projects / GitHub / MotorolaMobilityLLC / kernel-slsi.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: avoid wrong decrypted data from disk
[GitHub/MotorolaMobilityLLC/kernel-slsi.git] / fs / f2fs / file.c
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/uio.h>
24 #include <linux/uuid.h>
25 #include <linux/file.h>
26
27 #include "f2fs.h"
28 #include "node.h"
29 #include "segment.h"
30 #include "xattr.h"
31 #include "acl.h"
32 #include "gc.h"
33 #include "trace.h"
34 #include <trace/events/f2fs.h>
35
36 static int f2fs_filemap_fault(struct vm_fault *vmf)
37 {
38 struct inode *inode = file_inode(vmf->vma->vm_file);
39 int err;
40
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 err = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
44
45 return err;
46 }
47
48 static int 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;
54 int err;
55
56 if (unlikely(f2fs_cp_error(sbi))) {
57 err = -EIO;
58 goto err;
59 }
60
61 sb_start_pagefault(inode->i_sb);
62
63 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
64
65 /* block allocation */
66 f2fs_lock_op(sbi);
67 set_new_dnode(&dn, inode, NULL, NULL, 0);
68 err = f2fs_reserve_block(&dn, page->index);
69 if (err) {
70 f2fs_unlock_op(sbi);
71 goto out;
72 }
73 f2fs_put_dnode(&dn);
74 f2fs_unlock_op(sbi);
75
76 f2fs_balance_fs(sbi, dn.node_changed);
77
78 file_update_time(vmf->vma->vm_file);
79 down_read(&F2FS_I(inode)->i_mmap_sem);
80 lock_page(page);
81 if (unlikely(page->mapping != inode->i_mapping ||
82 page_offset(page) > i_size_read(inode) ||
83 !PageUptodate(page))) {
84 unlock_page(page);
85 err = -EFAULT;
86 goto out_sem;
87 }
88
89 /*
90 * check to see if the page is mapped already (no holes)
91 */
92 if (PageMappedToDisk(page))
93 goto mapped;
94
95 /* page is wholly or partially inside EOF */
96 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
97 i_size_read(inode)) {
98 unsigned offset;
99 offset = i_size_read(inode) & ~PAGE_MASK;
100 zero_user_segment(page, offset, PAGE_SIZE);
101 }
102 set_page_dirty(page);
103 if (!PageUptodate(page))
104 SetPageUptodate(page);
105
106 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
107
108 trace_f2fs_vm_page_mkwrite(page, DATA);
109 mapped:
110 /* fill the page */
111 f2fs_wait_on_page_writeback(page, DATA, false);
112
113 /* wait for GCed page writeback via META_MAPPING */
114 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
115
116 out_sem:
117 up_read(&F2FS_I(inode)->i_mmap_sem);
118 out:
119 sb_end_pagefault(inode->i_sb);
120 f2fs_update_time(sbi, REQ_TIME);
121 err:
122 return block_page_mkwrite_return(err);
123 }
124
125 static const struct vm_operations_struct f2fs_file_vm_ops = {
126 .fault = f2fs_filemap_fault,
127 .map_pages = filemap_map_pages,
128 .page_mkwrite = f2fs_vm_page_mkwrite,
129 };
130
131 static int get_parent_ino(struct inode *inode, nid_t *pino)
132 {
133 struct dentry *dentry;
134
135 inode = igrab(inode);
136 dentry = d_find_any_alias(inode);
137 iput(inode);
138 if (!dentry)
139 return 0;
140
141 *pino = parent_ino(dentry);
142 dput(dentry);
143 return 1;
144 }
145
146 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
147 {
148 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
149 enum cp_reason_type cp_reason = CP_NO_NEEDED;
150
151 if (!S_ISREG(inode->i_mode))
152 cp_reason = CP_NON_REGULAR;
153 else if (inode->i_nlink != 1)
154 cp_reason = CP_HARDLINK;
155 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
156 cp_reason = CP_SB_NEED_CP;
157 else if (file_wrong_pino(inode))
158 cp_reason = CP_WRONG_PINO;
159 else if (!space_for_roll_forward(sbi))
160 cp_reason = CP_NO_SPC_ROLL;
161 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
162 cp_reason = CP_NODE_NEED_CP;
163 else if (test_opt(sbi, FASTBOOT))
164 cp_reason = CP_FASTBOOT_MODE;
165 else if (F2FS_OPTION(sbi).active_logs == 2)
166 cp_reason = CP_SPEC_LOG_NUM;
167 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
168 need_dentry_mark(sbi, inode->i_ino) &&
169 exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
170 cp_reason = CP_RECOVER_DIR;
171
172 return cp_reason;
173 }
174
175 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
176 {
177 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
178 bool ret = false;
179 /* But we need to avoid that there are some inode updates */
180 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
181 ret = true;
182 f2fs_put_page(i, 0);
183 return ret;
184 }
185
186 static void try_to_fix_pino(struct inode *inode)
187 {
188 struct f2fs_inode_info *fi = F2FS_I(inode);
189 nid_t pino;
190
191 down_write(&fi->i_sem);
192 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
193 get_parent_ino(inode, &pino)) {
194 f2fs_i_pino_write(inode, pino);
195 file_got_pino(inode);
196 }
197 up_write(&fi->i_sem);
198 }
199
200 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
201 int datasync, bool atomic)
202 {
203 struct inode *inode = file->f_mapping->host;
204 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
205 nid_t ino = inode->i_ino;
206 int ret = 0;
207 enum cp_reason_type cp_reason = 0;
208 struct writeback_control wbc = {
209 .sync_mode = WB_SYNC_ALL,
210 .nr_to_write = LONG_MAX,
211 .for_reclaim = 0,
212 };
213
214 if (unlikely(f2fs_readonly(inode->i_sb)))
215 return 0;
216
217 trace_f2fs_sync_file_enter(inode);
218
219 if (S_ISDIR(inode->i_mode))
220 goto go_write;
221
222 /* if fdatasync is triggered, let's do in-place-update */
223 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
224 set_inode_flag(inode, FI_NEED_IPU);
225 ret = file_write_and_wait_range(file, start, end);
226 clear_inode_flag(inode, FI_NEED_IPU);
227
228 if (ret) {
229 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
230 return ret;
231 }
232
233 /* if the inode is dirty, let's recover all the time */
234 if (!f2fs_skip_inode_update(inode, datasync)) {
235 f2fs_write_inode(inode, NULL);
236 goto go_write;
237 }
238
239 /*
240 * if there is no written data, don't waste time to write recovery info.
241 */
242 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
243 !exist_written_data(sbi, ino, APPEND_INO)) {
244
245 /* it may call write_inode just prior to fsync */
246 if (need_inode_page_update(sbi, ino))
247 goto go_write;
248
249 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
250 exist_written_data(sbi, ino, UPDATE_INO))
251 goto flush_out;
252 goto out;
253 }
254 go_write:
255 /*
256 * Both of fdatasync() and fsync() are able to be recovered from
257 * sudden-power-off.
258 */
259 down_read(&F2FS_I(inode)->i_sem);
260 cp_reason = need_do_checkpoint(inode);
261 up_read(&F2FS_I(inode)->i_sem);
262
263 if (cp_reason) {
264 /* all the dirty node pages should be flushed for POR */
265 ret = f2fs_sync_fs(inode->i_sb, 1);
266
267 /*
268 * We've secured consistency through sync_fs. Following pino
269 * will be used only for fsynced inodes after checkpoint.
270 */
271 try_to_fix_pino(inode);
272 clear_inode_flag(inode, FI_APPEND_WRITE);
273 clear_inode_flag(inode, FI_UPDATE_WRITE);
274 goto out;
275 }
276 sync_nodes:
277 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
278 if (ret)
279 goto out;
280
281 /* if cp_error was enabled, we should avoid infinite loop */
282 if (unlikely(f2fs_cp_error(sbi))) {
283 ret = -EIO;
284 goto out;
285 }
286
287 if (need_inode_block_update(sbi, ino)) {
288 f2fs_mark_inode_dirty_sync(inode, true);
289 f2fs_write_inode(inode, NULL);
290 goto sync_nodes;
291 }
292
293 /*
294 * If it's atomic_write, it's just fine to keep write ordering. So
295 * here we don't need to wait for node write completion, since we use
296 * node chain which serializes node blocks. If one of node writes are
297 * reordered, we can see simply broken chain, resulting in stopping
298 * roll-forward recovery. It means we'll recover all or none node blocks
299 * given fsync mark.
