projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: set fsync mark only for the last dnode
[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_SHIFT) >
78 i_size_read(inode)) {
79 unsigned offset;
80 offset = i_size_read(inode) & ~PAGE_MASK;
81 zero_user_segment(page, offset, PAGE_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 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
186 int datasync, bool atomic)
187 {
188 struct inode *inode = file->f_mapping->host;
189 struct f2fs_inode_info *fi = F2FS_I(inode);
190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
191 nid_t ino = inode->i_ino;
192 int ret = 0;
193 bool need_cp = false;
194 struct writeback_control wbc = {
195 .sync_mode = WB_SYNC_ALL,
196 .nr_to_write = LONG_MAX,
197 .for_reclaim = 0,
198 };
199
200 if (unlikely(f2fs_readonly(inode->i_sb)))
201 return 0;
202
203 trace_f2fs_sync_file_enter(inode);
204
205 /* if fdatasync is triggered, let's do in-place-update */
206 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
207 set_inode_flag(fi, FI_NEED_IPU);
208 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
209 clear_inode_flag(fi, FI_NEED_IPU);
210
211 if (ret) {
212 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
213 return ret;
214 }
215
216 /* if the inode is dirty, let's recover all the time */
217 if (!datasync) {
218 f2fs_write_inode(inode, NULL);
219 goto go_write;
220 }
221
222 /*
223 * if there is no written data, don't waste time to write recovery info.
224 */
225 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
226 !exist_written_data(sbi, ino, APPEND_INO)) {
227
228 /* it may call write_inode just prior to fsync */
229 if (need_inode_page_update(sbi, ino))
230 goto go_write;
231
232 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
233 exist_written_data(sbi, ino, UPDATE_INO))
234 goto flush_out;
235 goto out;
236 }
237 go_write:
238 /*
239 * Both of fdatasync() and fsync() are able to be recovered from
240 * sudden-power-off.
241 */
242 down_read(&fi->i_sem);
243 need_cp = need_do_checkpoint(inode);
244 up_read(&fi->i_sem);
245
246 if (need_cp) {
247 /* all the dirty node pages should be flushed for POR */
248 ret = f2fs_sync_fs(inode->i_sb, 1);
249
250 /*
251 * We've secured consistency through sync_fs. Following pino
252 * will be used only for fsynced inodes after checkpoint.
253 */
254 try_to_fix_pino(inode);
255 clear_inode_flag(fi, FI_APPEND_WRITE);
256 clear_inode_flag(fi, FI_UPDATE_WRITE);
257 goto out;
258 }
259 sync_nodes:
260 ret = fsync_node_pages(sbi, ino, &wbc, atomic);
261 if (ret)
262 goto out;
263
264 /* if cp_error was enabled, we should avoid infinite loop */
265 if (unlikely(f2fs_cp_error(sbi))) {
266 ret = -EIO;
267 goto out;
268 }
269
270 if (need_inode_block_update(sbi, ino)) {
271 mark_inode_dirty_sync(inode);
272 f2fs_write_inode(inode, NULL);
273 goto sync_nodes;
274 }
275
276 ret = wait_on_node_pages_writeback(sbi, ino);
277 if (ret)
278 goto out;
279
280 /* once recovery info is written, don't need to tack this */
281 remove_ino_entry(sbi, ino, APPEND_INO);
282 clear_inode_flag(fi, FI_APPEND_WRITE);
283 flush_out:
284 remove_ino_entry(sbi, ino, UPDATE_INO);
285 clear_inode_flag(fi, FI_UPDATE_WRITE);
286 ret = f2fs_issue_flush(sbi);
287 f2fs_update_time(sbi, REQ_TIME);
288 out:
289 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
290 f2fs_trace_ios(NULL, 1);
291 return ret;
292 }
293
294 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
295 {
296 return f2fs_do_sync_file(file, start, end, datasync, false);
297 }
298
299 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
300 pgoff_t pgofs, int whence)
301 {
302 struct pagevec pvec;
303 int nr_pages;
304
305 if (whence != SEEK_DATA)
306 return 0;
307
308 /* find first dirty page index */
309 pagevec_init(&pvec, 0);
310 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
311 PAGECACHE_TAG_DIRTY, 1);
312 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
313 pagevec_release(&pvec);
314 return pgofs;
315 }
316
317 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
318 int whence)
319 {
320 switch (whence) {
321 case SEEK_DATA:
322 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
323 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
324 return true;
325 break;
326 case SEEK_HOLE:
327 if (blkaddr == NULL_ADDR)
328 return true;
329 break;
330 }
331 return false;
332 }
333
334 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
335 {
336 struct inode *inode = file->f_mapping->host;
337 loff_t maxbytes = inode->i_sb->s_maxbytes;
338 struct dnode_of_data dn;
339 pgoff_t pgofs, end_offset, dirty;
340 loff_t data_ofs = offset;
341 loff_t isize;
342 int err = 0;
343
344 inode_lock(inode);
345
346 isize = i_size_read(inode);
347 if (offset >= isize)
348 goto fail;
349
350 /* handle inline data case */
351 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
352 if (whence == SEEK_HOLE)
353 data_ofs = isize;
354 goto found;
355 }
356
357 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
358
359 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
360
361 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
362 set_new_dnode(&dn, inode, NULL, NULL, 0);
363 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
364 if (err && err != -ENOENT) {
365 goto fail;
366 } else if (err == -ENOENT) {
367 /* direct node does not exists */
368 if (whence == SEEK_DATA) {
369 pgofs = get_next_page_offset(&dn, pgofs);
370 continue;
371 } else {
372 goto found;
373 }
374 }
375
376 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
377
378 /* find data/hole in dnode block */
379 for (; dn.ofs_in_node < end_offset;
380 dn.ofs_in_node++, pgofs++,
381 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
382 block_t blkaddr;
