]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - fs/f2fs/file.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: flush inode metadata when checkpoint is doing
[mirror_ubuntu-artful-kernel.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 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 f2fs_i_pino_write(inode, pino);
175 file_got_pino(inode);
176 up_write(&fi->i_sem);
177
178 f2fs_write_inode(inode, NULL);
179 } else {
180 up_write(&fi->i_sem);
181 }
182 }
183
184 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
185 int datasync, bool atomic)
186 {
187 struct inode *inode = file->f_mapping->host;
188 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
189 nid_t ino = inode->i_ino;
190 int ret = 0;
191 bool need_cp = false;
192 struct writeback_control wbc = {
193 .sync_mode = WB_SYNC_ALL,
194 .nr_to_write = LONG_MAX,
195 .for_reclaim = 0,
196 };
197
198 if (unlikely(f2fs_readonly(inode->i_sb)))
199 return 0;
200
201 trace_f2fs_sync_file_enter(inode);
202
203 /* if fdatasync is triggered, let's do in-place-update */
204 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
205 set_inode_flag(inode, FI_NEED_IPU);
206 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
207 clear_inode_flag(inode, FI_NEED_IPU);
208
209 if (ret) {
210 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
211 return ret;
212 }
213
214 /* if the inode is dirty, let's recover all the time */
215 if (!datasync) {
216 f2fs_write_inode(inode, NULL);
217 goto go_write;
218 }
219
220 /*
221 * if there is no written data, don't waste time to write recovery info.
222 */
223 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
224 !exist_written_data(sbi, ino, APPEND_INO)) {
225
226 /* it may call write_inode just prior to fsync */
227 if (need_inode_page_update(sbi, ino))
228 goto go_write;
229
230 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
231 exist_written_data(sbi, ino, UPDATE_INO))
232 goto flush_out;
233 goto out;
234 }
235 go_write:
236 /*
237 * Both of fdatasync() and fsync() are able to be recovered from
238 * sudden-power-off.
239 */
240 down_read(&F2FS_I(inode)->i_sem);
241 need_cp = need_do_checkpoint(inode);
242 up_read(&F2FS_I(inode)->i_sem);
243
244 if (need_cp) {
245 /* all the dirty node pages should be flushed for POR */
246 ret = f2fs_sync_fs(inode->i_sb, 1);
247
248 /*
249 * We've secured consistency through sync_fs. Following pino
250 * will be used only for fsynced inodes after checkpoint.
251 */
252 try_to_fix_pino(inode);
253 clear_inode_flag(inode, FI_APPEND_WRITE);
254 clear_inode_flag(inode, FI_UPDATE_WRITE);
255 goto out;
256 }
257 sync_nodes:
258 ret = fsync_node_pages(sbi, ino, &wbc, atomic);
259 if (ret)
260 goto out;
261
262 /* if cp_error was enabled, we should avoid infinite loop */
263 if (unlikely(f2fs_cp_error(sbi))) {
264 ret = -EIO;
265 goto out;
266 }
267
268 if (need_inode_block_update(sbi, ino)) {
269 mark_inode_dirty_sync(inode);
270 f2fs_write_inode(inode, NULL);
271 goto sync_nodes;
272 }
273
274 ret = wait_on_node_pages_writeback(sbi, ino);
275 if (ret)
276 goto out;
277
278 /* once recovery info is written, don't need to tack this */
279 remove_ino_entry(sbi, ino, APPEND_INO);
280 clear_inode_flag(inode, FI_APPEND_WRITE);
281 flush_out:
282 remove_ino_entry(sbi, ino, UPDATE_INO);
283 clear_inode_flag(inode, FI_UPDATE_WRITE);
284 ret = f2fs_issue_flush(sbi);
285 f2fs_update_time(sbi, REQ_TIME);
286 out:
287 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
288 f2fs_trace_ios(NULL, 1);
289 return ret;
290 }
291
292 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
293 {
294 return f2fs_do_sync_file(file, start, end, datasync, false);
295 }
296
297 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
298 pgoff_t pgofs, int whence)
299 {
300 struct pagevec pvec;
301 int nr_pages;
302
303 if (whence != SEEK_DATA)
304 return 0;
305
306 /* find first dirty page index */
307 pagevec_init(&pvec, 0);
308 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
309 PAGECACHE_TAG_DIRTY, 1);
310 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
311 pagevec_release(&pvec);
312 return pgofs;
313 }
314
315 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
316 int whence)
317 {
318 switch (whence) {
319 case SEEK_DATA:
320 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
321 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
322 return true;
323 break;
324 case SEEK_HOLE:
325 if (blkaddr == NULL_ADDR)
326 return true;
327 break;
328 }
329 return false;
330 }
331
332 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
333 {
334 struct inode *inode = file->f_mapping->host;
335 loff_t maxbytes = inode->i_sb->s_maxbytes;
336 struct dnode_of_data dn;
337 pgoff_t pgofs, end_offset, dirty;
338 loff_t data_ofs = offset;
339 loff_t isize;
340 int err = 0;
341
342 inode_lock(inode);
343
344 isize = i_size_read(inode);
345 if (offset >= isize)
346 goto fail;
347
348 /* handle inline data case */
349 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
350 if (whence == SEEK_HOLE)
351 data_ofs = isize;
352 goto found;
353 }
354
355 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
356
357 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
358
359 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
360 set_new_dnode(&dn, inode, NULL, NULL, 0);
361 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
362 if (err && err != -ENOENT) {
363 goto fail;
364 } else if (err == -ENOENT) {
365 /* direct node does not exists */
366 if (whence == SEEK_DATA) {
367 pgofs = get_next_page_offset(&dn, pgofs);
