• 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/fadvise.h>
26 #include <linux/iomap.h>
27 
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "acl.h"
33 #include "gc.h"
34 #include "iostat.h"
35 #include <trace/events/f2fs.h>
36 #include <trace/events/android_fs.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 	f2fs_down_read(&F2FS_I(inode)->i_mmap_sem);
45 	ret = filemap_fault(vmf);
46 	f2fs_up_read(&F2FS_I(inode)->i_mmap_sem);
47 
48 	if (ret & VM_FAULT_LOCKED)
49 		f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
50 							F2FS_BLKSIZE);
51 
52 	trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
53 
54 	return ret;
55 }
56 
f2fs_vm_page_mkwrite(struct vm_fault * vmf)57 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
58 {
59 	struct page *page = vmf->page;
60 	struct inode *inode = file_inode(vmf->vma->vm_file);
61 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
62 	struct dnode_of_data dn;
63 	bool need_alloc = true;
64 	int err = 0;
65 
66 	if (unlikely(IS_IMMUTABLE(inode)))
67 		return VM_FAULT_SIGBUS;
68 
69 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
70 		return VM_FAULT_SIGBUS;
71 
72 	if (unlikely(f2fs_cp_error(sbi))) {
73 		err = -EIO;
74 		goto err;
75 	}
76 
77 	if (!f2fs_is_checkpoint_ready(sbi)) {
78 		err = -ENOSPC;
79 		goto err;
80 	}
81 
82 	err = f2fs_convert_inline_inode(inode);
83 	if (err)
84 		goto err;
85 
86 #ifdef CONFIG_F2FS_FS_COMPRESSION
87 	if (f2fs_compressed_file(inode)) {
88 		int ret = f2fs_is_compressed_cluster(inode, page->index);
89 
90 		if (ret < 0) {
91 			err = ret;
92 			goto err;
93 		} else if (ret) {
94 			need_alloc = false;
95 		}
96 	}
97 #endif
98 	/* should do out of any locked page */
99 	if (need_alloc)
100 		f2fs_balance_fs(sbi, true);
101 
102 	sb_start_pagefault(inode->i_sb);
103 
104 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
105 
106 	file_update_time(vmf->vma->vm_file);
107 	f2fs_down_read(&F2FS_I(inode)->i_mmap_sem);
108 	lock_page(page);
109 	if (unlikely(page->mapping != inode->i_mapping ||
110 			page_offset(page) > i_size_read(inode) ||
111 			!PageUptodate(page))) {
112 		unlock_page(page);
113 		err = -EFAULT;
114 		goto out_sem;
115 	}
116 
117 	if (need_alloc) {
118 		/* block allocation */
119 		f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
120 		set_new_dnode(&dn, inode, NULL, NULL, 0);
121 		err = f2fs_get_block(&dn, page->index);
122 		f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
123 	}
124 
125 #ifdef CONFIG_F2FS_FS_COMPRESSION
126 	if (!need_alloc) {
127 		set_new_dnode(&dn, inode, NULL, NULL, 0);
128 		err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
129 		f2fs_put_dnode(&dn);
130 	}
131 #endif
132 	if (err) {
133 		unlock_page(page);
134 		goto out_sem;
135 	}
136 
137 	f2fs_wait_on_page_writeback(page, DATA, false, true);
138 
139 	/* wait for GCed page writeback via META_MAPPING */
140 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
141 
142 	/*
143 	 * check to see if the page is mapped already (no holes)
144 	 */
145 	if (PageMappedToDisk(page))
146 		goto out_sem;
147 
148 	/* page is wholly or partially inside EOF */
149 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
150 						i_size_read(inode)) {
151 		loff_t offset;
152 
153 		offset = i_size_read(inode) & ~PAGE_MASK;
154 		zero_user_segment(page, offset, PAGE_SIZE);
155 	}
156 	set_page_dirty(page);
157 	if (!PageUptodate(page))
158 		SetPageUptodate(page);
159 
160 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
161 	f2fs_update_time(sbi, REQ_TIME);
162 
163 	trace_f2fs_vm_page_mkwrite(page, DATA);
164 out_sem:
165 	f2fs_up_read(&F2FS_I(inode)->i_mmap_sem);
166 
167 	sb_end_pagefault(inode->i_sb);
168 err:
169 	return block_page_mkwrite_return(err);
170 }
171 
172 static const struct vm_operations_struct f2fs_file_vm_ops = {
173 	.fault		= f2fs_filemap_fault,
174 	.map_pages	= filemap_map_pages,
175 	.page_mkwrite	= f2fs_vm_page_mkwrite,
176 #ifdef CONFIG_SPECULATIVE_PAGE_FAULT
177 	.allow_speculation = filemap_allow_speculation,
178 #endif
179 };
180 
get_parent_ino(struct inode * inode,nid_t * pino)181 static int get_parent_ino(struct inode *inode, nid_t *pino)
182 {
183 	struct dentry *dentry;
184 
185 	/*
186 	 * Make sure to get the non-deleted alias.  The alias associated with
187 	 * the open file descriptor being fsync()'ed may be deleted already.
188 	 */
189 	dentry = d_find_alias(inode);
190 	if (!dentry)
191 		return 0;
192 
193 	*pino = parent_ino(dentry);
194 	dput(dentry);
195 	return 1;
196 }
197 
need_do_checkpoint(struct inode * inode)198 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
199 {
200 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
201 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
202 
203 	if (!S_ISREG(inode->i_mode))
204 		cp_reason = CP_NON_REGULAR;
205 	else if (f2fs_compressed_file(inode))
206 		cp_reason = CP_COMPRESSED;
207 	else if (inode->i_nlink != 1)
208 		cp_reason = CP_HARDLINK;
209 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
210 		cp_reason = CP_SB_NEED_CP;
211 	else if (file_wrong_pino(inode))
212 		cp_reason = CP_WRONG_PINO;
213 	else if (!f2fs_space_for_roll_forward(sbi))
214 		cp_reason = CP_NO_SPC_ROLL;
215 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
216 		cp_reason = CP_NODE_NEED_CP;
217 	else if (test_opt(sbi, FASTBOOT))
218 		cp_reason = CP_FASTBOOT_MODE;
219 	else if (F2FS_OPTION(sbi).active_logs == 2)
220 		cp_reason = CP_SPEC_LOG_NUM;
221 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
222 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
223 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
224 							TRANS_DIR_INO))
225 		cp_reason = CP_RECOVER_DIR;
226 
227 	return cp_reason;
228 }
229 
need_inode_page_update(struct f2fs_sb_info * sbi,nid_t ino)230 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
231 {
232 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
233 	bool ret = false;
234 	/* But we need to avoid that there are some inode updates */
235 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
236 		ret = true;
237 	f2fs_put_page(i, 0);
238 	return ret;
239 }
240 
try_to_fix_pino(struct inode * inode)241 static void try_to_fix_pino(struct inode *inode)
242 {
243 	struct f2fs_inode_info *fi = F2FS_I(inode);
244 	nid_t pino;
245 
246 	f2fs_down_write(&fi->i_sem);
247 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
248 			get_parent_ino(inode, &pino)) {
249 		f2fs_i_pino_write(inode, pino);
250 		file_got_pino(inode);
251 	}
252 	f2fs_up_write(&fi->i_sem);
253 }
254 
f2fs_do_sync_file(struct file * file,loff_t start,loff_t end,int datasync,bool atomic)255 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
256 						int datasync, bool atomic)
257 {
258 	struct inode *inode = file->f_mapping->host;
259 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
260 	nid_t ino = inode->i_ino;
261 	int ret = 0;
262 	enum cp_reason_type cp_reason = 0;
263 	struct writeback_control wbc = {
264 		.sync_mode = WB_SYNC_ALL,
265 		.nr_to_write = LONG_MAX,
266 		.for_reclaim = 0,
267 	};
268 	unsigned int seq_id = 0;
269 
270 	if (unlikely(f2fs_readonly(inode->i_sb)))
271 		return 0;
272 
273 	trace_f2fs_sync_file_enter(inode);
274 
275 	if (S_ISDIR(inode->i_mode))
276 		goto go_write;
277 
278 	/* if fdatasync is triggered, let's do in-place-update */
279 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
280 		set_inode_flag(inode, FI_NEED_IPU);
281 	ret = file_write_and_wait_range(file, start, end);
282 	clear_inode_flag(inode, FI_NEED_IPU);
283 
284 	if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
285 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
286 		return ret;
287 	}
288 
289 	/* if the inode is dirty, let's recover all the time */
290 	if (!f2fs_skip_inode_update(inode, datasync)) {
291 		f2fs_write_inode(inode, NULL);
292 		goto go_write;
293 	}
294 
295 	/*
296 	 * if there is no written data, don't waste time to write recovery info.
297 	 */
298 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
299 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
300 
301 		/* it may call write_inode just prior to fsync */
302 		if (need_inode_page_update(sbi, ino))
303 			goto go_write;
304 
305 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
306 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
307 			goto flush_out;
308 		goto out;
309 	} else {
310 		/*
311 		 * for OPU case, during fsync(), node can be persisted before
312 		 * data when lower device doesn't support write barrier, result
313 		 * in data corruption after SPO.
314 		 * So for strict fsync mode, force to use atomic write sematics
315 		 * to keep write order in between data/node and last node to
316 		 * avoid potential data corruption.
317 		 */
318 		if (F2FS_OPTION(sbi).fsync_mode ==
319 				FSYNC_MODE_STRICT && !atomic)
320 			atomic = true;
321 	}
322 go_write:
323 	/*
324 	 * Both of fdatasync() and fsync() are able to be recovered from
325 	 * sudden-power-off.
326 	 */
327 	f2fs_down_read(&F2FS_I(inode)->i_sem);
328 	cp_reason = need_do_checkpoint(inode);
329 	f2fs_up_read(&F2FS_I(inode)->i_sem);
330 
331 	if (cp_reason) {
332 		/* all the dirty node pages should be flushed for POR */
333 		ret = f2fs_sync_fs(inode->i_sb, 1);
334 
335 		/*
336 		 * We've secured consistency through sync_fs. Following pino
337 		 * will be used only for fsynced inodes after checkpoint.
338 		 */
339 		try_to_fix_pino(inode);
340 		clear_inode_flag(inode, FI_APPEND_WRITE);
341 		clear_inode_flag(inode, FI_UPDATE_WRITE);
342 		goto out;
343 	}
344 sync_nodes:
345 	atomic_inc(&sbi->wb_sync_req[NODE]);
346 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
347 	atomic_dec(&sbi->wb_sync_req[NODE]);
348 	if (ret)
349 		goto out;
350 
351 	/* if cp_error was enabled, we should avoid infinite loop */
352 	if (unlikely(f2fs_cp_error(sbi))) {
353 		ret = -EIO;
354 		goto out;
355 	}
356 
357 	if (f2fs_need_inode_block_update(sbi, ino)) {
358 		f2fs_mark_inode_dirty_sync(inode, true);
359 		f2fs_write_inode(inode, NULL);
360 		goto sync_nodes;
361 	}
362 
363 	/*
364 	 * If it's atomic_write, it's just fine to keep write ordering. So
365 	 * here we don't need to wait for node write completion, since we use
366 	 * node chain which serializes node blocks. If one of node writes are
367 	 * reordered, we can see simply broken chain, resulting in stopping
368 	 * roll-forward recovery. It means we'll recover all or none node blocks
369 	 * given fsync mark.
370 	 */
371 	if (!atomic) {
372 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
373 		if (ret)
374 			goto out;
375 	}
376 
377 	/* once recovery info is written, don't need to tack this */
378 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
379 	clear_inode_flag(inode, FI_APPEND_WRITE);
380 flush_out:
381 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
382 		ret = f2fs_issue_flush(sbi, inode->i_ino);
383 	if (!ret) {
384 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
385 		clear_inode_flag(inode, FI_UPDATE_WRITE);
386 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
387 	}
388 	f2fs_update_time(sbi, REQ_TIME);
389 out:
390 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
391 	return ret;
392 }
393 
f2fs_sync_file(struct file * file,loff_t start,loff_t end,int datasync)394 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
395 {
396 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
397 		return -EIO;
398 	return f2fs_do_sync_file(file, start, end, datasync, false);
399 }
400 
__found_offset(struct address_space * mapping,block_t blkaddr,pgoff_t index,int whence)401 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
402 				pgoff_t index, int whence)
403 {
404 	switch (whence) {
405 	case SEEK_DATA:
406 		if (__is_valid_data_blkaddr(blkaddr))
407 			return true;
408 		if (blkaddr == NEW_ADDR &&
409 		    xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
410 			return true;
411 		break;
412 	case SEEK_HOLE:
413 		if (blkaddr == NULL_ADDR)
414 			return true;
415 		break;
416 	}
417 	return false;
418 }
419 
f2fs_seek_block(struct file * file,loff_t offset,int whence)420 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
421 {
422 	struct inode *inode = file->f_mapping->host;
423 	loff_t maxbytes = inode->i_sb->s_maxbytes;
424 	struct dnode_of_data dn;
425 	pgoff_t pgofs, end_offset;
426 	loff_t data_ofs = offset;
427 	loff_t isize;
428 	int err = 0;
429 
430 	inode_lock(inode);
431 
432 	isize = i_size_read(inode);
433 	if (offset >= isize)
434 		goto fail;
435 
436 	/* handle inline data case */
437 	if (f2fs_has_inline_data(inode)) {
438 		if (whence == SEEK_HOLE) {
439 			data_ofs = isize;
440 			goto found;
441 		} else if (whence == SEEK_DATA) {
442 			data_ofs = offset;
443 			goto found;
444 		}
445 	}
446 
447 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
448 
449 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
450 		set_new_dnode(&dn, inode, NULL, NULL, 0);
451 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
452 		if (err && err != -ENOENT) {
453 			goto fail;
454 		} else if (err == -ENOENT) {
455 			/* direct node does not exists */
456 			if (whence == SEEK_DATA) {
457 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
458 				continue;
459 			} else {
460 				goto found;
461 			}
462 		}
463 
464 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
465 
466 		/* find data/hole in dnode block */
467 		for (; dn.ofs_in_node < end_offset;
468 				dn.ofs_in_node++, pgofs++,
469 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
470 			block_t blkaddr;
471 
472 			blkaddr = f2fs_data_blkaddr(&dn);
473 
474 			if (__is_valid_data_blkaddr(blkaddr) &&
475 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
476 					blkaddr, DATA_GENERIC_ENHANCE)) {
477 				f2fs_put_dnode(&dn);
478 				goto fail;
479 			}
480 
481 			if (__found_offset(file->f_mapping, blkaddr,
482 							pgofs, whence)) {
483 				f2fs_put_dnode(&dn);
484 				goto found;
485 			}
486 		}
487 		f2fs_put_dnode(&dn);
488 	}
489 
490 	if (whence == SEEK_DATA)
491 		goto fail;
492 found:
493 	if (whence == SEEK_HOLE && data_ofs > isize)
494 		data_ofs = isize;
495 	inode_unlock(inode);
496 	return vfs_setpos(file, data_ofs, maxbytes);
497 fail:
498 	inode_unlock(inode);
499 	return -ENXIO;
500 }
501 
f2fs_llseek(struct file * file,loff_t offset,int whence)502 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
503 {
504 	struct inode *inode = file->f_mapping->host;
505 	loff_t maxbytes = inode->i_sb->s_maxbytes;
506 
507 	if (f2fs_compressed_file(inode))
508 		maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
509 
510 	switch (whence) {
511 	case SEEK_SET:
512 	case SEEK_CUR:
513 	case SEEK_END:
514 		return generic_file_llseek_size(file, offset, whence,
515 						maxbytes, i_size_read(inode));
516 	case SEEK_DATA:
517 	case SEEK_HOLE:
518 		if (offset < 0)
519 			return -ENXIO;
520 		return f2fs_seek_block(file, offset, whence);
521 	}
522 
523 	return -EINVAL;
524 }
525 
f2fs_file_mmap(struct file * file,struct vm_area_struct * vma)526 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
527 {
528 	struct inode *inode = file_inode(file);
529 
530 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
531 		return -EIO;
532 
533 	if (!f2fs_is_compress_backend_ready(inode))
534 		return -EOPNOTSUPP;
535 
536 	file_accessed(file);
537 	vma->vm_ops = &f2fs_file_vm_ops;
538 	set_inode_flag(inode, FI_MMAP_FILE);
539 	return 0;
540 }
541 
f2fs_file_open(struct inode * inode,struct file * filp)542 static int f2fs_file_open(struct inode *inode, struct file *filp)
543 {
544 	int err = fscrypt_file_open(inode, filp);
545 
546 	if (err)
547 		return err;
548 
549 	if (!f2fs_is_compress_backend_ready(inode))
550 		return -EOPNOTSUPP;
551 
552 	err = fsverity_file_open(inode, filp);
553 	if (err)
554 		return err;
555 
556 	filp->f_mode |= FMODE_NOWAIT;
557 
558 	return dquot_file_open(inode, filp);
559 }
560 
f2fs_truncate_data_blocks_range(struct dnode_of_data * dn,int count)561 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
562 {
563 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
564 	struct f2fs_node *raw_node;
565 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
566 	__le32 *addr;
567 	int base = 0;
568 	bool compressed_cluster = false;
569 	int cluster_index = 0, valid_blocks = 0;
570 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
571 	bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
572 
573 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
574 		base = get_extra_isize(dn->inode);
575 
576 	raw_node = F2FS_NODE(dn->node_page);
577 	addr = blkaddr_in_node(raw_node) + base + ofs;
578 
579 	/* Assumption: truncateion starts with cluster */
580 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
581 		block_t blkaddr = le32_to_cpu(*addr);
582 
583 		if (f2fs_compressed_file(dn->inode) &&
584 					!(cluster_index & (cluster_size - 1))) {
585 			if (compressed_cluster)
586 				f2fs_i_compr_blocks_update(dn->inode,
587 							valid_blocks, false);
588 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
589 			valid_blocks = 0;
590 		}
591 
592 		if (blkaddr == NULL_ADDR)
593 			continue;
594 
595 		dn->data_blkaddr = NULL_ADDR;
596 		f2fs_set_data_blkaddr(dn);
597 
598 		if (__is_valid_data_blkaddr(blkaddr)) {
599 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
600 					DATA_GENERIC_ENHANCE))
601 				continue;
602 			if (compressed_cluster)
603 				valid_blocks++;
604 		}
605 
606 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
607 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
608 
609 		f2fs_invalidate_blocks(sbi, blkaddr);
610 
611 		if (!released || blkaddr != COMPRESS_ADDR)
612 			nr_free++;
613 	}
614 
615 	if (compressed_cluster)
616 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
617 
618 	if (nr_free) {
619 		pgoff_t fofs;
620 		/*
621 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
622 		 * we will invalidate all blkaddr in the whole range.