300 */
301 if (!atomic) {
302 ret = wait_on_node_pages_writeback(sbi, ino);
303 if (ret)
304 goto out;
305 }
306
307 /* once recovery info is written, don't need to tack this */
308 remove_ino_entry(sbi, ino, APPEND_INO);
309 clear_inode_flag(inode, FI_APPEND_WRITE);
310 flush_out:
311 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
312 ret = f2fs_issue_flush(sbi, inode->i_ino);
313 if (!ret) {
314 remove_ino_entry(sbi, ino, UPDATE_INO);
315 clear_inode_flag(inode, FI_UPDATE_WRITE);
316 remove_ino_entry(sbi, ino, FLUSH_INO);
317 }
318 f2fs_update_time(sbi, REQ_TIME);
319 out:
320 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
321 f2fs_trace_ios(NULL, 1);
322 return ret;
323 }
324
325 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
326 {
327 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
328 return -EIO;
329 return f2fs_do_sync_file(file, start, end, datasync, false);
330 }
331
332 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
333 pgoff_t pgofs, int whence)
334 {
335 struct pagevec pvec;
336 int nr_pages;
337
338 if (whence != SEEK_DATA)
339 return 0;
340
341 /* find first dirty page index */
342 pagevec_init(&pvec, 0);
343 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
344 PAGECACHE_TAG_DIRTY, 1);
345 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
346 pagevec_release(&pvec);
347 return pgofs;
348 }
349
350 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
351 pgoff_t dirty, pgoff_t pgofs, int whence)
352 {
353 switch (whence) {
354 case SEEK_DATA:
355 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
356 is_valid_data_blkaddr(sbi, blkaddr))
357 return true;
358 break;
359 case SEEK_HOLE:
360 if (blkaddr == NULL_ADDR)
361 return true;
362 break;
363 }
364 return false;
365 }
366
367 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
368 {
369 struct inode *inode = file->f_mapping->host;
370 loff_t maxbytes = inode->i_sb->s_maxbytes;
371 struct dnode_of_data dn;
372 pgoff_t pgofs, end_offset, dirty;
373 loff_t data_ofs = offset;
374 loff_t isize;
375 int err = 0;
376
377 inode_lock(inode);
378
379 isize = i_size_read(inode);
380 if (offset >= isize)
381 goto fail;
382
383 /* handle inline data case */
384 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
385 if (whence == SEEK_HOLE)
386 data_ofs = isize;
387 goto found;
388 }
389
390 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
391
392 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
393
394 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
395 set_new_dnode(&dn, inode, NULL, NULL, 0);
396 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
397 if (err && err != -ENOENT) {
398 goto fail;
399 } else if (err == -ENOENT) {
400 /* direct node does not exists */
401 if (whence == SEEK_DATA) {
402 pgofs = get_next_page_offset(&dn, pgofs);
403 continue;
404 } else {
405 goto found;
406 }
407 }
408
409 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
410
411 /* find data/hole in dnode block */
412 for (; dn.ofs_in_node < end_offset;
413 dn.ofs_in_node++, pgofs++,
414 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
415 block_t blkaddr;
416 blkaddr = datablock_addr(dn.inode,
417 dn.node_page, dn.ofs_in_node);
418
419 if (__is_valid_data_blkaddr(blkaddr) &&
420 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
421 blkaddr, DATA_GENERIC)) {
422 f2fs_put_dnode(&dn);
423 goto fail;
424 }
425
426 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
427 pgofs, whence)) {
428 f2fs_put_dnode(&dn);
429 goto found;
430 }
431 }
432 f2fs_put_dnode(&dn);
433 }
434
435 if (whence == SEEK_DATA)
436 goto fail;
437 found:
438 if (whence == SEEK_HOLE && data_ofs > isize)
439 data_ofs = isize;
440 inode_unlock(inode);
441 return vfs_setpos(file, data_ofs, maxbytes);
442 fail:
443 inode_unlock(inode);
444 return -ENXIO;
445 }
446
447 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
448 {
449 struct inode *inode = file->f_mapping->host;
450 loff_t maxbytes = inode->i_sb->s_maxbytes;
451
452 switch (whence) {
453 case SEEK_SET:
454 case SEEK_CUR:
455 case SEEK_END:
456 return generic_file_llseek_size(file, offset, whence,
457 maxbytes, i_size_read(inode));
458 case SEEK_DATA:
459 case SEEK_HOLE:
460 if (offset < 0)
461 return -ENXIO;
462 return f2fs_seek_block(file, offset, whence);
463 }
464
465 return -EINVAL;
466 }
467
468 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
469 {
470 struct inode *inode = file_inode(file);
471 int err;
472
473 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
474 return -EIO;
475
476 /* we don't need to use inline_data strictly */
477 err = f2fs_convert_inline_inode(inode);
478 if (err)
479 return err;
480
481 file_accessed(file);
482 vma->vm_ops = &f2fs_file_vm_ops;
483 return 0;
484 }
485
486 static int f2fs_file_open(struct inode *inode, struct file *filp)
487 {
488 int err = fscrypt_file_open(inode, filp);
489
490 if (err)
491 return err;
492
493 filp->f_mode |= FMODE_NOWAIT;
494
495 return dquot_file_open(inode, filp);
496 }
497
498 void truncate_data_blocks_range(struct dnode_of_data *dn, int count)
499 {
500 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
501 struct f2fs_node *raw_node;
502 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
503 __le32 *addr;
504 int base = 0;
505
506 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
507 base = get_extra_isize(dn->inode);
508
509 raw_node = F2FS_NODE(dn->node_page);
510 addr = blkaddr_in_node(raw_node) + base + ofs;
511
512 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
513 block_t blkaddr = le32_to_cpu(*addr);
514 if (blkaddr == NULL_ADDR)
515 continue;
516
517 dn->data_blkaddr = NULL_ADDR;
518 set_data_blkaddr(dn);
519
520 if (__is_valid_data_blkaddr(blkaddr) &&
521 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
522 continue;
523
524 invalidate_blocks(sbi, blkaddr);
525 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
526 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
527 nr_free++;
528 }
529
530 if (nr_free) {
531 pgoff_t fofs;
532 /*
533 * once we invalidate valid blkaddr in range [ofs, ofs + count],
534 * we will invalidate all blkaddr in the whole range.
535 */
536 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
537 dn->inode) + ofs;
538 f2fs_update_extent_cache_range(dn, fofs, 0, len);
539 dec_valid_block_count(sbi, dn->inode, nr_free);
540 }
541 dn->ofs_in_node = ofs;
542
543 f2fs_update_time(sbi, REQ_TIME);
544 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
545 dn->ofs_in_node, nr_free);
546 }
547
548 void truncate_data_blocks(struct dnode_of_data *dn)
549 {
550 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
551 }
552
553 static int truncate_partial_data_page(struct inode *inode, u64 from,
554 bool cache_only)
555 {
556 unsigned offset = from & (PAGE_SIZE - 1);
557 pgoff_t index = from >> PAGE_SHIFT;
558 struct address_space *mapping = inode->i_mapping;
559 struct page *page;
560
561 if (!offset && !cache_only)
562 return 0;
563
564 if (cache_only) {
565 page = find_lock_page(mapping, index);
566 if (page && PageUptodate(page))
567 goto truncate_out;
568 f2fs_put_page(page, 1);
569 return 0;
570 }
571
572 page = get_lock_data_page(inode, index, true);
573 if (IS_ERR(page))
574 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
575 truncate_out:
576 f2fs_wait_on_page_writeback(page, DATA, true);
577 zero_user(page, offset, PAGE_SIZE - offset);
578
579 /* An encrypted inode should have a key and truncate the last page. */
580 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
581 if (!cache_only)
582 set_page_dirty(page);
583 f2fs_put_page(page, 1);
584 return 0;
585 }
586
587 int truncate_blocks(struct inode *inode, u64 from, bool lock)
588 {
589 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
590 struct dnode_of_data dn;
591 pgoff_t free_from;
592 int count = 0, err = 0;
593 struct page *ipage;
594 bool truncate_page = false;
595
596 trace_f2fs_truncate_blocks_enter(inode, from);
597
598 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
599
600 if (free_from >= sbi->max_file_blocks)
601 goto free_partial;
602
603 if (lock)
604 f2fs_lock_op(sbi);
605
606 ipage = get_node_page(sbi, inode->i_ino);
607 if (IS_ERR(ipage)) {
608 err = PTR_ERR(ipage);
609 goto out;
610 }
611
612 if (f2fs_has_inline_data(inode)) {
613 truncate_inline_inode(inode, ipage, from);
614 f2fs_put_page(ipage, 1);
615 truncate_page = true;
616 goto out;
617 }
618
619 set_new_dnode(&dn, inode, ipage, NULL, 0);
620 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
621 if (err) {
622 if (err == -ENOENT)
623 goto free_next;
624 goto out;
625 }
626
627 count = ADDRS_PER_PAGE(dn.node_page, inode);
628
629 count -= dn.ofs_in_node;
630 f2fs_bug_on(sbi, count < 0);
631
632 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
633 truncate_data_blocks_range(&dn, count);
634 free_from += count;
635 }
636
637 f2fs_put_dnode(&dn);
638 free_next:
639 err = truncate_inode_blocks(inode, free_from);
640 out:
641 if (lock)
642 f2fs_unlock_op(sbi);
643 free_partial:
644 /* lastly zero out the first data page */
645 if (!err)
646 err = truncate_partial_data_page(inode, from, truncate_page);
647
648 trace_f2fs_truncate_blocks_exit(inode, err);
649 return err;
650 }
651
652 int f2fs_truncate(struct inode *inode)
653 {
654 int err;
655
656 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
657 return -EIO;
658
659 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
660 S_ISLNK(inode->i_mode)))
661 return 0;
662
663 trace_f2fs_truncate(inode);
664
665 #ifdef CONFIG_F2FS_FAULT_INJECTION
666 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
667 f2fs_show_injection_info(FAULT_TRUNCATE);
668 return -EIO;
669 }
670 #endif
671 /* we should check inline_data size */
672 if (!f2fs_may_inline_data(inode)) {
673 err = f2fs_convert_inline_inode(inode);
674 if (err)
675 return err;
676 }
677
678 err = truncate_blocks(inode, i_size_read(inode), true);
679 if (err)
680 return err;
681
682 inode->i_mtime = inode->i_ctime = current_time(inode);
683 f2fs_mark_inode_dirty_sync(inode, false);
684 return 0;
685 }
686
687 int f2fs_getattr(const struct path *path, struct kstat *stat,
688 u32 request_mask, unsigned int query_flags)
689 {
690 struct inode *inode = d_inode(path->dentry);
691 struct f2fs_inode_info *fi = F2FS_I(inode);
692 struct f2fs_inode *ri;
693 unsigned int flags;
694
695 if (f2fs_has_extra_attr(inode) &&
696 f2fs_sb_has_inode_crtime(inode->i_sb) &&
697 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
698 stat->result_mask |= STATX_BTIME;
699 stat->btime.tv_sec = fi->i_crtime.tv_sec;
700 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
701 }
702
703 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
704 if (flags & FS_APPEND_FL)
705 stat->attributes |= STATX_ATTR_APPEND;
706 if (flags & FS_COMPR_FL)
707 stat->attributes |= STATX_ATTR_COMPRESSED;
708 if (f2fs_encrypted_inode(inode))
709 stat->attributes |= STATX_ATTR_ENCRYPTED;
710 if (flags & FS_IMMUTABLE_FL)
711 stat->attributes |= STATX_ATTR_IMMUTABLE;
712 if (flags & FS_NODUMP_FL)
713 stat->attributes |= STATX_ATTR_NODUMP;
714
715 stat->attributes_mask |= (STATX_ATTR_APPEND |
716 STATX_ATTR_COMPRESSED |
717 STATX_ATTR_ENCRYPTED |
718 STATX_ATTR_IMMUTABLE |
719 STATX_ATTR_NODUMP);
720
721 generic_fillattr(inode, stat);
722
723 /* we need to show initial sectors used for inline_data/dentries */
724 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
725 f2fs_has_inline_dentry(inode))
726 stat->blocks += (stat->size + 511) >> 9;
727
728 return 0;
729 }
730
731 #ifdef CONFIG_F2FS_FS_POSIX_ACL
732 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
733 {
734 unsigned int ia_valid = attr->ia_valid;
735
736 if (ia_valid & ATTR_UID)
737 inode->i_uid = attr->ia_uid;
738 if (ia_valid & ATTR_GID)
739 inode->i_gid = attr->ia_gid;
740 if (ia_valid & ATTR_ATIME)
741 inode->i_atime = timespec_trunc(attr->ia_atime,
742 inode->i_sb->s_time_gran);
743 if (ia_valid & ATTR_MTIME)
744 inode->i_mtime = timespec_trunc(attr->ia_mtime,
745 inode->i_sb->s_time_gran);
746 if (ia_valid & ATTR_CTIME)
747 inode->i_ctime = timespec_trunc(attr->ia_ctime,
748 inode->i_sb->s_time_gran);
749 if (ia_valid & ATTR_MODE) {
750 umode_t mode = attr->ia_mode;
751
752 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
753 mode &= ~S_ISGID;
754 set_acl_inode(inode, mode);
755 }
756 }
757 #else
758 #define __setattr_copy setattr_copy
759 #endif
760
761 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
762 {
763 struct inode *inode = d_inode(dentry);
764 int err;
765 bool size_changed = false;
766
767 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
768 return -EIO;
769
770 err = setattr_prepare(dentry, attr);
771 if (err)
772 return err;
773
774 err = fscrypt_prepare_setattr(dentry, attr);
775 if (err)
776 return err;
777
778 if (is_quota_modification(inode, attr)) {
779 err = dquot_initialize(inode);
780 if (err)
781 return err;
782 }
783 if ((attr->ia_valid & ATTR_UID &&
784 !uid_eq(attr->ia_uid, inode->i_uid)) ||
785 (attr->ia_valid & ATTR_GID &&
786 !gid_eq(attr->ia_gid, inode->i_gid))) {
787 err = dquot_transfer(inode, attr);
788 if (err)
789 return err;
790 }
791
792 if (attr->ia_valid & ATTR_SIZE) {
793 if (attr->ia_size <= i_size_read(inode)) {
794 down_write(&F2FS_I(inode)->i_mmap_sem);
795 truncate_setsize(inode, attr->ia_size);
796 err = f2fs_truncate(inode);
797 up_write(&F2FS_I(inode)->i_mmap_sem);
798 if (err)
799 return err;
800 } else {
801 /*
802 * do not trim all blocks after i_size if target size is
803 * larger than i_size.
