xref: /linux/fs/f2fs/file.c (revision 86f5536004a61a0c797c14a248fc976f03f55cd5)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/file.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/writeback.h>
12 #include <linux/blkdev.h>
13 #include <linux/falloc.h>
14 #include <linux/types.h>
15 #include <linux/compat.h>
16 #include <linux/uaccess.h>
17 #include <linux/mount.h>
18 #include <linux/pagevec.h>
19 #include <linux/uio.h>
20 #include <linux/uuid.h>
21 #include <linux/file.h>
22 #include <linux/nls.h>
23 #include <linux/sched/signal.h>
24 #include <linux/fileattr.h>
25 #include <linux/fadvise.h>
26 #include <linux/iomap.h>
27 
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "acl.h"
33 #include "gc.h"
34 #include "iostat.h"
35 #include <trace/events/f2fs.h>
36 #include <uapi/linux/f2fs.h>
37 
38 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
39 {
40 	struct inode *inode = file_inode(vmf->vma->vm_file);
41 	vm_flags_t flags = vmf->vma->vm_flags;
42 	vm_fault_t ret;
43 
44 	ret = filemap_fault(vmf);
45 	if (ret & VM_FAULT_LOCKED)
46 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
47 					APP_MAPPED_READ_IO, F2FS_BLKSIZE);
48 
49 	trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret);
50 
51 	return ret;
52 }
53 
54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
55 {
56 	struct folio *folio = page_folio(vmf->page);
57 	struct inode *inode = file_inode(vmf->vma->vm_file);
58 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
59 	struct dnode_of_data dn;
60 	bool need_alloc = !f2fs_is_pinned_file(inode);
61 	int err = 0;
62 	vm_fault_t ret;
63 
64 	if (unlikely(IS_IMMUTABLE(inode)))
65 		return VM_FAULT_SIGBUS;
66 
67 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
68 		err = -EIO;
69 		goto out;
70 	}
71 
72 	if (unlikely(f2fs_cp_error(sbi))) {
73 		err = -EIO;
74 		goto out;
75 	}
76 
77 	if (!f2fs_is_checkpoint_ready(sbi)) {
78 		err = -ENOSPC;
79 		goto out;
80 	}
81 
82 	err = f2fs_convert_inline_inode(inode);
83 	if (err)
84 		goto out;
85 
86 #ifdef CONFIG_F2FS_FS_COMPRESSION
87 	if (f2fs_compressed_file(inode)) {
88 		int ret = f2fs_is_compressed_cluster(inode, folio->index);
89 
90 		if (ret < 0) {
91 			err = ret;
92 			goto out;
93 		} else if (ret) {
94 			need_alloc = false;
95 		}
96 	}
97 #endif
98 	/* should do out of any locked page */
99 	if (need_alloc)
100 		f2fs_balance_fs(sbi, true);
101 
102 	sb_start_pagefault(inode->i_sb);
103 
104 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
105 
106 	file_update_time(vmf->vma->vm_file);
107 	filemap_invalidate_lock_shared(inode->i_mapping);
108 	folio_lock(folio);
109 	if (unlikely(folio->mapping != inode->i_mapping ||
110 			folio_pos(folio) > i_size_read(inode) ||
111 			!folio_test_uptodate(folio))) {
112 		folio_unlock(folio);
113 		err = -EFAULT;
114 		goto out_sem;
115 	}
116 
117 	set_new_dnode(&dn, inode, NULL, NULL, 0);
118 	if (need_alloc) {
119 		/* block allocation */
120 		err = f2fs_get_block_locked(&dn, folio->index);
121 	} else {
122 		err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE);
123 		f2fs_put_dnode(&dn);
124 		if (f2fs_is_pinned_file(inode) &&
125 		    !__is_valid_data_blkaddr(dn.data_blkaddr))
126 			err = -EIO;
127 	}
128 
129 	if (err) {
130 		folio_unlock(folio);
131 		goto out_sem;
132 	}
133 
134 	f2fs_wait_on_page_writeback(folio_page(folio, 0), DATA, false, true);
135 
136 	/* wait for GCed page writeback via META_MAPPING */
137 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
138 
139 	/*
140 	 * check to see if the page is mapped already (no holes)
141 	 */
142 	if (folio_test_mappedtodisk(folio))
143 		goto out_sem;
144 
145 	/* page is wholly or partially inside EOF */
146 	if (((loff_t)(folio->index + 1) << PAGE_SHIFT) >
147 						i_size_read(inode)) {
148 		loff_t offset;
149 
150 		offset = i_size_read(inode) & ~PAGE_MASK;
151 		folio_zero_segment(folio, offset, folio_size(folio));
152 	}
153 	folio_mark_dirty(folio);
154 
155 	f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
156 	f2fs_update_time(sbi, REQ_TIME);
157 
158 out_sem:
159 	filemap_invalidate_unlock_shared(inode->i_mapping);
160 
161 	sb_end_pagefault(inode->i_sb);
162 out:
163 	ret = vmf_fs_error(err);
164 
165 	trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret);
166 	return ret;
167 }
168 
169 static const struct vm_operations_struct f2fs_file_vm_ops = {
170 	.fault		= f2fs_filemap_fault,
171 	.map_pages	= filemap_map_pages,
172 	.page_mkwrite	= f2fs_vm_page_mkwrite,
173 };
174 
175 static int get_parent_ino(struct inode *inode, nid_t *pino)
176 {
177 	struct dentry *dentry;
178 
179 	/*
180 	 * Make sure to get the non-deleted alias.  The alias associated with
181 	 * the open file descriptor being fsync()'ed may be deleted already.
182 	 */
183 	dentry = d_find_alias(inode);
184 	if (!dentry)
185 		return 0;
186 
187 	*pino = d_parent_ino(dentry);
188 	dput(dentry);
189 	return 1;
190 }
191 
192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
193 {
194 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
195 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
196 
197 	if (!S_ISREG(inode->i_mode))
198 		cp_reason = CP_NON_REGULAR;
199 	else if (f2fs_compressed_file(inode))
200 		cp_reason = CP_COMPRESSED;
201 	else if (inode->i_nlink != 1)
202 		cp_reason = CP_HARDLINK;
203 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
204 		cp_reason = CP_SB_NEED_CP;
205 	else if (file_wrong_pino(inode))
206 		cp_reason = CP_WRONG_PINO;
207 	else if (!f2fs_space_for_roll_forward(sbi))
208 		cp_reason = CP_NO_SPC_ROLL;
209 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
210 		cp_reason = CP_NODE_NEED_CP;
211 	else if (test_opt(sbi, FASTBOOT))
212 		cp_reason = CP_FASTBOOT_MODE;
213 	else if (F2FS_OPTION(sbi).active_logs == 2)
214 		cp_reason = CP_SPEC_LOG_NUM;
215 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
216 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
217 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
218 							TRANS_DIR_INO))
219 		cp_reason = CP_RECOVER_DIR;
220 	else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
221 							XATTR_DIR_INO))
222 		cp_reason = CP_XATTR_DIR;
223 
224 	return cp_reason;
225 }
226 
227 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
228 {
229 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
230 	bool ret = false;
231 	/* But we need to avoid that there are some inode updates */
232 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
233 		ret = true;
234 	f2fs_put_page(i, 0);
235 	return ret;
236 }
237 
238 static void try_to_fix_pino(struct inode *inode)
239 {
240 	struct f2fs_inode_info *fi = F2FS_I(inode);
241 	nid_t pino;
242 
243 	f2fs_down_write(&fi->i_sem);
244 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
245 			get_parent_ino(inode, &pino)) {
246 		f2fs_i_pino_write(inode, pino);
247 		file_got_pino(inode);
248 	}
249 	f2fs_up_write(&fi->i_sem);
250 }
251 
252 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
253 						int datasync, bool atomic)
254 {
255 	struct inode *inode = file->f_mapping->host;
256 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
257 	nid_t ino = inode->i_ino;
258 	int ret = 0;
259 	enum cp_reason_type cp_reason = 0;
260 	struct writeback_control wbc = {
261 		.sync_mode = WB_SYNC_ALL,
262 		.nr_to_write = LONG_MAX,
263 		.for_reclaim = 0,
264 	};
265 	unsigned int seq_id = 0;
266 
267 	if (unlikely(f2fs_readonly(inode->i_sb)))
268 		return 0;
269 
270 	trace_f2fs_sync_file_enter(inode);
271 
272 	if (S_ISDIR(inode->i_mode))
273 		goto go_write;
274 
275 	/* if fdatasync is triggered, let's do in-place-update */
276 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
277 		set_inode_flag(inode, FI_NEED_IPU);
278 	ret = file_write_and_wait_range(file, start, end);
279 	clear_inode_flag(inode, FI_NEED_IPU);
280 
281 	if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
282 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
283 		return ret;
284 	}
285 
286 	/* if the inode is dirty, let's recover all the time */
287 	if (!f2fs_skip_inode_update(inode, datasync)) {
288 		f2fs_write_inode(inode, NULL);
289 		goto go_write;
290 	}
291 
292 	/*
293 	 * if there is no written data, don't waste time to write recovery info.
294 	 */
295 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
296 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
297 
298 		/* it may call write_inode just prior to fsync */
299 		if (need_inode_page_update(sbi, ino))
300 			goto go_write;
301 
302 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
303 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
304 			goto flush_out;
305 		goto out;
306 	} else {
307 		/*
308 		 * for OPU case, during fsync(), node can be persisted before
309 		 * data when lower device doesn't support write barrier, result
310 		 * in data corruption after SPO.
311 		 * So for strict fsync mode, force to use atomic write semantics
312 		 * to keep write order in between data/node and last node to
313 		 * avoid potential data corruption.
314 		 */
315 		if (F2FS_OPTION(sbi).fsync_mode ==
316 				FSYNC_MODE_STRICT && !atomic)
317 			atomic = true;
318 	}
319 go_write:
320 	/*
321 	 * Both of fdatasync() and fsync() are able to be recovered from
322 	 * sudden-power-off.
323 	 */
324 	f2fs_down_read(&F2FS_I(inode)->i_sem);
325 	cp_reason = need_do_checkpoint(inode);
326 	f2fs_up_read(&F2FS_I(inode)->i_sem);
327 
328 	if (cp_reason) {
329 		/* all the dirty node pages should be flushed for POR */
330 		ret = f2fs_sync_fs(inode->i_sb, 1);
331 
332 		/*
333 		 * We've secured consistency through sync_fs. Following pino
334 		 * will be used only for fsynced inodes after checkpoint.
335 		 */
336 		try_to_fix_pino(inode);
337 		clear_inode_flag(inode, FI_APPEND_WRITE);
338 		clear_inode_flag(inode, FI_UPDATE_WRITE);
339 		goto out;
340 	}
341 sync_nodes:
342 	atomic_inc(&sbi->wb_sync_req[NODE]);
343 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
344 	atomic_dec(&sbi->wb_sync_req[NODE]);
345 	if (ret)
346 		goto out;
347 
348 	/* if cp_error was enabled, we should avoid infinite loop */
349 	if (unlikely(f2fs_cp_error(sbi))) {
350 		ret = -EIO;
351 		goto out;
352 	}
353 
354 	if (f2fs_need_inode_block_update(sbi, ino)) {
355 		f2fs_mark_inode_dirty_sync(inode, true);
356 		f2fs_write_inode(inode, NULL);
357 		goto sync_nodes;
358 	}
359 
360 	/*
361 	 * If it's atomic_write, it's just fine to keep write ordering. So
362 	 * here we don't need to wait for node write completion, since we use
363 	 * node chain which serializes node blocks. If one of node writes are
364 	 * reordered, we can see simply broken chain, resulting in stopping
365 	 * roll-forward recovery. It means we'll recover all or none node blocks
366 	 * given fsync mark.
367 	 */
368 	if (!atomic) {
369 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
370 		if (ret)
371 			goto out;
372 	}
373 
374 	/* once recovery info is written, don't need to tack this */
375 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
376 	clear_inode_flag(inode, FI_APPEND_WRITE);
377 flush_out:
378 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
379 		ret = f2fs_issue_flush(sbi, inode->i_ino);
380 	if (!ret) {
381 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
382 		clear_inode_flag(inode, FI_UPDATE_WRITE);
383 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
384 	}
385 	f2fs_update_time(sbi, REQ_TIME);
386 out:
387 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
388 	return ret;
389 }
390 
391 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
392 {
393 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
394 		return -EIO;
395 	return f2fs_do_sync_file(file, start, end, datasync, false);
396 }
397 
398 static bool __found_offset(struct address_space *mapping,
399 		struct dnode_of_data *dn, pgoff_t index, int whence)
400 {
401 	block_t blkaddr = f2fs_data_blkaddr(dn);
402 	struct inode *inode = mapping->host;
403 	bool compressed_cluster = false;
404 
405 	if (f2fs_compressed_file(inode)) {
406 		block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_page,
407 		    ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size));
408 
409 		compressed_cluster = first_blkaddr == COMPRESS_ADDR;
410 	}
411 
412 	switch (whence) {
413 	case SEEK_DATA:
414 		if (__is_valid_data_blkaddr(blkaddr))
415 			return true;
416 		if (blkaddr == NEW_ADDR &&
417 		    xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
418 			return true;
419 		if (compressed_cluster)
420 			return true;
421 		break;
422 	case SEEK_HOLE:
423 		if (compressed_cluster)
424 			return false;
425 		if (blkaddr == NULL_ADDR)
426 			return true;
427 		break;
428 	}
429 	return false;
430 }
431 
432 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
433 {
434 	struct inode *inode = file->f_mapping->host;
435 	loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
436 	struct dnode_of_data dn;
437 	pgoff_t pgofs, end_offset;
438 	loff_t data_ofs = offset;
439 	loff_t isize;
440 	int err = 0;
441 
442 	inode_lock_shared(inode);
443 
444 	isize = i_size_read(inode);
445 	if (offset >= isize)
446 		goto fail;
447 
448 	/* handle inline data case */
449 	if (f2fs_has_inline_data(inode)) {
450 		if (whence == SEEK_HOLE) {
451 			data_ofs = isize;
452 			goto found;
453 		} else if (whence == SEEK_DATA) {
454 			data_ofs = offset;
455 			goto found;
456 		}
457 	}
458 
459 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
460 
461 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
462 		set_new_dnode(&dn, inode, NULL, NULL, 0);
463 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
464 		if (err && err != -ENOENT) {
465 			goto fail;
466 		} else if (err == -ENOENT) {
467 			/* direct node does not exists */
468 			if (whence == SEEK_DATA) {
469 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
470 				continue;
471 			} else {
472 				goto found;
473 			}
474 		}
475 
476 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
477 
478 		/* find data/hole in dnode block */
479 		for (; dn.ofs_in_node < end_offset;
480 				dn.ofs_in_node++, pgofs++,
481 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
482 			block_t blkaddr;
483 
484 			blkaddr = f2fs_data_blkaddr(&dn);
485 
486 			if (__is_valid_data_blkaddr(blkaddr) &&
487 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
488 					blkaddr, DATA_GENERIC_ENHANCE)) {
489 				f2fs_put_dnode(&dn);
490 				goto fail;
491 			}
492 
493 			if (__found_offset(file->f_mapping, &dn,
494 							pgofs, whence)) {
495 				f2fs_put_dnode(&dn);
496 				goto found;
497 			}
498 		}
499 		f2fs_put_dnode(&dn);
500 	}
501 
502 	if (whence == SEEK_DATA)
503 		goto fail;
504 found:
505 	if (whence == SEEK_HOLE && data_ofs > isize)
506 		data_ofs = isize;
507 	inode_unlock_shared(inode);
508 	return vfs_setpos(file, data_ofs, maxbytes);
509 fail:
510 	inode_unlock_shared(inode);
511 	return -ENXIO;
512 }
513 
514 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
515 {
516 	struct inode *inode = file->f_mapping->host;
517 	loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
518 
519 	switch (whence) {
520 	case SEEK_SET:
521 	case SEEK_CUR:
522 	case SEEK_END:
523 		return generic_file_llseek_size(file, offset, whence,
524 						maxbytes, i_size_read(inode));
525 	case SEEK_DATA:
526 	case SEEK_HOLE:
527 		if (offset < 0)
528 			return -ENXIO;
529 		return f2fs_seek_block(file, offset, whence);
530 	}
531 
532 	return -EINVAL;
533 }
534 
535 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
536 {
537 	struct inode *inode = file_inode(file);
538 
539 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
540 		return -EIO;
541 
542 	if (!f2fs_is_compress_backend_ready(inode))
543 		return -EOPNOTSUPP;
544 
545 	file_accessed(file);
546 	vma->vm_ops = &f2fs_file_vm_ops;
547 
548 	f2fs_down_read(&F2FS_I(inode)->i_sem);
549 	set_inode_flag(inode, FI_MMAP_FILE);
550 	f2fs_up_read(&F2FS_I(inode)->i_sem);
551 
552 	return 0;
553 }
554 
555 static int finish_preallocate_blocks(struct inode *inode)
556 {
557 	int ret;
558 
559 	inode_lock(inode);
560 	if (is_inode_flag_set(inode, FI_OPENED_FILE)) {
561 		inode_unlock(inode);
562 		return 0;
563 	}
564 
565 	if (!file_should_truncate(inode)) {
566 		set_inode_flag(inode, FI_OPENED_FILE);
567 		inode_unlock(inode);
568 		return 0;
569 	}
570 
571 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
572 	filemap_invalidate_lock(inode->i_mapping);
573 
574 	truncate_setsize(inode, i_size_read(inode));
575 	ret = f2fs_truncate(inode);
576 
577 	filemap_invalidate_unlock(inode->i_mapping);
578 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
579 
580 	if (!ret)
581 		set_inode_flag(inode, FI_OPENED_FILE);
582 
583 	inode_unlock(inode);
584 	if (ret)
585 		return ret;
586 
587 	file_dont_truncate(inode);
588 	return 0;
589 }
590 
591 static int f2fs_file_open(struct inode *inode, struct file *filp)
592 {
593 	int err = fscrypt_file_open(inode, filp);
594 
595 	if (err)
596 		return err;
597 
598 	if (!f2fs_is_compress_backend_ready(inode))
599 		return -EOPNOTSUPP;
600 
601 	err = fsverity_file_open(inode, filp);
602 	if (err)
603 		return err;
604 
605 	filp->f_mode |= FMODE_NOWAIT;
606 	filp->f_mode |= FMODE_CAN_ODIRECT;
607 
608 	err = dquot_file_open(inode, filp);
609 	if (err)
610 		return err;
611 
612 	return finish_preallocate_blocks(inode);
613 }
614 
615 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
616 {
617 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
618 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
619 	__le32 *addr;
620 	bool compressed_cluster = false;
621 	int cluster_index = 0, valid_blocks = 0;
622 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
623 	bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
624 	block_t blkstart;
625 	int blklen = 0;
626 
627 	addr = get_dnode_addr(dn->inode, dn->node_page) + ofs;
628 	blkstart = le32_to_cpu(*addr);
629 
630 	/* Assumption: truncation starts with cluster */
631 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
632 		block_t blkaddr = le32_to_cpu(*addr);
633 
634 		if (f2fs_compressed_file(dn->inode) &&
635 					!(cluster_index & (cluster_size - 1))) {
636 			if (compressed_cluster)
637 				f2fs_i_compr_blocks_update(dn->inode,
638 							valid_blocks, false);
639 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
640 			valid_blocks = 0;
641 		}
642 
643 		if (blkaddr == NULL_ADDR)
644 			goto next;
645 
646 		f2fs_set_data_blkaddr(dn, NULL_ADDR);
647 
648 		if (__is_valid_data_blkaddr(blkaddr)) {
649 			if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE))
650 				goto next;
651 			if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr,
652 						DATA_GENERIC_ENHANCE))
653 				goto next;
654 			if (compressed_cluster)
655 				valid_blocks++;
656 		}
657 
658 		if (blkstart + blklen == blkaddr) {
659 			blklen++;
660 		} else {
661 			f2fs_invalidate_blocks(sbi, blkstart, blklen);
662 			blkstart = blkaddr;
663 			blklen = 1;
664 		}
665 
666 		if (!released || blkaddr != COMPRESS_ADDR)
667 			nr_free++;
668 
669 		continue;
670 
671 next:
672 		if (blklen)
673 			f2fs_invalidate_blocks(sbi, blkstart, blklen);
674 
675 		blkstart = le32_to_cpu(*(addr + 1));
676 		blklen = 0;
677 	}
678 
679 	if (blklen)
680 		f2fs_invalidate_blocks(sbi, blkstart, blklen);
681 
682 	if (compressed_cluster)
683 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
684 
685 	if (nr_free) {
686 		pgoff_t fofs;
687 		/*
688 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
689 		 * we will invalidate all blkaddr in the whole range.