623 		 */
624 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
625 							dn->inode) + ofs;
626 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
627 		dec_valid_block_count(sbi, dn->inode, nr_free);
628 	}
629 	dn->ofs_in_node = ofs;
630 
631 	f2fs_update_time(sbi, REQ_TIME);
632 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
633 					 dn->ofs_in_node, nr_free);
634 }
635 
f2fs_truncate_data_blocks(struct dnode_of_data * dn)636 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
637 {
638 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
639 }
640 
truncate_partial_data_page(struct inode * inode,u64 from,bool cache_only)641 static int truncate_partial_data_page(struct inode *inode, u64 from,
642 								bool cache_only)
643 {
644 	loff_t offset = from & (PAGE_SIZE - 1);
645 	pgoff_t index = from >> PAGE_SHIFT;
646 	struct address_space *mapping = inode->i_mapping;
647 	struct page *page;
648 
649 	if (!offset && !cache_only)
650 		return 0;
651 
652 	if (cache_only) {
653 		page = find_lock_page(mapping, index);
654 		if (page && PageUptodate(page))
655 			goto truncate_out;
656 		f2fs_put_page(page, 1);
657 		return 0;
658 	}
659 
660 	page = f2fs_get_lock_data_page(inode, index, true);
661 	if (IS_ERR(page))
662 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
663 truncate_out:
664 	f2fs_wait_on_page_writeback(page, DATA, true, true);
665 	zero_user(page, offset, PAGE_SIZE - offset);
666 
667 	/* An encrypted inode should have a key and truncate the last page. */
668 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
669 	if (!cache_only)
670 		set_page_dirty(page);
671 	f2fs_put_page(page, 1);
672 	return 0;
673 }
674 
f2fs_do_truncate_blocks(struct inode * inode,u64 from,bool lock)675 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
676 {
677 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
678 	struct dnode_of_data dn;
679 	pgoff_t free_from;
680 	int count = 0, err = 0;
681 	struct page *ipage;
682 	bool truncate_page = false;
683 
684 	trace_f2fs_truncate_blocks_enter(inode, from);
685 
686 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
687 
688 	if (free_from >= max_file_blocks(inode))
689 		goto free_partial;
690 
691 	if (lock)
692 		f2fs_lock_op(sbi);
693 
694 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
695 	if (IS_ERR(ipage)) {
696 		err = PTR_ERR(ipage);
697 		goto out;
698 	}
699 
700 	if (f2fs_has_inline_data(inode)) {
701 		f2fs_truncate_inline_inode(inode, ipage, from);
702 		f2fs_put_page(ipage, 1);
703 		truncate_page = true;
704 		goto out;
705 	}
706 
707 	set_new_dnode(&dn, inode, ipage, NULL, 0);
708 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
709 	if (err) {
710 		if (err == -ENOENT)
711 			goto free_next;
712 		goto out;
713 	}
714 
715 	count = ADDRS_PER_PAGE(dn.node_page, inode);
716 
717 	count -= dn.ofs_in_node;
718 	f2fs_bug_on(sbi, count < 0);
719 
720 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
721 		f2fs_truncate_data_blocks_range(&dn, count);
722 		free_from += count;
723 	}
724 
725 	f2fs_put_dnode(&dn);
726 free_next:
727 	err = f2fs_truncate_inode_blocks(inode, free_from);
728 out:
729 	if (lock)
730 		f2fs_unlock_op(sbi);
731 free_partial:
732 	/* lastly zero out the first data page */
733 	if (!err)
734 		err = truncate_partial_data_page(inode, from, truncate_page);
735 
736 	trace_f2fs_truncate_blocks_exit(inode, err);
737 	return err;
738 }
739 
f2fs_truncate_blocks(struct inode * inode,u64 from,bool lock)740 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
741 {
742 	u64 free_from = from;
743 	int err;
744 
745 #ifdef CONFIG_F2FS_FS_COMPRESSION
746 	/*
747 	 * for compressed file, only support cluster size
748 	 * aligned truncation.
749 	 */
750 	if (f2fs_compressed_file(inode))
751 		free_from = round_up(from,
752 				F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
753 #endif
754 
755 	err = f2fs_do_truncate_blocks(inode, free_from, lock);
756 	if (err)
757 		return err;
758 
759 #ifdef CONFIG_F2FS_FS_COMPRESSION
760 	/*
761 	 * For compressed file, after release compress blocks, don't allow write
762 	 * direct, but we should allow write direct after truncate to zero.
763 	 */
764 	if (f2fs_compressed_file(inode) && !free_from
765 			&& is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
766 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
767 
768 	if (from != free_from) {
769 		err = f2fs_truncate_partial_cluster(inode, from, lock);
770 		if (err)
771 			return err;
772 	}
773 #endif
774 
775 	return 0;
776 }
777 
f2fs_truncate(struct inode * inode)778 int f2fs_truncate(struct inode *inode)
779 {
780 	int err;
781 
782 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
783 		return -EIO;
784 
785 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
786 				S_ISLNK(inode->i_mode)))
787 		return 0;
788 
789 	trace_f2fs_truncate(inode);
790 
791 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
792 		f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
793 		return -EIO;
794 	}
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_getattr(const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)816 int f2fs_getattr(const struct path *path, struct kstat *stat,
817 		 u32 request_mask, unsigned int query_flags)
818 {
819 	struct inode *inode = d_inode(path->dentry);
820 	struct f2fs_inode_info *fi = F2FS_I(inode);
821 	struct f2fs_inode *ri = NULL;
822 	unsigned int flags;
823 
824 	if (f2fs_has_extra_attr(inode) &&
825 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
826 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
827 		stat->result_mask |= STATX_BTIME;
828 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
829 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
830 	}
831 
832 	flags = fi->i_flags;
833 	if (flags & F2FS_COMPR_FL)
834 		stat->attributes |= STATX_ATTR_COMPRESSED;
835 	if (flags & F2FS_APPEND_FL)
836 		stat->attributes |= STATX_ATTR_APPEND;
837 	if (IS_ENCRYPTED(inode))
838 		stat->attributes |= STATX_ATTR_ENCRYPTED;
839 	if (flags & F2FS_IMMUTABLE_FL)
840 		stat->attributes |= STATX_ATTR_IMMUTABLE;
841 	if (flags & F2FS_NODUMP_FL)
842 		stat->attributes |= STATX_ATTR_NODUMP;
843 	if (IS_VERITY(inode))
844 		stat->attributes |= STATX_ATTR_VERITY;
845 
846 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
847 				  STATX_ATTR_APPEND |
848 				  STATX_ATTR_ENCRYPTED |
849 				  STATX_ATTR_IMMUTABLE |
850 				  STATX_ATTR_NODUMP |
851 				  STATX_ATTR_VERITY);
852 
853 	generic_fillattr(inode, stat);
854 
855 	/* we need to show initial sectors used for inline_data/dentries */
856 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
857 					f2fs_has_inline_dentry(inode))
858 		stat->blocks += (stat->size + 511) >> 9;
859 
860 	return 0;
861 }
862 
863 #ifdef CONFIG_F2FS_FS_POSIX_ACL
__setattr_copy(struct inode * inode,const struct iattr * attr)864 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
865 {
866 	unsigned int ia_valid = attr->ia_valid;
867 
868 	if (ia_valid & ATTR_UID)
869 		inode->i_uid = attr->ia_uid;
870 	if (ia_valid & ATTR_GID)
871 		inode->i_gid = attr->ia_gid;
872 	if (ia_valid & ATTR_ATIME)
873 		inode->i_atime = attr->ia_atime;
874 	if (ia_valid & ATTR_MTIME)
875 		inode->i_mtime = attr->ia_mtime;
876 	if (ia_valid & ATTR_CTIME)
877 		inode->i_ctime = attr->ia_ctime;
878 	if (ia_valid & ATTR_MODE) {
879 		umode_t mode = attr->ia_mode;
880 
881 		if (!in_group_p(inode->i_gid) &&
882 			!capable_wrt_inode_uidgid(inode, CAP_FSETID))
883 			mode &= ~S_ISGID;
884 		set_acl_inode(inode, mode);
885 	}
886 }
887 #else
888 #define __setattr_copy setattr_copy
889 #endif
890 
f2fs_setattr(struct dentry * dentry,struct iattr * attr)891 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
892 {
893 	struct inode *inode = d_inode(dentry);
894 	int err;
895 
896 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
897 		return -EIO;
898 
899 	if (unlikely(IS_IMMUTABLE(inode)))
900 		return -EPERM;
901 
902 	if (unlikely(IS_APPEND(inode) &&
903 			(attr->ia_valid & (ATTR_MODE | ATTR_UID |
904 				  ATTR_GID | ATTR_TIMES_SET))))
905 		return -EPERM;
906 
907 	if ((attr->ia_valid & ATTR_SIZE) &&
908 		!f2fs_is_compress_backend_ready(inode))
909 		return -EOPNOTSUPP;
910 
911 	err = setattr_prepare(dentry, attr);
912 	if (err)
913 		return err;
914 
915 	err = fscrypt_prepare_setattr(dentry, attr);
916 	if (err)
917 		return err;
918 
919 	err = fsverity_prepare_setattr(dentry, attr);
920 	if (err)
921 		return err;
922 
923 	if (is_quota_modification(inode, attr)) {
924 		err = f2fs_dquot_initialize(inode);
925 		if (err)
926 			return err;
927 	}
928 	if ((attr->ia_valid & ATTR_UID &&
929 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
930 		(attr->ia_valid & ATTR_GID &&
931 		!gid_eq(attr->ia_gid, inode->i_gid))) {
932 		f2fs_lock_op(F2FS_I_SB(inode));
933 		err = dquot_transfer(inode, attr);
934 		if (err) {
935 			set_sbi_flag(F2FS_I_SB(inode),
936 					SBI_QUOTA_NEED_REPAIR);
937 			f2fs_unlock_op(F2FS_I_SB(inode));
938 			return err;
939 		}
940 		/*
941 		 * update uid/gid under lock_op(), so that dquot and inode can
942 		 * be updated atomically.
943 		 */
944 		if (attr->ia_valid & ATTR_UID)
945 			inode->i_uid = attr->ia_uid;
946 		if (attr->ia_valid & ATTR_GID)
947 			inode->i_gid = attr->ia_gid;
948 		f2fs_mark_inode_dirty_sync(inode, true);
949 		f2fs_unlock_op(F2FS_I_SB(inode));
950 	}
951 
952 	if (attr->ia_valid & ATTR_SIZE) {
953 		loff_t old_size = i_size_read(inode);
954 
955 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
956 			/*
957 			 * should convert inline inode before i_size_write to
958 			 * keep smaller than inline_data size with inline flag.
959 			 */
960 			err = f2fs_convert_inline_inode(inode);
961 			if (err)
962 				return err;
963 		}
964 
965 		f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
966 		f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
967 
968 		truncate_setsize(inode, attr->ia_size);
969 
970 		if (attr->ia_size <= old_size)
971 			err = f2fs_truncate(inode);
972 		/*
973 		 * do not trim all blocks after i_size if target size is
974 		 * larger than i_size.
975 		 */
976 		f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
977 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
978 		if (err)
979 			return err;
980 
981 		spin_lock(&F2FS_I(inode)->i_size_lock);
982 		inode->i_mtime = inode->i_ctime = current_time(inode);
983 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
984 		spin_unlock(&F2FS_I(inode)->i_size_lock);
985 	}
986 
987 	__setattr_copy(inode, attr);
988 
989 	if (attr->ia_valid & ATTR_MODE) {
990 		err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
991 
992 		if (is_inode_flag_set(inode, FI_ACL_MODE)) {
993 			if (!err)
994 				inode->i_mode = F2FS_I(inode)->i_acl_mode;
995 			clear_inode_flag(inode, FI_ACL_MODE);
996 		}
997 	}
998 
999 	/* file size may changed here */
1000 	f2fs_mark_inode_dirty_sync(inode, true);
1001 
1002 	/* inode change will produce dirty node pages flushed by checkpoint */
1003 	f2fs_balance_fs(F2FS_I_SB(inode), true);
1004 
1005 	return err;
1006 }
1007 
1008 const struct inode_operations f2fs_file_inode_operations = {
1009 	.getattr	= f2fs_getattr,
1010 	.setattr	= f2fs_setattr,
1011 	.get_acl	= f2fs_get_acl,
1012 	.set_acl	= f2fs_set_acl,
1013 	.listxattr	= f2fs_listxattr,
1014 	.fiemap		= f2fs_fiemap,
1015 };
1016 
fill_zero(struct inode * inode,pgoff_t index,loff_t start,loff_t len)1017 static int fill_zero(struct inode *inode, pgoff_t index,
1018 					loff_t start, loff_t len)
1019 {
1020 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1021 	struct page *page;
1022 
1023 	if (!len)
1024 		return 0;
1025 
1026 	f2fs_balance_fs(sbi, true);
1027 
1028 	f2fs_lock_op(sbi);
1029 	page = f2fs_get_new_data_page(inode, NULL, index, false);
1030 	f2fs_unlock_op(sbi);
1031 
1032 	if (IS_ERR(page))
1033 		return PTR_ERR(page);
1034 
1035 	f2fs_wait_on_page_writeback(page, DATA, true, true);
1036 	zero_user(page, start, len);
1037 	set_page_dirty(page);
1038 	f2fs_put_page(page, 1);
1039 	return 0;
1040 }
1041 
f2fs_truncate_hole(struct inode * inode,pgoff_t pg_start,pgoff_t pg_end)1042 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1043 {
1044 	int err;
1045 
1046 	while (pg_start < pg_end) {
1047 		struct dnode_of_data dn;
1048 		pgoff_t end_offset, count;
1049 
1050 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1051 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1052 		if (err) {
1053 			if (err == -ENOENT) {
1054 				pg_start = f2fs_get_next_page_offset(&dn,
1055 								pg_start);
1056 				continue;
1057 			}
1058 			return err;
1059 		}
1060 
1061 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1062 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1063 
1064 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1065 
1066 		f2fs_truncate_data_blocks_range(&dn, count);
1067 		f2fs_put_dnode(&dn);
1068 
1069 		pg_start += count;
1070 	}
1071 	return 0;
1072 }
1073 
punch_hole(struct inode * inode,loff_t offset,loff_t len)1074 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1075 {
1076 	pgoff_t pg_start, pg_end;
1077 	loff_t off_start, off_end;
1078 	int ret;
1079 
1080 	ret = f2fs_convert_inline_inode(inode);
1081 	if (ret)
1082 		return ret;
1083 
1084 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1085 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1086 
1087 	off_start = offset & (PAGE_SIZE - 1);
1088 	off_end = (offset + len) & (PAGE_SIZE - 1);
1089 
1090 	if (pg_start == pg_end) {
1091 		ret = fill_zero(inode, pg_start, off_start,
1092 						off_end - off_start);
1093 		if (ret)
1094 			return ret;
1095 	} else {
1096 		if (off_start) {
1097 			ret = fill_zero(inode, pg_start++, off_start,
1098 						PAGE_SIZE - off_start);
1099 			if (ret)
1100 				return ret;
1101 		}
1102 		if (off_end) {
1103 			ret = fill_zero(inode, pg_end, 0, off_end);
1104 			if (ret)
1105 				return ret;
1106 		}
1107 
1108 		if (pg_start < pg_end) {
1109 			loff_t blk_start, blk_end;
1110 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1111 
1112 			f2fs_balance_fs(sbi, true);
1113 
1114 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1115 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1116 
1117 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1118 			f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
1119 
1120 			truncate_pagecache_range(inode, blk_start, blk_end - 1);
1121 
1122 			f2fs_lock_op(sbi);
1123 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1124 			f2fs_unlock_op(sbi);
1125 
1126 			f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
1127 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1128 		}
1129 	}
1130 
1131 	return ret;
1132 }
1133 
__read_out_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,pgoff_t len)1134 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1135 				int *do_replace, pgoff_t off, pgoff_t len)
1136 {
1137 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1138 	struct dnode_of_data dn;
1139 	int ret, done, i;
1140 
1141 next_dnode:
1142 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1143 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1144 	if (ret && ret != -ENOENT) {
1145 		return ret;
1146 	} else if (ret == -ENOENT) {
1147 		if (dn.max_level == 0)
1148 			return -ENOENT;
1149 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1150 						dn.ofs_in_node, len);
1151 		blkaddr += done;
1152 		do_replace += done;
1153 		goto next;
1154 	}
1155 
1156 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1157 							dn.ofs_in_node, len);
1158 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1159 		*blkaddr = f2fs_data_blkaddr(&dn);
1160 
1161 		if (__is_valid_data_blkaddr(*blkaddr) &&
1162 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1163 					DATA_GENERIC_ENHANCE)) {
1164 			f2fs_put_dnode(&dn);
1165 			return -EFSCORRUPTED;
1166 		}
1167 
1168 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1169 
1170 			if (f2fs_lfs_mode(sbi)) {
1171 				f2fs_put_dnode(&dn);
1172 				return -EOPNOTSUPP;
1173 			}
1174 
1175 			/* do not invalidate this block address */
1176 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1177 			*do_replace = 1;
1178 		}
1179 	}
1180 	f2fs_put_dnode(&dn);
1181 next:
1182 	len -= done;
1183 	off += done;
1184 	if (len)
1185 		goto next_dnode;
1186 	return 0;
1187 }
1188 
__roll_back_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,int len)1189 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1190 				int *do_replace, pgoff_t off, int len)
1191 {
1192 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1193 	struct dnode_of_data dn;
1194 	int ret, i;
1195 
1196 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1197 		if (*do_replace == 0)
1198 			continue;
1199 
1200 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1201 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1202 		if (ret) {
1203 			dec_valid_block_count(sbi, inode, 1);
1204 			f2fs_invalidate_blocks(sbi, *blkaddr);
1205 		} else {
1206 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1207 		}
1208 		f2fs_put_dnode(&dn);
1209 	}
1210 	return 0;
1211 }
1212 
__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)1213 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1214 			block_t *blkaddr, int *do_replace,
1215 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1216 {
1217 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1218 	pgoff_t i = 0;
1219 	int ret;
1220 
1221 	while (i < len) {
1222 		if (blkaddr[i] == NULL_ADDR && !full) {
1223 			i++;
1224 			continue;
1225 		}
1226 
1227 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1228 			struct dnode_of_data dn;
1229 			struct node_info ni;
1230 			size_t new_size;
1231 			pgoff_t ilen;
1232 
1233 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1234 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1235 			if (ret)
1236 				return ret;
1237 
1238 			ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1239 			if (ret) {
1240 				f2fs_put_dnode(&dn);
1241 				return ret;
1242 			}
1243 
1244 			ilen = min((pgoff_t)
1245 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1246 						dn.ofs_in_node, len - i);
1247 			do {
1248 				dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1249 				f2fs_truncate_data_blocks_range(&dn, 1);
1250 
1251 				if (do_replace[i]) {
1252 					f2fs_i_blocks_write(src_inode,
1253 							1, false, false);
1254 					f2fs_i_blocks_write(dst_inode,
1255 							1, true, false);
1256 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1257 					blkaddr[i], ni.version, true, false);
1258 
1259 					do_replace[i] = 0;
1260 				}
1261 				dn.ofs_in_node++;
1262 				i++;
1263 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1264 				if (dst_inode->i_size < new_size)
1265 					f2fs_i_size_write(dst_inode, new_size);
1266 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1267 
1268 			f2fs_put_dnode(&dn);
1269 		} else {
1270 			struct page *psrc, *pdst;
1271 
1272 			psrc = f2fs_get_lock_data_page(src_inode,
1273 							src + i, true);
1274 			if (IS_ERR(psrc))
1275 				return PTR_ERR(psrc);
1276 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1277 								true);
1278 			if (IS_ERR(pdst)) {
1279 				f2fs_put_page(psrc, 1);
1280 				return PTR_ERR(pdst);
1281 			}
1282 			f2fs_copy_page(psrc, pdst);
1283 			set_page_dirty(pdst);
1284 			f2fs_put_page(pdst, 1);
1285 			f2fs_put_page(psrc, 1);
1286 
1287 			ret = f2fs_truncate_hole(src_inode,
1288 						src + i, src + i + 1);
1289 			if (ret)
1290 				return ret;
1291 			i++;
1292 		}
1293 	}
1294 	return 0;
1295 }
1296 
__exchange_data_block(struct inode * src_inode,struct inode * dst_inode,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1297 static int __exchange_data_block(struct inode *src_inode,
1298 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1299 			pgoff_t len, bool full)
1300 {
1301 	block_t *src_blkaddr;
1302 	int *do_replace;
1303 	pgoff_t olen;
1304 	int ret;
1305 
1306 	while (len) {
1307 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1308 
1309 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1310 					array_size(olen, sizeof(block_t)),
1311 					GFP_NOFS);
1312 		if (!src_blkaddr)
1313 			return -ENOMEM;
1314 
1315 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1316 					array_size(olen, sizeof(int)),
1317 					GFP_NOFS);
1318 		if (!do_replace) {
1319 			kvfree(src_blkaddr);
1320 			return -ENOMEM;
1321 		}
1322 
1323 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1324 					do_replace, src, olen);
1325 		if (ret)
1326 			goto roll_back;
1327 
1328 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1329 					do_replace, src, dst, olen, full);
1330 		if (ret)
1331 			goto roll_back;
1332 
1333 		src += olen;
1334 		dst += olen;
1335 		len -= olen;
1336 
1337 		kvfree(src_blkaddr);
1338 		kvfree(do_replace);
1339 	}
1340 	return 0;
1341 
1342 roll_back:
1343 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1344 	kvfree(src_blkaddr);
1345 	kvfree(do_replace);
1346 	return ret;
1347 }
1348 
f2fs_do_collapse(struct inode * inode,loff_t offset,loff_t len)1349 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1350 {
1351 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1352 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1353 	pgoff_t start = offset >> PAGE_SHIFT;
1354 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1355 	int ret;
1356 
1357 	f2fs_balance_fs(sbi, true);
1358 
1359 	/* avoid gc operation during block exchange */
1360 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1361 	f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
1362 
1363 	f2fs_lock_op(sbi);
1364 	f2fs_drop_extent_tree(inode);
1365 	truncate_pagecache(inode, offset);
1366 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1367 	f2fs_unlock_op(sbi);
1368 
1369 	f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
1370 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1371 	return ret;
1372 }
1373 
f2fs_collapse_range(struct inode * inode,loff_t offset,loff_t len)1374 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1375 {
1376 	loff_t new_size;
1377 	int ret;
1378 
1379 	if (offset + len >= i_size_read(inode))
1380 		return -EINVAL;
1381 
1382 	/* collapse range should be aligned to block size of f2fs. */
1383 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1384 		return -EINVAL;
1385 
1386 	ret = f2fs_convert_inline_inode(inode);
1387 	if (ret)
1388 		return ret;
1389 
1390 	/* write out all dirty pages from offset */
1391 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1392 	if (ret)
1393 		return ret;
1394 
1395 	ret = f2fs_do_collapse(inode, offset, len);
1396 	if (ret)
1397 		return ret;
1398 
1399 	/* write out all moved pages, if possible */
1400 	f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
1401 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1402 	truncate_pagecache(inode, offset);
1403 
1404 	new_size = i_size_read(inode) - len;
1405 	ret = f2fs_truncate_blocks(inode, new_size, true);
1406 	f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
1407 	if (!ret)
1408 		f2fs_i_size_write(inode, new_size);
1409 	return ret;
1410 }
1411 
f2fs_do_zero_range(struct dnode_of_data * dn,pgoff_t start,pgoff_t end)1412 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1413 								pgoff_t end)
1414 {
1415 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1416 	pgoff_t index = start;
1417 	unsigned int ofs_in_node = dn->ofs_in_node;
1418 	blkcnt_t count = 0;
1419 	int ret;
1420 
1421 	for (; index < end; index++, dn->ofs_in_node++) {
1422 		if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1423 			count++;
1424 	}
1425 
1426 	dn->ofs_in_node = ofs_in_node;
1427 	ret = f2fs_reserve_new_blocks(dn, count);
1428 	if (ret)
1429 		return ret;
1430 
1431 	dn->ofs_in_node = ofs_in_node;
1432 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1433 		dn->data_blkaddr = f2fs_data_blkaddr(dn);
1434 		/*
1435 		 * f2fs_reserve_new_blocks will not guarantee entire block
1436 		 * allocation.
