projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: introduce f2fs_update_data_blkaddr for cleanup
[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 /* do not invalidate this block address */
861 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
862 do_replace = true;
863 }
864 f2fs_put_dnode(&dn);
865 }
866
867 if (new_addr == NULL_ADDR)
868 return full ? truncate_hole(inode, dst, dst + 1) : 0;
869
870 if (do_replace) {
871 struct page *ipage = get_node_page(sbi, inode->i_ino);
872 struct node_info ni;
873
874 if (IS_ERR(ipage)) {
875 ret = PTR_ERR(ipage);
876 goto err_out;
877 }
878
879 set_new_dnode(&dn, inode, ipage, NULL, 0);
880 ret = f2fs_reserve_block(&dn, dst);
881 if (ret)
882 goto err_out;
883
884 truncate_data_blocks_range(&dn, 1);
885
886 get_node_info(sbi, dn.nid, &ni);
887 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
888 ni.version, true, false);
889 f2fs_put_dnode(&dn);
890 } else {
891 struct page *psrc, *pdst;
892
893 psrc = get_lock_data_page(inode, src, true);
894 if (IS_ERR(psrc))
895 return PTR_ERR(psrc);
896 pdst = get_new_data_page(inode, NULL, dst, true);
897 if (IS_ERR(pdst)) {
898 f2fs_put_page(psrc, 1);
899 return PTR_ERR(pdst);
900 }
901 f2fs_copy_page(psrc, pdst);
902 set_page_dirty(pdst);
903 f2fs_put_page(pdst, 1);
904 f2fs_put_page(psrc, 1);
905
906 return truncate_hole(inode, src, src + 1);
907 }
908 return 0;
909
910 err_out:
911 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
912 f2fs_update_data_blkaddr(&dn, new_addr);
913 f2fs_put_dnode(&dn);
914 }
915 return ret;
916 }
917
918 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
919 {
920 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
921 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
922 int ret = 0;
923
924 for (; end < nrpages; start++, end++) {
925 f2fs_balance_fs(sbi, true);
926 f2fs_lock_op(sbi);
927 ret = __exchange_data_block(inode, end, start, true);
928 f2fs_unlock_op(sbi);
929 if (ret)
930 break;
931 }
932 return ret;
933 }
934
935 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
936 {
937 pgoff_t pg_start, pg_end;
938 loff_t new_size;
939 int ret;
940
941 if (offset + len >= i_size_read(inode))
942 return -EINVAL;
943
944 /* collapse range should be aligned to block size of f2fs. */
945 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
946 return -EINVAL;
947
948 ret = f2fs_convert_inline_inode(inode);
949 if (ret)
950 return ret;
951
952 pg_start = offset >> PAGE_CACHE_SHIFT;
953 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
954
955 /* write out all dirty pages from offset */
956 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
957 if (ret)
958 return ret;
959
960 truncate_pagecache(inode, offset);
961
962 ret = f2fs_do_collapse(inode, pg_start, pg_end);
963 if (ret)
964 return ret;
965
966 /* write out all moved pages, if possible */
967 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
968 truncate_pagecache(inode, offset);
969
970 new_size = i_size_read(inode) - len;
971 truncate_pagecache(inode, new_size);
972
973 ret = truncate_blocks(inode, new_size, true);
974 if (!ret)
975 i_size_write(inode, new_size);
976
977 return ret;
978 }
979
980 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
981 int mode)
982 {
983 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
984 struct address_space *mapping = inode->i_mapping;
985 pgoff_t index, pg_start, pg_end;
986 loff_t new_size = i_size_read(inode);
987 loff_t off_start, off_end;
988 int ret = 0;
989
990 ret = inode_newsize_ok(inode, (len + offset));
991 if (ret)
992 return ret;
993
994 ret = f2fs_convert_inline_inode(inode);
995 if (ret)
996 return ret;