690 		 */
691 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
692 							dn->inode) + ofs;
693 		f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
694 		f2fs_update_age_extent_cache_range(dn, fofs, len);
695 		dec_valid_block_count(sbi, dn->inode, nr_free);
696 	}
697 	dn->ofs_in_node = ofs;
698 
699 	f2fs_update_time(sbi, REQ_TIME);
700 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
701 					 dn->ofs_in_node, nr_free);
702 }
703 
704 static int truncate_partial_data_page(struct inode *inode, u64 from,
705 								bool cache_only)
706 {
707 	loff_t offset = from & (PAGE_SIZE - 1);
708 	pgoff_t index = from >> PAGE_SHIFT;
709 	struct address_space *mapping = inode->i_mapping;
710 	struct page *page;
711 
712 	if (!offset && !cache_only)
713 		return 0;
714 
715 	if (cache_only) {
716 		page = find_lock_page(mapping, index);
717 		if (page && PageUptodate(page))
718 			goto truncate_out;
719 		f2fs_put_page(page, 1);
720 		return 0;
721 	}
722 
723 	page = f2fs_get_lock_data_page(inode, index, true);
724 	if (IS_ERR(page))
725 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
726 truncate_out:
727 	f2fs_wait_on_page_writeback(page, DATA, true, true);
728 	zero_user(page, offset, PAGE_SIZE - offset);
729 
730 	/* An encrypted inode should have a key and truncate the last page. */
731 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
732 	if (!cache_only)
733 		set_page_dirty(page);
734 	f2fs_put_page(page, 1);
735 	return 0;
736 }
737 
738 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
739 {
740 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
741 	struct dnode_of_data dn;
742 	pgoff_t free_from;
743 	int count = 0, err = 0;
744 	struct page *ipage;
745 	bool truncate_page = false;
746 
747 	trace_f2fs_truncate_blocks_enter(inode, from);
748 
749 	if (IS_DEVICE_ALIASING(inode) && from) {
750 		err = -EINVAL;
751 		goto out_err;
752 	}
753 
754 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
755 
756 	if (free_from >= max_file_blocks(inode))
757 		goto free_partial;
758 
759 	if (lock)
760 		f2fs_lock_op(sbi);
761 
762 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
763 	if (IS_ERR(ipage)) {
764 		err = PTR_ERR(ipage);
765 		goto out;
766 	}
767 
768 	if (IS_DEVICE_ALIASING(inode)) {
769 		struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
770 		struct extent_info ei = et->largest;
771 
772 		f2fs_invalidate_blocks(sbi, ei.blk, ei.len);
773 
774 		dec_valid_block_count(sbi, inode, ei.len);
775 		f2fs_update_time(sbi, REQ_TIME);
776 
777 		f2fs_put_page(ipage, 1);
778 		goto out;
779 	}
780 
781 	if (f2fs_has_inline_data(inode)) {
782 		f2fs_truncate_inline_inode(inode, ipage, from);
783 		f2fs_put_page(ipage, 1);
784 		truncate_page = true;
785 		goto out;
786 	}
787 
788 	set_new_dnode(&dn, inode, ipage, NULL, 0);
789 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
790 	if (err) {
791 		if (err == -ENOENT)
792 			goto free_next;
793 		goto out;
794 	}
795 
796 	count = ADDRS_PER_PAGE(dn.node_page, inode);
797 
798 	count -= dn.ofs_in_node;
799 	f2fs_bug_on(sbi, count < 0);
800 
801 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
802 		f2fs_truncate_data_blocks_range(&dn, count);
803 		free_from += count;
804 	}
805 
806 	f2fs_put_dnode(&dn);
807 free_next:
808 	err = f2fs_truncate_inode_blocks(inode, free_from);
809 out:
810 	if (lock)
811 		f2fs_unlock_op(sbi);
812 free_partial:
813 	/* lastly zero out the first data page */
814 	if (!err)
815 		err = truncate_partial_data_page(inode, from, truncate_page);
816 out_err:
817 	trace_f2fs_truncate_blocks_exit(inode, err);
818 	return err;
819 }
820 
821 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
822 {
823 	u64 free_from = from;
824 	int err;
825 
826 #ifdef CONFIG_F2FS_FS_COMPRESSION
827 	/*
828 	 * for compressed file, only support cluster size
829 	 * aligned truncation.
830 	 */
831 	if (f2fs_compressed_file(inode))
832 		free_from = round_up(from,
833 				F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
834 #endif
835 
836 	err = f2fs_do_truncate_blocks(inode, free_from, lock);
837 	if (err)
838 		return err;
839 
840 #ifdef CONFIG_F2FS_FS_COMPRESSION
841 	/*
842 	 * For compressed file, after release compress blocks, don't allow write
843 	 * direct, but we should allow write direct after truncate to zero.
844 	 */
845 	if (f2fs_compressed_file(inode) && !free_from
846 			&& is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
847 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
848 
849 	if (from != free_from) {
850 		err = f2fs_truncate_partial_cluster(inode, from, lock);
851 		if (err)
852 			return err;
853 	}
854 #endif
855 
856 	return 0;
857 }
858 
859 int f2fs_truncate(struct inode *inode)
860 {
861 	int err;
862 
863 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
864 		return -EIO;
865 
866 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
867 				S_ISLNK(inode->i_mode)))
868 		return 0;
869 
870 	trace_f2fs_truncate(inode);
871 
872 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
873 		return -EIO;
874 
875 	err = f2fs_dquot_initialize(inode);
876 	if (err)
877 		return err;
878 
879 	/* we should check inline_data size */
880 	if (!f2fs_may_inline_data(inode)) {
881 		err = f2fs_convert_inline_inode(inode);
882 		if (err)
883 			return err;
884 	}
885 
886 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
887 	if (err)
888 		return err;
889 
890 	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
891 	f2fs_mark_inode_dirty_sync(inode, false);
892 	return 0;
893 }
894 
895 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
896 {
897 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
898 
899 	if (!fscrypt_dio_supported(inode))
900 		return true;
901 	if (fsverity_active(inode))
902 		return true;
903 	if (f2fs_compressed_file(inode))
904 		return true;
905 	/*
906 	 * only force direct read to use buffered IO, for direct write,
907 	 * it expects inline data conversion before committing IO.
908 	 */
909 	if (f2fs_has_inline_data(inode) && rw == READ)
910 		return true;
911 
912 	/* disallow direct IO if any of devices has unaligned blksize */
913 	if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
914 		return true;
915 	/*
916 	 * for blkzoned device, fallback direct IO to buffered IO, so
917 	 * all IOs can be serialized by log-structured write.
918 	 */
919 	if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) &&
920 	    !f2fs_is_pinned_file(inode))
921 		return true;
922 	if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
923 		return true;
924 
925 	return false;
926 }
927 
928 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path,
929 		 struct kstat *stat, u32 request_mask, unsigned int query_flags)
930 {
931 	struct inode *inode = d_inode(path->dentry);
932 	struct f2fs_inode_info *fi = F2FS_I(inode);
933 	struct f2fs_inode *ri = NULL;
934 	unsigned int flags;
935 
936 	if (f2fs_has_extra_attr(inode) &&
937 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
938 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
939 		stat->result_mask |= STATX_BTIME;
940 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
941 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
942 	}
943 
944 	/*
945 	 * Return the DIO alignment restrictions if requested.  We only return
946 	 * this information when requested, since on encrypted files it might
947 	 * take a fair bit of work to get if the file wasn't opened recently.
948 	 *
949 	 * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
950 	 * cannot represent that, so in that case we report no DIO support.
951 	 */
952 	if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
953 		unsigned int bsize = i_blocksize(inode);
954 
955 		stat->result_mask |= STATX_DIOALIGN;
956 		if (!f2fs_force_buffered_io(inode, WRITE)) {
957 			stat->dio_mem_align = bsize;
958 			stat->dio_offset_align = bsize;
959 		}
960 	}
961 
962 	flags = fi->i_flags;
963 	if (flags & F2FS_COMPR_FL)
964 		stat->attributes |= STATX_ATTR_COMPRESSED;
965 	if (flags & F2FS_APPEND_FL)
966 		stat->attributes |= STATX_ATTR_APPEND;
967 	if (IS_ENCRYPTED(inode))
968 		stat->attributes |= STATX_ATTR_ENCRYPTED;
969 	if (flags & F2FS_IMMUTABLE_FL)
970 		stat->attributes |= STATX_ATTR_IMMUTABLE;
971 	if (flags & F2FS_NODUMP_FL)
972 		stat->attributes |= STATX_ATTR_NODUMP;
973 	if (IS_VERITY(inode))
974 		stat->attributes |= STATX_ATTR_VERITY;
975 
976 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
977 				  STATX_ATTR_APPEND |
978 				  STATX_ATTR_ENCRYPTED |
979 				  STATX_ATTR_IMMUTABLE |
980 				  STATX_ATTR_NODUMP |
981 				  STATX_ATTR_VERITY);
982 
983 	generic_fillattr(idmap, request_mask, inode, stat);
984 
985 	/* we need to show initial sectors used for inline_data/dentries */
986 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
987 					f2fs_has_inline_dentry(inode))
988 		stat->blocks += (stat->size + 511) >> 9;
989 
990 	return 0;
991 }
992 
993 #ifdef CONFIG_F2FS_FS_POSIX_ACL
994 static void __setattr_copy(struct mnt_idmap *idmap,
995 			   struct inode *inode, const struct iattr *attr)
996 {
997 	unsigned int ia_valid = attr->ia_valid;
998 
999 	i_uid_update(idmap, attr, inode);
1000 	i_gid_update(idmap, attr, inode);
1001 	if (ia_valid & ATTR_ATIME)
1002 		inode_set_atime_to_ts(inode, attr->ia_atime);
1003 	if (ia_valid & ATTR_MTIME)
1004 		inode_set_mtime_to_ts(inode, attr->ia_mtime);
1005 	if (ia_valid & ATTR_CTIME)
1006 		inode_set_ctime_to_ts(inode, attr->ia_ctime);
1007 	if (ia_valid & ATTR_MODE) {
1008 		umode_t mode = attr->ia_mode;
1009 
1010 		if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode)))
1011 			mode &= ~S_ISGID;
1012 		set_acl_inode(inode, mode);
1013 	}
1014 }
1015 #else
1016 #define __setattr_copy setattr_copy
1017 #endif
1018 
1019 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
1020 		 struct iattr *attr)
1021 {
1022 	struct inode *inode = d_inode(dentry);
1023 	struct f2fs_inode_info *fi = F2FS_I(inode);
1024 	int err;
1025 
1026 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1027 		return -EIO;
1028 
1029 	if (unlikely(IS_IMMUTABLE(inode)))
1030 		return -EPERM;
1031 
1032 	if (unlikely(IS_APPEND(inode) &&
1033 			(attr->ia_valid & (ATTR_MODE | ATTR_UID |
1034 				  ATTR_GID | ATTR_TIMES_SET))))
1035 		return -EPERM;
1036 
1037 	if ((attr->ia_valid & ATTR_SIZE)) {
1038 		if (!f2fs_is_compress_backend_ready(inode) ||
1039 				IS_DEVICE_ALIASING(inode))
1040 			return -EOPNOTSUPP;
1041 		if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) &&
1042 			!IS_ALIGNED(attr->ia_size,
1043 			F2FS_BLK_TO_BYTES(fi->i_cluster_size)))
1044 			return -EINVAL;
1045 	}
1046 
1047 	err = setattr_prepare(idmap, dentry, attr);
1048 	if (err)
1049 		return err;
1050 
1051 	err = fscrypt_prepare_setattr(dentry, attr);
1052 	if (err)
1053 		return err;
1054 
1055 	err = fsverity_prepare_setattr(dentry, attr);
1056 	if (err)
1057 		return err;
1058 
1059 	if (is_quota_modification(idmap, inode, attr)) {
1060 		err = f2fs_dquot_initialize(inode);
1061 		if (err)
1062 			return err;
1063 	}
1064 	if (i_uid_needs_update(idmap, attr, inode) ||
1065 	    i_gid_needs_update(idmap, attr, inode)) {
1066 		f2fs_lock_op(F2FS_I_SB(inode));
1067 		err = dquot_transfer(idmap, inode, attr);
1068 		if (err) {
1069 			set_sbi_flag(F2FS_I_SB(inode),
1070 					SBI_QUOTA_NEED_REPAIR);
1071 			f2fs_unlock_op(F2FS_I_SB(inode));
1072 			return err;
1073 		}
1074 		/*
1075 		 * update uid/gid under lock_op(), so that dquot and inode can
1076 		 * be updated atomically.
1077 		 */
1078 		i_uid_update(idmap, attr, inode);
1079 		i_gid_update(idmap, attr, inode);
1080 		f2fs_mark_inode_dirty_sync(inode, true);
1081 		f2fs_unlock_op(F2FS_I_SB(inode));
1082 	}
1083 
1084 	if (attr->ia_valid & ATTR_SIZE) {
1085 		loff_t old_size = i_size_read(inode);
1086 
1087 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
1088 			/*
1089 			 * should convert inline inode before i_size_write to
1090 			 * keep smaller than inline_data size with inline flag.
1091 			 */
1092 			err = f2fs_convert_inline_inode(inode);
1093 			if (err)
1094 				return err;
1095 		}
1096 
1097 		/*
1098 		 * wait for inflight dio, blocks should be removed after
1099 		 * IO completion.
1100 		 */
1101 		if (attr->ia_size < old_size)
1102 			inode_dio_wait(inode);
1103 
1104 		f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
1105 		filemap_invalidate_lock(inode->i_mapping);
1106 
1107 		truncate_setsize(inode, attr->ia_size);
1108 
1109 		if (attr->ia_size <= old_size)
1110 			err = f2fs_truncate(inode);
1111 		/*
1112 		 * do not trim all blocks after i_size if target size is
1113 		 * larger than i_size.
