projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: call SetPageUptodate if needed
[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/uuid.h>
24
25 #include "f2fs.h"
26 #include "node.h"
27 #include "segment.h"
28 #include "xattr.h"
29 #include "acl.h"
30 #include "gc.h"
31 #include "trace.h"
32 #include <trace/events/f2fs.h>
33
34 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
35 struct vm_fault *vmf)
36 {
37 struct page *page = vmf->page;
38 struct inode *inode = file_inode(vma->vm_file);
39 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
40 struct dnode_of_data dn;
41 int err;
42
43 sb_start_pagefault(inode->i_sb);
44
45 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
46
47 /* block allocation */
48 f2fs_lock_op(sbi);
49 set_new_dnode(&dn, inode, NULL, NULL, 0);
50 err = f2fs_reserve_block(&dn, page->index);
51 if (err) {
52 f2fs_unlock_op(sbi);
53 goto out;
54 }
55 f2fs_put_dnode(&dn);
56 f2fs_unlock_op(sbi);
57
58 f2fs_balance_fs(sbi, dn.node_changed);
59
60 file_update_time(vma->vm_file);
61 lock_page(page);
62 if (unlikely(page->mapping != inode->i_mapping ||
63 page_offset(page) > i_size_read(inode) ||
64 !PageUptodate(page))) {
65 unlock_page(page);
66 err = -EFAULT;
67 goto out;
68 }
69
70 /*
71 * check to see if the page is mapped already (no holes)
72 */
73 if (PageMappedToDisk(page))
74 goto mapped;
75
76 /* page is wholly or partially inside EOF */
77 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
78 i_size_read(inode)) {
79 unsigned offset;
80 offset = i_size_read(inode) & ~PAGE_MASK;
81 zero_user_segment(page, offset, PAGE_SIZE);
82 }
83 set_page_dirty(page);
84 if (!PageUptodate(page))
85 SetPageUptodate(page);
86
87 trace_f2fs_vm_page_mkwrite(page, DATA);
88 mapped:
89 /* fill the page */
90 f2fs_wait_on_page_writeback(page, DATA, false);
91
92 /* wait for GCed encrypted page writeback */
93 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
94 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
95
96 /* if gced page is attached, don't write to cold segment */
97 clear_cold_data(page);
98 out:
99 sb_end_pagefault(inode->i_sb);
100 f2fs_update_time(sbi, REQ_TIME);
101 return block_page_mkwrite_return(err);
102 }
103
104 static const struct vm_operations_struct f2fs_file_vm_ops = {
105 .fault = filemap_fault,
106 .map_pages = filemap_map_pages,
107 .page_mkwrite = f2fs_vm_page_mkwrite,
108 };
109
110 static int get_parent_ino(struct inode *inode, nid_t *pino)
111 {
112 struct dentry *dentry;
113
114 inode = igrab(inode);
115 dentry = d_find_any_alias(inode);
116 iput(inode);
117 if (!dentry)
118 return 0;
119
120 if (update_dent_inode(inode, inode, &dentry->d_name)) {
121 dput(dentry);
122 return 0;
123 }
124
125 *pino = parent_ino(dentry);
126 dput(dentry);
127 return 1;
128 }
129
130 static inline bool need_do_checkpoint(struct inode *inode)
131 {
132 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
133 bool need_cp = false;
134
135 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
136 need_cp = true;
137 else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
138 need_cp = true;
139 else if (file_wrong_pino(inode))
140 need_cp = true;
141 else if (!space_for_roll_forward(sbi))
142 need_cp = true;
143 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
144 need_cp = true;
145 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
146 need_cp = true;
147 else if (test_opt(sbi, FASTBOOT))
148 need_cp = true;
149 else if (sbi->active_logs == 2)
150 need_cp = true;
151
152 return need_cp;
153 }
154
155 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
156 {
157 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
158 bool ret = false;
159 /* But we need to avoid that there are some inode updates */
160 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
161 ret = true;
162 f2fs_put_page(i, 0);
163 return ret;
164 }
165
166 static void try_to_fix_pino(struct inode *inode)
167 {
168 struct f2fs_inode_info *fi = F2FS_I(inode);
169 nid_t pino;
170
171 down_write(&fi->i_sem);
172 fi->xattr_ver = 0;
173 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
174 get_parent_ino(inode, &pino)) {
175 f2fs_i_pino_write(inode, pino);
176 file_got_pino(inode);
177 }
178 up_write(&fi->i_sem);
179 }
180
181 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
182 int datasync, bool atomic)
183 {
184 struct inode *inode = file->f_mapping->host;
185 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
186 nid_t ino = inode->i_ino;
187 int ret = 0;
188 bool need_cp = false;
189 struct writeback_control wbc = {
190 .sync_mode = WB_SYNC_ALL,
191 .nr_to_write = LONG_MAX,
192 .for_reclaim = 0,
193 };
194
195 if (unlikely(f2fs_readonly(inode->i_sb)))
196 return 0;
197
198 trace_f2fs_sync_file_enter(inode);
199
200 /* if fdatasync is triggered, let's do in-place-update */
201 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
202 set_inode_flag(inode, FI_NEED_IPU);
203 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
204 clear_inode_flag(inode, FI_NEED_IPU);
205
206 if (ret) {
207 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
208 return ret;
209 }
210
211 /* if the inode is dirty, let's recover all the time */
212 if (!datasync && !f2fs_skip_inode_update(inode)) {
213 f2fs_write_inode(inode, NULL);
214 goto go_write;
215 }
216
217 /*
218 * if there is no written data, don't waste time to write recovery info.
219 */
220 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
221 !exist_written_data(sbi, ino, APPEND_INO)) {
222
223 /* it may call write_inode just prior to fsync */
224 if (need_inode_page_update(sbi, ino))
225 goto go_write;
226
227 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
228 exist_written_data(sbi, ino, UPDATE_INO))
229 goto flush_out;
230 goto out;
231 }
232 go_write:
233 /*
234 * Both of fdatasync() and fsync() are able to be recovered from
235 * sudden-power-off.
236 */
237 down_read(&F2FS_I(inode)->i_sem);
238 need_cp = need_do_checkpoint(inode);
239 up_read(&F2FS_I(inode)->i_sem);
240
241 if (need_cp) {
242 /* all the dirty node pages should be flushed for POR */
243 ret = f2fs_sync_fs(inode->i_sb, 1);
244
245 /*
246 * We've secured consistency through sync_fs. Following pino
247 * will be used only for fsynced inodes after checkpoint.
