projects / GitHub / moto-9609 / android_kernel_motorola_exynos9610.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge 4.14.31 into android-4.14
[GitHub/moto-9609/android_kernel_motorola_exynos9610.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/uio.h>
24 #include <linux/uuid.h>
25 #include <linux/file.h>
26
27 #include "f2fs.h"
28 #include "node.h"
29 #include "segment.h"
30 #include "xattr.h"
31 #include "acl.h"
32 #include "gc.h"
33 #include "trace.h"
34 #include <trace/events/f2fs.h>
35
36 static int f2fs_filemap_fault(struct vm_fault *vmf)
37 {
38 struct inode *inode = file_inode(vmf->vma->vm_file);
39 int err;
40
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 err = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
44
45 return err;
46 }
47
48 static int f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 {
50 struct page *page = vmf->page;
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 struct dnode_of_data dn;
54 int err;
55
56 if (unlikely(f2fs_cp_error(sbi))) {
57 err = -EIO;
58 goto err;
59 }
60
61 sb_start_pagefault(inode->i_sb);
62
63 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
64
65 /* block allocation */
66 f2fs_lock_op(sbi);
67 set_new_dnode(&dn, inode, NULL, NULL, 0);
68 err = f2fs_reserve_block(&dn, page->index);
69 if (err) {
70 f2fs_unlock_op(sbi);
71 goto out;
72 }
73 f2fs_put_dnode(&dn);
74 f2fs_unlock_op(sbi);
75
76 f2fs_balance_fs(sbi, dn.node_changed);
77
78 file_update_time(vmf->vma->vm_file);
79 down_read(&F2FS_I(inode)->i_mmap_sem);
80 lock_page(page);
81 if (unlikely(page->mapping != inode->i_mapping ||
82 page_offset(page) > i_size_read(inode) ||
83 !PageUptodate(page))) {
84 unlock_page(page);
85 err = -EFAULT;
86 goto out_sem;
87 }
88
89 /*
90 * check to see if the page is mapped already (no holes)
91 */
92 if (PageMappedToDisk(page))
93 goto mapped;
94
95 /* page is wholly or partially inside EOF */
96 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
97 i_size_read(inode)) {
98 unsigned offset;
99 offset = i_size_read(inode) & ~PAGE_MASK;
100 zero_user_segment(page, offset, PAGE_SIZE);
101 }
102 set_page_dirty(page);
103 if (!PageUptodate(page))
104 SetPageUptodate(page);
105
106 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
107
108 trace_f2fs_vm_page_mkwrite(page, DATA);
109 mapped:
110 /* fill the page */
111 f2fs_wait_on_page_writeback(page, DATA, false);
112
113 /* wait for GCed encrypted page writeback */
114 if (f2fs_encrypted_file(inode))
115 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
116
117 out_sem:
118 up_read(&F2FS_I(inode)->i_mmap_sem);
119 out:
120 sb_end_pagefault(inode->i_sb);
121 f2fs_update_time(sbi, REQ_TIME);
122 err:
123 return block_page_mkwrite_return(err);
124 }
125
126 static const struct vm_operations_struct f2fs_file_vm_ops = {
127 .fault = f2fs_filemap_fault,
128 .map_pages = filemap_map_pages,
129 .page_mkwrite = f2fs_vm_page_mkwrite,
130 };
131
132 static int get_parent_ino(struct inode *inode, nid_t *pino)
133 {
134 struct dentry *dentry;
135
136 inode = igrab(inode);
137 dentry = d_find_any_alias(inode);
138 iput(inode);
139 if (!dentry)
140 return 0;
141
142 *pino = parent_ino(dentry);
143 dput(dentry);
144 return 1;
145 }
146
147 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
148 {
149 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
150 enum cp_reason_type cp_reason = CP_NO_NEEDED;
151
152 if (!S_ISREG(inode->i_mode))
153 cp_reason = CP_NON_REGULAR;
154 else if (inode->i_nlink != 1)
155 cp_reason = CP_HARDLINK;
156 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
157 cp_reason = CP_SB_NEED_CP;
158 else if (file_wrong_pino(inode))
159 cp_reason = CP_WRONG_PINO;
160 else if (!space_for_roll_forward(sbi))
161 cp_reason = CP_NO_SPC_ROLL;
162 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
163 cp_reason = CP_NODE_NEED_CP;
164 else if (test_opt(sbi, FASTBOOT))
165 cp_reason = CP_FASTBOOT_MODE;
