?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ia64: salinfo: use a waitqueue instead a sema down/up combo
[linux/fpc-iii.git] / fs / f2fs / file.c
blobf4c0086655c4bf448d76e017d2d6e47e0f2f6de4
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 SetPageUptodate(page);
85
86 trace_f2fs_vm_page_mkwrite(page, DATA);
87 mapped:
88 /* fill the page */
89 f2fs_wait_on_page_writeback(page, DATA, false);
90
91 /* wait for GCed encrypted page writeback */
92 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
93 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
94
95 /* if gced page is attached, don't write to cold segment */
96 clear_cold_data(page);
97 out:
98 sb_end_pagefault(inode->i_sb);
99 f2fs_update_time(sbi, REQ_TIME);
100 return block_page_mkwrite_return(err);
101 }
102
103 static const struct vm_operations_struct f2fs_file_vm_ops = {
104 .fault = filemap_fault,
105 .map_pages = filemap_map_pages,
106 .page_mkwrite = f2fs_vm_page_mkwrite,
107 };
108
109 static int get_parent_ino(struct inode *inode, nid_t *pino)
110 {
111 struct dentry *dentry;
112
113 inode = igrab(inode);
114 dentry = d_find_any_alias(inode);
115 iput(inode);
116 if (!dentry)
117 return 0;
118
119 if (update_dent_inode(inode, inode, &dentry->d_name)) {
120 dput(dentry);
121 return 0;
122 }
123
124 *pino = parent_ino(dentry);
125 dput(dentry);
126 return 1;
127 }
128
129 static inline bool need_do_checkpoint(struct inode *inode)
130 {
131 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
132 bool need_cp = false;
133
134 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
135 need_cp = true;
136 else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
137 need_cp = true;
138 else if (file_wrong_pino(inode))
139 need_cp = true;
140 else if (!space_for_roll_forward(sbi))
141 need_cp = true;
142 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
143 need_cp = true;
144 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
145 need_cp = true;
146 else if (test_opt(sbi, FASTBOOT))
147 need_cp = true;
148 else if (sbi->active_logs == 2)
149 need_cp = true;
150
151 return need_cp;
152 }
153
154 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
155 {
156 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
157 bool ret = false;
158 /* But we need to avoid that there are some inode updates */
159 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
160 ret = true;
161 f2fs_put_page(i, 0);
162 return ret;
163 }
164
165 static void try_to_fix_pino(struct inode *inode)
166 {
167 struct f2fs_inode_info *fi = F2FS_I(inode);
168 nid_t pino;
169
170 down_write(&fi->i_sem);
171 fi->xattr_ver = 0;
172 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
173 get_parent_ino(inode, &pino)) {
174 fi->i_pino = pino;
175 file_got_pino(inode);
176 up_write(&fi->i_sem);
177
178 mark_inode_dirty_sync(inode);
179 f2fs_write_inode(inode, NULL);
180 } else {
181 up_write(&fi->i_sem);
182 }
183 }
184
185 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
186 int datasync, bool atomic)
187 {
188 struct inode *inode = file->f_mapping->host;
189 struct f2fs_inode_info *fi = F2FS_I(inode);
190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
191 nid_t ino = inode->i_ino;
192 int ret = 0;
193 bool need_cp = false;
194 struct writeback_control wbc = {
195 .sync_mode = WB_SYNC_ALL,
196 .nr_to_write = LONG_MAX,
197 .for_reclaim = 0,
198 };
199
200 if (unlikely(f2fs_readonly(inode->i_sb)))
201 return 0;
202
203 trace_f2fs_sync_file_enter(inode);
204
205 /* if fdatasync is triggered, let's do in-place-update */
206 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
207 set_inode_flag(fi, FI_NEED_IPU);
208 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
209 clear_inode_flag(fi, FI_NEED_IPU);
210
211 if (ret) {
212 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
213 return ret;
214 }
215
216 /* if the inode is dirty, let's recover all the time */
217 if (!datasync) {
218 f2fs_write_inode(inode, NULL);
219 goto go_write;
220 }
221
222 /*
223 * if there is no written data, don't waste time to write recovery info.
224 */
225 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
226 !exist_written_data(sbi, ino, APPEND_INO)) {
227
228 /* it may call write_inode just prior to fsync */
229 if (need_inode_page_update(sbi, ino))
230 goto go_write;
231
232 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
233 exist_written_data(sbi, ino, UPDATE_INO))
234 goto flush_out;
235 goto out;
236 }
237 go_write:
238 /*
239 * Both of fdatasync() and fsync() are able to be recovered from
240 * sudden-power-off.
241 */
242 down_read(&fi->i_sem);
243 need_cp = need_do_checkpoint(inode);
244 up_read(&fi->i_sem);
245
246 if (need_cp) {
247 /* all the dirty node pages should be flushed for POR */
248 ret = f2fs_sync_fs(inode->i_sb, 1);
249
250 /*
251 * We've secured consistency through sync_fs. Following pino
252 * will be used only for fsynced inodes after checkpoint.
