy
1058 #endif
1059 
1060 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
1061 		 struct iattr *attr)
1062 {
1063 	struct inode *inode = d_inode(dentry);
1064 	struct f2fs_inode_info *fi = F2FS_I(inode);
1065 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1066 	int err;
1067 
1068 	if (unlikely(f2fs_cp_error(sbi)))
1069 		return -EIO;
1070 
1071 	err = setattr_prepare(idmap, dentry, attr);
1072 	if (err)
1073 		return err;
1074 
1075 	err = fscrypt_prepare_setattr(dentry, attr);
1076 	if (err)
1077 		return err;
1078 
1079 	err = fsverity_prepare_setattr(dentry, attr);
1080 	if (err)
1081 		return err;
1082 
1083 	if (unlikely(IS_IMMUTABLE(inode)))
1084 		return -EPERM;
1085 
1086 	if (unlikely(IS_APPEND(inode) &&
1087 			(attr->ia_valid & (ATTR_MODE | ATTR_UID |
1088 				  ATTR_GID | ATTR_TIMES_SET))))
1089 		return -EPERM;
1090 
1091 	if ((attr->ia_valid & ATTR_SIZE)) {
1092 		if (!f2fs_is_compress_backend_ready(inode) ||
1093 				IS_DEVICE_ALIASING(inode))
1094 			return -EOPNOTSUPP;
1095 		if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) &&
1096 			!IS_ALIGNED(attr->ia_size,
1097 			F2FS_BLK_TO_BYTES(fi->i_cluster_size)))
1098 			return -EINVAL;
1099 		/*
1100 		 * To prevent scattered pin block generation, we don't allow
1101 		 * smaller/equal size unaligned truncation for pinned file.
1102 		 * We only support overwrite IO to pinned file, so don't
1103 		 * care about larger size truncation.
1104 		 */
1105 		if (f2fs_is_pinned_file(inode) &&
1106 			attr->ia_size <= i_size_read(inode) &&
1107 			!IS_ALIGNED(attr->ia_size,
1108 			F2FS_BLK_TO_BYTES(CAP_BLKS_PER_SEC(sbi))))
1109 			return -EINVAL;
1110 	}
1111 
1112 	if (is_quota_modification(idmap, inode, attr)) {
1113 		err = f2fs_dquot_initialize(inode);
1114 		if (err)
1115 			return err;
1116 	}
1117 	if (i_uid_needs_update(idmap, attr, inode) ||
1118 	    i_gid_needs_update(idmap, attr, inode)) {
1119 		f2fs_lock_op(sbi);
1120 		err = dquot_transfer(idmap, inode, attr);
1121 		if (err) {
1122 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
1123 			f2fs_unlock_op(sbi);
1124 			return err;
1125 		}
1126 		/*
1127 		 * update uid/gid under lock_op(), so that dquot and inode can
1128 		 * be updated atomically.
1129 		 */
1130 		i_uid_update(idmap, attr, inode);
1131 		i_gid_update(idmap, attr, inode);
1132 		f2fs_mark_inode_dirty_sync(inode, true);
1133 		f2fs_unlock_op(sbi);
1134 	}
1135 
1136 	if (attr->ia_valid & ATTR_SIZE) {
1137 		loff_t old_size = i_size_read(inode);
1138 
1139 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
1140 			/*
1141 			 * should convert inline inode before i_size_write to
1142 			 * keep smaller than inline_data size with inline flag.
1143 			 */
1144 			err = f2fs_convert_inline_inode(inode);
1145 			if (err)
1146 				return err;
1147 		}
1148 
1149 		/*
1150 		 * wait for inflight dio, blocks should be removed after
1151 		 * IO completion.
1152 		 */
1153 		if (attr->ia_size < old_size)
1154 			inode_dio_wait(inode);
1155 
1156 		f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
1157 		filemap_invalidate_lock(inode->i_mapping);
1158 
1159 		if (attr->ia_size > old_size)
1160 			f2fs_zero_post_eof_page(inode, attr->ia_size, false);
1161 		truncate_setsize(inode, attr->ia_size);
1162 
1163 		if (attr->ia_size <= old_size)
1164 			err = f2fs_truncate(inode);
1165 		/*
1166 		 * do not trim all blocks after i_size if target size is
1167 		 * larger than i_size.
1168 		 */
1169 		filemap_invalidate_unlock(inode->i_mapping);
1170 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
1171 		if (err)
1172 			return err;
1173 
1174 		spin_lock(&fi->i_size_lock);
1175 		inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1176 		fi->last_disk_size = i_size_read(inode);
1177 		spin_unlock(&fi->i_size_lock);
1178 	}
1179 
1180 	__setattr_copy(idmap, inode, attr);
1181 
1182 	if (attr->ia_valid & ATTR_MODE) {
1183 		err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
1184 
1185 		if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1186 			if (!err)
1187 				inode->i_mode = fi->i_acl_mode;
1188 			clear_inode_flag(inode, FI_ACL_MODE);
1189 		}
1190 	}
1191 
1192 	/* file size may changed here */
1193 	f2fs_mark_inode_dirty_sync(inode, true);
1194 
1195 	/* inode change will produce dirty node pages flushed by checkpoint */
1196 	f2fs_balance_fs(sbi, true);
1197 
1198 	return err;
1199 }
1200 
1201 const struct inode_operations f2fs_file_inode_operations = {
1202 	.getattr	= f2fs_getattr,
1203 	.setattr	= f2fs_setattr,
1204 	.get_inode_acl	= f2fs_get_acl,
1205 	.set_acl	= f2fs_set_acl,
1206 	.listxattr	= f2fs_listxattr,
1207 	.fiemap		= f2fs_fiemap,
1208 	.fileattr_get	= f2fs_fileattr_get,
1209 	.fileattr_set	= f2fs_fileattr_set,
1210 };
1211 
1212 static int fill_zero(struct inode *inode, pgoff_t index,
1213 					loff_t start, loff_t len)
1214 {
1215 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1216 	struct folio *folio;
1217 
1218 	if (!len)
1219 		return 0;
1220 
1221 	f2fs_balance_fs(sbi, true);
1222 
1223 	f2fs_lock_op(sbi);
1224 	folio = f2fs_get_new_data_folio(inode, NULL, index, false);
1225 	f2fs_unlock_op(sbi);
1226 
1227 	if (IS_ERR(folio))
1228 		return PTR_ERR(folio);
1229 
1230 	f2fs_folio_wait_writeback(folio, DATA, true, true);
1231 	folio_zero_range(folio, start, len);
1232 	folio_mark_dirty(folio);
1233 	f2fs_folio_put(folio, true);
1234 	return 0;
1235 }
1236 
1237 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1238 {
1239 	int err;
1240 
1241 	while (pg_start < pg_end) {
1242 		struct dnode_of_data dn;
1243 		pgoff_t end_offset, count;
1244 
1245 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1246 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1247 		if (err) {
1248 			if (err == -ENOENT) {
1249 				pg_start = f2fs_get_next_page_offset(&dn,
1250 								pg_start);
1251 				continue;
1252 			}
1253 			return err;
1254 		}
1255 
1256 		end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
1257 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1258 
1259 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1260 
1261 		f2fs_truncate_data_blocks_range(&dn, count);
1262 		f2fs_put_dnode(&dn);
1263 
1264 		pg_start += count;
1265 	}
1266 	return 0;
1267 }
1268 
1269 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
1270 {
1271 	pgoff_t pg_start, pg_end;
1272 	loff_t off_start, off_end;
1273 	int ret;
1274 
1275 	ret = f2fs_convert_inline_inode(inode);
1276 	if (ret)
1277 		return ret;
1278 
1279 	f2fs_zero_post_eof_page(inode, offset + len, true);
1280 
1281 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1282 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1283 
1284 	off_start = offset & (PAGE_SIZE - 1);
1285 	off_end = (offset + len) & (PAGE_SIZE - 1);
1286 
1287 	if (pg_start == pg_end) {
1288 		ret = fill_zero(inode, pg_start, off_start,
1289 						off_end - off_start);
1290 		if (ret)
1291 			return ret;
1292 	} else {
1293 		if (off_start) {
1294 			ret = fill_zero(inode, pg_start++, off_start,
1295 						PAGE_SIZE - off_start);
1296 			if (ret)
1297 				return ret;
1298 		}
1299 		if (off_end) {
1300 			ret = fill_zero(inode, pg_end, 0, off_end);
1301 			if (ret)
1302 				return ret;
1303 		}
1304 
1305 		if (pg_start < pg_end) {
1306 			loff_t blk_start, blk_end;
1307 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1308 
1309 			f2fs_balance_fs(sbi, true);
1310 
1311 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1312 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1313 
1314 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1315 			filemap_invalidate_lock(inode->i_mapping);
1316 
1317 			truncate_pagecache_range(inode, blk_start, blk_end - 1);
1318 
1319 			f2fs_lock_op(sbi);
1320 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1321 			f2fs_unlock_op(sbi);
1322 
1323 			filemap_invalidate_unlock(inode->i_mapping);
1324 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1325 		}
1326 	}
1327 
1328 	return ret;
1329 }
1330 
1331 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1332 				int *do_replace, pgoff_t off, pgoff_t len)
1333 {
1334 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1335 	struct dnode_of_data dn;
1336 	int ret, done, i;
1337 
1338 next_dnode:
1339 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1340 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1341 	if (ret && ret != -ENOENT) {
1342 		return ret;
1343 	} else if (ret == -ENOENT) {
1344 		if (dn.max_level == 0)
1345 			return -ENOENT;
1346 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1347 						dn.ofs_in_node, len);
1348 		blkaddr += done;
1349 		do_replace += done;
1350 		goto next;
1351 	}
1352 
1353 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_folio, inode) -
1354 							dn.ofs_in_node, len);
1355 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1356 		*blkaddr = f2fs_data_blkaddr(&dn);
1357 
1358 		if (__is_valid_data_blkaddr(*blkaddr) &&
1359 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1360 					DATA_GENERIC_ENHANCE)) {
1361 			f2fs_put_dnode(&dn);
1362 			return -EFSCORRUPTED;
1363 		}
1364 
1365 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1366 
1367 			if (f2fs_lfs_mode(sbi)) {
1368 				f2fs_put_dnode(&dn);
1369 				return -EOPNOTSUPP;
1370 			}
1371 
1372 			/* do not invalidate this block address */
1373 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1374 			*do_replace = 1;
1375 		}
1376 	}
1377 	f2fs_put_dnode(&dn);
1378 next:
1379 	len -= done;
1380 	off += done;
1381 	if (len)
1382 		goto next_dnode;
1383 	return 0;
1384 }
1385 
1386 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1387 				int *do_replace, pgoff_t off, int len)
1388 {
1389 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1390 	struct dnode_of_data dn;
1391 	int ret, i;
1392 
1393 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1394 		if (*do_replace == 0)
1395 			continue;
1396 
1397 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1398 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1399 		if (ret) {
1400 			dec_valid_block_count(sbi, inode, 1);
1401 			f2fs_invalidate_blocks(sbi, *blkaddr, 1);
1402 		} else {
1403 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1404 		}
1405 		f2fs_put_dnode(&dn);
1406 	}
1407 	return 0;
1408 }
1409 
1410 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1411 			block_t *blkaddr, int *do_replace,
1412 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1413 {
1414 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1415 	pgoff_t i = 0;
1416 	int ret;
1417 
1418 	while (i < len) {
1419 		if (blkaddr[i] == NULL_ADDR && !full) {
1420 			i++;
1421 			continue;
1422 		}
1423 
1424 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1425 			struct dnode_of_data dn;
1426 			struct node_info ni;
1427 			size_t new_size;
1428 			pgoff_t ilen;
1429 
1430 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1431 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1432 			if (ret)
1433 				return ret;
1434 
1435 			ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1436 			if (ret) {
1437 				f2fs_put_dnode(&dn);
1438 				return ret;
1439 			}
1440 
1441 			ilen = min((pgoff_t)
1442 				ADDRS_PER_PAGE(dn.node_folio, dst_inode) -
1443 						dn.ofs_in_node, len - i);
1444 			do {
1445 				dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1446 				f2fs_truncate_data_blocks_range(&dn, 1);
1447 
1448 				if (do_replace[i]) {
1449 					f2fs_i_blocks_write(src_inode,
1450 							1, false, false);
1451 					f2fs_i_blocks_write(dst_inode,
1452 							1, true, false);
1453 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1454 					blkaddr[i], ni.version, true, false);
1455 
1456 					do_replace[i] = 0;
1457 				}
1458 				dn.ofs_in_node++;
1459 				i++;
1460 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1461 				if (dst_inode->i_size < new_size)
1462 					f2fs_i_size_write(dst_inode, new_size);
1463 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1464 
1465 			f2fs_put_dnode(&dn);
1466 		} else {
1467 			struct folio *fsrc, *fdst;
1468 
1469 			fsrc = f2fs_get_lock_data_folio(src_inode,
1470 							src + i, true);
1471 			if (IS_ERR(fsrc))
1472 				return PTR_ERR(fsrc);
1473 			fdst = f2fs_get_new_data_folio(dst_inode, NULL, dst + i,
1474 								true);
1475 			if (IS_ERR(fdst)) {
1476 				f2fs_folio_put(fsrc, true);
1477 				return PTR_ERR(fdst);
1478 			}
1479 
1480 			f2fs_folio_wait_writeback(fdst, DATA, true, true);
1481 
1482 			memcpy_folio(fdst, 0, fsrc, 0, PAGE_SIZE);
1483 			folio_mark_dirty(fdst);
1484 			folio_set_f2fs_gcing(fdst);
1485 			f2fs_folio_put(fdst, true);
1486 			f2fs_folio_put(fsrc, true);
1487 
1488 			ret = f2fs_truncate_hole(src_inode,
1489 						src + i, src + i + 1);
1490 			if (ret)
1491 				return ret;
1492 			i++;
1493 		}
1494 	}
1495 	return 0;
1496 }
1497 
1498 static int __exchange_data_block(struct inode *src_inode,
1499 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1500 			pgoff_t len, bool full)
1501 {
1502 	block_t *src_blkaddr;
1503 	int *do_replace;
1504 	pgoff_t olen;
1505 	int ret;
1506 
1507 	while (len) {
1508 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1509 
1510 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1511 					array_size(olen, sizeof(block_t)),
1512 					GFP_NOFS);
1513 		if (!src_blkaddr)
1514 			return -ENOMEM;
1515 
1516 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1517 					array_size(olen, sizeof(int)),
1518 					GFP_NOFS);
1519 		if (!do_replace) {
1520 			kvfree(src_blkaddr);
1521 			return -ENOMEM;
1522 		}
1523 
1524 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1525 					do_replace, src, olen);
1526 		if (ret)
1527 			goto roll_back;
1528 
1529 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1530 					do_replace, src, dst, olen, full);
1531 		if (ret)
1532 			goto roll_back;
1533 
1534 		src += olen;
1535 		dst += olen;
1536 		len -= olen;
1537 
1538 		kvfree(src_blkaddr);
1539 		kvfree(do_replace);
1540 	}
1541 	return 0;
1542 
1543 roll_back:
1544 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1545 	kvfree(src_blkaddr);
1546 	kvfree(do_replace);
1547 	return ret;
1548 }
1549 
1550 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1551 {
1552 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1553 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1554 	pgoff_t start = offset >> PAGE_SHIFT;
1555 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1556 	int ret;
1557 
1558 	f2fs_balance_fs(sbi, true);
1559 
1560 	/* avoid gc operation during block exchange */
1561 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1562 	filemap_invalidate_lock(inode->i_mapping);
1563 
1564 	f2fs_zero_post_eof_page(inode, offset + len, false);
1565 
1566 	f2fs_lock_op(sbi);
1567 	f2fs_drop_extent_tree(inode);
1568 	truncate_pagecache(inode, offset);
1569 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1570 	f2fs_unlock_op(sbi);
1571 
1572 	filemap_invalidate_unlock(inode->i_mapping);
1573 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1574 	return ret;
1575 }
1576 
1577 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1578 {
1579 	loff_t new_size;
1580 	int ret;
1581 
1582 	if (offset + len >= i_size_read(inode))
1583 		return -EINVAL;
1584 
1585 	/* collapse range should be aligned to block size of f2fs. */
1586 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1587 		return -EINVAL;
1588 
1589 	ret = f2fs_convert_inline_inode(inode);
1590 	if (ret)
1591 		return ret;
1592 
1593 	/* write out all dirty pages from offset */
1594 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1595 	if (ret)
1596 		return ret;
1597 
1598 	ret = f2fs_do_collapse(inode, offset, len);
1599 	if (ret)
1600 		return ret;
1601 
1602 	/* write out all moved pages, if possible */
1603 	filemap_invalidate_lock(inode->i_mapping);
1604 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1605 	truncate_pagecache(inode, offset);
1606 
1607 	new_size = i_size_read(inode) - len;
1608 	ret = f2fs_truncate_blocks(inode, new_size, true);
1609 	filemap_invalidate_unlock(inode->i_mapping);
1610 	if (!ret)
1611 		f2fs_i_size_write(inode, new_size);
1612 	return ret;
1613 }
1614 
1615 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1616 								pgoff_t end)
1617 {
1618 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1619 	pgoff_t index = start;
1620 	unsigned int ofs_in_node = dn->ofs_in_node;
1621 	blkcnt_t count = 0;
1622 	int ret;
1623 
1624 	for (; index < end; index++, dn->ofs_in_node++) {
1625 		if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1626 			count++;
1627 	}
1628 
1629 	dn->ofs_in_node = ofs_in_node;
1630 	ret = f2fs_reserve_new_blocks(dn, count);
1631 	if (ret)
1632 		return ret;
1633 
1634 	dn->ofs_in_node = ofs_in_node;
1635 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1636 		dn->data_blkaddr = f2fs_data_blkaddr(dn);
1637 		/*
1638 		 * f2fs_reserve_new_blocks will not guarantee entire block
1639 		 * allocation.