997
998 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
999 if (ret)
1000 return ret;
1001
1002 truncate_pagecache_range(inode, offset, offset + len - 1);
1003
1004 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1005 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1006
1007 off_start = offset & (PAGE_CACHE_SIZE - 1);
1008 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1009
1010 if (pg_start == pg_end) {
1011 ret = fill_zero(inode, pg_start, off_start,
1012 off_end - off_start);
1013 if (ret)
1014 return ret;
1015
1016 if (offset + len > new_size)
1017 new_size = offset + len;
1018 new_size = max_t(loff_t, new_size, offset + len);
1019 } else {
1020 if (off_start) {
1021 ret = fill_zero(inode, pg_start++, off_start,
1022 PAGE_CACHE_SIZE - off_start);
1023 if (ret)
1024 return ret;
1025
1026 new_size = max_t(loff_t, new_size,
1027 (loff_t)pg_start << PAGE_CACHE_SHIFT);
1028 }
1029
1030 for (index = pg_start; index < pg_end; index++) {
1031 struct dnode_of_data dn;
1032 struct page *ipage;
1033
1034 f2fs_lock_op(sbi);
1035
1036 ipage = get_node_page(sbi, inode->i_ino);
1037 if (IS_ERR(ipage)) {
1038 ret = PTR_ERR(ipage);
1039 f2fs_unlock_op(sbi);
1040 goto out;
1041 }
1042
1043 set_new_dnode(&dn, inode, ipage, NULL, 0);
1044 ret = f2fs_reserve_block(&dn, index);
1045 if (ret) {
1046 f2fs_unlock_op(sbi);
1047 goto out;
1048 }
1049
1050 if (dn.data_blkaddr != NEW_ADDR) {
1051 invalidate_blocks(sbi, dn.data_blkaddr);
1052 f2fs_update_data_blkaddr(&dn, NEW_ADDR);
1053 }
1054 f2fs_put_dnode(&dn);
1055 f2fs_unlock_op(sbi);
1056
1057 new_size = max_t(loff_t, new_size,
1058 (loff_t)(index + 1) << PAGE_CACHE_SHIFT);
1059 }
1060
1061 if (off_end) {
1062 ret = fill_zero(inode, pg_end, 0, off_end);
1063 if (ret)
1064 goto out;
1065
1066 new_size = max_t(loff_t, new_size, offset + len);
1067 }
1068 }
1069
1070 out:
1071 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1072 i_size_write(inode, new_size);
1073 mark_inode_dirty(inode);
1074 update_inode_page(inode);
1075 }
1076
1077 return ret;
1078 }
1079
1080 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1081 {
1082 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1083 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1084 loff_t new_size;
1085 int ret = 0;
1086
1087 new_size = i_size_read(inode) + len;
1088 if (new_size > inode->i_sb->s_maxbytes)
1089 return -EFBIG;
1090
1091 if (offset >= i_size_read(inode))
1092 return -EINVAL;
1093
1094 /* insert range should be aligned to block size of f2fs. */
1095 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1096 return -EINVAL;
1097
1098 ret = f2fs_convert_inline_inode(inode);
1099 if (ret)
1100 return ret;
1101
1102 f2fs_balance_fs(sbi, true);
1103
1104 ret = truncate_blocks(inode, i_size_read(inode), true);
1105 if (ret)
1106 return ret;
1107
1108 /* write out all dirty pages from offset */
1109 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1110 if (ret)
1111 return ret;
1112
1113 truncate_pagecache(inode, offset);
1114
1115 pg_start = offset >> PAGE_CACHE_SHIFT;
1116 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
1117 delta = pg_end - pg_start;
1118 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1119
1120 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1121 f2fs_lock_op(sbi);
1122 ret = __exchange_data_block(inode, idx, idx + delta, false);
1123 f2fs_unlock_op(sbi);
1124 if (ret)
1125 break;
1126 }
1127
1128 /* write out all moved pages, if possible */
1129 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1130 truncate_pagecache(inode, offset);
1131
1132 if (!ret)
1133 i_size_write(inode, new_size);
1134 return ret;
1135 }
1136
