projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
fs crypto: move per-file encryption from f2fs tree to fs/crypto
[deliverable/linux.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
25 #include "f2fs.h"
26 #include "node.h"
27 #include "segment.h"
28 #include "xattr.h"
29 #include "acl.h"
30 #include "gc.h"
31 #include "trace.h"
32 #include <trace/events/f2fs.h>
33
34 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
35 struct vm_fault *vmf)
36 {
37 struct page *page = vmf->page;
38 struct inode *inode = file_inode(vma->vm_file);
39 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
40 struct dnode_of_data dn;
41 int err;
42
43 sb_start_pagefault(inode->i_sb);
44
45 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
46
47 /* block allocation */
48 f2fs_lock_op(sbi);
49 set_new_dnode(&dn, inode, NULL, NULL, 0);
50 err = f2fs_reserve_block(&dn, page->index);
51 if (err) {
52 f2fs_unlock_op(sbi);
53 goto out;
54 }
55 f2fs_put_dnode(&dn);
56 f2fs_unlock_op(sbi);
57
58 f2fs_balance_fs(sbi, dn.node_changed);
59
60 file_update_time(vma->vm_file);
61 lock_page(page);
62 if (unlikely(page->mapping != inode->i_mapping ||
63 page_offset(page) > i_size_read(inode) ||
64 !PageUptodate(page))) {
65 unlock_page(page);
66 err = -EFAULT;
67 goto out;
68 }
69
70 /*
71 * check to see if the page is mapped already (no holes)
72 */
73 if (PageMappedToDisk(page))
74 goto mapped;
75
76 /* page is wholly or partially inside EOF */
77 if (((loff_t)(page->index + 1) << PAGE_CACHE_SHIFT) >
78 i_size_read(inode)) {
79 unsigned offset;
80 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
81 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
82 }
83 set_page_dirty(page);
84 SetPageUptodate(page);
85
86 trace_f2fs_vm_page_mkwrite(page, DATA);
87 mapped:
88 /* fill the page */
89 f2fs_wait_on_page_writeback(page, DATA, false);
90
91 /* wait for GCed encrypted page writeback */
92 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
93 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
94
95 /* if gced page is attached, don't write to cold segment */
96 clear_cold_data(page);
97 out:
98 sb_end_pagefault(inode->i_sb);
99 f2fs_update_time(sbi, REQ_TIME);
100 return block_page_mkwrite_return(err);
101 }
102
103 static const struct vm_operations_struct f2fs_file_vm_ops = {
104 .fault = filemap_fault,
105 .map_pages = filemap_map_pages,
106 .page_mkwrite = f2fs_vm_page_mkwrite,
107 };
108
109 static int get_parent_ino(struct inode *inode, nid_t *pino)
110 {
111 struct dentry *dentry;
112
113 inode = igrab(inode);
114 dentry = d_find_any_alias(inode);
115 iput(inode);
116 if (!dentry)
117 return 0;
118
119 if (update_dent_inode(inode, inode, &dentry->d_name)) {
120 dput(dentry);
121 return 0;
122 }
123
124 *pino = parent_ino(dentry);
125 dput(dentry);
126 return 1;
127 }
128
129 static inline bool need_do_checkpoint(struct inode *inode)
130 {
131 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
132 bool need_cp = false;
133
134 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
135 need_cp = true;
136 else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
137 need_cp = true;
138 else if (file_wrong_pino(inode))
139 need_cp = true;
140 else if (!space_for_roll_forward(sbi))
141 need_cp = true;
142 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
143 need_cp = true;
144 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
145 need_cp = true;
146 else if (test_opt(sbi, FASTBOOT))
147 need_cp = true;
148 else if (sbi->active_logs == 2)
149 need_cp = true;
150
151 return need_cp;
152 }
153
154 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
155 {
156 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
157 bool ret = false;
158 /* But we need to avoid that there are some inode updates */
159 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
160 ret = true;
161 f2fs_put_page(i, 0);
162 return ret;
163 }
164
165 static void try_to_fix_pino(struct inode *inode)
166 {
167 struct f2fs_inode_info *fi = F2FS_I(inode);
168 nid_t pino;
169
170 down_write(&fi->i_sem);
171 fi->xattr_ver = 0;
172 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
173 get_parent_ino(inode, &pino)) {
174 fi->i_pino = pino;
175 file_got_pino(inode);
176 up_write(&fi->i_sem);
177
178 mark_inode_dirty_sync(inode);
179 f2fs_write_inode(inode, NULL);
180 } else {
181 up_write(&fi->i_sem);
182 }
183 }
184
185 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
186 {
187 struct inode *inode = file->f_mapping->host;
188 struct f2fs_inode_info *fi = F2FS_I(inode);
189 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
190 nid_t ino = inode->i_ino;
191 int ret = 0;
192 bool need_cp = false;
193 struct writeback_control wbc = {
194 .sync_mode = WB_SYNC_ALL,
195 .nr_to_write = LONG_MAX,
196 .for_reclaim = 0,
197 };
198
199 if (unlikely(f2fs_readonly(inode->i_sb)))
200 return 0;
201
202 trace_f2fs_sync_file_enter(inode);
203
204 /* if fdatasync is triggered, let's do in-place-update */
205 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
206 set_inode_flag(fi, FI_NEED_IPU);
207 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
208 clear_inode_flag(fi, FI_NEED_IPU);
209
210 if (ret) {
211 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
212 return ret;
213 }
214
215 /* if the inode is dirty, let's recover all the time */
216 if (!datasync) {
217 f2fs_write_inode(inode, NULL);
218 goto go_write;
219 }
220
221 /*
222 * if there is no written data, don't waste time to write recovery info.
223 */
224 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
225 !exist_written_data(sbi, ino, APPEND_INO)) {
226
227 /* it may call write_inode just prior to fsync */
228 if (need_inode_page_update(sbi, ino))
229 goto go_write;
230
231 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
232 exist_written_data(sbi, ino, UPDATE_INO))
233 goto flush_out;
234 goto out;
235 }
236 go_write:
237 /*
238 * Both of fdatasync() and fsync() are able to be recovered from
239 * sudden-power-off.
240 */
241 down_read(&fi->i_sem);
242 need_cp = need_do_checkpoint(inode);
243 up_read(&fi->i_sem);
244
245 if (need_cp) {
246 /* all the dirty node pages should be flushed for POR */
247 ret = f2fs_sync_fs(inode->i_sb, 1);
248
249 /*
250 * We've secured consistency through sync_fs. Following pino
251 * will be used only for fsynced inodes after checkpoint.