383 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
384
385 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
386 f2fs_put_dnode(&dn);
387 goto found;
388 }
389 }
390 f2fs_put_dnode(&dn);
391 }
392
393 if (whence == SEEK_DATA)
394 goto fail;
395 found:
396 if (whence == SEEK_HOLE && data_ofs > isize)
397 data_ofs = isize;
398 inode_unlock(inode);
399 return vfs_setpos(file, data_ofs, maxbytes);
400 fail:
401 inode_unlock(inode);
402 return -ENXIO;
403 }
404
405 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
406 {
407 struct inode *inode = file->f_mapping->host;
408 loff_t maxbytes = inode->i_sb->s_maxbytes;
409
410 switch (whence) {
411 case SEEK_SET:
412 case SEEK_CUR:
413 case SEEK_END:
414 return generic_file_llseek_size(file, offset, whence,
415 maxbytes, i_size_read(inode));
416 case SEEK_DATA:
417 case SEEK_HOLE:
418 if (offset < 0)
419 return -ENXIO;
420 return f2fs_seek_block(file, offset, whence);
421 }
422
423 return -EINVAL;
424 }
425
426 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
427 {
428 struct inode *inode = file_inode(file);
429 int err;
430
431 if (f2fs_encrypted_inode(inode)) {
432 err = fscrypt_get_encryption_info(inode);
433 if (err)
434 return 0;
435 if (!f2fs_encrypted_inode(inode))
436 return -ENOKEY;
437 }
438
439 /* we don't need to use inline_data strictly */
440 err = f2fs_convert_inline_inode(inode);
441 if (err)
442 return err;
443
444 file_accessed(file);
445 vma->vm_ops = &f2fs_file_vm_ops;
446 return 0;
447 }
448
449 static int f2fs_file_open(struct inode *inode, struct file *filp)
450 {
451 int ret = generic_file_open(inode, filp);
452 struct dentry *dir;
453
454 if (!ret && f2fs_encrypted_inode(inode)) {
455 ret = fscrypt_get_encryption_info(inode);
456 if (ret)
457 return -EACCES;
458 if (!fscrypt_has_encryption_key(inode))
459 return -ENOKEY;
460 }
461 dir = dget_parent(file_dentry(filp));
462 if (f2fs_encrypted_inode(d_inode(dir)) &&
463 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
464 dput(dir);
465 return -EPERM;
466 }
467 dput(dir);
468 return ret;
469 }
470
471 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
472 {
473 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
474 struct f2fs_node *raw_node;
475 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
476 __le32 *addr;
477
478 raw_node = F2FS_NODE(dn->node_page);
479 addr = blkaddr_in_node(raw_node) + ofs;
480
481 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
482 block_t blkaddr = le32_to_cpu(*addr);
483 if (blkaddr == NULL_ADDR)
484 continue;
485
486 dn->data_blkaddr = NULL_ADDR;
487 set_data_blkaddr(dn);
488 invalidate_blocks(sbi, blkaddr);
489 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
490 clear_inode_flag(F2FS_I(dn->inode),
491 FI_FIRST_BLOCK_WRITTEN);
492 nr_free++;
493 }
494
495 if (nr_free) {
496 pgoff_t fofs;
497 /*
498 * once we invalidate valid blkaddr in range [ofs, ofs + count],
499 * we will invalidate all blkaddr in the whole range.
500 */
501 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
502 dn->inode) + ofs;
503 f2fs_update_extent_cache_range(dn, fofs, 0, len);
504 dec_valid_block_count(sbi, dn->inode, nr_free);
505 sync_inode_page(dn);
506 }
507 dn->ofs_in_node = ofs;
508
509 f2fs_update_time(sbi, REQ_TIME);
510 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
511 dn->ofs_in_node, nr_free);
512 return nr_free;
513 }
514
515 void truncate_data_blocks(struct dnode_of_data *dn)
516 {
517 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
518 }
519
520 static int truncate_partial_data_page(struct inode *inode, u64 from,
521 bool cache_only)
522 {
523 unsigned offset = from & (PAGE_SIZE - 1);
524 pgoff_t index = from >> PAGE_SHIFT;
525 struct address_space *mapping = inode->i_mapping;
526 struct page *page;
527
528 if (!offset && !cache_only)
529 return 0;
530
531 if (cache_only) {
532 page = f2fs_grab_cache_page(mapping, index, false);
533 if (page && PageUptodate(page))
534 goto truncate_out;
535 f2fs_put_page(page, 1);
536 return 0;
537 }
538
539 page = get_lock_data_page(inode, index, true);
540 if (IS_ERR(page))
541 return 0;
542 truncate_out:
543 f2fs_wait_on_page_writeback(page, DATA, true);
544 zero_user(page, offset, PAGE_SIZE - offset);
545 if (!cache_only || !f2fs_encrypted_inode(inode) ||
546 !S_ISREG(inode->i_mode))
547 set_page_dirty(page);
548 f2fs_put_page(page, 1);
549 return 0;
550 }
551
552 int truncate_blocks(struct inode *inode, u64 from, bool lock)
553 {
554 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
555 unsigned int blocksize = inode->i_sb->s_blocksize;
556 struct dnode_of_data dn;
557 pgoff_t free_from;
558 int count = 0, err = 0;
559 struct page *ipage;
560 bool truncate_page = false;
561
562 trace_f2fs_truncate_blocks_enter(inode, from);
563
564 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
565
566 if (lock)
567 f2fs_lock_op(sbi);
568
569 ipage = get_node_page(sbi, inode->i_ino);
570 if (IS_ERR(ipage)) {
571 err = PTR_ERR(ipage);
572 goto out;
573 }
574
575 if (f2fs_has_inline_data(inode)) {
576 if (truncate_inline_inode(ipage, from))
577 set_page_dirty(ipage);
578 f2fs_put_page(ipage, 1);
579 truncate_page = true;
580 goto out;
581 }
582
583 set_new_dnode(&dn, inode, ipage, NULL, 0);
584 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
585 if (err) {
586 if (err == -ENOENT)
587 goto free_next;
588 goto out;
589 }
590
591 count = ADDRS_PER_PAGE(dn.node_page, inode);
592
593 count -= dn.ofs_in_node;
594 f2fs_bug_on(sbi, count < 0);
595
596 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
597 truncate_data_blocks_range(&dn, count);
598 free_from += count;
599 }
600
601 f2fs_put_dnode(&dn);
602 free_next:
603 err = truncate_inode_blocks(inode, free_from);
604 out:
605 if (lock)
606 f2fs_unlock_op(sbi);
