]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - fs/f2fs/file.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: require key for truncate(2) of encrypted file
[mirror_ubuntu-artful-kernel.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 sb_start_pagefault(inode->i_sb);
57
58 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
59
60 /* block allocation */
61 f2fs_lock_op(sbi);
62 set_new_dnode(&dn, inode, NULL, NULL, 0);
63 err = f2fs_reserve_block(&dn, page->index);
64 if (err) {
65 f2fs_unlock_op(sbi);
66 goto out;
67 }
68 f2fs_put_dnode(&dn);
69 f2fs_unlock_op(sbi);
70
71 f2fs_balance_fs(sbi, dn.node_changed);
72
73 file_update_time(vmf->vma->vm_file);
74 down_read(&F2FS_I(inode)->i_mmap_sem);
75 lock_page(page);
76 if (unlikely(page->mapping != inode->i_mapping ||
77 page_offset(page) > i_size_read(inode) ||
78 !PageUptodate(page))) {
79 unlock_page(page);
80 err = -EFAULT;
81 goto out_sem;
82 }
83
84 /*
85 * check to see if the page is mapped already (no holes)
86 */
87 if (PageMappedToDisk(page))
88 goto mapped;
89
90 /* page is wholly or partially inside EOF */
91 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
92 i_size_read(inode)) {
93 unsigned offset;
94 offset = i_size_read(inode) & ~PAGE_MASK;
95 zero_user_segment(page, offset, PAGE_SIZE);
96 }
97 set_page_dirty(page);
98 if (!PageUptodate(page))
99 SetPageUptodate(page);
100
101 trace_f2fs_vm_page_mkwrite(page, DATA);
102 mapped:
103 /* fill the page */
104 f2fs_wait_on_page_writeback(page, DATA, false);
105
106 /* wait for GCed encrypted page writeback */
107 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
108 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
109
110 out_sem:
111 up_read(&F2FS_I(inode)->i_mmap_sem);
112 out:
113 sb_end_pagefault(inode->i_sb);
114 f2fs_update_time(sbi, REQ_TIME);
115 return block_page_mkwrite_return(err);
116 }
117
118 static const struct vm_operations_struct f2fs_file_vm_ops = {
119 .fault = f2fs_filemap_fault,
120 .map_pages = filemap_map_pages,
121 .page_mkwrite = f2fs_vm_page_mkwrite,
122 };
123
124 static int get_parent_ino(struct inode *inode, nid_t *pino)
125 {
126 struct dentry *dentry;
127
128 inode = igrab(inode);
129 dentry = d_find_any_alias(inode);
130 iput(inode);
131 if (!dentry)
132 return 0;
133
134 *pino = parent_ino(dentry);
135 dput(dentry);
136 return 1;
137 }
138
139 static inline bool need_do_checkpoint(struct inode *inode)
140 {
141 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
142 bool need_cp = false;
143
144 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
145 need_cp = true;
146 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
147 need_cp = true;
148 else if (file_wrong_pino(inode))
149 need_cp = true;
150 else if (!space_for_roll_forward(sbi))
151 need_cp = true;
152 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
153 need_cp = true;
154 else if (test_opt(sbi, FASTBOOT))
155 need_cp = true;
156 else if (sbi->active_logs == 2)
157 need_cp = true;
158
159 return need_cp;
160 }
161
162 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
163 {
164 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
165 bool ret = false;
166 /* But we need to avoid that there are some inode updates */
167 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
168 ret = true;
169 f2fs_put_page(i, 0);
170 return ret;
171 }
172
173 static void try_to_fix_pino(struct inode *inode)
174 {
175 struct f2fs_inode_info *fi = F2FS_I(inode);
176 nid_t pino;
177
178 down_write(&fi->i_sem);
179 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
180 get_parent_ino(inode, &pino)) {
181 f2fs_i_pino_write(inode, pino);
182 file_got_pino(inode);
183 }
184 up_write(&fi->i_sem);
185 }
186
187 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
188 int datasync, bool atomic)
189 {
190 struct inode *inode = file->f_mapping->host;
191 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
192 nid_t ino = inode->i_ino;
193 int ret = 0;
194 bool need_cp = false;
195 struct writeback_control wbc = {
196 .sync_mode = WB_SYNC_ALL,
197 .nr_to_write = LONG_MAX,
198 .for_reclaim = 0,
199 };
200
201 if (unlikely(f2fs_readonly(inode->i_sb)))
202 return 0;
203
204 trace_f2fs_sync_file_enter(inode);
205
206 /* if fdatasync is triggered, let's do in-place-update */
207 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
208 set_inode_flag(inode, FI_NEED_IPU);
209 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
210 clear_inode_flag(inode, FI_NEED_IPU);
211
212 if (ret) {
213 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
214 return ret;
215 }
216
217 /* if the inode is dirty, let's recover all the time */
218 if (!f2fs_skip_inode_update(inode, datasync)) {
219 f2fs_write_inode(inode, NULL);
220 goto go_write;
221 }
222
223 /*
224 * if there is no written data, don't waste time to write recovery info.
225 */
226 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
227 !exist_written_data(sbi, ino, APPEND_INO)) {
228
229 /* it may call write_inode just prior to fsync */
230 if (need_inode_page_update(sbi, ino))
231 goto go_write;
232
233 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
234 exist_written_data(sbi, ino, UPDATE_INO))
235 goto flush_out;
236 goto out;
237 }
238 go_write:
239 /*
240 * Both of fdatasync() and fsync() are able to be recovered from
241 * sudden-power-off.
242 */
243 down_read(&F2FS_I(inode)->i_sem);
244 need_cp = need_do_checkpoint(inode);
245 up_read(&F2FS_I(inode)->i_sem);
246
247 if (need_cp) {
248 /* all the dirty node pages should be flushed for POR */
249 ret = f2fs_sync_fs(inode->i_sb, 1);
250
251 /*
252 * We've secured consistency through sync_fs. Following pino
253 * will be used only for fsynced inodes after checkpoint.