1437 		 */
1438 		if (dn->data_blkaddr == NULL_ADDR) {
1439 			ret = -ENOSPC;
1440 			break;
1441 		}
1442 
1443 		if (dn->data_blkaddr == NEW_ADDR)
1444 			continue;
1445 
1446 		if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1447 					DATA_GENERIC_ENHANCE)) {
1448 			ret = -EFSCORRUPTED;
1449 			break;
1450 		}
1451 
1452 		f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1453 		dn->data_blkaddr = NEW_ADDR;
1454 		f2fs_set_data_blkaddr(dn);
1455 	}
1456 
1457 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1458 
1459 	return ret;
1460 }
1461 
f2fs_zero_range(struct inode * inode,loff_t offset,loff_t len,int mode)1462 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1463 								int mode)
1464 {
1465 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1466 	struct address_space *mapping = inode->i_mapping;
1467 	pgoff_t index, pg_start, pg_end;
1468 	loff_t new_size = i_size_read(inode);
1469 	loff_t off_start, off_end;
1470 	int ret = 0;
1471 
1472 	ret = inode_newsize_ok(inode, (len + offset));
1473 	if (ret)
1474 		return ret;
1475 
1476 	ret = f2fs_convert_inline_inode(inode);
1477 	if (ret)
1478 		return ret;
1479 
1480 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1481 	if (ret)
1482 		return ret;
1483 
1484 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1485 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1486 
1487 	off_start = offset & (PAGE_SIZE - 1);
1488 	off_end = (offset + len) & (PAGE_SIZE - 1);
1489 
1490 	if (pg_start == pg_end) {
1491 		ret = fill_zero(inode, pg_start, off_start,
1492 						off_end - off_start);
1493 		if (ret)
1494 			return ret;
1495 
1496 		new_size = max_t(loff_t, new_size, offset + len);
1497 	} else {
1498 		if (off_start) {
1499 			ret = fill_zero(inode, pg_start++, off_start,
1500 						PAGE_SIZE - off_start);
1501 			if (ret)
1502 				return ret;
1503 
1504 			new_size = max_t(loff_t, new_size,
1505 					(loff_t)pg_start << PAGE_SHIFT);
1506 		}
1507 
1508 		for (index = pg_start; index < pg_end;) {
1509 			struct dnode_of_data dn;
1510 			unsigned int end_offset;
1511 			pgoff_t end;
1512 
1513 			f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1514 			f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
1515 
1516 			truncate_pagecache_range(inode,
1517 				(loff_t)index << PAGE_SHIFT,
1518 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1519 
1520 			f2fs_lock_op(sbi);
1521 
1522 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1523 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1524 			if (ret) {
1525 				f2fs_unlock_op(sbi);
1526 				f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
1527 				f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1528 				goto out;
1529 			}
1530 
1531 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1532 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1533 
1534 			ret = f2fs_do_zero_range(&dn, index, end);
1535 			f2fs_put_dnode(&dn);
1536 
1537 			f2fs_unlock_op(sbi);
1538 			f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
1539 			f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1540 
1541 			f2fs_balance_fs(sbi, dn.node_changed);
1542 
1543 			if (ret)
1544 				goto out;
1545 
1546 			index = end;
1547 			new_size = max_t(loff_t, new_size,
1548 					(loff_t)index << PAGE_SHIFT);
1549 		}
1550 
1551 		if (off_end) {
1552 			ret = fill_zero(inode, pg_end, 0, off_end);
1553 			if (ret)
1554 				goto out;
1555 
1556 			new_size = max_t(loff_t, new_size, offset + len);
1557 		}
1558 	}
1559 
1560 out:
1561 	if (new_size > i_size_read(inode)) {
1562 		if (mode & FALLOC_FL_KEEP_SIZE)
1563 			file_set_keep_isize(inode);
1564 		else
1565 			f2fs_i_size_write(inode, new_size);
1566 	}
1567 	return ret;
1568 }
1569 
f2fs_insert_range(struct inode * inode,loff_t offset,loff_t len)1570 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1571 {
1572 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1573 	pgoff_t nr, pg_start, pg_end, delta, idx;
1574 	loff_t new_size;
1575 	int ret = 0;
1576 
1577 	new_size = i_size_read(inode) + len;
1578 	ret = inode_newsize_ok(inode, new_size);
1579 	if (ret)
1580 		return ret;
1581 
1582 	if (offset >= i_size_read(inode))
1583 		return -EINVAL;
1584 
1585 	/* insert range should be aligned to block size of f2fs. */
1586 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1587 		return -EINVAL;
1588 
1589 	ret = f2fs_convert_inline_inode(inode);
1590 	if (ret)
1591 		return ret;
1592 
1593 	f2fs_balance_fs(sbi, true);
1594 
1595 	f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
1596 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1597 	f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
1598 	if (ret)
1599 		return ret;
1600 
1601 	/* write out all dirty pages from offset */
1602 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1603 	if (ret)
1604 		return ret;
1605 
1606 	pg_start = offset >> PAGE_SHIFT;
1607 	pg_end = (offset + len) >> PAGE_SHIFT;
1608 	delta = pg_end - pg_start;
1609 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1610 
1611 	/* avoid gc operation during block exchange */
1612 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1613 	f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
1614 	truncate_pagecache(inode, offset);
1615 
1616 	while (!ret && idx > pg_start) {
1617 		nr = idx - pg_start;
1618 		if (nr > delta)
1619 			nr = delta;
1620 		idx -= nr;
1621 
1622 		f2fs_lock_op(sbi);
1623 		f2fs_drop_extent_tree(inode);
1624 
1625 		ret = __exchange_data_block(inode, inode, idx,
1626 					idx + delta, nr, false);
1627 		f2fs_unlock_op(sbi);
1628 	}
1629 	f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
1630 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1631 
1632 	/* write out all moved pages, if possible */
1633 	f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
1634 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1635 	truncate_pagecache(inode, offset);
1636 	f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
1637 
1638 	if (!ret)
1639 		f2fs_i_size_write(inode, new_size);
1640 	return ret;
1641 }
1642 
expand_inode_data(struct inode * inode,loff_t offset,loff_t len,int mode)1643 static int expand_inode_data(struct inode *inode, loff_t offset,
1644 					loff_t len, int mode)
1645 {
1646 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1647 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1648 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1649 			.m_may_create = true };
1650 	pgoff_t pg_start, pg_end;
1651 	loff_t new_size = i_size_read(inode);
1652 	loff_t off_end;
1653 	block_t expanded = 0;
1654 	int err;
1655 
1656 	err = inode_newsize_ok(inode, (len + offset));
1657 	if (err)
1658 		return err;
1659 
1660 	err = f2fs_convert_inline_inode(inode);
1661 	if (err)
1662 		return err;
1663 
1664 	f2fs_balance_fs(sbi, true);
1665 
1666 	pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1667 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1668 	off_end = (offset + len) & (PAGE_SIZE - 1);
1669 
1670 	map.m_lblk = pg_start;
1671 	map.m_len = pg_end - pg_start;
1672 	if (off_end)
1673 		map.m_len++;
1674 
1675 	if (!map.m_len)
1676 		return 0;
1677 
1678 	if (f2fs_is_pinned_file(inode)) {
1679 		block_t sec_blks = BLKS_PER_SEC(sbi);
1680 		block_t sec_len = roundup(map.m_len, sec_blks);
1681 
1682 		map.m_len = sec_blks;
1683 next_alloc:
1684 		if (has_not_enough_free_secs(sbi, 0,
1685 			GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1686 			f2fs_down_write(&sbi->gc_lock);
1687 			err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1688 			if (err && err != -ENODATA && err != -EAGAIN)
1689 				goto out_err;
1690 		}
1691 
1692 		f2fs_down_write(&sbi->pin_sem);
1693 
1694 		f2fs_lock_op(sbi);
1695 		f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1696 		f2fs_unlock_op(sbi);
1697 
1698 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1699 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1700 		file_dont_truncate(inode);
1701 
1702 		f2fs_up_write(&sbi->pin_sem);
1703 
1704 		expanded += map.m_len;
1705 		sec_len -= map.m_len;
1706 		map.m_lblk += map.m_len;
1707 		if (!err && sec_len)
1708 			goto next_alloc;
1709 
1710 		map.m_len = expanded;
1711 	} else {
1712 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1713 		expanded = map.m_len;
1714 	}
1715 out_err:
1716 	if (err) {
1717 		pgoff_t last_off;
1718 
1719 		if (!expanded)
1720 			return err;
1721 
1722 		last_off = pg_start + expanded - 1;
1723 
1724 		/* update new size to the failed position */
1725 		new_size = (last_off == pg_end) ? offset + len :
1726 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1727 	} else {
1728 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1729 	}
1730 
1731 	if (new_size > i_size_read(inode)) {
1732 		if (mode & FALLOC_FL_KEEP_SIZE)
1733 			file_set_keep_isize(inode);
1734 		else
1735 			f2fs_i_size_write(inode, new_size);
1736 	}
1737 
1738 	return err;
1739 }
1740 
f2fs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)1741 static long f2fs_fallocate(struct file *file, int mode,
1742 				loff_t offset, loff_t len)
1743 {
1744 	struct inode *inode = file_inode(file);
1745 	long ret = 0;
1746 
1747 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1748 		return -EIO;
1749 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1750 		return -ENOSPC;
1751 	if (!f2fs_is_compress_backend_ready(inode))
1752 		return -EOPNOTSUPP;
1753 
1754 	/* f2fs only support ->fallocate for regular file */
1755 	if (!S_ISREG(inode->i_mode))
1756 		return -EINVAL;
1757 
1758 	if (IS_ENCRYPTED(inode) &&
1759 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1760 		return -EOPNOTSUPP;
1761 
1762 	/*
1763 	 * Pinned file should not support partial trucation since the block
1764 	 * can be used by applications.
1765 	 */
1766 	if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1767 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1768 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1769 		return -EOPNOTSUPP;
1770 
1771 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1772 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1773 			FALLOC_FL_INSERT_RANGE))
1774 		return -EOPNOTSUPP;
1775 
1776 	inode_lock(inode);
1777 
1778 	ret = file_modified(file);
1779 	if (ret)
1780 		goto out;
1781 
1782 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1783 		if (offset >= inode->i_size)
1784 			goto out;
1785 
1786 		ret = punch_hole(inode, offset, len);
1787 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1788 		ret = f2fs_collapse_range(inode, offset, len);
1789 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1790 		ret = f2fs_zero_range(inode, offset, len, mode);
1791 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1792 		ret = f2fs_insert_range(inode, offset, len);
1793 	} else {
1794 		ret = expand_inode_data(inode, offset, len, mode);
1795 	}
1796 
1797 	if (!ret) {
1798 		inode->i_mtime = inode->i_ctime = current_time(inode);
1799 		f2fs_mark_inode_dirty_sync(inode, false);
1800 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1801 	}
1802 
1803 out:
1804 	inode_unlock(inode);
1805 
1806 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1807 	return ret;
1808 }
1809 
f2fs_release_file(struct inode * inode,struct file * filp)1810 static int f2fs_release_file(struct inode *inode, struct file *filp)
1811 {
1812 	/*
1813 	 * f2fs_relase_file is called at every close calls. So we should
1814 	 * not drop any inmemory pages by close called by other process.
1815 	 */
1816 	if (!(filp->f_mode & FMODE_WRITE) ||
1817 			atomic_read(&inode->i_writecount) != 1)
1818 		return 0;
1819 
1820 	/* some remained atomic pages should discarded */
1821 	if (f2fs_is_atomic_file(inode))
1822 		f2fs_drop_inmem_pages(inode);
1823 	if (f2fs_is_volatile_file(inode)) {
1824 		set_inode_flag(inode, FI_DROP_CACHE);
1825 		filemap_fdatawrite(inode->i_mapping);
1826 		clear_inode_flag(inode, FI_DROP_CACHE);
1827 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1828 		stat_dec_volatile_write(inode);
1829 	}
1830 	return 0;
1831 }
1832 
f2fs_file_flush(struct file * file,fl_owner_t id)1833 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1834 {
1835 	struct inode *inode = file_inode(file);
1836 
1837 	/*
1838 	 * If the process doing a transaction is crashed, we should do
1839 	 * roll-back. Otherwise, other reader/write can see corrupted database
1840 	 * until all the writers close its file. Since this should be done
1841 	 * before dropping file lock, it needs to do in ->flush.