252 */
253 try_to_fix_pino(inode);
254 clear_inode_flag(fi, FI_APPEND_WRITE);
255 clear_inode_flag(fi, FI_UPDATE_WRITE);
256 goto out;
257 }
258 sync_nodes:
259 sync_node_pages(sbi, ino, &wbc);
260
261 /* if cp_error was enabled, we should avoid infinite loop */
262 if (unlikely(f2fs_cp_error(sbi))) {
263 ret = -EIO;
264 goto out;
265 }
266
267 if (need_inode_block_update(sbi, ino)) {
268 mark_inode_dirty_sync(inode);
269 f2fs_write_inode(inode, NULL);
270 goto sync_nodes;
271 }
272
273 ret = wait_on_node_pages_writeback(sbi, ino);
274 if (ret)
275 goto out;
276
277 /* once recovery info is written, don't need to tack this */
278 remove_ino_entry(sbi, ino, APPEND_INO);
279 clear_inode_flag(fi, FI_APPEND_WRITE);
280 flush_out:
281 remove_ino_entry(sbi, ino, UPDATE_INO);
282 clear_inode_flag(fi, FI_UPDATE_WRITE);
283 ret = f2fs_issue_flush(sbi);
284 f2fs_update_time(sbi, REQ_TIME);
285 out:
286 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
287 f2fs_trace_ios(NULL, 1);
288 return ret;
289 }
290
291 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
292 pgoff_t pgofs, int whence)
293 {
294 struct pagevec pvec;
295 int nr_pages;
296
297 if (whence != SEEK_DATA)
298 return 0;
299
300 /* find first dirty page index */
301 pagevec_init(&pvec, 0);
302 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
303 PAGECACHE_TAG_DIRTY, 1);
304 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
305 pagevec_release(&pvec);
306 return pgofs;
307 }
308
309 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
310 int whence)
311 {
312 switch (whence) {
313 case SEEK_DATA:
314 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
315 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
316 return true;
317 break;
318 case SEEK_HOLE:
319 if (blkaddr == NULL_ADDR)
320 return true;
321 break;
322 }
323 return false;
324 }
325
326 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
327 {
328 struct inode *inode = file->f_mapping->host;
329 loff_t maxbytes = inode->i_sb->s_maxbytes;
330 struct dnode_of_data dn;
331 pgoff_t pgofs, end_offset, dirty;
332 loff_t data_ofs = offset;
333 loff_t isize;
334 int err = 0;
335
336 inode_lock(inode);
337
338 isize = i_size_read(inode);
339 if (offset >= isize)
340 goto fail;
341
342 /* handle inline data case */
343 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
344 if (whence == SEEK_HOLE)
345 data_ofs = isize;
346 goto found;
347 }
348
349 pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
350
351 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
352
353 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
354 set_new_dnode(&dn, inode, NULL, NULL, 0);
355 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
356 if (err && err != -ENOENT) {
357 goto fail;
358 } else if (err == -ENOENT) {
359 /* direct node does not exists */
360 if (whence == SEEK_DATA) {
361 pgofs = get_next_page_offset(&dn, pgofs);
362 continue;
363 } else {
364 goto found;
365 }
366 }
367
368 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
369
370 /* find data/hole in dnode block */
371 for (; dn.ofs_in_node < end_offset;
372 dn.ofs_in_node++, pgofs++,
373 data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
374 block_t blkaddr;
375 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
376
377 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
378 f2fs_put_dnode(&dn);
379 goto found;
380 }
381 }
382 f2fs_put_dnode(&dn);
383 }
384
385 if (whence == SEEK_DATA)
386 goto fail;
387 found:
388 if (whence == SEEK_HOLE && data_ofs > isize)
389 data_ofs = isize;
390 inode_unlock(inode);
391 return vfs_setpos(file, data_ofs, maxbytes);
392 fail:
393 inode_unlock(inode);
394 return -ENXIO;
395 }
396
397 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
398 {
399 struct inode *inode = file->f_mapping->host;
400 loff_t maxbytes = inode->i_sb->s_maxbytes;
401
402 switch (whence) {
403 case SEEK_SET:
404 case SEEK_CUR:
405 case SEEK_END:
406 return generic_file_llseek_size(file, offset, whence,
407 maxbytes, i_size_read(inode));
408 case SEEK_DATA:
409 case SEEK_HOLE:
410 if (offset < 0)
411 return -ENXIO;
412 return f2fs_seek_block(file, offset, whence);
413 }
414
415 return -EINVAL;
416 }
417
418 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
419 {
420 struct inode *inode = file_inode(file);
421 int err;
422
423 if (f2fs_encrypted_inode(inode)) {
424 err = fscrypt_get_encryption_info(inode);
425 if (err)
426 return 0;
427 if (!f2fs_encrypted_inode(inode))
428 return -ENOKEY;
429 }
430
431 /* we don't need to use inline_data strictly */
432 err = f2fs_convert_inline_inode(inode);
433 if (err)
434 return err;
435
436 file_accessed(file);
437 vma->vm_ops = &f2fs_file_vm_ops;
438 return 0;
439 }
440
441 static int f2fs_file_open(struct inode *inode, struct file *filp)
442 {
443 int ret = generic_file_open(inode, filp);
444
445 if (!ret && f2fs_encrypted_inode(inode)) {
446 ret = fscrypt_get_encryption_info(inode);
447 if (ret)
448 return -EACCES;
449 if (!fscrypt_has_encryption_key(inode))
450 return -ENOKEY;
451 }
452 return ret;
453 }
454
455 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
456 {
457 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
458 struct f2fs_node *raw_node;
459 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
460 __le32 *addr;
461
462 raw_node = F2FS_NODE(dn->node_page);
463 addr = blkaddr_in_node(raw_node) + ofs;
464
465 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
466 block_t blkaddr = le32_to_cpu(*addr);
467 if (blkaddr == NULL_ADDR)
468 continue;
469
470 dn->data_blkaddr = NULL_ADDR;
471 set_data_blkaddr(dn);
472 invalidate_blocks(sbi, blkaddr);
473 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
474 clear_inode_flag(F2FS_I(dn->inode),
475 FI_FIRST_BLOCK_WRITTEN);
476 nr_free++;
477 }
478
479 if (nr_free) {
480 pgoff_t fofs;
481 /*
482 * once we invalidate valid blkaddr in range [ofs, ofs + count],
483 * we will invalidate all blkaddr in the whole range.