368 continue;
369 } else {
370 goto found;
371 }
372 }
373
374 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
375
376 /* find data/hole in dnode block */
377 for (; dn.ofs_in_node < end_offset;
378 dn.ofs_in_node++, pgofs++,
379 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
380 block_t blkaddr;
381 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
382
383 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
384 f2fs_put_dnode(&dn);
385 goto found;
386 }
387 }
388 f2fs_put_dnode(&dn);
389 }
390
391 if (whence == SEEK_DATA)
392 goto fail;
393 found:
394 if (whence == SEEK_HOLE && data_ofs > isize)
395 data_ofs = isize;
396 inode_unlock(inode);
397 return vfs_setpos(file, data_ofs, maxbytes);
398 fail:
399 inode_unlock(inode);
400 return -ENXIO;
401 }
402
403 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
404 {
405 struct inode *inode = file->f_mapping->host;
406 loff_t maxbytes = inode->i_sb->s_maxbytes;
407
408 switch (whence) {
409 case SEEK_SET:
410 case SEEK_CUR:
411 case SEEK_END:
412 return generic_file_llseek_size(file, offset, whence,
413 maxbytes, i_size_read(inode));
414 case SEEK_DATA:
415 case SEEK_HOLE:
416 if (offset < 0)
417 return -ENXIO;
418 return f2fs_seek_block(file, offset, whence);
419 }
420
421 return -EINVAL;
422 }
423
424 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
425 {
426 struct inode *inode = file_inode(file);
427 int err;
428
429 if (f2fs_encrypted_inode(inode)) {
430 err = fscrypt_get_encryption_info(inode);
431 if (err)
432 return 0;
433 if (!f2fs_encrypted_inode(inode))
434 return -ENOKEY;
435 }
436
437 /* we don't need to use inline_data strictly */
438 err = f2fs_convert_inline_inode(inode);
439 if (err)
440 return err;
441
442 file_accessed(file);
443 vma->vm_ops = &f2fs_file_vm_ops;
444 return 0;
445 }
446
447 static int f2fs_file_open(struct inode *inode, struct file *filp)
448 {
449 int ret = generic_file_open(inode, filp);
450 struct dentry *dir;
451
452 if (!ret && f2fs_encrypted_inode(inode)) {
453 ret = fscrypt_get_encryption_info(inode);
454 if (ret)
455 return -EACCES;
456 if (!fscrypt_has_encryption_key(inode))
457 return -ENOKEY;
458 }
459 dir = dget_parent(file_dentry(filp));
460 if (f2fs_encrypted_inode(d_inode(dir)) &&
461 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
462 dput(dir);
463 return -EPERM;
464 }
465 dput(dir);
466 return ret;
467 }
468
469 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
470 {
471 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
472 struct f2fs_node *raw_node;
473 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
474 __le32 *addr;
475
476 raw_node = F2FS_NODE(dn->node_page);
477 addr = blkaddr_in_node(raw_node) + ofs;
478
479 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
480 block_t blkaddr = le32_to_cpu(*addr);
481 if (blkaddr == NULL_ADDR)
482 continue;
483
484 dn->data_blkaddr = NULL_ADDR;
485 set_data_blkaddr(dn);
486 invalidate_blocks(sbi, blkaddr);
487 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
488 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
489 nr_free++;
490 }
491
492 if (nr_free) {
493 pgoff_t fofs;
494 /*
495 * once we invalidate valid blkaddr in range [ofs, ofs + count],
496 * we will invalidate all blkaddr in the whole range.
497 */
498 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
499 dn->inode) + ofs;
500 f2fs_update_extent_cache_range(dn, fofs, 0, len);
501 dec_valid_block_count(sbi, dn->inode, nr_free);
502 sync_inode_page(dn);
503 }
504 dn->ofs_in_node = ofs;
505
506 f2fs_update_time(sbi, REQ_TIME);
507 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
508 dn->ofs_in_node, nr_free);
509 return nr_free;
510 }
511
512 void truncate_data_blocks(struct dnode_of_data *dn)
513 {
514 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
515 }
516
517 static int truncate_partial_data_page(struct inode *inode, u64 from,
518 bool cache_only)
519 {
520 unsigned offset = from & (PAGE_SIZE - 1);
521 pgoff_t index = from >> PAGE_SHIFT;
522 struct address_space *mapping = inode->i_mapping;
523 struct page *page;
524
525 if (!offset && !cache_only)
526 return 0;
527
528 if (cache_only) {
529 page = f2fs_grab_cache_page(mapping, index, false);
530 if (page && PageUptodate(page))
531 goto truncate_out;
532 f2fs_put_page(page, 1);
533 return 0;
534 }
535
536 page = get_lock_data_page(inode, index, true);
537 if (IS_ERR(page))
538 return 0;
539 truncate_out:
540 f2fs_wait_on_page_writeback(page, DATA, true);
541 zero_user(page, offset, PAGE_SIZE - offset);
542 if (!cache_only || !f2fs_encrypted_inode(inode) ||
543 !S_ISREG(inode->i_mode))
544 set_page_dirty(page);
545 f2fs_put_page(page, 1);
546 return 0;
547 }
548
549 int truncate_blocks(struct inode *inode, u64 from, bool lock)
550 {
551 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
552 unsigned int blocksize = inode->i_sb->s_blocksize;
553 struct dnode_of_data dn;
554 pgoff_t free_from;
555 int count = 0, err = 0;
556 struct page *ipage;
557 bool truncate_page = false;
558
559 trace_f2fs_truncate_blocks_enter(inode, from);
560
561 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
562
563 if (free_from >= sbi->max_file_blocks)
564 goto free_partial;
565
566 if (lock)
567 f2fs_lock_op(sbi);
568
569 ipage = get_node_page(sbi, inode->i_ino);
570 if (IS_ERR(ipage)) {
571 err = PTR_ERR(ipage);
572 goto out;
573 }
574
575 if (f2fs_has_inline_data(inode)) {
576 if (truncate_inline_inode(ipage, from))
577 set_page_dirty(ipage);
578 f2fs_put_page(ipage, 1);
579 truncate_page = true;
580 goto out;
581 }
582
