projects / GitHub / exynos8895 / android_kernel_samsung_universal8895.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: catch up to v4.14-rc1
[GitHub/exynos8895/android_kernel_samsung_universal8895.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/random.h>
24 #include <linux/uio.h>
25 #include <linux/uuid.h>
26 #include <linux/file.h>
27
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "acl.h"
33 #include "gc.h"
34 #include "trace.h"
35 #include <trace/events/f2fs.h>
36
37 static int f2fs_filemap_fault(struct vm_area_struct *vma,
38 struct vm_fault *vmf)
39 {
40 struct inode *inode = file_inode(vma->vm_file);
41 int err;
42
43 down_read(&F2FS_I(inode)->i_mmap_sem);
44 err = filemap_fault(vma, vmf);
45 up_read(&F2FS_I(inode)->i_mmap_sem);
46
47 return err;
48 }
49
50 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
51 struct vm_fault *vmf)
52 {
53 struct page *page = vmf->page;
54 struct inode *inode = file_inode(vma->vm_file);
55 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
56 struct dnode_of_data dn;
57 int err;
58
59 sb_start_pagefault(inode->i_sb);
60
61 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
62
63 /* block allocation */
64 f2fs_lock_op(sbi);
65 set_new_dnode(&dn, inode, NULL, NULL, 0);
66 err = f2fs_reserve_block(&dn, page->index);
67 if (err) {
68 f2fs_unlock_op(sbi);
69 goto out;
70 }
71 f2fs_put_dnode(&dn);
72 f2fs_unlock_op(sbi);
73
74 f2fs_balance_fs(sbi, dn.node_changed);
75
76 file_update_time(vma->vm_file);
77 down_read(&F2FS_I(inode)->i_mmap_sem);
78 lock_page(page);
79 if (unlikely(page->mapping != inode->i_mapping ||
80 page_offset(page) > i_size_read(inode) ||
81 !PageUptodate(page))) {
82 unlock_page(page);
83 err = -EFAULT;
84 goto out_sem;
85 }
86
87 /*
88 * check to see if the page is mapped already (no holes)
89 */
90 if (PageMappedToDisk(page))
91 goto mapped;
92
93 /* page is wholly or partially inside EOF */
94 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
95 i_size_read(inode)) {
96 unsigned offset;
97 offset = i_size_read(inode) & ~PAGE_MASK;
98 zero_user_segment(page, offset, PAGE_SIZE);
99 }
100 set_page_dirty(page);
101 if (!PageUptodate(page))
102 SetPageUptodate(page);
103
104 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
105
106 trace_f2fs_vm_page_mkwrite(page, DATA);
107 mapped:
108 /* fill the page */
109 f2fs_wait_on_page_writeback(page, DATA, false);
110
111 /* wait for GCed encrypted page writeback */
112 if (f2fs_encrypted_file(inode))
113 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
114
115 out_sem:
116 up_read(&F2FS_I(inode)->i_mmap_sem);
117 out:
118 sb_end_pagefault(inode->i_sb);
119 f2fs_update_time(sbi, REQ_TIME);
120 return block_page_mkwrite_return(err);
121 }
122
123 static const struct vm_operations_struct f2fs_file_vm_ops = {
124 .fault = f2fs_filemap_fault,
125 .map_pages = filemap_map_pages,
126 .page_mkwrite = f2fs_vm_page_mkwrite,
127 };
128
129 static int get_parent_ino(struct inode *inode, nid_t *pino)
130 {
131 struct dentry *dentry;
132
133 inode = igrab(inode);
134 dentry = d_find_any_alias(inode);
135 iput(inode);
136 if (!dentry)
137 return 0;
138
139 *pino = parent_ino(dentry);
140 dput(dentry);
141 return 1;
142 }
143
144 static inline bool need_do_checkpoint(struct inode *inode)
145 {
146 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
147 bool need_cp = false;
148
149 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
150 need_cp = true;
151 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
152 need_cp = true;
153 else if (file_wrong_pino(inode))
154 need_cp = true;
155 else if (!space_for_roll_forward(sbi))
156 need_cp = true;
157 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
158 need_cp = true;
159 else if (test_opt(sbi, FASTBOOT))
160 need_cp = true;
161 else if (sbi->active_logs == 2)
162 need_cp = true;
163
164 return need_cp;
165 }
166
167 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
168 {
169 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
170 bool ret = false;
171 /* But we need to avoid that there are some inode updates */
172 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
173 ret = true;
174 f2fs_put_page(i, 0);
175 return ret;
176 }
177
178 static void try_to_fix_pino(struct inode *inode)
179 {
180 struct f2fs_inode_info *fi = F2FS_I(inode);
181 nid_t pino;
182
183 down_write(&fi->i_sem);
184 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
185 get_parent_ino(inode, &pino)) {
186 f2fs_i_pino_write(inode, pino);
187 file_got_pino(inode);
188 }
189 up_write(&fi->i_sem);
190 }
191
192 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
193 int datasync, bool atomic)
194 {
195 struct inode *inode = file->f_mapping->host;
196 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
197 nid_t ino = inode->i_ino;
198 int ret = 0;
199 bool need_cp = false;
200 struct writeback_control wbc = {
201 .sync_mode = WB_SYNC_ALL,
202 .nr_to_write = LONG_MAX,
203 .for_reclaim = 0,
204 };
205
206 if (unlikely(f2fs_readonly(inode->i_sb)))
207 return 0;
208
209 trace_f2fs_sync_file_enter(inode);
210
211 /* if fdatasync is triggered, let's do in-place-update */
212 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
213 set_inode_flag(inode, FI_NEED_IPU);
214 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
215 clear_inode_flag(inode, FI_NEED_IPU);
216
217 if (ret) {
218 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
219 return ret;
220 }
221
222 /* if the inode is dirty, let's recover all the time */
223 if (!f2fs_skip_inode_update(inode, datasync)) {
224 f2fs_write_inode(inode, NULL);
225 goto go_write;
226 }
227
228 /*
229 * if there is no written data, don't waste time to write recovery info.
230 */
231 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
232 !exist_written_data(sbi, ino, APPEND_INO)) {
233
234 /* it may call write_inode just prior to fsync */
235 if (need_inode_page_update(sbi, ino))
236 goto go_write;
237
238 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
239 exist_written_data(sbi, ino, UPDATE_INO))
240 goto flush_out;
241 goto out;
242 }
243 go_write:
244 /*
245 * Both of fdatasync() and fsync() are able to be recovered from
246 * sudden-power-off.
247 */
248 down_read(&F2FS_I(inode)->i_sem);
249 need_cp = need_do_checkpoint(inode);
250 up_read(&F2FS_I(inode)->i_sem);
251
252 if (need_cp) {
253 /* all the dirty node pages should be flushed for POR */
254 ret = f2fs_sync_fs(inode->i_sb, 1);
255
256 /*
257 * We've secured consistency through sync_fs. Following pino
258 * will be used only for fsynced inodes after checkpoint.
259 */
260 try_to_fix_pino(inode);
261 clear_inode_flag(inode, FI_APPEND_WRITE);
262 clear_inode_flag(inode, FI_UPDATE_WRITE);
263 goto out;
264 }
265 sync_nodes:
266 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
267 if (ret)
268 goto out;
269
270 /* if cp_error was enabled, we should avoid infinite loop */
271 if (unlikely(f2fs_cp_error(sbi))) {
272 ret = -EIO;
273 goto out;
274 }
275
276 if (need_inode_block_update(sbi, ino)) {
277 f2fs_mark_inode_dirty_sync(inode, true);
278 f2fs_write_inode(inode, NULL);
279 goto sync_nodes;
280 }
281
282 /*
283 * If it's atomic_write, it's just fine to keep write ordering. So
284 * here we don't need to wait for node write completion, since we use
285 * node chain which serializes node blocks. If one of node writes are
286 * reordered, we can see simply broken chain, resulting in stopping
287 * roll-forward recovery. It means we'll recover all or none node blocks
288 * given fsync mark.