484 */
485 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
486 dn->inode) + ofs;
487 f2fs_update_extent_cache_range(dn, fofs, 0, len);
488 dec_valid_block_count(sbi, dn->inode, nr_free);
489 sync_inode_page(dn);
490 }
491 dn->ofs_in_node = ofs;
492
493 f2fs_update_time(sbi, REQ_TIME);
494 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
495 dn->ofs_in_node, nr_free);
496 return nr_free;
497 }
498
499 void truncate_data_blocks(struct dnode_of_data *dn)
500 {
501 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
502 }
503
504 static int truncate_partial_data_page(struct inode *inode, u64 from,
505 bool cache_only)
506 {
507 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
508 pgoff_t index = from >> PAGE_CACHE_SHIFT;
509 struct address_space *mapping = inode->i_mapping;
510 struct page *page;
511
512 if (!offset && !cache_only)
513 return 0;
514
515 if (cache_only) {
516 page = f2fs_grab_cache_page(mapping, index, false);
517 if (page && PageUptodate(page))
518 goto truncate_out;
519 f2fs_put_page(page, 1);
520 return 0;
521 }
522
523 page = get_lock_data_page(inode, index, true);
524 if (IS_ERR(page))
525 return 0;
526 truncate_out:
527 f2fs_wait_on_page_writeback(page, DATA, true);
528 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
529 if (!cache_only || !f2fs_encrypted_inode(inode) ||
530 !S_ISREG(inode->i_mode))
531 set_page_dirty(page);
532 f2fs_put_page(page, 1);
533 return 0;
534 }
535
536 int truncate_blocks(struct inode *inode, u64 from, bool lock)
537 {
538 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
539 unsigned int blocksize = inode->i_sb->s_blocksize;
540 struct dnode_of_data dn;
541 pgoff_t free_from;
542 int count = 0, err = 0;
543 struct page *ipage;
544 bool truncate_page = false;
545
546 trace_f2fs_truncate_blocks_enter(inode, from);
547
548 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
549
550 if (lock)
551 f2fs_lock_op(sbi);
552
553 ipage = get_node_page(sbi, inode->i_ino);
554 if (IS_ERR(ipage)) {
555 err = PTR_ERR(ipage);
556 goto out;
557 }
558
559 if (f2fs_has_inline_data(inode)) {
560 if (truncate_inline_inode(ipage, from))
561 set_page_dirty(ipage);
562 f2fs_put_page(ipage, 1);
563 truncate_page = true;
564 goto out;
565 }
566
567 set_new_dnode(&dn, inode, ipage, NULL, 0);
568 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
569 if (err) {
570 if (err == -ENOENT)
571 goto free_next;
572 goto out;
573 }
574
575 count = ADDRS_PER_PAGE(dn.node_page, inode);
576
577 count -= dn.ofs_in_node;
578 f2fs_bug_on(sbi, count < 0);
579
580 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
581 truncate_data_blocks_range(&dn, count);
582 free_from += count;
583 }
584
585 f2fs_put_dnode(&dn);
586 free_next:
587 err = truncate_inode_blocks(inode, free_from);
588 out:
589 if (lock)
590 f2fs_unlock_op(sbi);
591
592 /* lastly zero out the first data page */
593 if (!err)
594 err = truncate_partial_data_page(inode, from, truncate_page);
595
596 trace_f2fs_truncate_blocks_exit(inode, err);
597 return err;
598 }
599
600 int f2fs_truncate(struct inode *inode, bool lock)
601 {
602 int err;
603
604 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
605 S_ISLNK(inode->i_mode)))
606 return 0;
607
608 trace_f2fs_truncate(inode);
609
610 /* we should check inline_data size */
611 if (!f2fs_may_inline_data(inode)) {
612 err = f2fs_convert_inline_inode(inode);
613 if (err)
614 return err;
615 }
616
617 err = truncate_blocks(inode, i_size_read(inode), lock);
618 if (err)
619 return err;
620
621 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
622 mark_inode_dirty(inode);
623 return 0;
624 }
625
626 int f2fs_getattr(struct vfsmount *mnt,
627 struct dentry *dentry, struct kstat *stat)
628 {
629 struct inode *inode = d_inode(dentry);
630 generic_fillattr(inode, stat);
631 stat->blocks <<= 3;
632 return 0;
633 }
634
635 #ifdef CONFIG_F2FS_FS_POSIX_ACL
636 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
637 {
638 struct f2fs_inode_info *fi = F2FS_I(inode);
639 unsigned int ia_valid = attr->ia_valid;
640
641 if (ia_valid & ATTR_UID)
642 inode->i_uid = attr->ia_uid;
643 if (ia_valid & ATTR_GID)
644 inode->i_gid = attr->ia_gid;
645 if (ia_valid & ATTR_ATIME)
646 inode->i_atime = timespec_trunc(attr->ia_atime,
647 inode->i_sb->s_time_gran);
648 if (ia_valid & ATTR_MTIME)
649 inode->i_mtime = timespec_trunc(attr->ia_mtime,
650 inode->i_sb->s_time_gran);
651 if (ia_valid & ATTR_CTIME)
652 inode->i_ctime = timespec_trunc(attr->ia_ctime,
653 inode->i_sb->s_time_gran);
654 if (ia_valid & ATTR_MODE) {
655 umode_t mode = attr->ia_mode;
656
657 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
658 mode &= ~S_ISGID;
659 set_acl_inode(fi, mode);
660 }
661 }
662 #else
663 #define __setattr_copy setattr_copy
664 #endif
665
666 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
667 {
668 struct inode *inode = d_inode(dentry);
669 struct f2fs_inode_info *fi = F2FS_I(inode);
670 int err;
671
672 err = inode_change_ok(inode, attr);
673 if (err)
674 return err;
675
676 if (attr->ia_valid & ATTR_SIZE) {
677 if (f2fs_encrypted_inode(inode) &&
678 fscrypt_get_encryption_info(inode))
679 return -EACCES;
680
681 if (attr->ia_size <= i_size_read(inode)) {
682 truncate_setsize(inode, attr->ia_size);
683 err = f2fs_truncate(inode, true);
684 if (err)
685 return err;
686 f2fs_balance_fs(F2FS_I_SB(inode), true);
687 } else {
688 /*
689 * do not trim all blocks after i_size if target size is
690 * larger than i_size.