248 */
249 try_to_fix_pino(inode);
250 clear_inode_flag(inode, FI_APPEND_WRITE);
251 clear_inode_flag(inode, FI_UPDATE_WRITE);
252 goto out;
253 }
254 sync_nodes:
255 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
256 if (ret)
257 goto out;
258
259 /* if cp_error was enabled, we should avoid infinite loop */
260 if (unlikely(f2fs_cp_error(sbi))) {
261 ret = -EIO;
262 goto out;
263 }
264
265 if (need_inode_block_update(sbi, ino)) {
266 mark_inode_dirty_sync(inode);
267 f2fs_write_inode(inode, NULL);
268 goto sync_nodes;
269 }
270
271 ret = wait_on_node_pages_writeback(sbi, ino);
272 if (ret)
273 goto out;
274
275 /* once recovery info is written, don't need to tack this */
276 remove_ino_entry(sbi, ino, APPEND_INO);
277 clear_inode_flag(inode, FI_APPEND_WRITE);
278 flush_out:
279 remove_ino_entry(sbi, ino, UPDATE_INO);
280 clear_inode_flag(inode, FI_UPDATE_WRITE);
281 ret = f2fs_issue_flush(sbi);
282 f2fs_update_time(sbi, REQ_TIME);
283 out:
284 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
285 f2fs_trace_ios(NULL, 1);
286 return ret;
287 }
288
289 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
290 {
291 return f2fs_do_sync_file(file, start, end, datasync, false);
292 }
293
294 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
295 pgoff_t pgofs, int whence)
296 {
297 struct pagevec pvec;
298 int nr_pages;
299
300 if (whence != SEEK_DATA)
301 return 0;
302
303 /* find first dirty page index */
304 pagevec_init(&pvec, 0);
305 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
306 PAGECACHE_TAG_DIRTY, 1);
307 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
308 pagevec_release(&pvec);
309 return pgofs;
310 }
311
312 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
313 int whence)
314 {
315 switch (whence) {
316 case SEEK_DATA:
317 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
318 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
319 return true;
320 break;
321 case SEEK_HOLE:
322 if (blkaddr == NULL_ADDR)
323 return true;
324 break;
325 }
326 return false;
327 }
328
329 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
330 {
331 struct inode *inode = file->f_mapping->host;
332 loff_t maxbytes = inode->i_sb->s_maxbytes;
333 struct dnode_of_data dn;
334 pgoff_t pgofs, end_offset, dirty;
335 loff_t data_ofs = offset;
336 loff_t isize;
337 int err = 0;
338
339 inode_lock(inode);
340
341 isize = i_size_read(inode);
342 if (offset >= isize)
343 goto fail;
344
345 /* handle inline data case */
346 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
347 if (whence == SEEK_HOLE)
348 data_ofs = isize;
349 goto found;
350 }
351
352 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
353
354 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
355
356 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
357 set_new_dnode(&dn, inode, NULL, NULL, 0);
358 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
359 if (err && err != -ENOENT) {
360 goto fail;
361 } else if (err == -ENOENT) {
362 /* direct node does not exists */
363 if (whence == SEEK_DATA) {
364 pgofs = get_next_page_offset(&dn, pgofs);
365 continue;
366 } else {
367 goto found;
368 }
369 }
370
371 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
372
373 /* find data/hole in dnode block */
374 for (; dn.ofs_in_node < end_offset;
375 dn.ofs_in_node++, pgofs++,
376 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
377 block_t blkaddr;
378 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
379
380 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
381 f2fs_put_dnode(&dn);
382 goto found;
383 }
384 }
385 f2fs_put_dnode(&dn);
386 }
387
388 if (whence == SEEK_DATA)
389 goto fail;
390 found:
391 if (whence == SEEK_HOLE && data_ofs > isize)
392 data_ofs = isize;
393 inode_unlock(inode);
394 return vfs_setpos(file, data_ofs, maxbytes);
395 fail:
396 inode_unlock(inode);
397 return -ENXIO;
398 }
399
400 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
401 {
402 struct inode *inode = file->f_mapping->host;
403 loff_t maxbytes = inode->i_sb->s_maxbytes;
404
405 switch (whence) {
406 case SEEK_SET:
407 case SEEK_CUR:
408 case SEEK_END:
409 return generic_file_llseek_size(file, offset, whence,
410 maxbytes, i_size_read(inode));
411 case SEEK_DATA:
412 case SEEK_HOLE:
413 if (offset < 0)
414 return -ENXIO;
415 return f2fs_seek_block(file, offset, whence);
416 }
417
418 return -EINVAL;
419 }
420
421 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
422 {
423 struct inode *inode = file_inode(file);
424 int err;
425
426 if (f2fs_encrypted_inode(inode)) {
427 err = fscrypt_get_encryption_info(inode);
428 if (err)
429 return 0;
430 if (!f2fs_encrypted_inode(inode))
431 return -ENOKEY;
432 }
433
434 /* we don't need to use inline_data strictly */
435 err = f2fs_convert_inline_inode(inode);
436 if (err)
437 return err;
438
439 file_accessed(file);
440 vma->vm_ops = &f2fs_file_vm_ops;
441 return 0;
442 }
443
444 static int f2fs_file_open(struct inode *inode, struct file *filp)
445 {
446 int ret = generic_file_open(inode, filp);
447 struct dentry *dir;
448
449 if (!ret && f2fs_encrypted_inode(inode)) {
450 ret = fscrypt_get_encryption_info(inode);
451 if (ret)
452 return -EACCES;
453 if (!fscrypt_has_encryption_key(inode))
454 return -ENOKEY;
455 }
456 dir = dget_parent(file_dentry(filp));
457 if (f2fs_encrypted_inode(d_inode(dir)) &&
458 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
459 dput(dir);
460 return -EPERM;
461 }
462 dput(dir);
463 return ret;
464 }
465
466 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
467 {
468 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
469 struct f2fs_node *raw_node;
470 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
471 __le32 *addr;
472
473 raw_node = F2FS_NODE(dn->node_page);
474 addr = blkaddr_in_node(raw_node) + ofs;
475
476 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
477 block_t blkaddr = le32_to_cpu(*addr);
478 if (blkaddr == NULL_ADDR)
479 continue;
480
481 dn->data_blkaddr = NULL_ADDR;
482 set_data_blkaddr(dn);
483 invalidate_blocks(sbi, blkaddr);
484 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
485 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
486 nr_free++;
487 }
488
489 if (nr_free) {
490 pgoff_t fofs;
491 /*
492 * once we invalidate valid blkaddr in range [ofs, ofs + count],
493 * we will invalidate all blkaddr in the whole range.
