]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - fs/f2fs/file.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:


888ce47657790fd1497fa4b0c6640663276c419b
[mirror_ubuntu-bionic-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/random.h>
24
25 #include "f2fs.h"
26 #include "node.h"
27 #include "segment.h"
28 #include "xattr.h"
29 #include "acl.h"
30 #include "gc.h"
31 #include "trace.h"
32 #include <trace/events/f2fs.h>
33
34 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
35 struct vm_fault *vmf)
36 {
37 struct page *page = vmf->page;
38 struct inode *inode = file_inode(vma->vm_file);
39 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
40 struct dnode_of_data dn;
41 int err;
42
43 sb_start_pagefault(inode->i_sb);
44
45 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
46
47 /* block allocation */
48 f2fs_lock_op(sbi);
49 set_new_dnode(&dn, inode, NULL, NULL, 0);
50 err = f2fs_reserve_block(&dn, page->index);
51 if (err) {
52 f2fs_unlock_op(sbi);
53 goto out;
54 }
55 f2fs_put_dnode(&dn);
56 f2fs_unlock_op(sbi);
57
58 if (dn.node_changed)
59 f2fs_balance_fs(sbi);
60
61 file_update_time(vma->vm_file);
62 lock_page(page);
63 if (unlikely(page->mapping != inode->i_mapping ||
64 page_offset(page) > i_size_read(inode) ||
65 !PageUptodate(page))) {
66 unlock_page(page);
67 err = -EFAULT;
68 goto out;
69 }
70
71 /*
72 * check to see if the page is mapped already (no holes)
73 */
74 if (PageMappedToDisk(page))
75 goto mapped;
76
77 /* page is wholly or partially inside EOF */
78 if (((loff_t)(page->index + 1) << PAGE_CACHE_SHIFT) >
79 i_size_read(inode)) {
80 unsigned offset;
81 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
82 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
83 }
84 set_page_dirty(page);
85 SetPageUptodate(page);
86
87 trace_f2fs_vm_page_mkwrite(page, DATA);
88 mapped:
89 /* fill the page */
90 f2fs_wait_on_page_writeback(page, DATA);
91
92 /* wait for GCed encrypted page writeback */
93 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
94 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
95
96 /* if gced page is attached, don't write to cold segment */
97 clear_cold_data(page);
98 out:
99 sb_end_pagefault(inode->i_sb);
100 return block_page_mkwrite_return(err);
101 }
102
103 static const struct vm_operations_struct f2fs_file_vm_ops = {
104 .fault = filemap_fault,
105 .map_pages = filemap_map_pages,
106 .page_mkwrite = f2fs_vm_page_mkwrite,
107 };
108
109 static int get_parent_ino(struct inode *inode, nid_t *pino)
110 {
111 struct dentry *dentry;
112
113 inode = igrab(inode);
114 dentry = d_find_any_alias(inode);
115 iput(inode);
116 if (!dentry)
117 return 0;
118
119 if (update_dent_inode(inode, inode, &dentry->d_name)) {
120 dput(dentry);
121 return 0;
122 }
123
124 *pino = parent_ino(dentry);
125 dput(dentry);
126 return 1;
127 }
128
129 static inline bool need_do_checkpoint(struct inode *inode)
130 {
131 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
132 bool need_cp = false;
133
134 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
135 need_cp = true;
136 else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
137 need_cp = true;
138 else if (file_wrong_pino(inode))
139 need_cp = true;
140 else if (!space_for_roll_forward(sbi))
141 need_cp = true;
142 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
143 need_cp = true;
144 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
145 need_cp = true;
146 else if (test_opt(sbi, FASTBOOT))
147 need_cp = true;
148 else if (sbi->active_logs == 2)
149 need_cp = true;
150
151 return need_cp;
152 }
153
154 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
155 {
156 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
157 bool ret = false;
158 /* But we need to avoid that there are some inode updates */
159 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
160 ret = true;
161 f2fs_put_page(i, 0);
162 return ret;
163 }
164
165 static void try_to_fix_pino(struct inode *inode)
166 {
167 struct f2fs_inode_info *fi = F2FS_I(inode);
168 nid_t pino;
169
170 down_write(&fi->i_sem);
171 fi->xattr_ver = 0;
172 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
173 get_parent_ino(inode, &pino)) {
174 fi->i_pino = pino;
175 file_got_pino(inode);
176 up_write(&fi->i_sem);
177
178 mark_inode_dirty_sync(inode);
179 f2fs_write_inode(inode, NULL);
180 } else {
181 up_write(&fi->i_sem);
182 }
183 }
184
185 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
186 {
187 struct inode *inode = file->f_mapping->host;
188 struct f2fs_inode_info *fi = F2FS_I(inode);
189 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
190 nid_t ino = inode->i_ino;
191 int ret = 0;
192 bool need_cp = false;
193 struct writeback_control wbc = {
194 .sync_mode = WB_SYNC_ALL,
195 .nr_to_write = LONG_MAX,
196 .for_reclaim = 0,
197 };
198
199 if (unlikely(f2fs_readonly(inode->i_sb)))
200 return 0;
201
202 trace_f2fs_sync_file_enter(inode);
203
204 /* if fdatasync is triggered, let's do in-place-update */
205 if (get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
206 set_inode_flag(fi, FI_NEED_IPU);
207 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
208 clear_inode_flag(fi, FI_NEED_IPU);
209
210 if (ret) {
211 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
212 return ret;
213 }
214
215 /* if the inode is dirty, let's recover all the time */
216 if (!datasync) {
217 f2fs_write_inode(inode, NULL);
218 goto go_write;
219 }
220
221 /*
222 * if there is no written data, don't waste time to write recovery info.
223 */
224 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
225 !exist_written_data(sbi, ino, APPEND_INO)) {
226
227 /* it may call write_inode just prior to fsync */
228 if (need_inode_page_update(sbi, ino))
229 goto go_write;
230
231 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
232 exist_written_data(sbi, ino, UPDATE_INO))
233 goto flush_out;
234 goto out;
235 }
236 go_write:
237 /*
238 * Both of fdatasync() and fsync() are able to be recovered from
239 * sudden-power-off.
240 */
241 down_read(&fi->i_sem);
242 need_cp = need_do_checkpoint(inode);
243 up_read(&fi->i_sem);
244
245 if (need_cp) {
246 /* all the dirty node pages should be flushed for POR */
247 ret = f2fs_sync_fs(inode->i_sb, 1);
248
249 /*
250 * We've secured consistency through sync_fs. Following pino
251 * will be used only for fsynced inodes after checkpoint.