691 */
692 truncate_setsize(inode, attr->ia_size);
693
694 /* should convert inline inode here */
695 if (!f2fs_may_inline_data(inode)) {
696 err = f2fs_convert_inline_inode(inode);
697 if (err)
698 return err;
699 }
700 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
701 }
702 }
703
704 __setattr_copy(inode, attr);
705
706 if (attr->ia_valid & ATTR_MODE) {
707 err = posix_acl_chmod(inode, get_inode_mode(inode));
708 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
709 inode->i_mode = fi->i_acl_mode;
710 clear_inode_flag(fi, FI_ACL_MODE);
711 }
712 }
713
714 mark_inode_dirty(inode);
715 return err;
716 }
717
718 const struct inode_operations f2fs_file_inode_operations = {
719 .getattr = f2fs_getattr,
720 .setattr = f2fs_setattr,
721 .get_acl = f2fs_get_acl,
722 .set_acl = f2fs_set_acl,
723 #ifdef CONFIG_F2FS_FS_XATTR
724 .setxattr = generic_setxattr,
725 .getxattr = generic_getxattr,
726 .listxattr = f2fs_listxattr,
727 .removexattr = generic_removexattr,
728 #endif
729 .fiemap = f2fs_fiemap,
730 };
731
732 static int fill_zero(struct inode *inode, pgoff_t index,
733 loff_t start, loff_t len)
734 {
735 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
736 struct page *page;
737
738 if (!len)
739 return 0;
740
741 f2fs_balance_fs(sbi, true);
742
743 f2fs_lock_op(sbi);
744 page = get_new_data_page(inode, NULL, index, false);
745 f2fs_unlock_op(sbi);
746
747 if (IS_ERR(page))
748 return PTR_ERR(page);
749
750 f2fs_wait_on_page_writeback(page, DATA, true);
751 zero_user(page, start, len);
752 set_page_dirty(page);
753 f2fs_put_page(page, 1);
754 return 0;
755 }
756
757 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
758 {
759 int err;
760
761 while (pg_start < pg_end) {
762 struct dnode_of_data dn;
763 pgoff_t end_offset, count;
764
765 set_new_dnode(&dn, inode, NULL, NULL, 0);
766 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
767 if (err) {
768 if (err == -ENOENT) {
769 pg_start++;
770 continue;
771 }
772 return err;
773 }
774
775 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
776 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
777
778 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
779
780 truncate_data_blocks_range(&dn, count);
781 f2fs_put_dnode(&dn);
782
783 pg_start += count;
784 }
785 return 0;
786 }
787
788 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
789 {
790 pgoff_t pg_start, pg_end;
791 loff_t off_start, off_end;
792 int ret;
793
794 ret = f2fs_convert_inline_inode(inode);
795 if (ret)
796 return ret;
797
798 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
799 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
800
801 off_start = offset & (PAGE_CACHE_SIZE - 1);
802 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
803
804 if (pg_start == pg_end) {
805 ret = fill_zero(inode, pg_start, off_start,
806 off_end - off_start);
807 if (ret)
808 return ret;
809 } else {
810 if (off_start) {
811 ret = fill_zero(inode, pg_start++, off_start,
812 PAGE_CACHE_SIZE - off_start);
813 if (ret)
814 return ret;
815 }
816 if (off_end) {
817 ret = fill_zero(inode, pg_end, 0, off_end);
818 if (ret)
819 return ret;
820 }
821
822 if (pg_start < pg_end) {
823 struct address_space *mapping = inode->i_mapping;
824 loff_t blk_start, blk_end;
825 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
826
827 f2fs_balance_fs(sbi, true);
828
829 blk_start = (loff_t)pg_start << PAGE_CACHE_SHIFT;
830 blk_end = (loff_t)pg_end << PAGE_CACHE_SHIFT;
831 truncate_inode_pages_range(mapping, blk_start,
832 blk_end - 1);
833
834 f2fs_lock_op(sbi);
835 ret = truncate_hole(inode, pg_start, pg_end);
836 f2fs_unlock_op(sbi);
837 }
838 }
839
840 return ret;
841 }
842
843 static int __exchange_data_block(struct inode *inode, pgoff_t src,
844 pgoff_t dst, bool full)
845 {
846 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
847 struct dnode_of_data dn;
848 block_t new_addr;
849 bool do_replace = false;
850 int ret;
851
852 set_new_dnode(&dn, inode, NULL, NULL, 0);
853 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
854 if (ret && ret != -ENOENT) {
855 return ret;
856 } else if (ret == -ENOENT) {
857 new_addr = NULL_ADDR;
858 } else {
859 new_addr = dn.data_blkaddr;
860 if (!is_checkpointed_data(sbi, new_addr)) {
861 /* do not invalidate this block address */
862 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
863 do_replace = true;
864 }
865 f2fs_put_dnode(&dn);
866 }
867
868 if (new_addr == NULL_ADDR)
869 return full ? truncate_hole(inode, dst, dst + 1) : 0;
870
871 if (do_replace) {
872 struct page *ipage = get_node_page(sbi, inode->i_ino);
873 struct node_info ni;
874
875 if (IS_ERR(ipage)) {
876 ret = PTR_ERR(ipage);
877 goto err_out;
878 }
879
880 set_new_dnode(&dn, inode, ipage, NULL, 0);
881 ret = f2fs_reserve_block(&dn, dst);
882 if (ret)
883 goto err_out;
884
885 truncate_data_blocks_range(&dn, 1);
886
887 get_node_info(sbi, dn.nid, &ni);
888 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
889 ni.version, true, false);
890 f2fs_put_dnode(&dn);
891 } else {
892 struct page *psrc, *pdst;
893
894 psrc = get_lock_data_page(inode, src, true);
895 if (IS_ERR(psrc))
896 return PTR_ERR(psrc);
897 pdst = get_new_data_page(inode, NULL, dst, true);
898 if (IS_ERR(pdst)) {
899 f2fs_put_page(psrc, 1);
900 return PTR_ERR(pdst);
901 }
902 f2fs_copy_page(psrc, pdst);
903 set_page_dirty(pdst);
904 f2fs_put_page(pdst, 1);
905 f2fs_put_page(psrc, 1);
906
907 return truncate_hole(inode, src, src + 1);
908 }
909 return 0;
910
911 err_out:
912 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
913 f2fs_update_data_blkaddr(&dn, new_addr);
914 f2fs_put_dnode(&dn);
915 }
916 return ret;
917 }
918
919 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
920 {
921 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
922 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
923 int ret = 0;
924
925 for (; end < nrpages; start++, end++) {
926 f2fs_balance_fs(sbi, true);
927 f2fs_lock_op(sbi);
928 ret = __exchange_data_block(inode, end, start, true);
929 f2fs_unlock_op(sbi);
930 if (ret)
931 break;
932 }
933 return ret;
934 }
935
936 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
937 {
938 pgoff_t pg_start, pg_end;
939 loff_t new_size;
940 int ret;
941
942 if (offset + len >= i_size_read(inode))
943 return -EINVAL;
944
945 /* collapse range should be aligned to block size of f2fs. */
946 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
947 return -EINVAL;
948