607
608 /* lastly zero out the first data page */
609 if (!err)
610 err = truncate_partial_data_page(inode, from, truncate_page);
611
612 trace_f2fs_truncate_blocks_exit(inode, err);
613 return err;
614 }
615
616 int f2fs_truncate(struct inode *inode, bool lock)
617 {
618 int err;
619
620 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
621 S_ISLNK(inode->i_mode)))
622 return 0;
623
624 trace_f2fs_truncate(inode);
625
626 /* we should check inline_data size */
627 if (!f2fs_may_inline_data(inode)) {
628 err = f2fs_convert_inline_inode(inode);
629 if (err)
630 return err;
631 }
632
633 err = truncate_blocks(inode, i_size_read(inode), lock);
634 if (err)
635 return err;
636
637 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
638 mark_inode_dirty(inode);
639 return 0;
640 }
641
642 int f2fs_getattr(struct vfsmount *mnt,
643 struct dentry *dentry, struct kstat *stat)
644 {
645 struct inode *inode = d_inode(dentry);
646 generic_fillattr(inode, stat);
647 stat->blocks <<= 3;
648 return 0;
649 }
650
651 #ifdef CONFIG_F2FS_FS_POSIX_ACL
652 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
653 {
654 struct f2fs_inode_info *fi = F2FS_I(inode);
655 unsigned int ia_valid = attr->ia_valid;
656
657 if (ia_valid & ATTR_UID)
658 inode->i_uid = attr->ia_uid;
659 if (ia_valid & ATTR_GID)
660 inode->i_gid = attr->ia_gid;
661 if (ia_valid & ATTR_ATIME)
662 inode->i_atime = timespec_trunc(attr->ia_atime,
663 inode->i_sb->s_time_gran);
664 if (ia_valid & ATTR_MTIME)
665 inode->i_mtime = timespec_trunc(attr->ia_mtime,
666 inode->i_sb->s_time_gran);
667 if (ia_valid & ATTR_CTIME)
668 inode->i_ctime = timespec_trunc(attr->ia_ctime,
669 inode->i_sb->s_time_gran);
670 if (ia_valid & ATTR_MODE) {
671 umode_t mode = attr->ia_mode;
672
673 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
674 mode &= ~S_ISGID;
675 set_acl_inode(fi, mode);
676 }
677 }
678 #else
679 #define __setattr_copy setattr_copy
680 #endif
681
682 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
683 {
684 struct inode *inode = d_inode(dentry);
685 struct f2fs_inode_info *fi = F2FS_I(inode);
686 int err;
687
688 err = inode_change_ok(inode, attr);
689 if (err)
690 return err;
691
692 if (attr->ia_valid & ATTR_SIZE) {
693 if (f2fs_encrypted_inode(inode) &&
694 fscrypt_get_encryption_info(inode))
695 return -EACCES;
696
697 if (attr->ia_size <= i_size_read(inode)) {
698 truncate_setsize(inode, attr->ia_size);
699 err = f2fs_truncate(inode, true);
700 if (err)
701 return err;
702 f2fs_balance_fs(F2FS_I_SB(inode), true);
703 } else {
704 /*
705 * do not trim all blocks after i_size if target size is
706 * larger than i_size.
707 */
708 truncate_setsize(inode, attr->ia_size);
709
710 /* should convert inline inode here */
711 if (!f2fs_may_inline_data(inode)) {
712 err = f2fs_convert_inline_inode(inode);
713 if (err)
714 return err;
715 }
716 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
717 }
718 }
719
720 __setattr_copy(inode, attr);
721
722 if (attr->ia_valid & ATTR_MODE) {
723 err = posix_acl_chmod(inode, get_inode_mode(inode));
724 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
725 inode->i_mode = fi->i_acl_mode;
726 clear_inode_flag(fi, FI_ACL_MODE);
727 }
728 }
729
730 mark_inode_dirty(inode);
731 return err;
732 }
733
734 const struct inode_operations f2fs_file_inode_operations = {
735 .getattr = f2fs_getattr,
736 .setattr = f2fs_setattr,
737 .get_acl = f2fs_get_acl,
738 .set_acl = f2fs_set_acl,
739 #ifdef CONFIG_F2FS_FS_XATTR
740 .setxattr = generic_setxattr,
741 .getxattr = generic_getxattr,
742 .listxattr = f2fs_listxattr,
743 .removexattr = generic_removexattr,
744 #endif
745 .fiemap = f2fs_fiemap,
746 };
747
748 static int fill_zero(struct inode *inode, pgoff_t index,
749 loff_t start, loff_t len)
750 {
751 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
752 struct page *page;
753
754 if (!len)
755 return 0;
756
757 f2fs_balance_fs(sbi, true);
758
759 f2fs_lock_op(sbi);
760 page = get_new_data_page(inode, NULL, index, false);
761 f2fs_unlock_op(sbi);
762
763 if (IS_ERR(page))
764 return PTR_ERR(page);
765
766 f2fs_wait_on_page_writeback(page, DATA, true);
767 zero_user(page, start, len);
768 set_page_dirty(page);
769 f2fs_put_page(page, 1);
770 return 0;
771 }
772
773 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
774 {
775 int err;
776
777 while (pg_start < pg_end) {
778 struct dnode_of_data dn;
779 pgoff_t end_offset, count;
780
781 set_new_dnode(&dn, inode, NULL, NULL, 0);
782 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
783 if (err) {
784 if (err == -ENOENT) {
785 pg_start++;
786 continue;
787 }
788 return err;
789 }
790
791 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
792 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
793
794 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
795
796 truncate_data_blocks_range(&dn, count);
797 f2fs_put_dnode(&dn);
798
799 pg_start += count;
800 }
801 return 0;
802 }
803
804 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
805 {
806 pgoff_t pg_start, pg_end;
807 loff_t off_start, off_end;
808 int ret;
809
810 ret = f2fs_convert_inline_inode(inode);
811 if (ret)
812 return ret;
813
814 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
815 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
816
817 off_start = offset & (PAGE_SIZE - 1);
818 off_end = (offset + len) & (PAGE_SIZE - 1);
819
820 if (pg_start == pg_end) {
821 ret = fill_zero(inode, pg_start, off_start,
822 off_end - off_start);
823 if (ret)
824 return ret;
825 } else {
826 if (off_start) {
827 ret = fill_zero(inode, pg_start++, off_start,
828 PAGE_SIZE - off_start);
829 if (ret)
830 return ret;
831 }
832 if (off_end) {
833 ret = fill_zero(inode, pg_end, 0, off_end);
834 if (ret)
835 return ret;