253 */
254 try_to_fix_pino(inode);
255 clear_inode_flag(fi, FI_APPEND_WRITE);
256 clear_inode_flag(fi, FI_UPDATE_WRITE);
257 goto out;
258 }
259 sync_nodes:
260 ret = fsync_node_pages(sbi, ino, &wbc, atomic);
261 if (ret)
262 goto out;
263
264 /* if cp_error was enabled, we should avoid infinite loop */
265 if (unlikely(f2fs_cp_error(sbi))) {
266 ret = -EIO;
267 goto out;
268 }
269
270 if (need_inode_block_update(sbi, ino)) {
271 mark_inode_dirty_sync(inode);
272 f2fs_write_inode(inode, NULL);
273 goto sync_nodes;
274 }
275
276 ret = wait_on_node_pages_writeback(sbi, ino);
277 if (ret)
278 goto out;
279
280 /* once recovery info is written, don't need to tack this */
281 remove_ino_entry(sbi, ino, APPEND_INO);
282 clear_inode_flag(fi, FI_APPEND_WRITE);
283 flush_out:
284 remove_ino_entry(sbi, ino, UPDATE_INO);
285 clear_inode_flag(fi, FI_UPDATE_WRITE);
286 ret = f2fs_issue_flush(sbi);
287 f2fs_update_time(sbi, REQ_TIME);
288 out:
289 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
290 f2fs_trace_ios(NULL, 1);
291 return ret;
292 }
293
294 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
295 {
296 return f2fs_do_sync_file(file, start, end, datasync, false);
297 }
298
299 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
300 pgoff_t pgofs, int whence)
301 {
302 struct pagevec pvec;
303 int nr_pages;
304
305 if (whence != SEEK_DATA)
306 return 0;
307
308 /* find first dirty page index */
309 pagevec_init(&pvec, 0);
310 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
311 PAGECACHE_TAG_DIRTY, 1);
312 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
313 pagevec_release(&pvec);
314 return pgofs;
315 }
316
317 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
318 int whence)
319 {
320 switch (whence) {
321 case SEEK_DATA:
322 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
323 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
324 return true;
325 break;
326 case SEEK_HOLE:
327 if (blkaddr == NULL_ADDR)
328 return true;
329 break;
330 }
331 return false;
332 }
333
334 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
335 {
336 struct inode *inode = file->f_mapping->host;
337 loff_t maxbytes = inode->i_sb->s_maxbytes;
338 struct dnode_of_data dn;
339 pgoff_t pgofs, end_offset, dirty;
340 loff_t data_ofs = offset;
341 loff_t isize;
342 int err = 0;
343
344 inode_lock(inode);
345
346 isize = i_size_read(inode);
347 if (offset >= isize)
348 goto fail;
349
350 /* handle inline data case */
351 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
352 if (whence == SEEK_HOLE)
353 data_ofs = isize;
354 goto found;
355 }
356
357 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
358
359 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
360
361 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
362 set_new_dnode(&dn, inode, NULL, NULL, 0);
363 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
364 if (err && err != -ENOENT) {
365 goto fail;
366 } else if (err == -ENOENT) {
367 /* direct node does not exists */
368 if (whence == SEEK_DATA) {
369 pgofs = get_next_page_offset(&dn, pgofs);
370 continue;
371 } else {
372 goto found;
373 }
374 }
375
376 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
377
378 /* find data/hole in dnode block */
379 for (; dn.ofs_in_node < end_offset;
380 dn.ofs_in_node++, pgofs++,
381 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
382 block_t blkaddr;
383 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
384
385 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
386 f2fs_put_dnode(&dn);
387 goto found;
388 }
389 }
390 f2fs_put_dnode(&dn);
391 }
392
393 if (whence == SEEK_DATA)
394 goto fail;
395 found:
396 if (whence == SEEK_HOLE && data_ofs > isize)
397 data_ofs = isize;
398 inode_unlock(inode);
399 return vfs_setpos(file, data_ofs, maxbytes);
400 fail:
401 inode_unlock(inode);
402 return -ENXIO;
403 }
404
405 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
406 {
407 struct inode *inode = file->f_mapping->host;
408 loff_t maxbytes = inode->i_sb->s_maxbytes;
409
410 switch (whence) {
411 case SEEK_SET:
412 case SEEK_CUR:
413 case SEEK_END:
414 return generic_file_llseek_size(file, offset, whence,
415 maxbytes, i_size_read(inode));
416 case SEEK_DATA:
417 case SEEK_HOLE:
418 if (offset < 0)
419 return -ENXIO;
420 return f2fs_seek_block(file, offset, whence);
421 }
422
423 return -EINVAL;
424 }
425
426 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
427 {
428 struct inode *inode = file_inode(file);
429 int err;
430
431 if (f2fs_encrypted_inode(inode)) {
432 err = fscrypt_get_encryption_info(inode);
433 if (err)
434 return 0;
435 if (!f2fs_encrypted_inode(inode))
436 return -ENOKEY;
437 }
438
439 /* we don't need to use inline_data strictly */
440 err = f2fs_convert_inline_inode(inode);
441 if (err)
442 return err;
443
444 file_accessed(file);
445 vma->vm_ops = &f2fs_file_vm_ops;
446 return 0;
447 }
448
449 static int f2fs_file_open(struct inode *inode, struct file *filp)
450 {
451 int ret = generic_file_open(inode, filp);
452 struct dentry *dir;
453
454 if (!ret && f2fs_encrypted_inode(inode)) {
455 ret = fscrypt_get_encryption_info(inode);
456 if (ret)
457 return -EACCES;
458 if (!fscrypt_has_encryption_key(inode))
459 return -ENOKEY;
460 }
461 dir = dget_parent(file_dentry(filp));
462 if (f2fs_encrypted_inode(d_inode(dir)) &&
463 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
464 dput(dir);
465 return -EPERM;
466 }
467 dput(dir);
468 return ret;
469 }
470
471 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
472 {
473 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
474 struct f2fs_node *raw_node;
475 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
476 __le32 *addr;
477
478 raw_node = F2FS_NODE(dn->node_page);
479 addr = blkaddr_in_node(raw_node) + ofs;
480
481 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
482 block_t blkaddr = le32_to_cpu(*addr);
483 if (blkaddr == NULL_ADDR)
484 continue;
485
486 dn->data_blkaddr = NULL_ADDR;
487 set_data_blkaddr(dn);
488 invalidate_blocks(sbi, blkaddr);
489 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
490 clear_inode_flag(F2FS_I(dn->inode),
491 FI_FIRST_BLOCK_WRITTEN);
492 nr_free++;
493 }
494
495 if (nr_free) {
496 pgoff_t fofs;
497 /*
498 * once we invalidate valid blkaddr in range [ofs, ofs + count],
499 * we will invalidate all blkaddr in the whole range.
500 */
501 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
502 dn->inode) + ofs;
503 f2fs_update_extent_cache_range(dn, fofs, 0, len);
504 dec_valid_block_count(sbi, dn->inode, nr_free);
505 sync_inode_page(dn);
506 }
507 dn->ofs_in_node = ofs;
508
