xref: /linux/fs/f2fs/file.c (revision 663a917475530feff868a4f2bda286ea4171f420)
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 
625 	addr = get_dnode_addr(dn->inode, dn->node_page) + ofs;
626 
627 	/* Assumption: truncation starts with cluster */
628 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
629 		block_t blkaddr = le32_to_cpu(*addr);
630 
631 		if (f2fs_compressed_file(dn->inode) &&
632 					!(cluster_index & (cluster_size - 1))) {
633 			if (compressed_cluster)
634 				f2fs_i_compr_blocks_update(dn->inode,
635 							valid_blocks, false);
636 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
637 			valid_blocks = 0;
638 		}
639 
640 		if (blkaddr == NULL_ADDR)
641 			continue;
642 
643 		f2fs_set_data_blkaddr(dn, NULL_ADDR);
644 
645 		if (__is_valid_data_blkaddr(blkaddr)) {
646 			if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE))
647 				continue;
648 			if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr,
649 						DATA_GENERIC_ENHANCE))
650 				continue;
651 			if (compressed_cluster)
652 				valid_blocks++;
653 		}
654 
655 		f2fs_invalidate_blocks(sbi, blkaddr);
656 
657 		if (!released || blkaddr != COMPRESS_ADDR)
658 			nr_free++;
659 	}
660 
661 	if (compressed_cluster)
662 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
663 
664 	if (nr_free) {
665 		pgoff_t fofs;
666 		/*
667 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
668 		 * we will invalidate all blkaddr in the whole range.
669 		 */
670 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
671 							dn->inode) + ofs;
672 		f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
673 		f2fs_update_age_extent_cache_range(dn, fofs, len);
674 		dec_valid_block_count(sbi, dn->inode, nr_free);
675 	}
676 	dn->ofs_in_node = ofs;
677 
678 	f2fs_update_time(sbi, REQ_TIME);
679 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
680 					 dn->ofs_in_node, nr_free);
681 }
682 
683 static int truncate_partial_data_page(struct inode *inode, u64 from,
684 								bool cache_only)
685 {
686 	loff_t offset = from & (PAGE_SIZE - 1);
687 	pgoff_t index = from >> PAGE_SHIFT;
688 	struct address_space *mapping = inode->i_mapping;
689 	struct page *page;
690 
691 	if (!offset && !cache_only)
692 		return 0;
693 
694 	if (cache_only) {
695 		page = find_lock_page(mapping, index);
696 		if (page && PageUptodate(page))
697 			goto truncate_out;
698 		f2fs_put_page(page, 1);
699 		return 0;
700 	}
701 
702 	page = f2fs_get_lock_data_page(inode, index, true);
703 	if (IS_ERR(page))
704 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
705 truncate_out:
706 	f2fs_wait_on_page_writeback(page, DATA, true, true);
707 	zero_user(page, offset, PAGE_SIZE - offset);
708 
709 	/* An encrypted inode should have a key and truncate the last page. */
710 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
711 	if (!cache_only)
712 		set_page_dirty(page);
713 	f2fs_put_page(page, 1);
714 	return 0;
715 }
716 
717 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
718 {
719 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
720 	struct dnode_of_data dn;
721 	pgoff_t free_from;
722 	int count = 0, err = 0;
723 	struct page *ipage;
724 	bool truncate_page = false;
725 
726 	trace_f2fs_truncate_blocks_enter(inode, from);
727 
728 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
729 
730 	if (free_from >= max_file_blocks(inode))
731 		goto free_partial;
732 
733 	if (lock)
734 		f2fs_lock_op(sbi);
735 
736 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
737 	if (IS_ERR(ipage)) {
738 		err = PTR_ERR(ipage);
739 		goto out;
740 	}
741 
742 	if (f2fs_has_inline_data(inode)) {
743 		f2fs_truncate_inline_inode(inode, ipage, from);
744 		f2fs_put_page(ipage, 1);
745 		truncate_page = true;
746 		goto out;
747 	}
748 
749 	set_new_dnode(&dn, inode, ipage, NULL, 0);
750 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
751 	if (err) {
752 		if (err == -ENOENT)
753 			goto free_next;
754 		goto out;
755 	}
756 
757 	count = ADDRS_PER_PAGE(dn.node_page, inode);
758 
759 	count -= dn.ofs_in_node;
760 	f2fs_bug_on(sbi, count < 0);
761 
762 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
763 		f2fs_truncate_data_blocks_range(&dn, count);
764 		free_from += count;
765 	}
766 
767 	f2fs_put_dnode(&dn);
768 free_next:
769 	err = f2fs_truncate_inode_blocks(inode, free_from);
770 out:
771 	if (lock)
772 		f2fs_unlock_op(sbi);
773 free_partial:
774 	/* lastly zero out the first data page */
775 	if (!err)
776 		err = truncate_partial_data_page(inode, from, truncate_page);
777 
778 	trace_f2fs_truncate_blocks_exit(inode, err);
779 	return err;
780 }
781 
782 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
783 {
784 	u64 free_from = from;
785 	int err;
786 
787 #ifdef CONFIG_F2FS_FS_COMPRESSION
788 	/*
789 	 * for compressed file, only support cluster size
790 	 * aligned truncation.
791 	 */
792 	if (f2fs_compressed_file(inode))
793 		free_from = round_up(from,
794 				F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
795 #endif
796 
797 	err = f2fs_do_truncate_blocks(inode, free_from, lock);
798 	if (err)
799 		return err;
800 
801 #ifdef CONFIG_F2FS_FS_COMPRESSION
802 	/*
803 	 * For compressed file, after release compress blocks, don't allow write
804 	 * direct, but we should allow write direct after truncate to zero.
805 	 */
806 	if (f2fs_compressed_file(inode) && !free_from
807 			&& is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
808 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
809 
810 	if (from != free_from) {
811 		err = f2fs_truncate_partial_cluster(inode, from, lock);
812 		if (err)
813 			return err;
814 	}
815 #endif
816 
817 	return 0;
818 }
819 
820 int f2fs_truncate(struct inode *inode)
821 {
822 	int err;
823 
824 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
825 		return -EIO;
826 
827 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
828 				S_ISLNK(inode->i_mode)))
829 		return 0;
830 
831 	trace_f2fs_truncate(inode);
832 
833 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
834 		return -EIO;
835 
836 	err = f2fs_dquot_initialize(inode);
837 	if (err)
838 		return err;
839 
840 	/* we should check inline_data size */
841 	if (!f2fs_may_inline_data(inode)) {
842 		err = f2fs_convert_inline_inode(inode);
843 		if (err)
844 			return err;
845 	}
846 
847 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
848 	if (err)
849 		return err;
850 
851 	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
852 	f2fs_mark_inode_dirty_sync(inode, false);
853 	return 0;
854 }
855 
856 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
857 {
858 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
859 
860 	if (!fscrypt_dio_supported(inode))
861 		return true;
862 	if (fsverity_active(inode))
863 		return true;
864 	if (f2fs_compressed_file(inode))
865 		return true;
866 	if (f2fs_has_inline_data(inode))
867 		return true;
868 
869 	/* disallow direct IO if any of devices has unaligned blksize */
870 	if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
871 		return true;
872 	/*
873 	 * for blkzoned device, fallback direct IO to buffered IO, so
874 	 * all IOs can be serialized by log-structured write.
875 	 */
876 	if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) &&
877 	    !f2fs_is_pinned_file(inode))
878 		return true;
879 	if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
880 		return true;
881 
882 	return false;
883 }
884 
885 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path,
886 		 struct kstat *stat, u32 request_mask, unsigned int query_flags)
887 {
888 	struct inode *inode = d_inode(path->dentry);
889 	struct f2fs_inode_info *fi = F2FS_I(inode);
890 	struct f2fs_inode *ri = NULL;
891 	unsigned int flags;
892 
893 	if (f2fs_has_extra_attr(inode) &&
894 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
895 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
896 		stat->result_mask |= STATX_BTIME;
897 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
898 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
899 	}
900 
901 	/*
902 	 * Return the DIO alignment restrictions if requested.  We only return
903 	 * this information when requested, since on encrypted files it might
904 	 * take a fair bit of work to get if the file wasn't opened recently.
905 	 *
906 	 * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
907 	 * cannot represent that, so in that case we report no DIO support.
908 	 */
909 	if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
910 		unsigned int bsize = i_blocksize(inode);
911 
912 		stat->result_mask |= STATX_DIOALIGN;
913 		if (!f2fs_force_buffered_io(inode, WRITE)) {
914 			stat->dio_mem_align = bsize;
915 			stat->dio_offset_align = bsize;
916 		}
917 	}
918 
919 	flags = fi->i_flags;
920 	if (flags & F2FS_COMPR_FL)
921 		stat->attributes |= STATX_ATTR_COMPRESSED;
922 	if (flags & F2FS_APPEND_FL)
923 		stat->attributes |= STATX_ATTR_APPEND;
924 	if (IS_ENCRYPTED(inode))
925 		stat->attributes |= STATX_ATTR_ENCRYPTED;
926 	if (flags & F2FS_IMMUTABLE_FL)
927 		stat->attributes |= STATX_ATTR_IMMUTABLE;
928 	if (flags & F2FS_NODUMP_FL)
929 		stat->attributes |= STATX_ATTR_NODUMP;
930 	if (IS_VERITY(inode))
931 		stat->attributes |= STATX_ATTR_VERITY;
932 
933 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
934 				  STATX_ATTR_APPEND |
935 				  STATX_ATTR_ENCRYPTED |
936 				  STATX_ATTR_IMMUTABLE |
937 				  STATX_ATTR_NODUMP |
938 				  STATX_ATTR_VERITY);
939 
940 	generic_fillattr(idmap, request_mask, inode, stat);
941 
942 	/* we need to show initial sectors used for inline_data/dentries */
943 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
944 					f2fs_has_inline_dentry(inode))
945 		stat->blocks += (stat->size + 511) >> 9;
946 
947 	return 0;
948 }
949 
950 #ifdef CONFIG_F2FS_FS_POSIX_ACL
951 static void __setattr_copy(struct mnt_idmap *idmap,
952 			   struct inode *inode, const struct iattr *attr)
953 {
954 	unsigned int ia_valid = attr->ia_valid;
955 
956 	i_uid_update(idmap, attr, inode);
957 	i_gid_update(idmap, attr, inode);
958 	if (ia_valid & ATTR_ATIME)
959 		inode_set_atime_to_ts(inode, attr->ia_atime);
960 	if (ia_valid & ATTR_MTIME)
961 		inode_set_mtime_to_ts(inode, attr->ia_mtime);
962 	if (ia_valid & ATTR_CTIME)
963 		inode_set_ctime_to_ts(inode, attr->ia_ctime);
964 	if (ia_valid & ATTR_MODE) {
965 		umode_t mode = attr->ia_mode;
966 
967 		if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode)))
968 			mode &= ~S_ISGID;
969 		set_acl_inode(inode, mode);
970 	}
971 }
972 #else
973 #define __setattr_copy setattr_copy
974 #endif
975 
976 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
977 		 struct iattr *attr)
978 {
979 	struct inode *inode = d_inode(dentry);
980 	struct f2fs_inode_info *fi = F2FS_I(inode);
981 	int err;
982 
983 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
984 		return -EIO;
985 
986 	if (unlikely(IS_IMMUTABLE(inode)))
987 		return -EPERM;
988 
989 	if (unlikely(IS_APPEND(inode) &&
990 			(attr->ia_valid & (ATTR_MODE | ATTR_UID |
991 				  ATTR_GID | ATTR_TIMES_SET))))
992 		return -EPERM;
993 
994 	if ((attr->ia_valid & ATTR_SIZE)) {
995 		if (!f2fs_is_compress_backend_ready(inode))
996 			return -EOPNOTSUPP;
997 		if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) &&
998 			!IS_ALIGNED(attr->ia_size,
999 			F2FS_BLK_TO_BYTES(fi->i_cluster_size)))
1000 			return -EINVAL;
1001 	}
1002 
1003 	err = setattr_prepare(idmap, dentry, attr);
1004 	if (err)
1005 		return err;
1006 
1007 	err = fscrypt_prepare_setattr(dentry, attr);
1008 	if (err)
1009 		return err;
1010 
1011 	err = fsverity_prepare_setattr(dentry, attr);
1012 	if (err)
1013 		return err;
1014 
1015 	if (is_quota_modification(idmap, inode, attr)) {
1016 		err = f2fs_dquot_initialize(inode);
1017 		if (err)
1018 			return err;
1019 	}
1020 	if (i_uid_needs_update(idmap, attr, inode) ||
1021 	    i_gid_needs_update(idmap, attr, inode)) {
1022 		f2fs_lock_op(F2FS_I_SB(inode));
1023 		err = dquot_transfer(idmap, inode, attr);
1024 		if (err) {
1025 			set_sbi_flag(F2FS_I_SB(inode),
1026 					SBI_QUOTA_NEED_REPAIR);
1027 			f2fs_unlock_op(F2FS_I_SB(inode));
1028 			return err;
1029 		}
1030 		/*
1031 		 * update uid/gid under lock_op(), so that dquot and inode can
1032 		 * be updated atomically.
1033 		 */
1034 		i_uid_update(idmap, attr, inode);
1035 		i_gid_update(idmap, attr, inode);
1036 		f2fs_mark_inode_dirty_sync(inode, true);
1037 		f2fs_unlock_op(F2FS_I_SB(inode));
1038 	}
1039 
1040 	if (attr->ia_valid & ATTR_SIZE) {
1041 		loff_t old_size = i_size_read(inode);
1042 
1043 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
1044 			/*
1045 			 * should convert inline inode before i_size_write to
1046 			 * keep smaller than inline_data size with inline flag.
1047 			 */
1048 			err = f2fs_convert_inline_inode(inode);
1049 			if (err)
1050 				return err;
1051 		}
1052 
1053 		/*
1054 		 * wait for inflight dio, blocks should be removed after
1055 		 * IO completion.
1056 		 */
1057 		if (attr->ia_size < old_size)
1058 			inode_dio_wait(inode);
1059 
1060 		f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
1061 		filemap_invalidate_lock(inode->i_mapping);
1062 
1063 		truncate_setsize(inode, attr->ia_size);
1064 
1065 		if (attr->ia_size <= old_size)
1066 			err = f2fs_truncate(inode);
1067 		/*
1068 		 * do not trim all blocks after i_size if target size is
1069 		 * larger than i_size.