1114 		 */
1115 		filemap_invalidate_unlock(inode->i_mapping);
1116 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
1117 		if (err)
1118 			return err;
1119 
1120 		spin_lock(&fi->i_size_lock);
1121 		inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1122 		fi->last_disk_size = i_size_read(inode);
1123 		spin_unlock(&fi->i_size_lock);
1124 	}
1125 
1126 	__setattr_copy(idmap, inode, attr);
1127 
1128 	if (attr->ia_valid & ATTR_MODE) {
1129 		err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
1130 
1131 		if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1132 			if (!err)
1133 				inode->i_mode = fi->i_acl_mode;
1134 			clear_inode_flag(inode, FI_ACL_MODE);
1135 		}
1136 	}
1137 
1138 	/* file size may changed here */
1139 	f2fs_mark_inode_dirty_sync(inode, true);
1140 
1141 	/* inode change will produce dirty node pages flushed by checkpoint */
1142 	f2fs_balance_fs(F2FS_I_SB(inode), true);
1143 
1144 	return err;
1145 }
1146 
1147 const struct inode_operations f2fs_file_inode_operations = {
1148 	.getattr	= f2fs_getattr,
1149 	.setattr	= f2fs_setattr,
1150 	.get_inode_acl	= f2fs_get_acl,
1151 	.set_acl	= f2fs_set_acl,
1152 	.listxattr	= f2fs_listxattr,
1153 	.fiemap		= f2fs_fiemap,
1154 	.fileattr_get	= f2fs_fileattr_get,
1155 	.fileattr_set	= f2fs_fileattr_set,
1156 };
1157 
1158 static int fill_zero(struct inode *inode, pgoff_t index,
1159 					loff_t start, loff_t len)
1160 {
1161 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1162 	struct page *page;
1163 
1164 	if (!len)
1165 		return 0;
1166 
1167 	f2fs_balance_fs(sbi, true);
1168 
1169 	f2fs_lock_op(sbi);
1170 	page = f2fs_get_new_data_page(inode, NULL, index, false);
1171 	f2fs_unlock_op(sbi);
1172 
1173 	if (IS_ERR(page))
1174 		return PTR_ERR(page);
1175 
1176 	f2fs_wait_on_page_writeback(page, DATA, true, true);
1177 	zero_user(page, start, len);
1178 	set_page_dirty(page);
1179 	f2fs_put_page(page, 1);
1180 	return 0;
1181 }
1182 
1183 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1184 {
1185 	int err;
1186 
1187 	while (pg_start < pg_end) {
1188 		struct dnode_of_data dn;
1189 		pgoff_t end_offset, count;
1190 
1191 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1192 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1193 		if (err) {
1194 			if (err == -ENOENT) {
1195 				pg_start = f2fs_get_next_page_offset(&dn,
1196 								pg_start);
1197 				continue;
1198 			}
1199 			return err;
1200 		}
1201 
1202 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1203 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1204 
1205 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1206 
1207 		f2fs_truncate_data_blocks_range(&dn, count);
1208 		f2fs_put_dnode(&dn);
1209 
1210 		pg_start += count;
1211 	}
1212 	return 0;
1213 }
1214 
1215 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
1216 {
1217 	pgoff_t pg_start, pg_end;
1218 	loff_t off_start, off_end;
1219 	int ret;
1220 
1221 	ret = f2fs_convert_inline_inode(inode);
1222 	if (ret)
1223 		return ret;
1224 
1225 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1226 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1227 
1228 	off_start = offset & (PAGE_SIZE - 1);
1229 	off_end = (offset + len) & (PAGE_SIZE - 1);
1230 
1231 	if (pg_start == pg_end) {
1232 		ret = fill_zero(inode, pg_start, off_start,
1233 						off_end - off_start);
1234 		if (ret)
1235 			return ret;
1236 	} else {
1237 		if (off_start) {
1238 			ret = fill_zero(inode, pg_start++, off_start,
1239 						PAGE_SIZE - off_start);
1240 			if (ret)
1241 				return ret;
1242 		}
1243 		if (off_end) {
1244 			ret = fill_zero(inode, pg_end, 0, off_end);
1245 			if (ret)
1246 				return ret;
1247 		}
1248 
1249 		if (pg_start < pg_end) {
1250 			loff_t blk_start, blk_end;
1251 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1252 
1253 			f2fs_balance_fs(sbi, true);
1254 
1255 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1256 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1257 
1258 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1259 			filemap_invalidate_lock(inode->i_mapping);
1260 
1261 			truncate_pagecache_range(inode, blk_start, blk_end - 1);
1262 
1263 			f2fs_lock_op(sbi);
1264 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1265 			f2fs_unlock_op(sbi);
1266 
1267 			filemap_invalidate_unlock(inode->i_mapping);
1268 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1269 		}
1270 	}
1271 
1272 	return ret;
1273 }
1274 
1275 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1276 				int *do_replace, pgoff_t off, pgoff_t len)
1277 {
1278 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1279 	struct dnode_of_data dn;
1280 	int ret, done, i;
1281 
1282 next_dnode:
1283 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1284 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1285 	if (ret && ret != -ENOENT) {
1286 		return ret;
1287 	} else if (ret == -ENOENT) {
1288 		if (dn.max_level == 0)
1289 			return -ENOENT;
1290 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1291 						dn.ofs_in_node, len);
1292 		blkaddr += done;
1293 		do_replace += done;
1294 		goto next;
1295 	}
1296 
1297 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1298 							dn.ofs_in_node, len);
1299 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1300 		*blkaddr = f2fs_data_blkaddr(&dn);
1301 
1302 		if (__is_valid_data_blkaddr(*blkaddr) &&
1303 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1304 					DATA_GENERIC_ENHANCE)) {
1305 			f2fs_put_dnode(&dn);
1306 			return -EFSCORRUPTED;
1307 		}
1308 
1309 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1310 
1311 			if (f2fs_lfs_mode(sbi)) {
1312 				f2fs_put_dnode(&dn);
1313 				return -EOPNOTSUPP;
1314 			}
1315 
1316 			/* do not invalidate this block address */
1317 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1318 			*do_replace = 1;
1319 		}
1320 	}
1321 	f2fs_put_dnode(&dn);
1322 next:
1323 	len -= done;
1324 	off += done;
1325 	if (len)
1326 		goto next_dnode;
1327 	return 0;
1328 }
1329 
1330 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1331 				int *do_replace, pgoff_t off, int len)
1332 {
1333 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1334 	struct dnode_of_data dn;
1335 	int ret, i;
1336 
1337 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1338 		if (*do_replace == 0)
1339 			continue;
1340 
1341 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1342 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1343 		if (ret) {
1344 			dec_valid_block_count(sbi, inode, 1);
1345 			f2fs_invalidate_blocks(sbi, *blkaddr, 1);
1346 		} else {
1347 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1348 		}
1349 		f2fs_put_dnode(&dn);
1350 	}
1351 	return 0;
1352 }
1353 
1354 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1355 			block_t *blkaddr, int *do_replace,
1356 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1357 {
1358 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1359 	pgoff_t i = 0;
1360 	int ret;
1361 
1362 	while (i < len) {
1363 		if (blkaddr[i] == NULL_ADDR && !full) {
1364 			i++;
1365 			continue;
1366 		}
1367 
1368 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1369 			struct dnode_of_data dn;
1370 			struct node_info ni;
1371 			size_t new_size;
1372 			pgoff_t ilen;
1373 
1374 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1375 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1376 			if (ret)
1377 				return ret;
1378 
1379 			ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1380 			if (ret) {
1381 				f2fs_put_dnode(&dn);
1382 				return ret;
1383 			}
1384 
1385 			ilen = min((pgoff_t)
1386 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1387 						dn.ofs_in_node, len - i);
1388 			do {
1389 				dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1390 				f2fs_truncate_data_blocks_range(&dn, 1);
1391 
1392 				if (do_replace[i]) {
1393 					f2fs_i_blocks_write(src_inode,
1394 							1, false, false);
1395 					f2fs_i_blocks_write(dst_inode,
1396 							1, true, false);
1397 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1398 					blkaddr[i], ni.version, true, false);
1399 
1400 					do_replace[i] = 0;
1401 				}
1402 				dn.ofs_in_node++;
1403 				i++;
1404 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1405 				if (dst_inode->i_size < new_size)
1406 					f2fs_i_size_write(dst_inode, new_size);
1407 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1408 
1409 			f2fs_put_dnode(&dn);
1410 		} else {
1411 			struct page *psrc, *pdst;
1412 
1413 			psrc = f2fs_get_lock_data_page(src_inode,
1414 							src + i, true);
1415 			if (IS_ERR(psrc))
1416 				return PTR_ERR(psrc);
1417 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1418 								true);
1419 			if (IS_ERR(pdst)) {
1420 				f2fs_put_page(psrc, 1);
1421 				return PTR_ERR(pdst);
1422 			}
1423 
1424 			f2fs_wait_on_page_writeback(pdst, DATA, true, true);
1425 
1426 			memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1427 			set_page_dirty(pdst);
1428 			set_page_private_gcing(pdst);
1429 			f2fs_put_page(pdst, 1);
1430 			f2fs_put_page(psrc, 1);
1431 
1432 			ret = f2fs_truncate_hole(src_inode,
1433 						src + i, src + i + 1);
1434 			if (ret)
1435 				return ret;
1436 			i++;
1437 		}
1438 	}
1439 	return 0;
1440 }
1441 
1442 static int __exchange_data_block(struct inode *src_inode,
1443 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1444 			pgoff_t len, bool full)
1445 {
1446 	block_t *src_blkaddr;
1447 	int *do_replace;
1448 	pgoff_t olen;
1449 	int ret;
1450 
1451 	while (len) {
1452 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1453 
1454 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1455 					array_size(olen, sizeof(block_t)),
1456 					GFP_NOFS);
1457 		if (!src_blkaddr)
1458 			return -ENOMEM;
1459 
1460 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1461 					array_size(olen, sizeof(int)),
1462 					GFP_NOFS);
1463 		if (!do_replace) {
1464 			kvfree(src_blkaddr);
1465 			return -ENOMEM;
1466 		}
1467 
1468 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1469 					do_replace, src, olen);
1470 		if (ret)
1471 			goto roll_back;
1472 
1473 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1474 					do_replace, src, dst, olen, full);
1475 		if (ret)
1476 			goto roll_back;
1477 
1478 		src += olen;
1479 		dst += olen;
1480 		len -= olen;
1481 
1482 		kvfree(src_blkaddr);
1483 		kvfree(do_replace);
1484 	}
1485 	return 0;
1486 
1487 roll_back:
1488 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1489 	kvfree(src_blkaddr);
1490 	kvfree(do_replace);
1491 	return ret;
1492 }
1493 
1494 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1495 {
1496 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1497 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1498 	pgoff_t start = offset >> PAGE_SHIFT;
1499 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1500 	int ret;
1501 
1502 	f2fs_balance_fs(sbi, true);
1503 
1504 	/* avoid gc operation during block exchange */
1505 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1506 	filemap_invalidate_lock(inode->i_mapping);
1507 
1508 	f2fs_lock_op(sbi);
1509 	f2fs_drop_extent_tree(inode);
1510 	truncate_pagecache(inode, offset);
1511 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1512 	f2fs_unlock_op(sbi);
1513 
1514 	filemap_invalidate_unlock(inode->i_mapping);
1515 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1516 	return ret;
1517 }
1518 
1519 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1520 {
1521 	loff_t new_size;
1522 	int ret;
1523 
1524 	if (offset + len >= i_size_read(inode))
1525 		return -EINVAL;
1526 
1527 	/* collapse range should be aligned to block size of f2fs. */
1528 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1529 		return -EINVAL;
1530 
1531 	ret = f2fs_convert_inline_inode(inode);
1532 	if (ret)
1533 		return ret;
1534 
1535 	/* write out all dirty pages from offset */
1536 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1537 	if (ret)
1538 		return ret;
1539 
1540 	ret = f2fs_do_collapse(inode, offset, len);
1541 	if (ret)
1542 		return ret;
1543 
1544 	/* write out all moved pages, if possible */
1545 	filemap_invalidate_lock(inode->i_mapping);
1546 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1547 	truncate_pagecache(inode, offset);
1548 
1549 	new_size = i_size_read(inode) - len;
1550 	ret = f2fs_truncate_blocks(inode, new_size, true);
1551 	filemap_invalidate_unlock(inode->i_mapping);
1552 	if (!ret)
1553 		f2fs_i_size_write(inode, new_size);
1554 	return ret;
1555 }
1556 
1557 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1558 								pgoff_t end)
1559 {
1560 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1561 	pgoff_t index = start;
1562 	unsigned int ofs_in_node = dn->ofs_in_node;
1563 	blkcnt_t count = 0;
1564 	int ret;
1565 
1566 	for (; index < end; index++, dn->ofs_in_node++) {
1567 		if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1568 			count++;
1569 	}
1570 
1571 	dn->ofs_in_node = ofs_in_node;
1572 	ret = f2fs_reserve_new_blocks(dn, count);
1573 	if (ret)
1574 		return ret;
1575 
1576 	dn->ofs_in_node = ofs_in_node;
1577 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1578 		dn->data_blkaddr = f2fs_data_blkaddr(dn);
1579 		/*
1580 		 * f2fs_reserve_new_blocks will not guarantee entire block
1581 		 * allocation.
1582 		 */
1583 		if (dn->data_blkaddr == NULL_ADDR) {
1584 			ret = -ENOSPC;
1585 			break;
1586 		}
1587 
1588 		if (dn->data_blkaddr == NEW_ADDR)
1589 			continue;
1590 
1591 		if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1592 					DATA_GENERIC_ENHANCE)) {
1593 			ret = -EFSCORRUPTED;
1594 			break;
1595 		}
1596 
1597 		f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1);
1598 		f2fs_set_data_blkaddr(dn, NEW_ADDR);
1599 	}
1600 
1601 	f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
1602 	f2fs_update_age_extent_cache_range(dn, start, index - start);
1603 
1604 	return ret;
1605 }
1606 
1607 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1608 								int mode)
1609 {
1610 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1611 	struct address_space *mapping = inode->i_mapping;
1612 	pgoff_t index, pg_start, pg_end;
1613 	loff_t new_size = i_size_read(inode);
1614 	loff_t off_start, off_end;
1615 	int ret = 0;
1616 
1617 	ret = inode_newsize_ok(inode, (len + offset));
1618 	if (ret)
1619 		return ret;
1620 
1621 	ret = f2fs_convert_inline_inode(inode);
1622 	if (ret)
1623 		return ret;
1624 
1625 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1626 	if (ret)
1627 		return ret;
1628 
1629 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1630 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1631 
1632 	off_start = offset & (PAGE_SIZE - 1);
1633 	off_end = (offset + len) & (PAGE_SIZE - 1);
1634 
1635 	if (pg_start == pg_end) {
1636 		ret = fill_zero(inode, pg_start, off_start,
1637 						off_end - off_start);
1638 		if (ret)
1639 			return ret;
1640 
1641 		new_size = max_t(loff_t, new_size, offset + len);
1642 	} else {
1643 		if (off_start) {
1644 			ret = fill_zero(inode, pg_start++, off_start,
1645 						PAGE_SIZE - off_start);
1646 			if (ret)
1647 				return ret;
1648 
1649 			new_size = max_t(loff_t, new_size,
1650 					(loff_t)pg_start << PAGE_SHIFT);
1651 		}
1652 
1653 		for (index = pg_start; index < pg_end;) {
1654 			struct dnode_of_data dn;
1655 			unsigned int end_offset;
1656 			pgoff_t end;
1657 
1658 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1659 			filemap_invalidate_lock(mapping);
1660 
1661 			truncate_pagecache_range(inode,
1662 				(loff_t)index << PAGE_SHIFT,
1663 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1664 
1665 			f2fs_lock_op(sbi);
1666 
1667 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1668 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1669 			if (ret) {
1670 				f2fs_unlock_op(sbi);
1671 				filemap_invalidate_unlock(mapping);
1672 				f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1673 				goto out;
1674 			}
1675 
1676 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1677 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1678 
1679 			ret = f2fs_do_zero_range(&dn, index, end);
1680 			f2fs_put_dnode(&dn);
1681 
1682 			f2fs_unlock_op(sbi);
1683 			filemap_invalidate_unlock(mapping);
1684 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1685 
1686 			f2fs_balance_fs(sbi, dn.node_changed);
1687 
1688 			if (ret)
1689 				goto out;
1690 
1691 			index = end;
1692 			new_size = max_t(loff_t, new_size,
1693 					(loff_t)index << PAGE_SHIFT);
1694 		}
1695 
1696 		if (off_end) {
1697 			ret = fill_zero(inode, pg_end, 0, off_end);
1698 			if (ret)
1699 				goto out;
1700 
1701 			new_size = max_t(loff_t, new_size, offset + len);
1702 		}
1703 	}
1704 
1705 out:
1706 	if (new_size > i_size_read(inode)) {
1707 		if (mode & FALLOC_FL_KEEP_SIZE)
1708 			file_set_keep_isize(inode);
1709 		else
1710 			f2fs_i_size_write(inode, new_size);
1711 	}
1712 	return ret;
1713 }
1714 
1715 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1716 {
1717 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1718 	struct address_space *mapping = inode->i_mapping;
1719 	pgoff_t nr, pg_start, pg_end, delta, idx;
1720 	loff_t new_size;
1721 	int ret = 0;
1722 
1723 	new_size = i_size_read(inode) + len;
1724 	ret = inode_newsize_ok(inode, new_size);
1725 	if (ret)
1726 		return ret;
1727 
1728 	if (offset >= i_size_read(inode))
1729 		return -EINVAL;
1730 
1731 	/* insert range should be aligned to block size of f2fs. */
1732 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1733 		return -EINVAL;
1734 
1735 	ret = f2fs_convert_inline_inode(inode);
1736 	if (ret)
1737 		return ret;
1738 
1739 	f2fs_balance_fs(sbi, true);
1740 
1741 	filemap_invalidate_lock(mapping);
1742 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1743 	filemap_invalidate_unlock(mapping);
1744 	if (ret)
1745 		return ret;
1746 
1747 	/* write out all dirty pages from offset */
1748 	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1749 	if (ret)
1750 		return ret;
1751 
1752 	pg_start = offset >> PAGE_SHIFT;
1753 	pg_end = (offset + len) >> PAGE_SHIFT;
1754 	delta = pg_end - pg_start;
1755 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1756 
1757 	/* avoid gc operation during block exchange */
1758 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1759 	filemap_invalidate_lock(mapping);
1760 	truncate_pagecache(inode, offset);
1761 
1762 	while (!ret && idx > pg_start) {
1763 		nr = idx - pg_start;
1764 		if (nr > delta)
1765 			nr = delta;
1766 		idx -= nr;
1767 
1768 		f2fs_lock_op(sbi);
1769 		f2fs_drop_extent_tree(inode);
1770 
1771 		ret = __exchange_data_block(inode, inode, idx,
1772 					idx + delta, nr, false);
1773 		f2fs_unlock_op(sbi);
1774 	}
1775 	filemap_invalidate_unlock(mapping);
1776 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1777 	if (ret)
1778 		return ret;
1779 
1780 	/* write out all moved pages, if possible */
1781 	filemap_invalidate_lock(mapping);
1782 	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1783 	truncate_pagecache(inode, offset);
1784 	filemap_invalidate_unlock(mapping);
1785 
1786 	if (!ret)
1787 		f2fs_i_size_write(inode, new_size);
1788 	return ret;
1789 }
1790 
1791 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
1792 					loff_t len, int mode)
1793 {
1794 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1795 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1796 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1797 			.m_may_create = true };
1798 	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1799 			.init_gc_type = FG_GC,
1800 			.should_migrate_blocks = false,
1801 			.err_gc_skipped = true,
1802 			.nr_free_secs = 0 };
1803 	pgoff_t pg_start, pg_end;
1804 	loff_t new_size;
1805 	loff_t off_end;
1806 	block_t expanded = 0;
1807 	int err;
1808 
1809 	err = inode_newsize_ok(inode, (len + offset));
1810 	if (err)
1811 		return err;
1812 
1813 	err = f2fs_convert_inline_inode(inode);
1814 	if (err)
1815 		return err;
1816 
1817 	f2fs_balance_fs(sbi, true);
1818 
1819 	pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1820 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1821 	off_end = (offset + len) & (PAGE_SIZE - 1);
1822 
1823 	map.m_lblk = pg_start;
1824 	map.m_len = pg_end - pg_start;
1825 	if (off_end)
1826 		map.m_len++;
1827 
1828 	if (!map.m_len)
1829 		return 0;
1830 
1831 	if (f2fs_is_pinned_file(inode)) {
1832 		block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1833 		block_t sec_len = roundup(map.m_len, sec_blks);
1834 
1835 		map.m_len = sec_blks;
1836 next_alloc:
1837 		if (has_not_enough_free_secs(sbi, 0, f2fs_sb_has_blkzoned(sbi) ?
1838 			ZONED_PIN_SEC_REQUIRED_COUNT :
1839 			GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1840 			f2fs_down_write(&sbi->gc_lock);
1841 			stat_inc_gc_call_count(sbi, FOREGROUND);
1842 			err = f2fs_gc(sbi, &gc_control);
1843 			if (err && err != -ENODATA)
1844 				goto out_err;
1845 		}
1846 
1847 		f2fs_down_write(&sbi->pin_sem);
1848 
1849 		err = f2fs_allocate_pinning_section(sbi);
1850 		if (err) {
1851 			f2fs_up_write(&sbi->pin_sem);
1852 			goto out_err;
1853 		}
1854 
1855 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1856 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
1857 		file_dont_truncate(inode);
1858 
1859 		f2fs_up_write(&sbi->pin_sem);
1860 
1861 		expanded += map.m_len;
1862 		sec_len -= map.m_len;
1863 		map.m_lblk += map.m_len;
1864 		if (!err && sec_len)
1865 			goto next_alloc;
1866 
1867 		map.m_len = expanded;
1868 	} else {
1869 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
1870 		expanded = map.m_len;
1871 	}
1872 out_err:
1873 	if (err) {
1874 		pgoff_t last_off;
1875 
1876 		if (!expanded)
1877 			return err;
1878 
1879 		last_off = pg_start + expanded - 1;
1880 
1881 		/* update new size to the failed position */
1882 		new_size = (last_off == pg_end) ? offset + len :
1883 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1884 	} else {
1885 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1886 	}
1887 
1888 	if (new_size > i_size_read(inode)) {
1889 		if (mode & FALLOC_FL_KEEP_SIZE)
1890 			file_set_keep_isize(inode);
1891 		else
1892 			f2fs_i_size_write(inode, new_size);
1893 	}
1894 
1895 	return err;
1896 }
1897 
1898 static long f2fs_fallocate(struct file *file, int mode,
1899 				loff_t offset, loff_t len)
1900 {
1901 	struct inode *inode = file_inode(file);
1902 	long ret = 0;
1903 
1904 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1905 		return -EIO;
1906 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1907 		return -ENOSPC;
1908 	if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode))
1909 		return -EOPNOTSUPP;
1910 
1911 	/* f2fs only support ->fallocate for regular file */
1912 	if (!S_ISREG(inode->i_mode))
1913 		return -EINVAL;
1914 
1915 	if (IS_ENCRYPTED(inode) &&
1916 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1917 		return -EOPNOTSUPP;
1918 
1919 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1920 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1921 			FALLOC_FL_INSERT_RANGE))
1922 		return -EOPNOTSUPP;
1923 
1924 	inode_lock(inode);
1925 
1926 	/*
1927 	 * Pinned file should not support partial truncation since the block
1928 	 * can be used by applications.
1929 	 */
1930 	if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1931 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1932 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) {
1933 		ret = -EOPNOTSUPP;
1934 		goto out;
1935 	}
1936 
1937 	ret = file_modified(file);
1938 	if (ret)
1939 		goto out;
1940 
1941 	/*
1942 	 * wait for inflight dio, blocks should be removed after IO
1943 	 * completion.
1944 	 */
1945 	inode_dio_wait(inode);
1946 
1947 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1948 		if (offset >= inode->i_size)
1949 			goto out;
1950 
1951 		ret = f2fs_punch_hole(inode, offset, len);
1952 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1953 		ret = f2fs_collapse_range(inode, offset, len);
1954 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1955 		ret = f2fs_zero_range(inode, offset, len, mode);
1956 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1957 		ret = f2fs_insert_range(inode, offset, len);
1958 	} else {
1959 		ret = f2fs_expand_inode_data(inode, offset, len, mode);
1960 	}
1961 
1962 	if (!ret) {
1963 		inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1964 		f2fs_mark_inode_dirty_sync(inode, false);
1965 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1966 	}
1967 
1968 out:
1969 	inode_unlock(inode);
1970 
1971 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1972 	return ret;
1973 }
1974 
1975 static int f2fs_release_file(struct inode *inode, struct file *filp)
1976 {
1977 	/*
1978 	 * f2fs_release_file is called at every close calls. So we should
1979 	 * not drop any inmemory pages by close called by other process.
1980 	 */
1981 	if (!(filp->f_mode & FMODE_WRITE) ||
1982 			atomic_read(&inode->i_writecount) != 1)
1983 		return 0;
1984 
1985 	inode_lock(inode);
1986 	f2fs_abort_atomic_write(inode, true);
1987 	inode_unlock(inode);
1988 
1989 	return 0;
1990 }
1991 
1992 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1993 {
1994 	struct inode *inode = file_inode(file);
1995 
1996 	/*
1997 	 * If the process doing a transaction is crashed, we should do
1998 	 * roll-back. Otherwise, other reader/write can see corrupted database
1999 	 * until all the writers close its file. Since this should be done
2000 	 * before dropping file lock, it needs to do in ->flush.