252 */
253 try_to_fix_pino(inode);
254 clear_inode_flag(fi, FI_APPEND_WRITE);
255 clear_inode_flag(fi, FI_UPDATE_WRITE);
256 goto out;
257 }
258 sync_nodes:
259 sync_node_pages(sbi, ino, &wbc);
260
261 /* if cp_error was enabled, we should avoid infinite loop */
262 if (unlikely(f2fs_cp_error(sbi)))
263 goto out;
264
265 if (need_inode_block_update(sbi, ino)) {
266 mark_inode_dirty_sync(inode);
267 f2fs_write_inode(inode, NULL);
268
269 f2fs_balance_fs(sbi);
270 goto sync_nodes;
271 }
272
273 ret = wait_on_node_pages_writeback(sbi, ino);
274 if (ret)
275 goto out;
276
277 /* once recovery info is written, don't need to tack this */
278 remove_ino_entry(sbi, ino, APPEND_INO);
279 clear_inode_flag(fi, FI_APPEND_WRITE);
280 flush_out:
281 remove_ino_entry(sbi, ino, UPDATE_INO);
282 clear_inode_flag(fi, FI_UPDATE_WRITE);
283 ret = f2fs_issue_flush(sbi);
284 out:
285 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
286 f2fs_trace_ios(NULL, 1);
287 return ret;
288 }
289
290 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
291 pgoff_t pgofs, int whence)
292 {
293 struct pagevec pvec;
294 int nr_pages;
295
296 if (whence != SEEK_DATA)
297 return 0;
298
299 /* find first dirty page index */
300 pagevec_init(&pvec, 0);
301 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
302 PAGECACHE_TAG_DIRTY, 1);
303 pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
304 pagevec_release(&pvec);
305 return pgofs;
306 }
307
308 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
309 int whence)
310 {
311 switch (whence) {
312 case SEEK_DATA:
313 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
314 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
315 return true;
316 break;
317 case SEEK_HOLE:
318 if (blkaddr == NULL_ADDR)
319 return true;
320 break;
321 }
322 return false;
323 }
324
325 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
326 {
327 struct inode *inode = file->f_mapping->host;
328 loff_t maxbytes = inode->i_sb->s_maxbytes;
329 struct dnode_of_data dn;
330 pgoff_t pgofs, end_offset, dirty;
331 loff_t data_ofs = offset;
332 loff_t isize;
333 int err = 0;
334
335 mutex_lock(&inode->i_mutex);
336
337 isize = i_size_read(inode);
338 if (offset >= isize)
339 goto fail;
340
341 /* handle inline data case */
342 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
343 if (whence == SEEK_HOLE)
344 data_ofs = isize;
345 goto found;
346 }
347
348 pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
349
350 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
351
352 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
353 set_new_dnode(&dn, inode, NULL, NULL, 0);
354 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
355 if (err && err != -ENOENT) {
356 goto fail;
357 } else if (err == -ENOENT) {
358 /* direct node does not exists */
359 if (whence == SEEK_DATA) {
360 pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
361 F2FS_I(inode));
362 continue;
363 } else {
364 goto found;
365 }
366 }
367
368 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
369
370 /* find data/hole in dnode block */
371 for (; dn.ofs_in_node < end_offset;
372 dn.ofs_in_node++, pgofs++,
373 data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
374 block_t blkaddr;
375 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
376
377 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
378 f2fs_put_dnode(&dn);
379 goto found;
380 }
381 }
382 f2fs_put_dnode(&dn);
383 }
384
385 if (whence == SEEK_DATA)
386 goto fail;
387 found:
388 if (whence == SEEK_HOLE && data_ofs > isize)
389 data_ofs = isize;
390 mutex_unlock(&inode->i_mutex);
391 return vfs_setpos(file, data_ofs, maxbytes);
392 fail:
393 mutex_unlock(&inode->i_mutex);
394 return -ENXIO;
395 }
396
397 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
398 {
399 struct inode *inode = file->f_mapping->host;
400 loff_t maxbytes = inode->i_sb->s_maxbytes;
401
402 switch (whence) {
403 case SEEK_SET:
404 case SEEK_CUR:
405 case SEEK_END:
406 return generic_file_llseek_size(file, offset, whence,
407 maxbytes, i_size_read(inode));
408 case SEEK_DATA:
409 case SEEK_HOLE:
410 if (offset < 0)
411 return -ENXIO;
412 return f2fs_seek_block(file, offset, whence);
413 }
414
415 return -EINVAL;
416 }
417
418 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
419 {
420 struct inode *inode = file_inode(file);
421 int err;
422
423 if (f2fs_encrypted_inode(inode)) {
424 err = f2fs_get_encryption_info(inode);
425 if (err)
426 return 0;
427 }
428
429 /* we don't need to use inline_data strictly */
430 err = f2fs_convert_inline_inode(inode);
431 if (err)
432 return err;
433
434 file_accessed(file);
435 vma->vm_ops = &f2fs_file_vm_ops;
436 return 0;
437 }
438
439 static int f2fs_file_open(struct inode *inode, struct file *filp)
440 {
441 int ret = generic_file_open(inode, filp);
442
443 if (!ret && f2fs_encrypted_inode(inode)) {
444 ret = f2fs_get_encryption_info(inode);
445 if (ret)
446 ret = -EACCES;
447 }
448 return ret;
449 }
450
451 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
452 {
453 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
454 struct f2fs_node *raw_node;
455 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
456 __le32 *addr;
457
458 raw_node = F2FS_NODE(dn->node_page);
459 addr = blkaddr_in_node(raw_node) + ofs;
460
461 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
462 block_t blkaddr = le32_to_cpu(*addr);
463 if (blkaddr == NULL_ADDR)
464 continue;
465
466 dn->data_blkaddr = NULL_ADDR;
467 set_data_blkaddr(dn);
468 invalidate_blocks(sbi, blkaddr);
469 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
470 clear_inode_flag(F2FS_I(dn->inode),
471 FI_FIRST_BLOCK_WRITTEN);
472 nr_free++;
473 }
474
475 if (nr_free) {
476 pgoff_t fofs;
477 /*
478 * once we invalidate valid blkaddr in range [ofs, ofs + count],
479 * we will invalidate all blkaddr in the whole range.