494 */
495 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
496 dn->inode) + ofs;
497 f2fs_update_extent_cache_range(dn, fofs, 0, len);
498 dec_valid_block_count(sbi, dn->inode, nr_free);
499 }
500 dn->ofs_in_node = ofs;
501
502 f2fs_update_time(sbi, REQ_TIME);
503 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
504 dn->ofs_in_node, nr_free);
505 return nr_free;
506 }
507
508 void truncate_data_blocks(struct dnode_of_data *dn)
509 {
510 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
511 }
512
513 static int truncate_partial_data_page(struct inode *inode, u64 from,
514 bool cache_only)
515 {
516 unsigned offset = from & (PAGE_SIZE - 1);
517 pgoff_t index = from >> PAGE_SHIFT;
518 struct address_space *mapping = inode->i_mapping;
519 struct page *page;
520
521 if (!offset && !cache_only)
522 return 0;
523
524 if (cache_only) {
525 page = f2fs_grab_cache_page(mapping, index, false);
526 if (page && PageUptodate(page))
527 goto truncate_out;
528 f2fs_put_page(page, 1);
529 return 0;
530 }
531
532 page = get_lock_data_page(inode, index, true);
533 if (IS_ERR(page))
534 return 0;
535 truncate_out:
536 f2fs_wait_on_page_writeback(page, DATA, true);
537 zero_user(page, offset, PAGE_SIZE - offset);
538 if (!cache_only || !f2fs_encrypted_inode(inode) ||
539 !S_ISREG(inode->i_mode))
540 set_page_dirty(page);
541 f2fs_put_page(page, 1);
542 return 0;
543 }
544
545 int truncate_blocks(struct inode *inode, u64 from, bool lock)
546 {
547 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
548 unsigned int blocksize = inode->i_sb->s_blocksize;
549 struct dnode_of_data dn;
550 pgoff_t free_from;
551 int count = 0, err = 0;
552 struct page *ipage;
553 bool truncate_page = false;
554
555 trace_f2fs_truncate_blocks_enter(inode, from);
556
557 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
558
559 if (free_from >= sbi->max_file_blocks)
560 goto free_partial;
561
562 if (lock)
563 f2fs_lock_op(sbi);
564
565 ipage = get_node_page(sbi, inode->i_ino);
566 if (IS_ERR(ipage)) {
567 err = PTR_ERR(ipage);
568 goto out;
569 }
570
571 if (f2fs_has_inline_data(inode)) {
572 if (truncate_inline_inode(ipage, from))
573 set_page_dirty(ipage);
574 f2fs_put_page(ipage, 1);
575 truncate_page = true;
576 goto out;
577 }
578
579 set_new_dnode(&dn, inode, ipage, NULL, 0);
580 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
581 if (err) {
582 if (err == -ENOENT)
583 goto free_next;
584 goto out;
585 }
586
587 count = ADDRS_PER_PAGE(dn.node_page, inode);
588
589 count -= dn.ofs_in_node;
590 f2fs_bug_on(sbi, count < 0);
591
592 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
593 truncate_data_blocks_range(&dn, count);
594 free_from += count;
595 }
596
597 f2fs_put_dnode(&dn);
598 free_next:
599 err = truncate_inode_blocks(inode, free_from);
600 out:
601 if (lock)
602 f2fs_unlock_op(sbi);
603 free_partial:
604 /* lastly zero out the first data page */
605 if (!err)
606 err = truncate_partial_data_page(inode, from, truncate_page);
607
608 trace_f2fs_truncate_blocks_exit(inode, err);
609 return err;
610 }
611
612 int f2fs_truncate(struct inode *inode)
613 {
614 int err;
615
616 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
617 S_ISLNK(inode->i_mode)))
618 return 0;
619
620 trace_f2fs_truncate(inode);
621
622 /* we should check inline_data size */
623 if (!f2fs_may_inline_data(inode)) {
624 err = f2fs_convert_inline_inode(inode);
625 if (err)
626 return err;
627 }
628
629 err = truncate_blocks(inode, i_size_read(inode), true);
630 if (err)
631 return err;
632
633 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
634 mark_inode_dirty_sync(inode);
635 return 0;
636 }
637
638 int f2fs_getattr(struct vfsmount *mnt,
639 struct dentry *dentry, struct kstat *stat)
640 {
641 struct inode *inode = d_inode(dentry);
642 generic_fillattr(inode, stat);
643 stat->blocks <<= 3;
644 return 0;
645 }
646
647 #ifdef CONFIG_F2FS_FS_POSIX_ACL
648 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
649 {
650 unsigned int ia_valid = attr->ia_valid;
651
652 if (ia_valid & ATTR_UID)
653 inode->i_uid = attr->ia_uid;
654 if (ia_valid & ATTR_GID)
655 inode->i_gid = attr->ia_gid;
656 if (ia_valid & ATTR_ATIME)
657 inode->i_atime = timespec_trunc(attr->ia_atime,
658 inode->i_sb->s_time_gran);
659 if (ia_valid & ATTR_MTIME)
660 inode->i_mtime = timespec_trunc(attr->ia_mtime,
661 inode->i_sb->s_time_gran);
662 if (ia_valid & ATTR_CTIME)
663 inode->i_ctime = timespec_trunc(attr->ia_ctime,
664 inode->i_sb->s_time_gran);
665 if (ia_valid & ATTR_MODE) {
666 umode_t mode = attr->ia_mode;
667
668 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
669 mode &= ~S_ISGID;
670 set_acl_inode(inode, mode);
671 }
672 }
673 #else
674 #define __setattr_copy setattr_copy
675 #endif
676
677 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
678 {
679 struct inode *inode = d_inode(dentry);
680 int err;
681
682 err = inode_change_ok(inode, attr);
683 if (err)
684 return err;
685
686 if (attr->ia_valid & ATTR_SIZE) {
687 if (f2fs_encrypted_inode(inode) &&
688 fscrypt_get_encryption_info(inode))
689 return -EACCES;
690
691 if (attr->ia_size <= i_size_read(inode)) {
692 truncate_setsize(inode, attr->ia_size);
693 err = f2fs_truncate(inode);
694 if (err)
695 return err;
696 f2fs_balance_fs(F2FS_I_SB(inode), true);
697 } else {
698 /*
699 * do not trim all blocks after i_size if target size is
700 * larger than i_size.
701 */
702 truncate_setsize(inode, attr->ia_size);
703
704 /* should convert inline inode here */
705 if (!f2fs_may_inline_data(inode)) {
706 err = f2fs_convert_inline_inode(inode);
707 if (err)
708 return err;
709 }
710 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
711 }
712 }
713
714 __setattr_copy(inode, attr);
715
716 if (attr->ia_valid & ATTR_MODE) {
717 err = posix_acl_chmod(inode, get_inode_mode(inode));
718 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
719 inode->i_mode = F2FS_I(inode)->i_acl_mode;
720 clear_inode_flag(inode, FI_ACL_MODE);
721 }
722 }
723
724 mark_inode_dirty_sync(inode);
725 return err;
726 }
727
728 const struct inode_operations f2fs_file_inode_operations = {
729 .getattr = f2fs_getattr,
730 .setattr = f2fs_setattr,
731 .get_acl = f2fs_get_acl,
732 .set_acl = f2fs_set_acl,
733 #ifdef CONFIG_F2FS_FS_XATTR
734 .setxattr = generic_setxattr,
735 .getxattr = generic_getxattr,
736 .listxattr = f2fs_listxattr,
737 .removexattr = generic_removexattr,
738 #endif
739 .fiemap = f2fs_fiemap,
740 };
741
742 static int fill_zero(struct inode *inode, pgoff_t index,
743 loff_t start, loff_t len)
744 {
745 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
746 struct page *page;
747
748 if (!len)
749 return 0;
750
751 f2fs_balance_fs(sbi, true);
752
753 f2fs_lock_op(sbi);
754 page = get_new_data_page(inode, NULL, index, false);