254 */
255 try_to_fix_pino(inode);
256 clear_inode_flag(inode, FI_APPEND_WRITE);
257 clear_inode_flag(inode, FI_UPDATE_WRITE);
258 goto out;
259 }
260 sync_nodes:
261 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
262 if (ret)
263 goto out;
264
265 /* if cp_error was enabled, we should avoid infinite loop */
266 if (unlikely(f2fs_cp_error(sbi))) {
267 ret = -EIO;
268 goto out;
269 }
270
271 if (need_inode_block_update(sbi, ino)) {
272 f2fs_mark_inode_dirty_sync(inode, true);
273 f2fs_write_inode(inode, NULL);
274 goto sync_nodes;
275 }
276
277 ret = wait_on_node_pages_writeback(sbi, ino);
278 if (ret)
279 goto out;
280
281 /* once recovery info is written, don't need to tack this */
282 remove_ino_entry(sbi, ino, APPEND_INO);
283 clear_inode_flag(inode, FI_APPEND_WRITE);
284 flush_out:
285 remove_ino_entry(sbi, ino, UPDATE_INO);
286 clear_inode_flag(inode, FI_UPDATE_WRITE);
287 if (!atomic)
288 ret = f2fs_issue_flush(sbi);
289 f2fs_update_time(sbi, REQ_TIME);
290 out:
291 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
292 f2fs_trace_ios(NULL, 1);
293 return ret;
294 }
295
296 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
297 {
298 return f2fs_do_sync_file(file, start, end, datasync, false);
299 }
300
301 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
302 pgoff_t pgofs, int whence)
303 {
304 struct pagevec pvec;
305 int nr_pages;
306
307 if (whence != SEEK_DATA)
308 return 0;
309
310 /* find first dirty page index */
311 pagevec_init(&pvec, 0);
312 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
313 PAGECACHE_TAG_DIRTY, 1);
314 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
315 pagevec_release(&pvec);
316 return pgofs;
317 }
318
319 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
320 int whence)
321 {
322 switch (whence) {
323 case SEEK_DATA:
324 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
325 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
326 return true;
327 break;
328 case SEEK_HOLE:
329 if (blkaddr == NULL_ADDR)
330 return true;
331 break;
332 }
333 return false;
334 }
335
336 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
337 {
338 struct inode *inode = file->f_mapping->host;
339 loff_t maxbytes = inode->i_sb->s_maxbytes;
340 struct dnode_of_data dn;
341 pgoff_t pgofs, end_offset, dirty;
342 loff_t data_ofs = offset;
343 loff_t isize;
344 int err = 0;
345
346 inode_lock(inode);
347
348 isize = i_size_read(inode);
349 if (offset >= isize)
350 goto fail;
351
352 /* handle inline data case */
353 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
354 if (whence == SEEK_HOLE)
355 data_ofs = isize;
356 goto found;
357 }
358
359 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
360
361 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
362
363 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
364 set_new_dnode(&dn, inode, NULL, NULL, 0);
365 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
366 if (err && err != -ENOENT) {
367 goto fail;
368 } else if (err == -ENOENT) {
369 /* direct node does not exists */
370 if (whence == SEEK_DATA) {
371 pgofs = get_next_page_offset(&dn, pgofs);
372 continue;
373 } else {
374 goto found;
375 }
376 }
377
378 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
379
380 /* find data/hole in dnode block */
381 for (; dn.ofs_in_node < end_offset;
382 dn.ofs_in_node++, pgofs++,
383 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
384 block_t blkaddr;
385 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
386
387 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
388 f2fs_put_dnode(&dn);
389 goto found;
390 }
391 }
392 f2fs_put_dnode(&dn);
393 }
394
395 if (whence == SEEK_DATA)
396 goto fail;
397 found:
398 if (whence == SEEK_HOLE && data_ofs > isize)
399 data_ofs = isize;
400 inode_unlock(inode);
401 return vfs_setpos(file, data_ofs, maxbytes);
402 fail:
403 inode_unlock(inode);
404 return -ENXIO;
405 }
406
407 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
408 {
409 struct inode *inode = file->f_mapping->host;
410 loff_t maxbytes = inode->i_sb->s_maxbytes;
411
412 switch (whence) {
413 case SEEK_SET:
414 case SEEK_CUR:
415 case SEEK_END:
416 return generic_file_llseek_size(file, offset, whence,
417 maxbytes, i_size_read(inode));
418 case SEEK_DATA:
419 case SEEK_HOLE:
420 if (offset < 0)
421 return -ENXIO;
422 return f2fs_seek_block(file, offset, whence);
423 }
424
425 return -EINVAL;
426 }
427
428 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
429 {
430 struct inode *inode = file_inode(file);
431 int err;
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
443 static 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
465 int 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
507 void truncate_data_blocks(struct dnode_of_data *dn)
508 {
509 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
510 }
511
512 static 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 PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
534 truncate_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
544 int 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 truncate_inline_inode(inode, ipage, from);
572 f2fs_put_page(ipage, 1);
573 truncate_page = true;
574 goto out;
575 }
576
577 set_new_dnode(&dn, inode, ipage, NULL, 0);
578 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
579 if (err) {
580 if (err == -ENOENT)
581 goto free_next;
582 goto out;
583 }
584
585 count = ADDRS_PER_PAGE(dn.node_page, inode);
586
587 count -= dn.ofs_in_node;
588 f2fs_bug_on(sbi, count < 0);
589
590 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
591 truncate_data_blocks_range(&dn, count);
592 free_from += count;
593 }
594
595 f2fs_put_dnode(&dn);
596 free_next:
597 err = truncate_inode_blocks(inode, free_from);
598 out:
599 if (lock)
600 f2fs_unlock_op(sbi);
601 free_partial:
602 /* lastly zero out the first data page */
603 if (!err)
604 err = truncate_partial_data_page(inode, from, truncate_page);
605
606 trace_f2fs_truncate_blocks_exit(inode, err);
607 return err;
608 }
609
610 int f2fs_truncate(struct inode *inode)
611 {
612 int err;
613
614 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
615 S_ISLNK(inode->i_mode)))
616 return 0;
617
618 trace_f2fs_truncate(inode);
619
620 #ifdef CONFIG_F2FS_FAULT_INJECTION
621 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
622 f2fs_show_injection_info(FAULT_TRUNCATE);
623 return -EIO;
624 }
625 #endif
626 /* we should check inline_data size */
627 if (!f2fs_may_inline_data(inode)) {
628 err = f2fs_convert_inline_inode(inode);
629 if (err)
630 return err;
631 }
632
633 err = truncate_blocks(inode, i_size_read(inode), true);
634 if (err)
635 return err;
636
637 inode->i_mtime = inode->i_ctime = current_time(inode);
638 f2fs_mark_inode_dirty_sync(inode, false);
639 return 0;
640 }
641
642 int f2fs_getattr(const struct path *path, struct kstat *stat,
643 u32 request_mask, unsigned int query_flags)
644 {
645 struct inode *inode = d_inode(path->dentry);
646 struct f2fs_inode_info *fi = F2FS_I(inode);
647 unsigned int flags;
648
649 flags = fi->i_flags & FS_FL_USER_VISIBLE;
650 if (flags & FS_APPEND_FL)
651 stat->attributes |= STATX_ATTR_APPEND;
652 if (flags & FS_COMPR_FL)
653 stat->attributes |= STATX_ATTR_COMPRESSED;
654 if (f2fs_encrypted_inode(inode))
655 stat->attributes |= STATX_ATTR_ENCRYPTED;
656 if (flags & FS_IMMUTABLE_FL)
657 stat->attributes |= STATX_ATTR_IMMUTABLE;
658 if (flags & FS_NODUMP_FL)
659 stat->attributes |= STATX_ATTR_NODUMP;
660
661 stat->attributes_mask |= (STATX_ATTR_APPEND |
662 STATX_ATTR_COMPRESSED |
663 STATX_ATTR_ENCRYPTED |
664 STATX_ATTR_IMMUTABLE |
665 STATX_ATTR_NODUMP);
666
667 generic_fillattr(inode, stat);
668 stat->blocks <<= 3;
669 return 0;
670 }
671
672 #ifdef CONFIG_F2FS_FS_POSIX_ACL
673 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
674 {
675 unsigned int ia_valid = attr->ia_valid;
676
677 if (ia_valid & ATTR_UID)
678 inode->i_uid = attr->ia_uid;
679 if (ia_valid & ATTR_GID)
680 inode->i_gid = attr->ia_gid;
681 if (ia_valid & ATTR_ATIME)
682 inode->i_atime = timespec_trunc(attr->ia_atime,
683 inode->i_sb->s_time_gran);
684 if (ia_valid & ATTR_MTIME)
685 inode->i_mtime = timespec_trunc(attr->ia_mtime,
686 inode->i_sb->s_time_gran);
687 if (ia_valid & ATTR_CTIME)
688 inode->i_ctime = timespec_trunc(attr->ia_ctime,
689 inode->i_sb->s_time_gran);
690 if (ia_valid & ATTR_MODE) {
691 umode_t mode = attr->ia_mode;
692
693 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
694 mode &= ~S_ISGID;
695 set_acl_inode(inode, mode);
696 }
697 }
698 #else
699 #define __setattr_copy setattr_copy
700 #endif
701
702 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
703 {
704 struct inode *inode = d_inode(dentry);
705 int err;
706 bool size_changed = false;
707
708 err = setattr_prepare(dentry, attr);
709 if (err)
710 return err;
711
712 if (attr->ia_valid & ATTR_SIZE) {
713 if (f2fs_encrypted_inode(inode)) {
714 err = fscrypt_get_encryption_info(inode);
715 if (err)
716 return err;
717 if (!fscrypt_has_encryption_key(inode))
718 return -ENOKEY;
719 }
720
721 if (attr->ia_size <= i_size_read(inode)) {
722 down_write(&F2FS_I(inode)->i_mmap_sem);
723 truncate_setsize(inode, attr->ia_size);
724 err = f2fs_truncate(inode);
725 up_write(&F2FS_I(inode)->i_mmap_sem);
726 if (err)
727 return err;
728 } else {
729 /*
730 * do not trim all blocks after i_size if target size is
731 * larger than i_size.