583 set_new_dnode(&dn, inode, ipage, NULL, 0);
584 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
585 if (err) {
586 if (err == -ENOENT)
587 goto free_next;
588 goto out;
589 }
590
591 count = ADDRS_PER_PAGE(dn.node_page, inode);
592
593 count -= dn.ofs_in_node;
594 f2fs_bug_on(sbi, count < 0);
595
596 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
597 truncate_data_blocks_range(&dn, count);
598 free_from += count;
599 }
600
601 f2fs_put_dnode(&dn);
602 free_next:
603 err = truncate_inode_blocks(inode, free_from);
604 out:
605 if (lock)
606 f2fs_unlock_op(sbi);
607 free_partial:
608 /* lastly zero out the first data page */
609 if (!err)
610 err = truncate_partial_data_page(inode, from, truncate_page);
611
612 trace_f2fs_truncate_blocks_exit(inode, err);
613 return err;
614 }
615
616 int f2fs_truncate(struct inode *inode, bool lock)
617 {
618 int err;
619
620 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
621 S_ISLNK(inode->i_mode)))
622 return 0;
623
624 trace_f2fs_truncate(inode);
625
626 /* we should check inline_data size */
627 if (!f2fs_may_inline_data(inode)) {
628 err = f2fs_convert_inline_inode(inode);
629 if (err)
630 return err;
631 }
632
633 err = truncate_blocks(inode, i_size_read(inode), lock);
634 if (err)
635 return err;
636
637 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
638 mark_inode_dirty_sync(inode);
639 return 0;
640 }
641
642 int f2fs_getattr(struct vfsmount *mnt,
643 struct dentry *dentry, struct kstat *stat)
644 {
645 struct inode *inode = d_inode(dentry);
646 generic_fillattr(inode, stat);
647 stat->blocks <<= 3;
648 return 0;
649 }
650
651 #ifdef CONFIG_F2FS_FS_POSIX_ACL
652 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
653 {
654 unsigned int ia_valid = attr->ia_valid;
655
656 if (ia_valid & ATTR_UID)
657 inode->i_uid = attr->ia_uid;
658 if (ia_valid & ATTR_GID)
659 inode->i_gid = attr->ia_gid;
660 if (ia_valid & ATTR_ATIME)
661 inode->i_atime = timespec_trunc(attr->ia_atime,
662 inode->i_sb->s_time_gran);
663 if (ia_valid & ATTR_MTIME)
664 inode->i_mtime = timespec_trunc(attr->ia_mtime,
665 inode->i_sb->s_time_gran);
666 if (ia_valid & ATTR_CTIME)
667 inode->i_ctime = timespec_trunc(attr->ia_ctime,
668 inode->i_sb->s_time_gran);
669 if (ia_valid & ATTR_MODE) {
670 umode_t mode = attr->ia_mode;
671
672 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
673 mode &= ~S_ISGID;
674 set_acl_inode(inode, mode);
675 }
676 }
677 #else
678 #define __setattr_copy setattr_copy
679 #endif
680
681 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
682 {
683 struct inode *inode = d_inode(dentry);
684 int err;
685
686 err = inode_change_ok(inode, attr);
687 if (err)
688 return err;
689
690 if (attr->ia_valid & ATTR_SIZE) {
691 if (f2fs_encrypted_inode(inode) &&
692 fscrypt_get_encryption_info(inode))
693 return -EACCES;
694
695 if (attr->ia_size <= i_size_read(inode)) {
696 truncate_setsize(inode, attr->ia_size);
697 err = f2fs_truncate(inode, true);
698 if (err)
699 return err;
700 f2fs_balance_fs(F2FS_I_SB(inode), true);
701 } else {
702 /*
703 * do not trim all blocks after i_size if target size is
704 * larger than i_size.
705 */
706 truncate_setsize(inode, attr->ia_size);
707
708 /* should convert inline inode here */
709 if (!f2fs_may_inline_data(inode)) {
710 err = f2fs_convert_inline_inode(inode);
711 if (err)
712 return err;
713 }
714 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
715 }
716 }
717
718 __setattr_copy(inode, attr);
719
720 if (attr->ia_valid & ATTR_MODE) {
721 err = posix_acl_chmod(inode, get_inode_mode(inode));
722 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
723 inode->i_mode = F2FS_I(inode)->i_acl_mode;
724 clear_inode_flag(inode, FI_ACL_MODE);
725 }
726 }
727
728 mark_inode_dirty_sync(inode);
729 return err;
730 }
731
732 const struct inode_operations f2fs_file_inode_operations = {
733 .getattr = f2fs_getattr,
734 .setattr = f2fs_setattr,
735 .get_acl = f2fs_get_acl,
736 .set_acl = f2fs_set_acl,
737 #ifdef CONFIG_F2FS_FS_XATTR
738 .setxattr = generic_setxattr,
739 .getxattr = generic_getxattr,
740 .listxattr = f2fs_listxattr,
741 .removexattr = generic_removexattr,
742 #endif
743 .fiemap = f2fs_fiemap,
744 };
745
746 static int fill_zero(struct inode *inode, pgoff_t index,
747 loff_t start, loff_t len)
748 {
749 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
750 struct page *page;
751
752 if (!len)
753 return 0;
754
755 f2fs_balance_fs(sbi, true);
756
757 f2fs_lock_op(sbi);
758 page = get_new_data_page(inode, NULL, index, false);
759 f2fs_unlock_op(sbi);
760
761 if (IS_ERR(page))
762 return PTR_ERR(page);
763
764 f2fs_wait_on_page_writeback(page, DATA, true);
765 zero_user(page, start, len);
766 set_page_dirty(page);
767 f2fs_put_page(page, 1);
768 return 0;
769 }
770
771 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
772 {
773 int err;
774
775 while (pg_start < pg_end) {
776 struct dnode_of_data dn;
777 pgoff_t end_offset, count;
778
779 set_new_dnode(&dn, inode, NULL, NULL, 0);
780 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
781 if (err) {
782 if (err == -ENOENT) {
783 pg_start++;
784 continue;
785 }
786 return err;
787 }
788
789 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
790 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
791
792 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
793
794 truncate_data_blocks_range(&dn, count);
795 f2fs_put_dnode(&dn);
796
797 pg_start += count;
798 }
799 return 0;
800 }
801
802 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
803 {
804 pgoff_t pg_start, pg_end;
805 loff_t off_start, off_end;