1137 static int expand_inode_data(struct inode *inode, loff_t offset,
1138 loff_t len, int mode)
1139 {
1140 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1141 pgoff_t index, pg_start, pg_end;
1142 loff_t new_size = i_size_read(inode);
1143 loff_t off_start, off_end;
1144 int ret = 0;
1145
1146 ret = inode_newsize_ok(inode, (len + offset));
1147 if (ret)
1148 return ret;
1149
1150 ret = f2fs_convert_inline_inode(inode);
1151 if (ret)
1152 return ret;
1153
1154 f2fs_balance_fs(sbi, true);
1155
1156 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1157 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1158
1159 off_start = offset & (PAGE_CACHE_SIZE - 1);
1160 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1161
1162 f2fs_lock_op(sbi);
1163
1164 for (index = pg_start; index <= pg_end; index++) {
1165 struct dnode_of_data dn;
1166
1167 if (index == pg_end && !off_end)
1168 goto noalloc;
1169
1170 set_new_dnode(&dn, inode, NULL, NULL, 0);
1171 ret = f2fs_reserve_block(&dn, index);
1172 if (ret)
1173 break;
1174 noalloc:
1175 if (pg_start == pg_end)
1176 new_size = offset + len;
1177 else if (index == pg_start && off_start)
1178 new_size = (loff_t)(index + 1) << PAGE_CACHE_SHIFT;
1179 else if (index == pg_end)
1180 new_size = ((loff_t)index << PAGE_CACHE_SHIFT) +
1181 off_end;
1182 else
1183 new_size += PAGE_CACHE_SIZE;
1184 }
1185
1186 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1187 i_size_read(inode) < new_size) {
1188 i_size_write(inode, new_size);
1189 mark_inode_dirty(inode);
1190 update_inode_page(inode);
1191 }
1192 f2fs_unlock_op(sbi);
1193
1194 return ret;
1195 }
1196
1197 static long f2fs_fallocate(struct file *file, int mode,
1198 loff_t offset, loff_t len)
1199 {
1200 struct inode *inode = file_inode(file);
1201 long ret = 0;
1202
1203 /* f2fs only support ->fallocate for regular file */
1204 if (!S_ISREG(inode->i_mode))
1205 return -EINVAL;
1206
1207 if (f2fs_encrypted_inode(inode) &&
1208 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1209 return -EOPNOTSUPP;
1210
1211 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1212 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1213 FALLOC_FL_INSERT_RANGE))
1214 return -EOPNOTSUPP;
1215
1216 inode_lock(inode);
1217
1218 if (mode & FALLOC_FL_PUNCH_HOLE) {
1219 if (offset >= inode->i_size)
1220 goto out;
1221
1222 ret = punch_hole(inode, offset, len);
1223 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1224 ret = f2fs_collapse_range(inode, offset, len);
1225 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1226 ret = f2fs_zero_range(inode, offset, len, mode);
1227 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1228 ret = f2fs_insert_range(inode, offset, len);
1229 } else {
1230 ret = expand_inode_data(inode, offset, len, mode);
1231 }
1232
1233 if (!ret) {
1234 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1235 mark_inode_dirty(inode);
1236 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1237 }
1238
1239 out:
1240 inode_unlock(inode);
1241
1242 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1243 return ret;
1244 }
1245
1246 static int f2fs_release_file(struct inode *inode, struct file *filp)
1247 {
1248 /* some remained atomic pages should discarded */
1249 if (f2fs_is_atomic_file(inode))
1250 drop_inmem_pages(inode);
1251 if (f2fs_is_volatile_file(inode)) {
1252 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1253 filemap_fdatawrite(inode->i_mapping);
1254 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1255 }
1256 return 0;
1257 }
1258
1259 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1260 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1261