804 */
805 down_write(&F2FS_I(inode)->i_mmap_sem);
806 truncate_setsize(inode, attr->ia_size);
807 up_write(&F2FS_I(inode)->i_mmap_sem);
808
809 /* should convert inline inode here */
810 if (!f2fs_may_inline_data(inode)) {
811 err = f2fs_convert_inline_inode(inode);
812 if (err)
813 return err;
814 }
815 inode->i_mtime = inode->i_ctime = current_time(inode);
816 }
817
818 down_write(&F2FS_I(inode)->i_sem);
819 F2FS_I(inode)->last_disk_size = i_size_read(inode);
820 up_write(&F2FS_I(inode)->i_sem);
821
822 size_changed = true;
823 }
824
825 __setattr_copy(inode, attr);
826
827 if (attr->ia_valid & ATTR_MODE) {
828 err = posix_acl_chmod(inode, get_inode_mode(inode));
829 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
830 inode->i_mode = F2FS_I(inode)->i_acl_mode;
831 clear_inode_flag(inode, FI_ACL_MODE);
832 }
833 }
834
835 /* file size may changed here */
836 f2fs_mark_inode_dirty_sync(inode, size_changed);
837
838 /* inode change will produce dirty node pages flushed by checkpoint */
839 f2fs_balance_fs(F2FS_I_SB(inode), true);
840
841 return err;
842 }
843
844 const struct inode_operations f2fs_file_inode_operations = {
845 .getattr = f2fs_getattr,
846 .setattr = f2fs_setattr,
847 .get_acl = f2fs_get_acl,
848 .set_acl = f2fs_set_acl,
849 #ifdef CONFIG_F2FS_FS_XATTR
850 .listxattr = f2fs_listxattr,
851 #endif
852 .fiemap = f2fs_fiemap,
853 };
854
855 static int fill_zero(struct inode *inode, pgoff_t index,
856 loff_t start, loff_t len)
857 {
858 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
859 struct page *page;
860
861 if (!len)
862 return 0;
863
864 f2fs_balance_fs(sbi, true);
865
866 f2fs_lock_op(sbi);
867 page = get_new_data_page(inode, NULL, index, false);
868 f2fs_unlock_op(sbi);
869
870 if (IS_ERR(page))
871 return PTR_ERR(page);
872
873 f2fs_wait_on_page_writeback(page, DATA, true);
874 zero_user(page, start, len);
875 set_page_dirty(page);
876 f2fs_put_page(page, 1);
877 return 0;
878 }
879
880 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
881 {
882 int err;
883
884 while (pg_start < pg_end) {
885 struct dnode_of_data dn;
886 pgoff_t end_offset, count;
887
888 set_new_dnode(&dn, inode, NULL, NULL, 0);
889 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
890 if (err) {
891 if (err == -ENOENT) {
892 pg_start = get_next_page_offset(&dn, pg_start);
893 continue;
894 }
895 return err;
896 }
897
898 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
899 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
900
901 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
902
903 truncate_data_blocks_range(&dn, count);
904 f2fs_put_dnode(&dn);
905
906 pg_start += count;
907 }
908 return 0;
909 }
910
911 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
912 {
913 pgoff_t pg_start, pg_end;
914 loff_t off_start, off_end;
915 int ret;
916
917 ret = f2fs_convert_inline_inode(inode);
918 if (ret)
919 return ret;
920
921 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
922 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
923
924 off_start = offset & (PAGE_SIZE - 1);
925 off_end = (offset + len) & (PAGE_SIZE - 1);
926
927 if (pg_start == pg_end) {
928 ret = fill_zero(inode, pg_start, off_start,
929 off_end - off_start);
930 if (ret)
931 return ret;
932 } else {
933 if (off_start) {
934 ret = fill_zero(inode, pg_start++, off_start,
935 PAGE_SIZE - off_start);
936 if (ret)
937 return ret;
938 }
939 if (off_end) {
940 ret = fill_zero(inode, pg_end, 0, off_end);
941 if (ret)
942 return ret;
943 }
944
945 if (pg_start < pg_end) {
946 struct address_space *mapping = inode->i_mapping;
947 loff_t blk_start, blk_end;
948 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
949
950 f2fs_balance_fs(sbi, true);
951
952 blk_start = (loff_t)pg_start << PAGE_SHIFT;
953 blk_end = (loff_t)pg_end << PAGE_SHIFT;
954 down_write(&F2FS_I(inode)->i_mmap_sem);
955 truncate_inode_pages_range(mapping, blk_start,
956 blk_end - 1);
957
958 f2fs_lock_op(sbi);
959 ret = truncate_hole(inode, pg_start, pg_end);
960 f2fs_unlock_op(sbi);
961 up_write(&F2FS_I(inode)->i_mmap_sem);
962 }
963 }
964
965 return ret;
966 }
967
968 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
969 int *do_replace, pgoff_t off, pgoff_t len)
970 {
971 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
972 struct dnode_of_data dn;
973 int ret, done, i;
974
975 next_dnode:
976 set_new_dnode(&dn, inode, NULL, NULL, 0);
977 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
978 if (ret && ret != -ENOENT) {
979 return ret;
980 } else if (ret == -ENOENT) {
981 if (dn.max_level == 0)
982 return -ENOENT;
983 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
984 blkaddr += done;
985 do_replace += done;
986 goto next;
987 }
988
989 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
990 dn.ofs_in_node, len);
991 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
992 *blkaddr = datablock_addr(dn.inode,
993 dn.node_page, dn.ofs_in_node);
994 if (!is_checkpointed_data(sbi, *blkaddr)) {
995
996 if (test_opt(sbi, LFS)) {
997 f2fs_put_dnode(&dn);
998 return -ENOTSUPP;
999 }
1000
1001 /* do not invalidate this block address */
1002 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1003 *do_replace = 1;
1004 }
1005 }
1006 f2fs_put_dnode(&dn);
1007 next:
1008 len -= done;
1009 off += done;
1010 if (len)
1011 goto next_dnode;
1012 return 0;
1013 }
1014
1015 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1016 int *do_replace, pgoff_t off, int len)
1017 {
1018 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1019 struct dnode_of_data dn;
1020 int ret, i;
1021
1022 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1023 if (*do_replace == 0)
1024 continue;
1025
1026 set_new_dnode(&dn, inode, NULL, NULL, 0);
1027 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1028 if (ret) {
1029 dec_valid_block_count(sbi, inode, 1);
1030 invalidate_blocks(sbi, *blkaddr);
1031 } else {
1032 f2fs_update_data_blkaddr(&dn, *blkaddr);
1033 }
1034 f2fs_put_dnode(&dn);
1035 }
1036 return 0;
1037 }
1038
1039 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1040 block_t *blkaddr, int *do_replace,
1041 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1042 {
1043 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1044 pgoff_t i = 0;
1045 int ret;
1046
1047 while (i < len) {
1048 if (blkaddr[i] == NULL_ADDR && !full) {
1049 i++;
1050 continue;
1051 }
1052
1053 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1054 struct dnode_of_data dn;
1055 struct node_info ni;
1056 size_t new_size;
1057 pgoff_t ilen;
1058
1059 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1060 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1061 if (ret)
1062 return ret;
1063
1064 get_node_info(sbi, dn.nid, &ni);
1065 ilen = min((pgoff_t)
1066 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1067 dn.ofs_in_node, len - i);
1068 do {
1069 dn.data_blkaddr = datablock_addr(dn.inode,
1070 dn.node_page, dn.ofs_in_node);
1071 truncate_data_blocks_range(&dn, 1);
1072
1073 if (do_replace[i]) {
1074 f2fs_i_blocks_write(src_inode,
1075 1, false, false);
1076 f2fs_i_blocks_write(dst_inode,
1077 1, true, false);
1078 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1079 blkaddr[i], ni.version, true, false);
1080
1081 do_replace[i] = 0;
1082 }
1083 dn.ofs_in_node++;
1084 i++;
1085 new_size = (dst + i) << PAGE_SHIFT;
1086 if (dst_inode->i_size < new_size)
1087 f2fs_i_size_write(dst_inode, new_size);
1088 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1089
1090 f2fs_put_dnode(&dn);
1091 } else {
1092 struct page *psrc, *pdst;
1093
1094 psrc = get_lock_data_page(src_inode, src + i, true);
1095 if (IS_ERR(psrc))
1096 return PTR_ERR(psrc);
1097 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1098 true);
1099 if (IS_ERR(pdst)) {
1100 f2fs_put_page(psrc, 1);
1101 return PTR_ERR(pdst);
1102 }
1103 f2fs_copy_page(psrc, pdst);
1104 set_page_dirty(pdst);
1105 f2fs_put_page(pdst, 1);
1106 f2fs_put_page(psrc, 1);
1107
1108 ret = truncate_hole(src_inode, src + i, src + i + 1);
1109 if (ret)
1110 return ret;
1111 i++;
1112 }
1113 }
1114 return 0;
1115 }
1116
1117 static int __exchange_data_block(struct inode *src_inode,
1118 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1119 pgoff_t len, bool full)
1120 {
1121 block_t *src_blkaddr;
1122 int *do_replace;
1123 pgoff_t olen;
1124 int ret;
1125
1126 while (len) {
1127 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1128
1129 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1130 sizeof(block_t) * olen, GFP_KERNEL);
1131 if (!src_blkaddr)
1132 return -ENOMEM;
1133
1134 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1135 sizeof(int) * olen, GFP_KERNEL);
1136 if (!do_replace) {
1137 kvfree(src_blkaddr);
1138 return -ENOMEM;
1139 }
1140
1141 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1142 do_replace, src, olen);
1143 if (ret)
1144 goto roll_back;
1145
1146 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1147 do_replace, src, dst, olen, full);
1148 if (ret)
1149 goto roll_back;
1150
1151 src += olen;
1152 dst += olen;
1153 len -= olen;
1154
1155 kvfree(src_blkaddr);
1156 kvfree(do_replace);
1157 }
1158 return 0;
1159
1160 roll_back:
1161 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1162 kvfree(src_blkaddr);
1163 kvfree(do_replace);
1164 return ret;
1165 }
1166
1167 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1168 {
1169 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1170 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1171 int ret;
1172
1173 f2fs_balance_fs(sbi, true);
1174 f2fs_lock_op(sbi);
1175
1176 f2fs_drop_extent_tree(inode);
1177
1178 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1179 f2fs_unlock_op(sbi);
1180 return ret;
1181 }
1182
1183 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1184 {
1185 pgoff_t pg_start, pg_end;
1186 loff_t new_size;
1187 int ret;
1188
1189 if (offset + len >= i_size_read(inode))
1190 return -EINVAL;
1191
1192 /* collapse range should be aligned to block size of f2fs. */
1193 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1194 return -EINVAL;
1195
1196 ret = f2fs_convert_inline_inode(inode);
1197 if (ret)
1198 return ret;
1199
1200 pg_start = offset >> PAGE_SHIFT;
1201 pg_end = (offset + len) >> PAGE_SHIFT;
1202
1203 /* avoid gc operation during block exchange */
1204 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1205
1206 down_write(&F2FS_I(inode)->i_mmap_sem);
1207 /* write out all dirty pages from offset */
1208 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1209 if (ret)
1210 goto out_unlock;
1211
1212 truncate_pagecache(inode, offset);
1213
1214 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1215 if (ret)
1216 goto out_unlock;
1217
1218 /* write out all moved pages, if possible */
1219 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1220 truncate_pagecache(inode, offset);
1221
1222 new_size = i_size_read(inode) - len;
1223 truncate_pagecache(inode, new_size);
1224
1225 ret = truncate_blocks(inode, new_size, true);
1226 if (!ret)
1227 f2fs_i_size_write(inode, new_size);
1228 out_unlock:
1229 up_write(&F2FS_I(inode)->i_mmap_sem);
1230 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1231 return ret;
1232 }
1233
1234 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1235 pgoff_t end)
1236 {
1237 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1238 pgoff_t index = start;
1239 unsigned int ofs_in_node = dn->ofs_in_node;
1240 blkcnt_t count = 0;
1241 int ret;
1242
1243 for (; index < end; index++, dn->ofs_in_node++) {
1244 if (datablock_addr(dn->inode, dn->node_page,
1245 dn->ofs_in_node) == NULL_ADDR)
1246 count++;
1247 }
1248
1249 dn->ofs_in_node = ofs_in_node;
1250 ret = reserve_new_blocks(dn, count);
1251 if (ret)
1252 return ret;
1253
1254 dn->ofs_in_node = ofs_in_node;
1255 for (index = start; index < end; index++, dn->ofs_in_node++) {
1256 dn->data_blkaddr = datablock_addr(dn->inode,
1257 dn->node_page, dn->ofs_in_node);
1258 /*
1259 * reserve_new_blocks will not guarantee entire block
1260 * allocation.