806 int ret;
807
808 ret = f2fs_convert_inline_inode(inode);
809 if (ret)
810 return ret;
811
812 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
813 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
814
815 off_start = offset & (PAGE_SIZE - 1);
816 off_end = (offset + len) & (PAGE_SIZE - 1);
817
818 if (pg_start == pg_end) {
819 ret = fill_zero(inode, pg_start, off_start,
820 off_end - off_start);
821 if (ret)
822 return ret;
823 } else {
824 if (off_start) {
825 ret = fill_zero(inode, pg_start++, off_start,
826 PAGE_SIZE - off_start);
827 if (ret)
828 return ret;
829 }
830 if (off_end) {
831 ret = fill_zero(inode, pg_end, 0, off_end);
832 if (ret)
833 return ret;
834 }
835
836 if (pg_start < pg_end) {
837 struct address_space *mapping = inode->i_mapping;
838 loff_t blk_start, blk_end;
839 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
840
841 f2fs_balance_fs(sbi, true);
842
843 blk_start = (loff_t)pg_start << PAGE_SHIFT;
844 blk_end = (loff_t)pg_end << PAGE_SHIFT;
845 truncate_inode_pages_range(mapping, blk_start,
846 blk_end - 1);
847
848 f2fs_lock_op(sbi);
849 ret = truncate_hole(inode, pg_start, pg_end);
850 f2fs_unlock_op(sbi);
851 }
852 }
853
854 return ret;
855 }
856
857 static int __exchange_data_block(struct inode *inode, pgoff_t src,
858 pgoff_t dst, bool full)
859 {
860 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
861 struct dnode_of_data dn;
862 block_t new_addr;
863 bool do_replace = false;
864 int ret;
865
866 set_new_dnode(&dn, inode, NULL, NULL, 0);
867 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
868 if (ret && ret != -ENOENT) {
869 return ret;
870 } else if (ret == -ENOENT) {
871 new_addr = NULL_ADDR;
872 } else {
873 new_addr = dn.data_blkaddr;
874 if (!is_checkpointed_data(sbi, new_addr)) {
875 /* do not invalidate this block address */
876 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
877 do_replace = true;
878 }
879 f2fs_put_dnode(&dn);
880 }
881
882 if (new_addr == NULL_ADDR)
883 return full ? truncate_hole(inode, dst, dst + 1) : 0;
884
885 if (do_replace) {
886 struct page *ipage = get_node_page(sbi, inode->i_ino);
887 struct node_info ni;
888
889 if (IS_ERR(ipage)) {
890 ret = PTR_ERR(ipage);
891 goto err_out;
892 }
893
894 set_new_dnode(&dn, inode, ipage, NULL, 0);
895 ret = f2fs_reserve_block(&dn, dst);
896 if (ret)
897 goto err_out;
898
899 truncate_data_blocks_range(&dn, 1);
900
901 get_node_info(sbi, dn.nid, &ni);
902 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
903 ni.version, true, false);
904 f2fs_put_dnode(&dn);
905 } else {
906 struct page *psrc, *pdst;
907
908 psrc = get_lock_data_page(inode, src, true);
909 if (IS_ERR(psrc))
910 return PTR_ERR(psrc);
911 pdst = get_new_data_page(inode, NULL, dst, true);
912 if (IS_ERR(pdst)) {
913 f2fs_put_page(psrc, 1);
914 return PTR_ERR(pdst);
915 }
916 f2fs_copy_page(psrc, pdst);
917 set_page_dirty(pdst);
918 f2fs_put_page(pdst, 1);
919 f2fs_put_page(psrc, 1);
920
921 return truncate_hole(inode, src, src + 1);
922 }
923 return 0;
924
925 err_out:
926 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
927 f2fs_update_data_blkaddr(&dn, new_addr);
928 f2fs_put_dnode(&dn);
929 }
930 return ret;
931 }
932
933 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
934 {
935 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
936 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
937 int ret = 0;
938
939 for (; end < nrpages; start++, end++) {
940 f2fs_balance_fs(sbi, true);
941 f2fs_lock_op(sbi);
942 ret = __exchange_data_block(inode, end, start, true);
943 f2fs_unlock_op(sbi);
944 if (ret)
945 break;
946 }
947 return ret;
948 }
949
950 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
951 {
952 pgoff_t pg_start, pg_end;
953 loff_t new_size;
954 int ret;
955
956 if (offset + len >= i_size_read(inode))
957 return -EINVAL;
958
959 /* collapse range should be aligned to block size of f2fs. */
960 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
961 return -EINVAL;
962
963 ret = f2fs_convert_inline_inode(inode);
964 if (ret)
965 return ret;
966
967 pg_start = offset >> PAGE_SHIFT;
968 pg_end = (offset + len) >> PAGE_SHIFT;
969
970 /* write out all dirty pages from offset */
971 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
972 if (ret)
973 return ret;
974
975 truncate_pagecache(inode, offset);
976
977 ret = f2fs_do_collapse(inode, pg_start, pg_end);
978 if (ret)
979 return ret;
980
981 /* write out all moved pages, if possible */
982 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
983 truncate_pagecache(inode, offset);
984
985 new_size = i_size_read(inode) - len;
986 truncate_pagecache(inode, new_size);
987
988 ret = truncate_blocks(inode, new_size, true);
989 if (!ret)
990 f2fs_i_size_write(inode, new_size);
991
992 return ret;
993 }
994
995 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
996 pgoff_t end)
997 {
998 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
999 pgoff_t index = start;
1000 unsigned int ofs_in_node = dn->ofs_in_node;
1001 blkcnt_t count = 0;
1002 int ret;
1003
1004 for (; index < end; index++, dn->ofs_in_node++) {
1005 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1006 count++;
1007 }
1008
1009 dn->ofs_in_node = ofs_in_node;
1010 ret = reserve_new_blocks(dn, count);
1011 if (ret)
1012 return ret;
1013
1014 dn->ofs_in_node = ofs_in_node;
1015 for (index = start; index < end; index++, dn->ofs_in_node++) {
1016 dn->data_blkaddr =
1017 datablock_addr(dn->node_page, dn->ofs_in_node);
1018 /*
1019 * reserve_new_blocks will not guarantee entire block
1020 * allocation.