836 }
837
838 if (pg_start < pg_end) {
839 struct address_space *mapping = inode->i_mapping;
840 loff_t blk_start, blk_end;
841 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
842
843 f2fs_balance_fs(sbi, true);
844
845 blk_start = (loff_t)pg_start << PAGE_SHIFT;
846 blk_end = (loff_t)pg_end << PAGE_SHIFT;
847 truncate_inode_pages_range(mapping, blk_start,
848 blk_end - 1);
849
850 f2fs_lock_op(sbi);
851 ret = truncate_hole(inode, pg_start, pg_end);
852 f2fs_unlock_op(sbi);
853 }
854 }
855
856 return ret;
857 }
858
859 static int __exchange_data_block(struct inode *inode, pgoff_t src,
860 pgoff_t dst, bool full)
861 {
862 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
863 struct dnode_of_data dn;
864 block_t new_addr;
865 bool do_replace = false;
866 int ret;
867
868 set_new_dnode(&dn, inode, NULL, NULL, 0);
869 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
870 if (ret && ret != -ENOENT) {
871 return ret;
872 } else if (ret == -ENOENT) {
873 new_addr = NULL_ADDR;
874 } else {
875 new_addr = dn.data_blkaddr;
876 if (!is_checkpointed_data(sbi, new_addr)) {
877 /* do not invalidate this block address */
878 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
879 do_replace = true;
880 }
881 f2fs_put_dnode(&dn);
882 }
883
884 if (new_addr == NULL_ADDR)
885 return full ? truncate_hole(inode, dst, dst + 1) : 0;
886
887 if (do_replace) {
888 struct page *ipage = get_node_page(sbi, inode->i_ino);
889 struct node_info ni;
890
891 if (IS_ERR(ipage)) {
892 ret = PTR_ERR(ipage);
893 goto err_out;
894 }
895
896 set_new_dnode(&dn, inode, ipage, NULL, 0);
897 ret = f2fs_reserve_block(&dn, dst);
898 if (ret)
899 goto err_out;
900
901 truncate_data_blocks_range(&dn, 1);
902
903 get_node_info(sbi, dn.nid, &ni);
904 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
905 ni.version, true, false);
906 f2fs_put_dnode(&dn);
907 } else {
908 struct page *psrc, *pdst;
909
910 psrc = get_lock_data_page(inode, src, true);
911 if (IS_ERR(psrc))
912 return PTR_ERR(psrc);
913 pdst = get_new_data_page(inode, NULL, dst, true);
914 if (IS_ERR(pdst)) {
915 f2fs_put_page(psrc, 1);
916 return PTR_ERR(pdst);
917 }
918 f2fs_copy_page(psrc, pdst);
919 set_page_dirty(pdst);
920 f2fs_put_page(pdst, 1);
921 f2fs_put_page(psrc, 1);
922
923 return truncate_hole(inode, src, src + 1);
924 }
925 return 0;
926
927 err_out:
928 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
929 f2fs_update_data_blkaddr(&dn, new_addr);
930 f2fs_put_dnode(&dn);
931 }
932 return ret;
933 }
934
935 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
936 {
937 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
938 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
939 int ret = 0;
940
941 for (; end < nrpages; start++, end++) {
942 f2fs_balance_fs(sbi, true);
943 f2fs_lock_op(sbi);
944 ret = __exchange_data_block(inode, end, start, true);
945 f2fs_unlock_op(sbi);
946 if (ret)
947 break;
948 }
949 return ret;
950 }
951
952 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
953 {
954 pgoff_t pg_start, pg_end;
955 loff_t new_size;
956 int ret;
957
958 if (offset + len >= i_size_read(inode))
959 return -EINVAL;
960
961 /* collapse range should be aligned to block size of f2fs. */
962 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
963 return -EINVAL;
964
965 ret = f2fs_convert_inline_inode(inode);
966 if (ret)
967 return ret;
968
969 pg_start = offset >> PAGE_SHIFT;
970 pg_end = (offset + len) >> PAGE_SHIFT;
971
972 /* write out all dirty pages from offset */
973 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
974 if (ret)
975 return ret;
976
977 truncate_pagecache(inode, offset);
978
979 ret = f2fs_do_collapse(inode, pg_start, pg_end);
980 if (ret)
981 return ret;
982
983 /* write out all moved pages, if possible */
984 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
985 truncate_pagecache(inode, offset);
986
987 new_size = i_size_read(inode) - len;
988 truncate_pagecache(inode, new_size);
989
990 ret = truncate_blocks(inode, new_size, true);
991 if (!ret)
992 i_size_write(inode, new_size);
993
994 return ret;
995 }
996
997 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
998 int mode)
999 {
1000 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1001 struct address_space *mapping = inode->i_mapping;
1002 pgoff_t index, pg_start, pg_end;
1003 loff_t new_size = i_size_read(inode);
1004 loff_t off_start, off_end;
1005 int ret = 0;
1006
1007 ret = inode_newsize_ok(inode, (len + offset));
1008 if (ret)
1009 return ret;
1010
1011 ret = f2fs_convert_inline_inode(inode);
1012 if (ret)
1013 return ret;
1014
1015 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1016 if (ret)
1017 return ret;
1018
1019 truncate_pagecache_range(inode, offset, offset + len - 1);
1020
1021 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1022 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1023
1024 off_start = offset & (PAGE_SIZE - 1);
1025 off_end = (offset + len) & (PAGE_SIZE - 1);
1026
1027 if (pg_start == pg_end) {
1028 ret = fill_zero(inode, pg_start, off_start,
1029 off_end - off_start);
1030 if (ret)
1031 return ret;
1032
1033 if (offset + len > new_size)
1034 new_size = offset + len;
1035 new_size = max_t(loff_t, new_size, offset + len);
1036 } else {
1037 if (off_start) {
1038 ret = fill_zero(inode, pg_start++, off_start,
1039 PAGE_SIZE - off_start);
1040 if (ret)
1041 return ret;
1042
1043 new_size = max_t(loff_t, new_size,
1044 (loff_t)pg_start << PAGE_SHIFT);
1045 }
1046
1047 for (index = pg_start; index < pg_end; index++) {
1048 struct dnode_of_data dn;
1049 struct page *ipage;
1050
1051 f2fs_lock_op(sbi);
1052
1053 ipage = get_node_page(sbi, inode->i_ino);
1054 if (IS_ERR(ipage)) {
1055 ret = PTR_ERR(ipage);
1056 f2fs_unlock_op(sbi);
