xref: /linux/fs/f2fs/file.c (revision a4f174dee4ae842e07cab7eeec194a3e60925c8d)
1 /*
2  * fs/f2fs/file.c
3  *
4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5  *             http://www.samsung.com/
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  */
11 #include <linux/fs.h>
12 #include <linux/f2fs_fs.h>
13 #include <linux/stat.h>
14 #include <linux/buffer_head.h>
15 #include <linux/writeback.h>
16 #include <linux/blkdev.h>
17 #include <linux/falloc.h>
18 #include <linux/types.h>
19 #include <linux/compat.h>
20 #include <linux/uaccess.h>
21 #include <linux/mount.h>
22 #include <linux/pagevec.h>
23 
24 #include "f2fs.h"
25 #include "node.h"
26 #include "segment.h"
27 #include "xattr.h"
28 #include "acl.h"
29 #include <trace/events/f2fs.h>
30 
31 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
32 						struct vm_fault *vmf)
33 {
34 	struct page *page = vmf->page;
35 	struct inode *inode = file_inode(vma->vm_file);
36 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
37 	struct dnode_of_data dn;
38 	int err;
39 
40 	f2fs_balance_fs(sbi);
41 
42 	sb_start_pagefault(inode->i_sb);
43 
44 	/* force to convert with normal data indices */
45 	err = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1, page);
46 	if (err)
47 		goto out;
48 
49 	/* block allocation */
50 	f2fs_lock_op(sbi);
51 	set_new_dnode(&dn, inode, NULL, NULL, 0);
52 	err = f2fs_reserve_block(&dn, page->index);
53 	f2fs_unlock_op(sbi);
54 	if (err)
55 		goto out;
56 
57 	file_update_time(vma->vm_file);
58 	lock_page(page);
59 	if (unlikely(page->mapping != inode->i_mapping ||
60 			page_offset(page) > i_size_read(inode) ||
61 			!PageUptodate(page))) {
62 		unlock_page(page);
63 		err = -EFAULT;
64 		goto out;
65 	}
66 
67 	/*
68 	 * check to see if the page is mapped already (no holes)
69 	 */
70 	if (PageMappedToDisk(page))
71 		goto mapped;
72 
73 	/* page is wholly or partially inside EOF */
74 	if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
75 		unsigned offset;
76 		offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
77 		zero_user_segment(page, offset, PAGE_CACHE_SIZE);
78 	}
79 	set_page_dirty(page);
80 	SetPageUptodate(page);
81 
82 	trace_f2fs_vm_page_mkwrite(page, DATA);
83 mapped:
84 	/* fill the page */
85 	f2fs_wait_on_page_writeback(page, DATA);
86 out:
87 	sb_end_pagefault(inode->i_sb);
88 	return block_page_mkwrite_return(err);
89 }
90 
91 static const struct vm_operations_struct f2fs_file_vm_ops = {
92 	.fault		= filemap_fault,
93 	.map_pages	= filemap_map_pages,
94 	.page_mkwrite	= f2fs_vm_page_mkwrite,
95 	.remap_pages	= generic_file_remap_pages,
96 };
97 
98 static int get_parent_ino(struct inode *inode, nid_t *pino)
99 {
100 	struct dentry *dentry;
101 
102 	inode = igrab(inode);
103 	dentry = d_find_any_alias(inode);
104 	iput(inode);
105 	if (!dentry)
106 		return 0;
107 
108 	if (update_dent_inode(inode, &dentry->d_name)) {
109 		dput(dentry);
110 		return 0;
111 	}
112 
113 	*pino = parent_ino(dentry);
114 	dput(dentry);
115 	return 1;
116 }
117 
118 static inline bool need_do_checkpoint(struct inode *inode)
119 {
120 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
121 	bool need_cp = false;
122 
123 	if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
124 		need_cp = true;
125 	else if (file_wrong_pino(inode))
126 		need_cp = true;
127 	else if (!space_for_roll_forward(sbi))
128 		need_cp = true;