1070 		 */
1071 		filemap_invalidate_unlock(inode->i_mapping);
1072 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
1073 		if (err)
1074 			return err;
1075 
1076 		spin_lock(&fi->i_size_lock);
1077 		inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1078 		fi->last_disk_size = i_size_read(inode);
1079 		spin_unlock(&fi->i_size_lock);
1080 	}
1081 
1082 	__setattr_copy(idmap, inode, attr);
1083 
1084 	if (attr->ia_valid & ATTR_MODE) {
1085 		err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
1086 
1087 		if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1088 			if (!err)
1089 				inode->i_mode = fi->i_acl_mode;
1090 			clear_inode_flag(inode, FI_ACL_MODE);
1091 		}
1092 	}
1093 
1094 	/* file size may changed here */
1095 	f2fs_mark_inode_dirty_sync(inode, true);
1096 
1097 	/* inode change will produce dirty node pages flushed by checkpoint */
1098 	f2fs_balance_fs(F2FS_I_SB(inode), true);
1099 
1100 	return err;
1101 }
1102 
1103 const struct inode_operations f2fs_file_inode_operations = {
1104 	.getattr	= f2fs_getattr,
1105 	.setattr	= f2fs_setattr,
1106 	.get_inode_acl	= f2fs_get_acl,
1107 	.set_acl	= f2fs_set_acl,
1108 	.listxattr	= f2fs_listxattr,
1109 	.fiemap		= f2fs_fiemap,
1110 	.fileattr_get	= f2fs_fileattr_get,
1111 	.fileattr_set	= f2fs_fileattr_set,
1112 };
1113 
1114 static int fill_zero(struct inode *inode, pgoff_t index,
1115 					loff_t start, loff_t len)
1116 {
1117 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1118 	struct page *page;
1119 
1120 	if (!len)
1121 		return 0;
1122 
1123 	f2fs_balance_fs(sbi, true);
1124 
1125 	f2fs_lock_op(sbi);
1126 	page = f2fs_get_new_data_page(inode, NULL, index, false);
1127 	f2fs_unlock_op(sbi);
1128 
1129 	if (IS_ERR(page))
1130 		return PTR_ERR(page);
1131 
1132 	f2fs_wait_on_page_writeback(page, DATA, true, true);
1133 	zero_user(page, start, len);
1134 	set_page_dirty(page);
1135 	f2fs_put_page(page, 1);
1136 	return 0;
1137 }
1138 
1139 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1140 {
1141 	int err;
1142 
1143 	while (pg_start < pg_end) {
1144 		struct dnode_of_data dn;
1145 		pgoff_t end_offset, count;
1146 
1147 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1148 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1149 		if (err) {
1150 			if (err == -ENOENT) {
1151 				pg_start = f2fs_get_next_page_offset(&dn,
1152 								pg_start);
1153 				continue;
1154 			}
1155 			return err;
1156 		}
1157 
1158 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1159 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1160 
1161 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1162 
1163 		f2fs_truncate_data_blocks_range(&dn, count);
1164 		f2fs_put_dnode(&dn);
1165 
1166 		pg_start += count;
1167 	}
1168 	return 0;
1169 }
1170 
1171 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
1172 {
1173 	pgoff_t pg_start, pg_end;
1174 	loff_t off_start, off_end;
1175 	int ret;
1176 
1177 	ret = f2fs_convert_inline_inode(inode);
1178 	if (ret)
1179 		return ret;
1180 
1181 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1182 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1183 
1184 	off_start = offset & (PAGE_SIZE - 1);
1185 	off_end = (offset + len) & (PAGE_SIZE - 1);
1186 
1187 	if (pg_start == pg_end) {
1188 		ret = fill_zero(inode, pg_start, off_start,
1189 						off_end - off_start);
1190 		if (ret)
1191 			return ret;
1192 	} else {
1193 		if (off_start) {
1194 			ret = fill_zero(inode, pg_start++, off_start,
1195 						PAGE_SIZE - off_start);
1196 			if (ret)
1197 				return ret;
1198 		}
1199 		if (off_end) {
1200 			ret = fill_zero(inode, pg_end, 0, off_end);
1201 			if (ret)
1202 				return ret;
1203 		}
1204 
1205 		if (pg_start < pg_end) {
1206 			loff_t blk_start, blk_end;
1207 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1208 
1209 			f2fs_balance_fs(sbi, true);
1210 
1211 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1212 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1213 
1214 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1215 			filemap_invalidate_lock(inode->i_mapping);
1216 
1217 			truncate_pagecache_range(inode, blk_start, blk_end - 1);
1218 
1219 			f2fs_lock_op(sbi);
1220 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1221 			f2fs_unlock_op(sbi);
1222 
1223 			filemap_invalidate_unlock(inode->i_mapping);
1224 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1225 		}
1226 	}
1227 
1228 	return ret;
1229 }
1230 
1231 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1232 				int *do_replace, pgoff_t off, pgoff_t len)
1233 {
1234 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1235 	struct dnode_of_data dn;
1236 	int ret, done, i;
1237 
1238 next_dnode:
1239 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1240 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1241 	if (ret && ret != -ENOENT) {
1242 		return ret;
1243 	} else if (ret == -ENOENT) {
1244 		if (dn.max_level == 0)
1245 			return -ENOENT;
1246 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1247 						dn.ofs_in_node, len);
1248 		blkaddr += done;
1249 		do_replace += done;
1250 		goto next;
1251 	}
1252 
1253 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1254 							dn.ofs_in_node, len);
1255 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1256 		*blkaddr = f2fs_data_blkaddr(&dn);
1257 
1258 		if (__is_valid_data_blkaddr(*blkaddr) &&
1259 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1260 					DATA_GENERIC_ENHANCE)) {
1261 			f2fs_put_dnode(&dn);
1262 			return -EFSCORRUPTED;
1263 		}
1264 
1265 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1266 
1267 			if (f2fs_lfs_mode(sbi)) {
1268 				f2fs_put_dnode(&dn);
1269 				return -EOPNOTSUPP;
1270 			}
1271 
1272 			/* do not invalidate this block address */
1273 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1274 			*do_replace = 1;
1275 		}
1276 	}
1277 	f2fs_put_dnode(&dn);
1278 next:
1279 	len -= done;
1280 	off += done;
1281 	if (len)
1282 		goto next_dnode;
1283 	return 0;
1284 }
1285 
1286 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1287 				int *do_replace, pgoff_t off, int len)
1288 {
1289 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1290 	struct dnode_of_data dn;
1291 	int ret, i;
1292 
1293 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1294 		if (*do_replace == 0)
1295 			continue;
1296 
1297 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1298 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1299 		if (ret) {
1300 			dec_valid_block_count(sbi, inode, 1);
1301 			f2fs_invalidate_blocks(sbi, *blkaddr);
1302 		} else {
1303 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1304 		}
1305 		f2fs_put_dnode(&dn);
1306 	}
1307 	return 0;
1308 }
1309 
1310 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1311 			block_t *blkaddr, int *do_replace,
1312 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1313 {
1314 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1315 	pgoff_t i = 0;
1316 	int ret;
1317 
1318 	while (i < len) {
1319 		if (blkaddr[i] == NULL_ADDR && !full) {
1320 			i++;
1321 			continue;
1322 		}
1323 
1324 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1325 			struct dnode_of_data dn;
1326 			struct node_info ni;
1327 			size_t new_size;
1328 			pgoff_t ilen;
1329 
1330 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1331 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1332 			if (ret)
1333 				return ret;
1334 
1335 			ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1336 			if (ret) {
1337 				f2fs_put_dnode(&dn);
1338 				return ret;
1339 			}
1340 
1341 			ilen = min((pgoff_t)
1342 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1343 						dn.ofs_in_node, len - i);
1344 			do {
1345 				dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1346 				f2fs_truncate_data_blocks_range(&dn, 1);
1347 
1348 				if (do_replace[i]) {
1349 					f2fs_i_blocks_write(src_inode,
1350 							1, false, false);
1351 					f2fs_i_blocks_write(dst_inode,
1352 							1, true, false);
1353 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1354 					blkaddr[i], ni.version, true, false);
1355 
1356 					do_replace[i] = 0;
1357 				}
1358 				dn.ofs_in_node++;
1359 				i++;
1360 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1361 				if (dst_inode->i_size < new_size)
1362 					f2fs_i_size_write(dst_inode, new_size);
1363 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1364 
1365 			f2fs_put_dnode(&dn);
1366 		} else {
1367 			struct page *psrc, *pdst;
1368 
1369 			psrc = f2fs_get_lock_data_page(src_inode,
1370 							src + i, true);
1371 			if (IS_ERR(psrc))
1372 				return PTR_ERR(psrc);
1373 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1374 								true);
1375 			if (IS_ERR(pdst)) {
1376 				f2fs_put_page(psrc, 1);
1377 				return PTR_ERR(pdst);
1378 			}
1379 
1380 			f2fs_wait_on_page_writeback(pdst, DATA, true, true);
1381 
1382 			memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1383 			set_page_dirty(pdst);
1384 			set_page_private_gcing(pdst);
1385 			f2fs_put_page(pdst, 1);
1386 			f2fs_put_page(psrc, 1);
1387 
1388 			ret = f2fs_truncate_hole(src_inode,
1389 						src + i, src + i + 1);
1390 			if (ret)
1391 				return ret;
1392 			i++;
1393 		}
1394 	}
1395 	return 0;
1396 }
1397 
1398 static int __exchange_data_block(struct inode *src_inode,
1399 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1400 			pgoff_t len, bool full)
1401 {
1402 	block_t *src_blkaddr;
1403 	int *do_replace;
1404 	pgoff_t olen;
1405 	int ret;
1406 
1407 	while (len) {
1408 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1409 
1410 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1411 					array_size(olen, sizeof(block_t)),
1412 					GFP_NOFS);
1413 		if (!src_blkaddr)
1414 			return -ENOMEM;
1415 
1416 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1417 					array_size(olen, sizeof(int)),
1418 					GFP_NOFS);
1419 		if (!do_replace) {
1420 			kvfree(src_blkaddr);
1421 			return -ENOMEM;
1422 		}
1423 
1424 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1425 					do_replace, src, olen);
1426 		if (ret)
1427 			goto roll_back;
1428 
1429 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1430 					do_replace, src, dst, olen, full);
1431 		if (ret)
1432 			goto roll_back;
1433 
1434 		src += olen;
1435 		dst += olen;
1436 		len -= olen;
1437 
1438 		kvfree(src_blkaddr);
1439 		kvfree(do_replace);
1440 	}
1441 	return 0;
1442 
1443 roll_back:
1444 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1445 	kvfree(src_blkaddr);
1446 	kvfree(do_replace);
1447 	return ret;
1448 }
1449 
1450 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1451 {
1452 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1453 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1454 	pgoff_t start = offset >> PAGE_SHIFT;
1455 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1456 	int ret;
1457 
1458 	f2fs_balance_fs(sbi, true);
1459 
1460 	/* avoid gc operation during block exchange */
1461 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1462 	filemap_invalidate_lock(inode->i_mapping);
1463 
1464 	f2fs_lock_op(sbi);
1465 	f2fs_drop_extent_tree(inode);
1466 	truncate_pagecache(inode, offset);
1467 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1468 	f2fs_unlock_op(sbi);
1469 
1470 	filemap_invalidate_unlock(inode->i_mapping);
1471 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1472 	return ret;
1473 }
1474 
1475 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1476 {
1477 	loff_t new_size;
1478 	int ret;
1479 
1480 	if (offset + len >= i_size_read(inode))
1481 		return -EINVAL;
1482 
1483 	/* collapse range should be aligned to block size of f2fs. */
1484 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1485 		return -EINVAL;
1486 
1487 	ret = f2fs_convert_inline_inode(inode);
1488 	if (ret)
1489 		return ret;
1490 
1491 	/* write out all dirty pages from offset */
1492 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1493 	if (ret)
1494 		return ret;
1495 
1496 	ret = f2fs_do_collapse(inode, offset, len);
1497 	if (ret)
1498 		return ret;
1499 
1500 	/* write out all moved pages, if possible */
1501 	filemap_invalidate_lock(inode->i_mapping);
1502 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1503 	truncate_pagecache(inode, offset);
1504 
1505 	new_size = i_size_read(inode) - len;
1506 	ret = f2fs_truncate_blocks(inode, new_size, true);
1507 	filemap_invalidate_unlock(inode->i_mapping);
1508 	if (!ret)
1509 		f2fs_i_size_write(inode, new_size);
1510 	return ret;
1511 }
1512 
1513 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1514 								pgoff_t end)
1515 {
1516 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1517 	pgoff_t index = start;
1518 	unsigned int ofs_in_node = dn->ofs_in_node;
1519 	blkcnt_t count = 0;
1520 	int ret;
1521 
1522 	for (; index < end; index++, dn->ofs_in_node++) {
1523 		if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1524 			count++;
1525 	}
1526 
1527 	dn->ofs_in_node = ofs_in_node;
1528 	ret = f2fs_reserve_new_blocks(dn, count);
1529 	if (ret)
1530 		return ret;
1531 
1532 	dn->ofs_in_node = ofs_in_node;
1533 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1534 		dn->data_blkaddr = f2fs_data_blkaddr(dn);
1535 		/*
1536 		 * f2fs_reserve_new_blocks will not guarantee entire block
1537 		 * allocation.
1538 		 */
1539 		if (dn->data_blkaddr == NULL_ADDR) {
1540 			ret = -ENOSPC;
1541 			break;
1542 		}
1543 
1544 		if (dn->data_blkaddr == NEW_ADDR)
1545 			continue;
1546 
1547 		if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1548 					DATA_GENERIC_ENHANCE)) {
1549 			ret = -EFSCORRUPTED;
1550 			break;
1551 		}
1552 
1553 		f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1554 		f2fs_set_data_blkaddr(dn, NEW_ADDR);
1555 	}
1556 
1557 	f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
1558 	f2fs_update_age_extent_cache_range(dn, start, index - start);
1559 
1560 	return ret;
1561 }
1562 
1563 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1564 								int mode)
1565 {
1566 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1567 	struct address_space *mapping = inode->i_mapping;
1568 	pgoff_t index, pg_start, pg_end;
1569 	loff_t new_size = i_size_read(inode);
1570 	loff_t off_start, off_end;
1571 	int ret = 0;
1572 
1573 	ret = inode_newsize_ok(inode, (len + offset));
1574 	if (ret)
1575 		return ret;
1576 
1577 	ret = f2fs_convert_inline_inode(inode);
1578 	if (ret)
1579 		return ret;
1580 
1581 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1582 	if (ret)
1583 		return ret;
1584 
1585 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1586 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1587 
1588 	off_start = offset & (PAGE_SIZE - 1);
1589 	off_end = (offset + len) & (PAGE_SIZE - 1);
1590 
1591 	if (pg_start == pg_end) {
1592 		ret = fill_zero(inode, pg_start, off_start,
1593 						off_end - off_start);
1594 		if (ret)
1595 			return ret;
1596 
1597 		new_size = max_t(loff_t, new_size, offset + len);
1598 	} else {
1599 		if (off_start) {
1600 			ret = fill_zero(inode, pg_start++, off_start,
1601 						PAGE_SIZE - off_start);
1602 			if (ret)
1603 				return ret;
1604 
1605 			new_size = max_t(loff_t, new_size,
1606 					(loff_t)pg_start << PAGE_SHIFT);
1607 		}
1608 
1609 		for (index = pg_start; index < pg_end;) {
1610 			struct dnode_of_data dn;
1611 			unsigned int end_offset;
1612 			pgoff_t end;
1613 
1614 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1615 			filemap_invalidate_lock(mapping);
1616 
1617 			truncate_pagecache_range(inode,
1618 				(loff_t)index << PAGE_SHIFT,
1619 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1620 
1621 			f2fs_lock_op(sbi);
1622 
1623 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1624 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1625 			if (ret) {
1626 				f2fs_unlock_op(sbi);
1627 				filemap_invalidate_unlock(mapping);
1628 				f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1629 				goto out;
1630 			}
1631 
1632 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1633 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1634 
1635 			ret = f2fs_do_zero_range(&dn, index, end);
1636 			f2fs_put_dnode(&dn);
1637 
1638 			f2fs_unlock_op(sbi);
1639 			filemap_invalidate_unlock(mapping);
1640 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1641 
1642 			f2fs_balance_fs(sbi, dn.node_changed);
1643 
1644 			if (ret)
1645 				goto out;
1646 
1647 			index = end;
1648 			new_size = max_t(loff_t, new_size,
1649 					(loff_t)index << PAGE_SHIFT);
1650 		}
1651 
1652 		if (off_end) {
1653 			ret = fill_zero(inode, pg_end, 0, off_end);
1654 			if (ret)
1655 				goto out;
1656 
1657 			new_size = max_t(loff_t, new_size, offset + len);
1658 		}
1659 	}
1660 
1661 out:
1662 	if (new_size > i_size_read(inode)) {
1663 		if (mode & FALLOC_FL_KEEP_SIZE)
1664 			file_set_keep_isize(inode);
1665 		else
1666 			f2fs_i_size_write(inode, new_size);
1667 	}
1668 	return ret;
1669 }
1670 
1671 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1672 {
1673 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1674 	struct address_space *mapping = inode->i_mapping;
1675 	pgoff_t nr, pg_start, pg_end, delta, idx;
1676 	loff_t new_size;
1677 	int ret = 0;
1678 
1679 	new_size = i_size_read(inode) + len;
1680 	ret = inode_newsize_ok(inode, new_size);
1681 	if (ret)
1682 		return ret;
1683 
1684 	if (offset >= i_size_read(inode))
1685 		return -EINVAL;
1686 
1687 	/* insert range should be aligned to block size of f2fs. */
1688 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1689 		return -EINVAL;
1690 
1691 	ret = f2fs_convert_inline_inode(inode);
1692 	if (ret)
1693 		return ret;
1694 
1695 	f2fs_balance_fs(sbi, true);
1696 
1697 	filemap_invalidate_lock(mapping);
1698 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1699 	filemap_invalidate_unlock(mapping);
1700 	if (ret)
1701 		return ret;
1702 
1703 	/* write out all dirty pages from offset */
1704 	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1705 	if (ret)
1706 		return ret;
1707 
1708 	pg_start = offset >> PAGE_SHIFT;
1709 	pg_end = (offset + len) >> PAGE_SHIFT;
1710 	delta = pg_end - pg_start;
1711 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1712 
1713 	/* avoid gc operation during block exchange */
1714 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1715 	filemap_invalidate_lock(mapping);
1716 	truncate_pagecache(inode, offset);
1717 
1718 	while (!ret && idx > pg_start) {
1719 		nr = idx - pg_start;
1720 		if (nr > delta)
1721 			nr = delta;
1722 		idx -= nr;
1723 
1724 		f2fs_lock_op(sbi);
1725 		f2fs_drop_extent_tree(inode);
1726 
1727 		ret = __exchange_data_block(inode, inode, idx,
1728 					idx + delta, nr, false);
1729 		f2fs_unlock_op(sbi);
1730 	}
1731 	filemap_invalidate_unlock(mapping);
1732 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1733 	if (ret)
1734 		return ret;
1735 
1736 	/* write out all moved pages, if possible */
1737 	filemap_invalidate_lock(mapping);
1738 	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1739 	truncate_pagecache(inode, offset);
1740 	filemap_invalidate_unlock(mapping);
1741 
1742 	if (!ret)
1743 		f2fs_i_size_write(inode, new_size);
1744 	return ret;
1745 }
1746 
1747 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
1748 					loff_t len, int mode)
1749 {
1750 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1751 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1752 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1753 			.m_may_create = true };
1754 	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1755 			.init_gc_type = FG_GC,
1756 			.should_migrate_blocks = false,
1757 			.err_gc_skipped = true,
1758 			.nr_free_secs = 0 };
1759 	pgoff_t pg_start, pg_end;
1760 	loff_t new_size;
1761 	loff_t off_end;
1762 	block_t expanded = 0;
1763 	int err;
1764 
1765 	err = inode_newsize_ok(inode, (len + offset));
1766 	if (err)
1767 		return err;
1768 
1769 	err = f2fs_convert_inline_inode(inode);
1770 	if (err)
1771 		return err;
1772 
1773 	f2fs_balance_fs(sbi, true);
1774 
1775 	pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1776 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1777 	off_end = (offset + len) & (PAGE_SIZE - 1);
1778 
1779 	map.m_lblk = pg_start;
1780 	map.m_len = pg_end - pg_start;
1781 	if (off_end)
1782 		map.m_len++;
1783 
1784 	if (!map.m_len)
1785 		return 0;
1786 
1787 	if (f2fs_is_pinned_file(inode)) {
1788 		block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1789 		block_t sec_len = roundup(map.m_len, sec_blks);
1790 
1791 		map.m_len = sec_blks;
1792 next_alloc:
1793 		if (has_not_enough_free_secs(sbi, 0,
1794 			GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1795 			f2fs_down_write(&sbi->gc_lock);
1796 			stat_inc_gc_call_count(sbi, FOREGROUND);
1797 			err = f2fs_gc(sbi, &gc_control);
1798 			if (err && err != -ENODATA)
1799 				goto out_err;
1800 		}
1801 
1802 		f2fs_down_write(&sbi->pin_sem);
1803 
1804 		err = f2fs_allocate_pinning_section(sbi);
1805 		if (err) {
1806 			f2fs_up_write(&sbi->pin_sem);
1807 			goto out_err;
1808 		}
1809 
1810 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1811 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
1812 		file_dont_truncate(inode);
1813 
1814 		f2fs_up_write(&sbi->pin_sem);
1815 
1816 		expanded += map.m_len;
1817 		sec_len -= map.m_len;
1818 		map.m_lblk += map.m_len;
1819 		if (!err && sec_len)
1820 			goto next_alloc;
1821 
1822 		map.m_len = expanded;
1823 	} else {
1824 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
1825 		expanded = map.m_len;
1826 	}
1827 out_err:
1828 	if (err) {
1829 		pgoff_t last_off;
1830 
1831 		if (!expanded)
1832 			return err;
1833 
1834 		last_off = pg_start + expanded - 1;
1835 
1836 		/* update new size to the failed position */
1837 		new_size = (last_off == pg_end) ? offset + len :
1838 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1839 	} else {
1840 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1841 	}
1842 
1843 	if (new_size > i_size_read(inode)) {
1844 		if (mode & FALLOC_FL_KEEP_SIZE)
1845 			file_set_keep_isize(inode);
1846 		else
1847 			f2fs_i_size_write(inode, new_size);
1848 	}
1849 
1850 	return err;
1851 }
1852 
1853 static long f2fs_fallocate(struct file *file, int mode,
1854 				loff_t offset, loff_t len)
1855 {
1856 	struct inode *inode = file_inode(file);
1857 	long ret = 0;
1858 
1859 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1860 		return -EIO;
1861 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1862 		return -ENOSPC;
1863 	if (!f2fs_is_compress_backend_ready(inode))
1864 		return -EOPNOTSUPP;
1865 
1866 	/* f2fs only support ->fallocate for regular file */
1867 	if (!S_ISREG(inode->i_mode))
1868 		return -EINVAL;
1869 
1870 	if (IS_ENCRYPTED(inode) &&
1871 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1872 		return -EOPNOTSUPP;
1873 
1874 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1875 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1876 			FALLOC_FL_INSERT_RANGE))
1877 		return -EOPNOTSUPP;
1878 
1879 	inode_lock(inode);
1880 
1881 	/*
1882 	 * Pinned file should not support partial truncation since the block
1883 	 * can be used by applications.