755 f2fs_unlock_op(sbi);
756
757 if (IS_ERR(page))
758 return PTR_ERR(page);
759
760 f2fs_wait_on_page_writeback(page, DATA, true);
761 zero_user(page, start, len);
762 set_page_dirty(page);
763 f2fs_put_page(page, 1);
764 return 0;
765 }
766
767 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
768 {
769 int err;
770
771 while (pg_start < pg_end) {
772 struct dnode_of_data dn;
773 pgoff_t end_offset, count;
774
775 set_new_dnode(&dn, inode, NULL, NULL, 0);
776 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
777 if (err) {
778 if (err == -ENOENT) {
779 pg_start++;
780 continue;
781 }
782 return err;
783 }
784
785 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
786 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
787
788 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
789
790 truncate_data_blocks_range(&dn, count);
791 f2fs_put_dnode(&dn);
792
793 pg_start += count;
794 }
795 return 0;
796 }
797
798 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
799 {
800 pgoff_t pg_start, pg_end;
801 loff_t off_start, off_end;
802 int ret;
803
804 ret = f2fs_convert_inline_inode(inode);
805 if (ret)
806 return ret;
807
808 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
809 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
810
811 off_start = offset & (PAGE_SIZE - 1);
812 off_end = (offset + len) & (PAGE_SIZE - 1);
813
814 if (pg_start == pg_end) {
815 ret = fill_zero(inode, pg_start, off_start,
816 off_end - off_start);
817 if (ret)
818 return ret;
819 } else {
820 if (off_start) {
821 ret = fill_zero(inode, pg_start++, off_start,
822 PAGE_SIZE - off_start);
823 if (ret)
824 return ret;
825 }
826 if (off_end) {
827 ret = fill_zero(inode, pg_end, 0, off_end);
828 if (ret)
829 return ret;
830 }
831
832 if (pg_start < pg_end) {
833 struct address_space *mapping = inode->i_mapping;
834 loff_t blk_start, blk_end;
835 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
836
837 f2fs_balance_fs(sbi, true);
838
839 blk_start = (loff_t)pg_start << PAGE_SHIFT;
840 blk_end = (loff_t)pg_end << PAGE_SHIFT;
841 truncate_inode_pages_range(mapping, blk_start,
842 blk_end - 1);
843
844 f2fs_lock_op(sbi);
845 ret = truncate_hole(inode, pg_start, pg_end);
846 f2fs_unlock_op(sbi);
847 }
848 }
849
850 return ret;
851 }
852
853 static int __exchange_data_block(struct inode *inode, pgoff_t src,
854 pgoff_t dst, bool full)
855 {
856 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
857 struct dnode_of_data dn;
858 block_t new_addr;
859 bool do_replace = false;
860 int ret;
861
862 set_new_dnode(&dn, inode, NULL, NULL, 0);
863 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
864 if (ret && ret != -ENOENT) {
865 return ret;
866 } else if (ret == -ENOENT) {
867 new_addr = NULL_ADDR;
868 } else {
869 new_addr = dn.data_blkaddr;
870 if (!is_checkpointed_data(sbi, new_addr)) {
871 /* do not invalidate this block address */
872 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
873 do_replace = true;
874 }
875 f2fs_put_dnode(&dn);
876 }
877
878 if (new_addr == NULL_ADDR)
879 return full ? truncate_hole(inode, dst, dst + 1) : 0;
880
881 if (do_replace) {
882 struct page *ipage;
883 struct node_info ni;
884
885 if (test_opt(sbi, LFS)) {
886 ret = -ENOTSUPP;
887 goto err_out;
888 }
889
890 ipage = get_node_page(sbi, inode->i_ino);
891 if (IS_ERR(ipage)) {
892 ret = PTR_ERR(ipage);
893 goto err_out;
894 }
895
896 set_new_dnode(&dn, inode, ipage, NULL, 0);
897 ret = f2fs_reserve_block(&dn, dst);
898 if (ret)
899 goto err_out;
900
901 truncate_data_blocks_range(&dn, 1);
902
903 get_node_info(sbi, dn.nid, &ni);
904 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
905 ni.version, true, false);
906 f2fs_put_dnode(&dn);
907 } else {
908 struct page *psrc, *pdst;
909
910 psrc = get_lock_data_page(inode, src, true);
911 if (IS_ERR(psrc))
912 return PTR_ERR(psrc);
913 pdst = get_new_data_page(inode, NULL, dst, true);
914 if (IS_ERR(pdst)) {
915 f2fs_put_page(psrc, 1);
916 return PTR_ERR(pdst);
917 }
918 f2fs_copy_page(psrc, pdst);
919 set_page_dirty(pdst);
920 f2fs_put_page(pdst, 1);
921 f2fs_put_page(psrc, 1);
922
923 return truncate_hole(inode, src, src + 1);
924 }
925 return 0;
926
927 err_out:
928 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
929 f2fs_update_data_blkaddr(&dn, new_addr);
930 f2fs_put_dnode(&dn);
931 }
932 return ret;
933 }
934
935 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
936 {
937 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
938 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
939 int ret = 0;
940
941 for (; end < nrpages; start++, end++) {
942 f2fs_balance_fs(sbi, true);
943 f2fs_lock_op(sbi);
944 ret = __exchange_data_block(inode, end, start, true);
945 f2fs_unlock_op(sbi);
946 if (ret)
947 break;
948 }
949 return ret;
950 }
951
952 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
953 {
954 pgoff_t pg_start, pg_end;
955 loff_t new_size;
956 int ret;
957
958 if (offset + len >= i_size_read(inode))
959 return -EINVAL;
960
961 /* collapse range should be aligned to block size of f2fs. */
962 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
963 return -EINVAL;
964
965 ret = f2fs_convert_inline_inode(inode);
966 if (ret)
967 return ret;
968
969 pg_start = offset >> PAGE_SHIFT;
970 pg_end = (offset + len) >> PAGE_SHIFT;
971
972 /* write out all dirty pages from offset */
973 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
974 if (ret)
975 return ret;
976
977 truncate_pagecache(inode, offset);
978
979 ret = f2fs_do_collapse(inode, pg_start, pg_end);
980 if (ret)
981 return ret;
982
983 /* write out all moved pages, if possible */
984 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
985 truncate_pagecache(inode, offset);
986
987 new_size = i_size_read(inode) - len;
988 truncate_pagecache(inode, new_size);
989
990 ret = truncate_blocks(inode, new_size, true);
991 if (!ret)
992 f2fs_i_size_write(inode, new_size);
993
994 return ret;
995 }
996
997 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
998 pgoff_t end)
999 {
1000 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1001 pgoff_t index = start;
1002 unsigned int ofs_in_node = dn->ofs_in_node;
1003 blkcnt_t count = 0;
1004 int ret;
1005
1006 for (; index < end; index++, dn->ofs_in_node++) {
1007 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1008 count++;
1009 }
1010
1011 dn->ofs_in_node = ofs_in_node;
1012 ret = reserve_new_blocks(dn, count);
1013 if (ret)
1014 return ret;
1015
1016 dn->ofs_in_node = ofs_in_node;
1017 for (index = start; index < end; index++, dn->ofs_in_node++) {
1018 dn->data_blkaddr =
1019 datablock_addr(dn->node_page, dn->ofs_in_node);
1020 /*
1021 * reserve_new_blocks will not guarantee entire block
1022 * allocation.