732 */
733 down_write(&F2FS_I(inode)->i_mmap_sem);
734 truncate_setsize(inode, attr->ia_size);
735 up_write(&F2FS_I(inode)->i_mmap_sem);
736
737 /* should convert inline inode here */
738 if (!f2fs_may_inline_data(inode)) {
739 err = f2fs_convert_inline_inode(inode);
740 if (err)
741 return err;
742 }
743 inode->i_mtime = inode->i_ctime = current_time(inode);
744 }
745
746 size_changed = true;
747 }
748
749 __setattr_copy(inode, attr);
750
751 if (attr->ia_valid & ATTR_MODE) {
752 err = posix_acl_chmod(inode, get_inode_mode(inode));
753 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
754 inode->i_mode = F2FS_I(inode)->i_acl_mode;
755 clear_inode_flag(inode, FI_ACL_MODE);
756 }
757 }
758
759 /* file size may changed here */
760 f2fs_mark_inode_dirty_sync(inode, size_changed);
761
762 /* inode change will produce dirty node pages flushed by checkpoint */
763 f2fs_balance_fs(F2FS_I_SB(inode), true);
764
765 return err;
766 }
767
768 const struct inode_operations f2fs_file_inode_operations = {
769 .getattr = f2fs_getattr,
770 .setattr = f2fs_setattr,
771 .get_acl = f2fs_get_acl,
772 .set_acl = f2fs_set_acl,
773 #ifdef CONFIG_F2FS_FS_XATTR
774 .listxattr = f2fs_listxattr,
775 #endif
776 .fiemap = f2fs_fiemap,
777 };
778
779 static int fill_zero(struct inode *inode, pgoff_t index,
780 loff_t start, loff_t len)
781 {
782 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
783 struct page *page;
784
785 if (!len)
786 return 0;
787
788 f2fs_balance_fs(sbi, true);
789
790 f2fs_lock_op(sbi);
791 page = get_new_data_page(inode, NULL, index, false);
792 f2fs_unlock_op(sbi);
793
794 if (IS_ERR(page))
795 return PTR_ERR(page);
796
797 f2fs_wait_on_page_writeback(page, DATA, true);
798 zero_user(page, start, len);
799 set_page_dirty(page);
800 f2fs_put_page(page, 1);
801 return 0;
802 }
803
804 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
805 {
806 int err;
807
808 while (pg_start < pg_end) {
809 struct dnode_of_data dn;
810 pgoff_t end_offset, count;
811
812 set_new_dnode(&dn, inode, NULL, NULL, 0);
813 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
814 if (err) {
815 if (err == -ENOENT) {
816 pg_start++;
817 continue;
818 }
819 return err;
820 }
821
822 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
823 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
824
825 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
826
827 truncate_data_blocks_range(&dn, count);
828 f2fs_put_dnode(&dn);
829
830 pg_start += count;
831 }
832 return 0;
833 }
834
835 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
836 {
837 pgoff_t pg_start, pg_end;
838 loff_t off_start, off_end;
839 int ret;
840
841 ret = f2fs_convert_inline_inode(inode);
842 if (ret)
843 return ret;
844
845 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
846 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
847
848 off_start = offset & (PAGE_SIZE - 1);
849 off_end = (offset + len) & (PAGE_SIZE - 1);
850
851 if (pg_start == pg_end) {
852 ret = fill_zero(inode, pg_start, off_start,
853 off_end - off_start);
854 if (ret)
855 return ret;
856 } else {
857 if (off_start) {
858 ret = fill_zero(inode, pg_start++, off_start,
859 PAGE_SIZE - off_start);
860 if (ret)
861 return ret;
862 }
863 if (off_end) {
864 ret = fill_zero(inode, pg_end, 0, off_end);
865 if (ret)
866 return ret;
867 }
868
869 if (pg_start < pg_end) {
870 struct address_space *mapping = inode->i_mapping;
871 loff_t blk_start, blk_end;
872 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
873
874 f2fs_balance_fs(sbi, true);
875
876 blk_start = (loff_t)pg_start << PAGE_SHIFT;
877 blk_end = (loff_t)pg_end << PAGE_SHIFT;
878 down_write(&F2FS_I(inode)->i_mmap_sem);
879 truncate_inode_pages_range(mapping, blk_start,
880 blk_end - 1);
881
882 f2fs_lock_op(sbi);
883 ret = truncate_hole(inode, pg_start, pg_end);
884 f2fs_unlock_op(sbi);
885 up_write(&F2FS_I(inode)->i_mmap_sem);
886 }
887 }
888
889 return ret;
890 }
891
892 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
893 int *do_replace, pgoff_t off, pgoff_t len)
894 {
895 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
896 struct dnode_of_data dn;
897 int ret, done, i;
898
899 next_dnode:
900 set_new_dnode(&dn, inode, NULL, NULL, 0);
901 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
902 if (ret && ret != -ENOENT) {
903 return ret;
904 } else if (ret == -ENOENT) {
905 if (dn.max_level == 0)
906 return -ENOENT;
907 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
908 blkaddr += done;
909 do_replace += done;
910 goto next;
911 }
912
913 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
914 dn.ofs_in_node, len);
915 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
916 *blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
917 if (!is_checkpointed_data(sbi, *blkaddr)) {
918
919 if (test_opt(sbi, LFS)) {
920 f2fs_put_dnode(&dn);
921 return -ENOTSUPP;
922 }
923
924 /* do not invalidate this block address */
925 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
926 *do_replace = 1;
927 }
928 }
929 f2fs_put_dnode(&dn);
930 next:
931 len -= done;
932 off += done;
933 if (len)
934 goto next_dnode;
935 return 0;
936 }
937
938 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
939 int *do_replace, pgoff_t off, int len)
940 {
941 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
942 struct dnode_of_data dn;
943 int ret, i;
944
945 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
946 if (*do_replace == 0)
947 continue;
948
949 set_new_dnode(&dn, inode, NULL, NULL, 0);
950 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
951 if (ret) {
952 dec_valid_block_count(sbi, inode, 1);
953 invalidate_blocks(sbi, *blkaddr);
954 } else {
955 f2fs_update_data_blkaddr(&dn, *blkaddr);
956 }
957 f2fs_put_dnode(&dn);
958 }
959 return 0;
960 }
961
962 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
963 block_t *blkaddr, int *do_replace,
964 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
965 {
966 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
967 pgoff_t i = 0;
968 int ret;
969
970 while (i < len) {
971 if (blkaddr[i] == NULL_ADDR && !full) {
972 i++;
973 continue;
974 }
975
976 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
977 struct dnode_of_data dn;
978 struct node_info ni;
979 size_t new_size;
980 pgoff_t ilen;
981
982 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
983 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
984 if (ret)
985 return ret;
986
987 get_node_info(sbi, dn.nid, &ni);
988 ilen = min((pgoff_t)
989 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
990 dn.ofs_in_node, len - i);
991 do {
992 dn.data_blkaddr = datablock_addr(dn.node_page,
993 dn.ofs_in_node);
994 truncate_data_blocks_range(&dn, 1);
995
996 if (do_replace[i]) {
997 f2fs_i_blocks_write(src_inode,
998 1, false);
999 f2fs_i_blocks_write(dst_inode,
1000 1, true);
1001 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1002 blkaddr[i], ni.version, true, false);
1003
1004 do_replace[i] = 0;
1005 }
1006 dn.ofs_in_node++;
1007 i++;
1008 new_size = (dst + i) << PAGE_SHIFT;
1009 if (dst_inode->i_size < new_size)
1010 f2fs_i_size_write(dst_inode, new_size);
1011 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1012
1013 f2fs_put_dnode(&dn);
1014 } else {
1015 struct page *psrc, *pdst;
1016
1017 psrc = get_lock_data_page(src_inode, src + i, true);
1018 if (IS_ERR(psrc))
1019 return PTR_ERR(psrc);
1020 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1021 true);
1022 if (IS_ERR(pdst)) {
1023 f2fs_put_page(psrc, 1);
1024 return PTR_ERR(pdst);
1025 }
1026 f2fs_copy_page(psrc, pdst);
1027 set_page_dirty(pdst);
1028 f2fs_put_page(pdst, 1);
1029 f2fs_put_page(psrc, 1);
1030
1031 ret = truncate_hole(src_inode, src + i, src + i + 1);
1032 if (ret)
1033 return ret;
1034 i++;
1035 }
1036 }
1037 return 0;
1038 }
1039
1040 static int __exchange_data_block(struct inode *src_inode,
1041 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1042 pgoff_t len, bool full)