1884 	 */
1885 	if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1886 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1887 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) {
1888 		ret = -EOPNOTSUPP;
1889 		goto out;
1890 	}
1891 
1892 	ret = file_modified(file);
1893 	if (ret)
1894 		goto out;
1895 
1896 	/*
1897 	 * wait for inflight dio, blocks should be removed after IO
1898 	 * completion.
1899 	 */
1900 	inode_dio_wait(inode);
1901 
1902 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1903 		if (offset >= inode->i_size)
1904 			goto out;
1905 
1906 		ret = f2fs_punch_hole(inode, offset, len);
1907 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1908 		ret = f2fs_collapse_range(inode, offset, len);
1909 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1910 		ret = f2fs_zero_range(inode, offset, len, mode);
1911 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1912 		ret = f2fs_insert_range(inode, offset, len);
1913 	} else {
1914 		ret = f2fs_expand_inode_data(inode, offset, len, mode);
1915 	}
1916 
1917 	if (!ret) {
1918 		inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1919 		f2fs_mark_inode_dirty_sync(inode, false);
1920 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1921 	}
1922 
1923 out:
1924 	inode_unlock(inode);
1925 
1926 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1927 	return ret;
1928 }
1929 
1930 static int f2fs_release_file(struct inode *inode, struct file *filp)
1931 {
1932 	/*
1933 	 * f2fs_release_file is called at every close calls. So we should
1934 	 * not drop any inmemory pages by close called by other process.
1935 	 */
1936 	if (!(filp->f_mode & FMODE_WRITE) ||
1937 			atomic_read(&inode->i_writecount) != 1)
1938 		return 0;
1939 
1940 	inode_lock(inode);
1941 	f2fs_abort_atomic_write(inode, true);
1942 	inode_unlock(inode);
1943 
1944 	return 0;
1945 }
1946 
1947 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1948 {
1949 	struct inode *inode = file_inode(file);
1950 
1951 	/*
1952 	 * If the process doing a transaction is crashed, we should do
1953 	 * roll-back. Otherwise, other reader/write can see corrupted database
1954 	 * until all the writers close its file. Since this should be done
1955 	 * before dropping file lock, it needs to do in ->flush.
1956 	 */
1957 	if (F2FS_I(inode)->atomic_write_task == current &&
1958 				(current->flags & PF_EXITING)) {
1959 		inode_lock(inode);
1960 		f2fs_abort_atomic_write(inode, true);
1961 		inode_unlock(inode);
1962 	}
1963 
1964 	return 0;
1965 }
1966 
1967 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1968 {
1969 	struct f2fs_inode_info *fi = F2FS_I(inode);
1970 	u32 masked_flags = fi->i_flags & mask;
1971 
1972 	/* mask can be shrunk by flags_valid selector */
1973 	iflags &= mask;
1974 
1975 	/* Is it quota file? Do not allow user to mess with it */
1976 	if (IS_NOQUOTA(inode))
1977 		return -EPERM;
1978 
1979 	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1980 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1981 			return -EOPNOTSUPP;
1982 		if (!f2fs_empty_dir(inode))
1983 			return -ENOTEMPTY;
1984 	}
1985 
1986 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1987 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1988 			return -EOPNOTSUPP;
1989 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1990 			return -EINVAL;
1991 	}
1992 
1993 	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1994 		if (masked_flags & F2FS_COMPR_FL) {
1995 			if (!f2fs_disable_compressed_file(inode))
1996 				return -EINVAL;
1997 		} else {
1998 			/* try to convert inline_data to support compression */
1999 			int err = f2fs_convert_inline_inode(inode);
2000 			if (err)
2001 				return err;
2002 
2003 			f2fs_down_write(&fi->i_sem);
2004 			if (!f2fs_may_compress(inode) ||
2005 					(S_ISREG(inode->i_mode) &&
2006 					F2FS_HAS_BLOCKS(inode))) {
2007 				f2fs_up_write(&fi->i_sem);
2008 				return -EINVAL;
2009 			}
2010 			err = set_compress_context(inode);
2011 			f2fs_up_write(&fi->i_sem);
2012 
2013 			if (err)
2014 				return err;
2015 		}
2016 	}
2017 
2018 	fi->i_flags = iflags | (fi->i_flags & ~mask);
2019 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
2020 					(fi->i_flags & F2FS_NOCOMP_FL));
2021 
2022 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
2023 		set_inode_flag(inode, FI_PROJ_INHERIT);
2024 	else
2025 		clear_inode_flag(inode, FI_PROJ_INHERIT);
2026 
2027 	inode_set_ctime_current(inode);
2028 	f2fs_set_inode_flags(inode);
2029 	f2fs_mark_inode_dirty_sync(inode, true);
2030 	return 0;
2031 }
2032 
2033 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
2034 
2035 /*
2036  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
2037  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
2038  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
2039  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
2040  *
2041  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
2042  * FS_IOC_FSSETXATTR is done by the VFS.
2043  */
2044 
2045 static const struct {
2046 	u32 iflag;
2047 	u32 fsflag;
2048 } f2fs_fsflags_map[] = {
2049 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
2050 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
2051 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
2052 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
2053 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
2054 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
2055 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
2056 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
2057 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
2058 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
2059 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
2060 };
2061 
2062 #define F2FS_GETTABLE_FS_FL (		\
2063 		FS_COMPR_FL |		\
2064 		FS_SYNC_FL |		\
2065 		FS_IMMUTABLE_FL |	\
2066 		FS_APPEND_FL |		\
2067 		FS_NODUMP_FL |		\
2068 		FS_NOATIME_FL |		\
2069 		FS_NOCOMP_FL |		\
2070 		FS_INDEX_FL |		\
2071 		FS_DIRSYNC_FL |		\
2072 		FS_PROJINHERIT_FL |	\
2073 		FS_ENCRYPT_FL |		\
2074 		FS_INLINE_DATA_FL |	\
2075 		FS_NOCOW_FL |		\
2076 		FS_VERITY_FL |		\
2077 		FS_CASEFOLD_FL)
2078 
2079 #define F2FS_SETTABLE_FS_FL (		\
2080 		FS_COMPR_FL |		\
2081 		FS_SYNC_FL |		\
2082 		FS_IMMUTABLE_FL |	\
2083 		FS_APPEND_FL |		\
2084 		FS_NODUMP_FL |		\
2085 		FS_NOATIME_FL |		\
2086 		FS_NOCOMP_FL |		\
2087 		FS_DIRSYNC_FL |		\
2088 		FS_PROJINHERIT_FL |	\
2089 		FS_CASEFOLD_FL)
2090 
2091 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2092 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2093 {
2094 	u32 fsflags = 0;
2095 	int i;
2096 
2097 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2098 		if (iflags & f2fs_fsflags_map[i].iflag)
2099 			fsflags |= f2fs_fsflags_map[i].fsflag;
2100 
2101 	return fsflags;
2102 }
2103 
2104 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2105 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2106 {
2107 	u32 iflags = 0;
2108 	int i;
2109 
2110 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2111 		if (fsflags & f2fs_fsflags_map[i].fsflag)
2112 			iflags |= f2fs_fsflags_map[i].iflag;
2113 
2114 	return iflags;
2115 }
2116 
2117 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2118 {
2119 	struct inode *inode = file_inode(filp);
2120 
2121 	return put_user(inode->i_generation, (int __user *)arg);
2122 }
2123 
2124 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2125 {
2126 	struct inode *inode = file_inode(filp);
2127 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2128 	struct f2fs_inode_info *fi = F2FS_I(inode);
2129 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2130 	loff_t isize;
2131 	int ret;
2132 
2133 	if (!(filp->f_mode & FMODE_WRITE))
2134 		return -EBADF;
2135 
2136 	if (!inode_owner_or_capable(idmap, inode))
2137 		return -EACCES;
2138 
2139 	if (!S_ISREG(inode->i_mode))
2140 		return -EINVAL;
2141 
2142 	if (filp->f_flags & O_DIRECT)
2143 		return -EINVAL;
2144 
2145 	ret = mnt_want_write_file(filp);
2146 	if (ret)
2147 		return ret;
2148 
2149 	inode_lock(inode);
2150 
2151 	if (!f2fs_disable_compressed_file(inode) ||
2152 			f2fs_is_pinned_file(inode)) {
2153 		ret = -EINVAL;
2154 		goto out;
2155 	}
2156 
2157 	if (f2fs_is_atomic_file(inode))
2158 		goto out;
2159 
2160 	ret = f2fs_convert_inline_inode(inode);
2161 	if (ret)
2162 		goto out;
2163 
2164 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2165 	f2fs_down_write(&fi->i_gc_rwsem[READ]);
2166 
2167 	/*
2168 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2169 	 * f2fs_is_atomic_file.
2170 	 */
2171 	if (get_dirty_pages(inode))
2172 		f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2173 			  inode->i_ino, get_dirty_pages(inode));
2174 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2175 	if (ret)
2176 		goto out_unlock;
2177 
2178 	/* Check if the inode already has a COW inode */
2179 	if (fi->cow_inode == NULL) {
2180 		/* Create a COW inode for atomic write */
2181 		struct dentry *dentry = file_dentry(filp);
2182 		struct inode *dir = d_inode(dentry->d_parent);
2183 
2184 		ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode);
2185 		if (ret)
2186 			goto out_unlock;
2187 
2188 		set_inode_flag(fi->cow_inode, FI_COW_FILE);
2189 		clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2190 
2191 		/* Set the COW inode's atomic_inode to the atomic inode */
2192 		F2FS_I(fi->cow_inode)->atomic_inode = inode;
2193 	} else {
2194 		/* Reuse the already created COW inode */
2195 		f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode));
2196 
2197 		invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1);
2198 
2199 		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2200 		if (ret)
2201 			goto out_unlock;
2202 	}
2203 
2204 	f2fs_write_inode(inode, NULL);
2205 
2206 	stat_inc_atomic_inode(inode);
2207 
2208 	set_inode_flag(inode, FI_ATOMIC_FILE);
2209 
2210 	isize = i_size_read(inode);
2211 	fi->original_i_size = isize;
2212 	if (truncate) {
2213 		set_inode_flag(inode, FI_ATOMIC_REPLACE);
2214 		truncate_inode_pages_final(inode->i_mapping);
2215 		f2fs_i_size_write(inode, 0);
2216 		isize = 0;
2217 	}
2218 	f2fs_i_size_write(fi->cow_inode, isize);
2219 
2220 out_unlock:
2221 	f2fs_up_write(&fi->i_gc_rwsem[READ]);
2222 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2223 	if (ret)
2224 		goto out;
2225 
2226 	f2fs_update_time(sbi, REQ_TIME);
2227 	fi->atomic_write_task = current;
2228 	stat_update_max_atomic_write(inode);
2229 	fi->atomic_write_cnt = 0;
2230 out:
2231 	inode_unlock(inode);
2232 	mnt_drop_write_file(filp);
2233 	return ret;
2234 }
2235 
2236 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2237 {
2238 	struct inode *inode = file_inode(filp);
2239 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2240 	int ret;
2241 
2242 	if (!(filp->f_mode & FMODE_WRITE))
2243 		return -EBADF;
2244 
2245 	if (!inode_owner_or_capable(idmap, inode))
2246 		return -EACCES;
2247 
2248 	ret = mnt_want_write_file(filp);
2249 	if (ret)
2250 		return ret;
2251 
2252 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2253 
2254 	inode_lock(inode);
2255 
2256 	if (f2fs_is_atomic_file(inode)) {
2257 		ret = f2fs_commit_atomic_write(inode);
2258 		if (!ret)
2259 			ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2260 
2261 		f2fs_abort_atomic_write(inode, ret);
2262 	} else {
2263 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2264 	}
2265 
2266 	inode_unlock(inode);
2267 	mnt_drop_write_file(filp);
2268 	return ret;
2269 }
2270 
2271 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2272 {
2273 	struct inode *inode = file_inode(filp);
2274 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
2275 	int ret;
2276 
2277 	if (!(filp->f_mode & FMODE_WRITE))
2278 		return -EBADF;
2279 
2280 	if (!inode_owner_or_capable(idmap, inode))
2281 		return -EACCES;
2282 
2283 	ret = mnt_want_write_file(filp);
2284 	if (ret)
2285 		return ret;
2286 
2287 	inode_lock(inode);
2288 
2289 	f2fs_abort_atomic_write(inode, true);
2290 
2291 	inode_unlock(inode);
2292 
2293 	mnt_drop_write_file(filp);
2294 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2295 	return ret;
2296 }
2297 
2298 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
2299 						bool readonly, bool need_lock)
2300 {
2301 	struct super_block *sb = sbi->sb;
2302 	int ret = 0;
2303 
2304 	switch (flag) {
2305 	case F2FS_GOING_DOWN_FULLSYNC:
2306 		ret = bdev_freeze(sb->s_bdev);
2307 		if (ret)
2308 			goto out;
2309 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2310 		bdev_thaw(sb->s_bdev);
2311 		break;
2312 	case F2FS_GOING_DOWN_METASYNC:
2313 		/* do checkpoint only */
2314 		ret = f2fs_sync_fs(sb, 1);
2315 		if (ret) {
2316 			if (ret == -EIO)
2317 				ret = 0;
2318 			goto out;
2319 		}
2320 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2321 		break;
2322 	case F2FS_GOING_DOWN_NOSYNC:
2323 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2324 		break;
2325 	case F2FS_GOING_DOWN_METAFLUSH:
2326 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2327 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2328 		break;
2329 	case F2FS_GOING_DOWN_NEED_FSCK:
2330 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2331 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2332 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2333 		/* do checkpoint only */
2334 		ret = f2fs_sync_fs(sb, 1);
2335 		if (ret == -EIO)
2336 			ret = 0;
2337 		goto out;
2338 	default:
2339 		ret = -EINVAL;
2340 		goto out;
2341 	}
2342 
2343 	if (readonly)
2344 		goto out;
2345 
2346 	/* grab sb->s_umount to avoid racing w/ remount() */
2347 	if (need_lock)
2348 		down_read(&sbi->sb->s_umount);
2349 
2350 	f2fs_stop_gc_thread(sbi);
2351 	f2fs_stop_discard_thread(sbi);
2352 
2353 	f2fs_drop_discard_cmd(sbi);
2354 	clear_opt(sbi, DISCARD);
2355 
2356 	if (need_lock)
2357 		up_read(&sbi->sb->s_umount);
2358 
2359 	f2fs_update_time(sbi, REQ_TIME);
2360 out:
2361 
2362 	trace_f2fs_shutdown(sbi, flag, ret);
2363 
2364 	return ret;
2365 }
2366 
2367 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2368 {
2369 	struct inode *inode = file_inode(filp);
2370 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2371 	__u32 in;
2372 	int ret;
2373 	bool need_drop = false, readonly = false;
2374 
2375 	if (!capable(CAP_SYS_ADMIN))
2376 		return -EPERM;
2377 
2378 	if (get_user(in, (__u32 __user *)arg))
2379 		return -EFAULT;
2380 
2381 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2382 		ret = mnt_want_write_file(filp);
2383 		if (ret) {
2384 			if (ret != -EROFS)
2385 				return ret;
2386 
2387 			/* fallback to nosync shutdown for readonly fs */
2388 			in = F2FS_GOING_DOWN_NOSYNC;
2389 			readonly = true;
2390 		} else {
2391 			need_drop = true;
2392 		}
2393 	}
2394 
2395 	ret = f2fs_do_shutdown(sbi, in, readonly, true);
2396 
2397 	if (need_drop)
2398 		mnt_drop_write_file(filp);
2399 
2400 	return ret;
2401 }
2402 
2403 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2404 {
2405 	struct inode *inode = file_inode(filp);
2406 	struct super_block *sb = inode->i_sb;
2407 	struct fstrim_range range;
2408 	int ret;
2409 
2410 	if (!capable(CAP_SYS_ADMIN))
2411 		return -EPERM;
2412 
2413 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2414 		return -EOPNOTSUPP;
2415 
2416 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2417 				sizeof(range)))
2418 		return -EFAULT;
2419 
2420 	ret = mnt_want_write_file(filp);
2421 	if (ret)
2422 		return ret;
2423 
2424 	range.minlen = max((unsigned int)range.minlen,
2425 			   bdev_discard_granularity(sb->s_bdev));
2426 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2427 	mnt_drop_write_file(filp);
2428 	if (ret < 0)
2429 		return ret;
2430 
2431 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2432 				sizeof(range)))
2433 		return -EFAULT;
2434 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2435 	return 0;
2436 }
2437 
2438 static bool uuid_is_nonzero(__u8 u[16])
2439 {
2440 	int i;
2441 
2442 	for (i = 0; i < 16; i++)
2443 		if (u[i])
2444 			return true;
2445 	return false;
2446 }
2447 
2448 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2449 {
2450 	struct inode *inode = file_inode(filp);
2451 	int ret;
2452 
2453 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2454 		return -EOPNOTSUPP;
2455 
2456 	ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2457 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2458 	return ret;
2459 }
2460 
2461 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2462 {
2463 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2464 		return -EOPNOTSUPP;
2465 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2466 }
2467 
2468 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2469 {
2470 	struct inode *inode = file_inode(filp);
2471 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2472 	u8 encrypt_pw_salt[16];
2473 	int err;
2474 
2475 	if (!f2fs_sb_has_encrypt(sbi))
2476 		return -EOPNOTSUPP;
2477 
2478 	err = mnt_want_write_file(filp);
2479 	if (err)
2480 		return err;
2481 
2482 	f2fs_down_write(&sbi->sb_lock);
2483 
2484 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2485 		goto got_it;
2486 
2487 	/* update superblock with uuid */
2488 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2489 
2490 	err = f2fs_commit_super(sbi, false);
2491 	if (err) {
2492 		/* undo new data */
2493 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2494 		goto out_err;
2495 	}
2496 got_it:
2497 	memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2498 out_err:
2499 	f2fs_up_write(&sbi->sb_lock);