1262 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1263 {
1264 if (S_ISDIR(mode))
1265 return flags;
1266 else if (S_ISREG(mode))
1267 return flags & F2FS_REG_FLMASK;
1268 else
1269 return flags & F2FS_OTHER_FLMASK;
1270 }
1271
1272 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1273 {
1274 struct inode *inode = file_inode(filp);
1275 struct f2fs_inode_info *fi = F2FS_I(inode);
1276 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1277 return put_user(flags, (int __user *)arg);
1278 }
1279
1280 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1281 {
1282 struct inode *inode = file_inode(filp);
1283 struct f2fs_inode_info *fi = F2FS_I(inode);
1284 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1285 unsigned int oldflags;
1286 int ret;
1287
1288 ret = mnt_want_write_file(filp);
1289 if (ret)
1290 return ret;
1291
1292 if (!inode_owner_or_capable(inode)) {
1293 ret = -EACCES;
1294 goto out;
1295 }
1296
1297 if (get_user(flags, (int __user *)arg)) {
1298 ret = -EFAULT;
1299 goto out;
1300 }
1301
1302 flags = f2fs_mask_flags(inode->i_mode, flags);
1303
1304 inode_lock(inode);
1305
1306 oldflags = fi->i_flags;
1307
1308 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1309 if (!capable(CAP_LINUX_IMMUTABLE)) {
1310 inode_unlock(inode);
1311 ret = -EPERM;
1312 goto out;
1313 }
1314 }
1315
1316 flags = flags & FS_FL_USER_MODIFIABLE;
1317 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1318 fi->i_flags = flags;
1319 inode_unlock(inode);
1320
1321 f2fs_set_inode_flags(inode);
1322 inode->i_ctime = CURRENT_TIME;
1323 mark_inode_dirty(inode);
1324 out:
1325 mnt_drop_write_file(filp);
1326 return ret;
1327 }
1328
1329 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1330 {
1331 struct inode *inode = file_inode(filp);
1332
1333 return put_user(inode->i_generation, (int __user *)arg);
1334 }
1335
1336 static int f2fs_ioc_start_atomic_write(struct file *filp)
1337 {
1338 struct inode *inode = file_inode(filp);
1339 int ret;
1340
1341 if (!inode_owner_or_capable(inode))
1342 return -EACCES;
1343
1344 if (f2fs_is_atomic_file(inode))
1345 return 0;
1346
1347 ret = f2fs_convert_inline_inode(inode);
1348 if (ret)
1349 return ret;
1350
1351 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1352 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1353
1354 return 0;
1355 }
1356
1357 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1358 {
1359 struct inode *inode = file_inode(filp);
1360 int ret;
1361
1362 if (!inode_owner_or_capable(inode))
1363 return -EACCES;
1364
1365 if (f2fs_is_volatile_file(inode))
1366 return 0;
1367
1368 ret = mnt_want_write_file(filp);
1369 if (ret)
1370 return ret;
1371
1372 if (f2fs_is_atomic_file(inode)) {
1373 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1374 ret = commit_inmem_pages(inode);
1375 if (ret) {
1376 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1377 goto err_out;
1378 }
1379 }
1380
1381 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1382 err_out:
1383 mnt_drop_write_file(filp);
1384 return ret;
1385 }
1386
1387 static int f2fs_ioc_start_volatile_write(struct file *filp)
1388 {
1389 struct inode *inode = file_inode(filp);
1390 int ret;
1391
1392 if (!inode_owner_or_capable(inode))
1393 return -EACCES;
1394
1395 if (f2fs_is_volatile_file(inode))
1396 return 0;
1397
1398 ret = f2fs_convert_inline_inode(inode);
1399 if (ret)
1400 return ret;
1401
1402 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1403 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1404 return 0;
1405 }
1406
1407 static int f2fs_ioc_release_volatile_write(struct file *filp)
1408 {
1409 struct inode *inode = file_inode(filp);
