Loading...
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#include <linux/uuid.h>
24#include <linux/file.h>
25
26#include "f2fs.h"
27#include "node.h"
28#include "segment.h"
29#include "xattr.h"
30#include "acl.h"
31#include "gc.h"
32#include "trace.h"
33#include <trace/events/f2fs.h>
34
35static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
36 struct vm_fault *vmf)
37{
38 struct page *page = vmf->page;
39 struct inode *inode = file_inode(vma->vm_file);
40 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
41 struct dnode_of_data dn;
42 int err;
43
44 sb_start_pagefault(inode->i_sb);
45
46 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
47
48 /* block allocation */
49 f2fs_lock_op(sbi);
50 set_new_dnode(&dn, inode, NULL, NULL, 0);
51 err = f2fs_reserve_block(&dn, page->index);
52 if (err) {
53 f2fs_unlock_op(sbi);
54 goto out;
55 }
56 f2fs_put_dnode(&dn);
57 f2fs_unlock_op(sbi);
58
59 f2fs_balance_fs(sbi, dn.node_changed);
60
61 file_update_time(vma->vm_file);
62 lock_page(page);
63 if (unlikely(page->mapping != inode->i_mapping ||
64 page_offset(page) > i_size_read(inode) ||
65 !PageUptodate(page))) {
66 unlock_page(page);
67 err = -EFAULT;
68 goto out;
69 }
70
71 /*
72 * check to see if the page is mapped already (no holes)
73 */
74 if (PageMappedToDisk(page))
75 goto mapped;
76
77 /* page is wholly or partially inside EOF */
78 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
79 i_size_read(inode)) {
80 unsigned offset;
81 offset = i_size_read(inode) & ~PAGE_MASK;
82 zero_user_segment(page, offset, PAGE_SIZE);
83 }
84 set_page_dirty(page);
85 if (!PageUptodate(page))
86 SetPageUptodate(page);
87
88 trace_f2fs_vm_page_mkwrite(page, DATA);
89mapped:
90 /* fill the page */
91 f2fs_wait_on_page_writeback(page, DATA, false);
92
93 /* wait for GCed encrypted page writeback */
94 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
95 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
96
97out:
98 sb_end_pagefault(inode->i_sb);
99 f2fs_update_time(sbi, REQ_TIME);
100 return block_page_mkwrite_return(err);
101}
102
103static const struct vm_operations_struct f2fs_file_vm_ops = {
104 .fault = filemap_fault,
105 .map_pages = filemap_map_pages,
106 .page_mkwrite = f2fs_vm_page_mkwrite,
107};
108
109static int get_parent_ino(struct inode *inode, nid_t *pino)
110{
111 struct dentry *dentry;
112
113 inode = igrab(inode);
114 dentry = d_find_any_alias(inode);
115 iput(inode);
116 if (!dentry)
117 return 0;
118
119 if (update_dent_inode(inode, inode, &dentry->d_name)) {
120 dput(dentry);
121 return 0;
122 }
123
124 *pino = parent_ino(dentry);
125 dput(dentry);
126 return 1;
127}
128
129static inline bool need_do_checkpoint(struct inode *inode)
130{
131 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
132 bool need_cp = false;
133
134 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
135 need_cp = true;
136 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
137 need_cp = true;
138 else if (file_wrong_pino(inode))
139 need_cp = true;
140 else if (!space_for_roll_forward(sbi))
141 need_cp = true;
142 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
143 need_cp = true;
144 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
145 need_cp = true;
146 else if (test_opt(sbi, FASTBOOT))
147 need_cp = true;
148 else if (sbi->active_logs == 2)
149 need_cp = true;
150
151 return need_cp;
152}
153
154static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
155{
156 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
157 bool ret = false;
158 /* But we need to avoid that there are some inode updates */
159 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
160 ret = true;
161 f2fs_put_page(i, 0);
162 return ret;
163}
164
165static void try_to_fix_pino(struct inode *inode)
166{
167 struct f2fs_inode_info *fi = F2FS_I(inode);
168 nid_t pino;
169
170 down_write(&fi->i_sem);
171 fi->xattr_ver = 0;
172 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
173 get_parent_ino(inode, &pino)) {
174 f2fs_i_pino_write(inode, pino);
175 file_got_pino(inode);
176 }
177 up_write(&fi->i_sem);
178}
179
180static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
181 int datasync, bool atomic)
182{
183 struct inode *inode = file->f_mapping->host;
184 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
185 nid_t ino = inode->i_ino;
186 int ret = 0;
187 bool need_cp = false;
188 struct writeback_control wbc = {
189 .sync_mode = WB_SYNC_ALL,
190 .nr_to_write = LONG_MAX,
191 .for_reclaim = 0,
192 };
193
194 if (unlikely(f2fs_readonly(inode->i_sb)))
195 return 0;
196
197 trace_f2fs_sync_file_enter(inode);
198
199 /* if fdatasync is triggered, let's do in-place-update */
200 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
201 set_inode_flag(inode, FI_NEED_IPU);
202 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
203 clear_inode_flag(inode, FI_NEED_IPU);
204
205 if (ret) {
206 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
207 return ret;
208 }
209
210 /* if the inode is dirty, let's recover all the time */
211 if (!f2fs_skip_inode_update(inode, datasync)) {
212 f2fs_write_inode(inode, NULL);
213 goto go_write;
214 }
215
216 /*
217 * if there is no written data, don't waste time to write recovery info.
218 */
219 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
220 !exist_written_data(sbi, ino, APPEND_INO)) {
221
222 /* it may call write_inode just prior to fsync */
223 if (need_inode_page_update(sbi, ino))
224 goto go_write;
225
226 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
227 exist_written_data(sbi, ino, UPDATE_INO))
228 goto flush_out;
229 goto out;
230 }
231go_write:
232 /*
233 * Both of fdatasync() and fsync() are able to be recovered from
234 * sudden-power-off.
235 */
236 down_read(&F2FS_I(inode)->i_sem);
237 need_cp = need_do_checkpoint(inode);
238 up_read(&F2FS_I(inode)->i_sem);
239
240 if (need_cp) {
241 /* all the dirty node pages should be flushed for POR */
242 ret = f2fs_sync_fs(inode->i_sb, 1);
243
244 /*
245 * We've secured consistency through sync_fs. Following pino
246 * will be used only for fsynced inodes after checkpoint.
247 */
248 try_to_fix_pino(inode);
249 clear_inode_flag(inode, FI_APPEND_WRITE);
250 clear_inode_flag(inode, FI_UPDATE_WRITE);
251 goto out;
252 }
253sync_nodes:
254 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
255 if (ret)
256 goto out;
257
258 /* if cp_error was enabled, we should avoid infinite loop */
259 if (unlikely(f2fs_cp_error(sbi))) {
260 ret = -EIO;
261 goto out;
262 }
263
264 if (need_inode_block_update(sbi, ino)) {
265 f2fs_mark_inode_dirty_sync(inode, true);
266 f2fs_write_inode(inode, NULL);
267 goto sync_nodes;
268 }
269
270 ret = wait_on_node_pages_writeback(sbi, ino);
271 if (ret)
272 goto out;
273
274 /* once recovery info is written, don't need to tack this */
275 remove_ino_entry(sbi, ino, APPEND_INO);
276 clear_inode_flag(inode, FI_APPEND_WRITE);
277flush_out:
278 remove_ino_entry(sbi, ino, UPDATE_INO);
279 clear_inode_flag(inode, FI_UPDATE_WRITE);
280 ret = f2fs_issue_flush(sbi);
281 f2fs_update_time(sbi, REQ_TIME);
282out:
283 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
284 f2fs_trace_ios(NULL, 1);
285 return ret;
286}
287
288int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
289{
290 return f2fs_do_sync_file(file, start, end, datasync, false);
291}
292
293static pgoff_t __get_first_dirty_index(struct address_space *mapping,
294 pgoff_t pgofs, int whence)
295{
296 struct pagevec pvec;
297 int nr_pages;
298
299 if (whence != SEEK_DATA)
300 return 0;
301
302 /* find first dirty page index */
303 pagevec_init(&pvec, 0);
304 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
305 PAGECACHE_TAG_DIRTY, 1);
306 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
307 pagevec_release(&pvec);
308 return pgofs;
309}
310
311static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
312 int whence)
313{
314 switch (whence) {
315 case SEEK_DATA:
316 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
317 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
318 return true;
319 break;
320 case SEEK_HOLE:
321 if (blkaddr == NULL_ADDR)
322 return true;
323 break;
324 }
325 return false;
326}
327
328static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
329{
330 struct inode *inode = file->f_mapping->host;
331 loff_t maxbytes = inode->i_sb->s_maxbytes;
332 struct dnode_of_data dn;
333 pgoff_t pgofs, end_offset, dirty;
334 loff_t data_ofs = offset;
335 loff_t isize;
336 int err = 0;
337
338 inode_lock(inode);
339
340 isize = i_size_read(inode);
341 if (offset >= isize)
342 goto fail;
343
344 /* handle inline data case */
345 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
346 if (whence == SEEK_HOLE)
347 data_ofs = isize;
348 goto found;
349 }
350
351 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
352
353 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
354
355 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
356 set_new_dnode(&dn, inode, NULL, NULL, 0);
357 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
358 if (err && err != -ENOENT) {
359 goto fail;
360 } else if (err == -ENOENT) {
361 /* direct node does not exists */
362 if (whence == SEEK_DATA) {
363 pgofs = get_next_page_offset(&dn, pgofs);
364 continue;
365 } else {
366 goto found;
367 }
368 }
369
370 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
371
372 /* find data/hole in dnode block */
373 for (; dn.ofs_in_node < end_offset;
374 dn.ofs_in_node++, pgofs++,
375 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
376 block_t blkaddr;
377 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
378
379 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
380 f2fs_put_dnode(&dn);
381 goto found;
382 }
383 }
384 f2fs_put_dnode(&dn);
385 }
386
387 if (whence == SEEK_DATA)
388 goto fail;
389found:
390 if (whence == SEEK_HOLE && data_ofs > isize)
391 data_ofs = isize;
392 inode_unlock(inode);
393 return vfs_setpos(file, data_ofs, maxbytes);
394fail:
395 inode_unlock(inode);
396 return -ENXIO;
397}
398
399static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
400{
401 struct inode *inode = file->f_mapping->host;
402 loff_t maxbytes = inode->i_sb->s_maxbytes;
403
404 switch (whence) {
405 case SEEK_SET:
406 case SEEK_CUR:
407 case SEEK_END:
408 return generic_file_llseek_size(file, offset, whence,
409 maxbytes, i_size_read(inode));
410 case SEEK_DATA:
411 case SEEK_HOLE:
412 if (offset < 0)
413 return -ENXIO;
414 return f2fs_seek_block(file, offset, whence);
415 }
416
417 return -EINVAL;
418}
419
420static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
421{
422 struct inode *inode = file_inode(file);
423 int err;
424
425 if (f2fs_encrypted_inode(inode)) {
426 err = fscrypt_get_encryption_info(inode);
427 if (err)
428 return 0;
429 if (!f2fs_encrypted_inode(inode))
430 return -ENOKEY;
431 }
432
433 /* we don't need to use inline_data strictly */
434 err = f2fs_convert_inline_inode(inode);
435 if (err)
436 return err;
437
438 file_accessed(file);
439 vma->vm_ops = &f2fs_file_vm_ops;
440 return 0;
441}
442
443static int f2fs_file_open(struct inode *inode, struct file *filp)
444{
445 int ret = generic_file_open(inode, filp);
446 struct dentry *dir;
447
448 if (!ret && f2fs_encrypted_inode(inode)) {
449 ret = fscrypt_get_encryption_info(inode);
450 if (ret)
451 return -EACCES;
452 if (!fscrypt_has_encryption_key(inode))
453 return -ENOKEY;
454 }
455 dir = dget_parent(file_dentry(filp));
456 if (f2fs_encrypted_inode(d_inode(dir)) &&
457 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
458 dput(dir);
459 return -EPERM;
460 }
461 dput(dir);
462 return ret;
463}
464
465int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
466{
467 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
468 struct f2fs_node *raw_node;
469 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
470 __le32 *addr;
471
472 raw_node = F2FS_NODE(dn->node_page);
473 addr = blkaddr_in_node(raw_node) + ofs;
474
475 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
476 block_t blkaddr = le32_to_cpu(*addr);
477 if (blkaddr == NULL_ADDR)
478 continue;
479
480 dn->data_blkaddr = NULL_ADDR;
481 set_data_blkaddr(dn);
482 invalidate_blocks(sbi, blkaddr);
483 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
484 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
485 nr_free++;
486 }
487
488 if (nr_free) {
489 pgoff_t fofs;
490 /*
491 * once we invalidate valid blkaddr in range [ofs, ofs + count],
492 * we will invalidate all blkaddr in the whole range.
493 */
494 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
495 dn->inode) + ofs;
496 f2fs_update_extent_cache_range(dn, fofs, 0, len);
497 dec_valid_block_count(sbi, dn->inode, nr_free);
498 }
499 dn->ofs_in_node = ofs;
500
501 f2fs_update_time(sbi, REQ_TIME);
502 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
503 dn->ofs_in_node, nr_free);
504 return nr_free;
505}
506
507void truncate_data_blocks(struct dnode_of_data *dn)
508{
509 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
510}
511
512static int truncate_partial_data_page(struct inode *inode, u64 from,
513 bool cache_only)
514{
515 unsigned offset = from & (PAGE_SIZE - 1);
516 pgoff_t index = from >> PAGE_SHIFT;
517 struct address_space *mapping = inode->i_mapping;
518 struct page *page;
519
520 if (!offset && !cache_only)
521 return 0;
522
523 if (cache_only) {
524 page = find_lock_page(mapping, index);
525 if (page && PageUptodate(page))
526 goto truncate_out;
527 f2fs_put_page(page, 1);
528 return 0;
529 }
530
531 page = get_lock_data_page(inode, index, true);
532 if (IS_ERR(page))
533 return 0;
534truncate_out:
535 f2fs_wait_on_page_writeback(page, DATA, true);
536 zero_user(page, offset, PAGE_SIZE - offset);
537 if (!cache_only || !f2fs_encrypted_inode(inode) ||
538 !S_ISREG(inode->i_mode))
539 set_page_dirty(page);
540 f2fs_put_page(page, 1);
541 return 0;
542}
543
544int truncate_blocks(struct inode *inode, u64 from, bool lock)
545{
546 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
547 unsigned int blocksize = inode->i_sb->s_blocksize;
548 struct dnode_of_data dn;
549 pgoff_t free_from;
550 int count = 0, err = 0;
551 struct page *ipage;
552 bool truncate_page = false;
553
554 trace_f2fs_truncate_blocks_enter(inode, from);
555
556 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
557
558 if (free_from >= sbi->max_file_blocks)
559 goto free_partial;
560
561 if (lock)
562 f2fs_lock_op(sbi);
563
564 ipage = get_node_page(sbi, inode->i_ino);
565 if (IS_ERR(ipage)) {
566 err = PTR_ERR(ipage);
567 goto out;
568 }
569
570 if (f2fs_has_inline_data(inode)) {
571 if (truncate_inline_inode(ipage, from))
572 set_page_dirty(ipage);
573 f2fs_put_page(ipage, 1);
574 truncate_page = true;
575 goto out;
576 }
577
578 set_new_dnode(&dn, inode, ipage, NULL, 0);
579 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
580 if (err) {
581 if (err == -ENOENT)
582 goto free_next;
583 goto out;
584 }
585
586 count = ADDRS_PER_PAGE(dn.node_page, inode);
587
588 count -= dn.ofs_in_node;
589 f2fs_bug_on(sbi, count < 0);
590
591 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
592 truncate_data_blocks_range(&dn, count);
593 free_from += count;
594 }
595
596 f2fs_put_dnode(&dn);
597free_next:
598 err = truncate_inode_blocks(inode, free_from);
599out:
600 if (lock)
601 f2fs_unlock_op(sbi);
602free_partial:
603 /* lastly zero out the first data page */
604 if (!err)
605 err = truncate_partial_data_page(inode, from, truncate_page);
606
607 trace_f2fs_truncate_blocks_exit(inode, err);
608 return err;
609}
610
611int f2fs_truncate(struct inode *inode)
612{
613 int err;
614
615 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
616 S_ISLNK(inode->i_mode)))
617 return 0;
618
619 trace_f2fs_truncate(inode);
620
621 /* we should check inline_data size */
622 if (!f2fs_may_inline_data(inode)) {
623 err = f2fs_convert_inline_inode(inode);
624 if (err)
625 return err;
626 }
627
628 err = truncate_blocks(inode, i_size_read(inode), true);
629 if (err)
630 return err;
631
632 inode->i_mtime = inode->i_ctime = current_time(inode);
633 f2fs_mark_inode_dirty_sync(inode, false);
634 return 0;
635}
636
637int f2fs_getattr(struct vfsmount *mnt,
638 struct dentry *dentry, struct kstat *stat)
639{
640 struct inode *inode = d_inode(dentry);
641 generic_fillattr(inode, stat);
642 stat->blocks <<= 3;
643 return 0;
644}
645
646#ifdef CONFIG_F2FS_FS_POSIX_ACL
647static void __setattr_copy(struct inode *inode, const struct iattr *attr)
648{
649 unsigned int ia_valid = attr->ia_valid;
650
651 if (ia_valid & ATTR_UID)
652 inode->i_uid = attr->ia_uid;
653 if (ia_valid & ATTR_GID)
654 inode->i_gid = attr->ia_gid;
655 if (ia_valid & ATTR_ATIME)
656 inode->i_atime = timespec_trunc(attr->ia_atime,
657 inode->i_sb->s_time_gran);
658 if (ia_valid & ATTR_MTIME)
659 inode->i_mtime = timespec_trunc(attr->ia_mtime,
660 inode->i_sb->s_time_gran);
661 if (ia_valid & ATTR_CTIME)
662 inode->i_ctime = timespec_trunc(attr->ia_ctime,
663 inode->i_sb->s_time_gran);
664 if (ia_valid & ATTR_MODE) {
665 umode_t mode = attr->ia_mode;
666
667 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
668 mode &= ~S_ISGID;
669 set_acl_inode(inode, mode);
670 }
671}
672#else
673#define __setattr_copy setattr_copy
674#endif
675
676int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
677{
678 struct inode *inode = d_inode(dentry);
679 int err;
680 bool size_changed = false;
681
682 err = setattr_prepare(dentry, attr);
683 if (err)
684 return err;
685
686 if (attr->ia_valid & ATTR_SIZE) {
687 if (f2fs_encrypted_inode(inode) &&
688 fscrypt_get_encryption_info(inode))
689 return -EACCES;
690
691 if (attr->ia_size <= i_size_read(inode)) {
692 truncate_setsize(inode, attr->ia_size);
693 err = f2fs_truncate(inode);
694 if (err)
695 return err;
696 } else {
697 /*
698 * do not trim all blocks after i_size if target size is
699 * larger than i_size.