1043 {
1044 block_t *src_blkaddr;
1045 int *do_replace;
1046 pgoff_t olen;
1047 int ret;
1048
1049 while (len) {
1050 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1051
1052 src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1053 if (!src_blkaddr)
1054 return -ENOMEM;
1055
1056 do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1057 if (!do_replace) {
1058 kvfree(src_blkaddr);
1059 return -ENOMEM;
1060 }
1061
1062 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1063 do_replace, src, olen);
1064 if (ret)
1065 goto roll_back;
1066
1067 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1068 do_replace, src, dst, olen, full);
1069 if (ret)
1070 goto roll_back;
1071
1072 src += olen;
1073 dst += olen;
1074 len -= olen;
1075
1076 kvfree(src_blkaddr);
1077 kvfree(do_replace);
1078 }
1079 return 0;
1080
1081 roll_back:
1082 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1083 kvfree(src_blkaddr);
1084 kvfree(do_replace);
1085 return ret;
1086 }
1087
1088 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1089 {
1090 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1091 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1092 int ret;
1093
1094 f2fs_balance_fs(sbi, true);
1095 f2fs_lock_op(sbi);
1096
1097 f2fs_drop_extent_tree(inode);
1098
1099 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1100 f2fs_unlock_op(sbi);
1101 return ret;
1102 }
1103
1104 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1105 {
1106 pgoff_t pg_start, pg_end;
1107 loff_t new_size;
1108 int ret;
1109
1110 if (offset + len >= i_size_read(inode))
1111 return -EINVAL;
1112
1113 /* collapse range should be aligned to block size of f2fs. */
1114 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1115 return -EINVAL;
1116
1117 ret = f2fs_convert_inline_inode(inode);
1118 if (ret)
1119 return ret;
1120
1121 pg_start = offset >> PAGE_SHIFT;
1122 pg_end = (offset + len) >> PAGE_SHIFT;
1123
1124 down_write(&F2FS_I(inode)->i_mmap_sem);
1125 /* write out all dirty pages from offset */
1126 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1127 if (ret)
1128 goto out;
1129
1130 truncate_pagecache(inode, offset);
1131
1132 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1133 if (ret)
1134 goto out;
1135
1136 /* write out all moved pages, if possible */
1137 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1138 truncate_pagecache(inode, offset);
1139
1140 new_size = i_size_read(inode) - len;
1141 truncate_pagecache(inode, new_size);
1142
1143 ret = truncate_blocks(inode, new_size, true);
1144 if (!ret)
1145 f2fs_i_size_write(inode, new_size);
1146
1147 out:
1148 up_write(&F2FS_I(inode)->i_mmap_sem);
1149 return ret;
1150 }
1151
1152 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1153 pgoff_t end)
1154 {
1155 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1156 pgoff_t index = start;
1157 unsigned int ofs_in_node = dn->ofs_in_node;
1158 blkcnt_t count = 0;
1159 int ret;
1160
1161 for (; index < end; index++, dn->ofs_in_node++) {
1162 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1163 count++;
1164 }
1165
1166 dn->ofs_in_node = ofs_in_node;
1167 ret = reserve_new_blocks(dn, count);
1168 if (ret)
1169 return ret;
1170
1171 dn->ofs_in_node = ofs_in_node;
1172 for (index = start; index < end; index++, dn->ofs_in_node++) {
1173 dn->data_blkaddr =
1174 datablock_addr(dn->node_page, dn->ofs_in_node);
1175 /*
1176 * reserve_new_blocks will not guarantee entire block
1177 * allocation.
1178 */
1179 if (dn->data_blkaddr == NULL_ADDR) {
1180 ret = -ENOSPC;
1181 break;
1182 }
1183 if (dn->data_blkaddr != NEW_ADDR) {
1184 invalidate_blocks(sbi, dn->data_blkaddr);
1185 dn->data_blkaddr = NEW_ADDR;
1186 set_data_blkaddr(dn);
1187 }
1188 }
1189
1190 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1191
1192 return ret;
1193 }
1194
1195 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1196 int mode)
1197 {
1198 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1199 struct address_space *mapping = inode->i_mapping;
1200 pgoff_t index, pg_start, pg_end;
1201 loff_t new_size = i_size_read(inode);
1202 loff_t off_start, off_end;
1203 int ret = 0;
1204
1205 ret = inode_newsize_ok(inode, (len + offset));
1206 if (ret)
1207 return ret;
1208
1209 ret = f2fs_convert_inline_inode(inode);
1210 if (ret)
1211 return ret;
1212
1213 down_write(&F2FS_I(inode)->i_mmap_sem);
1214 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1215 if (ret)
1216 goto out_sem;
1217
1218 truncate_pagecache_range(inode, offset, offset + len - 1);
1219
1220 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1221 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1222
1223 off_start = offset & (PAGE_SIZE - 1);
1224 off_end = (offset + len) & (PAGE_SIZE - 1);
1225
1226 if (pg_start == pg_end) {
1227 ret = fill_zero(inode, pg_start, off_start,
1228 off_end - off_start);
1229 if (ret)
1230 goto out_sem;
1231
1232 new_size = max_t(loff_t, new_size, offset + len);
1233 } else {
1234 if (off_start) {
1235 ret = fill_zero(inode, pg_start++, off_start,
1236 PAGE_SIZE - off_start);
1237 if (ret)
1238 goto out_sem;
1239
1240 new_size = max_t(loff_t, new_size,
1241 (loff_t)pg_start << PAGE_SHIFT);
1242 }
1243
1244 for (index = pg_start; index < pg_end;) {
1245 struct dnode_of_data dn;
1246 unsigned int end_offset;
1247 pgoff_t end;
1248
1249 f2fs_lock_op(sbi);
1250
1251 set_new_dnode(&dn, inode, NULL, NULL, 0);
1252 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1253 if (ret) {
1254 f2fs_unlock_op(sbi);
1255 goto out;
1256 }
1257
1258 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1259 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1260
1261 ret = f2fs_do_zero_range(&dn, index, end);
1262 f2fs_put_dnode(&dn);
1263 f2fs_unlock_op(sbi);
1264
1265 f2fs_balance_fs(sbi, dn.node_changed);
1266
1267 if (ret)
1268 goto out;
1269
1270 index = end;
1271 new_size = max_t(loff_t, new_size,
1272 (loff_t)index << PAGE_SHIFT);
1273 }
1274
1275 if (off_end) {
1276 ret = fill_zero(inode, pg_end, 0, off_end);
1277 if (ret)
1278 goto out;
1279
1280 new_size = max_t(loff_t, new_size, offset + len);
1281 }
1282 }
1283
1284 out:
1285 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1286 f2fs_i_size_write(inode, new_size);
1287 out_sem:
1288 up_write(&F2FS_I(inode)->i_mmap_sem);
1289
1290 return ret;
1291 }
1292
1293 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1294 {
1295 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1296 pgoff_t nr, pg_start, pg_end, delta, idx;
1297 loff_t new_size;
1298 int ret = 0;
1299
1300 new_size = i_size_read(inode) + len;
1301 ret = inode_newsize_ok(inode, new_size);
1302 if (ret)
1303 return ret;
1304
1305 if (offset >= i_size_read(inode))
1306 return -EINVAL;
1307
1308 /* insert range should be aligned to block size of f2fs. */
1309 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1310 return -EINVAL;
1311
1312 ret = f2fs_convert_inline_inode(inode);
1313 if (ret)
1314 return ret;
1315
1316 f2fs_balance_fs(sbi, true);
1317
1318 down_write(&F2FS_I(inode)->i_mmap_sem);
1319 ret = truncate_blocks(inode, i_size_read(inode), true);
1320 if (ret)
1321 goto out;
1322
1323 /* write out all dirty pages from offset */
1324 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1325 if (ret)
1326 goto out;
1327
1328 truncate_pagecache(inode, offset);
1329
1330 pg_start = offset >> PAGE_SHIFT;
1331 pg_end = (offset + len) >> PAGE_SHIFT;
1332 delta = pg_end - pg_start;
1333 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1334
1335 while (!ret && idx > pg_start) {
1336 nr = idx - pg_start;
1337 if (nr > delta)
1338 nr = delta;
1339 idx -= nr;
1340
1341 f2fs_lock_op(sbi);
1342 f2fs_drop_extent_tree(inode);
1343
1344 ret = __exchange_data_block(inode, inode, idx,
1345 idx + delta, nr, false);
1346 f2fs_unlock_op(sbi);
1347 }
1348
1349 /* write out all moved pages, if possible */
1350 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1351 truncate_pagecache(inode, offset);