2500 	mnt_drop_write_file(filp);
2501 
2502 	if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2503 		err = -EFAULT;
2504 
2505 	return err;
2506 }
2507 
2508 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2509 					     unsigned long arg)
2510 {
2511 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2512 		return -EOPNOTSUPP;
2513 
2514 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2515 }
2516 
2517 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2518 {
2519 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2520 		return -EOPNOTSUPP;
2521 
2522 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2523 }
2524 
2525 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2526 {
2527 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2528 		return -EOPNOTSUPP;
2529 
2530 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2531 }
2532 
2533 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2534 						    unsigned long arg)
2535 {
2536 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2537 		return -EOPNOTSUPP;
2538 
2539 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2540 }
2541 
2542 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2543 					      unsigned long arg)
2544 {
2545 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2546 		return -EOPNOTSUPP;
2547 
2548 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2549 }
2550 
2551 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2552 {
2553 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2554 		return -EOPNOTSUPP;
2555 
2556 	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2557 }
2558 
2559 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2560 {
2561 	struct inode *inode = file_inode(filp);
2562 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2563 	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2564 			.no_bg_gc = false,
2565 			.should_migrate_blocks = false,
2566 			.nr_free_secs = 0 };
2567 	__u32 sync;
2568 	int ret;
2569 
2570 	if (!capable(CAP_SYS_ADMIN))
2571 		return -EPERM;
2572 
2573 	if (get_user(sync, (__u32 __user *)arg))
2574 		return -EFAULT;
2575 
2576 	if (f2fs_readonly(sbi->sb))
2577 		return -EROFS;
2578 
2579 	ret = mnt_want_write_file(filp);
2580 	if (ret)
2581 		return ret;
2582 
2583 	if (!sync) {
2584 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2585 			ret = -EBUSY;
2586 			goto out;
2587 		}
2588 	} else {
2589 		f2fs_down_write(&sbi->gc_lock);
2590 	}
2591 
2592 	gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2593 	gc_control.err_gc_skipped = sync;
2594 	stat_inc_gc_call_count(sbi, FOREGROUND);
2595 	ret = f2fs_gc(sbi, &gc_control);
2596 out:
2597 	mnt_drop_write_file(filp);
2598 	return ret;
2599 }
2600 
2601 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2602 {
2603 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2604 	struct f2fs_gc_control gc_control = {
2605 			.init_gc_type = range->sync ? FG_GC : BG_GC,
2606 			.no_bg_gc = false,
2607 			.should_migrate_blocks = false,
2608 			.err_gc_skipped = range->sync,
2609 			.nr_free_secs = 0 };
2610 	u64 end;
2611 	int ret;
2612 
2613 	if (!capable(CAP_SYS_ADMIN))
2614 		return -EPERM;
2615 	if (f2fs_readonly(sbi->sb))
2616 		return -EROFS;
2617 
2618 	end = range->start + range->len;
2619 	if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2620 					end >= MAX_BLKADDR(sbi))
2621 		return -EINVAL;
2622 
2623 	ret = mnt_want_write_file(filp);
2624 	if (ret)
2625 		return ret;
2626 
2627 do_more:
2628 	if (!range->sync) {
2629 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2630 			ret = -EBUSY;
2631 			goto out;
2632 		}
2633 	} else {
2634 		f2fs_down_write(&sbi->gc_lock);
2635 	}
2636 
2637 	gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2638 	stat_inc_gc_call_count(sbi, FOREGROUND);
2639 	ret = f2fs_gc(sbi, &gc_control);
2640 	if (ret) {
2641 		if (ret == -EBUSY)
2642 			ret = -EAGAIN;
2643 		goto out;
2644 	}
2645 	range->start += CAP_BLKS_PER_SEC(sbi);
2646 	if (range->start <= end)
2647 		goto do_more;
2648 out:
2649 	mnt_drop_write_file(filp);
2650 	return ret;
2651 }
2652 
2653 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2654 {
2655 	struct f2fs_gc_range range;
2656 
2657 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2658 							sizeof(range)))
2659 		return -EFAULT;
2660 	return __f2fs_ioc_gc_range(filp, &range);
2661 }
2662 
2663 static int f2fs_ioc_write_checkpoint(struct file *filp)
2664 {
2665 	struct inode *inode = file_inode(filp);
2666 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2667 	int ret;
2668 
2669 	if (!capable(CAP_SYS_ADMIN))
2670 		return -EPERM;
2671 
2672 	if (f2fs_readonly(sbi->sb))
2673 		return -EROFS;
2674 
2675 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2676 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2677 		return -EINVAL;
2678 	}
2679 
2680 	ret = mnt_want_write_file(filp);
2681 	if (ret)
2682 		return ret;
2683 
2684 	ret = f2fs_sync_fs(sbi->sb, 1);
2685 
2686 	mnt_drop_write_file(filp);
2687 	return ret;
2688 }
2689 
2690 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2691 					struct file *filp,
2692 					struct f2fs_defragment *range)
2693 {
2694 	struct inode *inode = file_inode(filp);
2695 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2696 					.m_seg_type = NO_CHECK_TYPE,
2697 					.m_may_create = false };
2698 	struct extent_info ei = {};
2699 	pgoff_t pg_start, pg_end, next_pgofs;
2700 	unsigned int total = 0, sec_num;
2701 	block_t blk_end = 0;
2702 	bool fragmented = false;
2703 	int err;
2704 
2705 	f2fs_balance_fs(sbi, true);
2706 
2707 	inode_lock(inode);
2708 	pg_start = range->start >> PAGE_SHIFT;
2709 	pg_end = min_t(pgoff_t,
2710 				(range->start + range->len) >> PAGE_SHIFT,
2711 				DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));
2712 
2713 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) ||
2714 		f2fs_is_atomic_file(inode)) {
2715 		err = -EINVAL;
2716 		goto unlock_out;
2717 	}
2718 
2719 	/* if in-place-update policy is enabled, don't waste time here */
2720 	set_inode_flag(inode, FI_OPU_WRITE);
2721 	if (f2fs_should_update_inplace(inode, NULL)) {
2722 		err = -EINVAL;
2723 		goto out;
2724 	}
2725 
2726 	/* writeback all dirty pages in the range */
2727 	err = filemap_write_and_wait_range(inode->i_mapping,
2728 						pg_start << PAGE_SHIFT,
2729 						(pg_end << PAGE_SHIFT) - 1);
2730 	if (err)
2731 		goto out;
2732 
2733 	/*
2734 	 * lookup mapping info in extent cache, skip defragmenting if physical
2735 	 * block addresses are continuous.
2736 	 */
2737 	if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2738 		if ((pgoff_t)ei.fofs + ei.len >= pg_end)
2739 			goto out;
2740 	}
2741 
2742 	map.m_lblk = pg_start;
2743 	map.m_next_pgofs = &next_pgofs;
2744 
2745 	/*
2746 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2747 	 * physical block addresses are continuous even if there are hole(s)
2748 	 * in logical blocks.
2749 	 */
2750 	while (map.m_lblk < pg_end) {
2751 		map.m_len = pg_end - map.m_lblk;
2752 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2753 		if (err)
2754 			goto out;
2755 
2756 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2757 			map.m_lblk = next_pgofs;
2758 			continue;
2759 		}
2760 
2761 		if (blk_end && blk_end != map.m_pblk)
2762 			fragmented = true;
2763 
2764 		/* record total count of block that we're going to move */
2765 		total += map.m_len;
2766 
2767 		blk_end = map.m_pblk + map.m_len;
2768 
2769 		map.m_lblk += map.m_len;
2770 	}
2771 
2772 	if (!fragmented) {
2773 		total = 0;
2774 		goto out;
2775 	}
2776 
2777 	sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2778 
2779 	/*
2780 	 * make sure there are enough free section for LFS allocation, this can
2781 	 * avoid defragment running in SSR mode when free section are allocated
2782 	 * intensively
2783 	 */
2784 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2785 		err = -EAGAIN;
2786 		goto out;
2787 	}
2788 
2789 	map.m_lblk = pg_start;
2790 	map.m_len = pg_end - pg_start;
2791 	total = 0;
2792 
2793 	while (map.m_lblk < pg_end) {
2794 		pgoff_t idx;
2795 		int cnt = 0;
2796 
2797 do_map:
2798 		map.m_len = pg_end - map.m_lblk;
2799 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2800 		if (err)
2801 			goto clear_out;
2802 
2803 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2804 			map.m_lblk = next_pgofs;
2805 			goto check;
2806 		}
2807 
2808 		set_inode_flag(inode, FI_SKIP_WRITES);
2809 
2810 		idx = map.m_lblk;
2811 		while (idx < map.m_lblk + map.m_len &&
2812 						cnt < BLKS_PER_SEG(sbi)) {
2813 			struct page *page;
2814 
2815 			page = f2fs_get_lock_data_page(inode, idx, true);
2816 			if (IS_ERR(page)) {
2817 				err = PTR_ERR(page);
2818 				goto clear_out;
2819 			}
2820 
2821 			f2fs_wait_on_page_writeback(page, DATA, true, true);
2822 
2823 			set_page_dirty(page);
2824 			set_page_private_gcing(page);
2825 			f2fs_put_page(page, 1);
2826 
2827 			idx++;
2828 			cnt++;
2829 			total++;
2830 		}
2831 
2832 		map.m_lblk = idx;
2833 check:
2834 		if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi))
2835 			goto do_map;
2836 
2837 		clear_inode_flag(inode, FI_SKIP_WRITES);
2838 
2839 		err = filemap_fdatawrite(inode->i_mapping);
2840 		if (err)
2841 			goto out;
2842 	}
2843 clear_out:
2844 	clear_inode_flag(inode, FI_SKIP_WRITES);
2845 out:
2846 	clear_inode_flag(inode, FI_OPU_WRITE);
2847 unlock_out:
2848 	inode_unlock(inode);
2849 	if (!err)
2850 		range->len = (u64)total << PAGE_SHIFT;
2851 	return err;
2852 }
2853 
2854 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2855 {
2856 	struct inode *inode = file_inode(filp);
2857 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2858 	struct f2fs_defragment range;
2859 	int err;
2860 
2861 	if (!capable(CAP_SYS_ADMIN))
2862 		return -EPERM;
2863 
2864 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2865 		return -EINVAL;
2866 
2867 	if (f2fs_readonly(sbi->sb))
2868 		return -EROFS;
2869 
2870 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2871 							sizeof(range)))
2872 		return -EFAULT;
2873 
2874 	/* verify alignment of offset & size */
2875 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2876 		return -EINVAL;
2877 
2878 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2879 					max_file_blocks(inode)))
2880 		return -EINVAL;
2881 
2882 	err = mnt_want_write_file(filp);
2883 	if (err)
2884 		return err;
2885 
2886 	err = f2fs_defragment_range(sbi, filp, &range);
2887 	mnt_drop_write_file(filp);
2888 
2889 	if (range.len)
2890 		f2fs_update_time(sbi, REQ_TIME);
2891 	if (err < 0)
2892 		return err;
2893 
2894 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2895 							sizeof(range)))
2896 		return -EFAULT;
2897 
2898 	return 0;
2899 }
2900 
2901 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2902 			struct file *file_out, loff_t pos_out, size_t len)
2903 {
2904 	struct inode *src = file_inode(file_in);
2905 	struct inode *dst = file_inode(file_out);
2906 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2907 	size_t olen = len, dst_max_i_size = 0;
2908 	size_t dst_osize;
2909 	int ret;
2910 
2911 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2912 				src->i_sb != dst->i_sb)
2913 		return -EXDEV;
2914 
2915 	if (unlikely(f2fs_readonly(src->i_sb)))
2916 		return -EROFS;
2917 
2918 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2919 		return -EINVAL;
2920 
2921 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2922 		return -EOPNOTSUPP;
2923 
2924 	if (pos_out < 0 || pos_in < 0)
2925 		return -EINVAL;
2926 
2927 	if (src == dst) {
2928 		if (pos_in == pos_out)
2929 			return 0;
2930 		if (pos_out > pos_in && pos_out < pos_in + len)
2931 			return -EINVAL;
2932 	}
2933 
2934 	inode_lock(src);
2935 	if (src != dst) {
2936 		ret = -EBUSY;
2937 		if (!inode_trylock(dst))
2938 			goto out;
2939 	}
2940 
2941 	if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) ||
2942 		f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) {
2943 		ret = -EOPNOTSUPP;
2944 		goto out_unlock;
2945 	}
2946 
2947 	if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) {
2948 		ret = -EINVAL;
2949 		goto out_unlock;
2950 	}
2951 
2952 	ret = -EINVAL;
2953 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2954 		goto out_unlock;
2955 	if (len == 0)
2956 		olen = len = src->i_size - pos_in;
2957 	if (pos_in + len == src->i_size)
2958 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2959 	if (len == 0) {
2960 		ret = 0;
2961 		goto out_unlock;
2962 	}
2963 
2964 	dst_osize = dst->i_size;
2965 	if (pos_out + olen > dst->i_size)
2966 		dst_max_i_size = pos_out + olen;
2967 
2968 	/* verify the end result is block aligned */
2969 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2970 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2971 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2972 		goto out_unlock;
2973 
2974 	ret = f2fs_convert_inline_inode(src);
2975 	if (ret)
2976 		goto out_unlock;
2977 
2978 	ret = f2fs_convert_inline_inode(dst);
2979 	if (ret)
2980 		goto out_unlock;
2981 
2982 	/* write out all dirty pages from offset */
2983 	ret = filemap_write_and_wait_range(src->i_mapping,
2984 					pos_in, pos_in + len);
2985 	if (ret)
2986 		goto out_unlock;
2987 
2988 	ret = filemap_write_and_wait_range(dst->i_mapping,
2989 					pos_out, pos_out + len);
2990 	if (ret)
2991 		goto out_unlock;
2992 
2993 	f2fs_balance_fs(sbi, true);
2994 
2995 	f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2996 	if (src != dst) {
2997 		ret = -EBUSY;
2998 		if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2999 			goto out_src;
3000 	}
3001 
3002 	f2fs_lock_op(sbi);
3003 	ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in),
3004 				F2FS_BYTES_TO_BLK(pos_out),
3005 				F2FS_BYTES_TO_BLK(len), false);
3006 
3007 	if (!ret) {
3008 		if (dst_max_i_size)
3009 			f2fs_i_size_write(dst, dst_max_i_size);
3010 		else if (dst_osize != dst->i_size)
3011 			f2fs_i_size_write(dst, dst_osize);
3012 	}
3013 	f2fs_unlock_op(sbi);
3014 
3015 	if (src != dst)
3016 		f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
3017 out_src:
3018 	f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3019 	if (ret)
3020 		goto out_unlock;
3021 
3022 	inode_set_mtime_to_ts(src, inode_set_ctime_current(src));
3023 	f2fs_mark_inode_dirty_sync(src, false);
3024 	if (src != dst) {
3025 		inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst));
3026 		f2fs_mark_inode_dirty_sync(dst, false);
3027 	}
3028 	f2fs_update_time(sbi, REQ_TIME);
3029 
3030 out_unlock:
3031 	if (src != dst)
3032 		inode_unlock(dst);
3033 out:
3034 	inode_unlock(src);
3035 	return ret;
3036 }
3037 
3038 static int __f2fs_ioc_move_range(struct file *filp,
3039 				struct f2fs_move_range *range)
3040 {
3041 	int err;
3042 
3043 	if (!(filp->f_mode & FMODE_READ) ||
3044 			!(filp->f_mode & FMODE_WRITE))
3045 		return -EBADF;
3046 
3047 	CLASS(fd, dst)(range->dst_fd);
3048 	if (fd_empty(dst))
3049 		return -EBADF;
3050 
3051 	if (!(fd_file(dst)->f_mode & FMODE_WRITE))
3052 		return -EBADF;
3053 
3054 	err = mnt_want_write_file(filp);
3055 	if (err)
3056 		return err;
3057 
3058 	err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst),
3059 					range->pos_out, range->len);
3060 
3061 	mnt_drop_write_file(filp);
3062 	return err;
3063 }
3064 
3065 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
3066 {
3067 	struct f2fs_move_range range;
3068 
3069 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
3070 							sizeof(range)))
3071 		return -EFAULT;
3072 	return __f2fs_ioc_move_range(filp, &range);
3073 }
3074 
3075 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
3076 {
3077 	struct inode *inode = file_inode(filp);
3078 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3079 	struct sit_info *sm = SIT_I(sbi);
3080 	unsigned int start_segno = 0, end_segno = 0;
3081 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
3082 	struct f2fs_flush_device range;
3083 	struct f2fs_gc_control gc_control = {
3084 			.init_gc_type = FG_GC,
3085 			.should_migrate_blocks = true,
3086 			.err_gc_skipped = true,
3087 			.nr_free_secs = 0 };
3088 	int ret;
3089 
3090 	if (!capable(CAP_SYS_ADMIN))
3091 		return -EPERM;
3092 
3093 	if (f2fs_readonly(sbi->sb))
3094 		return -EROFS;
3095 
3096 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3097 		return -EINVAL;
3098 
3099 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
3100 							sizeof(range)))
3101 		return -EFAULT;
3102 
3103 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
3104 			__is_large_section(sbi)) {
3105 		f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1",
3106 			  range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi));
3107 		return -EINVAL;
3108 	}
3109 
3110 	ret = mnt_want_write_file(filp);
3111 	if (ret)
3112 		return ret;
3113 
3114 	if (range.dev_num != 0)
3115 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
3116 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
3117 
3118 	start_segno = sm->last_victim[FLUSH_DEVICE];
3119 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
3120 		start_segno = dev_start_segno;
3121 	end_segno = min(start_segno + range.segments, dev_end_segno);
3122 
3123 	while (start_segno < end_segno) {
3124 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
3125 			ret = -EBUSY;
3126 			goto out;
3127 		}
3128 		sm->last_victim[GC_CB] = end_segno + 1;
3129 		sm->last_victim[GC_GREEDY] = end_segno + 1;
3130 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
3131 
3132 		gc_control.victim_segno = start_segno;