1261 */
1262 if (dn->data_blkaddr == NULL_ADDR) {
1263 ret = -ENOSPC;
1264 break;
1265 }
1266 if (dn->data_blkaddr != NEW_ADDR) {
1267 invalidate_blocks(sbi, dn->data_blkaddr);
1268 dn->data_blkaddr = NEW_ADDR;
1269 set_data_blkaddr(dn);
1270 }
1271 }
1272
1273 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1274
1275 return ret;
1276 }
1277
1278 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1279 int mode)
1280 {
1281 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1282 struct address_space *mapping = inode->i_mapping;
1283 pgoff_t index, pg_start, pg_end;
1284 loff_t new_size = i_size_read(inode);
1285 loff_t off_start, off_end;
1286 int ret = 0;
1287
1288 ret = inode_newsize_ok(inode, (len + offset));
1289 if (ret)
1290 return ret;
1291
1292 ret = f2fs_convert_inline_inode(inode);
1293 if (ret)
1294 return ret;
1295
1296 down_write(&F2FS_I(inode)->i_mmap_sem);
1297 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1298 if (ret)
1299 goto out_sem;
1300
1301 truncate_pagecache_range(inode, offset, offset + len - 1);
1302
1303 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1304 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1305
1306 off_start = offset & (PAGE_SIZE - 1);
1307 off_end = (offset + len) & (PAGE_SIZE - 1);
1308
1309 if (pg_start == pg_end) {
1310 ret = fill_zero(inode, pg_start, off_start,
1311 off_end - off_start);
1312 if (ret)
1313 goto out_sem;
1314
1315 new_size = max_t(loff_t, new_size, offset + len);
1316 } else {
1317 if (off_start) {
1318 ret = fill_zero(inode, pg_start++, off_start,
1319 PAGE_SIZE - off_start);
1320 if (ret)
1321 goto out_sem;
1322
1323 new_size = max_t(loff_t, new_size,
1324 (loff_t)pg_start << PAGE_SHIFT);
1325 }
1326
1327 for (index = pg_start; index < pg_end;) {
1328 struct dnode_of_data dn;
1329 unsigned int end_offset;
1330 pgoff_t end;
1331
1332 f2fs_lock_op(sbi);
1333
1334 set_new_dnode(&dn, inode, NULL, NULL, 0);
1335 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1336 if (ret) {
1337 f2fs_unlock_op(sbi);
1338 goto out;
1339 }
1340
1341 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1342 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1343
1344 ret = f2fs_do_zero_range(&dn, index, end);
1345 f2fs_put_dnode(&dn);
1346 f2fs_unlock_op(sbi);
1347
1348 f2fs_balance_fs(sbi, dn.node_changed);
1349
1350 if (ret)
1351 goto out;
1352
1353 index = end;
1354 new_size = max_t(loff_t, new_size,
1355 (loff_t)index << PAGE_SHIFT);
1356 }
1357
1358 if (off_end) {
1359 ret = fill_zero(inode, pg_end, 0, off_end);
1360 if (ret)
1361 goto out;
1362
1363 new_size = max_t(loff_t, new_size, offset + len);
1364 }
1365 }
1366
1367 out:
1368 if (new_size > i_size_read(inode)) {
1369 if (mode & FALLOC_FL_KEEP_SIZE)
1370 file_set_keep_isize(inode);
1371 else
1372 f2fs_i_size_write(inode, new_size);
1373 }
1374 out_sem:
1375 up_write(&F2FS_I(inode)->i_mmap_sem);
1376
1377 return ret;
1378 }
1379
1380 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1381 {
1382 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1383 pgoff_t nr, pg_start, pg_end, delta, idx;
1384 loff_t new_size;
1385 int ret = 0;
1386
1387 new_size = i_size_read(inode) + len;
1388 ret = inode_newsize_ok(inode, new_size);
1389 if (ret)
1390 return ret;
1391
1392 if (offset >= i_size_read(inode))
1393 return -EINVAL;
1394
1395 /* insert range should be aligned to block size of f2fs. */
1396 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1397 return -EINVAL;
1398
1399 ret = f2fs_convert_inline_inode(inode);
1400 if (ret)
1401 return ret;
1402
1403 f2fs_balance_fs(sbi, true);
1404
1405 /* avoid gc operation during block exchange */
1406 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1407
1408 down_write(&F2FS_I(inode)->i_mmap_sem);
1409 ret = truncate_blocks(inode, i_size_read(inode), true);
1410 if (ret)
1411 goto out;
1412
1413 /* write out all dirty pages from offset */
1414 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1415 if (ret)
1416 goto out;
1417
1418 truncate_pagecache(inode, offset);
1419
1420 pg_start = offset >> PAGE_SHIFT;
1421 pg_end = (offset + len) >> PAGE_SHIFT;
1422 delta = pg_end - pg_start;
1423 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1424
1425 while (!ret && idx > pg_start) {
1426 nr = idx - pg_start;
1427 if (nr > delta)
1428 nr = delta;
1429 idx -= nr;
1430
1431 f2fs_lock_op(sbi);
1432 f2fs_drop_extent_tree(inode);
1433
1434 ret = __exchange_data_block(inode, inode, idx,
1435 idx + delta, nr, false);
1436 f2fs_unlock_op(sbi);
1437 }
1438
1439 /* write out all moved pages, if possible */
1440 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1441 truncate_pagecache(inode, offset);
1442
1443 if (!ret)
1444 f2fs_i_size_write(inode, new_size);
1445 out:
1446 up_write(&F2FS_I(inode)->i_mmap_sem);
1447 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1448 return ret;
1449 }
1450
1451 static int expand_inode_data(struct inode *inode, loff_t offset,
1452 loff_t len, int mode)
1453 {
1454 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1455 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1456 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1457 pgoff_t pg_end;
1458 loff_t new_size = i_size_read(inode);
1459 loff_t off_end;
1460 int err;
1461
1462 err = inode_newsize_ok(inode, (len + offset));
1463 if (err)
1464 return err;
1465
1466 err = f2fs_convert_inline_inode(inode);
1467 if (err)
1468 return err;
1469
1470 f2fs_balance_fs(sbi, true);
1471
1472 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1473 off_end = (offset + len) & (PAGE_SIZE - 1);
1474
1475 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1476 map.m_len = pg_end - map.m_lblk;
1477 if (off_end)
1478 map.m_len++;
1479
1480 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1481 if (err) {
1482 pgoff_t last_off;
1483
1484 if (!map.m_len)
1485 return err;
1486
1487 last_off = map.m_lblk + map.m_len - 1;
1488
1489 /* update new size to the failed position */
1490 new_size = (last_off == pg_end) ? offset + len:
1491 (loff_t)(last_off + 1) << PAGE_SHIFT;
1492 } else {
1493 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1494 }
1495
1496 if (new_size > i_size_read(inode)) {
1497 if (mode & FALLOC_FL_KEEP_SIZE)
1498 file_set_keep_isize(inode);
1499 else
1500 f2fs_i_size_write(inode, new_size);
1501 }
1502
1503 return err;
1504 }
1505
1506 static long f2fs_fallocate(struct file *file, int mode,
1507 loff_t offset, loff_t len)
1508 {
1509 struct inode *inode = file_inode(file);
1510 long ret = 0;
1511
1512 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1513 return -EIO;
1514
1515 /* f2fs only support ->fallocate for regular file */
1516 if (!S_ISREG(inode->i_mode))
1517 return -EINVAL;
1518
1519 if (f2fs_encrypted_inode(inode) &&
1520 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1521 return -EOPNOTSUPP;
1522
1523 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1524 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1525 FALLOC_FL_INSERT_RANGE))
1526 return -EOPNOTSUPP;
1527
1528 inode_lock(inode);
1529
1530 if (mode & FALLOC_FL_PUNCH_HOLE) {
1531 if (offset >= inode->i_size)
1532 goto out;
1533
1534 ret = punch_hole(inode, offset, len);
1535 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1536 ret = f2fs_collapse_range(inode, offset, len);
1537 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1538 ret = f2fs_zero_range(inode, offset, len, mode);
1539 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1540 ret = f2fs_insert_range(inode, offset, len);
1541 } else {
1542 ret = expand_inode_data(inode, offset, len, mode);
1543 }
1544
1545 if (!ret) {
1546 inode->i_mtime = inode->i_ctime = current_time(inode);
1547 f2fs_mark_inode_dirty_sync(inode, false);
1548 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1549 }
1550
1551 out:
1552 inode_unlock(inode);
1553
1554 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1555 return ret;
1556 }
1557
1558 static int f2fs_release_file(struct inode *inode, struct file *filp)
1559 {
1560 /*
1561 * f2fs_relase_file is called at every close calls. So we should
1562 * not drop any inmemory pages by close called by other process.