1410
1411 if (!inode_owner_or_capable(inode))
1412 return -EACCES;
1413
1414 if (!f2fs_is_volatile_file(inode))
1415 return 0;
1416
1417 if (!f2fs_is_first_block_written(inode))
1418 return truncate_partial_data_page(inode, 0, true);
1419
1420 return punch_hole(inode, 0, F2FS_BLKSIZE);
1421 }
1422
1423 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1424 {
1425 struct inode *inode = file_inode(filp);
1426 int ret;
1427
1428 if (!inode_owner_or_capable(inode))
1429 return -EACCES;
1430
1431 ret = mnt_want_write_file(filp);
1432 if (ret)
1433 return ret;
1434
1435 if (f2fs_is_atomic_file(inode)) {
1436 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1437 drop_inmem_pages(inode);
1438 }
1439 if (f2fs_is_volatile_file(inode)) {
1440 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1441 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1442 }
1443
1444 mnt_drop_write_file(filp);
1445 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1446 return ret;
1447 }
1448
1449 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1450 {
1451 struct inode *inode = file_inode(filp);
1452 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1453 struct super_block *sb = sbi->sb;
1454 __u32 in;
1455
1456 if (!capable(CAP_SYS_ADMIN))
1457 return -EPERM;
1458
1459 if (get_user(in, (__u32 __user *)arg))
1460 return -EFAULT;
1461
1462 switch (in) {
1463 case F2FS_GOING_DOWN_FULLSYNC:
1464 sb = freeze_bdev(sb->s_bdev);
1465 if (sb && !IS_ERR(sb)) {
1466 f2fs_stop_checkpoint(sbi);
1467 thaw_bdev(sb->s_bdev, sb);
1468 }
1469 break;
1470 case F2FS_GOING_DOWN_METASYNC:
1471 /* do checkpoint only */
1472 f2fs_sync_fs(sb, 1);
1473 f2fs_stop_checkpoint(sbi);
1474 break;
1475 case F2FS_GOING_DOWN_NOSYNC:
1476 f2fs_stop_checkpoint(sbi);
1477 break;
1478 case F2FS_GOING_DOWN_METAFLUSH:
1479 sync_meta_pages(sbi, META, LONG_MAX);
1480 f2fs_stop_checkpoint(sbi);
1481 break;
1482 default:
1483 return -EINVAL;
1484 }
1485 f2fs_update_time(sbi, REQ_TIME);
1486 return 0;
1487 }
1488
1489 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1490 {
1491 struct inode *inode = file_inode(filp);
1492 struct super_block *sb = inode->i_sb;
1493 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1494 struct fstrim_range range;
1495 int ret;
1496
1497 if (!capable(CAP_SYS_ADMIN))
1498 return -EPERM;
1499
1500 if (!blk_queue_discard(q))
1501 return -EOPNOTSUPP;
1502
1503 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1504 sizeof(range)))
1505 return -EFAULT;
1506
1507 range.minlen = max((unsigned int)range.minlen,
1508 q->limits.discard_granularity);
1509 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1510 if (ret < 0)
1511 return ret;
1512
1513 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1514 sizeof(range)))
1515 return -EFAULT;
1516 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1517 return 0;
1518 }
1519
1520 static bool uuid_is_nonzero(__u8 u[16])
1521 {
1522 int i;
1523
1524 for (i = 0; i < 16; i++)
1525 if (u[i])
1526 return true;
1527 return false;
1528 }
1529
1530 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1531 {
1532 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1533 struct f2fs_encryption_policy policy;
1534 struct inode *inode = file_inode(filp);
1535
1536 if (copy_from_user(&policy, (struct f2fs_encryption_policy __user *)arg,
1537 sizeof(policy)))
1538 return -EFAULT;
1539
1540 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1541 return f2fs_process_policy(&policy, inode);
1542 #else
1543 return -EOPNOTSUPP;
1544 #endif
1545 }
1546
1547 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1548 {