949 ret = f2fs_convert_inline_inode(inode);
950 if (ret)
951 return ret;
952
953 pg_start = offset >> PAGE_CACHE_SHIFT;
954 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
955
956 /* write out all dirty pages from offset */
957 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
958 if (ret)
959 return ret;
960
961 truncate_pagecache(inode, offset);
962
963 ret = f2fs_do_collapse(inode, pg_start, pg_end);
964 if (ret)
965 return ret;
966
967 /* write out all moved pages, if possible */
968 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
969 truncate_pagecache(inode, offset);
970
971 new_size = i_size_read(inode) - len;
972 truncate_pagecache(inode, new_size);
973
974 ret = truncate_blocks(inode, new_size, true);
975 if (!ret)
976 i_size_write(inode, new_size);
977
978 return ret;
979 }
980
981 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
982 int mode)
983 {
984 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
985 struct address_space *mapping = inode->i_mapping;
986 pgoff_t index, pg_start, pg_end;
987 loff_t new_size = i_size_read(inode);
988 loff_t off_start, off_end;
989 int ret = 0;
990
991 ret = inode_newsize_ok(inode, (len + offset));
992 if (ret)
993 return ret;
994
995 ret = f2fs_convert_inline_inode(inode);
996 if (ret)
997 return ret;
998
999 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1000 if (ret)
1001 return ret;
1002
1003 truncate_pagecache_range(inode, offset, offset + len - 1);
1004
1005 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1006 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1007
1008 off_start = offset & (PAGE_CACHE_SIZE - 1);
1009 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1010
1011 if (pg_start == pg_end) {
1012 ret = fill_zero(inode, pg_start, off_start,
1013 off_end - off_start);
1014 if (ret)
1015 return ret;
1016
1017 if (offset + len > new_size)
1018 new_size = offset + len;
1019 new_size = max_t(loff_t, new_size, offset + len);
1020 } else {
1021 if (off_start) {
1022 ret = fill_zero(inode, pg_start++, off_start,
1023 PAGE_CACHE_SIZE - off_start);
1024 if (ret)
1025 return ret;
1026
1027 new_size = max_t(loff_t, new_size,
1028 (loff_t)pg_start << PAGE_CACHE_SHIFT);
1029 }
1030
1031 for (index = pg_start; index < pg_end; index++) {
1032 struct dnode_of_data dn;
1033 struct page *ipage;
1034
1035 f2fs_lock_op(sbi);
1036
1037 ipage = get_node_page(sbi, inode->i_ino);
1038 if (IS_ERR(ipage)) {
1039 ret = PTR_ERR(ipage);
1040 f2fs_unlock_op(sbi);
1041 goto out;
1042 }
1043
1044 set_new_dnode(&dn, inode, ipage, NULL, 0);
1045 ret = f2fs_reserve_block(&dn, index);
1046 if (ret) {
1047 f2fs_unlock_op(sbi);
1048 goto out;
1049 }
1050
1051 if (dn.data_blkaddr != NEW_ADDR) {
1052 invalidate_blocks(sbi, dn.data_blkaddr);
1053 f2fs_update_data_blkaddr(&dn, NEW_ADDR);
1054 }
1055 f2fs_put_dnode(&dn);
1056 f2fs_unlock_op(sbi);
1057
1058 new_size = max_t(loff_t, new_size,
1059 (loff_t)(index + 1) << PAGE_CACHE_SHIFT);
1060 }
1061
1062 if (off_end) {
1063 ret = fill_zero(inode, pg_end, 0, off_end);
1064 if (ret)
1065 goto out;
1066
1067 new_size = max_t(loff_t, new_size, offset + len);
1068 }
1069 }
1070
1071 out:
1072 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1073 i_size_write(inode, new_size);
1074 mark_inode_dirty(inode);
1075 update_inode_page(inode);
1076 }
1077
1078 return ret;
1079 }
1080
1081 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1082 {
1083 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1084 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1085 loff_t new_size;
1086 int ret = 0;
1087
1088 new_size = i_size_read(inode) + len;
1089 if (new_size > inode->i_sb->s_maxbytes)
1090 return -EFBIG;
1091
1092 if (offset >= i_size_read(inode))
1093 return -EINVAL;
1094
1095 /* insert range should be aligned to block size of f2fs. */
1096 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1097 return -EINVAL;
1098
1099 ret = f2fs_convert_inline_inode(inode);
1100 if (ret)
1101 return ret;
1102
1103 f2fs_balance_fs(sbi, true);
1104
1105 ret = truncate_blocks(inode, i_size_read(inode), true);
1106 if (ret)
1107 return ret;
1108
1109 /* write out all dirty pages from offset */
1110 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1111 if (ret)
1112 return ret;
1113
1114 truncate_pagecache(inode, offset);
1115
1116 pg_start = offset >> PAGE_CACHE_SHIFT;
1117 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
1118 delta = pg_end - pg_start;
1119 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1120
1121 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1122 f2fs_lock_op(sbi);
1123 ret = __exchange_data_block(inode, idx, idx + delta, false);
1124 f2fs_unlock_op(sbi);
1125 if (ret)
1126 break;
1127 }
1128
1129 /* write out all moved pages, if possible */
1130 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1131 truncate_pagecache(inode, offset);
1132
1133 if (!ret)
1134 i_size_write(inode, new_size);
1135 return ret;
1136 }
1137
1138 static int expand_inode_data(struct inode *inode, loff_t offset,
1139 loff_t len, int mode)
1140 {
1141 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1142 pgoff_t index, pg_start, pg_end;
1143 loff_t new_size = i_size_read(inode);
1144 loff_t off_start, off_end;
1145 int ret = 0;
1146
1147 ret = inode_newsize_ok(inode, (len + offset));
1148 if (ret)
1149 return ret;
1150
1151 ret = f2fs_convert_inline_inode(inode);
1152 if (ret)
1153 return ret;
1154
1155 f2fs_balance_fs(sbi, true);
1156
1157 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1158 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1159
1160 off_start = offset & (PAGE_CACHE_SIZE - 1);
1161 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1162
1163 f2fs_lock_op(sbi);
1164
1165 for (index = pg_start; index <= pg_end; index++) {
1166 struct dnode_of_data dn;
1167
1168 if (index == pg_end && !off_end)
1169 goto noalloc;
1170
1171 set_new_dnode(&dn, inode, NULL, NULL, 0);
1172 ret = f2fs_reserve_block(&dn, index);
1173 if (ret)
1174 break;
1175 noalloc:
1176 if (pg_start == pg_end)
1177 new_size = offset + len;
1178 else if (index == pg_start && off_start)
1179 new_size = (loff_t)(index + 1) << PAGE_CACHE_SHIFT;
1180 else if (index == pg_end)
1181 new_size = ((loff_t)index << PAGE_CACHE_SHIFT) +
1182 off_end;
1183 else
1184 new_size += PAGE_CACHE_SIZE;
1185 }
1186
1187 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1188 i_size_read(inode) < new_size) {
1189 i_size_write(inode, new_size);