1842 	 */
1843 	if (f2fs_is_atomic_file(inode) &&
1844 			F2FS_I(inode)->inmem_task == current)
1845 		f2fs_drop_inmem_pages(inode);
1846 	return 0;
1847 }
1848 
f2fs_setflags_common(struct inode * inode,u32 iflags,u32 mask)1849 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1850 {
1851 	struct f2fs_inode_info *fi = F2FS_I(inode);
1852 	u32 masked_flags = fi->i_flags & mask;
1853 
1854 	/* mask can be shrunk by flags_valid selector */
1855 	iflags &= mask;
1856 
1857 	/* Is it quota file? Do not allow user to mess with it */
1858 	if (IS_NOQUOTA(inode))
1859 		return -EPERM;
1860 
1861 	if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1862 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1863 			return -EOPNOTSUPP;
1864 		if (!f2fs_empty_dir(inode))
1865 			return -ENOTEMPTY;
1866 	}
1867 
1868 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1869 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1870 			return -EOPNOTSUPP;
1871 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1872 			return -EINVAL;
1873 	}
1874 
1875 	if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1876 		if (masked_flags & F2FS_COMPR_FL) {
1877 			if (!f2fs_disable_compressed_file(inode))
1878 				return -EINVAL;
1879 		} else {
1880 			if (!f2fs_may_compress(inode))
1881 				return -EINVAL;
1882 			if (S_ISREG(inode->i_mode) && inode->i_size)
1883 				return -EINVAL;
1884 			if (set_compress_context(inode))
1885 				return -EOPNOTSUPP;
1886 		}
1887 	}
1888 
1889 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1890 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1891 					(fi->i_flags & F2FS_NOCOMP_FL));
1892 
1893 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1894 		set_inode_flag(inode, FI_PROJ_INHERIT);
1895 	else
1896 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1897 
1898 	inode->i_ctime = current_time(inode);
1899 	f2fs_set_inode_flags(inode);
1900 	f2fs_mark_inode_dirty_sync(inode, true);
1901 	return 0;
1902 }
1903 
1904 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1905 
1906 /*
1907  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1908  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1909  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1910  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1911  */
1912 
1913 static const struct {
1914 	u32 iflag;
1915 	u32 fsflag;
1916 } f2fs_fsflags_map[] = {
1917 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
1918 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1919 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1920 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1921 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1922 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1923 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
1924 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1925 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1926 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1927 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
1928 };
1929 
1930 #define F2FS_GETTABLE_FS_FL (		\
1931 		FS_COMPR_FL |		\
1932 		FS_SYNC_FL |		\
1933 		FS_IMMUTABLE_FL |	\
1934 		FS_APPEND_FL |		\
1935 		FS_NODUMP_FL |		\
1936 		FS_NOATIME_FL |		\
1937 		FS_NOCOMP_FL |		\
1938 		FS_INDEX_FL |		\
1939 		FS_DIRSYNC_FL |		\
1940 		FS_PROJINHERIT_FL |	\
1941 		FS_ENCRYPT_FL |		\
1942 		FS_INLINE_DATA_FL |	\
1943 		FS_NOCOW_FL |		\
1944 		FS_VERITY_FL |		\
1945 		FS_CASEFOLD_FL)
1946 
1947 #define F2FS_SETTABLE_FS_FL (		\
1948 		FS_COMPR_FL |		\
1949 		FS_SYNC_FL |		\
1950 		FS_IMMUTABLE_FL |	\
1951 		FS_APPEND_FL |		\
1952 		FS_NODUMP_FL |		\
1953 		FS_NOATIME_FL |		\
1954 		FS_NOCOMP_FL |		\
1955 		FS_DIRSYNC_FL |		\
1956 		FS_PROJINHERIT_FL |	\
1957 		FS_CASEFOLD_FL)
1958 
1959 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
f2fs_iflags_to_fsflags(u32 iflags)1960 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1961 {
1962 	u32 fsflags = 0;
1963 	int i;
1964 
1965 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1966 		if (iflags & f2fs_fsflags_map[i].iflag)
1967 			fsflags |= f2fs_fsflags_map[i].fsflag;
1968 
1969 	return fsflags;
1970 }
1971 
1972 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
f2fs_fsflags_to_iflags(u32 fsflags)1973 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1974 {
1975 	u32 iflags = 0;
1976 	int i;
1977 
1978 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1979 		if (fsflags & f2fs_fsflags_map[i].fsflag)
1980 			iflags |= f2fs_fsflags_map[i].iflag;
1981 
1982 	return iflags;
1983 }
1984 
f2fs_ioc_getflags(struct file * filp,unsigned long arg)1985 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1986 {
1987 	struct inode *inode = file_inode(filp);
1988 	struct f2fs_inode_info *fi = F2FS_I(inode);
1989 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1990 
1991 	if (IS_ENCRYPTED(inode))
1992 		fsflags |= FS_ENCRYPT_FL;
1993 	if (IS_VERITY(inode))
1994 		fsflags |= FS_VERITY_FL;
1995 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1996 		fsflags |= FS_INLINE_DATA_FL;
1997 	if (is_inode_flag_set(inode, FI_PIN_FILE))
1998 		fsflags |= FS_NOCOW_FL;
1999 
2000 	fsflags &= F2FS_GETTABLE_FS_FL;
2001 
2002 	return put_user(fsflags, (int __user *)arg);
2003 }
2004 
f2fs_ioc_setflags(struct file * filp,unsigned long arg)2005 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
2006 {
2007 	struct inode *inode = file_inode(filp);
2008 	struct f2fs_inode_info *fi = F2FS_I(inode);
2009 	u32 fsflags, old_fsflags;
2010 	u32 iflags;
2011 	int ret;
2012 
2013 	if (!inode_owner_or_capable(inode))
2014 		return -EACCES;
2015 
2016 	if (get_user(fsflags, (int __user *)arg))
2017 		return -EFAULT;
2018 
2019 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
2020 		return -EOPNOTSUPP;
2021 	fsflags &= F2FS_SETTABLE_FS_FL;
2022 
2023 	iflags = f2fs_fsflags_to_iflags(fsflags);
2024 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2025 		return -EOPNOTSUPP;
2026 
2027 	ret = mnt_want_write_file(filp);
2028 	if (ret)
2029 		return ret;
2030 
2031 	inode_lock(inode);
2032 
2033 	old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
2034 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
2035 	if (ret)
2036 		goto out;
2037 
2038 	ret = f2fs_setflags_common(inode, iflags,
2039 			f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
2040 out:
2041 	inode_unlock(inode);
2042 	mnt_drop_write_file(filp);
2043 	return ret;
2044 }
2045 
f2fs_ioc_getversion(struct file * filp,unsigned long arg)2046 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2047 {
2048 	struct inode *inode = file_inode(filp);
2049 
2050 	return put_user(inode->i_generation, (int __user *)arg);
2051 }
2052 
f2fs_ioc_start_atomic_write(struct file * filp)2053 static int f2fs_ioc_start_atomic_write(struct file *filp)
2054 {
2055 	struct inode *inode = file_inode(filp);
2056 	struct f2fs_inode_info *fi = F2FS_I(inode);
2057 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2058 	int ret;
2059 
2060 	if (!inode_owner_or_capable(inode))
2061 		return -EACCES;
2062 
2063 	if (!S_ISREG(inode->i_mode))
2064 		return -EINVAL;
2065 
2066 	if (filp->f_flags & O_DIRECT)
2067 		return -EINVAL;
2068 
2069 	ret = mnt_want_write_file(filp);
2070 	if (ret)
2071 		return ret;
2072 
2073 	inode_lock(inode);
2074 
2075 	if (!f2fs_disable_compressed_file(inode)) {
2076 		ret = -EINVAL;
2077 		goto out;
2078 	}
2079 
2080 	if (f2fs_is_atomic_file(inode)) {
2081 		if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2082 			ret = -EINVAL;
2083 		goto out;
2084 	}
2085 
2086 	ret = f2fs_convert_inline_inode(inode);
2087 	if (ret)
2088 		goto out;
2089 
2090 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2091 
2092 	/*
2093 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2094 	 * f2fs_is_atomic_file.
2095 	 */
2096 	if (get_dirty_pages(inode))
2097 		f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2098 			  inode->i_ino, get_dirty_pages(inode));
2099 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2100 	if (ret) {
2101 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2102 		goto out;
2103 	}
2104 
2105 	spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2106 	if (list_empty(&fi->inmem_ilist))
2107 		list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2108 	sbi->atomic_files++;
2109 	spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2110 
2111 	/* add inode in inmem_list first and set atomic_file */
2112 	set_inode_flag(inode, FI_ATOMIC_FILE);
2113 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2114 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2115 
2116 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2117 	F2FS_I(inode)->inmem_task = current;
2118 	stat_update_max_atomic_write(inode);
2119 out:
2120 	inode_unlock(inode);
2121 	mnt_drop_write_file(filp);
2122 	return ret;
2123 }
2124 
f2fs_ioc_commit_atomic_write(struct file * filp)2125 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2126 {
2127 	struct inode *inode = file_inode(filp);
2128 	int ret;
2129 
2130 	if (!inode_owner_or_capable(inode))
2131 		return -EACCES;
2132 
2133 	ret = mnt_want_write_file(filp);
2134 	if (ret)
2135 		return ret;
2136 
2137 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2138 
2139 	inode_lock(inode);
2140 
2141 	if (f2fs_is_volatile_file(inode)) {
2142 		ret = -EINVAL;
2143 		goto err_out;
2144 	}
2145 
2146 	if (f2fs_is_atomic_file(inode)) {
2147 		ret = f2fs_commit_inmem_pages(inode);
2148 		if (ret)
2149 			goto err_out;
2150 
2151 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2152 		if (!ret)
2153 			f2fs_drop_inmem_pages(inode);
2154 	} else {
2155 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2156 	}
2157 err_out:
2158 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2159 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2160 		ret = -EINVAL;
2161 	}
2162 	inode_unlock(inode);
2163 	mnt_drop_write_file(filp);
2164 	return ret;
2165 }
2166 
f2fs_ioc_start_volatile_write(struct file * filp)2167 static int f2fs_ioc_start_volatile_write(struct file *filp)
2168 {
2169 	struct inode *inode = file_inode(filp);
2170 	int ret;
2171 
2172 	if (!inode_owner_or_capable(inode))
2173 		return -EACCES;
2174 
2175 	if (!S_ISREG(inode->i_mode))
2176 		return -EINVAL;
2177 
2178 	ret = mnt_want_write_file(filp);
2179 	if (ret)
2180 		return ret;
2181 
2182 	inode_lock(inode);
2183 
2184 	if (f2fs_is_volatile_file(inode))
2185 		goto out;
2186 
2187 	ret = f2fs_convert_inline_inode(inode);
2188 	if (ret)
2189 		goto out;
2190 
2191 	stat_inc_volatile_write(inode);
2192 	stat_update_max_volatile_write(inode);
2193 
2194 	set_inode_flag(inode, FI_VOLATILE_FILE);
2195 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2196 out:
2197 	inode_unlock(inode);
2198 	mnt_drop_write_file(filp);
2199 	return ret;
2200 }
2201 
f2fs_ioc_release_volatile_write(struct file * filp)2202 static int f2fs_ioc_release_volatile_write(struct file *filp)
2203 {
2204 	struct inode *inode = file_inode(filp);
2205 	int ret;
2206 
2207 	if (!inode_owner_or_capable(inode))
2208 		return -EACCES;
2209 
2210 	ret = mnt_want_write_file(filp);
2211 	if (ret)
2212 		return ret;
2213 
2214 	inode_lock(inode);
2215 
2216 	if (!f2fs_is_volatile_file(inode))
2217 		goto out;
2218 
2219 	if (!f2fs_is_first_block_written(inode)) {
2220 		ret = truncate_partial_data_page(inode, 0, true);
2221 		goto out;
2222 	}
2223 
2224 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2225 out:
2226 	inode_unlock(inode);
2227 	mnt_drop_write_file(filp);
2228 	return ret;
2229 }
2230 
f2fs_ioc_abort_volatile_write(struct file * filp)2231 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2232 {
2233 	struct inode *inode = file_inode(filp);
2234 	int ret;
2235 
2236 	if (!inode_owner_or_capable(inode))
2237 		return -EACCES;
2238 
2239 	ret = mnt_want_write_file(filp);
2240 	if (ret)
2241 		return ret;
2242 
2243 	inode_lock(inode);
2244 
2245 	if (f2fs_is_atomic_file(inode))
2246 		f2fs_drop_inmem_pages(inode);
2247 	if (f2fs_is_volatile_file(inode)) {
2248 		clear_inode_flag(inode, FI_VOLATILE_FILE);
2249 		stat_dec_volatile_write(inode);
2250 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2251 	}
2252 
2253 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2254 
2255 	inode_unlock(inode);
2256 
2257 	mnt_drop_write_file(filp);
2258 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2259 	return ret;
2260 }
2261 
f2fs_ioc_shutdown(struct file * filp,unsigned long arg)2262 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2263 {
2264 	struct inode *inode = file_inode(filp);
2265 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2266 	struct super_block *sb = sbi->sb;
2267 	__u32 in;
2268 	int ret = 0;
2269 
2270 	if (!capable(CAP_SYS_ADMIN))
2271 		return -EPERM;
2272 
2273 	if (get_user(in, (__u32 __user *)arg))
2274 		return -EFAULT;
2275 
2276 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2277 		ret = mnt_want_write_file(filp);
2278 		if (ret) {
2279 			if (ret == -EROFS) {
2280 				ret = 0;
2281 				f2fs_stop_checkpoint(sbi, false,
2282 						STOP_CP_REASON_SHUTDOWN);
2283 				set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2284 				trace_f2fs_shutdown(sbi, in, ret);
2285 			}
2286 			return ret;
2287 		}
2288 	}
2289 
2290 	switch (in) {
2291 	case F2FS_GOING_DOWN_FULLSYNC:
2292 		ret = freeze_bdev(sb->s_bdev);
2293 		if (ret)
2294 			goto out;
2295 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2296 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2297 		thaw_bdev(sb->s_bdev);
2298 		break;
2299 	case F2FS_GOING_DOWN_METASYNC:
2300 		/* do checkpoint only */
2301 		ret = f2fs_sync_fs(sb, 1);
2302 		if (ret)
2303 			goto out;
2304 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2305 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2306 		break;
2307 	case F2FS_GOING_DOWN_NOSYNC:
2308 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2309 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2310 		break;
2311 	case F2FS_GOING_DOWN_METAFLUSH:
2312 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2313 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2314 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2315 		break;
2316 	case F2FS_GOING_DOWN_NEED_FSCK:
2317 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2318 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2319 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2320 		/* do checkpoint only */
2321 		ret = f2fs_sync_fs(sb, 1);
2322 		goto out;
2323 	default:
2324 		ret = -EINVAL;
2325 		goto out;
2326 	}
2327 
2328 	f2fs_stop_gc_thread(sbi);
2329 	f2fs_stop_discard_thread(sbi);
2330 
2331 	f2fs_drop_discard_cmd(sbi);
2332 	clear_opt(sbi, DISCARD);
2333 
2334 	f2fs_update_time(sbi, REQ_TIME);
2335 out:
2336 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2337 		mnt_drop_write_file(filp);
2338 
2339 	trace_f2fs_shutdown(sbi, in, ret);
2340 
2341 	return ret;
2342 }
2343 
f2fs_ioc_fitrim(struct file * filp,unsigned long arg)2344 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2345 {
2346 	struct inode *inode = file_inode(filp);
2347 	struct super_block *sb = inode->i_sb;
2348 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
2349 	struct fstrim_range range;
2350 	int ret;
2351 
2352 	if (!capable(CAP_SYS_ADMIN))
2353 		return -EPERM;
2354 
2355 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2356 		return -EOPNOTSUPP;
2357 
2358 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2359 				sizeof(range)))
2360 		return -EFAULT;
2361 
2362 	ret = mnt_want_write_file(filp);
2363 	if (ret)
2364 		return ret;
2365 
2366 	range.minlen = max((unsigned int)range.minlen,
2367 				q->limits.discard_granularity);
2368 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2369 	mnt_drop_write_file(filp);
2370 	if (ret < 0)
2371 		return ret;
2372 
2373 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2374 				sizeof(range)))
2375 		return -EFAULT;
2376 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2377 	return 0;
2378 }
2379 
uuid_is_nonzero(__u8 u[16])2380 static bool uuid_is_nonzero(__u8 u[16])
2381 {
2382 	int i;
2383 
2384 	for (i = 0; i < 16; i++)
2385 		if (u[i])
2386 			return true;
2387 	return false;
2388 }
2389 
f2fs_ioc_set_encryption_policy(struct file * filp,unsigned long arg)2390 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2391 {
2392 	struct inode *inode = file_inode(filp);
2393 
2394 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2395 		return -EOPNOTSUPP;
2396 
2397 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2398 
2399 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2400 }
2401 
f2fs_ioc_get_encryption_policy(struct file * filp,unsigned long arg)2402 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2403 {
2404 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2405 		return -EOPNOTSUPP;
2406 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2407 }
2408 
f2fs_ioc_get_encryption_pwsalt(struct file * filp,unsigned long arg)2409 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2410 {
2411 	struct inode *inode = file_inode(filp);
2412 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2413 	int err;
2414 
2415 	if (!f2fs_sb_has_encrypt(sbi))
2416 		return -EOPNOTSUPP;
2417 
2418 	err = mnt_want_write_file(filp);
2419 	if (err)
2420 		return err;
2421 
2422 	f2fs_down_write(&sbi->sb_lock);
2423 
2424 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2425 		goto got_it;
2426 
2427 	/* update superblock with uuid */
2428 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2429 
2430 	err = f2fs_commit_super(sbi, false);
2431 	if (err) {
2432 		/* undo new data */
2433 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2434 		goto out_err;
2435 	}
2436 got_it:
2437 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2438 									16))
2439 		err = -EFAULT;
2440 out_err:
2441 	f2fs_up_write(&sbi->sb_lock);
2442 	mnt_drop_write_file(filp);
2443 	return err;
2444 }
2445 
f2fs_ioc_get_encryption_policy_ex(struct file * filp,unsigned long arg)2446 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2447 					     unsigned long arg)
2448 {
2449 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2450 		return -EOPNOTSUPP;
2451 
2452 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2453 }
2454 
f2fs_ioc_add_encryption_key(struct file * filp,unsigned long arg)2455 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2456 {
2457 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2458 		return -EOPNOTSUPP;
2459 
2460 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2461 }
2462 
f2fs_ioc_remove_encryption_key(struct file * filp,unsigned long arg)2463 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2464 {
2465 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2466 		return -EOPNOTSUPP;
2467 
2468 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2469 }
2470 
f2fs_ioc_remove_encryption_key_all_users(struct file * filp,unsigned long arg)2471 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2472 						    unsigned long arg)
2473 {
2474 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2475 		return -EOPNOTSUPP;
2476 
2477 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2478 }
2479 
f2fs_ioc_get_encryption_key_status(struct file * filp,unsigned long arg)2480 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2481 					      unsigned long arg)
2482 {
2483 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2484 		return -EOPNOTSUPP;
2485 
2486 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2487 }
2488 
f2fs_ioc_get_encryption_nonce(struct file * filp,unsigned long arg)2489 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2490 {
2491 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2492 		return -EOPNOTSUPP;
2493 
2494 	return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2495 }
2496 
f2fs_ioc_gc(struct file * filp,unsigned long arg)2497 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2498 {
2499 	struct inode *inode = file_inode(filp);
2500 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2501 	__u32 sync;
2502 	int ret;
2503 
2504 	if (!capable(CAP_SYS_ADMIN))
2505 		return -EPERM;
2506 
2507 	if (get_user(sync, (__u32 __user *)arg))
2508 		return -EFAULT;
2509 
2510 	if (f2fs_readonly(sbi->sb))
2511 		return -EROFS;
2512 
2513 	ret = mnt_want_write_file(filp);
2514 	if (ret)
2515 		return ret;
2516 
2517 	if (!sync) {
2518 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2519 			ret = -EBUSY;
2520 			goto out;
2521 		}
2522 	} else {
2523 		f2fs_down_write(&sbi->gc_lock);
2524 	}
2525 
2526 	ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO);
2527 out:
2528 	mnt_drop_write_file(filp);
2529 	return ret;
2530 }
2531 
__f2fs_ioc_gc_range(struct file * filp,struct f2fs_gc_range * range)2532 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2533 {
2534 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2535 	u64 end;
2536 	int ret;
2537 
2538 	if (!capable(CAP_SYS_ADMIN))
2539 		return -EPERM;
2540 	if (f2fs_readonly(sbi->sb))
2541 		return -EROFS;
2542 
2543 	end = range->start + range->len;
2544 	if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2545 					end >= MAX_BLKADDR(sbi))
2546 		return -EINVAL;
2547 
2548 	ret = mnt_want_write_file(filp);
2549 	if (ret)
2550 		return ret;
2551 
2552 do_more:
2553 	if (!range->sync) {
2554 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2555 			ret = -EBUSY;
2556 			goto out;
2557 		}
2558 	} else {
2559 		f2fs_down_write(&sbi->gc_lock);
2560 	}
2561 
2562 	ret = f2fs_gc(sbi, range->sync, true, false,
2563 				GET_SEGNO(sbi, range->start));
2564 	if (ret) {
2565 		if (ret == -EBUSY)
2566 			ret = -EAGAIN;
2567 		goto out;
2568 	}
2569 	range->start += BLKS_PER_SEC(sbi);
2570 	if (range->start <= end)
2571 		goto do_more;
2572 out:
2573 	mnt_drop_write_file(filp);
2574 	return ret;
2575 }
2576 
f2fs_ioc_gc_range(struct file * filp,unsigned long arg)2577 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2578 {
2579 	struct f2fs_gc_range range;
2580 
2581 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2582 							sizeof(range)))
2583 		return -EFAULT;
2584 	return __f2fs_ioc_gc_range(filp, &range);
2585 }
2586 
f2fs_ioc_write_checkpoint(struct file * filp,unsigned long arg)2587 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2588 {
2589 	struct inode *inode = file_inode(filp);
2590 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2591 	int ret;
2592 
2593 	if (!capable(CAP_SYS_ADMIN))
2594 		return -EPERM;
2595 
2596 	if (f2fs_readonly(sbi->sb))
2597 		return -EROFS;
2598 
2599 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2600 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2601 		return -EINVAL;
2602 	}
2603 
2604 	ret = mnt_want_write_file(filp);
2605 	if (ret)
2606 		return ret;
2607 
2608 	ret = f2fs_sync_fs(sbi->sb, 1);
2609 
2610 	mnt_drop_write_file(filp);
2611 	return ret;
2612 }
2613 
f2fs_defragment_range(struct f2fs_sb_info * sbi,struct file * filp,struct f2fs_defragment * range)2614 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2615 					struct file *filp,
2616 					struct f2fs_defragment *range)
2617 {
2618 	struct inode *inode = file_inode(filp);
2619 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2620 					.m_seg_type = NO_CHECK_TYPE,
2621 					.m_may_create = false };
2622 	struct extent_info ei = {0, 0, 0};
2623 	pgoff_t pg_start, pg_end, next_pgofs;
2624 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2625 	unsigned int total = 0, sec_num;
2626 	block_t blk_end = 0;
2627 	bool fragmented = false;
2628 	int err;
2629 
2630 	pg_start = range->start >> PAGE_SHIFT;
2631 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2632 
2633 	f2fs_balance_fs(sbi, true);
2634 
2635 	inode_lock(inode);
2636 
2637 	/* if in-place-update policy is enabled, don't waste time here */
2638 	set_inode_flag(inode, FI_OPU_WRITE);
2639 	if (f2fs_should_update_inplace(inode, NULL)) {
2640 		err = -EINVAL;
2641 		goto out;
2642 	}
2643 
2644 	/* writeback all dirty pages in the range */
2645 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2646 						range->start + range->len - 1);
2647 	if (err)
2648 		goto out;
2649 
2650 	/*
2651 	 * lookup mapping info in extent cache, skip defragmenting if physical
2652 	 * block addresses are continuous.