480 */
481 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
482 F2FS_I(dn->inode)) + ofs;
483 f2fs_update_extent_cache_range(dn, fofs, 0, len);
484 dec_valid_block_count(sbi, dn->inode, nr_free);
485 set_page_dirty(dn->node_page);
486 sync_inode_page(dn);
487 }
488 dn->ofs_in_node = ofs;
489
490 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
491 dn->ofs_in_node, nr_free);
492 return nr_free;
493 }
494
495 void truncate_data_blocks(struct dnode_of_data *dn)
496 {
497 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
498 }
499
500 static int truncate_partial_data_page(struct inode *inode, u64 from,
501 bool cache_only)
502 {
503 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
504 pgoff_t index = from >> PAGE_CACHE_SHIFT;
505 struct address_space *mapping = inode->i_mapping;
506 struct page *page;
507
508 if (!offset && !cache_only)
509 return 0;
510
511 if (cache_only) {
512 page = f2fs_grab_cache_page(mapping, index, false);
513 if (page && PageUptodate(page))
514 goto truncate_out;
515 f2fs_put_page(page, 1);
516 return 0;
517 }
518
519 page = get_lock_data_page(inode, index, true);
520 if (IS_ERR(page))
521 return 0;
522 truncate_out:
523 f2fs_wait_on_page_writeback(page, DATA);
524 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
525 if (!cache_only || !f2fs_encrypted_inode(inode) || !S_ISREG(inode->i_mode))
526 set_page_dirty(page);
527 f2fs_put_page(page, 1);
528 return 0;
529 }
530
531 int truncate_blocks(struct inode *inode, u64 from, bool lock)
532 {
533 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
534 unsigned int blocksize = inode->i_sb->s_blocksize;
535 struct dnode_of_data dn;
536 pgoff_t free_from;
537 int count = 0, err = 0;
538 struct page *ipage;
539 bool truncate_page = false;
540
541 trace_f2fs_truncate_blocks_enter(inode, from);
542
543 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
544
545 if (lock)
546 f2fs_lock_op(sbi);
547
548 ipage = get_node_page(sbi, inode->i_ino);
549 if (IS_ERR(ipage)) {
550 err = PTR_ERR(ipage);
551 goto out;
552 }
553
554 if (f2fs_has_inline_data(inode)) {
555 if (truncate_inline_inode(ipage, from))
556 set_page_dirty(ipage);
557 f2fs_put_page(ipage, 1);
558 truncate_page = true;
559 goto out;
560 }
561
562 set_new_dnode(&dn, inode, ipage, NULL, 0);
563 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
564 if (err) {
565 if (err == -ENOENT)
566 goto free_next;
567 goto out;
568 }
569
570 count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
571
572 count -= dn.ofs_in_node;
573 f2fs_bug_on(sbi, count < 0);
574
575 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
576 truncate_data_blocks_range(&dn, count);
577 free_from += count;
578 }
579
580 f2fs_put_dnode(&dn);
581 free_next:
582 err = truncate_inode_blocks(inode, free_from);
583 out:
584 if (lock)
585 f2fs_unlock_op(sbi);
586
587 /* lastly zero out the first data page */
588 if (!err)
589 err = truncate_partial_data_page(inode, from, truncate_page);
590
591 trace_f2fs_truncate_blocks_exit(inode, err);
592 return err;
593 }
594
595 int f2fs_truncate(struct inode *inode, bool lock)
596 {
597 int err;
598
599 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
600 S_ISLNK(inode->i_mode)))
601 return 0;
602
603 trace_f2fs_truncate(inode);
604
605 /* we should check inline_data size */
606 if (!f2fs_may_inline_data(inode)) {
607 err = f2fs_convert_inline_inode(inode);
608 if (err)
609 return err;
610 }
611
612 err = truncate_blocks(inode, i_size_read(inode), lock);
613 if (err)
614 return err;
615
616 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617 mark_inode_dirty(inode);
618 return 0;
619 }
620
621 int f2fs_getattr(struct vfsmount *mnt,
622 struct dentry *dentry, struct kstat *stat)
623 {
624 struct inode *inode = d_inode(dentry);
625 generic_fillattr(inode, stat);
626 stat->blocks <<= 3;
627 return 0;
628 }
629
630 #ifdef CONFIG_F2FS_FS_POSIX_ACL
631 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
632 {
633 struct f2fs_inode_info *fi = F2FS_I(inode);
634 unsigned int ia_valid = attr->ia_valid;
635
636 if (ia_valid & ATTR_UID)
637 inode->i_uid = attr->ia_uid;
638 if (ia_valid & ATTR_GID)
639 inode->i_gid = attr->ia_gid;
640 if (ia_valid & ATTR_ATIME)
641 inode->i_atime = timespec_trunc(attr->ia_atime,
642 inode->i_sb->s_time_gran);
643 if (ia_valid & ATTR_MTIME)
644 inode->i_mtime = timespec_trunc(attr->ia_mtime,
645 inode->i_sb->s_time_gran);
646 if (ia_valid & ATTR_CTIME)
647 inode->i_ctime = timespec_trunc(attr->ia_ctime,
648 inode->i_sb->s_time_gran);
649 if (ia_valid & ATTR_MODE) {
650 umode_t mode = attr->ia_mode;
651
652 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
653 mode &= ~S_ISGID;
654 set_acl_inode(fi, mode);
655 }
656 }
657 #else
658 #define __setattr_copy setattr_copy
659 #endif
660
661 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
662 {
663 struct inode *inode = d_inode(dentry);
664 struct f2fs_inode_info *fi = F2FS_I(inode);
665 int err;
666
667 err = inode_change_ok(inode, attr);
668 if (err)
669 return err;
670
671 if (attr->ia_valid & ATTR_SIZE) {
672 if (f2fs_encrypted_inode(inode) &&
673 f2fs_get_encryption_info(inode))
674 return -EACCES;
675
676 if (attr->ia_size <= i_size_read(inode)) {
677 truncate_setsize(inode, attr->ia_size);
678 err = f2fs_truncate(inode, true);
679 if (err)
680 return err;
681 f2fs_balance_fs(F2FS_I_SB(inode));
682 } else {
683 /*
684 * do not trim all blocks after i_size if target size is
685 * larger than i_size.