1640 		 */
1641 		if (dn->data_blkaddr == NULL_ADDR) {
1642 			ret = -ENOSPC;
1643 			break;
1644 		}
1645 
1646 		if (dn->data_blkaddr == NEW_ADDR)
1647 			continue;
1648 
1649 		if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1650 					DATA_GENERIC_ENHANCE)) {
1651 			ret = -EFSCORRUPTED;
1652 			break;
1653 		}
1654 
1655 		f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1);
1656 		f2fs_set_data_blkaddr(dn, NEW_ADDR);
1657 	}
1658 
1659 	if (index > start) {
1660 		f2fs_update_read_extent_cache_range(dn, start, 0,
1661 							index - start);
1662 		f2fs_update_age_extent_cache_range(dn, start, index - start);
1663 	}
1664 
1665 	return ret;
1666 }
1667 
1668 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1669 								int mode)
1670 {
1671 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1672 	struct address_space *mapping = inode->i_mapping;
1673 	pgoff_t index, pg_start, pg_end;
1674 	loff_t new_size = i_size_read(inode);
1675 	loff_t off_start, off_end;
1676 	int ret = 0;
1677 
1678 	ret = inode_newsize_ok(inode, (len + offset));
1679 	if (ret)
1680 		return ret;
1681 
1682 	ret = f2fs_convert_inline_inode(inode);
1683 	if (ret)
1684 		return ret;
1685 
1686 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1687 	if (ret)
1688 		return ret;
1689 
1690 	f2fs_zero_post_eof_page(inode, offset + len, true);
1691 
1692 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1693 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1694 
1695 	off_start = offset & (PAGE_SIZE - 1);
1696 	off_end = (offset + len) & (PAGE_SIZE - 1);
1697 
1698 	if (pg_start == pg_end) {
1699 		ret = fill_zero(inode, pg_start, off_start,
1700 						off_end - off_start);
1701 		if (ret)
1702 			return ret;
1703 
1704 		new_size = max_t(loff_t, new_size, offset + len);
1705 	} else {
1706 		if (off_start) {
1707 			ret = fill_zero(inode, pg_start++, off_start,
1708 						PAGE_SIZE - off_start);
1709 			if (ret)
1710 				return ret;
1711 
1712 			new_size = max_t(loff_t, new_size,
1713 					(loff_t)pg_start << PAGE_SHIFT);
1714 		}
1715 
1716 		for (index = pg_start; index < pg_end;) {
1717 			struct dnode_of_data dn;
1718 			unsigned int end_offset;
1719 			pgoff_t end;
1720 
1721 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1722 			filemap_invalidate_lock(mapping);
1723 
1724 			truncate_pagecache_range(inode,
1725 				(loff_t)index << PAGE_SHIFT,
1726 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1727 
1728 			f2fs_lock_op(sbi);
1729 
1730 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1731 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1732 			if (ret) {
1733 				f2fs_unlock_op(sbi);
1734 				filemap_invalidate_unlock(mapping);
1735 				f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1736 				goto out;
1737 			}
1738 
1739 			end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
1740 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1741 
1742 			ret = f2fs_do_zero_range(&dn, index, end);
1743 			f2fs_put_dnode(&dn);
1744 
1745 			f2fs_unlock_op(sbi);
1746 			filemap_invalidate_unlock(mapping);
1747 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1748 
1749 			f2fs_balance_fs(sbi, dn.node_changed);
1750 
1751 			if (ret)
1752 				goto out;
1753 
1754 			index = end;
1755 			new_size = max_t(loff_t, new_size,
1756 					(loff_t)index << PAGE_SHIFT);
1757 		}
1758 
1759 		if (off_end) {
1760 			ret = fill_zero(inode, pg_end, 0, off_end);
1761 			if (ret)
1762 				goto out;
1763 
1764 			new_size = max_t(loff_t, new_size, offset + len);
1765 		}
1766 	}
1767 
1768 out:
1769 	if (new_size > i_size_read(inode)) {
1770 		if (mode & FALLOC_FL_KEEP_SIZE)
1771 			file_set_keep_isize(inode);
1772 		else
1773 			f2fs_i_size_write(inode, new_size);
1774 	}
1775 	return ret;
1776 }
1777 
1778 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1779 {
1780 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1781 	struct address_space *mapping = inode->i_mapping;
1782 	pgoff_t nr, pg_start, pg_end, delta, idx;
1783 	loff_t new_size;
1784 	int ret = 0;
1785 
1786 	new_size = i_size_read(inode) + len;
1787 	ret = inode_newsize_ok(inode, new_size);
1788 	if (ret)
1789 		return ret;
1790 
1791 	if (offset >= i_size_read(inode))
1792 		return -EINVAL;
1793 
1794 	/* insert range should be aligned to block size of f2fs. */
1795 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1796 		return -EINVAL;
1797 
1798 	ret = f2fs_convert_inline_inode(inode);
1799 	if (ret)
1800 		return ret;
1801 
1802 	f2fs_balance_fs(sbi, true);
1803 
1804 	filemap_invalidate_lock(mapping);
1805 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1806 	filemap_invalidate_unlock(mapping);
1807 	if (ret)
1808 		return ret;
1809 
1810 	/* write out all dirty pages from offset */
1811 	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1812 	if (ret)
1813 		return ret;
1814 
1815 	pg_start = offset >> PAGE_SHIFT;
1816 	pg_end = (offset + len) >> PAGE_SHIFT;
1817 	delta = pg_end - pg_start;
1818 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1819 
1820 	/* avoid gc operation during block exchange */
1821 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1822 	filemap_invalidate_lock(mapping);
1823 
1824 	f2fs_zero_post_eof_page(inode, offset + len, false);
1825 	truncate_pagecache(inode, offset);
1826 
1827 	while (!ret && idx > pg_start) {
1828 		nr = idx - pg_start;
1829 		if (nr > delta)
1830 			nr = delta;
1831 		idx -= nr;
1832 
1833 		f2fs_lock_op(sbi);
1834 		f2fs_drop_extent_tree(inode);
1835 
1836 		ret = __exchange_data_block(inode, inode, idx,
1837 					idx + delta, nr, false);
1838 		f2fs_unlock_op(sbi);
1839 	}
1840 	filemap_invalidate_unlock(mapping);
1841 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1842 	if (ret)
1843 		return ret;
1844 
1845 	/* write out all moved pages, if possible */
1846 	filemap_invalidate_lock(mapping);
1847 	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1848 	truncate_pagecache(inode, offset);
1849 	filemap_invalidate_unlock(mapping);
1850 
1851 	if (!ret)
1852 		f2fs_i_size_write(inode, new_size);
1853 	return ret;
1854 }
1855 
1856 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
1857 					loff_t len, int mode)
1858 {
1859 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1860 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1861 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1862 			.m_may_create = true };
1863 	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1864 			.init_gc_type = FG_GC,
1865 			.should_migrate_blocks = false,
1866 			.err_gc_skipped = true,
1867 			.nr_free_secs = 0 };
1868 	pgoff_t pg_start, pg_end;
1869 	loff_t new_size;
1870 	loff_t off_end;
1871 	block_t expanded = 0;
1872 	int err;
1873 
1874 	err = inode_newsize_ok(inode, (len + offset));
1875 	if (err)
1876 		return err;
1877 
1878 	err = f2fs_convert_inline_inode(inode);
1879 	if (err)
1880 		return err;
1881 
1882 	f2fs_zero_post_eof_page(inode, offset + len, true);
1883 
1884 	f2fs_balance_fs(sbi, true);
1885 
1886 	pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1887 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1888 	off_end = (offset + len) & (PAGE_SIZE - 1);
1889 
1890 	map.m_lblk = pg_start;
1891 	map.m_len = pg_end - pg_start;
1892 	if (off_end)
1893 		map.m_len++;
1894 
1895 	if (!map.m_len)
1896 		return 0;
1897 
1898 	if (f2fs_is_pinned_file(inode)) {
1899 		block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1900 		block_t sec_len = roundup(map.m_len, sec_blks);
1901 
1902 		map.m_len = sec_blks;
1903 next_alloc:
1904 		f2fs_down_write(&sbi->pin_sem);
1905 
1906 		if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1907 			if (has_not_enough_free_secs(sbi, 0, 0)) {
1908 				f2fs_up_write(&sbi->pin_sem);
1909 				err = -ENOSPC;
1910 				f2fs_warn_ratelimited(sbi,
1911 					"ino:%lu, start:%lu, end:%lu, need to trigger GC to "
1912 					"reclaim enough free segment when checkpoint is enabled",
1913 					inode->i_ino, pg_start, pg_end);
1914 				goto out_err;
1915 			}
1916 		}
1917 
1918 		if (has_not_enough_free_secs(sbi, 0,
1919 				sbi->reserved_pin_section)) {
1920 			f2fs_down_write(&sbi->gc_lock);
1921 			stat_inc_gc_call_count(sbi, FOREGROUND);
1922 			err = f2fs_gc(sbi, &gc_control);
1923 			if (err && err != -ENODATA) {
1924 				f2fs_up_write(&sbi->pin_sem);
1925 				goto out_err;
1926 			}
1927 		}
1928 
1929 		err = f2fs_allocate_pinning_section(sbi);
1930 		if (err) {
1931 			f2fs_up_write(&sbi->pin_sem);
1932 			goto out_err;
1933 		}
1934 
1935 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1936 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
1937 		file_dont_truncate(inode);
1938 
1939 		f2fs_up_write(&sbi->pin_sem);
1940 
1941 		expanded += map.m_len;
1942 		sec_len -= map.m_len;
1943 		map.m_lblk += map.m_len;
1944 		if (!err && sec_len)
1945 			goto next_alloc;
1946 
1947 		map.m_len = expanded;
1948 	} else {
1949 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
1950 		expanded = map.m_len;
1951 	}
1952 out_err:
1953 	if (err) {
1954 		pgoff_t last_off;
1955 
1956 		if (!expanded)
1957 			return err;
1958 
1959 		last_off = pg_start + expanded - 1;
1960 
1961 		/* update new size to the failed position */
1962 		new_size = (last_off == pg_end) ? offset + len :
1963 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1964 	} else {
1965 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1966 	}
1967 
1968 	if (new_size > i_size_read(inode)) {
1969 		if (mode & FALLOC_FL_KEEP_SIZE)
1970 			file_set_keep_isize(inode);
1971 		else
1972 			f2fs_i_size_write(inode, new_size);
1973 	}
1974 
1975 	return err;
1976 }
1977 
1978 static long f2fs_fallocate(struct file *file, int mode,
1979 				loff_t offset, loff_t len)
1980 {
1981 	struct inode *inode = file_inode(file);
1982 	long ret = 0;
1983 
1984 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1985 		return -EIO;
1986 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1987 		return -ENOSPC;
1988 	if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode))
1989 		return -EOPNOTSUPP;
1990 
1991 	/* f2fs only support ->fallocate for regular file */
1992 	if (!S_ISREG(inode->i_mode))
1993 		return -EINVAL;
1994 
1995 	if (IS_ENCRYPTED(inode) &&
1996 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1997 		return -EOPNOTSUPP;
1998 
1999 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
2000 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
2001 			FALLOC_FL_INSERT_RANGE))
2002 		return -EOPNOTSUPP;
2003 
2004 	inode_lock(inode);
2005 
2006 	/*
2007 	 * Pinned file should not support partial truncation since the block
2008 	 * can be used by applications.
2009 	 */
2010 	if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
2011 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
2012 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) {
2013 		ret = -EOPNOTSUPP;
2014 		goto out;
2015 	}
2016 
2017 	ret = file_modified(file);
2018 	if (ret)
2019 		goto out;
2020 
2021 	/*
2022 	 * wait for inflight dio, blocks should be removed after IO
2023 	 * completion.
2024 	 */
2025 	inode_dio_wait(inode);
2026 
2027 	if (mode & FALLOC_FL_PUNCH_HOLE) {
2028 		if (offset >= inode->i_size)
2029 			goto out;
2030 
2031 		ret = f2fs_punch_hole(inode, offset, len);
2032 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
2033 		ret = f2fs_collapse_range(inode, offset, len);
2034 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
2035 		ret = f2fs_zero_range(inode, offset, len, mode);
2036 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
2037 		ret = f2fs_insert_range(inode, offset, len);
2038 	} else {
2039 		ret = f2fs_expand_inode_data(inode, offset, len, mode);
2040 	}
2041 
2042 	if (!ret) {
2043 		inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2044 		f2fs_mark_inode_dirty_sync(inode, false);
2045 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2046 	}
2047 
2048 out:
2049 	inode_unlock(inode);
2050 
2051 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
2052 	return ret;
2053 }
2054 
2055 static int f2fs_release_file(struct inode *inode, struct file *filp)
2056 {
2057 	if (atomic_dec_and_test(&F2FS_I(inode)->open_count))
2058 		f2fs_remove_donate_inode(inode);
2059 
2060 	/*
2061 	 * f2fs_release_file is called at every close calls. So we should
2062 	 * not drop any inmemory pages by close called by other process.
2063 	 */
2064 	if (!(filp->f_mode & FMODE_WRITE) ||
2065 			atomic_read(&inode->i_writecount) != 1)
2066 		return 0;
2067 
2068 	inode_lock(inode);
2069 	f2fs_abort_atomic_write(inode, true);
2070 	inode_unlock(inode);
2071 
2072 	return 0;
2073 }
2074 
2075 static int f2fs_file_flush(struct file *file, fl_owner_t id)
2076 {
2077 	struct inode *inode = file_inode(file);
2078 
2079 	/*
2080 	 * If the process doing a transaction is crashed, we should do
2081 	 * roll-back. Otherwise, other reader/write can see corrupted database
2082 	 * until all the writers close its file. Since this should be done
2083 	 * before dropping file lock, it needs to do in ->flush.
2084 	 */
2085 	if (F2FS_I(inode)->atomic_write_task == current &&
2086 				(current->flags & PF_EXITING)) {
2087 		inode_lock(inode);
2088 		f2fs_abort_atomic_write(inode, true);
2089 		inode_unlock(inode);
2090 	}
2091 
2092 	return 0;
2093 }
2094 
2095 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
2096 {
2097 	struct f2fs_inode_info *fi = F2FS_I(inode);
2098 	u32 masked_flags = fi->i_flags & mask;
2099 
2100 	/* mask can be shrunk by flags_valid selector */
2101 	iflags &= mask;
2102 
2103 	/* Is it quota file? Do not allow user to mess with it */
2104 	if (IS_NOQUOTA(inode))
2105 		return -EPERM;
2106 
2107 	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
2108 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
2109 			return -EOPNOTSUPP;
2110 		if (!f2fs_empty_dir(inode))
2111 			return -ENOTEMPTY;
2112 	}
2113 
2114 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
2115 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
2116 			return -EOPNOTSUPP;
2117 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
2118 			return -EINVAL;
2119 	}
2120 
2121 	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
2122 		if (masked_flags & F2FS_COMPR_FL) {
2123 			if (!f2fs_disable_compressed_file(inode))
2124 				return -EINVAL;
2125 		} else {
2126 			/* try to convert inline_data to support compression */
2127 			int err = f2fs_convert_inline_inode(inode);
2128 			if (err)
2129 				return err;
2130 
2131 			f2fs_down_write(&fi->i_sem);
2132 			if (!f2fs_may_compress(inode) ||
2133 				atomic_read(&fi->writeback) ||
2134 				(S_ISREG(inode->i_mode) &&
2135 				F2FS_HAS_BLOCKS(inode))) {
2136 				f2fs_up_write(&fi->i_sem);
2137 				return -EINVAL;
2138 			}
2139 			err = set_compress_context(inode);
2140 			f2fs_up_write(&fi->i_sem);
2141 
2142 			if (err)
2143 				return err;
2144 		}
2145 	}
2146 
2147 	fi->i_flags = iflags | (fi->i_flags & ~mask);
2148 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
2149 					(fi->i_flags & F2FS_NOCOMP_FL));
2150 
2151 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
2152 		set_inode_flag(inode, FI_PROJ_INHERIT);
2153 	else
2154 		clear_inode_flag(inode, FI_PROJ_INHERIT);
2155 
2156 	inode_set_ctime_current(inode);
2157 	f2fs_set_inode_flags(inode);
2158 	f2fs_mark_inode_dirty_sync(inode, true);
2159 	return 0;
2160 }
2161 
2162 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
2163 
2164 /*
2165  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
2166  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
2167  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
2168  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
2169  *
2170  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
2171  * FS_IOC_FSSETXATTR is done by the VFS.