1021 */
1022 if (dn->data_blkaddr == NULL_ADDR) {
1023 ret = -ENOSPC;
1024 break;
1025 }
1026 if (dn->data_blkaddr != NEW_ADDR) {
1027 invalidate_blocks(sbi, dn->data_blkaddr);
1028 dn->data_blkaddr = NEW_ADDR;
1029 set_data_blkaddr(dn);
1030 }
1031 }
1032
1033 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1034
1035 return ret;
1036 }
1037
1038 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1039 int mode)
1040 {
1041 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1042 struct address_space *mapping = inode->i_mapping;
1043 pgoff_t index, pg_start, pg_end;
1044 loff_t new_size = i_size_read(inode);
1045 loff_t off_start, off_end;
1046 int ret = 0;
1047
1048 ret = inode_newsize_ok(inode, (len + offset));
1049 if (ret)
1050 return ret;
1051
1052 ret = f2fs_convert_inline_inode(inode);
1053 if (ret)
1054 return ret;
1055
1056 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1057 if (ret)
1058 return ret;
1059
1060 truncate_pagecache_range(inode, offset, offset + len - 1);
1061
1062 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1063 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1064
1065 off_start = offset & (PAGE_SIZE - 1);
1066 off_end = (offset + len) & (PAGE_SIZE - 1);
1067
1068 if (pg_start == pg_end) {
1069 ret = fill_zero(inode, pg_start, off_start,
1070 off_end - off_start);
1071 if (ret)
1072 return ret;
1073
1074 if (offset + len > new_size)
1075 new_size = offset + len;
1076 new_size = max_t(loff_t, new_size, offset + len);
1077 } else {
1078 if (off_start) {
1079 ret = fill_zero(inode, pg_start++, off_start,
1080 PAGE_SIZE - off_start);
1081 if (ret)
1082 return ret;
1083
1084 new_size = max_t(loff_t, new_size,
1085 (loff_t)pg_start << PAGE_SHIFT);
1086 }
1087
1088 for (index = pg_start; index < pg_end;) {
1089 struct dnode_of_data dn;
1090 unsigned int end_offset;
1091 pgoff_t end;
1092
1093 f2fs_lock_op(sbi);
1094
1095 set_new_dnode(&dn, inode, NULL, NULL, 0);
1096 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1097 if (ret) {
1098 f2fs_unlock_op(sbi);
1099 goto out;
1100 }
1101
1102 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1103 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1104
1105 ret = f2fs_do_zero_range(&dn, index, end);
1106 f2fs_put_dnode(&dn);
1107 f2fs_unlock_op(sbi);
1108 if (ret)
1109 goto out;
1110
1111 index = end;
1112 new_size = max_t(loff_t, new_size,
1113 (loff_t)index << PAGE_SHIFT);
1114 }
1115
1116 if (off_end) {
1117 ret = fill_zero(inode, pg_end, 0, off_end);
1118 if (ret)
1119 goto out;
1120
1121 new_size = max_t(loff_t, new_size, offset + len);
1122 }
1123 }
1124
1125 out:
1126 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1127 f2fs_i_size_write(inode, new_size);
1128 update_inode_page(inode);
1129 }
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 update_inode_page(inode);
1238 }
1239
1240 return ret;
1241 }
1242
1243 static long f2fs_fallocate(struct file *file, int mode,
1244 loff_t offset, loff_t len)
1245 {
1246 struct inode *inode = file_inode(file);
1247 long ret = 0;
1248
1249 /* f2fs only support ->fallocate for regular file */
1250 if (!S_ISREG(inode->i_mode))
1251 return -EINVAL;
1252
1253 if (f2fs_encrypted_inode(inode) &&
1254 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1255 return -EOPNOTSUPP;
1256
1257 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1258 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1259 FALLOC_FL_INSERT_RANGE))
1260 return -EOPNOTSUPP;
1261
1262 inode_lock(inode);
1263
1264 if (mode & FALLOC_FL_PUNCH_HOLE) {
1265 if (offset >= inode->i_size)
1266 goto out;
1267
1268 ret = punch_hole(inode, offset, len);
1269 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1270 ret = f2fs_collapse_range(inode, offset, len);
1271 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1272 ret = f2fs_zero_range(inode, offset, len, mode);
1273 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1274 ret = f2fs_insert_range(inode, offset, len);
1275 } else {
1276 ret = expand_inode_data(inode, offset, len, mode);
1277 }
1278
1279 if (!ret) {
1280 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1281 mark_inode_dirty_sync(inode);
1282 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1283 }
1284
1285 out:
1286 inode_unlock(inode);
1287
1288 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1289 return ret;
1290 }
1291
1292 static int f2fs_release_file(struct inode *inode, struct file *filp)
1293 {
1294 /*
1295 * f2fs_relase_file is called at every close calls. So we should
1296 * not drop any inmemory pages by close called by other process.
1297 */
1298 if (!(filp->f_mode & FMODE_WRITE) ||
1299 atomic_read(&inode->i_writecount) != 1)
1300 return 0;
1301
1302 /* some remained atomic pages should discarded */
1303 if (f2fs_is_atomic_file(inode))
1304 drop_inmem_pages(inode);
1305 if (f2fs_is_volatile_file(inode)) {
1306 clear_inode_flag(inode, FI_VOLATILE_FILE);
1307 set_inode_flag(inode, FI_DROP_CACHE);
1308 filemap_fdatawrite(inode->i_mapping);
1309 clear_inode_flag(inode, FI_DROP_CACHE);
1310 }
1311 return 0;
1312 }
1313
1314 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1315 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1316
1317 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1318 {
1319 if (S_ISDIR(mode))
1320 return flags;
1321 else if (S_ISREG(mode))
1322 return flags & F2FS_REG_FLMASK;
1323 else
1324 return flags & F2FS_OTHER_FLMASK;
1325 }
1326
1327 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1328 {
1329 struct inode *inode = file_inode(filp);
1330 struct f2fs_inode_info *fi = F2FS_I(inode);
1331 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1332 return put_user(flags, (int __user *)arg);
1333 }
1334
1335 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1336 {
1337 struct inode *inode = file_inode(filp);