700 */
701 truncate_setsize(inode, attr->ia_size);
702
703 /* should convert inline inode here */
704 if (!f2fs_may_inline_data(inode)) {
705 err = f2fs_convert_inline_inode(inode);
706 if (err)
707 return err;
708 }
709 inode->i_mtime = inode->i_ctime = current_time(inode);
710 }
711
712 size_changed = true;
713 }
714
715 __setattr_copy(inode, attr);
716
717 if (attr->ia_valid & ATTR_MODE) {
718 err = posix_acl_chmod(inode, get_inode_mode(inode));
719 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
720 inode->i_mode = F2FS_I(inode)->i_acl_mode;
721 clear_inode_flag(inode, FI_ACL_MODE);
722 }
723 }
724
725 /* file size may changed here */
726 f2fs_mark_inode_dirty_sync(inode, size_changed);
727
728 /* inode change will produce dirty node pages flushed by checkpoint */
729 f2fs_balance_fs(F2FS_I_SB(inode), true);
730
731 return err;
732}
733
734const struct inode_operations f2fs_file_inode_operations = {
735 .getattr = f2fs_getattr,
736 .setattr = f2fs_setattr,
737 .get_acl = f2fs_get_acl,
738 .set_acl = f2fs_set_acl,
739#ifdef CONFIG_F2FS_FS_XATTR
740 .listxattr = f2fs_listxattr,
741#endif
742 .fiemap = f2fs_fiemap,
743};
744
745static int fill_zero(struct inode *inode, pgoff_t index,
746 loff_t start, loff_t len)
747{
748 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
749 struct page *page;
750
751 if (!len)
752 return 0;
753
754 f2fs_balance_fs(sbi, true);
755
756 f2fs_lock_op(sbi);
757 page = get_new_data_page(inode, NULL, index, false);
758 f2fs_unlock_op(sbi);
759
760 if (IS_ERR(page))
761 return PTR_ERR(page);
762
763 f2fs_wait_on_page_writeback(page, DATA, true);
764 zero_user(page, start, len);
765 set_page_dirty(page);
766 f2fs_put_page(page, 1);
767 return 0;
768}
769
770int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
771{
772 int err;
773
774 while (pg_start < pg_end) {
775 struct dnode_of_data dn;
776 pgoff_t end_offset, count;
777
778 set_new_dnode(&dn, inode, NULL, NULL, 0);
779 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
780 if (err) {
781 if (err == -ENOENT) {
782 pg_start++;
783 continue;
784 }
785 return err;
786 }
787
788 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
789 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
790
791 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
792
793 truncate_data_blocks_range(&dn, count);
794 f2fs_put_dnode(&dn);
795
796 pg_start += count;
797 }
798 return 0;
799}
800
801static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
802{
803 pgoff_t pg_start, pg_end;
804 loff_t off_start, off_end;
805 int ret;
806
807 ret = f2fs_convert_inline_inode(inode);
808 if (ret)
809 return ret;
810
811 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
812 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
813
814 off_start = offset & (PAGE_SIZE - 1);
815 off_end = (offset + len) & (PAGE_SIZE - 1);
816
817 if (pg_start == pg_end) {
818 ret = fill_zero(inode, pg_start, off_start,
819 off_end - off_start);
820 if (ret)
821 return ret;
822 } else {
823 if (off_start) {
824 ret = fill_zero(inode, pg_start++, off_start,
825 PAGE_SIZE - off_start);
826 if (ret)
827 return ret;
828 }
829 if (off_end) {
830 ret = fill_zero(inode, pg_end, 0, off_end);
831 if (ret)
832 return ret;
833 }
834
835 if (pg_start < pg_end) {
836 struct address_space *mapping = inode->i_mapping;
837 loff_t blk_start, blk_end;
838 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
839
840 f2fs_balance_fs(sbi, true);
841
842 blk_start = (loff_t)pg_start << PAGE_SHIFT;
843 blk_end = (loff_t)pg_end << PAGE_SHIFT;
844 truncate_inode_pages_range(mapping, blk_start,
845 blk_end - 1);
846
847 f2fs_lock_op(sbi);
848 ret = truncate_hole(inode, pg_start, pg_end);
849 f2fs_unlock_op(sbi);
850 }
851 }
852
853 return ret;
854}
855
856static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
857 int *do_replace, pgoff_t off, pgoff_t len)
858{
859 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
860 struct dnode_of_data dn;
861 int ret, done, i;
862
863next_dnode:
864 set_new_dnode(&dn, inode, NULL, NULL, 0);
865 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
866 if (ret && ret != -ENOENT) {
867 return ret;
868 } else if (ret == -ENOENT) {
869 if (dn.max_level == 0)
870 return -ENOENT;
871 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
872 blkaddr += done;
873 do_replace += done;
874 goto next;
875 }
876
877 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
878 dn.ofs_in_node, len);
879 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
880 *blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
881 if (!is_checkpointed_data(sbi, *blkaddr)) {
882
883 if (test_opt(sbi, LFS)) {
884 f2fs_put_dnode(&dn);
885 return -ENOTSUPP;
886 }
887
888 /* do not invalidate this block address */
889 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
890 *do_replace = 1;
891 }
892 }
893 f2fs_put_dnode(&dn);
894next:
895 len -= done;
896 off += done;
897 if (len)
898 goto next_dnode;
899 return 0;
900}
901
902static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
903 int *do_replace, pgoff_t off, int len)
904{
905 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
906 struct dnode_of_data dn;
907 int ret, i;
908
909 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
910 if (*do_replace == 0)
911 continue;
912
913 set_new_dnode(&dn, inode, NULL, NULL, 0);
914 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
915 if (ret) {
916 dec_valid_block_count(sbi, inode, 1);
917 invalidate_blocks(sbi, *blkaddr);
918 } else {
919 f2fs_update_data_blkaddr(&dn, *blkaddr);
920 }
921 f2fs_put_dnode(&dn);
922 }
923 return 0;
924}
925
926static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
927 block_t *blkaddr, int *do_replace,
928 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
929{
930 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
931 pgoff_t i = 0;
932 int ret;
933
934 while (i < len) {
935 if (blkaddr[i] == NULL_ADDR && !full) {
936 i++;
937 continue;
938 }
939
940 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
941 struct dnode_of_data dn;
942 struct node_info ni;
943 size_t new_size;
944 pgoff_t ilen;
945
946 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
947 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
948 if (ret)
949 return ret;
950
951 get_node_info(sbi, dn.nid, &ni);
952 ilen = min((pgoff_t)
953 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
954 dn.ofs_in_node, len - i);
955 do {
956 dn.data_blkaddr = datablock_addr(dn.node_page,
957 dn.ofs_in_node);
958 truncate_data_blocks_range(&dn, 1);
959
960 if (do_replace[i]) {
961 f2fs_i_blocks_write(src_inode,
962 1, false);
963 f2fs_i_blocks_write(dst_inode,
964 1, true);
965 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
966 blkaddr[i], ni.version, true, false);
967
968 do_replace[i] = 0;
969 }
970 dn.ofs_in_node++;
971 i++;
972 new_size = (dst + i) << PAGE_SHIFT;
973 if (dst_inode->i_size < new_size)
974 f2fs_i_size_write(dst_inode, new_size);
975 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
976
977 f2fs_put_dnode(&dn);
978 } else {
979 struct page *psrc, *pdst;
980
981 psrc = get_lock_data_page(src_inode, src + i, true);
982 if (IS_ERR(psrc))
983 return PTR_ERR(psrc);
984 pdst = get_new_data_page(dst_inode, NULL, dst + i,
985 true);
986 if (IS_ERR(pdst)) {
987 f2fs_put_page(psrc, 1);
988 return PTR_ERR(pdst);
989 }
990 f2fs_copy_page(psrc, pdst);
991 set_page_dirty(pdst);
992 f2fs_put_page(pdst, 1);
993 f2fs_put_page(psrc, 1);
994
995 ret = truncate_hole(src_inode, src + i, src + i + 1);
996 if (ret)
997 return ret;
998 i++;
999 }
1000 }
1001 return 0;
1002}
1003
1004static int __exchange_data_block(struct inode *src_inode,
1005 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1006 pgoff_t len, bool full)
1007{
1008 block_t *src_blkaddr;
1009 int *do_replace;
1010 pgoff_t olen;
1011 int ret;
1012
1013 while (len) {
1014 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1015
1016 src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1017 if (!src_blkaddr)
1018 return -ENOMEM;
1019
1020 do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1021 if (!do_replace) {
1022 kvfree(src_blkaddr);
1023 return -ENOMEM;
1024 }
1025
1026 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1027 do_replace, src, olen);
1028 if (ret)
1029 goto roll_back;
1030
1031 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1032 do_replace, src, dst, olen, full);
1033 if (ret)
1034 goto roll_back;
1035
1036 src += olen;
1037 dst += olen;
1038 len -= olen;
1039
1040 kvfree(src_blkaddr);
1041 kvfree(do_replace);
1042 }
1043 return 0;
1044
1045roll_back:
1046 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1047 kvfree(src_blkaddr);
1048 kvfree(do_replace);
1049 return ret;
1050}
1051
1052static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1053{
1054 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1055 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1056 int ret;
1057
1058 f2fs_balance_fs(sbi, true);
1059 f2fs_lock_op(sbi);
1060
1061 f2fs_drop_extent_tree(inode);
1062
1063 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1064 f2fs_unlock_op(sbi);
1065 return ret;
1066}
1067
1068static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1069{
1070 pgoff_t pg_start, pg_end;
1071 loff_t new_size;
1072 int ret;
1073
1074 if (offset + len >= i_size_read(inode))
1075 return -EINVAL;
1076
1077 /* collapse range should be aligned to block size of f2fs. */
1078 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1079 return -EINVAL;
1080
1081 ret = f2fs_convert_inline_inode(inode);
1082 if (ret)
1083 return ret;
1084
1085 pg_start = offset >> PAGE_SHIFT;
1086 pg_end = (offset + len) >> PAGE_SHIFT;
1087
1088 /* write out all dirty pages from offset */
1089 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1090 if (ret)
1091 return ret;
1092
1093 truncate_pagecache(inode, offset);
1094
1095 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1096 if (ret)
1097 return ret;
1098
1099 /* write out all moved pages, if possible */
1100 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1101 truncate_pagecache(inode, offset);
1102
1103 new_size = i_size_read(inode) - len;
1104 truncate_pagecache(inode, new_size);
1105
1106 ret = truncate_blocks(inode, new_size, true);
1107 if (!ret)
1108 f2fs_i_size_write(inode, new_size);
1109
1110 return ret;
1111}
1112
1113static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1114 pgoff_t end)
1115{
1116 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1117 pgoff_t index = start;
1118 unsigned int ofs_in_node = dn->ofs_in_node;
1119 blkcnt_t count = 0;
1120 int ret;
1121
1122 for (; index < end; index++, dn->ofs_in_node++) {
1123 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1124 count++;
1125 }
1126
1127 dn->ofs_in_node = ofs_in_node;
1128 ret = reserve_new_blocks(dn, count);
1129 if (ret)
1130 return ret;
1131
1132 dn->ofs_in_node = ofs_in_node;
1133 for (index = start; index < end; index++, dn->ofs_in_node++) {
1134 dn->data_blkaddr =
1135 datablock_addr(dn->node_page, dn->ofs_in_node);
1136 /*
1137 * reserve_new_blocks will not guarantee entire block
1138 * allocation.
1139 */
1140 if (dn->data_blkaddr == NULL_ADDR) {
1141 ret = -ENOSPC;
1142 break;
1143 }
1144 if (dn->data_blkaddr != NEW_ADDR) {
1145 invalidate_blocks(sbi, dn->data_blkaddr);
1146 dn->data_blkaddr = NEW_ADDR;
1147 set_data_blkaddr(dn);
1148 }
1149 }
1150
1151 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1152
1153 return ret;
1154}
1155
1156static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1157 int mode)
1158{
1159 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1160 struct address_space *mapping = inode->i_mapping;
1161 pgoff_t index, pg_start, pg_end;
1162 loff_t new_size = i_size_read(inode);
1163 loff_t off_start, off_end;
1164 int ret = 0;
1165
1166 ret = inode_newsize_ok(inode, (len + offset));
1167 if (ret)
1168 return ret;
1169
1170 ret = f2fs_convert_inline_inode(inode);
1171 if (ret)
1172 return ret;
1173
1174 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1175 if (ret)
1176 return ret;
1177
1178 truncate_pagecache_range(inode, offset, offset + len - 1);
1179
1180 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1181 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1182
1183 off_start = offset & (PAGE_SIZE - 1);
1184 off_end = (offset + len) & (PAGE_SIZE - 1);
1185
1186 if (pg_start == pg_end) {
1187 ret = fill_zero(inode, pg_start, off_start,
1188 off_end - off_start);
1189 if (ret)
1190 return ret;
1191
1192 if (offset + len > new_size)
1193 new_size = offset + len;
1194 new_size = max_t(loff_t, new_size, offset + len);
1195 } else {
1196 if (off_start) {
1197 ret = fill_zero(inode, pg_start++, off_start,
1198 PAGE_SIZE - off_start);
1199 if (ret)
1200 return ret;
1201
1202 new_size = max_t(loff_t, new_size,
1203 (loff_t)pg_start << PAGE_SHIFT);
1204 }
1205
1206 for (index = pg_start; index < pg_end;) {
1207 struct dnode_of_data dn;
1208 unsigned int end_offset;
1209 pgoff_t end;
1210
1211 f2fs_lock_op(sbi);
1212
1213 set_new_dnode(&dn, inode, NULL, NULL, 0);
1214 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1215 if (ret) {
1216 f2fs_unlock_op(sbi);
1217 goto out;
1218 }
1219
1220 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1221 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1222
1223 ret = f2fs_do_zero_range(&dn, index, end);
1224 f2fs_put_dnode(&dn);
1225 f2fs_unlock_op(sbi);
1226
1227 f2fs_balance_fs(sbi, dn.node_changed);
1228
1229 if (ret)
1230 goto out;
1231
1232 index = end;
1233 new_size = max_t(loff_t, new_size,
1234 (loff_t)index << PAGE_SHIFT);
1235 }
1236
1237 if (off_end) {
1238 ret = fill_zero(inode, pg_end, 0, off_end);
1239 if (ret)
1240 goto out;
1241
1242 new_size = max_t(loff_t, new_size, offset + len);
1243 }
1244 }
1245
1246out:
1247 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1248 f2fs_i_size_write(inode, new_size);
1249
1250 return ret;
1251}
1252
1253static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1254{
1255 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1256 pgoff_t nr, pg_start, pg_end, delta, idx;
1257 loff_t new_size;
1258 int ret = 0;
1259
1260 new_size = i_size_read(inode) + len;
1261 if (new_size > inode->i_sb->s_maxbytes)
1262 return -EFBIG;
1263
1264 if (offset >= i_size_read(inode))
1265 return -EINVAL;
1266
1267 /* insert range should be aligned to block size of f2fs. */
1268 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1269 return -EINVAL;
1270
1271 ret = f2fs_convert_inline_inode(inode);
1272 if (ret)
1273 return ret;
1274
1275 f2fs_balance_fs(sbi, true);
1276
1277 ret = truncate_blocks(inode, i_size_read(inode), true);
1278 if (ret)
1279 return ret;
1280
1281 /* write out all dirty pages from offset */
1282 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1283 if (ret)
1284 return ret;
1285
1286 truncate_pagecache(inode, offset);
1287
1288 pg_start = offset >> PAGE_SHIFT;
1289 pg_end = (offset + len) >> PAGE_SHIFT;
1290 delta = pg_end - pg_start;
1291 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1292
1293 while (!ret && idx > pg_start) {
1294 nr = idx - pg_start;
1295 if (nr > delta)
1296 nr = delta;
1297 idx -= nr;
1298
1299 f2fs_lock_op(sbi);
1300 f2fs_drop_extent_tree(inode);
1301
1302 ret = __exchange_data_block(inode, inode, idx,
1303 idx + delta, nr, false);
1304 f2fs_unlock_op(sbi);
1305 }
1306
1307 /* write out all moved pages, if possible */
1308 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1309 truncate_pagecache(inode, offset);
1310
1311 if (!ret)
1312 f2fs_i_size_write(inode, new_size);
1313 return ret;
1314}
1315
1316static int expand_inode_data(struct inode *inode, loff_t offset,
1317 loff_t len, int mode)
1318{
1319 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1320 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1321 pgoff_t pg_end;
1322 loff_t new_size = i_size_read(inode);
1323 loff_t off_end;
1324 int err;
1325
1326 err = inode_newsize_ok(inode, (len + offset));
1327 if (err)
1328 return err;
1329
1330 err = f2fs_convert_inline_inode(inode);
1331 if (err)
1332 return err;
1333
1334 f2fs_balance_fs(sbi, true);
1335
1336 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1337 off_end = (offset + len) & (PAGE_SIZE - 1);
1338
1339 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1340 map.m_len = pg_end - map.m_lblk;
1341 if (off_end)
1342 map.m_len++;
1343
1344 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1345 if (err) {
1346 pgoff_t last_off;
1347
1348 if (!map.m_len)
1349 return err;
1350
1351 last_off = map.m_lblk + map.m_len - 1;
1352
1353 /* update new size to the failed position */
1354 new_size = (last_off == pg_end) ? offset + len:
1355 (loff_t)(last_off + 1) << PAGE_SHIFT;
1356 } else {
1357 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1358 }
1359
1360 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1361 f2fs_i_size_write(inode, new_size);
1362
1363 return err;
1364}
1365
1366static long f2fs_fallocate(struct file *file, int mode,
1367 loff_t offset, loff_t len)
1368{
1369 struct inode *inode = file_inode(file);
1370 long ret = 0;
1371
1372 /* f2fs only support ->fallocate for regular file */
1373 if (!S_ISREG(inode->i_mode))
1374 return -EINVAL;
1375
1376 if (f2fs_encrypted_inode(inode) &&
1377 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1378 return -EOPNOTSUPP;
1379
1380 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1381 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1382 FALLOC_FL_INSERT_RANGE))
1383 return -EOPNOTSUPP;
1384
1385 inode_lock(inode);
1386
1387 if (mode & FALLOC_FL_PUNCH_HOLE) {
1388 if (offset >= inode->i_size)
1389 goto out;
1390
1391 ret = punch_hole(inode, offset, len);
1392 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1393 ret = f2fs_collapse_range(inode, offset, len);
1394 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1395 ret = f2fs_zero_range(inode, offset, len, mode);
1396 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1397 ret = f2fs_insert_range(inode, offset, len);
1398 } else {
1399 ret = expand_inode_data(inode, offset, len, mode);
1400 }
1401
1402 if (!ret) {
1403 inode->i_mtime = inode->i_ctime = current_time(inode);
1404 f2fs_mark_inode_dirty_sync(inode, false);
1405 if (mode & FALLOC_FL_KEEP_SIZE)
1406 file_set_keep_isize(inode);
1407 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1408 }
1409
1410out:
1411 inode_unlock(inode);
1412
1413 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1414 return ret;
1415}
1416
1417static int f2fs_release_file(struct inode *inode, struct file *filp)
1418{
1419 /*
1420 * f2fs_relase_file is called at every close calls. So we should
1421 * not drop any inmemory pages by close called by other process.