2172  */
2173 
2174 static const struct {
2175 	u32 iflag;
2176 	u32 fsflag;
2177 } f2fs_fsflags_map[] = {
2178 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
2179 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
2180 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
2181 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
2182 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
2183 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
2184 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
2185 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
2186 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
2187 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
2188 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
2189 };
2190 
2191 #define F2FS_GETTABLE_FS_FL (		\
2192 		FS_COMPR_FL |		\
2193 		FS_SYNC_FL |		\
2194 		FS_IMMUTABLE_FL |	\
2195 		FS_APPEND_FL |		\
2196 		FS_NODUMP_FL |		\
2197 		FS_NOATIME_FL |		\
2198 		FS_NOCOMP_FL |		\
2199 		FS_INDEX_FL |		\
2200 		FS_DIRSYNC_FL |		\
2201 		FS_PROJINHERIT_FL |	\
2202 		FS_ENCRYPT_FL |		\
2203 		FS_INLINE_DATA_FL |	\
2204 		FS_NOCOW_FL |		\
2205 		FS_VERITY_FL |		\
2206 		FS_CASEFOLD_FL)
2207 
2208 #define F2FS_SETTABLE_FS_FL (		\
2209 		FS_COMPR_FL |		\
2210 		FS_SYNC_FL |		\
2211 		FS_IMMUTABLE_FL |	\
2212 		FS_APPEND_FL |		\
2213 		FS_NODUMP_FL |		\
2214 		FS_NOATIME_FL |		\
2215 		FS_NOCOMP_FL |		\
2216 		FS_DIRSYNC_FL |		\
2217 		FS_PROJINHERIT_FL |	\
2218 		FS_CASEFOLD_FL)
2219 
2220 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2221 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2222 {
2223 	u32 fsflags = 0;
2224 	int i;
2225 
2226 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2227 		if (iflags & f2fs_fsflags_map[i].iflag)
2228 			fsflags |= f2fs_fsflags_map[i].fsflag;
2229 
2230 	return fsflags;
2231 }
2232 
2233 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2234 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2235 {
2236 	u32 iflags = 0;
2237 	int i;
2238 
2239 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2240 		if (fsflags & f2fs_fsflags_map[i].fsflag)
2241 			iflags |= f2fs_fsflags_map[i].iflag;
2242 
2243 	return iflags;
2244 }
2245 
2246 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2247 {
2248 	struct inode *inode = file_inode(filp);
2249 
2250 	return put_user(inode->i_generation, (int __user *)arg);
2251 }
2252 
2253 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2254 {
2255 	struct inode *inode = file_inode(filp);
2256 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2257 	struct f2fs_inode_info *fi = F2FS_I(inode);
2258 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2259 	loff_t isize;
2260 	int ret;
2261 
2262 	if (!(filp->f_mode & FMODE_WRITE))
2263 		return -EBADF;
2264 
2265 	if (!inode_owner_or_capable(idmap, inode))
2266 		return -EACCES;
2267 
2268 	if (!S_ISREG(inode->i_mode))
2269 		return -EINVAL;
2270 
2271 	if (filp->f_flags & O_DIRECT)
2272 		return -EINVAL;
2273 
2274 	ret = mnt_want_write_file(filp);
2275 	if (ret)
2276 		return ret;
2277 
2278 	inode_lock(inode);
2279 
2280 	if (!f2fs_disable_compressed_file(inode) ||
2281 			f2fs_is_pinned_file(inode)) {
2282 		ret = -EINVAL;
2283 		goto out;
2284 	}
2285 
2286 	if (f2fs_is_atomic_file(inode))
2287 		goto out;
2288 
2289 	ret = f2fs_convert_inline_inode(inode);
2290 	if (ret)
2291 		goto out;
2292 
2293 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2294 	f2fs_down_write(&fi->i_gc_rwsem[READ]);
2295 
2296 	/*
2297 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2298 	 * f2fs_is_atomic_file.
2299 	 */
2300 	if (get_dirty_pages(inode))
2301 		f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2302 			  inode->i_ino, get_dirty_pages(inode));
2303 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2304 	if (ret)
2305 		goto out_unlock;
2306 
2307 	/* Check if the inode already has a COW inode */
2308 	if (fi->cow_inode == NULL) {
2309 		/* Create a COW inode for atomic write */
2310 		struct dentry *dentry = file_dentry(filp);
2311 		struct inode *dir = d_inode(dentry->d_parent);
2312 
2313 		ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode);
2314 		if (ret)
2315 			goto out_unlock;
2316 
2317 		set_inode_flag(fi->cow_inode, FI_COW_FILE);
2318 		clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2319 
2320 		/* Set the COW inode's atomic_inode to the atomic inode */
2321 		F2FS_I(fi->cow_inode)->atomic_inode = inode;
2322 	} else {
2323 		/* Reuse the already created COW inode */
2324 		f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode));
2325 
2326 		invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1);
2327 
2328 		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2329 		if (ret)
2330 			goto out_unlock;
2331 	}
2332 
2333 	f2fs_write_inode(inode, NULL);
2334 
2335 	stat_inc_atomic_inode(inode);
2336 
2337 	set_inode_flag(inode, FI_ATOMIC_FILE);
2338 
2339 	isize = i_size_read(inode);
2340 	fi->original_i_size = isize;
2341 	if (truncate) {
2342 		set_inode_flag(inode, FI_ATOMIC_REPLACE);
2343 		truncate_inode_pages_final(inode->i_mapping);
2344 		f2fs_i_size_write(inode, 0);
2345 		isize = 0;
2346 	}
2347 	f2fs_i_size_write(fi->cow_inode, isize);
2348 
2349 out_unlock:
2350 	f2fs_up_write(&fi->i_gc_rwsem[READ]);
2351 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2352 	if (ret)
2353 		goto out;
2354 
2355 	f2fs_update_time(sbi, REQ_TIME);
2356 	fi->atomic_write_task = current;
2357 	stat_update_max_atomic_write(inode);
2358 	fi->atomic_write_cnt = 0;
2359 out:
2360 	inode_unlock(inode);
2361 	mnt_drop_write_file(filp);
2362 	return ret;
2363 }
2364 
2365 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2366 {
2367 	struct inode *inode = file_inode(filp);
2368 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2369 	int ret;
2370 
2371 	if (!(filp->f_mode & FMODE_WRITE))
2372 		return -EBADF;
2373 
2374 	if (!inode_owner_or_capable(idmap, inode))
2375 		return -EACCES;
2376 
2377 	ret = mnt_want_write_file(filp);
2378 	if (ret)
2379 		return ret;
2380 
2381 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2382 
2383 	inode_lock(inode);
2384 
2385 	if (f2fs_is_atomic_file(inode)) {
2386 		ret = f2fs_commit_atomic_write(inode);
2387 		if (!ret)
2388 			ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2389 
2390 		f2fs_abort_atomic_write(inode, ret);
2391 	} else {
2392 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2393 	}
2394 
2395 	inode_unlock(inode);
2396 	mnt_drop_write_file(filp);
2397 	return ret;
2398 }
2399 
2400 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2401 {
2402 	struct inode *inode = file_inode(filp);
2403 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2404 	int ret;
2405 
2406 	if (!(filp->f_mode & FMODE_WRITE))
2407 		return -EBADF;
2408 
2409 	if (!inode_owner_or_capable(idmap, inode))
2410 		return -EACCES;
2411 
2412 	ret = mnt_want_write_file(filp);
2413 	if (ret)
2414 		return ret;
2415 
2416 	inode_lock(inode);
2417 
2418 	f2fs_abort_atomic_write(inode, true);
2419 
2420 	inode_unlock(inode);
2421 
2422 	mnt_drop_write_file(filp);
2423 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2424 	return ret;
2425 }
2426 
2427 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
2428 						bool readonly, bool need_lock)
2429 {
2430 	struct super_block *sb = sbi->sb;
2431 	int ret = 0;
2432 
2433 	switch (flag) {
2434 	case F2FS_GOING_DOWN_FULLSYNC:
2435 		ret = bdev_freeze(sb->s_bdev);
2436 		if (ret)
2437 			goto out;
2438 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2439 		bdev_thaw(sb->s_bdev);
2440 		break;
2441 	case F2FS_GOING_DOWN_METASYNC:
2442 		/* do checkpoint only */
2443 		ret = f2fs_sync_fs(sb, 1);
2444 		if (ret) {
2445 			if (ret == -EIO)
2446 				ret = 0;
2447 			goto out;
2448 		}
2449 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2450 		break;
2451 	case F2FS_GOING_DOWN_NOSYNC:
2452 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2453 		break;
2454 	case F2FS_GOING_DOWN_METAFLUSH:
2455 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2456 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2457 		break;
2458 	case F2FS_GOING_DOWN_NEED_FSCK:
2459 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2460 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2461 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2462 		/* do checkpoint only */
2463 		ret = f2fs_sync_fs(sb, 1);
2464 		if (ret == -EIO)
2465 			ret = 0;
2466 		goto out;
2467 	default:
2468 		ret = -EINVAL;
2469 		goto out;
2470 	}
2471 
2472 	if (readonly)
2473 		goto out;
2474 
2475 	/*
2476 	 * grab sb->s_umount to avoid racing w/ remount() and other shutdown
2477 	 * paths.
2478 	 */
2479 	if (need_lock)
2480 		down_write(&sbi->sb->s_umount);
2481 
2482 	f2fs_stop_gc_thread(sbi);
2483 	f2fs_stop_discard_thread(sbi);
2484 
2485 	f2fs_drop_discard_cmd(sbi);
2486 	clear_opt(sbi, DISCARD);
2487 
2488 	if (need_lock)
2489 		up_write(&sbi->sb->s_umount);
2490 
2491 	f2fs_update_time(sbi, REQ_TIME);
2492 out:
2493 
2494 	trace_f2fs_shutdown(sbi, flag, ret);
2495 
2496 	return ret;
2497 }
2498 
2499 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2500 {
2501 	struct inode *inode = file_inode(filp);
2502 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2503 	__u32 in;
2504 	int ret;
2505 	bool need_drop = false, readonly = false;
2506 
2507 	if (!capable(CAP_SYS_ADMIN))
2508 		return -EPERM;
2509 
2510 	if (get_user(in, (__u32 __user *)arg))
2511 		return -EFAULT;
2512 
2513 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2514 		ret = mnt_want_write_file(filp);
2515 		if (ret) {
2516 			if (ret != -EROFS)
2517 				return ret;
2518 
2519 			/* fallback to nosync shutdown for readonly fs */
2520 			in = F2FS_GOING_DOWN_NOSYNC;
2521 			readonly = true;
2522 		} else {
2523 			need_drop = true;
2524 		}
2525 	}
2526 
2527 	ret = f2fs_do_shutdown(sbi, in, readonly, true);
2528 
2529 	if (need_drop)
2530 		mnt_drop_write_file(filp);
2531 
2532 	return ret;
2533 }
2534 
2535 static int f2fs_keep_noreuse_range(struct inode *inode,
2536 				loff_t offset, loff_t len)
2537 {
2538 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2539 	u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
2540 	u64 start, end;
2541 	int ret = 0;
2542 
2543 	if (!S_ISREG(inode->i_mode))
2544 		return 0;
2545 
2546 	if (offset >= max_bytes || len > max_bytes ||
2547 	    (offset + len) > max_bytes)
2548 		return 0;
2549 
2550 	start = offset >> PAGE_SHIFT;
2551 	end = DIV_ROUND_UP(offset + len, PAGE_SIZE);
2552 
2553 	inode_lock(inode);
2554 	if (f2fs_is_atomic_file(inode)) {
2555 		inode_unlock(inode);
2556 		return 0;
2557 	}
2558 
2559 	spin_lock(&sbi->inode_lock[DONATE_INODE]);
2560 	/* let's remove the range, if len = 0 */
2561 	if (!len) {
2562 		if (!list_empty(&F2FS_I(inode)->gdonate_list)) {
2563 			list_del_init(&F2FS_I(inode)->gdonate_list);
2564 			sbi->donate_files--;
2565 			if (is_inode_flag_set(inode, FI_DONATE_FINISHED))
2566 				ret = -EALREADY;
2567 			else
2568 				set_inode_flag(inode, FI_DONATE_FINISHED);
2569 		} else
2570 			ret = -ENOENT;
2571 	} else {
2572 		if (list_empty(&F2FS_I(inode)->gdonate_list)) {
2573 			list_add_tail(&F2FS_I(inode)->gdonate_list,
2574 					&sbi->inode_list[DONATE_INODE]);
2575 			sbi->donate_files++;
2576 		} else {
2577 			list_move_tail(&F2FS_I(inode)->gdonate_list,
2578 					&sbi->inode_list[DONATE_INODE]);
2579 		}
2580 		F2FS_I(inode)->donate_start = start;
2581 		F2FS_I(inode)->donate_end = end - 1;
2582 		clear_inode_flag(inode, FI_DONATE_FINISHED);
2583 	}
2584 	spin_unlock(&sbi->inode_lock[DONATE_INODE]);
2585 	inode_unlock(inode);
2586 
2587 	return ret;
2588 }
2589 
2590 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2591 {
2592 	struct inode *inode = file_inode(filp);
2593 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2594 	struct fstrim_range range;
2595 	int ret;
2596 
2597 	if (!capable(CAP_SYS_ADMIN))
2598 		return -EPERM;
2599 
2600 	if (!f2fs_hw_support_discard(sbi))
2601 		return -EOPNOTSUPP;
2602 
2603 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2604 				sizeof(range)))
2605 		return -EFAULT;
2606 
2607 	ret = mnt_want_write_file(filp);
2608 	if (ret)
2609 		return ret;
2610 
2611 	range.minlen = max_t(unsigned int, range.minlen,
2612 			f2fs_hw_discard_granularity(sbi));
2613 	ret = f2fs_trim_fs(sbi, &range);
2614 	mnt_drop_write_file(filp);
2615 	if (ret < 0)
2616 		return ret;
2617 
2618 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2619 				sizeof(range)))
2620 		return -EFAULT;
2621 	f2fs_update_time(sbi, REQ_TIME);
2622 	return 0;
2623 }
2624 
2625 static bool uuid_is_nonzero(__u8 u[16])
2626 {
2627 	int i;
2628 
2629 	for (i = 0; i < 16; i++)
2630 		if (u[i])
2631 			return true;
2632 	return false;
2633 }
2634 
2635 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2636 {
2637 	struct inode *inode = file_inode(filp);
2638 	int ret;
2639 
2640 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2641 		return -EOPNOTSUPP;
2642 
2643 	ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2644 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2645 	return ret;
2646 }
2647 
2648 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2649 {
2650 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2651 		return -EOPNOTSUPP;
2652 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2653 }
2654 
2655 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2656 {
2657 	struct inode *inode = file_inode(filp);
2658 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2659 	u8 encrypt_pw_salt[16];
2660 	int err;
2661 
2662 	if (!f2fs_sb_has_encrypt(sbi))
2663 		return -EOPNOTSUPP;
2664 
2665 	err = mnt_want_write_file(filp);
2666 	if (err)
2667 		return err;
2668 
2669 	f2fs_down_write(&sbi->sb_lock);
2670 
2671 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2672 		goto got_it;
2673 
2674 	/* update superblock with uuid */
2675 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2676 
2677 	err = f2fs_commit_super(sbi, false);
2678 	if (err) {
2679 		/* undo new data */
2680 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2681 		goto out_err;
2682 	}
2683 got_it:
2684 	memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2685 out_err:
2686 	f2fs_up_write(&sbi->sb_lock);
2687 	mnt_drop_write_file(filp);
2688 
2689 	if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2690 		err = -EFAULT;
2691 
2692 	return err;
2693 }
2694 
2695 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2696 					     unsigned long arg)
2697 {
2698 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2699 		return -EOPNOTSUPP;
2700 
2701 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2702 }
2703 
2704 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2705 {
2706 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2707 		return -EOPNOTSUPP;
2708 
2709 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2710 }
2711 
2712 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2713 {
2714 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2715 		return -EOPNOTSUPP;
2716 
2717 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2718 }
2719 
2720 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2721 						    unsigned long arg)
2722 {
2723 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2724 		return -EOPNOTSUPP;
2725 
2726 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2727 }
2728 
2729 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2730 					      unsigned long arg)
2731 {
2732 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2733 		return -EOPNOTSUPP;
2734 
2735 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2736 }
2737 
2738 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2739 {
2740 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2741 		return -EOPNOTSUPP;
2742 
2743 	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2744 }
2745 
2746 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2747 {
2748 	struct inode *inode = file_inode(filp);
2749 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2750 	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2751 			.no_bg_gc = false,
2752 			.should_migrate_blocks = false,
2753 			.nr_free_secs = 0 };
2754 	__u32 sync;
2755 	int ret;
2756 
2757 	if (!capable(CAP_SYS_ADMIN))
2758 		return -EPERM;
2759 
2760 	if (get_user(sync, (__u32 __user *)arg))
2761 		return -EFAULT;
2762 
2763 	if (f2fs_readonly(sbi->sb))
2764 		return -EROFS;
2765 
2766 	ret = mnt_want_write_file(filp);
2767 	if (ret)
2768 		return ret;
2769 
2770 	if (!sync) {
2771 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2772 			ret = -EBUSY;
2773 			goto out;
2774 		}
2775 	} else {
2776 		f2fs_down_write(&sbi->gc_lock);
2777 	}
2778 
2779 	gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2780 	gc_control.err_gc_skipped = sync;
2781 	stat_inc_gc_call_count(sbi, FOREGROUND);
2782 	ret = f2fs_gc(sbi, &gc_control);
2783 out:
2784 	mnt_drop_write_file(filp);
2785 	return ret;
2786 }
2787 
2788 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2789 {
2790 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2791 	struct f2fs_gc_control gc_control = {
2792 			.init_gc_type = range->sync ? FG_GC : BG_GC,
2793 			.no_bg_gc = false,
2794 			.should_migrate_blocks = false,
2795 			.err_gc_skipped = range->sync,
2796 			.nr_free_secs = 0 };
2797 	u64 end;
2798 	int ret;
2799 
2800 	if (!capable(CAP_SYS_ADMIN))
2801 		return -EPERM;
2802 	if (f2fs_readonly(sbi->sb))
2803 		return -EROFS;
2804 
2805 	end = range->start + range->len;
2806 	if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2807 					end >= MAX_BLKADDR(sbi))
2808 		return -EINVAL;
2809 
2810 	ret = mnt_want_write_file(filp);
2811 	if (ret)
2812 		return ret;
2813 
2814 do_more:
2815 	if (!range->sync) {
2816 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2817 			ret = -EBUSY;
2818 			goto out;
2819 		}
2820 	} else {
2821 		f2fs_down_write(&sbi->gc_lock);
2822 	}
2823 
2824 	gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2825 	stat_inc_gc_call_count(sbi, FOREGROUND);
2826 	ret = f2fs_gc(sbi, &gc_control);
2827 	if (ret) {
2828 		if (ret == -EBUSY)
2829 			ret = -EAGAIN;
2830 		goto out;
2831 	}
2832 	range->start += CAP_BLKS_PER_SEC(sbi);
2833 	if (range->start <= end)
2834 		goto do_more;
2835 out:
2836 	mnt_drop_write_file(filp);
2837 	return ret;
2838 }
2839 
2840 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2841 {
2842 	struct f2fs_gc_range range;
2843 
2844 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2845 							sizeof(range)))
2846 		return -EFAULT;
2847 	return __f2fs_ioc_gc_range(filp, &range);
2848 }
2849 
2850 static int f2fs_ioc_write_checkpoint(struct file *filp)
2851 {
2852 	struct inode *inode = file_inode(filp);
2853 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2854 	int ret;
2855 
2856 	if (!capable(CAP_SYS_ADMIN))
2857 		return -EPERM;
2858 
2859 	if (f2fs_readonly(sbi->sb))
2860 		return -EROFS;
2861 
2862 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2863 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2864 		return -EINVAL;
2865 	}
2866 
2867 	ret = mnt_want_write_file(filp);
2868 	if (ret)
2869 		return ret;
2870 
2871 	ret = f2fs_sync_fs(sbi->sb, 1);
2872 
2873 	mnt_drop_write_file(filp);
2874 	return ret;
2875 }
2876 
2877 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2878 					struct file *filp,
2879 					struct f2fs_defragment *range)
2880 {
2881 	struct inode *inode = file_inode(filp);
2882 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2883 					.m_seg_type = NO_CHECK_TYPE,
2884 					.m_may_create = false };
2885 	struct extent_info ei = {};
2886 	pgoff_t pg_start, pg_end, next_pgofs;
2887 	unsigned int total = 0, sec_num;
2888 	block_t blk_end = 0;
2889 	bool fragmented = false;
2890 	int err;
2891 
2892 	f2fs_balance_fs(sbi, true);
2893 
2894 	inode_lock(inode);
2895 	pg_start = range->start >> PAGE_SHIFT;
2896 	pg_end = min_t(pgoff_t,
2897 				(range->start + range->len) >> PAGE_SHIFT,
2898 				DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));
2899 
2900 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) ||
2901 		f2fs_is_atomic_file(inode)) {
2902 		err = -EINVAL;
2903 		goto unlock_out;
2904 	}
2905 
2906 	/* if in-place-update policy is enabled, don't waste time here */
2907 	set_inode_flag(inode, FI_OPU_WRITE);
2908 	if (f2fs_should_update_inplace(inode, NULL)) {
2909 		err = -EINVAL;
2910 		goto out;
2911 	}
2912 
2913 	/* writeback all dirty pages in the range */
2914 	err = filemap_write_and_wait_range(inode->i_mapping,
2915 						pg_start << PAGE_SHIFT,
2916 						(pg_end << PAGE_SHIFT) - 1);
2917 	if (err)
2918 		goto out;
2919 
2920 	/*
2921 	 * lookup mapping info in extent cache, skip defragmenting if physical
2922 	 * block addresses are continuous.