1352
1353 if (!ret)
1354 f2fs_i_size_write(inode, new_size);
1355 out:
1356 up_write(&F2FS_I(inode)->i_mmap_sem);
1357 return ret;
1358 }
1359
1360 static int expand_inode_data(struct inode *inode, loff_t offset,
1361 loff_t len, int mode)
1362 {
1363 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1364 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1365 pgoff_t pg_end;
1366 loff_t new_size = i_size_read(inode);
1367 loff_t off_end;
1368 int err;
1369
1370 err = inode_newsize_ok(inode, (len + offset));
1371 if (err)
1372 return err;
1373
1374 err = f2fs_convert_inline_inode(inode);
1375 if (err)
1376 return err;
1377
1378 f2fs_balance_fs(sbi, true);
1379
1380 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1381 off_end = (offset + len) & (PAGE_SIZE - 1);
1382
1383 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1384 map.m_len = pg_end - map.m_lblk;
1385 if (off_end)
1386 map.m_len++;
1387
1388 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1389 if (err) {
1390 pgoff_t last_off;
1391
1392 if (!map.m_len)
1393 return err;
1394
1395 last_off = map.m_lblk + map.m_len - 1;
1396
1397 /* update new size to the failed position */
1398 new_size = (last_off == pg_end) ? offset + len:
1399 (loff_t)(last_off + 1) << PAGE_SHIFT;
1400 } else {
1401 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1402 }
1403
1404 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1405 f2fs_i_size_write(inode, new_size);
1406
1407 return err;
1408 }
1409
1410 static long f2fs_fallocate(struct file *file, int mode,
1411 loff_t offset, loff_t len)
1412 {
1413 struct inode *inode = file_inode(file);
1414 long ret = 0;
1415
1416 /* f2fs only support ->fallocate for regular file */
1417 if (!S_ISREG(inode->i_mode))
1418 return -EINVAL;
1419
1420 if (f2fs_encrypted_inode(inode) &&
1421 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1422 return -EOPNOTSUPP;
1423
1424 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1425 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1426 FALLOC_FL_INSERT_RANGE))
1427 return -EOPNOTSUPP;
1428
1429 inode_lock(inode);
1430
1431 if (mode & FALLOC_FL_PUNCH_HOLE) {
1432 if (offset >= inode->i_size)
1433 goto out;
1434
1435 ret = punch_hole(inode, offset, len);
1436 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1437 ret = f2fs_collapse_range(inode, offset, len);
1438 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1439 ret = f2fs_zero_range(inode, offset, len, mode);
1440 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1441 ret = f2fs_insert_range(inode, offset, len);
1442 } else {
1443 ret = expand_inode_data(inode, offset, len, mode);
1444 }
1445
1446 if (!ret) {
1447 inode->i_mtime = inode->i_ctime = current_time(inode);
1448 f2fs_mark_inode_dirty_sync(inode, false);
1449 if (mode & FALLOC_FL_KEEP_SIZE)
1450 file_set_keep_isize(inode);
1451 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1452 }
1453
1454 out:
1455 inode_unlock(inode);
1456
1457 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1458 return ret;
1459 }
1460
1461 static int f2fs_release_file(struct inode *inode, struct file *filp)
1462 {
1463 /*
1464 * f2fs_relase_file is called at every close calls. So we should
1465 * not drop any inmemory pages by close called by other process.
1466 */
1467 if (!(filp->f_mode & FMODE_WRITE) ||
1468 atomic_read(&inode->i_writecount) != 1)
1469 return 0;
1470
1471 /* some remained atomic pages should discarded */
1472 if (f2fs_is_atomic_file(inode))
1473 drop_inmem_pages(inode);
1474 if (f2fs_is_volatile_file(inode)) {
1475 clear_inode_flag(inode, FI_VOLATILE_FILE);
1476 stat_dec_volatile_write(inode);
1477 set_inode_flag(inode, FI_DROP_CACHE);
1478 filemap_fdatawrite(inode->i_mapping);
1479 clear_inode_flag(inode, FI_DROP_CACHE);
1480 }
1481 return 0;
1482 }
1483
1484 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1485 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1486
1487 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1488 {
1489 if (S_ISDIR(mode))
1490 return flags;
1491 else if (S_ISREG(mode))
1492 return flags & F2FS_REG_FLMASK;
1493 else
1494 return flags & F2FS_OTHER_FLMASK;
1495 }
1496
1497 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1498 {
1499 struct inode *inode = file_inode(filp);
1500 struct f2fs_inode_info *fi = F2FS_I(inode);
1501 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1502 return put_user(flags, (int __user *)arg);
1503 }
1504
1505 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1506 {
1507 struct inode *inode = file_inode(filp);
1508 struct f2fs_inode_info *fi = F2FS_I(inode);
1509 unsigned int flags;
1510 unsigned int oldflags;
1511 int ret;
1512
1513 if (!inode_owner_or_capable(inode))
1514 return -EACCES;
1515
1516 if (get_user(flags, (int __user *)arg))
1517 return -EFAULT;
1518
1519 ret = mnt_want_write_file(filp);
1520 if (ret)
1521 return ret;
1522
1523 inode_lock(inode);
1524
1525 flags = f2fs_mask_flags(inode->i_mode, flags);
1526
1527 oldflags = fi->i_flags;
1528
1529 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1530 if (!capable(CAP_LINUX_IMMUTABLE)) {
1531 inode_unlock(inode);
1532 ret = -EPERM;
1533 goto out;
1534 }
1535 }
1536
1537 flags = flags & FS_FL_USER_MODIFIABLE;
1538 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1539 fi->i_flags = flags;
1540
1541 inode->i_ctime = current_time(inode);
1542 f2fs_set_inode_flags(inode);
1543 f2fs_mark_inode_dirty_sync(inode, false);
1544
1545 inode_unlock(inode);
1546 out:
1547 mnt_drop_write_file(filp);
1548 return ret;
1549 }
1550
1551 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1552 {
1553 struct inode *inode = file_inode(filp);
1554
1555 return put_user(inode->i_generation, (int __user *)arg);
1556 }
1557
1558 static int f2fs_ioc_start_atomic_write(struct file *filp)
1559 {
1560 struct inode *inode = file_inode(filp);
1561 int ret;
1562
1563 if (!inode_owner_or_capable(inode))
1564 return -EACCES;
1565
1566 if (!S_ISREG(inode->i_mode))
1567 return -EINVAL;
1568
1569 ret = mnt_want_write_file(filp);
1570 if (ret)
1571 return ret;
1572
1573 inode_lock(inode);
1574
1575 if (f2fs_is_atomic_file(inode))
1576 goto out;
1577
1578 ret = f2fs_convert_inline_inode(inode);
1579 if (ret)
1580 goto out;
1581
1582 set_inode_flag(inode, FI_ATOMIC_FILE);
1583 set_inode_flag(inode, FI_HOT_DATA);
1584 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1585
1586 if (!get_dirty_pages(inode))
1587 goto inc_stat;
1588
1589 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1590 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1591 inode->i_ino, get_dirty_pages(inode));
1592 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1593 if (ret) {
1594 clear_inode_flag(inode, FI_ATOMIC_FILE);
1595 goto out;
1596 }
1597
1598 inc_stat:
1599 stat_inc_atomic_write(inode);
1600 stat_update_max_atomic_write(inode);
1601 out:
1602 inode_unlock(inode);
1603 mnt_drop_write_file(filp);
1604 return ret;
1605 }
1606
1607 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1608 {
1609 struct inode *inode = file_inode(filp);
1610 int ret;
1611
1612 if (!inode_owner_or_capable(inode))
1613 return -EACCES;
1614
1615 ret = mnt_want_write_file(filp);
1616 if (ret)
1617 return ret;
1618
1619 inode_lock(inode);
1620
1621 if (f2fs_is_volatile_file(inode))
1622 goto err_out;
1623
1624 if (f2fs_is_atomic_file(inode)) {
1625 ret = commit_inmem_pages(inode);
1626 if (ret)
1627 goto err_out;
1628
1629 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1630 if (!ret) {
1631 clear_inode_flag(inode, FI_ATOMIC_FILE);
1632 stat_dec_atomic_write(inode);
1633 }
1634 } else {
1635 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1636 }
1637 err_out:
1638 inode_unlock(inode);
1639 mnt_drop_write_file(filp);
1640 return ret;
1641 }
1642
1643 static int f2fs_ioc_start_volatile_write(struct file *filp)
1644 {
1645 struct inode *inode = file_inode(filp);
1646 int ret;
