• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/file.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/buffer_head.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/types.h>
16 #include <linux/compat.h>
17 #include <linux/uaccess.h>
18 #include <linux/mount.h>
19 #include <linux/pagevec.h>
20 #include <linux/uio.h>
21 #include <linux/uuid.h>
22 #include <linux/file.h>
23 #include <linux/nls.h>
24 #include <linux/sched/signal.h>
25 #include <linux/fileattr.h>
26 #include <linux/fadvise.h>
27 #include <linux/iomap.h>
28 
29 #include "f2fs.h"
30 #include "node.h"
31 #include "segment.h"
32 #include "xattr.h"
33 #include "acl.h"
34 #include "gc.h"
35 #include "iostat.h"
36 #include <trace/events/f2fs.h>
37 #include <uapi/linux/f2fs.h>
38 
f2fs_filemap_fault(struct vm_fault * vmf)39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
40 {
41 	struct inode *inode = file_inode(vmf->vma->vm_file);
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, (unsigned long)ret);
50 
51 	return ret;
52 }
53 
f2fs_vm_page_mkwrite(struct vm_fault * vmf)54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
55 {
56 	struct page *page = 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 = true;
61 	int err = 0;
62 
63 	if (unlikely(IS_IMMUTABLE(inode)))
64 		return VM_FAULT_SIGBUS;
65 
66 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
67 		return VM_FAULT_SIGBUS;
68 
69 	if (unlikely(f2fs_cp_error(sbi))) {
70 		err = -EIO;
71 		goto err;
72 	}
73 
74 	if (!f2fs_is_checkpoint_ready(sbi)) {
75 		err = -ENOSPC;
76 		goto err;
77 	}
78 
79 	err = f2fs_convert_inline_inode(inode);
80 	if (err)
81 		goto err;
82 
83 #ifdef CONFIG_F2FS_FS_COMPRESSION
84 	if (f2fs_compressed_file(inode)) {
85 		int ret = f2fs_is_compressed_cluster(inode, page->index);
86 
87 		if (ret < 0) {
88 			err = ret;
89 			goto err;
90 		} else if (ret) {
91 			need_alloc = false;
92 		}
93 	}
94 #endif
95 	/* should do out of any locked page */
96 	if (need_alloc)
97 		f2fs_balance_fs(sbi, true);
98 
99 	sb_start_pagefault(inode->i_sb);
100 
101 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
102 
103 	file_update_time(vmf->vma->vm_file);
104 	filemap_invalidate_lock_shared(inode->i_mapping);
105 	lock_page(page);
106 	if (unlikely(page->mapping != inode->i_mapping ||
107 			page_offset(page) > i_size_read(inode) ||
108 			!PageUptodate(page))) {
109 		unlock_page(page);
110 		err = -EFAULT;
111 		goto out_sem;
112 	}
113 
114 	if (need_alloc) {
115 		/* block allocation */
116 		set_new_dnode(&dn, inode, NULL, NULL, 0);
117 		err = f2fs_get_block_locked(&dn, page->index);
118 	}
119 
120 #ifdef CONFIG_F2FS_FS_COMPRESSION
121 	if (!need_alloc) {
122 		set_new_dnode(&dn, inode, NULL, NULL, 0);
123 		err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
124 		f2fs_put_dnode(&dn);
125 	}
126 #endif
127 	if (err) {
128 		unlock_page(page);
129 		goto out_sem;
130 	}
131 
132 	f2fs_wait_on_page_writeback(page, DATA, false, true);
133 
134 	/* wait for GCed page writeback via META_MAPPING */
135 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
136 
137 	/*
138 	 * check to see if the page is mapped already (no holes)
139 	 */
140 	if (PageMappedToDisk(page))
141 		goto out_sem;
142 
143 	/* page is wholly or partially inside EOF */
144 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
145 						i_size_read(inode)) {
146 		loff_t offset;
147 
148 		offset = i_size_read(inode) & ~PAGE_MASK;
149 		zero_user_segment(page, offset, PAGE_SIZE);
150 	}
151 	set_page_dirty(page);
152 	if (!PageUptodate(page))
153 		SetPageUptodate(page);
154 
155 	f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
156 	f2fs_update_time(sbi, REQ_TIME);
157 
158 	trace_f2fs_vm_page_mkwrite(page, DATA);
159 out_sem:
160 	filemap_invalidate_unlock_shared(inode->i_mapping);
161 
162 	sb_end_pagefault(inode->i_sb);
163 err:
164 	return block_page_mkwrite_return(err);
165 }
166 
167 static const struct vm_operations_struct f2fs_file_vm_ops = {
168 	.fault		= f2fs_filemap_fault,
169 	.map_pages	= filemap_map_pages,
170 	.page_mkwrite	= f2fs_vm_page_mkwrite,
171 };
172 
get_parent_ino(struct inode * inode,nid_t * pino)173 static int get_parent_ino(struct inode *inode, nid_t *pino)
174 {
175 	struct dentry *dentry;
176 
177 	/*
178 	 * Make sure to get the non-deleted alias.  The alias associated with
179 	 * the open file descriptor being fsync()'ed may be deleted already.
180 	 */
181 	dentry = d_find_alias(inode);
182 	if (!dentry)
183 		return 0;
184 
185 	*pino = parent_ino(dentry);
186 	dput(dentry);
187 	return 1;
188 }
189 
need_do_checkpoint(struct inode * inode)190 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
191 {
192 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
193 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
194 
195 	if (!S_ISREG(inode->i_mode))
196 		cp_reason = CP_NON_REGULAR;
197 	else if (f2fs_compressed_file(inode))
198 		cp_reason = CP_COMPRESSED;
199 	else if (inode->i_nlink != 1)
200 		cp_reason = CP_HARDLINK;
201 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
202 		cp_reason = CP_SB_NEED_CP;
203 	else if (file_wrong_pino(inode))
204 		cp_reason = CP_WRONG_PINO;
205 	else if (!f2fs_space_for_roll_forward(sbi))
206 		cp_reason = CP_NO_SPC_ROLL;
207 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
208 		cp_reason = CP_NODE_NEED_CP;
209 	else if (test_opt(sbi, FASTBOOT))
210 		cp_reason = CP_FASTBOOT_MODE;
211 	else if (F2FS_OPTION(sbi).active_logs == 2)
212 		cp_reason = CP_SPEC_LOG_NUM;
213 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
214 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
215 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
216 							TRANS_DIR_INO))
217 		cp_reason = CP_RECOVER_DIR;
218 
219 	return cp_reason;
220 }
221 
need_inode_page_update(struct f2fs_sb_info * sbi,nid_t ino)222 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
223 {
224 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
225 	bool ret = false;
226 	/* But we need to avoid that there are some inode updates */
227 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
228 		ret = true;
229 	f2fs_put_page(i, 0);
230 	return ret;
231 }
232 
try_to_fix_pino(struct inode * inode)233 static void try_to_fix_pino(struct inode *inode)
234 {
235 	struct f2fs_inode_info *fi = F2FS_I(inode);
236 	nid_t pino;
237 
238 	f2fs_down_write(&fi->i_sem);
239 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
240 			get_parent_ino(inode, &pino)) {
241 		f2fs_i_pino_write(inode, pino);
242 		file_got_pino(inode);
243 	}
244 	f2fs_up_write(&fi->i_sem);
245 }
246 
f2fs_do_sync_file(struct file * file,loff_t start,loff_t end,int datasync,bool atomic)247 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
248 						int datasync, bool atomic)
249 {
250 	struct inode *inode = file->f_mapping->host;
251 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
252 	nid_t ino = inode->i_ino;
253 	int ret = 0;
254 	enum cp_reason_type cp_reason = 0;
255 	struct writeback_control wbc = {
256 		.sync_mode = WB_SYNC_ALL,
257 		.nr_to_write = LONG_MAX,
258 		.for_reclaim = 0,
259 	};
260 	unsigned int seq_id = 0;
261 
262 	if (unlikely(f2fs_readonly(inode->i_sb)))
263 		return 0;
264 
265 	trace_f2fs_sync_file_enter(inode);
266 
267 	if (S_ISDIR(inode->i_mode))
268 		goto go_write;
269 
270 	/* if fdatasync is triggered, let's do in-place-update */
271 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
272 		set_inode_flag(inode, FI_NEED_IPU);
273 	ret = file_write_and_wait_range(file, start, end);
274 	clear_inode_flag(inode, FI_NEED_IPU);
275 
276 	if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
277 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
278 		return ret;
279 	}
280 
281 	/* if the inode is dirty, let's recover all the time */
282 	if (!f2fs_skip_inode_update(inode, datasync)) {
283 		f2fs_write_inode(inode, NULL);
284 		goto go_write;
285 	}
286 
287 	/*
288 	 * if there is no written data, don't waste time to write recovery info.
289 	 */
290 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
291 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
292 
293 		/* it may call write_inode just prior to fsync */
294 		if (need_inode_page_update(sbi, ino))
295 			goto go_write;
296 
297 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
298 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
299 			goto flush_out;
300 		goto out;
301 	} else {
302 		/*
303 		 * for OPU case, during fsync(), node can be persisted before
304 		 * data when lower device doesn't support write barrier, result
305 		 * in data corruption after SPO.
306 		 * So for strict fsync mode, force to use atomic write semantics
307 		 * to keep write order in between data/node and last node to
308 		 * avoid potential data corruption.
309 		 */
310 		if (F2FS_OPTION(sbi).fsync_mode ==
311 				FSYNC_MODE_STRICT && !atomic)
312 			atomic = true;
313 	}
314 go_write:
315 	/*
316 	 * Both of fdatasync() and fsync() are able to be recovered from
317 	 * sudden-power-off.
318 	 */
319 	f2fs_down_read(&F2FS_I(inode)->i_sem);
320 	cp_reason = need_do_checkpoint(inode);
321 	f2fs_up_read(&F2FS_I(inode)->i_sem);
322 
323 	if (cp_reason) {
324 		/* all the dirty node pages should be flushed for POR */
325 		ret = f2fs_sync_fs(inode->i_sb, 1);
326 
327 		/*
328 		 * We've secured consistency through sync_fs. Following pino
329 		 * will be used only for fsynced inodes after checkpoint.
330 		 */
331 		try_to_fix_pino(inode);
332 		clear_inode_flag(inode, FI_APPEND_WRITE);
333 		clear_inode_flag(inode, FI_UPDATE_WRITE);
334 		goto out;
335 	}
336 sync_nodes:
337 	atomic_inc(&sbi->wb_sync_req[NODE]);
338 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
339 	atomic_dec(&sbi->wb_sync_req[NODE]);
340 	if (ret)
341 		goto out;
342 
343 	/* if cp_error was enabled, we should avoid infinite loop */
344 	if (unlikely(f2fs_cp_error(sbi))) {
345 		ret = -EIO;
346 		goto out;
347 	}
348 
349 	if (f2fs_need_inode_block_update(sbi, ino)) {
350 		f2fs_mark_inode_dirty_sync(inode, true);
351 		f2fs_write_inode(inode, NULL);
352 		goto sync_nodes;
353 	}
354 
355 	/*
356 	 * If it's atomic_write, it's just fine to keep write ordering. So
357 	 * here we don't need to wait for node write completion, since we use
358 	 * node chain which serializes node blocks. If one of node writes are
359 	 * reordered, we can see simply broken chain, resulting in stopping
360 	 * roll-forward recovery. It means we'll recover all or none node blocks
361 	 * given fsync mark.
362 	 */
363 	if (!atomic) {
364 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
365 		if (ret)
366 			goto out;
367 	}
368 
369 	/* once recovery info is written, don't need to tack this */
370 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
371 	clear_inode_flag(inode, FI_APPEND_WRITE);
372 flush_out:
373 	if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) ||
374 	    (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
375 		ret = f2fs_issue_flush(sbi, inode->i_ino);
376 	if (!ret) {
377 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
378 		clear_inode_flag(inode, FI_UPDATE_WRITE);
379 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
380 	}
381 	f2fs_update_time(sbi, REQ_TIME);
382 out:
383 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
384 	return ret;
385 }
386 
f2fs_sync_file(struct file * file,loff_t start,loff_t end,int datasync)387 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
388 {
389 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
390 		return -EIO;
391 	return f2fs_do_sync_file(file, start, end, datasync, false);
392 }
393 
__found_offset(struct address_space * mapping,block_t blkaddr,pgoff_t index,int whence)394 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
395 				pgoff_t index, int whence)
396 {
397 	switch (whence) {
398 	case SEEK_DATA:
399 		if (__is_valid_data_blkaddr(blkaddr))
400 			return true;
401 		if (blkaddr == NEW_ADDR &&
402 		    xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
403 			return true;
404 		break;
405 	case SEEK_HOLE:
406 		if (blkaddr == NULL_ADDR)
407 			return true;
408 		break;
409 	}
410 	return false;
411 }
412 
f2fs_seek_block(struct file * file,loff_t offset,int whence)413 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
414 {
415 	struct inode *inode = file->f_mapping->host;
416 	loff_t maxbytes = inode->i_sb->s_maxbytes;
417 	struct dnode_of_data dn;
418 	pgoff_t pgofs, end_offset;
419 	loff_t data_ofs = offset;
420 	loff_t isize;
421 	int err = 0;
422 
423 	inode_lock(inode);
424 
425 	isize = i_size_read(inode);
426 	if (offset >= isize)
427 		goto fail;
428 
429 	/* handle inline data case */
430 	if (f2fs_has_inline_data(inode)) {
431 		if (whence == SEEK_HOLE) {
432 			data_ofs = isize;
433 			goto found;
434 		} else if (whence == SEEK_DATA) {
435 			data_ofs = offset;
436 			goto found;
437 		}
438 	}
439 
440 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
441 
442 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
443 		set_new_dnode(&dn, inode, NULL, NULL, 0);
444 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
445 		if (err && err != -ENOENT) {
446 			goto fail;
447 		} else if (err == -ENOENT) {
448 			/* direct node does not exists */
449 			if (whence == SEEK_DATA) {
450 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
451 				continue;
452 			} else {
453 				goto found;
454 			}
455 		}
456 
457 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
458 
459 		/* find data/hole in dnode block */
460 		for (; dn.ofs_in_node < end_offset;
461 				dn.ofs_in_node++, pgofs++,
462 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
463 			block_t blkaddr;
464 
465 			blkaddr = f2fs_data_blkaddr(&dn);
466 
467 			if (__is_valid_data_blkaddr(blkaddr) &&
468 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
469 					blkaddr, DATA_GENERIC_ENHANCE)) {
470 				f2fs_put_dnode(&dn);
471 				goto fail;
472 			}
473 
474 			if (__found_offset(file->f_mapping, blkaddr,
475 							pgofs, whence)) {
476 				f2fs_put_dnode(&dn);
477 				goto found;
478 			}
479 		}
480 		f2fs_put_dnode(&dn);
481 	}
482 
483 	if (whence == SEEK_DATA)
484 		goto fail;
485 found:
486 	if (whence == SEEK_HOLE && data_ofs > isize)
487 		data_ofs = isize;
488 	inode_unlock(inode);
489 	return vfs_setpos(file, data_ofs, maxbytes);
490 fail:
491 	inode_unlock(inode);
492 	return -ENXIO;
493 }
494 
f2fs_llseek(struct file * file,loff_t offset,int whence)495 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
496 {
497 	struct inode *inode = file->f_mapping->host;
498 	loff_t maxbytes = inode->i_sb->s_maxbytes;
499 
500 	if (f2fs_compressed_file(inode))
501 		maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
502 
503 	switch (whence) {
504 	case SEEK_SET:
505 	case SEEK_CUR:
506 	case SEEK_END:
507 		return generic_file_llseek_size(file, offset, whence,
508 						maxbytes, i_size_read(inode));
509 	case SEEK_DATA:
510 	case SEEK_HOLE:
511 		if (offset < 0)
512 			return -ENXIO;
513 		return f2fs_seek_block(file, offset, whence);
514 	}
515 
516 	return -EINVAL;
517 }
518 
f2fs_file_mmap(struct file * file,struct vm_area_struct * vma)519 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
520 {
521 	struct inode *inode = file_inode(file);
522 
523 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
524 		return -EIO;
525 
526 	if (!f2fs_is_compress_backend_ready(inode))
527 		return -EOPNOTSUPP;
528 
529 	file_accessed(file);
530 	vma->vm_ops = &f2fs_file_vm_ops;
531 
532 	f2fs_down_read(&F2FS_I(inode)->i_sem);
533 	set_inode_flag(inode, FI_MMAP_FILE);
534 	f2fs_up_read(&F2FS_I(inode)->i_sem);
535 
536 	return 0;
537 }
538 
f2fs_file_open(struct inode * inode,struct file * filp)539 static int f2fs_file_open(struct inode *inode, struct file *filp)
540 {
541 	int err = fscrypt_file_open(inode, filp);
542 
543 	if (err)
544 		return err;
545 
546 	if (!f2fs_is_compress_backend_ready(inode))
547 		return -EOPNOTSUPP;
548 
549 	err = fsverity_file_open(inode, filp);
550 	if (err)
551 		return err;
552 
553 	filp->f_mode |= FMODE_NOWAIT;
554 
555 	return dquot_file_open(inode, filp);
556 }
557 
f2fs_truncate_data_blocks_range(struct dnode_of_data * dn,int count)558 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
559 {
560 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
561 	struct f2fs_node *raw_node;
562 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
563 	__le32 *addr;
564 	int base = 0;
565 	bool compressed_cluster = false;
566 	int cluster_index = 0, valid_blocks = 0;
567 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
568 	bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
569 
570 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
571 		base = get_extra_isize(dn->inode);
572 
573 	raw_node = F2FS_NODE(dn->node_page);
574 	addr = blkaddr_in_node(raw_node) + base + ofs;
575 
576 	/* Assumption: truncation starts with cluster */
577 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
578 		block_t blkaddr = le32_to_cpu(*addr);
579 
580 		if (f2fs_compressed_file(dn->inode) &&
581 					!(cluster_index & (cluster_size - 1))) {
582 			if (compressed_cluster)
583 				f2fs_i_compr_blocks_update(dn->inode,
584 							valid_blocks, false);
585 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
586 			valid_blocks = 0;
587 		}
588 
589 		if (blkaddr == NULL_ADDR)
590 			continue;
591 
592 		dn->data_blkaddr = NULL_ADDR;
593 		f2fs_set_data_blkaddr(dn);
594 
595 		if (__is_valid_data_blkaddr(blkaddr)) {
596 			if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE))
597 				continue;
598 			if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr,
599 						DATA_GENERIC_ENHANCE)) {
600 				f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
601 				continue;
602 			}
603 			if (compressed_cluster)
604 				valid_blocks++;
605 		}
606 
607 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
608 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
609 
610 		f2fs_invalidate_blocks(sbi, blkaddr);
611 
612 		if (!released || blkaddr != COMPRESS_ADDR)
613 			nr_free++;
614 	}
615 
616 	if (compressed_cluster)
617 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
618 
619 	if (nr_free) {
620 		pgoff_t fofs;
621 		/*
622 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
623 		 * we will invalidate all blkaddr in the whole range.