2923 	 */
2924 	if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2925 		if ((pgoff_t)ei.fofs + ei.len >= pg_end)
2926 			goto out;
2927 	}
2928 
2929 	map.m_lblk = pg_start;
2930 	map.m_next_pgofs = &next_pgofs;
2931 
2932 	/*
2933 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2934 	 * physical block addresses are continuous even if there are hole(s)
2935 	 * in logical blocks.
2936 	 */
2937 	while (map.m_lblk < pg_end) {
2938 		map.m_len = pg_end - map.m_lblk;
2939 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2940 		if (err)
2941 			goto out;
2942 
2943 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2944 			map.m_lblk = next_pgofs;
2945 			continue;
2946 		}
2947 
2948 		if (blk_end && blk_end != map.m_pblk)
2949 			fragmented = true;
2950 
2951 		/* record total count of block that we're going to move */
2952 		total += map.m_len;
2953 
2954 		blk_end = map.m_pblk + map.m_len;
2955 
2956 		map.m_lblk += map.m_len;
2957 	}
2958 
2959 	if (!fragmented) {
2960 		total = 0;
2961 		goto out;
2962 	}
2963 
2964 	sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2965 
2966 	/*
2967 	 * make sure there are enough free section for LFS allocation, this can
2968 	 * avoid defragment running in SSR mode when free section are allocated
2969 	 * intensively
2970 	 */
2971 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2972 		err = -EAGAIN;
2973 		goto out;
2974 	}
2975 
2976 	map.m_lblk = pg_start;
2977 	map.m_len = pg_end - pg_start;
2978 	total = 0;
2979 
2980 	while (map.m_lblk < pg_end) {
2981 		pgoff_t idx;
2982 		int cnt = 0;
2983 
2984 do_map:
2985 		map.m_len = pg_end - map.m_lblk;
2986 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2987 		if (err)
2988 			goto clear_out;
2989 
2990 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2991 			map.m_lblk = next_pgofs;
2992 			goto check;
2993 		}
2994 
2995 		set_inode_flag(inode, FI_SKIP_WRITES);
2996 
2997 		idx = map.m_lblk;
2998 		while (idx < map.m_lblk + map.m_len &&
2999 						cnt < BLKS_PER_SEG(sbi)) {
3000 			struct folio *folio;
3001 
3002 			folio = f2fs_get_lock_data_folio(inode, idx, true);
3003 			if (IS_ERR(folio)) {
3004 				err = PTR_ERR(folio);
3005 				goto clear_out;
3006 			}
3007 
3008 			f2fs_folio_wait_writeback(folio, DATA, true, true);
3009 
3010 			folio_mark_dirty(folio);
3011 			folio_set_f2fs_gcing(folio);
3012 			f2fs_folio_put(folio, true);
3013 
3014 			idx++;
3015 			cnt++;
3016 			total++;
3017 		}
3018 
3019 		map.m_lblk = idx;
3020 check:
3021 		if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi))
3022 			goto do_map;
3023 
3024 		clear_inode_flag(inode, FI_SKIP_WRITES);
3025 
3026 		err = filemap_fdatawrite(inode->i_mapping);
3027 		if (err)
3028 			goto out;
3029 	}
3030 clear_out:
3031 	clear_inode_flag(inode, FI_SKIP_WRITES);
3032 out:
3033 	clear_inode_flag(inode, FI_OPU_WRITE);
3034 unlock_out:
3035 	inode_unlock(inode);
3036 	if (!err)
3037 		range->len = (u64)total << PAGE_SHIFT;
3038 	return err;
3039 }
3040 
3041 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
3042 {
3043 	struct inode *inode = file_inode(filp);
3044 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3045 	struct f2fs_defragment range;
3046 	int err;
3047 
3048 	if (!capable(CAP_SYS_ADMIN))
3049 		return -EPERM;
3050 
3051 	if (!S_ISREG(inode->i_mode))
3052 		return -EINVAL;
3053 
3054 	if (f2fs_readonly(sbi->sb))
3055 		return -EROFS;
3056 
3057 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
3058 							sizeof(range)))
3059 		return -EFAULT;
3060 
3061 	/* verify alignment of offset & size */
3062 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
3063 		return -EINVAL;
3064 
3065 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
3066 					max_file_blocks(inode)))
3067 		return -EINVAL;
3068 
3069 	err = mnt_want_write_file(filp);
3070 	if (err)
3071 		return err;
3072 
3073 	err = f2fs_defragment_range(sbi, filp, &range);
3074 	mnt_drop_write_file(filp);
3075 
3076 	if (range.len)
3077 		f2fs_update_time(sbi, REQ_TIME);
3078 	if (err < 0)
3079 		return err;
3080 
3081 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
3082 							sizeof(range)))
3083 		return -EFAULT;
3084 
3085 	return 0;
3086 }
3087 
3088 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
3089 			struct file *file_out, loff_t pos_out, size_t len)
3090 {
3091 	struct inode *src = file_inode(file_in);
3092 	struct inode *dst = file_inode(file_out);
3093 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
3094 	size_t olen = len, dst_max_i_size = 0;
3095 	size_t dst_osize;
3096 	int ret;
3097 
3098 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
3099 				src->i_sb != dst->i_sb)
3100 		return -EXDEV;
3101 
3102 	if (unlikely(f2fs_readonly(src->i_sb)))
3103 		return -EROFS;
3104 
3105 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
3106 		return -EINVAL;
3107 
3108 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
3109 		return -EOPNOTSUPP;
3110 
3111 	if (pos_out < 0 || pos_in < 0)
3112 		return -EINVAL;
3113 
3114 	if (src == dst) {
3115 		if (pos_in == pos_out)
3116 			return 0;
3117 		if (pos_out > pos_in && pos_out < pos_in + len)
3118 			return -EINVAL;
3119 	}
3120 
3121 	inode_lock(src);
3122 	if (src != dst) {
3123 		ret = -EBUSY;
3124 		if (!inode_trylock(dst))
3125 			goto out;
3126 	}
3127 
3128 	if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) ||
3129 		f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) {
3130 		ret = -EOPNOTSUPP;
3131 		goto out_unlock;
3132 	}
3133 
3134 	if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) {
3135 		ret = -EINVAL;
3136 		goto out_unlock;
3137 	}
3138 
3139 	ret = -EINVAL;
3140 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
3141 		goto out_unlock;
3142 	if (len == 0)
3143 		olen = len = src->i_size - pos_in;
3144 	if (pos_in + len == src->i_size)
3145 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
3146 	if (len == 0) {
3147 		ret = 0;
3148 		goto out_unlock;
3149 	}
3150 
3151 	dst_osize = dst->i_size;
3152 	if (pos_out + olen > dst->i_size)
3153 		dst_max_i_size = pos_out + olen;
3154 
3155 	/* verify the end result is block aligned */
3156 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
3157 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
3158 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
3159 		goto out_unlock;
3160 
3161 	ret = f2fs_convert_inline_inode(src);
3162 	if (ret)
3163 		goto out_unlock;
3164 
3165 	ret = f2fs_convert_inline_inode(dst);
3166 	if (ret)
3167 		goto out_unlock;
3168 
3169 	/* write out all dirty pages from offset */
3170 	ret = filemap_write_and_wait_range(src->i_mapping,
3171 					pos_in, pos_in + len);
3172 	if (ret)
3173 		goto out_unlock;
3174 
3175 	ret = filemap_write_and_wait_range(dst->i_mapping,
3176 					pos_out, pos_out + len);
3177 	if (ret)
3178 		goto out_unlock;
3179 
3180 	f2fs_balance_fs(sbi, true);
3181 
3182 	f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3183 	if (src != dst) {
3184 		ret = -EBUSY;
3185 		if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
3186 			goto out_src;
3187 	}
3188 
3189 	f2fs_lock_op(sbi);
3190 	ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in),
3191 				F2FS_BYTES_TO_BLK(pos_out),
3192 				F2FS_BYTES_TO_BLK(len), false);
3193 
3194 	if (!ret) {
3195 		if (dst_max_i_size)
3196 			f2fs_i_size_write(dst, dst_max_i_size);
3197 		else if (dst_osize != dst->i_size)
3198 			f2fs_i_size_write(dst, dst_osize);
3199 	}
3200 	f2fs_unlock_op(sbi);
3201 
3202 	if (src != dst)
3203 		f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
3204 out_src:
3205 	f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3206 	if (ret)
3207 		goto out_unlock;
3208 
3209 	inode_set_mtime_to_ts(src, inode_set_ctime_current(src));
3210 	f2fs_mark_inode_dirty_sync(src, false);
3211 	if (src != dst) {
3212 		inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst));
3213 		f2fs_mark_inode_dirty_sync(dst, false);
3214 	}
3215 	f2fs_update_time(sbi, REQ_TIME);
3216 
3217 out_unlock:
3218 	if (src != dst)
3219 		inode_unlock(dst);
3220 out:
3221 	inode_unlock(src);
3222 	return ret;
3223 }
3224 
3225 static int __f2fs_ioc_move_range(struct file *filp,
3226 				struct f2fs_move_range *range)
3227 {
3228 	int err;
3229 
3230 	if (!(filp->f_mode & FMODE_READ) ||
3231 			!(filp->f_mode & FMODE_WRITE))
3232 		return -EBADF;
3233 
3234 	CLASS(fd, dst)(range->dst_fd);
3235 	if (fd_empty(dst))
3236 		return -EBADF;
3237 
3238 	if (!(fd_file(dst)->f_mode & FMODE_WRITE))
3239 		return -EBADF;
3240 
3241 	err = mnt_want_write_file(filp);
3242 	if (err)
3243 		return err;
3244 
3245 	err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst),
3246 					range->pos_out, range->len);
3247 
3248 	mnt_drop_write_file(filp);
3249 	return err;
3250 }
3251 
3252 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
3253 {
3254 	struct f2fs_move_range range;
3255 
3256 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
3257 							sizeof(range)))
3258 		return -EFAULT;
3259 	return __f2fs_ioc_move_range(filp, &range);
3260 }
3261 
3262 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
3263 {
3264 	struct inode *inode = file_inode(filp);
3265 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3266 	struct sit_info *sm = SIT_I(sbi);
3267 	unsigned int start_segno = 0, end_segno = 0;
3268 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
3269 	struct f2fs_flush_device range;
3270 	struct f2fs_gc_control gc_control = {
3271 			.init_gc_type = FG_GC,
3272 			.should_migrate_blocks = true,
3273 			.err_gc_skipped = true,
3274 			.nr_free_secs = 0 };
3275 	int ret;
3276 
3277 	if (!capable(CAP_SYS_ADMIN))
3278 		return -EPERM;
3279 
3280 	if (f2fs_readonly(sbi->sb))
3281 		return -EROFS;
3282 
3283 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3284 		return -EINVAL;
3285 
3286 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
3287 							sizeof(range)))
3288 		return -EFAULT;
3289 
3290 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
3291 			__is_large_section(sbi)) {
3292 		f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1",
3293 			  range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi));
3294 		return -EINVAL;
3295 	}
3296 
3297 	ret = mnt_want_write_file(filp);
3298 	if (ret)
3299 		return ret;
3300 
3301 	if (range.dev_num != 0)
3302 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
3303 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
3304 
3305 	start_segno = sm->last_victim[FLUSH_DEVICE];
3306 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
3307 		start_segno = dev_start_segno;
3308 	end_segno = min(start_segno + range.segments, dev_end_segno);
3309 
3310 	while (start_segno < end_segno) {
3311 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
3312 			ret = -EBUSY;
3313 			goto out;
3314 		}
3315 		sm->last_victim[GC_CB] = end_segno + 1;
3316 		sm->last_victim[GC_GREEDY] = end_segno + 1;
3317 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
3318 
3319 		gc_control.victim_segno = start_segno;
3320 		stat_inc_gc_call_count(sbi, FOREGROUND);
3321 		ret = f2fs_gc(sbi, &gc_control);
3322 		if (ret == -EAGAIN)
3323 			ret = 0;
3324 		else if (ret < 0)
3325 			break;
3326 		start_segno++;
3327 	}
3328 out:
3329 	mnt_drop_write_file(filp);
3330 	return ret;
3331 }
3332 
3333 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3334 {
3335 	struct inode *inode = file_inode(filp);
3336 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3337 
3338 	/* Must validate to set it with SQLite behavior in Android. */
3339 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3340 
3341 	return put_user(sb_feature, (u32 __user *)arg);
3342 }
3343 
3344 #ifdef CONFIG_QUOTA
3345 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3346 {
3347 	struct dquot *transfer_to[MAXQUOTAS] = {};
3348 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3349 	struct super_block *sb = sbi->sb;
3350 	int err;
3351 
3352 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3353 	if (IS_ERR(transfer_to[PRJQUOTA]))
3354 		return PTR_ERR(transfer_to[PRJQUOTA]);
3355 
3356 	err = __dquot_transfer(inode, transfer_to);
3357 	if (err)
3358 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3359 	dqput(transfer_to[PRJQUOTA]);
3360 	return err;
3361 }
3362 
3363 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3364 {
3365 	struct f2fs_inode_info *fi = F2FS_I(inode);
3366 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3367 	struct f2fs_inode *ri = NULL;
3368 	kprojid_t kprojid;
3369 	int err;
3370 
3371 	if (!f2fs_sb_has_project_quota(sbi)) {
3372 		if (projid != F2FS_DEF_PROJID)
3373 			return -EOPNOTSUPP;
3374 		else
3375 			return 0;
3376 	}
3377 
3378 	if (!f2fs_has_extra_attr(inode))
3379 		return -EOPNOTSUPP;
3380 
3381 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3382 
3383 	if (projid_eq(kprojid, fi->i_projid))
3384 		return 0;
3385 
3386 	err = -EPERM;
3387 	/* Is it quota file? Do not allow user to mess with it */
3388 	if (IS_NOQUOTA(inode))
3389 		return err;
3390 
3391 	if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3392 		return -EOVERFLOW;
3393 
3394 	err = f2fs_dquot_initialize(inode);
3395 	if (err)
3396 		return err;
3397 
3398 	f2fs_lock_op(sbi);
3399 	err = f2fs_transfer_project_quota(inode, kprojid);
3400 	if (err)
3401 		goto out_unlock;
3402 
3403 	fi->i_projid = kprojid;
3404 	inode_set_ctime_current(inode);
3405 	f2fs_mark_inode_dirty_sync(inode, true);
3406 out_unlock:
3407 	f2fs_unlock_op(sbi);
3408 	return err;
3409 }
3410 #else
3411 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3412 {
3413 	return 0;
3414 }
3415 
3416 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3417 {
3418 	if (projid != F2FS_DEF_PROJID)
3419 		return -EOPNOTSUPP;
3420 	return 0;
3421 }
3422 #endif
3423 
3424 int f2fs_fileattr_get(struct dentry *dentry, struct file_kattr *fa)
3425 {
3426 	struct inode *inode = d_inode(dentry);
3427 	struct f2fs_inode_info *fi = F2FS_I(inode);
3428 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3429 
3430 	if (IS_ENCRYPTED(inode))
3431 		fsflags |= FS_ENCRYPT_FL;
3432 	if (IS_VERITY(inode))
3433 		fsflags |= FS_VERITY_FL;
3434 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3435 		fsflags |= FS_INLINE_DATA_FL;
3436 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3437 		fsflags |= FS_NOCOW_FL;
3438 
3439 	fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3440 
3441 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3442 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3443 
3444 	return 0;
3445 }
3446 
3447 int f2fs_fileattr_set(struct mnt_idmap *idmap,
3448 		      struct dentry *dentry, struct file_kattr *fa)
3449 {