1647
1648 if (!inode_owner_or_capable(inode))
1649 return -EACCES;
1650
1651 if (!S_ISREG(inode->i_mode))
1652 return -EINVAL;
1653
1654 ret = mnt_want_write_file(filp);
1655 if (ret)
1656 return ret;
1657
1658 inode_lock(inode);
1659
1660 if (f2fs_is_volatile_file(inode))
1661 goto out;
1662
1663 ret = f2fs_convert_inline_inode(inode);
1664 if (ret)
1665 goto out;
1666
1667 stat_inc_volatile_write(inode);
1668 stat_update_max_volatile_write(inode);
1669
1670 set_inode_flag(inode, FI_VOLATILE_FILE);
1671 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1672 out:
1673 inode_unlock(inode);
1674 mnt_drop_write_file(filp);
1675 return ret;
1676 }
1677
1678 static int f2fs_ioc_release_volatile_write(struct file *filp)
1679 {
1680 struct inode *inode = file_inode(filp);
1681 int ret;
1682
1683 if (!inode_owner_or_capable(inode))
1684 return -EACCES;
1685
1686 ret = mnt_want_write_file(filp);
1687 if (ret)
1688 return ret;
1689
1690 inode_lock(inode);
1691
1692 if (!f2fs_is_volatile_file(inode))
1693 goto out;
1694
1695 if (!f2fs_is_first_block_written(inode)) {
1696 ret = truncate_partial_data_page(inode, 0, true);
1697 goto out;
1698 }
1699
1700 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1701 out:
1702 inode_unlock(inode);
1703 mnt_drop_write_file(filp);
1704 return ret;
1705 }
1706
1707 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1708 {
1709 struct inode *inode = file_inode(filp);
1710 int ret;
1711
1712 if (!inode_owner_or_capable(inode))
1713 return -EACCES;
1714
1715 ret = mnt_want_write_file(filp);
1716 if (ret)
1717 return ret;
1718
1719 inode_lock(inode);
1720
1721 if (f2fs_is_atomic_file(inode))
1722 drop_inmem_pages(inode);
1723 if (f2fs_is_volatile_file(inode)) {
1724 clear_inode_flag(inode, FI_VOLATILE_FILE);
1725 stat_dec_volatile_write(inode);
1726 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1727 }
1728
1729 inode_unlock(inode);
1730
1731 mnt_drop_write_file(filp);
1732 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1733 return ret;
1734 }
1735
1736 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1737 {
1738 struct inode *inode = file_inode(filp);
1739 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1740 struct super_block *sb = sbi->sb;
1741 __u32 in;
1742 int ret;
1743
1744 if (!capable(CAP_SYS_ADMIN))
1745 return -EPERM;
1746
1747 if (get_user(in, (__u32 __user *)arg))
1748 return -EFAULT;
1749
1750 ret = mnt_want_write_file(filp);
1751 if (ret)
1752 return ret;
1753
1754 switch (in) {
1755 case F2FS_GOING_DOWN_FULLSYNC:
1756 sb = freeze_bdev(sb->s_bdev);
1757 if (sb && !IS_ERR(sb)) {
1758 f2fs_stop_checkpoint(sbi, false);
1759 thaw_bdev(sb->s_bdev, sb);
1760 }
1761 break;
1762 case F2FS_GOING_DOWN_METASYNC:
1763 /* do checkpoint only */
1764 f2fs_sync_fs(sb, 1);
1765 f2fs_stop_checkpoint(sbi, false);
1766 break;
1767 case F2FS_GOING_DOWN_NOSYNC:
1768 f2fs_stop_checkpoint(sbi, false);
1769 break;
1770 case F2FS_GOING_DOWN_METAFLUSH:
1771 sync_meta_pages(sbi, META, LONG_MAX);
1772 f2fs_stop_checkpoint(sbi, false);
1773 break;
1774 default:
1775 ret = -EINVAL;
1776 goto out;
1777 }
1778 f2fs_update_time(sbi, REQ_TIME);
1779 out:
1780 mnt_drop_write_file(filp);
1781 return ret;
1782 }
1783
1784 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1785 {
1786 struct inode *inode = file_inode(filp);
1787 struct super_block *sb = inode->i_sb;
1788 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1789 struct fstrim_range range;
1790 int ret;
1791
1792 if (!capable(CAP_SYS_ADMIN))
1793 return -EPERM;
1794
1795 if (!blk_queue_discard(q))
1796 return -EOPNOTSUPP;
1797
1798 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1799 sizeof(range)))
1800 return -EFAULT;
1801
1802 ret = mnt_want_write_file(filp);
1803 if (ret)
1804 return ret;
1805
1806 range.minlen = max((unsigned int)range.minlen,
1807 q->limits.discard_granularity);
1808 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1809 mnt_drop_write_file(filp);
1810 if (ret < 0)
1811 return ret;
1812
1813 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1814 sizeof(range)))
1815 return -EFAULT;
1816 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1817 return 0;
1818 }
1819
1820 static bool uuid_is_nonzero(__u8 u[16])
1821 {
1822 int i;
1823
1824 for (i = 0; i < 16; i++)
1825 if (u[i])
1826 return true;
1827 return false;
1828 }
1829
1830 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1831 {
1832 struct inode *inode = file_inode(filp);
1833
1834 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1835
1836 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1837 }
1838
1839 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1840 {
1841 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1842 }
1843
1844 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1845 {
1846 struct inode *inode = file_inode(filp);
1847 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1848 int err;
1849
1850 if (!f2fs_sb_has_crypto(inode->i_sb))
1851 return -EOPNOTSUPP;
1852
1853 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1854 goto got_it;
1855
1856 err = mnt_want_write_file(filp);
1857 if (err)
1858 return err;
1859
1860 /* update superblock with uuid */
1861 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1862
1863 err = f2fs_commit_super(sbi, false);
1864 if (err) {
1865 /* undo new data */
1866 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1867 mnt_drop_write_file(filp);
1868 return err;
1869 }
1870 mnt_drop_write_file(filp);
1871 got_it:
1872 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1873 16))
1874 return -EFAULT;
1875 return 0;
1876 }
1877
1878 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1879 {
1880 struct inode *inode = file_inode(filp);
1881 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1882 __u32 sync;
1883 int ret;
1884
1885 if (!capable(CAP_SYS_ADMIN))
1886 return -EPERM;
1887
1888 if (get_user(sync, (__u32 __user *)arg))
1889 return -EFAULT;
1890
1891 if (f2fs_readonly(sbi->sb))
1892 return -EROFS;
1893
1894 ret = mnt_want_write_file(filp);
1895 if (ret)
1896 return ret;
1897
1898 if (!sync) {
1899 if (!mutex_trylock(&sbi->gc_mutex)) {
1900 ret = -EBUSY;
1901 goto out;
1902 }
1903 } else {
1904 mutex_lock(&sbi->gc_mutex);
1905 }
1906
1907 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
1908 out:
1909 mnt_drop_write_file(filp);
1910 return ret;
1911 }
1912
1913 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1914 {
1915 struct inode *inode = file_inode(filp);
1916 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1917 int ret;
1918
1919 if (!capable(CAP_SYS_ADMIN))
1920 return -EPERM;
1921
1922 if (f2fs_readonly(sbi->sb))
1923 return -EROFS;
1924
1925 ret = mnt_want_write_file(filp);
1926 if (ret)
1927 return ret;
1928
1929 ret = f2fs_sync_fs(sbi->sb, 1);
1930
1931 mnt_drop_write_file(filp);
1932 return ret;
1933 }
1934
1935 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1936 struct file *filp,
1937 struct f2fs_defragment *range)
1938 {
1939 struct inode *inode = file_inode(filp);
1940 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1941 struct extent_info ei = {0,0,0};
1942 pgoff_t pg_start, pg_end;
1943 unsigned int blk_per_seg = sbi->blocks_per_seg;
1944 unsigned int total = 0, sec_num;
1945 block_t blk_end = 0;
1946 bool fragmented = false;
1947 int err;
1948
1949 /* if in-place-update policy is enabled, don't waste time here */
1950 if (need_inplace_update_policy(inode, NULL))
1951 return -EINVAL;
1952
1953 pg_start = range->start >> PAGE_SHIFT;
1954 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1955
1956 f2fs_balance_fs(sbi, true);
1957
1958 inode_lock(inode);
1959
1960 /* writeback all dirty pages in the range */
1961 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1962 range->start + range->len - 1);
1963 if (err)
1964 goto out;
1965
1966 /*
1967 * lookup mapping info in extent cache, skip defragmenting if physical
1968 * block addresses are continuous.
