projects / deliverable / linux.git / blob
? search:


0e2a2bde26351c26eec3de13d6b43a9ffbb17940
[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 = f2fs_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 = f2fs_get_encryption_info(inode);
447 if (ret)
448 return -EACCES;
449 if (!f2fs_encrypted_inode(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) || !S_ISREG(inode->i_mode))
530 set_page_dirty(page);
531 f2fs_put_page(page, 1);
532 return 0;
533 }
534
535 int truncate_blocks(struct inode *inode, u64 from, bool lock)
536 {
537 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
538 unsigned int blocksize = inode->i_sb->s_blocksize;
539 struct dnode_of_data dn;
540 pgoff_t free_from;
541 int count = 0, err = 0;
542 struct page *ipage;
543 bool truncate_page = false;
544
545 trace_f2fs_truncate_blocks_enter(inode, from);
546
547 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
548
549 if (lock)
550 f2fs_lock_op(sbi);
551
552 ipage = get_node_page(sbi, inode->i_ino);
553 if (IS_ERR(ipage)) {
554 err = PTR_ERR(ipage);
555 goto out;
556 }
557
558 if (f2fs_has_inline_data(inode)) {
559 if (truncate_inline_inode(ipage, from))
560 set_page_dirty(ipage);
561 f2fs_put_page(ipage, 1);
562 truncate_page = true;
563 goto out;
564 }
565
566 set_new_dnode(&dn, inode, ipage, NULL, 0);
567 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
568 if (err) {
569 if (err == -ENOENT)
570 goto free_next;
571 goto out;
572 }
573
574 count = ADDRS_PER_PAGE(dn.node_page, inode);
575
576 count -= dn.ofs_in_node;
577 f2fs_bug_on(sbi, count < 0);
578
579 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
580 truncate_data_blocks_range(&dn, count);
581 free_from += count;
582 }
583
584 f2fs_put_dnode(&dn);
585 free_next:
586 err = truncate_inode_blocks(inode, free_from);
587 out:
588 if (lock)
589 f2fs_unlock_op(sbi);
590
591 /* lastly zero out the first data page */
592 if (!err)
593 err = truncate_partial_data_page(inode, from, truncate_page);
594
595 trace_f2fs_truncate_blocks_exit(inode, err);
596 return err;
597 }
598
599 int f2fs_truncate(struct inode *inode, bool lock)
600 {
601 int err;
602
603 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
604 S_ISLNK(inode->i_mode)))
605 return 0;
606
607 trace_f2fs_truncate(inode);
608
609 /* we should check inline_data size */
610 if (!f2fs_may_inline_data(inode)) {
611 err = f2fs_convert_inline_inode(inode);
612 if (err)
613 return err;
614 }
615
616 err = truncate_blocks(inode, i_size_read(inode), lock);
617 if (err)
618 return err;
619
620 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
621 mark_inode_dirty(inode);
622 return 0;
623 }
624
625 int f2fs_getattr(struct vfsmount *mnt,
626 struct dentry *dentry, struct kstat *stat)
627 {
628 struct inode *inode = d_inode(dentry);
629 generic_fillattr(inode, stat);
630 stat->blocks <<= 3;
631 return 0;
632 }
633
634 #ifdef CONFIG_F2FS_FS_POSIX_ACL
635 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
636 {
637 struct f2fs_inode_info *fi = F2FS_I(inode);
638 unsigned int ia_valid = attr->ia_valid;
639
640 if (ia_valid & ATTR_UID)
641 inode->i_uid = attr->ia_uid;
642 if (ia_valid & ATTR_GID)
643 inode->i_gid = attr->ia_gid;
644 if (ia_valid & ATTR_ATIME)
645 inode->i_atime = timespec_trunc(attr->ia_atime,
646 inode->i_sb->s_time_gran);
647 if (ia_valid & ATTR_MTIME)
648 inode->i_mtime = timespec_trunc(attr->ia_mtime,
649 inode->i_sb->s_time_gran);
650 if (ia_valid & ATTR_CTIME)
651 inode->i_ctime = timespec_trunc(attr->ia_ctime,
652 inode->i_sb->s_time_gran);
653 if (ia_valid & ATTR_MODE) {
654 umode_t mode = attr->ia_mode;
655
656 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
657 mode &= ~S_ISGID;
658 set_acl_inode(fi, mode);
659 }
660 }
661 #else
662 #define __setattr_copy setattr_copy
663 #endif
664
665 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
666 {
667 struct inode *inode = d_inode(dentry);
668 struct f2fs_inode_info *fi = F2FS_I(inode);
669 int err;
670
671 err = inode_change_ok(inode, attr);
672 if (err)
673 return err;
674
675 if (attr->ia_valid & ATTR_SIZE) {
676 if (f2fs_encrypted_inode(inode) &&
677 f2fs_get_encryption_info(inode))
678 return -EACCES;
679
680 if (attr->ia_size <= i_size_read(inode)) {
681 truncate_setsize(inode, attr->ia_size);
682 err = f2fs_truncate(inode, true);
683 if (err)
684 return err;
685 f2fs_balance_fs(F2FS_I_SB(inode), true);
686 } else {
687 /*
688 * do not trim all blocks after i_size if target size is
689 * larger than i_size.