289 */
290 if (!atomic) {
291 ret = wait_on_node_pages_writeback(sbi, ino);
292 if (ret)
293 goto out;
294 }
295
296 /* once recovery info is written, don't need to tack this */
297 remove_ino_entry(sbi, ino, APPEND_INO);
298 clear_inode_flag(inode, FI_APPEND_WRITE);
299 flush_out:
300 remove_ino_entry(sbi, ino, UPDATE_INO);
301 clear_inode_flag(inode, FI_UPDATE_WRITE);
302 if (!atomic)
303 ret = f2fs_issue_flush(sbi);
304 f2fs_update_time(sbi, REQ_TIME);
305 out:
306 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
307 f2fs_trace_ios(NULL, 1);
308 return ret;
309 }
310
311 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
312 {
313 return f2fs_do_sync_file(file, start, end, datasync, false);
314 }
315
316 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
317 pgoff_t pgofs, int whence)
318 {
319 struct pagevec pvec;
320 int nr_pages;
321
322 if (whence != SEEK_DATA)
323 return 0;
324
325 /* find first dirty page index */
326 pagevec_init(&pvec, 0);
327 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
328 PAGECACHE_TAG_DIRTY, 1);
329 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
330 pagevec_release(&pvec);
331 return pgofs;
332 }
333
334 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
335 int whence)
336 {
337 switch (whence) {
338 case SEEK_DATA:
339 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
340 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
341 return true;
342 break;
343 case SEEK_HOLE:
344 if (blkaddr == NULL_ADDR)
345 return true;
346 break;
347 }
348 return false;
349 }
350
351 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
352 {
353 struct inode *inode = file->f_mapping->host;
354 loff_t maxbytes = inode->i_sb->s_maxbytes;
355 struct dnode_of_data dn;
356 pgoff_t pgofs, end_offset, dirty;
357 loff_t data_ofs = offset;
358 loff_t isize;
359 int err = 0;
360
361 inode_lock(inode);
362
363 isize = i_size_read(inode);
364 if (offset >= isize)
365 goto fail;
366
367 /* handle inline data case */
368 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
369 if (whence == SEEK_HOLE)
370 data_ofs = isize;
371 goto found;
372 }
373
374 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
375
376 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
377
378 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
379 set_new_dnode(&dn, inode, NULL, NULL, 0);
380 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
381 if (err && err != -ENOENT) {
382 goto fail;
383 } else if (err == -ENOENT) {
384 /* direct node does not exists */
385 if (whence == SEEK_DATA) {
386 pgofs = get_next_page_offset(&dn, pgofs);
387 continue;
388 } else {
389 goto found;
390 }
391 }
392
393 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
394
395 /* find data/hole in dnode block */
396 for (; dn.ofs_in_node < end_offset;
397 dn.ofs_in_node++, pgofs++,
398 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
399 block_t blkaddr;
400 blkaddr = datablock_addr(dn.inode,
401 dn.node_page, dn.ofs_in_node);
402
403 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
404 f2fs_put_dnode(&dn);
405 goto found;
406 }
407 }
408 f2fs_put_dnode(&dn);
409 }
410
411 if (whence == SEEK_DATA)
412 goto fail;
413 found:
414 if (whence == SEEK_HOLE && data_ofs > isize)
415 data_ofs = isize;
416 inode_unlock(inode);
417 return vfs_setpos(file, data_ofs, maxbytes);
418 fail:
419 inode_unlock(inode);
420 return -ENXIO;
421 }
422
423 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
424 {
425 struct inode *inode = file->f_mapping->host;
426 loff_t maxbytes = inode->i_sb->s_maxbytes;
427
428 switch (whence) {
429 case SEEK_SET:
430 case SEEK_CUR:
431 case SEEK_END:
432 return generic_file_llseek_size(file, offset, whence,
433 maxbytes, i_size_read(inode));
434 case SEEK_DATA:
435 case SEEK_HOLE:
436 if (offset < 0)
437 return -ENXIO;
438 return f2fs_seek_block(file, offset, whence);
439 }
440
441 return -EINVAL;
442 }
443
444 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
445 {
446 struct inode *inode = file_inode(file);
447 int err;
448
449 /* we don't need to use inline_data strictly */
450 err = f2fs_convert_inline_inode(inode);
451 if (err)
452 return err;
453
454 file_accessed(file);
455 vma->vm_ops = &f2fs_file_vm_ops;
456 return 0;
457 }
458
459 static int f2fs_file_open(struct inode *inode, struct file *filp)
460 {
461 struct dentry *dir;
462
463 if (f2fs_encrypted_inode(inode)) {
464 int ret = fscrypt_get_encryption_info(inode);
465 if (ret)
466 return -EACCES;
467 if (!fscrypt_has_encryption_key(inode))
468 return -ENOKEY;
469 }
470 dir = dget_parent(file_dentry(filp));
471 if (f2fs_encrypted_inode(d_inode(dir)) &&
472 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
473 dput(dir);
474 return -EPERM;
475 }
476 dput(dir);
477 return dquot_file_open(inode, filp);
478 }
479
480 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
481 {
482 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
483 struct f2fs_node *raw_node;
484 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
485 __le32 *addr;
486 int base = 0;
487
488 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
489 base = get_extra_isize(dn->inode);
490
491 raw_node = F2FS_NODE(dn->node_page);
492 addr = blkaddr_in_node(raw_node) + base + ofs;
493
494 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
495 block_t blkaddr = le32_to_cpu(*addr);
496 if (blkaddr == NULL_ADDR)
497 continue;
498
499 dn->data_blkaddr = NULL_ADDR;
500 set_data_blkaddr(dn);
501 invalidate_blocks(sbi, blkaddr);
502 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
503 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
504 nr_free++;
505 }
506
507 if (nr_free) {
508 pgoff_t fofs;
509 /*
510 * once we invalidate valid blkaddr in range [ofs, ofs + count],
511 * we will invalidate all blkaddr in the whole range.
512 */
513 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
514 dn->inode) + ofs;
515 f2fs_update_extent_cache_range(dn, fofs, 0, len);
516 dec_valid_block_count(sbi, dn->inode, nr_free);
517 }
518 dn->ofs_in_node = ofs;
519
520 f2fs_update_time(sbi, REQ_TIME);
521 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
522 dn->ofs_in_node, nr_free);
523 return nr_free;
524 }
525
526 void truncate_data_blocks(struct dnode_of_data *dn)
527 {
528 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
529 }
530
531 static int truncate_partial_data_page(struct inode *inode, u64 from,
532 bool cache_only)
533 {
534 unsigned offset = from & (PAGE_SIZE - 1);
535 pgoff_t index = from >> PAGE_SHIFT;
536 struct address_space *mapping = inode->i_mapping;
537 struct page *page;
538
539 if (!offset && !cache_only)
540 return 0;
541
542 if (cache_only) {
543 page = find_lock_page(mapping, index);
544 if (page && PageUptodate(page))
545 goto truncate_out;
546 f2fs_put_page(page, 1);
547 return 0;
548 }
549
550 page = get_lock_data_page(inode, index, true);
551 if (IS_ERR(page))
552 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
553 truncate_out:
554 f2fs_wait_on_page_writeback(page, DATA, true);
555 zero_user(page, offset, PAGE_SIZE - offset);
556
557 /* An encrypted inode should have a key and truncate the last page. */
558 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
559 if (!cache_only)
560 set_page_dirty(page);
561 f2fs_put_page(page, 1);
562 return 0;
563 }
564
565 int truncate_blocks(struct inode *inode, u64 from, bool lock)
566 {
567 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
568 unsigned int blocksize = inode->i_sb->s_blocksize;
569 struct dnode_of_data dn;
570 pgoff_t free_from;
571 int count = 0, err = 0;
572 struct page *ipage;
573 bool truncate_page = false;
574
575 trace_f2fs_truncate_blocks_enter(inode, from);
576
577 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
578
579 if (free_from >= sbi->max_file_blocks)
580 goto free_partial;
581
582 if (lock)
583 f2fs_lock_op(sbi);
584
585 ipage = get_node_page(sbi, inode->i_ino);
586 if (IS_ERR(ipage)) {
587 err = PTR_ERR(ipage);
588 goto out;
589 }
590
591 if (f2fs_has_inline_data(inode)) {
592 truncate_inline_inode(inode, ipage, from);