1057 goto out;
1058 }
1059
1060 set_new_dnode(&dn, inode, ipage, NULL, 0);
1061 ret = f2fs_reserve_block(&dn, index);
1062 if (ret) {
1063 f2fs_unlock_op(sbi);
1064 goto out;
1065 }
1066
1067 if (dn.data_blkaddr != NEW_ADDR) {
1068 invalidate_blocks(sbi, dn.data_blkaddr);
1069 f2fs_update_data_blkaddr(&dn, NEW_ADDR);
1070 }
1071 f2fs_put_dnode(&dn);
1072 f2fs_unlock_op(sbi);
1073
1074 new_size = max_t(loff_t, new_size,
1075 (loff_t)(index + 1) << PAGE_SHIFT);
1076 }
1077
1078 if (off_end) {
1079 ret = fill_zero(inode, pg_end, 0, off_end);
1080 if (ret)
1081 goto out;
1082
1083 new_size = max_t(loff_t, new_size, offset + len);
1084 }
1085 }
1086
1087 out:
1088 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1089 i_size_write(inode, new_size);
1090 mark_inode_dirty(inode);
1091 update_inode_page(inode);
1092 }
1093
1094 return ret;
1095 }
1096
1097 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1098 {
1099 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1100 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1101 loff_t new_size;
1102 int ret = 0;
1103
1104 new_size = i_size_read(inode) + len;
1105 if (new_size > inode->i_sb->s_maxbytes)
1106 return -EFBIG;
1107
1108 if (offset >= i_size_read(inode))
1109 return -EINVAL;
1110
1111 /* insert range should be aligned to block size of f2fs. */
1112 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1113 return -EINVAL;
1114
1115 ret = f2fs_convert_inline_inode(inode);
1116 if (ret)
1117 return ret;
1118
1119 f2fs_balance_fs(sbi, true);
1120
1121 ret = truncate_blocks(inode, i_size_read(inode), true);
1122 if (ret)
1123 return ret;
1124
1125 /* write out all dirty pages from offset */
1126 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1127 if (ret)
1128 return ret;
1129
1130 truncate_pagecache(inode, offset);
1131
1132 pg_start = offset >> PAGE_SHIFT;
1133 pg_end = (offset + len) >> PAGE_SHIFT;
1134 delta = pg_end - pg_start;
1135 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1136
1137 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1138 f2fs_lock_op(sbi);
1139 ret = __exchange_data_block(inode, idx, idx + delta, false);
1140 f2fs_unlock_op(sbi);
1141 if (ret)
1142 break;
1143 }
1144
1145 /* write out all moved pages, if possible */
1146 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1147 truncate_pagecache(inode, offset);
1148
1149 if (!ret)
1150 i_size_write(inode, new_size);
1151 return ret;
1152 }
1153
1154 static int expand_inode_data(struct inode *inode, loff_t offset,
1155 loff_t len, int mode)
1156 {
1157 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1158 pgoff_t index, pg_start, pg_end;
1159 loff_t new_size = i_size_read(inode);
1160 loff_t off_start, off_end;
1161 int ret = 0;
1162
1163 ret = inode_newsize_ok(inode, (len + offset));
1164 if (ret)
1165 return ret;
1166
1167 ret = f2fs_convert_inline_inode(inode);
1168 if (ret)
1169 return ret;
1170
1171 f2fs_balance_fs(sbi, true);
1172
1173 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1174 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1175
1176 off_start = offset & (PAGE_SIZE - 1);
1177 off_end = (offset + len) & (PAGE_SIZE - 1);
1178
1179 f2fs_lock_op(sbi);
1180
1181 for (index = pg_start; index <= pg_end; index++) {
1182 struct dnode_of_data dn;
1183
1184 if (index == pg_end && !off_end)
1185 goto noalloc;
1186
1187 set_new_dnode(&dn, inode, NULL, NULL, 0);
1188 ret = f2fs_reserve_block(&dn, index);
1189 if (ret)
1190 break;
1191 noalloc:
1192 if (pg_start == pg_end)
1193 new_size = offset + len;
1194 else if (index == pg_start && off_start)
1195 new_size = (loff_t)(index + 1) << PAGE_SHIFT;
1196 else if (index == pg_end)
1197 new_size = ((loff_t)index << PAGE_SHIFT) +
1198 off_end;
1199 else
1200 new_size += PAGE_SIZE;
1201 }
1202
1203 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1204 i_size_read(inode) < new_size) {
1205 i_size_write(inode, new_size);
1206 mark_inode_dirty(inode);
1207 update_inode_page(inode);
1208 }
1209 f2fs_unlock_op(sbi);
1210
1211 return ret;
1212 }
1213
1214 static long f2fs_fallocate(struct file *file, int mode,
1215 loff_t offset, loff_t len)
1216 {
1217 struct inode *inode = file_inode(file);
1218 long ret = 0;
1219
1220 /* f2fs only support ->fallocate for regular file */
1221 if (!S_ISREG(inode->i_mode))
1222 return -EINVAL;
1223
1224 if (f2fs_encrypted_inode(inode) &&
1225 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1226 return -EOPNOTSUPP;
1227
1228 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1229 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1230 FALLOC_FL_INSERT_RANGE))
1231 return -EOPNOTSUPP;
1232
1233 inode_lock(inode);
1234
1235 if (mode & FALLOC_FL_PUNCH_HOLE) {
1236 if (offset >= inode->i_size)
1237 goto out;
1238
1239 ret = punch_hole(inode, offset, len);
1240 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1241 ret = f2fs_collapse_range(inode, offset, len);
1242 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1243 ret = f2fs_zero_range(inode, offset, len, mode);
1244 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1245 ret = f2fs_insert_range(inode, offset, len);
1246 } else {
1247 ret = expand_inode_data(inode, offset, len, mode);
1248 }
1249
1250 if (!ret) {
1251 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1252 mark_inode_dirty(inode);
1253 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1254 }
1255
1256 out:
1257 inode_unlock(inode);
1258
1259 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1260 return ret;
1261 }
1262
1263 static int f2fs_release_file(struct inode *inode, struct file *filp)
1264 {
1265 /*
1266 * f2fs_relase_file is called at every close calls. So we should
1267 * not drop any inmemory pages by close called by other process.