1023 */
1024 if (dn->data_blkaddr == NULL_ADDR) {
1025 ret = -ENOSPC;
1026 break;
1027 }
1028 if (dn->data_blkaddr != NEW_ADDR) {
1029 invalidate_blocks(sbi, dn->data_blkaddr);
1030 dn->data_blkaddr = NEW_ADDR;
1031 set_data_blkaddr(dn);
1032 }
1033 }
1034
1035 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1036
1037 return ret;
1038 }
1039
1040 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1041 int mode)
1042 {
1043 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1044 struct address_space *mapping = inode->i_mapping;
1045 pgoff_t index, pg_start, pg_end;
1046 loff_t new_size = i_size_read(inode);
1047 loff_t off_start, off_end;
1048 int ret = 0;
1049
1050 ret = inode_newsize_ok(inode, (len + offset));
1051 if (ret)
1052 return ret;
1053
1054 ret = f2fs_convert_inline_inode(inode);
1055 if (ret)
1056 return ret;
1057
1058 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1059 if (ret)
1060 return ret;
1061
1062 truncate_pagecache_range(inode, offset, offset + len - 1);
1063
1064 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1065 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1066
1067 off_start = offset & (PAGE_SIZE - 1);
1068 off_end = (offset + len) & (PAGE_SIZE - 1);
1069
1070 if (pg_start == pg_end) {
1071 ret = fill_zero(inode, pg_start, off_start,
1072 off_end - off_start);
1073 if (ret)
1074 return ret;
1075
1076 if (offset + len > new_size)
1077 new_size = offset + len;
1078 new_size = max_t(loff_t, new_size, offset + len);
1079 } else {
1080 if (off_start) {
1081 ret = fill_zero(inode, pg_start++, off_start,
1082 PAGE_SIZE - off_start);
1083 if (ret)
1084 return ret;
1085
1086 new_size = max_t(loff_t, new_size,
1087 (loff_t)pg_start << PAGE_SHIFT);
1088 }
1089
1090 for (index = pg_start; index < pg_end;) {
1091 struct dnode_of_data dn;
1092 unsigned int end_offset;
1093 pgoff_t end;
1094
1095 f2fs_lock_op(sbi);
1096
1097 set_new_dnode(&dn, inode, NULL, NULL, 0);
1098 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1099 if (ret) {
1100 f2fs_unlock_op(sbi);
1101 goto out;
1102 }
1103
1104 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1105 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1106
1107 ret = f2fs_do_zero_range(&dn, index, end);
1108 f2fs_put_dnode(&dn);
1109 f2fs_unlock_op(sbi);
1110 if (ret)
1111 goto out;
1112
1113 index = end;
1114 new_size = max_t(loff_t, new_size,
1115 (loff_t)index << PAGE_SHIFT);
1116 }
1117
1118 if (off_end) {
1119 ret = fill_zero(inode, pg_end, 0, off_end);
1120 if (ret)
1121 goto out;
1122
1123 new_size = max_t(loff_t, new_size, offset + len);
1124 }
1125 }
1126
1127 out:
1128 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1129 f2fs_i_size_write(inode, new_size);
1130
1131 return ret;
1132 }
1133
1134 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1135 {
1136 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1137 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1138 loff_t new_size;
1139 int ret = 0;
1140
1141 new_size = i_size_read(inode) + len;
1142 if (new_size > inode->i_sb->s_maxbytes)
1143 return -EFBIG;
1144
1145 if (offset >= i_size_read(inode))
1146 return -EINVAL;
1147
1148 /* insert range should be aligned to block size of f2fs. */
1149 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1150 return -EINVAL;
1151
1152 ret = f2fs_convert_inline_inode(inode);
1153 if (ret)
1154 return ret;
1155
1156 f2fs_balance_fs(sbi, true);
1157
1158 ret = truncate_blocks(inode, i_size_read(inode), true);
1159 if (ret)
1160 return ret;
1161
1162 /* write out all dirty pages from offset */
1163 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1164 if (ret)
1165 return ret;
1166
1167 truncate_pagecache(inode, offset);
1168
1169 pg_start = offset >> PAGE_SHIFT;
1170 pg_end = (offset + len) >> PAGE_SHIFT;
1171 delta = pg_end - pg_start;
1172 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1173
1174 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1175 f2fs_lock_op(sbi);
1176 ret = __exchange_data_block(inode, idx, idx + delta, false);
1177 f2fs_unlock_op(sbi);
1178 if (ret)
1179 break;
1180 }
1181
1182 /* write out all moved pages, if possible */
1183 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1184 truncate_pagecache(inode, offset);
1185
1186 if (!ret)
1187 f2fs_i_size_write(inode, new_size);
1188 return ret;
1189 }
1190
1191 static int expand_inode_data(struct inode *inode, loff_t offset,
1192 loff_t len, int mode)
1193 {
1194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1195 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1196 pgoff_t pg_end;
1197 loff_t new_size = i_size_read(inode);
1198 loff_t off_end;
1199 int ret;
1200
1201 ret = inode_newsize_ok(inode, (len + offset));
1202 if (ret)
1203 return ret;
1204
1205 ret = f2fs_convert_inline_inode(inode);
1206 if (ret)
1207 return ret;
1208
1209 f2fs_balance_fs(sbi, true);
1210
1211 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1212 off_end = (offset + len) & (PAGE_SIZE - 1);
1213
1214 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1215 map.m_len = pg_end - map.m_lblk;
1216 if (off_end)
1217 map.m_len++;
1218
1219 ret = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1220 if (ret) {
1221 pgoff_t last_off;
1222
1223 if (!map.m_len)
1224 return ret;
1225
1226 last_off = map.m_lblk + map.m_len - 1;
1227
1228 /* update new size to the failed position */
1229 new_size = (last_off == pg_end) ? offset + len:
1230 (loff_t)(last_off + 1) << PAGE_SHIFT;
1231 } else {
1232 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1233 }
1234
1235 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1236 f2fs_i_size_write(inode, new_size);
1237
1238 return ret;
1239 }
1240
1241 static long f2fs_fallocate(struct file *file, int mode,
1242 loff_t offset, loff_t len)
1243 {
1244 struct inode *inode = file_inode(file);
1245 long ret = 0;
1246
1247 /* f2fs only support ->fallocate for regular file */
1248 if (!S_ISREG(inode->i_mode))
1249 return -EINVAL;
1250
1251 if (f2fs_encrypted_inode(inode) &&
1252 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1253 return -EOPNOTSUPP;
1254
1255 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1256 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1257 FALLOC_FL_INSERT_RANGE))
1258 return -EOPNOTSUPP;
1259
1260 inode_lock(inode);
1261
1262 if (mode & FALLOC_FL_PUNCH_HOLE) {
1263 if (offset >= inode->i_size)
1264 goto out;
1265
1266 ret = punch_hole(inode, offset, len);
1267 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1268 ret = f2fs_collapse_range(inode, offset, len);
1269 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1270 ret = f2fs_zero_range(inode, offset, len, mode);
1271 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1272 ret = f2fs_insert_range(inode, offset, len);
1273 } else {
1274 ret = expand_inode_data(inode, offset, len, mode);
1275 }
1276
1277 if (!ret) {
1278 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1279 mark_inode_dirty_sync(inode);
1280 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1281 }
1282
1283 out:
1284 inode_unlock(inode);
1285
1286 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1287 return ret;
1288 }
1289
1290 static int f2fs_release_file(struct inode *inode, struct file *filp)
1291 {
1292 /*
1293 * f2fs_relase_file is called at every close calls. So we should
1294 * not drop any inmemory pages by close called by other process.