1190 mark_inode_dirty(inode);
1191 update_inode_page(inode);
1192 }
1193 f2fs_unlock_op(sbi);
1194
1195 return ret;
1196 }
1197
1198 static long f2fs_fallocate(struct file *file, int mode,
1199 loff_t offset, loff_t len)
1200 {
1201 struct inode *inode = file_inode(file);
1202 long ret = 0;
1203
1204 /* f2fs only support ->fallocate for regular file */
1205 if (!S_ISREG(inode->i_mode))
1206 return -EINVAL;
1207
1208 if (f2fs_encrypted_inode(inode) &&
1209 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1210 return -EOPNOTSUPP;
1211
1212 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1213 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1214 FALLOC_FL_INSERT_RANGE))
1215 return -EOPNOTSUPP;
1216
1217 inode_lock(inode);
1218
1219 if (mode & FALLOC_FL_PUNCH_HOLE) {
1220 if (offset >= inode->i_size)
1221 goto out;
1222
1223 ret = punch_hole(inode, offset, len);
1224 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1225 ret = f2fs_collapse_range(inode, offset, len);
1226 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1227 ret = f2fs_zero_range(inode, offset, len, mode);
1228 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1229 ret = f2fs_insert_range(inode, offset, len);
1230 } else {
1231 ret = expand_inode_data(inode, offset, len, mode);
1232 }
1233
1234 if (!ret) {
1235 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1236 mark_inode_dirty(inode);
1237 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1238 }
1239
1240 out:
1241 inode_unlock(inode);
1242
1243 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1244 return ret;
1245 }
1246
1247 static int f2fs_release_file(struct inode *inode, struct file *filp)
1248 {
1249 /* some remained atomic pages should discarded */
1250 if (f2fs_is_atomic_file(inode))
1251 drop_inmem_pages(inode);
1252 if (f2fs_is_volatile_file(inode)) {
1253 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1254 filemap_fdatawrite(inode->i_mapping);
1255 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1256 }
1257 return 0;
1258 }
1259
1260 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1261 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1262
1263 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1264 {
1265 if (S_ISDIR(mode))
1266 return flags;
1267 else if (S_ISREG(mode))
1268 return flags & F2FS_REG_FLMASK;
1269 else
1270 return flags & F2FS_OTHER_FLMASK;
1271 }
1272
1273 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1274 {
1275 struct inode *inode = file_inode(filp);
1276 struct f2fs_inode_info *fi = F2FS_I(inode);
1277 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1278 return put_user(flags, (int __user *)arg);
1279 }
1280
1281 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1282 {
1283 struct inode *inode = file_inode(filp);
1284 struct f2fs_inode_info *fi = F2FS_I(inode);
1285 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1286 unsigned int oldflags;
1287 int ret;
1288
1289 ret = mnt_want_write_file(filp);
1290 if (ret)
1291 return ret;
1292
1293 if (!inode_owner_or_capable(inode)) {
1294 ret = -EACCES;
1295 goto out;
1296 }
1297
1298 if (get_user(flags, (int __user *)arg)) {
1299 ret = -EFAULT;
1300 goto out;
1301 }
1302
1303 flags = f2fs_mask_flags(inode->i_mode, flags);
1304
1305 inode_lock(inode);
1306
1307 oldflags = fi->i_flags;
1308
1309 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1310 if (!capable(CAP_LINUX_IMMUTABLE)) {
1311 inode_unlock(inode);
1312 ret = -EPERM;
1313 goto out;
1314 }
1315 }
1316
1317 flags = flags & FS_FL_USER_MODIFIABLE;
1318 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1319 fi->i_flags = flags;
1320 inode_unlock(inode);
1321
1322 f2fs_set_inode_flags(inode);
1323 inode->i_ctime = CURRENT_TIME;
1324 mark_inode_dirty(inode);
1325 out:
1326 mnt_drop_write_file(filp);
1327 return ret;
1328 }
1329
1330 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1331 {
1332 struct inode *inode = file_inode(filp);
1333
1334 return put_user(inode->i_generation, (int __user *)arg);
1335 }
1336
1337 static int f2fs_ioc_start_atomic_write(struct file *filp)
1338 {
1339 struct inode *inode = file_inode(filp);
1340 int ret;
1341
1342 if (!inode_owner_or_capable(inode))
1343 return -EACCES;
1344
1345 if (f2fs_is_atomic_file(inode))
1346 return 0;
1347
1348 ret = f2fs_convert_inline_inode(inode);
1349 if (ret)
1350 return ret;
1351
1352 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1353 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1354
1355 return 0;
1356 }
1357
1358 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1359 {
1360 struct inode *inode = file_inode(filp);
1361 int ret;
1362
1363 if (!inode_owner_or_capable(inode))
1364 return -EACCES;
1365
1366 if (f2fs_is_volatile_file(inode))
1367 return 0;
1368
1369 ret = mnt_want_write_file(filp);
1370 if (ret)
1371 return ret;
1372
1373 if (f2fs_is_atomic_file(inode)) {
1374 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1375 ret = commit_inmem_pages(inode);
1376 if (ret) {
1377 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1378 goto err_out;
1379 }
1380 }
1381
1382 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1383 err_out:
1384 mnt_drop_write_file(filp);
1385 return ret;
1386 }
1387
1388 static int f2fs_ioc_start_volatile_write(struct file *filp)
1389 {
1390 struct inode *inode = file_inode(filp);
1391 int ret;
1392
1393 if (!inode_owner_or_capable(inode))
1394 return -EACCES;
1395
1396 if (f2fs_is_volatile_file(inode))
1397 return 0;
1398
1399 ret = f2fs_convert_inline_inode(inode);
1400 if (ret)
1401 return ret;
1402
1403 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1404 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1405 return 0;
1406 }
1407
1408 static int f2fs_ioc_release_volatile_write(struct file *filp)
1409 {
1410 struct inode *inode = file_inode(filp);
1411
1412 if (!inode_owner_or_capable(inode))
1413 return -EACCES;
1414
1415 if (!f2fs_is_volatile_file(inode))
1416 return 0;
1417
1418 if (!f2fs_is_first_block_written(inode))
1419 return truncate_partial_data_page(inode, 0, true);
1420
1421 return punch_hole(inode, 0, F2FS_BLKSIZE);
1422 }
1423
1424 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1425 {
1426 struct inode *inode = file_inode(filp);
1427 int ret;
1428
1429 if (!inode_owner_or_capable(inode))
1430 return -EACCES;
1431
1432 ret = mnt_want_write_file(filp);
1433 if (ret)
1434 return ret;
1435
1436 if (f2fs_is_atomic_file(inode)) {