1549 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1550 struct f2fs_encryption_policy policy;
1551 struct inode *inode = file_inode(filp);
1552 int err;
1553
1554 err = f2fs_get_policy(inode, &policy);
1555 if (err)
1556 return err;
1557
1558 if (copy_to_user((struct f2fs_encryption_policy __user *)arg, &policy,
1559 sizeof(policy)))
1560 return -EFAULT;
1561 return 0;
1562 #else
1563 return -EOPNOTSUPP;
1564 #endif
1565 }
1566
1567 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1568 {
1569 struct inode *inode = file_inode(filp);
1570 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1571 int err;
1572
1573 if (!f2fs_sb_has_crypto(inode->i_sb))
1574 return -EOPNOTSUPP;
1575
1576 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1577 goto got_it;
1578
1579 err = mnt_want_write_file(filp);
1580 if (err)
1581 return err;
1582
1583 /* update superblock with uuid */
1584 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1585
1586 err = f2fs_commit_super(sbi, false);
1587 if (err) {
1588 /* undo new data */
1589 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1590 mnt_drop_write_file(filp);
1591 return err;
1592 }
1593 mnt_drop_write_file(filp);
1594 got_it:
1595 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1596 16))
1597 return -EFAULT;
1598 return 0;
1599 }
1600
1601 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1602 {
1603 struct inode *inode = file_inode(filp);
1604 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1605 __u32 sync;
1606
1607 if (!capable(CAP_SYS_ADMIN))
1608 return -EPERM;
1609
1610 if (get_user(sync, (__u32 __user *)arg))
1611 return -EFAULT;
1612
1613 if (f2fs_readonly(sbi->sb))
1614 return -EROFS;
1615
1616 if (!sync) {
1617 if (!mutex_trylock(&sbi->gc_mutex))
1618 return -EBUSY;
1619 } else {
1620 mutex_lock(&sbi->gc_mutex);
1621 }
1622
1623 return f2fs_gc(sbi, sync);
1624 }
1625
1626 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1627 {
1628 struct inode *inode = file_inode(filp);
1629 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1630
1631 if (!capable(CAP_SYS_ADMIN))
1632 return -EPERM;
1633
1634 if (f2fs_readonly(sbi->sb))
1635 return -EROFS;
1636
1637 return f2fs_sync_fs(sbi->sb, 1);
1638 }
1639
1640 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1641 struct file *filp,
1642 struct f2fs_defragment *range)
1643 {
1644 struct inode *inode = file_inode(filp);
1645 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1646 struct extent_info ei;
1647 pgoff_t pg_start, pg_end;
1648 unsigned int blk_per_seg = sbi->blocks_per_seg;
1649 unsigned int total = 0, sec_num;
1650 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1651 block_t blk_end = 0;
1652 bool fragmented = false;
1653 int err;
1654
1655 /* if in-place-update policy is enabled, don't waste time here */
1656 if (need_inplace_update(inode))
1657 return -EINVAL;
1658
1659 pg_start = range->start >> PAGE_CACHE_SHIFT;
1660 pg_end = (range->start + range->len) >> PAGE_CACHE_SHIFT;
1661
1662 f2fs_balance_fs(sbi, true);
1663
1664 inode_lock(inode);
1665
1666 /* writeback all dirty pages in the range */
1667 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1668 range->start + range->len - 1);
1669 if (err)
1670 goto out;
1671
1672 /*
1673 * lookup mapping info in extent cache, skip defragmenting if physical
1674 * block addresses are continuous.
1675 */
1676 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1677 if (ei.fofs + ei.len >= pg_end)
1678 goto out;
1679 }
1680
1681 map.m_lblk = pg_start;
1682
1683 /*
1684 * lookup mapping info in dnode page cache, skip defragmenting if all
1685 * physical block addresses are continuous even if there are hole(s)
1686 * in logical blocks.