2653 	 */
2654 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2655 		if (ei.fofs + ei.len >= pg_end)
2656 			goto out;
2657 	}
2658 
2659 	map.m_lblk = pg_start;
2660 	map.m_next_pgofs = &next_pgofs;
2661 
2662 	/*
2663 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2664 	 * physical block addresses are continuous even if there are hole(s)
2665 	 * in logical blocks.
2666 	 */
2667 	while (map.m_lblk < pg_end) {
2668 		map.m_len = pg_end - map.m_lblk;
2669 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2670 		if (err)
2671 			goto out;
2672 
2673 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2674 			map.m_lblk = next_pgofs;
2675 			continue;
2676 		}
2677 
2678 		if (blk_end && blk_end != map.m_pblk)
2679 			fragmented = true;
2680 
2681 		/* record total count of block that we're going to move */
2682 		total += map.m_len;
2683 
2684 		blk_end = map.m_pblk + map.m_len;
2685 
2686 		map.m_lblk += map.m_len;
2687 	}
2688 
2689 	if (!fragmented) {
2690 		total = 0;
2691 		goto out;
2692 	}
2693 
2694 	sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2695 
2696 	/*
2697 	 * make sure there are enough free section for LFS allocation, this can
2698 	 * avoid defragment running in SSR mode when free section are allocated
2699 	 * intensively
2700 	 */
2701 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2702 		err = -EAGAIN;
2703 		goto out;
2704 	}
2705 
2706 	map.m_lblk = pg_start;
2707 	map.m_len = pg_end - pg_start;
2708 	total = 0;
2709 
2710 	while (map.m_lblk < pg_end) {
2711 		pgoff_t idx;
2712 		int cnt = 0;
2713 
2714 do_map:
2715 		map.m_len = pg_end - map.m_lblk;
2716 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2717 		if (err)
2718 			goto clear_out;
2719 
2720 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2721 			map.m_lblk = next_pgofs;
2722 			goto check;
2723 		}
2724 
2725 		set_inode_flag(inode, FI_SKIP_WRITES);
2726 
2727 		idx = map.m_lblk;
2728 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2729 			struct page *page;
2730 
2731 			page = f2fs_get_lock_data_page(inode, idx, true);
2732 			if (IS_ERR(page)) {
2733 				err = PTR_ERR(page);
2734 				goto clear_out;
2735 			}
2736 
2737 			set_page_dirty(page);
2738 			f2fs_put_page(page, 1);
2739 
2740 			idx++;
2741 			cnt++;
2742 			total++;
2743 		}
2744 
2745 		map.m_lblk = idx;
2746 check:
2747 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
2748 			goto do_map;
2749 
2750 		clear_inode_flag(inode, FI_SKIP_WRITES);
2751 
2752 		err = filemap_fdatawrite(inode->i_mapping);
2753 		if (err)
2754 			goto out;
2755 	}
2756 clear_out:
2757 	clear_inode_flag(inode, FI_SKIP_WRITES);
2758 out:
2759 	clear_inode_flag(inode, FI_OPU_WRITE);
2760 	inode_unlock(inode);
2761 	if (!err)
2762 		range->len = (u64)total << PAGE_SHIFT;
2763 	return err;
2764 }
2765 
f2fs_ioc_defragment(struct file * filp,unsigned long arg)2766 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2767 {
2768 	struct inode *inode = file_inode(filp);
2769 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2770 	struct f2fs_defragment range;
2771 	int err;
2772 
2773 	if (!capable(CAP_SYS_ADMIN))
2774 		return -EPERM;
2775 
2776 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2777 		return -EINVAL;
2778 
2779 	if (f2fs_readonly(sbi->sb))
2780 		return -EROFS;
2781 
2782 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2783 							sizeof(range)))
2784 		return -EFAULT;
2785 
2786 	/* verify alignment of offset & size */
2787 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2788 		return -EINVAL;
2789 
2790 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2791 					max_file_blocks(inode)))
2792 		return -EINVAL;
2793 
2794 	err = mnt_want_write_file(filp);
2795 	if (err)
2796 		return err;
2797 
2798 	err = f2fs_defragment_range(sbi, filp, &range);
2799 	mnt_drop_write_file(filp);
2800 
2801 	f2fs_update_time(sbi, REQ_TIME);
2802 	if (err < 0)
2803 		return err;
2804 
2805 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2806 							sizeof(range)))
2807 		return -EFAULT;
2808 
2809 	return 0;
2810 }
2811 
f2fs_move_file_range(struct file * file_in,loff_t pos_in,struct file * file_out,loff_t pos_out,size_t len)2812 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2813 			struct file *file_out, loff_t pos_out, size_t len)
2814 {
2815 	struct inode *src = file_inode(file_in);
2816 	struct inode *dst = file_inode(file_out);
2817 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2818 	size_t olen = len, dst_max_i_size = 0;
2819 	size_t dst_osize;
2820 	int ret;
2821 
2822 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2823 				src->i_sb != dst->i_sb)
2824 		return -EXDEV;
2825 
2826 	if (unlikely(f2fs_readonly(src->i_sb)))
2827 		return -EROFS;
2828 
2829 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2830 		return -EINVAL;
2831 
2832 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2833 		return -EOPNOTSUPP;
2834 
2835 	if (pos_out < 0 || pos_in < 0)
2836 		return -EINVAL;
2837 
2838 	if (src == dst) {
2839 		if (pos_in == pos_out)
2840 			return 0;
2841 		if (pos_out > pos_in && pos_out < pos_in + len)
2842 			return -EINVAL;
2843 	}
2844 
2845 	inode_lock(src);
2846 	if (src != dst) {
2847 		ret = -EBUSY;
2848 		if (!inode_trylock(dst))
2849 			goto out;
2850 	}
2851 
2852 	if (f2fs_compressed_file(src) || f2fs_compressed_file(dst)) {
2853 		ret = -EOPNOTSUPP;
2854 		goto out_unlock;
2855 	}
2856 
2857 	ret = -EINVAL;
2858 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2859 		goto out_unlock;
2860 	if (len == 0)
2861 		olen = len = src->i_size - pos_in;
2862 	if (pos_in + len == src->i_size)
2863 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2864 	if (len == 0) {
2865 		ret = 0;
2866 		goto out_unlock;
2867 	}
2868 
2869 	dst_osize = dst->i_size;
2870 	if (pos_out + olen > dst->i_size)
2871 		dst_max_i_size = pos_out + olen;
2872 
2873 	/* verify the end result is block aligned */
2874 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2875 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2876 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2877 		goto out_unlock;
2878 
2879 	ret = f2fs_convert_inline_inode(src);
2880 	if (ret)
2881 		goto out_unlock;
2882 
2883 	ret = f2fs_convert_inline_inode(dst);
2884 	if (ret)
2885 		goto out_unlock;
2886 
2887 	/* write out all dirty pages from offset */
2888 	ret = filemap_write_and_wait_range(src->i_mapping,
2889 					pos_in, pos_in + len);
2890 	if (ret)
2891 		goto out_unlock;
2892 
2893 	ret = filemap_write_and_wait_range(dst->i_mapping,
2894 					pos_out, pos_out + len);
2895 	if (ret)
2896 		goto out_unlock;
2897 
2898 	f2fs_balance_fs(sbi, true);
2899 
2900 	f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2901 	if (src != dst) {
2902 		ret = -EBUSY;
2903 		if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2904 			goto out_src;
2905 	}
2906 
2907 	f2fs_lock_op(sbi);
2908 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2909 				pos_out >> F2FS_BLKSIZE_BITS,
2910 				len >> F2FS_BLKSIZE_BITS, false);
2911 
2912 	if (!ret) {
2913 		if (dst_max_i_size)
2914 			f2fs_i_size_write(dst, dst_max_i_size);
2915 		else if (dst_osize != dst->i_size)
2916 			f2fs_i_size_write(dst, dst_osize);
2917 	}
2918 	f2fs_unlock_op(sbi);
2919 
2920 	if (src != dst)
2921 		f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2922 out_src:
2923 	f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2924 out_unlock:
2925 	if (src != dst)
2926 		inode_unlock(dst);
2927 out:
2928 	inode_unlock(src);
2929 	return ret;
2930 }
2931 
__f2fs_ioc_move_range(struct file * filp,struct f2fs_move_range * range)2932 static int __f2fs_ioc_move_range(struct file *filp,
2933 				struct f2fs_move_range *range)
2934 {
2935 	struct fd dst;
2936 	int err;
2937 
2938 	if (!(filp->f_mode & FMODE_READ) ||
2939 			!(filp->f_mode & FMODE_WRITE))
2940 		return -EBADF;
2941 
2942 	dst = fdget(range->dst_fd);
2943 	if (!dst.file)
2944 		return -EBADF;
2945 
2946 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2947 		err = -EBADF;
2948 		goto err_out;
2949 	}
2950 
2951 	err = mnt_want_write_file(filp);
2952 	if (err)
2953 		goto err_out;
2954 
2955 	err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2956 					range->pos_out, range->len);
2957 
2958 	mnt_drop_write_file(filp);
2959 err_out:
2960 	fdput(dst);
2961 	return err;
2962 }
2963 
f2fs_ioc_move_range(struct file * filp,unsigned long arg)2964 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2965 {
2966 	struct f2fs_move_range range;
2967 
2968 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2969 							sizeof(range)))
2970 		return -EFAULT;
2971 	return __f2fs_ioc_move_range(filp, &range);
2972 }
2973 
f2fs_ioc_flush_device(struct file * filp,unsigned long arg)2974 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2975 {
2976 	struct inode *inode = file_inode(filp);
2977 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2978 	struct sit_info *sm = SIT_I(sbi);
2979 	unsigned int start_segno = 0, end_segno = 0;
2980 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2981 	struct f2fs_flush_device range;
2982 	int ret;
2983 
2984 	if (!capable(CAP_SYS_ADMIN))
2985 		return -EPERM;
2986 
2987 	if (f2fs_readonly(sbi->sb))
2988 		return -EROFS;
2989 
2990 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2991 		return -EINVAL;
2992 
2993 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2994 							sizeof(range)))
2995 		return -EFAULT;
2996 
2997 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2998 			__is_large_section(sbi)) {
2999 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
3000 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
3001 		return -EINVAL;
3002 	}
3003 
3004 	ret = mnt_want_write_file(filp);
3005 	if (ret)
3006 		return ret;
3007 
3008 	if (range.dev_num != 0)
3009 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
3010 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
3011 
3012 	start_segno = sm->last_victim[FLUSH_DEVICE];
3013 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
3014 		start_segno = dev_start_segno;
3015 	end_segno = min(start_segno + range.segments, dev_end_segno);
3016 
3017 	while (start_segno < end_segno) {
3018 		if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
3019 			ret = -EBUSY;
3020 			goto out;
3021 		}
3022 		sm->last_victim[GC_CB] = end_segno + 1;
3023 		sm->last_victim[GC_GREEDY] = end_segno + 1;
3024 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
3025 		ret = f2fs_gc(sbi, true, true, true, start_segno);
3026 		if (ret == -EAGAIN)
3027 			ret = 0;
3028 		else if (ret < 0)
3029 			break;
3030 		start_segno++;
3031 	}
3032 out:
3033 	mnt_drop_write_file(filp);
3034 	return ret;
3035 }
3036 
f2fs_ioc_get_features(struct file * filp,unsigned long arg)3037 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3038 {
3039 	struct inode *inode = file_inode(filp);
3040 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3041 
3042 	/* Must validate to set it with SQLite behavior in Android. */
3043 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3044 
3045 	return put_user(sb_feature, (u32 __user *)arg);
3046 }
3047 
3048 #ifdef CONFIG_QUOTA
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3049 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3050 {
3051 	struct dquot *transfer_to[MAXQUOTAS] = {};
3052 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3053 	struct super_block *sb = sbi->sb;
3054 	int err;
3055 
3056 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3057 	if (IS_ERR(transfer_to[PRJQUOTA]))
3058 		return PTR_ERR(transfer_to[PRJQUOTA]);
3059 
3060 	err = __dquot_transfer(inode, transfer_to);
3061 	if (err)
3062 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3063 	dqput(transfer_to[PRJQUOTA]);
3064 	return err;
3065 }
3066 
f2fs_ioc_setproject(struct file * filp,__u32 projid)3067 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3068 {
3069 	struct inode *inode = file_inode(filp);
3070 	struct f2fs_inode_info *fi = F2FS_I(inode);
3071 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3072 	struct f2fs_inode *ri = NULL;
3073 	kprojid_t kprojid;
3074 	int err;
3075 
3076 	if (!f2fs_sb_has_project_quota(sbi)) {
3077 		if (projid != F2FS_DEF_PROJID)
3078 			return -EOPNOTSUPP;
3079 		else
3080 			return 0;
3081 	}
3082 
3083 	if (!f2fs_has_extra_attr(inode))
3084 		return -EOPNOTSUPP;
3085 
3086 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3087 
3088 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
3089 		return 0;
3090 
3091 	err = -EPERM;
3092 	/* Is it quota file? Do not allow user to mess with it */
3093 	if (IS_NOQUOTA(inode))
3094 		return err;
3095 
3096 	if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3097 		return -EOVERFLOW;
3098 
3099 	err = f2fs_dquot_initialize(inode);
3100 	if (err)
3101 		return err;
3102 
3103 	f2fs_lock_op(sbi);
3104 	err = f2fs_transfer_project_quota(inode, kprojid);
3105 	if (err)
3106 		goto out_unlock;
3107 
3108 	F2FS_I(inode)->i_projid = kprojid;
3109 	inode->i_ctime = current_time(inode);
3110 	f2fs_mark_inode_dirty_sync(inode, true);
3111 out_unlock:
3112 	f2fs_unlock_op(sbi);
3113 	return err;
3114 }
3115 #else
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3116 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3117 {
3118 	return 0;
3119 }
3120 
f2fs_ioc_setproject(struct file * filp,__u32 projid)3121 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3122 {
3123 	if (projid != F2FS_DEF_PROJID)
3124 		return -EOPNOTSUPP;
3125 	return 0;
3126 }
3127 #endif
3128 
3129 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
3130 
3131 /*
3132  * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
3133  * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
3134  * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
3135  */
3136 
3137 static const struct {
3138 	u32 iflag;
3139 	u32 xflag;
3140 } f2fs_xflags_map[] = {
3141 	{ F2FS_SYNC_FL,		FS_XFLAG_SYNC },
3142 	{ F2FS_IMMUTABLE_FL,	FS_XFLAG_IMMUTABLE },
3143 	{ F2FS_APPEND_FL,	FS_XFLAG_APPEND },
3144 	{ F2FS_NODUMP_FL,	FS_XFLAG_NODUMP },
3145 	{ F2FS_NOATIME_FL,	FS_XFLAG_NOATIME },
3146 	{ F2FS_PROJINHERIT_FL,	FS_XFLAG_PROJINHERIT },
3147 };
3148 
3149 #define F2FS_SUPPORTED_XFLAGS (		\
3150 		FS_XFLAG_SYNC |		\
3151 		FS_XFLAG_IMMUTABLE |	\
3152 		FS_XFLAG_APPEND |	\
3153 		FS_XFLAG_NODUMP |	\
3154 		FS_XFLAG_NOATIME |	\
3155 		FS_XFLAG_PROJINHERIT)
3156 
3157 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
f2fs_iflags_to_xflags(u32 iflags)3158 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
3159 {
3160 	u32 xflags = 0;
3161 	int i;
3162 
3163 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3164 		if (iflags & f2fs_xflags_map[i].iflag)
3165 			xflags |= f2fs_xflags_map[i].xflag;
3166 
3167 	return xflags;
3168 }
3169 
3170 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
f2fs_xflags_to_iflags(u32 xflags)3171 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
3172 {
3173 	u32 iflags = 0;
3174 	int i;
3175 
3176 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3177 		if (xflags & f2fs_xflags_map[i].xflag)
3178 			iflags |= f2fs_xflags_map[i].iflag;
3179 
3180 	return iflags;
3181 }
3182 
f2fs_fill_fsxattr(struct inode * inode,struct fsxattr * fa)3183 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
3184 {
3185 	struct f2fs_inode_info *fi = F2FS_I(inode);
3186 
3187 	simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
3188 
3189 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3190 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3191 }
3192 
f2fs_ioc_fsgetxattr(struct file * filp,unsigned long arg)3193 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
3194 {
3195 	struct inode *inode = file_inode(filp);
3196 	struct fsxattr fa;
3197 
3198 	f2fs_fill_fsxattr(inode, &fa);
3199 
3200 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
3201 		return -EFAULT;
3202 	return 0;
3203 }
3204 
f2fs_ioc_fssetxattr(struct file * filp,unsigned long arg)3205 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
3206 {
3207 	struct inode *inode = file_inode(filp);
3208 	struct fsxattr fa, old_fa;
3209 	u32 iflags;
3210 	int err;
3211 
3212 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
3213 		return -EFAULT;
3214 
3215 	/* Make sure caller has proper permission */
3216 	if (!inode_owner_or_capable(inode))
3217 		return -EACCES;
3218 
3219 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
3220 		return -EOPNOTSUPP;
3221 
3222 	iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
3223 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3224 		return -EOPNOTSUPP;
3225 
3226 	err = mnt_want_write_file(filp);
3227 	if (err)
3228 		return err;
3229 
3230 	inode_lock(inode);
3231 
3232 	f2fs_fill_fsxattr(inode, &old_fa);
3233 	err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
3234 	if (err)
3235 		goto out;
3236 
3237 	err = f2fs_setflags_common(inode, iflags,
3238 			f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
3239 	if (err)
3240 		goto out;
3241 
3242 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3243 out:
3244 	inode_unlock(inode);
3245 	mnt_drop_write_file(filp);
3246 	return err;
3247 }
3248 
f2fs_pin_file_control(struct inode * inode,bool inc)3249 int f2fs_pin_file_control(struct inode *inode, bool inc)
3250 {
3251 	struct f2fs_inode_info *fi = F2FS_I(inode);
3252 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3253 
3254 	/* Use i_gc_failures for normal file as a risk signal. */
3255 	if (inc)
3256 		f2fs_i_gc_failures_write(inode,
3257 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3258 
3259 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3260 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3261 			  __func__, inode->i_ino,
3262 			  fi->i_gc_failures[GC_FAILURE_PIN]);
3263 		clear_inode_flag(inode, FI_PIN_FILE);
3264 		return -EAGAIN;
3265 	}
3266 	return 0;
3267 }
3268 
f2fs_ioc_set_pin_file(struct file * filp,unsigned long arg)3269 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3270 {
3271 	struct inode *inode = file_inode(filp);
3272 	__u32 pin;
3273 	int ret = 0;
3274 
3275 	if (get_user(pin, (__u32 __user *)arg))