1295 */
1296 if (!(filp->f_mode & FMODE_WRITE) ||
1297 atomic_read(&inode->i_writecount) != 1)
1298 return 0;
1299
1300 /* some remained atomic pages should discarded */
1301 if (f2fs_is_atomic_file(inode))
1302 drop_inmem_pages(inode);
1303 if (f2fs_is_volatile_file(inode)) {
1304 clear_inode_flag(inode, FI_VOLATILE_FILE);
1305 set_inode_flag(inode, FI_DROP_CACHE);
1306 filemap_fdatawrite(inode->i_mapping);
1307 clear_inode_flag(inode, FI_DROP_CACHE);
1308 }
1309 return 0;
1310 }
1311
1312 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1313 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1314
1315 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1316 {
1317 if (S_ISDIR(mode))
1318 return flags;
1319 else if (S_ISREG(mode))
1320 return flags & F2FS_REG_FLMASK;
1321 else
1322 return flags & F2FS_OTHER_FLMASK;
1323 }
1324
1325 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1326 {
1327 struct inode *inode = file_inode(filp);
1328 struct f2fs_inode_info *fi = F2FS_I(inode);
1329 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1330 return put_user(flags, (int __user *)arg);
1331 }
1332
1333 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1334 {
1335 struct inode *inode = file_inode(filp);
1336 struct f2fs_inode_info *fi = F2FS_I(inode);
1337 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1338 unsigned int oldflags;
1339 int ret;
1340
1341 if (!inode_owner_or_capable(inode))
1342 return -EACCES;
1343
1344 if (get_user(flags, (int __user *)arg))
1345 return -EFAULT;
1346
1347 ret = mnt_want_write_file(filp);
1348 if (ret)
1349 return ret;
1350
1351 flags = f2fs_mask_flags(inode->i_mode, flags);
1352
1353 inode_lock(inode);
1354
1355 oldflags = fi->i_flags;
1356
1357 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1358 if (!capable(CAP_LINUX_IMMUTABLE)) {
1359 inode_unlock(inode);
1360 ret = -EPERM;
1361 goto out;
1362 }
1363 }
1364
1365 flags = flags & FS_FL_USER_MODIFIABLE;
1366 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1367 fi->i_flags = flags;
1368 inode_unlock(inode);
1369
1370 inode->i_ctime = CURRENT_TIME;
1371 f2fs_set_inode_flags(inode);
1372 out:
1373 mnt_drop_write_file(filp);
1374 return ret;
1375 }
1376
1377 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1378 {
1379 struct inode *inode = file_inode(filp);
1380
1381 return put_user(inode->i_generation, (int __user *)arg);
1382 }
1383
1384 static int f2fs_ioc_start_atomic_write(struct file *filp)
1385 {
1386 struct inode *inode = file_inode(filp);
1387 int ret;
1388
1389 if (!inode_owner_or_capable(inode))
1390 return -EACCES;
1391
1392 ret = mnt_want_write_file(filp);
1393 if (ret)
1394 return ret;
1395
1396 inode_lock(inode);
1397
1398 if (f2fs_is_atomic_file(inode))
1399 goto out;
1400
1401 ret = f2fs_convert_inline_inode(inode);
1402 if (ret)
1403 goto out;
1404
1405 set_inode_flag(inode, FI_ATOMIC_FILE);
1406 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1407
1408 if (!get_dirty_pages(inode))
1409 goto out;
1410
1411 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1412 "Unexpected flush for atomic writes: ino=%lu, npages=%lld",
1413 inode->i_ino, get_dirty_pages(inode));
1414 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1415 if (ret)
1416 clear_inode_flag(inode, FI_ATOMIC_FILE);
1417 out:
1418 inode_unlock(inode);
1419 mnt_drop_write_file(filp);
1420 return ret;
1421 }
1422
1423 static int f2fs_ioc_commit_atomic_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 inode_lock(inode);
1436
1437 if (f2fs_is_volatile_file(inode))
1438 goto err_out;
1439
1440 if (f2fs_is_atomic_file(inode)) {
1441 clear_inode_flag(inode, FI_ATOMIC_FILE);
1442 ret = commit_inmem_pages(inode);
1443 if (ret) {
1444 set_inode_flag(inode, FI_ATOMIC_FILE);
1445 goto err_out;
1446 }
1447 }
1448
1449 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1450 err_out:
1451 inode_unlock(inode);
1452 mnt_drop_write_file(filp);
1453 return ret;
1454 }
1455
1456 static int f2fs_ioc_start_volatile_write(struct file *filp)
1457 {
1458 struct inode *inode = file_inode(filp);
1459 int ret;
1460
1461 if (!inode_owner_or_capable(inode))
1462 return -EACCES;
1463
1464 ret = mnt_want_write_file(filp);
1465 if (ret)
1466 return ret;
1467
1468 inode_lock(inode);
1469
1470 if (f2fs_is_volatile_file(inode))
1471 goto out;
1472
1473 ret = f2fs_convert_inline_inode(inode);
1474 if (ret)
1475 goto out;
1476
1477 set_inode_flag(inode, FI_VOLATILE_FILE);
1478 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1479 out:
1480 inode_unlock(inode);
1481 mnt_drop_write_file(filp);
1482 return ret;
1483 }
1484
1485 static int f2fs_ioc_release_volatile_write(struct file *filp)
1486 {
1487 struct inode *inode = file_inode(filp);
1488 int ret;
1489
1490 if (!inode_owner_or_capable(inode))
1491 return -EACCES;
1492
1493 ret = mnt_want_write_file(filp);
1494 if (ret)
1495 return ret;
1496
1497 inode_lock(inode);
1498
1499 if (!f2fs_is_volatile_file(inode))
1500 goto out;
1501
1502 if (!f2fs_is_first_block_written(inode)) {
1503 ret = truncate_partial_data_page(inode, 0, true);
1504 goto out;
1505 }
1506
1507 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1508 out:
1509 inode_unlock(inode);
1510 mnt_drop_write_file(filp);
1511 return ret;
1512 }
1513
1514 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1515 {
1516 struct inode *inode = file_inode(filp);
1517 int ret;
1518
1519 if (!inode_owner_or_capable(inode))
1520 return -EACCES;
1521
1522 ret = mnt_want_write_file(filp);
1523 if (ret)
1524 return ret;
1525
1526 inode_lock(inode);
1527
1528 if (f2fs_is_atomic_file(inode))
1529 drop_inmem_pages(inode);
1530 if (f2fs_is_volatile_file(inode)) {
1531 clear_inode_flag(inode, FI_VOLATILE_FILE);
1532 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1533 }
1534
1535 inode_unlock(inode);
1536
1537 mnt_drop_write_file(filp);
1538 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1539 return ret;
1540 }
1541
1542 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1543 {
1544 struct inode *inode = file_inode(filp);
1545 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1546 struct super_block *sb = sbi->sb;
1547 __u32 in;
1548 int ret;
1549
1550 if (!capable(CAP_SYS_ADMIN))
1551 return -EPERM;
1552
1553 if (get_user(in, (__u32 __user *)arg))
1554 return -EFAULT;
1555
1556 ret = mnt_want_write_file(filp);
1557 if (ret)
1558 return ret;