593 f2fs_put_page(ipage, 1);
594 truncate_page = true;
595 goto out;
596 }
597
598 set_new_dnode(&dn, inode, ipage, NULL, 0);
599 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
600 if (err) {
601 if (err == -ENOENT)
602 goto free_next;
603 goto out;
604 }
605
606 count = ADDRS_PER_PAGE(dn.node_page, inode);
607
608 count -= dn.ofs_in_node;
609 f2fs_bug_on(sbi, count < 0);
610
611 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
612 truncate_data_blocks_range(&dn, count);
613 free_from += count;
614 }
615
616 f2fs_put_dnode(&dn);
617 free_next:
618 err = truncate_inode_blocks(inode, free_from);
619 out:
620 if (lock)
621 f2fs_unlock_op(sbi);
622 free_partial:
623 /* lastly zero out the first data page */
624 if (!err)
625 err = truncate_partial_data_page(inode, from, truncate_page);
626
627 trace_f2fs_truncate_blocks_exit(inode, err);
628 return err;
629 }
630
631 int f2fs_truncate(struct inode *inode)
632 {
633 int err;
634
635 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
636 S_ISLNK(inode->i_mode)))
637 return 0;
638
639 trace_f2fs_truncate(inode);
640
641 #ifdef CONFIG_F2FS_FAULT_INJECTION
642 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
643 f2fs_show_injection_info(FAULT_TRUNCATE);
644 return -EIO;
645 }
646 #endif
647 /* we should check inline_data size */
648 if (!f2fs_may_inline_data(inode)) {
649 err = f2fs_convert_inline_inode(inode);
650 if (err)
651 return err;
652 }
653
654 err = truncate_blocks(inode, i_size_read(inode), true);
655 if (err)
656 return err;
657
658 inode->i_mtime = inode->i_ctime = current_time(inode);
659 f2fs_mark_inode_dirty_sync(inode, false);
660 return 0;
661 }
662
663 int f2fs_getattr(struct vfsmount *mnt,
664 struct dentry *dentry, struct kstat *stat)
665 {
666 struct inode *inode = d_inode(dentry);
667 generic_fillattr(inode, stat);
668 return 0;
669 }
670
671 #ifdef CONFIG_F2FS_FS_POSIX_ACL
672 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
673 {
674 unsigned int ia_valid = attr->ia_valid;
675
676 if (ia_valid & ATTR_UID)
677 inode->i_uid = attr->ia_uid;
678 if (ia_valid & ATTR_GID)
679 inode->i_gid = attr->ia_gid;
680 if (ia_valid & ATTR_ATIME)
681 inode->i_atime = timespec_trunc(attr->ia_atime,
682 inode->i_sb->s_time_gran);
683 if (ia_valid & ATTR_MTIME)
684 inode->i_mtime = timespec_trunc(attr->ia_mtime,
685 inode->i_sb->s_time_gran);
686 if (ia_valid & ATTR_CTIME)
687 inode->i_ctime = timespec_trunc(attr->ia_ctime,
688 inode->i_sb->s_time_gran);
689 if (ia_valid & ATTR_MODE) {
690 umode_t mode = attr->ia_mode;
691
692 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
693 mode &= ~S_ISGID;
694 set_acl_inode(inode, mode);
695 }
696 }
697 #else
698 #define __setattr_copy setattr_copy
699 #endif
700
701 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
702 {
703 struct inode *inode = d_inode(dentry);
704 int err;
705 bool size_changed = false;
706
707 err = inode_change_ok(inode, attr);
708 if (err)
709 return err;
710
711 if (is_quota_modification(inode, attr)) {
712 err = dquot_initialize(inode);
713 if (err)
714 return err;
715 }
716 if ((attr->ia_valid & ATTR_UID &&
717 !uid_eq(attr->ia_uid, inode->i_uid)) ||
718 (attr->ia_valid & ATTR_GID &&
719 !gid_eq(attr->ia_gid, inode->i_gid))) {
720 err = dquot_transfer(inode, attr);
721 if (err)
722 return err;
723 }
724
725 if (attr->ia_valid & ATTR_SIZE) {
726 if (f2fs_encrypted_inode(inode)) {
727 err = fscrypt_get_encryption_info(inode);
728 if (err)
729 return err;
730 if (!fscrypt_has_encryption_key(inode))
731 return -ENOKEY;
732 }
733
734 if (attr->ia_size <= i_size_read(inode)) {
735 down_write(&F2FS_I(inode)->i_mmap_sem);
736 truncate_setsize(inode, attr->ia_size);
737 err = f2fs_truncate(inode);
738 up_write(&F2FS_I(inode)->i_mmap_sem);
739 if (err)
740 return err;
741 } else {
742 /*
743 * do not trim all blocks after i_size if target size is
744 * larger than i_size.
745 */
746 down_write(&F2FS_I(inode)->i_mmap_sem);
747 truncate_setsize(inode, attr->ia_size);
748 up_write(&F2FS_I(inode)->i_mmap_sem);
749
750 /* should convert inline inode here */
751 if (!f2fs_may_inline_data(inode)) {
752 err = f2fs_convert_inline_inode(inode);
753 if (err)
754 return err;
755 }
756 inode->i_mtime = inode->i_ctime = current_time(inode);
757 }
758
759 size_changed = true;
760 }
761
762 __setattr_copy(inode, attr);
763
764 if (attr->ia_valid & ATTR_MODE) {
765 err = posix_acl_chmod(inode, get_inode_mode(inode));
766 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
767 inode->i_mode = F2FS_I(inode)->i_acl_mode;
768 clear_inode_flag(inode, FI_ACL_MODE);
769 }
770 }
771
772 /* file size may changed here */
773 f2fs_mark_inode_dirty_sync(inode, size_changed);
774
775 /* inode change will produce dirty node pages flushed by checkpoint */
776 f2fs_balance_fs(F2FS_I_SB(inode), true);
777
778 return err;
779 }
780
781 const struct inode_operations f2fs_file_inode_operations = {
782 .getattr = f2fs_getattr,
783 .setattr = f2fs_setattr,
784 .get_acl = f2fs_get_acl,
785 .set_acl = f2fs_set_acl,
786 #ifdef CONFIG_F2FS_FS_XATTR
787 .setxattr = generic_setxattr,
788 .getxattr = generic_getxattr,
789 .listxattr = f2fs_listxattr,
790 .removexattr = generic_removexattr,
791 #endif
792 .fiemap = f2fs_fiemap,
793 };
794
795 static int fill_zero(struct inode *inode, pgoff_t index,
796 loff_t start, loff_t len)
797 {
798 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
799 struct page *page;
800
801 if (!len)
802 return 0;
803
804 f2fs_balance_fs(sbi, true);
805
806 f2fs_lock_op(sbi);
807 page = get_new_data_page(inode, NULL, index, false);
808 f2fs_unlock_op(sbi);
809
810 if (IS_ERR(page))
811 return PTR_ERR(page);
812
813 f2fs_wait_on_page_writeback(page, DATA, true);
814 zero_user(page, start, len);
815 set_page_dirty(page);
816 f2fs_put_page(page, 1);
817 return 0;
818 }
819
820 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
821 {
822 int err;
823
824 while (pg_start < pg_end) {
825 struct dnode_of_data dn;
826 pgoff_t end_offset, count;
827
828 set_new_dnode(&dn, inode, NULL, NULL, 0);
829 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
830 if (err) {
831 if (err == -ENOENT) {
832 pg_start++;
833 continue;
834 }
835 return err;
836 }
837
838 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
839 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
840
841 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
842
843 truncate_data_blocks_range(&dn, count);
844 f2fs_put_dnode(&dn);
845
846 pg_start += count;
847 }
848 return 0;
849 }
850
851 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
852 {
853 pgoff_t pg_start, pg_end;
854 loff_t off_start, off_end;
855 int ret;
856
857 ret = f2fs_convert_inline_inode(inode);
858 if (ret)
859 return ret;
860
861 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
862 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
863
864 off_start = offset & (PAGE_SIZE - 1);
865 off_end = (offset + len) & (PAGE_SIZE - 1);
866
867 if (pg_start == pg_end) {
868 ret = fill_zero(inode, pg_start, off_start,
869 off_end - off_start);
870 if (ret)
871 return ret;
872 } else {
873 if (off_start) {
874 ret = fill_zero(inode, pg_start++, off_start,
875 PAGE_SIZE - off_start);
876 if (ret)
877 return ret;
878 }
879 if (off_end) {
880 ret = fill_zero(inode, pg_end, 0, off_end);
881 if (ret)
882 return ret;
883 }
884
885 if (pg_start < pg_end) {
886 struct address_space *mapping = inode->i_mapping;
887 loff_t blk_start, blk_end;
888 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
889
890 f2fs_balance_fs(sbi, true);
891
892 blk_start = (loff_t)pg_start << PAGE_SHIFT;
893 blk_end = (loff_t)pg_end << PAGE_SHIFT;
894 down_write(&F2FS_I(inode)->i_mmap_sem);
895 truncate_inode_pages_range(mapping, blk_start,
896 blk_end - 1);
897
898 f2fs_lock_op(sbi);
899 ret = truncate_hole(inode, pg_start, pg_end);
900 f2fs_unlock_op(sbi);
901 up_write(&F2FS_I(inode)->i_mmap_sem);
902 }
903 }
904
905 return ret;
906 }
907
908 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
909 int *do_replace, pgoff_t off, pgoff_t len)
910 {
911 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