166 else if (sbi->active_logs == 2)
167 cp_reason = CP_SPEC_LOG_NUM;
168 else if (need_dentry_mark(sbi, inode->i_ino) &&
169 exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
170 cp_reason = CP_RECOVER_DIR;
171
172 return cp_reason;
173 }
174
175 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
176 {
177 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
178 bool ret = false;
179 /* But we need to avoid that there are some inode updates */
180 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
181 ret = true;
182 f2fs_put_page(i, 0);
183 return ret;
184 }
185
186 static void try_to_fix_pino(struct inode *inode)
187 {
188 struct f2fs_inode_info *fi = F2FS_I(inode);
189 nid_t pino;
190
191 down_write(&fi->i_sem);
192 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
193 get_parent_ino(inode, &pino)) {
194 f2fs_i_pino_write(inode, pino);
195 file_got_pino(inode);
196 }
197 up_write(&fi->i_sem);
198 }
199
200 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
201 int datasync, bool atomic)
202 {
203 struct inode *inode = file->f_mapping->host;
204 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
205 nid_t ino = inode->i_ino;
206 int ret = 0;
207 enum cp_reason_type cp_reason = 0;
208 struct writeback_control wbc = {
209 .sync_mode = WB_SYNC_ALL,
210 .nr_to_write = LONG_MAX,
211 .for_reclaim = 0,
212 };
213
214 if (unlikely(f2fs_readonly(inode->i_sb)))
215 return 0;
216
217 trace_f2fs_sync_file_enter(inode);
218
219 /* if fdatasync is triggered, let's do in-place-update */
220 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
221 set_inode_flag(inode, FI_NEED_IPU);
222 ret = file_write_and_wait_range(file, start, end);
223 clear_inode_flag(inode, FI_NEED_IPU);
224
225 if (ret) {
226 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
227 return ret;
228 }
229
230 /* if the inode is dirty, let's recover all the time */
231 if (!f2fs_skip_inode_update(inode, datasync)) {
232 f2fs_write_inode(inode, NULL);
233 goto go_write;
234 }
235
236 /*
237 * if there is no written data, don't waste time to write recovery info.
238 */
239 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
240 !exist_written_data(sbi, ino, APPEND_INO)) {
241
242 /* it may call write_inode just prior to fsync */
243 if (need_inode_page_update(sbi, ino))
244 goto go_write;
245
246 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
247 exist_written_data(sbi, ino, UPDATE_INO))
248 goto flush_out;
249 goto out;
250 }
251 go_write:
252 /*
253 * Both of fdatasync() and fsync() are able to be recovered from
254 * sudden-power-off.
255 */
256 down_read(&F2FS_I(inode)->i_sem);
257 cp_reason = need_do_checkpoint(inode);
258 up_read(&F2FS_I(inode)->i_sem);
259
260 if (cp_reason) {
261 /* all the dirty node pages should be flushed for POR */
262 ret = f2fs_sync_fs(inode->i_sb, 1);
263
264 /*
265 * We've secured consistency through sync_fs. Following pino
266 * will be used only for fsynced inodes after checkpoint.
267 */
268 try_to_fix_pino(inode);
269 clear_inode_flag(inode, FI_APPEND_WRITE);
270 clear_inode_flag(inode, FI_UPDATE_WRITE);
271 goto out;
272 }
273 sync_nodes:
274 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
275 if (ret)
276 goto out;
277
278 /* if cp_error was enabled, we should avoid infinite loop */
279 if (unlikely(f2fs_cp_error(sbi))) {
280 ret = -EIO;
281 goto out;
282 }
283
284 if (need_inode_block_update(sbi, ino)) {
285 f2fs_mark_inode_dirty_sync(inode, true);
286 f2fs_write_inode(inode, NULL);
287 goto sync_nodes;
288 }
289
290 /*
291 * If it's atomic_write, it's just fine to keep write ordering. So
292 * here we don't need to wait for node write completion, since we use
293 * node chain which serializes node blocks. If one of node writes are
294 * reordered, we can see simply broken chain, resulting in stopping
295 * roll-forward recovery. It means we'll recover all or none node blocks
296 * given fsync mark.