3450 	struct inode *inode = d_inode(dentry);
3451 	u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3452 	u32 iflags;
3453 	int err;
3454 
3455 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3456 		return -EIO;
3457 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3458 		return -ENOSPC;
3459 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
3460 		return -EOPNOTSUPP;
3461 	fsflags &= F2FS_SETTABLE_FS_FL;
3462 	if (!fa->flags_valid)
3463 		mask &= FS_COMMON_FL;
3464 
3465 	iflags = f2fs_fsflags_to_iflags(fsflags);
3466 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3467 		return -EOPNOTSUPP;
3468 
3469 	err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3470 	if (!err)
3471 		err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3472 
3473 	return err;
3474 }
3475 
3476 int f2fs_pin_file_control(struct inode *inode, bool inc)
3477 {
3478 	struct f2fs_inode_info *fi = F2FS_I(inode);
3479 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3480 
3481 	if (IS_DEVICE_ALIASING(inode))
3482 		return -EINVAL;
3483 
3484 	if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) {
3485 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3486 			  __func__, inode->i_ino, fi->i_gc_failures);
3487 		clear_inode_flag(inode, FI_PIN_FILE);
3488 		return -EAGAIN;
3489 	}
3490 
3491 	/* Use i_gc_failures for normal file as a risk signal. */
3492 	if (inc)
3493 		f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
3494 
3495 	return 0;
3496 }
3497 
3498 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3499 {
3500 	struct inode *inode = file_inode(filp);
3501 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3502 	__u32 pin;
3503 	int ret = 0;
3504 
3505 	if (get_user(pin, (__u32 __user *)arg))
3506 		return -EFAULT;
3507 
3508 	if (!S_ISREG(inode->i_mode))
3509 		return -EINVAL;
3510 
3511 	if (f2fs_readonly(sbi->sb))
3512 		return -EROFS;
3513 
3514 	if (!pin && IS_DEVICE_ALIASING(inode))
3515 		return -EOPNOTSUPP;
3516 
3517 	ret = mnt_want_write_file(filp);
3518 	if (ret)
3519 		return ret;
3520 
3521 	inode_lock(inode);
3522 
3523 	if (f2fs_is_atomic_file(inode)) {
3524 		ret = -EINVAL;
3525 		goto out;
3526 	}
3527 
3528 	if (!pin) {
3529 		clear_inode_flag(inode, FI_PIN_FILE);
3530 		f2fs_i_gc_failures_write(inode, 0);
3531 		goto done;
3532 	} else if (f2fs_is_pinned_file(inode)) {
3533 		goto done;
3534 	}
3535 
3536 	if (F2FS_HAS_BLOCKS(inode)) {
3537 		ret = -EFBIG;
3538 		goto out;
3539 	}
3540 
3541 	/* Let's allow file pinning on zoned device. */
3542 	if (!f2fs_sb_has_blkzoned(sbi) &&
3543 	    f2fs_should_update_outplace(inode, NULL)) {
3544 		ret = -EINVAL;
3545 		goto out;
3546 	}
3547 
3548 	if (f2fs_pin_file_control(inode, false)) {
3549 		ret = -EAGAIN;
3550 		goto out;
3551 	}
3552 
3553 	ret = f2fs_convert_inline_inode(inode);
3554 	if (ret)
3555 		goto out;
3556 
3557 	if (!f2fs_disable_compressed_file(inode)) {
3558 		ret = -EOPNOTSUPP;
3559 		goto out;
3560 	}
3561 
3562 	set_inode_flag(inode, FI_PIN_FILE);
3563 	ret = F2FS_I(inode)->i_gc_failures;
3564 done:
3565 	f2fs_update_time(sbi, REQ_TIME);
3566 out:
3567 	inode_unlock(inode);
3568 	mnt_drop_write_file(filp);
3569 	return ret;
3570 }
3571 
3572 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3573 {
3574 	struct inode *inode = file_inode(filp);
3575 	__u32 pin = 0;
3576 
3577 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3578 		pin = F2FS_I(inode)->i_gc_failures;
3579 	return put_user(pin, (u32 __user *)arg);
3580 }
3581 
3582 static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg)
3583 {
3584 	return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0,
3585 			(u32 __user *)arg);
3586 }
3587 
3588 static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg)
3589 {
3590 	struct inode *inode = file_inode(filp);
3591 	__u32 level;
3592 
3593 	if (get_user(level, (__u32 __user *)arg))
3594 		return -EFAULT;
3595 
3596 	if (!S_ISREG(inode->i_mode) || level >= F2FS_IOPRIO_MAX)
3597 		return -EINVAL;
3598 
3599 	inode_lock(inode);
3600 	F2FS_I(inode)->ioprio_hint = level;
3601 	inode_unlock(inode);
3602 	return 0;
3603 }
3604 
3605 int f2fs_precache_extents(struct inode *inode)
3606 {
3607 	struct f2fs_inode_info *fi = F2FS_I(inode);
3608 	struct f2fs_map_blocks map;
3609 	pgoff_t m_next_extent;
3610 	loff_t end;
3611 	int err;
3612 
3613 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3614 		return -EOPNOTSUPP;
3615 
3616 	map.m_lblk = 0;
3617 	map.m_pblk = 0;
3618 	map.m_next_pgofs = NULL;
3619 	map.m_next_extent = &m_next_extent;
3620 	map.m_seg_type = NO_CHECK_TYPE;
3621 	map.m_may_create = false;
3622 	end = F2FS_BLK_ALIGN(i_size_read(inode));
3623 
3624 	while (map.m_lblk < end) {
3625 		map.m_len = end - map.m_lblk;
3626 
3627 		f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3628 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
3629 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3630 		if (err || !map.m_len)
3631 			return err;
3632 
3633 		map.m_lblk = m_next_extent;
3634 	}
3635 
3636 	return 0;
3637 }
3638 
3639 static int f2fs_ioc_precache_extents(struct file *filp)
3640 {
3641 	return f2fs_precache_extents(file_inode(filp));
3642 }
3643 
3644 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3645 {
3646 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3647 	__u64 block_count;
3648 
3649 	if (!capable(CAP_SYS_ADMIN))
3650 		return -EPERM;
3651 
3652 	if (f2fs_readonly(sbi->sb))
3653 		return -EROFS;
3654 
3655 	if (copy_from_user(&block_count, (void __user *)arg,
3656 			   sizeof(block_count)))
3657 		return -EFAULT;
3658 
3659 	return f2fs_resize_fs(filp, block_count);
3660 }
3661 
3662 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3663 {
3664 	struct inode *inode = file_inode(filp);
3665 
3666 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3667 
3668 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3669 		f2fs_warn(F2FS_I_SB(inode),
3670 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3671 			  inode->i_ino);
3672 		return -EOPNOTSUPP;
3673 	}
3674 
3675 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3676 }
3677 
3678 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3679 {
3680 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3681 		return -EOPNOTSUPP;
3682 
3683 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3684 }
3685 
3686 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3687 {
3688 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3689 		return -EOPNOTSUPP;
3690 
3691 	return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3692 }
3693 
3694 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3695 {
3696 	struct inode *inode = file_inode(filp);
3697 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3698 	char *vbuf;
3699 	int count;
3700 	int err = 0;
3701 
3702 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3703 	if (!vbuf)
3704 		return -ENOMEM;
3705 
3706 	f2fs_down_read(&sbi->sb_lock);
3707 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3708 			ARRAY_SIZE(sbi->raw_super->volume_name),
3709 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3710 	f2fs_up_read(&sbi->sb_lock);
3711 
3712 	if (copy_to_user((char __user *)arg, vbuf,
3713 				min(FSLABEL_MAX, count)))
3714 		err = -EFAULT;
3715 
3716 	kfree(vbuf);
3717 	return err;
3718 }
3719 
3720 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3721 {
3722 	struct inode *inode = file_inode(filp);
3723 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3724 	char *vbuf;
3725 	int err = 0;
3726 
3727 	if (!capable(CAP_SYS_ADMIN))
3728 		return -EPERM;
3729 
3730 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3731 	if (IS_ERR(vbuf))
3732 		return PTR_ERR(vbuf);
3733 
3734 	err = mnt_want_write_file(filp);
3735 	if (err)
3736 		goto out;
3737 
3738 	f2fs_down_write(&sbi->sb_lock);
3739 
3740 	memset(sbi->raw_super->volume_name, 0,
3741 			sizeof(sbi->raw_super->volume_name));
3742 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3743 			sbi->raw_super->volume_name,
3744 			ARRAY_SIZE(sbi->raw_super->volume_name));
3745 
3746 	err = f2fs_commit_super(sbi, false);
3747 
3748 	f2fs_up_write(&sbi->sb_lock);
3749 
3750 	mnt_drop_write_file(filp);
3751 out:
3752 	kfree(vbuf);
3753 	return err;
3754 }
3755 
3756 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks)
3757 {
3758 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3759 		return -EOPNOTSUPP;
3760 
3761 	if (!f2fs_compressed_file(inode))
3762 		return -EINVAL;
3763 
3764 	*blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3765 
3766 	return 0;
3767 }
3768 
3769 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg)
3770 {
3771 	struct inode *inode = file_inode(filp);
3772 	__u64 blocks;
3773 	int ret;
3774 
3775 	ret = f2fs_get_compress_blocks(inode, &blocks);
3776 	if (ret < 0)
3777 		return ret;
3778 
3779 	return put_user(blocks, (u64 __user *)arg);
3780 }
3781 
3782 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3783 {
3784 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3785 	unsigned int released_blocks = 0;
3786 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3787 	block_t blkaddr;
3788 	int i;
3789 
3790 	for (i = 0; i < count; i++) {
3791 		blkaddr = data_blkaddr(dn->inode, dn->node_folio,
3792 						dn->ofs_in_node + i);
3793 
3794 		if (!__is_valid_data_blkaddr(blkaddr))
3795 			continue;
3796 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3797 					DATA_GENERIC_ENHANCE)))
3798 			return -EFSCORRUPTED;
3799 	}
3800 
3801 	while (count) {
3802 		int compr_blocks = 0;
3803 
3804 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3805 			blkaddr = f2fs_data_blkaddr(dn);
3806 
3807 			if (i == 0) {
3808 				if (blkaddr == COMPRESS_ADDR)
3809 					continue;
3810 				dn->ofs_in_node += cluster_size;
3811 				goto next;
3812 			}
3813 
3814 			if (__is_valid_data_blkaddr(blkaddr))
3815 				compr_blocks++;
3816 
3817 			if (blkaddr != NEW_ADDR)
3818 				continue;
3819 
3820 			f2fs_set_data_blkaddr(dn, NULL_ADDR);
3821 		}
3822 
3823 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3824 		dec_valid_block_count(sbi, dn->inode,
3825 					cluster_size - compr_blocks);
3826 
3827 		released_blocks += cluster_size - compr_blocks;
3828 next:
3829 		count -= cluster_size;
3830 	}
3831 
3832 	return released_blocks;
3833 }
3834 
3835 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3836 {
3837 	struct inode *inode = file_inode(filp);
3838 	struct f2fs_inode_info *fi = F2FS_I(inode);
3839 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3840 	pgoff_t page_idx = 0, last_idx;
3841 	unsigned int released_blocks = 0;
3842 	int ret;
3843 	int writecount;
3844 
3845 	if (!f2fs_sb_has_compression(sbi))
3846 		return -EOPNOTSUPP;
3847 
3848 	if (f2fs_readonly(sbi->sb))
3849 		return -EROFS;
3850 
3851 	ret = mnt_want_write_file(filp);
3852 	if (ret)
3853 		return ret;
3854 
3855 	f2fs_balance_fs(sbi, true);
3856 
3857 	inode_lock(inode);
3858 
3859 	writecount = atomic_read(&inode->i_writecount);
3860 	if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3861 			(!(filp->f_mode & FMODE_WRITE) && writecount)) {
3862 		ret = -EBUSY;
3863 		goto out;
3864 	}
3865 
3866 	if (!f2fs_compressed_file(inode) ||
3867 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3868 		ret = -EINVAL;
3869 		goto out;
3870 	}
3871 
3872 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3873 	if (ret)
3874 		goto out;
3875 
3876 	if (!atomic_read(&fi->i_compr_blocks)) {
3877 		ret = -EPERM;
3878 		goto out;
3879 	}
3880 
3881 	set_inode_flag(inode, FI_COMPRESS_RELEASED);
3882 	inode_set_ctime_current(inode);
3883 	f2fs_mark_inode_dirty_sync(inode, true);
3884 
3885 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3886 	filemap_invalidate_lock(inode->i_mapping);
3887 
3888 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3889 
3890 	while (page_idx < last_idx) {
3891 		struct dnode_of_data dn;
3892 		pgoff_t end_offset, count;
3893 
3894 		f2fs_lock_op(sbi);
3895 
3896 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3897 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3898 		if (ret) {
3899 			f2fs_unlock_op(sbi);
3900 			if (ret == -ENOENT) {
3901 				page_idx = f2fs_get_next_page_offset(&dn,
3902 								page_idx);
3903 				ret = 0;
3904 				continue;
3905 			}
3906 			break;
3907 		}
3908 
3909 		end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
3910 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3911 		count = round_up(count, fi->i_cluster_size);
3912 
3913 		ret = release_compress_blocks(&dn, count);
3914 
3915 		f2fs_put_dnode(&dn);
3916 
3917 		f2fs_unlock_op(sbi);
3918 
3919 		if (ret < 0)
3920 			break;
3921 
3922 		page_idx += count;
3923 		released_blocks += ret;
3924 	}
3925 
3926 	filemap_invalidate_unlock(inode->i_mapping);
3927 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3928 out:
3929 	if (released_blocks)
3930 		f2fs_update_time(sbi, REQ_TIME);
3931 	inode_unlock(inode);
3932 
3933 	mnt_drop_write_file(filp);
3934 
3935 	if (ret >= 0) {
3936 		ret = put_user(released_blocks, (u64 __user *)arg);
3937 	} else if (released_blocks &&
3938 			atomic_read(&fi->i_compr_blocks)) {
3939 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3940 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3941 			"iblocks=%llu, released=%u, compr_blocks=%u, "
3942 			"run fsck to fix.",
3943 			__func__, inode->i_ino, inode->i_blocks,
3944 			released_blocks,
3945 			atomic_read(&fi->i_compr_blocks));
3946 	}
3947 
3948 	return ret;
3949 }
3950 
3951 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count,
3952 		unsigned int *reserved_blocks)
3953 {
3954 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3955 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3956 	block_t blkaddr;
3957 	int i;
3958 
3959 	for (i = 0; i < count; i++) {
3960 		blkaddr = data_blkaddr(dn->inode, dn->node_folio,
3961 						dn->ofs_in_node + i);
3962 
3963 		if (!__is_valid_data_blkaddr(blkaddr))
3964 			continue;
3965 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3966 					DATA_GENERIC_ENHANCE)))
3967 			return -EFSCORRUPTED;
3968 	}
3969 
3970 	while (count) {
3971 		int compr_blocks = 0;
3972 		blkcnt_t reserved = 0;
3973 		blkcnt_t to_reserved;
3974 		int ret;
3975 
3976 		for (i = 0; i < cluster_size; i++) {
3977 			blkaddr = data_blkaddr(dn->inode, dn->node_folio,
3978 						dn->ofs_in_node + i);
3979 
3980 			if (i == 0) {
3981 				if (blkaddr != COMPRESS_ADDR) {
3982 					dn->ofs_in_node += cluster_size;
3983 					goto next;
3984 				}
3985 				continue;
3986 			}
3987 
3988 			/*
3989 			 * compressed cluster was not released due to it
3990 			 * fails in release_compress_blocks(), so NEW_ADDR
3991 			 * is a possible case.