129 	else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
130 		need_cp = true;
131 	else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
132 		need_cp = true;
133 
134 	return need_cp;
135 }
136 
137 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
138 {
139 	struct inode *inode = file->f_mapping->host;
140 	struct f2fs_inode_info *fi = F2FS_I(inode);
141 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
142 	nid_t ino = inode->i_ino;
143 	int ret = 0;
144 	bool need_cp = false;
145 	struct writeback_control wbc = {
146 		.sync_mode = WB_SYNC_ALL,
147 		.nr_to_write = LONG_MAX,
148 		.for_reclaim = 0,
149 	};
150 
151 	if (unlikely(f2fs_readonly(inode->i_sb)))
152 		return 0;
153 
154 	trace_f2fs_sync_file_enter(inode);
155 
156 	/* if fdatasync is triggered, let's do in-place-update */
157 	if (get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
158 		set_inode_flag(fi, FI_NEED_IPU);
159 	ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
160 	clear_inode_flag(fi, FI_NEED_IPU);
161 
162 	if (ret) {
163 		trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
164 		return ret;
165 	}
166 
167 	/*
168 	 * if there is no written data, don't waste time to write recovery info.
169 	 */
170 	if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
171 			!exist_written_data(sbi, ino, APPEND_INO)) {
172 		struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
173 
174 		/* But we need to avoid that there are some inode updates */
175 		if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino)) {
176 			f2fs_put_page(i, 0);
177 			goto go_write;
178 		}
179 		f2fs_put_page(i, 0);
180 
181 		if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
182 				exist_written_data(sbi, ino, UPDATE_INO))
183 			goto flush_out;
184 		goto out;
185 	}
186 go_write:
187 	/* guarantee free sections for fsync */
188 	f2fs_balance_fs(sbi);
189 
190 	/*
191 	 * Both of fdatasync() and fsync() are able to be recovered from
192 	 * sudden-power-off.
193 	 */
194 	down_read(&fi->i_sem);
195 	need_cp = need_do_checkpoint(inode);
196 	up_read(&fi->i_sem);
197 
198 	if (need_cp) {
199 		nid_t pino;
200 
201 		/* all the dirty node pages should be flushed for POR */
202 		ret = f2fs_sync_fs(inode->i_sb, 1);
203 
204 		down_write(&fi->i_sem);
205 		F2FS_I(inode)->xattr_ver = 0;
206 		if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
207 					get_parent_ino(inode, &pino)) {
208 			F2FS_I(inode)->i_pino = pino;
209 			file_got_pino(inode);
210 			up_write(&fi->i_sem);
211 			mark_inode_dirty_sync(inode);
212 			ret = f2fs_write_inode(inode, NULL);
213 			if (ret)
214 				goto out;
215 		} else {
216 			up_write(&fi->i_sem);
217 		}
218 	} else {
219 sync_nodes:
220 		sync_node_pages(sbi, ino, &wbc);
221 
222 		if (need_inode_block_update(sbi, ino)) {
223 			mark_inode_dirty_sync(inode);
224 			ret = f2fs_write_inode(inode, NULL);
225 			if (ret)
226 				goto out;
227 			goto sync_nodes;
228 		}
229 
230 		ret = wait_on_node_pages_writeback(sbi, ino);
231 		if (ret)
232 			goto out;
233 
234 		/* once recovery info is written, don't need to tack this */
235 		remove_dirty_inode(sbi, ino, APPEND_INO);
236 		clear_inode_flag(fi, FI_APPEND_WRITE);
237 flush_out:
238 		remove_dirty_inode(sbi, ino, UPDATE_INO);
239 		clear_inode_flag(fi, FI_UPDATE_WRITE);
240 		ret = f2fs_issue_flush(F2FS_I_SB(inode));
241 	}
242 out:
243 	trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
244 	return ret;
245 }
246 
247 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
248 						pgoff_t pgofs, int whence)
249 {
250 	struct pagevec pvec;
251 	int nr_pages;
252 
253 	if (whence != SEEK_DATA)
254 		return 0;
255 
256 	/* find first dirty page index */
257 	pagevec_init(&pvec, 0);
258 	nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
259 					PAGECACHE_TAG_DIRTY, 1);
260 	pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
261 	pagevec_release(&pvec);
262 	return pgofs;
263 }
264 
265 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
266 							int whence)
267 {
268 	switch (whence) {
269 	case SEEK_DATA:
270 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
271 			(blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
272 			return true;
273 		break;
274 	case SEEK_HOLE:
275 		if (blkaddr == NULL_ADDR)
276 			return true;
277 		break;
278 	}
279 	return false;
280 }
281 
282 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
283 {
284 	struct inode *inode = file->f_mapping->host;
285 	loff_t maxbytes = inode->i_sb->s_maxbytes;
286 	struct dnode_of_data dn;
287 	pgoff_t pgofs, end_offset, dirty;
288 	loff_t data_ofs = offset;
289 	loff_t isize;
290 	int err = 0;
291 
292 	mutex_lock(&inode->i_mutex);
293 
294 	isize = i_size_read(inode);
295 	if (offset >= isize)
296 		goto fail;
297 
298 	/* handle inline data case */
299 	if (f2fs_has_inline_data(inode)) {
300 		if (whence == SEEK_HOLE)
301 			data_ofs = isize;
302 		goto found;
303 	}
304 
305 	pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
306 
307 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
308 
309 	for (; data_ofs < isize; data_ofs = pgofs << PAGE_CACHE_SHIFT) {
310 		set_new_dnode(&dn, inode, NULL, NULL, 0);
311 		err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
312 		if (err && err != -ENOENT) {
313 			goto fail;
314 		} else if (err == -ENOENT) {
315 			/* direct node does not exists */
316 			if (whence == SEEK_DATA) {
317 				pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
318 							F2FS_I(inode));
319 				continue;
320 			} else {
321 				goto found;
322 			}
323 		}
324 
325 		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
326 
327 		/* find data/hole in dnode block */
328 		for (; dn.ofs_in_node < end_offset;
329 				dn.ofs_in_node++, pgofs++,
330 				data_ofs = pgofs << PAGE_CACHE_SHIFT) {
331 			block_t blkaddr;
332 			blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
333 
334 			if (__found_offset(blkaddr, dirty, pgofs, whence)) {
335 				f2fs_put_dnode(&dn);
336 				goto found;
337 			}
338 		}
339 		f2fs_put_dnode(&dn);
340 	}
341 
342 	if (whence == SEEK_DATA)
343 		goto fail;
344 found:
345 	if (whence == SEEK_HOLE && data_ofs > isize)
346 		data_ofs = isize;
347 	mutex_unlock(&inode->i_mutex);
348 	return vfs_setpos(file, data_ofs, maxbytes);
349 fail:
350 	mutex_unlock(&inode->i_mutex);
351 	return -ENXIO;
352 }
353 
354 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
355 {
356 	struct inode *inode = file->f_mapping->host;
357 	loff_t maxbytes = inode->i_sb->s_maxbytes;
358 
359 	switch (whence) {
360 	case SEEK_SET:
361 	case SEEK_CUR:
362 	case SEEK_END:
363 		return generic_file_llseek_size(file, offset, whence,
364 						maxbytes, i_size_read(inode));
365 	case SEEK_DATA:
366 	case SEEK_HOLE:
367 		if (offset < 0)