3133 		stat_inc_gc_call_count(sbi, FOREGROUND);
3134 		ret = f2fs_gc(sbi, &gc_control);
3135 		if (ret == -EAGAIN)
3136 			ret = 0;
3137 		else if (ret < 0)
3138 			break;
3139 		start_segno++;
3140 	}
3141 out:
3142 	mnt_drop_write_file(filp);
3143 	return ret;
3144 }
3145 
3146 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3147 {
3148 	struct inode *inode = file_inode(filp);
3149 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3150 
3151 	/* Must validate to set it with SQLite behavior in Android. */
3152 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3153 
3154 	return put_user(sb_feature, (u32 __user *)arg);
3155 }
3156 
3157 #ifdef CONFIG_QUOTA
3158 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3159 {
3160 	struct dquot *transfer_to[MAXQUOTAS] = {};
3161 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3162 	struct super_block *sb = sbi->sb;
3163 	int err;
3164 
3165 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3166 	if (IS_ERR(transfer_to[PRJQUOTA]))
3167 		return PTR_ERR(transfer_to[PRJQUOTA]);
3168 
3169 	err = __dquot_transfer(inode, transfer_to);
3170 	if (err)
3171 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3172 	dqput(transfer_to[PRJQUOTA]);
3173 	return err;
3174 }
3175 
3176 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3177 {
3178 	struct f2fs_inode_info *fi = F2FS_I(inode);
3179 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3180 	struct f2fs_inode *ri = NULL;
3181 	kprojid_t kprojid;
3182 	int err;
3183 
3184 	if (!f2fs_sb_has_project_quota(sbi)) {
3185 		if (projid != F2FS_DEF_PROJID)
3186 			return -EOPNOTSUPP;
3187 		else
3188 			return 0;
3189 	}
3190 
3191 	if (!f2fs_has_extra_attr(inode))
3192 		return -EOPNOTSUPP;
3193 
3194 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3195 
3196 	if (projid_eq(kprojid, fi->i_projid))
3197 		return 0;
3198 
3199 	err = -EPERM;
3200 	/* Is it quota file? Do not allow user to mess with it */
3201 	if (IS_NOQUOTA(inode))
3202 		return err;
3203 
3204 	if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3205 		return -EOVERFLOW;
3206 
3207 	err = f2fs_dquot_initialize(inode);
3208 	if (err)
3209 		return err;
3210 
3211 	f2fs_lock_op(sbi);
3212 	err = f2fs_transfer_project_quota(inode, kprojid);
3213 	if (err)
3214 		goto out_unlock;
3215 
3216 	fi->i_projid = kprojid;
3217 	inode_set_ctime_current(inode);
3218 	f2fs_mark_inode_dirty_sync(inode, true);
3219 out_unlock:
3220 	f2fs_unlock_op(sbi);
3221 	return err;
3222 }
3223 #else
3224 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3225 {
3226 	return 0;
3227 }
3228 
3229 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3230 {
3231 	if (projid != F2FS_DEF_PROJID)
3232 		return -EOPNOTSUPP;
3233 	return 0;
3234 }
3235 #endif
3236 
3237 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3238 {
3239 	struct inode *inode = d_inode(dentry);
3240 	struct f2fs_inode_info *fi = F2FS_I(inode);
3241 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3242 
3243 	if (IS_ENCRYPTED(inode))
3244 		fsflags |= FS_ENCRYPT_FL;
3245 	if (IS_VERITY(inode))
3246 		fsflags |= FS_VERITY_FL;
3247 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3248 		fsflags |= FS_INLINE_DATA_FL;
3249 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3250 		fsflags |= FS_NOCOW_FL;
3251 
3252 	fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3253 
3254 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3255 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3256 
3257 	return 0;
3258 }
3259 
3260 int f2fs_fileattr_set(struct mnt_idmap *idmap,
3261 		      struct dentry *dentry, struct fileattr *fa)
3262 {
3263 	struct inode *inode = d_inode(dentry);
3264 	u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3265 	u32 iflags;
3266 	int err;
3267 
3268 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3269 		return -EIO;
3270 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3271 		return -ENOSPC;
3272 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
3273 		return -EOPNOTSUPP;
3274 	fsflags &= F2FS_SETTABLE_FS_FL;
3275 	if (!fa->flags_valid)
3276 		mask &= FS_COMMON_FL;
3277 
3278 	iflags = f2fs_fsflags_to_iflags(fsflags);
3279 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3280 		return -EOPNOTSUPP;
3281 
3282 	err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3283 	if (!err)
3284 		err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3285 
3286 	return err;
3287 }
3288 
3289 int f2fs_pin_file_control(struct inode *inode, bool inc)
3290 {
3291 	struct f2fs_inode_info *fi = F2FS_I(inode);
3292 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3293 
3294 	if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) {
3295 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3296 			  __func__, inode->i_ino, fi->i_gc_failures);
3297 		clear_inode_flag(inode, FI_PIN_FILE);
3298 		return -EAGAIN;
3299 	}
3300 
3301 	/* Use i_gc_failures for normal file as a risk signal. */
3302 	if (inc)
3303 		f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
3304 
3305 	return 0;
3306 }
3307 
3308 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3309 {
3310 	struct inode *inode = file_inode(filp);
3311 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3312 	__u32 pin;
3313 	int ret = 0;
3314 
3315 	if (get_user(pin, (__u32 __user *)arg))
3316 		return -EFAULT;
3317 
3318 	if (!S_ISREG(inode->i_mode))
3319 		return -EINVAL;
3320 
3321 	if (f2fs_readonly(sbi->sb))
3322 		return -EROFS;
3323 
3324 	ret = mnt_want_write_file(filp);
3325 	if (ret)
3326 		return ret;
3327 
3328 	inode_lock(inode);
3329 
3330 	if (f2fs_is_atomic_file(inode)) {
3331 		ret = -EINVAL;
3332 		goto out;
3333 	}
3334 
3335 	if (!pin) {
3336 		clear_inode_flag(inode, FI_PIN_FILE);
3337 		f2fs_i_gc_failures_write(inode, 0);
3338 		goto done;
3339 	} else if (f2fs_is_pinned_file(inode)) {
3340 		goto done;
3341 	}
3342 
3343 	if (F2FS_HAS_BLOCKS(inode)) {
3344 		ret = -EFBIG;
3345 		goto out;
3346 	}
3347 
3348 	/* Let's allow file pinning on zoned device. */
3349 	if (!f2fs_sb_has_blkzoned(sbi) &&
3350 	    f2fs_should_update_outplace(inode, NULL)) {
3351 		ret = -EINVAL;
3352 		goto out;
3353 	}
3354 
3355 	if (f2fs_pin_file_control(inode, false)) {
3356 		ret = -EAGAIN;
3357 		goto out;
3358 	}
3359 
3360 	ret = f2fs_convert_inline_inode(inode);
3361 	if (ret)
3362 		goto out;
3363 
3364 	if (!f2fs_disable_compressed_file(inode)) {
3365 		ret = -EOPNOTSUPP;
3366 		goto out;
3367 	}
3368 
3369 	set_inode_flag(inode, FI_PIN_FILE);
3370 	ret = F2FS_I(inode)->i_gc_failures;
3371 done:
3372 	f2fs_update_time(sbi, REQ_TIME);
3373 out:
3374 	inode_unlock(inode);
3375 	mnt_drop_write_file(filp);
3376 	return ret;
3377 }
3378 
3379 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3380 {
3381 	struct inode *inode = file_inode(filp);
3382 	__u32 pin = 0;
3383 
3384 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3385 		pin = F2FS_I(inode)->i_gc_failures;
3386 	return put_user(pin, (u32 __user *)arg);
3387 }
3388 
3389 int f2fs_precache_extents(struct inode *inode)
3390 {
3391 	struct f2fs_inode_info *fi = F2FS_I(inode);
3392 	struct f2fs_map_blocks map;
3393 	pgoff_t m_next_extent;
3394 	loff_t end;
3395 	int err;
3396 
3397 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3398 		return -EOPNOTSUPP;
3399 
3400 	map.m_lblk = 0;
3401 	map.m_pblk = 0;
3402 	map.m_next_pgofs = NULL;
3403 	map.m_next_extent = &m_next_extent;
3404 	map.m_seg_type = NO_CHECK_TYPE;
3405 	map.m_may_create = false;
3406 	end = F2FS_BLK_ALIGN(i_size_read(inode));
3407 
3408 	while (map.m_lblk < end) {
3409 		map.m_len = end - map.m_lblk;
3410 
3411 		f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3412 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
3413 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3414 		if (err || !map.m_len)
3415 			return err;
3416 
3417 		map.m_lblk = m_next_extent;
3418 	}
3419 
3420 	return 0;
3421 }
3422 
3423 static int f2fs_ioc_precache_extents(struct file *filp)
3424 {
3425 	return f2fs_precache_extents(file_inode(filp));
3426 }
3427 
3428 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3429 {
3430 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3431 	__u64 block_count;
3432 
3433 	if (!capable(CAP_SYS_ADMIN))
3434 		return -EPERM;
3435 
3436 	if (f2fs_readonly(sbi->sb))
3437 		return -EROFS;
3438 
3439 	if (copy_from_user(&block_count, (void __user *)arg,
3440 			   sizeof(block_count)))
3441 		return -EFAULT;
3442 
3443 	return f2fs_resize_fs(filp, block_count);
3444 }
3445 
3446 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3447 {
3448 	struct inode *inode = file_inode(filp);
3449 
3450 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3451 
3452 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3453 		f2fs_warn(F2FS_I_SB(inode),
3454 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3455 			  inode->i_ino);
3456 		return -EOPNOTSUPP;
3457 	}
3458 
3459 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3460 }
3461 
3462 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3463 {
3464 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3465 		return -EOPNOTSUPP;
3466 
3467 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3468 }
3469 
3470 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3471 {
3472 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3473 		return -EOPNOTSUPP;
3474 
3475 	return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3476 }
3477 
3478 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3479 {
3480 	struct inode *inode = file_inode(filp);
3481 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3482 	char *vbuf;
3483 	int count;
3484 	int err = 0;
3485 
3486 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3487 	if (!vbuf)
3488 		return -ENOMEM;
3489 
3490 	f2fs_down_read(&sbi->sb_lock);
3491 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3492 			ARRAY_SIZE(sbi->raw_super->volume_name),
3493 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3494 	f2fs_up_read(&sbi->sb_lock);
3495 
3496 	if (copy_to_user((char __user *)arg, vbuf,
3497 				min(FSLABEL_MAX, count)))
3498 		err = -EFAULT;
3499 
3500 	kfree(vbuf);
3501 	return err;
3502 }
3503 
3504 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3505 {
3506 	struct inode *inode = file_inode(filp);
3507 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3508 	char *vbuf;
3509 	int err = 0;
3510 
3511 	if (!capable(CAP_SYS_ADMIN))
3512 		return -EPERM;
3513 
3514 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3515 	if (IS_ERR(vbuf))
3516 		return PTR_ERR(vbuf);
3517 
3518 	err = mnt_want_write_file(filp);
3519 	if (err)
3520 		goto out;
3521 
3522 	f2fs_down_write(&sbi->sb_lock);
3523 
3524 	memset(sbi->raw_super->volume_name, 0,
3525 			sizeof(sbi->raw_super->volume_name));
3526 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3527 			sbi->raw_super->volume_name,
3528 			ARRAY_SIZE(sbi->raw_super->volume_name));
3529 
3530 	err = f2fs_commit_super(sbi, false);
3531 
3532 	f2fs_up_write(&sbi->sb_lock);
3533 
3534 	mnt_drop_write_file(filp);
3535 out:
3536 	kfree(vbuf);
3537 	return err;
3538 }
3539 
3540 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks)
3541 {
3542 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3543 		return -EOPNOTSUPP;
3544 
3545 	if (!f2fs_compressed_file(inode))
3546 		return -EINVAL;
3547 
3548 	*blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3549 
3550 	return 0;
3551 }
3552 
3553 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg)
3554 {
3555 	struct inode *inode = file_inode(filp);
3556 	__u64 blocks;
3557 	int ret;
3558 
3559 	ret = f2fs_get_compress_blocks(inode, &blocks);
3560 	if (ret < 0)
3561 		return ret;
3562 
3563 	return put_user(blocks, (u64 __user *)arg);
3564 }
3565 
3566 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3567 {
3568 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3569 	unsigned int released_blocks = 0;
3570 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3571 	block_t blkaddr;
3572 	int i;
3573 
3574 	for (i = 0; i < count; i++) {
3575 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3576 						dn->ofs_in_node + i);
3577 
3578 		if (!__is_valid_data_blkaddr(blkaddr))
3579 			continue;
3580 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3581 					DATA_GENERIC_ENHANCE)))
3582 			return -EFSCORRUPTED;
3583 	}
3584 
3585 	while (count) {
3586 		int compr_blocks = 0;
3587 
3588 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3589 			blkaddr = f2fs_data_blkaddr(dn);
3590 
3591 			if (i == 0) {
3592 				if (blkaddr == COMPRESS_ADDR)
3593 					continue;
3594 				dn->ofs_in_node += cluster_size;
3595 				goto next;
3596 			}
3597 
3598 			if (__is_valid_data_blkaddr(blkaddr))
3599 				compr_blocks++;
3600 
3601 			if (blkaddr != NEW_ADDR)
3602 				continue;
3603 
3604 			f2fs_set_data_blkaddr(dn, NULL_ADDR);
3605 		}
3606 
3607 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3608 		dec_valid_block_count(sbi, dn->inode,
3609 					cluster_size - compr_blocks);
3610 
3611 		released_blocks += cluster_size - compr_blocks;
3612 next:
3613 		count -= cluster_size;
3614 	}
3615 
3616 	return released_blocks;
3617 }
3618 
3619 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3620 {
3621 	struct inode *inode = file_inode(filp);
3622 	struct f2fs_inode_info *fi = F2FS_I(inode);
3623 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3624 	pgoff_t page_idx = 0, last_idx;
3625 	unsigned int released_blocks = 0;
3626 	int ret;
3627 	int writecount;
3628 
3629 	if (!f2fs_sb_has_compression(sbi))
3630 		return -EOPNOTSUPP;
3631 
3632 	if (f2fs_readonly(sbi->sb))
3633 		return -EROFS;
3634 
3635 	ret = mnt_want_write_file(filp);
3636 	if (ret)
3637 		return ret;
3638 
3639 	f2fs_balance_fs(sbi, true);
3640 
3641 	inode_lock(inode);
3642 
3643 	writecount = atomic_read(&inode->i_writecount);
3644 	if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3645 			(!(filp->f_mode & FMODE_WRITE) && writecount)) {
3646 		ret = -EBUSY;
3647 		goto out;
3648 	}
3649 
3650 	if (!f2fs_compressed_file(inode) ||
3651 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3652 		ret = -EINVAL;
3653 		goto out;
3654 	}
3655 
3656 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3657 	if (ret)
3658 		goto out;
3659 
3660 	if (!atomic_read(&fi->i_compr_blocks)) {
3661 		ret = -EPERM;
3662 		goto out;
3663 	}
3664 
3665 	set_inode_flag(inode, FI_COMPRESS_RELEASED);
3666 	inode_set_ctime_current(inode);
3667 	f2fs_mark_inode_dirty_sync(inode, true);
3668 
3669 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3670 	filemap_invalidate_lock(inode->i_mapping);
3671 
3672 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3673 
3674 	while (page_idx < last_idx) {
3675 		struct dnode_of_data dn;
3676 		pgoff_t end_offset, count;
3677 
3678 		f2fs_lock_op(sbi);
3679 
3680 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3681 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3682 		if (ret) {
3683 			f2fs_unlock_op(sbi);
3684 			if (ret == -ENOENT) {
3685 				page_idx = f2fs_get_next_page_offset(&dn,
3686 								page_idx);
3687 				ret = 0;
3688 				continue;
3689 			}
3690 			break;
3691 		}
3692 
3693 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3694 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3695 		count = round_up(count, fi->i_cluster_size);
3696 
3697 		ret = release_compress_blocks(&dn, count);
3698 
3699 		f2fs_put_dnode(&dn);
3700 
3701 		f2fs_unlock_op(sbi);
3702 
3703 		if (ret < 0)
3704 			break;
3705 
3706 		page_idx += count;
3707 		released_blocks += ret;
3708 	}
3709 
3710 	filemap_invalidate_unlock(inode->i_mapping);
3711 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3712 out:
3713 	if (released_blocks)
3714 		f2fs_update_time(sbi, REQ_TIME);
3715 	inode_unlock(inode);
3716 
3717 	mnt_drop_write_file(filp);
3718 
3719 	if (ret >= 0) {
3720 		ret = put_user(released_blocks, (u64 __user *)arg);
3721 	} else if (released_blocks &&
3722 			atomic_read(&fi->i_compr_blocks)) {
3723 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3724 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3725 			"iblocks=%llu, released=%u, compr_blocks=%u, "
3726 			"run fsck to fix.",
3727 			__func__, inode->i_ino, inode->i_blocks,
3728 			released_blocks,
3729 			atomic_read(&fi->i_compr_blocks));
3730 	}
3731 
3732 	return ret;
3733 }
3734 
3735 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count,
3736 		unsigned int *reserved_blocks)
3737 {
3738 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3739 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3740 	block_t blkaddr;
3741 	int i;
3742 
3743 	for (i = 0; i < count; i++) {
3744 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3745 						dn->ofs_in_node + i);
3746 
3747 		if (!__is_valid_data_blkaddr(blkaddr))
3748 			continue;
3749 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3750 					DATA_GENERIC_ENHANCE)))
3751 			return -EFSCORRUPTED;
3752 	}
3753 
3754 	while (count) {
3755 		int compr_blocks = 0;
3756 		blkcnt_t reserved = 0;
3757 		blkcnt_t to_reserved;
3758 		int ret;
3759 
3760 		for (i = 0; i < cluster_size; i++) {
3761 			blkaddr = data_blkaddr(dn->inode, dn->node_page,
3762 						dn->ofs_in_node + i);
3763 
3764 			if (i == 0) {
3765 				if (blkaddr != COMPRESS_ADDR) {
3766 					dn->ofs_in_node += cluster_size;
3767 					goto next;
3768 				}
3769 				continue;
3770 			}
3771 
3772 			/*
3773 			 * compressed cluster was not released due to it
3774 			 * fails in release_compress_blocks(), so NEW_ADDR
3775 			 * is a possible case.