3992 			 */
3993 			if (blkaddr == NEW_ADDR) {
3994 				reserved++;
3995 				continue;
3996 			}
3997 			if (__is_valid_data_blkaddr(blkaddr)) {
3998 				compr_blocks++;
3999 				continue;
4000 			}
4001 		}
4002 
4003 		to_reserved = cluster_size - compr_blocks - reserved;
4004 
4005 		/* for the case all blocks in cluster were reserved */
4006 		if (reserved && to_reserved == 1) {
4007 			dn->ofs_in_node += cluster_size;
4008 			goto next;
4009 		}
4010 
4011 		ret = inc_valid_block_count(sbi, dn->inode,
4012 						&to_reserved, false);
4013 		if (unlikely(ret))
4014 			return ret;
4015 
4016 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
4017 			if (f2fs_data_blkaddr(dn) == NULL_ADDR)
4018 				f2fs_set_data_blkaddr(dn, NEW_ADDR);
4019 		}
4020 
4021 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
4022 
4023 		*reserved_blocks += to_reserved;
4024 next:
4025 		count -= cluster_size;
4026 	}
4027 
4028 	return 0;
4029 }
4030 
4031 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
4032 {
4033 	struct inode *inode = file_inode(filp);
4034 	struct f2fs_inode_info *fi = F2FS_I(inode);
4035 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4036 	pgoff_t page_idx = 0, last_idx;
4037 	unsigned int reserved_blocks = 0;
4038 	int ret;
4039 
4040 	if (!f2fs_sb_has_compression(sbi))
4041 		return -EOPNOTSUPP;
4042 
4043 	if (f2fs_readonly(sbi->sb))
4044 		return -EROFS;
4045 
4046 	ret = mnt_want_write_file(filp);
4047 	if (ret)
4048 		return ret;
4049 
4050 	f2fs_balance_fs(sbi, true);
4051 
4052 	inode_lock(inode);
4053 
4054 	if (!f2fs_compressed_file(inode) ||
4055 		!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4056 		ret = -EINVAL;
4057 		goto unlock_inode;
4058 	}
4059 
4060 	if (atomic_read(&fi->i_compr_blocks))
4061 		goto unlock_inode;
4062 
4063 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
4064 	filemap_invalidate_lock(inode->i_mapping);
4065 
4066 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4067 
4068 	while (page_idx < last_idx) {
4069 		struct dnode_of_data dn;
4070 		pgoff_t end_offset, count;
4071 
4072 		f2fs_lock_op(sbi);
4073 
4074 		set_new_dnode(&dn, inode, NULL, NULL, 0);
4075 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
4076 		if (ret) {
4077 			f2fs_unlock_op(sbi);
4078 			if (ret == -ENOENT) {
4079 				page_idx = f2fs_get_next_page_offset(&dn,
4080 								page_idx);
4081 				ret = 0;
4082 				continue;
4083 			}
4084 			break;
4085 		}
4086 
4087 		end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
4088 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
4089 		count = round_up(count, fi->i_cluster_size);
4090 
4091 		ret = reserve_compress_blocks(&dn, count, &reserved_blocks);
4092 
4093 		f2fs_put_dnode(&dn);
4094 
4095 		f2fs_unlock_op(sbi);
4096 
4097 		if (ret < 0)
4098 			break;
4099 
4100 		page_idx += count;
4101 	}
4102 
4103 	filemap_invalidate_unlock(inode->i_mapping);
4104 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
4105 
4106 	if (!ret) {
4107 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
4108 		inode_set_ctime_current(inode);
4109 		f2fs_mark_inode_dirty_sync(inode, true);
4110 	}
4111 unlock_inode:
4112 	if (reserved_blocks)
4113 		f2fs_update_time(sbi, REQ_TIME);
4114 	inode_unlock(inode);
4115 	mnt_drop_write_file(filp);
4116 
4117 	if (!ret) {
4118 		ret = put_user(reserved_blocks, (u64 __user *)arg);
4119 	} else if (reserved_blocks &&
4120 			atomic_read(&fi->i_compr_blocks)) {
4121 		set_sbi_flag(sbi, SBI_NEED_FSCK);
4122 		f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx "
4123 			"iblocks=%llu, reserved=%u, compr_blocks=%u, "
4124 			"run fsck to fix.",
4125 			__func__, inode->i_ino, inode->i_blocks,
4126 			reserved_blocks,
4127 			atomic_read(&fi->i_compr_blocks));
4128 	}
4129 
4130 	return ret;
4131 }
4132 
4133 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
4134 		pgoff_t off, block_t block, block_t len, u32 flags)
4135 {
4136 	sector_t sector = SECTOR_FROM_BLOCK(block);
4137 	sector_t nr_sects = SECTOR_FROM_BLOCK(len);
4138 	int ret = 0;
4139 
4140 	if (flags & F2FS_TRIM_FILE_DISCARD) {
4141 		if (bdev_max_secure_erase_sectors(bdev))
4142 			ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
4143 					GFP_NOFS);
4144 		else
4145 			ret = blkdev_issue_discard(bdev, sector, nr_sects,
4146 					GFP_NOFS);
4147 	}
4148 
4149 	if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
4150 		if (IS_ENCRYPTED(inode))
4151 			ret = fscrypt_zeroout_range(inode, off, block, len);
4152 		else
4153 			ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
4154 					GFP_NOFS, 0);
4155 	}
4156 
4157 	return ret;
4158 }
4159 
4160 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
4161 {
4162 	struct inode *inode = file_inode(filp);
4163 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4164 	struct address_space *mapping = inode->i_mapping;
4165 	struct block_device *prev_bdev = NULL;
4166 	struct f2fs_sectrim_range range;
4167 	pgoff_t index, pg_end, prev_index = 0;
4168 	block_t prev_block = 0, len = 0;
4169 	loff_t end_addr;
4170 	bool to_end = false;
4171 	int ret = 0;
4172 
4173 	if (!(filp->f_mode & FMODE_WRITE))
4174 		return -EBADF;
4175 
4176 	if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
4177 				sizeof(range)))
4178 		return -EFAULT;
4179 
4180 	if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
4181 			!S_ISREG(inode->i_mode))
4182 		return -EINVAL;
4183 
4184 	if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
4185 			!f2fs_hw_support_discard(sbi)) ||
4186 			((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
4187 			 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
4188 		return -EOPNOTSUPP;
4189 
4190 	ret = mnt_want_write_file(filp);
4191 	if (ret)
4192 		return ret;
4193 	inode_lock(inode);
4194 
4195 	if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
4196 			range.start >= inode->i_size) {
4197 		ret = -EINVAL;
4198 		goto err;
4199 	}
4200 
4201 	if (range.len == 0)
4202 		goto err;
4203 
4204 	if (inode->i_size - range.start > range.len) {
4205 		end_addr = range.start + range.len;
4206 	} else {
4207 		end_addr = range.len == (u64)-1 ?
4208 			sbi->sb->s_maxbytes : inode->i_size;
4209 		to_end = true;
4210 	}
4211 
4212 	if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
4213 			(!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
4214 		ret = -EINVAL;
4215 		goto err;
4216 	}
4217 
4218 	index = F2FS_BYTES_TO_BLK(range.start);
4219 	pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
4220 
4221 	ret = f2fs_convert_inline_inode(inode);
4222 	if (ret)
4223 		goto err;
4224 
4225 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4226 	filemap_invalidate_lock(mapping);
4227 
4228 	ret = filemap_write_and_wait_range(mapping, range.start,
4229 			to_end ? LLONG_MAX : end_addr - 1);
4230 	if (ret)
4231 		goto out;
4232 
4233 	truncate_inode_pages_range(mapping, range.start,
4234 			to_end ? -1 : end_addr - 1);
4235 
4236 	while (index < pg_end) {
4237 		struct dnode_of_data dn;
4238 		pgoff_t end_offset, count;
4239 		int i;
4240 
4241 		set_new_dnode(&dn, inode, NULL, NULL, 0);
4242 		ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4243 		if (ret) {
4244 			if (ret == -ENOENT) {
4245 				index = f2fs_get_next_page_offset(&dn, index);
4246 				continue;
4247 			}
4248 			goto out;
4249 		}
4250 
4251 		end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
4252 		count = min(end_offset - dn.ofs_in_node, pg_end - index);
4253 		for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
4254 			struct block_device *cur_bdev;
4255 			block_t blkaddr = f2fs_data_blkaddr(&dn);
4256 
4257 			if (!__is_valid_data_blkaddr(blkaddr))
4258 				continue;
4259 
4260 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
4261 						DATA_GENERIC_ENHANCE)) {
4262 				ret = -EFSCORRUPTED;
4263 				f2fs_put_dnode(&dn);
4264 				goto out;
4265 			}
4266 
4267 			cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
4268 			if (f2fs_is_multi_device(sbi)) {
4269 				int di = f2fs_target_device_index(sbi, blkaddr);
4270 
4271 				blkaddr -= FDEV(di).start_blk;
4272 			}
4273 
4274 			if (len) {
4275 				if (prev_bdev == cur_bdev &&
4276 						index == prev_index + len &&
4277 						blkaddr == prev_block + len) {
4278 					len++;
4279 				} else {
4280 					ret = f2fs_secure_erase(prev_bdev,
4281 						inode, prev_index, prev_block,
4282 						len, range.flags);
4283 					if (ret) {
4284 						f2fs_put_dnode(&dn);
4285 						goto out;
4286 					}
4287 
4288 					len = 0;
4289 				}
4290 			}
4291 
4292 			if (!len) {
4293 				prev_bdev = cur_bdev;
4294 				prev_index = index;
4295 				prev_block = blkaddr;
4296 				len = 1;
4297 			}
4298 		}
4299 
4300 		f2fs_put_dnode(&dn);
4301 
4302 		if (fatal_signal_pending(current)) {
4303 			ret = -EINTR;
4304 			goto out;
4305 		}
4306 		cond_resched();
4307 	}
4308 
4309 	if (len)
4310 		ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
4311 				prev_block, len, range.flags);
4312 	f2fs_update_time(sbi, REQ_TIME);
4313 out:
4314 	filemap_invalidate_unlock(mapping);
4315 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4316 err:
4317 	inode_unlock(inode);
4318 	mnt_drop_write_file(filp);
4319 
4320 	return ret;
4321 }
4322 
4323 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
4324 {
4325 	struct inode *inode = file_inode(filp);
4326 	struct f2fs_comp_option option;
4327 
4328 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4329 		return -EOPNOTSUPP;
4330 
4331 	inode_lock_shared(inode);
4332 
4333 	if (!f2fs_compressed_file(inode)) {
4334 		inode_unlock_shared(inode);
4335 		return -ENODATA;
4336 	}
4337 
4338 	option.algorithm = F2FS_I(inode)->i_compress_algorithm;
4339 	option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
4340 
4341 	inode_unlock_shared(inode);
4342 
4343 	if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
4344 				sizeof(option)))
4345 		return -EFAULT;
4346 
4347 	return 0;
4348 }
4349 
4350 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4351 {
4352 	struct inode *inode = file_inode(filp);
4353 	struct f2fs_inode_info *fi = F2FS_I(inode);
4354 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4355 	struct f2fs_comp_option option;
4356 	int ret = 0;
4357 
4358 	if (!f2fs_sb_has_compression(sbi))
4359 		return -EOPNOTSUPP;
4360 
4361 	if (!(filp->f_mode & FMODE_WRITE))
4362 		return -EBADF;
4363 
4364 	if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4365 				sizeof(option)))
4366 		return -EFAULT;
4367 
4368 	if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
4369 		option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
4370 		option.algorithm >= COMPRESS_MAX)
4371 		return -EINVAL;
4372 
4373 	ret = mnt_want_write_file(filp);
4374 	if (ret)
4375 		return ret;
4376 	inode_lock(inode);
4377 
4378 	f2fs_down_write(&F2FS_I(inode)->i_sem);
4379 	if (!f2fs_compressed_file(inode)) {
4380 		ret = -EINVAL;
4381 		goto out;
4382 	}
4383 
4384 	if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4385 		ret = -EBUSY;
4386 		goto out;
4387 	}
4388 
4389 	if (F2FS_HAS_BLOCKS(inode)) {
4390 		ret = -EFBIG;
4391 		goto out;
4392 	}
4393 
4394 	fi->i_compress_algorithm = option.algorithm;
4395 	fi->i_log_cluster_size = option.log_cluster_size;
4396 	fi->i_cluster_size = BIT(option.log_cluster_size);
4397 	/* Set default level */
4398 	if (fi->i_compress_algorithm == COMPRESS_ZSTD)
4399 		fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
4400 	else
4401 		fi->i_compress_level = 0;
4402 	/* Adjust mount option level */
4403 	if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm &&
4404 	    F2FS_OPTION(sbi).compress_level)
4405 		fi->i_compress_level = F2FS_OPTION(sbi).compress_level;
4406 	f2fs_mark_inode_dirty_sync(inode, true);
4407 
4408 	if (!f2fs_is_compress_backend_ready(inode))
4409 		f2fs_warn(sbi, "compression algorithm is successfully set, "
4410 			"but current kernel doesn't support this algorithm.");
4411 out:
4412 	f2fs_up_write(&fi->i_sem);
4413 	inode_unlock(inode);
4414 	mnt_drop_write_file(filp);
4415 
4416 	return ret;
4417 }
4418 
4419 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4420 {
4421 	DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
4422 	struct address_space *mapping = inode->i_mapping;
4423 	struct folio *folio;
4424 	pgoff_t redirty_idx = page_idx;
4425 	int page_len = 0, ret = 0;
4426 
4427 	page_cache_ra_unbounded(&ractl, len, 0);
4428 
4429 	do {
4430 		folio = read_cache_folio(mapping, page_idx, NULL, NULL);
4431 		if (IS_ERR(folio)) {
4432 			ret = PTR_ERR(folio);
4433 			break;
4434 		}
4435 		page_len += folio_nr_pages(folio) - (page_idx - folio->index);
4436 		page_idx = folio_next_index(folio);
4437 	} while (page_len < len);
4438 
4439 	do {
4440 		folio = filemap_lock_folio(mapping, redirty_idx);
4441 
4442 		/* It will never fail, when folio has pinned above */
4443 		f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(folio));
4444 
4445 		f2fs_folio_wait_writeback(folio, DATA, true, true);
4446 
4447 		folio_mark_dirty(folio);
4448 		folio_set_f2fs_gcing(folio);
4449 		redirty_idx = folio_next_index(folio);
4450 		folio_unlock(folio);
4451 		folio_put_refs(folio, 2);
4452 	} while (redirty_idx < page_idx);
4453 
4454 	return ret;
4455 }
4456 
4457 static int f2fs_ioc_decompress_file(struct file *filp)
4458 {
4459 	struct inode *inode = file_inode(filp);
4460 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4461 	struct f2fs_inode_info *fi = F2FS_I(inode);
4462 	pgoff_t page_idx = 0, last_idx, cluster_idx;
4463 	int ret;
4464 
4465 	if (!f2fs_sb_has_compression(sbi) ||
4466 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4467 		return -EOPNOTSUPP;
4468 
4469 	if (!(filp->f_mode & FMODE_WRITE))
4470 		return -EBADF;
4471 
4472 	f2fs_balance_fs(sbi, true);
4473 
4474 	ret = mnt_want_write_file(filp);
4475 	if (ret)
4476 		return ret;
4477 	inode_lock(inode);
4478 
4479 	if (!f2fs_is_compress_backend_ready(inode)) {
4480 		ret = -EOPNOTSUPP;
4481 		goto out;
4482 	}
4483 
4484 	if (!f2fs_compressed_file(inode) ||
4485 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4486 		ret = -EINVAL;
4487 		goto out;
4488 	}
4489 
4490 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4491 	if (ret)
4492 		goto out;
4493 
4494 	if (!atomic_read(&fi->i_compr_blocks))
4495 		goto out;
4496 
4497 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4498 	last_idx >>= fi->i_log_cluster_size;
4499 
4500 	for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4501 		page_idx = cluster_idx << fi->i_log_cluster_size;
4502 
4503 		if (!f2fs_is_compressed_cluster(inode, page_idx))
4504 			continue;
4505 