1437 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1438 drop_inmem_pages(inode);
1439 }
1440 if (f2fs_is_volatile_file(inode)) {
1441 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1442 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1443 }
1444
1445 mnt_drop_write_file(filp);
1446 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1447 return ret;
1448 }
1449
1450 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1451 {
1452 struct inode *inode = file_inode(filp);
1453 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1454 struct super_block *sb = sbi->sb;
1455 __u32 in;
1456
1457 if (!capable(CAP_SYS_ADMIN))
1458 return -EPERM;
1459
1460 if (get_user(in, (__u32 __user *)arg))
1461 return -EFAULT;
1462
1463 switch (in) {
1464 case F2FS_GOING_DOWN_FULLSYNC:
1465 sb = freeze_bdev(sb->s_bdev);
1466 if (sb && !IS_ERR(sb)) {
1467 f2fs_stop_checkpoint(sbi);
1468 thaw_bdev(sb->s_bdev, sb);
1469 }
1470 break;
1471 case F2FS_GOING_DOWN_METASYNC:
1472 /* do checkpoint only */
1473 f2fs_sync_fs(sb, 1);
1474 f2fs_stop_checkpoint(sbi);
1475 break;
1476 case F2FS_GOING_DOWN_NOSYNC:
1477 f2fs_stop_checkpoint(sbi);
1478 break;
1479 case F2FS_GOING_DOWN_METAFLUSH:
1480 sync_meta_pages(sbi, META, LONG_MAX);
1481 f2fs_stop_checkpoint(sbi);
1482 break;
1483 default:
1484 return -EINVAL;
1485 }
1486 f2fs_update_time(sbi, REQ_TIME);
1487 return 0;
1488 }
1489
1490 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1491 {
1492 struct inode *inode = file_inode(filp);
1493 struct super_block *sb = inode->i_sb;
1494 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1495 struct fstrim_range range;
1496 int ret;
1497
1498 if (!capable(CAP_SYS_ADMIN))
1499 return -EPERM;
1500
1501 if (!blk_queue_discard(q))
1502 return -EOPNOTSUPP;
1503
1504 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1505 sizeof(range)))
1506 return -EFAULT;
1507
1508 range.minlen = max((unsigned int)range.minlen,
1509 q->limits.discard_granularity);
1510 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1511 if (ret < 0)
1512 return ret;
1513
1514 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1515 sizeof(range)))
1516 return -EFAULT;
1517 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1518 return 0;
1519 }
1520
1521 static bool uuid_is_nonzero(__u8 u[16])
1522 {
1523 int i;
1524
1525 for (i = 0; i < 16; i++)
1526 if (u[i])
1527 return true;
1528 return false;
1529 }
1530
1531 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1532 {
1533 struct fscrypt_policy policy;
1534 struct inode *inode = file_inode(filp);
1535
1536 if (copy_from_user(&policy, (struct fscrypt_policy __user *)arg,
1537 sizeof(policy)))
1538 return -EFAULT;
1539
1540 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1541 return fscrypt_process_policy(inode, &policy);
1542 }
1543
1544 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1545 {
1546 struct fscrypt_policy policy;
1547 struct inode *inode = file_inode(filp);
1548 int err;
1549
1550 err = fscrypt_get_policy(inode, &policy);
1551 if (err)
1552 return err;
1553
1554 if (copy_to_user((struct fscrypt_policy __user *)arg, &policy, sizeof(policy)))
1555 return -EFAULT;
1556 return 0;
1557 }
1558
1559 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1560 {
1561 struct inode *inode = file_inode(filp);
1562 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1563 int err;
1564
1565 if (!f2fs_sb_has_crypto(inode->i_sb))
1566 return -EOPNOTSUPP;
1567
1568 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1569 goto got_it;
1570
1571 err = mnt_want_write_file(filp);
1572 if (err)
1573 return err;
1574
1575 /* update superblock with uuid */
1576 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1577
1578 err = f2fs_commit_super(sbi, false);
1579 if (err) {
1580 /* undo new data */
1581 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1582 mnt_drop_write_file(filp);
1583 return err;
1584 }
1585 mnt_drop_write_file(filp);
1586 got_it:
1587 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1588 16))
1589 return -EFAULT;
1590 return 0;
1591 }
1592
1593 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1594 {
1595 struct inode *inode = file_inode(filp);
1596 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1597 __u32 sync;
1598
1599 if (!capable(CAP_SYS_ADMIN))
1600 return -EPERM;
1601
1602 if (get_user(sync, (__u32 __user *)arg))
1603 return -EFAULT;
1604
1605 if (f2fs_readonly(sbi->sb))
1606 return -EROFS;
1607
1608 if (!sync) {
1609 if (!mutex_trylock(&sbi->gc_mutex))
1610 return -EBUSY;
1611 } else {
1612 mutex_lock(&sbi->gc_mutex);
1613 }
1614
1615 return f2fs_gc(sbi, sync);
1616 }
1617
1618 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1619 {
1620 struct inode *inode = file_inode(filp);
1621 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1622
1623 if (!capable(CAP_SYS_ADMIN))
1624 return -EPERM;
1625
1626 if (f2fs_readonly(sbi->sb))
1627 return -EROFS;
1628
1629 return f2fs_sync_fs(sbi->sb, 1);
1630 }
1631
1632 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1633 struct file *filp,
1634 struct f2fs_defragment *range)
1635 {
1636 struct inode *inode = file_inode(filp);
1637 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1638 struct extent_info ei;
1639 pgoff_t pg_start, pg_end;
1640 unsigned int blk_per_seg = sbi->blocks_per_seg;
1641 unsigned int total = 0, sec_num;
1642 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1643 block_t blk_end = 0;
1644 bool fragmented = false;
1645 int err;
1646
1647 /* if in-place-update policy is enabled, don't waste time here */
1648 if (need_inplace_update(inode))
1649 return -EINVAL;
1650
1651 pg_start = range->start >> PAGE_CACHE_SHIFT;
1652 pg_end = (range->start + range->len) >> PAGE_CACHE_SHIFT;
1653
1654 f2fs_balance_fs(sbi, true);
1655
1656 inode_lock(inode);
1657
1658 /* writeback all dirty pages in the range */
1659 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1660 range->start + range->len - 1);
1661 if (err)
1662 goto out;
1663
1664 /*
1665 * lookup mapping info in extent cache, skip defragmenting if physical
1666 * block addresses are continuous.
1667 */
1668 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1669 if (ei.fofs + ei.len >= pg_end)
1670 goto out;
1671 }
1672
1673 map.m_lblk = pg_start;
1674
1675 /*
1676 * lookup mapping info in dnode page cache, skip defragmenting if all
1677 * physical block addresses are continuous even if there are hole(s)
1678 * in logical blocks.