509 f2fs_update_time(sbi, REQ_TIME);
510 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
511 dn->ofs_in_node, nr_free);
512 return nr_free;
513 }
514
515 void truncate_data_blocks(struct dnode_of_data *dn)
516 {
517 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
518 }
519
520 static int truncate_partial_data_page(struct inode *inode, u64 from,
521 bool cache_only)
522 {
523 unsigned offset = from & (PAGE_SIZE - 1);
524 pgoff_t index = from >> PAGE_SHIFT;
525 struct address_space *mapping = inode->i_mapping;
526 struct page *page;
527
528 if (!offset && !cache_only)
529 return 0;
530
531 if (cache_only) {
532 page = f2fs_grab_cache_page(mapping, index, false);
533 if (page && PageUptodate(page))
534 goto truncate_out;
535 f2fs_put_page(page, 1);
536 return 0;
537 }
538
539 page = get_lock_data_page(inode, index, true);
540 if (IS_ERR(page))
541 return 0;
542 truncate_out:
543 f2fs_wait_on_page_writeback(page, DATA, true);
544 zero_user(page, offset, PAGE_SIZE - offset);
545 if (!cache_only || !f2fs_encrypted_inode(inode) ||
546 !S_ISREG(inode->i_mode))
547 set_page_dirty(page);
548 f2fs_put_page(page, 1);
549 return 0;
550 }
551
552 int truncate_blocks(struct inode *inode, u64 from, bool lock)
553 {
554 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
555 unsigned int blocksize = inode->i_sb->s_blocksize;
556 struct dnode_of_data dn;
557 pgoff_t free_from;
558 int count = 0, err = 0;
559 struct page *ipage;
560 bool truncate_page = false;
561
562 trace_f2fs_truncate_blocks_enter(inode, from);
563
564 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
565
566 if (free_from >= sbi->max_file_blocks)
567 goto free_partial;
568
569 if (lock)
570 f2fs_lock_op(sbi);
571
572 ipage = get_node_page(sbi, inode->i_ino);
573 if (IS_ERR(ipage)) {
574 err = PTR_ERR(ipage);
575 goto out;
576 }
577
578 if (f2fs_has_inline_data(inode)) {
579 if (truncate_inline_inode(ipage, from))
580 set_page_dirty(ipage);
581 f2fs_put_page(ipage, 1);
582 truncate_page = true;
583 goto out;
584 }
585
586 set_new_dnode(&dn, inode, ipage, NULL, 0);
587 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
588 if (err) {
589 if (err == -ENOENT)
590 goto free_next;
591 goto out;
592 }
593
594 count = ADDRS_PER_PAGE(dn.node_page, inode);
595
596 count -= dn.ofs_in_node;
597 f2fs_bug_on(sbi, count < 0);
598
599 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
600 truncate_data_blocks_range(&dn, count);
601 free_from += count;
602 }
603
604 f2fs_put_dnode(&dn);
605 free_next:
606 err = truncate_inode_blocks(inode, free_from);
607 out:
608 if (lock)
609 f2fs_unlock_op(sbi);
610 free_partial:
611 /* lastly zero out the first data page */
612 if (!err)
613 err = truncate_partial_data_page(inode, from, truncate_page);
614
615 trace_f2fs_truncate_blocks_exit(inode, err);
616 return err;
617 }
618
619 int f2fs_truncate(struct inode *inode, bool lock)
620 {
621 int err;
622
623 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
624 S_ISLNK(inode->i_mode)))
625 return 0;
626
627 trace_f2fs_truncate(inode);
628
629 /* we should check inline_data size */
630 if (!f2fs_may_inline_data(inode)) {
631 err = f2fs_convert_inline_inode(inode);
632 if (err)
633 return err;
634 }
635
636 err = truncate_blocks(inode, i_size_read(inode), lock);
637 if (err)
638 return err;
639
640 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
641 mark_inode_dirty(inode);
642 return 0;
643 }
644
645 int f2fs_getattr(struct vfsmount *mnt,
646 struct dentry *dentry, struct kstat *stat)
647 {
648 struct inode *inode = d_inode(dentry);
649 generic_fillattr(inode, stat);
650 stat->blocks <<= 3;
651 return 0;
652 }
653
654 #ifdef CONFIG_F2FS_FS_POSIX_ACL
655 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
656 {
657 struct f2fs_inode_info *fi = F2FS_I(inode);
658 unsigned int ia_valid = attr->ia_valid;
659
660 if (ia_valid & ATTR_UID)
661 inode->i_uid = attr->ia_uid;
662 if (ia_valid & ATTR_GID)
663 inode->i_gid = attr->ia_gid;
664 if (ia_valid & ATTR_ATIME)
665 inode->i_atime = timespec_trunc(attr->ia_atime,
666 inode->i_sb->s_time_gran);
667 if (ia_valid & ATTR_MTIME)
668 inode->i_mtime = timespec_trunc(attr->ia_mtime,
669 inode->i_sb->s_time_gran);
670 if (ia_valid & ATTR_CTIME)
671 inode->i_ctime = timespec_trunc(attr->ia_ctime,
672 inode->i_sb->s_time_gran);
673 if (ia_valid & ATTR_MODE) {
674 umode_t mode = attr->ia_mode;
675
676 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
677 mode &= ~S_ISGID;
678 set_acl_inode(fi, mode);
679 }
680 }
681 #else
682 #define __setattr_copy setattr_copy
683 #endif
684
685 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
686 {
687 struct inode *inode = d_inode(dentry);
688 struct f2fs_inode_info *fi = F2FS_I(inode);
689 int err;
690
691 err = inode_change_ok(inode, attr);
692 if (err)
693 return err;
694
695 if (attr->ia_valid & ATTR_SIZE) {
696 if (f2fs_encrypted_inode(inode) &&
697 fscrypt_get_encryption_info(inode))
698 return -EACCES;
699
700 if (attr->ia_size <= i_size_read(inode)) {
701 truncate_setsize(inode, attr->ia_size);
702 err = f2fs_truncate(inode, true);
703 if (err)
704 return err;
705 f2fs_balance_fs(F2FS_I_SB(inode), true);
706 } else {
707 /*
708 * do not trim all blocks after i_size if target size is
709 * larger than i_size.
710 */
711 truncate_setsize(inode, attr->ia_size);
712
713 /* should convert inline inode here */
714 if (!f2fs_may_inline_data(inode)) {
715 err = f2fs_convert_inline_inode(inode);
716 if (err)
717 return err;
718 }
719 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
720 }
721 }
722
723 __setattr_copy(inode, attr);
724
725 if (attr->ia_valid & ATTR_MODE) {
726 err = posix_acl_chmod(inode, get_inode_mode(inode));
727 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
728 inode->i_mode = fi->i_acl_mode;
729 clear_inode_flag(fi, FI_ACL_MODE);
730 }
731 }
732
733 mark_inode_dirty(inode);
734 return err;
735 }
736
737 const struct inode_operations f2fs_file_inode_operations = {
738 .getattr = f2fs_getattr,
739 .setattr = f2fs_setattr,
740 .get_acl = f2fs_get_acl,
741 .set_acl = f2fs_set_acl,
742 #ifdef CONFIG_F2FS_FS_XATTR
743 .setxattr = generic_setxattr,
744 .getxattr = generic_getxattr,
745 .listxattr = f2fs_listxattr,
746 .removexattr = generic_removexattr,
747 #endif
748 .fiemap = f2fs_fiemap,
749 };
750
751 static int fill_zero(struct inode *inode, pgoff_t index,
752 loff_t start, loff_t len)
753 {
754 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
755 struct page *page;
756
757 if (!len)
758 return 0;
759
760 f2fs_balance_fs(sbi, true);
761
762 f2fs_lock_op(sbi);
763 page = get_new_data_page(inode, NULL, index, false);