2001 	 */
2002 	if (F2FS_I(inode)->atomic_write_task == current &&
2003 				(current->flags & PF_EXITING)) {
2004 		inode_lock(inode);
2005 		f2fs_abort_atomic_write(inode, true);
2006 		inode_unlock(inode);
2007 	}
2008 
2009 	return 0;
2010 }
2011 
2012 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
2013 {
2014 	struct f2fs_inode_info *fi = F2FS_I(inode);
2015 	u32 masked_flags = fi->i_flags & mask;
2016 
2017 	/* mask can be shrunk by flags_valid selector */
2018 	iflags &= mask;
2019 
2020 	/* Is it quota file? Do not allow user to mess with it */
2021 	if (IS_NOQUOTA(inode))
2022 		return -EPERM;
2023 
2024 	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
2025 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
2026 			return -EOPNOTSUPP;
2027 		if (!f2fs_empty_dir(inode))
2028 			return -ENOTEMPTY;
2029 	}
2030 
2031 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
2032 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
2033 			return -EOPNOTSUPP;
2034 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
2035 			return -EINVAL;
2036 	}
2037 
2038 	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
2039 		if (masked_flags & F2FS_COMPR_FL) {
2040 			if (!f2fs_disable_compressed_file(inode))
2041 				return -EINVAL;
2042 		} else {
2043 			/* try to convert inline_data to support compression */
2044 			int err = f2fs_convert_inline_inode(inode);
2045 			if (err)
2046 				return err;
2047 
2048 			f2fs_down_write(&fi->i_sem);
2049 			if (!f2fs_may_compress(inode) ||
2050 					(S_ISREG(inode->i_mode) &&
2051 					F2FS_HAS_BLOCKS(inode))) {
2052 				f2fs_up_write(&fi->i_sem);
2053 				return -EINVAL;
2054 			}
2055 			err = set_compress_context(inode);
2056 			f2fs_up_write(&fi->i_sem);
2057 
2058 			if (err)
2059 				return err;
2060 		}
2061 	}
2062 
2063 	fi->i_flags = iflags | (fi->i_flags & ~mask);
2064 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
2065 					(fi->i_flags & F2FS_NOCOMP_FL));
2066 
2067 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
2068 		set_inode_flag(inode, FI_PROJ_INHERIT);
2069 	else
2070 		clear_inode_flag(inode, FI_PROJ_INHERIT);
2071 
2072 	inode_set_ctime_current(inode);
2073 	f2fs_set_inode_flags(inode);
2074 	f2fs_mark_inode_dirty_sync(inode, true);
2075 	return 0;
2076 }
2077 
2078 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
2079 
2080 /*
2081  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
2082  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
2083  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
2084  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
2085  *
2086  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
2087  * FS_IOC_FSSETXATTR is done by the VFS.
2088  */
2089 
2090 static const struct {
2091 	u32 iflag;
2092 	u32 fsflag;
2093 } f2fs_fsflags_map[] = {
2094 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
2095 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
2096 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
2097 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
2098 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
2099 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
2100 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
2101 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
2102 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
2103 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
2104 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
2105 };
2106 
2107 #define F2FS_GETTABLE_FS_FL (		\
2108 		FS_COMPR_FL |		\
2109 		FS_SYNC_FL |		\
2110 		FS_IMMUTABLE_FL |	\
2111 		FS_APPEND_FL |		\
2112 		FS_NODUMP_FL |		\
2113 		FS_NOATIME_FL |		\
2114 		FS_NOCOMP_FL |		\
2115 		FS_INDEX_FL |		\
2116 		FS_DIRSYNC_FL |		\
2117 		FS_PROJINHERIT_FL |	\
2118 		FS_ENCRYPT_FL |		\
2119 		FS_INLINE_DATA_FL |	\
2120 		FS_NOCOW_FL |		\
2121 		FS_VERITY_FL |		\
2122 		FS_CASEFOLD_FL)
2123 
2124 #define F2FS_SETTABLE_FS_FL (		\
2125 		FS_COMPR_FL |		\
2126 		FS_SYNC_FL |		\
2127 		FS_IMMUTABLE_FL |	\
2128 		FS_APPEND_FL |		\
2129 		FS_NODUMP_FL |		\
2130 		FS_NOATIME_FL |		\
2131 		FS_NOCOMP_FL |		\
2132 		FS_DIRSYNC_FL |		\
2133 		FS_PROJINHERIT_FL |	\
2134 		FS_CASEFOLD_FL)
2135 
2136 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2137 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2138 {
2139 	u32 fsflags = 0;
2140 	int i;
2141 
2142 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2143 		if (iflags & f2fs_fsflags_map[i].iflag)
2144 			fsflags |= f2fs_fsflags_map[i].fsflag;
2145 
2146 	return fsflags;
2147 }
2148 
2149 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2150 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2151 {
2152 	u32 iflags = 0;
2153 	int i;
2154 
2155 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2156 		if (fsflags & f2fs_fsflags_map[i].fsflag)
2157 			iflags |= f2fs_fsflags_map[i].iflag;
2158 
2159 	return iflags;
2160 }
2161 
2162 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2163 {
2164 	struct inode *inode = file_inode(filp);
2165 
2166 	return put_user(inode->i_generation, (int __user *)arg);
2167 }
2168 
2169 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2170 {
2171 	struct inode *inode = file_inode(filp);
2172 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2173 	struct f2fs_inode_info *fi = F2FS_I(inode);
2174 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2175 	loff_t isize;
2176 	int ret;
2177 
2178 	if (!(filp->f_mode & FMODE_WRITE))
2179 		return -EBADF;
2180 
2181 	if (!inode_owner_or_capable(idmap, inode))
2182 		return -EACCES;
2183 
2184 	if (!S_ISREG(inode->i_mode))
2185 		return -EINVAL;
2186 
2187 	if (filp->f_flags & O_DIRECT)
2188 		return -EINVAL;
2189 
2190 	ret = mnt_want_write_file(filp);
2191 	if (ret)
2192 		return ret;
2193 
2194 	inode_lock(inode);
2195 
2196 	if (!f2fs_disable_compressed_file(inode) ||
2197 			f2fs_is_pinned_file(inode)) {
2198 		ret = -EINVAL;
2199 		goto out;
2200 	}
2201 
2202 	if (f2fs_is_atomic_file(inode))
2203 		goto out;
2204 
2205 	ret = f2fs_convert_inline_inode(inode);
2206 	if (ret)
2207 		goto out;
2208 
2209 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2210 	f2fs_down_write(&fi->i_gc_rwsem[READ]);
2211 
2212 	/*
2213 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2214 	 * f2fs_is_atomic_file.
2215 	 */
2216 	if (get_dirty_pages(inode))
2217 		f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2218 			  inode->i_ino, get_dirty_pages(inode));
2219 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2220 	if (ret)
2221 		goto out_unlock;
2222 
2223 	/* Check if the inode already has a COW inode */
2224 	if (fi->cow_inode == NULL) {
2225 		/* Create a COW inode for atomic write */
2226 		struct dentry *dentry = file_dentry(filp);
2227 		struct inode *dir = d_inode(dentry->d_parent);
2228 
2229 		ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode);
2230 		if (ret)
2231 			goto out_unlock;
2232 
2233 		set_inode_flag(fi->cow_inode, FI_COW_FILE);
2234 		clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2235 
2236 		/* Set the COW inode's atomic_inode to the atomic inode */
2237 		F2FS_I(fi->cow_inode)->atomic_inode = inode;
2238 	} else {
2239 		/* Reuse the already created COW inode */
2240 		f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode));
2241 
2242 		invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1);
2243 
2244 		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2245 		if (ret)
2246 			goto out_unlock;
2247 	}
2248 
2249 	f2fs_write_inode(inode, NULL);
2250 
2251 	stat_inc_atomic_inode(inode);
2252 
2253 	set_inode_flag(inode, FI_ATOMIC_FILE);
2254 
2255 	isize = i_size_read(inode);
2256 	fi->original_i_size = isize;
2257 	if (truncate) {
2258 		set_inode_flag(inode, FI_ATOMIC_REPLACE);
2259 		truncate_inode_pages_final(inode->i_mapping);
2260 		f2fs_i_size_write(inode, 0);
2261 		isize = 0;
2262 	}
2263 	f2fs_i_size_write(fi->cow_inode, isize);
2264 
2265 out_unlock:
2266 	f2fs_up_write(&fi->i_gc_rwsem[READ]);
2267 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2268 	if (ret)
2269 		goto out;
2270 
2271 	f2fs_update_time(sbi, REQ_TIME);
2272 	fi->atomic_write_task = current;
2273 	stat_update_max_atomic_write(inode);
2274 	fi->atomic_write_cnt = 0;
2275 out:
2276 	inode_unlock(inode);
2277 	mnt_drop_write_file(filp);
2278 	return ret;
2279 }
2280 
2281 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2282 {
2283 	struct inode *inode = file_inode(filp);
2284 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2285 	int ret;
2286 
2287 	if (!(filp->f_mode & FMODE_WRITE))
2288 		return -EBADF;
2289 
2290 	if (!inode_owner_or_capable(idmap, inode))
2291 		return -EACCES;
2292 
2293 	ret = mnt_want_write_file(filp);
2294 	if (ret)
2295 		return ret;
2296 
2297 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2298 
2299 	inode_lock(inode);
2300 
2301 	if (f2fs_is_atomic_file(inode)) {
2302 		ret = f2fs_commit_atomic_write(inode);
2303 		if (!ret)
2304 			ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2305 
2306 		f2fs_abort_atomic_write(inode, ret);
2307 	} else {
2308 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2309 	}
2310 
2311 	inode_unlock(inode);
2312 	mnt_drop_write_file(filp);
2313 	return ret;
2314 }
2315 
2316 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2317 {
2318 	struct inode *inode = file_inode(filp);
2319 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2320 	int ret;
2321 
2322 	if (!(filp->f_mode & FMODE_WRITE))
2323 		return -EBADF;
2324 
2325 	if (!inode_owner_or_capable(idmap, inode))
2326 		return -EACCES;
2327 
2328 	ret = mnt_want_write_file(filp);
2329 	if (ret)
2330 		return ret;
2331 
2332 	inode_lock(inode);
2333 
2334 	f2fs_abort_atomic_write(inode, true);
2335 
2336 	inode_unlock(inode);
2337 
2338 	mnt_drop_write_file(filp);
2339 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2340 	return ret;
2341 }
2342 
2343 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
2344 						bool readonly, bool need_lock)
2345 {
2346 	struct super_block *sb = sbi->sb;
2347 	int ret = 0;
2348 
2349 	switch (flag) {
2350 	case F2FS_GOING_DOWN_FULLSYNC:
2351 		ret = bdev_freeze(sb->s_bdev);
2352 		if (ret)
2353 			goto out;
2354 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2355 		bdev_thaw(sb->s_bdev);
2356 		break;
2357 	case F2FS_GOING_DOWN_METASYNC:
2358 		/* do checkpoint only */
2359 		ret = f2fs_sync_fs(sb, 1);
2360 		if (ret) {
2361 			if (ret == -EIO)
2362 				ret = 0;
2363 			goto out;
2364 		}
2365 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2366 		break;
2367 	case F2FS_GOING_DOWN_NOSYNC:
2368 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2369 		break;
2370 	case F2FS_GOING_DOWN_METAFLUSH:
2371 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2372 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2373 		break;
2374 	case F2FS_GOING_DOWN_NEED_FSCK:
2375 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2376 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2377 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2378 		/* do checkpoint only */
2379 		ret = f2fs_sync_fs(sb, 1);
2380 		if (ret == -EIO)
2381 			ret = 0;
2382 		goto out;
2383 	default:
2384 		ret = -EINVAL;
2385 		goto out;
2386 	}
2387 
2388 	if (readonly)
2389 		goto out;
2390 
2391 	/*
2392 	 * grab sb->s_umount to avoid racing w/ remount() and other shutdown
2393 	 * paths.
2394 	 */
2395 	if (need_lock)
2396 		down_write(&sbi->sb->s_umount);
2397 
2398 	f2fs_stop_gc_thread(sbi);
2399 	f2fs_stop_discard_thread(sbi);
2400 
2401 	f2fs_drop_discard_cmd(sbi);
2402 	clear_opt(sbi, DISCARD);
2403 
2404 	if (need_lock)
2405 		up_write(&sbi->sb->s_umount);
2406 
2407 	f2fs_update_time(sbi, REQ_TIME);
2408 out:
2409 
2410 	trace_f2fs_shutdown(sbi, flag, ret);
2411 
2412 	return ret;
2413 }
2414 
2415 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2416 {
2417 	struct inode *inode = file_inode(filp);
2418 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2419 	__u32 in;
2420 	int ret;
2421 	bool need_drop = false, readonly = false;
2422 
2423 	if (!capable(CAP_SYS_ADMIN))
2424 		return -EPERM;
2425 
2426 	if (get_user(in, (__u32 __user *)arg))
2427 		return -EFAULT;
2428 
2429 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2430 		ret = mnt_want_write_file(filp);
2431 		if (ret) {
2432 			if (ret != -EROFS)
2433 				return ret;
2434 
2435 			/* fallback to nosync shutdown for readonly fs */
2436 			in = F2FS_GOING_DOWN_NOSYNC;
2437 			readonly = true;
2438 		} else {
2439 			need_drop = true;
2440 		}
2441 	}
2442 
2443 	ret = f2fs_do_shutdown(sbi, in, readonly, true);
2444 
2445 	if (need_drop)
2446 		mnt_drop_write_file(filp);
2447 
2448 	return ret;
2449 }
2450 
2451 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2452 {
2453 	struct inode *inode = file_inode(filp);
2454 	struct super_block *sb = inode->i_sb;
2455 	struct fstrim_range range;
2456 	int ret;
2457 
2458 	if (!capable(CAP_SYS_ADMIN))
2459 		return -EPERM;
2460 
2461 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2462 		return -EOPNOTSUPP;
2463 
2464 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2465 				sizeof(range)))
2466 		return -EFAULT;
2467 
2468 	ret = mnt_want_write_file(filp);
2469 	if (ret)
2470 		return ret;
2471 
2472 	range.minlen = max((unsigned int)range.minlen,
2473 			   bdev_discard_granularity(sb->s_bdev));
2474 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2475 	mnt_drop_write_file(filp);
2476 	if (ret < 0)
2477 		return ret;
2478 
2479 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2480 				sizeof(range)))
2481 		return -EFAULT;
2482 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2483 	return 0;
2484 }
2485 
2486 static bool uuid_is_nonzero(__u8 u[16])
2487 {
2488 	int i;
2489 
2490 	for (i = 0; i < 16; i++)
2491 		if (u[i])
2492 			return true;
2493 	return false;
2494 }
2495 
2496 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2497 {
2498 	struct inode *inode = file_inode(filp);
2499 	int ret;
2500 
2501 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2502 		return -EOPNOTSUPP;
2503 
2504 	ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2505 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2506 	return ret;
2507 }
2508 
2509 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2510 {
2511 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2512 		return -EOPNOTSUPP;
2513 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2514 }
2515 
2516 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2517 {
2518 	struct inode *inode = file_inode(filp);
2519 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2520 	u8 encrypt_pw_salt[16];
2521 	int err;
2522 
2523 	if (!f2fs_sb_has_encrypt(sbi))
2524 		return -EOPNOTSUPP;
2525 
2526 	err = mnt_want_write_file(filp);
2527 	if (err)
2528 		return err;
2529 
2530 	f2fs_down_write(&sbi->sb_lock);
2531 
2532 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2533 		goto got_it;
2534 
2535 	/* update superblock with uuid */
2536 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2537 
2538 	err = f2fs_commit_super(sbi, false);
2539 	if (err) {
2540 		/* undo new data */
2541 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2542 		goto out_err;
2543 	}
2544 got_it:
2545 	memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2546 out_err:
2547 	f2fs_up_write(&sbi->sb_lock);
2548 	mnt_drop_write_file(filp);
2549 
2550 	if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2551 		err = -EFAULT;
2552 
2553 	return err;
2554 }
2555 
2556 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2557 					     unsigned long arg)
2558 {
2559 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2560 		return -EOPNOTSUPP;
2561 
2562 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2563 }
2564 
2565 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2566 {
2567 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2568 		return -EOPNOTSUPP;
2569 
2570 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2571 }
2572 
2573 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2574 {
2575 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2576 		return -EOPNOTSUPP;
2577 
2578 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2579 }
2580 
2581 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2582 						    unsigned long arg)
2583 {
2584 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2585 		return -EOPNOTSUPP;
2586 
2587 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2588 }
2589 
2590 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2591 					      unsigned long arg)
2592 {
2593 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2594 		return -EOPNOTSUPP;
2595 
2596 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2597 }
2598 
2599 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2600 {
2601 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2602 		return -EOPNOTSUPP;
2603 
2604 	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2605 }
2606 
2607 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2608 {
2609 	struct inode *inode = file_inode(filp);
2610 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2611 	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2612 			.no_bg_gc = false,
2613 			.should_migrate_blocks = false,
2614 			.nr_free_secs = 0 };
2615 	__u32 sync;
2616 	int ret;
2617 
2618 	if (!capable(CAP_SYS_ADMIN))
2619 		return -EPERM;
2620 
2621 	if (get_user(sync, (__u32 __user *)arg))
2622 		return -EFAULT;
2623 
2624 	if (f2fs_readonly(sbi->sb))
2625 		return -EROFS;
2626 
2627 	ret = mnt_want_write_file(filp);
2628 	if (ret)
2629 		return ret;
2630 
2631 	if (!sync) {
2632 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2633 			ret = -EBUSY;
2634 			goto out;
2635 		}
2636 	} else {
2637 		f2fs_down_write(&sbi->gc_lock);
2638 	}
2639 
2640 	gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2641 	gc_control.err_gc_skipped = sync;
2642 	stat_inc_gc_call_count(sbi, FOREGROUND);
2643 	ret = f2fs_gc(sbi, &gc_control);
2644 out:
2645 	mnt_drop_write_file(filp);
2646 	return ret;
2647 }
2648 
2649 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2650 {
2651 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2652 	struct f2fs_gc_control gc_control = {
2653 			.init_gc_type = range->sync ? FG_GC : BG_GC,
2654 			.no_bg_gc = false,
2655 			.should_migrate_blocks = false,
2656 			.err_gc_skipped = range->sync,
2657 			.nr_free_secs = 0 };
2658 	u64 end;
2659 	int ret;
2660 
2661 	if (!capable(CAP_SYS_ADMIN))
2662 		return -EPERM;
2663 	if (f2fs_readonly(sbi->sb))
2664 		return -EROFS;
2665 
2666 	end = range->start + range->len;
2667 	if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2668 					end >= MAX_BLKADDR(sbi))
2669 		return -EINVAL;
2670 
2671 	ret = mnt_want_write_file(filp);
2672 	if (ret)
2673 		return ret;
2674 
2675 do_more:
2676 	if (!range->sync) {
2677 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2678 			ret = -EBUSY;
2679 			goto out;
2680 		}
2681 	} else {
2682 		f2fs_down_write(&sbi->gc_lock);
2683 	}
2684 
2685 	gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2686 	stat_inc_gc_call_count(sbi, FOREGROUND);
2687 	ret = f2fs_gc(sbi, &gc_control);
2688 	if (ret) {
2689 		if (ret == -EBUSY)
2690 			ret = -EAGAIN;
2691 		goto out;
2692 	}
2693 	range->start += CAP_BLKS_PER_SEC(sbi);
2694 	if (range->start <= end)
2695 		goto do_more;
2696 out:
2697 	mnt_drop_write_file(filp);
2698 	return ret;
2699 }
2700 
2701 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2702 {
2703 	struct f2fs_gc_range range;
2704 
2705 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2706 							sizeof(range)))
2707 		return -EFAULT;
2708 	return __f2fs_ioc_gc_range(filp, &range);
2709 }
2710 
2711 static int f2fs_ioc_write_checkpoint(struct file *filp)
2712 {
2713 	struct inode *inode = file_inode(filp);
2714 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2715 	int ret;
2716 
2717 	if (!capable(CAP_SYS_ADMIN))
2718 		return -EPERM;
2719 
2720 	if (f2fs_readonly(sbi->sb))
2721 		return -EROFS;
2722 
2723 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2724 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2725 		return -EINVAL;
2726 	}
2727 
2728 	ret = mnt_want_write_file(filp);
2729 	if (ret)
2730 		return ret;
2731 
2732 	ret = f2fs_sync_fs(sbi->sb, 1);
2733 
2734 	mnt_drop_write_file(filp);
2735 	return ret;
2736 }
2737 
2738 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2739 					struct file *filp,
2740 					struct f2fs_defragment *range)
2741 {
2742 	struct inode *inode = file_inode(filp);
2743 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2744 					.m_seg_type = NO_CHECK_TYPE,
2745 					.m_may_create = false };
2746 	struct extent_info ei = {};
2747 	pgoff_t pg_start, pg_end, next_pgofs;
2748 	unsigned int total = 0, sec_num;
2749 	block_t blk_end = 0;
2750 	bool fragmented = false;
2751 	int err;
2752 
2753 	f2fs_balance_fs(sbi, true);
2754 
2755 	inode_lock(inode);
2756 	pg_start = range->start >> PAGE_SHIFT;
2757 	pg_end = min_t(pgoff_t,
2758 				(range->start + range->len) >> PAGE_SHIFT,
2759 				DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));
2760 
2761 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) ||
2762 		f2fs_is_atomic_file(inode)) {
2763 		err = -EINVAL;
2764 		goto unlock_out;
2765 	}
2766 
2767 	/* if in-place-update policy is enabled, don't waste time here */
2768 	set_inode_flag(inode, FI_OPU_WRITE);
2769 	if (f2fs_should_update_inplace(inode, NULL)) {
2770 		err = -EINVAL;
2771 		goto out;
2772 	}
2773 
2774 	/* writeback all dirty pages in the range */
2775 	err = filemap_write_and_wait_range(inode->i_mapping,
2776 						pg_start << PAGE_SHIFT,
2777 						(pg_end << PAGE_SHIFT) - 1);
2778 	if (err)
2779 		goto out;
2780 
2781 	/*
2782 	 * lookup mapping info in extent cache, skip defragmenting if physical
2783 	 * block addresses are continuous.