368 			return -ENXIO;
369 		return f2fs_seek_block(file, offset, whence);
370 	}
371 
372 	return -EINVAL;
373 }
374 
375 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
376 {
377 	file_accessed(file);
378 	vma->vm_ops = &f2fs_file_vm_ops;
379 	return 0;
380 }
381 
382 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
383 {
384 	int nr_free = 0, ofs = dn->ofs_in_node;
385 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
386 	struct f2fs_node *raw_node;
387 	__le32 *addr;
388 
389 	raw_node = F2FS_NODE(dn->node_page);
390 	addr = blkaddr_in_node(raw_node) + ofs;
391 
392 	for (; count > 0; count--, addr++, dn->ofs_in_node++) {
393 		block_t blkaddr = le32_to_cpu(*addr);
394 		if (blkaddr == NULL_ADDR)
395 			continue;
396 
397 		update_extent_cache(NULL_ADDR, dn);
398 		invalidate_blocks(sbi, blkaddr);
399 		nr_free++;
400 	}
401 	if (nr_free) {
402 		dec_valid_block_count(sbi, dn->inode, nr_free);
403 		set_page_dirty(dn->node_page);
404 		sync_inode_page(dn);
405 	}
406 	dn->ofs_in_node = ofs;
407 
408 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
409 					 dn->ofs_in_node, nr_free);
410 	return nr_free;
411 }
412 
413 void truncate_data_blocks(struct dnode_of_data *dn)
414 {
415 	truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
416 }
417 
418 static void truncate_partial_data_page(struct inode *inode, u64 from)
419 {
420 	unsigned offset = from & (PAGE_CACHE_SIZE - 1);
421 	struct page *page;
422 
423 	if (f2fs_has_inline_data(inode))
424 		return truncate_inline_data(inode, from);
425 
426 	if (!offset)
427 		return;
428 
429 	page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false);
430 	if (IS_ERR(page))
431 		return;
432 
433 	lock_page(page);
434 	if (unlikely(!PageUptodate(page) ||
435 			page->mapping != inode->i_mapping))
436 		goto out;
437 
438 	f2fs_wait_on_page_writeback(page, DATA);
439 	zero_user(page, offset, PAGE_CACHE_SIZE - offset);
440 	set_page_dirty(page);
441 
442 out:
443 	f2fs_put_page(page, 1);
444 }
445 
446 int truncate_blocks(struct inode *inode, u64 from, bool lock)
447 {
448 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
449 	unsigned int blocksize = inode->i_sb->s_blocksize;
450 	struct dnode_of_data dn;
451 	pgoff_t free_from;
452 	int count = 0, err = 0;
453 
454 	trace_f2fs_truncate_blocks_enter(inode, from);
455 
456 	if (f2fs_has_inline_data(inode))
457 		goto done;
458 
459 	free_from = (pgoff_t)
460 			((from + blocksize - 1) >> (sbi->log_blocksize));
461 
462 	if (lock)
463 		f2fs_lock_op(sbi);
464 
465 	set_new_dnode(&dn, inode, NULL, NULL, 0);
466 	err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
467 	if (err) {
468 		if (err == -ENOENT)
469 			goto free_next;
470 		if (lock)
471 			f2fs_unlock_op(sbi);
472 		trace_f2fs_truncate_blocks_exit(inode, err);
473 		return err;
474 	}
475 
476 	count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
477 
478 	count -= dn.ofs_in_node;
479 	f2fs_bug_on(sbi, count < 0);
480 
481 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
482 		truncate_data_blocks_range(&dn, count);
483 		free_from += count;
484 	}
485 
486 	f2fs_put_dnode(&dn);
487 free_next:
488 	err = truncate_inode_blocks(inode, free_from);
489 	if (lock)
490 		f2fs_unlock_op(sbi);
491 done:
492 	/* lastly zero out the first data page */
493 	truncate_partial_data_page(inode, from);
494 
495 	trace_f2fs_truncate_blocks_exit(inode, err);