3276 		return -EFAULT;
3277 
3278 	if (!S_ISREG(inode->i_mode))
3279 		return -EINVAL;
3280 
3281 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3282 		return -EROFS;
3283 
3284 	ret = mnt_want_write_file(filp);
3285 	if (ret)
3286 		return ret;
3287 
3288 	inode_lock(inode);
3289 
3290 	if (!pin) {
3291 		clear_inode_flag(inode, FI_PIN_FILE);
3292 		f2fs_i_gc_failures_write(inode, 0);
3293 		goto done;
3294 	}
3295 
3296 	if (f2fs_should_update_outplace(inode, NULL)) {
3297 		ret = -EINVAL;
3298 		goto out;
3299 	}
3300 
3301 	if (f2fs_pin_file_control(inode, false)) {
3302 		ret = -EAGAIN;
3303 		goto out;
3304 	}
3305 
3306 	ret = f2fs_convert_inline_inode(inode);
3307 	if (ret)
3308 		goto out;
3309 
3310 	if (!f2fs_disable_compressed_file(inode)) {
3311 		ret = -EOPNOTSUPP;
3312 		goto out;
3313 	}
3314 
3315 	set_inode_flag(inode, FI_PIN_FILE);
3316 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3317 done:
3318 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3319 out:
3320 	inode_unlock(inode);
3321 	mnt_drop_write_file(filp);
3322 	return ret;
3323 }
3324 
f2fs_ioc_get_pin_file(struct file * filp,unsigned long arg)3325 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3326 {
3327 	struct inode *inode = file_inode(filp);
3328 	__u32 pin = 0;
3329 
3330 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3331 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3332 	return put_user(pin, (u32 __user *)arg);
3333 }
3334 
f2fs_precache_extents(struct inode * inode)3335 int f2fs_precache_extents(struct inode *inode)
3336 {
3337 	struct f2fs_inode_info *fi = F2FS_I(inode);
3338 	struct f2fs_map_blocks map;
3339 	pgoff_t m_next_extent;
3340 	loff_t end;
3341 	int err;
3342 
3343 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3344 		return -EOPNOTSUPP;
3345 
3346 	map.m_lblk = 0;
3347 	map.m_pblk = 0;
3348 	map.m_next_pgofs = NULL;
3349 	map.m_next_extent = &m_next_extent;
3350 	map.m_seg_type = NO_CHECK_TYPE;
3351 	map.m_may_create = false;
3352 	end = max_file_blocks(inode);
3353 
3354 	while (map.m_lblk < end) {
3355 		map.m_len = end - map.m_lblk;
3356 
3357 		f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3358 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3359 		f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3360 		if (err)
3361 			return err;
3362 
3363 		map.m_lblk = m_next_extent;
3364 	}
3365 
3366 	return 0;
3367 }
3368 
f2fs_ioc_precache_extents(struct file * filp,unsigned long arg)3369 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3370 {
3371 	return f2fs_precache_extents(file_inode(filp));
3372 }
3373 
f2fs_ioc_resize_fs(struct file * filp,unsigned long arg)3374 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3375 {
3376 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3377 	__u64 block_count;
3378 
3379 	if (!capable(CAP_SYS_ADMIN))
3380 		return -EPERM;
3381 
3382 	if (f2fs_readonly(sbi->sb))
3383 		return -EROFS;
3384 
3385 	if (copy_from_user(&block_count, (void __user *)arg,
3386 			   sizeof(block_count)))
3387 		return -EFAULT;
3388 
3389 	return f2fs_resize_fs(filp, block_count);
3390 }
3391 
f2fs_ioc_enable_verity(struct file * filp,unsigned long arg)3392 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3393 {
3394 	struct inode *inode = file_inode(filp);
3395 
3396 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3397 
3398 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3399 		f2fs_warn(F2FS_I_SB(inode),
3400 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3401 			  inode->i_ino);
3402 		return -EOPNOTSUPP;
3403 	}
3404 
3405 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3406 }
3407 
f2fs_ioc_measure_verity(struct file * filp,unsigned long arg)3408 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3409 {
3410 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3411 		return -EOPNOTSUPP;
3412 
3413 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3414 }
3415 
f2fs_ioc_read_verity_metadata(struct file * filp,unsigned long arg)3416 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3417 {
3418 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3419 		return -EOPNOTSUPP;
3420 
3421 	return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3422 }
3423 
f2fs_ioc_getfslabel(struct file * filp,unsigned long arg)3424 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3425 {
3426 	struct inode *inode = file_inode(filp);
3427 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3428 	char *vbuf;
3429 	int count;
3430 	int err = 0;
3431 
3432 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3433 	if (!vbuf)
3434 		return -ENOMEM;
3435 
3436 	f2fs_down_read(&sbi->sb_lock);
3437 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3438 			ARRAY_SIZE(sbi->raw_super->volume_name),
3439 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3440 	f2fs_up_read(&sbi->sb_lock);
3441 
3442 	if (copy_to_user((char __user *)arg, vbuf,
3443 				min(FSLABEL_MAX, count)))
3444 		err = -EFAULT;
3445 
3446 	kfree(vbuf);
3447 	return err;
3448 }
3449 
f2fs_ioc_setfslabel(struct file * filp,unsigned long arg)3450 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3451 {
3452 	struct inode *inode = file_inode(filp);
3453 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3454 	char *vbuf;
3455 	int err = 0;
3456 
3457 	if (!capable(CAP_SYS_ADMIN))
3458 		return -EPERM;
3459 
3460 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3461 	if (IS_ERR(vbuf))
3462 		return PTR_ERR(vbuf);
3463 
3464 	err = mnt_want_write_file(filp);
3465 	if (err)
3466 		goto out;
3467 
3468 	f2fs_down_write(&sbi->sb_lock);
3469 
3470 	memset(sbi->raw_super->volume_name, 0,
3471 			sizeof(sbi->raw_super->volume_name));
3472 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3473 			sbi->raw_super->volume_name,
3474 			ARRAY_SIZE(sbi->raw_super->volume_name));
3475 
3476 	err = f2fs_commit_super(sbi, false);
3477 
3478 	f2fs_up_write(&sbi->sb_lock);
3479 
3480 	mnt_drop_write_file(filp);
3481 out:
3482 	kfree(vbuf);
3483 	return err;
3484 }
3485 
f2fs_get_compress_blocks(struct file * filp,unsigned long arg)3486 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3487 {
3488 	struct inode *inode = file_inode(filp);
3489 	__u64 blocks;
3490 
3491 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3492 		return -EOPNOTSUPP;
3493 
3494 	if (!f2fs_compressed_file(inode))
3495 		return -EINVAL;
3496 
3497 	blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3498 	return put_user(blocks, (u64 __user *)arg);
3499 }
3500 
release_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3501 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3502 {
3503 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3504 	unsigned int released_blocks = 0;
3505 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3506 	block_t blkaddr;
3507 	int i;
3508 
3509 	for (i = 0; i < count; i++) {
3510 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3511 						dn->ofs_in_node + i);
3512 
3513 		if (!__is_valid_data_blkaddr(blkaddr))
3514 			continue;
3515 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3516 					DATA_GENERIC_ENHANCE)))
3517 			return -EFSCORRUPTED;
3518 	}
3519 
3520 	while (count) {
3521 		int compr_blocks = 0;
3522 
3523 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3524 			blkaddr = f2fs_data_blkaddr(dn);
3525 
3526 			if (i == 0) {
3527 				if (blkaddr == COMPRESS_ADDR)
3528 					continue;
3529 				dn->ofs_in_node += cluster_size;
3530 				goto next;
3531 			}
3532 
3533 			if (__is_valid_data_blkaddr(blkaddr))
3534 				compr_blocks++;
3535 
3536 			if (blkaddr != NEW_ADDR)
3537 				continue;
3538 
3539 			dn->data_blkaddr = NULL_ADDR;
3540 			f2fs_set_data_blkaddr(dn);
3541 		}
3542 
3543 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3544 		dec_valid_block_count(sbi, dn->inode,
3545 					cluster_size - compr_blocks);
3546 
3547 		released_blocks += cluster_size - compr_blocks;
3548 next:
3549 		count -= cluster_size;
3550 	}
3551 
3552 	return released_blocks;
3553 }
3554 
f2fs_release_compress_blocks(struct file * filp,unsigned long arg)3555 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3556 {
3557 	struct inode *inode = file_inode(filp);
3558 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3559 	pgoff_t page_idx = 0, last_idx;
3560 	unsigned int released_blocks = 0;
3561 	int ret;
3562 	int writecount;
3563 
3564 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3565 		return -EOPNOTSUPP;
3566 
3567 	if (!f2fs_compressed_file(inode))
3568 		return -EINVAL;
3569 
3570 	if (f2fs_readonly(sbi->sb))
3571 		return -EROFS;
3572 
3573 	ret = mnt_want_write_file(filp);
3574 	if (ret)
3575 		return ret;
3576 
3577 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3578 
3579 	inode_lock(inode);
3580 
3581 	writecount = atomic_read(&inode->i_writecount);
3582 	if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3583 			(!(filp->f_mode & FMODE_WRITE) && writecount)) {
3584 		ret = -EBUSY;
3585 		goto out;
3586 	}
3587 
3588 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3589 		ret = -EINVAL;
3590 		goto out;
3591 	}
3592 
3593 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3594 	if (ret)
3595 		goto out;
3596 
3597 	set_inode_flag(inode, FI_COMPRESS_RELEASED);
3598 	inode->i_ctime = current_time(inode);
3599 	f2fs_mark_inode_dirty_sync(inode, true);
3600 
3601 	if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3602 		goto out;
3603 
3604 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3605 	f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
3606 
3607 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3608 
3609 	while (page_idx < last_idx) {
3610 		struct dnode_of_data dn;
3611 		pgoff_t end_offset, count;
3612 
3613 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3614 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3615 		if (ret) {
3616 			if (ret == -ENOENT) {
3617 				page_idx = f2fs_get_next_page_offset(&dn,
3618 								page_idx);
3619 				ret = 0;
3620 				continue;
3621 			}
3622 			break;
3623 		}
3624 
3625 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3626 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3627 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3628 
3629 		ret = release_compress_blocks(&dn, count);
3630 
3631 		f2fs_put_dnode(&dn);
3632 
3633 		if (ret < 0)
3634 			break;
3635 
3636 		page_idx += count;
3637 		released_blocks += ret;
3638 	}
3639 
3640 	f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
3641 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3642 out:
3643 	inode_unlock(inode);
3644 
3645 	mnt_drop_write_file(filp);
3646 
3647 	if (ret >= 0) {
3648 		ret = put_user(released_blocks, (u64 __user *)arg);
3649 	} else if (released_blocks &&
3650 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3651 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3652 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3653 			"iblocks=%llu, released=%u, compr_blocks=%u, "
3654 			"run fsck to fix.",
3655 			__func__, inode->i_ino, inode->i_blocks,
3656 			released_blocks,
3657 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
3658 	}
3659 
3660 	return ret;
3661 }
3662 
reserve_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3663 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3664 {
3665 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3666 	unsigned int reserved_blocks = 0;
3667 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3668 	block_t blkaddr;
3669 	int i;
3670 
3671 	for (i = 0; i < count; i++) {
3672 		blkaddr = data_blkaddr(dn->inode, dn->node_page,
3673 						dn->ofs_in_node + i);
3674 
3675 		if (!__is_valid_data_blkaddr(blkaddr))
3676 			continue;
3677 		if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3678 					DATA_GENERIC_ENHANCE)))
3679 			return -EFSCORRUPTED;
3680 	}
3681 
3682 	while (count) {
3683 		int compr_blocks = 0;
3684 		blkcnt_t reserved;
3685 		int ret;
3686 
3687 		for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3688 			blkaddr = f2fs_data_blkaddr(dn);
3689 
3690 			if (i == 0) {
3691 				if (blkaddr == COMPRESS_ADDR)
3692 					continue;
3693 				dn->ofs_in_node += cluster_size;
3694 				goto next;
3695 			}
3696 
3697 			if (__is_valid_data_blkaddr(blkaddr)) {
3698 				compr_blocks++;
3699 				continue;
3700 			}
3701 
3702 			dn->data_blkaddr = NEW_ADDR;
3703 			f2fs_set_data_blkaddr(dn);
3704 		}
3705 
3706 		reserved = cluster_size - compr_blocks;
3707 		ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3708 		if (ret)
3709 			return ret;
3710 
3711 		if (reserved != cluster_size - compr_blocks)
3712 			return -ENOSPC;
3713 
3714 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3715 
3716 		reserved_blocks += reserved;
3717 next:
3718 		count -= cluster_size;
3719 	}
3720 
3721 	return reserved_blocks;
3722 }
3723 
f2fs_reserve_compress_blocks(struct file * filp,unsigned long arg)3724 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3725 {
3726 	struct inode *inode = file_inode(filp);
3727 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3728 	pgoff_t page_idx = 0, last_idx;
3729 	unsigned int reserved_blocks = 0;
3730 	int ret;
3731 
3732 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3733 		return -EOPNOTSUPP;
3734 
3735 	if (!f2fs_compressed_file(inode))
3736 		return -EINVAL;
3737 
3738 	if (f2fs_readonly(sbi->sb))
3739 		return -EROFS;
3740 
3741 	ret = mnt_want_write_file(filp);
3742 	if (ret)
3743 		return ret;
3744 
3745 	if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3746 		goto out;
3747 
3748 	f2fs_balance_fs(F2FS_I_SB(inode), true);
3749 
3750 	inode_lock(inode);
3751 
3752 	if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3753 		ret = -EINVAL;
3754 		goto unlock_inode;
3755 	}
3756 
3757 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3758 	f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
3759 
3760 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3761 
3762 	while (page_idx < last_idx) {
3763 		struct dnode_of_data dn;
3764 		pgoff_t end_offset, count;
3765 
3766 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3767 		ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3768 		if (ret) {
3769 			if (ret == -ENOENT) {
3770 				page_idx = f2fs_get_next_page_offset(&dn,
3771 								page_idx);
3772 				ret = 0;
3773 				continue;
3774 			}
3775 			break;
3776 		}
3777 
3778 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3779 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3780 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
3781 
3782 		ret = reserve_compress_blocks(&dn, count);
3783 
3784 		f2fs_put_dnode(&dn);
3785 
3786 		if (ret < 0)
3787 			break;
3788 
3789 		page_idx += count;
3790 		reserved_blocks += ret;
3791 	}
3792 
3793 	f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
3794 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3795 
3796 	if (ret >= 0) {
3797 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3798 		inode->i_ctime = current_time(inode);
3799 		f2fs_mark_inode_dirty_sync(inode, true);
3800 	}
3801 unlock_inode:
3802 	inode_unlock(inode);
3803 out:
3804 	mnt_drop_write_file(filp);
3805 
3806 	if (ret >= 0) {
3807 		ret = put_user(reserved_blocks, (u64 __user *)arg);
3808 	} else if (reserved_blocks &&
3809 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3810 		set_sbi_flag(sbi, SBI_NEED_FSCK);
3811 		f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3812 			"iblocks=%llu, reserved=%u, compr_blocks=%u, "
3813 			"run fsck to fix.",
3814 			__func__, inode->i_ino, inode->i_blocks,
3815 			reserved_blocks,
3816 			atomic_read(&F2FS_I(inode)->i_compr_blocks));
3817 	}
3818 
3819 	return ret;
3820 }
3821 
f2fs_secure_erase(struct block_device * bdev,struct inode * inode,pgoff_t off,block_t block,block_t len,u32 flags)3822 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3823 		pgoff_t off, block_t block, block_t len, u32 flags)
3824 {
3825 	struct request_queue *q = bdev_get_queue(bdev);
3826 	sector_t sector = SECTOR_FROM_BLOCK(block);
3827 	sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3828 	int ret = 0;
3829 
3830 	if (!q)
3831 		return -ENXIO;
3832 
3833 	if (flags & F2FS_TRIM_FILE_DISCARD)
3834 		ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3835 						blk_queue_secure_erase(q) ?
3836 						BLKDEV_DISCARD_SECURE : 0);
3837 
3838 	if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3839 		if (IS_ENCRYPTED(inode))
3840 			ret = fscrypt_zeroout_range(inode, off, block, len);
3841 		else
3842 			ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3843 					GFP_NOFS, 0);
3844 	}
3845 
3846 	return ret;
3847 }
3848 
f2fs_sec_trim_file(struct file * filp,unsigned long arg)3849 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3850 {
3851 	struct inode *inode = file_inode(filp);
3852 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3853 	struct address_space *mapping = inode->i_mapping;
3854 	struct block_device *prev_bdev = NULL;
3855 	struct f2fs_sectrim_range range;
3856 	pgoff_t index, pg_end, prev_index = 0;
3857 	block_t prev_block = 0, len = 0;
3858 	loff_t end_addr;
3859 	bool to_end = false;
3860 	int ret = 0;
3861 
3862 	if (!(filp->f_mode & FMODE_WRITE))
3863 		return -EBADF;
3864 
3865 	if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3866 				sizeof(range)))
3867 		return -EFAULT;
3868 
3869 	if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3870 			!S_ISREG(inode->i_mode))
3871 		return -EINVAL;
3872 
3873 	if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3874 			!f2fs_hw_support_discard(sbi)) ||
3875 			((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3876 			 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3877 		return -EOPNOTSUPP;
3878 
3879 	file_start_write(filp);
3880 	inode_lock(inode);
3881 
3882 	if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3883 			range.start >= inode->i_size) {
3884 		ret = -EINVAL;
3885 		goto err;
3886 	}
3887 
3888 	if (range.len == 0)
3889 		goto err;
3890 
3891 	if (inode->i_size - range.start > range.len) {
3892 		end_addr = range.start + range.len;
3893 	} else {
3894 		end_addr = range.len == (u64)-1 ?