686 */
687 truncate_setsize(inode, attr->ia_size);
688
689 /* should convert inline inode here */
690 if (!f2fs_may_inline_data(inode)) {
691 err = f2fs_convert_inline_inode(inode);
692 if (err)
693 return err;
694 }
695 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
696 }
697 }
698
699 __setattr_copy(inode, attr);
700
701 if (attr->ia_valid & ATTR_MODE) {
702 err = posix_acl_chmod(inode, get_inode_mode(inode));
703 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
704 inode->i_mode = fi->i_acl_mode;
705 clear_inode_flag(fi, FI_ACL_MODE);
706 }
707 }
708
709 mark_inode_dirty(inode);
710 return err;
711 }
712
713 const struct inode_operations f2fs_file_inode_operations = {
714 .getattr = f2fs_getattr,
715 .setattr = f2fs_setattr,
716 .get_acl = f2fs_get_acl,
717 .set_acl = f2fs_set_acl,
718 #ifdef CONFIG_F2FS_FS_XATTR
719 .setxattr = generic_setxattr,
720 .getxattr = generic_getxattr,
721 .listxattr = f2fs_listxattr,
722 .removexattr = generic_removexattr,
723 #endif
724 .fiemap = f2fs_fiemap,
725 };
726
727 static int fill_zero(struct inode *inode, pgoff_t index,
728 loff_t start, loff_t len)
729 {
730 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
731 struct page *page;
732
733 if (!len)
734 return 0;
735
736 f2fs_balance_fs(sbi);
737
738 f2fs_lock_op(sbi);
739 page = get_new_data_page(inode, NULL, index, false);
740 f2fs_unlock_op(sbi);
741
742 if (IS_ERR(page))
743 return PTR_ERR(page);
744
745 f2fs_wait_on_page_writeback(page, DATA);
746 zero_user(page, start, len);
747 set_page_dirty(page);
748 f2fs_put_page(page, 1);
749 return 0;
750 }
751
752 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
753 {
754 int err;
755
756 while (pg_start < pg_end) {
757 struct dnode_of_data dn;
758 pgoff_t end_offset, count;
759
760 set_new_dnode(&dn, inode, NULL, NULL, 0);
761 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
762 if (err) {
763 if (err == -ENOENT) {
764 pg_start++;
765 continue;
766 }
767 return err;
768 }
769
770 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
771 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
772
773 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
774
775 truncate_data_blocks_range(&dn, count);
776 f2fs_put_dnode(&dn);
777
778 pg_start += count;
779 }
780 return 0;
781 }
782
783 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
784 {
785 pgoff_t pg_start, pg_end;
786 loff_t off_start, off_end;
787 int ret;
788
789 ret = f2fs_convert_inline_inode(inode);
790 if (ret)
791 return ret;
792
793 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
794 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
795
796 off_start = offset & (PAGE_CACHE_SIZE - 1);
797 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
798
799 if (pg_start == pg_end) {
800 ret = fill_zero(inode, pg_start, off_start,
801 off_end - off_start);
802 if (ret)
803 return ret;
804 } else {
805 if (off_start) {
806 ret = fill_zero(inode, pg_start++, off_start,
807 PAGE_CACHE_SIZE - off_start);
808 if (ret)
809 return ret;
810 }
811 if (off_end) {
812 ret = fill_zero(inode, pg_end, 0, off_end);
813 if (ret)
814 return ret;
815 }
816
817 if (pg_start < pg_end) {
818 struct address_space *mapping = inode->i_mapping;
819 loff_t blk_start, blk_end;
820 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
821
822 f2fs_balance_fs(sbi);
823
824 blk_start = (loff_t)pg_start << PAGE_CACHE_SHIFT;
825 blk_end = (loff_t)pg_end << PAGE_CACHE_SHIFT;
826 truncate_inode_pages_range(mapping, blk_start,
827 blk_end - 1);
828
829 f2fs_lock_op(sbi);
830 ret = truncate_hole(inode, pg_start, pg_end);
831 f2fs_unlock_op(sbi);
832 }
833 }
834
835 return ret;
836 }
837
838 static int __exchange_data_block(struct inode *inode, pgoff_t src,
839 pgoff_t dst, bool full)
840 {
841 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
842 struct dnode_of_data dn;
843 block_t new_addr;
844 bool do_replace = false;
845 int ret;
846
847 set_new_dnode(&dn, inode, NULL, NULL, 0);
848 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
849 if (ret && ret != -ENOENT) {
850 return ret;
851 } else if (ret == -ENOENT) {
852 new_addr = NULL_ADDR;
853 } else {
854 new_addr = dn.data_blkaddr;
855 if (!is_checkpointed_data(sbi, new_addr)) {
856 dn.data_blkaddr = NULL_ADDR;
857 /* do not invalidate this block address */
858 set_data_blkaddr(&dn);
859 f2fs_update_extent_cache(&dn);
860 do_replace = true;
861 }
862 f2fs_put_dnode(&dn);
863 }
864
865 if (new_addr == NULL_ADDR)
866 return full ? truncate_hole(inode, dst, dst + 1) : 0;
867
868 if (do_replace) {
869 struct page *ipage = get_node_page(sbi, inode->i_ino);
870 struct node_info ni;
871
872 if (IS_ERR(ipage)) {
873 ret = PTR_ERR(ipage);
874 goto err_out;
875 }
876
877 set_new_dnode(&dn, inode, ipage, NULL, 0);
878 ret = f2fs_reserve_block(&dn, dst);
879 if (ret)
880 goto err_out;
881
882 truncate_data_blocks_range(&dn, 1);
883
884 get_node_info(sbi, dn.nid, &ni);
885 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
886 ni.version, true);
887 f2fs_put_dnode(&dn);
888 } else {
889 struct page *psrc, *pdst;
890
891 psrc = get_lock_data_page(inode, src, true);
892 if (IS_ERR(psrc))
893 return PTR_ERR(psrc);
894 pdst = get_new_data_page(inode, NULL, dst, false);
895 if (IS_ERR(pdst)) {
896 f2fs_put_page(psrc, 1);
897 return PTR_ERR(pdst);
898 }
899 f2fs_copy_page(psrc, pdst);
900 set_page_dirty(pdst);
901 f2fs_put_page(pdst, 1);
902 f2fs_put_page(psrc, 1);
903
904 return truncate_hole(inode, src, src + 1);
905 }
906 return 0;
907
908 err_out:
909 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
910 dn.data_blkaddr = new_addr;
911 set_data_blkaddr(&dn);
912 f2fs_update_extent_cache(&dn);
913 f2fs_put_dnode(&dn);
914 }
915 return ret;
916 }
917
918 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
919 {
920 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
921 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
922 int ret = 0;
923
924 for (; end < nrpages; start++, end++) {
925 f2fs_balance_fs(sbi);
926 f2fs_lock_op(sbi);
927 ret = __exchange_data_block(inode, end, start, true);
928 f2fs_unlock_op(sbi);
929 if (ret)
930 break;
931 }
932 return ret;
933 }
934
935 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
936 {
937 pgoff_t pg_start, pg_end;
938 loff_t new_size;
939 int ret;
940
941 if (offset + len >= i_size_read(inode))
942 return -EINVAL;
943
944 /* collapse range should be aligned to block size of f2fs. */