624 		 */
625 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
626 							dn->inode) + ofs;
627 		f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
628 		f2fs_update_age_extent_cache_range(dn, fofs, len);
629 		dec_valid_block_count(sbi, dn->inode, nr_free);
630 	}
631 	dn->ofs_in_node = ofs;
632 
633 	f2fs_update_time(sbi, REQ_TIME);
634 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
635 					 dn->ofs_in_node, nr_free);
636 }
637 
f2fs_truncate_data_blocks(struct dnode_of_data * dn)638 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
639 {
640 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
641 }
642 
truncate_partial_data_page(struct inode * inode,u64 from,bool cache_only)643 static int truncate_partial_data_page(struct inode *inode, u64 from,
644 								bool cache_only)
645 {
646 	loff_t offset = from & (PAGE_SIZE - 1);
647 	pgoff_t index = from >> PAGE_SHIFT;
648 	struct address_space *mapping = inode->i_mapping;
649 	struct page *page;
650 
651 	if (!offset && !cache_only)
652 		return 0;
653 
654 	if (cache_only) {
655 		page = find_lock_page(mapping, index);
656 		if (page && PageUptodate(page))
657 			goto truncate_out;
658 		f2fs_put_page(page, 1);
659 		return 0;
660 	}
661 
662 	page = f2fs_get_lock_data_page(inode, index, true);
663 	if (IS_ERR(page))
664 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
665 truncate_out:
666 	f2fs_wait_on_page_writeback(page, DATA, true, true);
667 	zero_user(page, offset, PAGE_SIZE - offset);
668 
669 	/* An encrypted inode should have a key and truncate the last page. */
670 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
671 	if (!cache_only)
672 		set_page_dirty(page);
673 	f2fs_put_page(page, 1);
674 	return 0;
675 }
676 
f2fs_do_truncate_blocks(struct inode * inode,u64 from,bool lock)677 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
678 {
679 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
680 	struct dnode_of_data dn;
681 	pgoff_t free_from;
682 	int count = 0, err = 0;
683 	struct page *ipage;
684 	bool truncate_page = false;
685 
686 	trace_f2fs_truncate_blocks_enter(inode, from);
687 
688 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
689 
690 	if (free_from >= max_file_blocks(inode))
691 		goto free_partial;
692 
693 	if (lock)
694 		f2fs_lock_op(sbi);
695 
696 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
697 	if (IS_ERR(ipage)) {
698 		err = PTR_ERR(ipage);
699 		goto out;
700 	}
701 
702 	if (f2fs_has_inline_data(inode)) {
703 		f2fs_truncate_inline_inode(inode, ipage, from);
704 		f2fs_put_page(ipage, 1);
705 		truncate_page = true;
706 		goto out;
707 	}
708 
709 	set_new_dnode(&dn, inode, ipage, NULL, 0);
710 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
711 	if (err) {
712 		if (err == -ENOENT)
713 			goto free_next;
714 		goto out;
715 	}
716 
717 	count = ADDRS_PER_PAGE(dn.node_page, inode);
718 
719 	count -= dn.ofs_in_node;
720 	f2fs_bug_on(sbi, count < 0);
721 
722 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
723 		f2fs_truncate_data_blocks_range(&dn, count);
724 		free_from += count;
725 	}
726 
727 	f2fs_put_dnode(&dn);
728 free_next:
729 	err = f2fs_truncate_inode_blocks(inode, free_from);
730 out:
731 	if (lock)
732 		f2fs_unlock_op(sbi);
733 free_partial:
734 	/* lastly zero out the first data page */
735 	if (!err)
736 		err = truncate_partial_data_page(inode, from, truncate_page);
737 
738 	trace_f2fs_truncate_blocks_exit(inode, err);
739 	return err;
740 }
741 
f2fs_truncate_blocks(struct inode * inode,u64 from,bool lock)742 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
743 {
744 	u64 free_from = from;
745 	int err;
746 
747 #ifdef CONFIG_F2FS_FS_COMPRESSION
748 	/*
749 	 * for compressed file, only support cluster size
750 	 * aligned truncation.
751 	 */
752 	if (f2fs_compressed_file(inode))
753 		free_from = round_up(from,
754 				F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
755 #endif
756 
757 	err = f2fs_do_truncate_blocks(inode, free_from, lock);
758 	if (err)
759 		return err;
760 
761 #ifdef CONFIG_F2FS_FS_COMPRESSION
762 	/*
763 	 * For compressed file, after release compress blocks, don't allow write
764 	 * direct, but we should allow write direct after truncate to zero.
765 	 */
766 	if (f2fs_compressed_file(inode) && !free_from
767 			&& is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
768 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
769 
770 	if (from != free_from) {
771 		err = f2fs_truncate_partial_cluster(inode, from, lock);
772 		if (err)
773 			return err;
774 	}
775 #endif
776 
777 	return 0;
778 }
779 
f2fs_truncate(struct inode * inode)780 int f2fs_truncate(struct inode *inode)
781 {
782 	int err;
783 
784 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
785 		return -EIO;
786 
787 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
788 				S_ISLNK(inode->i_mode)))
789 		return 0;
790 
791 	trace_f2fs_truncate(inode);
792 
793 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
794 		return -EIO;
795 
796 	err = f2fs_dquot_initialize(inode);
797 	if (err)
798 		return err;
799 
800 	/* we should check inline_data size */
801 	if (!f2fs_may_inline_data(inode)) {
802 		err = f2fs_convert_inline_inode(inode);
803 		if (err)
804 			return err;
805 	}
806 
807 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
808 	if (err)
809 		return err;
810 
811 	inode->i_mtime = inode->i_ctime = current_time(inode);
812 	f2fs_mark_inode_dirty_sync(inode, false);
813 	return 0;
814 }
815 
f2fs_force_buffered_io(struct inode * inode,int rw)816 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
817 {
818 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
819 
820 	if (!fscrypt_dio_supported(inode))
821 		return true;
822 	if (fsverity_active(inode))
823 		return true;
824 	if (f2fs_compressed_file(inode))
825 		return true;
826 
827 	/* disallow direct IO if any of devices has unaligned blksize */
828 	if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
829 		return true;
830 	/*
831 	 * for blkzoned device, fallback direct IO to buffered IO, so
832 	 * all IOs can be serialized by log-structured write.
833 	 */
834 	if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE))
835 		return true;
836 	if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi))
837 		return true;
838 	if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
839 		return true;
840 
841 	return false;
842 }
843 
f2fs_getattr(struct user_namespace * mnt_userns,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)844 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
845 		 struct kstat *stat, u32 request_mask, unsigned int query_flags)
846 {
847 	struct inode *inode = d_inode(path->dentry);
848 	struct f2fs_inode_info *fi = F2FS_I(inode);
849 	struct f2fs_inode *ri = NULL;
850 	unsigned int flags;
851 
852 	if (f2fs_has_extra_attr(inode) &&
853 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
854 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
855 		stat->result_mask |= STATX_BTIME;
856 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
857 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
858 	}
859 
860 	/*
861 	 * Return the DIO alignment restrictions if requested.  We only return
862 	 * this information when requested, since on encrypted files it might
863 	 * take a fair bit of work to get if the file wasn't opened recently.
864 	 *
865 	 * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
866 	 * cannot represent that, so in that case we report no DIO support.
867 	 */
868 	if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
869 		unsigned int bsize = i_blocksize(inode);
870 
871 		stat->result_mask |= STATX_DIOALIGN;
872 		if (!f2fs_force_buffered_io(inode, WRITE)) {
873 			stat->dio_mem_align = bsize;
874 			stat->dio_offset_align = bsize;
875 		}
876 	}
877 
878 	flags = fi->i_flags;
879 	if (flags & F2FS_COMPR_FL)
880 		stat->attributes |= STATX_ATTR_COMPRESSED;
881 	if (flags & F2FS_APPEND_FL)
882 		stat->attributes |= STATX_ATTR_APPEND;
883 	if (IS_ENCRYPTED(inode))
884 		stat->attributes |= STATX_ATTR_ENCRYPTED;
885 	if (flags & F2FS_IMMUTABLE_FL)
886 		stat->attributes |= STATX_ATTR_IMMUTABLE;
887 	if (flags & F2FS_NODUMP_FL)
888 		stat->attributes |= STATX_ATTR_NODUMP;
889 	if (IS_VERITY(inode))
890 		stat->attributes |= STATX_ATTR_VERITY;
891 
892 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
893 				  STATX_ATTR_APPEND |
894 				  STATX_ATTR_ENCRYPTED |
895 				  STATX_ATTR_IMMUTABLE |
896 				  STATX_ATTR_NODUMP |
897 				  STATX_ATTR_VERITY);
898 
899 	generic_fillattr(mnt_userns, inode, stat);
900 
901 	/* we need to show initial sectors used for inline_data/dentries */
902 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
903 					f2fs_has_inline_dentry(inode))
904 		stat->blocks += (stat->size + 511) >> 9;
905 
906 	return 0;
907 }
908 
909 #ifdef CONFIG_F2FS_FS_POSIX_ACL
__setattr_copy(struct user_namespace * mnt_userns,struct inode * inode,const struct iattr * attr)910 static void __setattr_copy(struct user_namespace *mnt_userns,
911 			   struct inode *inode, const struct iattr *attr)
912 {
913 	unsigned int ia_valid = attr->ia_valid;
914 
915 	i_uid_update(mnt_userns, attr, inode);
916 	i_gid_update(mnt_userns, attr, inode);
917 	if (ia_valid & ATTR_ATIME)
918 		inode->i_atime = attr->ia_atime;
919 	if (ia_valid & ATTR_MTIME)
920 		inode->i_mtime = attr->ia_mtime;
921 	if (ia_valid & ATTR_CTIME)
922 		inode->i_ctime = attr->ia_ctime;
923 	if (ia_valid & ATTR_MODE) {
924 		umode_t mode = attr->ia_mode;
925 		vfsgid_t vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
926 
927 		if (!vfsgid_in_group_p(vfsgid) &&
928 		    !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
929 			mode &= ~S_ISGID;
930 		set_acl_inode(inode, mode);
931 	}
932 }
933 #else
934 #define __setattr_copy setattr_copy
935 #endif
936 
f2fs_setattr(struct user_namespace * mnt_userns,struct dentry * dentry,struct iattr * attr)937 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
938 		 struct iattr *attr)
939 {
940 	struct inode *inode = d_inode(dentry);
941 	int err;
942 
943 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
944 		return -EIO;
945 
946 	if (unlikely(IS_IMMUTABLE(inode)))
947 		return -EPERM;
948 
949 	if (unlikely(IS_APPEND(inode) &&
950 			(attr->ia_valid & (ATTR_MODE | ATTR_UID |
951 				  ATTR_GID | ATTR_TIMES_SET))))
952 		return -EPERM;
953 
954 	if ((attr->ia_valid & ATTR_SIZE) &&
955 		!f2fs_is_compress_backend_ready(inode))
956 		return -EOPNOTSUPP;
957 
958 	err = setattr_prepare(mnt_userns, dentry, attr);
959 	if (err)
960 		return err;
961 
962 	err = fscrypt_prepare_setattr(dentry, attr);
963 	if (err)
964 		return err;
965 
966 	err = fsverity_prepare_setattr(dentry, attr);
967 	if (err)
968 		return err;
969 
970 	if (is_quota_modification(mnt_userns, inode, attr)) {
971 		err = f2fs_dquot_initialize(inode);
972 		if (err)
973 			return err;
974 	}
975 	if (i_uid_needs_update(mnt_userns, attr, inode) ||
976 	    i_gid_needs_update(mnt_userns, attr, inode)) {
977 		f2fs_lock_op(F2FS_I_SB(inode));
978 		err = dquot_transfer(mnt_userns, inode, attr);
979 		if (err) {
980 			set_sbi_flag(F2FS_I_SB(inode),
981 					SBI_QUOTA_NEED_REPAIR);
982 			f2fs_unlock_op(F2FS_I_SB(inode));
983 			return err;
984 		}
985 		/*
986 		 * update uid/gid under lock_op(), so that dquot and inode can
987 		 * be updated atomically.
988 		 */
989 		i_uid_update(mnt_userns, attr, inode);
990 		i_gid_update(mnt_userns, attr, inode);
991 		f2fs_mark_inode_dirty_sync(inode, true);
992 		f2fs_unlock_op(F2FS_I_SB(inode));
993 	}
994 
995 	if (attr->ia_valid & ATTR_SIZE) {
996 		loff_t old_size = i_size_read(inode);
997 
998 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
999 			/*
1000 			 * should convert inline inode before i_size_write to
1001 			 * keep smaller than inline_data size with inline flag.
1002 			 */
1003 			err = f2fs_convert_inline_inode(inode);
1004 			if (err)
1005 				return err;
1006 		}
1007 
1008 		f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1009 		filemap_invalidate_lock(inode->i_mapping);
1010 
1011 		truncate_setsize(inode, attr->ia_size);
1012 
1013 		if (attr->ia_size <= old_size)
1014 			err = f2fs_truncate(inode);
1015 		/*
1016 		 * do not trim all blocks after i_size if target size is
1017 		 * larger than i_size.
1018 		 */
1019 		filemap_invalidate_unlock(inode->i_mapping);
1020 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1021 		if (err)
1022 			return err;
1023 
1024 		spin_lock(&F2FS_I(inode)->i_size_lock);
1025 		inode->i_mtime = inode->i_ctime = current_time(inode);
1026 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
1027 		spin_unlock(&F2FS_I(inode)->i_size_lock);
1028 	}
1029 
1030 	__setattr_copy(mnt_userns, inode, attr);
1031 
1032 	if (attr->ia_valid & ATTR_MODE) {
1033 		err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
1034 
1035 		if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1036 			if (!err)
1037 				inode->i_mode = F2FS_I(inode)->i_acl_mode;
1038 			clear_inode_flag(inode, FI_ACL_MODE);
1039 		}
1040 	}
1041 
1042 	/* file size may changed here */
1043 	f2fs_mark_inode_dirty_sync(inode, true);
1044 
1045 	/* inode change will produce dirty node pages flushed by checkpoint */
1046 	f2fs_balance_fs(F2FS_I_SB(inode), true);
1047 
1048 	return err;
1049 }
1050 
1051 const struct inode_operations f2fs_file_inode_operations = {
1052 	.getattr	= f2fs_getattr,
1053 	.setattr	= f2fs_setattr,
1054 	.get_acl	= f2fs_get_acl,
1055 	.set_acl	= f2fs_set_acl,
1056 	.listxattr	= f2fs_listxattr,
1057 	.fiemap		= f2fs_fiemap,
1058 	.fileattr_get	= f2fs_fileattr_get,
1059 	.fileattr_set	= f2fs_fileattr_set,
1060 };
1061 
fill_zero(struct inode * inode,pgoff_t index,loff_t start,loff_t len)1062 static int fill_zero(struct inode *inode, pgoff_t index,
1063 					loff_t start, loff_t len)
1064 {
1065 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1066 	struct page *page;
1067 
1068 	if (!len)
1069 		return 0;
1070 
1071 	f2fs_balance_fs(sbi, true);
1072 
1073 	f2fs_lock_op(sbi);
1074 	page = f2fs_get_new_data_page(inode, NULL, index, false);
1075 	f2fs_unlock_op(sbi);
1076 
1077 	if (IS_ERR(page))
1078 		return PTR_ERR(page);
1079 
1080 	f2fs_wait_on_page_writeback(page, DATA, true, true);
1081 	zero_user(page, start, len);
1082 	set_page_dirty(page);
1083 	f2fs_put_page(page, 1);
1084 	return 0;
1085 }
1086 
f2fs_truncate_hole(struct inode * inode,pgoff_t pg_start,pgoff_t pg_end)1087 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1088 {
1089 	int err;
1090 
1091 	while (pg_start < pg_end) {
1092 		struct dnode_of_data dn;
1093 		pgoff_t end_offset, count;
1094 
1095 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1096 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1097 		if (err) {
1098 			if (err == -ENOENT) {
1099 				pg_start = f2fs_get_next_page_offset(&dn,
1100 								pg_start);
1101 				continue;
1102 			}
1103 			return err;
1104 		}
1105 
1106 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1107 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1108 
1109 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1110 
1111 		f2fs_truncate_data_blocks_range(&dn, count);
1112 		f2fs_put_dnode(&dn);
1113 
1114 		pg_start += count;
1115 	}
1116 	return 0;
1117 }
1118 
f2fs_punch_hole(struct inode * inode,loff_t offset,loff_t len)1119 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
1120 {
1121 	pgoff_t pg_start, pg_end;
1122 	loff_t off_start, off_end;
1123 	int ret;
1124 
1125 	ret = f2fs_convert_inline_inode(inode);
1126 	if (ret)
1127 		return ret;
1128 
1129 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1130 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1131 
1132 	off_start = offset & (PAGE_SIZE - 1);
1133 	off_end = (offset + len) & (PAGE_SIZE - 1);
1134 
1135 	if (pg_start == pg_end) {
1136 		ret = fill_zero(inode, pg_start, off_start,
1137 						off_end - off_start);
1138 		if (ret)
1139 			return ret;
1140 	} else {
1141 		if (off_start) {
1142 			ret = fill_zero(inode, pg_start++, off_start,
1143 						PAGE_SIZE - off_start);
1144 			if (ret)
1145 				return ret;
1146 		}
1147 		if (off_end) {
1148 			ret = fill_zero(inode, pg_end, 0, off_end);
1149 			if (ret)
1150 				return ret;
1151 		}
1152 
1153 		if (pg_start < pg_end) {
1154 			loff_t blk_start, blk_end;
1155 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1156 
1157 			f2fs_balance_fs(sbi, true);
1158 
1159 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1160 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1161 
1162 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1163 			filemap_invalidate_lock(inode->i_mapping);
1164 
1165 			truncate_pagecache_range(inode, blk_start, blk_end - 1);
1166 
1167 			f2fs_lock_op(sbi);
1168 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1169 			f2fs_unlock_op(sbi);
1170 
1171 			filemap_invalidate_unlock(inode->i_mapping);
1172 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1173 		}
1174 	}
1175 
1176 	return ret;
1177 }
1178 
__read_out_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,pgoff_t len)1179 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1180 				int *do_replace, pgoff_t off, pgoff_t len)
1181 {
1182 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1183 	struct dnode_of_data dn;
1184 	int ret, done, i;
1185 
1186 next_dnode:
1187 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1188 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1189 	if (ret && ret != -ENOENT) {
1190 		return ret;
1191 	} else if (ret == -ENOENT) {
1192 		if (dn.max_level == 0)
1193 			return -ENOENT;
1194 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1195 						dn.ofs_in_node, len);
1196 		blkaddr += done;
1197 		do_replace += done;
1198 		goto next;
1199 	}
1200 
1201 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1202 							dn.ofs_in_node, len);
1203 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1204 		*blkaddr = f2fs_data_blkaddr(&dn);
1205 
1206 		if (__is_valid_data_blkaddr(*blkaddr) &&
1207 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1208 					DATA_GENERIC_ENHANCE)) {
1209 			f2fs_put_dnode(&dn);
1210 			f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1211 			return -EFSCORRUPTED;
1212 		}
1213 
1214 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1215 
1216 			if (f2fs_lfs_mode(sbi)) {
1217 				f2fs_put_dnode(&dn);
1218 				return -EOPNOTSUPP;
1219 			}
1220 
1221 			/* do not invalidate this block address */
1222 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1223 			*do_replace = 1;
1224 		}
1225 	}
1226 	f2fs_put_dnode(&dn);
1227 next:
1228 	len -= done;
1229 	off += done;
1230 	if (len)
1231 		goto next_dnode;
1232 	return 0;
1233 }
1234 
__roll_back_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,int len)1235 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1236 				int *do_replace, pgoff_t off, int len)
1237 {
1238 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1239 	struct dnode_of_data dn;
1240 	int ret, i;
1241 
1242 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1243 		if (*do_replace == 0)
1244 			continue;
1245 
1246 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1247 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1248 		if (ret) {
1249 			dec_valid_block_count(sbi, inode, 1);
1250 			f2fs_invalidate_blocks(sbi, *blkaddr);
1251 		} else {
1252 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1253 		}