3895 			sbi->sb->s_maxbytes : inode->i_size;
3896 		to_end = true;
3897 	}
3898 
3899 	if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3900 			(!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3901 		ret = -EINVAL;
3902 		goto err;
3903 	}
3904 
3905 	index = F2FS_BYTES_TO_BLK(range.start);
3906 	pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3907 
3908 	ret = f2fs_convert_inline_inode(inode);
3909 	if (ret)
3910 		goto err;
3911 
3912 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3913 	f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
3914 
3915 	ret = filemap_write_and_wait_range(mapping, range.start,
3916 			to_end ? LLONG_MAX : end_addr - 1);
3917 	if (ret)
3918 		goto out;
3919 
3920 	truncate_inode_pages_range(mapping, range.start,
3921 			to_end ? -1 : end_addr - 1);
3922 
3923 	while (index < pg_end) {
3924 		struct dnode_of_data dn;
3925 		pgoff_t end_offset, count;
3926 		int i;
3927 
3928 		set_new_dnode(&dn, inode, NULL, NULL, 0);
3929 		ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3930 		if (ret) {
3931 			if (ret == -ENOENT) {
3932 				index = f2fs_get_next_page_offset(&dn, index);
3933 				continue;
3934 			}
3935 			goto out;
3936 		}
3937 
3938 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3939 		count = min(end_offset - dn.ofs_in_node, pg_end - index);
3940 		for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3941 			struct block_device *cur_bdev;
3942 			block_t blkaddr = f2fs_data_blkaddr(&dn);
3943 
3944 			if (!__is_valid_data_blkaddr(blkaddr))
3945 				continue;
3946 
3947 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3948 						DATA_GENERIC_ENHANCE)) {
3949 				ret = -EFSCORRUPTED;
3950 				f2fs_put_dnode(&dn);
3951 				goto out;
3952 			}
3953 
3954 			cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3955 			if (f2fs_is_multi_device(sbi)) {
3956 				int di = f2fs_target_device_index(sbi, blkaddr);
3957 
3958 				blkaddr -= FDEV(di).start_blk;
3959 			}
3960 
3961 			if (len) {
3962 				if (prev_bdev == cur_bdev &&
3963 						index == prev_index + len &&
3964 						blkaddr == prev_block + len) {
3965 					len++;
3966 				} else {
3967 					ret = f2fs_secure_erase(prev_bdev,
3968 						inode, prev_index, prev_block,
3969 						len, range.flags);
3970 					if (ret) {
3971 						f2fs_put_dnode(&dn);
3972 						goto out;
3973 					}
3974 
3975 					len = 0;
3976 				}
3977 			}
3978 
3979 			if (!len) {
3980 				prev_bdev = cur_bdev;
3981 				prev_index = index;
3982 				prev_block = blkaddr;
3983 				len = 1;
3984 			}
3985 		}
3986 
3987 		f2fs_put_dnode(&dn);
3988 
3989 		if (fatal_signal_pending(current)) {
3990 			ret = -EINTR;
3991 			goto out;
3992 		}
3993 		cond_resched();
3994 	}
3995 
3996 	if (len)
3997 		ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3998 				prev_block, len, range.flags);
3999 out:
4000 	f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
4001 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4002 err:
4003 	inode_unlock(inode);
4004 	file_end_write(filp);
4005 
4006 	return ret;
4007 }
4008 
f2fs_ioc_get_compress_option(struct file * filp,unsigned long arg)4009 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
4010 {
4011 	struct inode *inode = file_inode(filp);
4012 	struct f2fs_comp_option option;
4013 
4014 	if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4015 		return -EOPNOTSUPP;
4016 
4017 	inode_lock_shared(inode);
4018 
4019 	if (!f2fs_compressed_file(inode)) {
4020 		inode_unlock_shared(inode);
4021 		return -ENODATA;
4022 	}
4023 
4024 	option.algorithm = F2FS_I(inode)->i_compress_algorithm;
4025 	option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
4026 
4027 	inode_unlock_shared(inode);
4028 
4029 	if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
4030 				sizeof(option)))
4031 		return -EFAULT;
4032 
4033 	return 0;
4034 }
4035 
f2fs_ioc_set_compress_option(struct file * filp,unsigned long arg)4036 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4037 {
4038 	struct inode *inode = file_inode(filp);
4039 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4040 	struct f2fs_comp_option option;
4041 	int ret = 0;
4042 
4043 	if (!f2fs_sb_has_compression(sbi))
4044 		return -EOPNOTSUPP;
4045 
4046 	if (!(filp->f_mode & FMODE_WRITE))
4047 		return -EBADF;
4048 
4049 	if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4050 				sizeof(option)))
4051 		return -EFAULT;
4052 
4053 	if (!f2fs_compressed_file(inode) ||
4054 			option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
4055 			option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
4056 			option.algorithm >= COMPRESS_MAX)
4057 		return -EINVAL;
4058 
4059 	file_start_write(filp);
4060 	inode_lock(inode);
4061 
4062 	if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4063 		ret = -EBUSY;
4064 		goto out;
4065 	}
4066 
4067 	if (inode->i_size != 0) {
4068 		ret = -EFBIG;
4069 		goto out;
4070 	}
4071 
4072 	F2FS_I(inode)->i_compress_algorithm = option.algorithm;
4073 	F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
4074 	F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
4075 	f2fs_mark_inode_dirty_sync(inode, true);
4076 
4077 	if (!f2fs_is_compress_backend_ready(inode))
4078 		f2fs_warn(sbi, "compression algorithm is successfully set, "
4079 			"but current kernel doesn't support this algorithm.");
4080 out:
4081 	inode_unlock(inode);
4082 	file_end_write(filp);
4083 
4084 	return ret;
4085 }
4086 
redirty_blocks(struct inode * inode,pgoff_t page_idx,int len)4087 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4088 {
4089 	DEFINE_READAHEAD(ractl, NULL, inode->i_mapping, page_idx);
4090 	struct address_space *mapping = inode->i_mapping;
4091 	struct page *page;
4092 	pgoff_t redirty_idx = page_idx;
4093 	int i, page_len = 0, ret = 0;
4094 
4095 	page_cache_ra_unbounded(&ractl, len, 0);
4096 
4097 	for (i = 0; i < len; i++, page_idx++) {
4098 		page = read_cache_page(mapping, page_idx, NULL, NULL);
4099 		if (IS_ERR(page)) {
4100 			ret = PTR_ERR(page);
4101 			break;
4102 		}
4103 		page_len++;
4104 	}
4105 
4106 	for (i = 0; i < page_len; i++, redirty_idx++) {
4107 		page = find_lock_page(mapping, redirty_idx);
4108 		if (!page) {
4109 			ret = -ENOMEM;
4110 			break;
4111 		}
4112 		set_page_dirty(page);
4113 		f2fs_put_page(page, 1);
4114 		f2fs_put_page(page, 0);
4115 	}
4116 
4117 	return ret;
4118 }
4119 
f2fs_ioc_decompress_file(struct file * filp,unsigned long arg)4120 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
4121 {
4122 	struct inode *inode = file_inode(filp);
4123 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4124 	struct f2fs_inode_info *fi = F2FS_I(inode);
4125 	pgoff_t page_idx = 0, last_idx;
4126 	unsigned int blk_per_seg = sbi->blocks_per_seg;
4127 	int cluster_size = F2FS_I(inode)->i_cluster_size;
4128 	int count, ret;
4129 
4130 	if (!f2fs_sb_has_compression(sbi) ||
4131 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4132 		return -EOPNOTSUPP;
4133 
4134 	if (!(filp->f_mode & FMODE_WRITE))
4135 		return -EBADF;
4136 
4137 	if (!f2fs_compressed_file(inode))
4138 		return -EINVAL;
4139 
4140 	f2fs_balance_fs(F2FS_I_SB(inode), true);
4141 
4142 	file_start_write(filp);
4143 	inode_lock(inode);
4144 
4145 	if (!f2fs_is_compress_backend_ready(inode)) {
4146 		ret = -EOPNOTSUPP;
4147 		goto out;
4148 	}
4149 
4150 	if (f2fs_is_mmap_file(inode)) {
4151 		ret = -EBUSY;
4152 		goto out;
4153 	}
4154 
4155 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4156 	if (ret)
4157 		goto out;
4158 
4159 	if (!atomic_read(&fi->i_compr_blocks))
4160 		goto out;
4161 
4162 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4163 
4164 	count = last_idx - page_idx;
4165 	while (count) {
4166 		int len = min(cluster_size, count);
4167 
4168 		ret = redirty_blocks(inode, page_idx, len);
4169 		if (ret < 0)
4170 			break;
4171 
4172 		if (get_dirty_pages(inode) >= blk_per_seg)
4173 			filemap_fdatawrite(inode->i_mapping);
4174 
4175 		count -= len;
4176 		page_idx += len;
4177 	}
4178 
4179 	if (!ret)
4180 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4181 							LLONG_MAX);
4182 
4183 	if (ret)
4184 		f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4185 			  __func__, ret);
4186 out:
4187 	inode_unlock(inode);
4188 	file_end_write(filp);
4189 
4190 	return ret;
4191 }
4192 
f2fs_ioc_compress_file(struct file * filp,unsigned long arg)4193 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4194 {
4195 	struct inode *inode = file_inode(filp);
4196 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4197 	pgoff_t page_idx = 0, last_idx;
4198 	unsigned int blk_per_seg = sbi->blocks_per_seg;
4199 	int cluster_size = F2FS_I(inode)->i_cluster_size;
4200 	int count, ret;
4201 
4202 	if (!f2fs_sb_has_compression(sbi) ||
4203 			F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4204 		return -EOPNOTSUPP;
4205 
4206 	if (!(filp->f_mode & FMODE_WRITE))
4207 		return -EBADF;
4208 
4209 	if (!f2fs_compressed_file(inode))
4210 		return -EINVAL;
4211 
4212 	f2fs_balance_fs(F2FS_I_SB(inode), true);
4213 
4214 	file_start_write(filp);
4215 	inode_lock(inode);
4216 
4217 	if (!f2fs_is_compress_backend_ready(inode)) {
4218 		ret = -EOPNOTSUPP;
4219 		goto out;
4220 	}
4221 
4222 	if (f2fs_is_mmap_file(inode)) {
4223 		ret = -EBUSY;
4224 		goto out;
4225 	}
4226 
4227 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4228 	if (ret)
4229 		goto out;
4230 
4231 	set_inode_flag(inode, FI_ENABLE_COMPRESS);
4232 
4233 	last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4234 
4235 	count = last_idx - page_idx;
4236 	while (count) {
4237 		int len = min(cluster_size, count);
4238 
4239 		ret = redirty_blocks(inode, page_idx, len);
4240 		if (ret < 0)
4241 			break;
4242 
4243 		if (get_dirty_pages(inode) >= blk_per_seg)
4244 			filemap_fdatawrite(inode->i_mapping);
4245 
4246 		count -= len;
4247 		page_idx += len;
4248 	}
4249 
4250 	if (!ret)
4251 		ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4252 							LLONG_MAX);
4253 
4254 	clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4255 
4256 	if (ret)
4257 		f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4258 			  __func__, ret);
4259 out:
4260 	inode_unlock(inode);
4261 	file_end_write(filp);
4262 
4263 	return ret;
4264 }
4265 
__f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4266 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4267 {
4268 	switch (cmd) {
4269 	case FS_IOC_GETFLAGS:
4270 		return f2fs_ioc_getflags(filp, arg);
4271 	case FS_IOC_SETFLAGS:
4272 		return f2fs_ioc_setflags(filp, arg);
4273 	case FS_IOC_GETVERSION:
4274 		return f2fs_ioc_getversion(filp, arg);
4275 	case F2FS_IOC_START_ATOMIC_WRITE:
4276 		return f2fs_ioc_start_atomic_write(filp);
4277 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4278 		return f2fs_ioc_commit_atomic_write(filp);
4279 	case F2FS_IOC_START_VOLATILE_WRITE:
4280 		return f2fs_ioc_start_volatile_write(filp);
4281 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4282 		return f2fs_ioc_release_volatile_write(filp);
4283 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
4284 		return f2fs_ioc_abort_volatile_write(filp);
4285 	case F2FS_IOC_SHUTDOWN:
4286 		return f2fs_ioc_shutdown(filp, arg);
4287 	case FITRIM:
4288 		return f2fs_ioc_fitrim(filp, arg);
4289 	case FS_IOC_SET_ENCRYPTION_POLICY:
4290 		return f2fs_ioc_set_encryption_policy(filp, arg);
4291 	case FS_IOC_GET_ENCRYPTION_POLICY:
4292 		return f2fs_ioc_get_encryption_policy(filp, arg);
4293 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4294 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4295 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4296 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4297 	case FS_IOC_ADD_ENCRYPTION_KEY:
4298 		return f2fs_ioc_add_encryption_key(filp, arg);
4299 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4300 		return f2fs_ioc_remove_encryption_key(filp, arg);
4301 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4302 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4303 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4304 		return f2fs_ioc_get_encryption_key_status(filp, arg);
4305 	case FS_IOC_GET_ENCRYPTION_NONCE:
4306 		return f2fs_ioc_get_encryption_nonce(filp, arg);
4307 	case F2FS_IOC_GARBAGE_COLLECT:
4308 		return f2fs_ioc_gc(filp, arg);
4309 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4310 		return f2fs_ioc_gc_range(filp, arg);
4311 	case F2FS_IOC_WRITE_CHECKPOINT:
4312 		return f2fs_ioc_write_checkpoint(filp, arg);
4313 	case F2FS_IOC_DEFRAGMENT:
4314 		return f2fs_ioc_defragment(filp, arg);
4315 	case F2FS_IOC_MOVE_RANGE:
4316 		return f2fs_ioc_move_range(filp, arg);
4317 	case F2FS_IOC_FLUSH_DEVICE:
4318 		return f2fs_ioc_flush_device(filp, arg);
4319 	case F2FS_IOC_GET_FEATURES:
4320 		return f2fs_ioc_get_features(filp, arg);
4321 	case FS_IOC_FSGETXATTR:
4322 		return f2fs_ioc_fsgetxattr(filp, arg);
4323 	case FS_IOC_FSSETXATTR:
4324 		return f2fs_ioc_fssetxattr(filp, arg);
4325 	case F2FS_IOC_GET_PIN_FILE:
4326 		return f2fs_ioc_get_pin_file(filp, arg);
4327 	case F2FS_IOC_SET_PIN_FILE:
4328 		return f2fs_ioc_set_pin_file(filp, arg);
4329 	case F2FS_IOC_PRECACHE_EXTENTS:
4330 		return f2fs_ioc_precache_extents(filp, arg);
4331 	case F2FS_IOC_RESIZE_FS:
4332 		return f2fs_ioc_resize_fs(filp, arg);
4333 	case FS_IOC_ENABLE_VERITY:
4334 		return f2fs_ioc_enable_verity(filp, arg);
4335 	case FS_IOC_MEASURE_VERITY:
4336 		return f2fs_ioc_measure_verity(filp, arg);
4337 	case FS_IOC_READ_VERITY_METADATA:
4338 		return f2fs_ioc_read_verity_metadata(filp, arg);
4339 	case FS_IOC_GETFSLABEL:
4340 		return f2fs_ioc_getfslabel(filp, arg);
4341 	case FS_IOC_SETFSLABEL:
4342 		return f2fs_ioc_setfslabel(filp, arg);
4343 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4344 		return f2fs_get_compress_blocks(filp, arg);
4345 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4346 		return f2fs_release_compress_blocks(filp, arg);
4347 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4348 		return f2fs_reserve_compress_blocks(filp, arg);
4349 	case F2FS_IOC_SEC_TRIM_FILE:
4350 		return f2fs_sec_trim_file(filp, arg);
4351 	case F2FS_IOC_GET_COMPRESS_OPTION:
4352 		return f2fs_ioc_get_compress_option(filp, arg);
4353 	case F2FS_IOC_SET_COMPRESS_OPTION:
4354 		return f2fs_ioc_set_compress_option(filp, arg);
4355 	case F2FS_IOC_DECOMPRESS_FILE:
4356 		return f2fs_ioc_decompress_file(filp, arg);
4357 	case F2FS_IOC_COMPRESS_FILE:
4358 		return f2fs_ioc_compress_file(filp, arg);
4359 	default:
4360 		return -ENOTTY;
4361 	}
4362 }
4363 
f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4364 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4365 {
4366 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4367 		return -EIO;
4368 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4369 		return -ENOSPC;
4370 
4371 	return __f2fs_ioctl(filp, cmd, arg);
4372 }
4373 
4374 /*
4375  * Return %true if the given read or write request should use direct I/O, or
4376  * %false if it should use buffered I/O.
4377  */
f2fs_should_use_dio(struct inode * inode,struct kiocb * iocb,struct iov_iter * iter)4378 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4379 				struct iov_iter *iter)
4380 {
4381 	unsigned int align;
4382 
4383 	if (!(iocb->ki_flags & IOCB_DIRECT))
4384 		return false;
4385 
4386 	if (f2fs_force_buffered_io(inode, iocb, iter))
4387 		return false;
4388 
4389 	/*
4390 	 * Direct I/O not aligned to the disk's logical_block_size will be
4391 	 * attempted, but will fail with -EINVAL.
4392 	 *
4393 	 * f2fs additionally requires that direct I/O be aligned to the
4394 	 * filesystem block size, which is often a stricter requirement.
4395 	 * However, f2fs traditionally falls back to buffered I/O on requests
4396 	 * that are logical_block_size-aligned but not fs-block aligned.
4397 	 *
4398 	 * The below logic implements this behavior.
4399 	 */
4400 	align = iocb->ki_pos | iov_iter_alignment(iter);
4401 	if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4402 	    IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4403 		return false;
4404 
4405 	return true;
4406 }
4407 
f2fs_dio_read_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4408 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4409 				unsigned int flags)
4410 {
4411 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4412 
4413 	dec_page_count(sbi, F2FS_DIO_READ);
4414 	if (error)
4415 		return error;
4416 	f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size);
4417 	return 0;
4418 }
4419 
4420 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4421 	.end_io = f2fs_dio_read_end_io,
4422 };
4423 
f2fs_dio_read_iter(struct kiocb * iocb,struct iov_iter * to)4424 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4425 {
4426 	struct file *file = iocb->ki_filp;
4427 	struct inode *inode = file_inode(file);
4428 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4429 	struct f2fs_inode_info *fi = F2FS_I(inode);
4430 	const loff_t pos = iocb->ki_pos;
4431 	const size_t count = iov_iter_count(to);
4432 	struct iomap_dio *dio;
4433 	ssize_t ret;
4434 
4435 	if (count == 0)
4436 		return 0; /* skip atime update */
4437 
4438 	trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4439 	if (trace_android_fs_dataread_start_enabled()) {
4440 		char *path, pathbuf[MAX_TRACE_PATHBUF_LEN];
4441 
4442 		path = android_fstrace_get_pathname(pathbuf,
4443 						    MAX_TRACE_PATHBUF_LEN,
4444 						    inode);
4445 		trace_android_fs_dataread_start(inode, pos,
4446 						count, current->pid, path,
4447 						current->comm);
4448 	}
4449 
4450 	if (iocb->ki_flags & IOCB_NOWAIT) {
4451 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4452 			ret = -EAGAIN;
4453 			goto out;
4454 		}
4455 	} else {
4456 		f2fs_down_read(&fi->i_gc_rwsem[READ]);
4457 	}
4458 
4459 	/*
4460 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4461 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4462 	 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4463 	 */
4464 	inc_page_count(sbi, F2FS_DIO_READ);
4465 	dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4466 			     &f2fs_iomap_dio_read_ops, is_sync_kiocb(iocb));
4467 	if (IS_ERR_OR_NULL(dio)) {
4468 		ret = PTR_ERR_OR_ZERO(dio);
4469 		if (ret != -EIOCBQUEUED)
4470 			dec_page_count(sbi, F2FS_DIO_READ);
4471 	} else {
4472 		ret = iomap_dio_complete(dio);
4473 	}
4474 
4475 	f2fs_up_read(&fi->i_gc_rwsem[READ]);
4476 
4477 	file_accessed(file);
4478 out:
4479 	if (trace_android_fs_dataread_start_enabled())
4480 		trace_android_fs_dataread_end(inode, pos, count);
4481 	trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4482 	return ret;
4483 }
4484 
f2fs_file_read_iter(struct kiocb * iocb,struct iov_iter * to)4485 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4486 {
4487 	struct inode *inode = file_inode(iocb->ki_filp);
4488 	ssize_t ret = 0;
4489 
4490 	if (!f2fs_is_compress_backend_ready(inode))
4491 		return -EOPNOTSUPP;
4492 
4493 	if (f2fs_should_use_dio(inode, iocb, to))
4494 		return f2fs_dio_read_iter(iocb, to);
4495 
4496 	ret = generic_file_buffered_read(iocb, to, ret);
4497 	if (ret > 0)
4498 		f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret);
4499 	return ret;
4500 }
4501 
f2fs_write_checks(struct kiocb * iocb,struct iov_iter * from)4502 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4503 {
4504 	struct file *file = iocb->ki_filp;
4505 	struct inode *inode = file_inode(file);
4506 	ssize_t count;
4507 	int err;
4508 
4509 	if (IS_IMMUTABLE(inode))
4510 		return -EPERM;
4511 
4512 	if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4513 		return -EPERM;
4514 
4515 	count = generic_write_checks(iocb, from);
4516 	if (count <= 0)
4517 		return count;
4518 
4519 	err = file_modified(file);
4520 	if (err)
4521 		return err;
4522 	return count;
4523 }
4524 
4525 /*
4526  * Preallocate blocks for a write request, if it is possible and helpful to do
4527  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4528  * blocks were preallocated, or a negative errno value if something went
4529  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4530  * requested blocks (not just some of them) have been allocated.