945 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
946 return -EINVAL;
947
948 ret = f2fs_convert_inline_inode(inode);
949 if (ret)
950 return ret;
951
952 pg_start = offset >> PAGE_CACHE_SHIFT;
953 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
954
955 /* write out all dirty pages from offset */
956 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
957 if (ret)
958 return ret;
959
960 truncate_pagecache(inode, offset);
961
962 ret = f2fs_do_collapse(inode, pg_start, pg_end);
963 if (ret)
964 return ret;
965
966 /* write out all moved pages, if possible */
967 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
968 truncate_pagecache(inode, offset);
969
970 new_size = i_size_read(inode) - len;
971 truncate_pagecache(inode, new_size);
972
973 ret = truncate_blocks(inode, new_size, true);
974 if (!ret)
975 i_size_write(inode, new_size);
976
977 return ret;
978 }
979
980 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
981 int mode)
982 {
983 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
984 struct address_space *mapping = inode->i_mapping;
985 pgoff_t index, pg_start, pg_end;
986 loff_t new_size = i_size_read(inode);
987 loff_t off_start, off_end;
988 int ret = 0;
989
990 ret = inode_newsize_ok(inode, (len + offset));
991 if (ret)
992 return ret;
993
994 ret = f2fs_convert_inline_inode(inode);
995 if (ret)
996 return ret;
997
998 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
999 if (ret)
1000 return ret;
1001
1002 truncate_pagecache_range(inode, offset, offset + len - 1);
1003
1004 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1005 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1006
1007 off_start = offset & (PAGE_CACHE_SIZE - 1);
1008 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1009
1010 if (pg_start == pg_end) {
1011 ret = fill_zero(inode, pg_start, off_start,
1012 off_end - off_start);
1013 if (ret)
1014 return ret;
1015
1016 if (offset + len > new_size)
1017 new_size = offset + len;
1018 new_size = max_t(loff_t, new_size, offset + len);
1019 } else {
1020 if (off_start) {
1021 ret = fill_zero(inode, pg_start++, off_start,
1022 PAGE_CACHE_SIZE - off_start);
1023 if (ret)
1024 return ret;
1025
1026 new_size = max_t(loff_t, new_size,
1027 (loff_t)pg_start << PAGE_CACHE_SHIFT);
1028 }
1029
1030 for (index = pg_start; index < pg_end; index++) {
1031 struct dnode_of_data dn;
1032 struct page *ipage;
1033
1034 f2fs_lock_op(sbi);
1035
1036 ipage = get_node_page(sbi, inode->i_ino);
1037 if (IS_ERR(ipage)) {
1038 ret = PTR_ERR(ipage);
1039 f2fs_unlock_op(sbi);
1040 goto out;
1041 }
1042
1043 set_new_dnode(&dn, inode, ipage, NULL, 0);
1044 ret = f2fs_reserve_block(&dn, index);
1045 if (ret) {
1046 f2fs_unlock_op(sbi);
1047 goto out;
1048 }
1049
1050 if (dn.data_blkaddr != NEW_ADDR) {
1051 invalidate_blocks(sbi, dn.data_blkaddr);
1052
1053 dn.data_blkaddr = NEW_ADDR;
1054 set_data_blkaddr(&dn);
1055
1056 dn.data_blkaddr = NULL_ADDR;
1057 f2fs_update_extent_cache(&dn);
1058 }
1059 f2fs_put_dnode(&dn);
1060 f2fs_unlock_op(sbi);
1061
1062 new_size = max_t(loff_t, new_size,
1063 (loff_t)(index + 1) << PAGE_CACHE_SHIFT);
1064 }
1065
1066 if (off_end) {
1067 ret = fill_zero(inode, pg_end, 0, off_end);
1068 if (ret)
1069 goto out;
1070
1071 new_size = max_t(loff_t, new_size, offset + len);
1072 }
1073 }
1074
1075 out:
1076 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1077 i_size_write(inode, new_size);
1078 mark_inode_dirty(inode);
1079 update_inode_page(inode);
1080 }
1081
1082 return ret;
1083 }
1084
1085 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1086 {
1087 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1088 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1089 loff_t new_size;
1090 int ret = 0;
1091
1092 new_size = i_size_read(inode) + len;
1093 if (new_size > inode->i_sb->s_maxbytes)
1094 return -EFBIG;
1095
1096 if (offset >= i_size_read(inode))
1097 return -EINVAL;
1098
1099 /* insert range should be aligned to block size of f2fs. */
1100 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1101 return -EINVAL;
1102
1103 ret = f2fs_convert_inline_inode(inode);
1104 if (ret)
1105 return ret;
1106
1107 f2fs_balance_fs(sbi);
1108
1109 ret = truncate_blocks(inode, i_size_read(inode), true);
1110 if (ret)
1111 return ret;
1112
1113 /* write out all dirty pages from offset */
1114 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1115 if (ret)
1116 return ret;
1117
1118 truncate_pagecache(inode, offset);
1119
1120 pg_start = offset >> PAGE_CACHE_SHIFT;
1121 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
1122 delta = pg_end - pg_start;
1123 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1124
1125 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1126 f2fs_lock_op(sbi);
1127 ret = __exchange_data_block(inode, idx, idx + delta, false);
1128 f2fs_unlock_op(sbi);
1129 if (ret)
1130 break;
1131 }
1132
1133 /* write out all moved pages, if possible */
1134 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1135 truncate_pagecache(inode, offset);
1136
1137 if (!ret)
1138 i_size_write(inode, new_size);
1139 return ret;
1140 }
1141
1142 static int expand_inode_data(struct inode *inode, loff_t offset,
1143 loff_t len, int mode)
1144 {
1145 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1146 pgoff_t index, pg_start, pg_end;
1147 loff_t new_size = i_size_read(inode);
1148 loff_t off_start, off_end;
1149 int ret = 0;
1150
1151 ret = inode_newsize_ok(inode, (len + offset));
1152 if (ret)
1153 return ret;
1154
1155 ret = f2fs_convert_inline_inode(inode);
1156 if (ret)
1157 return ret;
1158
1159 f2fs_balance_fs(sbi);
1160
1161 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
1162 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
1163
1164 off_start = offset & (PAGE_CACHE_SIZE - 1);
1165 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
1166
1167 f2fs_lock_op(sbi);
1168
1169 for (index = pg_start; index <= pg_end; index++) {
1170 struct dnode_of_data dn;
1171
1172 if (index == pg_end && !off_end)
1173 goto noalloc;
1174
1175 set_new_dnode(&dn, inode, NULL, NULL, 0);
1176 ret = f2fs_reserve_block(&dn, index);
1177 if (ret)
1178 break;
1179 noalloc:
1180 if (pg_start == pg_end)
1181 new_size = offset + len;
1182 else if (index == pg_start && off_start)
1183 new_size = (loff_t)(index + 1) << PAGE_CACHE_SHIFT;
1184 else if (index == pg_end)
1185 new_size = ((loff_t)index << PAGE_CACHE_SHIFT) +
1186 off_end;
1187 else
1188 new_size += PAGE_CACHE_SIZE;