912 struct dnode_of_data dn;
913 int ret, done, i;
914
915 next_dnode:
916 set_new_dnode(&dn, inode, NULL, NULL, 0);
917 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
918 if (ret && ret != -ENOENT) {
919 return ret;
920 } else if (ret == -ENOENT) {
921 if (dn.max_level == 0)
922 return -ENOENT;
923 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
924 blkaddr += done;
925 do_replace += done;
926 goto next;
927 }
928
929 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
930 dn.ofs_in_node, len);
931 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
932 *blkaddr = datablock_addr(dn.inode,
933 dn.node_page, dn.ofs_in_node);
934 if (!is_checkpointed_data(sbi, *blkaddr)) {
935
936 if (test_opt(sbi, LFS)) {
937 f2fs_put_dnode(&dn);
938 return -ENOTSUPP;
939 }
940
941 /* do not invalidate this block address */
942 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
943 *do_replace = 1;
944 }
945 }
946 f2fs_put_dnode(&dn);
947 next:
948 len -= done;
949 off += done;
950 if (len)
951 goto next_dnode;
952 return 0;
953 }
954
955 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
956 int *do_replace, pgoff_t off, int len)
957 {
958 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
959 struct dnode_of_data dn;
960 int ret, i;
961
962 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
963 if (*do_replace == 0)
964 continue;
965
966 set_new_dnode(&dn, inode, NULL, NULL, 0);
967 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
968 if (ret) {
969 dec_valid_block_count(sbi, inode, 1);
970 invalidate_blocks(sbi, *blkaddr);
971 } else {
972 f2fs_update_data_blkaddr(&dn, *blkaddr);
973 }
974 f2fs_put_dnode(&dn);
975 }
976 return 0;
977 }
978
979 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
980 block_t *blkaddr, int *do_replace,
981 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
982 {
983 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
984 pgoff_t i = 0;
985 int ret;
986
987 while (i < len) {
988 if (blkaddr[i] == NULL_ADDR && !full) {
989 i++;
990 continue;
991 }
992
993 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
994 struct dnode_of_data dn;
995 struct node_info ni;
996 size_t new_size;
997 pgoff_t ilen;
998
999 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1000 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1001 if (ret)
1002 return ret;
1003
1004 get_node_info(sbi, dn.nid, &ni);
1005 ilen = min((pgoff_t)
1006 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1007 dn.ofs_in_node, len - i);
1008 do {
1009 dn.data_blkaddr = datablock_addr(dn.inode,
1010 dn.node_page, dn.ofs_in_node);
1011 truncate_data_blocks_range(&dn, 1);
1012
1013 if (do_replace[i]) {
1014 f2fs_i_blocks_write(src_inode,
1015 1, false, false);
1016 f2fs_i_blocks_write(dst_inode,
1017 1, true, false);
1018 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1019 blkaddr[i], ni.version, true, false);
1020
1021 do_replace[i] = 0;
1022 }
1023 dn.ofs_in_node++;
1024 i++;
1025 new_size = (dst + i) << PAGE_SHIFT;
1026 if (dst_inode->i_size < new_size)
1027 f2fs_i_size_write(dst_inode, new_size);
1028 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1029
1030 f2fs_put_dnode(&dn);
1031 } else {
1032 struct page *psrc, *pdst;
1033
1034 psrc = get_lock_data_page(src_inode, src + i, true);
1035 if (IS_ERR(psrc))
1036 return PTR_ERR(psrc);
1037 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1038 true);
1039 if (IS_ERR(pdst)) {
1040 f2fs_put_page(psrc, 1);
1041 return PTR_ERR(pdst);
1042 }
1043 f2fs_copy_page(psrc, pdst);
1044 set_page_dirty(pdst);
1045 f2fs_put_page(pdst, 1);
1046 f2fs_put_page(psrc, 1);
1047
1048 ret = truncate_hole(src_inode, src + i, src + i + 1);
1049 if (ret)
1050 return ret;
1051 i++;
1052 }
1053 }
1054 return 0;
1055 }
1056
1057 static int __exchange_data_block(struct inode *src_inode,
1058 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1059 pgoff_t len, bool full)
1060 {
1061 block_t *src_blkaddr;
1062 int *do_replace;
1063 pgoff_t olen;
1064 int ret;
1065
1066 while (len) {
1067 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1068
1069 src_blkaddr = kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1070 if (!src_blkaddr)
1071 return -ENOMEM;
1072
1073 do_replace = kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1074 if (!do_replace) {
1075 kvfree(src_blkaddr);
1076 return -ENOMEM;
1077 }
1078
1079 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1080 do_replace, src, olen);
1081 if (ret)
1082 goto roll_back;
1083
1084 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1085 do_replace, src, dst, olen, full);
1086 if (ret)
1087 goto roll_back;
1088
1089 src += olen;
1090 dst += olen;
1091 len -= olen;
1092
1093 kvfree(src_blkaddr);
1094 kvfree(do_replace);
1095 }
1096 return 0;
1097
1098 roll_back:
1099 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1100 kvfree(src_blkaddr);
1101 kvfree(do_replace);
1102 return ret;
1103 }
1104
1105 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1106 {
1107 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1108 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1109 int ret;
1110
1111 f2fs_balance_fs(sbi, true);
1112 f2fs_lock_op(sbi);
1113
1114 f2fs_drop_extent_tree(inode);
1115
1116 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1117 f2fs_unlock_op(sbi);
1118 return ret;
1119 }
1120
1121 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1122 {
1123 pgoff_t pg_start, pg_end;
1124 loff_t new_size;
1125 int ret;
1126
1127 if (offset + len >= i_size_read(inode))
1128 return -EINVAL;
1129
1130 /* collapse range should be aligned to block size of f2fs. */
1131 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1132 return -EINVAL;
1133
1134 ret = f2fs_convert_inline_inode(inode);
1135 if (ret)
1136 return ret;
1137
1138 pg_start = offset >> PAGE_SHIFT;
1139 pg_end = (offset + len) >> PAGE_SHIFT;
1140
1141 down_write(&F2FS_I(inode)->i_mmap_sem);
1142 /* write out all dirty pages from offset */
1143 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1144 if (ret)
1145 goto out;
1146
1147 truncate_pagecache(inode, offset);
1148
1149 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1150 if (ret)
1151 goto out;
1152
1153 /* write out all moved pages, if possible */
1154 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1155 truncate_pagecache(inode, offset);
1156
1157 new_size = i_size_read(inode) - len;
1158 truncate_pagecache(inode, new_size);
1159
1160 ret = truncate_blocks(inode, new_size, true);
1161 if (!ret)
1162 f2fs_i_size_write(inode, new_size);
1163
1164 out:
1165 up_write(&F2FS_I(inode)->i_mmap_sem);
1166 return ret;
1167 }
1168
1169 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1170 pgoff_t end)
1171 {
1172 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1173 pgoff_t index = start;
1174 unsigned int ofs_in_node = dn->ofs_in_node;
1175 blkcnt_t count = 0;
1176 int ret;
1177
1178 for (; index < end; index++, dn->ofs_in_node++) {
1179 if (datablock_addr(dn->inode, dn->node_page,
1180 dn->ofs_in_node) == NULL_ADDR)
1181 count++;
1182 }
1183
1184 dn->ofs_in_node = ofs_in_node;
1185 ret = reserve_new_blocks(dn, count);
1186 if (ret)
1187 return ret;
1188
1189 dn->ofs_in_node = ofs_in_node;
1190 for (index = start; index < end; index++, dn->ofs_in_node++) {
1191 dn->data_blkaddr = datablock_addr(dn->inode,
1192 dn->node_page, dn->ofs_in_node);
1193 /*
1194 * reserve_new_blocks will not guarantee entire block
1195 * allocation.
1196 */
1197 if (dn->data_blkaddr == NULL_ADDR) {
1198 ret = -ENOSPC;
1199 break;
1200 }
1201 if (dn->data_blkaddr != NEW_ADDR) {
1202 invalidate_blocks(sbi, dn->data_blkaddr);
1203 dn->data_blkaddr = NEW_ADDR;
1204 set_data_blkaddr(dn);
1205 }
1206 }
1207
1208 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1209
1210 return ret;
1211 }
1212
1213 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1214 int mode)
1215 {
1216 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1217 struct address_space *mapping = inode->i_mapping;
1218 pgoff_t index, pg_start, pg_end;
1219 loff_t new_size = i_size_read(inode);