764 f2fs_unlock_op(sbi);
765
766 if (IS_ERR(page))
767 return PTR_ERR(page);
768
769 f2fs_wait_on_page_writeback(page, DATA, true);
770 zero_user(page, start, len);
771 set_page_dirty(page);
772 f2fs_put_page(page, 1);
773 return 0;
774 }
775
776 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
777 {
778 int err;
779
780 while (pg_start < pg_end) {
781 struct dnode_of_data dn;
782 pgoff_t end_offset, count;
783
784 set_new_dnode(&dn, inode, NULL, NULL, 0);
785 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
786 if (err) {
787 if (err == -ENOENT) {
788 pg_start++;
789 continue;
790 }
791 return err;
792 }
793
794 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
795 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
796
797 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
798
799 truncate_data_blocks_range(&dn, count);
800 f2fs_put_dnode(&dn);
801
802 pg_start += count;
803 }
804 return 0;
805 }
806
807 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
808 {
809 pgoff_t pg_start, pg_end;
810 loff_t off_start, off_end;
811 int ret;
812
813 ret = f2fs_convert_inline_inode(inode);
814 if (ret)
815 return ret;
816
817 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
818 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
819
820 off_start = offset & (PAGE_SIZE - 1);
821 off_end = (offset + len) & (PAGE_SIZE - 1);
822
823 if (pg_start == pg_end) {
824 ret = fill_zero(inode, pg_start, off_start,
825 off_end - off_start);
826 if (ret)
827 return ret;
828 } else {
829 if (off_start) {
830 ret = fill_zero(inode, pg_start++, off_start,
831 PAGE_SIZE - off_start);
832 if (ret)
833 return ret;
834 }
835 if (off_end) {
836 ret = fill_zero(inode, pg_end, 0, off_end);
837 if (ret)
838 return ret;
839 }
840
841 if (pg_start < pg_end) {
842 struct address_space *mapping = inode->i_mapping;
843 loff_t blk_start, blk_end;
844 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
845
846 f2fs_balance_fs(sbi, true);
847
848 blk_start = (loff_t)pg_start << PAGE_SHIFT;
849 blk_end = (loff_t)pg_end << PAGE_SHIFT;
850 truncate_inode_pages_range(mapping, blk_start,
851 blk_end - 1);
852
853 f2fs_lock_op(sbi);
854 ret = truncate_hole(inode, pg_start, pg_end);
855 f2fs_unlock_op(sbi);
856 }
857 }
858
859 return ret;
860 }
861
862 static int __exchange_data_block(struct inode *inode, pgoff_t src,
863 pgoff_t dst, bool full)
864 {
865 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
866 struct dnode_of_data dn;
867 block_t new_addr;
868 bool do_replace = false;
869 int ret;
870
871 set_new_dnode(&dn, inode, NULL, NULL, 0);
872 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
873 if (ret && ret != -ENOENT) {
874 return ret;
875 } else if (ret == -ENOENT) {
876 new_addr = NULL_ADDR;
877 } else {
878 new_addr = dn.data_blkaddr;
879 if (!is_checkpointed_data(sbi, new_addr)) {
880 /* do not invalidate this block address */
881 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
882 do_replace = true;
883 }
884 f2fs_put_dnode(&dn);
885 }
886
887 if (new_addr == NULL_ADDR)
888 return full ? truncate_hole(inode, dst, dst + 1) : 0;
889
890 if (do_replace) {
891 struct page *ipage = get_node_page(sbi, inode->i_ino);
892 struct node_info ni;
893
894 if (IS_ERR(ipage)) {
895 ret = PTR_ERR(ipage);
896 goto err_out;
897 }
898
899 set_new_dnode(&dn, inode, ipage, NULL, 0);
900 ret = f2fs_reserve_block(&dn, dst);
901 if (ret)
902 goto err_out;
903
904 truncate_data_blocks_range(&dn, 1);
905
906 get_node_info(sbi, dn.nid, &ni);
907 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
908 ni.version, true, false);
909 f2fs_put_dnode(&dn);
910 } else {
911 struct page *psrc, *pdst;
912
913 psrc = get_lock_data_page(inode, src, true);
914 if (IS_ERR(psrc))
915 return PTR_ERR(psrc);
916 pdst = get_new_data_page(inode, NULL, dst, true);
917 if (IS_ERR(pdst)) {
918 f2fs_put_page(psrc, 1);
919 return PTR_ERR(pdst);
920 }
921 f2fs_copy_page(psrc, pdst);
922 set_page_dirty(pdst);
923 f2fs_put_page(pdst, 1);
924 f2fs_put_page(psrc, 1);
925
926 return truncate_hole(inode, src, src + 1);
927 }
928 return 0;
929
930 err_out:
931 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
932 f2fs_update_data_blkaddr(&dn, new_addr);
933 f2fs_put_dnode(&dn);
934 }
935 return ret;
936 }
937
938 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
939 {
940 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
941 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
942 int ret = 0;
943
944 for (; end < nrpages; start++, end++) {
945 f2fs_balance_fs(sbi, true);
946 f2fs_lock_op(sbi);
947 ret = __exchange_data_block(inode, end, start, true);
948 f2fs_unlock_op(sbi);
949 if (ret)
950 break;
951 }
952 return ret;
953 }
954
955 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
956 {
957 pgoff_t pg_start, pg_end;
958 loff_t new_size;
959 int ret;
960
961 if (offset + len >= i_size_read(inode))
962 return -EINVAL;
963
964 /* collapse range should be aligned to block size of f2fs. */
965 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
966 return -EINVAL;
967
968 ret = f2fs_convert_inline_inode(inode);
969 if (ret)
970 return ret;
971
972 pg_start = offset >> PAGE_SHIFT;
973 pg_end = (offset + len) >> PAGE_SHIFT;
974
975 /* write out all dirty pages from offset */
976 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
977 if (ret)
978 return ret;
979
980 truncate_pagecache(inode, offset);
981
982 ret = f2fs_do_collapse(inode, pg_start, pg_end);
983 if (ret)
984 return ret;
985
986 /* write out all moved pages, if possible */
987 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
988 truncate_pagecache(inode, offset);
989
990 new_size = i_size_read(inode) - len;
991 truncate_pagecache(inode, new_size);
992
993 ret = truncate_blocks(inode, new_size, true);
994 if (!ret)
995 i_size_write(inode, new_size);
996
997 return ret;
998 }
999
1000 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1001 pgoff_t end)
1002 {
1003 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1004 pgoff_t index = start;
1005 unsigned int ofs_in_node = dn->ofs_in_node;
1006 blkcnt_t count = 0;
1007 int ret;
1008
1009 for (; index < end; index++, dn->ofs_in_node++) {
1010 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1011 count++;
1012 }
1013
1014 dn->ofs_in_node = ofs_in_node;
1015 ret = reserve_new_blocks(dn, count);
1016 if (ret)
1017 return ret;
1018
1019 dn->ofs_in_node = ofs_in_node;
1020 for (index = start; index < end; index++, dn->ofs_in_node++) {
1021 dn->data_blkaddr =
1022 datablock_addr(dn->node_page, dn->ofs_in_node);
1023 /*
1024 * reserve_new_blocks will not guarantee entire block
1025 * allocation.