1563 */
1564 if (!(filp->f_mode & FMODE_WRITE) ||
1565 atomic_read(&inode->i_writecount) != 1)
1566 return 0;
1567
1568 /* some remained atomic pages should discarded */
1569 if (f2fs_is_atomic_file(inode))
1570 drop_inmem_pages(inode);
1571 if (f2fs_is_volatile_file(inode)) {
1572 clear_inode_flag(inode, FI_VOLATILE_FILE);
1573 stat_dec_volatile_write(inode);
1574 set_inode_flag(inode, FI_DROP_CACHE);
1575 filemap_fdatawrite(inode->i_mapping);
1576 clear_inode_flag(inode, FI_DROP_CACHE);
1577 }
1578 return 0;
1579 }
1580
1581 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1582 {
1583 struct inode *inode = file_inode(file);
1584
1585 /*
1586 * If the process doing a transaction is crashed, we should do
1587 * roll-back. Otherwise, other reader/write can see corrupted database
1588 * until all the writers close its file. Since this should be done
1589 * before dropping file lock, it needs to do in ->flush.
1590 */
1591 if (f2fs_is_atomic_file(inode) &&
1592 F2FS_I(inode)->inmem_task == current)
1593 drop_inmem_pages(inode);
1594 return 0;
1595 }
1596
1597 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1598 {
1599 struct inode *inode = file_inode(filp);
1600 struct f2fs_inode_info *fi = F2FS_I(inode);
1601 unsigned int flags = fi->i_flags &
1602 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1603 return put_user(flags, (int __user *)arg);
1604 }
1605
1606 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1607 {
1608 struct f2fs_inode_info *fi = F2FS_I(inode);
1609 unsigned int oldflags;
1610
1611 /* Is it quota file? Do not allow user to mess with it */
1612 if (IS_NOQUOTA(inode))
1613 return -EPERM;
1614
1615 flags = f2fs_mask_flags(inode->i_mode, flags);
1616
1617 oldflags = fi->i_flags;
1618
1619 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1620 if (!capable(CAP_LINUX_IMMUTABLE))
1621 return -EPERM;
1622
1623 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1624 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1625 fi->i_flags = flags;
1626
1627 if (fi->i_flags & FS_PROJINHERIT_FL)
1628 set_inode_flag(inode, FI_PROJ_INHERIT);
1629 else
1630 clear_inode_flag(inode, FI_PROJ_INHERIT);
1631
1632 inode->i_ctime = current_time(inode);
1633 f2fs_set_inode_flags(inode);
1634 f2fs_mark_inode_dirty_sync(inode, false);
1635 return 0;
1636 }
1637
1638 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1639 {
1640 struct inode *inode = file_inode(filp);
1641 unsigned int flags;
1642 int ret;
1643
1644 if (!inode_owner_or_capable(inode))
1645 return -EACCES;
1646
1647 if (get_user(flags, (int __user *)arg))
1648 return -EFAULT;
1649
1650 ret = mnt_want_write_file(filp);
1651 if (ret)
1652 return ret;
1653
1654 inode_lock(inode);
1655
1656 ret = __f2fs_ioc_setflags(inode, flags);
1657
1658 inode_unlock(inode);
1659 mnt_drop_write_file(filp);
1660 return ret;
1661 }
1662
1663 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1664 {
1665 struct inode *inode = file_inode(filp);
1666
1667 return put_user(inode->i_generation, (int __user *)arg);
1668 }
1669
1670 static int f2fs_ioc_start_atomic_write(struct file *filp)
1671 {
1672 struct inode *inode = file_inode(filp);
1673 int ret;
1674
1675 if (!inode_owner_or_capable(inode))
1676 return -EACCES;
1677
1678 if (!S_ISREG(inode->i_mode))
1679 return -EINVAL;
1680
1681 ret = mnt_want_write_file(filp);
1682 if (ret)
1683 return ret;
1684
1685 inode_lock(inode);
1686
1687 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1688
1689 if (f2fs_is_atomic_file(inode))
1690 goto out;
1691
1692 ret = f2fs_convert_inline_inode(inode);
1693 if (ret)
1694 goto out;
1695
1696 set_inode_flag(inode, FI_ATOMIC_FILE);
1697 set_inode_flag(inode, FI_HOT_DATA);
1698 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1699
1700 if (!get_dirty_pages(inode))
1701 goto inc_stat;
1702
1703 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1704 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1705 inode->i_ino, get_dirty_pages(inode));
1706 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1707 if (ret) {
1708 clear_inode_flag(inode, FI_ATOMIC_FILE);
1709 clear_inode_flag(inode, FI_HOT_DATA);
1710 goto out;
1711 }
1712
1713 inc_stat:
1714 F2FS_I(inode)->inmem_task = current;
1715 stat_inc_atomic_write(inode);
1716 stat_update_max_atomic_write(inode);
1717 out:
1718 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1719 inode_unlock(inode);
1720 mnt_drop_write_file(filp);
1721 return ret;
1722 }
1723
1724 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1725 {
1726 struct inode *inode = file_inode(filp);
1727 int ret;
1728
1729 if (!inode_owner_or_capable(inode))
1730 return -EACCES;
1731
1732 ret = mnt_want_write_file(filp);
1733 if (ret)
1734 return ret;
1735
1736 inode_lock(inode);
1737
1738 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1739
1740 if (f2fs_is_volatile_file(inode))
1741 goto err_out;
1742
1743 if (f2fs_is_atomic_file(inode)) {
1744 ret = commit_inmem_pages(inode);
1745 if (ret)
1746 goto err_out;
1747
1748 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1749 if (!ret) {
1750 clear_inode_flag(inode, FI_ATOMIC_FILE);
1751 clear_inode_flag(inode, FI_HOT_DATA);
1752 stat_dec_atomic_write(inode);
1753 }
1754 } else {
1755 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1756 }
1757 err_out:
1758 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1759 inode_unlock(inode);
1760 mnt_drop_write_file(filp);
1761 return ret;
1762 }
1763
1764 static int f2fs_ioc_start_volatile_write(struct file *filp)
1765 {
1766 struct inode *inode = file_inode(filp);
1767 int ret;
1768
1769 if (!inode_owner_or_capable(inode))
1770 return -EACCES;
1771
1772 if (!S_ISREG(inode->i_mode))
1773 return -EINVAL;
1774
1775 ret = mnt_want_write_file(filp);
1776 if (ret)
1777 return ret;
1778
1779 inode_lock(inode);
1780
1781 if (f2fs_is_volatile_file(inode))
1782 goto out;
1783
1784 ret = f2fs_convert_inline_inode(inode);
1785 if (ret)
1786 goto out;
1787
1788 stat_inc_volatile_write(inode);
1789 stat_update_max_volatile_write(inode);
1790
1791 set_inode_flag(inode, FI_VOLATILE_FILE);
1792 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1793 out:
1794 inode_unlock(inode);
1795 mnt_drop_write_file(filp);
1796 return ret;
1797 }
1798
1799 static int f2fs_ioc_release_volatile_write(struct file *filp)
1800 {
1801 struct inode *inode = file_inode(filp);
1802 int ret;
1803
1804 if (!inode_owner_or_capable(inode))
1805 return -EACCES;
1806
1807 ret = mnt_want_write_file(filp);
1808 if (ret)
1809 return ret;
1810
1811 inode_lock(inode);
1812
1813 if (!f2fs_is_volatile_file(inode))
1814 goto out;
1815
1816 if (!f2fs_is_first_block_written(inode)) {
1817 ret = truncate_partial_data_page(inode, 0, true);
1818 goto out;
1819 }
1820
1821 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1822 out:
1823 inode_unlock(inode);
1824 mnt_drop_write_file(filp);
1825 return ret;
1826 }
1827
1828 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1829 {
1830 struct inode *inode = file_inode(filp);
1831 int ret;
1832
1833 if (!inode_owner_or_capable(inode))
1834 return -EACCES;
1835
1836 ret = mnt_want_write_file(filp);
1837 if (ret)
1838 return ret;
1839
1840 inode_lock(inode);
1841
1842 if (f2fs_is_atomic_file(inode))
1843 drop_inmem_pages(inode);
1844 if (f2fs_is_volatile_file(inode)) {
1845 clear_inode_flag(inode, FI_VOLATILE_FILE);
1846 stat_dec_volatile_write(inode);
1847 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1848 }
1849
1850 inode_unlock(inode);
1851
1852 mnt_drop_write_file(filp);
1853 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1854 return ret;
1855 }
1856
1857 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1858 {
1859 struct inode *inode = file_inode(filp);
1860 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1861 struct super_block *sb = sbi->sb;
1862 __u32 in;
1863 int ret = 0;
1864
1865 if (!capable(CAP_SYS_ADMIN))
1866 return -EPERM;
1867
1868 if (get_user(in, (__u32 __user *)arg))
1869 return -EFAULT;
1870
1871 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1872 ret = mnt_want_write_file(filp);
1873 if (ret)
1874 return ret;
1875 }
1876
1877 switch (in) {
1878 case F2FS_GOING_DOWN_FULLSYNC:
1879 sb = freeze_bdev(sb->s_bdev);
1880 if (IS_ERR(sb)) {
1881 ret = PTR_ERR(sb);
1882 goto out;
1883 }
1884 if (sb) {
1885 f2fs_stop_checkpoint(sbi, false);
1886 thaw_bdev(sb->s_bdev, sb);
1887 }
1888 break;
1889 case F2FS_GOING_DOWN_METASYNC:
1890 /* do checkpoint only */
1891 ret = f2fs_sync_fs(sb, 1);
1892 if (ret)
1893 goto out;
1894 f2fs_stop_checkpoint(sbi, false);
1895 break;
1896 case F2FS_GOING_DOWN_NOSYNC:
1897 f2fs_stop_checkpoint(sbi, false);
1898 break;
1899 case F2FS_GOING_DOWN_METAFLUSH:
1900 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1901 f2fs_stop_checkpoint(sbi, false);
1902 break;
1903 default:
1904 ret = -EINVAL;
1905 goto out;
1906 }
1907
1908 stop_gc_thread(sbi);
1909 stop_discard_thread(sbi);
1910
1911 drop_discard_cmd(sbi);
1912 clear_opt(sbi, DISCARD);
1913
1914 f2fs_update_time(sbi, REQ_TIME);
1915 out:
1916 if (in != F2FS_GOING_DOWN_FULLSYNC)
1917 mnt_drop_write_file(filp);
1918 return ret;
1919 }
1920
1921 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1922 {
1923 struct inode *inode = file_inode(filp);
1924 struct super_block *sb = inode->i_sb;
1925 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1926 struct fstrim_range range;
1927 int ret;
1928
1929 if (!capable(CAP_SYS_ADMIN))
1930 return -EPERM;
1931
1932 if (!blk_queue_discard(q))
1933 return -EOPNOTSUPP;
1934
1935 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1936 sizeof(range)))
1937 return -EFAULT;
1938
1939 ret = mnt_want_write_file(filp);
1940 if (ret)
1941 return ret;
1942
1943 range.minlen = max((unsigned int)range.minlen,
1944 q->limits.discard_granularity);
1945 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1946 mnt_drop_write_file(filp);