1422 */
1423 if (!(filp->f_mode & FMODE_WRITE) ||
1424 atomic_read(&inode->i_writecount) != 1)
1425 return 0;
1426
1427 /* some remained atomic pages should discarded */
1428 if (f2fs_is_atomic_file(inode))
1429 drop_inmem_pages(inode);
1430 if (f2fs_is_volatile_file(inode)) {
1431 clear_inode_flag(inode, FI_VOLATILE_FILE);
1432 set_inode_flag(inode, FI_DROP_CACHE);
1433 filemap_fdatawrite(inode->i_mapping);
1434 clear_inode_flag(inode, FI_DROP_CACHE);
1435 }
1436 return 0;
1437}
1438
1439#define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1440#define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1441
1442static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1443{
1444 if (S_ISDIR(mode))
1445 return flags;
1446 else if (S_ISREG(mode))
1447 return flags & F2FS_REG_FLMASK;
1448 else
1449 return flags & F2FS_OTHER_FLMASK;
1450}
1451
1452static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1453{
1454 struct inode *inode = file_inode(filp);
1455 struct f2fs_inode_info *fi = F2FS_I(inode);
1456 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1457 return put_user(flags, (int __user *)arg);
1458}
1459
1460static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1461{
1462 struct inode *inode = file_inode(filp);
1463 struct f2fs_inode_info *fi = F2FS_I(inode);
1464 unsigned int flags;
1465 unsigned int oldflags;
1466 int ret;
1467
1468 if (!inode_owner_or_capable(inode))
1469 return -EACCES;
1470
1471 if (get_user(flags, (int __user *)arg))
1472 return -EFAULT;
1473
1474 ret = mnt_want_write_file(filp);
1475 if (ret)
1476 return ret;
1477
1478 flags = f2fs_mask_flags(inode->i_mode, flags);
1479
1480 inode_lock(inode);
1481
1482 oldflags = fi->i_flags;
1483
1484 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1485 if (!capable(CAP_LINUX_IMMUTABLE)) {
1486 inode_unlock(inode);
1487 ret = -EPERM;
1488 goto out;
1489 }
1490 }
1491
1492 flags = flags & FS_FL_USER_MODIFIABLE;
1493 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1494 fi->i_flags = flags;
1495 inode_unlock(inode);
1496
1497 inode->i_ctime = current_time(inode);
1498 f2fs_set_inode_flags(inode);
1499out:
1500 mnt_drop_write_file(filp);
1501 return ret;
1502}
1503
1504static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1505{
1506 struct inode *inode = file_inode(filp);
1507
1508 return put_user(inode->i_generation, (int __user *)arg);
1509}
1510
1511static int f2fs_ioc_start_atomic_write(struct file *filp)
1512{
1513 struct inode *inode = file_inode(filp);
1514 int ret;
1515
1516 if (!inode_owner_or_capable(inode))
1517 return -EACCES;
1518
1519 ret = mnt_want_write_file(filp);
1520 if (ret)
1521 return ret;
1522
1523 inode_lock(inode);
1524
1525 if (f2fs_is_atomic_file(inode))
1526 goto out;
1527
1528 ret = f2fs_convert_inline_inode(inode);
1529 if (ret)
1530 goto out;
1531
1532 set_inode_flag(inode, FI_ATOMIC_FILE);
1533 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1534
1535 if (!get_dirty_pages(inode))
1536 goto out;
1537
1538 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1539 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1540 inode->i_ino, get_dirty_pages(inode));
1541 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1542 if (ret)
1543 clear_inode_flag(inode, FI_ATOMIC_FILE);
1544out:
1545 inode_unlock(inode);
1546 mnt_drop_write_file(filp);
1547 return ret;
1548}
1549
1550static int f2fs_ioc_commit_atomic_write(struct file *filp)
1551{
1552 struct inode *inode = file_inode(filp);
1553 int ret;
1554
1555 if (!inode_owner_or_capable(inode))
1556 return -EACCES;
1557
1558 ret = mnt_want_write_file(filp);
1559 if (ret)
1560 return ret;
1561
1562 inode_lock(inode);
1563
1564 if (f2fs_is_volatile_file(inode))
1565 goto err_out;
1566
1567 if (f2fs_is_atomic_file(inode)) {
1568 clear_inode_flag(inode, FI_ATOMIC_FILE);
1569 ret = commit_inmem_pages(inode);
1570 if (ret) {
1571 set_inode_flag(inode, FI_ATOMIC_FILE);
1572 goto err_out;
1573 }
1574 }
1575
1576 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1577err_out:
1578 inode_unlock(inode);
1579 mnt_drop_write_file(filp);
1580 return ret;
1581}
1582
1583static int f2fs_ioc_start_volatile_write(struct file *filp)
1584{
1585 struct inode *inode = file_inode(filp);
1586 int ret;
1587
1588 if (!inode_owner_or_capable(inode))
1589 return -EACCES;
1590
1591 ret = mnt_want_write_file(filp);
1592 if (ret)
1593 return ret;
1594
1595 inode_lock(inode);
1596
1597 if (f2fs_is_volatile_file(inode))
1598 goto out;
1599
1600 ret = f2fs_convert_inline_inode(inode);
1601 if (ret)
1602 goto out;
1603
1604 set_inode_flag(inode, FI_VOLATILE_FILE);
1605 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1606out:
1607 inode_unlock(inode);
1608 mnt_drop_write_file(filp);
1609 return ret;
1610}
1611
1612static int f2fs_ioc_release_volatile_write(struct file *filp)
1613{
1614 struct inode *inode = file_inode(filp);
1615 int ret;
1616
1617 if (!inode_owner_or_capable(inode))
1618 return -EACCES;
1619
1620 ret = mnt_want_write_file(filp);
1621 if (ret)
1622 return ret;
1623
1624 inode_lock(inode);
1625
1626 if (!f2fs_is_volatile_file(inode))
1627 goto out;
1628
1629 if (!f2fs_is_first_block_written(inode)) {
1630 ret = truncate_partial_data_page(inode, 0, true);
1631 goto out;
1632 }
1633
1634 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1635out:
1636 inode_unlock(inode);
1637 mnt_drop_write_file(filp);
1638 return ret;
1639}
1640
1641static int f2fs_ioc_abort_volatile_write(struct file *filp)
1642{
1643 struct inode *inode = file_inode(filp);
1644 int ret;
1645
1646 if (!inode_owner_or_capable(inode))
1647 return -EACCES;
1648
1649 ret = mnt_want_write_file(filp);
1650 if (ret)
1651 return ret;
1652
1653 inode_lock(inode);
1654
1655 if (f2fs_is_atomic_file(inode))
1656 drop_inmem_pages(inode);
1657 if (f2fs_is_volatile_file(inode)) {
1658 clear_inode_flag(inode, FI_VOLATILE_FILE);
1659 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1660 }
1661
1662 inode_unlock(inode);
1663
1664 mnt_drop_write_file(filp);
1665 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1666 return ret;
1667}
1668
1669static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1670{
1671 struct inode *inode = file_inode(filp);
1672 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1673 struct super_block *sb = sbi->sb;
1674 __u32 in;
1675 int ret;
1676
1677 if (!capable(CAP_SYS_ADMIN))
1678 return -EPERM;
1679
1680 if (get_user(in, (__u32 __user *)arg))
1681 return -EFAULT;
1682
1683 ret = mnt_want_write_file(filp);
1684 if (ret)
1685 return ret;
1686
1687 switch (in) {
1688 case F2FS_GOING_DOWN_FULLSYNC:
1689 sb = freeze_bdev(sb->s_bdev);
1690 if (sb && !IS_ERR(sb)) {
1691 f2fs_stop_checkpoint(sbi, false);
1692 thaw_bdev(sb->s_bdev, sb);
1693 }
1694 break;
1695 case F2FS_GOING_DOWN_METASYNC:
1696 /* do checkpoint only */
1697 f2fs_sync_fs(sb, 1);
1698 f2fs_stop_checkpoint(sbi, false);
1699 break;
1700 case F2FS_GOING_DOWN_NOSYNC:
1701 f2fs_stop_checkpoint(sbi, false);
1702 break;
1703 case F2FS_GOING_DOWN_METAFLUSH:
1704 sync_meta_pages(sbi, META, LONG_MAX);
1705 f2fs_stop_checkpoint(sbi, false);
1706 break;
1707 default:
1708 ret = -EINVAL;
1709 goto out;
1710 }
1711 f2fs_update_time(sbi, REQ_TIME);
1712out:
1713 mnt_drop_write_file(filp);
1714 return ret;
1715}
1716
1717static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1718{
1719 struct inode *inode = file_inode(filp);
1720 struct super_block *sb = inode->i_sb;
1721 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1722 struct fstrim_range range;
1723 int ret;
1724
1725 if (!capable(CAP_SYS_ADMIN))
1726 return -EPERM;
1727
1728 if (!blk_queue_discard(q))
1729 return -EOPNOTSUPP;
1730
1731 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1732 sizeof(range)))
1733 return -EFAULT;
1734
1735 ret = mnt_want_write_file(filp);
1736 if (ret)
1737 return ret;
1738
1739 range.minlen = max((unsigned int)range.minlen,
1740 q->limits.discard_granularity);
1741 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1742 mnt_drop_write_file(filp);
1743 if (ret < 0)
1744 return ret;
1745
1746 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1747 sizeof(range)))
1748 return -EFAULT;
1749 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1750 return 0;
1751}
1752
1753static bool uuid_is_nonzero(__u8 u[16])
1754{
1755 int i;
1756
1757 for (i = 0; i < 16; i++)
1758 if (u[i])
1759 return true;
1760 return false;
1761}
1762
1763static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1764{
1765 struct inode *inode = file_inode(filp);
1766
1767 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1768
1769 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1770}
1771
1772static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1773{
1774 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1775}
1776
1777static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1778{
1779 struct inode *inode = file_inode(filp);
1780 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1781 int err;
1782
1783 if (!f2fs_sb_has_crypto(inode->i_sb))
1784 return -EOPNOTSUPP;
1785
1786 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1787 goto got_it;
1788
1789 err = mnt_want_write_file(filp);
1790 if (err)
1791 return err;
1792
1793 /* update superblock with uuid */
1794 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1795
1796 err = f2fs_commit_super(sbi, false);
1797 if (err) {
1798 /* undo new data */
1799 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1800 mnt_drop_write_file(filp);
1801 return err;
1802 }
1803 mnt_drop_write_file(filp);
1804got_it:
1805 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1806 16))
1807 return -EFAULT;
1808 return 0;
1809}
1810
1811static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1812{
1813 struct inode *inode = file_inode(filp);
1814 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1815 __u32 sync;
1816 int ret;
1817
1818 if (!capable(CAP_SYS_ADMIN))
1819 return -EPERM;
1820
1821 if (get_user(sync, (__u32 __user *)arg))
1822 return -EFAULT;
1823
1824 if (f2fs_readonly(sbi->sb))
1825 return -EROFS;
1826
1827 ret = mnt_want_write_file(filp);
1828 if (ret)
1829 return ret;
1830
1831 if (!sync) {
1832 if (!mutex_trylock(&sbi->gc_mutex)) {
1833 ret = -EBUSY;
1834 goto out;
1835 }
1836 } else {
1837 mutex_lock(&sbi->gc_mutex);
1838 }
1839
1840 ret = f2fs_gc(sbi, sync, true);
1841out:
1842 mnt_drop_write_file(filp);
1843 return ret;
1844}
1845
1846static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1847{
1848 struct inode *inode = file_inode(filp);
1849 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1850 int ret;
1851
1852 if (!capable(CAP_SYS_ADMIN))
1853 return -EPERM;
1854
1855 if (f2fs_readonly(sbi->sb))
1856 return -EROFS;
1857
1858 ret = mnt_want_write_file(filp);
1859 if (ret)
1860 return ret;
1861
1862 ret = f2fs_sync_fs(sbi->sb, 1);
1863
1864 mnt_drop_write_file(filp);
1865 return ret;
1866}
1867
1868static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1869 struct file *filp,
1870 struct f2fs_defragment *range)
1871{
1872 struct inode *inode = file_inode(filp);
1873 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1874 struct extent_info ei;
1875 pgoff_t pg_start, pg_end;
1876 unsigned int blk_per_seg = sbi->blocks_per_seg;
1877 unsigned int total = 0, sec_num;
1878 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1879 block_t blk_end = 0;
1880 bool fragmented = false;
1881 int err;
1882
1883 /* if in-place-update policy is enabled, don't waste time here */
1884 if (need_inplace_update(inode))
1885 return -EINVAL;
1886
1887 pg_start = range->start >> PAGE_SHIFT;
1888 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1889
1890 f2fs_balance_fs(sbi, true);
1891
1892 inode_lock(inode);
1893
1894 /* writeback all dirty pages in the range */
1895 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1896 range->start + range->len - 1);
1897 if (err)
1898 goto out;
1899
1900 /*
1901 * lookup mapping info in extent cache, skip defragmenting if physical
1902 * block addresses are continuous.
1903 */
1904 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1905 if (ei.fofs + ei.len >= pg_end)
1906 goto out;
1907 }
1908
1909 map.m_lblk = pg_start;
1910
1911 /*
1912 * lookup mapping info in dnode page cache, skip defragmenting if all
1913 * physical block addresses are continuous even if there are hole(s)
1914 * in logical blocks.
1915 */
1916 while (map.m_lblk < pg_end) {
1917 map.m_len = pg_end - map.m_lblk;
1918 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1919 if (err)
1920 goto out;
1921
1922 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1923 map.m_lblk++;
1924 continue;
1925 }
1926
1927 if (blk_end && blk_end != map.m_pblk) {
1928 fragmented = true;
1929 break;
1930 }
1931 blk_end = map.m_pblk + map.m_len;
1932
1933 map.m_lblk += map.m_len;
1934 }
1935
1936 if (!fragmented)
1937 goto out;
1938
1939 map.m_lblk = pg_start;
1940 map.m_len = pg_end - pg_start;
1941
1942 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1943
1944 /*
1945 * make sure there are enough free section for LFS allocation, this can
1946 * avoid defragment running in SSR mode when free section are allocated
1947 * intensively
1948 */
1949 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
1950 err = -EAGAIN;
1951 goto out;
1952 }
1953
1954 while (map.m_lblk < pg_end) {
1955 pgoff_t idx;
1956 int cnt = 0;
1957
1958do_map:
1959 map.m_len = pg_end - map.m_lblk;
1960 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1961 if (err)
1962 goto clear_out;
1963
1964 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1965 map.m_lblk++;
1966 continue;
1967 }
1968
1969 set_inode_flag(inode, FI_DO_DEFRAG);
1970
1971 idx = map.m_lblk;
1972 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1973 struct page *page;
1974
1975 page = get_lock_data_page(inode, idx, true);
1976 if (IS_ERR(page)) {
1977 err = PTR_ERR(page);
1978 goto clear_out;
1979 }
1980
1981 set_page_dirty(page);
1982 f2fs_put_page(page, 1);
1983
1984 idx++;
1985 cnt++;
1986 total++;
1987 }
1988
1989 map.m_lblk = idx;
1990
1991 if (idx < pg_end && cnt < blk_per_seg)
1992 goto do_map;
1993
1994 clear_inode_flag(inode, FI_DO_DEFRAG);
1995
1996 err = filemap_fdatawrite(inode->i_mapping);
1997 if (err)
1998 goto out;
1999 }
2000clear_out:
2001 clear_inode_flag(inode, FI_DO_DEFRAG);
2002out:
2003 inode_unlock(inode);
2004 if (!err)
2005 range->len = (u64)total << PAGE_SHIFT;
2006 return err;
2007}
2008
2009static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2010{
2011 struct inode *inode = file_inode(filp);
2012 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2013 struct f2fs_defragment range;
2014 int err;
2015
2016 if (!capable(CAP_SYS_ADMIN))
2017 return -EPERM;
2018
2019 if (!S_ISREG(inode->i_mode))
2020 return -EINVAL;
2021
2022 err = mnt_want_write_file(filp);
2023 if (err)
2024 return err;
2025
2026 if (f2fs_readonly(sbi->sb)) {
2027 err = -EROFS;
2028 goto out;
2029 }
2030
2031 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2032 sizeof(range))) {
2033 err = -EFAULT;
2034 goto out;
2035 }
2036
2037 /* verify alignment of offset & size */
2038 if (range.start & (F2FS_BLKSIZE - 1) ||
2039 range.len & (F2FS_BLKSIZE - 1)) {
2040 err = -EINVAL;
2041 goto out;
2042 }
2043
2044 err = f2fs_defragment_range(sbi, filp, &range);
2045 f2fs_update_time(sbi, REQ_TIME);
2046 if (err < 0)
2047 goto out;
2048
2049 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2050 sizeof(range)))
2051 err = -EFAULT;
2052out:
2053 mnt_drop_write_file(filp);
2054 return err;
2055}
2056
2057static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2058 struct file *file_out, loff_t pos_out, size_t len)
2059{
2060 struct inode *src = file_inode(file_in);
2061 struct inode *dst = file_inode(file_out);
2062 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2063 size_t olen = len, dst_max_i_size = 0;
2064 size_t dst_osize;
2065 int ret;
2066
2067 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2068 src->i_sb != dst->i_sb)
2069 return -EXDEV;
2070
2071 if (unlikely(f2fs_readonly(src->i_sb)))
2072 return -EROFS;
2073
2074 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2075 return -EINVAL;
2076
2077 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2078 return -EOPNOTSUPP;
2079
2080 if (src == dst) {
2081 if (pos_in == pos_out)
2082 return 0;
2083 if (pos_out > pos_in && pos_out < pos_in + len)
2084 return -EINVAL;
2085 }
2086
2087 inode_lock(src);
2088 if (src != dst) {
2089 if (!inode_trylock(dst)) {
2090 ret = -EBUSY;
2091 goto out;
2092 }
2093 }
2094
2095 ret = -EINVAL;
2096 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2097 goto out_unlock;
2098 if (len == 0)
2099 olen = len = src->i_size - pos_in;
2100 if (pos_in + len == src->i_size)
2101 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2102 if (len == 0) {
2103 ret = 0;
2104 goto out_unlock;
2105 }
2106
2107 dst_osize = dst->i_size;
2108 if (pos_out + olen > dst->i_size)
2109 dst_max_i_size = pos_out + olen;
2110
2111 /* verify the end result is block aligned */
2112 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2113 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2114 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2115 goto out_unlock;
2116
2117 ret = f2fs_convert_inline_inode(src);
2118 if (ret)
2119 goto out_unlock;
2120
2121 ret = f2fs_convert_inline_inode(dst);
2122 if (ret)
2123 goto out_unlock;
2124
2125 /* write out all dirty pages from offset */
2126 ret = filemap_write_and_wait_range(src->i_mapping,
2127 pos_in, pos_in + len);
2128 if (ret)
2129 goto out_unlock;
2130
2131 ret = filemap_write_and_wait_range(dst->i_mapping,
2132 pos_out, pos_out + len);
2133 if (ret)
2134 goto out_unlock;
2135
2136 f2fs_balance_fs(sbi, true);
2137 f2fs_lock_op(sbi);
2138 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2139 pos_out >> F2FS_BLKSIZE_BITS,
2140 len >> F2FS_BLKSIZE_BITS, false);
2141
2142 if (!ret) {
2143 if (dst_max_i_size)
2144 f2fs_i_size_write(dst, dst_max_i_size);
2145 else if (dst_osize != dst->i_size)
2146 f2fs_i_size_write(dst, dst_osize);
2147 }
2148 f2fs_unlock_op(sbi);
2149out_unlock:
2150 if (src != dst)
2151 inode_unlock(dst);
2152out:
2153 inode_unlock(src);
2154 return ret;
2155}
2156
2157static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2158{
2159 struct f2fs_move_range range;
2160 struct fd dst;
2161 int err;
2162
2163 if (!(filp->f_mode & FMODE_READ) ||
2164 !(filp->f_mode & FMODE_WRITE))
2165 return -EBADF;
2166
2167 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2168 sizeof(range)))
2169 return -EFAULT;
2170
2171 dst = fdget(range.dst_fd);
2172 if (!dst.file)
2173 return -EBADF;
2174
2175 if (!(dst.file->f_mode & FMODE_WRITE)) {
2176 err = -EBADF;
2177 goto err_out;
2178 }
2179
2180 err = mnt_want_write_file(filp);
2181 if (err)