1679 */
1680 while (map.m_lblk < pg_end) {
1681 map.m_len = pg_end - map.m_lblk;
1682 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1683 if (err)
1684 goto out;
1685
1686 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1687 map.m_lblk++;
1688 continue;
1689 }
1690
1691 if (blk_end && blk_end != map.m_pblk) {
1692 fragmented = true;
1693 break;
1694 }
1695 blk_end = map.m_pblk + map.m_len;
1696
1697 map.m_lblk += map.m_len;
1698 }
1699
1700 if (!fragmented)
1701 goto out;
1702
1703 map.m_lblk = pg_start;
1704 map.m_len = pg_end - pg_start;
1705
1706 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1707
1708 /*
1709 * make sure there are enough free section for LFS allocation, this can
1710 * avoid defragment running in SSR mode when free section are allocated
1711 * intensively
1712 */
1713 if (has_not_enough_free_secs(sbi, sec_num)) {
1714 err = -EAGAIN;
1715 goto out;
1716 }
1717
1718 while (map.m_lblk < pg_end) {
1719 pgoff_t idx;
1720 int cnt = 0;
1721
1722 do_map:
1723 map.m_len = pg_end - map.m_lblk;
1724 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1725 if (err)
1726 goto clear_out;
1727
1728 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1729 map.m_lblk++;
1730 continue;
1731 }
1732
1733 set_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1734
1735 idx = map.m_lblk;
1736 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1737 struct page *page;
1738
1739 page = get_lock_data_page(inode, idx, true);
1740 if (IS_ERR(page)) {
1741 err = PTR_ERR(page);
1742 goto clear_out;
1743 }
1744
1745 set_page_dirty(page);
1746 f2fs_put_page(page, 1);
1747
1748 idx++;
1749 cnt++;
1750 total++;
1751 }
1752
1753 map.m_lblk = idx;
1754
1755 if (idx < pg_end && cnt < blk_per_seg)
1756 goto do_map;
1757
1758 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1759
1760 err = filemap_fdatawrite(inode->i_mapping);
1761 if (err)
1762 goto out;
1763 }
1764 clear_out:
1765 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1766 out:
1767 inode_unlock(inode);
1768 if (!err)
1769 range->len = (u64)total << PAGE_CACHE_SHIFT;
1770 return err;
1771 }
1772
1773 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1774 {
1775 struct inode *inode = file_inode(filp);
1776 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1777 struct f2fs_defragment range;
1778 int err;
1779
1780 if (!capable(CAP_SYS_ADMIN))
1781 return -EPERM;
1782
1783 if (!S_ISREG(inode->i_mode))
1784 return -EINVAL;
1785
1786 err = mnt_want_write_file(filp);
1787 if (err)
1788 return err;
1789
1790 if (f2fs_readonly(sbi->sb)) {
1791 err = -EROFS;
1792 goto out;
1793 }
1794
1795 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1796 sizeof(range))) {
1797 err = -EFAULT;
1798 goto out;
1799 }
1800
1801 /* verify alignment of offset & size */
1802 if (range.start & (F2FS_BLKSIZE - 1) ||
1803 range.len & (F2FS_BLKSIZE - 1)) {
1804 err = -EINVAL;
1805 goto out;
1806 }
1807
1808 err = f2fs_defragment_range(sbi, filp, &range);
1809 f2fs_update_time(sbi, REQ_TIME);
1810 if (err < 0)
1811 goto out;
1812
1813 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1814 sizeof(range)))
1815 err = -EFAULT;
1816 out:
1817 mnt_drop_write_file(filp);
1818 return err;
1819 }
1820
1821 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1822 {
1823 switch (cmd) {
1824 case F2FS_IOC_GETFLAGS:
1825 return f2fs_ioc_getflags(filp, arg);
1826 case F2FS_IOC_SETFLAGS:
1827 return f2fs_ioc_setflags(filp, arg);
1828 case F2FS_IOC_GETVERSION:
1829 return f2fs_ioc_getversion(filp, arg);
1830 case F2FS_IOC_START_ATOMIC_WRITE:
1831 return f2fs_ioc_start_atomic_write(filp);
1832 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1833 return f2fs_ioc_commit_atomic_write(filp);
1834 case F2FS_IOC_START_VOLATILE_WRITE:
1835 return f2fs_ioc_start_volatile_write(filp);
1836 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1837 return f2fs_ioc_release_volatile_write(filp);
1838 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1839 return f2fs_ioc_abort_volatile_write(filp);
1840 case F2FS_IOC_SHUTDOWN:
1841 return f2fs_ioc_shutdown(filp, arg);
1842 case FITRIM:
1843 return f2fs_ioc_fitrim(filp, arg);
1844 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1845 return f2fs_ioc_set_encryption_policy(filp, arg);
1846 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1847 return f2fs_ioc_get_encryption_policy(filp, arg);
1848 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1849 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1850 case F2FS_IOC_GARBAGE_COLLECT:
1851 return f2fs_ioc_gc(filp, arg);
1852 case F2FS_IOC_WRITE_CHECKPOINT:
1853 return f2fs_ioc_write_checkpoint(filp, arg);
1854 case F2FS_IOC_DEFRAGMENT:
1855 return f2fs_ioc_defragment(filp, arg);
1856 default:
1857 return -ENOTTY;
1858 }
1859 }
1860
1861 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1862 {
1863 struct file *file = iocb->ki_filp;
1864 struct inode *inode = file_inode(file);
1865 ssize_t ret;
1866
1867 if (f2fs_encrypted_inode(inode) &&
1868 !fscrypt_has_encryption_key(inode) &&
1869 fscrypt_get_encryption_info(inode))
1870 return -EACCES;
1871
1872 inode_lock(inode);
1873 ret = generic_write_checks(iocb, from);
1874 if (ret > 0) {
1875 ret = f2fs_preallocate_blocks(iocb, from);
1876 if (!ret)
1877 ret = __generic_file_write_iter(iocb, from);
1878 }
1879 inode_unlock(inode);
1880
1881 if (ret > 0) {
1882 ssize_t err;
1883
1884 err = generic_write_sync(file, iocb->ki_pos - ret, ret);
1885 if (err < 0)
1886 ret = err;
1887 }
1888 return ret;
1889 }
1890
1891 #ifdef CONFIG_COMPAT
1892 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1893 {
1894 switch (cmd) {
1895 case F2FS_IOC32_GETFLAGS:
1896 cmd = F2FS_IOC_GETFLAGS;
1897 break;
1898 case F2FS_IOC32_SETFLAGS:
1899 cmd = F2FS_IOC_SETFLAGS;
1900 break;
1901 case F2FS_IOC32_GETVERSION:
1902 cmd = F2FS_IOC_GETVERSION;
1903 break;
1904 case F2FS_IOC_START_ATOMIC_WRITE:
1905 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1906 case F2FS_IOC_START_VOLATILE_WRITE:
1907 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1908 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1909 case F2FS_IOC_SHUTDOWN:
1910 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1911 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1912 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1913 case F2FS_IOC_GARBAGE_COLLECT:
1914 case F2FS_IOC_WRITE_CHECKPOINT:
1915 case F2FS_IOC_DEFRAGMENT:
1916 break;
1917 default:
1918 return -ENOIOCTLCMD;
1919 }
1920 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1921 }
1922 #endif
1923
1924 const struct file_operations f2fs_file_operations = {
1925 .llseek = f2fs_llseek,
1926 .read_iter = generic_file_read_iter,
1927 .write_iter = f2fs_file_write_iter,
1928 .open = f2fs_file_open,
1929 .release = f2fs_release_file,
1930 .mmap = f2fs_file_mmap,
1931 .fsync = f2fs_sync_file,
1932 .fallocate = f2fs_fallocate,
1933 .unlocked_ioctl = f2fs_ioctl,
1934 #ifdef CONFIG_COMPAT
1935 .compat_ioctl = f2fs_compat_ioctl,
1936 #endif
1937 .splice_read = generic_file_splice_read,
1938 .splice_write = iter_file_splice_write,
1939 };
Linux kernel - Deliverables
This page took 0.070089 seconds and 5 git commands to generate.