690 */
691 truncate_setsize(inode, attr->ia_size);
692
693 /* should convert inline inode here */
694 if (!f2fs_may_inline_data(inode)) {
695 err = f2fs_convert_inline_inode(inode);
696 if (err)
697 return err;
698 }
699 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
700 }
701 }
702
703 __setattr_copy(inode, attr);
704
705 if (attr->ia_valid & ATTR_MODE) {
706 err = posix_acl_chmod(inode, get_inode_mode(inode));
707 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
708 inode->i_mode = fi->i_acl_mode;
709 clear_inode_flag(fi, FI_ACL_MODE);
710 }
711 }
712
713 mark_inode_dirty(inode);
714 return err;
715 }
716
717 const struct inode_operations f2fs_file_inode_operations = {
718 .getattr = f2fs_getattr,
719 .setattr = f2fs_setattr,
720 .get_acl = f2fs_get_acl,
721 .set_acl = f2fs_set_acl,
722 #ifdef CONFIG_F2FS_FS_XATTR
723 .setxattr = generic_setxattr,
724 .getxattr = generic_getxattr,
725 .listxattr = f2fs_listxattr,
726 .removexattr = generic_removexattr,
727 #endif
728 .fiemap = f2fs_fiemap,
729 };
730
731 static int fill_zero(struct inode *inode, pgoff_t index,
732 loff_t start, loff_t len)
733 {
734 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
735 struct page *page;
736
737 if (!len)
738 return 0;
739
740 f2fs_balance_fs(sbi, true);
741
742 f2fs_lock_op(sbi);
743 page = get_new_data_page(inode, NULL, index, false);
744 f2fs_unlock_op(sbi);
745
746 if (IS_ERR(page))
747 return PTR_ERR(page);
748
749 f2fs_wait_on_page_writeback(page, DATA, true);
750 zero_user(page, start, len);
751 set_page_dirty(page);
752 f2fs_put_page(page, 1);
753 return 0;
754 }
755
756 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
757 {
758 int err;
759
760 while (pg_start < pg_end) {
761 struct dnode_of_data dn;
762 pgoff_t end_offset, count;
763
764 set_new_dnode(&dn, inode, NULL, NULL, 0);
765 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
766 if (err) {
767 if (err == -ENOENT) {
768 pg_start++;
769 continue;
770 }
771 return err;
772 }
773
774 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
775 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
776
777 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
778
779 truncate_data_blocks_range(&dn, count);
780 f2fs_put_dnode(&dn);
781
782 pg_start += count;
783 }
784 return 0;
785 }
786
787 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
788 {
789 pgoff_t pg_start, pg_end;
790 loff_t off_start, off_end;
791 int ret;
792
793 ret = f2fs_convert_inline_inode(inode);
794 if (ret)
795 return ret;
796
797 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
798 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
799
800 off_start = offset & (PAGE_CACHE_SIZE - 1);
801 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
802
803 if (pg_start == pg_end) {
804 ret = fill_zero(inode, pg_start, off_start,
805 off_end - off_start);
806 if (ret)
807 return ret;
808 } else {
809 if (off_start) {
810 ret = fill_zero(inode, pg_start++, off_start,
811 PAGE_CACHE_SIZE - off_start);
812 if (ret)
813 return ret;
814 }
815 if (off_end) {
816 ret = fill_zero(inode, pg_end, 0, off_end);
817 if (ret)
818 return ret;
819 }
820
821 if (pg_start < pg_end) {
822 struct address_space *mapping = inode->i_mapping;
823 loff_t blk_start, blk_end;
824 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
825
826 f2fs_balance_fs(sbi, true);
827
828 blk_start = (loff_t)pg_start << PAGE_CACHE_SHIFT;
829 blk_end = (loff_t)pg_end << PAGE_CACHE_SHIFT;
830 truncate_inode_pages_range(mapping, blk_start,
831 blk_end - 1);
832
833 f2fs_lock_op(sbi);
834 ret = truncate_hole(inode, pg_start, pg_end);
835 f2fs_unlock_op(sbi);
836 }
837 }
838
839 return ret;
840 }
841
842 static int __exchange_data_block(struct inode *inode, pgoff_t src,
843 pgoff_t dst, bool full)
844 {
845 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
846 struct dnode_of_data dn;
847 block_t new_addr;
848 bool do_replace = false;
849 int ret;
850
851 set_new_dnode(&dn, inode, NULL, NULL, 0);
852 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
853 if (ret && ret != -ENOENT) {
854 return ret;
855 } else if (ret == -ENOENT) {
856 new_addr = NULL_ADDR;
857 } else {
858 new_addr = dn.data_blkaddr;
859 if (!is_checkpointed_data(sbi, new_addr)) {
860 dn.data_blkaddr = NULL_ADDR;
861 /* do not invalidate this block address */
862 set_data_blkaddr(&dn);
863 f2fs_update_extent_cache(&dn);
864 do_replace = true;
865 }
866 f2fs_put_dnode(&dn);
867 }
868
869 if (new_addr == NULL_ADDR)
870 return full ? truncate_hole(inode, dst, dst + 1) : 0;
871
872 if (do_replace) {
873 struct page *ipage = get_node_page(sbi, inode->i_ino);
874 struct node_info ni;
875
876 if (IS_ERR(ipage)) {
877 ret = PTR_ERR(ipage);
878 goto err_out;
879 }
880