496 	return err;
497 }
498 
499 void f2fs_truncate(struct inode *inode)
500 {
501 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
502 				S_ISLNK(inode->i_mode)))
503 		return;
504 
505 	trace_f2fs_truncate(inode);
506 
507 	if (!truncate_blocks(inode, i_size_read(inode), true)) {
508 		inode->i_mtime = inode->i_ctime = CURRENT_TIME;
509 		mark_inode_dirty(inode);
510 	}
511 }
512 
513 int f2fs_getattr(struct vfsmount *mnt,
514 			 struct dentry *dentry, struct kstat *stat)
515 {
516 	struct inode *inode = dentry->d_inode;
517 	generic_fillattr(inode, stat);
518 	stat->blocks <<= 3;
519 	return 0;
520 }
521 
522 #ifdef CONFIG_F2FS_FS_POSIX_ACL
523 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
524 {
525 	struct f2fs_inode_info *fi = F2FS_I(inode);
526 	unsigned int ia_valid = attr->ia_valid;
527 
528 	if (ia_valid & ATTR_UID)
529 		inode->i_uid = attr->ia_uid;
530 	if (ia_valid & ATTR_GID)
531 		inode->i_gid = attr->ia_gid;
532 	if (ia_valid & ATTR_ATIME)
533 		inode->i_atime = timespec_trunc(attr->ia_atime,
534 						inode->i_sb->s_time_gran);
535 	if (ia_valid & ATTR_MTIME)
536 		inode->i_mtime = timespec_trunc(attr->ia_mtime,
537 						inode->i_sb->s_time_gran);
538 	if (ia_valid & ATTR_CTIME)
539 		inode->i_ctime = timespec_trunc(attr->ia_ctime,
540 						inode->i_sb->s_time_gran);
541 	if (ia_valid & ATTR_MODE) {
542 		umode_t mode = attr->ia_mode;
543 
544 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
545 			mode &= ~S_ISGID;
546 		set_acl_inode(fi, mode);
547 	}
548 }
549 #else
550 #define __setattr_copy setattr_copy
551 #endif
552 
553 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
554 {
555 	struct inode *inode = dentry->d_inode;
556 	struct f2fs_inode_info *fi = F2FS_I(inode);
557 	int err;
558 
559 	err = inode_change_ok(inode, attr);
560 	if (err)
561 		return err;
562 
563 	if (attr->ia_valid & ATTR_SIZE) {
564 		err = f2fs_convert_inline_data(inode, attr->ia_size, NULL);
565 		if (err)
566 			return err;
567 
568 		if (attr->ia_size != i_size_read(inode)) {
569 			truncate_setsize(inode, attr->ia_size);
570 			f2fs_truncate(inode);
571 			f2fs_balance_fs(F2FS_I_SB(inode));
572 		} else {
573 			/*
574 			 * giving a chance to truncate blocks past EOF which
575 			 * are fallocated with FALLOC_FL_KEEP_SIZE.
576 			 */
577 			f2fs_truncate(inode);
578 		}
579 	}
580 
581 	__setattr_copy(inode, attr);
582 
583 	if (attr->ia_valid & ATTR_MODE) {
584 		err = posix_acl_chmod(inode, get_inode_mode(inode));
585 		if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
586 			inode->i_mode = fi->i_acl_mode;
587 			clear_inode_flag(fi, FI_ACL_MODE);
588 		}
589 	}
590 
591 	mark_inode_dirty(inode);
592 	return err;
593 }
594 
595 const struct inode_operations f2fs_file_inode_operations = {
596 	.getattr	= f2fs_getattr,
597 	.setattr	= f2fs_setattr,
598 	.get_acl	= f2fs_get_acl,
599 	.set_acl	= f2fs_set_acl,
600 #ifdef CONFIG_F2FS_FS_XATTR
601 	.setxattr	= generic_setxattr,
602 	.getxattr	= generic_getxattr,
603 	.listxattr	= f2fs_listxattr,
604 	.removexattr	= generic_removexattr,
605 #endif
606 	.fiemap		= f2fs_fiemap,
607 };
608 
609 static void fill_zero(struct inode *inode, pgoff_t index,
610 					loff_t start, loff_t len)
611 {
612 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
613 	struct page *page;
614 
615 	if (!len)
616 		return;
617 
618 	f2fs_balance_fs(sbi);
619 