1026 */
1027 if (dn->data_blkaddr == NULL_ADDR) {
1028 ret = -ENOSPC;
1029 break;
1030 }
1031 if (dn->data_blkaddr != NEW_ADDR) {
1032 invalidate_blocks(sbi, dn->data_blkaddr);
1033 dn->data_blkaddr = NEW_ADDR;
1034 set_data_blkaddr(dn);
1035 }
1036 }
1037
1038 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1039
1040 return ret;
1041 }
1042
1043 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1044 int mode)
1045 {
1046 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1047 struct address_space *mapping = inode->i_mapping;
1048 pgoff_t index, pg_start, pg_end;
1049 loff_t new_size = i_size_read(inode);
1050 loff_t off_start, off_end;
1051 int ret = 0;
1052
1053 ret = inode_newsize_ok(inode, (len + offset));
1054 if (ret)
1055 return ret;
1056
1057 ret = f2fs_convert_inline_inode(inode);
1058 if (ret)
1059 return ret;
1060
1061 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1062 if (ret)
1063 return ret;
1064
1065 truncate_pagecache_range(inode, offset, offset + len - 1);
1066
1067 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1068 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1069
1070 off_start = offset & (PAGE_SIZE - 1);
1071 off_end = (offset + len) & (PAGE_SIZE - 1);
1072
1073 if (pg_start == pg_end) {
1074 ret = fill_zero(inode, pg_start, off_start,
1075 off_end - off_start);
1076 if (ret)
1077 return ret;
1078
1079 if (offset + len > new_size)
1080 new_size = offset + len;
1081 new_size = max_t(loff_t, new_size, offset + len);
1082 } else {
1083 if (off_start) {
1084 ret = fill_zero(inode, pg_start++, off_start,
1085 PAGE_SIZE - off_start);
1086 if (ret)
1087 return ret;
1088
1089 new_size = max_t(loff_t, new_size,
1090 (loff_t)pg_start << PAGE_SHIFT);
1091 }
1092
1093 for (index = pg_start; index < pg_end;) {
1094 struct dnode_of_data dn;
1095 unsigned int end_offset;
1096 pgoff_t end;
1097
1098 f2fs_lock_op(sbi);
1099
1100 set_new_dnode(&dn, inode, NULL, NULL, 0);
1101 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1102 if (ret) {
1103 f2fs_unlock_op(sbi);
1104 goto out;
1105 }
1106
1107 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1108 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1109
1110 ret = f2fs_do_zero_range(&dn, index, end);
1111 f2fs_put_dnode(&dn);
1112 f2fs_unlock_op(sbi);
1113 if (ret)
1114 goto out;
1115
1116 index = end;
1117 new_size = max_t(loff_t, new_size,
1118 (loff_t)index << PAGE_SHIFT);
1119 }
1120
1121 if (off_end) {
1122 ret = fill_zero(inode, pg_end, 0, off_end);
1123 if (ret)
1124 goto out;
1125
1126 new_size = max_t(loff_t, new_size, offset + len);
1127 }
1128 }
1129
1130 out:
1131 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1132 i_size_write(inode, new_size);
1133 mark_inode_dirty(inode);
1134 update_inode_page(inode);
1135 }
1136
1137 return ret;
1138 }
1139
1140 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1141 {
1142 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1143 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1144 loff_t new_size;
1145 int ret = 0;
1146
1147 new_size = i_size_read(inode) + len;
1148 if (new_size > inode->i_sb->s_maxbytes)
1149 return -EFBIG;
1150
1151 if (offset >= i_size_read(inode))
1152 return -EINVAL;
1153
1154 /* insert range should be aligned to block size of f2fs. */
1155 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1156 return -EINVAL;
1157
1158 ret = f2fs_convert_inline_inode(inode);
1159 if (ret)
1160 return ret;
1161
1162 f2fs_balance_fs(sbi, true);
1163
1164 ret = truncate_blocks(inode, i_size_read(inode), true);
1165 if (ret)
1166 return ret;
1167
1168 /* write out all dirty pages from offset */
1169 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1170 if (ret)
1171 return ret;
1172
1173 truncate_pagecache(inode, offset);
1174
1175 pg_start = offset >> PAGE_SHIFT;
1176 pg_end = (offset + len) >> PAGE_SHIFT;
1177 delta = pg_end - pg_start;
1178 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1179
1180 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1181 f2fs_lock_op(sbi);
1182 ret = __exchange_data_block(inode, idx, idx + delta, false);
1183 f2fs_unlock_op(sbi);
1184 if (ret)
1185 break;
1186 }
1187
1188 /* write out all moved pages, if possible */
1189 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1190 truncate_pagecache(inode, offset);
1191
1192 if (!ret)
1193 i_size_write(inode, new_size);
1194 return ret;
1195 }
1196
1197 static int expand_inode_data(struct inode *inode, loff_t offset,
1198 loff_t len, int mode)
1199 {
1200 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1201 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1202 pgoff_t pg_end;
1203 loff_t new_size = i_size_read(inode);
1204 loff_t off_end;
1205 int ret;
1206
1207 ret = inode_newsize_ok(inode, (len + offset));
1208 if (ret)
1209 return ret;
1210
1211 ret = f2fs_convert_inline_inode(inode);
1212 if (ret)
1213 return ret;
1214
1215 f2fs_balance_fs(sbi, true);
1216
1217 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1218 off_end = (offset + len) & (PAGE_SIZE - 1);
1219
1220 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1221 map.m_len = pg_end - map.m_lblk;
1222 if (off_end)
1223 map.m_len++;
1224
1225 ret = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1226 if (ret) {
1227 pgoff_t last_off;
1228
1229 if (!map.m_len)
1230 return ret;
1231
1232 last_off = map.m_lblk + map.m_len - 1;
1233
1234 /* update new size to the failed position */
1235 new_size = (last_off == pg_end) ? offset + len:
1236 (loff_t)(last_off + 1) << PAGE_SHIFT;
1237 } else {
1238 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1239 }
1240
1241 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1242 i_size_write(inode, new_size);
1243 mark_inode_dirty(inode);
1244 update_inode_page(inode);
1245 }
1246
1247 return ret;
1248 }
1249
1250 static long f2fs_fallocate(struct file *file, int mode,
1251 loff_t offset, loff_t len)
1252 {
1253 struct inode *inode = file_inode(file);
1254 long ret = 0;
1255
1256 /* f2fs only support ->fallocate for regular file */
1257 if (!S_ISREG(inode->i_mode))
1258 return -EINVAL;
1259
1260 if (f2fs_encrypted_inode(inode) &&
1261 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1262 return -EOPNOTSUPP;
1263
1264 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1265 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1266 FALLOC_FL_INSERT_RANGE))
1267 return -EOPNOTSUPP;
1268
1269 inode_lock(inode);
1270
1271 if (mode & FALLOC_FL_PUNCH_HOLE) {
1272 if (offset >= inode->i_size)
1273 goto out;
1274
1275 ret = punch_hole(inode, offset, len);
1276 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1277 ret = f2fs_collapse_range(inode, offset, len);
1278 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1279 ret = f2fs_zero_range(inode, offset, len, mode);
1280 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1281 ret = f2fs_insert_range(inode, offset, len);
1282 } else {
1283 ret = expand_inode_data(inode, offset, len, mode);
1284 }
1285
1286 if (!ret) {
1287 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1288 mark_inode_dirty(inode);
1289 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1290 }
1291
1292 out:
1293 inode_unlock(inode);
1294
1295 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1296 return ret;
1297 }
1298
1299 static int f2fs_release_file(struct inode *inode, struct file *filp)
1300 {
1301 /*
1302 * f2fs_relase_file is called at every close calls. So we should
1303 * not drop any inmemory pages by close called by other process.