881 set_new_dnode(&dn, inode, ipage, NULL, 0);
882 ret = f2fs_reserve_block(&dn, dst);
883 if (ret)
884 goto err_out;
885
886 truncate_data_blocks_range(&dn, 1);
887
888 get_node_info(sbi, dn.nid, &ni);
889 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
890 ni.version, true, false);
891 f2fs_put_dnode(&dn);
892 } else {
893 struct page *psrc, *pdst;
894
895 psrc = get_lock_data_page(inode, src, true);
896 if (IS_ERR(psrc))
897 return PTR_ERR(psrc);
898 pdst = get_new_data_page(inode, NULL, dst, true);
899 if (IS_ERR(pdst)) {
900 f2fs_put_page(psrc, 1);
901 return PTR_ERR(pdst);
902 }
903 f2fs_copy_page(psrc, pdst);
904 set_page_dirty(pdst);
905 f2fs_put_page(pdst, 1);
906 f2fs_put_page(psrc, 1);
907
908 return truncate_hole(inode, src, src + 1);
909 }
910 return 0;
911
912 err_out:
913 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
914 dn.data_blkaddr = new_addr;
915 set_data_blkaddr(&dn);
916 f2fs_update_extent_cache(&dn);
917 f2fs_put_dnode(&dn);
918 }
919 return ret;
920 }
921
922 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
923 {
924 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
925 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
926 int ret = 0;
927
928 for (; end < nrpages; start++, end++) {
929 f2fs_balance_fs(sbi, true);
930 f2fs_lock_op(sbi);
931 ret = __exchange_data_block(inode, end, start, true);
932 f2fs_unlock_op(sbi);
933 if (ret)
934 break;
935 }
936 return ret;
937 }
938
939 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
940 {
941 pgoff_t pg_start, pg_end;
942 loff_t new_size;
943 int ret;
944
945 if (offset + len >= i_size_read(inode))
946 return -EINVAL;
947
948 /* collapse range should be aligned to block size of f2fs. */
949 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
950 return -EINVAL;
951
952 ret = f2fs_convert_inline_inode(inode);
953 if (ret)
954 return ret;
955
956 pg_start = offset >> PAGE_CACHE_SHIFT;
957 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
958
959 /* write out all dirty pages from offset */
960 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
961 if (ret)
962 return ret;
963
964 truncate_pagecache(inode, offset);
965
966 ret = f2fs_do_collapse(inode, pg_start, pg_end);
967 if (ret)
968 return ret;
969
970 /* write out all moved pages, if possible */
971 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
972 truncate_pagecache(inode, offset);
973
974 new_size = i_size_read(inode) - len;
975 truncate_pagecache(inode, new_size);
976
977 ret = truncate_blocks(inode, new_size, true);
978 if (!ret)
979 i_size_write(inode, new_size);
980
981 return ret;
982 }
983
984 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
985 int mode)
986 {
987 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
988 struct address_space *mapping = inode->i_mapping;
989 pgoff_t index, pg_start, pg_end;
990 loff_t new_size = i_size_read(inode);
991 loff_t off_start, off_end;
992 int ret = 0;
993
994 ret = inode_newsize_ok(inode, (len + offset));
995 if (ret)
996 return ret;
997
998 ret = f2fs_convert_inline_inode(inode);
999 if (ret)
1000 return ret;
1001
1002 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1003 if (ret)
1004 return ret;
1005
1006 truncate_pagecache_range(inode, offset, offset + len - 1);
1007
1008 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1009 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1010
1011 off_start = offset & (PAGE_CACHE_SIZE - 1);
1012 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1013
1014 if (pg_start == pg_end) {
1015 ret = fill_zero(inode, pg_start, off_start,
1016 off_end - off_start);
1017 if (ret)
1018 return ret;
1019
1020 if (offset + len > new_size)
1021 new_size = offset + len;
1022 new_size = max_t(loff_t, new_size, offset + len);
1023 } else {
1024 if (off_start) {
1025 ret = fill_zero(inode, pg_start++, off_start,
1026 PAGE_CACHE_SIZE - off_start);
1027 if (ret)
1028 return ret;
1029
1030 new_size = max_t(loff_t, new_size,
1031 (loff_t)pg_start << PAGE_CACHE_SHIFT);
1032 }
1033
1034 for (index = pg_start; index < pg_end; index++) {
1035 struct dnode_of_data dn;
1036 struct page *ipage;
1037
1038 f2fs_lock_op(sbi);
1039
1040 ipage = get_node_page(sbi, inode->i_ino);
1041 if (IS_ERR(ipage)) {
1042 ret = PTR_ERR(ipage);
1043 f2fs_unlock_op(sbi);
1044 goto out;
1045 }
1046
1047 set_new_dnode(&dn, inode, ipage, NULL, 0);
1048 ret = f2fs_reserve_block(&dn, index);
1049 if (ret) {
1050 f2fs_unlock_op(sbi);
1051 goto out;
1052 }
1053
1054 if (dn.data_blkaddr != NEW_ADDR) {
1055 invalidate_blocks(sbi, dn.data_blkaddr);
1056
1057 dn.data_blkaddr = NEW_ADDR;
1058 set_data_blkaddr(&dn);
1059
1060 dn.data_blkaddr = NULL_ADDR;
1061 f2fs_update_extent_cache(&dn);
1062 }
1063 f2fs_put_dnode(&dn);
1064 f2fs_unlock_op(sbi);
1065