2182 goto err_out;
2183
2184 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2185 range.pos_out, range.len);
2186
2187 mnt_drop_write_file(filp);
2188
2189 if (copy_to_user((struct f2fs_move_range __user *)arg,
2190 &range, sizeof(range)))
2191 err = -EFAULT;
2192err_out:
2193 fdput(dst);
2194 return err;
2195}
2196
2197long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2198{
2199 switch (cmd) {
2200 case F2FS_IOC_GETFLAGS:
2201 return f2fs_ioc_getflags(filp, arg);
2202 case F2FS_IOC_SETFLAGS:
2203 return f2fs_ioc_setflags(filp, arg);
2204 case F2FS_IOC_GETVERSION:
2205 return f2fs_ioc_getversion(filp, arg);
2206 case F2FS_IOC_START_ATOMIC_WRITE:
2207 return f2fs_ioc_start_atomic_write(filp);
2208 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2209 return f2fs_ioc_commit_atomic_write(filp);
2210 case F2FS_IOC_START_VOLATILE_WRITE:
2211 return f2fs_ioc_start_volatile_write(filp);
2212 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2213 return f2fs_ioc_release_volatile_write(filp);
2214 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2215 return f2fs_ioc_abort_volatile_write(filp);
2216 case F2FS_IOC_SHUTDOWN:
2217 return f2fs_ioc_shutdown(filp, arg);
2218 case FITRIM:
2219 return f2fs_ioc_fitrim(filp, arg);
2220 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2221 return f2fs_ioc_set_encryption_policy(filp, arg);
2222 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2223 return f2fs_ioc_get_encryption_policy(filp, arg);
2224 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2225 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2226 case F2FS_IOC_GARBAGE_COLLECT:
2227 return f2fs_ioc_gc(filp, arg);
2228 case F2FS_IOC_WRITE_CHECKPOINT:
2229 return f2fs_ioc_write_checkpoint(filp, arg);
2230 case F2FS_IOC_DEFRAGMENT:
2231 return f2fs_ioc_defragment(filp, arg);
2232 case F2FS_IOC_MOVE_RANGE:
2233 return f2fs_ioc_move_range(filp, arg);
2234 default:
2235 return -ENOTTY;
2236 }
2237}
2238
2239static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2240{
2241 struct file *file = iocb->ki_filp;
2242 struct inode *inode = file_inode(file);
2243 struct blk_plug plug;
2244 ssize_t ret;
2245
2246 if (f2fs_encrypted_inode(inode) &&
2247 !fscrypt_has_encryption_key(inode) &&
2248 fscrypt_get_encryption_info(inode))
2249 return -EACCES;
2250
2251 inode_lock(inode);
2252 ret = generic_write_checks(iocb, from);
2253 if (ret > 0) {
2254 int err = f2fs_preallocate_blocks(iocb, from);
2255
2256 if (err) {
2257 inode_unlock(inode);
2258 return err;
2259 }
2260 blk_start_plug(&plug);
2261 ret = __generic_file_write_iter(iocb, from);
2262 blk_finish_plug(&plug);
2263 }
2264 inode_unlock(inode);
2265
2266 if (ret > 0)
2267 ret = generic_write_sync(iocb, ret);
2268 return ret;
2269}
2270
2271#ifdef CONFIG_COMPAT
2272long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2273{
2274 switch (cmd) {
2275 case F2FS_IOC32_GETFLAGS:
2276 cmd = F2FS_IOC_GETFLAGS;
2277 break;
2278 case F2FS_IOC32_SETFLAGS:
2279 cmd = F2FS_IOC_SETFLAGS;
2280 break;
2281 case F2FS_IOC32_GETVERSION:
2282 cmd = F2FS_IOC_GETVERSION;
2283 break;
2284 case F2FS_IOC_START_ATOMIC_WRITE:
2285 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2286 case F2FS_IOC_START_VOLATILE_WRITE:
2287 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2288 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2289 case F2FS_IOC_SHUTDOWN:
2290 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2291 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2292 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2293 case F2FS_IOC_GARBAGE_COLLECT:
2294 case F2FS_IOC_WRITE_CHECKPOINT:
2295 case F2FS_IOC_DEFRAGMENT:
2296 break;
2297 case F2FS_IOC_MOVE_RANGE:
2298 break;
2299 default:
2300 return -ENOIOCTLCMD;
2301 }
2302 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2303}
2304#endif
2305
2306const struct file_operations f2fs_file_operations = {
2307 .llseek = f2fs_llseek,
2308 .read_iter = generic_file_read_iter,
2309 .write_iter = f2fs_file_write_iter,
2310 .open = f2fs_file_open,
2311 .release = f2fs_release_file,
2312 .mmap = f2fs_file_mmap,
2313 .fsync = f2fs_sync_file,
2314 .fallocate = f2fs_fallocate,
2315 .unlocked_ioctl = f2fs_ioctl,
2316#ifdef CONFIG_COMPAT
2317 .compat_ioctl = f2fs_compat_ioctl,
2318#endif
2319 .splice_read = generic_file_splice_read,
2320 .splice_write = iter_file_splice_write,
2321};
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
25#include "f2fs.h"
26#include "node.h"
27#include "segment.h"
28#include "xattr.h"
29#include "acl.h"
30#include "gc.h"
31#include "trace.h"
32#include <trace/events/f2fs.h>
33
34static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
35{
36 struct inode *inode = file_inode(vmf->vma->vm_file);
37 vm_fault_t ret;
38
39 down_read(&F2FS_I(inode)->i_mmap_sem);
40 ret = filemap_fault(vmf);
41 up_read(&F2FS_I(inode)->i_mmap_sem);
42
43 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
44
45 return ret;
46}
47
48static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49{
50 struct page *page = vmf->page;
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 struct dnode_of_data dn = { .node_changed = false };
54 int err;
55
56 if (unlikely(f2fs_cp_error(sbi))) {
57 err = -EIO;
58 goto err;
59 }
60
61 if (!f2fs_is_checkpoint_ready(sbi)) {
62 err = -ENOSPC;
63 goto err;
64 }
65
66 sb_start_pagefault(inode->i_sb);
67
68 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
69
70 file_update_time(vmf->vma->vm_file);
71 down_read(&F2FS_I(inode)->i_mmap_sem);
72 lock_page(page);
73 if (unlikely(page->mapping != inode->i_mapping ||
74 page_offset(page) > i_size_read(inode) ||
75 !PageUptodate(page))) {
76 unlock_page(page);
77 err = -EFAULT;
78 goto out_sem;
79 }
80
81 /* block allocation */
82 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
83 set_new_dnode(&dn, inode, NULL, NULL, 0);
84 err = f2fs_get_block(&dn, page->index);
85 f2fs_put_dnode(&dn);
86 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
87 if (err) {
88 unlock_page(page);
89 goto out_sem;
90 }
91
92 /* fill the page */
93 f2fs_wait_on_page_writeback(page, DATA, false, true);
94
95 /* wait for GCed page writeback via META_MAPPING */
96 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
97
98 /*
99 * check to see if the page is mapped already (no holes)
100 */
101 if (PageMappedToDisk(page))
102 goto out_sem;
103
104 /* page is wholly or partially inside EOF */
105 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
106 i_size_read(inode)) {
107 loff_t offset;
108
109 offset = i_size_read(inode) & ~PAGE_MASK;
110 zero_user_segment(page, offset, PAGE_SIZE);
111 }
112 set_page_dirty(page);
113 if (!PageUptodate(page))
114 SetPageUptodate(page);
115
116 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
117 f2fs_update_time(sbi, REQ_TIME);
118
119 trace_f2fs_vm_page_mkwrite(page, DATA);
120out_sem:
121 up_read(&F2FS_I(inode)->i_mmap_sem);
122
123 f2fs_balance_fs(sbi, dn.node_changed);
124
125 sb_end_pagefault(inode->i_sb);
126err:
127 return block_page_mkwrite_return(err);
128}
129
130static const struct vm_operations_struct f2fs_file_vm_ops = {
131 .fault = f2fs_filemap_fault,
132 .map_pages = filemap_map_pages,
133 .page_mkwrite = f2fs_vm_page_mkwrite,
134};
135
136static int get_parent_ino(struct inode *inode, nid_t *pino)
137{
138 struct dentry *dentry;
139
140 inode = igrab(inode);
141 dentry = d_find_any_alias(inode);
142 iput(inode);
143 if (!dentry)
144 return 0;
145
146 *pino = parent_ino(dentry);
147 dput(dentry);
148 return 1;
149}
150
151static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
152{
153 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
154 enum cp_reason_type cp_reason = CP_NO_NEEDED;
155
156 if (!S_ISREG(inode->i_mode))
157 cp_reason = CP_NON_REGULAR;
158 else if (inode->i_nlink != 1)
159 cp_reason = CP_HARDLINK;
160 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
161 cp_reason = CP_SB_NEED_CP;
162 else if (file_wrong_pino(inode))
163 cp_reason = CP_WRONG_PINO;
164 else if (!f2fs_space_for_roll_forward(sbi))
165 cp_reason = CP_NO_SPC_ROLL;
166 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
167 cp_reason = CP_NODE_NEED_CP;
168 else if (test_opt(sbi, FASTBOOT))
169 cp_reason = CP_FASTBOOT_MODE;
170 else if (F2FS_OPTION(sbi).active_logs == 2)
171 cp_reason = CP_SPEC_LOG_NUM;
172 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
173 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
174 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
175 TRANS_DIR_INO))
176 cp_reason = CP_RECOVER_DIR;
177
178 return cp_reason;
179}
180
181static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
182{
183 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
184 bool ret = false;
185 /* But we need to avoid that there are some inode updates */
186 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
187 ret = true;
188 f2fs_put_page(i, 0);
189 return ret;
190}
191
192static void try_to_fix_pino(struct inode *inode)
193{
194 struct f2fs_inode_info *fi = F2FS_I(inode);
195 nid_t pino;
196
197 down_write(&fi->i_sem);
198 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
199 get_parent_ino(inode, &pino)) {
200 f2fs_i_pino_write(inode, pino);
201 file_got_pino(inode);
202 }
203 up_write(&fi->i_sem);
204}
205
206static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
207 int datasync, bool atomic)
208{
209 struct inode *inode = file->f_mapping->host;
210 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
211 nid_t ino = inode->i_ino;
212 int ret = 0;
213 enum cp_reason_type cp_reason = 0;
214 struct writeback_control wbc = {
215 .sync_mode = WB_SYNC_ALL,
216 .nr_to_write = LONG_MAX,
217 .for_reclaim = 0,
218 };
219 unsigned int seq_id = 0;
220
221 if (unlikely(f2fs_readonly(inode->i_sb) ||
222 is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
223 return 0;
224
225 trace_f2fs_sync_file_enter(inode);
226
227 if (S_ISDIR(inode->i_mode))
228 goto go_write;
229
230 /* if fdatasync is triggered, let's do in-place-update */
231 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
232 set_inode_flag(inode, FI_NEED_IPU);
233 ret = file_write_and_wait_range(file, start, end);
234 clear_inode_flag(inode, FI_NEED_IPU);
235
236 if (ret) {
237 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
238 return ret;
239 }
240
241 /* if the inode is dirty, let's recover all the time */
242 if (!f2fs_skip_inode_update(inode, datasync)) {
243 f2fs_write_inode(inode, NULL);
244 goto go_write;
245 }
246
247 /*
248 * if there is no written data, don't waste time to write recovery info.
249 */
250 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
251 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
252
253 /* it may call write_inode just prior to fsync */
254 if (need_inode_page_update(sbi, ino))
255 goto go_write;
256
257 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
258 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
259 goto flush_out;
260 goto out;
261 }
262go_write:
263 /*
264 * Both of fdatasync() and fsync() are able to be recovered from
265 * sudden-power-off.
266 */
267 down_read(&F2FS_I(inode)->i_sem);
268 cp_reason = need_do_checkpoint(inode);
269 up_read(&F2FS_I(inode)->i_sem);
270
271 if (cp_reason) {
272 /* all the dirty node pages should be flushed for POR */
273 ret = f2fs_sync_fs(inode->i_sb, 1);
274
275 /*
276 * We've secured consistency through sync_fs. Following pino
277 * will be used only for fsynced inodes after checkpoint.
278 */
279 try_to_fix_pino(inode);
280 clear_inode_flag(inode, FI_APPEND_WRITE);
281 clear_inode_flag(inode, FI_UPDATE_WRITE);
282 goto out;
283 }
284sync_nodes:
285 atomic_inc(&sbi->wb_sync_req[NODE]);
286 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
287 atomic_dec(&sbi->wb_sync_req[NODE]);
288 if (ret)
289 goto out;
290
291 /* if cp_error was enabled, we should avoid infinite loop */
292 if (unlikely(f2fs_cp_error(sbi))) {
293 ret = -EIO;
294 goto out;
295 }
296
297 if (f2fs_need_inode_block_update(sbi, ino)) {
298 f2fs_mark_inode_dirty_sync(inode, true);
299 f2fs_write_inode(inode, NULL);
300 goto sync_nodes;
301 }
302
303 /*
304 * If it's atomic_write, it's just fine to keep write ordering. So
305 * here we don't need to wait for node write completion, since we use
306 * node chain which serializes node blocks. If one of node writes are
307 * reordered, we can see simply broken chain, resulting in stopping
308 * roll-forward recovery. It means we'll recover all or none node blocks
309 * given fsync mark.
310 */
311 if (!atomic) {
312 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
313 if (ret)
314 goto out;
315 }
316
317 /* once recovery info is written, don't need to tack this */
318 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
319 clear_inode_flag(inode, FI_APPEND_WRITE);
320flush_out:
321 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
322 ret = f2fs_issue_flush(sbi, inode->i_ino);
323 if (!ret) {
324 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
325 clear_inode_flag(inode, FI_UPDATE_WRITE);
326 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
327 }
328 f2fs_update_time(sbi, REQ_TIME);
329out:
330 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
331 f2fs_trace_ios(NULL, 1);
332 return ret;
333}
334
335int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
336{
337 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
338 return -EIO;
339 return f2fs_do_sync_file(file, start, end, datasync, false);
340}
341
342static pgoff_t __get_first_dirty_index(struct address_space *mapping,
343 pgoff_t pgofs, int whence)
344{
345 struct page *page;
346 int nr_pages;
347
348 if (whence != SEEK_DATA)
349 return 0;
350
351 /* find first dirty page index */
352 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
353 1, &page);
354 if (!nr_pages)
355 return ULONG_MAX;
356 pgofs = page->index;
357 put_page(page);
358 return pgofs;
359}
360
361static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
362 pgoff_t dirty, pgoff_t pgofs, int whence)
363{
364 switch (whence) {
365 case SEEK_DATA:
366 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
367 __is_valid_data_blkaddr(blkaddr))
368 return true;
369 break;
370 case SEEK_HOLE:
371 if (blkaddr == NULL_ADDR)
372 return true;
373 break;
374 }
375 return false;
376}
377
378static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
379{
380 struct inode *inode = file->f_mapping->host;
381 loff_t maxbytes = inode->i_sb->s_maxbytes;
382 struct dnode_of_data dn;
383 pgoff_t pgofs, end_offset, dirty;
384 loff_t data_ofs = offset;
385 loff_t isize;
386 int err = 0;
387
388 inode_lock(inode);
389
390 isize = i_size_read(inode);
391 if (offset >= isize)
392 goto fail;
393
394 /* handle inline data case */
395 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
396 if (whence == SEEK_HOLE)
397 data_ofs = isize;
398 goto found;
399 }
400
401 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
402
403 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
404
405 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
406 set_new_dnode(&dn, inode, NULL, NULL, 0);
407 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
408 if (err && err != -ENOENT) {
409 goto fail;
410 } else if (err == -ENOENT) {
411 /* direct node does not exists */
412 if (whence == SEEK_DATA) {
413 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
414 continue;
415 } else {
416 goto found;
417 }
418 }
419
420 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
421
422 /* find data/hole in dnode block */
423 for (; dn.ofs_in_node < end_offset;
424 dn.ofs_in_node++, pgofs++,
425 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
426 block_t blkaddr;
427
428 blkaddr = datablock_addr(dn.inode,
429 dn.node_page, dn.ofs_in_node);
430
431 if (__is_valid_data_blkaddr(blkaddr) &&
432 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
433 blkaddr, DATA_GENERIC_ENHANCE)) {
434 f2fs_put_dnode(&dn);
435 goto fail;
436 }
437
438 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
439 pgofs, whence)) {
440 f2fs_put_dnode(&dn);
441 goto found;
442 }
443 }
444 f2fs_put_dnode(&dn);
445 }
446
447 if (whence == SEEK_DATA)
448 goto fail;
449found:
450 if (whence == SEEK_HOLE && data_ofs > isize)
451 data_ofs = isize;
452 inode_unlock(inode);
453 return vfs_setpos(file, data_ofs, maxbytes);
454fail:
455 inode_unlock(inode);
456 return -ENXIO;
457}
458
459static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
460{
461 struct inode *inode = file->f_mapping->host;
462 loff_t maxbytes = inode->i_sb->s_maxbytes;
463
464 switch (whence) {
465 case SEEK_SET:
466 case SEEK_CUR:
467 case SEEK_END:
468 return generic_file_llseek_size(file, offset, whence,
469 maxbytes, i_size_read(inode));
470 case SEEK_DATA:
471 case SEEK_HOLE:
472 if (offset < 0)
473 return -ENXIO;
474 return f2fs_seek_block(file, offset, whence);
475 }
476
477 return -EINVAL;
478}
479
480static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
481{
482 struct inode *inode = file_inode(file);
483 int err;
484
485 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
486 return -EIO;
487
488 /* we don't need to use inline_data strictly */
489 err = f2fs_convert_inline_inode(inode);
490 if (err)
491 return err;
492
493 file_accessed(file);
494 vma->vm_ops = &f2fs_file_vm_ops;
495 return 0;
496}
497
498static int f2fs_file_open(struct inode *inode, struct file *filp)
499{
500 int err = fscrypt_file_open(inode, filp);
501
502 if (err)
503 return err;
504
505 err = fsverity_file_open(inode, filp);
506 if (err)
507 return err;
508
509 filp->f_mode |= FMODE_NOWAIT;
510
511 return dquot_file_open(inode, filp);
512}
513
514void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
515{
516 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
517 struct f2fs_node *raw_node;
518 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
519 __le32 *addr;
520 int base = 0;
521
522 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
523 base = get_extra_isize(dn->inode);
524
525 raw_node = F2FS_NODE(dn->node_page);
526 addr = blkaddr_in_node(raw_node) + base + ofs;
527
528 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
529 block_t blkaddr = le32_to_cpu(*addr);
530
531 if (blkaddr == NULL_ADDR)
532 continue;
533
534 dn->data_blkaddr = NULL_ADDR;
535 f2fs_set_data_blkaddr(dn);
536
537 if (__is_valid_data_blkaddr(blkaddr) &&
538 !f2fs_is_valid_blkaddr(sbi, blkaddr,
539 DATA_GENERIC_ENHANCE))
540 continue;
541
542 f2fs_invalidate_blocks(sbi, blkaddr);
543 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
544 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
545 nr_free++;
546 }
547
548 if (nr_free) {
549 pgoff_t fofs;
550 /*
551 * once we invalidate valid blkaddr in range [ofs, ofs + count],
552 * we will invalidate all blkaddr in the whole range.