1268 */
1269 if (!(filp->f_mode & FMODE_WRITE) ||
1270 atomic_read(&inode->i_writecount) != 1)
1271 return 0;
1272
1273 /* some remained atomic pages should discarded */
1274 if (f2fs_is_atomic_file(inode))
1275 drop_inmem_pages(inode);
1276 if (f2fs_is_volatile_file(inode)) {
1277 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1278 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1279 filemap_fdatawrite(inode->i_mapping);
1280 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1281 }
1282 return 0;
1283 }
1284
1285 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1286 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1287
1288 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1289 {
1290 if (S_ISDIR(mode))
1291 return flags;
1292 else if (S_ISREG(mode))
1293 return flags & F2FS_REG_FLMASK;
1294 else
1295 return flags & F2FS_OTHER_FLMASK;
1296 }
1297
1298 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1299 {
1300 struct inode *inode = file_inode(filp);
1301 struct f2fs_inode_info *fi = F2FS_I(inode);
1302 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1303 return put_user(flags, (int __user *)arg);
1304 }
1305
1306 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1307 {
1308 struct inode *inode = file_inode(filp);
1309 struct f2fs_inode_info *fi = F2FS_I(inode);
1310 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1311 unsigned int oldflags;
1312 int ret;
1313
1314 ret = mnt_want_write_file(filp);
1315 if (ret)
1316 return ret;
1317
1318 if (!inode_owner_or_capable(inode)) {
1319 ret = -EACCES;
1320 goto out;
1321 }
1322
1323 if (get_user(flags, (int __user *)arg)) {
1324 ret = -EFAULT;
1325 goto out;
1326 }
1327
1328 flags = f2fs_mask_flags(inode->i_mode, flags);
1329
1330 inode_lock(inode);
1331
1332 oldflags = fi->i_flags;
1333
1334 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1335 if (!capable(CAP_LINUX_IMMUTABLE)) {
1336 inode_unlock(inode);
1337 ret = -EPERM;
1338 goto out;
1339 }
1340 }
1341
1342 flags = flags & FS_FL_USER_MODIFIABLE;
1343 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1344 fi->i_flags = flags;
1345 inode_unlock(inode);
1346
1347 f2fs_set_inode_flags(inode);
1348 inode->i_ctime = CURRENT_TIME;
1349 mark_inode_dirty(inode);
1350 out:
1351 mnt_drop_write_file(filp);
1352 return ret;
1353 }
1354
1355 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1356 {
1357 struct inode *inode = file_inode(filp);
1358
1359 return put_user(inode->i_generation, (int __user *)arg);
1360 }
1361
1362 static int f2fs_ioc_start_atomic_write(struct file *filp)
1363 {
1364 struct inode *inode = file_inode(filp);
1365 int ret;
1366
1367 if (!inode_owner_or_capable(inode))
1368 return -EACCES;
1369
1370 if (f2fs_is_atomic_file(inode))
1371 return 0;
1372
1373 ret = f2fs_convert_inline_inode(inode);
1374 if (ret)
1375 return ret;
1376
1377 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1378 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1379
1380 if (!get_dirty_pages(inode))
1381 return 0;
1382
1383 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1384 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1385 inode->i_ino, get_dirty_pages(inode));
1386 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1387 if (ret)
1388 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1389 return ret;
1390 }
1391
1392 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1393 {
1394 struct inode *inode = file_inode(filp);
1395 int ret;
1396
1397 if (!inode_owner_or_capable(inode))
1398 return -EACCES;
1399
1400 if (f2fs_is_volatile_file(inode))
1401 return 0;
1402
1403 ret = mnt_want_write_file(filp);
1404 if (ret)
1405 return ret;
1406
1407 if (f2fs_is_atomic_file(inode)) {
1408 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1409 ret = commit_inmem_pages(inode);
1410 if (ret) {
1411 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1412 goto err_out;
1413 }
1414 }
1415
1416 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1417 err_out:
1418 mnt_drop_write_file(filp);
1419 return ret;
1420 }
1421
1422 static int f2fs_ioc_start_volatile_write(struct file *filp)
1423 {
1424 struct inode *inode = file_inode(filp);
1425 int ret;
1426
1427 if (!inode_owner_or_capable(inode))
1428 return -EACCES;
1429
1430 if (f2fs_is_volatile_file(inode))
1431 return 0;
1432
1433 ret = f2fs_convert_inline_inode(inode);
1434 if (ret)
1435 return ret;
1436
1437 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1438 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1439 return 0;
1440 }
1441
1442 static int f2fs_ioc_release_volatile_write(struct file *filp)
1443 {
1444 struct inode *inode = file_inode(filp);
1445
1446 if (!inode_owner_or_capable(inode))
1447 return -EACCES;
1448
1449 if (!f2fs_is_volatile_file(inode))
1450 return 0;
1451
1452 if (!f2fs_is_first_block_written(inode))
1453 return truncate_partial_data_page(inode, 0, true);
1454
1455 return punch_hole(inode, 0, F2FS_BLKSIZE);
1456 }
1457
1458 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1459 {
1460 struct inode *inode = file_inode(filp);
1461 int ret;
1462
1463 if (!inode_owner_or_capable(inode))
1464 return -EACCES;
1465
1466 ret = mnt_want_write_file(filp);
1467 if (ret)
1468 return ret;
1469
1470 if (f2fs_is_atomic_file(inode))
1471 drop_inmem_pages(inode);
1472 if (f2fs_is_volatile_file(inode)) {
1473 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1474 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1475 }
1476
1477 mnt_drop_write_file(filp);
1478 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1479 return ret;
1480 }
1481
1482 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1483 {
1484 struct inode *inode = file_inode(filp);
1485 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1486 struct super_block *sb = sbi->sb;
1487 __u32 in;
1488