1304 */
1305 if (!(filp->f_mode & FMODE_WRITE) ||
1306 atomic_read(&inode->i_writecount) != 1)
1307 return 0;
1308
1309 /* some remained atomic pages should discarded */
1310 if (f2fs_is_atomic_file(inode))
1311 drop_inmem_pages(inode);
1312 if (f2fs_is_volatile_file(inode)) {
1313 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1314 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1315 filemap_fdatawrite(inode->i_mapping);
1316 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1317 }
1318 return 0;
1319 }
1320
1321 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1322 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1323
1324 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1325 {
1326 if (S_ISDIR(mode))
1327 return flags;
1328 else if (S_ISREG(mode))
1329 return flags & F2FS_REG_FLMASK;
1330 else
1331 return flags & F2FS_OTHER_FLMASK;
1332 }
1333
1334 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1335 {
1336 struct inode *inode = file_inode(filp);
1337 struct f2fs_inode_info *fi = F2FS_I(inode);
1338 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1339 return put_user(flags, (int __user *)arg);
1340 }
1341
1342 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1343 {
1344 struct inode *inode = file_inode(filp);
1345 struct f2fs_inode_info *fi = F2FS_I(inode);
1346 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1347 unsigned int oldflags;
1348 int ret;
1349
1350 if (!inode_owner_or_capable(inode))
1351 return -EACCES;
1352
1353 if (get_user(flags, (int __user *)arg))
1354 return -EFAULT;
1355
1356 ret = mnt_want_write_file(filp);
1357 if (ret)
1358 return ret;
1359
1360 flags = f2fs_mask_flags(inode->i_mode, flags);
1361
1362 inode_lock(inode);
1363
1364 oldflags = fi->i_flags;
1365
1366 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1367 if (!capable(CAP_LINUX_IMMUTABLE)) {
1368 inode_unlock(inode);
1369 ret = -EPERM;
1370 goto out;
1371 }
1372 }
1373
1374 flags = flags & FS_FL_USER_MODIFIABLE;
1375 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1376 fi->i_flags = flags;
1377 inode_unlock(inode);
1378
1379 f2fs_set_inode_flags(inode);
1380 inode->i_ctime = CURRENT_TIME;
1381 mark_inode_dirty(inode);
1382 out:
1383 mnt_drop_write_file(filp);
1384 return ret;
1385 }
1386
1387 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1388 {
1389 struct inode *inode = file_inode(filp);
1390
1391 return put_user(inode->i_generation, (int __user *)arg);
1392 }
1393
1394 static int f2fs_ioc_start_atomic_write(struct file *filp)
1395 {
1396 struct inode *inode = file_inode(filp);
1397 int ret;
1398
1399 if (!inode_owner_or_capable(inode))
1400 return -EACCES;
1401
1402 ret = mnt_want_write_file(filp);
1403 if (ret)
1404 return ret;
1405
1406 inode_lock(inode);
1407
1408 if (f2fs_is_atomic_file(inode))
1409 goto out;
1410
1411 ret = f2fs_convert_inline_inode(inode);
1412 if (ret)
1413 goto out;
1414
1415 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1416 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1417
1418 if (!get_dirty_pages(inode))
1419 goto out;
1420
1421 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1422 "Unexpected flush for atomic writes: ino=%lu, npages=%lld",
1423 inode->i_ino, get_dirty_pages(inode));
1424 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1425 if (ret)
1426 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1427 out:
1428 inode_unlock(inode);
1429 mnt_drop_write_file(filp);
1430 return ret;
1431 }
1432
1433 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1434 {
1435 struct inode *inode = file_inode(filp);
1436 int ret;
1437
1438 if (!inode_owner_or_capable(inode))
1439 return -EACCES;
1440
1441 ret = mnt_want_write_file(filp);
1442 if (ret)
1443 return ret;
1444
1445 inode_lock(inode);
1446
1447 if (f2fs_is_volatile_file(inode))
1448 goto err_out;
1449
1450 if (f2fs_is_atomic_file(inode)) {
1451 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1452 ret = commit_inmem_pages(inode);
1453 if (ret) {
1454 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1455 goto err_out;
1456 }
1457 }
1458
1459 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1460 err_out:
1461 inode_unlock(inode);
1462 mnt_drop_write_file(filp);
1463 return ret;
1464 }
1465
1466 static int f2fs_ioc_start_volatile_write(struct file *filp)
1467 {
1468 struct inode *inode = file_inode(filp);
1469 int ret;
1470
1471 if (!inode_owner_or_capable(inode))
1472 return -EACCES;
1473
1474 ret = mnt_want_write_file(filp);
1475 if (ret)
1476 return ret;
1477
1478 inode_lock(inode);
1479
1480 if (f2fs_is_volatile_file(inode))
1481 goto out;
1482
1483 ret = f2fs_convert_inline_inode(inode);
1484 if (ret)
1485 goto out;
1486
1487 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1488 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1489 out:
1490 inode_unlock(inode);
1491 mnt_drop_write_file(filp);
1492 return ret;
1493 }
1494
1495 static int f2fs_ioc_release_volatile_write(struct file *filp)
1496 {
1497 struct inode *inode = file_inode(filp);
1498 int ret;
1499
1500 if (!inode_owner_or_capable(inode))
1501 return -EACCES;
1502
1503 ret = mnt_want_write_file(filp);
1504 if (ret)
1505 return ret;
1506
1507 inode_lock(inode);
1508
1509 if (!f2fs_is_volatile_file(inode))
1510 goto out;
1511
1512 if (!f2fs_is_first_block_written(inode)) {
1513 ret = truncate_partial_data_page(inode, 0, true);
1514 goto out;
1515 }
1516
1517 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1518 out:
1519 inode_unlock(inode);
1520 mnt_drop_write_file(filp);
1521 return ret;
1522 }
1523
1524 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1525 {
1526 struct inode *inode = file_inode(filp);
1527 int ret;
1528
1529 if (!inode_owner_or_capable(inode))
1530 return -EACCES;
1531
1532 ret = mnt_want_write_file(filp);
1533 if (ret)
1534 return ret;
1535
1536 inode_lock(inode);
1537
1538 if (f2fs_is_atomic_file(inode))
1539 drop_inmem_pages(inode);
1540 if (f2fs_is_volatile_file(inode)) {
1541 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1542 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1543 }
1544
1545 inode_unlock(inode);
1546
1547 mnt_drop_write_file(filp);
1548 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1549 return ret;
1550 }
1551
1552 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1553 {
1554 struct inode *inode = file_inode(filp);
1555 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1556 struct super_block *sb = sbi->sb;
1557 __u32 in;
1558 int ret;
1559
1560 if (!capable(CAP_SYS_ADMIN))