2784 	 */
2785 	if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2786 		if ((pgoff_t)ei.fofs + ei.len >= pg_end)
2787 			goto out;
2788 	}
2789 
2790 	map.m_lblk = pg_start;
2791 	map.m_next_pgofs = &next_pgofs;
2792 
2793 	/*
2794 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2795 	 * physical block addresses are continuous even if there are hole(s)
2796 	 * in logical blocks.
2797 	 */
2798 	while (map.m_lblk < pg_end) {
2799 		map.m_len = pg_end - map.m_lblk;
2800 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2801 		if (err)
2802 			goto out;
2803 
2804 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2805 			map.m_lblk = next_pgofs;
2806 			continue;
2807 		}
2808 
2809 		if (blk_end && blk_end != map.m_pblk)
2810 			fragmented = true;
2811 
2812 		/* record total count of block that we're going to move */
2813 		total += map.m_len;
2814 
2815 		blk_end = map.m_pblk + map.m_len;
2816 
2817 		map.m_lblk += map.m_len;
2818 	}
2819 
2820 	if (!fragmented) {
2821 		total = 0;
2822 		goto out;
2823 	}
2824 
2825 	sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2826 
2827 	/*
2828 	 * make sure there are enough free section for LFS allocation, this can
2829 	 * avoid defragment running in SSR mode when free section are allocated
2830 	 * intensively
2831 	 */
2832 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2833 		err = -EAGAIN;
2834 		goto out;
2835 	}
2836 
2837 	map.m_lblk = pg_start;
2838 	map.m_len = pg_end - pg_start;
2839 	total = 0;
2840 
2841 	while (map.m_lblk < pg_end) {
2842 		pgoff_t idx;
2843 		int cnt = 0;
2844 
2845 do_map:
2846 		map.m_len = pg_end - map.m_lblk;
2847 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2848 		if (err)
2849 			goto clear_out;
2850 
2851 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2852 			map.m_lblk = next_pgofs;
2853 			goto check;
2854 		}
2855 
2856 		set_inode_flag(inode, FI_SKIP_WRITES);
2857 
2858 		idx = map.m_lblk;
2859 		while (idx < map.m_lblk + map.m_len &&
2860 						cnt < BLKS_PER_SEG(sbi)) {
2861 			struct page *page;
2862 
2863 			page = f2fs_get_lock_data_page(inode, idx, true);
2864 			if (IS_ERR(page)) {
2865 				err = PTR_ERR(page);
2866 				goto clear_out;
2867 			}
2868 
2869 			f2fs_wait_on_page_writeback(page, DATA, true, true);
2870 
2871 			set_page_dirty(page);
2872 			set_page_private_gcing(page);
2873 			f2fs_put_page(page, 1);
2874 
2875 			idx++;
2876 			cnt++;
2877 			total++;
2878 		}
2879 
2880 		map.m_lblk = idx;
2881 check:
2882 		if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi))
2883 			goto do_map;
2884 
2885 		clear_inode_flag(inode, FI_SKIP_WRITES);
2886 
2887 		err = filemap_fdatawrite(inode->i_mapping);
2888 		if (err)
2889 			goto out;
2890 	}
2891 clear_out:
2892 	clear_inode_flag(inode, FI_SKIP_WRITES);
2893 out:
2894 	clear_inode_flag(inode, FI_OPU_WRITE);
2895 unlock_out:
2896 	inode_unlock(inode);
2897 	if (!err)
2898 		range->len = (u64)total << PAGE_SHIFT;
2899 	return err;
2900 }
2901 
2902 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2903 {
2904 	struct inode *inode = file_inode(filp);
2905 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2906 	struct f2fs_defragment range;
2907 	int err;
2908 
2909 	if (!capable(CAP_SYS_ADMIN))
2910 		return -EPERM;
2911 
2912 	if (!S_ISREG(inode->i_mode))
2913 		return -EINVAL;
2914 
2915 	if (f2fs_readonly(sbi->sb))
2916 		return -EROFS;
2917 
2918 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2919 							sizeof(range)))
2920 		return -EFAULT;
2921 
2922 	/* verify alignment of offset & size */
2923 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2924 		return -EINVAL;
2925 
2926 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2927 					max_file_blocks(inode)))
2928 		return -EINVAL;
2929 
2930 	err = mnt_want_write_file(filp);
2931 	if (err)
2932 		return err;
2933 
2934 	err = f2fs_defragment_range(sbi, filp, &range);
2935 	mnt_drop_write_file(filp);
2936 
2937 	if (range.len)
2938 		f2fs_update_time(sbi, REQ_TIME);
2939 	if (err < 0)
2940 		return err;
2941 
2942 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2943 							sizeof(range)))
2944 		return -EFAULT;
2945 
2946 	return 0;
2947 }
2948 
2949 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2950 			struct file *file_out, loff_t pos_out, size_t len)
2951 {
2952 	struct inode *src = file_inode(file_in);
2953 	struct inode *dst = file_inode(file_out);
2954 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2955 	size_t olen = len, dst_max_i_size = 0;
2956 	size_t dst_osize;
2957 	int ret;
2958 
2959 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2960 				src->i_sb != dst->i_sb)
2961 		return -EXDEV;
2962 
2963 	if (unlikely(f2fs_readonly(src->i_sb)))
2964 		return -EROFS;
2965 
2966 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2967 		return -EINVAL;
2968 
2969 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2970 		return -EOPNOTSUPP;
2971 
2972 	if (pos_out < 0 || pos_in < 0)
2973 		return -EINVAL;
2974 
2975 	if (src == dst) {
2976 		if (pos_in == pos_out)
2977 			return 0;
2978 		if (pos_out > pos_in && pos_out < pos_in + len)
2979 			return -EINVAL;
2980 	}
2981 
2982 	inode_lock(src);
2983 	if (src != dst) {
2984 		ret = -EBUSY;
2985 		if (!inode_trylock(dst))
2986 			goto out;
2987 	}
2988 
2989 	if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) ||
2990 		f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) {
2991 		ret = -EOPNOTSUPP;
2992 		goto out_unlock;
2993 	}
2994 
2995 	if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) {
2996 		ret = -EINVAL;
2997 		goto out_unlock;
2998 	}
2999 
3000 	ret = -EINVAL;
3001 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
3002 		goto out_unlock;
3003 	if (len == 0)
3004 		olen = len = src->i_size - pos_in;
3005 	if (pos_in + len == src->i_size)
3006 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
3007 	if (len == 0) {
3008 		ret = 0;
3009 		goto out_unlock;
3010 	}
3011 
3012 	dst_osize = dst->i_size;
3013 	if (pos_out + olen > dst->i_size)
3014 		dst_max_i_size = pos_out + olen;
3015 
3016 	/* verify the end result is block aligned */
3017 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
3018 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
3019 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
3020 		goto out_unlock;
3021 
3022 	ret = f2fs_convert_inline_inode(src);
3023 	if (ret)
3024 		goto out_unlock;
3025 
3026 	ret = f2fs_convert_inline_inode(dst);
3027 	if (ret)
3028 		goto out_unlock;
3029 
3030 	/* write out all dirty pages from offset */
3031 	ret = filemap_write_and_wait_range(src->i_mapping,
3032 					pos_in, pos_in + len);
3033 	if (ret)
3034 		goto out_unlock;
3035 
3036 	ret = filemap_write_and_wait_range(dst->i_mapping,
3037 					pos_out, pos_out + len);
3038 	if (ret)
3039 		goto out_unlock;
3040 
3041 	f2fs_balance_fs(sbi, true);
3042 
3043 	f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3044 	if (src != dst) {
3045 		ret = -EBUSY;
3046 		if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
3047 			goto out_src;
3048 	}
3049 
3050 	f2fs_lock_op(sbi);
3051 	ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in),
3052 				F2FS_BYTES_TO_BLK(pos_out),
3053 				F2FS_BYTES_TO_BLK(len), false);
3054 
3055 	if (!ret) {
3056 		if (dst_max_i_size)
3057 			f2fs_i_size_write(dst, dst_max_i_size);
3058 		else if (dst_osize != dst->i_size)
3059 			f2fs_i_size_write(dst, dst_osize);
3060 	}
3061 	f2fs_unlock_op(sbi);
3062 
3063 	if (src != dst)
3064 		f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
3065 out_src:
3066 	f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3067 	if (ret)
3068 		goto out_unlock;
3069 
3070 	inode_set_mtime_to_ts(src, inode_set_ctime_current(src));
3071 	f2fs_mark_inode_dirty_sync(src, false);
3072 	if (src != dst) {
3073 		inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst));
3074 		f2fs_mark_inode_dirty_sync(dst, false);
3075 	}
3076 	f2fs_update_time(sbi, REQ_TIME);
3077 
3078 out_unlock:
3079 	if (src != dst)
3080 		inode_unlock(dst);
3081 out:
3082 	inode_unlock(src);
3083 	return ret;
3084 }
3085 
3086 static int __f2fs_ioc_move_range(struct file *filp,
3087 				struct f2fs_move_range *range)
3088 {
3089 	int err;
3090 
3091 	if (!(filp->f_mode & FMODE_READ) ||
3092 			!(filp->f_mode & FMODE_WRITE))
3093 		return -EBADF;
3094 
3095 	CLASS(fd, dst)(range->dst_fd);
3096 	if (fd_empty(dst))
3097 		return -EBADF;
3098 
3099 	if (!(fd_file(dst)->f_mode & FMODE_WRITE))
3100 		return -EBADF;
3101 
3102 	err = mnt_want_write_file(filp);
3103 	if (err)
3104 		return err;
3105 
3106 	err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst),
3107 					range->pos_out, range->len);
3108 
3109 	mnt_drop_write_file(filp);
3110 	return err;
3111 }
3112 
3113 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
3114 {
3115 	struct f2fs_move_range range;
3116 
3117 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
3118 							sizeof(range)))
3119 		return -EFAULT;
3120 	return __f2fs_ioc_move_range(filp, &range);
3121 }
3122 
3123 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
3124 {
3125 	struct inode *inode = file_inode(filp);
3126 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3127 	struct sit_info *sm = SIT_I(sbi);
3128 	unsigned int start_segno = 0, end_segno = 0;
3129 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
3130 	struct f2fs_flush_device range;
3131 	struct f2fs_gc_control gc_control = {
3132 			.init_gc_type = FG_GC,
3133 			.should_migrate_blocks = true,
3134 			.err_gc_skipped = true,
3135 			.nr_free_secs = 0 };
3136 	int ret;
3137 
3138 	if (!capable(CAP_SYS_ADMIN))
3139 		return -EPERM;
3140 
3141 	if (f2fs_readonly(sbi->sb))
3142 		return -EROFS;
3143 
3144 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3145 		return -EINVAL;
3146 
3147 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
3148 							sizeof(range)))
3149 		return -EFAULT;
3150 
3151 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
3152 			__is_large_section(sbi)) {
3153 		f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1",
3154 			  range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi));
3155 		return -EINVAL;
3156 	}
3157 
3158 	ret = mnt_want_write_file(filp);
3159 	if (ret)
3160 		return ret;
3161 
3162 	if (range.dev_num != 0)
3163 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
3164 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
3165 
3166 	start_segno = sm->last_victim[FLUSH_DEVICE];
3167 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
3168 		start_segno = dev_start_segno;
3169 	end_segno = min(start_segno + range.segments, dev_end_segno);
3170 
3171 	while (start_segno < end_segno) {
3172 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
3173 			ret = -EBUSY;
3174 			goto out;
3175 		}
3176 		sm->last_victim[GC_CB] = end_segno + 1;
3177 		sm->last_victim[GC_GREEDY] = end_segno + 1;
3178 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
3179 
3180 		gc_control.victim_segno = start_segno;
3181 		stat_inc_gc_call_count(sbi, FOREGROUND);
3182 		ret = f2fs_gc(sbi, &gc_control);
3183 		if (ret == -EAGAIN)
3184 			ret = 0;
3185 		else if (ret < 0)
3186 			break;
3187 		start_segno++;
3188 	}
3189 out:
3190 	mnt_drop_write_file(filp);
3191 	return ret;
3192 }
3193 
3194 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3195 {
3196 	struct inode *inode = file_inode(filp);
3197 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3198 
3199 	/* Must validate to set it with SQLite behavior in Android. */
3200 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3201 
3202 	return put_user(sb_feature, (u32 __user *)arg);
3203 }
3204 
3205 #ifdef CONFIG_QUOTA
3206 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3207 {
3208 	struct dquot *transfer_to[MAXQUOTAS] = {};
3209 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3210 	struct super_block *sb = sbi->sb;
3211 	int err;
3212 
3213 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3214 	if (IS_ERR(transfer_to[PRJQUOTA]))
3215 		return PTR_ERR(transfer_to[PRJQUOTA]);
3216 
3217 	err = __dquot_transfer(inode, transfer_to);
3218 	if (err)
3219 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3220 	dqput(transfer_to[PRJQUOTA]);
3221 	return err;
3222 }
3223 
3224 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3225 {
3226 	struct f2fs_inode_info *fi = F2FS_I(inode);
3227 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3228 	struct f2fs_inode *ri = NULL;
3229 	kprojid_t kprojid;
3230 	int err;
3231 
3232 	if (!f2fs_sb_has_project_quota(sbi)) {
3233 		if (projid != F2FS_DEF_PROJID)
3234 			return -EOPNOTSUPP;
3235 		else
3236 			return 0;
3237 	}
3238 
3239 	if (!f2fs_has_extra_attr(inode))
3240 		return -EOPNOTSUPP;
3241 
3242 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3243 
3244 	if (projid_eq(kprojid, fi->i_projid))
3245 		return 0;
3246 
3247 	err = -EPERM;
3248 	/* Is it quota file? Do not allow user to mess with it */
3249 	if (IS_NOQUOTA(inode))
3250 		return err;
3251 
3252 	if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3253 		return -EOVERFLOW;
3254 
3255 	err = f2fs_dquot_initialize(inode);
3256 	if (err)
3257 		return err;
3258 
3259 	f2fs_lock_op(sbi);
3260 	err = f2fs_transfer_project_quota(inode, kprojid);
3261 	if (err)
3262 		goto out_unlock;
3263 
3264 	fi->i_projid = kprojid;
3265 	inode_set_ctime_current(inode);
3266 	f2fs_mark_inode_dirty_sync(inode, true);
3267 out_unlock:
3268 	f2fs_unlock_op(sbi);
3269 	return err;
3270 }
3271 #else
3272 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3273 {
3274 	return 0;
3275 }
3276 
3277 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3278 {
3279 	if (projid != F2FS_DEF_PROJID)
3280 		return -EOPNOTSUPP;
3281 	return 0;
3282 }
3283 #endif
3284 
3285 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3286 {
3287 	struct inode *inode = d_inode(dentry);
3288 	struct f2fs_inode_info *fi = F2FS_I(inode);
3289 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3290 
3291 	if (IS_ENCRYPTED(inode))
3292 		fsflags |= FS_ENCRYPT_FL;
3293 	if (IS_VERITY(inode))
3294 		fsflags |= FS_VERITY_FL;
3295 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3296 		fsflags |= FS_INLINE_DATA_FL;
3297 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3298 		fsflags |= FS_NOCOW_FL;
3299 
3300 	fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3301 
3302 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3303 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3304 
3305 	return 0;
3306 }
3307 
3308 int f2fs_fileattr_set(struct mnt_idmap *idmap,
3309 		      struct dentry *dentry, struct fileattr *fa)
3310 {
3311 	struct inode *inode = d_inode(dentry);
3312 	u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3313 	u32 iflags;
3314 	int err;
3315 
3316 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3317 		return -EIO;
3318 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3319 		return -ENOSPC;
3320 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
3321 		return -EOPNOTSUPP;
3322 	fsflags &= F2FS_SETTABLE_FS_FL;
3323 	if (!fa->flags_valid)
3324 		mask &= FS_COMMON_FL;
3325 
3326 	iflags = f2fs_fsflags_to_iflags(fsflags);
3327 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3328 		return -EOPNOTSUPP;
3329 
3330 	err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3331 	if (!err)
3332 		err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3333 
3334 	return err;
3335 }
3336 
3337 int f2fs_pin_file_control(struct inode *inode, bool inc)
3338 {
3339 	struct f2fs_inode_info *fi = F2FS_I(inode);
3340 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3341 
3342 	if (IS_DEVICE_ALIASING(inode))
3343 		return -EINVAL;
3344 
3345 	if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) {
3346 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3347 			  __func__, inode->i_ino, fi->i_gc_failures);
3348 		clear_inode_flag(inode, FI_PIN_FILE);
3349 		return -EAGAIN;
3350 	}
3351 
3352 	/* Use i_gc_failures for normal file as a risk signal. */
3353 	if (inc)
3354 		f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
3355 
3356 	return 0;
3357 }
3358 
3359 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3360 {
3361 	struct inode *inode = file_inode(filp);
3362 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3363 	__u32 pin;
3364 	int ret = 0;
3365 
3366 	if (get_user(pin, (__u32 __user *)arg))
3367 		return -EFAULT;
3368 
3369 	if (!S_ISREG(inode->i_mode))
3370 		return -EINVAL;
3371 
3372 	if (f2fs_readonly(sbi->sb))
3373 		return -EROFS;
3374 
3375 	if (!pin && IS_DEVICE_ALIASING(inode))
3376 		return -EOPNOTSUPP;
3377 
3378 	ret = mnt_want_write_file(filp);
3379 	if (ret)
3380 		return ret;
3381 
3382 	inode_lock(inode);
3383 
3384 	if (f2fs_is_atomic_file(inode)) {
3385 		ret = -EINVAL;
3386 		goto out;
3387 	}
3388 
3389 	if (!pin) {
3390 		clear_inode_flag(inode, FI_PIN_FILE);
3391 		f2fs_i_gc_failures_write(inode, 0);
3392 		goto done;
3393 	} else if (f2fs_is_pinned_file(inode)) {
3394 		goto done;
3395 	}
3396 
3397 	if (F2FS_HAS_BLOCKS(inode)) {
3398 		ret = -EFBIG;
3399 		goto out;
3400 	}
3401 
3402 	/* Let's allow file pinning on zoned device. */
3403 	if (!f2fs_sb_has_blkzoned(sbi) &&
3404 	    f2fs_should_update_outplace(inode, NULL)) {
3405 		ret = -EINVAL;
3406 		goto out;
3407 	}
3408 
3409 	if (f2fs_pin_file_control(inode, false)) {
3410 		ret = -EAGAIN;
3411 		goto out;
3412 	}
3413 
3414 	ret = f2fs_convert_inline_inode(inode);
3415 	if (ret)
3416 		goto out;
3417 
3418 	if (!f2fs_disable_compressed_file(inode)) {