1687 */
1688 while (map.m_lblk < pg_end) {
1689 map.m_len = pg_end - map.m_lblk;
1690 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1691 if (err)
1692 goto out;
1693
1694 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1695 map.m_lblk++;
1696 continue;
1697 }
1698
1699 if (blk_end && blk_end != map.m_pblk) {
1700 fragmented = true;
1701 break;
1702 }
1703 blk_end = map.m_pblk + map.m_len;
1704
1705 map.m_lblk += map.m_len;
1706 }
1707
1708 if (!fragmented)
1709 goto out;
1710
1711 map.m_lblk = pg_start;
1712 map.m_len = pg_end - pg_start;
1713
1714 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1715
1716 /*
1717 * make sure there are enough free section for LFS allocation, this can
1718 * avoid defragment running in SSR mode when free section are allocated
1719 * intensively
1720 */
1721 if (has_not_enough_free_secs(sbi, sec_num)) {
1722 err = -EAGAIN;
1723 goto out;
1724 }
1725
1726 while (map.m_lblk < pg_end) {
1727 pgoff_t idx;
1728 int cnt = 0;
1729
1730 do_map:
1731 map.m_len = pg_end - map.m_lblk;
1732 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1733 if (err)
1734 goto clear_out;
1735
1736 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1737 map.m_lblk++;
1738 continue;
1739 }
1740
1741 set_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1742
1743 idx = map.m_lblk;
1744 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1745 struct page *page;
1746
1747 page = get_lock_data_page(inode, idx, true);
1748 if (IS_ERR(page)) {
1749 err = PTR_ERR(page);
1750 goto clear_out;
1751 }
1752
1753 set_page_dirty(page);
1754 f2fs_put_page(page, 1);
1755
1756 idx++;
1757 cnt++;
1758 total++;
1759 }
1760
1761 map.m_lblk = idx;
1762
1763 if (idx < pg_end && cnt < blk_per_seg)
1764 goto do_map;
1765
1766 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1767
1768 err = filemap_fdatawrite(inode->i_mapping);
1769 if (err)
1770 goto out;
1771 }
1772 clear_out:
1773 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1774 out:
1775 inode_unlock(inode);
1776 if (!err)
1777 range->len = (u64)total << PAGE_CACHE_SHIFT;
1778 return err;
1779 }
1780
1781 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1782 {
1783 struct inode *inode = file_inode(filp);
1784 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1785 struct f2fs_defragment range;
1786 int err;
1787
1788 if (!capable(CAP_SYS_ADMIN))
1789 return -EPERM;
1790
1791 if (!S_ISREG(inode->i_mode))
1792 return -EINVAL;
1793
1794 err = mnt_want_write_file(filp);
1795 if (err)
1796 return err;
1797
1798 if (f2fs_readonly(sbi->sb)) {
1799 err = -EROFS;
1800 goto out;
1801 }
1802
1803 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1804 sizeof(range))) {
1805 err = -EFAULT;
1806 goto out;
1807 }
1808
1809 /* verify alignment of offset & size */
1810 if (range.start & (F2FS_BLKSIZE - 1) ||
1811 range.len & (F2FS_BLKSIZE - 1)) {
1812 err = -EINVAL;
1813 goto out;
1814 }
1815
1816 err = f2fs_defragment_range(sbi, filp, &range);
1817 f2fs_update_time(sbi, REQ_TIME);
1818 if (err < 0)
1819 goto out;
1820
1821 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1822 sizeof(range)))
1823 err = -EFAULT;
1824 out:
1825 mnt_drop_write_file(filp);
1826 return err;
1827 }
1828
1829 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1830 {
1831 switch (cmd) {
1832 case F2FS_IOC_GETFLAGS:
1833 return f2fs_ioc_getflags(filp, arg);
1834 case F2FS_IOC_SETFLAGS:
1835 return f2fs_ioc_setflags(filp, arg);
1836 case F2FS_IOC_GETVERSION:
1837 return f2fs_ioc_getversion(filp, arg);
1838 case F2FS_IOC_START_ATOMIC_WRITE:
1839 return f2fs_ioc_start_atomic_write(filp);