1489 if (!capable(CAP_SYS_ADMIN))
1490 return -EPERM;
1491
1492 if (get_user(in, (__u32 __user *)arg))
1493 return -EFAULT;
1494
1495 switch (in) {
1496 case F2FS_GOING_DOWN_FULLSYNC:
1497 sb = freeze_bdev(sb->s_bdev);
1498 if (sb && !IS_ERR(sb)) {
1499 f2fs_stop_checkpoint(sbi);
1500 thaw_bdev(sb->s_bdev, sb);
1501 }
1502 break;
1503 case F2FS_GOING_DOWN_METASYNC:
1504 /* do checkpoint only */
1505 f2fs_sync_fs(sb, 1);
1506 f2fs_stop_checkpoint(sbi);
1507 break;
1508 case F2FS_GOING_DOWN_NOSYNC:
1509 f2fs_stop_checkpoint(sbi);
1510 break;
1511 case F2FS_GOING_DOWN_METAFLUSH:
1512 sync_meta_pages(sbi, META, LONG_MAX);
1513 f2fs_stop_checkpoint(sbi);
1514 break;
1515 default:
1516 return -EINVAL;
1517 }
1518 f2fs_update_time(sbi, REQ_TIME);
1519 return 0;
1520 }
1521
1522 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1523 {
1524 struct inode *inode = file_inode(filp);
1525 struct super_block *sb = inode->i_sb;
1526 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1527 struct fstrim_range range;
1528 int ret;
1529
1530 if (!capable(CAP_SYS_ADMIN))
1531 return -EPERM;
1532
1533 if (!blk_queue_discard(q))
1534 return -EOPNOTSUPP;
1535
1536 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1537 sizeof(range)))
1538 return -EFAULT;
1539
1540 range.minlen = max((unsigned int)range.minlen,
1541 q->limits.discard_granularity);
1542 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1543 if (ret < 0)
1544 return ret;
1545
1546 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1547 sizeof(range)))
1548 return -EFAULT;
1549 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1550 return 0;
1551 }
1552
1553 static bool uuid_is_nonzero(__u8 u[16])
1554 {
1555 int i;
1556
1557 for (i = 0; i < 16; i++)
1558 if (u[i])
1559 return true;
1560 return false;
1561 }
1562
1563 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1564 {
1565 struct fscrypt_policy policy;
1566 struct inode *inode = file_inode(filp);
1567
1568 if (copy_from_user(&policy, (struct fscrypt_policy __user *)arg,
1569 sizeof(policy)))
1570 return -EFAULT;
1571
1572 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1573 return fscrypt_process_policy(inode, &policy);
1574 }
1575
1576 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1577 {
1578 struct fscrypt_policy policy;
1579 struct inode *inode = file_inode(filp);
1580 int err;
1581
1582 err = fscrypt_get_policy(inode, &policy);
1583 if (err)
1584 return err;
1585
1586 if (copy_to_user((struct fscrypt_policy __user *)arg, &policy, sizeof(policy)))
1587 return -EFAULT;
1588 return 0;
1589 }
1590
1591 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1592 {
1593 struct inode *inode = file_inode(filp);
1594 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1595 int err;
1596
1597 if (!f2fs_sb_has_crypto(inode->i_sb))
1598 return -EOPNOTSUPP;
1599
1600 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1601 goto got_it;
1602
1603 err = mnt_want_write_file(filp);
1604 if (err)
1605 return err;
1606
1607 /* update superblock with uuid */
1608 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1609
1610 err = f2fs_commit_super(sbi, false);
1611 if (err) {
1612 /* undo new data */
1613 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1614 mnt_drop_write_file(filp);
1615 return err;
1616 }
1617 mnt_drop_write_file(filp);
1618 got_it:
1619 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1620 16))
1621 return -EFAULT;
1622 return 0;
1623 }
1624
1625 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1626 {
1627 struct inode *inode = file_inode(filp);
1628 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1629 __u32 sync;
1630
1631 if (!capable(CAP_SYS_ADMIN))
1632 return -EPERM;
1633
1634 if (get_user(sync, (__u32 __user *)arg))
1635 return -EFAULT;
1636
1637 if (f2fs_readonly(sbi->sb))
1638 return -EROFS;
1639
1640 if (!sync) {
1641 if (!mutex_trylock(&sbi->gc_mutex))
1642 return -EBUSY;
1643 } else {
1644 mutex_lock(&sbi->gc_mutex);
1645 }
1646
1647 return f2fs_gc(sbi, sync);
1648 }
1649
1650 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1651 {
1652 struct inode *inode = file_inode(filp);
1653 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1654
1655 if (!capable(CAP_SYS_ADMIN))
1656 return -EPERM;
1657
1658 if (f2fs_readonly(sbi->sb))
1659 return -EROFS;
1660
1661 return f2fs_sync_fs(sbi->sb, 1);
1662 }
1663
1664 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1665 struct file *filp,
1666 struct f2fs_defragment *range)
1667 {
1668 struct inode *inode = file_inode(filp);
1669 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1670 struct extent_info ei;
1671 pgoff_t pg_start, pg_end;
1672 unsigned int blk_per_seg = sbi->blocks_per_seg;
1673 unsigned int total = 0, sec_num;
1674 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1675 block_t blk_end = 0;
1676 bool fragmented = false;
1677 int err;
1678
1679 /* if in-place-update policy is enabled, don't waste time here */
1680 if (need_inplace_update(inode))
1681 return -EINVAL;
1682
1683 pg_start = range->start >> PAGE_SHIFT;
1684 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1685
1686 f2fs_balance_fs(sbi, true);
1687
1688 inode_lock(inode);
1689
1690 /* writeback all dirty pages in the range */
1691 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1692 range->start + range->len - 1);
1693 if (err)
1694 goto out;
1695
1696 /*
1697 * lookup mapping info in extent cache, skip defragmenting if physical
1698 * block addresses are continuous.
1699 */
1700 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1701 if (ei.fofs + ei.len >= pg_end)
1702 goto out;
1703 }
1704
1705 map.m_lblk = pg_start;
1706
1707 /*
1708 * lookup mapping info in dnode page cache, skip defragmenting if all
1709 * physical block addresses are continuous even if there are hole(s)
1710 * in logical blocks.