3419 		ret = -EOPNOTSUPP;
3420 		goto out;
3421 	}
3422 
3423 	set_inode_flag(inode, FI_PIN_FILE);
3424 	ret = F2FS_I(inode)->i_gc_failures;
3425 done:
3426 	f2fs_update_time(sbi, REQ_TIME);
3427 out:
3428 	inode_unlock(inode);
3429 	mnt_drop_write_file(filp);
3430 	return ret;
3431 }
3432 
3433 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3434 {
3435 	struct inode *inode = file_inode(filp);
3436 	__u32 pin = 0;
3437 
3438 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3439 		pin = F2FS_I(inode)->i_gc_failures;
3440 	return put_user(pin, (u32 __user *)arg);
3441 }
3442 
3443 static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg)
3444 {
3445 	return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0,
3446 			(u32 __user *)arg);
3447 }
3448 
3449 int f2fs_precache_extents(struct inode *inode)
3450 {
3451 	struct f2fs_inode_info *fi = F2FS_I(inode);
3452 	struct f2fs_map_blocks map;
3453 	pgoff_t m_next_extent;
3454 	loff_t end;
3455 	int err;
3456 
3457 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3458 		return -EOPNOTSUPP;
3459 
3460 	map.m_lblk = 0;
3461 	map.m_pblk = 0;
3462 	map.m_next_pgofs = NULL;
3463 	map.m_next_extent = &m_next_extent;
3464 	map.m_seg_type = NO_CHECK_TYPE;
3465 	map.m_may_create = false;
3466 	end = F2FS_BLK_ALIGN(i_size_read(inode));
3467 
3468 	while (map.m_lblk < end) {
3469 		map.m_len = end - map.m_lblk;
3470 
3471 		f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3472 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
3473 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3474 		if (err || !map.m_len)
3475 			return err;
3476 
3477 		map.m_lblk = m_next_extent;
3478 	}
3479 
3480 	return 0;
3481 }
3482 
3483 static int f2fs_ioc_precache_extents(struct file *filp)
3484 {
3485 	return f2fs_precache_extents(file_inode(filp));
3486 }
3487 
3488 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3489 {
3490 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3491 	__u64 block_count;
3492 
3493 	if (!capable(CAP_SYS_ADMIN))
3494 		return -EPERM;
3495 
3496 	if (f2fs_readonly(sbi->sb))
3497 		return -EROFS;
3498 
3499 	if (copy_from_user(&block_count, (void __user *)arg,
3500 			   sizeof(block_count)))
3501 		return -EFAULT;
3502 
3503 	return f2fs_resize_fs(filp, block_count);
3504 }
3505 
3506 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3507 {
3508 	struct inode *inode = file_inode(filp);
3509 
3510 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3511 
3512 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3513 		f2fs_warn(F2FS_I_SB(inode),
3514 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3515 			  inode->i_ino);
3516 		return -EOPNOTSUPP;
3517 	}
3518 
3519 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3520 }
3521 
3522 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3523 {
3524 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3525 		return -EOPNOTSUPP;
3526 
3527 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3528 }
3529 
3530 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3531 {
3532 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3533 		return -EOPNOTSUPP;
3534 
3535 	return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3536 }
3537 
3538 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3539 {
3540 	struct inode *inode = file_inode(filp);
3541 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3542 	char *vbuf;
3543 	int count;
3544 	int err = 0;
3545 
3546 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3547 	if (!vbuf)
3548 		return -ENOMEM;
3549 
3550 	f2fs_down_read(&sbi->sb_lock);
3551 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3552 			ARRAY_SIZE(sbi->raw_super->volume_name),
3553 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3554 	f2fs_up_read(&sbi->sb_lock);
3555 
3556 	if (copy_to_user((char __user *)arg, vbuf,
3557 				min(FSLABEL_MAX, count)))
3558 		err = -EFAULT;
3559 
3560 	kfree(vbuf);
3561 	return err;
3562 }
3563 
3564 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3565 {
3566 	struct inode *inode = file_inode(filp);
3567 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3568 	char *vbuf;
3569 	int err = 0;
3570 
3571 	if (!capable(CAP_SYS_ADMIN))
3572 		return -EPERM;
3573 
3574 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3575 	if (IS_ERR(vbuf))
3576 		return PTR_ERR(vbuf);
3577 
3578 	err = mnt_want_write_file(filp);
3579 	if (err)
3580 		goto out;
3581 
3582 	f2fs_down_write(&sbi->sb_lock);
3583 
3584 	memset(sbi->raw_super->volume_name, 0,
3585 			sizeof(sbi->raw_super->volume_name));
3586 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3587 			sbi->raw_super->volume_name,
3588 			ARRAY_SIZE(sbi->raw_super->volume_name));
3589 
3590 	err = f2fs_commit_super(sbi, false);
3591 
3592 	f2fs_up_write(&sbi->sb_lock);
3593 
3594 	mnt_drop_write_file(filp);
3595 out:
3596 	kfree(vbuf);
3597 	return err;
3598 }
3599 
3600 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks)
3601 {
3602 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3603 		return -EOPNOTSUPP;
3604 
3605 	if (!f2fs_compressed_file(inode))
3606 		return -EINVAL;
3607 
3608 	*blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3609 
3610 	return 0;
3611 }
3612 
3613 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg)
3614 {
3615 	struct inode *inode = file_inode(filp);
3616 	__u64 blocks;
3617 	int ret;
3618 
3619 	ret = f2fs_get_compress_blocks(inode, &blocks);
3620 	if (ret < 0)
3621 		return ret;
3622 
3623 	return put_user(blocks, (u64 __user *)arg);
3624 }
3625 
3626 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3627 {
3628 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3629 	unsigned int released_blocks = 0;
3630 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3631 	block_t blkaddr;
3632 	int i;
3633 
3634 	for (i = 0; i < count; i++) {
3635 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3636 						dn->ofs_in_node + i);
3637 
3638 		if (!__is_valid_data_blkaddr(blkaddr))
3639 			continue;
3640 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3641 					DATA_GENERIC_ENHANCE)))
3642 			return -EFSCORRUPTED;
3643 	}
3644 
3645 	while (count) {
3646 		int compr_blocks = 0;
3647 
3648 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3649 			blkaddr = f2fs_data_blkaddr(dn);
3650 
3651 			if (i == 0) {
3652 				if (blkaddr == COMPRESS_ADDR)
3653 					continue;
3654 				dn->ofs_in_node += cluster_size;
3655 				goto next;
3656 			}
3657 
3658 			if (__is_valid_data_blkaddr(blkaddr))
3659 				compr_blocks++;
3660 
3661 			if (blkaddr != NEW_ADDR)
3662 				continue;
3663 
3664 			f2fs_set_data_blkaddr(dn, NULL_ADDR);
3665 		}
3666 
3667 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3668 		dec_valid_block_count(sbi, dn->inode,
3669 					cluster_size - compr_blocks);
3670 
3671 		released_blocks += cluster_size - compr_blocks;
3672 next:
3673 		count -= cluster_size;
3674 	}
3675 
3676 	return released_blocks;
3677 }
3678 
3679 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3680 {
3681 	struct inode *inode = file_inode(filp);
3682 	struct f2fs_inode_info *fi = F2FS_I(inode);
3683 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3684 	pgoff_t page_idx = 0, last_idx;
3685 	unsigned int released_blocks = 0;
3686 	int ret;
3687 	int writecount;
3688 
3689 	if (!f2fs_sb_has_compression(sbi))
3690 		return -EOPNOTSUPP;
3691 
3692 	if (f2fs_readonly(sbi->sb))
3693 		return -EROFS;
3694 
3695 	ret = mnt_want_write_file(filp);
3696 	if (ret)
3697 		return ret;
3698 
3699 	f2fs_balance_fs(sbi, true);
3700 
3701 	inode_lock(inode);
3702 
3703 	writecount = atomic_read(&inode->i_writecount);
3704 	if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3705 			(!(filp->f_mode & FMODE_WRITE) && writecount)) {
3706 		ret = -EBUSY;
3707 		goto out;
3708 	}
3709 
3710 	if (!f2fs_compressed_file(inode) ||
3711 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3712 		ret = -EINVAL;
3713 		goto out;
3714 	}
3715 
3716 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3717 	if (ret)
3718 		goto out;
3719 
3720 	if (!atomic_read(&fi->i_compr_blocks)) {
3721 		ret = -EPERM;
3722 		goto out;
3723 	}
3724 
3725 	set_inode_flag(inode, FI_COMPRESS_RELEASED);
3726 	inode_set_ctime_current(inode);
3727 	f2fs_mark_inode_dirty_sync(inode, true);
3728 
3729 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3730 	filemap_invalidate_lock(inode->i_mapping);
3731 
3732 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3733 
3734 	while (page_idx < last_idx) {
3735 		struct dnode_of_data dn;
3736 		pgoff_t end_offset, count;
3737 
3738 		f2fs_lock_op(sbi);
3739 
3740 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3741 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3742 		if (ret) {
3743 			f2fs_unlock_op(sbi);
3744 			if (ret == -ENOENT) {
3745 				page_idx = f2fs_get_next_page_offset(&dn,
3746 								page_idx);
3747 				ret = 0;
3748 				continue;
3749 			}
3750 			break;
3751 		}
3752 
3753 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3754 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3755 		count = round_up(count, fi->i_cluster_size);
3756 
3757 		ret = release_compress_blocks(&dn, count);
3758 
3759 		f2fs_put_dnode(&dn);
3760 
3761 		f2fs_unlock_op(sbi);
3762 
3763 		if (ret < 0)
3764 			break;
3765 
3766 		page_idx += count;
3767 		released_blocks += ret;
3768 	}
3769 
3770 	filemap_invalidate_unlock(inode->i_mapping);
3771 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3772 out:
3773 	if (released_blocks)
3774 		f2fs_update_time(sbi, REQ_TIME);
3775 	inode_unlock(inode);
3776 
3777 	mnt_drop_write_file(filp);
3778 
3779 	if (ret >= 0) {
3780 		ret = put_user(released_blocks, (u64 __user *)arg);
3781 	} else if (released_blocks &&
3782 			atomic_read(&fi->i_compr_blocks)) {
3783 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3784 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3785 			"iblocks=%llu, released=%u, compr_blocks=%u, "
3786 			"run fsck to fix.",
3787 			__func__, inode->i_ino, inode->i_blocks,
3788 			released_blocks,
3789 			atomic_read(&fi->i_compr_blocks));
3790 	}
3791 
3792 	return ret;
3793 }
3794 
3795 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count,
3796 		unsigned int *reserved_blocks)
3797 {
3798 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3799 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3800 	block_t blkaddr;
3801 	int i;
3802 
3803 	for (i = 0; i < count; i++) {
3804 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3805 						dn->ofs_in_node + i);
3806 
3807 		if (!__is_valid_data_blkaddr(blkaddr))
3808 			continue;
3809 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3810 					DATA_GENERIC_ENHANCE)))
3811 			return -EFSCORRUPTED;
3812 	}
3813 
3814 	while (count) {
3815 		int compr_blocks = 0;
3816 		blkcnt_t reserved = 0;
3817 		blkcnt_t to_reserved;
3818 		int ret;
3819 
3820 		for (i = 0; i < cluster_size; i++) {
3821 			blkaddr = data_blkaddr(dn->inode, dn->node_page,
3822 						dn->ofs_in_node + i);
3823 
3824 			if (i == 0) {
3825 				if (blkaddr != COMPRESS_ADDR) {
3826 					dn->ofs_in_node += cluster_size;
3827 					goto next;
3828 				}
3829 				continue;
3830 			}
3831 
3832 			/*
3833 			 * compressed cluster was not released due to it
3834 			 * fails in release_compress_blocks(), so NEW_ADDR
3835 			 * is a possible case.
3836 			 */
3837 			if (blkaddr == NEW_ADDR) {
3838 				reserved++;
3839 				continue;
3840 			}
3841 			if (__is_valid_data_blkaddr(blkaddr)) {
3842 				compr_blocks++;
3843 				continue;
3844 			}
3845 		}
3846 
3847 		to_reserved = cluster_size - compr_blocks - reserved;
3848 
3849 		/* for the case all blocks in cluster were reserved */
3850 		if (reserved && to_reserved == 1) {
3851 			dn->ofs_in_node += cluster_size;
3852 			goto next;
3853 		}
3854 
3855 		ret = inc_valid_block_count(sbi, dn->inode,
3856 						&to_reserved, false);
3857 		if (unlikely(ret))
3858 			return ret;
3859 
3860 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3861 			if (f2fs_data_blkaddr(dn) == NULL_ADDR)
3862 				f2fs_set_data_blkaddr(dn, NEW_ADDR);
3863 		}
3864 
3865 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3866 
3867 		*reserved_blocks += to_reserved;
3868 next:
3869 		count -= cluster_size;
3870 	}
3871 
3872 	return 0;
3873 }
3874 
3875 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3876 {
3877 	struct inode *inode = file_inode(filp);
3878 	struct f2fs_inode_info *fi = F2FS_I(inode);
3879 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3880 	pgoff_t page_idx = 0, last_idx;
3881 	unsigned int reserved_blocks = 0;
3882 	int ret;
3883 
3884 	if (!f2fs_sb_has_compression(sbi))
3885 		return -EOPNOTSUPP;
3886 
3887 	if (f2fs_readonly(sbi->sb))
3888 		return -EROFS;
3889 
3890 	ret = mnt_want_write_file(filp);
3891 	if (ret)
3892 		return ret;
3893 
3894 	f2fs_balance_fs(sbi, true);
3895 
3896 	inode_lock(inode);
3897 
3898 	if (!f2fs_compressed_file(inode) ||
3899 		!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3900 		ret = -EINVAL;
3901 		goto unlock_inode;
3902 	}
3903 
3904 	if (atomic_read(&fi->i_compr_blocks))
3905 		goto unlock_inode;
3906 
3907 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3908 	filemap_invalidate_lock(inode->i_mapping);
3909 
3910 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3911 
3912 	while (page_idx < last_idx) {
3913 		struct dnode_of_data dn;
3914 		pgoff_t end_offset, count;
3915 
3916 		f2fs_lock_op(sbi);
3917 
3918 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3919 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3920 		if (ret) {
3921 			f2fs_unlock_op(sbi);
3922 			if (ret == -ENOENT) {
3923 				page_idx = f2fs_get_next_page_offset(&dn,
3924 								page_idx);
3925 				ret = 0;
3926 				continue;
3927 			}
3928 			break;
3929 		}
3930 
3931 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3932 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3933 		count = round_up(count, fi->i_cluster_size);
3934 
3935 		ret = reserve_compress_blocks(&dn, count, &reserved_blocks);
3936 
3937 		f2fs_put_dnode(&dn);
3938 
3939 		f2fs_unlock_op(sbi);
3940 
3941 		if (ret < 0)
3942 			break;
3943 
3944 		page_idx += count;
3945 	}
3946 
3947 	filemap_invalidate_unlock(inode->i_mapping);
3948 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3949 
3950 	if (!ret) {
3951 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3952 		inode_set_ctime_current(inode);
3953 		f2fs_mark_inode_dirty_sync(inode, true);
3954 	}
3955 unlock_inode:
3956 	if (reserved_blocks)
3957 		f2fs_update_time(sbi, REQ_TIME);
3958 	inode_unlock(inode);
3959 	mnt_drop_write_file(filp);
3960 
3961 	if (!ret) {
3962 		ret = put_user(reserved_blocks, (u64 __user *)arg);
3963 	} else if (reserved_blocks &&
3964 			atomic_read(&fi->i_compr_blocks)) {
3965 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3966 		f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx "
3967 			"iblocks=%llu, reserved=%u, compr_blocks=%u, "
3968 			"run fsck to fix.",
3969 			__func__, inode->i_ino, inode->i_blocks,
3970 			reserved_blocks,
3971 			atomic_read(&fi->i_compr_blocks));
3972 	}
3973 
3974 	return ret;
3975 }
3976 
3977 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3978 		pgoff_t off, block_t block, block_t len, u32 flags)
3979 {
3980 	sector_t sector = SECTOR_FROM_BLOCK(block);
3981 	sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3982 	int ret = 0;
3983 
3984 	if (flags & F2FS_TRIM_FILE_DISCARD) {
3985 		if (bdev_max_secure_erase_sectors(bdev))
3986 			ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3987 					GFP_NOFS);
3988 		else
3989 			ret = blkdev_issue_discard(bdev, sector, nr_sects,
3990 					GFP_NOFS);
3991 	}
3992 
3993 	if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3994 		if (IS_ENCRYPTED(inode))
3995 			ret = fscrypt_zeroout_range(inode, off, block, len);
3996 		else
3997 			ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3998 					GFP_NOFS, 0);
3999 	}
4000 
4001 	return ret;
4002 }
4003 
4004 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
4005 {
4006 	struct inode *inode = file_inode(filp);
4007 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4008 	struct address_space *mapping = inode->i_mapping;
4009 	struct block_device *prev_bdev = NULL;
4010 	struct f2fs_sectrim_range range;
4011 	pgoff_t index, pg_end, prev_index = 0;
4012 	block_t prev_block = 0, len = 0;
4013 	loff_t end_addr;
4014 	bool to_end = false;
4015 	int ret = 0;
4016 
4017 	if (!(filp->f_mode & FMODE_WRITE))
4018 		return -EBADF;
4019 
4020 	if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
4021 				sizeof(range)))
4022 		return -EFAULT;
4023 
4024 	if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
4025 			!S_ISREG(inode->i_mode))
4026 		return -EINVAL;
4027 
4028 	if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
4029 			!f2fs_hw_support_discard(sbi)) ||
4030 			((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
4031 			 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
4032 		return -EOPNOTSUPP;
4033 
4034 	ret = mnt_want_write_file(filp);
4035 	if (ret)
4036 		return ret;
4037 	inode_lock(inode);
4038 
4039 	if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
4040 			range.start >= inode->i_size) {
4041 		ret = -EINVAL;
4042 		goto err;
4043 	}
4044 
4045 	if (range.len == 0)
4046 		goto err;
4047 
4048 	if (inode->i_size - range.start > range.len) {
4049 		end_addr = range.start + range.len;
4050 	} else {
4051 		end_addr = range.len == (u64)-1 ?