1254 		f2fs_put_dnode(&dn);
1255 	}
1256 	return 0;
1257 }
1258 
__clone_blkaddrs(struct inode * src_inode,struct inode * dst_inode,block_t * blkaddr,int * do_replace,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1259 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1260 			block_t *blkaddr, int *do_replace,
1261 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1262 {
1263 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1264 	pgoff_t i = 0;
1265 	int ret;
1266 
1267 	while (i < len) {
1268 		if (blkaddr[i] == NULL_ADDR && !full) {
1269 			i++;
1270 			continue;
1271 		}
1272 
1273 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1274 			struct dnode_of_data dn;
1275 			struct node_info ni;
1276 			size_t new_size;
1277 			pgoff_t ilen;
1278 
1279 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1280 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1281 			if (ret)
1282 				return ret;
1283 
1284 			ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1285 			if (ret) {
1286 				f2fs_put_dnode(&dn);
1287 				return ret;
1288 			}
1289 
1290 			ilen = min((pgoff_t)
1291 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1292 						dn.ofs_in_node, len - i);
1293 			do {
1294 				dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1295 				f2fs_truncate_data_blocks_range(&dn, 1);
1296 
1297 				if (do_replace[i]) {
1298 					f2fs_i_blocks_write(src_inode,
1299 							1, false, false);
1300 					f2fs_i_blocks_write(dst_inode,
1301 							1, true, false);
1302 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1303 					blkaddr[i], ni.version, true, false);
1304 
1305 					do_replace[i] = 0;
1306 				}
1307 				dn.ofs_in_node++;
1308 				i++;
1309 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1310 				if (dst_inode->i_size < new_size)
1311 					f2fs_i_size_write(dst_inode, new_size);
1312 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1313 
1314 			f2fs_put_dnode(&dn);
1315 		} else {
1316 			struct page *psrc, *pdst;
1317 
1318 			psrc = f2fs_get_lock_data_page(src_inode,
1319 							src + i, true);
1320 			if (IS_ERR(psrc))
1321 				return PTR_ERR(psrc);
1322 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1323 								true);
1324 			if (IS_ERR(pdst)) {
1325 				f2fs_put_page(psrc, 1);
1326 				return PTR_ERR(pdst);
1327 			}
1328 			memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1329 			set_page_dirty(pdst);
1330 			set_page_private_gcing(pdst);
1331 			f2fs_put_page(pdst, 1);
1332 			f2fs_put_page(psrc, 1);
1333 
1334 			ret = f2fs_truncate_hole(src_inode,
1335 						src + i, src + i + 1);
1336 			if (ret)
1337 				return ret;
1338 			i++;
1339 		}
1340 	}
1341 	return 0;
1342 }
1343 
__exchange_data_block(struct inode * src_inode,struct inode * dst_inode,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1344 static int __exchange_data_block(struct inode *src_inode,
1345 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1346 			pgoff_t len, bool full)
1347 {
1348 	block_t *src_blkaddr;
1349 	int *do_replace;
1350 	pgoff_t olen;
1351 	int ret;
1352 
1353 	while (len) {
1354 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1355 
1356 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1357 					array_size(olen, sizeof(block_t)),
1358 					GFP_NOFS);
1359 		if (!src_blkaddr)
1360 			return -ENOMEM;
1361 
1362 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1363 					array_size(olen, sizeof(int)),
1364 					GFP_NOFS);
1365 		if (!do_replace) {
1366 			kvfree(src_blkaddr);
1367 			return -ENOMEM;
1368 		}
1369 
1370 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1371 					do_replace, src, olen);
1372 		if (ret)
1373 			goto roll_back;
1374 
1375 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1376 					do_replace, src, dst, olen, full);
1377 		if (ret)
1378 			goto roll_back;
1379 
1380 		src += olen;
1381 		dst += olen;
1382 		len -= olen;
1383 
1384 		kvfree(src_blkaddr);
1385 		kvfree(do_replace);
1386 	}
1387 	return 0;
1388 
1389 roll_back:
1390 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1391 	kvfree(src_blkaddr);
1392 	kvfree(do_replace);
1393 	return ret;
1394 }
1395 
f2fs_do_collapse(struct inode * inode,loff_t offset,loff_t len)1396 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1397 {
1398 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1399 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1400 	pgoff_t start = offset >> PAGE_SHIFT;
1401 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1402 	int ret;
1403 
1404 	f2fs_balance_fs(sbi, true);
1405 
1406 	/* avoid gc operation during block exchange */
1407 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1408 	filemap_invalidate_lock(inode->i_mapping);
1409 
1410 	f2fs_lock_op(sbi);
1411 	f2fs_drop_extent_tree(inode);
1412 	truncate_pagecache(inode, offset);
1413 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1414 	f2fs_unlock_op(sbi);
1415 
1416 	filemap_invalidate_unlock(inode->i_mapping);
1417 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1418 	return ret;
1419 }
1420 
f2fs_collapse_range(struct inode * inode,loff_t offset,loff_t len)1421 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1422 {
1423 	loff_t new_size;
1424 	int ret;
1425 
1426 	if (offset + len >= i_size_read(inode))
1427 		return -EINVAL;
1428 
1429 	/* collapse range should be aligned to block size of f2fs. */
1430 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1431 		return -EINVAL;
1432 
1433 	ret = f2fs_convert_inline_inode(inode);
1434 	if (ret)
1435 		return ret;
1436 
1437 	/* write out all dirty pages from offset */
1438 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1439 	if (ret)
1440 		return ret;
1441 
1442 	ret = f2fs_do_collapse(inode, offset, len);
1443 	if (ret)
1444 		return ret;
1445 
1446 	/* write out all moved pages, if possible */
1447 	filemap_invalidate_lock(inode->i_mapping);
1448 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1449 	truncate_pagecache(inode, offset);
1450 
1451 	new_size = i_size_read(inode) - len;
1452 	ret = f2fs_truncate_blocks(inode, new_size, true);
1453 	filemap_invalidate_unlock(inode->i_mapping);
1454 	if (!ret)
1455 		f2fs_i_size_write(inode, new_size);
1456 	return ret;
1457 }
1458 
f2fs_do_zero_range(struct dnode_of_data * dn,pgoff_t start,pgoff_t end)1459 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1460 								pgoff_t end)
1461 {
1462 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1463 	pgoff_t index = start;
1464 	unsigned int ofs_in_node = dn->ofs_in_node;
1465 	blkcnt_t count = 0;
1466 	int ret;
1467 
1468 	for (; index < end; index++, dn->ofs_in_node++) {
1469 		if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1470 			count++;
1471 	}
1472 
1473 	dn->ofs_in_node = ofs_in_node;
1474 	ret = f2fs_reserve_new_blocks(dn, count);
1475 	if (ret)
1476 		return ret;
1477 
1478 	dn->ofs_in_node = ofs_in_node;
1479 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1480 		dn->data_blkaddr = f2fs_data_blkaddr(dn);
1481 		/*
1482 		 * f2fs_reserve_new_blocks will not guarantee entire block
1483 		 * allocation.
1484 		 */
1485 		if (dn->data_blkaddr == NULL_ADDR) {
1486 			ret = -ENOSPC;
1487 			break;
1488 		}
1489 
1490 		if (dn->data_blkaddr == NEW_ADDR)
1491 			continue;
1492 
1493 		if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1494 					DATA_GENERIC_ENHANCE)) {
1495 			ret = -EFSCORRUPTED;
1496 			f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1497 			break;
1498 		}
1499 
1500 		f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1501 		dn->data_blkaddr = NEW_ADDR;
1502 		f2fs_set_data_blkaddr(dn);
1503 	}
1504 
1505 	f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
1506 	f2fs_update_age_extent_cache_range(dn, start, index - start);
1507 
1508 	return ret;
1509 }
1510 
f2fs_zero_range(struct inode * inode,loff_t offset,loff_t len,int mode)1511 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1512 								int mode)
1513 {
1514 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1515 	struct address_space *mapping = inode->i_mapping;
1516 	pgoff_t index, pg_start, pg_end;
1517 	loff_t new_size = i_size_read(inode);
1518 	loff_t off_start, off_end;
1519 	int ret = 0;
1520 
1521 	ret = inode_newsize_ok(inode, (len + offset));
1522 	if (ret)
1523 		return ret;
1524 
1525 	ret = f2fs_convert_inline_inode(inode);
1526 	if (ret)
1527 		return ret;
1528 
1529 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1530 	if (ret)
1531 		return ret;
1532 
1533 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1534 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1535 
1536 	off_start = offset & (PAGE_SIZE - 1);
1537 	off_end = (offset + len) & (PAGE_SIZE - 1);
1538 
1539 	if (pg_start == pg_end) {
1540 		ret = fill_zero(inode, pg_start, off_start,
1541 						off_end - off_start);
1542 		if (ret)
1543 			return ret;
1544 
1545 		new_size = max_t(loff_t, new_size, offset + len);
1546 	} else {
1547 		if (off_start) {
1548 			ret = fill_zero(inode, pg_start++, off_start,
1549 						PAGE_SIZE - off_start);
1550 			if (ret)
1551 				return ret;
1552 
1553 			new_size = max_t(loff_t, new_size,
1554 					(loff_t)pg_start << PAGE_SHIFT);
1555 		}
1556 
1557 		for (index = pg_start; index < pg_end;) {
1558 			struct dnode_of_data dn;
1559 			unsigned int end_offset;
1560 			pgoff_t end;
1561 
1562 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1563 			filemap_invalidate_lock(mapping);
1564 
1565 			truncate_pagecache_range(inode,
1566 				(loff_t)index << PAGE_SHIFT,
1567 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1568 
1569 			f2fs_lock_op(sbi);
1570 
1571 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1572 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1573 			if (ret) {
1574 				f2fs_unlock_op(sbi);
1575 				filemap_invalidate_unlock(mapping);
1576 				f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1577 				goto out;
1578 			}
1579 
1580 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1581 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1582 
1583 			ret = f2fs_do_zero_range(&dn, index, end);
1584 			f2fs_put_dnode(&dn);
1585 
1586 			f2fs_unlock_op(sbi);
1587 			filemap_invalidate_unlock(mapping);
1588 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1589 
1590 			f2fs_balance_fs(sbi, dn.node_changed);
1591 
1592 			if (ret)
1593 				goto out;
1594 
1595 			index = end;
1596 			new_size = max_t(loff_t, new_size,
1597 					(loff_t)index << PAGE_SHIFT);
1598 		}
1599 
1600 		if (off_end) {
1601 			ret = fill_zero(inode, pg_end, 0, off_end);
1602 			if (ret)
1603 				goto out;
1604 
1605 			new_size = max_t(loff_t, new_size, offset + len);
1606 		}
1607 	}
1608 
1609 out:
1610 	if (new_size > i_size_read(inode)) {
1611 		if (mode & FALLOC_FL_KEEP_SIZE)
1612 			file_set_keep_isize(inode);
1613 		else
1614 			f2fs_i_size_write(inode, new_size);
1615 	}
1616 	return ret;
1617 }
1618 
f2fs_insert_range(struct inode * inode,loff_t offset,loff_t len)1619 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1620 {
1621 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1622 	struct address_space *mapping = inode->i_mapping;
1623 	pgoff_t nr, pg_start, pg_end, delta, idx;
1624 	loff_t new_size;
1625 	int ret = 0;
1626 
1627 	new_size = i_size_read(inode) + len;
1628 	ret = inode_newsize_ok(inode, new_size);
1629 	if (ret)
1630 		return ret;
1631 
1632 	if (offset >= i_size_read(inode))
1633 		return -EINVAL;
1634 
1635 	/* insert range should be aligned to block size of f2fs. */
1636 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1637 		return -EINVAL;
1638 
1639 	ret = f2fs_convert_inline_inode(inode);
1640 	if (ret)
1641 		return ret;
1642 
1643 	f2fs_balance_fs(sbi, true);
1644 
1645 	filemap_invalidate_lock(mapping);
1646 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1647 	filemap_invalidate_unlock(mapping);
1648 	if (ret)
1649 		return ret;
1650 
1651 	/* write out all dirty pages from offset */
1652 	ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1653 	if (ret)
1654 		return ret;
1655 
1656 	pg_start = offset >> PAGE_SHIFT;
1657 	pg_end = (offset + len) >> PAGE_SHIFT;
1658 	delta = pg_end - pg_start;
1659 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1660 
1661 	/* avoid gc operation during block exchange */
1662 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1663 	filemap_invalidate_lock(mapping);
1664 	truncate_pagecache(inode, offset);
1665 
1666 	while (!ret && idx > pg_start) {
1667 		nr = idx - pg_start;
1668 		if (nr > delta)
1669 			nr = delta;
1670 		idx -= nr;
1671 
1672 		f2fs_lock_op(sbi);
1673 		f2fs_drop_extent_tree(inode);
1674 
1675 		ret = __exchange_data_block(inode, inode, idx,
1676 					idx + delta, nr, false);
1677 		f2fs_unlock_op(sbi);
1678 	}
1679 	filemap_invalidate_unlock(mapping);
1680 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1681 
1682 	/* write out all moved pages, if possible */
1683 	filemap_invalidate_lock(mapping);
1684 	filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1685 	truncate_pagecache(inode, offset);
1686 	filemap_invalidate_unlock(mapping);
1687 
1688 	if (!ret)
1689 		f2fs_i_size_write(inode, new_size);
1690 	return ret;
1691 }
1692 
f2fs_expand_inode_data(struct inode * inode,loff_t offset,loff_t len,int mode)1693 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
1694 					loff_t len, int mode)
1695 {
1696 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1697 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1698 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1699 			.m_may_create = true };
1700 	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1701 			.init_gc_type = FG_GC,
1702 			.should_migrate_blocks = false,
1703 			.err_gc_skipped = true,
1704 			.nr_free_secs = 0 };
1705 	pgoff_t pg_start, pg_end;
1706 	loff_t new_size;
1707 	loff_t off_end;
1708 	block_t expanded = 0;
1709 	int err;
1710 
1711 	err = inode_newsize_ok(inode, (len + offset));
1712 	if (err)
1713 		return err;
1714 
1715 	err = f2fs_convert_inline_inode(inode);
1716 	if (err)
1717 		return err;
1718 
1719 	f2fs_balance_fs(sbi, true);
1720 
1721 	pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1722 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1723 	off_end = (offset + len) & (PAGE_SIZE - 1);
1724 
1725 	map.m_lblk = pg_start;
1726 	map.m_len = pg_end - pg_start;
1727 	if (off_end)
1728 		map.m_len++;
1729 
1730 	if (!map.m_len)
1731 		return 0;
1732 
1733 	if (f2fs_is_pinned_file(inode)) {
1734 		block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1735 		block_t sec_len = roundup(map.m_len, sec_blks);
1736 
1737 		map.m_len = sec_blks;
1738 next_alloc:
1739 		if (has_not_enough_free_secs(sbi, 0,
1740 			GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1741 			f2fs_down_write(&sbi->gc_lock);
1742 			err = f2fs_gc(sbi, &gc_control);
1743 			if (err && err != -ENODATA)
1744 				goto out_err;
1745 		}
1746 
1747 		f2fs_down_write(&sbi->pin_sem);
1748 
1749 		f2fs_lock_op(sbi);
1750 		f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1751 		f2fs_unlock_op(sbi);
1752 
1753 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1754 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
1755 		file_dont_truncate(inode);
1756 
1757 		f2fs_up_write(&sbi->pin_sem);
1758 
1759 		expanded += map.m_len;
1760 		sec_len -= map.m_len;
1761 		map.m_lblk += map.m_len;
1762 		if (!err && sec_len)
1763 			goto next_alloc;
1764 
1765 		map.m_len = expanded;
1766 	} else {
1767 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
1768 		expanded = map.m_len;
1769 	}
1770 out_err:
1771 	if (err) {
1772 		pgoff_t last_off;
1773 
1774 		if (!expanded)
1775 			return err;
1776 
1777 		last_off = pg_start + expanded - 1;
1778 
1779 		/* update new size to the failed position */
1780 		new_size = (last_off == pg_end) ? offset + len :
1781 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1782 	} else {
1783 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1784 	}
1785 
1786 	if (new_size > i_size_read(inode)) {
1787 		if (mode & FALLOC_FL_KEEP_SIZE)
1788 			file_set_keep_isize(inode);
1789 		else
1790 			f2fs_i_size_write(inode, new_size);
1791 	}
1792 
1793 	return err;
1794 }
1795 
f2fs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)1796 static long f2fs_fallocate(struct file *file, int mode,
1797 				loff_t offset, loff_t len)
1798 {
1799 	struct inode *inode = file_inode(file);
1800 	long ret = 0;
1801 
1802 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1803 		return -EIO;
1804 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1805 		return -ENOSPC;
1806 	if (!f2fs_is_compress_backend_ready(inode))
1807 		return -EOPNOTSUPP;
1808 
1809 	/* f2fs only support ->fallocate for regular file */
1810 	if (!S_ISREG(inode->i_mode))
1811 		return -EINVAL;
1812 
1813 	if (IS_ENCRYPTED(inode) &&
1814 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1815 		return -EOPNOTSUPP;
1816 
1817 	/*
1818 	 * Pinned file should not support partial truncation since the block
1819 	 * can be used by applications.
1820 	 */
1821 	if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1822 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1823 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1824 		return -EOPNOTSUPP;
1825 
1826 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1827 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1828 			FALLOC_FL_INSERT_RANGE))
1829 		return -EOPNOTSUPP;
1830 
1831 	inode_lock(inode);
1832 
1833 	ret = file_modified(file);
1834 	if (ret)
1835 		goto out;
1836 
1837 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1838 		if (offset >= inode->i_size)
1839 			goto out;
1840 
1841 		ret = f2fs_punch_hole(inode, offset, len);
1842 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1843 		ret = f2fs_collapse_range(inode, offset, len);
1844 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1845 		ret = f2fs_zero_range(inode, offset, len, mode);
1846 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1847 		ret = f2fs_insert_range(inode, offset, len);
1848 	} else {
1849 		ret = f2fs_expand_inode_data(inode, offset, len, mode);
1850 	}
1851 
1852 	if (!ret) {
1853 		inode->i_mtime = inode->i_ctime = current_time(inode);
1854 		f2fs_mark_inode_dirty_sync(inode, false);
1855 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1856 	}
1857 
1858 out:
1859 	inode_unlock(inode);
1860 
1861 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1862 	return ret;
1863 }
1864 
f2fs_release_file(struct inode * inode,struct file * filp)1865 static int f2fs_release_file(struct inode *inode, struct file *filp)
1866 {
1867 	/*
1868 	 * f2fs_release_file is called at every close calls. So we should
1869 	 * not drop any inmemory pages by close called by other process.
1870 	 */
1871 	if (!(filp->f_mode & FMODE_WRITE) ||
1872 			atomic_read(&inode->i_writecount) != 1)
1873 		return 0;
1874 
1875 	inode_lock(inode);
1876 	f2fs_abort_atomic_write(inode, true);
1877 	inode_unlock(inode);
1878 
1879 	return 0;
1880 }
1881 
f2fs_file_flush(struct file * file,fl_owner_t id)1882 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1883 {
1884 	struct inode *inode = file_inode(file);
1885 
1886 	/*
1887 	 * If the process doing a transaction is crashed, we should do
1888 	 * roll-back. Otherwise, other reader/write can see corrupted database
1889 	 * until all the writers close its file. Since this should be done
1890 	 * before dropping file lock, it needs to do in ->flush.