1559
1560 switch (in) {
1561 case F2FS_GOING_DOWN_FULLSYNC:
1562 sb = freeze_bdev(sb->s_bdev);
1563 if (sb && !IS_ERR(sb)) {
1564 f2fs_stop_checkpoint(sbi, false);
1565 thaw_bdev(sb->s_bdev, sb);
1566 }
1567 break;
1568 case F2FS_GOING_DOWN_METASYNC:
1569 /* do checkpoint only */
1570 f2fs_sync_fs(sb, 1);
1571 f2fs_stop_checkpoint(sbi, false);
1572 break;
1573 case F2FS_GOING_DOWN_NOSYNC:
1574 f2fs_stop_checkpoint(sbi, false);
1575 break;
1576 case F2FS_GOING_DOWN_METAFLUSH:
1577 sync_meta_pages(sbi, META, LONG_MAX);
1578 f2fs_stop_checkpoint(sbi, false);
1579 break;
1580 default:
1581 ret = -EINVAL;
1582 goto out;
1583 }
1584 f2fs_update_time(sbi, REQ_TIME);
1585 out:
1586 mnt_drop_write_file(filp);
1587 return ret;
1588 }
1589
1590 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1591 {
1592 struct inode *inode = file_inode(filp);
1593 struct super_block *sb = inode->i_sb;
1594 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1595 struct fstrim_range range;
1596 int ret;
1597
1598 if (!capable(CAP_SYS_ADMIN))
1599 return -EPERM;
1600
1601 if (!blk_queue_discard(q))
1602 return -EOPNOTSUPP;
1603
1604 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1605 sizeof(range)))
1606 return -EFAULT;
1607
1608 ret = mnt_want_write_file(filp);
1609 if (ret)
1610 return ret;
1611
1612 range.minlen = max((unsigned int)range.minlen,
1613 q->limits.discard_granularity);
1614 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1615 mnt_drop_write_file(filp);
1616 if (ret < 0)
1617 return ret;
1618
1619 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1620 sizeof(range)))
1621 return -EFAULT;
1622 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1623 return 0;
1624 }
1625
1626 static bool uuid_is_nonzero(__u8 u[16])
1627 {
1628 int i;
1629
1630 for (i = 0; i < 16; i++)
1631 if (u[i])
1632 return true;
1633 return false;
1634 }
1635
1636 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1637 {
1638 struct fscrypt_policy policy;
1639 struct inode *inode = file_inode(filp);
1640 int ret;
1641
1642 if (copy_from_user(&policy, (struct fscrypt_policy __user *)arg,
1643 sizeof(policy)))
1644 return -EFAULT;
1645
1646 ret = mnt_want_write_file(filp);
1647 if (ret)
1648 return ret;
1649
1650 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1651 ret = fscrypt_process_policy(inode, &policy);
1652
1653 mnt_drop_write_file(filp);
1654 return ret;
1655 }
1656
1657 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1658 {
1659 struct fscrypt_policy policy;
1660 struct inode *inode = file_inode(filp);
1661 int err;
1662
1663 err = fscrypt_get_policy(inode, &policy);
1664 if (err)
1665 return err;
1666
1667 if (copy_to_user((struct fscrypt_policy __user *)arg, &policy, sizeof(policy)))
1668 return -EFAULT;
1669 return 0;
1670 }
1671
1672 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1673 {
1674 struct inode *inode = file_inode(filp);
1675 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1676 int err;
1677
1678 if (!f2fs_sb_has_crypto(inode->i_sb))
1679 return -EOPNOTSUPP;
1680
1681 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1682 goto got_it;
1683
1684 err = mnt_want_write_file(filp);
1685 if (err)
1686 return err;
1687
1688 /* update superblock with uuid */
1689 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1690
1691 err = f2fs_commit_super(sbi, false);
1692 if (err) {
1693 /* undo new data */
1694 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1695 mnt_drop_write_file(filp);
1696 return err;
1697 }
1698 mnt_drop_write_file(filp);
1699 got_it:
1700 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1701 16))
1702 return -EFAULT;
1703 return 0;
1704 }
1705
1706 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1707 {
1708 struct inode *inode = file_inode(filp);
1709 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1710 __u32 sync;
1711 int ret;
1712
1713 if (!capable(CAP_SYS_ADMIN))
1714 return -EPERM;
1715
1716 if (get_user(sync, (__u32 __user *)arg))
1717 return -EFAULT;
1718
1719 if (f2fs_readonly(sbi->sb))
1720 return -EROFS;
1721
1722 ret = mnt_want_write_file(filp);
1723 if (ret)
1724 return ret;
1725
1726 if (!sync) {
1727 if (!mutex_trylock(&sbi->gc_mutex)) {
1728 ret = -EBUSY;
1729 goto out;
1730 }
1731 } else {
1732 mutex_lock(&sbi->gc_mutex);
1733 }
1734
1735 ret = f2fs_gc(sbi, sync);
1736 out:
1737 mnt_drop_write_file(filp);
1738 return ret;
1739 }
1740
1741 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1742 {
1743 struct inode *inode = file_inode(filp);
1744 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1745 int ret;
1746
1747 if (!capable(CAP_SYS_ADMIN))
1748 return -EPERM;
1749
1750 if (f2fs_readonly(sbi->sb))
1751 return -EROFS;
1752
1753 ret = mnt_want_write_file(filp);
1754 if (ret)
1755 return ret;
1756
1757 ret = f2fs_sync_fs(sbi->sb, 1);
1758
1759 mnt_drop_write_file(filp);
1760 return ret;
1761 }
1762
1763 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1764 struct file *filp,
1765 struct f2fs_defragment *range)
1766 {
1767 struct inode *inode = file_inode(filp);
1768 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1769 struct extent_info ei;
1770 pgoff_t pg_start, pg_end;
1771 unsigned int blk_per_seg = sbi->blocks_per_seg;
1772 unsigned int total = 0, sec_num;
1773 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1774 block_t blk_end = 0;
1775 bool fragmented = false;
1776 int err;
1777
1778 /* if in-place-update policy is enabled, don't waste time here */
1779 if (need_inplace_update(inode))
1780 return -EINVAL;
1781
1782 pg_start = range->start >> PAGE_SHIFT;
1783 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1784
1785 f2fs_balance_fs(sbi, true);
1786
1787 inode_lock(inode);
1788
1789 /* writeback all dirty pages in the range */
1790 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1791 range->start + range->len - 1);
1792 if (err)
1793 goto out;
1794
1795 /*
1796 * lookup mapping info in extent cache, skip defragmenting if physical
1797 * block addresses are continuous.
1798 */
1799 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1800 if (ei.fofs + ei.len >= pg_end)
1801 goto out;
1802 }
1803
1804 map.m_lblk = pg_start;
1805
1806 /*
1807 * lookup mapping info in dnode page cache, skip defragmenting if all
1808 * physical block addresses are continuous even if there are hole(s)
1809 * in logical blocks.