3776 			 */
3777 			if (blkaddr == NEW_ADDR) {
3778 				reserved++;
3779 				continue;
3780 			}
3781 			if (__is_valid_data_blkaddr(blkaddr)) {
3782 				compr_blocks++;
3783 				continue;
3784 			}
3785 		}
3786 
3787 		to_reserved = cluster_size - compr_blocks - reserved;
3788 
3789 		/* for the case all blocks in cluster were reserved */
3790 		if (to_reserved == 1) {
3791 			dn->ofs_in_node += cluster_size;
3792 			goto next;
3793 		}
3794 
3795 		ret = inc_valid_block_count(sbi, dn->inode,
3796 						&to_reserved, false);
3797 		if (unlikely(ret))
3798 			return ret;
3799 
3800 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3801 			if (f2fs_data_blkaddr(dn) == NULL_ADDR)
3802 				f2fs_set_data_blkaddr(dn, NEW_ADDR);
3803 		}
3804 
3805 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3806 
3807 		*reserved_blocks += to_reserved;
3808 next:
3809 		count -= cluster_size;
3810 	}
3811 
3812 	return 0;
3813 }
3814 
3815 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3816 {
3817 	struct inode *inode = file_inode(filp);
3818 	struct f2fs_inode_info *fi = F2FS_I(inode);
3819 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3820 	pgoff_t page_idx = 0, last_idx;
3821 	unsigned int reserved_blocks = 0;
3822 	int ret;
3823 
3824 	if (!f2fs_sb_has_compression(sbi))
3825 		return -EOPNOTSUPP;
3826 
3827 	if (f2fs_readonly(sbi->sb))
3828 		return -EROFS;
3829 
3830 	ret = mnt_want_write_file(filp);
3831 	if (ret)
3832 		return ret;
3833 
3834 	f2fs_balance_fs(sbi, true);
3835 
3836 	inode_lock(inode);
3837 
3838 	if (!f2fs_compressed_file(inode) ||
3839 		!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3840 		ret = -EINVAL;
3841 		goto unlock_inode;
3842 	}
3843 
3844 	if (atomic_read(&fi->i_compr_blocks))
3845 		goto unlock_inode;
3846 
3847 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3848 	filemap_invalidate_lock(inode->i_mapping);
3849 
3850 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3851 
3852 	while (page_idx < last_idx) {
3853 		struct dnode_of_data dn;
3854 		pgoff_t end_offset, count;
3855 
3856 		f2fs_lock_op(sbi);
3857 
3858 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3859 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3860 		if (ret) {
3861 			f2fs_unlock_op(sbi);
3862 			if (ret == -ENOENT) {
3863 				page_idx = f2fs_get_next_page_offset(&dn,
3864 								page_idx);
3865 				ret = 0;
3866 				continue;
3867 			}
3868 			break;
3869 		}
3870 
3871 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3872 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3873 		count = round_up(count, fi->i_cluster_size);
3874 
3875 		ret = reserve_compress_blocks(&dn, count, &reserved_blocks);
3876 
3877 		f2fs_put_dnode(&dn);
3878 
3879 		f2fs_unlock_op(sbi);
3880 
3881 		if (ret < 0)
3882 			break;
3883 
3884 		page_idx += count;
3885 	}
3886 
3887 	filemap_invalidate_unlock(inode->i_mapping);
3888 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3889 
3890 	if (!ret) {
3891 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3892 		inode_set_ctime_current(inode);
3893 		f2fs_mark_inode_dirty_sync(inode, true);
3894 	}
3895 unlock_inode:
3896 	if (reserved_blocks)
3897 		f2fs_update_time(sbi, REQ_TIME);
3898 	inode_unlock(inode);
3899 	mnt_drop_write_file(filp);
3900 
3901 	if (!ret) {
3902 		ret = put_user(reserved_blocks, (u64 __user *)arg);
3903 	} else if (reserved_blocks &&
3904 			atomic_read(&fi->i_compr_blocks)) {
3905 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3906 		f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx "
3907 			"iblocks=%llu, reserved=%u, compr_blocks=%u, "
3908 			"run fsck to fix.",
3909 			__func__, inode->i_ino, inode->i_blocks,
3910 			reserved_blocks,
3911 			atomic_read(&fi->i_compr_blocks));
3912 	}
3913 
3914 	return ret;
3915 }
3916 
3917 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3918 		pgoff_t off, block_t block, block_t len, u32 flags)
3919 {
3920 	sector_t sector = SECTOR_FROM_BLOCK(block);
3921 	sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3922 	int ret = 0;
3923 
3924 	if (flags & F2FS_TRIM_FILE_DISCARD) {
3925 		if (bdev_max_secure_erase_sectors(bdev))
3926 			ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3927 					GFP_NOFS);
3928 		else
3929 			ret = blkdev_issue_discard(bdev, sector, nr_sects,
3930 					GFP_NOFS);
3931 	}
3932 
3933 	if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3934 		if (IS_ENCRYPTED(inode))
3935 			ret = fscrypt_zeroout_range(inode, off, block, len);
3936 		else
3937 			ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3938 					GFP_NOFS, 0);
3939 	}
3940 
3941 	return ret;
3942 }
3943 
3944 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3945 {
3946 	struct inode *inode = file_inode(filp);
3947 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3948 	struct address_space *mapping = inode->i_mapping;
3949 	struct block_device *prev_bdev = NULL;
3950 	struct f2fs_sectrim_range range;
3951 	pgoff_t index, pg_end, prev_index = 0;
3952 	block_t prev_block = 0, len = 0;
3953 	loff_t end_addr;
3954 	bool to_end = false;
3955 	int ret = 0;
3956 
3957 	if (!(filp->f_mode & FMODE_WRITE))
3958 		return -EBADF;
3959 
3960 	if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3961 				sizeof(range)))
3962 		return -EFAULT;
3963 
3964 	if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3965 			!S_ISREG(inode->i_mode))
3966 		return -EINVAL;
3967 
3968 	if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3969 			!f2fs_hw_support_discard(sbi)) ||
3970 			((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3971 			 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3972 		return -EOPNOTSUPP;
3973 
3974 	ret = mnt_want_write_file(filp);
3975 	if (ret)
3976 		return ret;
3977 	inode_lock(inode);
3978 
3979 	if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3980 			range.start >= inode->i_size) {
3981 		ret = -EINVAL;
3982 		goto err;
3983 	}
3984 
3985 	if (range.len == 0)
3986 		goto err;
3987 
3988 	if (inode->i_size - range.start > range.len) {
3989 		end_addr = range.start + range.len;
3990 	} else {
3991 		end_addr = range.len == (u64)-1 ?
3992 			sbi->sb->s_maxbytes : inode->i_size;
3993 		to_end = true;
3994 	}
3995 
3996 	if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3997 			(!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3998 		ret = -EINVAL;
3999 		goto err;
4000 	}
4001 
4002 	index = F2FS_BYTES_TO_BLK(range.start);
4003 	pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
4004 
4005 	ret = f2fs_convert_inline_inode(inode);
4006 	if (ret)
4007 		goto err;
4008 
4009 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4010 	filemap_invalidate_lock(mapping);
4011 
4012 	ret = filemap_write_and_wait_range(mapping, range.start,
4013 			to_end ? LLONG_MAX : end_addr - 1);
4014 	if (ret)
4015 		goto out;
4016 
4017 	truncate_inode_pages_range(mapping, range.start,
4018 			to_end ? -1 : end_addr - 1);
4019 
4020 	while (index < pg_end) {
4021 		struct dnode_of_data dn;
4022 		pgoff_t end_offset, count;
4023 		int i;
4024 
4025 		set_new_dnode(&dn, inode, NULL, NULL, 0);
4026 		ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4027 		if (ret) {
4028 			if (ret == -ENOENT) {
4029 				index = f2fs_get_next_page_offset(&dn, index);
4030 				continue;
4031 			}
4032 			goto out;
4033 		}
4034 
4035 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4036 		count = min(end_offset - dn.ofs_in_node, pg_end - index);
4037 		for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
4038 			struct block_device *cur_bdev;
4039 			block_t blkaddr = f2fs_data_blkaddr(&dn);
4040 
4041 			if (!__is_valid_data_blkaddr(blkaddr))
4042 				continue;
4043 
4044 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
4045 						DATA_GENERIC_ENHANCE)) {
4046 				ret = -EFSCORRUPTED;
4047 				f2fs_put_dnode(&dn);
4048 				goto out;
4049 			}
4050 
4051 			cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
4052 			if (f2fs_is_multi_device(sbi)) {
4053 				int di = f2fs_target_device_index(sbi, blkaddr);
4054 
4055 				blkaddr -= FDEV(di).start_blk;
4056 			}
4057 
4058 			if (len) {
4059 				if (prev_bdev == cur_bdev &&
4060 						index == prev_index + len &&
4061 						blkaddr == prev_block + len) {
4062 					len++;
4063 				} else {
4064 					ret = f2fs_secure_erase(prev_bdev,
4065 						inode, prev_index, prev_block,
4066 						len, range.flags);
4067 					if (ret) {
4068 						f2fs_put_dnode(&dn);
4069 						goto out;
4070 					}
4071 
4072 					len = 0;
4073 				}
4074 			}
4075 
4076 			if (!len) {
4077 				prev_bdev = cur_bdev;
4078 				prev_index = index;
4079 				prev_block = blkaddr;
4080 				len = 1;
4081 			}
4082 		}
4083 
4084 		f2fs_put_dnode(&dn);
4085 
4086 		if (fatal_signal_pending(current)) {
4087 			ret = -EINTR;
4088 			goto out;
4089 		}
4090 		cond_resched();
4091 	}
4092 
4093 	if (len)
4094 		ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
4095 				prev_block, len, range.flags);
4096 	f2fs_update_time(sbi, REQ_TIME);
4097 out:
4098 	filemap_invalidate_unlock(mapping);
4099 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4100 err:
4101 	inode_unlock(inode);
4102 	mnt_drop_write_file(filp);
4103 
4104 	return ret;
4105 }
4106 
4107 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
4108 {
4109 	struct inode *inode = file_inode(filp);
4110 	struct f2fs_comp_option option;
4111 
4112 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4113 		return -EOPNOTSUPP;
4114 
4115 	inode_lock_shared(inode);
4116 
4117 	if (!f2fs_compressed_file(inode)) {
4118 		inode_unlock_shared(inode);
4119 		return -ENODATA;
4120 	}
4121 
4122 	option.algorithm = F2FS_I(inode)->i_compress_algorithm;
4123 	option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
4124 
4125 	inode_unlock_shared(inode);
4126 
4127 	if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
4128 				sizeof(option)))
4129 		return -EFAULT;
4130 
4131 	return 0;
4132 }
4133 
4134 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4135 {
4136 	struct inode *inode = file_inode(filp);
4137 	struct f2fs_inode_info *fi = F2FS_I(inode);
4138 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4139 	struct f2fs_comp_option option;
4140 	int ret = 0;
4141 
4142 	if (!f2fs_sb_has_compression(sbi))
4143 		return -EOPNOTSUPP;
4144 
4145 	if (!(filp->f_mode & FMODE_WRITE))
4146 		return -EBADF;
4147 
4148 	if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4149 				sizeof(option)))
4150 		return -EFAULT;
4151 
4152 	if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
4153 		option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
4154 		option.algorithm >= COMPRESS_MAX)
4155 		return -EINVAL;
4156 
4157 	ret = mnt_want_write_file(filp);
4158 	if (ret)
4159 		return ret;
4160 	inode_lock(inode);
4161 
4162 	f2fs_down_write(&F2FS_I(inode)->i_sem);
4163 	if (!f2fs_compressed_file(inode)) {
4164 		ret = -EINVAL;
4165 		goto out;
4166 	}
4167 
4168 	if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4169 		ret = -EBUSY;
4170 		goto out;
4171 	}
4172 
4173 	if (F2FS_HAS_BLOCKS(inode)) {
4174 		ret = -EFBIG;
4175 		goto out;
4176 	}
4177 
4178 	fi->i_compress_algorithm = option.algorithm;
4179 	fi->i_log_cluster_size = option.log_cluster_size;
4180 	fi->i_cluster_size = BIT(option.log_cluster_size);
4181 	/* Set default level */
4182 	if (fi->i_compress_algorithm == COMPRESS_ZSTD)
4183 		fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
4184 	else
4185 		fi->i_compress_level = 0;
4186 	/* Adjust mount option level */
4187 	if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm &&
4188 	    F2FS_OPTION(sbi).compress_level)
4189 		fi->i_compress_level = F2FS_OPTION(sbi).compress_level;
4190 	f2fs_mark_inode_dirty_sync(inode, true);
4191 
4192 	if (!f2fs_is_compress_backend_ready(inode))
4193 		f2fs_warn(sbi, "compression algorithm is successfully set, "
4194 			"but current kernel doesn't support this algorithm.");
4195 out:
4196 	f2fs_up_write(&fi->i_sem);
4197 	inode_unlock(inode);
4198 	mnt_drop_write_file(filp);
4199 
4200 	return ret;
4201 }
4202 
4203 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4204 {
4205 	DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
4206 	struct address_space *mapping = inode->i_mapping;
4207 	struct page *page;
4208 	pgoff_t redirty_idx = page_idx;
4209 	int i, page_len = 0, ret = 0;
4210 
4211 	page_cache_ra_unbounded(&ractl, len, 0);
4212 
4213 	for (i = 0; i < len; i++, page_idx++) {
4214 		page = read_cache_page(mapping, page_idx, NULL, NULL);
4215 		if (IS_ERR(page)) {
4216 			ret = PTR_ERR(page);
4217 			break;
4218 		}
4219 		page_len++;
4220 	}
4221 
4222 	for (i = 0; i < page_len; i++, redirty_idx++) {
4223 		page = find_lock_page(mapping, redirty_idx);
4224 
4225 		/* It will never fail, when page has pinned above */
4226 		f2fs_bug_on(F2FS_I_SB(inode), !page);
4227 
4228 		f2fs_wait_on_page_writeback(page, DATA, true, true);
4229 
4230 		set_page_dirty(page);
4231 		set_page_private_gcing(page);
4232 		f2fs_put_page(page, 1);
4233 		f2fs_put_page(page, 0);
4234 	}
4235 
4236 	return ret;
4237 }
4238 
4239 static int f2fs_ioc_decompress_file(struct file *filp)
4240 {
4241 	struct inode *inode = file_inode(filp);
4242 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4243 	struct f2fs_inode_info *fi = F2FS_I(inode);
4244 	pgoff_t page_idx = 0, last_idx, cluster_idx;
4245 	int ret;
4246 
4247 	if (!f2fs_sb_has_compression(sbi) ||
4248 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4249 		return -EOPNOTSUPP;
4250 
4251 	if (!(filp->f_mode & FMODE_WRITE))
4252 		return -EBADF;
4253 
4254 	f2fs_balance_fs(sbi, true);
4255 
4256 	ret = mnt_want_write_file(filp);
4257 	if (ret)
4258 		return ret;
4259 	inode_lock(inode);
4260 
4261 	if (!f2fs_is_compress_backend_ready(inode)) {
4262 		ret = -EOPNOTSUPP;
4263 		goto out;
4264 	}
4265 
4266 	if (!f2fs_compressed_file(inode) ||
4267 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4268 		ret = -EINVAL;
4269 		goto out;
4270 	}
4271 
4272 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4273 	if (ret)
4274 		goto out;
4275 
4276 	if (!atomic_read(&fi->i_compr_blocks))
4277 		goto out;
4278 
4279 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4280 	last_idx >>= fi->i_log_cluster_size;
4281 
4282 	for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4283 		page_idx = cluster_idx << fi->i_log_cluster_size;
4284 
4285 		if (!f2fs_is_compressed_cluster(inode, page_idx))