553 */
554 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
555 dn->inode) + ofs;
556 f2fs_update_extent_cache_range(dn, fofs, 0, len);
557 dec_valid_block_count(sbi, dn->inode, nr_free);
558 }
559 dn->ofs_in_node = ofs;
560
561 f2fs_update_time(sbi, REQ_TIME);
562 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
563 dn->ofs_in_node, nr_free);
564}
565
566void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
567{
568 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
569}
570
571static int truncate_partial_data_page(struct inode *inode, u64 from,
572 bool cache_only)
573{
574 loff_t offset = from & (PAGE_SIZE - 1);
575 pgoff_t index = from >> PAGE_SHIFT;
576 struct address_space *mapping = inode->i_mapping;
577 struct page *page;
578
579 if (!offset && !cache_only)
580 return 0;
581
582 if (cache_only) {
583 page = find_lock_page(mapping, index);
584 if (page && PageUptodate(page))
585 goto truncate_out;
586 f2fs_put_page(page, 1);
587 return 0;
588 }
589
590 page = f2fs_get_lock_data_page(inode, index, true);
591 if (IS_ERR(page))
592 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
593truncate_out:
594 f2fs_wait_on_page_writeback(page, DATA, true, true);
595 zero_user(page, offset, PAGE_SIZE - offset);
596
597 /* An encrypted inode should have a key and truncate the last page. */
598 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
599 if (!cache_only)
600 set_page_dirty(page);
601 f2fs_put_page(page, 1);
602 return 0;
603}
604
605int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
606{
607 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
608 struct dnode_of_data dn;
609 pgoff_t free_from;
610 int count = 0, err = 0;
611 struct page *ipage;
612 bool truncate_page = false;
613
614 trace_f2fs_truncate_blocks_enter(inode, from);
615
616 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
617
618 if (free_from >= sbi->max_file_blocks)
619 goto free_partial;
620
621 if (lock)
622 f2fs_lock_op(sbi);
623
624 ipage = f2fs_get_node_page(sbi, inode->i_ino);
625 if (IS_ERR(ipage)) {
626 err = PTR_ERR(ipage);
627 goto out;
628 }
629
630 if (f2fs_has_inline_data(inode)) {
631 f2fs_truncate_inline_inode(inode, ipage, from);
632 f2fs_put_page(ipage, 1);
633 truncate_page = true;
634 goto out;
635 }
636
637 set_new_dnode(&dn, inode, ipage, NULL, 0);
638 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
639 if (err) {
640 if (err == -ENOENT)
641 goto free_next;
642 goto out;
643 }
644
645 count = ADDRS_PER_PAGE(dn.node_page, inode);
646
647 count -= dn.ofs_in_node;
648 f2fs_bug_on(sbi, count < 0);
649
650 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
651 f2fs_truncate_data_blocks_range(&dn, count);
652 free_from += count;
653 }
654
655 f2fs_put_dnode(&dn);
656free_next:
657 err = f2fs_truncate_inode_blocks(inode, free_from);
658out:
659 if (lock)
660 f2fs_unlock_op(sbi);
661free_partial:
662 /* lastly zero out the first data page */
663 if (!err)
664 err = truncate_partial_data_page(inode, from, truncate_page);
665
666 trace_f2fs_truncate_blocks_exit(inode, err);
667 return err;
668}
669
670int f2fs_truncate(struct inode *inode)
671{
672 int err;
673
674 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
675 return -EIO;
676
677 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
678 S_ISLNK(inode->i_mode)))
679 return 0;
680
681 trace_f2fs_truncate(inode);
682
683 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
684 f2fs_show_injection_info(FAULT_TRUNCATE);
685 return -EIO;
686 }
687
688 /* we should check inline_data size */
689 if (!f2fs_may_inline_data(inode)) {
690 err = f2fs_convert_inline_inode(inode);
691 if (err)
692 return err;
693 }
694
695 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
696 if (err)
697 return err;
698
699 inode->i_mtime = inode->i_ctime = current_time(inode);
700 f2fs_mark_inode_dirty_sync(inode, false);
701 return 0;
702}
703
704int f2fs_getattr(const struct path *path, struct kstat *stat,
705 u32 request_mask, unsigned int query_flags)
706{
707 struct inode *inode = d_inode(path->dentry);
708 struct f2fs_inode_info *fi = F2FS_I(inode);
709 struct f2fs_inode *ri;
710 unsigned int flags;
711
712 if (f2fs_has_extra_attr(inode) &&
713 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
714 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
715 stat->result_mask |= STATX_BTIME;
716 stat->btime.tv_sec = fi->i_crtime.tv_sec;
717 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
718 }
719
720 flags = fi->i_flags;
721 if (flags & F2FS_APPEND_FL)
722 stat->attributes |= STATX_ATTR_APPEND;
723 if (IS_ENCRYPTED(inode))
724 stat->attributes |= STATX_ATTR_ENCRYPTED;
725 if (flags & F2FS_IMMUTABLE_FL)
726 stat->attributes |= STATX_ATTR_IMMUTABLE;
727 if (flags & F2FS_NODUMP_FL)
728 stat->attributes |= STATX_ATTR_NODUMP;
729
730 stat->attributes_mask |= (STATX_ATTR_APPEND |
731 STATX_ATTR_ENCRYPTED |
732 STATX_ATTR_IMMUTABLE |
733 STATX_ATTR_NODUMP);
734
735 generic_fillattr(inode, stat);
736
737 /* we need to show initial sectors used for inline_data/dentries */
738 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
739 f2fs_has_inline_dentry(inode))
740 stat->blocks += (stat->size + 511) >> 9;
741
742 return 0;
743}
744
745#ifdef CONFIG_F2FS_FS_POSIX_ACL
746static void __setattr_copy(struct inode *inode, const struct iattr *attr)
747{
748 unsigned int ia_valid = attr->ia_valid;
749
750 if (ia_valid & ATTR_UID)
751 inode->i_uid = attr->ia_uid;
752 if (ia_valid & ATTR_GID)
753 inode->i_gid = attr->ia_gid;
754 if (ia_valid & ATTR_ATIME) {
755 inode->i_atime = timestamp_truncate(attr->ia_atime,
756 inode);
757 }
758 if (ia_valid & ATTR_MTIME) {
759 inode->i_mtime = timestamp_truncate(attr->ia_mtime,
760 inode);
761 }
762 if (ia_valid & ATTR_CTIME) {
763 inode->i_ctime = timestamp_truncate(attr->ia_ctime,
764 inode);
765 }
766 if (ia_valid & ATTR_MODE) {
767 umode_t mode = attr->ia_mode;
768
769 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
770 mode &= ~S_ISGID;
771 set_acl_inode(inode, mode);
772 }
773}
774#else
775#define __setattr_copy setattr_copy
776#endif
777
778int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
779{
780 struct inode *inode = d_inode(dentry);
781 int err;
782
783 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
784 return -EIO;
785
786 err = setattr_prepare(dentry, attr);
787 if (err)
788 return err;
789
790 err = fscrypt_prepare_setattr(dentry, attr);
791 if (err)
792 return err;
793
794 err = fsverity_prepare_setattr(dentry, attr);
795 if (err)
796 return err;
797
798 if (is_quota_modification(inode, attr)) {
799 err = dquot_initialize(inode);
800 if (err)
801 return err;
802 }
803 if ((attr->ia_valid & ATTR_UID &&
804 !uid_eq(attr->ia_uid, inode->i_uid)) ||
805 (attr->ia_valid & ATTR_GID &&
806 !gid_eq(attr->ia_gid, inode->i_gid))) {
807 f2fs_lock_op(F2FS_I_SB(inode));
808 err = dquot_transfer(inode, attr);
809 if (err) {
810 set_sbi_flag(F2FS_I_SB(inode),
811 SBI_QUOTA_NEED_REPAIR);
812 f2fs_unlock_op(F2FS_I_SB(inode));
813 return err;
814 }
815 /*
816 * update uid/gid under lock_op(), so that dquot and inode can
817 * be updated atomically.
818 */
819 if (attr->ia_valid & ATTR_UID)
820 inode->i_uid = attr->ia_uid;
821 if (attr->ia_valid & ATTR_GID)
822 inode->i_gid = attr->ia_gid;
823 f2fs_mark_inode_dirty_sync(inode, true);
824 f2fs_unlock_op(F2FS_I_SB(inode));
825 }
826
827 if (attr->ia_valid & ATTR_SIZE) {
828 loff_t old_size = i_size_read(inode);
829
830 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
831 /*
832 * should convert inline inode before i_size_write to
833 * keep smaller than inline_data size with inline flag.
834 */
835 err = f2fs_convert_inline_inode(inode);
836 if (err)
837 return err;
838 }
839
840 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
841 down_write(&F2FS_I(inode)->i_mmap_sem);
842
843 truncate_setsize(inode, attr->ia_size);
844
845 if (attr->ia_size <= old_size)
846 err = f2fs_truncate(inode);
847 /*
848 * do not trim all blocks after i_size if target size is
849 * larger than i_size.
850 */
851 up_write(&F2FS_I(inode)->i_mmap_sem);
852 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
853 if (err)
854 return err;
855
856 down_write(&F2FS_I(inode)->i_sem);
857 inode->i_mtime = inode->i_ctime = current_time(inode);
858 F2FS_I(inode)->last_disk_size = i_size_read(inode);
859 up_write(&F2FS_I(inode)->i_sem);
860 }
861
862 __setattr_copy(inode, attr);
863
864 if (attr->ia_valid & ATTR_MODE) {
865 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
866 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
867 inode->i_mode = F2FS_I(inode)->i_acl_mode;
868 clear_inode_flag(inode, FI_ACL_MODE);
869 }
870 }
871
872 /* file size may changed here */
873 f2fs_mark_inode_dirty_sync(inode, true);
874
875 /* inode change will produce dirty node pages flushed by checkpoint */
876 f2fs_balance_fs(F2FS_I_SB(inode), true);
877
878 return err;
879}
880
881const struct inode_operations f2fs_file_inode_operations = {
882 .getattr = f2fs_getattr,
883 .setattr = f2fs_setattr,
884 .get_acl = f2fs_get_acl,
885 .set_acl = f2fs_set_acl,
886#ifdef CONFIG_F2FS_FS_XATTR
887 .listxattr = f2fs_listxattr,
888#endif
889 .fiemap = f2fs_fiemap,
890};
891
892static int fill_zero(struct inode *inode, pgoff_t index,
893 loff_t start, loff_t len)
894{
895 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
896 struct page *page;
897
898 if (!len)
899 return 0;
900
901 f2fs_balance_fs(sbi, true);
902
903 f2fs_lock_op(sbi);
904 page = f2fs_get_new_data_page(inode, NULL, index, false);
905 f2fs_unlock_op(sbi);
906
907 if (IS_ERR(page))
908 return PTR_ERR(page);
909
910 f2fs_wait_on_page_writeback(page, DATA, true, true);
911 zero_user(page, start, len);
912 set_page_dirty(page);
913 f2fs_put_page(page, 1);
914 return 0;
915}
916
917int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
918{
919 int err;
920
921 while (pg_start < pg_end) {
922 struct dnode_of_data dn;
923 pgoff_t end_offset, count;
924
925 set_new_dnode(&dn, inode, NULL, NULL, 0);
926 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
927 if (err) {
928 if (err == -ENOENT) {
929 pg_start = f2fs_get_next_page_offset(&dn,
930 pg_start);
931 continue;
932 }
933 return err;
934 }
935
936 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
937 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
938
939 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
940
941 f2fs_truncate_data_blocks_range(&dn, count);
942 f2fs_put_dnode(&dn);
943
944 pg_start += count;
945 }
946 return 0;
947}
948
949static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
950{
951 pgoff_t pg_start, pg_end;
952 loff_t off_start, off_end;
953 int ret;
954
955 ret = f2fs_convert_inline_inode(inode);
956 if (ret)
957 return ret;
958
959 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
960 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
961
962 off_start = offset & (PAGE_SIZE - 1);
963 off_end = (offset + len) & (PAGE_SIZE - 1);
964
965 if (pg_start == pg_end) {
966 ret = fill_zero(inode, pg_start, off_start,
967 off_end - off_start);
968 if (ret)
969 return ret;
970 } else {
971 if (off_start) {
972 ret = fill_zero(inode, pg_start++, off_start,
973 PAGE_SIZE - off_start);
974 if (ret)
975 return ret;
976 }
977 if (off_end) {
978 ret = fill_zero(inode, pg_end, 0, off_end);
979 if (ret)
980 return ret;
981 }
982
983 if (pg_start < pg_end) {
984 struct address_space *mapping = inode->i_mapping;
985 loff_t blk_start, blk_end;
986 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
987
988 f2fs_balance_fs(sbi, true);
989
990 blk_start = (loff_t)pg_start << PAGE_SHIFT;
991 blk_end = (loff_t)pg_end << PAGE_SHIFT;
992
993 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
994 down_write(&F2FS_I(inode)->i_mmap_sem);
995
996 truncate_inode_pages_range(mapping, blk_start,
997 blk_end - 1);
998
999 f2fs_lock_op(sbi);
1000 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1001 f2fs_unlock_op(sbi);
1002
1003 up_write(&F2FS_I(inode)->i_mmap_sem);
1004 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1005 }
1006 }
1007
1008 return ret;
1009}
1010
1011static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1012 int *do_replace, pgoff_t off, pgoff_t len)
1013{
1014 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1015 struct dnode_of_data dn;
1016 int ret, done, i;
1017
1018next_dnode:
1019 set_new_dnode(&dn, inode, NULL, NULL, 0);
1020 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1021 if (ret && ret != -ENOENT) {
1022 return ret;
1023 } else if (ret == -ENOENT) {
1024 if (dn.max_level == 0)
1025 return -ENOENT;
1026 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - dn.ofs_in_node,
1027 len);
1028 blkaddr += done;
1029 do_replace += done;
1030 goto next;
1031 }
1032
1033 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1034 dn.ofs_in_node, len);
1035 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1036 *blkaddr = datablock_addr(dn.inode,
1037 dn.node_page, dn.ofs_in_node);
1038
1039 if (__is_valid_data_blkaddr(*blkaddr) &&
1040 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1041 DATA_GENERIC_ENHANCE)) {
1042 f2fs_put_dnode(&dn);
1043 return -EFSCORRUPTED;
1044 }
1045
1046 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1047
1048 if (test_opt(sbi, LFS)) {
1049 f2fs_put_dnode(&dn);
1050 return -EOPNOTSUPP;
1051 }
1052
1053 /* do not invalidate this block address */
1054 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1055 *do_replace = 1;
1056 }
1057 }
1058 f2fs_put_dnode(&dn);
1059next:
1060 len -= done;
1061 off += done;
1062 if (len)
1063 goto next_dnode;
1064 return 0;
1065}
1066
1067static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1068 int *do_replace, pgoff_t off, int len)
1069{
1070 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1071 struct dnode_of_data dn;
1072 int ret, i;
1073
1074 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1075 if (*do_replace == 0)
1076 continue;
1077
1078 set_new_dnode(&dn, inode, NULL, NULL, 0);
1079 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1080 if (ret) {
1081 dec_valid_block_count(sbi, inode, 1);
1082 f2fs_invalidate_blocks(sbi, *blkaddr);
1083 } else {
1084 f2fs_update_data_blkaddr(&dn, *blkaddr);
1085 }
1086 f2fs_put_dnode(&dn);
1087 }
1088 return 0;
1089}
1090
1091static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1092 block_t *blkaddr, int *do_replace,
1093 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1094{
1095 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1096 pgoff_t i = 0;
1097 int ret;
1098
1099 while (i < len) {
1100 if (blkaddr[i] == NULL_ADDR && !full) {
1101 i++;
1102 continue;
1103 }
1104
1105 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1106 struct dnode_of_data dn;
1107 struct node_info ni;
1108 size_t new_size;
1109 pgoff_t ilen;
1110
1111 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1112 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1113 if (ret)
1114 return ret;
1115
1116 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1117 if (ret) {
1118 f2fs_put_dnode(&dn);
1119 return ret;
1120 }
1121
1122 ilen = min((pgoff_t)
1123 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1124 dn.ofs_in_node, len - i);
1125 do {
1126 dn.data_blkaddr = datablock_addr(dn.inode,
1127 dn.node_page, dn.ofs_in_node);
1128 f2fs_truncate_data_blocks_range(&dn, 1);
1129
1130 if (do_replace[i]) {
1131 f2fs_i_blocks_write(src_inode,
1132 1, false, false);
1133 f2fs_i_blocks_write(dst_inode,
1134 1, true, false);
1135 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1136 blkaddr[i], ni.version, true, false);
1137
1138 do_replace[i] = 0;
1139 }
1140 dn.ofs_in_node++;
1141 i++;
1142 new_size = (dst + i) << PAGE_SHIFT;
1143 if (dst_inode->i_size < new_size)
1144 f2fs_i_size_write(dst_inode, new_size);
1145 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1146
1147 f2fs_put_dnode(&dn);
1148 } else {
1149 struct page *psrc, *pdst;
1150
1151 psrc = f2fs_get_lock_data_page(src_inode,
1152 src + i, true);
1153 if (IS_ERR(psrc))
1154 return PTR_ERR(psrc);
1155 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1156 true);
1157 if (IS_ERR(pdst)) {
1158 f2fs_put_page(psrc, 1);
1159 return PTR_ERR(pdst);
1160 }
1161 f2fs_copy_page(psrc, pdst);
1162 set_page_dirty(pdst);
1163 f2fs_put_page(pdst, 1);
1164 f2fs_put_page(psrc, 1);
1165
1166 ret = f2fs_truncate_hole(src_inode,
1167 src + i, src + i + 1);
1168 if (ret)
1169 return ret;
1170 i++;
1171 }
1172 }
1173 return 0;
1174}
1175
1176static int __exchange_data_block(struct inode *src_inode,
1177 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1178 pgoff_t len, bool full)
1179{
1180 block_t *src_blkaddr;
1181 int *do_replace;
1182 pgoff_t olen;
1183 int ret;
1184
1185 while (len) {
1186 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1187
1188 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1189 array_size(olen, sizeof(block_t)),
1190 GFP_KERNEL);
1191 if (!src_blkaddr)
1192 return -ENOMEM;
1193
1194 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1195 array_size(olen, sizeof(int)),
1196 GFP_KERNEL);
1197 if (!do_replace) {
1198 kvfree(src_blkaddr);
1199 return -ENOMEM;
1200 }
1201
1202 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1203 do_replace, src, olen);
1204 if (ret)
1205 goto roll_back;
1206
1207 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1208 do_replace, src, dst, olen, full);
1209 if (ret)
1210 goto roll_back;
1211
1212 src += olen;
1213 dst += olen;
1214 len -= olen;
1215
1216 kvfree(src_blkaddr);
1217 kvfree(do_replace);
1218 }
1219 return 0;
1220
1221roll_back:
1222 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1223 kvfree(src_blkaddr);
1224 kvfree(do_replace);
1225 return ret;
1226}
1227
1228static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1229{
1230 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1231 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1232 pgoff_t start = offset >> PAGE_SHIFT;
1233 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1234 int ret;
1235
1236 f2fs_balance_fs(sbi, true);
1237
1238 /* avoid gc operation during block exchange */
1239 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1240 down_write(&F2FS_I(inode)->i_mmap_sem);
1241
1242 f2fs_lock_op(sbi);
1243 f2fs_drop_extent_tree(inode);
1244 truncate_pagecache(inode, offset);
1245 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1246 f2fs_unlock_op(sbi);
1247
1248 up_write(&F2FS_I(inode)->i_mmap_sem);
1249 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1250 return ret;
1251}
1252
1253static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1254{
1255 loff_t new_size;
1256 int ret;
1257
1258 if (offset + len >= i_size_read(inode))
1259 return -EINVAL;
1260
1261 /* collapse range should be aligned to block size of f2fs. */
1262 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1263 return -EINVAL;
1264
1265 ret = f2fs_convert_inline_inode(inode);
1266 if (ret)
1267 return ret;
1268
1269 /* write out all dirty pages from offset */
1270 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1271 if (ret)
1272 return ret;
1273
1274 ret = f2fs_do_collapse(inode, offset, len);
1275 if (ret)
1276 return ret;
1277
1278 /* write out all moved pages, if possible */
1279 down_write(&F2FS_I(inode)->i_mmap_sem);
1280 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1281 truncate_pagecache(inode, offset);
1282
1283 new_size = i_size_read(inode) - len;
1284 truncate_pagecache(inode, new_size);
1285
1286 ret = f2fs_truncate_blocks(inode, new_size, true);
1287 up_write(&F2FS_I(inode)->i_mmap_sem);
1288 if (!ret)
1289 f2fs_i_size_write(inode, new_size);
1290 return ret;
1291}
1292
1293static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1294 pgoff_t end)
1295{
1296 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1297 pgoff_t index = start;
1298 unsigned int ofs_in_node = dn->ofs_in_node;
1299 blkcnt_t count = 0;
1300 int ret;
1301
1302 for (; index < end; index++, dn->ofs_in_node++) {
1303 if (datablock_addr(dn->inode, dn->node_page,
1304 dn->ofs_in_node) == NULL_ADDR)
1305 count++;
1306 }
1307
1308 dn->ofs_in_node = ofs_in_node;
1309 ret = f2fs_reserve_new_blocks(dn, count);
1310 if (ret)
1311 return ret;
1312
1313 dn->ofs_in_node = ofs_in_node;
1314 for (index = start; index < end; index++, dn->ofs_in_node++) {
1315 dn->data_blkaddr = datablock_addr(dn->inode,
1316 dn->node_page, dn->ofs_in_node);
1317 /*
1318 * f2fs_reserve_new_blocks will not guarantee entire block
1319 * allocation.