1947 if (ret < 0)
1948 return ret;
1949
1950 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1951 sizeof(range)))
1952 return -EFAULT;
1953 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1954 return 0;
1955 }
1956
1957 static bool uuid_is_nonzero(__u8 u[16])
1958 {
1959 int i;
1960
1961 for (i = 0; i < 16; i++)
1962 if (u[i])
1963 return true;
1964 return false;
1965 }
1966
1967 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1968 {
1969 struct inode *inode = file_inode(filp);
1970
1971 if (!f2fs_sb_has_encrypt(inode->i_sb))
1972 return -EOPNOTSUPP;
1973
1974 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1975
1976 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1977 }
1978
1979 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1980 {
1981 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
1982 return -EOPNOTSUPP;
1983 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1984 }
1985
1986 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1987 {
1988 struct inode *inode = file_inode(filp);
1989 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1990 int err;
1991
1992 if (!f2fs_sb_has_encrypt(inode->i_sb))
1993 return -EOPNOTSUPP;
1994
1995 err = mnt_want_write_file(filp);
1996 if (err)
1997 return err;
1998
1999 down_write(&sbi->sb_lock);
2000
2001 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2002 goto got_it;
2003
2004 /* update superblock with uuid */
2005 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2006
2007 err = f2fs_commit_super(sbi, false);
2008 if (err) {
2009 /* undo new data */
2010 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2011 goto out_err;
2012 }
2013 got_it:
2014 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2015 16))
2016 err = -EFAULT;
2017 out_err:
2018 up_write(&sbi->sb_lock);
2019 mnt_drop_write_file(filp);
2020 return err;
2021 }
2022
2023 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2024 {
2025 struct inode *inode = file_inode(filp);
2026 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2027 __u32 sync;
2028 int ret;
2029
2030 if (!capable(CAP_SYS_ADMIN))
2031 return -EPERM;
2032
2033 if (get_user(sync, (__u32 __user *)arg))
2034 return -EFAULT;
2035
2036 if (f2fs_readonly(sbi->sb))
2037 return -EROFS;
2038
2039 ret = mnt_want_write_file(filp);
2040 if (ret)
2041 return ret;
2042
2043 if (!sync) {
2044 if (!mutex_trylock(&sbi->gc_mutex)) {
2045 ret = -EBUSY;
2046 goto out;
2047 }
2048 } else {
2049 mutex_lock(&sbi->gc_mutex);
2050 }
2051
2052 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2053 out:
2054 mnt_drop_write_file(filp);
2055 return ret;
2056 }
2057
2058 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2059 {
2060 struct inode *inode = file_inode(filp);
2061 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2062 struct f2fs_gc_range range;
2063 u64 end;
2064 int ret;
2065
2066 if (!capable(CAP_SYS_ADMIN))
2067 return -EPERM;
2068
2069 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2070 sizeof(range)))
2071 return -EFAULT;
2072
2073 if (f2fs_readonly(sbi->sb))
2074 return -EROFS;
2075
2076 ret = mnt_want_write_file(filp);
2077 if (ret)
2078 return ret;
2079
2080 end = range.start + range.len;
2081 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2082 ret = -EINVAL;
2083 goto out;
2084 }
2085 do_more:
2086 if (!range.sync) {
2087 if (!mutex_trylock(&sbi->gc_mutex)) {
2088 ret = -EBUSY;
2089 goto out;
2090 }
2091 } else {
2092 mutex_lock(&sbi->gc_mutex);
2093 }
2094
2095 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2096 range.start += sbi->blocks_per_seg;
2097 if (range.start <= end)
2098 goto do_more;
2099 out:
2100 mnt_drop_write_file(filp);
2101 return ret;
2102 }
2103
2104 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2105 {
2106 struct inode *inode = file_inode(filp);
2107 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2108 int ret;
2109
2110 if (!capable(CAP_SYS_ADMIN))
2111 return -EPERM;
2112
2113 if (f2fs_readonly(sbi->sb))
2114 return -EROFS;
2115
2116 ret = mnt_want_write_file(filp);
2117 if (ret)
2118 return ret;
2119
2120 ret = f2fs_sync_fs(sbi->sb, 1);
2121
2122 mnt_drop_write_file(filp);
2123 return ret;
2124 }
2125
2126 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2127 struct file *filp,
2128 struct f2fs_defragment *range)
2129 {
2130 struct inode *inode = file_inode(filp);
2131 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2132 .m_seg_type = NO_CHECK_TYPE };
2133 struct extent_info ei = {0,0,0};
2134 pgoff_t pg_start, pg_end, next_pgofs;
2135 unsigned int blk_per_seg = sbi->blocks_per_seg;
2136 unsigned int total = 0, sec_num;
2137 block_t blk_end = 0;
2138 bool fragmented = false;
2139 int err;
2140
2141 /* if in-place-update policy is enabled, don't waste time here */
2142 if (should_update_inplace(inode, NULL))
2143 return -EINVAL;
2144
2145 pg_start = range->start >> PAGE_SHIFT;
2146 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2147
2148 f2fs_balance_fs(sbi, true);
2149
2150 inode_lock(inode);
2151
2152 /* writeback all dirty pages in the range */
2153 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2154 range->start + range->len - 1);
2155 if (err)
2156 goto out;
2157
2158 /*
2159 * lookup mapping info in extent cache, skip defragmenting if physical
2160 * block addresses are continuous.
2161 */
2162 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2163 if (ei.fofs + ei.len >= pg_end)
2164 goto out;
2165 }
2166
2167 map.m_lblk = pg_start;
2168 map.m_next_pgofs = &next_pgofs;
2169
2170 /*
2171 * lookup mapping info in dnode page cache, skip defragmenting if all
2172 * physical block addresses are continuous even if there are hole(s)
2173 * in logical blocks.
2174 */
2175 while (map.m_lblk < pg_end) {
2176 map.m_len = pg_end - map.m_lblk;
2177 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2178 if (err)
2179 goto out;
2180
2181 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2182 map.m_lblk = next_pgofs;
2183 continue;
2184 }
2185
2186 if (blk_end && blk_end != map.m_pblk)
2187 fragmented = true;
2188
2189 /* record total count of block that we're going to move */
2190 total += map.m_len;
2191
2192 blk_end = map.m_pblk + map.m_len;
2193
2194 map.m_lblk += map.m_len;
2195 }
2196
2197 if (!fragmented)
2198 goto out;
2199
2200 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2201
2202 /*
2203 * make sure there are enough free section for LFS allocation, this can
2204 * avoid defragment running in SSR mode when free section are allocated
2205 * intensively
2206 */
2207 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2208 err = -EAGAIN;
2209 goto out;
2210 }
2211
2212 map.m_lblk = pg_start;
2213 map.m_len = pg_end - pg_start;
2214 total = 0;
2215
2216 while (map.m_lblk < pg_end) {
2217 pgoff_t idx;
2218 int cnt = 0;
2219
2220 do_map:
2221 map.m_len = pg_end - map.m_lblk;
2222 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2223 if (err)
2224 goto clear_out;
2225
2226 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2227 map.m_lblk = next_pgofs;
2228 continue;
2229 }
2230
2231 set_inode_flag(inode, FI_DO_DEFRAG);
2232
2233 idx = map.m_lblk;
2234 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2235 struct page *page;
2236
2237 page = get_lock_data_page(inode, idx, true);
2238 if (IS_ERR(page)) {
2239 err = PTR_ERR(page);
2240 goto clear_out;
2241 }
2242
2243 set_page_dirty(page);
2244 f2fs_put_page(page, 1);
2245
2246 idx++;
2247 cnt++;
2248 total++;
2249 }
2250
2251 map.m_lblk = idx;
2252
2253 if (idx < pg_end && cnt < blk_per_seg)
2254 goto do_map;
2255
2256 clear_inode_flag(inode, FI_DO_DEFRAG);
2257
2258 err = filemap_fdatawrite(inode->i_mapping);
2259 if (err)
2260 goto out;
2261 }
2262 clear_out:
2263 clear_inode_flag(inode, FI_DO_DEFRAG);
2264 out:
2265 inode_unlock(inode);
2266 if (!err)
2267 range->len = (u64)total << PAGE_SHIFT;
2268 return err;
2269 }
2270
2271 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2272 {
2273 struct inode *inode = file_inode(filp);
2274 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2275 struct f2fs_defragment range;
2276 int err;
2277
2278 if (!capable(CAP_SYS_ADMIN))
2279 return -EPERM;
2280
2281 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2282 return -EINVAL;
2283
2284 if (f2fs_readonly(sbi->sb))
2285 return -EROFS;
2286
2287 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2288 sizeof(range)))
2289 return -EFAULT;
2290
2291 /* verify alignment of offset & size */
2292 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2293 return -EINVAL;
2294
2295 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2296 sbi->max_file_blocks))
2297 return -EINVAL;
2298
2299 err = mnt_want_write_file(filp);
2300 if (err)
2301 return err;
2302
2303 err = f2fs_defragment_range(sbi, filp, &range);
2304 mnt_drop_write_file(filp);
2305
2306 f2fs_update_time(sbi, REQ_TIME);
2307 if (err < 0)
2308 return err;
2309
2310 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2311 sizeof(range)))
2312 return -EFAULT;
2313
2314 return 0;
2315 }
2316
2317 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2318 struct file *file_out, loff_t pos_out, size_t len)
2319 {
2320 struct inode *src = file_inode(file_in);
2321 struct inode *dst = file_inode(file_out);
2322 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2323 size_t olen = len, dst_max_i_size = 0;
2324 size_t dst_osize;
2325 int ret;
2326
2327 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2328 src->i_sb != dst->i_sb)
2329 return -EXDEV;
2330
2331 if (unlikely(f2fs_readonly(src->i_sb)))
2332 return -EROFS;
2333
2334 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2335 return -EINVAL;
2336
2337 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2338 return -EOPNOTSUPP;
2339
2340 if (src == dst) {
2341 if (pos_in == pos_out)
2342 return 0;
2343 if (pos_out > pos_in && pos_out < pos_in + len)
2344 return -EINVAL;