4286 			continue;
4287 
4288 		ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4289 		if (ret < 0)
4290 			break;
4291 
4292 		if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4293 			ret = filemap_fdatawrite(inode->i_mapping);
4294 			if (ret < 0)
4295 				break;
4296 		}
4297 
4298 		cond_resched();
4299 		if (fatal_signal_pending(current)) {
4300 			ret = -EINTR;
4301 			break;
4302 		}
4303 	}
4304 
4305 	if (!ret)
4306 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4307 							LLONG_MAX);
4308 
4309 	if (ret)
4310 		f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4311 			  __func__, ret);
4312 	f2fs_update_time(sbi, REQ_TIME);
4313 out:
4314 	inode_unlock(inode);
4315 	mnt_drop_write_file(filp);
4316 
4317 	return ret;
4318 }
4319 
4320 static int f2fs_ioc_compress_file(struct file *filp)
4321 {
4322 	struct inode *inode = file_inode(filp);
4323 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4324 	struct f2fs_inode_info *fi = F2FS_I(inode);
4325 	pgoff_t page_idx = 0, last_idx, cluster_idx;
4326 	int ret;
4327 
4328 	if (!f2fs_sb_has_compression(sbi) ||
4329 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4330 		return -EOPNOTSUPP;
4331 
4332 	if (!(filp->f_mode & FMODE_WRITE))
4333 		return -EBADF;
4334 
4335 	f2fs_balance_fs(sbi, true);
4336 
4337 	ret = mnt_want_write_file(filp);
4338 	if (ret)
4339 		return ret;
4340 	inode_lock(inode);
4341 
4342 	if (!f2fs_is_compress_backend_ready(inode)) {
4343 		ret = -EOPNOTSUPP;
4344 		goto out;
4345 	}
4346 
4347 	if (!f2fs_compressed_file(inode) ||
4348 		is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4349 		ret = -EINVAL;
4350 		goto out;
4351 	}
4352 
4353 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4354 	if (ret)
4355 		goto out;
4356 
4357 	set_inode_flag(inode, FI_ENABLE_COMPRESS);
4358 
4359 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4360 	last_idx >>= fi->i_log_cluster_size;
4361 
4362 	for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4363 		page_idx = cluster_idx << fi->i_log_cluster_size;
4364 
4365 		if (f2fs_is_sparse_cluster(inode, page_idx))
4366 			continue;
4367 
4368 		ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4369 		if (ret < 0)
4370 			break;
4371 
4372 		if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4373 			ret = filemap_fdatawrite(inode->i_mapping);
4374 			if (ret < 0)
4375 				break;
4376 		}
4377 
4378 		cond_resched();
4379 		if (fatal_signal_pending(current)) {
4380 			ret = -EINTR;
4381 			break;
4382 		}
4383 	}
4384 
4385 	if (!ret)
4386 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4387 							LLONG_MAX);
4388 
4389 	clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4390 
4391 	if (ret)
4392 		f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4393 			  __func__, ret);
4394 	f2fs_update_time(sbi, REQ_TIME);
4395 out:
4396 	inode_unlock(inode);
4397 	mnt_drop_write_file(filp);
4398 
4399 	return ret;
4400 }
4401 
4402 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4403 {
4404 	switch (cmd) {
4405 	case FS_IOC_GETVERSION:
4406 		return f2fs_ioc_getversion(filp, arg);
4407 	case F2FS_IOC_START_ATOMIC_WRITE:
4408 		return f2fs_ioc_start_atomic_write(filp, false);
4409 	case F2FS_IOC_START_ATOMIC_REPLACE:
4410 		return f2fs_ioc_start_atomic_write(filp, true);
4411 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4412 		return f2fs_ioc_commit_atomic_write(filp);
4413 	case F2FS_IOC_ABORT_ATOMIC_WRITE:
4414 		return f2fs_ioc_abort_atomic_write(filp);
4415 	case F2FS_IOC_START_VOLATILE_WRITE:
4416 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4417 		return -EOPNOTSUPP;
4418 	case F2FS_IOC_SHUTDOWN:
4419 		return f2fs_ioc_shutdown(filp, arg);
4420 	case FITRIM:
4421 		return f2fs_ioc_fitrim(filp, arg);
4422 	case FS_IOC_SET_ENCRYPTION_POLICY:
4423 		return f2fs_ioc_set_encryption_policy(filp, arg);
4424 	case FS_IOC_GET_ENCRYPTION_POLICY:
4425 		return f2fs_ioc_get_encryption_policy(filp, arg);
4426 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4427 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4428 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4429 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4430 	case FS_IOC_ADD_ENCRYPTION_KEY:
4431 		return f2fs_ioc_add_encryption_key(filp, arg);
4432 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4433 		return f2fs_ioc_remove_encryption_key(filp, arg);
4434 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4435 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4436 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4437 		return f2fs_ioc_get_encryption_key_status(filp, arg);
4438 	case FS_IOC_GET_ENCRYPTION_NONCE:
4439 		return f2fs_ioc_get_encryption_nonce(filp, arg);
4440 	case F2FS_IOC_GARBAGE_COLLECT:
4441 		return f2fs_ioc_gc(filp, arg);
4442 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4443 		return f2fs_ioc_gc_range(filp, arg);
4444 	case F2FS_IOC_WRITE_CHECKPOINT:
4445 		return f2fs_ioc_write_checkpoint(filp);
4446 	case F2FS_IOC_DEFRAGMENT:
4447 		return f2fs_ioc_defragment(filp, arg);
4448 	case F2FS_IOC_MOVE_RANGE:
4449 		return f2fs_ioc_move_range(filp, arg);
4450 	case F2FS_IOC_FLUSH_DEVICE:
4451 		return f2fs_ioc_flush_device(filp, arg);
4452 	case F2FS_IOC_GET_FEATURES:
4453 		return f2fs_ioc_get_features(filp, arg);
4454 	case F2FS_IOC_GET_PIN_FILE:
4455 		return f2fs_ioc_get_pin_file(filp, arg);
4456 	case F2FS_IOC_SET_PIN_FILE:
4457 		return f2fs_ioc_set_pin_file(filp, arg);
4458 	case F2FS_IOC_PRECACHE_EXTENTS:
4459 		return f2fs_ioc_precache_extents(filp);
4460 	case F2FS_IOC_RESIZE_FS:
4461 		return f2fs_ioc_resize_fs(filp, arg);
4462 	case FS_IOC_ENABLE_VERITY:
4463 		return f2fs_ioc_enable_verity(filp, arg);
4464 	case FS_IOC_MEASURE_VERITY:
4465 		return f2fs_ioc_measure_verity(filp, arg);
4466 	case FS_IOC_READ_VERITY_METADATA:
4467 		return f2fs_ioc_read_verity_metadata(filp, arg);
4468 	case FS_IOC_GETFSLABEL:
4469 		return f2fs_ioc_getfslabel(filp, arg);
4470 	case FS_IOC_SETFSLABEL:
4471 		return f2fs_ioc_setfslabel(filp, arg);
4472 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4473 		return f2fs_ioc_get_compress_blocks(filp, arg);
4474 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4475 		return f2fs_release_compress_blocks(filp, arg);
4476 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4477 		return f2fs_reserve_compress_blocks(filp, arg);
4478 	case F2FS_IOC_SEC_TRIM_FILE:
4479 		return f2fs_sec_trim_file(filp, arg);
4480 	case F2FS_IOC_GET_COMPRESS_OPTION:
4481 		return f2fs_ioc_get_compress_option(filp, arg);
4482 	case F2FS_IOC_SET_COMPRESS_OPTION:
4483 		return f2fs_ioc_set_compress_option(filp, arg);
4484 	case F2FS_IOC_DECOMPRESS_FILE:
4485 		return f2fs_ioc_decompress_file(filp);
4486 	case F2FS_IOC_COMPRESS_FILE:
4487 		return f2fs_ioc_compress_file(filp);
4488 	default:
4489 		return -ENOTTY;
4490 	}
4491 }
4492 
4493 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4494 {
4495 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4496 		return -EIO;
4497 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4498 		return -ENOSPC;
4499 
4500 	return __f2fs_ioctl(filp, cmd, arg);
4501 }
4502 
4503 /*
4504  * Return %true if the given read or write request should use direct I/O, or
4505  * %false if it should use buffered I/O.
4506  */
4507 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4508 				struct iov_iter *iter)
4509 {
4510 	unsigned int align;
4511 
4512 	if (!(iocb->ki_flags & IOCB_DIRECT))
4513 		return false;
4514 
4515 	if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4516 		return false;
4517 
4518 	/*
4519 	 * Direct I/O not aligned to the disk's logical_block_size will be
4520 	 * attempted, but will fail with -EINVAL.
4521 	 *
4522 	 * f2fs additionally requires that direct I/O be aligned to the
4523 	 * filesystem block size, which is often a stricter requirement.
4524 	 * However, f2fs traditionally falls back to buffered I/O on requests
4525 	 * that are logical_block_size-aligned but not fs-block aligned.
4526 	 *
4527 	 * The below logic implements this behavior.
4528 	 */
4529 	align = iocb->ki_pos | iov_iter_alignment(iter);
4530 	if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4531 	    IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4532 		return false;
4533 
4534 	return true;
4535 }
4536 
4537 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4538 				unsigned int flags)
4539 {
4540 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4541 
4542 	dec_page_count(sbi, F2FS_DIO_READ);
4543 	if (error)
4544 		return error;
4545 	f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4546 	return 0;
4547 }
4548 
4549 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4550 	.end_io = f2fs_dio_read_end_io,
4551 };
4552 
4553 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4554 {
4555 	struct file *file = iocb->ki_filp;
4556 	struct inode *inode = file_inode(file);
4557 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4558 	struct f2fs_inode_info *fi = F2FS_I(inode);
4559 	const loff_t pos = iocb->ki_pos;
4560 	const size_t count = iov_iter_count(to);
4561 	struct iomap_dio *dio;
4562 	ssize_t ret;
4563 
4564 	if (count == 0)
4565 		return 0; /* skip atime update */
4566 
4567 	trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4568 
4569 	if (iocb->ki_flags & IOCB_NOWAIT) {
4570 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4571 			ret = -EAGAIN;
4572 			goto out;
4573 		}
4574 	} else {
4575 		f2fs_down_read(&fi->i_gc_rwsem[READ]);
4576 	}
4577 
4578 	/* dio is not compatible w/ atomic file */
4579 	if (f2fs_is_atomic_file(inode)) {
4580 		f2fs_up_read(&fi->i_gc_rwsem[READ]);
4581 		ret = -EOPNOTSUPP;
4582 		goto out;
4583 	}
4584 
4585 	/*
4586 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4587 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4588 	 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4589 	 */
4590 	inc_page_count(sbi, F2FS_DIO_READ);
4591 	dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4592 			     &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4593 	if (IS_ERR_OR_NULL(dio)) {
4594 		ret = PTR_ERR_OR_ZERO(dio);
4595 		if (ret != -EIOCBQUEUED)
4596 			dec_page_count(sbi, F2FS_DIO_READ);
4597 	} else {
4598 		ret = iomap_dio_complete(dio);
4599 	}
4600 
4601 	f2fs_up_read(&fi->i_gc_rwsem[READ]);
4602 
4603 	file_accessed(file);
4604 out:
4605 	trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4606 	return ret;
4607 }
4608 
4609 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count,
4610 				    int rw)
4611 {
4612 	struct inode *inode = file_inode(file);
4613 	char *buf, *path;
4614 
4615 	buf = f2fs_getname(F2FS_I_SB(inode));
4616 	if (!buf)
4617 		return;
4618 	path = dentry_path_raw(file_dentry(file), buf, PATH_MAX);
4619 	if (IS_ERR(path))
4620 		goto free_buf;
4621 	if (rw == WRITE)
4622 		trace_f2fs_datawrite_start(inode, pos, count,
4623 				current->pid, path, current->comm);
4624 	else
4625 		trace_f2fs_dataread_start(inode, pos, count,
4626 				current->pid, path, current->comm);
4627 free_buf:
4628 	f2fs_putname(buf);
4629 }
4630 
4631 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4632 {
4633 	struct inode *inode = file_inode(iocb->ki_filp);
4634 	const loff_t pos = iocb->ki_pos;
4635 	ssize_t ret;
4636 
4637 	if (!f2fs_is_compress_backend_ready(inode))
4638 		return -EOPNOTSUPP;
4639 
4640 	if (trace_f2fs_dataread_start_enabled())
4641 		f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
4642 					iov_iter_count(to), READ);
4643 
4644 	/* In LFS mode, if there is inflight dio, wait for its completion */
4645 	if (f2fs_lfs_mode(F2FS_I_SB(inode)) &&
4646 	    get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE))
4647 		inode_dio_wait(inode);
4648 
4649 	if (f2fs_should_use_dio(inode, iocb, to)) {
4650 		ret = f2fs_dio_read_iter(iocb, to);
4651 	} else {
4652 		ret = filemap_read(iocb, to, 0);
4653 		if (ret > 0)
4654 			f2fs_update_iostat(F2FS_I_SB(inode), inode,
4655 						APP_BUFFERED_READ_IO, ret);
4656 	}
4657 	if (trace_f2fs_dataread_end_enabled())
4658 		trace_f2fs_dataread_end(inode, pos, ret);
4659 	return ret;
4660 }
4661 
4662 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos,
4663 				     struct pipe_inode_info *pipe,
4664 				     size_t len, unsigned int flags)
4665 {
4666 	struct inode *inode = file_inode(in);
4667 	const loff_t pos = *ppos;
4668 	ssize_t ret;
4669 
4670 	if (!f2fs_is_compress_backend_ready(inode))
4671 		return -EOPNOTSUPP;
4672 
4673 	if (trace_f2fs_dataread_start_enabled())
4674 		f2fs_trace_rw_file_path(in, pos, len, READ);
4675 
4676 	ret = filemap_splice_read(in, ppos, pipe, len, flags);
4677 	if (ret > 0)
4678 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
4679 				   APP_BUFFERED_READ_IO, ret);
4680 
4681 	if (trace_f2fs_dataread_end_enabled())
4682 		trace_f2fs_dataread_end(inode, pos, ret);
4683 	return ret;
4684 }
4685 
4686 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4687 {
4688 	struct file *file = iocb->ki_filp;
4689 	struct inode *inode = file_inode(file);
4690 	ssize_t count;
4691 	int err;
4692 
4693 	if (IS_IMMUTABLE(inode))
4694 		return -EPERM;
4695 
4696 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4697 		return -EPERM;
4698 
4699 	count = generic_write_checks(iocb, from);
4700 	if (count <= 0)
4701 		return count;
4702 
4703 	err = file_modified(file);
4704 	if (err)
4705 		return err;
4706 	return count;
4707 }
4708 
4709 /*
4710  * Preallocate blocks for a write request, if it is possible and helpful to do
4711  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4712  * blocks were preallocated, or a negative errno value if something went
4713  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4714  * requested blocks (not just some of them) have been allocated.
4715  */
4716 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4717 				   bool dio)
4718 {
4719 	struct inode *inode = file_inode(iocb->ki_filp);
4720 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4721 	const loff_t pos = iocb->ki_pos;
4722 	const size_t count = iov_iter_count(iter);
4723 	struct f2fs_map_blocks map = {};
4724 	int flag;
4725 	int ret;
4726 
4727 	/* If it will be an out-of-place direct write, don't bother. */
4728 	if (dio && f2fs_lfs_mode(sbi))
4729 		return 0;
4730 	/*
4731 	 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4732 	 * buffered IO, if DIO meets any holes.