1320 */
1321 if (dn->data_blkaddr == NULL_ADDR) {
1322 ret = -ENOSPC;
1323 break;
1324 }
1325 if (dn->data_blkaddr != NEW_ADDR) {
1326 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1327 dn->data_blkaddr = NEW_ADDR;
1328 f2fs_set_data_blkaddr(dn);
1329 }
1330 }
1331
1332 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1333
1334 return ret;
1335}
1336
1337static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1338 int mode)
1339{
1340 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1341 struct address_space *mapping = inode->i_mapping;
1342 pgoff_t index, pg_start, pg_end;
1343 loff_t new_size = i_size_read(inode);
1344 loff_t off_start, off_end;
1345 int ret = 0;
1346
1347 ret = inode_newsize_ok(inode, (len + offset));
1348 if (ret)
1349 return ret;
1350
1351 ret = f2fs_convert_inline_inode(inode);
1352 if (ret)
1353 return ret;
1354
1355 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1356 if (ret)
1357 return ret;
1358
1359 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1360 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1361
1362 off_start = offset & (PAGE_SIZE - 1);
1363 off_end = (offset + len) & (PAGE_SIZE - 1);
1364
1365 if (pg_start == pg_end) {
1366 ret = fill_zero(inode, pg_start, off_start,
1367 off_end - off_start);
1368 if (ret)
1369 return ret;
1370
1371 new_size = max_t(loff_t, new_size, offset + len);
1372 } else {
1373 if (off_start) {
1374 ret = fill_zero(inode, pg_start++, off_start,
1375 PAGE_SIZE - off_start);
1376 if (ret)
1377 return ret;
1378
1379 new_size = max_t(loff_t, new_size,
1380 (loff_t)pg_start << PAGE_SHIFT);
1381 }
1382
1383 for (index = pg_start; index < pg_end;) {
1384 struct dnode_of_data dn;
1385 unsigned int end_offset;
1386 pgoff_t end;
1387
1388 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1389 down_write(&F2FS_I(inode)->i_mmap_sem);
1390
1391 truncate_pagecache_range(inode,
1392 (loff_t)index << PAGE_SHIFT,
1393 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1394
1395 f2fs_lock_op(sbi);
1396
1397 set_new_dnode(&dn, inode, NULL, NULL, 0);
1398 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1399 if (ret) {
1400 f2fs_unlock_op(sbi);
1401 up_write(&F2FS_I(inode)->i_mmap_sem);
1402 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1403 goto out;
1404 }
1405
1406 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1407 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1408
1409 ret = f2fs_do_zero_range(&dn, index, end);
1410 f2fs_put_dnode(&dn);
1411
1412 f2fs_unlock_op(sbi);
1413 up_write(&F2FS_I(inode)->i_mmap_sem);
1414 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1415
1416 f2fs_balance_fs(sbi, dn.node_changed);
1417
1418 if (ret)
1419 goto out;
1420
1421 index = end;
1422 new_size = max_t(loff_t, new_size,
1423 (loff_t)index << PAGE_SHIFT);
1424 }
1425
1426 if (off_end) {
1427 ret = fill_zero(inode, pg_end, 0, off_end);
1428 if (ret)
1429 goto out;
1430
1431 new_size = max_t(loff_t, new_size, offset + len);
1432 }
1433 }
1434
1435out:
1436 if (new_size > i_size_read(inode)) {
1437 if (mode & FALLOC_FL_KEEP_SIZE)
1438 file_set_keep_isize(inode);
1439 else
1440 f2fs_i_size_write(inode, new_size);
1441 }
1442 return ret;
1443}
1444
1445static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1446{
1447 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1448 pgoff_t nr, pg_start, pg_end, delta, idx;
1449 loff_t new_size;
1450 int ret = 0;
1451
1452 new_size = i_size_read(inode) + len;
1453 ret = inode_newsize_ok(inode, new_size);
1454 if (ret)
1455 return ret;
1456
1457 if (offset >= i_size_read(inode))
1458 return -EINVAL;
1459
1460 /* insert range should be aligned to block size of f2fs. */
1461 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1462 return -EINVAL;
1463
1464 ret = f2fs_convert_inline_inode(inode);
1465 if (ret)
1466 return ret;
1467
1468 f2fs_balance_fs(sbi, true);
1469
1470 down_write(&F2FS_I(inode)->i_mmap_sem);
1471 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1472 up_write(&F2FS_I(inode)->i_mmap_sem);
1473 if (ret)
1474 return ret;
1475
1476 /* write out all dirty pages from offset */
1477 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1478 if (ret)
1479 return ret;
1480
1481 pg_start = offset >> PAGE_SHIFT;
1482 pg_end = (offset + len) >> PAGE_SHIFT;
1483 delta = pg_end - pg_start;
1484 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1485
1486 /* avoid gc operation during block exchange */
1487 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1488 down_write(&F2FS_I(inode)->i_mmap_sem);
1489 truncate_pagecache(inode, offset);
1490
1491 while (!ret && idx > pg_start) {
1492 nr = idx - pg_start;
1493 if (nr > delta)
1494 nr = delta;
1495 idx -= nr;
1496
1497 f2fs_lock_op(sbi);
1498 f2fs_drop_extent_tree(inode);
1499
1500 ret = __exchange_data_block(inode, inode, idx,
1501 idx + delta, nr, false);
1502 f2fs_unlock_op(sbi);
1503 }
1504 up_write(&F2FS_I(inode)->i_mmap_sem);
1505 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1506
1507 /* write out all moved pages, if possible */
1508 down_write(&F2FS_I(inode)->i_mmap_sem);
1509 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1510 truncate_pagecache(inode, offset);
1511 up_write(&F2FS_I(inode)->i_mmap_sem);
1512
1513 if (!ret)
1514 f2fs_i_size_write(inode, new_size);
1515 return ret;
1516}
1517
1518static int expand_inode_data(struct inode *inode, loff_t offset,
1519 loff_t len, int mode)
1520{
1521 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1522 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1523 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1524 .m_may_create = true };
1525 pgoff_t pg_end;
1526 loff_t new_size = i_size_read(inode);
1527 loff_t off_end;
1528 int err;
1529
1530 err = inode_newsize_ok(inode, (len + offset));
1531 if (err)
1532 return err;
1533
1534 err = f2fs_convert_inline_inode(inode);
1535 if (err)
1536 return err;
1537
1538 f2fs_balance_fs(sbi, true);
1539
1540 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1541 off_end = (offset + len) & (PAGE_SIZE - 1);
1542
1543 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1544 map.m_len = pg_end - map.m_lblk;
1545 if (off_end)
1546 map.m_len++;
1547
1548 if (f2fs_is_pinned_file(inode))
1549 map.m_seg_type = CURSEG_COLD_DATA;
1550
1551 err = f2fs_map_blocks(inode, &map, 1, (f2fs_is_pinned_file(inode) ?
1552 F2FS_GET_BLOCK_PRE_DIO :
1553 F2FS_GET_BLOCK_PRE_AIO));
1554 if (err) {
1555 pgoff_t last_off;
1556
1557 if (!map.m_len)
1558 return err;
1559
1560 last_off = map.m_lblk + map.m_len - 1;
1561
1562 /* update new size to the failed position */
1563 new_size = (last_off == pg_end) ? offset + len :
1564 (loff_t)(last_off + 1) << PAGE_SHIFT;
1565 } else {
1566 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1567 }
1568
1569 if (new_size > i_size_read(inode)) {
1570 if (mode & FALLOC_FL_KEEP_SIZE)
1571 file_set_keep_isize(inode);
1572 else
1573 f2fs_i_size_write(inode, new_size);
1574 }
1575
1576 return err;
1577}
1578
1579static long f2fs_fallocate(struct file *file, int mode,
1580 loff_t offset, loff_t len)
1581{
1582 struct inode *inode = file_inode(file);
1583 long ret = 0;
1584
1585 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1586 return -EIO;
1587 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1588 return -ENOSPC;
1589
1590 /* f2fs only support ->fallocate for regular file */
1591 if (!S_ISREG(inode->i_mode))
1592 return -EINVAL;
1593
1594 if (IS_ENCRYPTED(inode) &&
1595 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1596 return -EOPNOTSUPP;
1597
1598 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1599 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1600 FALLOC_FL_INSERT_RANGE))
1601 return -EOPNOTSUPP;
1602
1603 inode_lock(inode);
1604
1605 if (mode & FALLOC_FL_PUNCH_HOLE) {
1606 if (offset >= inode->i_size)
1607 goto out;
1608
1609 ret = punch_hole(inode, offset, len);
1610 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1611 ret = f2fs_collapse_range(inode, offset, len);
1612 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1613 ret = f2fs_zero_range(inode, offset, len, mode);
1614 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1615 ret = f2fs_insert_range(inode, offset, len);
1616 } else {
1617 ret = expand_inode_data(inode, offset, len, mode);
1618 }
1619
1620 if (!ret) {
1621 inode->i_mtime = inode->i_ctime = current_time(inode);
1622 f2fs_mark_inode_dirty_sync(inode, false);
1623 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1624 }
1625
1626out:
1627 inode_unlock(inode);
1628
1629 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1630 return ret;
1631}
1632
1633static int f2fs_release_file(struct inode *inode, struct file *filp)
1634{
1635 /*
1636 * f2fs_relase_file is called at every close calls. So we should
1637 * not drop any inmemory pages by close called by other process.
1638 */
1639 if (!(filp->f_mode & FMODE_WRITE) ||
1640 atomic_read(&inode->i_writecount) != 1)
1641 return 0;
1642
1643 /* some remained atomic pages should discarded */
1644 if (f2fs_is_atomic_file(inode))
1645 f2fs_drop_inmem_pages(inode);
1646 if (f2fs_is_volatile_file(inode)) {
1647 set_inode_flag(inode, FI_DROP_CACHE);
1648 filemap_fdatawrite(inode->i_mapping);
1649 clear_inode_flag(inode, FI_DROP_CACHE);
1650 clear_inode_flag(inode, FI_VOLATILE_FILE);
1651 stat_dec_volatile_write(inode);
1652 }
1653 return 0;
1654}
1655
1656static int f2fs_file_flush(struct file *file, fl_owner_t id)
1657{
1658 struct inode *inode = file_inode(file);
1659
1660 /*
1661 * If the process doing a transaction is crashed, we should do
1662 * roll-back. Otherwise, other reader/write can see corrupted database
1663 * until all the writers close its file. Since this should be done
1664 * before dropping file lock, it needs to do in ->flush.
1665 */
1666 if (f2fs_is_atomic_file(inode) &&
1667 F2FS_I(inode)->inmem_task == current)
1668 f2fs_drop_inmem_pages(inode);
1669 return 0;
1670}
1671
1672static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1673{
1674 struct f2fs_inode_info *fi = F2FS_I(inode);
1675
1676 /* Is it quota file? Do not allow user to mess with it */
1677 if (IS_NOQUOTA(inode))
1678 return -EPERM;
1679
1680 if ((iflags ^ fi->i_flags) & F2FS_CASEFOLD_FL) {
1681 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1682 return -EOPNOTSUPP;
1683 if (!f2fs_empty_dir(inode))
1684 return -ENOTEMPTY;
1685 }
1686
1687 fi->i_flags = iflags | (fi->i_flags & ~mask);
1688
1689 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1690 set_inode_flag(inode, FI_PROJ_INHERIT);
1691 else
1692 clear_inode_flag(inode, FI_PROJ_INHERIT);
1693
1694 inode->i_ctime = current_time(inode);
1695 f2fs_set_inode_flags(inode);
1696 f2fs_mark_inode_dirty_sync(inode, true);
1697 return 0;
1698}
1699
1700/* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1701
1702/*
1703 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1704 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1705 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1706 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1707 */
1708
1709static const struct {
1710 u32 iflag;
1711 u32 fsflag;
1712} f2fs_fsflags_map[] = {
1713 { F2FS_SYNC_FL, FS_SYNC_FL },
1714 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1715 { F2FS_APPEND_FL, FS_APPEND_FL },
1716 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1717 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1718 { F2FS_INDEX_FL, FS_INDEX_FL },
1719 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1720 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1721 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1722};
1723
1724#define F2FS_GETTABLE_FS_FL ( \
1725 FS_SYNC_FL | \
1726 FS_IMMUTABLE_FL | \
1727 FS_APPEND_FL | \
1728 FS_NODUMP_FL | \
1729 FS_NOATIME_FL | \
1730 FS_INDEX_FL | \
1731 FS_DIRSYNC_FL | \
1732 FS_PROJINHERIT_FL | \
1733 FS_ENCRYPT_FL | \
1734 FS_INLINE_DATA_FL | \
1735 FS_NOCOW_FL | \
1736 FS_VERITY_FL | \
1737 FS_CASEFOLD_FL)
1738
1739#define F2FS_SETTABLE_FS_FL ( \
1740 FS_SYNC_FL | \
1741 FS_IMMUTABLE_FL | \
1742 FS_APPEND_FL | \
1743 FS_NODUMP_FL | \
1744 FS_NOATIME_FL | \
1745 FS_DIRSYNC_FL | \
1746 FS_PROJINHERIT_FL | \
1747 FS_CASEFOLD_FL)
1748
1749/* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1750static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1751{
1752 u32 fsflags = 0;
1753 int i;
1754
1755 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1756 if (iflags & f2fs_fsflags_map[i].iflag)
1757 fsflags |= f2fs_fsflags_map[i].fsflag;
1758
1759 return fsflags;
1760}
1761
1762/* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1763static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1764{
1765 u32 iflags = 0;
1766 int i;
1767
1768 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1769 if (fsflags & f2fs_fsflags_map[i].fsflag)
1770 iflags |= f2fs_fsflags_map[i].iflag;
1771
1772 return iflags;
1773}
1774
1775static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1776{
1777 struct inode *inode = file_inode(filp);
1778 struct f2fs_inode_info *fi = F2FS_I(inode);
1779 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1780
1781 if (IS_ENCRYPTED(inode))
1782 fsflags |= FS_ENCRYPT_FL;
1783 if (IS_VERITY(inode))
1784 fsflags |= FS_VERITY_FL;
1785 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1786 fsflags |= FS_INLINE_DATA_FL;
1787 if (is_inode_flag_set(inode, FI_PIN_FILE))
1788 fsflags |= FS_NOCOW_FL;
1789
1790 fsflags &= F2FS_GETTABLE_FS_FL;
1791
1792 return put_user(fsflags, (int __user *)arg);
1793}
1794
1795static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1796{
1797 struct inode *inode = file_inode(filp);
1798 struct f2fs_inode_info *fi = F2FS_I(inode);
1799 u32 fsflags, old_fsflags;
1800 u32 iflags;
1801 int ret;
1802
1803 if (!inode_owner_or_capable(inode))
1804 return -EACCES;
1805
1806 if (get_user(fsflags, (int __user *)arg))
1807 return -EFAULT;
1808
1809 if (fsflags & ~F2FS_GETTABLE_FS_FL)
1810 return -EOPNOTSUPP;
1811 fsflags &= F2FS_SETTABLE_FS_FL;
1812
1813 iflags = f2fs_fsflags_to_iflags(fsflags);
1814 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1815 return -EOPNOTSUPP;
1816
1817 ret = mnt_want_write_file(filp);
1818 if (ret)
1819 return ret;
1820
1821 inode_lock(inode);
1822
1823 old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1824 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1825 if (ret)
1826 goto out;
1827
1828 ret = f2fs_setflags_common(inode, iflags,
1829 f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1830out:
1831 inode_unlock(inode);
1832 mnt_drop_write_file(filp);
1833 return ret;
1834}
1835
1836static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1837{
1838 struct inode *inode = file_inode(filp);
1839
1840 return put_user(inode->i_generation, (int __user *)arg);
1841}
1842
1843static int f2fs_ioc_start_atomic_write(struct file *filp)
1844{
1845 struct inode *inode = file_inode(filp);
1846 struct f2fs_inode_info *fi = F2FS_I(inode);
1847 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1848 int ret;
1849
1850 if (!inode_owner_or_capable(inode))
1851 return -EACCES;
1852
1853 if (!S_ISREG(inode->i_mode))
1854 return -EINVAL;
1855
1856 if (filp->f_flags & O_DIRECT)
1857 return -EINVAL;
1858
1859 ret = mnt_want_write_file(filp);
1860 if (ret)
1861 return ret;
1862
1863 inode_lock(inode);
1864
1865 if (f2fs_is_atomic_file(inode)) {
1866 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1867 ret = -EINVAL;
1868 goto out;
1869 }
1870
1871 ret = f2fs_convert_inline_inode(inode);
1872 if (ret)
1873 goto out;
1874
1875 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1876
1877 /*
1878 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1879 * f2fs_is_atomic_file.