1220 loff_t off_start, off_end;
1221 int ret = 0;
1222
1223 ret = inode_newsize_ok(inode, (len + offset));
1224 if (ret)
1225 return ret;
1226
1227 ret = f2fs_convert_inline_inode(inode);
1228 if (ret)
1229 return ret;
1230
1231 down_write(&F2FS_I(inode)->i_mmap_sem);
1232 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1233 if (ret)
1234 goto out_sem;
1235
1236 truncate_pagecache_range(inode, offset, offset + len - 1);
1237
1238 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1239 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1240
1241 off_start = offset & (PAGE_SIZE - 1);
1242 off_end = (offset + len) & (PAGE_SIZE - 1);
1243
1244 if (pg_start == pg_end) {
1245 ret = fill_zero(inode, pg_start, off_start,
1246 off_end - off_start);
1247 if (ret)
1248 goto out_sem;
1249
1250 new_size = max_t(loff_t, new_size, offset + len);
1251 } else {
1252 if (off_start) {
1253 ret = fill_zero(inode, pg_start++, off_start,
1254 PAGE_SIZE - off_start);
1255 if (ret)
1256 goto out_sem;
1257
1258 new_size = max_t(loff_t, new_size,
1259 (loff_t)pg_start << PAGE_SHIFT);
1260 }
1261
1262 for (index = pg_start; index < pg_end;) {
1263 struct dnode_of_data dn;
1264 unsigned int end_offset;
1265 pgoff_t end;
1266
1267 f2fs_lock_op(sbi);
1268
1269 set_new_dnode(&dn, inode, NULL, NULL, 0);
1270 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1271 if (ret) {
1272 f2fs_unlock_op(sbi);
1273 goto out;
1274 }
1275
1276 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1277 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1278
1279 ret = f2fs_do_zero_range(&dn, index, end);
1280 f2fs_put_dnode(&dn);
1281 f2fs_unlock_op(sbi);
1282
1283 f2fs_balance_fs(sbi, dn.node_changed);
1284
1285 if (ret)
1286 goto out;
1287
1288 index = end;
1289 new_size = max_t(loff_t, new_size,
1290 (loff_t)index << PAGE_SHIFT);
1291 }
1292
1293 if (off_end) {
1294 ret = fill_zero(inode, pg_end, 0, off_end);
1295 if (ret)
1296 goto out;
1297
1298 new_size = max_t(loff_t, new_size, offset + len);
1299 }
1300 }
1301
1302 out:
1303 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1304 f2fs_i_size_write(inode, new_size);
1305 out_sem:
1306 up_write(&F2FS_I(inode)->i_mmap_sem);
1307
1308 return ret;
1309 }
1310
1311 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1312 {
1313 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1314 pgoff_t nr, pg_start, pg_end, delta, idx;
1315 loff_t new_size;
1316 int ret = 0;
1317
1318 new_size = i_size_read(inode) + len;
1319 ret = inode_newsize_ok(inode, new_size);
1320 if (ret)
1321 return ret;
1322
1323 if (offset >= i_size_read(inode))
1324 return -EINVAL;
1325
1326 /* insert range should be aligned to block size of f2fs. */
1327 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1328 return -EINVAL;
1329
1330 ret = f2fs_convert_inline_inode(inode);
1331 if (ret)
1332 return ret;
1333
1334 f2fs_balance_fs(sbi, true);
1335
1336 down_write(&F2FS_I(inode)->i_mmap_sem);
1337 ret = truncate_blocks(inode, i_size_read(inode), true);
1338 if (ret)
1339 goto out;
1340
1341 /* write out all dirty pages from offset */
1342 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1343 if (ret)
1344 goto out;
1345
1346 truncate_pagecache(inode, offset);
1347
1348 pg_start = offset >> PAGE_SHIFT;
1349 pg_end = (offset + len) >> PAGE_SHIFT;
1350 delta = pg_end - pg_start;
1351 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1352
1353 while (!ret && idx > pg_start) {
1354 nr = idx - pg_start;
1355 if (nr > delta)
1356 nr = delta;
1357 idx -= nr;
1358
1359 f2fs_lock_op(sbi);
1360 f2fs_drop_extent_tree(inode);
1361
1362 ret = __exchange_data_block(inode, inode, idx,
1363 idx + delta, nr, false);
1364 f2fs_unlock_op(sbi);
1365 }
1366
1367 /* write out all moved pages, if possible */
1368 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1369 truncate_pagecache(inode, offset);
1370
1371 if (!ret)
1372 f2fs_i_size_write(inode, new_size);
1373 out:
1374 up_write(&F2FS_I(inode)->i_mmap_sem);
1375 return ret;
1376 }
1377
1378 static int expand_inode_data(struct inode *inode, loff_t offset,
1379 loff_t len, int mode)
1380 {
1381 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1382 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1383 pgoff_t pg_end;
1384 loff_t new_size = i_size_read(inode);
1385 loff_t off_end;
1386 int err;
1387
1388 err = inode_newsize_ok(inode, (len + offset));
1389 if (err)
1390 return err;
1391
1392 err = f2fs_convert_inline_inode(inode);
1393 if (err)
1394 return err;
1395
1396 f2fs_balance_fs(sbi, true);
1397
1398 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1399 off_end = (offset + len) & (PAGE_SIZE - 1);
1400
1401 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1402 map.m_len = pg_end - map.m_lblk;
1403 if (off_end)
1404 map.m_len++;
1405
1406 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1407 if (err) {
1408 pgoff_t last_off;
1409
1410 if (!map.m_len)
1411 return err;
1412
1413 last_off = map.m_lblk + map.m_len - 1;
1414
1415 /* update new size to the failed position */
1416 new_size = (last_off == pg_end) ? offset + len:
1417 (loff_t)(last_off + 1) << PAGE_SHIFT;
1418 } else {
1419 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1420 }
1421
1422 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1423 f2fs_i_size_write(inode, new_size);
1424
1425 return err;
1426 }
1427
1428 static long f2fs_fallocate(struct file *file, int mode,
1429 loff_t offset, loff_t len)
1430 {
1431 struct inode *inode = file_inode(file);
1432 long ret = 0;
1433
1434 /* f2fs only support ->fallocate for regular file */
1435 if (!S_ISREG(inode->i_mode))
1436 return -EINVAL;
1437
1438 if (f2fs_encrypted_inode(inode) &&
1439 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1440 return -EOPNOTSUPP;
1441
1442 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1443 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1444 FALLOC_FL_INSERT_RANGE))
1445 return -EOPNOTSUPP;
1446
1447 inode_lock(inode);
1448
1449 if (mode & FALLOC_FL_PUNCH_HOLE) {
1450 if (offset >= inode->i_size)
1451 goto out;
1452
1453 ret = punch_hole(inode, offset, len);
1454 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1455 ret = f2fs_collapse_range(inode, offset, len);
1456 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1457 ret = f2fs_zero_range(inode, offset, len, mode);
1458 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1459 ret = f2fs_insert_range(inode, offset, len);
1460 } else {
1461 ret = expand_inode_data(inode, offset, len, mode);
1462 }
1463
1464 if (!ret) {
1465 inode->i_mtime = inode->i_ctime = current_time(inode);
1466 f2fs_mark_inode_dirty_sync(inode, false);
1467 if (mode & FALLOC_FL_KEEP_SIZE)
1468 file_set_keep_isize(inode);
1469 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1470 }
1471
1472 out:
1473 inode_unlock(inode);
1474
1475 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1476 return ret;
1477 }
1478
1479 static int f2fs_release_file(struct inode *inode, struct file *filp)
1480 {
1481 /*
1482 * f2fs_relase_file is called at every close calls. So we should
1483 * not drop any inmemory pages by close called by other process.
1484 */
1485 if (!(filp->f_mode & FMODE_WRITE) ||
1486 atomic_read(&inode->i_writecount) != 1)
1487 return 0;
1488
1489 /* some remained atomic pages should discarded */
1490 if (f2fs_is_atomic_file(inode))
1491 drop_inmem_pages(inode);
1492 if (f2fs_is_volatile_file(inode)) {
1493 clear_inode_flag(inode, FI_VOLATILE_FILE);
1494 stat_dec_volatile_write(inode);
1495 set_inode_flag(inode, FI_DROP_CACHE);
1496 filemap_fdatawrite(inode->i_mapping);
1497 clear_inode_flag(inode, FI_DROP_CACHE);
1498 }
1499 return 0;
1500 }
1501
1502 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1503 {
1504 struct inode *inode = file_inode(file);
1505
1506 /*
1507 * If the process doing a transaction is crashed, we should do
1508 * roll-back. Otherwise, other reader/write can see corrupted database
1509 * until all the writers close its file. Since this should be done
1510 * before dropping file lock, it needs to do in ->flush.