1810 */
1811 while (map.m_lblk < pg_end) {
1812 map.m_len = pg_end - map.m_lblk;
1813 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1814 if (err)
1815 goto out;
1816
1817 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1818 map.m_lblk++;
1819 continue;
1820 }
1821
1822 if (blk_end && blk_end != map.m_pblk) {
1823 fragmented = true;
1824 break;
1825 }
1826 blk_end = map.m_pblk + map.m_len;
1827
1828 map.m_lblk += map.m_len;
1829 }
1830
1831 if (!fragmented)
1832 goto out;
1833
1834 map.m_lblk = pg_start;
1835 map.m_len = pg_end - pg_start;
1836
1837 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1838
1839 /*
1840 * make sure there are enough free section for LFS allocation, this can
1841 * avoid defragment running in SSR mode when free section are allocated
1842 * intensively
1843 */
1844 if (has_not_enough_free_secs(sbi, sec_num)) {
1845 err = -EAGAIN;
1846 goto out;
1847 }
1848
1849 while (map.m_lblk < pg_end) {
1850 pgoff_t idx;
1851 int cnt = 0;
1852
1853 do_map:
1854 map.m_len = pg_end - map.m_lblk;
1855 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1856 if (err)
1857 goto clear_out;
1858
1859 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1860 map.m_lblk++;
1861 continue;
1862 }
1863
1864 set_inode_flag(inode, FI_DO_DEFRAG);
1865
1866 idx = map.m_lblk;
1867 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1868 struct page *page;
1869
1870 page = get_lock_data_page(inode, idx, true);
1871 if (IS_ERR(page)) {
1872 err = PTR_ERR(page);
1873 goto clear_out;
1874 }
1875
1876 set_page_dirty(page);
1877 f2fs_put_page(page, 1);
1878
1879 idx++;
1880 cnt++;
1881 total++;
1882 }
1883
1884 map.m_lblk = idx;
1885
1886 if (idx < pg_end && cnt < blk_per_seg)
1887 goto do_map;
1888
1889 clear_inode_flag(inode, FI_DO_DEFRAG);
1890
1891 err = filemap_fdatawrite(inode->i_mapping);
1892 if (err)
1893 goto out;
1894 }
1895 clear_out:
1896 clear_inode_flag(inode, FI_DO_DEFRAG);
1897 out:
1898 inode_unlock(inode);
1899 if (!err)
1900 range->len = (u64)total << PAGE_SHIFT;
1901 return err;
1902 }
1903
1904 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1905 {
1906 struct inode *inode = file_inode(filp);
1907 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1908 struct f2fs_defragment range;
1909 int err;
1910
1911 if (!capable(CAP_SYS_ADMIN))
1912 return -EPERM;
1913
1914 if (!S_ISREG(inode->i_mode))
1915 return -EINVAL;
1916
1917 err = mnt_want_write_file(filp);
1918 if (err)
1919 return err;
1920
1921 if (f2fs_readonly(sbi->sb)) {
1922 err = -EROFS;
1923 goto out;
1924 }
1925
1926 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1927 sizeof(range))) {
1928 err = -EFAULT;
1929 goto out;
1930 }
1931
1932 /* verify alignment of offset & size */
1933 if (range.start & (F2FS_BLKSIZE - 1) ||
1934 range.len & (F2FS_BLKSIZE - 1)) {
1935 err = -EINVAL;
1936 goto out;
1937 }
1938
1939 err = f2fs_defragment_range(sbi, filp, &range);
1940 f2fs_update_time(sbi, REQ_TIME);
1941 if (err < 0)
1942 goto out;
1943
1944 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1945 sizeof(range)))
1946 err = -EFAULT;
1947 out:
1948 mnt_drop_write_file(filp);
1949 return err;
1950 }
1951
1952 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1953 {
1954 switch (cmd) {
1955 case F2FS_IOC_GETFLAGS:
1956 return f2fs_ioc_getflags(filp, arg);
1957 case F2FS_IOC_SETFLAGS:
1958 return f2fs_ioc_setflags(filp, arg);
1959 case F2FS_IOC_GETVERSION:
1960 return f2fs_ioc_getversion(filp, arg);
1961 case F2FS_IOC_START_ATOMIC_WRITE:
1962 return f2fs_ioc_start_atomic_write(filp);
1963 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1964 return f2fs_ioc_commit_atomic_write(filp);
1965 case F2FS_IOC_START_VOLATILE_WRITE:
1966 return f2fs_ioc_start_volatile_write(filp);
1967 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1968 return f2fs_ioc_release_volatile_write(filp);
1969 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1970 return f2fs_ioc_abort_volatile_write(filp);
1971 case F2FS_IOC_SHUTDOWN:
1972 return f2fs_ioc_shutdown(filp, arg);
1973 case FITRIM:
1974 return f2fs_ioc_fitrim(filp, arg);
1975 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1976 return f2fs_ioc_set_encryption_policy(filp, arg);
1977 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1978 return f2fs_ioc_get_encryption_policy(filp, arg);
1979 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1980 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1981 case F2FS_IOC_GARBAGE_COLLECT:
1982 return f2fs_ioc_gc(filp, arg);
1983 case F2FS_IOC_WRITE_CHECKPOINT:
1984 return f2fs_ioc_write_checkpoint(filp, arg);
1985 case F2FS_IOC_DEFRAGMENT:
1986 return f2fs_ioc_defragment(filp, arg);
1987 default:
1988 return -ENOTTY;
1989 }
1990 }
1991
1992 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1993 {
1994 struct file *file = iocb->ki_filp;
1995 struct inode *inode = file_inode(file);
1996 ssize_t ret;
1997
1998 if (f2fs_encrypted_inode(inode) &&
1999 !fscrypt_has_encryption_key(inode) &&
2000 fscrypt_get_encryption_info(inode))
2001 return -EACCES;
2002
2003 inode_lock(inode);
2004 ret = generic_write_checks(iocb, from);
2005 if (ret > 0) {
2006 ret = f2fs_preallocate_blocks(iocb, from);
2007 if (!ret)
2008 ret = __generic_file_write_iter(iocb, from);
2009 }
2010 inode_unlock(inode);
2011
2012 if (ret > 0)
2013 ret = generic_write_sync(iocb, ret);
2014 return ret;
2015 }
2016
2017 #ifdef CONFIG_COMPAT
2018 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2019 {
2020 switch (cmd) {
2021 case F2FS_IOC32_GETFLAGS:
2022 cmd = F2FS_IOC_GETFLAGS;
2023 break;
2024 case F2FS_IOC32_SETFLAGS:
2025 cmd = F2FS_IOC_SETFLAGS;
2026 break;
2027 case F2FS_IOC32_GETVERSION:
2028 cmd = F2FS_IOC_GETVERSION;
2029 break;
2030 case F2FS_IOC_START_ATOMIC_WRITE:
2031 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2032 case F2FS_IOC_START_VOLATILE_WRITE:
2033 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2034 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2035 case F2FS_IOC_SHUTDOWN:
2036 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2037 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2038 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2039 case F2FS_IOC_GARBAGE_COLLECT:
2040 case F2FS_IOC_WRITE_CHECKPOINT:
2041 case F2FS_IOC_DEFRAGMENT:
2042 break;
2043 default:
2044 return -ENOIOCTLCMD;
2045 }
2046 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2047 }
2048 #endif
2049
2050 const struct file_operations f2fs_file_operations = {
2051 .llseek = f2fs_llseek,
2052 .read_iter = generic_file_read_iter,
2053 .write_iter = f2fs_file_write_iter,
2054 .open = f2fs_file_open,
2055 .release = f2fs_release_file,
2056 .mmap = f2fs_file_mmap,
2057 .fsync = f2fs_sync_file,
2058 .fallocate = f2fs_fallocate,
2059 .unlocked_ioctl = f2fs_ioctl,
2060 #ifdef CONFIG_COMPAT
2061 .compat_ioctl = f2fs_compat_ioctl,
2062 #endif
2063 .splice_read = generic_file_splice_read,
2064 .splice_write = iter_file_splice_write,
2065 };
Linux kernel - Deliverables
This page took 0.07313 seconds and 5 git commands to generate.