1066 new_size = max_t(loff_t, new_size,
1067 (loff_t)(index + 1) << PAGE_CACHE_SHIFT);
1068 }
1069
1070 if (off_end) {
1071 ret = fill_zero(inode, pg_end, 0, off_end);
1072 if (ret)
1073 goto out;
1074
1075 new_size = max_t(loff_t, new_size, offset + len);
1076 }
1077 }
1078
1079 out:
1080 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1081 i_size_write(inode, new_size);
1082 mark_inode_dirty(inode);
1083 update_inode_page(inode);
1084 }
1085
1086 return ret;
1087 }
1088
1089 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1090 {
1091 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1092 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1093 loff_t new_size;
1094 int ret = 0;
1095
1096 new_size = i_size_read(inode) + len;
1097 if (new_size > inode->i_sb->s_maxbytes)
1098 return -EFBIG;
1099
1100 if (offset >= i_size_read(inode))
1101 return -EINVAL;
1102
1103 /* insert range should be aligned to block size of f2fs. */
1104 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1105 return -EINVAL;
1106
1107 ret = f2fs_convert_inline_inode(inode);
1108 if (ret)
1109 return ret;
1110
1111 f2fs_balance_fs(sbi, true);
1112
1113 ret = truncate_blocks(inode, i_size_read(inode), true);
1114 if (ret)
1115 return ret;
1116
1117 /* write out all dirty pages from offset */
1118 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1119 if (ret)
1120 return ret;
1121
1122 truncate_pagecache(inode, offset);
1123
1124 pg_start = offset >> PAGE_CACHE_SHIFT;
1125 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
1126 delta = pg_end - pg_start;
1127 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1128
1129 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1130 f2fs_lock_op(sbi);
1131 ret = __exchange_data_block(inode, idx, idx + delta, false);
1132 f2fs_unlock_op(sbi);
1133 if (ret)
1134 break;
1135 }
1136
1137 /* write out all moved pages, if possible */
1138 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1139 truncate_pagecache(inode, offset);
1140
1141 if (!ret)
1142 i_size_write(inode, new_size);
1143 return ret;
1144 }
1145
1146 static int expand_inode_data(struct inode *inode, loff_t offset,
1147 loff_t len, int mode)
1148 {
1149 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1150 pgoff_t index, pg_start, pg_end;
1151 loff_t new_size = i_size_read(inode);
1152 loff_t off_start, off_end;
1153 int ret = 0;
1154
1155 ret = inode_newsize_ok(inode, (len + offset));
1156 if (ret)
1157 return ret;
1158
1159 ret = f2fs_convert_inline_inode(inode);
1160 if (ret)
1161 return ret;
1162
1163 f2fs_balance_fs(sbi, true);
1164
1165 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1166 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1167
1168 off_start = offset & (PAGE_CACHE_SIZE - 1);
1169 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1170
1171 f2fs_lock_op(sbi);
1172
1173 for (index = pg_start; index <= pg_end; index++) {
1174 struct dnode_of_data dn;
1175
1176 if (index == pg_end && !off_end)
1177 goto noalloc;
1178
1179 set_new_dnode(&dn, inode, NULL, NULL, 0);
1180 ret = f2fs_reserve_block(&dn, index);
1181 if (ret)
1182 break;
1183 noalloc:
1184 if (pg_start == pg_end)
1185 new_size = offset + len;
1186 else if (index == pg_start && off_start)
1187 new_size = (loff_t)(index + 1) << PAGE_CACHE_SHIFT;
1188 else if (index == pg_end)
1189 new_size = ((loff_t)index << PAGE_CACHE_SHIFT) +
1190 off_end;
1191 else
1192 new_size += PAGE_CACHE_SIZE;
1193 }
1194
1195 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1196 i_size_read(inode) < new_size) {
1197 i_size_write(inode, new_size);
1198 mark_inode_dirty(inode);
1199 update_inode_page(inode);
1200 }
1201 f2fs_unlock_op(sbi);
1202
1203 return ret;
1204 }
1205
1206 static long f2fs_fallocate(struct file *file, int mode,
1207 loff_t offset, loff_t len)
1208 {
1209 struct inode *inode = file_inode(file);
1210 long ret = 0;
1211
1212 /* f2fs only support ->fallocate for regular file */
1213 if (!S_ISREG(inode->i_mode))
1214 return -EINVAL;
1215
1216 if (f2fs_encrypted_inode(inode) &&
1217 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1218 return -EOPNOTSUPP;
1219
1220 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1221 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1222 FALLOC_FL_INSERT_RANGE))
1223 return -EOPNOTSUPP;
1224
1225 inode_lock(inode);
1226
1227 if (mode & FALLOC_FL_PUNCH_HOLE) {
1228 if (offset >= inode->i_size)
1229 goto out;
1230
1231 ret = punch_hole(inode, offset, len);
1232 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1233 ret = f2fs_collapse_range(inode, offset, len);
1234 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1235 ret = f2fs_zero_range(inode, offset, len, mode);
1236 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1237 ret = f2fs_insert_range(inode, offset, len);