620 	f2fs_lock_op(sbi);
621 	page = get_new_data_page(inode, NULL, index, false);
622 	f2fs_unlock_op(sbi);
623 
624 	if (!IS_ERR(page)) {
625 		f2fs_wait_on_page_writeback(page, DATA);
626 		zero_user(page, start, len);
627 		set_page_dirty(page);
628 		f2fs_put_page(page, 1);
629 	}
630 }
631 
632 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
633 {
634 	pgoff_t index;
635 	int err;
636 
637 	for (index = pg_start; index < pg_end; index++) {
638 		struct dnode_of_data dn;
639 
640 		set_new_dnode(&dn, inode, NULL, NULL, 0);
641 		err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
642 		if (err) {
643 			if (err == -ENOENT)
644 				continue;
645 			return err;
646 		}
647 
648 		if (dn.data_blkaddr != NULL_ADDR)
649 			truncate_data_blocks_range(&dn, 1);
650 		f2fs_put_dnode(&dn);
651 	}
652 	return 0;
653 }
654 
655 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
656 {
657 	pgoff_t pg_start, pg_end;
658 	loff_t off_start, off_end;
659 	int ret = 0;
660 
661 	if (!S_ISREG(inode->i_mode))
662 		return -EOPNOTSUPP;
663 
664 	/* skip punching hole beyond i_size */
665 	if (offset >= inode->i_size)
666 		return ret;
667 
668 	ret = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1, NULL);
669 	if (ret)
670 		return ret;
671 
672 	pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
673 	pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
674 
675 	off_start = offset & (PAGE_CACHE_SIZE - 1);
676 	off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
677 
678 	if (pg_start == pg_end) {
679 		fill_zero(inode, pg_start, off_start,
680 						off_end - off_start);
681 	} else {
682 		if (off_start)
683 			fill_zero(inode, pg_start++, off_start,
684 					PAGE_CACHE_SIZE - off_start);
685 		if (off_end)
686 			fill_zero(inode, pg_end, 0, off_end);
687 
688 		if (pg_start < pg_end) {
689 			struct address_space *mapping = inode->i_mapping;
690 			loff_t blk_start, blk_end;
691 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
692 
693 			f2fs_balance_fs(sbi);
694 
695 			blk_start = pg_start << PAGE_CACHE_SHIFT;
696 			blk_end = pg_end << PAGE_CACHE_SHIFT;
697 			truncate_inode_pages_range(mapping, blk_start,
698 					blk_end - 1);
699 
700 			f2fs_lock_op(sbi);
701 			ret = truncate_hole(inode, pg_start, pg_end);
702 			f2fs_unlock_op(sbi);
703 		}
704 	}
705 
706 	return ret;
707 }
708 
709 static int expand_inode_data(struct inode *inode, loff_t offset,
710 					loff_t len, int mode)
711 {
712 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
713 	pgoff_t index, pg_start, pg_end;
714 	loff_t new_size = i_size_read(inode);
715 	loff_t off_start, off_end;
716 	int ret = 0;
717 
718 	f2fs_balance_fs(sbi);
719 
720 	ret = inode_newsize_ok(inode, (len + offset));
721 	if (ret)
722 		return ret;
723 
724 	ret = f2fs_convert_inline_data(inode, offset + len, NULL);
725 	if (ret)
726 		return ret;
727 
728 	pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
729 	pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
730 
731 	off_start = offset & (PAGE_CACHE_SIZE - 1);
732 	off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
733 
734 	f2fs_lock_op(sbi);
735 
736 	for (index = pg_start; index <= pg_end; index++) {
737 		struct dnode_of_data dn;
738 
739 		if (index == pg_end && !off_end)
740 			goto noalloc;
741 
742 		set_new_dnode(&dn, inode, NULL, NULL, 0);
743 		ret = f2fs_reserve_block(&dn, index);