1891 	 */
1892 	if (F2FS_I(inode)->atomic_write_task == current &&
1893 				(current->flags & PF_EXITING)) {
1894 		inode_lock(inode);
1895 		f2fs_abort_atomic_write(inode, true);
1896 		inode_unlock(inode);
1897 	}
1898 
1899 	return 0;
1900 }
1901 
f2fs_setflags_common(struct inode * inode,u32 iflags,u32 mask)1902 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1903 {
1904 	struct f2fs_inode_info *fi = F2FS_I(inode);
1905 	u32 masked_flags = fi->i_flags & mask;
1906 
1907 	/* mask can be shrunk by flags_valid selector */
1908 	iflags &= mask;
1909 
1910 	/* Is it quota file? Do not allow user to mess with it */
1911 	if (IS_NOQUOTA(inode))
1912 		return -EPERM;
1913 
1914 	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1915 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1916 			return -EOPNOTSUPP;
1917 		if (!f2fs_empty_dir(inode))
1918 			return -ENOTEMPTY;
1919 	}
1920 
1921 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1922 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1923 			return -EOPNOTSUPP;
1924 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1925 			return -EINVAL;
1926 	}
1927 
1928 	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1929 		if (masked_flags & F2FS_COMPR_FL) {
1930 			if (!f2fs_disable_compressed_file(inode))
1931 				return -EINVAL;
1932 		} else {
1933 			/* try to convert inline_data to support compression */
1934 			int err = f2fs_convert_inline_inode(inode);
1935 			if (err)
1936 				return err;
1937 
1938 			f2fs_down_write(&F2FS_I(inode)->i_sem);
1939 			if (!f2fs_may_compress(inode) ||
1940 					(S_ISREG(inode->i_mode) &&
1941 					F2FS_HAS_BLOCKS(inode))) {
1942 				f2fs_up_write(&F2FS_I(inode)->i_sem);
1943 				return -EINVAL;
1944 			}
1945 			err = set_compress_context(inode);
1946 			f2fs_up_write(&F2FS_I(inode)->i_sem);
1947 
1948 			if (err)
1949 				return err;
1950 		}
1951 	}
1952 
1953 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1954 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1955 					(fi->i_flags & F2FS_NOCOMP_FL));
1956 
1957 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1958 		set_inode_flag(inode, FI_PROJ_INHERIT);
1959 	else
1960 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1961 
1962 	inode->i_ctime = current_time(inode);
1963 	f2fs_set_inode_flags(inode);
1964 	f2fs_mark_inode_dirty_sync(inode, true);
1965 	return 0;
1966 }
1967 
1968 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1969 
1970 /*
1971  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1972  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1973  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1974  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1975  *
1976  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1977  * FS_IOC_FSSETXATTR is done by the VFS.
1978  */
1979 
1980 static const struct {
1981 	u32 iflag;
1982 	u32 fsflag;
1983 } f2fs_fsflags_map[] = {
1984 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
1985 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1986 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1987 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1988 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1989 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1990 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
1991 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1992 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1993 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1994 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
1995 };
1996 
1997 #define F2FS_GETTABLE_FS_FL (		\
1998 		FS_COMPR_FL |		\
1999 		FS_SYNC_FL |		\
2000 		FS_IMMUTABLE_FL |	\
2001 		FS_APPEND_FL |		\
2002 		FS_NODUMP_FL |		\
2003 		FS_NOATIME_FL |		\
2004 		FS_NOCOMP_FL |		\
2005 		FS_INDEX_FL |		\
2006 		FS_DIRSYNC_FL |		\
2007 		FS_PROJINHERIT_FL |	\
2008 		FS_ENCRYPT_FL |		\
2009 		FS_INLINE_DATA_FL |	\
2010 		FS_NOCOW_FL |		\
2011 		FS_VERITY_FL |		\
2012 		FS_CASEFOLD_FL)
2013 
2014 #define F2FS_SETTABLE_FS_FL (		\
2015 		FS_COMPR_FL |		\
2016 		FS_SYNC_FL |		\
2017 		FS_IMMUTABLE_FL |	\
2018 		FS_APPEND_FL |		\
2019 		FS_NODUMP_FL |		\
2020 		FS_NOATIME_FL |		\
2021 		FS_NOCOMP_FL |		\
2022 		FS_DIRSYNC_FL |		\
2023 		FS_PROJINHERIT_FL |	\
2024 		FS_CASEFOLD_FL)
2025 
2026 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
f2fs_iflags_to_fsflags(u32 iflags)2027 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2028 {
2029 	u32 fsflags = 0;
2030 	int i;
2031 
2032 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2033 		if (iflags & f2fs_fsflags_map[i].iflag)
2034 			fsflags |= f2fs_fsflags_map[i].fsflag;
2035 
2036 	return fsflags;
2037 }
2038 
2039 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
f2fs_fsflags_to_iflags(u32 fsflags)2040 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2041 {
2042 	u32 iflags = 0;
2043 	int i;
2044 
2045 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2046 		if (fsflags & f2fs_fsflags_map[i].fsflag)
2047 			iflags |= f2fs_fsflags_map[i].iflag;
2048 
2049 	return iflags;
2050 }
2051 
f2fs_ioc_getversion(struct file * filp,unsigned long arg)2052 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2053 {
2054 	struct inode *inode = file_inode(filp);
2055 
2056 	return put_user(inode->i_generation, (int __user *)arg);
2057 }
2058 
f2fs_ioc_start_atomic_write(struct file * filp,bool truncate)2059 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2060 {
2061 	struct inode *inode = file_inode(filp);
2062 	struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2063 	struct f2fs_inode_info *fi = F2FS_I(inode);
2064 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2065 	struct inode *pinode;
2066 	loff_t isize;
2067 	int ret;
2068 
2069 	if (!inode_owner_or_capable(mnt_userns, inode))
2070 		return -EACCES;
2071 
2072 	if (!S_ISREG(inode->i_mode))
2073 		return -EINVAL;
2074 
2075 	if (filp->f_flags & O_DIRECT)
2076 		return -EINVAL;
2077 
2078 	ret = mnt_want_write_file(filp);
2079 	if (ret)
2080 		return ret;
2081 
2082 	inode_lock(inode);
2083 
2084 	if (!f2fs_disable_compressed_file(inode)) {
2085 		ret = -EINVAL;
2086 		goto out;
2087 	}
2088 
2089 	if (f2fs_is_atomic_file(inode))
2090 		goto out;
2091 
2092 	ret = f2fs_convert_inline_inode(inode);
2093 	if (ret)
2094 		goto out;
2095 
2096 	f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2097 
2098 	/*
2099 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2100 	 * f2fs_is_atomic_file.
2101 	 */
2102 	if (get_dirty_pages(inode))
2103 		f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2104 			  inode->i_ino, get_dirty_pages(inode));
2105 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2106 	if (ret) {
2107 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2108 		goto out;
2109 	}
2110 
2111 	/* Check if the inode already has a COW inode */
2112 	if (fi->cow_inode == NULL) {
2113 		/* Create a COW inode for atomic write */
2114 		pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2115 		if (IS_ERR(pinode)) {
2116 			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2117 			ret = PTR_ERR(pinode);
2118 			goto out;
2119 		}
2120 
2121 		ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2122 		iput(pinode);
2123 		if (ret) {
2124 			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2125 			goto out;
2126 		}
2127 
2128 		set_inode_flag(fi->cow_inode, FI_COW_FILE);
2129 		clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2130 	} else {
2131 		/* Reuse the already created COW inode */
2132 		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2133 		if (ret) {
2134 			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2135 			goto out;
2136 		}
2137 	}
2138 
2139 	f2fs_write_inode(inode, NULL);
2140 
2141 	stat_inc_atomic_inode(inode);
2142 
2143 	set_inode_flag(inode, FI_ATOMIC_FILE);
2144 
2145 	isize = i_size_read(inode);
2146 	fi->original_i_size = isize;
2147 	if (truncate) {
2148 		set_inode_flag(inode, FI_ATOMIC_REPLACE);
2149 		truncate_inode_pages_final(inode->i_mapping);
2150 		f2fs_i_size_write(inode, 0);
2151 		isize = 0;
2152 	}
2153 	f2fs_i_size_write(fi->cow_inode, isize);
2154 
2155 	f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2156 
2157 	f2fs_update_time(sbi, REQ_TIME);
2158 	fi->atomic_write_task = current;
2159 	stat_update_max_atomic_write(inode);
2160 	fi->atomic_write_cnt = 0;
2161 out:
2162 	inode_unlock(inode);
2163 	mnt_drop_write_file(filp);
2164 	return ret;
2165 }
2166 
f2fs_ioc_commit_atomic_write(struct file * filp)2167 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2168 {
2169 	struct inode *inode = file_inode(filp);
2170 	struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2171 	int ret;
2172 
2173 	if (!inode_owner_or_capable(mnt_userns, inode))
2174 		return -EACCES;
2175 
2176 	ret = mnt_want_write_file(filp);
2177 	if (ret)
2178 		return ret;
2179 
2180 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2181 
2182 	inode_lock(inode);
2183 
2184 	if (f2fs_is_atomic_file(inode)) {
2185 		ret = f2fs_commit_atomic_write(inode);
2186 		if (!ret)
2187 			ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2188 
2189 		f2fs_abort_atomic_write(inode, ret);
2190 	} else {
2191 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2192 	}
2193 
2194 	inode_unlock(inode);
2195 	mnt_drop_write_file(filp);
2196 	return ret;
2197 }
2198 
f2fs_ioc_abort_atomic_write(struct file * filp)2199 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2200 {
2201 	struct inode *inode = file_inode(filp);
2202 	struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2203 	int ret;
2204 
2205 	if (!inode_owner_or_capable(mnt_userns, inode))
2206 		return -EACCES;
2207 
2208 	ret = mnt_want_write_file(filp);
2209 	if (ret)
2210 		return ret;
2211 
2212 	inode_lock(inode);
2213 
2214 	f2fs_abort_atomic_write(inode, true);
2215 
2216 	inode_unlock(inode);
2217 
2218 	mnt_drop_write_file(filp);
2219 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2220 	return ret;
2221 }
2222 
f2fs_ioc_shutdown(struct file * filp,unsigned long arg)2223 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2224 {
2225 	struct inode *inode = file_inode(filp);
2226 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2227 	struct super_block *sb = sbi->sb;
2228 	__u32 in;
2229 	int ret = 0;
2230 
2231 	if (!capable(CAP_SYS_ADMIN))
2232 		return -EPERM;
2233 
2234 	if (get_user(in, (__u32 __user *)arg))
2235 		return -EFAULT;
2236 
2237 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2238 		ret = mnt_want_write_file(filp);
2239 		if (ret) {
2240 			if (ret == -EROFS) {
2241 				ret = 0;
2242 				f2fs_stop_checkpoint(sbi, false,
2243 						STOP_CP_REASON_SHUTDOWN);
2244 				set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2245 				trace_f2fs_shutdown(sbi, in, ret);
2246 			}
2247 			return ret;
2248 		}
2249 	}
2250 
2251 	switch (in) {
2252 	case F2FS_GOING_DOWN_FULLSYNC:
2253 		ret = freeze_bdev(sb->s_bdev);
2254 		if (ret)
2255 			goto out;
2256 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2257 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2258 		thaw_bdev(sb->s_bdev);
2259 		break;
2260 	case F2FS_GOING_DOWN_METASYNC:
2261 		/* do checkpoint only */
2262 		ret = f2fs_sync_fs(sb, 1);
2263 		if (ret)
2264 			goto out;
2265 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2266 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2267 		break;
2268 	case F2FS_GOING_DOWN_NOSYNC:
2269 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2270 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2271 		break;
2272 	case F2FS_GOING_DOWN_METAFLUSH:
2273 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2274 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2275 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2276 		break;
2277 	case F2FS_GOING_DOWN_NEED_FSCK:
2278 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2279 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2280 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2281 		/* do checkpoint only */
2282 		ret = f2fs_sync_fs(sb, 1);
2283 		goto out;
2284 	default:
2285 		ret = -EINVAL;
2286 		goto out;
2287 	}
2288 
2289 	f2fs_stop_gc_thread(sbi);
2290 	f2fs_stop_discard_thread(sbi);
2291 
2292 	f2fs_drop_discard_cmd(sbi);
2293 	clear_opt(sbi, DISCARD);
2294 
2295 	f2fs_update_time(sbi, REQ_TIME);
2296 out:
2297 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2298 		mnt_drop_write_file(filp);
2299 
2300 	trace_f2fs_shutdown(sbi, in, ret);
2301 
2302 	return ret;
2303 }
2304 
f2fs_ioc_fitrim(struct file * filp,unsigned long arg)2305 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2306 {
2307 	struct inode *inode = file_inode(filp);
2308 	struct super_block *sb = inode->i_sb;
2309 	struct fstrim_range range;
2310 	int ret;
2311 
2312 	if (!capable(CAP_SYS_ADMIN))
2313 		return -EPERM;
2314 
2315 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2316 		return -EOPNOTSUPP;
2317 
2318 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2319 				sizeof(range)))
2320 		return -EFAULT;
2321 
2322 	ret = mnt_want_write_file(filp);
2323 	if (ret)
2324 		return ret;
2325 
2326 	range.minlen = max((unsigned int)range.minlen,
2327 			   bdev_discard_granularity(sb->s_bdev));
2328 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2329 	mnt_drop_write_file(filp);
2330 	if (ret < 0)
2331 		return ret;
2332 
2333 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2334 				sizeof(range)))
2335 		return -EFAULT;
2336 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2337 	return 0;
2338 }
2339 
uuid_is_nonzero(__u8 u[16])2340 static bool uuid_is_nonzero(__u8 u[16])
2341 {
2342 	int i;
2343 
2344 	for (i = 0; i < 16; i++)
2345 		if (u[i])
2346 			return true;
2347 	return false;
2348 }
2349 
f2fs_ioc_set_encryption_policy(struct file * filp,unsigned long arg)2350 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2351 {
2352 	struct inode *inode = file_inode(filp);
2353 
2354 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2355 		return -EOPNOTSUPP;
2356 
2357 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2358 
2359 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2360 }
2361 
f2fs_ioc_get_encryption_policy(struct file * filp,unsigned long arg)2362 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2363 {
2364 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2365 		return -EOPNOTSUPP;
2366 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2367 }
2368 
f2fs_ioc_get_encryption_pwsalt(struct file * filp,unsigned long arg)2369 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2370 {
2371 	struct inode *inode = file_inode(filp);
2372 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2373 	u8 encrypt_pw_salt[16];
2374 	int err;
2375 
2376 	if (!f2fs_sb_has_encrypt(sbi))
2377 		return -EOPNOTSUPP;
2378 
2379 	err = mnt_want_write_file(filp);
2380 	if (err)
2381 		return err;
2382 
2383 	f2fs_down_write(&sbi->sb_lock);
2384 
2385 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2386 		goto got_it;
2387 
2388 	/* update superblock with uuid */
2389 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2390 
2391 	err = f2fs_commit_super(sbi, false);
2392 	if (err) {
2393 		/* undo new data */
2394 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2395 		goto out_err;
2396 	}
2397 got_it:
2398 	memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2399 out_err:
2400 	f2fs_up_write(&sbi->sb_lock);
2401 	mnt_drop_write_file(filp);
2402 
2403 	if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2404 		err = -EFAULT;
2405 
2406 	return err;
2407 }
2408 
f2fs_ioc_get_encryption_policy_ex(struct file * filp,unsigned long arg)2409 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2410 					     unsigned long arg)
2411 {
2412 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2413 		return -EOPNOTSUPP;
2414 
2415 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2416 }
2417 
f2fs_ioc_add_encryption_key(struct file * filp,unsigned long arg)2418 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2419 {
2420 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2421 		return -EOPNOTSUPP;
2422 
2423 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2424 }
2425 
f2fs_ioc_remove_encryption_key(struct file * filp,unsigned long arg)2426 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2427 {
2428 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2429 		return -EOPNOTSUPP;
2430 
2431 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2432 }
2433 
f2fs_ioc_remove_encryption_key_all_users(struct file * filp,unsigned long arg)2434 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2435 						    unsigned long arg)
2436 {
2437 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2438 		return -EOPNOTSUPP;
2439 
2440 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2441 }
2442 
f2fs_ioc_get_encryption_key_status(struct file * filp,unsigned long arg)2443 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2444 					      unsigned long arg)
2445 {
2446 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2447 		return -EOPNOTSUPP;
2448 
2449 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2450 }
2451 
f2fs_ioc_get_encryption_nonce(struct file * filp,unsigned long arg)2452 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2453 {
2454 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2455 		return -EOPNOTSUPP;
2456 
2457 	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2458 }
2459 
f2fs_ioc_gc(struct file * filp,unsigned long arg)2460 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2461 {
2462 	struct inode *inode = file_inode(filp);
2463 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2464 	struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2465 			.no_bg_gc = false,
2466 			.should_migrate_blocks = false,
2467 			.nr_free_secs = 0 };
2468 	__u32 sync;
2469 	int ret;
2470 
2471 	if (!capable(CAP_SYS_ADMIN))
2472 		return -EPERM;
2473 
2474 	if (get_user(sync, (__u32 __user *)arg))
2475 		return -EFAULT;
2476 
2477 	if (f2fs_readonly(sbi->sb))
2478 		return -EROFS;
2479 
2480 	ret = mnt_want_write_file(filp);
2481 	if (ret)
2482 		return ret;
2483 
2484 	if (!sync) {
2485 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2486 			ret = -EBUSY;
2487 			goto out;
2488 		}
2489 	} else {
2490 		f2fs_down_write(&sbi->gc_lock);
2491 	}
2492 
2493 	gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2494 	gc_control.err_gc_skipped = sync;
2495 	ret = f2fs_gc(sbi, &gc_control);
2496 out:
2497 	mnt_drop_write_file(filp);
2498 	return ret;
2499 }
2500 
__f2fs_ioc_gc_range(struct file * filp,struct f2fs_gc_range * range)2501 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2502 {
2503 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2504 	struct f2fs_gc_control gc_control = {
2505 			.init_gc_type = range->sync ? FG_GC : BG_GC,
2506 			.no_bg_gc = false,
2507 			.should_migrate_blocks = false,
2508 			.err_gc_skipped = range->sync,
2509 			.nr_free_secs = 0 };
2510 	u64 end;
2511 	int ret;
2512 
2513 	if (!capable(CAP_SYS_ADMIN))
2514 		return -EPERM;
2515 	if (f2fs_readonly(sbi->sb))
2516 		return -EROFS;
2517 
2518 	end = range->start + range->len;
2519 	if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2520 					end >= MAX_BLKADDR(sbi))
2521 		return -EINVAL;
2522 
2523 	ret = mnt_want_write_file(filp);
2524 	if (ret)
2525 		return ret;
2526 
2527 do_more:
2528 	if (!range->sync) {
2529 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2530 			ret = -EBUSY;
2531 			goto out;
2532 		}
2533 	} else {
2534 		f2fs_down_write(&sbi->gc_lock);
2535 	}
2536 
2537 	gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2538 	ret = f2fs_gc(sbi, &gc_control);
2539 	if (ret) {
2540 		if (ret == -EBUSY)
2541 			ret = -EAGAIN;
2542 		goto out;
2543 	}
2544 	range->start += CAP_BLKS_PER_SEC(sbi);
2545 	if (range->start <= end)
2546 		goto do_more;
2547 out:
2548 	mnt_drop_write_file(filp);
2549 	return ret;
2550 }
2551 
f2fs_ioc_gc_range(struct file * filp,unsigned long arg)2552 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2553 {
2554 	struct f2fs_gc_range range;
2555 
2556 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2557 							sizeof(range)))
2558 		return -EFAULT;
2559 	return __f2fs_ioc_gc_range(filp, &range);
2560 }
2561 
f2fs_ioc_write_checkpoint(struct file * filp)2562 static int f2fs_ioc_write_checkpoint(struct file *filp)
2563 {
2564 	struct inode *inode = file_inode(filp);
2565 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2566 	int ret;
2567 
2568 	if (!capable(CAP_SYS_ADMIN))
2569 		return -EPERM;
2570 
2571 	if (f2fs_readonly(sbi->sb))
2572 		return -EROFS;
2573 
2574 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2575 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2576 		return -EINVAL;
2577 	}
2578 
2579 	ret = mnt_want_write_file(filp);
2580 	if (ret)
2581 		return ret;
2582 
2583 	ret = f2fs_sync_fs(sbi->sb, 1);
2584 
2585 	mnt_drop_write_file(filp);
2586 	return ret;
2587 }
2588 
f2fs_defragment_range(struct f2fs_sb_info * sbi,struct file * filp,struct f2fs_defragment * range)2589 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2590 					struct file *filp,
2591 					struct f2fs_defragment *range)
2592 {
2593 	struct inode *inode = file_inode(filp);
2594 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2595 					.m_seg_type = NO_CHECK_TYPE,
2596 					.m_may_create = false };
2597 	struct extent_info ei = {};
2598 	pgoff_t pg_start, pg_end, next_pgofs;
2599 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2600 	unsigned int total = 0, sec_num;
2601 	block_t blk_end = 0;
2602 	bool fragmented = false;
2603 	int err;
2604 
2605 	pg_start = range->start >> PAGE_SHIFT;
2606 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2607 
2608 	f2fs_balance_fs(sbi, true);
2609 
2610 	inode_lock(inode);
2611 
2612 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
2613 		err = -EINVAL;
2614 		goto unlock_out;
2615 	}
2616 
2617 	/* if in-place-update policy is enabled, don't waste time here */
2618 	set_inode_flag(inode, FI_OPU_WRITE);
2619 	if (f2fs_should_update_inplace(inode, NULL)) {
2620 		err = -EINVAL;
2621 		goto out;
2622 	}
2623 
2624 	/* writeback all dirty pages in the range */
2625 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2626 						range->start + range->len - 1);
2627 	if (err)
2628 		goto out;
2629 
2630 	/*
2631 	 * lookup mapping info in extent cache, skip defragmenting if physical
2632 	 * block addresses are continuous.