1840 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1841 return f2fs_ioc_commit_atomic_write(filp);
1842 case F2FS_IOC_START_VOLATILE_WRITE:
1843 return f2fs_ioc_start_volatile_write(filp);
1844 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1845 return f2fs_ioc_release_volatile_write(filp);
1846 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1847 return f2fs_ioc_abort_volatile_write(filp);
1848 case F2FS_IOC_SHUTDOWN:
1849 return f2fs_ioc_shutdown(filp, arg);
1850 case FITRIM:
1851 return f2fs_ioc_fitrim(filp, arg);
1852 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1853 return f2fs_ioc_set_encryption_policy(filp, arg);
1854 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1855 return f2fs_ioc_get_encryption_policy(filp, arg);
1856 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1857 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1858 case F2FS_IOC_GARBAGE_COLLECT:
1859 return f2fs_ioc_gc(filp, arg);
1860 case F2FS_IOC_WRITE_CHECKPOINT:
1861 return f2fs_ioc_write_checkpoint(filp, arg);
1862 case F2FS_IOC_DEFRAGMENT:
1863 return f2fs_ioc_defragment(filp, arg);
1864 default:
1865 return -ENOTTY;
1866 }
1867 }
1868
1869 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1870 {
1871 struct file *file = iocb->ki_filp;
1872 struct inode *inode = file_inode(file);
1873 ssize_t ret;
1874
1875 if (f2fs_encrypted_inode(inode) &&
1876 !f2fs_has_encryption_key(inode) &&
1877 f2fs_get_encryption_info(inode))
1878 return -EACCES;
1879
1880 inode_lock(inode);
1881 ret = generic_write_checks(iocb, from);
1882 if (ret > 0) {
1883 ret = f2fs_preallocate_blocks(iocb, from);
1884 if (!ret)
1885 ret = __generic_file_write_iter(iocb, from);
1886 }
1887 inode_unlock(inode);
1888
1889 if (ret > 0) {
1890 ssize_t err;
1891
1892 err = generic_write_sync(file, iocb->ki_pos - ret, ret);
1893 if (err < 0)
1894 ret = err;
1895 }
1896 return ret;
1897 }
1898
1899 #ifdef CONFIG_COMPAT
1900 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1901 {
1902 switch (cmd) {
1903 case F2FS_IOC32_GETFLAGS:
1904 cmd = F2FS_IOC_GETFLAGS;
1905 break;
1906 case F2FS_IOC32_SETFLAGS:
1907 cmd = F2FS_IOC_SETFLAGS;
1908 break;
1909 case F2FS_IOC32_GETVERSION:
1910 cmd = F2FS_IOC_GETVERSION;
1911 break;
1912 case F2FS_IOC_START_ATOMIC_WRITE:
1913 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1914 case F2FS_IOC_START_VOLATILE_WRITE:
1915 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1916 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1917 case F2FS_IOC_SHUTDOWN:
1918 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1919 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1920 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1921 case F2FS_IOC_GARBAGE_COLLECT:
1922 case F2FS_IOC_WRITE_CHECKPOINT:
1923 case F2FS_IOC_DEFRAGMENT:
1924 break;
1925 default:
1926 return -ENOIOCTLCMD;
1927 }
1928 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1929 }
1930 #endif
1931
1932 const struct file_operations f2fs_file_operations = {
1933 .llseek = f2fs_llseek,
1934 .read_iter = generic_file_read_iter,
1935 .write_iter = f2fs_file_write_iter,
1936 .open = f2fs_file_open,
1937 .release = f2fs_release_file,
1938 .mmap = f2fs_file_mmap,
1939 .fsync = f2fs_sync_file,
1940 .fallocate = f2fs_fallocate,
1941 .unlocked_ioctl = f2fs_ioctl,
1942 #ifdef CONFIG_COMPAT
1943 .compat_ioctl = f2fs_compat_ioctl,
1944 #endif
1945 .splice_read = generic_file_splice_read,
1946 .splice_write = iter_file_splice_write,
1947 };
Linux kernel - Deliverables
This page took 0.079644 seconds and 6 git commands to generate.