4531  */
f2fs_preallocate_blocks(struct kiocb * iocb,struct iov_iter * iter,bool dio)4532 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4533 				   bool dio)
4534 {
4535 	struct inode *inode = file_inode(iocb->ki_filp);
4536 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4537 	const loff_t pos = iocb->ki_pos;
4538 	const size_t count = iov_iter_count(iter);
4539 	struct f2fs_map_blocks map = {};
4540 	int flag;
4541 	int ret;
4542 
4543 	/* If it will be an out-of-place direct write, don't bother. */
4544 	if (dio && f2fs_lfs_mode(sbi))
4545 		return 0;
4546 	/*
4547 	 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4548 	 * buffered IO, if DIO meets any holes.
4549 	 */
4550 	if (dio && i_size_read(inode) &&
4551 		(F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4552 		return 0;
4553 
4554 	/* No-wait I/O can't allocate blocks. */
4555 	if (iocb->ki_flags & IOCB_NOWAIT)
4556 		return 0;
4557 
4558 	/* If it will be a short write, don't bother. */
4559 	if (iov_iter_fault_in_readable(iter, count))
4560 		return 0;
4561 
4562 	if (f2fs_has_inline_data(inode)) {
4563 		/* If the data will fit inline, don't bother. */
4564 		if (pos + count <= MAX_INLINE_DATA(inode))
4565 			return 0;
4566 		ret = f2fs_convert_inline_inode(inode);
4567 		if (ret)
4568 			return ret;
4569 	}
4570 
4571 	/* Do not preallocate blocks that will be written partially in 4KB. */
4572 	map.m_lblk = F2FS_BLK_ALIGN(pos);
4573 	map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4574 	if (map.m_len > map.m_lblk)
4575 		map.m_len -= map.m_lblk;
4576 	else
4577 		map.m_len = 0;
4578 	map.m_may_create = true;
4579 	if (dio) {
4580 		map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4581 		flag = F2FS_GET_BLOCK_PRE_DIO;
4582 	} else {
4583 		map.m_seg_type = NO_CHECK_TYPE;
4584 		flag = F2FS_GET_BLOCK_PRE_AIO;
4585 	}
4586 
4587 	ret = f2fs_map_blocks(inode, &map, 1, flag);
4588 	/* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4589 	if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4590 		return ret;
4591 	if (ret == 0)
4592 		set_inode_flag(inode, FI_PREALLOCATED_ALL);
4593 	return map.m_len;
4594 }
4595 
f2fs_buffered_write_iter(struct kiocb * iocb,struct iov_iter * from)4596 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4597 					struct iov_iter *from)
4598 {
4599 	struct file *file = iocb->ki_filp;
4600 	struct inode *inode = file_inode(file);
4601 	ssize_t ret;
4602 
4603 	if (iocb->ki_flags & IOCB_NOWAIT)
4604 		return -EOPNOTSUPP;
4605 
4606 	current->backing_dev_info = inode_to_bdi(inode);
4607 	ret = generic_perform_write(file, from, iocb->ki_pos);
4608 	current->backing_dev_info = NULL;
4609 
4610 	if (ret > 0) {
4611 		iocb->ki_pos += ret;
4612 		f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret);
4613 	}
4614 	return ret;
4615 }
4616 
f2fs_dio_write_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4617 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4618 				 unsigned int flags)
4619 {
4620 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4621 
4622 	dec_page_count(sbi, F2FS_DIO_WRITE);
4623 	if (error)
4624 		return error;
4625 	f2fs_update_iostat(sbi, APP_DIRECT_IO, size);
4626 	return 0;
4627 }
4628 
4629 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4630 	.end_io = f2fs_dio_write_end_io,
4631 };
4632 
f2fs_dio_write_iter(struct kiocb * iocb,struct iov_iter * from,bool * may_need_sync)4633 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4634 				   bool *may_need_sync)
4635 {
4636 	struct file *file = iocb->ki_filp;
4637 	struct inode *inode = file_inode(file);
4638 	struct f2fs_inode_info *fi = F2FS_I(inode);
4639 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4640 	const bool do_opu = f2fs_lfs_mode(sbi);
4641 	const int whint_mode = F2FS_OPTION(sbi).whint_mode;
4642 	const loff_t pos = iocb->ki_pos;
4643 	const ssize_t count = iov_iter_count(from);
4644 	const enum rw_hint hint = iocb->ki_hint;
4645 	unsigned int dio_flags;
4646 	struct iomap_dio *dio;
4647 	ssize_t ret;
4648 
4649 	trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4650 	if (trace_android_fs_datawrite_start_enabled()) {
4651 		char *path, pathbuf[MAX_TRACE_PATHBUF_LEN];
4652 
4653 		path = android_fstrace_get_pathname(pathbuf,
4654 						    MAX_TRACE_PATHBUF_LEN,
4655 						    inode);
4656 		trace_android_fs_datawrite_start(inode, pos, count,
4657 						 current->pid, path,
4658 						 current->comm);
4659 	}
4660 
4661 	if (iocb->ki_flags & IOCB_NOWAIT) {
4662 		/* f2fs_convert_inline_inode() and block allocation can block */
4663 		if (f2fs_has_inline_data(inode) ||
4664 		    !f2fs_overwrite_io(inode, pos, count)) {
4665 			ret = -EAGAIN;
4666 			goto out;
4667 		}
4668 
4669 		if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4670 			ret = -EAGAIN;
4671 			goto out;
4672 		}
4673 		if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4674 			f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4675 			ret = -EAGAIN;
4676 			goto out;
4677 		}
4678 	} else {
4679 		ret = f2fs_convert_inline_inode(inode);
4680 		if (ret)
4681 			goto out;
4682 
4683 		f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4684 		if (do_opu)
4685 			f2fs_down_read(&fi->i_gc_rwsem[READ]);
4686 	}
4687 	if (whint_mode == WHINT_MODE_OFF)
4688 		iocb->ki_hint = WRITE_LIFE_NOT_SET;
4689 
4690 	/*
4691 	 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4692 	 * the higher-level function iomap_dio_rw() in order to ensure that the
4693 	 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4694 	 */
4695 	inc_page_count(sbi, F2FS_DIO_WRITE);
4696 	dio_flags = 0;
4697 	if (pos + count > inode->i_size)
4698 		dio_flags = 1;	/* IOMAP_DIO_FORCE_WAIT */
4699 	dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4700 			     &f2fs_iomap_dio_write_ops,
4701 			     dio_flags || is_sync_kiocb(iocb));
4702 	if (IS_ERR_OR_NULL(dio)) {
4703 		ret = PTR_ERR_OR_ZERO(dio);
4704 		if (ret == -ENOTBLK)
4705 			ret = 0;
4706 		if (ret != -EIOCBQUEUED)
4707 			dec_page_count(sbi, F2FS_DIO_WRITE);
4708 	} else {
4709 		ret = iomap_dio_complete(dio);
4710 	}
4711 
4712 	if (whint_mode == WHINT_MODE_OFF)
4713 		iocb->ki_hint = hint;
4714 	if (do_opu)
4715 		f2fs_up_read(&fi->i_gc_rwsem[READ]);
4716 	f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4717 
4718 	if (ret < 0)
4719 		goto out;
4720 	if (pos + ret > inode->i_size)
4721 		f2fs_i_size_write(inode, pos + ret);
4722 	if (!do_opu)
4723 		set_inode_flag(inode, FI_UPDATE_WRITE);
4724 
4725 	if (iov_iter_count(from)) {
4726 		ssize_t ret2;
4727 		loff_t bufio_start_pos = iocb->ki_pos;
4728 
4729 		/*
4730 		 * The direct write was partial, so we need to fall back to a
4731 		 * buffered write for the remainder.
4732 		 */
4733 
4734 		ret2 = f2fs_buffered_write_iter(iocb, from);
4735 		if (iov_iter_count(from))
4736 			f2fs_write_failed(inode, iocb->ki_pos);
4737 		if (ret2 < 0)
4738 			goto out;
4739 
4740 		/*
4741 		 * Ensure that the pagecache pages are written to disk and
4742 		 * invalidated to preserve the expected O_DIRECT semantics.
4743 		 */
4744 		if (ret2 > 0) {
4745 			loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4746 
4747 			ret += ret2;
4748 
4749 			ret2 = filemap_write_and_wait_range(file->f_mapping,
4750 							    bufio_start_pos,
4751 							    bufio_end_pos);
4752 			if (ret2 < 0)
4753 				goto out;
4754 			invalidate_mapping_pages(file->f_mapping,
4755 						 bufio_start_pos >> PAGE_SHIFT,
4756 						 bufio_end_pos >> PAGE_SHIFT);
4757 		}
4758 	} else {
4759 		/* iomap_dio_rw() already handled the generic_write_sync(). */
4760 		*may_need_sync = false;
4761 	}
4762 out:
4763 	if (trace_android_fs_datawrite_start_enabled())
4764 		trace_android_fs_datawrite_end(inode, pos, count);
4765 	trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4766 	return ret;
4767 }
4768 
f2fs_file_write_iter(struct kiocb * iocb,struct iov_iter * from)4769 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4770 {
4771 	struct inode *inode = file_inode(iocb->ki_filp);
4772 	const loff_t orig_pos = iocb->ki_pos;
4773 	const size_t orig_count = iov_iter_count(from);
4774 	loff_t target_size;
4775 	bool dio;
4776 	bool may_need_sync = true;
4777 	int preallocated;
4778 	ssize_t ret;
4779 
4780 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4781 		ret = -EIO;
4782 		goto out;
4783 	}
4784 
4785 	if (!f2fs_is_compress_backend_ready(inode)) {
4786 		ret = -EOPNOTSUPP;
4787 		goto out;
4788 	}
4789 
4790 	if (iocb->ki_flags & IOCB_NOWAIT) {
4791 		if (!inode_trylock(inode)) {
4792 			ret = -EAGAIN;
4793 			goto out;
4794 		}
4795 	} else {
4796 		inode_lock(inode);
4797 	}
4798 
4799 	ret = f2fs_write_checks(iocb, from);
4800 	if (ret <= 0)
4801 		goto out_unlock;
4802 
4803 	/* Determine whether we will do a direct write or a buffered write. */
4804 	dio = f2fs_should_use_dio(inode, iocb, from);
4805 
4806 	/* Possibly preallocate the blocks for the write. */
4807 	target_size = iocb->ki_pos + iov_iter_count(from);
4808 	preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4809 	if (preallocated < 0)
4810 		ret = preallocated;
4811 	else
4812 		/* Do the actual write. */
4813 		ret = dio ?
4814 			f2fs_dio_write_iter(iocb, from, &may_need_sync):
4815 			f2fs_buffered_write_iter(iocb, from);
4816 
4817 	/* Don't leave any preallocated blocks around past i_size. */
4818 	if (preallocated && i_size_read(inode) < target_size) {
4819 		f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4820 		f2fs_down_write(&F2FS_I(inode)->i_mmap_sem);
4821 		if (!f2fs_truncate(inode))
4822 			file_dont_truncate(inode);
4823 		f2fs_up_write(&F2FS_I(inode)->i_mmap_sem);
4824 		f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4825 	} else {
4826 		file_dont_truncate(inode);
4827 	}
4828 
4829 	clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4830 out_unlock:
4831 	inode_unlock(inode);
4832 out:
4833 	trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4834 	if (ret > 0 && may_need_sync)
4835 		ret = generic_write_sync(iocb, ret);
4836 	return ret;
4837 }
4838 
f2fs_file_fadvise(struct file * filp,loff_t offset,loff_t len,int advice)4839 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4840 		int advice)
4841 {
4842 	struct address_space *mapping;
4843 	struct backing_dev_info *bdi;
4844 	struct inode *inode = file_inode(filp);
4845 	int err;
4846 
4847 	if (advice == POSIX_FADV_SEQUENTIAL) {
4848 		if (S_ISFIFO(inode->i_mode))
4849 			return -ESPIPE;
4850 
4851 		mapping = filp->f_mapping;
4852 		if (!mapping || len < 0)
4853 			return -EINVAL;
4854 
4855 		bdi = inode_to_bdi(mapping->host);
4856 		filp->f_ra.ra_pages = bdi->ra_pages *
4857 			F2FS_I_SB(inode)->seq_file_ra_mul;
4858 		spin_lock(&filp->f_lock);
4859 		filp->f_mode &= ~FMODE_RANDOM;
4860 		spin_unlock(&filp->f_lock);
4861 		return 0;
4862 	}
4863 
4864 	err = generic_fadvise(filp, offset, len, advice);
4865 	if (!err && advice == POSIX_FADV_DONTNEED &&
4866 		test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4867 		f2fs_compressed_file(inode))
4868 		f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4869 
4870 	return err;
4871 }
4872 
4873 #ifdef CONFIG_COMPAT
4874 struct compat_f2fs_gc_range {
4875 	u32 sync;
4876 	compat_u64 start;
4877 	compat_u64 len;
4878 };
4879 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE	_IOW(F2FS_IOCTL_MAGIC, 11,\
4880 						struct compat_f2fs_gc_range)
4881 
f2fs_compat_ioc_gc_range(struct file * file,unsigned long arg)4882 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4883 {
4884 	struct compat_f2fs_gc_range __user *urange;
4885 	struct f2fs_gc_range range;
4886 	int err;
4887 
4888 	urange = compat_ptr(arg);
4889 	err = get_user(range.sync, &urange->sync);
4890 	err |= get_user(range.start, &urange->start);
4891 	err |= get_user(range.len, &urange->len);
4892 	if (err)
4893 		return -EFAULT;
4894 
4895 	return __f2fs_ioc_gc_range(file, &range);
4896 }
4897 
4898 struct compat_f2fs_move_range {
4899 	u32 dst_fd;
4900 	compat_u64 pos_in;
4901 	compat_u64 pos_out;
4902 	compat_u64 len;
4903 };
4904 #define F2FS_IOC32_MOVE_RANGE		_IOWR(F2FS_IOCTL_MAGIC, 9,	\
4905 					struct compat_f2fs_move_range)
4906 
f2fs_compat_ioc_move_range(struct file * file,unsigned long arg)4907 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4908 {
4909 	struct compat_f2fs_move_range __user *urange;
4910 	struct f2fs_move_range range;
4911 	int err;
4912 
4913 	urange = compat_ptr(arg);
4914 	err = get_user(range.dst_fd, &urange->dst_fd);
4915 	err |= get_user(range.pos_in, &urange->pos_in);
4916 	err |= get_user(range.pos_out, &urange->pos_out);
4917 	err |= get_user(range.len, &urange->len);
4918 	if (err)
4919 		return -EFAULT;
4920 
4921 	return __f2fs_ioc_move_range(file, &range);
4922 }
4923 
f2fs_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)4924 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4925 {
4926 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4927 		return -EIO;
4928 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4929 		return -ENOSPC;
4930 
4931 	switch (cmd) {
4932 	case FS_IOC32_GETFLAGS:
4933 		cmd = FS_IOC_GETFLAGS;
4934 		break;
4935 	case FS_IOC32_SETFLAGS:
4936 		cmd = FS_IOC_SETFLAGS;
4937 		break;
4938 	case FS_IOC32_GETVERSION:
4939 		cmd = FS_IOC_GETVERSION;
4940 		break;
4941 	case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4942 		return f2fs_compat_ioc_gc_range(file, arg);
4943 	case F2FS_IOC32_MOVE_RANGE:
4944 		return f2fs_compat_ioc_move_range(file, arg);
4945 	case F2FS_IOC_START_ATOMIC_WRITE:
4946 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4947 	case F2FS_IOC_START_VOLATILE_WRITE:
4948 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4949 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
4950 	case F2FS_IOC_SHUTDOWN:
4951 	case FITRIM:
4952 	case FS_IOC_SET_ENCRYPTION_POLICY:
4953 	case FS_IOC_GET_ENCRYPTION_PWSALT:
4954 	case FS_IOC_GET_ENCRYPTION_POLICY:
4955 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4956 	case FS_IOC_ADD_ENCRYPTION_KEY:
4957 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
4958 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4959 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4960 	case FS_IOC_GET_ENCRYPTION_NONCE:
4961 	case F2FS_IOC_GARBAGE_COLLECT:
4962 	case F2FS_IOC_WRITE_CHECKPOINT:
4963 	case F2FS_IOC_DEFRAGMENT:
4964 	case F2FS_IOC_FLUSH_DEVICE:
4965 	case F2FS_IOC_GET_FEATURES:
4966 	case FS_IOC_FSGETXATTR:
4967 	case FS_IOC_FSSETXATTR:
4968 	case F2FS_IOC_GET_PIN_FILE:
4969 	case F2FS_IOC_SET_PIN_FILE:
4970 	case F2FS_IOC_PRECACHE_EXTENTS:
4971 	case F2FS_IOC_RESIZE_FS:
4972 	case FS_IOC_ENABLE_VERITY:
4973 	case FS_IOC_MEASURE_VERITY:
4974 	case FS_IOC_READ_VERITY_METADATA:
4975 	case FS_IOC_GETFSLABEL:
4976 	case FS_IOC_SETFSLABEL:
4977 	case F2FS_IOC_GET_COMPRESS_BLOCKS:
4978 	case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4979 	case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4980 	case F2FS_IOC_SEC_TRIM_FILE:
4981 	case F2FS_IOC_GET_COMPRESS_OPTION:
4982 	case F2FS_IOC_SET_COMPRESS_OPTION:
4983 	case F2FS_IOC_DECOMPRESS_FILE:
4984 	case F2FS_IOC_COMPRESS_FILE:
4985 		break;
4986 	default:
4987 		return -ENOIOCTLCMD;
4988 	}
4989 	return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4990 }
4991 #endif
4992 
4993 const struct file_operations f2fs_file_operations = {
4994 	.llseek		= f2fs_llseek,
4995 	.read_iter	= f2fs_file_read_iter,
4996 	.write_iter	= f2fs_file_write_iter,
4997 	.open		= f2fs_file_open,
4998 	.release	= f2fs_release_file,
4999 	.mmap		= f2fs_file_mmap,
5000 	.flush		= f2fs_file_flush,
5001 	.fsync		= f2fs_sync_file,
5002 	.fallocate	= f2fs_fallocate,
5003 	.unlocked_ioctl	= f2fs_ioctl,
5004 #ifdef CONFIG_COMPAT
5005 	.compat_ioctl	= f2fs_compat_ioctl,
5006 #endif
5007 	.splice_read	= generic_file_splice_read,
5008 	.splice_write	= iter_file_splice_write,
5009 	.fadvise	= f2fs_file_fadvise,
5010 };
5011