1338 struct f2fs_inode_info *fi = F2FS_I(inode);
1339 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1340 unsigned int oldflags;
1341 int ret;
1342
1343 if (!inode_owner_or_capable(inode))
1344 return -EACCES;
1345
1346 if (get_user(flags, (int __user *)arg))
1347 return -EFAULT;
1348
1349 ret = mnt_want_write_file(filp);
1350 if (ret)
1351 return ret;
1352
1353 flags = f2fs_mask_flags(inode->i_mode, flags);
1354
1355 inode_lock(inode);
1356
1357 oldflags = fi->i_flags;
1358
1359 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1360 if (!capable(CAP_LINUX_IMMUTABLE)) {
1361 inode_unlock(inode);
1362 ret = -EPERM;
1363 goto out;
1364 }
1365 }
1366
1367 flags = flags & FS_FL_USER_MODIFIABLE;
1368 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1369 fi->i_flags = flags;
1370 inode_unlock(inode);
1371
1372 inode->i_ctime = CURRENT_TIME;
1373 f2fs_set_inode_flags(inode);
1374 out:
1375 mnt_drop_write_file(filp);
1376 return ret;
1377 }
1378
1379 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1380 {
1381 struct inode *inode = file_inode(filp);
1382
1383 return put_user(inode->i_generation, (int __user *)arg);
1384 }
1385
1386 static int f2fs_ioc_start_atomic_write(struct file *filp)
1387 {
1388 struct inode *inode = file_inode(filp);
1389 int ret;
1390
1391 if (!inode_owner_or_capable(inode))
1392 return -EACCES;
1393
1394 ret = mnt_want_write_file(filp);
1395 if (ret)
1396 return ret;
1397
1398 inode_lock(inode);
1399
1400 if (f2fs_is_atomic_file(inode))
1401 goto out;
1402
1403 ret = f2fs_convert_inline_inode(inode);
1404 if (ret)
1405 goto out;
1406
1407 set_inode_flag(inode, FI_ATOMIC_FILE);
1408 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1409
1410 if (!get_dirty_pages(inode))
1411 goto out;
1412
1413 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1414 "Unexpected flush for atomic writes: ino=%lu, npages=%lld",
1415 inode->i_ino, get_dirty_pages(inode));
1416 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1417 if (ret)
1418 clear_inode_flag(inode, FI_ATOMIC_FILE);
1419 out:
1420 inode_unlock(inode);
1421 mnt_drop_write_file(filp);
1422 return ret;
1423 }
1424
1425 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1426 {
1427 struct inode *inode = file_inode(filp);
1428 int ret;
1429
1430 if (!inode_owner_or_capable(inode))
1431 return -EACCES;
1432
1433 ret = mnt_want_write_file(filp);
1434 if (ret)
1435 return ret;
1436
1437 inode_lock(inode);
1438
1439 if (f2fs_is_volatile_file(inode))
1440 goto err_out;
1441
1442 if (f2fs_is_atomic_file(inode)) {
1443 clear_inode_flag(inode, FI_ATOMIC_FILE);
1444 ret = commit_inmem_pages(inode);
1445 if (ret) {
1446 set_inode_flag(inode, FI_ATOMIC_FILE);
1447 goto err_out;
1448 }
1449 }
1450
1451 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1452 err_out:
1453 inode_unlock(inode);
1454 mnt_drop_write_file(filp);
1455 return ret;
1456 }
1457
1458 static int f2fs_ioc_start_volatile_write(struct file *filp)
1459 {
1460 struct inode *inode = file_inode(filp);
1461 int ret;
1462
1463 if (!inode_owner_or_capable(inode))
1464 return -EACCES;
1465
1466 ret = mnt_want_write_file(filp);
1467 if (ret)
1468 return ret;
1469
1470 inode_lock(inode);
1471
1472 if (f2fs_is_volatile_file(inode))
1473 goto out;
1474
1475 ret = f2fs_convert_inline_inode(inode);
1476 if (ret)
1477 goto out;
1478
1479 set_inode_flag(inode, FI_VOLATILE_FILE);
1480 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1481 out:
1482 inode_unlock(inode);
1483 mnt_drop_write_file(filp);
1484 return ret;
1485 }
1486
1487 static int f2fs_ioc_release_volatile_write(struct file *filp)
1488 {
1489 struct inode *inode = file_inode(filp);
1490 int ret;
1491
1492 if (!inode_owner_or_capable(inode))
1493 return -EACCES;
1494
1495 ret = mnt_want_write_file(filp);
1496 if (ret)
1497 return ret;
1498
1499 inode_lock(inode);
1500
1501 if (!f2fs_is_volatile_file(inode))
1502 goto out;
1503
1504 if (!f2fs_is_first_block_written(inode)) {
1505 ret = truncate_partial_data_page(inode, 0, true);
1506 goto out;
1507 }
1508
1509 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1510 out:
1511 inode_unlock(inode);
1512 mnt_drop_write_file(filp);
1513 return ret;
1514 }
1515
1516 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1517 {
1518 struct inode *inode = file_inode(filp);
1519 int ret;
1520
1521 if (!inode_owner_or_capable(inode))
1522 return -EACCES;
1523
1524 ret = mnt_want_write_file(filp);
1525 if (ret)
1526 return ret;
1527
1528 inode_lock(inode);
1529
1530 if (f2fs_is_atomic_file(inode))
1531 drop_inmem_pages(inode);
1532 if (f2fs_is_volatile_file(inode)) {
1533 clear_inode_flag(inode, FI_VOLATILE_FILE);
1534 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1535 }
1536
1537 inode_unlock(inode);
1538
1539 mnt_drop_write_file(filp);
1540 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1541 return ret;
1542 }
1543
1544 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1545 {
1546 struct inode *inode = file_inode(filp);
1547 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1548 struct super_block *sb = sbi->sb;
1549 __u32 in;
1550 int ret;
1551
1552 if (!capable(CAP_SYS_ADMIN))
1553 return -EPERM;
1554
1555 if (get_user(in, (__u32 __user *)arg))
1556 return -EFAULT;
1557
1558 ret = mnt_want_write_file(filp);
1559 if (ret)
1560 return ret;
1561
1562 switch (in) {
1563 case F2FS_GOING_DOWN_FULLSYNC:
1564 sb = freeze_bdev(sb->s_bdev);
1565 if (sb && !IS_ERR(sb)) {
1566 f2fs_stop_checkpoint(sbi, false);
1567 thaw_bdev(sb->s_bdev, sb);
1568 }
1569 break;
1570 case F2FS_GOING_DOWN_METASYNC:
1571 /* do checkpoint only */
1572 f2fs_sync_fs(sb, 1);
1573 f2fs_stop_checkpoint(sbi, false);
1574 break;
1575 case F2FS_GOING_DOWN_NOSYNC:
1576 f2fs_stop_checkpoint(sbi, false);
1577 break;
1578 case F2FS_GOING_DOWN_METAFLUSH:
1579 sync_meta_pages(sbi, META, LONG_MAX);