2345 }
2346
2347 inode_lock(src);
2348 down_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2349 if (src != dst) {
2350 ret = -EBUSY;
2351 if (!inode_trylock(dst))
2352 goto out;
2353 if (!down_write_trylock(&F2FS_I(dst)->dio_rwsem[WRITE])) {
2354 inode_unlock(dst);
2355 goto out;
2356 }
2357 }
2358
2359 ret = -EINVAL;
2360 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2361 goto out_unlock;
2362 if (len == 0)
2363 olen = len = src->i_size - pos_in;
2364 if (pos_in + len == src->i_size)
2365 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2366 if (len == 0) {
2367 ret = 0;
2368 goto out_unlock;
2369 }
2370
2371 dst_osize = dst->i_size;
2372 if (pos_out + olen > dst->i_size)
2373 dst_max_i_size = pos_out + olen;
2374
2375 /* verify the end result is block aligned */
2376 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2377 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2378 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2379 goto out_unlock;
2380
2381 ret = f2fs_convert_inline_inode(src);
2382 if (ret)
2383 goto out_unlock;
2384
2385 ret = f2fs_convert_inline_inode(dst);
2386 if (ret)
2387 goto out_unlock;
2388
2389 /* write out all dirty pages from offset */
2390 ret = filemap_write_and_wait_range(src->i_mapping,
2391 pos_in, pos_in + len);
2392 if (ret)
2393 goto out_unlock;
2394
2395 ret = filemap_write_and_wait_range(dst->i_mapping,
2396 pos_out, pos_out + len);
2397 if (ret)
2398 goto out_unlock;
2399
2400 f2fs_balance_fs(sbi, true);
2401 f2fs_lock_op(sbi);
2402 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2403 pos_out >> F2FS_BLKSIZE_BITS,
2404 len >> F2FS_BLKSIZE_BITS, false);
2405
2406 if (!ret) {
2407 if (dst_max_i_size)
2408 f2fs_i_size_write(dst, dst_max_i_size);
2409 else if (dst_osize != dst->i_size)
2410 f2fs_i_size_write(dst, dst_osize);
2411 }
2412 f2fs_unlock_op(sbi);
2413 out_unlock:
2414 if (src != dst) {
2415 up_write(&F2FS_I(dst)->dio_rwsem[WRITE]);
2416 inode_unlock(dst);
2417 }
2418 out:
2419 up_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2420 inode_unlock(src);
2421 return ret;
2422 }
2423
2424 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2425 {
2426 struct f2fs_move_range range;
2427 struct fd dst;
2428 int err;
2429
2430 if (!(filp->f_mode & FMODE_READ) ||
2431 !(filp->f_mode & FMODE_WRITE))
2432 return -EBADF;
2433
2434 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2435 sizeof(range)))
2436 return -EFAULT;
2437
2438 dst = fdget(range.dst_fd);
2439 if (!dst.file)
2440 return -EBADF;
2441
2442 if (!(dst.file->f_mode & FMODE_WRITE)) {
2443 err = -EBADF;
2444 goto err_out;
2445 }
2446
2447 err = mnt_want_write_file(filp);
2448 if (err)
2449 goto err_out;
2450
2451 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2452 range.pos_out, range.len);
2453
2454 mnt_drop_write_file(filp);
2455 if (err)
2456 goto err_out;
2457
2458 if (copy_to_user((struct f2fs_move_range __user *)arg,
2459 &range, sizeof(range)))
2460 err = -EFAULT;
2461 err_out:
2462 fdput(dst);
2463 return err;
2464 }
2465
2466 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2467 {
2468 struct inode *inode = file_inode(filp);
2469 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2470 struct sit_info *sm = SIT_I(sbi);
2471 unsigned int start_segno = 0, end_segno = 0;
2472 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2473 struct f2fs_flush_device range;
2474 int ret;
2475
2476 if (!capable(CAP_SYS_ADMIN))
2477 return -EPERM;
2478
2479 if (f2fs_readonly(sbi->sb))
2480 return -EROFS;
2481
2482 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2483 sizeof(range)))
2484 return -EFAULT;
2485
2486 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2487 sbi->segs_per_sec != 1) {
2488 f2fs_msg(sbi->sb, KERN_WARNING,
2489 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2490 range.dev_num, sbi->s_ndevs,
2491 sbi->segs_per_sec);
2492 return -EINVAL;
2493 }
2494
2495 ret = mnt_want_write_file(filp);
2496 if (ret)
2497 return ret;
2498
2499 if (range.dev_num != 0)
2500 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2501 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2502
2503 start_segno = sm->last_victim[FLUSH_DEVICE];
2504 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2505 start_segno = dev_start_segno;
2506 end_segno = min(start_segno + range.segments, dev_end_segno);
2507
2508 while (start_segno < end_segno) {
2509 if (!mutex_trylock(&sbi->gc_mutex)) {
2510 ret = -EBUSY;
2511 goto out;
2512 }
2513 sm->last_victim[GC_CB] = end_segno + 1;
2514 sm->last_victim[GC_GREEDY] = end_segno + 1;
2515 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2516 ret = f2fs_gc(sbi, true, true, start_segno);
2517 if (ret == -EAGAIN)
2518 ret = 0;
2519 else if (ret < 0)
2520 break;
2521 start_segno++;
2522 }
2523 out:
2524 mnt_drop_write_file(filp);
2525 return ret;
2526 }
2527
2528 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2529 {
2530 struct inode *inode = file_inode(filp);
2531 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2532
2533 /* Must validate to set it with SQLite behavior in Android. */
2534 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2535
2536 return put_user(sb_feature, (u32 __user *)arg);
2537 }
2538
2539 #ifdef CONFIG_QUOTA
2540 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2541 {
2542 struct inode *inode = file_inode(filp);
2543 struct f2fs_inode_info *fi = F2FS_I(inode);
2544 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2545 struct super_block *sb = sbi->sb;
2546 struct dquot *transfer_to[MAXQUOTAS] = {};
2547 struct page *ipage;
2548 kprojid_t kprojid;
2549 int err;
2550
2551 if (!f2fs_sb_has_project_quota(sb)) {
2552 if (projid != F2FS_DEF_PROJID)
2553 return -EOPNOTSUPP;
2554 else
2555 return 0;
2556 }
2557
2558 if (!f2fs_has_extra_attr(inode))
2559 return -EOPNOTSUPP;
2560
2561 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2562
2563 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2564 return 0;
2565
2566 err = mnt_want_write_file(filp);
2567 if (err)
2568 return err;
2569
2570 err = -EPERM;
2571 inode_lock(inode);
2572
2573 /* Is it quota file? Do not allow user to mess with it */
2574 if (IS_NOQUOTA(inode))
2575 goto out_unlock;
2576
2577 ipage = get_node_page(sbi, inode->i_ino);
2578 if (IS_ERR(ipage)) {
2579 err = PTR_ERR(ipage);
2580 goto out_unlock;
2581 }
2582
2583 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2584 i_projid)) {
2585 err = -EOVERFLOW;
2586 f2fs_put_page(ipage, 1);
2587 goto out_unlock;
2588 }
2589 f2fs_put_page(ipage, 1);
2590
2591 err = dquot_initialize(inode);
2592 if (err)
2593 goto out_unlock;
2594
2595 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2596 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2597 err = __dquot_transfer(inode, transfer_to);
2598 dqput(transfer_to[PRJQUOTA]);
2599 if (err)
2600 goto out_dirty;
2601 }
2602
2603 F2FS_I(inode)->i_projid = kprojid;
2604 inode->i_ctime = current_time(inode);
2605 out_dirty:
2606 f2fs_mark_inode_dirty_sync(inode, true);
2607 out_unlock:
2608 inode_unlock(inode);
2609 mnt_drop_write_file(filp);
2610 return err;
2611 }
2612 #else
2613 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2614 {
2615 if (projid != F2FS_DEF_PROJID)
2616 return -EOPNOTSUPP;
2617 return 0;
2618 }
2619 #endif
2620
2621 /* Transfer internal flags to xflags */
2622 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2623 {
2624 __u32 xflags = 0;
2625
2626 if (iflags & FS_SYNC_FL)
2627 xflags |= FS_XFLAG_SYNC;
2628 if (iflags & FS_IMMUTABLE_FL)
2629 xflags |= FS_XFLAG_IMMUTABLE;
2630 if (iflags & FS_APPEND_FL)
2631 xflags |= FS_XFLAG_APPEND;
2632 if (iflags & FS_NODUMP_FL)
2633 xflags |= FS_XFLAG_NODUMP;
2634 if (iflags & FS_NOATIME_FL)
2635 xflags |= FS_XFLAG_NOATIME;
2636 if (iflags & FS_PROJINHERIT_FL)
2637 xflags |= FS_XFLAG_PROJINHERIT;
2638 return xflags;
2639 }
2640
2641 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2642 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2643 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2644
2645 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2646 #define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2647 FS_IMMUTABLE_FL | \
2648 FS_APPEND_FL | \
2649 FS_NODUMP_FL | \
2650 FS_NOATIME_FL | \
2651 FS_PROJINHERIT_FL)
2652
2653 /* Transfer xflags flags to internal */
2654 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2655 {
2656 unsigned long iflags = 0;
2657
2658 if (xflags & FS_XFLAG_SYNC)
2659 iflags |= FS_SYNC_FL;
2660 if (xflags & FS_XFLAG_IMMUTABLE)
2661 iflags |= FS_IMMUTABLE_FL;
2662 if (xflags & FS_XFLAG_APPEND)
2663 iflags |= FS_APPEND_FL;
2664 if (xflags & FS_XFLAG_NODUMP)
2665 iflags |= FS_NODUMP_FL;
2666 if (xflags & FS_XFLAG_NOATIME)
2667 iflags |= FS_NOATIME_FL;
2668 if (xflags & FS_XFLAG_PROJINHERIT)
2669 iflags |= FS_PROJINHERIT_FL;
2670
2671 return iflags;
2672 }
2673
2674 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2675 {
2676 struct inode *inode = file_inode(filp);
2677 struct f2fs_inode_info *fi = F2FS_I(inode);
2678 struct fsxattr fa;
2679
2680 memset(&fa, 0, sizeof(struct fsxattr));
2681 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2682 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2683
2684 if (f2fs_sb_has_project_quota(inode->i_sb))
2685 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2686 fi->i_projid);
2687
2688 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2689 return -EFAULT;
2690 return 0;
2691 }
2692
2693 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2694 {
2695 struct inode *inode = file_inode(filp);
2696 struct f2fs_inode_info *fi = F2FS_I(inode);
2697 struct fsxattr fa;
2698 unsigned int flags;
2699 int err;
2700
2701 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2702 return -EFAULT;
2703
2704 /* Make sure caller has proper permission */
2705 if (!inode_owner_or_capable(inode))
2706 return -EACCES;