1511 */
1512 if (f2fs_is_atomic_file(inode) &&
1513 F2FS_I(inode)->inmem_task == current)
1514 drop_inmem_pages(inode);
1515 return 0;
1516 }
1517
1518 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1519 {
1520 struct inode *inode = file_inode(filp);
1521 struct f2fs_inode_info *fi = F2FS_I(inode);
1522 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1523 return put_user(flags, (int __user *)arg);
1524 }
1525
1526 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1527 {
1528 struct inode *inode = file_inode(filp);
1529 struct f2fs_inode_info *fi = F2FS_I(inode);
1530 unsigned int flags;
1531 unsigned int oldflags;
1532 int ret;
1533
1534 if (!inode_owner_or_capable(inode))
1535 return -EACCES;
1536
1537 if (get_user(flags, (int __user *)arg))
1538 return -EFAULT;
1539
1540 ret = mnt_want_write_file(filp);
1541 if (ret)
1542 return ret;
1543
1544 inode_lock(inode);
1545
1546 /* Is it quota file? Do not allow user to mess with it */
1547 if (IS_NOQUOTA(inode)) {
1548 ret = -EPERM;
1549 goto unlock_out;
1550 }
1551
1552 flags = f2fs_mask_flags(inode->i_mode, flags);
1553
1554 oldflags = fi->i_flags;
1555
1556 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1557 if (!capable(CAP_LINUX_IMMUTABLE)) {
1558 ret = -EPERM;
1559 goto unlock_out;
1560 }
1561 }
1562
1563 flags = flags & FS_FL_USER_MODIFIABLE;
1564 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1565 fi->i_flags = flags;
1566
1567 inode->i_ctime = current_time(inode);
1568 f2fs_set_inode_flags(inode);
1569 f2fs_mark_inode_dirty_sync(inode, false);
1570 unlock_out:
1571 inode_unlock(inode);
1572 mnt_drop_write_file(filp);
1573 return ret;
1574 }
1575
1576 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1577 {
1578 struct inode *inode = file_inode(filp);
1579
1580 return put_user(inode->i_generation, (int __user *)arg);
1581 }
1582
1583 static int f2fs_ioc_start_atomic_write(struct file *filp)
1584 {
1585 struct inode *inode = file_inode(filp);
1586 int ret;
1587
1588 if (!inode_owner_or_capable(inode))
1589 return -EACCES;
1590
1591 if (!S_ISREG(inode->i_mode))
1592 return -EINVAL;
1593
1594 ret = mnt_want_write_file(filp);
1595 if (ret)
1596 return ret;
1597
1598 inode_lock(inode);
1599
1600 if (f2fs_is_atomic_file(inode))
1601 goto out;
1602
1603 ret = f2fs_convert_inline_inode(inode);
1604 if (ret)
1605 goto out;
1606
1607 set_inode_flag(inode, FI_ATOMIC_FILE);
1608 set_inode_flag(inode, FI_HOT_DATA);
1609 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1610
1611 if (!get_dirty_pages(inode))
1612 goto inc_stat;
1613
1614 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1615 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1616 inode->i_ino, get_dirty_pages(inode));
1617 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1618 if (ret) {
1619 clear_inode_flag(inode, FI_ATOMIC_FILE);
1620 clear_inode_flag(inode, FI_HOT_DATA);
1621 goto out;
1622 }
1623
1624 inc_stat:
1625 F2FS_I(inode)->inmem_task = current;
1626 stat_inc_atomic_write(inode);
1627 stat_update_max_atomic_write(inode);
1628 out:
1629 inode_unlock(inode);
1630 mnt_drop_write_file(filp);
1631 return ret;
1632 }
1633
1634 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1635 {
1636 struct inode *inode = file_inode(filp);
1637 int ret;
1638
1639 if (!inode_owner_or_capable(inode))
1640 return -EACCES;
1641
1642 ret = mnt_want_write_file(filp);
1643 if (ret)
1644 return ret;
1645
1646 inode_lock(inode);
1647
1648 if (f2fs_is_volatile_file(inode))
1649 goto err_out;
1650
1651 if (f2fs_is_atomic_file(inode)) {
1652 ret = commit_inmem_pages(inode);
1653 if (ret)
1654 goto err_out;
1655
1656 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1657 if (!ret) {
1658 clear_inode_flag(inode, FI_ATOMIC_FILE);
1659 clear_inode_flag(inode, FI_HOT_DATA);
1660 stat_dec_atomic_write(inode);
1661 }
1662 } else {
1663 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1664 }
1665 err_out:
1666 inode_unlock(inode);
1667 mnt_drop_write_file(filp);
1668 return ret;
1669 }
1670
1671 static int f2fs_ioc_start_volatile_write(struct file *filp)
1672 {
1673 struct inode *inode = file_inode(filp);
1674 int ret;
1675
1676 if (!inode_owner_or_capable(inode))
1677 return -EACCES;
1678
1679 if (!S_ISREG(inode->i_mode))
1680 return -EINVAL;
1681
1682 ret = mnt_want_write_file(filp);
1683 if (ret)
1684 return ret;
1685
1686 inode_lock(inode);
1687
1688 if (f2fs_is_volatile_file(inode))
1689 goto out;
1690
1691 ret = f2fs_convert_inline_inode(inode);
1692 if (ret)
1693 goto out;
1694
1695 stat_inc_volatile_write(inode);
1696 stat_update_max_volatile_write(inode);
1697
1698 set_inode_flag(inode, FI_VOLATILE_FILE);
1699 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1700 out:
1701 inode_unlock(inode);
1702 mnt_drop_write_file(filp);
1703 return ret;
1704 }
1705
1706 static int f2fs_ioc_release_volatile_write(struct file *filp)
1707 {
1708 struct inode *inode = file_inode(filp);
1709 int ret;
1710
1711 if (!inode_owner_or_capable(inode))
1712 return -EACCES;
1713
1714 ret = mnt_want_write_file(filp);
1715 if (ret)
1716 return ret;
1717
1718 inode_lock(inode);
1719
1720 if (!f2fs_is_volatile_file(inode))
1721 goto out;
1722
1723 if (!f2fs_is_first_block_written(inode)) {
1724 ret = truncate_partial_data_page(inode, 0, true);
1725 goto out;
1726 }
1727
1728 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1729 out:
1730 inode_unlock(inode);
1731 mnt_drop_write_file(filp);
1732 return ret;
1733 }
1734
1735 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1736 {
1737 struct inode *inode = file_inode(filp);
1738 int ret;
1739
1740 if (!inode_owner_or_capable(inode))
1741 return -EACCES;
1742
1743 ret = mnt_want_write_file(filp);
1744 if (ret)
1745 return ret;
1746
1747 inode_lock(inode);
1748
1749 if (f2fs_is_atomic_file(inode))
1750 drop_inmem_pages(inode);
1751 if (f2fs_is_volatile_file(inode)) {
1752 clear_inode_flag(inode, FI_VOLATILE_FILE);
1753 stat_dec_volatile_write(inode);
1754 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1755 }
1756
1757 inode_unlock(inode);
1758
1759 mnt_drop_write_file(filp);
1760 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1761 return ret;
1762 }
1763
1764 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1765 {
1766 struct inode *inode = file_inode(filp);
1767 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1768 struct super_block *sb = sbi->sb;
1769 __u32 in;
1770 int ret;
1771
1772 if (!capable(CAP_SYS_ADMIN))
1773 return -EPERM;
1774
1775 if (get_user(in, (__u32 __user *)arg))
1776 return -EFAULT;
1777
1778 ret = mnt_want_write_file(filp);
1779 if (ret)
1780 return ret;
1781
1782 switch (in) {
1783 case F2FS_GOING_DOWN_FULLSYNC:
1784 sb = freeze_bdev(sb->s_bdev);
1785 if (sb && !IS_ERR(sb)) {
1786 f2fs_stop_checkpoint(sbi, false);
1787 thaw_bdev(sb->s_bdev, sb);
1788 }
1789 break;
1790 case F2FS_GOING_DOWN_METASYNC:
1791 /* do checkpoint only */
1792 f2fs_sync_fs(sb, 1);
1793 f2fs_stop_checkpoint(sbi, false);
1794 break;
1795 case F2FS_GOING_DOWN_NOSYNC:
1796 f2fs_stop_checkpoint(sbi, false);
1797 break;
1798 case F2FS_GOING_DOWN_METAFLUSH:
1799 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1800 f2fs_stop_checkpoint(sbi, false);
1801 break;
1802 default:
1803 ret = -EINVAL;
1804 goto out;
1805 }
1806 f2fs_update_time(sbi, REQ_TIME);
1807 out:
1808 mnt_drop_write_file(filp);
1809 return ret;
1810 }
1811
1812 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1813 {
1814 struct inode *inode = file_inode(filp);
1815 struct super_block *sb = inode->i_sb;
1816 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1817 struct fstrim_range range;
1818 int ret;
1819
1820 if (!capable(CAP_SYS_ADMIN))
1821 return -EPERM;
1822
1823 if (!blk_queue_discard(q))
1824 return -EOPNOTSUPP;
1825
1826 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1827 sizeof(range)))
1828 return -EFAULT;
1829
1830 ret = mnt_want_write_file(filp);
1831 if (ret)
1832 return ret;
1833
1834 range.minlen = max((unsigned int)range.minlen,
1835 q->limits.discard_granularity);
1836 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1837 mnt_drop_write_file(filp);
1838 if (ret < 0)
1839 return ret;
1840
1841 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1842 sizeof(range)))
1843 return -EFAULT;
1844 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1845 return 0;
1846 }
1847
1848 static bool uuid_is_nonzero(__u8 u[16])
1849 {
1850 int i;
1851
1852 for (i = 0; i < 16; i++)
1853 if (u[i])
1854 return true;
1855 return false;
1856 }
1857
1858 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1859 {
1860 struct inode *inode = file_inode(filp);
1861
1862 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1863
1864 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1865 }
1866
1867 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1868 {
1869 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1870 }
1871
1872 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1873 {
1874 struct inode *inode = file_inode(filp);
1875 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1876 int err;