1580 f2fs_stop_checkpoint(sbi, false);
1581 break;
1582 default:
1583 ret = -EINVAL;
1584 goto out;
1585 }
1586 f2fs_update_time(sbi, REQ_TIME);
1587 out:
1588 mnt_drop_write_file(filp);
1589 return ret;
1590 }
1591
1592 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1593 {
1594 struct inode *inode = file_inode(filp);
1595 struct super_block *sb = inode->i_sb;
1596 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1597 struct fstrim_range range;
1598 int ret;
1599
1600 if (!capable(CAP_SYS_ADMIN))
1601 return -EPERM;
1602
1603 if (!blk_queue_discard(q))
1604 return -EOPNOTSUPP;
1605
1606 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1607 sizeof(range)))
1608 return -EFAULT;
1609
1610 ret = mnt_want_write_file(filp);
1611 if (ret)
1612 return ret;
1613
1614 range.minlen = max((unsigned int)range.minlen,
1615 q->limits.discard_granularity);
1616 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1617 mnt_drop_write_file(filp);
1618 if (ret < 0)
1619 return ret;
1620
1621 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1622 sizeof(range)))
1623 return -EFAULT;
1624 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1625 return 0;
1626 }
1627
1628 static bool uuid_is_nonzero(__u8 u[16])
1629 {
1630 int i;
1631
1632 for (i = 0; i < 16; i++)
1633 if (u[i])
1634 return true;
1635 return false;
1636 }
1637
1638 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1639 {
1640 struct fscrypt_policy policy;
1641 struct inode *inode = file_inode(filp);
1642 int ret;
1643
1644 if (copy_from_user(&policy, (struct fscrypt_policy __user *)arg,
1645 sizeof(policy)))
1646 return -EFAULT;
1647
1648 ret = mnt_want_write_file(filp);
1649 if (ret)
1650 return ret;
1651
1652 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1653 ret = fscrypt_process_policy(inode, &policy);
1654
1655 mnt_drop_write_file(filp);
1656 return ret;
1657 }
1658
1659 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1660 {
1661 struct fscrypt_policy policy;
1662 struct inode *inode = file_inode(filp);
1663 int err;
1664
1665 err = fscrypt_get_policy(inode, &policy);
1666 if (err)
1667 return err;
1668
1669 if (copy_to_user((struct fscrypt_policy __user *)arg, &policy, sizeof(policy)))
1670 return -EFAULT;
1671 return 0;
1672 }
1673
1674 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1675 {
1676 struct inode *inode = file_inode(filp);
1677 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1678 int err;
1679
1680 if (!f2fs_sb_has_crypto(inode->i_sb))
1681 return -EOPNOTSUPP;
1682
1683 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1684 goto got_it;
1685
1686 err = mnt_want_write_file(filp);
1687 if (err)
1688 return err;
1689
1690 /* update superblock with uuid */
1691 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1692
1693 err = f2fs_commit_super(sbi, false);
1694 if (err) {
1695 /* undo new data */
1696 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1697 mnt_drop_write_file(filp);
1698 return err;
1699 }
1700 mnt_drop_write_file(filp);
1701 got_it:
1702 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1703 16))
1704 return -EFAULT;
1705 return 0;
1706 }
1707
1708 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1709 {
1710 struct inode *inode = file_inode(filp);
1711 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1712 __u32 sync;
1713 int ret;
1714
1715 if (!capable(CAP_SYS_ADMIN))
1716 return -EPERM;
1717
1718 if (get_user(sync, (__u32 __user *)arg))
1719 return -EFAULT;
1720
1721 if (f2fs_readonly(sbi->sb))
1722 return -EROFS;
1723
1724 ret = mnt_want_write_file(filp);
1725 if (ret)
1726 return ret;
1727
1728 if (!sync) {
1729 if (!mutex_trylock(&sbi->gc_mutex)) {
1730 ret = -EBUSY;
1731 goto out;
1732 }
1733 } else {
1734 mutex_lock(&sbi->gc_mutex);
1735 }
1736
1737 ret = f2fs_gc(sbi, sync);
1738 out:
1739 mnt_drop_write_file(filp);
1740 return ret;
1741 }
1742
1743 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1744 {
1745 struct inode *inode = file_inode(filp);
1746 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1747 int ret;
1748
1749 if (!capable(CAP_SYS_ADMIN))
1750 return -EPERM;
1751
1752 if (f2fs_readonly(sbi->sb))
1753 return -EROFS;
1754
1755 ret = mnt_want_write_file(filp);
1756 if (ret)
1757 return ret;
1758
1759 ret = f2fs_sync_fs(sbi->sb, 1);
1760
1761 mnt_drop_write_file(filp);
1762 return ret;
1763 }
1764
1765 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1766 struct file *filp,
1767 struct f2fs_defragment *range)
1768 {
1769 struct inode *inode = file_inode(filp);
1770 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1771 struct extent_info ei;
1772 pgoff_t pg_start, pg_end;
1773 unsigned int blk_per_seg = sbi->blocks_per_seg;
1774 unsigned int total = 0, sec_num;
1775 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1776 block_t blk_end = 0;
1777 bool fragmented = false;
1778 int err;
1779
1780 /* if in-place-update policy is enabled, don't waste time here */
1781 if (need_inplace_update(inode))
1782 return -EINVAL;
1783
1784 pg_start = range->start >> PAGE_SHIFT;
1785 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1786
1787 f2fs_balance_fs(sbi, true);
1788
1789 inode_lock(inode);
1790
1791 /* writeback all dirty pages in the range */
1792 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1793 range->start + range->len - 1);
1794 if (err)
1795 goto out;
1796
1797 /*
1798 * lookup mapping info in extent cache, skip defragmenting if physical
1799 * block addresses are continuous.
1800 */
1801 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1802 if (ei.fofs + ei.len >= pg_end)
1803 goto out;
1804 }
1805
1806 map.m_lblk = pg_start;
1807
1808 /*
1809 * lookup mapping info in dnode page cache, skip defragmenting if all
1810 * physical block addresses are continuous even if there are hole(s)
1811 * in logical blocks.