1189 }
1190
1191 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1192 i_size_read(inode) < new_size) {
1193 i_size_write(inode, new_size);
1194 mark_inode_dirty(inode);
1195 update_inode_page(inode);
1196 }
1197 f2fs_unlock_op(sbi);
1198
1199 return ret;
1200 }
1201
1202 static long f2fs_fallocate(struct file *file, int mode,
1203 loff_t offset, loff_t len)
1204 {
1205 struct inode *inode = file_inode(file);
1206 long ret = 0;
1207
1208 /* f2fs only support ->fallocate for regular file */
1209 if (!S_ISREG(inode->i_mode))
1210 return -EINVAL;
1211
1212 if (f2fs_encrypted_inode(inode) &&
1213 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1214 return -EOPNOTSUPP;
1215
1216 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1217 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1218 FALLOC_FL_INSERT_RANGE))
1219 return -EOPNOTSUPP;
1220
1221 mutex_lock(&inode->i_mutex);
1222
1223 if (mode & FALLOC_FL_PUNCH_HOLE) {
1224 if (offset >= inode->i_size)
1225 goto out;
1226
1227 ret = punch_hole(inode, offset, len);
1228 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1229 ret = f2fs_collapse_range(inode, offset, len);
1230 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1231 ret = f2fs_zero_range(inode, offset, len, mode);
1232 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1233 ret = f2fs_insert_range(inode, offset, len);
1234 } else {
1235 ret = expand_inode_data(inode, offset, len, mode);
1236 }
1237
1238 if (!ret) {
1239 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1240 mark_inode_dirty(inode);
1241 }
1242
1243 out:
1244 mutex_unlock(&inode->i_mutex);
1245
1246 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1247 return ret;
1248 }
1249
1250 static int f2fs_release_file(struct inode *inode, struct file *filp)
1251 {
1252 /* some remained atomic pages should discarded */
1253 if (f2fs_is_atomic_file(inode))
1254 commit_inmem_pages(inode, true);
1255 if (f2fs_is_volatile_file(inode)) {
1256 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1257 filemap_fdatawrite(inode->i_mapping);
1258 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1259 }
1260 return 0;
1261 }
1262
1263 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1264 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1265
1266 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1267 {
1268 if (S_ISDIR(mode))
1269 return flags;
1270 else if (S_ISREG(mode))
1271 return flags & F2FS_REG_FLMASK;
1272 else
1273 return flags & F2FS_OTHER_FLMASK;
1274 }
1275
1276 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1277 {
1278 struct inode *inode = file_inode(filp);
1279 struct f2fs_inode_info *fi = F2FS_I(inode);
1280 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1281 return put_user(flags, (int __user *)arg);
1282 }
1283
1284 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1285 {
1286 struct inode *inode = file_inode(filp);
1287 struct f2fs_inode_info *fi = F2FS_I(inode);
1288 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1289 unsigned int oldflags;
1290 int ret;
1291
1292 ret = mnt_want_write_file(filp);
1293 if (ret)
1294 return ret;
1295
1296 if (!inode_owner_or_capable(inode)) {
1297 ret = -EACCES;
1298 goto out;
1299 }
1300
1301 if (get_user(flags, (int __user *)arg)) {
1302 ret = -EFAULT;
1303 goto out;
1304 }
1305
1306 flags = f2fs_mask_flags(inode->i_mode, flags);
1307
1308 mutex_lock(&inode->i_mutex);
1309
1310 oldflags = fi->i_flags;
1311
1312 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1313 if (!capable(CAP_LINUX_IMMUTABLE)) {
1314 mutex_unlock(&inode->i_mutex);
1315 ret = -EPERM;
1316 goto out;
1317 }
1318 }
1319
1320 flags = flags & FS_FL_USER_MODIFIABLE;
1321 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1322 fi->i_flags = flags;
1323 mutex_unlock(&inode->i_mutex);
1324
1325 f2fs_set_inode_flags(inode);
1326 inode->i_ctime = CURRENT_TIME;
1327 mark_inode_dirty(inode);
1328 out:
1329 mnt_drop_write_file(filp);
1330 return ret;
1331 }
1332
1333 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1334 {
1335 struct inode *inode = file_inode(filp);
1336
1337 return put_user(inode->i_generation, (int __user *)arg);
1338 }
1339
1340 static int f2fs_ioc_start_atomic_write(struct file *filp)
1341 {
1342 struct inode *inode = file_inode(filp);
1343 int ret;
1344
1345 if (!inode_owner_or_capable(inode))
1346 return -EACCES;
1347
1348 if (f2fs_is_atomic_file(inode))
1349 return 0;
1350
1351 ret = f2fs_convert_inline_inode(inode);
1352 if (ret)
1353 return ret;
1354
1355 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1356 return 0;
1357 }
1358
1359 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1360 {
1361 struct inode *inode = file_inode(filp);
1362 int ret;
1363
1364 if (!inode_owner_or_capable(inode))
1365 return -EACCES;
1366
1367 if (f2fs_is_volatile_file(inode))
1368 return 0;
1369
1370 ret = mnt_want_write_file(filp);
1371 if (ret)
1372 return ret;
1373
1374 if (f2fs_is_atomic_file(inode)) {
1375 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1376 ret = commit_inmem_pages(inode, false);
1377 if (ret)
1378 goto err_out;
1379 }
1380
1381 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1382 err_out:
1383 mnt_drop_write_file(filp);
1384 return ret;
1385 }
1386
1387 static int f2fs_ioc_start_volatile_write(struct file *filp)
1388 {
1389 struct inode *inode = file_inode(filp);
1390 int ret;
1391
1392 if (!inode_owner_or_capable(inode))
1393 return -EACCES;
1394
1395 if (f2fs_is_volatile_file(inode))
1396 return 0;
1397
1398 ret = f2fs_convert_inline_inode(inode);
1399 if (ret)
1400 return ret;
1401
1402 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1403 return 0;
1404 }
1405
1406 static int f2fs_ioc_release_volatile_write(struct file *filp)
1407 {
1408 struct inode *inode = file_inode(filp);
1409
1410 if (!inode_owner_or_capable(inode))
1411 return -EACCES;
1412
1413 if (!f2fs_is_volatile_file(inode))
1414 return 0;
1415
1416 if (!f2fs_is_first_block_written(inode))
1417 return truncate_partial_data_page(inode, 0, true);
1418
1419 return punch_hole(inode, 0, F2FS_BLKSIZE);
1420 }
1421
1422 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1423 {
1424 struct inode *inode = file_inode(filp);
1425 int ret;
1426
1427 if (!inode_owner_or_capable(inode))
1428 return -EACCES;
1429
1430 ret = mnt_want_write_file(filp);
1431 if (ret)
1432 return ret;
1433