1877
1878 if (!f2fs_sb_has_crypto(inode->i_sb))
1879 return -EOPNOTSUPP;
1880
1881 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1882 goto got_it;
1883
1884 err = mnt_want_write_file(filp);
1885 if (err)
1886 return err;
1887
1888 /* update superblock with uuid */
1889 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1890
1891 err = f2fs_commit_super(sbi, false);
1892 if (err) {
1893 /* undo new data */
1894 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1895 mnt_drop_write_file(filp);
1896 return err;
1897 }
1898 mnt_drop_write_file(filp);
1899 got_it:
1900 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1901 16))
1902 return -EFAULT;
1903 return 0;
1904 }
1905
1906 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1907 {
1908 struct inode *inode = file_inode(filp);
1909 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1910 __u32 sync;
1911 int ret;
1912
1913 if (!capable(CAP_SYS_ADMIN))
1914 return -EPERM;
1915
1916 if (get_user(sync, (__u32 __user *)arg))
1917 return -EFAULT;
1918
1919 if (f2fs_readonly(sbi->sb))
1920 return -EROFS;
1921
1922 ret = mnt_want_write_file(filp);
1923 if (ret)
1924 return ret;
1925
1926 if (!sync) {
1927 if (!mutex_trylock(&sbi->gc_mutex)) {
1928 ret = -EBUSY;
1929 goto out;
1930 }
1931 } else {
1932 mutex_lock(&sbi->gc_mutex);
1933 }
1934
1935 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
1936 out:
1937 mnt_drop_write_file(filp);
1938 return ret;
1939 }
1940
1941 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
1942 {
1943 struct inode *inode = file_inode(filp);
1944 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1945 struct f2fs_gc_range range;
1946 u64 end;
1947 int ret;
1948
1949 if (!capable(CAP_SYS_ADMIN))
1950 return -EPERM;
1951
1952 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
1953 sizeof(range)))
1954 return -EFAULT;
1955
1956 if (f2fs_readonly(sbi->sb))
1957 return -EROFS;
1958
1959 ret = mnt_want_write_file(filp);
1960 if (ret)
1961 return ret;
1962
1963 end = range.start + range.len;
1964 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi))
1965 return -EINVAL;
1966 do_more:
1967 if (!range.sync) {
1968 if (!mutex_trylock(&sbi->gc_mutex)) {
1969 ret = -EBUSY;
1970 goto out;
1971 }
1972 } else {
1973 mutex_lock(&sbi->gc_mutex);
1974 }
1975
1976 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
1977 range.start += sbi->blocks_per_seg;
1978 if (range.start <= end)
1979 goto do_more;
1980 out:
1981 mnt_drop_write_file(filp);
1982 return ret;
1983 }
1984
1985 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1986 {
1987 struct inode *inode = file_inode(filp);
1988 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1989 int ret;
1990
1991 if (!capable(CAP_SYS_ADMIN))
1992 return -EPERM;
1993
1994 if (f2fs_readonly(sbi->sb))
1995 return -EROFS;
1996
1997 ret = mnt_want_write_file(filp);
1998 if (ret)
1999 return ret;
2000
2001 ret = f2fs_sync_fs(sbi->sb, 1);
2002
2003 mnt_drop_write_file(filp);
2004 return ret;
2005 }
2006
2007 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2008 struct file *filp,
2009 struct f2fs_defragment *range)
2010 {
2011 struct inode *inode = file_inode(filp);
2012 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
2013 struct extent_info ei = {0,0,0};
2014 pgoff_t pg_start, pg_end;
2015 unsigned int blk_per_seg = sbi->blocks_per_seg;
2016 unsigned int total = 0, sec_num;
2017 block_t blk_end = 0;
2018 bool fragmented = false;
2019 int err;
2020
2021 /* if in-place-update policy is enabled, don't waste time here */
2022 if (need_inplace_update_policy(inode, NULL))
2023 return -EINVAL;
2024
2025 pg_start = range->start >> PAGE_SHIFT;
2026 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2027
2028 f2fs_balance_fs(sbi, true);
2029
2030 inode_lock(inode);
2031
2032 /* writeback all dirty pages in the range */
2033 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2034 range->start + range->len - 1);
2035 if (err)
2036 goto out;
2037
2038 /*
2039 * lookup mapping info in extent cache, skip defragmenting if physical
2040 * block addresses are continuous.
2041 */
2042 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2043 if (ei.fofs + ei.len >= pg_end)
2044 goto out;
2045 }
2046
2047 map.m_lblk = pg_start;
2048
2049 /*
2050 * lookup mapping info in dnode page cache, skip defragmenting if all
2051 * physical block addresses are continuous even if there are hole(s)
2052 * in logical blocks.
2053 */
2054 while (map.m_lblk < pg_end) {
2055 map.m_len = pg_end - map.m_lblk;
2056 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2057 if (err)
2058 goto out;
2059
2060 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2061 map.m_lblk++;
2062 continue;
2063 }
2064
2065 if (blk_end && blk_end != map.m_pblk) {
2066 fragmented = true;
2067 break;
2068 }
2069 blk_end = map.m_pblk + map.m_len;
2070
2071 map.m_lblk += map.m_len;
2072 }
2073
2074 if (!fragmented)
2075 goto out;
2076
2077 map.m_lblk = pg_start;
2078 map.m_len = pg_end - pg_start;
2079
2080 sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2081
2082 /*
2083 * make sure there are enough free section for LFS allocation, this can
2084 * avoid defragment running in SSR mode when free section are allocated
2085 * intensively
2086 */
2087 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2088 err = -EAGAIN;
2089 goto out;
2090 }
2091
2092 while (map.m_lblk < pg_end) {
2093 pgoff_t idx;
2094 int cnt = 0;
2095
2096 do_map:
2097 map.m_len = pg_end - map.m_lblk;
2098 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2099 if (err)
2100 goto clear_out;
2101
2102 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2103 map.m_lblk++;
2104 continue;
2105 }
2106
2107 set_inode_flag(inode, FI_DO_DEFRAG);
2108
2109 idx = map.m_lblk;
2110 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2111 struct page *page;
2112
2113 page = get_lock_data_page(inode, idx, true);
2114 if (IS_ERR(page)) {
2115 err = PTR_ERR(page);
2116 goto clear_out;
2117 }
2118
2119 set_page_dirty(page);
2120 f2fs_put_page(page, 1);
2121
2122 idx++;
2123 cnt++;
2124 total++;
2125 }
2126
2127 map.m_lblk = idx;
2128
2129 if (idx < pg_end && cnt < blk_per_seg)
2130 goto do_map;
2131
2132 clear_inode_flag(inode, FI_DO_DEFRAG);
2133
2134 err = filemap_fdatawrite(inode->i_mapping);
2135 if (err)
2136 goto out;
2137 }
2138 clear_out:
2139 clear_inode_flag(inode, FI_DO_DEFRAG);
2140 out:
2141 inode_unlock(inode);
2142 if (!err)
2143 range->len = (u64)total << PAGE_SHIFT;
2144 return err;
2145 }
2146
2147 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2148 {
2149 struct inode *inode = file_inode(filp);
2150 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2151 struct f2fs_defragment range;
2152 int err;
2153
2154 if (!capable(CAP_SYS_ADMIN))
2155 return -EPERM;
2156
2157 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2158 return -EINVAL;
2159
2160 if (f2fs_readonly(sbi->sb))
2161 return -EROFS;
2162
2163 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2164 sizeof(range)))
2165 return -EFAULT;
2166
2167 /* verify alignment of offset & size */
2168 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2169 return -EINVAL;
2170
2171 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2172 sbi->max_file_blocks))
2173 return -EINVAL;
2174
2175 err = mnt_want_write_file(filp);
2176 if (err)
2177 return err;
2178
2179 err = f2fs_defragment_range(sbi, filp, &range);
2180 mnt_drop_write_file(filp);
2181
2182 f2fs_update_time(sbi, REQ_TIME);
2183 if (err < 0)
2184 return err;
2185
2186 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2187 sizeof(range)))
2188 return -EFAULT;
2189
2190 return 0;
2191 }
2192
2193 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2194 struct file *file_out, loff_t pos_out, size_t len)
2195 {
2196 struct inode *src = file_inode(file_in);
2197 struct inode *dst = file_inode(file_out);
2198 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2199 size_t olen = len, dst_max_i_size = 0;
2200 size_t dst_osize;
2201 int ret;
2202
2203 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2204 src->i_sb != dst->i_sb)
2205 return -EXDEV;
2206
2207 if (unlikely(f2fs_readonly(src->i_sb)))
2208 return -EROFS;
2209
2210 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2211 return -EINVAL;
2212
2213 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2214 return -EOPNOTSUPP;
2215
2216 if (src == dst) {
2217 if (pos_in == pos_out)
2218 return 0;
2219 if (pos_out > pos_in && pos_out < pos_in + len)
2220 return -EINVAL;
2221 }
2222
2223 inode_lock(src);
2224 if (src != dst) {
2225 if (!inode_trylock(dst)) {
2226 ret = -EBUSY;
2227 goto out;
2228 }
2229 }
2230
2231 ret = -EINVAL;
2232 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2233 goto out_unlock;
2234 if (len == 0)
2235 olen = len = src->i_size - pos_in;
2236 if (pos_in + len == src->i_size)
2237 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2238 if (len == 0) {
2239 ret = 0;
2240 goto out_unlock;
2241 }
2242
2243 dst_osize = dst->i_size;
2244 if (pos_out + olen > dst->i_size)