1238 } else {
1239 ret = expand_inode_data(inode, offset, len, mode);
1240 }
1241
1242 if (!ret) {
1243 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1244 mark_inode_dirty(inode);
1245 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1246 }
1247
1248 out:
1249 inode_unlock(inode);
1250
1251 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1252 return ret;
1253 }
1254
1255 static int f2fs_release_file(struct inode *inode, struct file *filp)
1256 {
1257 /* some remained atomic pages should discarded */
1258 if (f2fs_is_atomic_file(inode))
1259 drop_inmem_pages(inode);
1260 if (f2fs_is_volatile_file(inode)) {
1261 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1262 filemap_fdatawrite(inode->i_mapping);
1263 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1264 }
1265 return 0;
1266 }
1267
1268 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1269 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1270
1271 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1272 {
1273 if (S_ISDIR(mode))
1274 return flags;
1275 else if (S_ISREG(mode))
1276 return flags & F2FS_REG_FLMASK;
1277 else
1278 return flags & F2FS_OTHER_FLMASK;
1279 }
1280
1281 static int f2fs_ioc_getflags(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 return put_user(flags, (int __user *)arg);
1287 }
1288
1289 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1290 {
1291 struct inode *inode = file_inode(filp);
1292 struct f2fs_inode_info *fi = F2FS_I(inode);
1293 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1294 unsigned int oldflags;
1295 int ret;
1296
1297 ret = mnt_want_write_file(filp);
1298 if (ret)
1299 return ret;
1300
1301 if (!inode_owner_or_capable(inode)) {
1302 ret = -EACCES;
1303 goto out;
1304 }
1305
1306 if (get_user(flags, (int __user *)arg)) {
1307 ret = -EFAULT;
1308 goto out;
1309 }
1310
1311 flags = f2fs_mask_flags(inode->i_mode, flags);
1312
1313 inode_lock(inode);
1314
1315 oldflags = fi->i_flags;
1316
1317 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1318 if (!capable(CAP_LINUX_IMMUTABLE)) {
1319 inode_unlock(inode);
1320 ret = -EPERM;
1321 goto out;
1322 }
1323 }
1324
1325 flags = flags & FS_FL_USER_MODIFIABLE;
1326 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1327 fi->i_flags = flags;
1328 inode_unlock(inode);
1329
1330 f2fs_set_inode_flags(inode);
1331 inode->i_ctime = CURRENT_TIME;
1332 mark_inode_dirty(inode);
1333 out:
1334 mnt_drop_write_file(filp);
1335 return ret;
1336 }
1337
1338 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1339 {
1340 struct inode *inode = file_inode(filp);
1341
1342 return put_user(inode->i_generation, (int __user *)arg);
1343 }
1344
1345 static int f2fs_ioc_start_atomic_write(struct file *filp)
1346 {
1347 struct inode *inode = file_inode(filp);
1348 int ret;
1349
1350 if (!inode_owner_or_capable(inode))
1351 return -EACCES;
1352
1353 if (f2fs_is_atomic_file(inode))
1354 return 0;
1355
1356 ret = f2fs_convert_inline_inode(inode);
1357 if (ret)
1358 return ret;
1359
1360 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1361 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1362
1363 return 0;
1364 }
1365
1366 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1367 {
1368 struct inode *inode = file_inode(filp);
1369 int ret;
1370
1371 if (!inode_owner_or_capable(inode))
1372 return -EACCES;
1373
1374 if (f2fs_is_volatile_file(inode))
1375 return 0;
1376
1377 ret = mnt_want_write_file(filp);
1378 if (ret)
1379 return ret;
1380
1381 if (f2fs_is_atomic_file(inode)) {
1382 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1383 ret = commit_inmem_pages(inode);
1384 if (ret) {
1385 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1386 goto err_out;
1387 }
1388 }
1389
1390 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1391 err_out:
1392 mnt_drop_write_file(filp);
1393 return ret;
1394 }
1395
1396 static int f2fs_ioc_start_volatile_write(struct file *filp)
1397 {
1398 struct inode *inode = file_inode(filp);
1399 int ret;
1400
1401 if (!inode_owner_or_capable(inode))
1402 return -EACCES;
1403
1404 if (f2fs_is_volatile_file(inode))
1405 return 0;
1406
1407 ret = f2fs_convert_inline_inode(inode);
1408 if (ret)
1409 return ret;
1410
1411 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1412 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1413 return 0;
1414 }
1415
1416 static int f2fs_ioc_release_volatile_write(struct file *filp)
1417 {
1418 struct inode *inode = file_inode(filp);
1419
1420 if (!inode_owner_or_capable(inode))
1421 return -EACCES;
1422
1423 if (!f2fs_is_volatile_file(inode))
1424 return 0;
1425
1426 if (!f2fs_is_first_block_written(inode))
1427 return truncate_partial_data_page(inode, 0, true);