744 		if (ret)
745 			break;
746 noalloc:
747 		if (pg_start == pg_end)
748 			new_size = offset + len;
749 		else if (index == pg_start && off_start)
750 			new_size = (index + 1) << PAGE_CACHE_SHIFT;
751 		else if (index == pg_end)
752 			new_size = (index << PAGE_CACHE_SHIFT) + off_end;
753 		else
754 			new_size += PAGE_CACHE_SIZE;
755 	}
756 
757 	if (!(mode & FALLOC_FL_KEEP_SIZE) &&
758 		i_size_read(inode) < new_size) {
759 		i_size_write(inode, new_size);
760 		mark_inode_dirty(inode);
761 		update_inode_page(inode);
762 	}
763 	f2fs_unlock_op(sbi);
764 
765 	return ret;
766 }
767 
768 static long f2fs_fallocate(struct file *file, int mode,
769 				loff_t offset, loff_t len)
770 {
771 	struct inode *inode = file_inode(file);
772 	long ret;
773 
774 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
775 		return -EOPNOTSUPP;
776 
777 	mutex_lock(&inode->i_mutex);
778 
779 	if (mode & FALLOC_FL_PUNCH_HOLE)
780 		ret = punch_hole(inode, offset, len);
781 	else
782 		ret = expand_inode_data(inode, offset, len, mode);
783 
784 	if (!ret) {
785 		inode->i_mtime = inode->i_ctime = CURRENT_TIME;
786 		mark_inode_dirty(inode);
787 	}
788 
789 	mutex_unlock(&inode->i_mutex);
790 
791 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
792 	return ret;
793 }
794 
795 #define F2FS_REG_FLMASK		(~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
796 #define F2FS_OTHER_FLMASK	(FS_NODUMP_FL | FS_NOATIME_FL)
797 
798 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
799 {
800 	if (S_ISDIR(mode))
801 		return flags;
802 	else if (S_ISREG(mode))
803 		return flags & F2FS_REG_FLMASK;
804 	else
805 		return flags & F2FS_OTHER_FLMASK;
806 }
807 
808 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
809 {
810 	struct inode *inode = file_inode(filp);
811 	struct f2fs_inode_info *fi = F2FS_I(inode);
812 	unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
813 	return put_user(flags, (int __user *)arg);
814 }
815 
816 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
817 {
818 	struct inode *inode = file_inode(filp);
819 	struct f2fs_inode_info *fi = F2FS_I(inode);
820 	unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
821 	unsigned int oldflags;
822 	int ret;
823 
824 	ret = mnt_want_write_file(filp);
825 	if (ret)
826 		return ret;
827 
828 	if (!inode_owner_or_capable(inode)) {
829 		ret = -EACCES;
830 		goto out;
831 	}
832 
833 	if (get_user(flags, (int __user *)arg)) {
834 		ret = -EFAULT;
835 		goto out;
836 	}
837 
838 	flags = f2fs_mask_flags(inode->i_mode, flags);
839 
840 	mutex_lock(&inode->i_mutex);
841 
842 	oldflags = fi->i_flags;
843 
844 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
845 		if (!capable(CAP_LINUX_IMMUTABLE)) {
846 			mutex_unlock(&inode->i_mutex);
847 			ret = -EPERM;
848 			goto out;
849 		}
850 	}
851 
852 	flags = flags & FS_FL_USER_MODIFIABLE;
853 	flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
854 	fi->i_flags = flags;
855 	mutex_unlock(&inode->i_mutex);
856 
857 	f2fs_set_inode_flags(inode);
858 	inode->i_ctime = CURRENT_TIME;
859 	mark_inode_dirty(inode);
860 out:
861 	mnt_drop_write_file(filp);
862 	return ret;
863 }
864 
865 static int f2fs_ioc_start_atomic_write(struct file *filp)
866 {
867 	struct inode *inode = file_inode(filp);
868 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
869 
870 	if (!inode_owner_or_capable(inode))
871 		return -EACCES;
872 