1969 */
1970 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1971 if (ei.fofs + ei.len >= pg_end)
1972 goto out;
1973 }
1974
1975 map.m_lblk = pg_start;
1976
1977 /*
1978 * lookup mapping info in dnode page cache, skip defragmenting if all
1979 * physical block addresses are continuous even if there are hole(s)
1980 * in logical blocks.
1981 */
1982 while (map.m_lblk < pg_end) {
1983 map.m_len = pg_end - map.m_lblk;
1984 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1985 if (err)
1986 goto out;
1987
1988 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1989 map.m_lblk++;
1990 continue;
1991 }
1992
1993 if (blk_end && blk_end != map.m_pblk) {
1994 fragmented = true;
1995 break;
1996 }
1997 blk_end = map.m_pblk + map.m_len;
1998
1999 map.m_lblk += map.m_len;
2000 }
2001
2002 if (!fragmented)
2003 goto out;
2004
2005 map.m_lblk = pg_start;
2006 map.m_len = pg_end - pg_start;
2007
2008 sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2009
2010 /*
2011 * make sure there are enough free section for LFS allocation, this can
2012 * avoid defragment running in SSR mode when free section are allocated
2013 * intensively
2014 */
2015 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2016 err = -EAGAIN;
2017 goto out;
2018 }
2019
2020 while (map.m_lblk < pg_end) {
2021 pgoff_t idx;
2022 int cnt = 0;
2023
2024 do_map:
2025 map.m_len = pg_end - map.m_lblk;
2026 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
2027 if (err)
2028 goto clear_out;
2029
2030 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2031 map.m_lblk++;
2032 continue;
2033 }
2034
2035 set_inode_flag(inode, FI_DO_DEFRAG);
2036
2037 idx = map.m_lblk;
2038 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2039 struct page *page;
2040
2041 page = get_lock_data_page(inode, idx, true);
2042 if (IS_ERR(page)) {
2043 err = PTR_ERR(page);
2044 goto clear_out;
2045 }
2046
2047 set_page_dirty(page);
2048 f2fs_put_page(page, 1);
2049
2050 idx++;
2051 cnt++;
2052 total++;
2053 }
2054
2055 map.m_lblk = idx;
2056
2057 if (idx < pg_end && cnt < blk_per_seg)
2058 goto do_map;
2059
2060 clear_inode_flag(inode, FI_DO_DEFRAG);
2061
2062 err = filemap_fdatawrite(inode->i_mapping);
2063 if (err)
2064 goto out;
2065 }
2066 clear_out:
2067 clear_inode_flag(inode, FI_DO_DEFRAG);
2068 out:
2069 inode_unlock(inode);
2070 if (!err)
2071 range->len = (u64)total << PAGE_SHIFT;
2072 return err;
2073 }
2074
2075 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2076 {
2077 struct inode *inode = file_inode(filp);
2078 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2079 struct f2fs_defragment range;
2080 int err;
2081
2082 if (!capable(CAP_SYS_ADMIN))
2083 return -EPERM;
2084
2085 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2086 return -EINVAL;
2087
2088 if (f2fs_readonly(sbi->sb))
2089 return -EROFS;
2090
2091 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2092 sizeof(range)))
2093 return -EFAULT;
2094
2095 /* verify alignment of offset & size */
2096 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2097 return -EINVAL;
2098
2099 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2100 sbi->max_file_blocks))
2101 return -EINVAL;
2102
2103 err = mnt_want_write_file(filp);
2104 if (err)
2105 return err;
2106
2107 err = f2fs_defragment_range(sbi, filp, &range);
2108 mnt_drop_write_file(filp);
2109
2110 f2fs_update_time(sbi, REQ_TIME);
2111 if (err < 0)
2112 return err;
2113
2114 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2115 sizeof(range)))
2116 return -EFAULT;
2117
2118 return 0;
2119 }
2120
2121 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2122 struct file *file_out, loff_t pos_out, size_t len)
2123 {
2124 struct inode *src = file_inode(file_in);
2125 struct inode *dst = file_inode(file_out);
2126 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2127 size_t olen = len, dst_max_i_size = 0;
2128 size_t dst_osize;
2129 int ret;
2130
2131 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2132 src->i_sb != dst->i_sb)
2133 return -EXDEV;
2134
2135 if (unlikely(f2fs_readonly(src->i_sb)))
2136 return -EROFS;
2137
2138 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2139 return -EINVAL;
2140
2141 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2142 return -EOPNOTSUPP;
2143
2144 if (src == dst) {
2145 if (pos_in == pos_out)
2146 return 0;
2147 if (pos_out > pos_in && pos_out < pos_in + len)
2148 return -EINVAL;
2149 }
2150
2151 inode_lock(src);
2152 if (src != dst) {
2153 if (!inode_trylock(dst)) {
2154 ret = -EBUSY;
2155 goto out;
2156 }
2157 }
2158
2159 ret = -EINVAL;
2160 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2161 goto out_unlock;
2162 if (len == 0)
2163 olen = len = src->i_size - pos_in;
2164 if (pos_in + len == src->i_size)
2165 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2166 if (len == 0) {
2167 ret = 0;
2168 goto out_unlock;
2169 }
2170
2171 dst_osize = dst->i_size;
2172 if (pos_out + olen > dst->i_size)
2173 dst_max_i_size = pos_out + olen;
2174
2175 /* verify the end result is block aligned */
2176 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2177 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2178 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2179 goto out_unlock;
2180
2181 ret = f2fs_convert_inline_inode(src);
2182 if (ret)
2183 goto out_unlock;
2184
2185 ret = f2fs_convert_inline_inode(dst);
2186 if (ret)
2187 goto out_unlock;
2188
2189 /* write out all dirty pages from offset */
2190 ret = filemap_write_and_wait_range(src->i_mapping,
2191 pos_in, pos_in + len);
2192 if (ret)
2193 goto out_unlock;
2194
2195 ret = filemap_write_and_wait_range(dst->i_mapping,
2196 pos_out, pos_out + len);
2197 if (ret)
2198 goto out_unlock;
2199
2200 f2fs_balance_fs(sbi, true);
2201 f2fs_lock_op(sbi);
2202 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2203 pos_out >> F2FS_BLKSIZE_BITS,
2204 len >> F2FS_BLKSIZE_BITS, false);
2205
2206 if (!ret) {
2207 if (dst_max_i_size)
2208 f2fs_i_size_write(dst, dst_max_i_size);
2209 else if (dst_osize != dst->i_size)
2210 f2fs_i_size_write(dst, dst_osize);
2211 }
2212 f2fs_unlock_op(sbi);
2213 out_unlock:
2214 if (src != dst)
2215 inode_unlock(dst);
2216 out:
2217 inode_unlock(src);
2218 return ret;
2219 }
2220
2221 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2222 {
2223 struct f2fs_move_range range;
2224 struct fd dst;
2225 int err;
2226
2227 if (!(filp->f_mode & FMODE_READ) ||
2228 !(filp->f_mode & FMODE_WRITE))
2229 return -EBADF;
2230
2231 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2232 sizeof(range)))
2233 return -EFAULT;
2234