1711 */
1712 while (map.m_lblk < pg_end) {
1713 map.m_len = pg_end - map.m_lblk;
1714 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1715 if (err)
1716 goto out;
1717
1718 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1719 map.m_lblk++;
1720 continue;
1721 }
1722
1723 if (blk_end && blk_end != map.m_pblk) {
1724 fragmented = true;
1725 break;
1726 }
1727 blk_end = map.m_pblk + map.m_len;
1728
1729 map.m_lblk += map.m_len;
1730 }
1731
1732 if (!fragmented)
1733 goto out;
1734
1735 map.m_lblk = pg_start;
1736 map.m_len = pg_end - pg_start;
1737
1738 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1739
1740 /*
1741 * make sure there are enough free section for LFS allocation, this can
1742 * avoid defragment running in SSR mode when free section are allocated
1743 * intensively
1744 */
1745 if (has_not_enough_free_secs(sbi, sec_num)) {
1746 err = -EAGAIN;
1747 goto out;
1748 }
1749
1750 while (map.m_lblk < pg_end) {
1751 pgoff_t idx;
1752 int cnt = 0;
1753
1754 do_map:
1755 map.m_len = pg_end - map.m_lblk;
1756 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1757 if (err)
1758 goto clear_out;
1759
1760 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1761 map.m_lblk++;
1762 continue;
1763 }
1764
1765 set_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1766
1767 idx = map.m_lblk;
1768 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1769 struct page *page;
1770
1771 page = get_lock_data_page(inode, idx, true);
1772 if (IS_ERR(page)) {
1773 err = PTR_ERR(page);
1774 goto clear_out;
1775 }
1776
1777 set_page_dirty(page);
1778 f2fs_put_page(page, 1);
1779
1780 idx++;
1781 cnt++;
1782 total++;
1783 }
1784
1785 map.m_lblk = idx;
1786
1787 if (idx < pg_end && cnt < blk_per_seg)
1788 goto do_map;
1789
1790 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1791
1792 err = filemap_fdatawrite(inode->i_mapping);
1793 if (err)
1794 goto out;
1795 }
1796 clear_out:
1797 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1798 out:
1799 inode_unlock(inode);
1800 if (!err)
1801 range->len = (u64)total << PAGE_SHIFT;
1802 return err;
1803 }
1804
1805 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1806 {
1807 struct inode *inode = file_inode(filp);
1808 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1809 struct f2fs_defragment range;
1810 int err;
1811
1812 if (!capable(CAP_SYS_ADMIN))
1813 return -EPERM;
1814
1815 if (!S_ISREG(inode->i_mode))
1816 return -EINVAL;
1817
1818 err = mnt_want_write_file(filp);
1819 if (err)
1820 return err;
1821
1822 if (f2fs_readonly(sbi->sb)) {
1823 err = -EROFS;
1824 goto out;
1825 }
1826
1827 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1828 sizeof(range))) {
1829 err = -EFAULT;
1830 goto out;
1831 }
1832
1833 /* verify alignment of offset & size */
1834 if (range.start & (F2FS_BLKSIZE - 1) ||
1835 range.len & (F2FS_BLKSIZE - 1)) {
1836 err = -EINVAL;
1837 goto out;
1838 }
1839
1840 err = f2fs_defragment_range(sbi, filp, &range);
1841 f2fs_update_time(sbi, REQ_TIME);
1842 if (err < 0)
1843 goto out;
1844
1845 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1846 sizeof(range)))
1847 err = -EFAULT;
1848 out:
1849 mnt_drop_write_file(filp);
1850 return err;
1851 }
1852
1853 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1854 {
1855 switch (cmd) {
1856 case F2FS_IOC_GETFLAGS:
1857 return f2fs_ioc_getflags(filp, arg);
1858 case F2FS_IOC_SETFLAGS:
1859 return f2fs_ioc_setflags(filp, arg);
1860 case F2FS_IOC_GETVERSION:
1861 return f2fs_ioc_getversion(filp, arg);
1862 case F2FS_IOC_START_ATOMIC_WRITE:
1863 return f2fs_ioc_start_atomic_write(filp);
1864 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1865 return f2fs_ioc_commit_atomic_write(filp);
1866 case F2FS_IOC_START_VOLATILE_WRITE:
1867 return f2fs_ioc_start_volatile_write(filp);
1868 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1869 return f2fs_ioc_release_volatile_write(filp);
1870 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1871 return f2fs_ioc_abort_volatile_write(filp);
1872 case F2FS_IOC_SHUTDOWN:
1873 return f2fs_ioc_shutdown(filp, arg);
1874 case FITRIM:
1875 return f2fs_ioc_fitrim(filp, arg);
1876 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1877 return f2fs_ioc_set_encryption_policy(filp, arg);
1878 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1879 return f2fs_ioc_get_encryption_policy(filp, arg);
1880 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1881 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1882 case F2FS_IOC_GARBAGE_COLLECT:
1883 return f2fs_ioc_gc(filp, arg);
1884 case F2FS_IOC_WRITE_CHECKPOINT:
1885 return f2fs_ioc_write_checkpoint(filp, arg);
1886 case F2FS_IOC_DEFRAGMENT:
1887 return f2fs_ioc_defragment(filp, arg);
1888 default:
1889 return -ENOTTY;
1890 }
1891 }
1892
1893 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1894 {
1895 struct file *file = iocb->ki_filp;
1896 struct inode *inode = file_inode(file);
1897 ssize_t ret;
1898
1899 if (f2fs_encrypted_inode(inode) &&
1900 !fscrypt_has_encryption_key(inode) &&
1901 fscrypt_get_encryption_info(inode))
1902 return -EACCES;
1903
1904 inode_lock(inode);
1905 ret = generic_write_checks(iocb, from);
1906 if (ret > 0) {
1907 ret = f2fs_preallocate_blocks(iocb, from);
1908 if (!ret)
1909 ret = __generic_file_write_iter(iocb, from);
1910 }
1911 inode_unlock(inode);
1912
1913 if (ret > 0) {
1914 ssize_t err;
1915
1916 err = generic_write_sync(file, iocb->ki_pos - ret, ret);
1917 if (err < 0)
1918 ret = err;
1919 }
1920 return ret;
1921 }
1922
1923 #ifdef CONFIG_COMPAT
1924 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1925 {
1926 switch (cmd) {
1927 case F2FS_IOC32_GETFLAGS:
1928 cmd = F2FS_IOC_GETFLAGS;
1929 break;
1930 case F2FS_IOC32_SETFLAGS:
1931 cmd = F2FS_IOC_SETFLAGS;
1932 break;
1933 case F2FS_IOC32_GETVERSION:
1934 cmd = F2FS_IOC_GETVERSION;
1935 break;
1936 case F2FS_IOC_START_ATOMIC_WRITE:
1937 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1938 case F2FS_IOC_START_VOLATILE_WRITE:
1939 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1940 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1941 case F2FS_IOC_SHUTDOWN:
1942 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1943 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1944 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1945 case F2FS_IOC_GARBAGE_COLLECT:
1946 case F2FS_IOC_WRITE_CHECKPOINT:
1947 case F2FS_IOC_DEFRAGMENT:
1948 break;
1949 default:
1950 return -ENOIOCTLCMD;
1951 }
1952 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1953 }
1954 #endif
1955
1956 const struct file_operations f2fs_file_operations = {
1957 .llseek = f2fs_llseek,
1958 .read_iter = generic_file_read_iter,
1959 .write_iter = f2fs_file_write_iter,
1960 .open = f2fs_file_open,
1961 .release = f2fs_release_file,
1962 .mmap = f2fs_file_mmap,
1963 .fsync = f2fs_sync_file,
1964 .fallocate = f2fs_fallocate,
1965 .unlocked_ioctl = f2fs_ioctl,
1966 #ifdef CONFIG_COMPAT
1967 .compat_ioctl = f2fs_compat_ioctl,
1968 #endif
1969 .splice_read = generic_file_splice_read,
1970 .splice_write = iter_file_splice_write,
1971 };
Linux kernel - Deliverables
This page took 0.075203 seconds and 5 git commands to generate.