4052 			sbi->sb->s_maxbytes : inode->i_size;
4053 		to_end = true;
4054 	}
4055 
4056 	if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
4057 			(!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
4058 		ret = -EINVAL;
4059 		goto err;
4060 	}
4061 
4062 	index = F2FS_BYTES_TO_BLK(range.start);
4063 	pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
4064 
4065 	ret = f2fs_convert_inline_inode(inode);
4066 	if (ret)
4067 		goto err;
4068 
4069 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4070 	filemap_invalidate_lock(mapping);
4071 
4072 	ret = filemap_write_and_wait_range(mapping, range.start,
4073 			to_end ? LLONG_MAX : end_addr - 1);
4074 	if (ret)
4075 		goto out;
4076 
4077 	truncate_inode_pages_range(mapping, range.start,
4078 			to_end ? -1 : end_addr - 1);
4079 
4080 	while (index < pg_end) {
4081 		struct dnode_of_data dn;
4082 		pgoff_t end_offset, count;
4083 		int i;
4084 
4085 		set_new_dnode(&dn, inode, NULL, NULL, 0);
4086 		ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4087 		if (ret) {
4088 			if (ret == -ENOENT) {
4089 				index = f2fs_get_next_page_offset(&dn, index);
4090 				continue;
4091 			}
4092 			goto out;
4093 		}
4094 
4095 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4096 		count = min(end_offset - dn.ofs_in_node, pg_end - index);
4097 		for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
4098 			struct block_device *cur_bdev;
4099 			block_t blkaddr = f2fs_data_blkaddr(&dn);
4100 
4101 			if (!__is_valid_data_blkaddr(blkaddr))
4102 				continue;
4103 
4104 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
4105 						DATA_GENERIC_ENHANCE)) {
4106 				ret = -EFSCORRUPTED;
4107 				f2fs_put_dnode(&dn);
4108 				goto out;
4109 			}
4110 
4111 			cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
4112 			if (f2fs_is_multi_device(sbi)) {
4113 				int di = f2fs_target_device_index(sbi, blkaddr);
4114 
4115 				blkaddr -= FDEV(di).start_blk;
4116 			}
4117 
4118 			if (len) {
4119 				if (prev_bdev == cur_bdev &&
4120 						index == prev_index + len &&
4121 						blkaddr == prev_block + len) {
4122 					len++;
4123 				} else {
4124 					ret = f2fs_secure_erase(prev_bdev,
4125 						inode, prev_index, prev_block,
4126 						len, range.flags);
4127 					if (ret) {
4128 						f2fs_put_dnode(&dn);
4129 						goto out;
4130 					}
4131 
4132 					len = 0;
4133 				}
4134 			}
4135 
4136 			if (!len) {
4137 				prev_bdev = cur_bdev;
4138 				prev_index = index;
4139 				prev_block = blkaddr;
4140 				len = 1;
4141 			}
4142 		}
4143 
4144 		f2fs_put_dnode(&dn);
4145 
4146 		if (fatal_signal_pending(current)) {
4147 			ret = -EINTR;
4148 			goto out;
4149 		}
4150 		cond_resched();
4151 	}
4152 
4153 	if (len)
4154 		ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
4155 				prev_block, len, range.flags);
4156 	f2fs_update_time(sbi, REQ_TIME);
4157 out:
4158 	filemap_invalidate_unlock(mapping);
4159 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4160 err:
4161 	inode_unlock(inode);
4162 	mnt_drop_write_file(filp);
4163 
4164 	return ret;
4165 }
4166 
4167 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
4168 {
4169 	struct inode *inode = file_inode(filp);
4170 	struct f2fs_comp_option option;
4171 
4172 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4173 		return -EOPNOTSUPP;
4174 
4175 	inode_lock_shared(inode);
4176 
4177 	if (!f2fs_compressed_file(inode)) {
4178 		inode_unlock_shared(inode);
4179 		return -ENODATA;
4180 	}
4181 
4182 	option.algorithm = F2FS_I(inode)->i_compress_algorithm;
4183 	option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
4184 
4185 	inode_unlock_shared(inode);
4186 
4187 	if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
4188 				sizeof(option)))
4189 		return -EFAULT;
4190 
4191 	return 0;
4192 }
4193 
4194 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4195 {
4196 	struct inode *inode = file_inode(filp);
4197 	struct f2fs_inode_info *fi = F2FS_I(inode);
4198 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4199 	struct f2fs_comp_option option;
4200 	int ret = 0;
4201 
4202 	if (!f2fs_sb_has_compression(sbi))
4203 		return -EOPNOTSUPP;
4204 
4205 	if (!(filp->f_mode & FMODE_WRITE))
4206 		return -EBADF;
4207 
4208 	if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4209 				sizeof(option)))
4210 		return -EFAULT;
4211 
4212 	if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
4213 		option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
4214 		option.algorithm >= COMPRESS_MAX)
4215 		return -EINVAL;
4216 
4217 	ret = mnt_want_write_file(filp);
4218 	if (ret)
4219 		return ret;
4220 	inode_lock(inode);
4221 
4222 	f2fs_down_write(&F2FS_I(inode)->i_sem);
4223 	if (!f2fs_compressed_file(inode)) {
4224 		ret = -EINVAL;
4225 		goto out;
4226 	}
4227 
4228 	if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4229 		ret = -EBUSY;
4230 		goto out;
4231 	}
4232 
4233 	if (F2FS_HAS_BLOCKS(inode)) {
4234 		ret = -EFBIG;
4235 		goto out;
4236 	}
4237 
4238 	fi->i_compress_algorithm = option.algorithm;
4239 	fi->i_log_cluster_size = option.log_cluster_size;
4240 	fi->i_cluster_size = BIT(option.log_cluster_size);
4241 	/* Set default level */
4242 	if (fi->i_compress_algorithm == COMPRESS_ZSTD)
4243 		fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
4244 	else
4245 		fi->i_compress_level = 0;
4246 	/* Adjust mount option level */
4247 	if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm &&
4248 	    F2FS_OPTION(sbi).compress_level)
4249 		fi->i_compress_level = F2FS_OPTION(sbi).compress_level;
4250 	f2fs_mark_inode_dirty_sync(inode, true);
4251 
4252 	if (!f2fs_is_compress_backend_ready(inode))
4253 		f2fs_warn(sbi, "compression algorithm is successfully set, "
4254 			"but current kernel doesn't support this algorithm.");
4255 out:
4256 	f2fs_up_write(&fi->i_sem);
4257 	inode_unlock(inode);
4258 	mnt_drop_write_file(filp);
4259 
4260 	return ret;
4261 }
4262 
4263 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4264 {
4265 	DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
4266 	struct address_space *mapping = inode->i_mapping;
4267 	struct page *page;
4268 	pgoff_t redirty_idx = page_idx;
4269 	int i, page_len = 0, ret = 0;
4270 
4271 	page_cache_ra_unbounded(&ractl, len, 0);
4272 
4273 	for (i = 0; i < len; i++, page_idx++) {
4274 		page = read_cache_page(mapping, page_idx, NULL, NULL);
4275 		if (IS_ERR(page)) {
4276 			ret = PTR_ERR(page);
4277 			break;
4278 		}
4279 		page_len++;
4280 	}
4281 
4282 	for (i = 0; i < page_len; i++, redirty_idx++) {
4283 		page = find_lock_page(mapping, redirty_idx);
4284 
4285 		/* It will never fail, when page has pinned above */
4286 		f2fs_bug_on(F2FS_I_SB(inode), !page);
4287 
4288 		f2fs_wait_on_page_writeback(page, DATA, true, true);
4289 
4290 		set_page_dirty(page);
4291 		set_page_private_gcing(page);
4292 		f2fs_put_page(page, 1);
4293 		f2fs_put_page(page, 0);
4294 	}
4295 
4296 	return ret;
4297 }
4298 
4299 static int f2fs_ioc_decompress_file(struct file *filp)
4300 {
4301 	struct inode *inode = file_inode(filp);
4302 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4303 	struct f2fs_inode_info *fi = F2FS_I(inode);
4304 	pgoff_t page_idx = 0, last_idx, cluster_idx;
4305 	int ret;
4306 
4307 	if (!f2fs_sb_has_compression(sbi) ||
4308 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4309 		return -EOPNOTSUPP;
4310 
4311 	if (!(filp->f_mode & FMODE_WRITE))
4312 		return -EBADF;
4313 
4314 	f2fs_balance_fs(sbi, true);
4315 
4316 	ret = mnt_want_write_file(filp);
4317 	if (ret)
4318 		return ret;
4319 	inode_lock(inode);
4320 
4321 	if (!f2fs_is_compress_backend_ready(inode)) {
4322 		ret = -EOPNOTSUPP;
4323 		goto out;
4324 	}
4325 
4326 	if (!f2fs_compressed_file(inode) ||
4327 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4328 		ret = -EINVAL;
4329 		goto out;
4330 	}
4331 
4332 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4333 	if (ret)
4334 		goto out;
4335 
4336 	if (!atomic_read(&fi->i_compr_blocks))
4337 		goto out;
4338 
4339 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4340 	last_idx >>= fi->i_log_cluster_size;
4341 
4342 	for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4343 		page_idx = cluster_idx << fi->i_log_cluster_size;
4344 
4345 		if (!f2fs_is_compressed_cluster(inode, page_idx))
4346 			continue;
4347 
4348 		ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4349 		if (ret < 0)
4350 			break;
4351 
4352 		if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4353 			ret = filemap_fdatawrite(inode->i_mapping);
4354 			if (ret < 0)
4355 				break;
4356 		}
4357 
4358 		cond_resched();
4359 		if (fatal_signal_pending(current)) {
4360 			ret = -EINTR;
4361 			break;
4362 		}
4363 	}
4364 
4365 	if (!ret)
4366 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4367 							LLONG_MAX);
4368 
4369 	if (ret)
4370 		f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4371 			  __func__, ret);
4372 	f2fs_update_time(sbi, REQ_TIME);
4373 out:
4374 	inode_unlock(inode);
4375 	mnt_drop_write_file(filp);
4376 
4377 	return ret;
4378 }
4379 
4380 static int f2fs_ioc_compress_file(struct file *filp)
4381 {
4382 	struct inode *inode = file_inode(filp);
4383 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4384 	struct f2fs_inode_info *fi = F2FS_I(inode);
4385 	pgoff_t page_idx = 0, last_idx, cluster_idx;
4386 	int ret;
4387 
4388 	if (!f2fs_sb_has_compression(sbi) ||
4389 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4390 		return -EOPNOTSUPP;
4391 
4392 	if (!(filp->f_mode & FMODE_WRITE))
4393 		return -EBADF;
4394 
4395 	f2fs_balance_fs(sbi, true);
4396 
4397 	ret = mnt_want_write_file(filp);
4398 	if (ret)
4399 		return ret;
4400 	inode_lock(inode);
4401 
4402 	if (!f2fs_is_compress_backend_ready(inode)) {
4403 		ret = -EOPNOTSUPP;
4404 		goto out;
4405 	}
4406 
4407 	if (!f2fs_compressed_file(inode) ||
4408 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4409 		ret = -EINVAL;
4410 		goto out;
4411 	}
4412 
4413 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4414 	if (ret)
4415 		goto out;
4416 
4417 	set_inode_flag(inode, FI_ENABLE_COMPRESS);
4418 
4419 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4420 	last_idx >>= fi->i_log_cluster_size;
4421 
4422 	for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4423 		page_idx = cluster_idx << fi->i_log_cluster_size;
4424 
4425 		if (f2fs_is_sparse_cluster(inode, page_idx))
4426 			continue;
4427 
4428 		ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4429 		if (ret < 0)
4430 			break;
4431 
4432 		if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4433 			ret = filemap_fdatawrite(inode->i_mapping);
4434 			if (ret < 0)
4435 				break;
4436 		}
4437 
4438 		cond_resched();
4439 		if (fatal_signal_pending(current)) {
4440 			ret = -EINTR;
4441 			break;
4442 		}
4443 	}
4444 
4445 	if (!ret)
4446 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4447 							LLONG_MAX);
4448 
4449 	clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4450 
4451 	if (ret)
4452 		f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4453 			  __func__, ret);
4454 	f2fs_update_time(sbi, REQ_TIME);
4455 out:
4456 	inode_unlock(inode);
4457 	mnt_drop_write_file(filp);
4458 
4459 	return ret;
4460 }
4461 
4462 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4463 {
4464 	switch (cmd) {
4465 	case FS_IOC_GETVERSION:
4466 		return f2fs_ioc_getversion(filp, arg);
4467 	case F2FS_IOC_START_ATOMIC_WRITE:
4468 		return f2fs_ioc_start_atomic_write(filp, false);
4469 	case F2FS_IOC_START_ATOMIC_REPLACE:
4470 		return f2fs_ioc_start_atomic_write(filp, true);
4471 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4472 		return f2fs_ioc_commit_atomic_write(filp);
4473 	case F2FS_IOC_ABORT_ATOMIC_WRITE:
4474 		return f2fs_ioc_abort_atomic_write(filp);
4475 	case F2FS_IOC_START_VOLATILE_WRITE:
4476 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4477 		return -EOPNOTSUPP;
4478 	case F2FS_IOC_SHUTDOWN:
4479 		return f2fs_ioc_shutdown(filp, arg);
4480 	case FITRIM:
4481 		return f2fs_ioc_fitrim(filp, arg);
4482 	case FS_IOC_SET_ENCRYPTION_POLICY:
4483 		return f2fs_ioc_set_encryption_policy(filp, arg);
4484 	case FS_IOC_GET_ENCRYPTION_POLICY:
4485 		return f2fs_ioc_get_encryption_policy(filp, arg);
4486 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4487 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4488 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4489 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4490 	case FS_IOC_ADD_ENCRYPTION_KEY:
4491 		return f2fs_ioc_add_encryption_key(filp, arg);
4492 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4493 		return f2fs_ioc_remove_encryption_key(filp, arg);
4494 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4495 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4496 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4497 		return f2fs_ioc_get_encryption_key_status(filp, arg);
4498 	case FS_IOC_GET_ENCRYPTION_NONCE:
4499 		return f2fs_ioc_get_encryption_nonce(filp, arg);
4500 	case F2FS_IOC_GARBAGE_COLLECT:
4501 		return f2fs_ioc_gc(filp, arg);
4502 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4503 		return f2fs_ioc_gc_range(filp, arg);
4504 	case F2FS_IOC_WRITE_CHECKPOINT:
4505 		return f2fs_ioc_write_checkpoint(filp);
4506 	case F2FS_IOC_DEFRAGMENT:
4507 		return f2fs_ioc_defragment(filp, arg);
4508 	case F2FS_IOC_MOVE_RANGE:
4509 		return f2fs_ioc_move_range(filp, arg);
4510 	case F2FS_IOC_FLUSH_DEVICE:
4511 		return f2fs_ioc_flush_device(filp, arg);
4512 	case F2FS_IOC_GET_FEATURES:
4513 		return f2fs_ioc_get_features(filp, arg);
4514 	case F2FS_IOC_GET_PIN_FILE:
4515 		return f2fs_ioc_get_pin_file(filp, arg);
4516 	case F2FS_IOC_SET_PIN_FILE:
4517 		return f2fs_ioc_set_pin_file(filp, arg);
4518 	case F2FS_IOC_PRECACHE_EXTENTS:
4519 		return f2fs_ioc_precache_extents(filp);
4520 	case F2FS_IOC_RESIZE_FS:
4521 		return f2fs_ioc_resize_fs(filp, arg);
4522 	case FS_IOC_ENABLE_VERITY:
4523 		return f2fs_ioc_enable_verity(filp, arg);
4524 	case FS_IOC_MEASURE_VERITY:
4525 		return f2fs_ioc_measure_verity(filp, arg);
4526 	case FS_IOC_READ_VERITY_METADATA:
4527 		return f2fs_ioc_read_verity_metadata(filp, arg);
4528 	case FS_IOC_GETFSLABEL:
4529 		return f2fs_ioc_getfslabel(filp, arg);
4530 	case FS_IOC_SETFSLABEL:
4531 		return f2fs_ioc_setfslabel(filp, arg);
4532 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4533 		return f2fs_ioc_get_compress_blocks(filp, arg);
4534 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4535 		return f2fs_release_compress_blocks(filp, arg);
4536 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4537 		return f2fs_reserve_compress_blocks(filp, arg);
4538 	case F2FS_IOC_SEC_TRIM_FILE:
4539 		return f2fs_sec_trim_file(filp, arg);
4540 	case F2FS_IOC_GET_COMPRESS_OPTION:
4541 		return f2fs_ioc_get_compress_option(filp, arg);
4542 	case F2FS_IOC_SET_COMPRESS_OPTION:
4543 		return f2fs_ioc_set_compress_option(filp, arg);
4544 	case F2FS_IOC_DECOMPRESS_FILE:
4545 		return f2fs_ioc_decompress_file(filp);
4546 	case F2FS_IOC_COMPRESS_FILE:
4547 		return f2fs_ioc_compress_file(filp);
4548 	case F2FS_IOC_GET_DEV_ALIAS_FILE:
4549 		return f2fs_ioc_get_dev_alias_file(filp, arg);
4550 	default:
4551 		return -ENOTTY;
4552 	}
4553 }
4554 
4555 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4556 {
4557 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4558 		return -EIO;
4559 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4560 		return -ENOSPC;
4561 
4562 	return __f2fs_ioctl(filp, cmd, arg);
4563 }
4564 
4565 /*
4566  * Return %true if the given read or write request should use direct I/O, or
4567  * %false if it should use buffered I/O.
4568  */
4569 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4570 				struct iov_iter *iter)
4571 {
4572 	unsigned int align;
4573 
4574 	if (!(iocb->ki_flags & IOCB_DIRECT))
4575 		return false;
4576 
4577 	if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4578 		return false;
4579 
4580 	/*
4581 	 * Direct I/O not aligned to the disk's logical_block_size will be
4582 	 * attempted, but will fail with -EINVAL.
4583 	 *
4584 	 * f2fs additionally requires that direct I/O be aligned to the
4585 	 * filesystem block size, which is often a stricter requirement.
4586 	 * However, f2fs traditionally falls back to buffered I/O on requests
4587 	 * that are logical_block_size-aligned but not fs-block aligned.
4588 	 *
4589 	 * The below logic implements this behavior.
4590 	 */
4591 	align = iocb->ki_pos | iov_iter_alignment(iter);
4592 	if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4593 	    IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4594 		return false;
4595 
4596 	return true;
4597 }
4598 
4599 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4600 				unsigned int flags)
4601 {
4602 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4603 
4604 	dec_page_count(sbi, F2FS_DIO_READ);
4605 	if (error)
4606 		return error;
4607 	f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4608 	return 0;
4609 }
4610 
4611 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4612 	.end_io = f2fs_dio_read_end_io,
4613 };
4614 
4615 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4616 {
4617 	struct file *file = iocb->ki_filp;
4618 	struct inode *inode = file_inode(file);
4619 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4620 	struct f2fs_inode_info *fi = F2FS_I(inode);
4621 	const loff_t pos = iocb->ki_pos;
4622 	const size_t count = iov_iter_count(to);
4623 	struct iomap_dio *dio;
4624 	ssize_t ret;
4625 
4626 	if (count == 0)
4627 		return 0; /* skip atime update */
4628 
4629 	trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4630 
4631 	if (iocb->ki_flags & IOCB_NOWAIT) {
4632 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4633 			ret = -EAGAIN;
4634 			goto out;
4635 		}
4636 	} else {
4637 		f2fs_down_read(&fi->i_gc_rwsem[READ]);
4638 	}
4639 
4640 	/* dio is not compatible w/ atomic file */
4641 	if (f2fs_is_atomic_file(inode)) {
4642 		f2fs_up_read(&fi->i_gc_rwsem[READ]);
4643 		ret = -EOPNOTSUPP;
4644 		goto out;
4645 	}
4646 
4647 	/*
4648 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4649 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4650 	 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4651 	 */
4652 	inc_page_count(sbi, F2FS_DIO_READ);
4653 	dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4654 			     &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4655 	if (IS_ERR_OR_NULL(dio)) {
4656 		ret = PTR_ERR_OR_ZERO(dio);
4657 		if (ret != -EIOCBQUEUED)
4658 			dec_page_count(sbi, F2FS_DIO_READ);
4659 	} else {
4660 		ret = iomap_dio_complete(dio);
4661 	}
4662 
4663 	f2fs_up_read(&fi->i_gc_rwsem[READ]);
4664 
4665 	file_accessed(file);
4666 out:
4667 	trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4668 	return ret;
4669 }
4670 
4671 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count,
4672 				    int rw)
4673 {
4674 	struct inode *inode = file_inode(file);
4675 	char *buf, *path;
4676 
4677 	buf = f2fs_getname(F2FS_I_SB(inode));
4678 	if (!buf)
4679 		return;
4680 	path = dentry_path_raw(file_dentry(file), buf, PATH_MAX);
4681 	if (IS_ERR(path))
4682 		goto free_buf;
4683 	if (rw == WRITE)
4684 		trace_f2fs_datawrite_start(inode, pos, count,
4685 				current->pid, path, current->comm);
4686 	else
4687 		trace_f2fs_dataread_start(inode, pos, count,
4688 				current->pid, path, current->comm);
4689 free_buf:
4690 	f2fs_putname(buf);
4691 }
4692 
4693 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4694 {
4695 	struct inode *inode = file_inode(iocb->ki_filp);
4696 	const loff_t pos = iocb->ki_pos;
4697 	ssize_t ret;
4698 
4699 	if (!f2fs_is_compress_backend_ready(inode))
4700 		return -EOPNOTSUPP;
4701 
4702 	if (trace_f2fs_dataread_start_enabled())
4703 		f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
4704 					iov_iter_count(to), READ);
4705 
4706 	/* In LFS mode, if there is inflight dio, wait for its completion */
4707 	if (f2fs_lfs_mode(F2FS_I_SB(inode)) &&
4708 	    get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE))
4709 		inode_dio_wait(inode);
4710 
4711 	if (f2fs_should_use_dio(inode, iocb, to)) {
4712 		ret = f2fs_dio_read_iter(iocb, to);
4713 	} else {
4714 		ret = filemap_read(iocb, to, 0);
4715 		if (ret > 0)
4716 			f2fs_update_iostat(F2FS_I_SB(inode), inode,
4717 						APP_BUFFERED_READ_IO, ret);
4718 	}
4719 	if (trace_f2fs_dataread_end_enabled())
4720 		trace_f2fs_dataread_end(inode, pos, ret);
4721 	return ret;
4722 }
4723 
4724 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos,
4725 				     struct pipe_inode_info *pipe,
4726 				     size_t len, unsigned int flags)
4727 {
4728 	struct inode *inode = file_inode(in);
4729 	const loff_t pos = *ppos;
4730 	ssize_t ret;
4731 
4732 	if (!f2fs_is_compress_backend_ready(inode))
4733 		return -EOPNOTSUPP;
4734 
4735 	if (trace_f2fs_dataread_start_enabled())
4736 		f2fs_trace_rw_file_path(in, pos, len, READ);
4737 
4738 	ret = filemap_splice_read(in, ppos, pipe, len, flags);
4739 	if (ret > 0)
4740 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
4741 				   APP_BUFFERED_READ_IO, ret);
4742 
4743 	if (trace_f2fs_dataread_end_enabled())
4744 		trace_f2fs_dataread_end(inode, pos, ret);
4745 	return ret;
4746 }
4747 
4748 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4749 {
4750 	struct file *file = iocb->ki_filp;
4751 	struct inode *inode = file_inode(file);
4752 	ssize_t count;
4753 	int err;
4754 
4755 	if (IS_IMMUTABLE(inode))
4756 		return -EPERM;
4757 
4758 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4759 		return -EPERM;
4760 
4761 	count = generic_write_checks(iocb, from);
4762 	if (count <= 0)
4763 		return count;
4764 
4765 	err = file_modified(file);
4766 	if (err)
4767 		return err;
4768 	return count;
4769 }
4770 
4771 /*
4772  * Preallocate blocks for a write request, if it is possible and helpful to do
4773  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4774  * blocks were preallocated, or a negative errno value if something went
4775  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4776  * requested blocks (not just some of them) have been allocated.