2235 dst = fdget(range.dst_fd);
2236 if (!dst.file)
2237 return -EBADF;
2238
2239 if (!(dst.file->f_mode & FMODE_WRITE)) {
2240 err = -EBADF;
2241 goto err_out;
2242 }
2243
2244 err = mnt_want_write_file(filp);
2245 if (err)
2246 goto err_out;
2247
2248 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2249 range.pos_out, range.len);
2250
2251 mnt_drop_write_file(filp);
2252 if (err)
2253 goto err_out;
2254
2255 if (copy_to_user((struct f2fs_move_range __user *)arg,
2256 &range, sizeof(range)))
2257 err = -EFAULT;
2258 err_out:
2259 fdput(dst);
2260 return err;
2261 }
2262
2263 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2264 {
2265 struct inode *inode = file_inode(filp);
2266 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2267 struct sit_info *sm = SIT_I(sbi);
2268 unsigned int start_segno = 0, end_segno = 0;
2269 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2270 struct f2fs_flush_device range;
2271 int ret;
2272
2273 if (!capable(CAP_SYS_ADMIN))
2274 return -EPERM;
2275
2276 if (f2fs_readonly(sbi->sb))
2277 return -EROFS;
2278
2279 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2280 sizeof(range)))
2281 return -EFAULT;
2282
2283 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2284 sbi->segs_per_sec != 1) {
2285 f2fs_msg(sbi->sb, KERN_WARNING,
2286 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2287 range.dev_num, sbi->s_ndevs,
2288 sbi->segs_per_sec);
2289 return -EINVAL;
2290 }
2291
2292 ret = mnt_want_write_file(filp);
2293 if (ret)
2294 return ret;
2295
2296 if (range.dev_num != 0)
2297 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2298 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2299
2300 start_segno = sm->last_victim[FLUSH_DEVICE];
2301 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2302 start_segno = dev_start_segno;
2303 end_segno = min(start_segno + range.segments, dev_end_segno);
2304
2305 while (start_segno < end_segno) {
2306 if (!mutex_trylock(&sbi->gc_mutex)) {
2307 ret = -EBUSY;
2308 goto out;
2309 }
2310 sm->last_victim[GC_CB] = end_segno + 1;
2311 sm->last_victim[GC_GREEDY] = end_segno + 1;
2312 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2313 ret = f2fs_gc(sbi, true, true, start_segno);
2314 if (ret == -EAGAIN)
2315 ret = 0;
2316 else if (ret < 0)
2317 break;
2318 start_segno++;
2319 }
2320 out:
2321 mnt_drop_write_file(filp);
2322 return ret;
2323 }
2324
2325
2326 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2327 {
2328 switch (cmd) {
2329 case F2FS_IOC_GETFLAGS:
2330 return f2fs_ioc_getflags(filp, arg);
2331 case F2FS_IOC_SETFLAGS:
2332 return f2fs_ioc_setflags(filp, arg);
2333 case F2FS_IOC_GETVERSION:
2334 return f2fs_ioc_getversion(filp, arg);
2335 case F2FS_IOC_START_ATOMIC_WRITE:
2336 return f2fs_ioc_start_atomic_write(filp);
2337 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2338 return f2fs_ioc_commit_atomic_write(filp);
2339 case F2FS_IOC_START_VOLATILE_WRITE:
2340 return f2fs_ioc_start_volatile_write(filp);
2341 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2342 return f2fs_ioc_release_volatile_write(filp);
2343 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2344 return f2fs_ioc_abort_volatile_write(filp);
2345 case F2FS_IOC_SHUTDOWN:
2346 return f2fs_ioc_shutdown(filp, arg);
2347 case FITRIM:
2348 return f2fs_ioc_fitrim(filp, arg);
2349 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2350 return f2fs_ioc_set_encryption_policy(filp, arg);
2351 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2352 return f2fs_ioc_get_encryption_policy(filp, arg);
2353 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2354 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2355 case F2FS_IOC_GARBAGE_COLLECT:
2356 return f2fs_ioc_gc(filp, arg);
2357 case F2FS_IOC_WRITE_CHECKPOINT:
2358 return f2fs_ioc_write_checkpoint(filp, arg);
2359 case F2FS_IOC_DEFRAGMENT:
2360 return f2fs_ioc_defragment(filp, arg);
2361 case F2FS_IOC_MOVE_RANGE:
2362 return f2fs_ioc_move_range(filp, arg);
2363 case F2FS_IOC_FLUSH_DEVICE:
2364 return f2fs_ioc_flush_device(filp, arg);
2365 default:
2366 return -ENOTTY;
2367 }
2368 }
2369
2370 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2371 {
2372 struct file *file = iocb->ki_filp;
2373 struct inode *inode = file_inode(file);
2374 struct blk_plug plug;
2375 ssize_t ret;
2376
2377 inode_lock(inode);
2378 ret = generic_write_checks(iocb, from);
2379 if (ret > 0) {
2380 int err;
2381
2382 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2383 set_inode_flag(inode, FI_NO_PREALLOC);
2384
2385 err = f2fs_preallocate_blocks(iocb, from);
2386 if (err) {
2387 inode_unlock(inode);
2388 return err;
2389 }
2390 blk_start_plug(&plug);
2391 ret = __generic_file_write_iter(iocb, from);
2392 blk_finish_plug(&plug);
2393 clear_inode_flag(inode, FI_NO_PREALLOC);
2394 }
2395 inode_unlock(inode);
2396
2397 if (ret > 0)
2398 ret = generic_write_sync(iocb, ret);
2399 return ret;
2400 }
2401
2402 #ifdef CONFIG_COMPAT
2403 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2404 {
2405 switch (cmd) {
2406 case F2FS_IOC32_GETFLAGS:
2407 cmd = F2FS_IOC_GETFLAGS;
2408 break;
2409 case F2FS_IOC32_SETFLAGS:
2410 cmd = F2FS_IOC_SETFLAGS;
2411 break;
2412 case F2FS_IOC32_GETVERSION:
2413 cmd = F2FS_IOC_GETVERSION;
2414 break;
2415 case F2FS_IOC_START_ATOMIC_WRITE:
2416 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2417 case F2FS_IOC_START_VOLATILE_WRITE:
2418 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2419 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2420 case F2FS_IOC_SHUTDOWN:
2421 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2422 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2423 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2424 case F2FS_IOC_GARBAGE_COLLECT:
2425 case F2FS_IOC_WRITE_CHECKPOINT:
2426 case F2FS_IOC_DEFRAGMENT:
2427 case F2FS_IOC_MOVE_RANGE:
2428 case F2FS_IOC_FLUSH_DEVICE:
2429 break;
2430 default:
2431 return -ENOIOCTLCMD;
2432 }
2433 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2434 }
2435 #endif
2436
2437 const struct file_operations f2fs_file_operations = {
2438 .llseek = f2fs_llseek,
2439 .read_iter = generic_file_read_iter,
2440 .write_iter = f2fs_file_write_iter,
2441 .open = f2fs_file_open,
2442 .release = f2fs_release_file,
2443 .mmap = f2fs_file_mmap,
2444 .fsync = f2fs_sync_file,
2445 .fallocate = f2fs_fallocate,
2446 .unlocked_ioctl = f2fs_ioctl,
2447 #ifdef CONFIG_COMPAT
2448 .compat_ioctl = f2fs_compat_ioctl,
2449 #endif
2450 .splice_read = generic_file_splice_read,
2451 .splice_write = iter_file_splice_write,
2452 };
Ubuntu Artful kernel mirror