1428
1429 return punch_hole(inode, 0, F2FS_BLKSIZE);
1430 }
1431
1432 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1433 {
1434 struct inode *inode = file_inode(filp);
1435 int ret;
1436
1437 if (!inode_owner_or_capable(inode))
1438 return -EACCES;
1439
1440 ret = mnt_want_write_file(filp);
1441 if (ret)
1442 return ret;
1443
1444 if (f2fs_is_atomic_file(inode)) {
1445 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1446 drop_inmem_pages(inode);
1447 }
1448 if (f2fs_is_volatile_file(inode)) {
1449 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1450 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1451 }
1452
1453 mnt_drop_write_file(filp);
1454 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1455 return ret;
1456 }
1457
1458 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1459 {
1460 struct inode *inode = file_inode(filp);
1461 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1462 struct super_block *sb = sbi->sb;
1463 __u32 in;
1464
1465 if (!capable(CAP_SYS_ADMIN))
1466 return -EPERM;
1467
1468 if (get_user(in, (__u32 __user *)arg))
1469 return -EFAULT;
1470
1471 switch (in) {
1472 case F2FS_GOING_DOWN_FULLSYNC:
1473 sb = freeze_bdev(sb->s_bdev);
1474 if (sb && !IS_ERR(sb)) {
1475 f2fs_stop_checkpoint(sbi);
1476 thaw_bdev(sb->s_bdev, sb);
1477 }
1478 break;
1479 case F2FS_GOING_DOWN_METASYNC:
1480 /* do checkpoint only */
1481 f2fs_sync_fs(sb, 1);
1482 f2fs_stop_checkpoint(sbi);
1483 break;
1484 case F2FS_GOING_DOWN_NOSYNC:
1485 f2fs_stop_checkpoint(sbi);
1486 break;
1487 case F2FS_GOING_DOWN_METAFLUSH:
1488 sync_meta_pages(sbi, META, LONG_MAX);
1489 f2fs_stop_checkpoint(sbi);
1490 break;
1491 default:
1492 return -EINVAL;
1493 }
1494 f2fs_update_time(sbi, REQ_TIME);
1495 return 0;
1496 }
1497
1498 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1499 {
1500 struct inode *inode = file_inode(filp);
1501 struct super_block *sb = inode->i_sb;
1502 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1503 struct fstrim_range range;
1504 int ret;
1505
1506 if (!capable(CAP_SYS_ADMIN))
1507 return -EPERM;
1508
1509 if (!blk_queue_discard(q))
1510 return -EOPNOTSUPP;
1511
1512 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1513 sizeof(range)))
1514 return -EFAULT;
1515
1516 range.minlen = max((unsigned int)range.minlen,
1517 q->limits.discard_granularity);
1518 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1519 if (ret < 0)
1520 return ret;
1521
1522 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1523 sizeof(range)))
1524 return -EFAULT;
1525 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1526 return 0;
1527 }
1528
1529 static bool uuid_is_nonzero(__u8 u[16])
1530 {
1531 int i;
1532
1533 for (i = 0; i < 16; i++)
1534 if (u[i])
1535 return true;
1536 return false;
1537 }
1538
1539 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1540 {
1541 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1542 struct f2fs_encryption_policy policy;
1543 struct inode *inode = file_inode(filp);
1544
1545 if (copy_from_user(&policy, (struct f2fs_encryption_policy __user *)arg,
1546 sizeof(policy)))
1547 return -EFAULT;
1548
1549 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1550 return f2fs_process_policy(&policy, inode);
1551 #else
1552 return -EOPNOTSUPP;
1553 #endif
1554 }
1555
1556 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1557 {
1558 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1559 struct f2fs_encryption_policy policy;
1560 struct inode *inode = file_inode(filp);
1561 int err;
1562
1563 err = f2fs_get_policy(inode, &policy);
1564 if (err)
1565 return err;
1566
1567 if (copy_to_user((struct f2fs_encryption_policy __user *)arg, &policy,
1568 sizeof(policy)))
1569 return -EFAULT;
1570 return 0;
1571 #else
1572 return -EOPNOTSUPP;
1573 #endif
1574 }
1575
1576 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1577 {
1578 struct inode *inode = file_inode(filp);
1579 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1580 int err;
1581
1582 if (!f2fs_sb_has_crypto(inode->i_sb))
1583 return -EOPNOTSUPP;
1584
1585 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1586 goto got_it;
1587
1588 err = mnt_want_write_file(filp);
1589 if (err)
1590 return err;
1591
1592 /* update superblock with uuid */
1593 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1594
1595 err = f2fs_commit_super(sbi, false);
1596 if (err) {
1597 /* undo new data */
1598 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1599 mnt_drop_write_file(filp);
1600 return err;
1601 }
1602 mnt_drop_write_file(filp);
1603 got_it:
1604 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1605 16))
1606 return -EFAULT;
1607 return 0;
1608 }
1609
1610 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1611 {