873 	f2fs_balance_fs(sbi);
874 
875 	set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
876 
877 	return f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1, NULL);
878 }
879 
880 static int f2fs_ioc_commit_atomic_write(struct file *filp)
881 {
882 	struct inode *inode = file_inode(filp);
883 	int ret;
884 
885 	if (!inode_owner_or_capable(inode))
886 		return -EACCES;
887 
888 	if (f2fs_is_volatile_file(inode))
889 		return 0;
890 
891 	ret = mnt_want_write_file(filp);
892 	if (ret)
893 		return ret;
894 
895 	if (f2fs_is_atomic_file(inode))
896 		commit_inmem_pages(inode, false);
897 
898 	ret = f2fs_sync_file(filp, 0, LONG_MAX, 0);
899 	mnt_drop_write_file(filp);
900 	return ret;
901 }
902 
903 static int f2fs_ioc_start_volatile_write(struct file *filp)
904 {
905 	struct inode *inode = file_inode(filp);
906 
907 	if (!inode_owner_or_capable(inode))
908 		return -EACCES;
909 
910 	set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
911 	return 0;
912 }
913 
914 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
915 {
916 	struct inode *inode = file_inode(filp);
917 	struct super_block *sb = inode->i_sb;
918 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
919 	struct fstrim_range range;
920 	int ret;
921 
922 	if (!capable(CAP_SYS_ADMIN))
923 		return -EPERM;
924 
925 	if (!blk_queue_discard(q))
926 		return -EOPNOTSUPP;
927 
928 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
929 				sizeof(range)))
930 		return -EFAULT;
931 
932 	range.minlen = max((unsigned int)range.minlen,
933 				q->limits.discard_granularity);
934 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
935 	if (ret < 0)
936 		return ret;
937 
938 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
939 				sizeof(range)))
940 		return -EFAULT;
941 	return 0;
942 }
943 
944 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
945 {
946 	switch (cmd) {
947 	case F2FS_IOC_GETFLAGS:
948 		return f2fs_ioc_getflags(filp, arg);
949 	case F2FS_IOC_SETFLAGS:
950 		return f2fs_ioc_setflags(filp, arg);
951 	case F2FS_IOC_START_ATOMIC_WRITE:
952 		return f2fs_ioc_start_atomic_write(filp);
953 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
954 		return f2fs_ioc_commit_atomic_write(filp);
955 	case F2FS_IOC_START_VOLATILE_WRITE:
956 		return f2fs_ioc_start_volatile_write(filp);
957 	case FITRIM:
958 		return f2fs_ioc_fitrim(filp, arg);
959 	default:
960 		return -ENOTTY;
961 	}
962 }
963 
964 #ifdef CONFIG_COMPAT
965 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
966 {
967 	switch (cmd) {
968 	case F2FS_IOC32_GETFLAGS:
969 		cmd = F2FS_IOC_GETFLAGS;
970 		break;
971 	case F2FS_IOC32_SETFLAGS:
972 		cmd = F2FS_IOC_SETFLAGS;
973 		break;
974 	default:
975 		return -ENOIOCTLCMD;
976 	}
977 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
978 }
979 #endif
980 
981 const struct file_operations f2fs_file_operations = {
982 	.llseek		= f2fs_llseek,
983 	.read		= new_sync_read,
984 	.write		= new_sync_write,
985 	.read_iter	= generic_file_read_iter,
986 	.write_iter	= generic_file_write_iter,
987 	.open		= generic_file_open,
988 	.mmap		= f2fs_file_mmap,
989 	.fsync		= f2fs_sync_file,
990 	.fallocate	= f2fs_fallocate,
991 	.unlocked_ioctl	= f2fs_ioctl,
992 #ifdef CONFIG_COMPAT
993 	.compat_ioctl	= f2fs_compat_ioctl,
994 #endif
995 	.splice_read	= generic_file_splice_read,
996 	.splice_write	= iter_file_splice_write,
997 };
998