1880 */
1881 if (get_dirty_pages(inode))
1882 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1883 inode->i_ino, get_dirty_pages(inode));
1884 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1885 if (ret) {
1886 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1887 goto out;
1888 }
1889
1890 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
1891 if (list_empty(&fi->inmem_ilist))
1892 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
1893 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
1894
1895 /* add inode in inmem_list first and set atomic_file */
1896 set_inode_flag(inode, FI_ATOMIC_FILE);
1897 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1898 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1899
1900 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1901 F2FS_I(inode)->inmem_task = current;
1902 stat_inc_atomic_write(inode);
1903 stat_update_max_atomic_write(inode);
1904out:
1905 inode_unlock(inode);
1906 mnt_drop_write_file(filp);
1907 return ret;
1908}
1909
1910static int f2fs_ioc_commit_atomic_write(struct file *filp)
1911{
1912 struct inode *inode = file_inode(filp);
1913 int ret;
1914
1915 if (!inode_owner_or_capable(inode))
1916 return -EACCES;
1917
1918 ret = mnt_want_write_file(filp);
1919 if (ret)
1920 return ret;
1921
1922 f2fs_balance_fs(F2FS_I_SB(inode), true);
1923
1924 inode_lock(inode);
1925
1926 if (f2fs_is_volatile_file(inode)) {
1927 ret = -EINVAL;
1928 goto err_out;
1929 }
1930
1931 if (f2fs_is_atomic_file(inode)) {
1932 ret = f2fs_commit_inmem_pages(inode);
1933 if (ret)
1934 goto err_out;
1935
1936 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1937 if (!ret)
1938 f2fs_drop_inmem_pages(inode);
1939 } else {
1940 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1941 }
1942err_out:
1943 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1944 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1945 ret = -EINVAL;
1946 }
1947 inode_unlock(inode);
1948 mnt_drop_write_file(filp);
1949 return ret;
1950}
1951
1952static int f2fs_ioc_start_volatile_write(struct file *filp)
1953{
1954 struct inode *inode = file_inode(filp);
1955 int ret;
1956
1957 if (!inode_owner_or_capable(inode))
1958 return -EACCES;
1959
1960 if (!S_ISREG(inode->i_mode))
1961 return -EINVAL;
1962
1963 ret = mnt_want_write_file(filp);
1964 if (ret)
1965 return ret;
1966
1967 inode_lock(inode);
1968
1969 if (f2fs_is_volatile_file(inode))
1970 goto out;
1971
1972 ret = f2fs_convert_inline_inode(inode);
1973 if (ret)
1974 goto out;
1975
1976 stat_inc_volatile_write(inode);
1977 stat_update_max_volatile_write(inode);
1978
1979 set_inode_flag(inode, FI_VOLATILE_FILE);
1980 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1981out:
1982 inode_unlock(inode);
1983 mnt_drop_write_file(filp);
1984 return ret;
1985}
1986
1987static int f2fs_ioc_release_volatile_write(struct file *filp)
1988{
1989 struct inode *inode = file_inode(filp);
1990 int ret;
1991
1992 if (!inode_owner_or_capable(inode))
1993 return -EACCES;
1994
1995 ret = mnt_want_write_file(filp);
1996 if (ret)
1997 return ret;
1998
1999 inode_lock(inode);
2000
2001 if (!f2fs_is_volatile_file(inode))
2002 goto out;
2003
2004 if (!f2fs_is_first_block_written(inode)) {
2005 ret = truncate_partial_data_page(inode, 0, true);
2006 goto out;
2007 }
2008
2009 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2010out:
2011 inode_unlock(inode);
2012 mnt_drop_write_file(filp);
2013 return ret;
2014}
2015
2016static int f2fs_ioc_abort_volatile_write(struct file *filp)
2017{
2018 struct inode *inode = file_inode(filp);
2019 int ret;
2020
2021 if (!inode_owner_or_capable(inode))
2022 return -EACCES;
2023
2024 ret = mnt_want_write_file(filp);
2025 if (ret)
2026 return ret;
2027
2028 inode_lock(inode);
2029
2030 if (f2fs_is_atomic_file(inode))
2031 f2fs_drop_inmem_pages(inode);
2032 if (f2fs_is_volatile_file(inode)) {
2033 clear_inode_flag(inode, FI_VOLATILE_FILE);
2034 stat_dec_volatile_write(inode);
2035 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2036 }
2037
2038 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2039
2040 inode_unlock(inode);
2041
2042 mnt_drop_write_file(filp);
2043 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2044 return ret;
2045}
2046
2047static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2048{
2049 struct inode *inode = file_inode(filp);
2050 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2051 struct super_block *sb = sbi->sb;
2052 __u32 in;
2053 int ret = 0;
2054
2055 if (!capable(CAP_SYS_ADMIN))
2056 return -EPERM;
2057
2058 if (get_user(in, (__u32 __user *)arg))
2059 return -EFAULT;
2060
2061 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2062 ret = mnt_want_write_file(filp);
2063 if (ret)
2064 return ret;
2065 }
2066
2067 switch (in) {
2068 case F2FS_GOING_DOWN_FULLSYNC:
2069 sb = freeze_bdev(sb->s_bdev);
2070 if (IS_ERR(sb)) {
2071 ret = PTR_ERR(sb);
2072 goto out;
2073 }
2074 if (sb) {
2075 f2fs_stop_checkpoint(sbi, false);
2076 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2077 thaw_bdev(sb->s_bdev, sb);
2078 }
2079 break;
2080 case F2FS_GOING_DOWN_METASYNC:
2081 /* do checkpoint only */
2082 ret = f2fs_sync_fs(sb, 1);
2083 if (ret)
2084 goto out;
2085 f2fs_stop_checkpoint(sbi, false);
2086 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2087 break;
2088 case F2FS_GOING_DOWN_NOSYNC:
2089 f2fs_stop_checkpoint(sbi, false);
2090 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2091 break;
2092 case F2FS_GOING_DOWN_METAFLUSH:
2093 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2094 f2fs_stop_checkpoint(sbi, false);
2095 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2096 break;
2097 case F2FS_GOING_DOWN_NEED_FSCK:
2098 set_sbi_flag(sbi, SBI_NEED_FSCK);
2099 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2100 set_sbi_flag(sbi, SBI_IS_DIRTY);
2101 /* do checkpoint only */
2102 ret = f2fs_sync_fs(sb, 1);
2103 goto out;
2104 default:
2105 ret = -EINVAL;
2106 goto out;
2107 }
2108
2109 f2fs_stop_gc_thread(sbi);
2110 f2fs_stop_discard_thread(sbi);
2111
2112 f2fs_drop_discard_cmd(sbi);
2113 clear_opt(sbi, DISCARD);
2114
2115 f2fs_update_time(sbi, REQ_TIME);
2116out:
2117 if (in != F2FS_GOING_DOWN_FULLSYNC)
2118 mnt_drop_write_file(filp);
2119
2120 trace_f2fs_shutdown(sbi, in, ret);
2121
2122 return ret;
2123}
2124
2125static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2126{
2127 struct inode *inode = file_inode(filp);
2128 struct super_block *sb = inode->i_sb;
2129 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2130 struct fstrim_range range;
2131 int ret;
2132
2133 if (!capable(CAP_SYS_ADMIN))
2134 return -EPERM;
2135
2136 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2137 return -EOPNOTSUPP;
2138
2139 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2140 sizeof(range)))
2141 return -EFAULT;
2142
2143 ret = mnt_want_write_file(filp);
2144 if (ret)
2145 return ret;
2146
2147 range.minlen = max((unsigned int)range.minlen,
2148 q->limits.discard_granularity);
2149 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2150 mnt_drop_write_file(filp);
2151 if (ret < 0)
2152 return ret;
2153
2154 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2155 sizeof(range)))
2156 return -EFAULT;
2157 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2158 return 0;
2159}
2160
2161static bool uuid_is_nonzero(__u8 u[16])
2162{
2163 int i;
2164
2165 for (i = 0; i < 16; i++)
2166 if (u[i])
2167 return true;
2168 return false;
2169}
2170
2171static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2172{
2173 struct inode *inode = file_inode(filp);
2174
2175 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2176 return -EOPNOTSUPP;
2177
2178 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2179
2180 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2181}
2182
2183static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2184{
2185 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2186 return -EOPNOTSUPP;
2187 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2188}
2189
2190static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2191{
2192 struct inode *inode = file_inode(filp);
2193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2194 int err;
2195
2196 if (!f2fs_sb_has_encrypt(sbi))
2197 return -EOPNOTSUPP;
2198
2199 err = mnt_want_write_file(filp);
2200 if (err)
2201 return err;
2202
2203 down_write(&sbi->sb_lock);
2204
2205 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2206 goto got_it;
2207
2208 /* update superblock with uuid */
2209 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2210
2211 err = f2fs_commit_super(sbi, false);
2212 if (err) {
2213 /* undo new data */
2214 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2215 goto out_err;
2216 }
2217got_it:
2218 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2219 16))
2220 err = -EFAULT;
2221out_err:
2222 up_write(&sbi->sb_lock);
2223 mnt_drop_write_file(filp);
2224 return err;
2225}
2226
2227static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2228 unsigned long arg)
2229{
2230 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2231 return -EOPNOTSUPP;
2232
2233 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2234}
2235
2236static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2237{
2238 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2239 return -EOPNOTSUPP;
2240
2241 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2242}
2243
2244static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2245{
2246 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2247 return -EOPNOTSUPP;
2248
2249 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2250}
2251
2252static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2253 unsigned long arg)
2254{
2255 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2256 return -EOPNOTSUPP;
2257
2258 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2259}
2260
2261static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2262 unsigned long arg)
2263{
2264 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2265 return -EOPNOTSUPP;
2266
2267 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2268}
2269
2270static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2271{
2272 struct inode *inode = file_inode(filp);
2273 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2274 __u32 sync;
2275 int ret;
2276
2277 if (!capable(CAP_SYS_ADMIN))
2278 return -EPERM;
2279
2280 if (get_user(sync, (__u32 __user *)arg))
2281 return -EFAULT;
2282
2283 if (f2fs_readonly(sbi->sb))
2284 return -EROFS;
2285
2286 ret = mnt_want_write_file(filp);
2287 if (ret)
2288 return ret;
2289
2290 if (!sync) {
2291 if (!mutex_trylock(&sbi->gc_mutex)) {
2292 ret = -EBUSY;
2293 goto out;
2294 }
2295 } else {
2296 mutex_lock(&sbi->gc_mutex);
2297 }
2298
2299 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2300out:
2301 mnt_drop_write_file(filp);
2302 return ret;
2303}
2304
2305static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2306{
2307 struct inode *inode = file_inode(filp);
2308 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2309 struct f2fs_gc_range range;
2310 u64 end;
2311 int ret;
2312
2313 if (!capable(CAP_SYS_ADMIN))
2314 return -EPERM;
2315
2316 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2317 sizeof(range)))
2318 return -EFAULT;
2319
2320 if (f2fs_readonly(sbi->sb))
2321 return -EROFS;
2322
2323 end = range.start + range.len;
2324 if (end < range.start || range.start < MAIN_BLKADDR(sbi) ||
2325 end >= MAX_BLKADDR(sbi))
2326 return -EINVAL;
2327
2328 ret = mnt_want_write_file(filp);
2329 if (ret)
2330 return ret;
2331
2332do_more:
2333 if (!range.sync) {
2334 if (!mutex_trylock(&sbi->gc_mutex)) {
2335 ret = -EBUSY;
2336 goto out;
2337 }
2338 } else {
2339 mutex_lock(&sbi->gc_mutex);
2340 }
2341
2342 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2343 range.start += BLKS_PER_SEC(sbi);
2344 if (range.start <= end)
2345 goto do_more;
2346out:
2347 mnt_drop_write_file(filp);
2348 return ret;
2349}
2350
2351static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2352{
2353 struct inode *inode = file_inode(filp);
2354 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2355 int ret;
2356
2357 if (!capable(CAP_SYS_ADMIN))
2358 return -EPERM;
2359
2360 if (f2fs_readonly(sbi->sb))
2361 return -EROFS;
2362
2363 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2364 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2365 return -EINVAL;
2366 }
2367
2368 ret = mnt_want_write_file(filp);
2369 if (ret)
2370 return ret;
2371
2372 ret = f2fs_sync_fs(sbi->sb, 1);
2373
2374 mnt_drop_write_file(filp);
2375 return ret;
2376}
2377
2378static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2379 struct file *filp,
2380 struct f2fs_defragment *range)
2381{
2382 struct inode *inode = file_inode(filp);
2383 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2384 .m_seg_type = NO_CHECK_TYPE ,
2385 .m_may_create = false };
2386 struct extent_info ei = {0, 0, 0};
2387 pgoff_t pg_start, pg_end, next_pgofs;
2388 unsigned int blk_per_seg = sbi->blocks_per_seg;
2389 unsigned int total = 0, sec_num;
2390 block_t blk_end = 0;
2391 bool fragmented = false;
2392 int err;
2393
2394 /* if in-place-update policy is enabled, don't waste time here */
2395 if (f2fs_should_update_inplace(inode, NULL))
2396 return -EINVAL;
2397
2398 pg_start = range->start >> PAGE_SHIFT;
2399 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2400
2401 f2fs_balance_fs(sbi, true);
2402
2403 inode_lock(inode);
2404
2405 /* writeback all dirty pages in the range */
2406 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2407 range->start + range->len - 1);
2408 if (err)
2409 goto out;
2410
2411 /*
2412 * lookup mapping info in extent cache, skip defragmenting if physical
2413 * block addresses are continuous.
2414 */
2415 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2416 if (ei.fofs + ei.len >= pg_end)
2417 goto out;
2418 }
2419
2420 map.m_lblk = pg_start;
2421 map.m_next_pgofs = &next_pgofs;
2422
2423 /*
2424 * lookup mapping info in dnode page cache, skip defragmenting if all
2425 * physical block addresses are continuous even if there are hole(s)
2426 * in logical blocks.
2427 */
2428 while (map.m_lblk < pg_end) {
2429 map.m_len = pg_end - map.m_lblk;
2430 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2431 if (err)
2432 goto out;
2433
2434 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2435 map.m_lblk = next_pgofs;
2436 continue;
2437 }
2438
2439 if (blk_end && blk_end != map.m_pblk)
2440 fragmented = true;
2441
2442 /* record total count of block that we're going to move */
2443 total += map.m_len;
2444
2445 blk_end = map.m_pblk + map.m_len;
2446
2447 map.m_lblk += map.m_len;
2448 }
2449
2450 if (!fragmented) {
2451 total = 0;
2452 goto out;
2453 }
2454
2455 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2456
2457 /*
2458 * make sure there are enough free section for LFS allocation, this can
2459 * avoid defragment running in SSR mode when free section are allocated
2460 * intensively
2461 */
2462 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2463 err = -EAGAIN;
2464 goto out;
2465 }
2466
2467 map.m_lblk = pg_start;
2468 map.m_len = pg_end - pg_start;
2469 total = 0;
2470
2471 while (map.m_lblk < pg_end) {
2472 pgoff_t idx;
2473 int cnt = 0;
2474
2475do_map:
2476 map.m_len = pg_end - map.m_lblk;
2477 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2478 if (err)
2479 goto clear_out;
2480
2481 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2482 map.m_lblk = next_pgofs;
2483 goto check;
2484 }
2485
2486 set_inode_flag(inode, FI_DO_DEFRAG);
2487
2488 idx = map.m_lblk;
2489 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2490 struct page *page;
2491
2492 page = f2fs_get_lock_data_page(inode, idx, true);
2493 if (IS_ERR(page)) {
2494 err = PTR_ERR(page);
2495 goto clear_out;
2496 }
2497
2498 set_page_dirty(page);
2499 f2fs_put_page(page, 1);
2500
2501 idx++;
2502 cnt++;
2503 total++;
2504 }
2505
2506 map.m_lblk = idx;
2507check:
2508 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2509 goto do_map;
2510
2511 clear_inode_flag(inode, FI_DO_DEFRAG);
2512
2513 err = filemap_fdatawrite(inode->i_mapping);
2514 if (err)
2515 goto out;
2516 }
2517clear_out:
2518 clear_inode_flag(inode, FI_DO_DEFRAG);
2519out:
2520 inode_unlock(inode);
2521 if (!err)
2522 range->len = (u64)total << PAGE_SHIFT;
2523 return err;
2524}
2525
2526static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2527{
2528 struct inode *inode = file_inode(filp);
2529 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2530 struct f2fs_defragment range;
2531 int err;
2532
2533 if (!capable(CAP_SYS_ADMIN))
2534 return -EPERM;
2535
2536 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2537 return -EINVAL;
2538
2539 if (f2fs_readonly(sbi->sb))
2540 return -EROFS;
2541
2542 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2543 sizeof(range)))
2544 return -EFAULT;
2545
2546 /* verify alignment of offset & size */
2547 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2548 return -EINVAL;
2549
2550 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2551 sbi->max_file_blocks))
2552 return -EINVAL;
2553
2554 err = mnt_want_write_file(filp);
2555 if (err)
2556 return err;
2557
2558 err = f2fs_defragment_range(sbi, filp, &range);
2559 mnt_drop_write_file(filp);
2560
2561 f2fs_update_time(sbi, REQ_TIME);
2562 if (err < 0)
2563 return err;
2564
2565 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2566 sizeof(range)))
2567 return -EFAULT;
2568
2569 return 0;
2570}
2571
2572static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2573 struct file *file_out, loff_t pos_out, size_t len)
2574{
2575 struct inode *src = file_inode(file_in);
2576 struct inode *dst = file_inode(file_out);
2577 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2578 size_t olen = len, dst_max_i_size = 0;
2579 size_t dst_osize;
2580 int ret;
2581
2582 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2583 src->i_sb != dst->i_sb)
2584 return -EXDEV;
2585
2586 if (unlikely(f2fs_readonly(src->i_sb)))
2587 return -EROFS;
2588
2589 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2590 return -EINVAL;
2591
2592 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2593 return -EOPNOTSUPP;
2594
2595 if (src == dst) {
2596 if (pos_in == pos_out)
2597 return 0;
2598 if (pos_out > pos_in && pos_out < pos_in + len)
2599 return -EINVAL;
2600 }
2601
2602 inode_lock(src);
2603 if (src != dst) {
2604 ret = -EBUSY;
2605 if (!inode_trylock(dst))
2606 goto out;
2607 }
2608
2609 ret = -EINVAL;
2610 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2611 goto out_unlock;
2612 if (len == 0)
2613 olen = len = src->i_size - pos_in;
2614 if (pos_in + len == src->i_size)
2615 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2616 if (len == 0) {
2617 ret = 0;
2618 goto out_unlock;
2619 }
2620
2621 dst_osize = dst->i_size;
2622 if (pos_out + olen > dst->i_size)
2623 dst_max_i_size = pos_out + olen;
2624
2625 /* verify the end result is block aligned */
2626 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2627 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2628 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2629 goto out_unlock;
2630
2631 ret = f2fs_convert_inline_inode(src);
2632 if (ret)
2633 goto out_unlock;
2634
2635 ret = f2fs_convert_inline_inode(dst);
2636 if (ret)
2637 goto out_unlock;
2638
2639 /* write out all dirty pages from offset */
2640 ret = filemap_write_and_wait_range(src->i_mapping,
2641 pos_in, pos_in + len);
2642 if (ret)
2643 goto out_unlock;
2644
2645 ret = filemap_write_and_wait_range(dst->i_mapping,
2646 pos_out, pos_out + len);
2647 if (ret)
2648 goto out_unlock;
2649
2650 f2fs_balance_fs(sbi, true);
2651
2652 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2653 if (src != dst) {
2654 ret = -EBUSY;
2655 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2656 goto out_src;
2657 }
2658
2659 f2fs_lock_op(sbi);
2660 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2661 pos_out >> F2FS_BLKSIZE_BITS,
2662 len >> F2FS_BLKSIZE_BITS, false);
2663
2664 if (!ret) {
2665 if (dst_max_i_size)
2666 f2fs_i_size_write(dst, dst_max_i_size);
2667 else if (dst_osize != dst->i_size)
2668 f2fs_i_size_write(dst, dst_osize);
2669 }
2670 f2fs_unlock_op(sbi);
2671
2672 if (src != dst)
2673 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2674out_src:
2675 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2676out_unlock:
2677 if (src != dst)
2678 inode_unlock(dst);
2679out:
2680 inode_unlock(src);
2681 return ret;
2682}
2683
2684static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2685{
2686 struct f2fs_move_range range;
2687 struct fd dst;
2688 int err;
2689
2690 if (!(filp->f_mode & FMODE_READ) ||
2691 !(filp->f_mode & FMODE_WRITE))
2692 return -EBADF;
2693
2694 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2695 sizeof(range)))
2696 return -EFAULT;
2697
2698 dst = fdget(range.dst_fd);
2699 if (!dst.file)
2700 return -EBADF;
2701
2702 if (!(dst.file->f_mode & FMODE_WRITE)) {
2703 err = -EBADF;
2704 goto err_out;
2705 }
2706
2707 err = mnt_want_write_file(filp);
2708 if (err)
2709 goto err_out;
2710
2711 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2712 range.pos_out, range.len);
2713
2714 mnt_drop_write_file(filp);
2715 if (err)
2716 goto err_out;
2717
2718 if (copy_to_user((struct f2fs_move_range __user *)arg,