1612 struct inode *inode = file_inode(filp);
1613 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1614 __u32 sync;
1615
1616 if (!capable(CAP_SYS_ADMIN))
1617 return -EPERM;
1618
1619 if (get_user(sync, (__u32 __user *)arg))
1620 return -EFAULT;
1621
1622 if (f2fs_readonly(sbi->sb))
1623 return -EROFS;
1624
1625 if (!sync) {
1626 if (!mutex_trylock(&sbi->gc_mutex))
1627 return -EBUSY;
1628 } else {
1629 mutex_lock(&sbi->gc_mutex);
1630 }
1631
1632 return f2fs_gc(sbi, sync);
1633 }
1634
1635 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1636 {
1637 struct inode *inode = file_inode(filp);
1638 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1639
1640 if (!capable(CAP_SYS_ADMIN))
1641 return -EPERM;
1642
1643 if (f2fs_readonly(sbi->sb))
1644 return -EROFS;
1645
1646 return f2fs_sync_fs(sbi->sb, 1);
1647 }
1648
1649 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1650 struct file *filp,
1651 struct f2fs_defragment *range)
1652 {
1653 struct inode *inode = file_inode(filp);
1654 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1655 struct extent_info ei;
1656 pgoff_t pg_start, pg_end;
1657 unsigned int blk_per_seg = sbi->blocks_per_seg;
1658 unsigned int total = 0, sec_num;
1659 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1660 block_t blk_end = 0;
1661 bool fragmented = false;
1662 int err;
1663
1664 /* if in-place-update policy is enabled, don't waste time here */
1665 if (need_inplace_update(inode))
1666 return -EINVAL;
1667
1668 pg_start = range->start >> PAGE_CACHE_SHIFT;
1669 pg_end = (range->start + range->len) >> PAGE_CACHE_SHIFT;
1670
1671 f2fs_balance_fs(sbi, true);
1672
1673 inode_lock(inode);
1674
1675 /* writeback all dirty pages in the range */
1676 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1677 range->start + range->len - 1);
1678 if (err)
1679 goto out;
1680
1681 /*
1682 * lookup mapping info in extent cache, skip defragmenting if physical
1683 * block addresses are continuous.
1684 */
1685 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1686 if (ei.fofs + ei.len >= pg_end)
1687 goto out;
1688 }
1689
1690 map.m_lblk = pg_start;
1691
1692 /*
1693 * lookup mapping info in dnode page cache, skip defragmenting if all
1694 * physical block addresses are continuous even if there are hole(s)
1695 * in logical blocks.
1696 */
1697 while (map.m_lblk < pg_end) {
1698 map.m_len = pg_end - map.m_lblk;
1699 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1700 if (err)
1701 goto out;
1702
1703 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1704 map.m_lblk++;
1705 continue;
1706 }
1707
1708 if (blk_end && blk_end != map.m_pblk) {
1709 fragmented = true;
1710 break;
1711 }
1712 blk_end = map.m_pblk + map.m_len;
1713
1714 map.m_lblk += map.m_len;
1715 }
1716
1717 if (!fragmented)
1718 goto out;
1719
1720 map.m_lblk = pg_start;
1721 map.m_len = pg_end - pg_start;
1722
1723 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1724
1725 /*
1726 * make sure there are enough free section for LFS allocation, this can
1727 * avoid defragment running in SSR mode when free section are allocated
1728 * intensively
1729 */
1730 if (has_not_enough_free_secs(sbi, sec_num)) {
1731 err = -EAGAIN;
1732 goto out;
1733 }
1734
1735 while (map.m_lblk < pg_end) {
1736 pgoff_t idx;
1737 int cnt = 0;
1738
1739 do_map:
1740 map.m_len = pg_end - map.m_lblk;
1741 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1742 if (err)
1743 goto clear_out;
1744
1745 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1746 map.m_lblk++;
1747 continue;
1748 }
1749
1750 set_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1751
1752 idx = map.m_lblk;
1753 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1754 struct page *page;
1755
1756 page = get_lock_data_page(inode, idx, true);
1757 if (IS_ERR(page)) {
1758 err = PTR_ERR(page);
1759 goto clear_out;
1760 }
1761
1762 set_page_dirty(page);
1763 f2fs_put_page(page, 1);
1764
1765 idx++;
1766 cnt++;
1767 total++;
1768 }
1769
1770 map.m_lblk = idx;
1771
1772 if (idx < pg_end && cnt < blk_per_seg)
1773 goto do_map;
1774
1775 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1776
1777 err = filemap_fdatawrite(inode->i_mapping);
1778 if (err)
1779 goto out;
1780 }
1781 clear_out:
1782 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1783 out:
1784 inode_unlock(inode);
1785 if (!err)
1786 range->len = (u64)total << PAGE_CACHE_SHIFT;
1787 return err;
1788 }
1789
1790 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1791 {
1792 struct inode *inode = file_inode(filp);
1793 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1794 struct f2fs_defragment range;
1795 int err;
1796
1797 if (!capable(CAP_SYS_ADMIN))
1798 return -EPERM;
1799
1800 if (!S_ISREG(inode->i_mode))