2633 	 */
2634 	if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2635 		if (ei.fofs + ei.len >= pg_end)
2636 			goto out;
2637 	}
2638 
2639 	map.m_lblk = pg_start;
2640 	map.m_next_pgofs = &next_pgofs;
2641 
2642 	/*
2643 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2644 	 * physical block addresses are continuous even if there are hole(s)
2645 	 * in logical blocks.
2646 	 */
2647 	while (map.m_lblk < pg_end) {
2648 		map.m_len = pg_end - map.m_lblk;
2649 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2650 		if (err)
2651 			goto out;
2652 
2653 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2654 			map.m_lblk = next_pgofs;
2655 			continue;
2656 		}
2657 
2658 		if (blk_end && blk_end != map.m_pblk)
2659 			fragmented = true;
2660 
2661 		/* record total count of block that we're going to move */
2662 		total += map.m_len;
2663 
2664 		blk_end = map.m_pblk + map.m_len;
2665 
2666 		map.m_lblk += map.m_len;
2667 	}
2668 
2669 	if (!fragmented) {
2670 		total = 0;
2671 		goto out;
2672 	}
2673 
2674 	sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2675 
2676 	/*
2677 	 * make sure there are enough free section for LFS allocation, this can
2678 	 * avoid defragment running in SSR mode when free section are allocated
2679 	 * intensively
2680 	 */
2681 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2682 		err = -EAGAIN;
2683 		goto out;
2684 	}
2685 
2686 	map.m_lblk = pg_start;
2687 	map.m_len = pg_end - pg_start;
2688 	total = 0;
2689 
2690 	while (map.m_lblk < pg_end) {
2691 		pgoff_t idx;
2692 		int cnt = 0;
2693 
2694 do_map:
2695 		map.m_len = pg_end - map.m_lblk;
2696 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2697 		if (err)
2698 			goto clear_out;
2699 
2700 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2701 			map.m_lblk = next_pgofs;
2702 			goto check;
2703 		}
2704 
2705 		set_inode_flag(inode, FI_SKIP_WRITES);
2706 
2707 		idx = map.m_lblk;
2708 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2709 			struct page *page;
2710 
2711 			page = f2fs_get_lock_data_page(inode, idx, true);
2712 			if (IS_ERR(page)) {
2713 				err = PTR_ERR(page);
2714 				goto clear_out;
2715 			}
2716 
2717 			set_page_dirty(page);
2718 			set_page_private_gcing(page);
2719 			f2fs_put_page(page, 1);
2720 
2721 			idx++;
2722 			cnt++;
2723 			total++;
2724 		}
2725 
2726 		map.m_lblk = idx;
2727 check:
2728 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
2729 			goto do_map;
2730 
2731 		clear_inode_flag(inode, FI_SKIP_WRITES);
2732 
2733 		err = filemap_fdatawrite(inode->i_mapping);
2734 		if (err)
2735 			goto out;
2736 	}
2737 clear_out:
2738 	clear_inode_flag(inode, FI_SKIP_WRITES);
2739 out:
2740 	clear_inode_flag(inode, FI_OPU_WRITE);
2741 unlock_out:
2742 	inode_unlock(inode);
2743 	if (!err)
2744 		range->len = (u64)total << PAGE_SHIFT;
2745 	return err;
2746 }
2747 
f2fs_ioc_defragment(struct file * filp,unsigned long arg)2748 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2749 {
2750 	struct inode *inode = file_inode(filp);
2751 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2752 	struct f2fs_defragment range;
2753 	int err;
2754 
2755 	if (!capable(CAP_SYS_ADMIN))
2756 		return -EPERM;
2757 
2758 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2759 		return -EINVAL;
2760 
2761 	if (f2fs_readonly(sbi->sb))
2762 		return -EROFS;
2763 
2764 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2765 							sizeof(range)))
2766 		return -EFAULT;
2767 
2768 	/* verify alignment of offset & size */
2769 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2770 		return -EINVAL;
2771 
2772 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2773 					max_file_blocks(inode)))
2774 		return -EINVAL;
2775 
2776 	err = mnt_want_write_file(filp);
2777 	if (err)
2778 		return err;
2779 
2780 	err = f2fs_defragment_range(sbi, filp, &range);
2781 	mnt_drop_write_file(filp);
2782 
2783 	f2fs_update_time(sbi, REQ_TIME);
2784 	if (err < 0)
2785 		return err;
2786 
2787 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2788 							sizeof(range)))
2789 		return -EFAULT;
2790 
2791 	return 0;
2792 }
2793 
f2fs_move_file_range(struct file * file_in,loff_t pos_in,struct file * file_out,loff_t pos_out,size_t len)2794 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2795 			struct file *file_out, loff_t pos_out, size_t len)
2796 {
2797 	struct inode *src = file_inode(file_in);
2798 	struct inode *dst = file_inode(file_out);
2799 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2800 	size_t olen = len, dst_max_i_size = 0;
2801 	size_t dst_osize;
2802 	int ret;
2803 
2804 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2805 				src->i_sb != dst->i_sb)
2806 		return -EXDEV;
2807 
2808 	if (unlikely(f2fs_readonly(src->i_sb)))
2809 		return -EROFS;
2810 
2811 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2812 		return -EINVAL;
2813 
2814 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2815 		return -EOPNOTSUPP;
2816 
2817 	if (pos_out < 0 || pos_in < 0)
2818 		return -EINVAL;
2819 
2820 	if (src == dst) {
2821 		if (pos_in == pos_out)
2822 			return 0;
2823 		if (pos_out > pos_in && pos_out < pos_in + len)
2824 			return -EINVAL;
2825 	}
2826 
2827 	inode_lock(src);
2828 	if (src != dst) {
2829 		ret = -EBUSY;
2830 		if (!inode_trylock(dst))
2831 			goto out;
2832 	}
2833 
2834 	if (f2fs_compressed_file(src) || f2fs_compressed_file(dst)) {
2835 		ret = -EOPNOTSUPP;
2836 		goto out_unlock;
2837 	}
2838 
2839 	ret = -EINVAL;
2840 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2841 		goto out_unlock;
2842 	if (len == 0)
2843 		olen = len = src->i_size - pos_in;
2844 	if (pos_in + len == src->i_size)
2845 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2846 	if (len == 0) {
2847 		ret = 0;
2848 		goto out_unlock;
2849 	}
2850 
2851 	dst_osize = dst->i_size;
2852 	if (pos_out + olen > dst->i_size)
2853 		dst_max_i_size = pos_out + olen;
2854 
2855 	/* verify the end result is block aligned */
2856 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2857 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2858 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2859 		goto out_unlock;
2860 
2861 	ret = f2fs_convert_inline_inode(src);
2862 	if (ret)
2863 		goto out_unlock;
2864 
2865 	ret = f2fs_convert_inline_inode(dst);
2866 	if (ret)
2867 		goto out_unlock;
2868 
2869 	/* write out all dirty pages from offset */
2870 	ret = filemap_write_and_wait_range(src->i_mapping,
2871 					pos_in, pos_in + len);
2872 	if (ret)
2873 		goto out_unlock;
2874 
2875 	ret = filemap_write_and_wait_range(dst->i_mapping,
2876 					pos_out, pos_out + len);
2877 	if (ret)
2878 		goto out_unlock;
2879 
2880 	f2fs_balance_fs(sbi, true);
2881 
2882 	f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2883 	if (src != dst) {
2884 		ret = -EBUSY;
2885 		if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2886 			goto out_src;
2887 	}
2888 
2889 	f2fs_lock_op(sbi);
2890 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2891 				pos_out >> F2FS_BLKSIZE_BITS,
2892 				len >> F2FS_BLKSIZE_BITS, false);
2893 
2894 	if (!ret) {
2895 		if (dst_max_i_size)
2896 			f2fs_i_size_write(dst, dst_max_i_size);
2897 		else if (dst_osize != dst->i_size)
2898 			f2fs_i_size_write(dst, dst_osize);
2899 	}
2900 	f2fs_unlock_op(sbi);
2901 
2902 	if (src != dst)
2903 		f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2904 out_src:
2905 	f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2906 out_unlock:
2907 	if (src != dst)
2908 		inode_unlock(dst);
2909 out:
2910 	inode_unlock(src);
2911 	return ret;
2912 }
2913 
__f2fs_ioc_move_range(struct file * filp,struct f2fs_move_range * range)2914 static int __f2fs_ioc_move_range(struct file *filp,
2915 				struct f2fs_move_range *range)
2916 {
2917 	struct fd dst;
2918 	int err;
2919 
2920 	if (!(filp->f_mode & FMODE_READ) ||
2921 			!(filp->f_mode & FMODE_WRITE))
2922 		return -EBADF;
2923 
2924 	dst = fdget(range->dst_fd);
2925 	if (!dst.file)
2926 		return -EBADF;
2927 
2928 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2929 		err = -EBADF;
2930 		goto err_out;
2931 	}
2932 
2933 	err = mnt_want_write_file(filp);
2934 	if (err)
2935 		goto err_out;
2936 
2937 	err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2938 					range->pos_out, range->len);
2939 
2940 	mnt_drop_write_file(filp);
2941 err_out:
2942 	fdput(dst);
2943 	return err;
2944 }
2945 
f2fs_ioc_move_range(struct file * filp,unsigned long arg)2946 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2947 {
2948 	struct f2fs_move_range range;
2949 
2950 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2951 							sizeof(range)))
2952 		return -EFAULT;
2953 	return __f2fs_ioc_move_range(filp, &range);
2954 }
2955 
f2fs_ioc_flush_device(struct file * filp,unsigned long arg)2956 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2957 {
2958 	struct inode *inode = file_inode(filp);
2959 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2960 	struct sit_info *sm = SIT_I(sbi);
2961 	unsigned int start_segno = 0, end_segno = 0;
2962 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2963 	struct f2fs_flush_device range;
2964 	struct f2fs_gc_control gc_control = {
2965 			.init_gc_type = FG_GC,
2966 			.should_migrate_blocks = true,
2967 			.err_gc_skipped = true,
2968 			.nr_free_secs = 0 };
2969 	int ret;
2970 
2971 	if (!capable(CAP_SYS_ADMIN))
2972 		return -EPERM;
2973 
2974 	if (f2fs_readonly(sbi->sb))
2975 		return -EROFS;
2976 
2977 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2978 		return -EINVAL;
2979 
2980 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2981 							sizeof(range)))
2982 		return -EFAULT;
2983 
2984 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2985 			__is_large_section(sbi)) {
2986 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2987 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2988 		return -EINVAL;
2989 	}
2990 
2991 	ret = mnt_want_write_file(filp);
2992 	if (ret)
2993 		return ret;
2994 
2995 	if (range.dev_num != 0)
2996 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2997 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2998 
2999 	start_segno = sm->last_victim[FLUSH_DEVICE];
3000 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
3001 		start_segno = dev_start_segno;
3002 	end_segno = min(start_segno + range.segments, dev_end_segno);
3003 
3004 	while (start_segno < end_segno) {
3005 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
3006 			ret = -EBUSY;
3007 			goto out;
3008 		}
3009 		sm->last_victim[GC_CB] = end_segno + 1;
3010 		sm->last_victim[GC_GREEDY] = end_segno + 1;
3011 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
3012 
3013 		gc_control.victim_segno = start_segno;
3014 		ret = f2fs_gc(sbi, &gc_control);
3015 		if (ret == -EAGAIN)
3016 			ret = 0;
3017 		else if (ret < 0)
3018 			break;
3019 		start_segno++;
3020 	}
3021 out:
3022 	mnt_drop_write_file(filp);
3023 	return ret;
3024 }
3025 
f2fs_ioc_get_features(struct file * filp,unsigned long arg)3026 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3027 {
3028 	struct inode *inode = file_inode(filp);
3029 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3030 
3031 	/* Must validate to set it with SQLite behavior in Android. */
3032 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3033 
3034 	return put_user(sb_feature, (u32 __user *)arg);
3035 }
3036 
3037 #ifdef CONFIG_QUOTA
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3038 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3039 {
3040 	struct dquot *transfer_to[MAXQUOTAS] = {};
3041 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3042 	struct super_block *sb = sbi->sb;
3043 	int err;
3044 
3045 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3046 	if (IS_ERR(transfer_to[PRJQUOTA]))
3047 		return PTR_ERR(transfer_to[PRJQUOTA]);
3048 
3049 	err = __dquot_transfer(inode, transfer_to);
3050 	if (err)
3051 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3052 	dqput(transfer_to[PRJQUOTA]);
3053 	return err;
3054 }
3055 
f2fs_ioc_setproject(struct inode * inode,__u32 projid)3056 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3057 {
3058 	struct f2fs_inode_info *fi = F2FS_I(inode);
3059 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3060 	struct f2fs_inode *ri = NULL;
3061 	kprojid_t kprojid;
3062 	int err;
3063 
3064 	if (!f2fs_sb_has_project_quota(sbi)) {
3065 		if (projid != F2FS_DEF_PROJID)
3066 			return -EOPNOTSUPP;
3067 		else
3068 			return 0;
3069 	}
3070 
3071 	if (!f2fs_has_extra_attr(inode))
3072 		return -EOPNOTSUPP;
3073 
3074 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3075 
3076 	if (projid_eq(kprojid, fi->i_projid))
3077 		return 0;
3078 
3079 	err = -EPERM;
3080 	/* Is it quota file? Do not allow user to mess with it */
3081 	if (IS_NOQUOTA(inode))
3082 		return err;
3083 
3084 	if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3085 		return -EOVERFLOW;
3086 
3087 	err = f2fs_dquot_initialize(inode);
3088 	if (err)
3089 		return err;
3090 
3091 	f2fs_lock_op(sbi);
3092 	err = f2fs_transfer_project_quota(inode, kprojid);
3093 	if (err)
3094 		goto out_unlock;
3095 
3096 	fi->i_projid = kprojid;
3097 	inode->i_ctime = current_time(inode);
3098 	f2fs_mark_inode_dirty_sync(inode, true);
3099 out_unlock:
3100 	f2fs_unlock_op(sbi);
3101 	return err;
3102 }
3103 #else
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3104 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3105 {
3106 	return 0;
3107 }
3108 
f2fs_ioc_setproject(struct inode * inode,__u32 projid)3109 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3110 {
3111 	if (projid != F2FS_DEF_PROJID)
3112 		return -EOPNOTSUPP;
3113 	return 0;
3114 }
3115 #endif
3116 
f2fs_fileattr_get(struct dentry * dentry,struct fileattr * fa)3117 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3118 {
3119 	struct inode *inode = d_inode(dentry);
3120 	struct f2fs_inode_info *fi = F2FS_I(inode);
3121 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3122 
3123 	if (IS_ENCRYPTED(inode))
3124 		fsflags |= FS_ENCRYPT_FL;
3125 	if (IS_VERITY(inode))
3126 		fsflags |= FS_VERITY_FL;
3127 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3128 		fsflags |= FS_INLINE_DATA_FL;
3129 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3130 		fsflags |= FS_NOCOW_FL;
3131 
3132 	fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3133 
3134 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3135 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3136 
3137 	return 0;
3138 }
3139 
f2fs_fileattr_set(struct user_namespace * mnt_userns,struct dentry * dentry,struct fileattr * fa)3140 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3141 		      struct dentry *dentry, struct fileattr *fa)
3142 {
3143 	struct inode *inode = d_inode(dentry);
3144 	u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3145 	u32 iflags;
3146 	int err;
3147 
3148 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3149 		return -EIO;
3150 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3151 		return -ENOSPC;
3152 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
3153 		return -EOPNOTSUPP;
3154 	fsflags &= F2FS_SETTABLE_FS_FL;
3155 	if (!fa->flags_valid)
3156 		mask &= FS_COMMON_FL;
3157 
3158 	iflags = f2fs_fsflags_to_iflags(fsflags);
3159 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3160 		return -EOPNOTSUPP;
3161 
3162 	err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3163 	if (!err)
3164 		err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3165 
3166 	return err;
3167 }
3168 
f2fs_pin_file_control(struct inode * inode,bool inc)3169 int f2fs_pin_file_control(struct inode *inode, bool inc)
3170 {
3171 	struct f2fs_inode_info *fi = F2FS_I(inode);
3172 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3173 
3174 	/* Use i_gc_failures for normal file as a risk signal. */
3175 	if (inc)
3176 		f2fs_i_gc_failures_write(inode,
3177 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3178 
3179 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3180 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3181 			  __func__, inode->i_ino,
3182 			  fi->i_gc_failures[GC_FAILURE_PIN]);
3183 		clear_inode_flag(inode, FI_PIN_FILE);
3184 		return -EAGAIN;
3185 	}
3186 	return 0;
3187 }
3188 
f2fs_ioc_set_pin_file(struct file * filp,unsigned long arg)3189 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3190 {
3191 	struct inode *inode = file_inode(filp);
3192 	__u32 pin;
3193 	int ret = 0;
3194 
3195 	if (get_user(pin, (__u32 __user *)arg))
3196 		return -EFAULT;
3197 
3198 	if (!S_ISREG(inode->i_mode))
3199 		return -EINVAL;
3200 
3201 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3202 		return -EROFS;
3203 
3204 	ret = mnt_want_write_file(filp);
3205 	if (ret)
3206 		return ret;
3207 
3208 	inode_lock(inode);
3209 
3210 	if (!pin) {
3211 		clear_inode_flag(inode, FI_PIN_FILE);
3212 		f2fs_i_gc_failures_write(inode, 0);
3213 		goto done;
3214 	}
3215 
3216 	if (f2fs_should_update_outplace(inode, NULL)) {
3217 		ret = -EINVAL;
3218 		goto out;
3219 	}
3220 
3221 	if (f2fs_pin_file_control(inode, false)) {
3222 		ret = -EAGAIN;
3223 		goto out;
3224 	}
3225 
3226 	ret = f2fs_convert_inline_inode(inode);
3227 	if (ret)
3228 		goto out;
3229 
3230 	if (!f2fs_disable_compressed_file(inode)) {
3231 		ret = -EOPNOTSUPP;
3232 		goto out;
3233 	}
3234 
3235 	set_inode_flag(inode, FI_PIN_FILE);
3236 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3237 done:
3238 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3239 out:
3240 	inode_unlock(inode);
3241 	mnt_drop_write_file(filp);