2719 &range, sizeof(range)))
2720 err = -EFAULT;
2721err_out:
2722 fdput(dst);
2723 return err;
2724}
2725
2726static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2727{
2728 struct inode *inode = file_inode(filp);
2729 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2730 struct sit_info *sm = SIT_I(sbi);
2731 unsigned int start_segno = 0, end_segno = 0;
2732 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2733 struct f2fs_flush_device range;
2734 int ret;
2735
2736 if (!capable(CAP_SYS_ADMIN))
2737 return -EPERM;
2738
2739 if (f2fs_readonly(sbi->sb))
2740 return -EROFS;
2741
2742 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2743 return -EINVAL;
2744
2745 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2746 sizeof(range)))
2747 return -EFAULT;
2748
2749 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2750 __is_large_section(sbi)) {
2751 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2752 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2753 return -EINVAL;
2754 }
2755
2756 ret = mnt_want_write_file(filp);
2757 if (ret)
2758 return ret;
2759
2760 if (range.dev_num != 0)
2761 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2762 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2763
2764 start_segno = sm->last_victim[FLUSH_DEVICE];
2765 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2766 start_segno = dev_start_segno;
2767 end_segno = min(start_segno + range.segments, dev_end_segno);
2768
2769 while (start_segno < end_segno) {
2770 if (!mutex_trylock(&sbi->gc_mutex)) {
2771 ret = -EBUSY;
2772 goto out;
2773 }
2774 sm->last_victim[GC_CB] = end_segno + 1;
2775 sm->last_victim[GC_GREEDY] = end_segno + 1;
2776 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2777 ret = f2fs_gc(sbi, true, true, start_segno);
2778 if (ret == -EAGAIN)
2779 ret = 0;
2780 else if (ret < 0)
2781 break;
2782 start_segno++;
2783 }
2784out:
2785 mnt_drop_write_file(filp);
2786 return ret;
2787}
2788
2789static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2790{
2791 struct inode *inode = file_inode(filp);
2792 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2793
2794 /* Must validate to set it with SQLite behavior in Android. */
2795 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2796
2797 return put_user(sb_feature, (u32 __user *)arg);
2798}
2799
2800#ifdef CONFIG_QUOTA
2801int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2802{
2803 struct dquot *transfer_to[MAXQUOTAS] = {};
2804 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2805 struct super_block *sb = sbi->sb;
2806 int err = 0;
2807
2808 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2809 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2810 err = __dquot_transfer(inode, transfer_to);
2811 if (err)
2812 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2813 dqput(transfer_to[PRJQUOTA]);
2814 }
2815 return err;
2816}
2817
2818static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2819{
2820 struct inode *inode = file_inode(filp);
2821 struct f2fs_inode_info *fi = F2FS_I(inode);
2822 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2823 struct page *ipage;
2824 kprojid_t kprojid;
2825 int err;
2826
2827 if (!f2fs_sb_has_project_quota(sbi)) {
2828 if (projid != F2FS_DEF_PROJID)
2829 return -EOPNOTSUPP;
2830 else
2831 return 0;
2832 }
2833
2834 if (!f2fs_has_extra_attr(inode))
2835 return -EOPNOTSUPP;
2836
2837 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2838
2839 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2840 return 0;
2841
2842 err = -EPERM;
2843 /* Is it quota file? Do not allow user to mess with it */
2844 if (IS_NOQUOTA(inode))
2845 return err;
2846
2847 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2848 if (IS_ERR(ipage))
2849 return PTR_ERR(ipage);
2850
2851 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2852 i_projid)) {
2853 err = -EOVERFLOW;
2854 f2fs_put_page(ipage, 1);
2855 return err;
2856 }
2857 f2fs_put_page(ipage, 1);
2858
2859 err = dquot_initialize(inode);
2860 if (err)
2861 return err;
2862
2863 f2fs_lock_op(sbi);
2864 err = f2fs_transfer_project_quota(inode, kprojid);
2865 if (err)
2866 goto out_unlock;
2867
2868 F2FS_I(inode)->i_projid = kprojid;
2869 inode->i_ctime = current_time(inode);
2870 f2fs_mark_inode_dirty_sync(inode, true);
2871out_unlock:
2872 f2fs_unlock_op(sbi);
2873 return err;
2874}
2875#else
2876int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2877{
2878 return 0;
2879}
2880
2881static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2882{
2883 if (projid != F2FS_DEF_PROJID)
2884 return -EOPNOTSUPP;
2885 return 0;
2886}
2887#endif
2888
2889/* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
2890
2891/*
2892 * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
2893 * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
2894 * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
2895 */
2896
2897static const struct {
2898 u32 iflag;
2899 u32 xflag;
2900} f2fs_xflags_map[] = {
2901 { F2FS_SYNC_FL, FS_XFLAG_SYNC },
2902 { F2FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE },
2903 { F2FS_APPEND_FL, FS_XFLAG_APPEND },
2904 { F2FS_NODUMP_FL, FS_XFLAG_NODUMP },
2905 { F2FS_NOATIME_FL, FS_XFLAG_NOATIME },
2906 { F2FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT },
2907};
2908
2909#define F2FS_SUPPORTED_XFLAGS ( \
2910 FS_XFLAG_SYNC | \
2911 FS_XFLAG_IMMUTABLE | \
2912 FS_XFLAG_APPEND | \
2913 FS_XFLAG_NODUMP | \
2914 FS_XFLAG_NOATIME | \
2915 FS_XFLAG_PROJINHERIT)
2916
2917/* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
2918static inline u32 f2fs_iflags_to_xflags(u32 iflags)
2919{
2920 u32 xflags = 0;
2921 int i;
2922
2923 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2924 if (iflags & f2fs_xflags_map[i].iflag)
2925 xflags |= f2fs_xflags_map[i].xflag;
2926
2927 return xflags;
2928}
2929
2930/* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
2931static inline u32 f2fs_xflags_to_iflags(u32 xflags)
2932{
2933 u32 iflags = 0;
2934 int i;
2935
2936 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
2937 if (xflags & f2fs_xflags_map[i].xflag)
2938 iflags |= f2fs_xflags_map[i].iflag;
2939
2940 return iflags;
2941}
2942
2943static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
2944{
2945 struct f2fs_inode_info *fi = F2FS_I(inode);
2946
2947 simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
2948
2949 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2950 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2951}
2952
2953static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2954{
2955 struct inode *inode = file_inode(filp);
2956 struct fsxattr fa;
2957
2958 f2fs_fill_fsxattr(inode, &fa);
2959
2960 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2961 return -EFAULT;
2962 return 0;
2963}
2964
2965static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2966{
2967 struct inode *inode = file_inode(filp);
2968 struct fsxattr fa, old_fa;
2969 u32 iflags;
2970 int err;
2971
2972 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2973 return -EFAULT;
2974
2975 /* Make sure caller has proper permission */
2976 if (!inode_owner_or_capable(inode))
2977 return -EACCES;
2978
2979 if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
2980 return -EOPNOTSUPP;
2981
2982 iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2983 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2984 return -EOPNOTSUPP;
2985
2986 err = mnt_want_write_file(filp);
2987 if (err)
2988 return err;
2989
2990 inode_lock(inode);
2991
2992 f2fs_fill_fsxattr(inode, &old_fa);
2993 err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
2994 if (err)
2995 goto out;
2996
2997 err = f2fs_setflags_common(inode, iflags,
2998 f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
2999 if (err)
3000 goto out;
3001
3002 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3003out:
3004 inode_unlock(inode);
3005 mnt_drop_write_file(filp);
3006 return err;
3007}
3008
3009int f2fs_pin_file_control(struct inode *inode, bool inc)
3010{
3011 struct f2fs_inode_info *fi = F2FS_I(inode);
3012 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3013
3014 /* Use i_gc_failures for normal file as a risk signal. */
3015 if (inc)
3016 f2fs_i_gc_failures_write(inode,
3017 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3018
3019 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3020 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3021 __func__, inode->i_ino,
3022 fi->i_gc_failures[GC_FAILURE_PIN]);
3023 clear_inode_flag(inode, FI_PIN_FILE);
3024 return -EAGAIN;
3025 }
3026 return 0;
3027}
3028
3029static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3030{
3031 struct inode *inode = file_inode(filp);
3032 __u32 pin;
3033 int ret = 0;
3034
3035 if (get_user(pin, (__u32 __user *)arg))
3036 return -EFAULT;
3037
3038 if (!S_ISREG(inode->i_mode))
3039 return -EINVAL;
3040
3041 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3042 return -EROFS;
3043
3044 ret = mnt_want_write_file(filp);
3045 if (ret)
3046 return ret;
3047
3048 inode_lock(inode);
3049
3050 if (f2fs_should_update_outplace(inode, NULL)) {
3051 ret = -EINVAL;
3052 goto out;
3053 }
3054
3055 if (!pin) {
3056 clear_inode_flag(inode, FI_PIN_FILE);
3057 f2fs_i_gc_failures_write(inode, 0);
3058 goto done;
3059 }
3060
3061 if (f2fs_pin_file_control(inode, false)) {
3062 ret = -EAGAIN;
3063 goto out;
3064 }
3065 ret = f2fs_convert_inline_inode(inode);
3066 if (ret)
3067 goto out;
3068
3069 set_inode_flag(inode, FI_PIN_FILE);
3070 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3071done:
3072 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3073out:
3074 inode_unlock(inode);
3075 mnt_drop_write_file(filp);
3076 return ret;
3077}
3078
3079static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3080{
3081 struct inode *inode = file_inode(filp);
3082 __u32 pin = 0;
3083
3084 if (is_inode_flag_set(inode, FI_PIN_FILE))
3085 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3086 return put_user(pin, (u32 __user *)arg);
3087}
3088
3089int f2fs_precache_extents(struct inode *inode)
3090{
3091 struct f2fs_inode_info *fi = F2FS_I(inode);
3092 struct f2fs_map_blocks map;
3093 pgoff_t m_next_extent;
3094 loff_t end;
3095 int err;
3096
3097 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3098 return -EOPNOTSUPP;
3099
3100 map.m_lblk = 0;
3101 map.m_next_pgofs = NULL;
3102 map.m_next_extent = &m_next_extent;
3103 map.m_seg_type = NO_CHECK_TYPE;
3104 map.m_may_create = false;
3105 end = F2FS_I_SB(inode)->max_file_blocks;
3106
3107 while (map.m_lblk < end) {
3108 map.m_len = end - map.m_lblk;
3109
3110 down_write(&fi->i_gc_rwsem[WRITE]);
3111 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3112 up_write(&fi->i_gc_rwsem[WRITE]);
3113 if (err)
3114 return err;
3115
3116 map.m_lblk = m_next_extent;
3117 }
3118
3119 return err;
3120}
3121
3122static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3123{
3124 return f2fs_precache_extents(file_inode(filp));
3125}
3126
3127static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3128{
3129 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3130 __u64 block_count;
3131 int ret;
3132
3133 if (!capable(CAP_SYS_ADMIN))
3134 return -EPERM;
3135
3136 if (f2fs_readonly(sbi->sb))
3137 return -EROFS;
3138
3139 if (copy_from_user(&block_count, (void __user *)arg,
3140 sizeof(block_count)))
3141 return -EFAULT;
3142
3143 ret = f2fs_resize_fs(sbi, block_count);
3144
3145 return ret;
3146}
3147
3148static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3149{
3150 struct inode *inode = file_inode(filp);
3151
3152 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3153
3154 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3155 f2fs_warn(F2FS_I_SB(inode),
3156 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3157 inode->i_ino);
3158 return -EOPNOTSUPP;
3159 }
3160
3161 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3162}
3163
3164static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3165{
3166 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3167 return -EOPNOTSUPP;
3168
3169 return fsverity_ioctl_measure(filp, (void __user *)arg);
3170}
3171
3172static int f2fs_get_volume_name(struct file *filp, unsigned long arg)
3173{
3174 struct inode *inode = file_inode(filp);
3175 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3176 char *vbuf;
3177 int count;
3178 int err = 0;
3179
3180 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3181 if (!vbuf)
3182 return -ENOMEM;
3183
3184 down_read(&sbi->sb_lock);
3185 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3186 ARRAY_SIZE(sbi->raw_super->volume_name),
3187 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3188 up_read(&sbi->sb_lock);
3189
3190 if (copy_to_user((char __user *)arg, vbuf,
3191 min(FSLABEL_MAX, count)))
3192 err = -EFAULT;
3193
3194 kvfree(vbuf);
3195 return err;
3196}
3197
3198static int f2fs_set_volume_name(struct file *filp, unsigned long arg)
3199{
3200 struct inode *inode = file_inode(filp);
3201 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3202 char *vbuf;
3203 int err = 0;
3204
3205 if (!capable(CAP_SYS_ADMIN))
3206 return -EPERM;
3207
3208 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3209 if (IS_ERR(vbuf))
3210 return PTR_ERR(vbuf);
3211
3212 err = mnt_want_write_file(filp);
3213 if (err)
3214 goto out;
3215
3216 down_write(&sbi->sb_lock);
3217
3218 memset(sbi->raw_super->volume_name, 0,
3219 sizeof(sbi->raw_super->volume_name));
3220 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3221 sbi->raw_super->volume_name,
3222 ARRAY_SIZE(sbi->raw_super->volume_name));
3223
3224 err = f2fs_commit_super(sbi, false);
3225
3226 up_write(&sbi->sb_lock);
3227
3228 mnt_drop_write_file(filp);
3229out:
3230 kfree(vbuf);
3231 return err;
3232}
3233
3234long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3235{
3236 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3237 return -EIO;
3238 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
3239 return -ENOSPC;
3240
3241 switch (cmd) {
3242 case F2FS_IOC_GETFLAGS:
3243 return f2fs_ioc_getflags(filp, arg);
3244 case F2FS_IOC_SETFLAGS:
3245 return f2fs_ioc_setflags(filp, arg);
3246 case F2FS_IOC_GETVERSION:
3247 return f2fs_ioc_getversion(filp, arg);
3248 case F2FS_IOC_START_ATOMIC_WRITE:
3249 return f2fs_ioc_start_atomic_write(filp);
3250 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3251 return f2fs_ioc_commit_atomic_write(filp);
3252 case F2FS_IOC_START_VOLATILE_WRITE:
3253 return f2fs_ioc_start_volatile_write(filp);
3254 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3255 return f2fs_ioc_release_volatile_write(filp);
3256 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3257 return f2fs_ioc_abort_volatile_write(filp);
3258 case F2FS_IOC_SHUTDOWN:
3259 return f2fs_ioc_shutdown(filp, arg);
3260 case FITRIM:
3261 return f2fs_ioc_fitrim(filp, arg);
3262 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3263 return f2fs_ioc_set_encryption_policy(filp, arg);
3264 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3265 return f2fs_ioc_get_encryption_policy(filp, arg);
3266 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3267 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3268 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3269 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3270 case FS_IOC_ADD_ENCRYPTION_KEY:
3271 return f2fs_ioc_add_encryption_key(filp, arg);
3272 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3273 return f2fs_ioc_remove_encryption_key(filp, arg);
3274 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3275 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3276 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3277 return f2fs_ioc_get_encryption_key_status(filp, arg);
3278 case F2FS_IOC_GARBAGE_COLLECT:
3279 return f2fs_ioc_gc(filp, arg);
3280 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3281 return f2fs_ioc_gc_range(filp, arg);
3282 case F2FS_IOC_WRITE_CHECKPOINT:
3283 return f2fs_ioc_write_checkpoint(filp, arg);
3284 case F2FS_IOC_DEFRAGMENT:
3285 return f2fs_ioc_defragment(filp, arg);
3286 case F2FS_IOC_MOVE_RANGE:
3287 return f2fs_ioc_move_range(filp, arg);
3288 case F2FS_IOC_FLUSH_DEVICE:
3289 return f2fs_ioc_flush_device(filp, arg);
3290 case F2FS_IOC_GET_FEATURES:
3291 return f2fs_ioc_get_features(filp, arg);
3292 case F2FS_IOC_FSGETXATTR:
3293 return f2fs_ioc_fsgetxattr(filp, arg);
3294 case F2FS_IOC_FSSETXATTR:
3295 return f2fs_ioc_fssetxattr(filp, arg);
3296 case F2FS_IOC_GET_PIN_FILE:
3297 return f2fs_ioc_get_pin_file(filp, arg);
3298 case F2FS_IOC_SET_PIN_FILE:
3299 return f2fs_ioc_set_pin_file(filp, arg);
3300 case F2FS_IOC_PRECACHE_EXTENTS:
3301 return f2fs_ioc_precache_extents(filp, arg);
3302 case F2FS_IOC_RESIZE_FS:
3303 return f2fs_ioc_resize_fs(filp, arg);
3304 case FS_IOC_ENABLE_VERITY:
3305 return f2fs_ioc_enable_verity(filp, arg);
3306 case FS_IOC_MEASURE_VERITY:
3307 return f2fs_ioc_measure_verity(filp, arg);
3308 case F2FS_IOC_GET_VOLUME_NAME:
3309 return f2fs_get_volume_name(filp, arg);
3310 case F2FS_IOC_SET_VOLUME_NAME:
3311 return f2fs_set_volume_name(filp, arg);
3312 default:
3313 return -ENOTTY;
3314 }
3315}
3316
3317static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3318{
3319 struct file *file = iocb->ki_filp;
3320 struct inode *inode = file_inode(file);
3321 ssize_t ret;
3322
3323 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3324 ret = -EIO;
3325 goto out;
3326 }
3327
3328 if (iocb->ki_flags & IOCB_NOWAIT) {
3329 if (!inode_trylock(inode)) {
3330 ret = -EAGAIN;
3331 goto out;
3332 }
3333 } else {
3334 inode_lock(inode);
3335 }
3336
3337 ret = generic_write_checks(iocb, from);
3338 if (ret > 0) {
3339 bool preallocated = false;
3340 size_t target_size = 0;
3341 int err;
3342
3343 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3344 set_inode_flag(inode, FI_NO_PREALLOC);
3345
3346 if ((iocb->ki_flags & IOCB_NOWAIT)) {
3347 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3348 iov_iter_count(from)) ||
3349 f2fs_has_inline_data(inode) ||
3350 f2fs_force_buffered_io(inode, iocb, from)) {
3351 clear_inode_flag(inode, FI_NO_PREALLOC);
3352 inode_unlock(inode);
3353 ret = -EAGAIN;
3354 goto out;
3355 }
3356 } else {
3357 preallocated = true;
3358 target_size = iocb->ki_pos + iov_iter_count(from);
3359
3360 err = f2fs_preallocate_blocks(iocb, from);
3361 if (err) {
3362 clear_inode_flag(inode, FI_NO_PREALLOC);
3363 inode_unlock(inode);
3364 ret = err;
3365 goto out;
3366 }
3367 }
3368 ret = __generic_file_write_iter(iocb, from);
3369 clear_inode_flag(inode, FI_NO_PREALLOC);
3370
3371 /* if we couldn't write data, we should deallocate blocks. */
3372 if (preallocated && i_size_read(inode) < target_size)
3373 f2fs_truncate(inode);
3374
3375 if (ret > 0)
3376 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3377 }
3378 inode_unlock(inode);
3379out:
3380 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3381 iov_iter_count(from), ret);
3382 if (ret > 0)
3383 ret = generic_write_sync(iocb, ret);
3384 return ret;
3385}
3386
3387#ifdef CONFIG_COMPAT
3388long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3389{
3390 switch (cmd) {
3391 case F2FS_IOC32_GETFLAGS:
3392 cmd = F2FS_IOC_GETFLAGS;
3393 break;
3394 case F2FS_IOC32_SETFLAGS:
3395 cmd = F2FS_IOC_SETFLAGS;
3396 break;
3397 case F2FS_IOC32_GETVERSION:
3398 cmd = F2FS_IOC_GETVERSION;
3399 break;
3400 case F2FS_IOC_START_ATOMIC_WRITE:
3401 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3402 case F2FS_IOC_START_VOLATILE_WRITE:
3403 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3404 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3405 case F2FS_IOC_SHUTDOWN:
3406 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3407 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3408 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3409 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3410 case FS_IOC_ADD_ENCRYPTION_KEY:
3411 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3412 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3413 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3414 case F2FS_IOC_GARBAGE_COLLECT:
3415 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3416 case F2FS_IOC_WRITE_CHECKPOINT:
3417 case F2FS_IOC_DEFRAGMENT:
3418 case F2FS_IOC_MOVE_RANGE:
3419 case F2FS_IOC_FLUSH_DEVICE:
3420 case F2FS_IOC_GET_FEATURES:
3421 case F2FS_IOC_FSGETXATTR:
3422 case F2FS_IOC_FSSETXATTR:
3423 case F2FS_IOC_GET_PIN_FILE:
3424 case F2FS_IOC_SET_PIN_FILE:
3425 case F2FS_IOC_PRECACHE_EXTENTS:
3426 case F2FS_IOC_RESIZE_FS:
3427 case FS_IOC_ENABLE_VERITY:
3428 case FS_IOC_MEASURE_VERITY:
3429 case F2FS_IOC_GET_VOLUME_NAME:
3430 case F2FS_IOC_SET_VOLUME_NAME:
3431 break;
3432 default:
3433 return -ENOIOCTLCMD;
3434 }
3435 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3436}
3437#endif
3438
3439const struct file_operations f2fs_file_operations = {
3440 .llseek = f2fs_llseek,
3441 .read_iter = generic_file_read_iter,
3442 .write_iter = f2fs_file_write_iter,
3443 .open = f2fs_file_open,
3444 .release = f2fs_release_file,
3445 .mmap = f2fs_file_mmap,
3446 .flush = f2fs_file_flush,
3447 .fsync = f2fs_sync_file,
3448 .fallocate = f2fs_fallocate,
3449 .unlocked_ioctl = f2fs_ioctl,
3450#ifdef CONFIG_COMPAT
3451 .compat_ioctl = f2fs_compat_ioctl,
3452#endif
3453 .splice_read = generic_file_splice_read,
3454 .splice_write = iter_file_splice_write,
3455};