2707
2708 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2709 return -EOPNOTSUPP;
2710
2711 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2712 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2713 return -EOPNOTSUPP;
2714
2715 err = mnt_want_write_file(filp);
2716 if (err)
2717 return err;
2718
2719 inode_lock(inode);
2720 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2721 (flags & F2FS_FL_XFLAG_VISIBLE);
2722 err = __f2fs_ioc_setflags(inode, flags);
2723 inode_unlock(inode);
2724 mnt_drop_write_file(filp);
2725 if (err)
2726 return err;
2727
2728 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2729 if (err)
2730 return err;
2731
2732 return 0;
2733 }
2734
2735 int f2fs_pin_file_control(struct inode *inode, bool inc)
2736 {
2737 struct f2fs_inode_info *fi = F2FS_I(inode);
2738 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2739
2740 /* Use i_gc_failures for normal file as a risk signal. */
2741 if (inc)
2742 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
2743
2744 if (fi->i_gc_failures > sbi->gc_pin_file_threshold) {
2745 f2fs_msg(sbi->sb, KERN_WARNING,
2746 "%s: Enable GC = ino %lx after %x GC trials\n",
2747 __func__, inode->i_ino, fi->i_gc_failures);
2748 clear_inode_flag(inode, FI_PIN_FILE);
2749 return -EAGAIN;
2750 }
2751 return 0;
2752 }
2753
2754 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2755 {
2756 struct inode *inode = file_inode(filp);
2757 __u32 pin;
2758 int ret = 0;
2759
2760 if (!inode_owner_or_capable(inode))
2761 return -EACCES;
2762
2763 if (get_user(pin, (__u32 __user *)arg))
2764 return -EFAULT;
2765
2766 if (!S_ISREG(inode->i_mode))
2767 return -EINVAL;
2768
2769 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2770 return -EROFS;
2771
2772 ret = mnt_want_write_file(filp);
2773 if (ret)
2774 return ret;
2775
2776 inode_lock(inode);
2777
2778 if (should_update_outplace(inode, NULL)) {
2779 ret = -EINVAL;
2780 goto out;
2781 }
2782
2783 if (!pin) {
2784 clear_inode_flag(inode, FI_PIN_FILE);
2785 F2FS_I(inode)->i_gc_failures = 1;
2786 goto done;
2787 }
2788
2789 if (f2fs_pin_file_control(inode, false)) {
2790 ret = -EAGAIN;
2791 goto out;
2792 }
2793 ret = f2fs_convert_inline_inode(inode);
2794 if (ret)
2795 goto out;
2796
2797 set_inode_flag(inode, FI_PIN_FILE);
2798 ret = F2FS_I(inode)->i_gc_failures;
2799 done:
2800 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2801 out:
2802 inode_unlock(inode);
2803 mnt_drop_write_file(filp);
2804 return ret;
2805 }
2806
2807 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2808 {
2809 struct inode *inode = file_inode(filp);
2810 __u32 pin = 0;
2811
2812 if (is_inode_flag_set(inode, FI_PIN_FILE))
2813 pin = F2FS_I(inode)->i_gc_failures;
2814 return put_user(pin, (u32 __user *)arg);
2815 }
2816
2817 int f2fs_precache_extents(struct inode *inode)
2818 {
2819 struct f2fs_inode_info *fi = F2FS_I(inode);
2820 struct f2fs_map_blocks map;
2821 pgoff_t m_next_extent;
2822 loff_t end;
2823 int err;
2824
2825 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2826 return -EOPNOTSUPP;
2827
2828 map.m_lblk = 0;
2829 map.m_next_pgofs = NULL;
2830 map.m_next_extent = &m_next_extent;
2831 map.m_seg_type = NO_CHECK_TYPE;
2832 end = F2FS_I_SB(inode)->max_file_blocks;
2833
2834 while (map.m_lblk < end) {
2835 map.m_len = end - map.m_lblk;
2836
2837 down_write(&fi->dio_rwsem[WRITE]);
2838 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2839 up_write(&fi->dio_rwsem[WRITE]);
2840 if (err)
2841 return err;
2842
2843 map.m_lblk = m_next_extent;
2844 }
2845
2846 return err;
2847 }
2848
2849 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2850 {
2851 return f2fs_precache_extents(file_inode(filp));
2852 }
2853
2854 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2855 {
2856 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2857 return -EIO;
2858
2859 switch (cmd) {
2860 case F2FS_IOC_GETFLAGS:
2861 return f2fs_ioc_getflags(filp, arg);
2862 case F2FS_IOC_SETFLAGS:
2863 return f2fs_ioc_setflags(filp, arg);
2864 case F2FS_IOC_GETVERSION:
2865 return f2fs_ioc_getversion(filp, arg);
2866 case F2FS_IOC_START_ATOMIC_WRITE:
2867 return f2fs_ioc_start_atomic_write(filp);
2868 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2869 return f2fs_ioc_commit_atomic_write(filp);
2870 case F2FS_IOC_START_VOLATILE_WRITE:
2871 return f2fs_ioc_start_volatile_write(filp);
2872 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2873 return f2fs_ioc_release_volatile_write(filp);
2874 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2875 return f2fs_ioc_abort_volatile_write(filp);
2876 case F2FS_IOC_SHUTDOWN:
2877 return f2fs_ioc_shutdown(filp, arg);
2878 case FITRIM:
2879 return f2fs_ioc_fitrim(filp, arg);
2880 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2881 return f2fs_ioc_set_encryption_policy(filp, arg);
2882 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2883 return f2fs_ioc_get_encryption_policy(filp, arg);
2884 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2885 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2886 case F2FS_IOC_GARBAGE_COLLECT:
2887 return f2fs_ioc_gc(filp, arg);
2888 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2889 return f2fs_ioc_gc_range(filp, arg);
2890 case F2FS_IOC_WRITE_CHECKPOINT:
2891 return f2fs_ioc_write_checkpoint(filp, arg);
2892 case F2FS_IOC_DEFRAGMENT:
2893 return f2fs_ioc_defragment(filp, arg);
2894 case F2FS_IOC_MOVE_RANGE:
2895 return f2fs_ioc_move_range(filp, arg);
2896 case F2FS_IOC_FLUSH_DEVICE:
2897 return f2fs_ioc_flush_device(filp, arg);
2898 case F2FS_IOC_GET_FEATURES:
2899 return f2fs_ioc_get_features(filp, arg);
2900 case F2FS_IOC_FSGETXATTR:
2901 return f2fs_ioc_fsgetxattr(filp, arg);
2902 case F2FS_IOC_FSSETXATTR:
2903 return f2fs_ioc_fssetxattr(filp, arg);
2904 case F2FS_IOC_GET_PIN_FILE:
2905 return f2fs_ioc_get_pin_file(filp, arg);
2906 case F2FS_IOC_SET_PIN_FILE:
2907 return f2fs_ioc_set_pin_file(filp, arg);
2908 case F2FS_IOC_PRECACHE_EXTENTS:
2909 return f2fs_ioc_precache_extents(filp, arg);
2910 default:
2911 return -ENOTTY;
2912 }
2913 }
2914
2915 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2916 {
2917 struct file *file = iocb->ki_filp;
2918 struct inode *inode = file_inode(file);
2919 struct blk_plug plug;
2920 ssize_t ret;
2921
2922 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2923 return -EIO;
2924
2925 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
2926 return -EINVAL;
2927
2928 if (!inode_trylock(inode)) {
2929 if (iocb->ki_flags & IOCB_NOWAIT)
2930 return -EAGAIN;
2931 inode_lock(inode);
2932 }
2933
2934 ret = generic_write_checks(iocb, from);
2935 if (ret > 0) {
2936 bool preallocated = false;
2937 size_t target_size = 0;
2938 int err;
2939
2940 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2941 set_inode_flag(inode, FI_NO_PREALLOC);
2942
2943 if ((iocb->ki_flags & IOCB_NOWAIT) &&
2944 (iocb->ki_flags & IOCB_DIRECT)) {
2945 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
2946 iov_iter_count(from)) ||
2947 f2fs_has_inline_data(inode) ||
2948 f2fs_force_buffered_io(inode, WRITE)) {
2949 inode_unlock(inode);
2950 return -EAGAIN;
2951 }
2952
2953 } else {
2954 preallocated = true;
2955 target_size = iocb->ki_pos + iov_iter_count(from);
2956
2957 err = f2fs_preallocate_blocks(iocb, from);
2958 if (err) {
2959 clear_inode_flag(inode, FI_NO_PREALLOC);
2960 inode_unlock(inode);
2961 return err;
2962 }
2963 }
2964 blk_start_plug(&plug);
2965 ret = __generic_file_write_iter(iocb, from);
2966 blk_finish_plug(&plug);
2967 clear_inode_flag(inode, FI_NO_PREALLOC);
2968
2969 /* if we couldn't write data, we should deallocate blocks. */
2970 if (preallocated && i_size_read(inode) < target_size)
2971 f2fs_truncate(inode);
2972
2973 if (ret > 0)
2974 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2975 }
2976 inode_unlock(inode);
2977
2978 if (ret > 0)
2979 ret = generic_write_sync(iocb, ret);
2980 return ret;
2981 }
2982
2983 #ifdef CONFIG_COMPAT
2984 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2985 {
2986 switch (cmd) {
2987 case F2FS_IOC32_GETFLAGS:
2988 cmd = F2FS_IOC_GETFLAGS;
2989 break;
2990 case F2FS_IOC32_SETFLAGS:
2991 cmd = F2FS_IOC_SETFLAGS;
2992 break;
2993 case F2FS_IOC32_GETVERSION:
2994 cmd = F2FS_IOC_GETVERSION;
2995 break;
2996 case F2FS_IOC_START_ATOMIC_WRITE:
2997 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2998 case F2FS_IOC_START_VOLATILE_WRITE:
2999 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3000 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3001 case F2FS_IOC_SHUTDOWN:
3002 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3003 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3004 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3005 case F2FS_IOC_GARBAGE_COLLECT:
3006 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3007 case F2FS_IOC_WRITE_CHECKPOINT:
3008 case F2FS_IOC_DEFRAGMENT:
3009 case F2FS_IOC_MOVE_RANGE:
3010 case F2FS_IOC_FLUSH_DEVICE:
3011 case F2FS_IOC_GET_FEATURES:
3012 case F2FS_IOC_FSGETXATTR:
3013 case F2FS_IOC_FSSETXATTR:
3014 case F2FS_IOC_GET_PIN_FILE:
3015 case F2FS_IOC_SET_PIN_FILE:
3016 case F2FS_IOC_PRECACHE_EXTENTS:
3017 break;
3018 default:
3019 return -ENOIOCTLCMD;
3020 }
3021 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3022 }
3023 #endif
3024
3025 const struct file_operations f2fs_file_operations = {
3026 .llseek = f2fs_llseek,
3027 .read_iter = generic_file_read_iter,
3028 .write_iter = f2fs_file_write_iter,
3029 .open = f2fs_file_open,
3030 .release = f2fs_release_file,
3031 .mmap = f2fs_file_mmap,
3032 .flush = f2fs_file_flush,
3033 .fsync = f2fs_sync_file,
3034 .fallocate = f2fs_fallocate,
3035 .unlocked_ioctl = f2fs_ioctl,
3036 #ifdef CONFIG_COMPAT
3037 .compat_ioctl = f2fs_compat_ioctl,
3038 #endif
3039 .splice_read = generic_file_splice_read,
3040 .splice_write = iter_file_splice_write,
3041 };
Motorola kernel-slsi