4733 	 */
4734 	if (dio && i_size_read(inode) &&
4735 		(F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4736 		return 0;
4737 
4738 	/* No-wait I/O can't allocate blocks. */
4739 	if (iocb->ki_flags & IOCB_NOWAIT)
4740 		return 0;
4741 
4742 	/* If it will be a short write, don't bother. */
4743 	if (fault_in_iov_iter_readable(iter, count))
4744 		return 0;
4745 
4746 	if (f2fs_has_inline_data(inode)) {
4747 		/* If the data will fit inline, don't bother. */
4748 		if (pos + count <= MAX_INLINE_DATA(inode))
4749 			return 0;
4750 		ret = f2fs_convert_inline_inode(inode);
4751 		if (ret)
4752 			return ret;
4753 	}
4754 
4755 	/* Do not preallocate blocks that will be written partially in 4KB. */
4756 	map.m_lblk = F2FS_BLK_ALIGN(pos);
4757 	map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4758 	if (map.m_len > map.m_lblk)
4759 		map.m_len -= map.m_lblk;
4760 	else
4761 		return 0;
4762 
4763 	map.m_may_create = true;
4764 	if (dio) {
4765 		map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi,
4766 						inode->i_write_hint);
4767 		flag = F2FS_GET_BLOCK_PRE_DIO;
4768 	} else {
4769 		map.m_seg_type = NO_CHECK_TYPE;
4770 		flag = F2FS_GET_BLOCK_PRE_AIO;
4771 	}
4772 
4773 	ret = f2fs_map_blocks(inode, &map, flag);
4774 	/* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4775 	if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4776 		return ret;
4777 	if (ret == 0)
4778 		set_inode_flag(inode, FI_PREALLOCATED_ALL);
4779 	return map.m_len;
4780 }
4781 
4782 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4783 					struct iov_iter *from)
4784 {
4785 	struct file *file = iocb->ki_filp;
4786 	struct inode *inode = file_inode(file);
4787 	ssize_t ret;
4788 
4789 	if (iocb->ki_flags & IOCB_NOWAIT)
4790 		return -EOPNOTSUPP;
4791 
4792 	ret = generic_perform_write(iocb, from);
4793 
4794 	if (ret > 0) {
4795 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
4796 						APP_BUFFERED_IO, ret);
4797 	}
4798 	return ret;
4799 }
4800 
4801 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4802 				 unsigned int flags)
4803 {
4804 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4805 
4806 	dec_page_count(sbi, F2FS_DIO_WRITE);
4807 	if (error)
4808 		return error;
4809 	f2fs_update_time(sbi, REQ_TIME);
4810 	f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4811 	return 0;
4812 }
4813 
4814 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter,
4815 					struct bio *bio, loff_t file_offset)
4816 {
4817 	struct inode *inode = iter->inode;
4818 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4819 	int seg_type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint);
4820 	enum temp_type temp = f2fs_get_segment_temp(seg_type);
4821 
4822 	bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp);
4823 	submit_bio(bio);
4824 }
4825 
4826 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4827 	.end_io		= f2fs_dio_write_end_io,
4828 	.submit_io	= f2fs_dio_write_submit_io,
4829 };
4830 
4831 static void f2fs_flush_buffered_write(struct address_space *mapping,
4832 				      loff_t start_pos, loff_t end_pos)
4833 {
4834 	int ret;
4835 
4836 	ret = filemap_write_and_wait_range(mapping, start_pos, end_pos);
4837 	if (ret < 0)
4838 		return;
4839 	invalidate_mapping_pages(mapping,
4840 				 start_pos >> PAGE_SHIFT,
4841 				 end_pos >> PAGE_SHIFT);
4842 }
4843 
4844 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4845 				   bool *may_need_sync)
4846 {
4847 	struct file *file = iocb->ki_filp;
4848 	struct inode *inode = file_inode(file);
4849 	struct f2fs_inode_info *fi = F2FS_I(inode);
4850 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4851 	const bool do_opu = f2fs_lfs_mode(sbi);
4852 	const loff_t pos = iocb->ki_pos;
4853 	const ssize_t count = iov_iter_count(from);
4854 	unsigned int dio_flags;
4855 	struct iomap_dio *dio;
4856 	ssize_t ret;
4857 
4858 	trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4859 
4860 	if (iocb->ki_flags & IOCB_NOWAIT) {
4861 		/* f2fs_convert_inline_inode() and block allocation can block */
4862 		if (f2fs_has_inline_data(inode) ||
4863 		    !f2fs_overwrite_io(inode, pos, count)) {
4864 			ret = -EAGAIN;
4865 			goto out;
4866 		}
4867 
4868 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4869 			ret = -EAGAIN;
4870 			goto out;
4871 		}
4872 		if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4873 			f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4874 			ret = -EAGAIN;
4875 			goto out;
4876 		}
4877 	} else {
4878 		ret = f2fs_convert_inline_inode(inode);
4879 		if (ret)
4880 			goto out;
4881 
4882 		f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4883 		if (do_opu)
4884 			f2fs_down_read(&fi->i_gc_rwsem[READ]);
4885 	}
4886 
4887 	/*
4888 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4889 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4890 	 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4891 	 */
4892 	inc_page_count(sbi, F2FS_DIO_WRITE);
4893 	dio_flags = 0;
4894 	if (pos + count > inode->i_size)
4895 		dio_flags |= IOMAP_DIO_FORCE_WAIT;
4896 	dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4897 			     &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4898 	if (IS_ERR_OR_NULL(dio)) {
4899 		ret = PTR_ERR_OR_ZERO(dio);
4900 		if (ret == -ENOTBLK)
4901 			ret = 0;
4902 		if (ret != -EIOCBQUEUED)
4903 			dec_page_count(sbi, F2FS_DIO_WRITE);
4904 	} else {
4905 		ret = iomap_dio_complete(dio);
4906 	}
4907 
4908 	if (do_opu)
4909 		f2fs_up_read(&fi->i_gc_rwsem[READ]);
4910 	f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4911 
4912 	if (ret < 0)
4913 		goto out;
4914 	if (pos + ret > inode->i_size)
4915 		f2fs_i_size_write(inode, pos + ret);
4916 	if (!do_opu)
4917 		set_inode_flag(inode, FI_UPDATE_WRITE);
4918 
4919 	if (iov_iter_count(from)) {
4920 		ssize_t ret2;
4921 		loff_t bufio_start_pos = iocb->ki_pos;
4922 
4923 		/*
4924 		 * The direct write was partial, so we need to fall back to a
4925 		 * buffered write for the remainder.
4926 		 */
4927 
4928 		ret2 = f2fs_buffered_write_iter(iocb, from);
4929 		if (iov_iter_count(from))
4930 			f2fs_write_failed(inode, iocb->ki_pos);
4931 		if (ret2 < 0)
4932 			goto out;
4933 
4934 		/*
4935 		 * Ensure that the pagecache pages are written to disk and
4936 		 * invalidated to preserve the expected O_DIRECT semantics.
4937 		 */
4938 		if (ret2 > 0) {
4939 			loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4940 
4941 			ret += ret2;
4942 
4943 			f2fs_flush_buffered_write(file->f_mapping,
4944 						  bufio_start_pos,
4945 						  bufio_end_pos);
4946 		}
4947 	} else {
4948 		/* iomap_dio_rw() already handled the generic_write_sync(). */
4949 		*may_need_sync = false;
4950 	}
4951 out:
4952 	trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4953 	return ret;
4954 }
4955 
4956 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4957 {
4958 	struct inode *inode = file_inode(iocb->ki_filp);
4959 	const loff_t orig_pos = iocb->ki_pos;
4960 	const size_t orig_count = iov_iter_count(from);
4961 	loff_t target_size;
4962 	bool dio;
4963 	bool may_need_sync = true;
4964 	int preallocated;
4965 	const loff_t pos = iocb->ki_pos;
4966 	const ssize_t count = iov_iter_count(from);
4967 	ssize_t ret;
4968 
4969 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4970 		ret = -EIO;
4971 		goto out;
4972 	}
4973 
4974 	if (!f2fs_is_compress_backend_ready(inode)) {
4975 		ret = -EOPNOTSUPP;
4976 		goto out;
4977 	}
4978 
4979 	if (iocb->ki_flags & IOCB_NOWAIT) {
4980 		if (!inode_trylock(inode)) {
4981 			ret = -EAGAIN;
4982 			goto out;
4983 		}
4984 	} else {
4985 		inode_lock(inode);
4986 	}
4987 
4988 	if (f2fs_is_pinned_file(inode) &&
4989 	    !f2fs_overwrite_io(inode, pos, count)) {
4990 		ret = -EIO;
4991 		goto out_unlock;
4992 	}
4993 
4994 	ret = f2fs_write_checks(iocb, from);
4995 	if (ret <= 0)
4996 		goto out_unlock;
4997 
4998 	/* Determine whether we will do a direct write or a buffered write. */
4999 	dio = f2fs_should_use_dio(inode, iocb, from);
5000 
5001 	/* dio is not compatible w/ atomic write */
5002 	if (dio && f2fs_is_atomic_file(inode)) {
5003 		ret = -EOPNOTSUPP;
5004 		goto out_unlock;
5005 	}
5006 
5007 	/* Possibly preallocate the blocks for the write. */
5008 	target_size = iocb->ki_pos + iov_iter_count(from);
5009 	preallocated = f2fs_preallocate_blocks(iocb, from, dio);
5010 	if (preallocated < 0) {
5011 		ret = preallocated;
5012 	} else {
5013 		if (trace_f2fs_datawrite_start_enabled())
5014 			f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
5015 						orig_count, WRITE);
5016 
5017 		/* Do the actual write. */
5018 		ret = dio ?
5019 			f2fs_dio_write_iter(iocb, from, &may_need_sync) :
5020 			f2fs_buffered_write_iter(iocb, from);
5021 
5022 		if (trace_f2fs_datawrite_end_enabled())
5023 			trace_f2fs_datawrite_end(inode, orig_pos, ret);
5024 	}
5025 
5026 	/* Don't leave any preallocated blocks around past i_size. */
5027 	if (preallocated && i_size_read(inode) < target_size) {
5028 		f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5029 		filemap_invalidate_lock(inode->i_mapping);
5030 		if (!f2fs_truncate(inode))
5031 			file_dont_truncate(inode);
5032 		filemap_invalidate_unlock(inode->i_mapping);
5033 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5034 	} else {
5035 		file_dont_truncate(inode);
5036 	}
5037 
5038 	clear_inode_flag(inode, FI_PREALLOCATED_ALL);
5039 out_unlock:
5040 	inode_unlock(inode);
5041 out:
5042 	trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
5043 
5044 	if (ret > 0 && may_need_sync)
5045 		ret = generic_write_sync(iocb, ret);
5046 
5047 	/* If buffered IO was forced, flush and drop the data from
5048 	 * the page cache to preserve O_DIRECT semantics
5049 	 */
5050 	if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
5051 		f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
5052 					  orig_pos,
5053 					  orig_pos + ret - 1);
5054 
5055 	return ret;
5056 }
5057 
5058 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
5059 		int advice)
5060 {
5061 	struct address_space *mapping;
5062 	struct backing_dev_info *bdi;
5063 	struct inode *inode = file_inode(filp);
5064 	int err;
5065 
5066 	if (advice == POSIX_FADV_SEQUENTIAL) {
5067 		if (S_ISFIFO(inode->i_mode))
5068 			return -ESPIPE;
5069 
5070 		mapping = filp->f_mapping;
5071 		if (!mapping || len < 0)
5072 			return -EINVAL;
5073 
5074 		bdi = inode_to_bdi(mapping->host);
5075 		filp->f_ra.ra_pages = bdi->ra_pages *
5076 			F2FS_I_SB(inode)->seq_file_ra_mul;
5077 		spin_lock(&filp->f_lock);
5078 		filp->f_mode &= ~FMODE_RANDOM;
5079 		spin_unlock(&filp->f_lock);
5080 		return 0;
5081 	} else if (advice == POSIX_FADV_WILLNEED && offset == 0) {
5082 		/* Load extent cache at the first readahead. */
5083 		f2fs_precache_extents(inode);
5084 	}
5085 
5086 	err = generic_fadvise(filp, offset, len, advice);
5087 	if (!err && advice == POSIX_FADV_DONTNEED &&
5088 		test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
5089 		f2fs_compressed_file(inode))
5090 		f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
5091 
5092 	return err;
5093 }
5094 
5095 #ifdef CONFIG_COMPAT
5096 struct compat_f2fs_gc_range {
5097 	u32 sync;
5098 	compat_u64 start;
5099 	compat_u64 len;
5100 };
5101 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE	_IOW(F2FS_IOCTL_MAGIC, 11,\
5102 						struct compat_f2fs_gc_range)
5103 
5104 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
5105 {
5106 	struct compat_f2fs_gc_range __user *urange;
5107 	struct f2fs_gc_range range;
5108 	int err;
5109 
5110 	urange = compat_ptr(arg);
5111 	err = get_user(range.sync, &urange->sync);
5112 	err |= get_user(range.start, &urange->start);
5113 	err |= get_user(range.len, &urange->len);
5114 	if (err)
5115 		return -EFAULT;
5116 
5117 	return __f2fs_ioc_gc_range(file, &range);
5118 }
5119 
5120 struct compat_f2fs_move_range {
5121 	u32 dst_fd;
5122 	compat_u64 pos_in;
5123 	compat_u64 pos_out;
5124 	compat_u64 len;
5125 };
5126 #define F2FS_IOC32_MOVE_RANGE		_IOWR(F2FS_IOCTL_MAGIC, 9,	\
5127 					struct compat_f2fs_move_range)
5128 
5129 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
5130 {
5131 	struct compat_f2fs_move_range __user *urange;
5132 	struct f2fs_move_range range;
5133 	int err;
5134 
5135 	urange = compat_ptr(arg);
5136 	err = get_user(range.dst_fd, &urange->dst_fd);
5137 	err |= get_user(range.pos_in, &urange->pos_in);
5138 	err |= get_user(range.pos_out, &urange->pos_out);
5139 	err |= get_user(range.len, &urange->len);
5140 	if (err)
5141 		return -EFAULT;
5142 
5143 	return __f2fs_ioc_move_range(file, &range);
5144 }
5145 
5146 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5147 {
5148 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
5149 		return -EIO;
5150 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
5151 		return -ENOSPC;
5152 
5153 	switch (cmd) {
5154 	case FS_IOC32_GETVERSION:
5155 		cmd = FS_IOC_GETVERSION;
5156 		break;
5157 	case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
5158 		return f2fs_compat_ioc_gc_range(file, arg);
5159 	case F2FS_IOC32_MOVE_RANGE:
5160 		return f2fs_compat_ioc_move_range(file, arg);
5161 	case F2FS_IOC_START_ATOMIC_WRITE:
5162 	case F2FS_IOC_START_ATOMIC_REPLACE:
5163 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
5164 	case F2FS_IOC_START_VOLATILE_WRITE:
5165 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
5166 	case F2FS_IOC_ABORT_ATOMIC_WRITE:
5167 	case F2FS_IOC_SHUTDOWN:
5168 	case FITRIM:
5169 	case FS_IOC_SET_ENCRYPTION_POLICY:
5170 	case FS_IOC_GET_ENCRYPTION_PWSALT:
5171 	case FS_IOC_GET_ENCRYPTION_POLICY:
5172 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
5173 	case FS_IOC_ADD_ENCRYPTION_KEY:
5174 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
5175 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
5176 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
5177 	case FS_IOC_GET_ENCRYPTION_NONCE:
5178 	case F2FS_IOC_GARBAGE_COLLECT:
5179 	case F2FS_IOC_WRITE_CHECKPOINT:
5180 	case F2FS_IOC_DEFRAGMENT:
5181 	case F2FS_IOC_FLUSH_DEVICE:
5182 	case F2FS_IOC_GET_FEATURES:
5183 	case F2FS_IOC_GET_PIN_FILE:
5184 	case F2FS_IOC_SET_PIN_FILE:
5185 	case F2FS_IOC_PRECACHE_EXTENTS:
5186 	case F2FS_IOC_RESIZE_FS:
5187 	case FS_IOC_ENABLE_VERITY:
5188 	case FS_IOC_MEASURE_VERITY:
5189 	case FS_IOC_READ_VERITY_METADATA:
5190 	case FS_IOC_GETFSLABEL:
5191 	case FS_IOC_SETFSLABEL:
5192 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
5193 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
5194 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
5195 	case F2FS_IOC_SEC_TRIM_FILE:
5196 	case F2FS_IOC_GET_COMPRESS_OPTION:
5197 	case F2FS_IOC_SET_COMPRESS_OPTION:
5198 	case F2FS_IOC_DECOMPRESS_FILE:
5199 	case F2FS_IOC_COMPRESS_FILE:
5200 		break;
5201 	default:
5202 		return -ENOIOCTLCMD;
5203 	}
5204 	return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5205 }
5206 #endif
5207 
5208 const struct file_operations f2fs_file_operations = {
5209 	.llseek		= f2fs_llseek,
5210 	.read_iter	= f2fs_file_read_iter,
5211 	.write_iter	= f2fs_file_write_iter,
5212 	.iopoll		= iocb_bio_iopoll,
5213 	.open		= f2fs_file_open,
5214 	.release	= f2fs_release_file,
5215 	.mmap		= f2fs_file_mmap,
5216 	.flush		= f2fs_file_flush,
5217 	.fsync		= f2fs_sync_file,
5218 	.fallocate	= f2fs_fallocate,
5219 	.unlocked_ioctl	= f2fs_ioctl,
5220 #ifdef CONFIG_COMPAT
5221 	.compat_ioctl	= f2fs_compat_ioctl,
5222 #endif
5223 	.splice_read	= f2fs_file_splice_read,
5224 	.splice_write	= iter_file_splice_write,
5225 	.fadvise	= f2fs_file_fadvise,
5226 	.fop_flags	= FOP_BUFFER_RASYNC,
5227 };
5228