1561 return -EPERM;
1562
1563 if (get_user(in, (__u32 __user *)arg))
1564 return -EFAULT;
1565
1566 ret = mnt_want_write_file(filp);
1567 if (ret)
1568 return ret;
1569
1570 switch (in) {
1571 case F2FS_GOING_DOWN_FULLSYNC:
1572 sb = freeze_bdev(sb->s_bdev);
1573 if (sb && !IS_ERR(sb)) {
1574 f2fs_stop_checkpoint(sbi, false);
1575 thaw_bdev(sb->s_bdev, sb);
1576 }
1577 break;
1578 case F2FS_GOING_DOWN_METASYNC:
1579 /* do checkpoint only */
1580 f2fs_sync_fs(sb, 1);
1581 f2fs_stop_checkpoint(sbi, false);
1582 break;
1583 case F2FS_GOING_DOWN_NOSYNC:
1584 f2fs_stop_checkpoint(sbi, false);
1585 break;
1586 case F2FS_GOING_DOWN_METAFLUSH:
1587 sync_meta_pages(sbi, META, LONG_MAX);
1588 f2fs_stop_checkpoint(sbi, false);
1589 break;
1590 default:
1591 ret = -EINVAL;
1592 goto out;
1593 }
1594 f2fs_update_time(sbi, REQ_TIME);
1595 out:
1596 mnt_drop_write_file(filp);
1597 return ret;
1598 }
1599
1600 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1601 {
1602 struct inode *inode = file_inode(filp);
1603 struct super_block *sb = inode->i_sb;
1604 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1605 struct fstrim_range range;
1606 int ret;
1607
1608 if (!capable(CAP_SYS_ADMIN))
1609 return -EPERM;
1610
1611 if (!blk_queue_discard(q))
1612 return -EOPNOTSUPP;
1613
1614 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1615 sizeof(range)))
1616 return -EFAULT;
1617
1618 ret = mnt_want_write_file(filp);
1619 if (ret)
1620 return ret;
1621
1622 range.minlen = max((unsigned int)range.minlen,
1623 q->limits.discard_granularity);
1624 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1625 mnt_drop_write_file(filp);
1626 if (ret < 0)
1627 return ret;
1628
1629 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1630 sizeof(range)))
1631 return -EFAULT;
1632 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1633 return 0;
1634 }
1635
1636 static bool uuid_is_nonzero(__u8 u[16])
1637 {
1638 int i;
1639
1640 for (i = 0; i < 16; i++)
1641 if (u[i])
1642 return true;
1643 return false;
1644 }
1645
1646 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1647 {
1648 struct fscrypt_policy policy;
1649 struct inode *inode = file_inode(filp);
1650 int ret;
1651
1652 if (copy_from_user(&policy, (struct fscrypt_policy __user *)arg,
1653 sizeof(policy)))
1654 return -EFAULT;
1655
1656 ret = mnt_want_write_file(filp);
1657 if (ret)
1658 return ret;
1659
1660 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1661 ret = fscrypt_process_policy(inode, &policy);
1662
1663 mnt_drop_write_file(filp);
1664 return ret;
1665 }
1666
1667 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1668 {
1669 struct fscrypt_policy policy;
1670 struct inode *inode = file_inode(filp);
1671 int err;
1672
1673 err = fscrypt_get_policy(inode, &policy);
1674 if (err)
1675 return err;
1676
1677 if (copy_to_user((struct fscrypt_policy __user *)arg, &policy, sizeof(policy)))
1678 return -EFAULT;
1679 return 0;
1680 }
1681
1682 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1683 {
1684 struct inode *inode = file_inode(filp);
1685 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1686 int err;
1687
1688 if (!f2fs_sb_has_crypto(inode->i_sb))
1689 return -EOPNOTSUPP;
1690
1691 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1692 goto got_it;
1693
1694 err = mnt_want_write_file(filp);
1695 if (err)
1696 return err;
1697
1698 /* update superblock with uuid */
1699 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1700
1701 err = f2fs_commit_super(sbi, false);
1702 if (err) {
1703 /* undo new data */
1704 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1705 mnt_drop_write_file(filp);
1706 return err;
1707 }
1708 mnt_drop_write_file(filp);
1709 got_it:
1710 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1711 16))
1712 return -EFAULT;
1713 return 0;
1714 }
1715
1716 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1717 {
1718 struct inode *inode = file_inode(filp);
1719 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1720 __u32 sync;
1721 int ret;
1722
1723 if (!capable(CAP_SYS_ADMIN))
1724 return -EPERM;
1725
1726 if (get_user(sync, (__u32 __user *)arg))
1727 return -EFAULT;
1728
1729 if (f2fs_readonly(sbi->sb))
1730 return -EROFS;
1731
1732 ret = mnt_want_write_file(filp);
1733 if (ret)
1734 return ret;
1735
1736 if (!sync) {
1737 if (!mutex_trylock(&sbi->gc_mutex)) {
1738 ret = -EBUSY;
1739 goto out;
1740 }
1741 } else {
1742 mutex_lock(&sbi->gc_mutex);
1743 }
1744
1745 ret = f2fs_gc(sbi, sync);
1746 out:
1747 mnt_drop_write_file(filp);
1748 return ret;
1749 }
1750
1751 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1752 {
1753 struct inode *inode = file_inode(filp);
1754 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1755 int ret;
1756
1757 if (!capable(CAP_SYS_ADMIN))
1758 return -EPERM;
1759
1760 if (f2fs_readonly(sbi->sb))
1761 return -EROFS;
1762
1763 ret = mnt_want_write_file(filp);
1764 if (ret)
1765 return ret;
1766
1767 ret = f2fs_sync_fs(sbi->sb, 1);
1768
1769 mnt_drop_write_file(filp);
1770 return ret;
1771 }
1772
1773 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1774 struct file *filp,
1775 struct f2fs_defragment *range)
1776 {
1777 struct inode *inode = file_inode(filp);
1778 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1779 struct extent_info ei;
1780 pgoff_t pg_start, pg_end;
1781 unsigned int blk_per_seg = sbi->blocks_per_seg;
1782 unsigned int total = 0, sec_num;
1783 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1784 block_t blk_end = 0;
1785 bool fragmented = false;
1786 int err;
1787
1788 /* if in-place-update policy is enabled, don't waste time here */
1789 if (need_inplace_update(inode))
1790 return -EINVAL;
1791
1792 pg_start = range->start >> PAGE_SHIFT;
1793 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1794
1795 f2fs_balance_fs(sbi, true);
1796
1797 inode_lock(inode);
1798
1799 /* writeback all dirty pages in the range */
1800 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1801 range->start + range->len - 1);
1802 if (err)
1803 goto out;
1804
1805 /*
1806 * lookup mapping info in extent cache, skip defragmenting if physical
1807 * block addresses are continuous.
1808 */
1809 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1810 if (ei.fofs + ei.len >= pg_end)
1811 goto out;
1812 }
1813
1814 map.m_lblk = pg_start;
1815
1816 /*
1817 * lookup mapping info in dnode page cache, skip defragmenting if all
1818 * physical block addresses are continuous even if there are hole(s)
1819 * in logical blocks.