3242 	return ret;
3243 }
3244 
f2fs_ioc_get_pin_file(struct file * filp,unsigned long arg)3245 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3246 {
3247 	struct inode *inode = file_inode(filp);
3248 	__u32 pin = 0;
3249 
3250 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3251 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3252 	return put_user(pin, (u32 __user *)arg);
3253 }
3254 
f2fs_precache_extents(struct inode * inode)3255 int f2fs_precache_extents(struct inode *inode)
3256 {
3257 	struct f2fs_inode_info *fi = F2FS_I(inode);
3258 	struct f2fs_map_blocks map;
3259 	pgoff_t m_next_extent;
3260 	loff_t end;
3261 	int err;
3262 
3263 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3264 		return -EOPNOTSUPP;
3265 
3266 	map.m_lblk = 0;
3267 	map.m_pblk = 0;
3268 	map.m_next_pgofs = NULL;
3269 	map.m_next_extent = &m_next_extent;
3270 	map.m_seg_type = NO_CHECK_TYPE;
3271 	map.m_may_create = false;
3272 	end = max_file_blocks(inode);
3273 
3274 	while (map.m_lblk < end) {
3275 		map.m_len = end - map.m_lblk;
3276 
3277 		f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3278 		err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
3279 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3280 		if (err)
3281 			return err;
3282 
3283 		map.m_lblk = m_next_extent;
3284 	}
3285 
3286 	return 0;
3287 }
3288 
f2fs_ioc_precache_extents(struct file * filp)3289 static int f2fs_ioc_precache_extents(struct file *filp)
3290 {
3291 	return f2fs_precache_extents(file_inode(filp));
3292 }
3293 
f2fs_ioc_resize_fs(struct file * filp,unsigned long arg)3294 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3295 {
3296 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3297 	__u64 block_count;
3298 
3299 	if (!capable(CAP_SYS_ADMIN))
3300 		return -EPERM;
3301 
3302 	if (f2fs_readonly(sbi->sb))
3303 		return -EROFS;
3304 
3305 	if (copy_from_user(&block_count, (void __user *)arg,
3306 			   sizeof(block_count)))
3307 		return -EFAULT;
3308 
3309 	return f2fs_resize_fs(filp, block_count);
3310 }
3311 
f2fs_ioc_enable_verity(struct file * filp,unsigned long arg)3312 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3313 {
3314 	struct inode *inode = file_inode(filp);
3315 
3316 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3317 
3318 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3319 		f2fs_warn(F2FS_I_SB(inode),
3320 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3321 			  inode->i_ino);
3322 		return -EOPNOTSUPP;
3323 	}
3324 
3325 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3326 }
3327 
f2fs_ioc_measure_verity(struct file * filp,unsigned long arg)3328 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3329 {
3330 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3331 		return -EOPNOTSUPP;
3332 
3333 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3334 }
3335 
f2fs_ioc_read_verity_metadata(struct file * filp,unsigned long arg)3336 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3337 {
3338 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3339 		return -EOPNOTSUPP;
3340 
3341 	return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3342 }
3343 
f2fs_ioc_getfslabel(struct file * filp,unsigned long arg)3344 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3345 {
3346 	struct inode *inode = file_inode(filp);
3347 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3348 	char *vbuf;
3349 	int count;
3350 	int err = 0;
3351 
3352 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3353 	if (!vbuf)
3354 		return -ENOMEM;
3355 
3356 	f2fs_down_read(&sbi->sb_lock);
3357 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3358 			ARRAY_SIZE(sbi->raw_super->volume_name),
3359 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3360 	f2fs_up_read(&sbi->sb_lock);
3361 
3362 	if (copy_to_user((char __user *)arg, vbuf,
3363 				min(FSLABEL_MAX, count)))
3364 		err = -EFAULT;
3365 
3366 	kfree(vbuf);
3367 	return err;
3368 }
3369 
f2fs_ioc_setfslabel(struct file * filp,unsigned long arg)3370 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3371 {
3372 	struct inode *inode = file_inode(filp);
3373 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3374 	char *vbuf;
3375 	int err = 0;
3376 
3377 	if (!capable(CAP_SYS_ADMIN))
3378 		return -EPERM;
3379 
3380 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3381 	if (IS_ERR(vbuf))
3382 		return PTR_ERR(vbuf);
3383 
3384 	err = mnt_want_write_file(filp);
3385 	if (err)
3386 		goto out;
3387 
3388 	f2fs_down_write(&sbi->sb_lock);
3389 
3390 	memset(sbi->raw_super->volume_name, 0,
3391 			sizeof(sbi->raw_super->volume_name));
3392 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3393 			sbi->raw_super->volume_name,
3394 			ARRAY_SIZE(sbi->raw_super->volume_name));
3395 
3396 	err = f2fs_commit_super(sbi, false);
3397 
3398 	f2fs_up_write(&sbi->sb_lock);
3399 
3400 	mnt_drop_write_file(filp);
3401 out:
3402 	kfree(vbuf);
3403 	return err;
3404 }
3405 
f2fs_get_compress_blocks(struct file * filp,unsigned long arg)3406 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3407 {
3408 	struct inode *inode = file_inode(filp);
3409 	__u64 blocks;
3410 
3411 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3412 		return -EOPNOTSUPP;
3413 
3414 	if (!f2fs_compressed_file(inode))
3415 		return -EINVAL;
3416 
3417 	blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3418 	return put_user(blocks, (u64 __user *)arg);
3419 }
3420 
release_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3421 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3422 {
3423 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3424 	unsigned int released_blocks = 0;
3425 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3426 	block_t blkaddr;
3427 	int i;
3428 
3429 	for (i = 0; i < count; i++) {
3430 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3431 						dn->ofs_in_node + i);
3432 
3433 		if (!__is_valid_data_blkaddr(blkaddr))
3434 			continue;
3435 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3436 					DATA_GENERIC_ENHANCE))) {
3437 			f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3438 			return -EFSCORRUPTED;
3439 		}
3440 	}
3441 
3442 	while (count) {
3443 		int compr_blocks = 0;
3444 
3445 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3446 			blkaddr = f2fs_data_blkaddr(dn);
3447 
3448 			if (i == 0) {
3449 				if (blkaddr == COMPRESS_ADDR)
3450 					continue;
3451 				dn->ofs_in_node += cluster_size;
3452 				goto next;
3453 			}
3454 
3455 			if (__is_valid_data_blkaddr(blkaddr))
3456 				compr_blocks++;
3457 
3458 			if (blkaddr != NEW_ADDR)
3459 				continue;
3460 
3461 			dn->data_blkaddr = NULL_ADDR;
3462 			f2fs_set_data_blkaddr(dn);
3463 		}
3464 
3465 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3466 		dec_valid_block_count(sbi, dn->inode,
3467 					cluster_size - compr_blocks);
3468 
3469 		released_blocks += cluster_size - compr_blocks;
3470 next:
3471 		count -= cluster_size;
3472 	}
3473 
3474 	return released_blocks;
3475 }
3476 
f2fs_release_compress_blocks(struct file * filp,unsigned long arg)3477 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3478 {
3479 	struct inode *inode = file_inode(filp);
3480 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3481 	pgoff_t page_idx = 0, last_idx;
3482 	unsigned int released_blocks = 0;
3483 	int ret;
3484 	int writecount;
3485 
3486 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3487 		return -EOPNOTSUPP;
3488 
3489 	if (!f2fs_compressed_file(inode))
3490 		return -EINVAL;
3491 
3492 	if (f2fs_readonly(sbi->sb))
3493 		return -EROFS;
3494 
3495 	ret = mnt_want_write_file(filp);
3496 	if (ret)
3497 		return ret;
3498 
3499 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3500 
3501 	inode_lock(inode);
3502 
3503 	writecount = atomic_read(&inode->i_writecount);
3504 	if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3505 			(!(filp->f_mode & FMODE_WRITE) && writecount)) {
3506 		ret = -EBUSY;
3507 		goto out;
3508 	}
3509 
3510 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3511 		ret = -EINVAL;
3512 		goto out;
3513 	}
3514 
3515 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3516 	if (ret)
3517 		goto out;
3518 
3519 	set_inode_flag(inode, FI_COMPRESS_RELEASED);
3520 	inode->i_ctime = current_time(inode);
3521 	f2fs_mark_inode_dirty_sync(inode, true);
3522 
3523 	if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3524 		goto out;
3525 
3526 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3527 	filemap_invalidate_lock(inode->i_mapping);
3528 
3529 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3530 
3531 	while (page_idx < last_idx) {
3532 		struct dnode_of_data dn;
3533 		pgoff_t end_offset, count;
3534 
3535 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3536 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3537 		if (ret) {
3538 			if (ret == -ENOENT) {
3539 				page_idx = f2fs_get_next_page_offset(&dn,
3540 								page_idx);
3541 				ret = 0;
3542 				continue;
3543 			}
3544 			break;
3545 		}
3546 
3547 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3548 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3549 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3550 
3551 		ret = release_compress_blocks(&dn, count);
3552 
3553 		f2fs_put_dnode(&dn);
3554 
3555 		if (ret < 0)
3556 			break;
3557 
3558 		page_idx += count;
3559 		released_blocks += ret;
3560 	}
3561 
3562 	filemap_invalidate_unlock(inode->i_mapping);
3563 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3564 out:
3565 	inode_unlock(inode);
3566 
3567 	mnt_drop_write_file(filp);
3568 
3569 	if (ret >= 0) {
3570 		ret = put_user(released_blocks, (u64 __user *)arg);
3571 	} else if (released_blocks &&
3572 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3573 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3574 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3575 			"iblocks=%llu, released=%u, compr_blocks=%u, "
3576 			"run fsck to fix.",
3577 			__func__, inode->i_ino, inode->i_blocks,
3578 			released_blocks,
3579 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
3580 	}
3581 
3582 	return ret;
3583 }
3584 
reserve_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3585 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3586 {
3587 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3588 	unsigned int reserved_blocks = 0;
3589 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3590 	block_t blkaddr;
3591 	int i;
3592 
3593 	for (i = 0; i < count; i++) {
3594 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3595 						dn->ofs_in_node + i);
3596 
3597 		if (!__is_valid_data_blkaddr(blkaddr))
3598 			continue;
3599 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3600 					DATA_GENERIC_ENHANCE))) {
3601 			f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3602 			return -EFSCORRUPTED;
3603 		}
3604 	}
3605 
3606 	while (count) {
3607 		int compr_blocks = 0;
3608 		blkcnt_t reserved;
3609 		int ret;
3610 
3611 		for (i = 0; i < cluster_size; i++) {
3612 			blkaddr = data_blkaddr(dn->inode, dn->node_page,
3613 						dn->ofs_in_node + i);
3614 
3615 			if (i == 0) {
3616 				if (blkaddr != COMPRESS_ADDR) {
3617 					dn->ofs_in_node += cluster_size;
3618 					goto next;
3619 				}
3620 				continue;
3621 			}
3622 
3623 			/*
3624 			 * compressed cluster was not released due to it
3625 			 * fails in release_compress_blocks(), so NEW_ADDR
3626 			 * is a possible case.
3627 			 */
3628 			if (blkaddr == NEW_ADDR ||
3629 				__is_valid_data_blkaddr(blkaddr)) {
3630 				compr_blocks++;
3631 				continue;
3632 			}
3633 		}
3634 
3635 		reserved = cluster_size - compr_blocks;
3636 
3637 		/* for the case all blocks in cluster were reserved */
3638 		if (reserved == 1)
3639 			goto next;
3640 
3641 		ret = inc_valid_block_count(sbi, dn->inode, &reserved, false);
3642 		if (unlikely(ret))
3643 			return ret;
3644 
3645 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3646 			if (f2fs_data_blkaddr(dn) == NULL_ADDR) {
3647 				dn->data_blkaddr = NEW_ADDR;
3648 				f2fs_set_data_blkaddr(dn);
3649 			}
3650 		}
3651 
3652 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3653 
3654 		reserved_blocks += reserved;
3655 next:
3656 		count -= cluster_size;
3657 	}
3658 
3659 	return reserved_blocks;
3660 }
3661 
f2fs_reserve_compress_blocks(struct file * filp,unsigned long arg)3662 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3663 {
3664 	struct inode *inode = file_inode(filp);
3665 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3666 	pgoff_t page_idx = 0, last_idx;
3667 	unsigned int reserved_blocks = 0;
3668 	int ret;
3669 
3670 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3671 		return -EOPNOTSUPP;
3672 
3673 	if (!f2fs_compressed_file(inode))
3674 		return -EINVAL;
3675 
3676 	if (f2fs_readonly(sbi->sb))
3677 		return -EROFS;
3678 
3679 	ret = mnt_want_write_file(filp);
3680 	if (ret)
3681 		return ret;
3682 
3683 	if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3684 		goto out;
3685 
3686 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3687 
3688 	inode_lock(inode);
3689 
3690 	if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3691 		ret = -EINVAL;
3692 		goto unlock_inode;
3693 	}
3694 
3695 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3696 	filemap_invalidate_lock(inode->i_mapping);
3697 
3698 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3699 
3700 	while (page_idx < last_idx) {
3701 		struct dnode_of_data dn;
3702 		pgoff_t end_offset, count;
3703 
3704 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3705 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3706 		if (ret) {
3707 			if (ret == -ENOENT) {
3708 				page_idx = f2fs_get_next_page_offset(&dn,
3709 								page_idx);
3710 				ret = 0;
3711 				continue;
3712 			}
3713 			break;
3714 		}
3715 
3716 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3717 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3718 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3719 
3720 		ret = reserve_compress_blocks(&dn, count);
3721 
3722 		f2fs_put_dnode(&dn);
3723 
3724 		if (ret < 0)
3725 			break;
3726 
3727 		page_idx += count;
3728 		reserved_blocks += ret;
3729 	}
3730 
3731 	filemap_invalidate_unlock(inode->i_mapping);
3732 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3733 
3734 	if (ret >= 0) {
3735 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3736 		inode->i_ctime = current_time(inode);
3737 		f2fs_mark_inode_dirty_sync(inode, true);
3738 	}
3739 unlock_inode:
3740 	inode_unlock(inode);
3741 out:
3742 	mnt_drop_write_file(filp);
3743 
3744 	if (ret >= 0) {
3745 		ret = put_user(reserved_blocks, (u64 __user *)arg);
3746 	} else if (reserved_blocks &&
3747 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3748 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3749 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3750 			"iblocks=%llu, reserved=%u, compr_blocks=%u, "
3751 			"run fsck to fix.",
3752 			__func__, inode->i_ino, inode->i_blocks,
3753 			reserved_blocks,
3754 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
3755 	}
3756 
3757 	return ret;
3758 }
3759 
f2fs_secure_erase(struct block_device * bdev,struct inode * inode,pgoff_t off,block_t block,block_t len,u32 flags)3760 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3761 		pgoff_t off, block_t block, block_t len, u32 flags)
3762 {
3763 	sector_t sector = SECTOR_FROM_BLOCK(block);
3764 	sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3765 	int ret = 0;
3766 
3767 	if (flags & F2FS_TRIM_FILE_DISCARD) {
3768 		if (bdev_max_secure_erase_sectors(bdev))
3769 			ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3770 					GFP_NOFS);
3771 		else
3772 			ret = blkdev_issue_discard(bdev, sector, nr_sects,
3773 					GFP_NOFS);
3774 	}
3775 
3776 	if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3777 		if (IS_ENCRYPTED(inode))
3778 			ret = fscrypt_zeroout_range(inode, off, block, len);
3779 		else
3780 			ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3781 					GFP_NOFS, 0);
3782 	}
3783 
3784 	return ret;
3785 }
3786 
f2fs_sec_trim_file(struct file * filp,unsigned long arg)3787 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3788 {
3789 	struct inode *inode = file_inode(filp);
3790 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3791 	struct address_space *mapping = inode->i_mapping;
3792 	struct block_device *prev_bdev = NULL;
3793 	struct f2fs_sectrim_range range;
3794 	pgoff_t index, pg_end, prev_index = 0;
3795 	block_t prev_block = 0, len = 0;
3796 	loff_t end_addr;
3797 	bool to_end = false;
3798 	int ret = 0;
3799 
3800 	if (!(filp->f_mode & FMODE_WRITE))
3801 		return -EBADF;
3802 
3803 	if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3804 				sizeof(range)))
3805 		return -EFAULT;
3806 
3807 	if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3808 			!S_ISREG(inode->i_mode))
3809 		return -EINVAL;
3810 
3811 	if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3812 			!f2fs_hw_support_discard(sbi)) ||
3813 			((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3814 			 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3815 		return -EOPNOTSUPP;
3816 
3817 	file_start_write(filp);
3818 	inode_lock(inode);
3819 
3820 	if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3821 			range.start >= inode->i_size) {
3822 		ret = -EINVAL;
3823 		goto err;
3824 	}
3825 
3826 	if (range.len == 0)
3827 		goto err;
3828 
3829 	if (inode->i_size - range.start > range.len) {
3830 		end_addr = range.start + range.len;
3831 	} else {
3832 		end_addr = range.len == (u64)-1 ?