4506 		ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4507 		if (ret < 0)
4508 			break;
4509 
4510 		if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4511 			ret = filemap_fdatawrite(inode->i_mapping);
4512 			if (ret < 0)
4513 				break;
4514 		}
4515 
4516 		cond_resched();
4517 		if (fatal_signal_pending(current)) {
4518 			ret = -EINTR;
4519 			break;
4520 		}
4521 	}
4522 
4523 	if (!ret)
4524 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4525 							LLONG_MAX);
4526 
4527 	if (ret)
4528 		f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4529 			  __func__, ret);
4530 	f2fs_update_time(sbi, REQ_TIME);
4531 out:
4532 	inode_unlock(inode);
4533 	mnt_drop_write_file(filp);
4534 
4535 	return ret;
4536 }
4537 
4538 static int f2fs_ioc_compress_file(struct file *filp)
4539 {
4540 	struct inode *inode = file_inode(filp);
4541 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4542 	struct f2fs_inode_info *fi = F2FS_I(inode);
4543 	pgoff_t page_idx = 0, last_idx, cluster_idx;
4544 	int ret;
4545 
4546 	if (!f2fs_sb_has_compression(sbi) ||
4547 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4548 		return -EOPNOTSUPP;
4549 
4550 	if (!(filp->f_mode & FMODE_WRITE))
4551 		return -EBADF;
4552 
4553 	f2fs_balance_fs(sbi, true);
4554 
4555 	ret = mnt_want_write_file(filp);
4556 	if (ret)
4557 		return ret;
4558 	inode_lock(inode);
4559 
4560 	if (!f2fs_is_compress_backend_ready(inode)) {
4561 		ret = -EOPNOTSUPP;
4562 		goto out;
4563 	}
4564 
4565 	if (!f2fs_compressed_file(inode) ||
4566 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4567 		ret = -EINVAL;
4568 		goto out;
4569 	}
4570 
4571 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4572 	if (ret)
4573 		goto out;
4574 
4575 	set_inode_flag(inode, FI_ENABLE_COMPRESS);
4576 
4577 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4578 	last_idx >>= fi->i_log_cluster_size;
4579 
4580 	for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4581 		page_idx = cluster_idx << fi->i_log_cluster_size;
4582 
4583 		if (f2fs_is_sparse_cluster(inode, page_idx))
4584 			continue;
4585 
4586 		ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4587 		if (ret < 0)
4588 			break;
4589 
4590 		if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4591 			ret = filemap_fdatawrite(inode->i_mapping);
4592 			if (ret < 0)
4593 				break;
4594 		}
4595 
4596 		cond_resched();
4597 		if (fatal_signal_pending(current)) {
4598 			ret = -EINTR;
4599 			break;
4600 		}
4601 	}
4602 
4603 	if (!ret)
4604 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4605 							LLONG_MAX);
4606 
4607 	clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4608 
4609 	if (ret)
4610 		f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4611 			  __func__, ret);
4612 	f2fs_update_time(sbi, REQ_TIME);
4613 out:
4614 	inode_unlock(inode);
4615 	mnt_drop_write_file(filp);
4616 
4617 	return ret;
4618 }
4619 
4620 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4621 {
4622 	switch (cmd) {
4623 	case FS_IOC_GETVERSION:
4624 		return f2fs_ioc_getversion(filp, arg);
4625 	case F2FS_IOC_START_ATOMIC_WRITE:
4626 		return f2fs_ioc_start_atomic_write(filp, false);
4627 	case F2FS_IOC_START_ATOMIC_REPLACE:
4628 		return f2fs_ioc_start_atomic_write(filp, true);
4629 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4630 		return f2fs_ioc_commit_atomic_write(filp);
4631 	case F2FS_IOC_ABORT_ATOMIC_WRITE:
4632 		return f2fs_ioc_abort_atomic_write(filp);
4633 	case F2FS_IOC_START_VOLATILE_WRITE:
4634 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4635 		return -EOPNOTSUPP;
4636 	case F2FS_IOC_SHUTDOWN:
4637 		return f2fs_ioc_shutdown(filp, arg);
4638 	case FITRIM:
4639 		return f2fs_ioc_fitrim(filp, arg);
4640 	case FS_IOC_SET_ENCRYPTION_POLICY:
4641 		return f2fs_ioc_set_encryption_policy(filp, arg);
4642 	case FS_IOC_GET_ENCRYPTION_POLICY:
4643 		return f2fs_ioc_get_encryption_policy(filp, arg);
4644 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4645 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4646 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4647 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4648 	case FS_IOC_ADD_ENCRYPTION_KEY:
4649 		return f2fs_ioc_add_encryption_key(filp, arg);
4650 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4651 		return f2fs_ioc_remove_encryption_key(filp, arg);
4652 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4653 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4654 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4655 		return f2fs_ioc_get_encryption_key_status(filp, arg);
4656 	case FS_IOC_GET_ENCRYPTION_NONCE:
4657 		return f2fs_ioc_get_encryption_nonce(filp, arg);
4658 	case F2FS_IOC_GARBAGE_COLLECT:
4659 		return f2fs_ioc_gc(filp, arg);
4660 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4661 		return f2fs_ioc_gc_range(filp, arg);
4662 	case F2FS_IOC_WRITE_CHECKPOINT:
4663 		return f2fs_ioc_write_checkpoint(filp);
4664 	case F2FS_IOC_DEFRAGMENT:
4665 		return f2fs_ioc_defragment(filp, arg);
4666 	case F2FS_IOC_MOVE_RANGE:
4667 		return f2fs_ioc_move_range(filp, arg);
4668 	case F2FS_IOC_FLUSH_DEVICE:
4669 		return f2fs_ioc_flush_device(filp, arg);
4670 	case F2FS_IOC_GET_FEATURES:
4671 		return f2fs_ioc_get_features(filp, arg);
4672 	case F2FS_IOC_GET_PIN_FILE:
4673 		return f2fs_ioc_get_pin_file(filp, arg);
4674 	case F2FS_IOC_SET_PIN_FILE:
4675 		return f2fs_ioc_set_pin_file(filp, arg);
4676 	case F2FS_IOC_PRECACHE_EXTENTS:
4677 		return f2fs_ioc_precache_extents(filp);
4678 	case F2FS_IOC_RESIZE_FS:
4679 		return f2fs_ioc_resize_fs(filp, arg);
4680 	case FS_IOC_ENABLE_VERITY:
4681 		return f2fs_ioc_enable_verity(filp, arg);
4682 	case FS_IOC_MEASURE_VERITY:
4683 		return f2fs_ioc_measure_verity(filp, arg);
4684 	case FS_IOC_READ_VERITY_METADATA:
4685 		return f2fs_ioc_read_verity_metadata(filp, arg);
4686 	case FS_IOC_GETFSLABEL:
4687 		return f2fs_ioc_getfslabel(filp, arg);
4688 	case FS_IOC_SETFSLABEL:
4689 		return f2fs_ioc_setfslabel(filp, arg);
4690 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4691 		return f2fs_ioc_get_compress_blocks(filp, arg);
4692 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4693 		return f2fs_release_compress_blocks(filp, arg);
4694 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4695 		return f2fs_reserve_compress_blocks(filp, arg);
4696 	case F2FS_IOC_SEC_TRIM_FILE:
4697 		return f2fs_sec_trim_file(filp, arg);
4698 	case F2FS_IOC_GET_COMPRESS_OPTION:
4699 		return f2fs_ioc_get_compress_option(filp, arg);
4700 	case F2FS_IOC_SET_COMPRESS_OPTION:
4701 		return f2fs_ioc_set_compress_option(filp, arg);
4702 	case F2FS_IOC_DECOMPRESS_FILE:
4703 		return f2fs_ioc_decompress_file(filp);
4704 	case F2FS_IOC_COMPRESS_FILE:
4705 		return f2fs_ioc_compress_file(filp);
4706 	case F2FS_IOC_GET_DEV_ALIAS_FILE:
4707 		return f2fs_ioc_get_dev_alias_file(filp, arg);
4708 	case F2FS_IOC_IO_PRIO:
4709 		return f2fs_ioc_io_prio(filp, arg);
4710 	default:
4711 		return -ENOTTY;
4712 	}
4713 }
4714 
4715 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4716 {
4717 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4718 		return -EIO;
4719 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4720 		return -ENOSPC;
4721 
4722 	return __f2fs_ioctl(filp, cmd, arg);
4723 }
4724 
4725 /*
4726  * Return %true if the given read or write request should use direct I/O, or
4727  * %false if it should use buffered I/O.
4728  */
4729 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4730 				struct iov_iter *iter)
4731 {
4732 	unsigned int align;
4733 
4734 	if (!(iocb->ki_flags & IOCB_DIRECT))
4735 		return false;
4736 
4737 	if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4738 		return false;
4739 
4740 	/*
4741 	 * Direct I/O not aligned to the disk's logical_block_size will be
4742 	 * attempted, but will fail with -EINVAL.
4743 	 *
4744 	 * f2fs additionally requires that direct I/O be aligned to the
4745 	 * filesystem block size, which is often a stricter requirement.
4746 	 * However, f2fs traditionally falls back to buffered I/O on requests
4747 	 * that are logical_block_size-aligned but not fs-block aligned.
4748 	 *
4749 	 * The below logic implements this behavior.
4750 	 */
4751 	align = iocb->ki_pos | iov_iter_alignment(iter);
4752 	if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4753 	    IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4754 		return false;
4755 
4756 	return true;
4757 }
4758 
4759 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4760 				unsigned int flags)
4761 {
4762 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4763 
4764 	dec_page_count(sbi, F2FS_DIO_READ);
4765 	if (error)
4766 		return error;
4767 	f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4768 	return 0;
4769 }
4770 
4771 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4772 	.end_io = f2fs_dio_read_end_io,
4773 };
4774 
4775 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4776 {
4777 	struct file *file = iocb->ki_filp;
4778 	struct inode *inode = file_inode(file);
4779 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4780 	struct f2fs_inode_info *fi = F2FS_I(inode);
4781 	const loff_t pos = iocb->ki_pos;
4782 	const size_t count = iov_iter_count(to);
4783 	struct iomap_dio *dio;
4784 	ssize_t ret;
4785 
4786 	if (count == 0)
4787 		return 0; /* skip atime update */
4788 
4789 	trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4790 
4791 	if (iocb->ki_flags & IOCB_NOWAIT) {
4792 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4793 			ret = -EAGAIN;
4794 			goto out;
4795 		}
4796 	} else {
4797 		f2fs_down_read(&fi->i_gc_rwsem[READ]);
4798 	}
4799 
4800 	/* dio is not compatible w/ atomic file */
4801 	if (f2fs_is_atomic_file(inode)) {
4802 		f2fs_up_read(&fi->i_gc_rwsem[READ]);
4803 		ret = -EOPNOTSUPP;
4804 		goto out;
4805 	}
4806 
4807 	/*
4808 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4809 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4810 	 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4811 	 */
4812 	inc_page_count(sbi, F2FS_DIO_READ);
4813 	dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4814 			     &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4815 	if (IS_ERR_OR_NULL(dio)) {
4816 		ret = PTR_ERR_OR_ZERO(dio);
4817 		if (ret != -EIOCBQUEUED)
4818 			dec_page_count(sbi, F2FS_DIO_READ);
4819 	} else {
4820 		ret = iomap_dio_complete(dio);
4821 	}
4822 
4823 	f2fs_up_read(&fi->i_gc_rwsem[READ]);
4824 
4825 	file_accessed(file);
4826 out:
4827 	trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4828 	return ret;
4829 }
4830 
4831 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count,
4832 				    int rw)
4833 {
4834 	struct inode *inode = file_inode(file);
4835 	char *buf, *path;
4836 
4837 	buf = f2fs_getname(F2FS_I_SB(inode));
4838 	if (!buf)
4839 		return;
4840 	path = dentry_path_raw(file_dentry(file), buf, PATH_MAX);
4841 	if (IS_ERR(path))
4842 		goto free_buf;
4843 	if (rw == WRITE)
4844 		trace_f2fs_datawrite_start(inode, pos, count,
4845 				current->pid, path, current->comm);
4846 	else
4847 		trace_f2fs_dataread_start(inode, pos, count,
4848 				current->pid, path, current->comm);
4849 free_buf:
4850 	f2fs_putname(buf);
4851 }
4852 
4853 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4854 {
4855 	struct inode *inode = file_inode(iocb->ki_filp);
4856 	const loff_t pos = iocb->ki_pos;
4857 	ssize_t ret;
4858 	bool dio;
4859 
4860 	if (!f2fs_is_compress_backend_ready(inode))
4861 		return -EOPNOTSUPP;
4862 
4863 	if (trace_f2fs_dataread_start_enabled())
4864 		f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
4865 					iov_iter_count(to), READ);
4866 
4867 	dio = f2fs_should_use_dio(inode, iocb, to);
4868 
4869 	/* In LFS mode, if there is inflight dio, wait for its completion */
4870 	if (f2fs_lfs_mode(F2FS_I_SB(inode)) &&
4871 	    get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE) &&
4872 		(!f2fs_is_pinned_file(inode) || !dio))
4873 		inode_dio_wait(inode);
4874 
4875 	if (dio) {
4876 		ret = f2fs_dio_read_iter(iocb, to);
4877 	} else {
4878 		ret = filemap_read(iocb, to, 0);
4879 		if (ret > 0)
4880 			f2fs_update_iostat(F2FS_I_SB(inode), inode,
4881 						APP_BUFFERED_READ_IO, ret);
4882 	}
4883 	trace_f2fs_dataread_end(inode, pos, ret);
4884 	return ret;
4885 }
4886 
4887 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos,
4888 				     struct pipe_inode_info *pipe,
4889 				     size_t len, unsigned int flags)
4890 {
4891 	struct inode *inode = file_inode(in);
4892 	const loff_t pos = *ppos;
4893 	ssize_t ret;
4894 
4895 	if (!f2fs_is_compress_backend_ready(inode))
4896 		return -EOPNOTSUPP;
4897 
4898 	if (trace_f2fs_dataread_start_enabled())
4899 		f2fs_trace_rw_file_path(in, pos, len, READ);
4900 
4901 	ret = filemap_splice_read(in, ppos, pipe, len, flags);
4902 	if (ret > 0)
4903 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
4904 				   APP_BUFFERED_READ_IO, ret);
4905 
4906 	trace_f2fs_dataread_end(inode, pos, ret);
4907 	return ret;
4908 }
4909 
4910 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4911 {
4912 	struct file *file = iocb->ki_filp;
4913 	struct inode *inode = file_inode(file);
4914 	ssize_t count;
4915 	int err;
4916 
4917 	if (IS_IMMUTABLE(inode))
4918 		return -EPERM;
4919 
4920 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4921 		return -EPERM;
4922 
4923 	count = generic_write_checks(iocb, from);
4924 	if (count <= 0)
4925 		return count;
4926 
4927 	err = file_modified(file);
4928 	if (err)
4929 		return err;
4930 
4931 	f2fs_zero_post_eof_page(inode,
4932 		iocb->ki_pos + iov_iter_count(from), true);
4933 	return count;
4934 }
4935 
4936 /*
4937  * Preallocate blocks for a write request, if it is possible and helpful to do
4938  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4939  * blocks were preallocated, or a negative errno value if something went
4940  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4941  * requested blocks (not just some of them) have been allocated.
4942  */
4943 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4944 				   bool dio)
4945 {
4946 	struct inode *inode = file_inode(iocb->ki_filp);
4947 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4948 	const loff_t pos = iocb->ki_pos;
4949 	const size_t count = iov_iter_count(iter);
4950 	struct f2fs_map_blocks map = {};
4951 	int flag;
4952 	int ret;
4953 
4954 	/* If it will be an out-of-place direct write, don't bother. */
4955 	if (dio && f2fs_lfs_mode(sbi))
4956 		return 0;
4957 	/*
4958 	 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4959 	 * buffered IO, if DIO meets any holes.