1820 */
1821 while (map.m_lblk < pg_end) {
1822 map.m_len = pg_end - map.m_lblk;
1823 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1824 if (err)
1825 goto out;
1826
1827 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1828 map.m_lblk++;
1829 continue;
1830 }
1831
1832 if (blk_end && blk_end != map.m_pblk) {
1833 fragmented = true;
1834 break;
1835 }
1836 blk_end = map.m_pblk + map.m_len;
1837
1838 map.m_lblk += map.m_len;
1839 }
1840
1841 if (!fragmented)
1842 goto out;
1843
1844 map.m_lblk = pg_start;
1845 map.m_len = pg_end - pg_start;
1846
1847 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1848
1849 /*
1850 * make sure there are enough free section for LFS allocation, this can
1851 * avoid defragment running in SSR mode when free section are allocated
1852 * intensively
1853 */
1854 if (has_not_enough_free_secs(sbi, sec_num)) {
1855 err = -EAGAIN;
1856 goto out;
1857 }
1858
1859 while (map.m_lblk < pg_end) {
1860 pgoff_t idx;
1861 int cnt = 0;
1862
1863 do_map:
1864 map.m_len = pg_end - map.m_lblk;
1865 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1866 if (err)
1867 goto clear_out;
1868
1869 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1870 map.m_lblk++;
1871 continue;
1872 }
1873
1874 set_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1875
1876 idx = map.m_lblk;
1877 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1878 struct page *page;
1879
1880 page = get_lock_data_page(inode, idx, true);
1881 if (IS_ERR(page)) {
1882 err = PTR_ERR(page);
1883 goto clear_out;
1884 }
1885
1886 set_page_dirty(page);
1887 f2fs_put_page(page, 1);
1888
1889 idx++;
1890 cnt++;
1891 total++;
1892 }
1893
1894 map.m_lblk = idx;
1895
1896 if (idx < pg_end && cnt < blk_per_seg)
1897 goto do_map;
1898
1899 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1900
1901 err = filemap_fdatawrite(inode->i_mapping);
1902 if (err)
1903 goto out;
1904 }
1905 clear_out:
1906 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1907 out:
1908 inode_unlock(inode);
1909 if (!err)
1910 range->len = (u64)total << PAGE_SHIFT;
1911 return err;
1912 }
1913
1914 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1915 {
1916 struct inode *inode = file_inode(filp);
1917 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1918 struct f2fs_defragment range;
1919 int err;
1920
1921 if (!capable(CAP_SYS_ADMIN))
1922 return -EPERM;
1923
1924 if (!S_ISREG(inode->i_mode))
1925 return -EINVAL;
1926
1927 err = mnt_want_write_file(filp);
1928 if (err)
1929 return err;
1930
1931 if (f2fs_readonly(sbi->sb)) {
1932 err = -EROFS;
1933 goto out;
1934 }
1935
1936 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1937 sizeof(range))) {
1938 err = -EFAULT;
1939 goto out;
1940 }
1941
1942 /* verify alignment of offset & size */
1943 if (range.start & (F2FS_BLKSIZE - 1) ||
1944 range.len & (F2FS_BLKSIZE - 1)) {
1945 err = -EINVAL;
1946 goto out;
1947 }
1948
1949 err = f2fs_defragment_range(sbi, filp, &range);
1950 f2fs_update_time(sbi, REQ_TIME);
1951 if (err < 0)
1952 goto out;
1953
1954 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1955 sizeof(range)))
1956 err = -EFAULT;
1957 out:
1958 mnt_drop_write_file(filp);
1959 return err;
1960 }
1961
1962 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1963 {
1964 switch (cmd) {
1965 case F2FS_IOC_GETFLAGS:
1966 return f2fs_ioc_getflags(filp, arg);
1967 case F2FS_IOC_SETFLAGS:
1968 return f2fs_ioc_setflags(filp, arg);
1969 case F2FS_IOC_GETVERSION:
1970 return f2fs_ioc_getversion(filp, arg);
1971 case F2FS_IOC_START_ATOMIC_WRITE:
1972 return f2fs_ioc_start_atomic_write(filp);
1973 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1974 return f2fs_ioc_commit_atomic_write(filp);
1975 case F2FS_IOC_START_VOLATILE_WRITE:
1976 return f2fs_ioc_start_volatile_write(filp);
1977 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1978 return f2fs_ioc_release_volatile_write(filp);
1979 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1980 return f2fs_ioc_abort_volatile_write(filp);
1981 case F2FS_IOC_SHUTDOWN:
1982 return f2fs_ioc_shutdown(filp, arg);
1983 case FITRIM:
1984 return f2fs_ioc_fitrim(filp, arg);
1985 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1986 return f2fs_ioc_set_encryption_policy(filp, arg);
1987 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1988 return f2fs_ioc_get_encryption_policy(filp, arg);
1989 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1990 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1991 case F2FS_IOC_GARBAGE_COLLECT:
1992 return f2fs_ioc_gc(filp, arg);
1993 case F2FS_IOC_WRITE_CHECKPOINT:
1994 return f2fs_ioc_write_checkpoint(filp, arg);
1995 case F2FS_IOC_DEFRAGMENT:
1996 return f2fs_ioc_defragment(filp, arg);
1997 default:
1998 return -ENOTTY;
1999 }
2000 }
2001
2002 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2003 {
2004 struct file *file = iocb->ki_filp;
2005 struct inode *inode = file_inode(file);
2006 ssize_t ret;
2007
2008 if (f2fs_encrypted_inode(inode) &&
2009 !fscrypt_has_encryption_key(inode) &&
2010 fscrypt_get_encryption_info(inode))
2011 return -EACCES;
2012
2013 inode_lock(inode);
2014 ret = generic_write_checks(iocb, from);
2015 if (ret > 0) {
2016 ret = f2fs_preallocate_blocks(iocb, from);
2017 if (!ret)
2018 ret = __generic_file_write_iter(iocb, from);
2019 }
2020 inode_unlock(inode);
2021
2022 if (ret > 0)
2023 ret = generic_write_sync(iocb, ret);
2024 return ret;
2025 }
2026
2027 #ifdef CONFIG_COMPAT
2028 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2029 {
2030 switch (cmd) {
2031 case F2FS_IOC32_GETFLAGS:
2032 cmd = F2FS_IOC_GETFLAGS;
2033 break;
2034 case F2FS_IOC32_SETFLAGS:
2035 cmd = F2FS_IOC_SETFLAGS;
2036 break;
2037 case F2FS_IOC32_GETVERSION:
2038 cmd = F2FS_IOC_GETVERSION;
2039 break;
2040 case F2FS_IOC_START_ATOMIC_WRITE:
2041 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2042 case F2FS_IOC_START_VOLATILE_WRITE:
2043 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2044 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2045 case F2FS_IOC_SHUTDOWN:
2046 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2047 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2048 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2049 case F2FS_IOC_GARBAGE_COLLECT:
2050 case F2FS_IOC_WRITE_CHECKPOINT:
2051 case F2FS_IOC_DEFRAGMENT:
2052 break;
2053 default:
2054 return -ENOIOCTLCMD;
2055 }
2056 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2057 }
2058 #endif
2059
2060 const struct file_operations f2fs_file_operations = {
2061 .llseek = f2fs_llseek,
2062 .read_iter = generic_file_read_iter,
2063 .write_iter = f2fs_file_write_iter,
2064 .open = f2fs_file_open,
2065 .release = f2fs_release_file,
2066 .mmap = f2fs_file_mmap,
2067 .fsync = f2fs_sync_file,
2068 .fallocate = f2fs_fallocate,
2069 .unlocked_ioctl = f2fs_ioctl,
2070 #ifdef CONFIG_COMPAT
2071 .compat_ioctl = f2fs_compat_ioctl,
2072 #endif
2073 .splice_read = generic_file_splice_read,
2074 .splice_write = iter_file_splice_write,
2075 };
Linux with added FPC-III support