1434 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1435 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1436 commit_inmem_pages(inode, true);
1437
1438 mnt_drop_write_file(filp);
1439 return ret;
1440 }
1441
1442 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1443 {
1444 struct inode *inode = file_inode(filp);
1445 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1446 struct super_block *sb = sbi->sb;
1447 __u32 in;
1448
1449 if (!capable(CAP_SYS_ADMIN))
1450 return -EPERM;
1451
1452 if (get_user(in, (__u32 __user *)arg))
1453 return -EFAULT;
1454
1455 switch (in) {
1456 case F2FS_GOING_DOWN_FULLSYNC:
1457 sb = freeze_bdev(sb->s_bdev);
1458 if (sb && !IS_ERR(sb)) {
1459 f2fs_stop_checkpoint(sbi);
1460 thaw_bdev(sb->s_bdev, sb);
1461 }
1462 break;
1463 case F2FS_GOING_DOWN_METASYNC:
1464 /* do checkpoint only */
1465 f2fs_sync_fs(sb, 1);
1466 f2fs_stop_checkpoint(sbi);
1467 break;
1468 case F2FS_GOING_DOWN_NOSYNC:
1469 f2fs_stop_checkpoint(sbi);
1470 break;
1471 case F2FS_GOING_DOWN_METAFLUSH:
1472 sync_meta_pages(sbi, META, LONG_MAX);
1473 f2fs_stop_checkpoint(sbi);
1474 break;
1475 default:
1476 return -EINVAL;
1477 }
1478 return 0;
1479 }
1480
1481 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1482 {
1483 struct inode *inode = file_inode(filp);
1484 struct super_block *sb = inode->i_sb;
1485 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1486 struct fstrim_range range;
1487 int ret;
1488
1489 if (!capable(CAP_SYS_ADMIN))
1490 return -EPERM;
1491
1492 if (!blk_queue_discard(q))
1493 return -EOPNOTSUPP;
1494
1495 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1496 sizeof(range)))
1497 return -EFAULT;
1498
1499 range.minlen = max((unsigned int)range.minlen,
1500 q->limits.discard_granularity);
1501 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1502 if (ret < 0)
1503 return ret;
1504
1505 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1506 sizeof(range)))
1507 return -EFAULT;
1508 return 0;
1509 }
1510
1511 static bool uuid_is_nonzero(__u8 u[16])
1512 {
1513 int i;
1514
1515 for (i = 0; i < 16; i++)
1516 if (u[i])
1517 return true;
1518 return false;
1519 }
1520
1521 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1522 {
1523 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1524 struct f2fs_encryption_policy policy;
1525 struct inode *inode = file_inode(filp);
1526
1527 if (copy_from_user(&policy, (struct f2fs_encryption_policy __user *)arg,
1528 sizeof(policy)))
1529 return -EFAULT;
1530
1531 return f2fs_process_policy(&policy, inode);
1532 #else
1533 return -EOPNOTSUPP;
1534 #endif
1535 }
1536
1537 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1538 {
1539 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1540 struct f2fs_encryption_policy policy;
1541 struct inode *inode = file_inode(filp);
1542 int err;
1543
1544 err = f2fs_get_policy(inode, &policy);
1545 if (err)
1546 return err;
1547
1548 if (copy_to_user((struct f2fs_encryption_policy __user *)arg, &policy,
1549 sizeof(policy)))
1550 return -EFAULT;
1551 return 0;
1552 #else
1553 return -EOPNOTSUPP;
1554 #endif
1555 }
1556
1557 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1558 {
1559 struct inode *inode = file_inode(filp);
1560 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1561 int err;
1562
1563 if (!f2fs_sb_has_crypto(inode->i_sb))
1564 return -EOPNOTSUPP;
1565
1566 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1567 goto got_it;
1568
1569 err = mnt_want_write_file(filp);
1570 if (err)
1571 return err;
1572
1573 /* update superblock with uuid */
1574 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1575
1576 err = f2fs_commit_super(sbi, false);
1577 if (err) {
1578 /* undo new data */
1579 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1580 mnt_drop_write_file(filp);
1581 return err;
1582 }
1583 mnt_drop_write_file(filp);
1584 got_it:
1585 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1586 16))
1587 return -EFAULT;
1588 return 0;
1589 }
1590
1591 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1592 {
1593 struct inode *inode = file_inode(filp);
1594 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1595 __u32 sync;
1596
1597 if (!capable(CAP_SYS_ADMIN))
1598 return -EPERM;
1599
1600 if (get_user(sync, (__u32 __user *)arg))
1601 return -EFAULT;
1602
1603 if (f2fs_readonly(sbi->sb))
1604 return -EROFS;
1605
1606 if (!sync) {
1607 if (!mutex_trylock(&sbi->gc_mutex))
1608 return -EBUSY;
1609 } else {
1610 mutex_lock(&sbi->gc_mutex);
1611 }
1612
1613 return f2fs_gc(sbi, sync);
1614 }
1615
1616 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1617 {
1618 struct inode *inode = file_inode(filp);
1619 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1620 struct cp_control cpc;
1621
1622 if (!capable(CAP_SYS_ADMIN))
1623 return -EPERM;
1624
1625 if (f2fs_readonly(sbi->sb))
1626 return -EROFS;
1627
1628 cpc.reason = __get_cp_reason(sbi);
1629
1630 mutex_lock(&sbi->gc_mutex);
1631 write_checkpoint(sbi, &cpc);
1632 mutex_unlock(&sbi->gc_mutex);
1633
1634 return 0;
1635 }
1636
1637 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1638 struct file *filp,
1639 struct f2fs_defragment *range)
1640 {
1641 struct inode *inode = file_inode(filp);
1642 struct f2fs_map_blocks map;
1643 struct extent_info ei;
1644 pgoff_t pg_start, pg_end;
1645 unsigned int blk_per_seg = sbi->blocks_per_seg;
1646 unsigned int total = 0, sec_num;
1647 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1648 block_t blk_end = 0;
1649 bool fragmented = false;
1650 int err;
1651
1652 /* if in-place-update policy is enabled, don't waste time here */
1653 if (need_inplace_update(inode))
1654 return -EINVAL;
1655
1656 pg_start = range->start >> PAGE_CACHE_SHIFT;
1657 pg_end = (range->start + range->len) >> PAGE_CACHE_SHIFT;
1658
1659 f2fs_balance_fs(sbi);
1660
1661 mutex_lock(&inode->i_mutex);
1662
1663 /* writeback all dirty pages in the range */
1664 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1665 range->start + range->len - 1);
1666 if (err)
1667 goto out;
1668
1669 /*
1670 * lookup mapping info in extent cache, skip defragmenting if physical
1671 * block addresses are continuous.
1672 */
1673 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1674 if (ei.fofs + ei.len >= pg_end)
1675 goto out;
1676 }
1677
1678 map.m_lblk = pg_start;
1679
1680 /*
1681 * lookup mapping info in dnode page cache, skip defragmenting if all
1682 * physical block addresses are continuous even if there are hole(s)
1683 * in logical blocks.