1812 */
1813 while (map.m_lblk < pg_end) {
1814 map.m_len = pg_end - map.m_lblk;
1815 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1816 if (err)
1817 goto out;
1818
1819 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1820 map.m_lblk++;
1821 continue;
1822 }
1823
1824 if (blk_end && blk_end != map.m_pblk) {
1825 fragmented = true;
1826 break;
1827 }
1828 blk_end = map.m_pblk + map.m_len;
1829
1830 map.m_lblk += map.m_len;
1831 }
1832
1833 if (!fragmented)
1834 goto out;
1835
1836 map.m_lblk = pg_start;
1837 map.m_len = pg_end - pg_start;
1838
1839 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1840
1841 /*
1842 * make sure there are enough free section for LFS allocation, this can
1843 * avoid defragment running in SSR mode when free section are allocated
1844 * intensively
1845 */
1846 if (has_not_enough_free_secs(sbi, sec_num)) {
1847 err = -EAGAIN;
1848 goto out;
1849 }
1850
1851 while (map.m_lblk < pg_end) {
1852 pgoff_t idx;
1853 int cnt = 0;
1854
1855 do_map:
1856 map.m_len = pg_end - map.m_lblk;
1857 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1858 if (err)
1859 goto clear_out;
1860
1861 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1862 map.m_lblk++;
1863 continue;
1864 }
1865
1866 set_inode_flag(inode, FI_DO_DEFRAG);
1867
1868 idx = map.m_lblk;
1869 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1870 struct page *page;
1871
1872 page = get_lock_data_page(inode, idx, true);
1873 if (IS_ERR(page)) {
1874 err = PTR_ERR(page);
1875 goto clear_out;
1876 }
1877
1878 set_page_dirty(page);
1879 f2fs_put_page(page, 1);
1880
1881 idx++;
1882 cnt++;
1883 total++;
1884 }
1885
1886 map.m_lblk = idx;
1887
1888 if (idx < pg_end && cnt < blk_per_seg)
1889 goto do_map;
1890
1891 clear_inode_flag(inode, FI_DO_DEFRAG);
1892
1893 err = filemap_fdatawrite(inode->i_mapping);
1894 if (err)
1895 goto out;
1896 }
1897 clear_out:
1898 clear_inode_flag(inode, FI_DO_DEFRAG);
1899 out:
1900 inode_unlock(inode);
1901 if (!err)
1902 range->len = (u64)total << PAGE_SHIFT;
1903 return err;
1904 }
1905
1906 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1907 {
1908 struct inode *inode = file_inode(filp);
1909 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1910 struct f2fs_defragment range;
1911 int err;
1912
1913 if (!capable(CAP_SYS_ADMIN))
1914 return -EPERM;
1915
1916 if (!S_ISREG(inode->i_mode))
1917 return -EINVAL;
1918
1919 err = mnt_want_write_file(filp);
1920 if (err)
1921 return err;
1922
1923 if (f2fs_readonly(sbi->sb)) {
1924 err = -EROFS;
1925 goto out;
1926 }
1927
1928 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1929 sizeof(range))) {
1930 err = -EFAULT;
1931 goto out;
1932 }
1933
1934 /* verify alignment of offset & size */
1935 if (range.start & (F2FS_BLKSIZE - 1) ||
1936 range.len & (F2FS_BLKSIZE - 1)) {
1937 err = -EINVAL;
1938 goto out;
1939 }
1940
1941 err = f2fs_defragment_range(sbi, filp, &range);
1942 f2fs_update_time(sbi, REQ_TIME);
1943 if (err < 0)
1944 goto out;
1945
1946 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1947 sizeof(range)))
1948 err = -EFAULT;
1949 out:
1950 mnt_drop_write_file(filp);
1951 return err;
1952 }
1953
1954 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1955 {
1956 switch (cmd) {
1957 case F2FS_IOC_GETFLAGS:
1958 return f2fs_ioc_getflags(filp, arg);
1959 case F2FS_IOC_SETFLAGS:
1960 return f2fs_ioc_setflags(filp, arg);
1961 case F2FS_IOC_GETVERSION:
1962 return f2fs_ioc_getversion(filp, arg);
1963 case F2FS_IOC_START_ATOMIC_WRITE:
1964 return f2fs_ioc_start_atomic_write(filp);
1965 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1966 return f2fs_ioc_commit_atomic_write(filp);
1967 case F2FS_IOC_START_VOLATILE_WRITE:
1968 return f2fs_ioc_start_volatile_write(filp);
1969 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1970 return f2fs_ioc_release_volatile_write(filp);
1971 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1972 return f2fs_ioc_abort_volatile_write(filp);
1973 case F2FS_IOC_SHUTDOWN:
1974 return f2fs_ioc_shutdown(filp, arg);
1975 case FITRIM:
1976 return f2fs_ioc_fitrim(filp, arg);
1977 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1978 return f2fs_ioc_set_encryption_policy(filp, arg);
1979 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1980 return f2fs_ioc_get_encryption_policy(filp, arg);
1981 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1982 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1983 case F2FS_IOC_GARBAGE_COLLECT:
1984 return f2fs_ioc_gc(filp, arg);
1985 case F2FS_IOC_WRITE_CHECKPOINT:
1986 return f2fs_ioc_write_checkpoint(filp, arg);
1987 case F2FS_IOC_DEFRAGMENT:
1988 return f2fs_ioc_defragment(filp, arg);
1989 default:
1990 return -ENOTTY;
1991 }
1992 }
1993
1994 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1995 {
1996 struct file *file = iocb->ki_filp;
1997 struct inode *inode = file_inode(file);
1998 ssize_t ret;
1999
2000 if (f2fs_encrypted_inode(inode) &&
2001 !fscrypt_has_encryption_key(inode) &&
2002 fscrypt_get_encryption_info(inode))
2003 return -EACCES;
2004
2005 inode_lock(inode);
2006 ret = generic_write_checks(iocb, from);
2007 if (ret > 0) {
2008 ret = f2fs_preallocate_blocks(iocb, from);
2009 if (!ret)
2010 ret = __generic_file_write_iter(iocb, from);
2011 }
2012 inode_unlock(inode);
2013
2014 if (ret > 0)
2015 ret = generic_write_sync(iocb, ret);
2016 return ret;
2017 }
2018
2019 #ifdef CONFIG_COMPAT
2020 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2021 {
2022 switch (cmd) {
2023 case F2FS_IOC32_GETFLAGS:
2024 cmd = F2FS_IOC_GETFLAGS;
2025 break;
2026 case F2FS_IOC32_SETFLAGS:
2027 cmd = F2FS_IOC_SETFLAGS;
2028 break;
2029 case F2FS_IOC32_GETVERSION:
2030 cmd = F2FS_IOC_GETVERSION;
2031 break;
2032 case F2FS_IOC_START_ATOMIC_WRITE:
2033 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2034 case F2FS_IOC_START_VOLATILE_WRITE:
2035 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2036 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2037 case F2FS_IOC_SHUTDOWN:
2038 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2039 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2040 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2041 case F2FS_IOC_GARBAGE_COLLECT:
2042 case F2FS_IOC_WRITE_CHECKPOINT:
2043 case F2FS_IOC_DEFRAGMENT:
2044 break;
2045 default:
2046 return -ENOIOCTLCMD;
2047 }
2048 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2049 }
2050 #endif
2051
2052 const struct file_operations f2fs_file_operations = {
2053 .llseek = f2fs_llseek,
2054 .read_iter = generic_file_read_iter,
2055 .write_iter = f2fs_file_write_iter,
2056 .open = f2fs_file_open,
2057 .release = f2fs_release_file,
2058 .mmap = f2fs_file_mmap,
2059 .fsync = f2fs_sync_file,
2060 .fallocate = f2fs_fallocate,
2061 .unlocked_ioctl = f2fs_ioctl,
2062 #ifdef CONFIG_COMPAT
2063 .compat_ioctl = f2fs_compat_ioctl,
2064 #endif
2065 .splice_read = generic_file_splice_read,
2066 .splice_write = iter_file_splice_write,
2067 };
Ubuntu Artful kernel mirror