4960 	 */
4961 	if (dio && i_size_read(inode) &&
4962 		(F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4963 		return 0;
4964 
4965 	/* No-wait I/O can't allocate blocks. */
4966 	if (iocb->ki_flags & IOCB_NOWAIT)
4967 		return 0;
4968 
4969 	/* If it will be a short write, don't bother. */
4970 	if (fault_in_iov_iter_readable(iter, count))
4971 		return 0;
4972 
4973 	if (f2fs_has_inline_data(inode)) {
4974 		/* If the data will fit inline, don't bother. */
4975 		if (pos + count <= MAX_INLINE_DATA(inode))
4976 			return 0;
4977 		ret = f2fs_convert_inline_inode(inode);
4978 		if (ret)
4979 			return ret;
4980 	}
4981 
4982 	/* Do not preallocate blocks that will be written partially in 4KB. */
4983 	map.m_lblk = F2FS_BLK_ALIGN(pos);
4984 	map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4985 	if (map.m_len > map.m_lblk)
4986 		map.m_len -= map.m_lblk;
4987 	else
4988 		return 0;
4989 
4990 	if (!IS_DEVICE_ALIASING(inode))
4991 		map.m_may_create = true;
4992 	if (dio) {
4993 		map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi,
4994 						inode->i_write_hint);
4995 		flag = F2FS_GET_BLOCK_PRE_DIO;
4996 	} else {
4997 		map.m_seg_type = NO_CHECK_TYPE;
4998 		flag = F2FS_GET_BLOCK_PRE_AIO;
4999 	}
5000 
5001 	ret = f2fs_map_blocks(inode, &map, flag);
5002 	/* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
5003 	if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
5004 		return ret;
5005 	if (ret == 0)
5006 		set_inode_flag(inode, FI_PREALLOCATED_ALL);
5007 	return map.m_len;
5008 }
5009 
5010 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
5011 					struct iov_iter *from)
5012 {
5013 	struct file *file = iocb->ki_filp;
5014 	struct inode *inode = file_inode(file);
5015 	ssize_t ret;
5016 
5017 	if (iocb->ki_flags & IOCB_NOWAIT)
5018 		return -EOPNOTSUPP;
5019 
5020 	ret = generic_perform_write(iocb, from);
5021 
5022 	if (ret > 0) {
5023 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
5024 						APP_BUFFERED_IO, ret);
5025 	}
5026 	return ret;
5027 }
5028 
5029 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
5030 				 unsigned int flags)
5031 {
5032 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
5033 
5034 	dec_page_count(sbi, F2FS_DIO_WRITE);
5035 	if (error)
5036 		return error;
5037 	f2fs_update_time(sbi, REQ_TIME);
5038 	f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
5039 	return 0;
5040 }
5041 
5042 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter,
5043 					struct bio *bio, loff_t file_offset)
5044 {
5045 	struct inode *inode = iter->inode;
5046 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5047 	enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint);
5048 	enum temp_type temp = f2fs_get_segment_temp(sbi, type);
5049 
5050 	bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp);
5051 	blk_crypto_submit_bio(bio);
5052 }
5053 
5054 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
5055 	.end_io		= f2fs_dio_write_end_io,
5056 	.submit_io	= f2fs_dio_write_submit_io,
5057 };
5058 
5059 static void f2fs_flush_buffered_write(struct address_space *mapping,
5060 				      loff_t start_pos, loff_t end_pos)
5061 {
5062 	int ret;
5063 
5064 	ret = filemap_write_and_wait_range(mapping, start_pos, end_pos);
5065 	if (ret < 0)
5066 		return;
5067 	invalidate_mapping_pages(mapping,
5068 				 start_pos >> PAGE_SHIFT,
5069 				 end_pos >> PAGE_SHIFT);
5070 }
5071 
5072 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
5073 				   bool *may_need_sync)
5074 {
5075 	struct file *file = iocb->ki_filp;
5076 	struct inode *inode = file_inode(file);
5077 	struct f2fs_inode_info *fi = F2FS_I(inode);
5078 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5079 	const bool do_opu = f2fs_lfs_mode(sbi);
5080 	const loff_t pos = iocb->ki_pos;
5081 	const ssize_t count = iov_iter_count(from);
5082 	unsigned int dio_flags;
5083 	struct iomap_dio *dio;
5084 	ssize_t ret;
5085 
5086 	trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
5087 
5088 	if (iocb->ki_flags & IOCB_NOWAIT) {
5089 		/* f2fs_convert_inline_inode() and block allocation can block */
5090 		if (f2fs_has_inline_data(inode) ||
5091 		    !f2fs_overwrite_io(inode, pos, count)) {
5092 			ret = -EAGAIN;
5093 			goto out;
5094 		}
5095 
5096 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
5097 			ret = -EAGAIN;
5098 			goto out;
5099 		}
5100 		if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
5101 			f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
5102 			ret = -EAGAIN;
5103 			goto out;
5104 		}
5105 	} else {
5106 		ret = f2fs_convert_inline_inode(inode);
5107 		if (ret)
5108 			goto out;
5109 
5110 		f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
5111 		if (do_opu)
5112 			f2fs_down_read(&fi->i_gc_rwsem[READ]);
5113 	}
5114 
5115 	/*
5116 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
5117 	 * the higher-level function iomap_dio_rw() in order to ensure that the
5118 	 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
5119 	 */
5120 	inc_page_count(sbi, F2FS_DIO_WRITE);
5121 	dio_flags = 0;
5122 	if (pos + count > inode->i_size)
5123 		dio_flags |= IOMAP_DIO_FORCE_WAIT;
5124 	dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
5125 			     &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
5126 	if (IS_ERR_OR_NULL(dio)) {
5127 		ret = PTR_ERR_OR_ZERO(dio);
5128 		if (ret == -ENOTBLK)
5129 			ret = 0;
5130 		if (ret != -EIOCBQUEUED)
5131 			dec_page_count(sbi, F2FS_DIO_WRITE);
5132 	} else {
5133 		ret = iomap_dio_complete(dio);
5134 	}
5135 
5136 	if (do_opu)
5137 		f2fs_up_read(&fi->i_gc_rwsem[READ]);
5138 	f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
5139 
5140 	if (ret < 0)
5141 		goto out;
5142 	if (pos + ret > inode->i_size)
5143 		f2fs_i_size_write(inode, pos + ret);
5144 	if (!do_opu)
5145 		set_inode_flag(inode, FI_UPDATE_WRITE);
5146 
5147 	if (iov_iter_count(from)) {
5148 		ssize_t ret2;
5149 		loff_t bufio_start_pos = iocb->ki_pos;
5150 
5151 		/*
5152 		 * The direct write was partial, so we need to fall back to a
5153 		 * buffered write for the remainder.
5154 		 */
5155 
5156 		ret2 = f2fs_buffered_write_iter(iocb, from);
5157 		if (iov_iter_count(from))
5158 			f2fs_write_failed(inode, iocb->ki_pos);
5159 		if (ret2 < 0)
5160 			goto out;
5161 
5162 		/*
5163 		 * Ensure that the pagecache pages are written to disk and
5164 		 * invalidated to preserve the expected O_DIRECT semantics.
5165 		 */
5166 		if (ret2 > 0) {
5167 			loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
5168 
5169 			ret += ret2;
5170 
5171 			f2fs_flush_buffered_write(file->f_mapping,
5172 						  bufio_start_pos,
5173 						  bufio_end_pos);
5174 		}
5175 	} else {
5176 		/* iomap_dio_rw() already handled the generic_write_sync(). */
5177 		*may_need_sync = false;
5178 	}
5179 out:
5180 	trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
5181 	return ret;
5182 }
5183 
5184 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
5185 {
5186 	struct inode *inode = file_inode(iocb->ki_filp);
5187 	const loff_t orig_pos = iocb->ki_pos;
5188 	const size_t orig_count = iov_iter_count(from);
5189 	loff_t target_size;
5190 	bool dio;
5191 	bool may_need_sync = true;
5192 	int preallocated;
5193 	const loff_t pos = iocb->ki_pos;
5194 	const ssize_t count = iov_iter_count(from);
5195 	ssize_t ret;
5196 
5197 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
5198 		ret = -EIO;
5199 		goto out;
5200 	}
5201 
5202 	if (!f2fs_is_compress_backend_ready(inode)) {
5203 		ret = -EOPNOTSUPP;
5204 		goto out;
5205 	}
5206 
5207 	if (iocb->ki_flags & IOCB_NOWAIT) {
5208 		if (!inode_trylock(inode)) {
5209 			ret = -EAGAIN;
5210 			goto out;
5211 		}
5212 	} else {
5213 		inode_lock(inode);
5214 	}
5215 
5216 	if (f2fs_is_pinned_file(inode) &&
5217 	    !f2fs_overwrite_io(inode, pos, count)) {
5218 		ret = -EIO;
5219 		goto out_unlock;
5220 	}
5221 
5222 	ret = f2fs_write_checks(iocb, from);
5223 	if (ret <= 0)
5224 		goto out_unlock;
5225 
5226 	/* Determine whether we will do a direct write or a buffered write. */
5227 	dio = f2fs_should_use_dio(inode, iocb, from);
5228 
5229 	/* dio is not compatible w/ atomic write */
5230 	if (dio && f2fs_is_atomic_file(inode)) {
5231 		ret = -EOPNOTSUPP;
5232 		goto out_unlock;
5233 	}
5234 
5235 	/* Possibly preallocate the blocks for the write. */
5236 	target_size = iocb->ki_pos + iov_iter_count(from);
5237 	preallocated = f2fs_preallocate_blocks(iocb, from, dio);
5238 	if (preallocated < 0) {
5239 		ret = preallocated;
5240 	} else {
5241 		if (trace_f2fs_datawrite_start_enabled())
5242 			f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
5243 						orig_count, WRITE);
5244 
5245 		/* Do the actual write. */
5246 		ret = dio ?
5247 			f2fs_dio_write_iter(iocb, from, &may_need_sync) :
5248 			f2fs_buffered_write_iter(iocb, from);
5249 
5250 		trace_f2fs_datawrite_end(inode, orig_pos, ret);
5251 	}
5252 
5253 	/* Don't leave any preallocated blocks around past i_size. */
5254 	if (preallocated && i_size_read(inode) < target_size) {
5255 		f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5256 		filemap_invalidate_lock(inode->i_mapping);
5257 		if (!f2fs_truncate(inode))
5258 			file_dont_truncate(inode);
5259 		filemap_invalidate_unlock(inode->i_mapping);
5260 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5261 	} else {
5262 		file_dont_truncate(inode);
5263 	}
5264 
5265 	clear_inode_flag(inode, FI_PREALLOCATED_ALL);
5266 out_unlock:
5267 	inode_unlock(inode);
5268 out:
5269 	trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
5270 
5271 	if (ret > 0 && may_need_sync)
5272 		ret = generic_write_sync(iocb, ret);
5273 
5274 	/* If buffered IO was forced, flush and drop the data from
5275 	 * the page cache to preserve O_DIRECT semantics
5276 	 */
5277 	if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
5278 		f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
5279 					  orig_pos,
5280 					  orig_pos + ret - 1);
5281 
5282 	return ret;
5283 }
5284 
5285 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
5286 		int advice)
5287 {
5288 	struct address_space *mapping;
5289 	struct backing_dev_info *bdi;
5290 	struct inode *inode = file_inode(filp);
5291 	int err;
5292 
5293 	trace_f2fs_fadvise(inode, offset, len, advice);
5294 
5295 	if (advice == POSIX_FADV_SEQUENTIAL) {
5296 		if (S_ISFIFO(inode->i_mode))
5297 			return -ESPIPE;
5298 
5299 		mapping = filp->f_mapping;
5300 		if (!mapping || len < 0)
5301 			return -EINVAL;
5302 
5303 		bdi = inode_to_bdi(mapping->host);
5304 		filp->f_ra.ra_pages = bdi->ra_pages *
5305 			F2FS_I_SB(inode)->seq_file_ra_mul;
5306 		spin_lock(&filp->f_lock);
5307 		filp->f_mode &= ~FMODE_RANDOM;
5308 		spin_unlock(&filp->f_lock);
5309 		return 0;
5310 	} else if (advice == POSIX_FADV_WILLNEED && offset == 0) {
5311 		/* Load extent cache at the first readahead. */
5312 		f2fs_precache_extents(inode);
5313 	}
5314 
5315 	err = generic_fadvise(filp, offset, len, advice);
5316 	if (err)
5317 		return err;
5318 
5319 	if (advice == POSIX_FADV_DONTNEED &&
5320 	    (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
5321 	     f2fs_compressed_file(inode)))
5322 		f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
5323 	else if (advice == POSIX_FADV_NOREUSE)
5324 		err = f2fs_keep_noreuse_range(inode, offset, len);
5325 	return err;
5326 }
5327 
5328 #ifdef CONFIG_COMPAT
5329 struct compat_f2fs_gc_range {
5330 	u32 sync;
5331 	compat_u64 start;
5332 	compat_u64 len;
5333 };
5334 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE	_IOW(F2FS_IOCTL_MAGIC, 11,\
5335 						struct compat_f2fs_gc_range)
5336 
5337 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
5338 {
5339 	struct compat_f2fs_gc_range __user *urange;
5340 	struct f2fs_gc_range range;
5341 	int err;
5342 
5343 	urange = compat_ptr(arg);
5344 	err = get_user(range.sync, &urange->sync);
5345 	err |= get_user(range.start, &urange->start);
5346 	err |= get_user(range.len, &urange->len);
5347 	if (err)
5348 		return -EFAULT;
5349 
5350 	return __f2fs_ioc_gc_range(file, &range);
5351 }
5352 
5353 struct compat_f2fs_move_range {
5354 	u32 dst_fd;
5355 	compat_u64 pos_in;
5356 	compat_u64 pos_out;
5357 	compat_u64 len;
5358 };
5359 #define F2FS_IOC32_MOVE_RANGE		_IOWR(F2FS_IOCTL_MAGIC, 9,	\
5360 					struct compat_f2fs_move_range)
5361 
5362 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
5363 {
5364 	struct compat_f2fs_move_range __user *urange;
5365 	struct f2fs_move_range range;
5366 	int err;
5367 
5368 	urange = compat_ptr(arg);
5369 	err = get_user(range.dst_fd, &urange->dst_fd);
5370 	err |= get_user(range.pos_in, &urange->pos_in);
5371 	err |= get_user(range.pos_out, &urange->pos_out);
5372 	err |= get_user(range.len, &urange->len);
5373 	if (err)
5374 		return -EFAULT;
5375 
5376 	return __f2fs_ioc_move_range(file, &range);
5377 }
5378 
5379 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5380 {
5381 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
5382 		return -EIO;
5383 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
5384 		return -ENOSPC;
5385 
5386 	switch (cmd) {
5387 	case FS_IOC32_GETVERSION:
5388 		cmd = FS_IOC_GETVERSION;
5389 		break;
5390 	case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
5391 		return f2fs_compat_ioc_gc_range(file, arg);
5392 	case F2FS_IOC32_MOVE_RANGE:
5393 		return f2fs_compat_ioc_move_range(file, arg);
5394 	case F2FS_IOC_START_ATOMIC_WRITE:
5395 	case F2FS_IOC_START_ATOMIC_REPLACE:
5396 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
5397 	case F2FS_IOC_START_VOLATILE_WRITE:
5398 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
5399 	case F2FS_IOC_ABORT_ATOMIC_WRITE:
5400 	case F2FS_IOC_SHUTDOWN:
5401 	case FITRIM:
5402 	case FS_IOC_SET_ENCRYPTION_POLICY:
5403 	case FS_IOC_GET_ENCRYPTION_PWSALT:
5404 	case FS_IOC_GET_ENCRYPTION_POLICY:
5405 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
5406 	case FS_IOC_ADD_ENCRYPTION_KEY:
5407 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
5408 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
5409 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
5410 	case FS_IOC_GET_ENCRYPTION_NONCE:
5411 	case F2FS_IOC_GARBAGE_COLLECT:
5412 	case F2FS_IOC_WRITE_CHECKPOINT:
5413 	case F2FS_IOC_DEFRAGMENT:
5414 	case F2FS_IOC_FLUSH_DEVICE:
5415 	case F2FS_IOC_GET_FEATURES:
5416 	case F2FS_IOC_GET_PIN_FILE:
5417 	case F2FS_IOC_SET_PIN_FILE:
5418 	case F2FS_IOC_PRECACHE_EXTENTS:
5419 	case F2FS_IOC_RESIZE_FS:
5420 	case FS_IOC_ENABLE_VERITY:
5421 	case FS_IOC_MEASURE_VERITY:
5422 	case FS_IOC_READ_VERITY_METADATA:
5423 	case FS_IOC_GETFSLABEL:
5424 	case FS_IOC_SETFSLABEL:
5425 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
5426 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
5427 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
5428 	case F2FS_IOC_SEC_TRIM_FILE:
5429 	case F2FS_IOC_GET_COMPRESS_OPTION:
5430 	case F2FS_IOC_SET_COMPRESS_OPTION:
5431 	case F2FS_IOC_DECOMPRESS_FILE:
5432 	case F2FS_IOC_COMPRESS_FILE:
5433 	case F2FS_IOC_GET_DEV_ALIAS_FILE:
5434 	case F2FS_IOC_IO_PRIO:
5435 		break;
5436 	default:
5437 		return -ENOIOCTLCMD;
5438 	}
5439 	return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5440 }
5441 #endif
5442 
5443 const struct file_operations f2fs_file_operations = {
5444 	.llseek		= f2fs_llseek,
5445 	.read_iter	= f2fs_file_read_iter,
5446 	.write_iter	= f2fs_file_write_iter,
5447 	.iopoll		= iocb_bio_iopoll,
5448 	.open		= f2fs_file_open,
5449 	.release	= f2fs_release_file,
5450 	.mmap_prepare	= f2fs_file_mmap_prepare,
5451 	.flush		= f2fs_file_flush,
5452 	.fsync		= f2fs_sync_file,
5453 	.fallocate	= f2fs_fallocate,
5454 	.unlocked_ioctl	= f2fs_ioctl,
5455 #ifdef CONFIG_COMPAT
5456 	.compat_ioctl	= f2fs_compat_ioctl,
5457 #endif
5458 	.splice_read	= f2fs_file_splice_read,
5459 	.splice_write	= iter_file_splice_write,
5460 	.fadvise	= f2fs_file_fadvise,
5461 	.fop_flags	= FOP_BUFFER_RASYNC,
5462 	.setlease	= generic_setlease,
5463 };
5464