1801 return -EINVAL;
1802
1803 err = mnt_want_write_file(filp);
1804 if (err)
1805 return err;
1806
1807 if (f2fs_readonly(sbi->sb)) {
1808 err = -EROFS;
1809 goto out;
1810 }
1811
1812 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1813 sizeof(range))) {
1814 err = -EFAULT;
1815 goto out;
1816 }
1817
1818 /* verify alignment of offset & size */
1819 if (range.start & (F2FS_BLKSIZE - 1) ||
1820 range.len & (F2FS_BLKSIZE - 1)) {
1821 err = -EINVAL;
1822 goto out;
1823 }
1824
1825 err = f2fs_defragment_range(sbi, filp, &range);
1826 f2fs_update_time(sbi, REQ_TIME);
1827 if (err < 0)
1828 goto out;
1829
1830 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1831 sizeof(range)))
1832 err = -EFAULT;
1833 out:
1834 mnt_drop_write_file(filp);
1835 return err;
1836 }
1837
1838 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1839 {
1840 switch (cmd) {
1841 case F2FS_IOC_GETFLAGS:
1842 return f2fs_ioc_getflags(filp, arg);
1843 case F2FS_IOC_SETFLAGS:
1844 return f2fs_ioc_setflags(filp, arg);
1845 case F2FS_IOC_GETVERSION:
1846 return f2fs_ioc_getversion(filp, arg);
1847 case F2FS_IOC_START_ATOMIC_WRITE:
1848 return f2fs_ioc_start_atomic_write(filp);
1849 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1850 return f2fs_ioc_commit_atomic_write(filp);
1851 case F2FS_IOC_START_VOLATILE_WRITE:
1852 return f2fs_ioc_start_volatile_write(filp);
1853 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1854 return f2fs_ioc_release_volatile_write(filp);
1855 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1856 return f2fs_ioc_abort_volatile_write(filp);
1857 case F2FS_IOC_SHUTDOWN:
1858 return f2fs_ioc_shutdown(filp, arg);
1859 case FITRIM:
1860 return f2fs_ioc_fitrim(filp, arg);
1861 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1862 return f2fs_ioc_set_encryption_policy(filp, arg);
1863 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1864 return f2fs_ioc_get_encryption_policy(filp, arg);
1865 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1866 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1867 case F2FS_IOC_GARBAGE_COLLECT:
1868 return f2fs_ioc_gc(filp, arg);
1869 case F2FS_IOC_WRITE_CHECKPOINT:
1870 return f2fs_ioc_write_checkpoint(filp, arg);
1871 case F2FS_IOC_DEFRAGMENT:
1872 return f2fs_ioc_defragment(filp, arg);
1873 default:
1874 return -ENOTTY;
1875 }
1876 }
1877
1878 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1879 {
1880 struct file *file = iocb->ki_filp;
1881 struct inode *inode = file_inode(file);
1882 ssize_t ret;
1883
1884 if (f2fs_encrypted_inode(inode) &&
1885 !f2fs_has_encryption_key(inode) &&
1886 f2fs_get_encryption_info(inode))
1887 return -EACCES;
1888
1889 inode_lock(inode);
1890 ret = generic_write_checks(iocb, from);
1891 if (ret > 0) {
1892 ret = f2fs_preallocate_blocks(iocb, from);
1893 if (!ret)
1894 ret = __generic_file_write_iter(iocb, from);
1895 }
1896 inode_unlock(inode);
1897
1898 if (ret > 0) {
1899 ssize_t err;
1900
1901 err = generic_write_sync(file, iocb->ki_pos - ret, ret);
1902 if (err < 0)
1903 ret = err;
1904 }
1905 return ret;
1906 }
1907
1908 #ifdef CONFIG_COMPAT
1909 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1910 {
1911 switch (cmd) {
1912 case F2FS_IOC32_GETFLAGS:
1913 cmd = F2FS_IOC_GETFLAGS;
1914 break;
1915 case F2FS_IOC32_SETFLAGS:
1916 cmd = F2FS_IOC_SETFLAGS;
1917 break;
1918 case F2FS_IOC32_GETVERSION:
1919 cmd = F2FS_IOC_GETVERSION;
1920 break;
1921 case F2FS_IOC_START_ATOMIC_WRITE:
1922 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1923 case F2FS_IOC_START_VOLATILE_WRITE:
1924 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1925 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1926 case F2FS_IOC_SHUTDOWN:
1927 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1928 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1929 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1930 case F2FS_IOC_GARBAGE_COLLECT:
1931 case F2FS_IOC_WRITE_CHECKPOINT:
1932 case F2FS_IOC_DEFRAGMENT:
1933 break;
1934 default:
1935 return -ENOIOCTLCMD;
1936 }
1937 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1938 }
1939 #endif
1940
1941 const struct file_operations f2fs_file_operations = {
1942 .llseek = f2fs_llseek,
1943 .read_iter = generic_file_read_iter,
1944 .write_iter = f2fs_file_write_iter,
1945 .open = f2fs_file_open,
1946 .release = f2fs_release_file,
1947 .mmap = f2fs_file_mmap,
1948 .fsync = f2fs_sync_file,
1949 .fallocate = f2fs_fallocate,
1950 .unlocked_ioctl = f2fs_ioctl,
1951 #ifdef CONFIG_COMPAT
1952 .compat_ioctl = f2fs_compat_ioctl,
1953 #endif
1954 .splice_read = generic_file_splice_read,
1955 .splice_write = iter_file_splice_write,
1956 };
Linux kernel - Deliverables
This page took 0.114527 seconds and 4 git commands to generate.