3833 			sbi->sb->s_maxbytes : inode->i_size;
3834 		to_end = true;
3835 	}
3836 
3837 	if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3838 			(!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3839 		ret = -EINVAL;
3840 		goto err;
3841 	}
3842 
3843 	index = F2FS_BYTES_TO_BLK(range.start);
3844 	pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3845 
3846 	ret = f2fs_convert_inline_inode(inode);
3847 	if (ret)
3848 		goto err;
3849 
3850 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3851 	filemap_invalidate_lock(mapping);
3852 
3853 	ret = filemap_write_and_wait_range(mapping, range.start,
3854 			to_end ? LLONG_MAX : end_addr - 1);
3855 	if (ret)
3856 		goto out;
3857 
3858 	truncate_inode_pages_range(mapping, range.start,
3859 			to_end ? -1 : end_addr - 1);
3860 
3861 	while (index < pg_end) {
3862 		struct dnode_of_data dn;
3863 		pgoff_t end_offset, count;
3864 		int i;
3865 
3866 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3867 		ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3868 		if (ret) {
3869 			if (ret == -ENOENT) {
3870 				index = f2fs_get_next_page_offset(&dn, index);
3871 				continue;
3872 			}
3873 			goto out;
3874 		}
3875 
3876 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3877 		count = min(end_offset - dn.ofs_in_node, pg_end - index);
3878 		for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3879 			struct block_device *cur_bdev;
3880 			block_t blkaddr = f2fs_data_blkaddr(&dn);
3881 
3882 			if (!__is_valid_data_blkaddr(blkaddr))
3883 				continue;
3884 
3885 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3886 						DATA_GENERIC_ENHANCE)) {
3887 				ret = -EFSCORRUPTED;
3888 				f2fs_put_dnode(&dn);
3889 				f2fs_handle_error(sbi,
3890 						ERROR_INVALID_BLKADDR);
3891 				goto out;
3892 			}
3893 
3894 			cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3895 			if (f2fs_is_multi_device(sbi)) {
3896 				int di = f2fs_target_device_index(sbi, blkaddr);
3897 
3898 				blkaddr -= FDEV(di).start_blk;
3899 			}
3900 
3901 			if (len) {
3902 				if (prev_bdev == cur_bdev &&
3903 						index == prev_index + len &&
3904 						blkaddr == prev_block + len) {
3905 					len++;
3906 				} else {
3907 					ret = f2fs_secure_erase(prev_bdev,
3908 						inode, prev_index, prev_block,
3909 						len, range.flags);
3910 					if (ret) {
3911 						f2fs_put_dnode(&dn);
3912 						goto out;
3913 					}
3914 
3915 					len = 0;
3916 				}
3917 			}
3918 
3919 			if (!len) {
3920 				prev_bdev = cur_bdev;
3921 				prev_index = index;
3922 				prev_block = blkaddr;
3923 				len = 1;
3924 			}
3925 		}
3926 
3927 		f2fs_put_dnode(&dn);
3928 
3929 		if (fatal_signal_pending(current)) {
3930 			ret = -EINTR;
3931 			goto out;
3932 		}
3933 		cond_resched();
3934 	}
3935 
3936 	if (len)
3937 		ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3938 				prev_block, len, range.flags);
3939 out:
3940 	filemap_invalidate_unlock(mapping);
3941 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3942 err:
3943 	inode_unlock(inode);
3944 	file_end_write(filp);
3945 
3946 	return ret;
3947 }
3948 
f2fs_ioc_get_compress_option(struct file * filp,unsigned long arg)3949 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3950 {
3951 	struct inode *inode = file_inode(filp);
3952 	struct f2fs_comp_option option;
3953 
3954 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3955 		return -EOPNOTSUPP;
3956 
3957 	inode_lock_shared(inode);
3958 
3959 	if (!f2fs_compressed_file(inode)) {
3960 		inode_unlock_shared(inode);
3961 		return -ENODATA;
3962 	}
3963 
3964 	option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3965 	option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3966 
3967 	inode_unlock_shared(inode);
3968 
3969 	if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3970 				sizeof(option)))
3971 		return -EFAULT;
3972 
3973 	return 0;
3974 }
3975 
f2fs_ioc_set_compress_option(struct file * filp,unsigned long arg)3976 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3977 {
3978 	struct inode *inode = file_inode(filp);
3979 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3980 	struct f2fs_comp_option option;
3981 	int ret = 0;
3982 
3983 	if (!f2fs_sb_has_compression(sbi))
3984 		return -EOPNOTSUPP;
3985 
3986 	if (!(filp->f_mode & FMODE_WRITE))
3987 		return -EBADF;
3988 
3989 	if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3990 				sizeof(option)))
3991 		return -EFAULT;
3992 
3993 	if (!f2fs_compressed_file(inode) ||
3994 			option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3995 			option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3996 			option.algorithm >= COMPRESS_MAX)
3997 		return -EINVAL;
3998 
3999 	file_start_write(filp);
4000 	inode_lock(inode);
4001 
4002 	f2fs_down_write(&F2FS_I(inode)->i_sem);
4003 	if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4004 		ret = -EBUSY;
4005 		goto out;
4006 	}
4007 
4008 	if (F2FS_HAS_BLOCKS(inode)) {
4009 		ret = -EFBIG;
4010 		goto out;
4011 	}
4012 
4013 	F2FS_I(inode)->i_compress_algorithm = option.algorithm;
4014 	F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
4015 	F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size);
4016 	/* Set default level */
4017 	if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD)
4018 		F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
4019 	else
4020 		F2FS_I(inode)->i_compress_level = 0;
4021 	/* Adjust mount option level */
4022 	if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm &&
4023 	    F2FS_OPTION(sbi).compress_level)
4024 		F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level;
4025 	f2fs_mark_inode_dirty_sync(inode, true);
4026 
4027 	if (!f2fs_is_compress_backend_ready(inode))
4028 		f2fs_warn(sbi, "compression algorithm is successfully set, "
4029 			"but current kernel doesn't support this algorithm.");
4030 out:
4031 	f2fs_up_write(&F2FS_I(inode)->i_sem);
4032 	inode_unlock(inode);
4033 	file_end_write(filp);
4034 
4035 	return ret;
4036 }
4037 
redirty_blocks(struct inode * inode,pgoff_t page_idx,int len)4038 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4039 {
4040 	DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
4041 	struct address_space *mapping = inode->i_mapping;
4042 	struct page *page;
4043 	pgoff_t redirty_idx = page_idx;
4044 	int i, page_len = 0, ret = 0;
4045 
4046 	page_cache_ra_unbounded(&ractl, len, 0);
4047 
4048 	for (i = 0; i < len; i++, page_idx++) {
4049 		page = read_cache_page(mapping, page_idx, NULL, NULL);
4050 		if (IS_ERR(page)) {
4051 			ret = PTR_ERR(page);
4052 			break;
4053 		}
4054 		page_len++;
4055 	}
4056 
4057 	for (i = 0; i < page_len; i++, redirty_idx++) {
4058 		page = find_lock_page(mapping, redirty_idx);
4059 
4060 		/* It will never fail, when page has pinned above */
4061 		f2fs_bug_on(F2FS_I_SB(inode), !page);
4062 
4063 		set_page_dirty(page);
4064 		set_page_private_gcing(page);
4065 		f2fs_put_page(page, 1);
4066 		f2fs_put_page(page, 0);
4067 	}
4068 
4069 	return ret;
4070 }
4071 
f2fs_ioc_decompress_file(struct file * filp)4072 static int f2fs_ioc_decompress_file(struct file *filp)
4073 {
4074 	struct inode *inode = file_inode(filp);
4075 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4076 	struct f2fs_inode_info *fi = F2FS_I(inode);
4077 	pgoff_t page_idx = 0, last_idx;
4078 	unsigned int blk_per_seg = sbi->blocks_per_seg;
4079 	int cluster_size = fi->i_cluster_size;
4080 	int count, ret;
4081 
4082 	if (!f2fs_sb_has_compression(sbi) ||
4083 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4084 		return -EOPNOTSUPP;
4085 
4086 	if (!(filp->f_mode & FMODE_WRITE))
4087 		return -EBADF;
4088 
4089 	if (!f2fs_compressed_file(inode))
4090 		return -EINVAL;
4091 
4092 	f2fs_balance_fs(F2FS_I_SB(inode), true);
4093 
4094 	file_start_write(filp);
4095 	inode_lock(inode);
4096 
4097 	if (!f2fs_is_compress_backend_ready(inode)) {
4098 		ret = -EOPNOTSUPP;
4099 		goto out;
4100 	}
4101 
4102 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4103 		ret = -EINVAL;
4104 		goto out;
4105 	}
4106 
4107 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4108 	if (ret)
4109 		goto out;
4110 
4111 	if (!atomic_read(&fi->i_compr_blocks))
4112 		goto out;
4113 
4114 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4115 
4116 	count = last_idx - page_idx;
4117 	while (count) {
4118 		int len = min(cluster_size, count);
4119 
4120 		ret = redirty_blocks(inode, page_idx, len);
4121 		if (ret < 0)
4122 			break;
4123 
4124 		if (get_dirty_pages(inode) >= blk_per_seg) {
4125 			ret = filemap_fdatawrite(inode->i_mapping);
4126 			if (ret < 0)
4127 				break;
4128 		}
4129 
4130 		count -= len;
4131 		page_idx += len;
4132 	}
4133 
4134 	if (!ret)
4135 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4136 							LLONG_MAX);
4137 
4138 	if (ret)
4139 		f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4140 			  __func__, ret);
4141 out:
4142 	inode_unlock(inode);
4143 	file_end_write(filp);
4144 
4145 	return ret;
4146 }
4147 
f2fs_ioc_compress_file(struct file * filp)4148 static int f2fs_ioc_compress_file(struct file *filp)
4149 {
4150 	struct inode *inode = file_inode(filp);
4151 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4152 	pgoff_t page_idx = 0, last_idx;
4153 	unsigned int blk_per_seg = sbi->blocks_per_seg;
4154 	int cluster_size = F2FS_I(inode)->i_cluster_size;
4155 	int count, ret;
4156 
4157 	if (!f2fs_sb_has_compression(sbi) ||
4158 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4159 		return -EOPNOTSUPP;
4160 
4161 	if (!(filp->f_mode & FMODE_WRITE))
4162 		return -EBADF;
4163 
4164 	if (!f2fs_compressed_file(inode))
4165 		return -EINVAL;
4166 
4167 	f2fs_balance_fs(F2FS_I_SB(inode), true);
4168 
4169 	file_start_write(filp);
4170 	inode_lock(inode);
4171 
4172 	if (!f2fs_is_compress_backend_ready(inode)) {
4173 		ret = -EOPNOTSUPP;
4174 		goto out;
4175 	}
4176 
4177 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4178 		ret = -EINVAL;
4179 		goto out;
4180 	}
4181 
4182 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4183 	if (ret)
4184 		goto out;
4185 
4186 	set_inode_flag(inode, FI_ENABLE_COMPRESS);
4187 
4188 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4189 
4190 	count = last_idx - page_idx;
4191 	while (count) {
4192 		int len = min(cluster_size, count);
4193 
4194 		ret = redirty_blocks(inode, page_idx, len);
4195 		if (ret < 0)
4196 			break;
4197 
4198 		if (get_dirty_pages(inode) >= blk_per_seg) {
4199 			ret = filemap_fdatawrite(inode->i_mapping);
4200 			if (ret < 0)
4201 				break;
4202 		}
4203 
4204 		count -= len;
4205 		page_idx += len;
4206 	}
4207 
4208 	if (!ret)
4209 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4210 							LLONG_MAX);
4211 
4212 	clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4213 
4214 	if (ret)
4215 		f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4216 			  __func__, ret);
4217 out:
4218 	inode_unlock(inode);
4219 	file_end_write(filp);
4220 
4221 	return ret;
4222 }
4223 
__f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4224 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4225 {
4226 	switch (cmd) {
4227 	case FS_IOC_GETVERSION:
4228 		return f2fs_ioc_getversion(filp, arg);
4229 	case F2FS_IOC_START_ATOMIC_WRITE:
4230 		return f2fs_ioc_start_atomic_write(filp, false);
4231 	case F2FS_IOC_START_ATOMIC_REPLACE:
4232 		return f2fs_ioc_start_atomic_write(filp, true);
4233 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4234 		return f2fs_ioc_commit_atomic_write(filp);
4235 	case F2FS_IOC_ABORT_ATOMIC_WRITE:
4236 		return f2fs_ioc_abort_atomic_write(filp);
4237 	case F2FS_IOC_START_VOLATILE_WRITE:
4238 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4239 		return -EOPNOTSUPP;
4240 	case F2FS_IOC_SHUTDOWN:
4241 		return f2fs_ioc_shutdown(filp, arg);
4242 	case FITRIM:
4243 		return f2fs_ioc_fitrim(filp, arg);
4244 	case FS_IOC_SET_ENCRYPTION_POLICY:
4245 		return f2fs_ioc_set_encryption_policy(filp, arg);
4246 	case FS_IOC_GET_ENCRYPTION_POLICY:
4247 		return f2fs_ioc_get_encryption_policy(filp, arg);
4248 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4249 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4250 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4251 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4252 	case FS_IOC_ADD_ENCRYPTION_KEY:
4253 		return f2fs_ioc_add_encryption_key(filp, arg);
4254 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4255 		return f2fs_ioc_remove_encryption_key(filp, arg);
4256 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4257 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4258 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4259 		return f2fs_ioc_get_encryption_key_status(filp, arg);
4260 	case FS_IOC_GET_ENCRYPTION_NONCE:
4261 		return f2fs_ioc_get_encryption_nonce(filp, arg);
4262 	case F2FS_IOC_GARBAGE_COLLECT:
4263 		return f2fs_ioc_gc(filp, arg);
4264 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4265 		return f2fs_ioc_gc_range(filp, arg);
4266 	case F2FS_IOC_WRITE_CHECKPOINT:
4267 		return f2fs_ioc_write_checkpoint(filp);
4268 	case F2FS_IOC_DEFRAGMENT:
4269 		return f2fs_ioc_defragment(filp, arg);
4270 	case F2FS_IOC_MOVE_RANGE:
4271 		return f2fs_ioc_move_range(filp, arg);
4272 	case F2FS_IOC_FLUSH_DEVICE:
4273 		return f2fs_ioc_flush_device(filp, arg);
4274 	case F2FS_IOC_GET_FEATURES:
4275 		return f2fs_ioc_get_features(filp, arg);
4276 	case F2FS_IOC_GET_PIN_FILE:
4277 		return f2fs_ioc_get_pin_file(filp, arg);
4278 	case F2FS_IOC_SET_PIN_FILE:
4279 		return f2fs_ioc_set_pin_file(filp, arg);
4280 	case F2FS_IOC_PRECACHE_EXTENTS:
4281 		return f2fs_ioc_precache_extents(filp);
4282 	case F2FS_IOC_RESIZE_FS:
4283 		return f2fs_ioc_resize_fs(filp, arg);
4284 	case FS_IOC_ENABLE_VERITY:
4285 		return f2fs_ioc_enable_verity(filp, arg);
4286 	case FS_IOC_MEASURE_VERITY:
4287 		return f2fs_ioc_measure_verity(filp, arg);
4288 	case FS_IOC_READ_VERITY_METADATA:
4289 		return f2fs_ioc_read_verity_metadata(filp, arg);
4290 	case FS_IOC_GETFSLABEL:
4291 		return f2fs_ioc_getfslabel(filp, arg);
4292 	case FS_IOC_SETFSLABEL:
4293 		return f2fs_ioc_setfslabel(filp, arg);
4294 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4295 		return f2fs_get_compress_blocks(filp, arg);
4296 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4297 		return f2fs_release_compress_blocks(filp, arg);
4298 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4299 		return f2fs_reserve_compress_blocks(filp, arg);
4300 	case F2FS_IOC_SEC_TRIM_FILE:
4301 		return f2fs_sec_trim_file(filp, arg);
4302 	case F2FS_IOC_GET_COMPRESS_OPTION:
4303 		return f2fs_ioc_get_compress_option(filp, arg);
4304 	case F2FS_IOC_SET_COMPRESS_OPTION:
4305 		return f2fs_ioc_set_compress_option(filp, arg);
4306 	case F2FS_IOC_DECOMPRESS_FILE:
4307 		return f2fs_ioc_decompress_file(filp);
4308 	case F2FS_IOC_COMPRESS_FILE:
4309 		return f2fs_ioc_compress_file(filp);
4310 	default:
4311 		return -ENOTTY;
4312 	}
4313 }
4314 
f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4315 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4316 {
4317 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4318 		return -EIO;
4319 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4320 		return -ENOSPC;
4321 
4322 	return __f2fs_ioctl(filp, cmd, arg);
4323 }
4324 
4325 /*
4326  * Return %true if the given read or write request should use direct I/O, or
4327  * %false if it should use buffered I/O.
4328  */
f2fs_should_use_dio(struct inode * inode,struct kiocb * iocb,struct iov_iter * iter)4329 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4330 				struct iov_iter *iter)
4331 {
4332 	unsigned int align;
4333 
4334 	if (!(iocb->ki_flags & IOCB_DIRECT))
4335 		return false;
4336 
4337 	if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4338 		return false;
4339 
4340 	/*
4341 	 * Direct I/O not aligned to the disk's logical_block_size will be
4342 	 * attempted, but will fail with -EINVAL.
4343 	 *
4344 	 * f2fs additionally requires that direct I/O be aligned to the
4345 	 * filesystem block size, which is often a stricter requirement.
4346 	 * However, f2fs traditionally falls back to buffered I/O on requests
4347 	 * that are logical_block_size-aligned but not fs-block aligned.
4348 	 *
4349 	 * The below logic implements this behavior.
4350 	 */
4351 	align = iocb->ki_pos | iov_iter_alignment(iter);
4352 	if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4353 	    IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4354 		return false;
4355 
4356 	return true;
4357 }
4358 
f2fs_dio_read_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4359 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4360 				unsigned int flags)
4361 {
4362 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4363 
4364 	dec_page_count(sbi, F2FS_DIO_READ);
4365 	if (error)
4366 		return error;
4367 	f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4368 	return 0;
4369 }
4370 
4371 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4372 	.end_io = f2fs_dio_read_end_io,
4373 };
4374 
f2fs_dio_read_iter(struct kiocb * iocb,struct iov_iter * to)4375 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4376 {
4377 	struct file *file = iocb->ki_filp;
4378 	struct inode *inode = file_inode(file);
4379 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4380 	struct f2fs_inode_info *fi = F2FS_I(inode);
4381 	const loff_t pos = iocb->ki_pos;
4382 	const size_t count = iov_iter_count(to);
4383 	struct iomap_dio *dio;
4384 	ssize_t ret;
4385 
4386 	if (count == 0)
4387 		return 0; /* skip atime update */
4388 
4389 	trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4390 
4391 	if (iocb->ki_flags & IOCB_NOWAIT) {
4392 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4393 			ret = -EAGAIN;
4394 			goto out;
4395 		}
4396 	} else {
4397 		f2fs_down_read(&fi->i_gc_rwsem[READ]);
4398 	}
4399 
4400 	/*
4401 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4402 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4403 	 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4404 	 */
4405 	inc_page_count(sbi, F2FS_DIO_READ);
4406 	dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4407 			     &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4408 	if (IS_ERR_OR_NULL(dio)) {
4409 		ret = PTR_ERR_OR_ZERO(dio);
4410 		if (ret != -EIOCBQUEUED)
4411 			dec_page_count(sbi, F2FS_DIO_READ);
4412 	} else {
4413 		ret = iomap_dio_complete(dio);
4414 	}
4415 
4416 	f2fs_up_read(&fi->i_gc_rwsem[READ]);
4417 
4418 	file_accessed(file);
4419 out:
4420 	trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4421 	return ret;
4422 }
4423 
f2fs_trace_rw_file_path(struct kiocb * iocb,size_t count,int rw)4424 static void f2fs_trace_rw_file_path(struct kiocb *iocb, size_t count, int rw)
4425 {
4426 	struct inode *inode = file_inode(iocb->ki_filp);
4427 	char *buf, *path;
4428 
4429 	buf = f2fs_getname(F2FS_I_SB(inode));
4430 	if (!buf)
4431 		return;
4432 	path = dentry_path_raw(file_dentry(iocb->ki_filp), buf, PATH_MAX);
4433 	if (IS_ERR(path))
4434 		goto free_buf;
4435 	if (rw == WRITE)
4436 		trace_f2fs_datawrite_start(inode, iocb->ki_pos, count,
4437 				current->pid, path, current->comm);
4438 	else
4439 		trace_f2fs_dataread_start(inode, iocb->ki_pos, count,
4440 				current->pid, path, current->comm);
4441 free_buf:
4442 	f2fs_putname(buf);
4443 }
4444 
f2fs_file_read_iter(struct kiocb * iocb,struct iov_iter * to)4445 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4446 {
4447 	struct inode *inode = file_inode(iocb->ki_filp);
4448 	const loff_t pos = iocb->ki_pos;
4449 	ssize_t ret;
4450 
4451 	if (!f2fs_is_compress_backend_ready(inode))
4452 		return -EOPNOTSUPP;
4453 
4454 	if (trace_f2fs_dataread_start_enabled())
4455 		f2fs_trace_rw_file_path(iocb, iov_iter_count(to), READ);
4456 
4457 	if (f2fs_should_use_dio(inode, iocb, to)) {
4458 		ret = f2fs_dio_read_iter(iocb, to);
4459 	} else {
4460 		ret = filemap_read(iocb, to, 0);
4461 		if (ret > 0)
4462 			f2fs_update_iostat(F2FS_I_SB(inode), inode,
4463 						APP_BUFFERED_READ_IO, ret);
4464 	}
4465 	if (trace_f2fs_dataread_end_enabled())
4466 		trace_f2fs_dataread_end(inode, pos, ret);
4467 	return ret;
4468 }
4469 
f2fs_write_checks(struct kiocb * iocb,struct iov_iter * from)4470 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4471 {
4472 	struct file *file = iocb->ki_filp;
4473 	struct inode *inode = file_inode(file);
4474 	ssize_t count;
4475 	int err;
4476 
4477 	if (IS_IMMUTABLE(inode))
4478 		return -EPERM;
4479 
4480 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4481 		return -EPERM;
4482 
4483 	count = generic_write_checks(iocb, from);
4484 	if (count <= 0)
4485 		return count;
4486 
4487 	err = file_modified(file);
4488 	if (err)
4489 		return err;
4490 	return count;
4491 }
4492 
4493 /*
4494  * Preallocate blocks for a write request, if it is possible and helpful to do
4495  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4496  * blocks were preallocated, or a negative errno value if something went
4497  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4498  * requested blocks (not just some of them) have been allocated.