4777  */
4778 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4779 				   bool dio)
4780 {
4781 	struct inode *inode = file_inode(iocb->ki_filp);
4782 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4783 	const loff_t pos = iocb->ki_pos;
4784 	const size_t count = iov_iter_count(iter);
4785 	struct f2fs_map_blocks map = {};
4786 	int flag;
4787 	int ret;
4788 
4789 	/* If it will be an out-of-place direct write, don't bother. */
4790 	if (dio && f2fs_lfs_mode(sbi))
4791 		return 0;
4792 	/*
4793 	 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4794 	 * buffered IO, if DIO meets any holes.
4795 	 */
4796 	if (dio && i_size_read(inode) &&
4797 		(F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4798 		return 0;
4799 
4800 	/* No-wait I/O can't allocate blocks. */
4801 	if (iocb->ki_flags & IOCB_NOWAIT)
4802 		return 0;
4803 
4804 	/* If it will be a short write, don't bother. */
4805 	if (fault_in_iov_iter_readable(iter, count))
4806 		return 0;
4807 
4808 	if (f2fs_has_inline_data(inode)) {
4809 		/* If the data will fit inline, don't bother. */
4810 		if (pos + count <= MAX_INLINE_DATA(inode))
4811 			return 0;
4812 		ret = f2fs_convert_inline_inode(inode);
4813 		if (ret)
4814 			return ret;
4815 	}
4816 
4817 	/* Do not preallocate blocks that will be written partially in 4KB. */
4818 	map.m_lblk = F2FS_BLK_ALIGN(pos);
4819 	map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4820 	if (map.m_len > map.m_lblk)
4821 		map.m_len -= map.m_lblk;
4822 	else
4823 		return 0;
4824 
4825 	if (!IS_DEVICE_ALIASING(inode))
4826 		map.m_may_create = true;
4827 	if (dio) {
4828 		map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi,
4829 						inode->i_write_hint);
4830 		flag = F2FS_GET_BLOCK_PRE_DIO;
4831 	} else {
4832 		map.m_seg_type = NO_CHECK_TYPE;
4833 		flag = F2FS_GET_BLOCK_PRE_AIO;
4834 	}
4835 
4836 	ret = f2fs_map_blocks(inode, &map, flag);
4837 	/* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4838 	if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4839 		return ret;
4840 	if (ret == 0)
4841 		set_inode_flag(inode, FI_PREALLOCATED_ALL);
4842 	return map.m_len;
4843 }
4844 
4845 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4846 					struct iov_iter *from)
4847 {
4848 	struct file *file = iocb->ki_filp;
4849 	struct inode *inode = file_inode(file);
4850 	ssize_t ret;
4851 
4852 	if (iocb->ki_flags & IOCB_NOWAIT)
4853 		return -EOPNOTSUPP;
4854 
4855 	ret = generic_perform_write(iocb, from);
4856 
4857 	if (ret > 0) {
4858 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
4859 						APP_BUFFERED_IO, ret);
4860 	}
4861 	return ret;
4862 }
4863 
4864 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4865 				 unsigned int flags)
4866 {
4867 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4868 
4869 	dec_page_count(sbi, F2FS_DIO_WRITE);
4870 	if (error)
4871 		return error;
4872 	f2fs_update_time(sbi, REQ_TIME);
4873 	f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4874 	return 0;
4875 }
4876 
4877 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter,
4878 					struct bio *bio, loff_t file_offset)
4879 {
4880 	struct inode *inode = iter->inode;
4881 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4882 	enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint);
4883 	enum temp_type temp = f2fs_get_segment_temp(sbi, type);
4884 
4885 	bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp);
4886 	submit_bio(bio);
4887 }
4888 
4889 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4890 	.end_io		= f2fs_dio_write_end_io,
4891 	.submit_io	= f2fs_dio_write_submit_io,
4892 };
4893 
4894 static void f2fs_flush_buffered_write(struct address_space *mapping,
4895 				      loff_t start_pos, loff_t end_pos)
4896 {
4897 	int ret;
4898 
4899 	ret = filemap_write_and_wait_range(mapping, start_pos, end_pos);
4900 	if (ret < 0)
4901 		return;
4902 	invalidate_mapping_pages(mapping,
4903 				 start_pos >> PAGE_SHIFT,
4904 				 end_pos >> PAGE_SHIFT);
4905 }
4906 
4907 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4908 				   bool *may_need_sync)
4909 {
4910 	struct file *file = iocb->ki_filp;
4911 	struct inode *inode = file_inode(file);
4912 	struct f2fs_inode_info *fi = F2FS_I(inode);
4913 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4914 	const bool do_opu = f2fs_lfs_mode(sbi);
4915 	const loff_t pos = iocb->ki_pos;
4916 	const ssize_t count = iov_iter_count(from);
4917 	unsigned int dio_flags;
4918 	struct iomap_dio *dio;
4919 	ssize_t ret;
4920 
4921 	trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4922 
4923 	if (iocb->ki_flags & IOCB_NOWAIT) {
4924 		/* f2fs_convert_inline_inode() and block allocation can block */
4925 		if (f2fs_has_inline_data(inode) ||
4926 		    !f2fs_overwrite_io(inode, pos, count)) {
4927 			ret = -EAGAIN;
4928 			goto out;
4929 		}
4930 
4931 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4932 			ret = -EAGAIN;
4933 			goto out;
4934 		}
4935 		if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4936 			f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4937 			ret = -EAGAIN;
4938 			goto out;
4939 		}
4940 	} else {
4941 		ret = f2fs_convert_inline_inode(inode);
4942 		if (ret)
4943 			goto out;
4944 
4945 		f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4946 		if (do_opu)
4947 			f2fs_down_read(&fi->i_gc_rwsem[READ]);
4948 	}
4949 
4950 	/*
4951 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4952 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4953 	 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4954 	 */
4955 	inc_page_count(sbi, F2FS_DIO_WRITE);
4956 	dio_flags = 0;
4957 	if (pos + count > inode->i_size)
4958 		dio_flags |= IOMAP_DIO_FORCE_WAIT;
4959 	dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4960 			     &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4961 	if (IS_ERR_OR_NULL(dio)) {
4962 		ret = PTR_ERR_OR_ZERO(dio);
4963 		if (ret == -ENOTBLK)
4964 			ret = 0;
4965 		if (ret != -EIOCBQUEUED)
4966 			dec_page_count(sbi, F2FS_DIO_WRITE);
4967 	} else {
4968 		ret = iomap_dio_complete(dio);
4969 	}
4970 
4971 	if (do_opu)
4972 		f2fs_up_read(&fi->i_gc_rwsem[READ]);
4973 	f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4974 
4975 	if (ret < 0)
4976 		goto out;
4977 	if (pos + ret > inode->i_size)
4978 		f2fs_i_size_write(inode, pos + ret);
4979 	if (!do_opu)
4980 		set_inode_flag(inode, FI_UPDATE_WRITE);
4981 
4982 	if (iov_iter_count(from)) {
4983 		ssize_t ret2;
4984 		loff_t bufio_start_pos = iocb->ki_pos;
4985 
4986 		/*
4987 		 * The direct write was partial, so we need to fall back to a
4988 		 * buffered write for the remainder.
4989 		 */
4990 
4991 		ret2 = f2fs_buffered_write_iter(iocb, from);
4992 		if (iov_iter_count(from))
4993 			f2fs_write_failed(inode, iocb->ki_pos);
4994 		if (ret2 < 0)
4995 			goto out;
4996 
4997 		/*
4998 		 * Ensure that the pagecache pages are written to disk and
4999 		 * invalidated to preserve the expected O_DIRECT semantics.
5000 		 */
5001 		if (ret2 > 0) {
5002 			loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
5003 
5004 			ret += ret2;
5005 
5006 			f2fs_flush_buffered_write(file->f_mapping,
5007 						  bufio_start_pos,
5008 						  bufio_end_pos);
5009 		}
5010 	} else {
5011 		/* iomap_dio_rw() already handled the generic_write_sync(). */
5012 		*may_need_sync = false;
5013 	}
5014 out:
5015 	trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
5016 	return ret;
5017 }
5018 
5019 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
5020 {
5021 	struct inode *inode = file_inode(iocb->ki_filp);
5022 	const loff_t orig_pos = iocb->ki_pos;
5023 	const size_t orig_count = iov_iter_count(from);
5024 	loff_t target_size;
5025 	bool dio;
5026 	bool may_need_sync = true;
5027 	int preallocated;
5028 	const loff_t pos = iocb->ki_pos;
5029 	const ssize_t count = iov_iter_count(from);
5030 	ssize_t ret;
5031 
5032 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
5033 		ret = -EIO;
5034 		goto out;
5035 	}
5036 
5037 	if (!f2fs_is_compress_backend_ready(inode)) {
5038 		ret = -EOPNOTSUPP;
5039 		goto out;
5040 	}
5041 
5042 	if (iocb->ki_flags & IOCB_NOWAIT) {
5043 		if (!inode_trylock(inode)) {
5044 			ret = -EAGAIN;
5045 			goto out;
5046 		}
5047 	} else {
5048 		inode_lock(inode);
5049 	}
5050 
5051 	if (f2fs_is_pinned_file(inode) &&
5052 	    !f2fs_overwrite_io(inode, pos, count)) {
5053 		ret = -EIO;
5054 		goto out_unlock;
5055 	}
5056 
5057 	ret = f2fs_write_checks(iocb, from);
5058 	if (ret <= 0)
5059 		goto out_unlock;
5060 
5061 	/* Determine whether we will do a direct write or a buffered write. */
5062 	dio = f2fs_should_use_dio(inode, iocb, from);
5063 
5064 	/* dio is not compatible w/ atomic write */
5065 	if (dio && f2fs_is_atomic_file(inode)) {
5066 		ret = -EOPNOTSUPP;
5067 		goto out_unlock;
5068 	}
5069 
5070 	/* Possibly preallocate the blocks for the write. */
5071 	target_size = iocb->ki_pos + iov_iter_count(from);
5072 	preallocated = f2fs_preallocate_blocks(iocb, from, dio);
5073 	if (preallocated < 0) {
5074 		ret = preallocated;
5075 	} else {
5076 		if (trace_f2fs_datawrite_start_enabled())
5077 			f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
5078 						orig_count, WRITE);
5079 
5080 		/* Do the actual write. */
5081 		ret = dio ?
5082 			f2fs_dio_write_iter(iocb, from, &may_need_sync) :
5083 			f2fs_buffered_write_iter(iocb, from);
5084 
5085 		if (trace_f2fs_datawrite_end_enabled())
5086 			trace_f2fs_datawrite_end(inode, orig_pos, ret);
5087 	}
5088 
5089 	/* Don't leave any preallocated blocks around past i_size. */
5090 	if (preallocated && i_size_read(inode) < target_size) {
5091 		f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5092 		filemap_invalidate_lock(inode->i_mapping);
5093 		if (!f2fs_truncate(inode))
5094 			file_dont_truncate(inode);
5095 		filemap_invalidate_unlock(inode->i_mapping);
5096 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5097 	} else {
5098 		file_dont_truncate(inode);
5099 	}
5100 
5101 	clear_inode_flag(inode, FI_PREALLOCATED_ALL);
5102 out_unlock:
5103 	inode_unlock(inode);
5104 out:
5105 	trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
5106 
5107 	if (ret > 0 && may_need_sync)
5108 		ret = generic_write_sync(iocb, ret);
5109 
5110 	/* If buffered IO was forced, flush and drop the data from
5111 	 * the page cache to preserve O_DIRECT semantics
5112 	 */
5113 	if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
5114 		f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
5115 					  orig_pos,
5116 					  orig_pos + ret - 1);
5117 
5118 	return ret;
5119 }
5120 
5121 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
5122 		int advice)
5123 {
5124 	struct address_space *mapping;
5125 	struct backing_dev_info *bdi;
5126 	struct inode *inode = file_inode(filp);
5127 	int err;
5128 
5129 	if (advice == POSIX_FADV_SEQUENTIAL) {
5130 		if (S_ISFIFO(inode->i_mode))
5131 			return -ESPIPE;
5132 
5133 		mapping = filp->f_mapping;
5134 		if (!mapping || len < 0)
5135 			return -EINVAL;
5136 
5137 		bdi = inode_to_bdi(mapping->host);
5138 		filp->f_ra.ra_pages = bdi->ra_pages *
5139 			F2FS_I_SB(inode)->seq_file_ra_mul;
5140 		spin_lock(&filp->f_lock);
5141 		filp->f_mode &= ~FMODE_RANDOM;
5142 		spin_unlock(&filp->f_lock);
5143 		return 0;
5144 	} else if (advice == POSIX_FADV_WILLNEED && offset == 0) {
5145 		/* Load extent cache at the first readahead. */
5146 		f2fs_precache_extents(inode);
5147 	}
5148 
5149 	err = generic_fadvise(filp, offset, len, advice);
5150 	if (!err && advice == POSIX_FADV_DONTNEED &&
5151 		test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
5152 		f2fs_compressed_file(inode))
5153 		f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
5154 
5155 	return err;
5156 }
5157 
5158 #ifdef CONFIG_COMPAT
5159 struct compat_f2fs_gc_range {
5160 	u32 sync;
5161 	compat_u64 start;
5162 	compat_u64 len;
5163 };
5164 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE	_IOW(F2FS_IOCTL_MAGIC, 11,\
5165 						struct compat_f2fs_gc_range)
5166 
5167 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
5168 {
5169 	struct compat_f2fs_gc_range __user *urange;
5170 	struct f2fs_gc_range range;
5171 	int err;
5172 
5173 	urange = compat_ptr(arg);
5174 	err = get_user(range.sync, &urange->sync);
5175 	err |= get_user(range.start, &urange->start);
5176 	err |= get_user(range.len, &urange->len);
5177 	if (err)
5178 		return -EFAULT;
5179 
5180 	return __f2fs_ioc_gc_range(file, &range);
5181 }
5182 
5183 struct compat_f2fs_move_range {
5184 	u32 dst_fd;
5185 	compat_u64 pos_in;
5186 	compat_u64 pos_out;
5187 	compat_u64 len;
5188 };
5189 #define F2FS_IOC32_MOVE_RANGE		_IOWR(F2FS_IOCTL_MAGIC, 9,	\
5190 					struct compat_f2fs_move_range)
5191 
5192 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
5193 {
5194 	struct compat_f2fs_move_range __user *urange;
5195 	struct f2fs_move_range range;
5196 	int err;
5197 
5198 	urange = compat_ptr(arg);
5199 	err = get_user(range.dst_fd, &urange->dst_fd);
5200 	err |= get_user(range.pos_in, &urange->pos_in);
5201 	err |= get_user(range.pos_out, &urange->pos_out);
5202 	err |= get_user(range.len, &urange->len);
5203 	if (err)
5204 		return -EFAULT;
5205 
5206 	return __f2fs_ioc_move_range(file, &range);
5207 }
5208 
5209 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5210 {
5211 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
5212 		return -EIO;
5213 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
5214 		return -ENOSPC;
5215 
5216 	switch (cmd) {
5217 	case FS_IOC32_GETVERSION:
5218 		cmd = FS_IOC_GETVERSION;
5219 		break;
5220 	case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
5221 		return f2fs_compat_ioc_gc_range(file, arg);
5222 	case F2FS_IOC32_MOVE_RANGE:
5223 		return f2fs_compat_ioc_move_range(file, arg);
5224 	case F2FS_IOC_START_ATOMIC_WRITE:
5225 	case F2FS_IOC_START_ATOMIC_REPLACE:
5226 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
5227 	case F2FS_IOC_START_VOLATILE_WRITE:
5228 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
5229 	case F2FS_IOC_ABORT_ATOMIC_WRITE:
5230 	case F2FS_IOC_SHUTDOWN:
5231 	case FITRIM:
5232 	case FS_IOC_SET_ENCRYPTION_POLICY:
5233 	case FS_IOC_GET_ENCRYPTION_PWSALT:
5234 	case FS_IOC_GET_ENCRYPTION_POLICY:
5235 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
5236 	case FS_IOC_ADD_ENCRYPTION_KEY:
5237 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
5238 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
5239 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
5240 	case FS_IOC_GET_ENCRYPTION_NONCE:
5241 	case F2FS_IOC_GARBAGE_COLLECT:
5242 	case F2FS_IOC_WRITE_CHECKPOINT:
5243 	case F2FS_IOC_DEFRAGMENT:
5244 	case F2FS_IOC_FLUSH_DEVICE:
5245 	case F2FS_IOC_GET_FEATURES:
5246 	case F2FS_IOC_GET_PIN_FILE:
5247 	case F2FS_IOC_SET_PIN_FILE:
5248 	case F2FS_IOC_PRECACHE_EXTENTS:
5249 	case F2FS_IOC_RESIZE_FS:
5250 	case FS_IOC_ENABLE_VERITY:
5251 	case FS_IOC_MEASURE_VERITY:
5252 	case FS_IOC_READ_VERITY_METADATA:
5253 	case FS_IOC_GETFSLABEL:
5254 	case FS_IOC_SETFSLABEL:
5255 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
5256 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
5257 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
5258 	case F2FS_IOC_SEC_TRIM_FILE:
5259 	case F2FS_IOC_GET_COMPRESS_OPTION:
5260 	case F2FS_IOC_SET_COMPRESS_OPTION:
5261 	case F2FS_IOC_DECOMPRESS_FILE:
5262 	case F2FS_IOC_COMPRESS_FILE:
5263 	case F2FS_IOC_GET_DEV_ALIAS_FILE:
5264 		break;
5265 	default:
5266 		return -ENOIOCTLCMD;
5267 	}
5268 	return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5269 }
5270 #endif
5271 
5272 const struct file_operations f2fs_file_operations = {
5273 	.llseek		= f2fs_llseek,
5274 	.read_iter	= f2fs_file_read_iter,
5275 	.write_iter	= f2fs_file_write_iter,
5276 	.iopoll		= iocb_bio_iopoll,
5277 	.open		= f2fs_file_open,
5278 	.release	= f2fs_release_file,
5279 	.mmap		= f2fs_file_mmap,
5280 	.flush		= f2fs_file_flush,
5281 	.fsync		= f2fs_sync_file,
5282 	.fallocate	= f2fs_fallocate,
5283 	.unlocked_ioctl	= f2fs_ioctl,
5284 #ifdef CONFIG_COMPAT
5285 	.compat_ioctl	= f2fs_compat_ioctl,
5286 #endif
5287 	.splice_read	= f2fs_file_splice_read,
5288 	.splice_write	= iter_file_splice_write,
5289 	.fadvise	= f2fs_file_fadvise,
5290 	.fop_flags	= FOP_BUFFER_RASYNC,
5291 };
5292