2245 dst_max_i_size = pos_out + olen;
2246
2247 /* verify the end result is block aligned */
2248 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2249 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2250 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2251 goto out_unlock;
2252
2253 ret = f2fs_convert_inline_inode(src);
2254 if (ret)
2255 goto out_unlock;
2256
2257 ret = f2fs_convert_inline_inode(dst);
2258 if (ret)
2259 goto out_unlock;
2260
2261 /* write out all dirty pages from offset */
2262 ret = filemap_write_and_wait_range(src->i_mapping,
2263 pos_in, pos_in + len);
2264 if (ret)
2265 goto out_unlock;
2266
2267 ret = filemap_write_and_wait_range(dst->i_mapping,
2268 pos_out, pos_out + len);
2269 if (ret)
2270 goto out_unlock;
2271
2272 f2fs_balance_fs(sbi, true);
2273 f2fs_lock_op(sbi);
2274 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2275 pos_out >> F2FS_BLKSIZE_BITS,
2276 len >> F2FS_BLKSIZE_BITS, false);
2277
2278 if (!ret) {
2279 if (dst_max_i_size)
2280 f2fs_i_size_write(dst, dst_max_i_size);
2281 else if (dst_osize != dst->i_size)
2282 f2fs_i_size_write(dst, dst_osize);
2283 }
2284 f2fs_unlock_op(sbi);
2285 out_unlock:
2286 if (src != dst)
2287 inode_unlock(dst);
2288 out:
2289 inode_unlock(src);
2290 return ret;
2291 }
2292
2293 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2294 {
2295 struct f2fs_move_range range;
2296 struct fd dst;
2297 int err;
2298
2299 if (!(filp->f_mode & FMODE_READ) ||
2300 !(filp->f_mode & FMODE_WRITE))
2301 return -EBADF;
2302
2303 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2304 sizeof(range)))
2305 return -EFAULT;
2306
2307 dst = fdget(range.dst_fd);
2308 if (!dst.file)
2309 return -EBADF;
2310
2311 if (!(dst.file->f_mode & FMODE_WRITE)) {
2312 err = -EBADF;
2313 goto err_out;
2314 }
2315
2316 err = mnt_want_write_file(filp);
2317 if (err)
2318 goto err_out;
2319
2320 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2321 range.pos_out, range.len);
2322
2323 mnt_drop_write_file(filp);
2324 if (err)
2325 goto err_out;
2326
2327 if (copy_to_user((struct f2fs_move_range __user *)arg,
2328 &range, sizeof(range)))
2329 err = -EFAULT;
2330 err_out:
2331 fdput(dst);
2332 return err;
2333 }
2334
2335 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2336 {
2337 struct inode *inode = file_inode(filp);
2338 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2339 struct sit_info *sm = SIT_I(sbi);
2340 unsigned int start_segno = 0, end_segno = 0;
2341 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2342 struct f2fs_flush_device range;
2343 int ret;
2344
2345 if (!capable(CAP_SYS_ADMIN))
2346 return -EPERM;
2347
2348 if (f2fs_readonly(sbi->sb))
2349 return -EROFS;
2350
2351 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2352 sizeof(range)))
2353 return -EFAULT;
2354
2355 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2356 sbi->segs_per_sec != 1) {
2357 f2fs_msg(sbi->sb, KERN_WARNING,
2358 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2359 range.dev_num, sbi->s_ndevs,
2360 sbi->segs_per_sec);
2361 return -EINVAL;
2362 }
2363
2364 ret = mnt_want_write_file(filp);
2365 if (ret)
2366 return ret;
2367
2368 if (range.dev_num != 0)
2369 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2370 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2371
2372 start_segno = sm->last_victim[FLUSH_DEVICE];
2373 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2374 start_segno = dev_start_segno;
2375 end_segno = min(start_segno + range.segments, dev_end_segno);
2376
2377 while (start_segno < end_segno) {
2378 if (!mutex_trylock(&sbi->gc_mutex)) {
2379 ret = -EBUSY;
2380 goto out;
2381 }
2382 sm->last_victim[GC_CB] = end_segno + 1;
2383 sm->last_victim[GC_GREEDY] = end_segno + 1;
2384 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2385 ret = f2fs_gc(sbi, true, true, start_segno);
2386 if (ret == -EAGAIN)
2387 ret = 0;
2388 else if (ret < 0)
2389 break;
2390 start_segno++;
2391 }
2392 out:
2393 mnt_drop_write_file(filp);
2394 return ret;
2395 }
2396
2397 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2398 {
2399 struct inode *inode = file_inode(filp);
2400 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2401
2402 /* Must validate to set it with SQLite behavior in Android. */
2403 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2404
2405 return put_user(sb_feature, (u32 __user *)arg);
2406 }
2407
2408 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2409 {
2410 switch (cmd) {
2411 case F2FS_IOC_GETFLAGS:
2412 return f2fs_ioc_getflags(filp, arg);
2413 case F2FS_IOC_SETFLAGS:
2414 return f2fs_ioc_setflags(filp, arg);
2415 case F2FS_IOC_GETVERSION:
2416 return f2fs_ioc_getversion(filp, arg);
2417 case F2FS_IOC_START_ATOMIC_WRITE:
2418 return f2fs_ioc_start_atomic_write(filp);
2419 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2420 return f2fs_ioc_commit_atomic_write(filp);
2421 case F2FS_IOC_START_VOLATILE_WRITE:
2422 return f2fs_ioc_start_volatile_write(filp);
2423 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2424 return f2fs_ioc_release_volatile_write(filp);
2425 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2426 return f2fs_ioc_abort_volatile_write(filp);
2427 case F2FS_IOC_SHUTDOWN:
2428 return f2fs_ioc_shutdown(filp, arg);
2429 case FITRIM:
2430 return f2fs_ioc_fitrim(filp, arg);
2431 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2432 return f2fs_ioc_set_encryption_policy(filp, arg);
2433 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2434 return f2fs_ioc_get_encryption_policy(filp, arg);
2435 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2436 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2437 case F2FS_IOC_GARBAGE_COLLECT:
2438 return f2fs_ioc_gc(filp, arg);
2439 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2440 return f2fs_ioc_gc_range(filp, arg);
2441 case F2FS_IOC_WRITE_CHECKPOINT:
2442 return f2fs_ioc_write_checkpoint(filp, arg);
2443 case F2FS_IOC_DEFRAGMENT:
2444 return f2fs_ioc_defragment(filp, arg);
2445 case F2FS_IOC_MOVE_RANGE:
2446 return f2fs_ioc_move_range(filp, arg);
2447 case F2FS_IOC_FLUSH_DEVICE:
2448 return f2fs_ioc_flush_device(filp, arg);
2449 case F2FS_IOC_GET_FEATURES:
2450 return f2fs_ioc_get_features(filp, arg);
2451 default:
2452 return -ENOTTY;
2453 }
2454 }
2455
2456 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2457 {
2458 struct file *file = iocb->ki_filp;
2459 struct inode *inode = file_inode(file);
2460 struct blk_plug plug;
2461 ssize_t ret;
2462
2463 inode_lock(inode);
2464 ret = generic_write_checks(iocb, from);
2465 if (ret > 0) {
2466 int err;
2467
2468 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2469 set_inode_flag(inode, FI_NO_PREALLOC);
2470
2471 err = f2fs_preallocate_blocks(iocb, from);
2472 if (err) {
2473 inode_unlock(inode);
2474 return err;
2475 }
2476 blk_start_plug(&plug);
2477 ret = __generic_file_write_iter(iocb, from);
2478 blk_finish_plug(&plug);
2479 clear_inode_flag(inode, FI_NO_PREALLOC);
2480
2481 if (ret > 0)
2482 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2483 }
2484 inode_unlock(inode);
2485
2486 if (ret > 0) {
2487 ssize_t err;
2488
2489 err = generic_write_sync(file, iocb->ki_pos - ret, ret);
2490 if (err < 0)
2491 ret = err;
2492 }
2493 return ret;
2494 }
2495
2496 #ifdef CONFIG_COMPAT
2497 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2498 {
2499 switch (cmd) {
2500 case F2FS_IOC32_GETFLAGS:
2501 cmd = F2FS_IOC_GETFLAGS;
2502 break;
2503 case F2FS_IOC32_SETFLAGS:
2504 cmd = F2FS_IOC_SETFLAGS;
2505 break;
2506 case F2FS_IOC32_GETVERSION:
2507 cmd = F2FS_IOC_GETVERSION;
2508 break;
2509 case F2FS_IOC_START_ATOMIC_WRITE:
2510 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2511 case F2FS_IOC_START_VOLATILE_WRITE:
2512 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2513 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2514 case F2FS_IOC_SHUTDOWN:
2515 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2516 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2517 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2518 case F2FS_IOC_GARBAGE_COLLECT:
2519 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2520 case F2FS_IOC_WRITE_CHECKPOINT:
2521 case F2FS_IOC_DEFRAGMENT:
2522 case F2FS_IOC_MOVE_RANGE:
2523 case F2FS_IOC_FLUSH_DEVICE:
2524 case F2FS_IOC_GET_FEATURES:
2525 break;
2526 default:
2527 return -ENOIOCTLCMD;
2528 }
2529 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2530 }
2531 #endif
2532
2533 const struct file_operations f2fs_file_operations = {
2534 .llseek = f2fs_llseek,
2535 .read_iter = generic_file_read_iter,
2536 .write_iter = f2fs_file_write_iter,
2537 .open = f2fs_file_open,
2538 .release = f2fs_release_file,
2539 .mmap = f2fs_file_mmap,
2540 .flush = f2fs_file_flush,
2541 .fsync = f2fs_sync_file,
2542 .fallocate = f2fs_fallocate,
2543 .unlocked_ioctl = f2fs_ioctl,
2544 #ifdef CONFIG_COMPAT
2545 .compat_ioctl = f2fs_compat_ioctl,
2546 #endif
2547 .splice_read = generic_file_splice_read,
2548 .splice_write = iter_file_splice_write,
2549 };