1684 */
1685 while (map.m_lblk < pg_end) {
1686 map.m_len = pg_end - map.m_lblk;
1687 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1688 if (err)
1689 goto out;
1690
1691 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1692 map.m_lblk++;
1693 continue;
1694 }
1695
1696 if (blk_end && blk_end != map.m_pblk) {
1697 fragmented = true;
1698 break;
1699 }
1700 blk_end = map.m_pblk + map.m_len;
1701
1702 map.m_lblk += map.m_len;
1703 }
1704
1705 if (!fragmented)
1706 goto out;
1707
1708 map.m_lblk = pg_start;
1709 map.m_len = pg_end - pg_start;
1710
1711 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1712
1713 /*
1714 * make sure there are enough free section for LFS allocation, this can
1715 * avoid defragment running in SSR mode when free section are allocated
1716 * intensively
1717 */
1718 if (has_not_enough_free_secs(sbi, sec_num)) {
1719 err = -EAGAIN;
1720 goto out;
1721 }
1722
1723 while (map.m_lblk < pg_end) {
1724 pgoff_t idx;
1725 int cnt = 0;
1726
1727 do_map:
1728 map.m_len = pg_end - map.m_lblk;
1729 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1730 if (err)
1731 goto clear_out;
1732
1733 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1734 map.m_lblk++;
1735 continue;
1736 }
1737
1738 set_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1739
1740 idx = map.m_lblk;
1741 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1742 struct page *page;
1743
1744 page = get_lock_data_page(inode, idx, true);
1745 if (IS_ERR(page)) {
1746 err = PTR_ERR(page);
1747 goto clear_out;
1748 }
1749
1750 set_page_dirty(page);
1751 f2fs_put_page(page, 1);
1752
1753 idx++;
1754 cnt++;
1755 total++;
1756 }
1757
1758 map.m_lblk = idx;
1759
1760 if (idx < pg_end && cnt < blk_per_seg)
1761 goto do_map;
1762
1763 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1764
1765 err = filemap_fdatawrite(inode->i_mapping);
1766 if (err)
1767 goto out;
1768 }
1769 clear_out:
1770 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1771 out:
1772 mutex_unlock(&inode->i_mutex);
1773 if (!err)
1774 range->len = (u64)total << PAGE_CACHE_SHIFT;
1775 return err;
1776 }
1777
1778 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1779 {
1780 struct inode *inode = file_inode(filp);
1781 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1782 struct f2fs_defragment range;
1783 int err;
1784
1785 if (!capable(CAP_SYS_ADMIN))
1786 return -EPERM;
1787
1788 if (!S_ISREG(inode->i_mode))
1789 return -EINVAL;
1790
1791 err = mnt_want_write_file(filp);
1792 if (err)
1793 return err;
1794
1795 if (f2fs_readonly(sbi->sb)) {
1796 err = -EROFS;
1797 goto out;
1798 }
1799
1800 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1801 sizeof(range))) {
1802 err = -EFAULT;
1803 goto out;
1804 }
1805
1806 /* verify alignment of offset & size */
1807 if (range.start & (F2FS_BLKSIZE - 1) ||
1808 range.len & (F2FS_BLKSIZE - 1)) {
1809 err = -EINVAL;
1810 goto out;
1811 }
1812
1813 err = f2fs_defragment_range(sbi, filp, &range);
1814 if (err < 0)
1815 goto out;
1816
1817 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1818 sizeof(range)))
1819 err = -EFAULT;
1820 out:
1821 mnt_drop_write_file(filp);
1822 return err;
1823 }
1824
1825 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1826 {
1827 switch (cmd) {
1828 case F2FS_IOC_GETFLAGS:
1829 return f2fs_ioc_getflags(filp, arg);
1830 case F2FS_IOC_SETFLAGS:
1831 return f2fs_ioc_setflags(filp, arg);
1832 case F2FS_IOC_GETVERSION:
1833 return f2fs_ioc_getversion(filp, arg);
1834 case F2FS_IOC_START_ATOMIC_WRITE:
1835 return f2fs_ioc_start_atomic_write(filp);
1836 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1837 return f2fs_ioc_commit_atomic_write(filp);
1838 case F2FS_IOC_START_VOLATILE_WRITE:
1839 return f2fs_ioc_start_volatile_write(filp);
1840 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1841 return f2fs_ioc_release_volatile_write(filp);
1842 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1843 return f2fs_ioc_abort_volatile_write(filp);
1844 case F2FS_IOC_SHUTDOWN:
1845 return f2fs_ioc_shutdown(filp, arg);
1846 case FITRIM:
1847 return f2fs_ioc_fitrim(filp, arg);
1848 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1849 return f2fs_ioc_set_encryption_policy(filp, arg);
1850 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1851 return f2fs_ioc_get_encryption_policy(filp, arg);
1852 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1853 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1854 case F2FS_IOC_GARBAGE_COLLECT:
1855 return f2fs_ioc_gc(filp, arg);
1856 case F2FS_IOC_WRITE_CHECKPOINT:
1857 return f2fs_ioc_write_checkpoint(filp, arg);
1858 case F2FS_IOC_DEFRAGMENT:
1859 return f2fs_ioc_defragment(filp, arg);
1860 default:
1861 return -ENOTTY;
1862 }
1863 }
1864
1865 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1866 {
1867 struct inode *inode = file_inode(iocb->ki_filp);
1868
1869 if (f2fs_encrypted_inode(inode) &&
1870 !f2fs_has_encryption_key(inode) &&
1871 f2fs_get_encryption_info(inode))
1872 return -EACCES;
1873
1874 return generic_file_write_iter(iocb, from);
1875 }
1876
1877 #ifdef CONFIG_COMPAT
1878 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1879 {
1880 switch (cmd) {
1881 case F2FS_IOC32_GETFLAGS:
1882 cmd = F2FS_IOC_GETFLAGS;
1883 break;
1884 case F2FS_IOC32_SETFLAGS:
1885 cmd = F2FS_IOC_SETFLAGS;
1886 break;
1887 case F2FS_IOC32_GETVERSION:
1888 cmd = F2FS_IOC_GETVERSION;
1889 break;
1890 case F2FS_IOC_START_ATOMIC_WRITE:
1891 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1892 case F2FS_IOC_START_VOLATILE_WRITE:
1893 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1894 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1895 case F2FS_IOC_SHUTDOWN:
1896 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1897 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1898 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1899 case F2FS_IOC_GARBAGE_COLLECT:
1900 case F2FS_IOC_WRITE_CHECKPOINT:
1901 case F2FS_IOC_DEFRAGMENT:
1902 break;
1903 default:
1904 return -ENOIOCTLCMD;
1905 }
1906 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1907 }
1908 #endif
1909
1910 const struct file_operations f2fs_file_operations = {
1911 .llseek = f2fs_llseek,
1912 .read_iter = generic_file_read_iter,
1913 .write_iter = f2fs_file_write_iter,
1914 .open = f2fs_file_open,
1915 .release = f2fs_release_file,
1916 .mmap = f2fs_file_mmap,
1917 .fsync = f2fs_sync_file,
1918 .fallocate = f2fs_fallocate,
1919 .unlocked_ioctl = f2fs_ioctl,
1920 #ifdef CONFIG_COMPAT
1921 .compat_ioctl = f2fs_compat_ioctl,
1922 #endif
1923 .splice_read = generic_file_splice_read,
1924 .splice_write = iter_file_splice_write,
1925 };
ubuntu-bionic-kernel mirror