4499  */
f2fs_preallocate_blocks(struct kiocb * iocb,struct iov_iter * iter,bool dio)4500 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4501 				   bool dio)
4502 {
4503 	struct inode *inode = file_inode(iocb->ki_filp);
4504 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4505 	const loff_t pos = iocb->ki_pos;
4506 	const size_t count = iov_iter_count(iter);
4507 	struct f2fs_map_blocks map = {};
4508 	int flag;
4509 	int ret;
4510 
4511 	/* If it will be an out-of-place direct write, don't bother. */
4512 	if (dio && f2fs_lfs_mode(sbi))
4513 		return 0;
4514 	/*
4515 	 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4516 	 * buffered IO, if DIO meets any holes.
4517 	 */
4518 	if (dio && i_size_read(inode) &&
4519 		(F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4520 		return 0;
4521 
4522 	/* No-wait I/O can't allocate blocks. */
4523 	if (iocb->ki_flags & IOCB_NOWAIT)
4524 		return 0;
4525 
4526 	/* If it will be a short write, don't bother. */
4527 	if (fault_in_iov_iter_readable(iter, count))
4528 		return 0;
4529 
4530 	if (f2fs_has_inline_data(inode)) {
4531 		/* If the data will fit inline, don't bother. */
4532 		if (pos + count <= MAX_INLINE_DATA(inode))
4533 			return 0;
4534 		ret = f2fs_convert_inline_inode(inode);
4535 		if (ret)
4536 			return ret;
4537 	}
4538 
4539 	/* Do not preallocate blocks that will be written partially in 4KB. */
4540 	map.m_lblk = F2FS_BLK_ALIGN(pos);
4541 	map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4542 	if (map.m_len > map.m_lblk)
4543 		map.m_len -= map.m_lblk;
4544 	else
4545 		map.m_len = 0;
4546 	map.m_may_create = true;
4547 	if (dio) {
4548 		map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4549 		flag = F2FS_GET_BLOCK_PRE_DIO;
4550 	} else {
4551 		map.m_seg_type = NO_CHECK_TYPE;
4552 		flag = F2FS_GET_BLOCK_PRE_AIO;
4553 	}
4554 
4555 	ret = f2fs_map_blocks(inode, &map, flag);
4556 	/* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4557 	if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4558 		return ret;
4559 	if (ret == 0)
4560 		set_inode_flag(inode, FI_PREALLOCATED_ALL);
4561 	return map.m_len;
4562 }
4563 
f2fs_buffered_write_iter(struct kiocb * iocb,struct iov_iter * from)4564 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4565 					struct iov_iter *from)
4566 {
4567 	struct file *file = iocb->ki_filp;
4568 	struct inode *inode = file_inode(file);
4569 	ssize_t ret;
4570 
4571 	if (iocb->ki_flags & IOCB_NOWAIT)
4572 		return -EOPNOTSUPP;
4573 
4574 	current->backing_dev_info = inode_to_bdi(inode);
4575 	ret = generic_perform_write(iocb, from);
4576 	current->backing_dev_info = NULL;
4577 
4578 	if (ret > 0) {
4579 		iocb->ki_pos += ret;
4580 		f2fs_update_iostat(F2FS_I_SB(inode), inode,
4581 						APP_BUFFERED_IO, ret);
4582 	}
4583 	return ret;
4584 }
4585 
f2fs_dio_write_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4586 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4587 				 unsigned int flags)
4588 {
4589 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4590 
4591 	dec_page_count(sbi, F2FS_DIO_WRITE);
4592 	if (error)
4593 		return error;
4594 	f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4595 	return 0;
4596 }
4597 
4598 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4599 	.end_io = f2fs_dio_write_end_io,
4600 };
4601 
f2fs_flush_buffered_write(struct address_space * mapping,loff_t start_pos,loff_t end_pos)4602 static void f2fs_flush_buffered_write(struct address_space *mapping,
4603 				      loff_t start_pos, loff_t end_pos)
4604 {
4605 	int ret;
4606 
4607 	ret = filemap_write_and_wait_range(mapping, start_pos, end_pos);
4608 	if (ret < 0)
4609 		return;
4610 	invalidate_mapping_pages(mapping,
4611 				 start_pos >> PAGE_SHIFT,
4612 				 end_pos >> PAGE_SHIFT);
4613 }
4614 
f2fs_dio_write_iter(struct kiocb * iocb,struct iov_iter * from,bool * may_need_sync)4615 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4616 				   bool *may_need_sync)
4617 {
4618 	struct file *file = iocb->ki_filp;
4619 	struct inode *inode = file_inode(file);
4620 	struct f2fs_inode_info *fi = F2FS_I(inode);
4621 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4622 	const bool do_opu = f2fs_lfs_mode(sbi);
4623 	const loff_t pos = iocb->ki_pos;
4624 	const ssize_t count = iov_iter_count(from);
4625 	unsigned int dio_flags;
4626 	struct iomap_dio *dio;
4627 	ssize_t ret;
4628 
4629 	trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4630 
4631 	if (iocb->ki_flags & IOCB_NOWAIT) {
4632 		/* f2fs_convert_inline_inode() and block allocation can block */
4633 		if (f2fs_has_inline_data(inode) ||
4634 		    !f2fs_overwrite_io(inode, pos, count)) {
4635 			ret = -EAGAIN;
4636 			goto out;
4637 		}
4638 
4639 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4640 			ret = -EAGAIN;
4641 			goto out;
4642 		}
4643 		if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4644 			f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4645 			ret = -EAGAIN;
4646 			goto out;
4647 		}
4648 	} else {
4649 		ret = f2fs_convert_inline_inode(inode);
4650 		if (ret)
4651 			goto out;
4652 
4653 		f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4654 		if (do_opu)
4655 			f2fs_down_read(&fi->i_gc_rwsem[READ]);
4656 	}
4657 
4658 	/*
4659 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4660 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4661 	 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4662 	 */
4663 	inc_page_count(sbi, F2FS_DIO_WRITE);
4664 	dio_flags = 0;
4665 	if (pos + count > inode->i_size)
4666 		dio_flags |= IOMAP_DIO_FORCE_WAIT;
4667 	dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4668 			     &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4669 	if (IS_ERR_OR_NULL(dio)) {
4670 		ret = PTR_ERR_OR_ZERO(dio);
4671 		if (ret == -ENOTBLK)
4672 			ret = 0;
4673 		if (ret != -EIOCBQUEUED)
4674 			dec_page_count(sbi, F2FS_DIO_WRITE);
4675 	} else {
4676 		ret = iomap_dio_complete(dio);
4677 	}
4678 
4679 	if (do_opu)
4680 		f2fs_up_read(&fi->i_gc_rwsem[READ]);
4681 	f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4682 
4683 	if (ret < 0)
4684 		goto out;
4685 	if (pos + ret > inode->i_size)
4686 		f2fs_i_size_write(inode, pos + ret);
4687 	if (!do_opu)
4688 		set_inode_flag(inode, FI_UPDATE_WRITE);
4689 
4690 	if (iov_iter_count(from)) {
4691 		ssize_t ret2;
4692 		loff_t bufio_start_pos = iocb->ki_pos;
4693 
4694 		/*
4695 		 * The direct write was partial, so we need to fall back to a
4696 		 * buffered write for the remainder.
4697 		 */
4698 
4699 		ret2 = f2fs_buffered_write_iter(iocb, from);
4700 		if (iov_iter_count(from))
4701 			f2fs_write_failed(inode, iocb->ki_pos);
4702 		if (ret2 < 0)
4703 			goto out;
4704 
4705 		/*
4706 		 * Ensure that the pagecache pages are written to disk and
4707 		 * invalidated to preserve the expected O_DIRECT semantics.
4708 		 */
4709 		if (ret2 > 0) {
4710 			loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4711 
4712 			ret += ret2;
4713 
4714 			f2fs_flush_buffered_write(file->f_mapping,
4715 						  bufio_start_pos,
4716 						  bufio_end_pos);
4717 		}
4718 	} else {
4719 		/* iomap_dio_rw() already handled the generic_write_sync(). */
4720 		*may_need_sync = false;
4721 	}
4722 out:
4723 	trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4724 	return ret;
4725 }
4726 
f2fs_file_write_iter(struct kiocb * iocb,struct iov_iter * from)4727 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4728 {
4729 	struct inode *inode = file_inode(iocb->ki_filp);
4730 	const loff_t orig_pos = iocb->ki_pos;
4731 	const size_t orig_count = iov_iter_count(from);
4732 	loff_t target_size;
4733 	bool dio;
4734 	bool may_need_sync = true;
4735 	int preallocated;
4736 	ssize_t ret;
4737 
4738 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4739 		ret = -EIO;
4740 		goto out;
4741 	}
4742 
4743 	if (!f2fs_is_compress_backend_ready(inode)) {
4744 		ret = -EOPNOTSUPP;
4745 		goto out;
4746 	}
4747 
4748 	if (iocb->ki_flags & IOCB_NOWAIT) {
4749 		if (!inode_trylock(inode)) {
4750 			ret = -EAGAIN;
4751 			goto out;
4752 		}
4753 	} else {
4754 		inode_lock(inode);
4755 	}
4756 
4757 	ret = f2fs_write_checks(iocb, from);
4758 	if (ret <= 0)
4759 		goto out_unlock;
4760 
4761 	/* Determine whether we will do a direct write or a buffered write. */
4762 	dio = f2fs_should_use_dio(inode, iocb, from);
4763 
4764 	/* Possibly preallocate the blocks for the write. */
4765 	target_size = iocb->ki_pos + iov_iter_count(from);
4766 	preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4767 	if (preallocated < 0) {
4768 		ret = preallocated;
4769 	} else {
4770 		if (trace_f2fs_datawrite_start_enabled())
4771 			f2fs_trace_rw_file_path(iocb, orig_count, WRITE);
4772 
4773 		/* Do the actual write. */
4774 		ret = dio ?
4775 			f2fs_dio_write_iter(iocb, from, &may_need_sync) :
4776 			f2fs_buffered_write_iter(iocb, from);
4777 
4778 		if (trace_f2fs_datawrite_end_enabled())
4779 			trace_f2fs_datawrite_end(inode, orig_pos, ret);
4780 	}
4781 
4782 	/* Don't leave any preallocated blocks around past i_size. */
4783 	if (preallocated && i_size_read(inode) < target_size) {
4784 		f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4785 		filemap_invalidate_lock(inode->i_mapping);
4786 		if (!f2fs_truncate(inode))
4787 			file_dont_truncate(inode);
4788 		filemap_invalidate_unlock(inode->i_mapping);
4789 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4790 	} else {
4791 		file_dont_truncate(inode);
4792 	}
4793 
4794 	clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4795 out_unlock:
4796 	inode_unlock(inode);
4797 out:
4798 	trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4799 
4800 	if (ret > 0 && may_need_sync)
4801 		ret = generic_write_sync(iocb, ret);
4802 
4803 	/* If buffered IO was forced, flush and drop the data from
4804 	 * the page cache to preserve O_DIRECT semantics
4805 	 */
4806 	if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
4807 		f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
4808 					  orig_pos,
4809 					  orig_pos + ret - 1);
4810 
4811 	return ret;
4812 }
4813 
f2fs_file_fadvise(struct file * filp,loff_t offset,loff_t len,int advice)4814 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4815 		int advice)
4816 {
4817 	struct address_space *mapping;
4818 	struct backing_dev_info *bdi;
4819 	struct inode *inode = file_inode(filp);
4820 	int err;
4821 
4822 	if (advice == POSIX_FADV_SEQUENTIAL) {
4823 		if (S_ISFIFO(inode->i_mode))
4824 			return -ESPIPE;
4825 
4826 		mapping = filp->f_mapping;
4827 		if (!mapping || len < 0)
4828 			return -EINVAL;
4829 
4830 		bdi = inode_to_bdi(mapping->host);
4831 		filp->f_ra.ra_pages = bdi->ra_pages *
4832 			F2FS_I_SB(inode)->seq_file_ra_mul;
4833 		spin_lock(&filp->f_lock);
4834 		filp->f_mode &= ~FMODE_RANDOM;
4835 		spin_unlock(&filp->f_lock);
4836 		return 0;
4837 	}
4838 
4839 	err = generic_fadvise(filp, offset, len, advice);
4840 	if (!err && advice == POSIX_FADV_DONTNEED &&
4841 		test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4842 		f2fs_compressed_file(inode))
4843 		f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4844 
4845 	return err;
4846 }
4847 
4848 #ifdef CONFIG_COMPAT
4849 struct compat_f2fs_gc_range {
4850 	u32 sync;
4851 	compat_u64 start;
4852 	compat_u64 len;
4853 };
4854 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE	_IOW(F2FS_IOCTL_MAGIC, 11,\
4855 						struct compat_f2fs_gc_range)
4856 
f2fs_compat_ioc_gc_range(struct file * file,unsigned long arg)4857 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4858 {
4859 	struct compat_f2fs_gc_range __user *urange;
4860 	struct f2fs_gc_range range;
4861 	int err;
4862 
4863 	urange = compat_ptr(arg);
4864 	err = get_user(range.sync, &urange->sync);
4865 	err |= get_user(range.start, &urange->start);
4866 	err |= get_user(range.len, &urange->len);
4867 	if (err)
4868 		return -EFAULT;
4869 
4870 	return __f2fs_ioc_gc_range(file, &range);
4871 }
4872 
4873 struct compat_f2fs_move_range {
4874 	u32 dst_fd;
4875 	compat_u64 pos_in;
4876 	compat_u64 pos_out;
4877 	compat_u64 len;
4878 };
4879 #define F2FS_IOC32_MOVE_RANGE		_IOWR(F2FS_IOCTL_MAGIC, 9,	\
4880 					struct compat_f2fs_move_range)
4881 
f2fs_compat_ioc_move_range(struct file * file,unsigned long arg)4882 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4883 {
4884 	struct compat_f2fs_move_range __user *urange;
4885 	struct f2fs_move_range range;
4886 	int err;
4887 
4888 	urange = compat_ptr(arg);
4889 	err = get_user(range.dst_fd, &urange->dst_fd);
4890 	err |= get_user(range.pos_in, &urange->pos_in);
4891 	err |= get_user(range.pos_out, &urange->pos_out);
4892 	err |= get_user(range.len, &urange->len);
4893 	if (err)
4894 		return -EFAULT;
4895 
4896 	return __f2fs_ioc_move_range(file, &range);
4897 }
4898 
f2fs_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)4899 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4900 {
4901 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4902 		return -EIO;
4903 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4904 		return -ENOSPC;
4905 
4906 	switch (cmd) {
4907 	case FS_IOC32_GETVERSION:
4908 		cmd = FS_IOC_GETVERSION;
4909 		break;
4910 	case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4911 		return f2fs_compat_ioc_gc_range(file, arg);
4912 	case F2FS_IOC32_MOVE_RANGE:
4913 		return f2fs_compat_ioc_move_range(file, arg);
4914 	case F2FS_IOC_START_ATOMIC_WRITE:
4915 	case F2FS_IOC_START_ATOMIC_REPLACE:
4916 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4917 	case F2FS_IOC_START_VOLATILE_WRITE:
4918 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4919 	case F2FS_IOC_ABORT_ATOMIC_WRITE:
4920 	case F2FS_IOC_SHUTDOWN:
4921 	case FITRIM:
4922 	case FS_IOC_SET_ENCRYPTION_POLICY:
4923 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4924 	case FS_IOC_GET_ENCRYPTION_POLICY:
4925 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4926 	case FS_IOC_ADD_ENCRYPTION_KEY:
4927 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4928 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4929 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4930 	case FS_IOC_GET_ENCRYPTION_NONCE:
4931 	case F2FS_IOC_GARBAGE_COLLECT:
4932 	case F2FS_IOC_WRITE_CHECKPOINT:
4933 	case F2FS_IOC_DEFRAGMENT:
4934 	case F2FS_IOC_FLUSH_DEVICE:
4935 	case F2FS_IOC_GET_FEATURES:
4936 	case F2FS_IOC_GET_PIN_FILE:
4937 	case F2FS_IOC_SET_PIN_FILE:
4938 	case F2FS_IOC_PRECACHE_EXTENTS:
4939 	case F2FS_IOC_RESIZE_FS:
4940 	case FS_IOC_ENABLE_VERITY:
4941 	case FS_IOC_MEASURE_VERITY:
4942 	case FS_IOC_READ_VERITY_METADATA:
4943 	case FS_IOC_GETFSLABEL:
4944 	case FS_IOC_SETFSLABEL:
4945 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4946 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4947 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4948 	case F2FS_IOC_SEC_TRIM_FILE:
4949 	case F2FS_IOC_GET_COMPRESS_OPTION:
4950 	case F2FS_IOC_SET_COMPRESS_OPTION:
4951 	case F2FS_IOC_DECOMPRESS_FILE:
4952 	case F2FS_IOC_COMPRESS_FILE:
4953 		break;
4954 	default:
4955 		return -ENOIOCTLCMD;
4956 	}
4957 	return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4958 }
4959 #endif
4960 
4961 const struct file_operations f2fs_file_operations = {
4962 	.llseek		= f2fs_llseek,
4963 	.read_iter	= f2fs_file_read_iter,
4964 	.write_iter	= f2fs_file_write_iter,
4965 	.iopoll		= iocb_bio_iopoll,
4966 	.open		= f2fs_file_open,
4967 	.release	= f2fs_release_file,
4968 	.mmap		= f2fs_file_mmap,
4969 	.flush		= f2fs_file_flush,
4970 	.fsync		= f2fs_sync_file,
4971 	.fallocate	= f2fs_fallocate,
4972 	.unlocked_ioctl	= f2fs_ioctl,
4973 #ifdef CONFIG_COMPAT
4974 	.compat_ioctl	= f2fs_compat_ioctl,
4975 #endif
4976 	.splice_read	= generic_file_splice_read,
4977 	.splice_write	= iter_file_splice_write,
4978 	.fadvise	= f2fs_file_fadvise,
4979 };
4980