297 */
298 if (!atomic) {
299 ret = wait_on_node_pages_writeback(sbi, ino);
300 if (ret)
301 goto out;
302 }
303
304 /* once recovery info is written, don't need to tack this */
305 remove_ino_entry(sbi, ino, APPEND_INO);
306 clear_inode_flag(inode, FI_APPEND_WRITE);
307 flush_out:
308 if (!atomic)
309 ret = f2fs_issue_flush(sbi, inode->i_ino);
310 if (!ret) {
311 remove_ino_entry(sbi, ino, UPDATE_INO);
312 clear_inode_flag(inode, FI_UPDATE_WRITE);
313 remove_ino_entry(sbi, ino, FLUSH_INO);
314 }
315 f2fs_update_time(sbi, REQ_TIME);
316 out:
317 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
318 f2fs_trace_ios(NULL, 1);
319 return ret;
320 }
321
322 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
323 {
324 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
325 return -EIO;
326 return f2fs_do_sync_file(file, start, end, datasync, false);
327 }
328
329 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
330 pgoff_t pgofs, int whence)
331 {
332 struct pagevec pvec;
333 int nr_pages;
334
335 if (whence != SEEK_DATA)
336 return 0;
337
338 /* find first dirty page index */
339 pagevec_init(&pvec, 0);
340 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
341 PAGECACHE_TAG_DIRTY, 1);
342 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
343 pagevec_release(&pvec);
344 return pgofs;
345 }
346
347 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
348 int whence)
349 {
350 switch (whence) {
351 case SEEK_DATA:
352 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
353 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
354 return true;
355 break;
356 case SEEK_HOLE:
357 if (blkaddr == NULL_ADDR)
358 return true;
359 break;
360 }
361 return false;
362 }
363
364 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
365 {
366 struct inode *inode = file->f_mapping->host;
367 loff_t maxbytes = inode->i_sb->s_maxbytes;
368 struct dnode_of_data dn;
369 pgoff_t pgofs, end_offset, dirty;
370 loff_t data_ofs = offset;
371 loff_t isize;
372 int err = 0;
373
374 inode_lock(inode);
375
376 isize = i_size_read(inode);
377 if (offset >= isize)
378 goto fail;
379
380 /* handle inline data case */
381 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
382 if (whence == SEEK_HOLE)
383 data_ofs = isize;
384 goto found;
385 }
386
387 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
388
389 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
390
391 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
392 set_new_dnode(&dn, inode, NULL, NULL, 0);
393 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
394 if (err && err != -ENOENT) {
395 goto fail;
396 } else if (err == -ENOENT) {
397 /* direct node does not exists */
398 if (whence == SEEK_DATA) {
399 pgofs = get_next_page_offset(&dn, pgofs);
400 continue;
401 } else {
402 goto found;
403 }
404 }
405
406 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
407
408 /* find data/hole in dnode block */
409 for (; dn.ofs_in_node < end_offset;
410 dn.ofs_in_node++, pgofs++,
411 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
412 block_t blkaddr;
413 blkaddr = datablock_addr(dn.inode,
414 dn.node_page, dn.ofs_in_node);
415
416 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
417 f2fs_put_dnode(&dn);
418 goto found;
419 }
420 }
421 f2fs_put_dnode(&dn);
422 }
423
424 if (whence == SEEK_DATA)
425 goto fail;
426 found:
427 if (whence == SEEK_HOLE && data_ofs > isize)
428 data_ofs = isize;
429 inode_unlock(inode);
430 return vfs_setpos(file, data_ofs, maxbytes);
431 fail:
432 inode_unlock(inode);
433 return -ENXIO;
434 }
435
436 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
437 {
438 struct inode *inode = file->f_mapping->host;
439 loff_t maxbytes = inode->i_sb->s_maxbytes;
440
441 switch (whence) {
442 case SEEK_SET:
443 case SEEK_CUR:
444 case SEEK_END:
445 return generic_file_llseek_size(file, offset, whence,
446 maxbytes, i_size_read(inode));
447 case SEEK_DATA:
448 case SEEK_HOLE:
449 if (offset < 0)
450 return -ENXIO;
451 return f2fs_seek_block(file, offset, whence);
452 }
453
454 return -EINVAL;
455 }
456
457 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
458 {
459 struct inode *inode = file_inode(file);
460 int err;
461
462 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
463 return -EIO;
464
465 /* we don't need to use inline_data strictly */
466 err = f2fs_convert_inline_inode(inode);
467 if (err)
468 return err;
469
470 file_accessed(file);
471 vma->vm_ops = &f2fs_file_vm_ops;
472 return 0;
473 }
474
475 static int f2fs_file_open(struct inode *inode, struct file *filp)
476 {
477 int err = fscrypt_file_open(inode, filp);
478
479 if (err)
480 return err;
481 return dquot_file_open(inode, filp);
482 }
483
484 void truncate_data_blocks_range(struct dnode_of_data *dn, int count)
485 {
486 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
487 struct f2fs_node *raw_node;
488 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
489 __le32 *addr;
490 int base = 0;
491
492 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
493 base = get_extra_isize(dn->inode);
494
495 raw_node = F2FS_NODE(dn->node_page);
496 addr = blkaddr_in_node(raw_node) + base + ofs;
497
498 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
499 block_t blkaddr = le32_to_cpu(*addr);
500 if (blkaddr == NULL_ADDR)
501 continue;
502
503 dn->data_blkaddr = NULL_ADDR;
504 set_data_blkaddr(dn);
505 invalidate_blocks(sbi, blkaddr);
506 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
507 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
508 nr_free++;
509 }
510
511 if (nr_free) {
512 pgoff_t fofs;
513 /*
514 * once we invalidate valid blkaddr in range [ofs, ofs + count],
515 * we will invalidate all blkaddr in the whole range.
516 */
517 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
518 dn->inode) + ofs;
519 f2fs_update_extent_cache_range(dn, fofs, 0, len);
520 dec_valid_block_count(sbi, dn->inode, nr_free);
521 }
522 dn->ofs_in_node = ofs;
523
524 f2fs_update_time(sbi, REQ_TIME);
525 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
526 dn->ofs_in_node, nr_free);
527 }
528
529 void truncate_data_blocks(struct dnode_of_data *dn)
530 {
531 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
532 }
533
534 static int truncate_partial_data_page(struct inode *inode, u64 from,
535 bool cache_only)
536 {
537 unsigned offset = from & (PAGE_SIZE - 1);
538 pgoff_t index = from >> PAGE_SHIFT;
539 struct address_space *mapping = inode->i_mapping;
540 struct page *page;
541
542 if (!offset && !cache_only)
543 return 0;
544
545 if (cache_only) {
546 page = find_lock_page(mapping, index);
547 if (page && PageUptodate(page))
548 goto truncate_out;
549 f2fs_put_page(page, 1);
550 return 0;
551 }
552
553 page = get_lock_data_page(inode, index, true);
554 if (IS_ERR(page))
555 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
556 truncate_out:
557 f2fs_wait_on_page_writeback(page, DATA, true);
558 zero_user(page, offset, PAGE_SIZE - offset);
559
560 /* An encrypted inode should have a key and truncate the last page. */
561 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
562 if (!cache_only)
563 set_page_dirty(page);
564 f2fs_put_page(page, 1);
565 return 0;
566 }
567
568 int truncate_blocks(struct inode *inode, u64 from, bool lock)
569 {
570 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
571 unsigned int blocksize = inode->i_sb->s_blocksize;
572 struct dnode_of_data dn;
573 pgoff_t free_from;
574 int count = 0, err = 0;
575 struct page *ipage;
576 bool truncate_page = false;
577
578 trace_f2fs_truncate_blocks_enter(inode, from);
579
580 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
581
582 if (free_from >= sbi->max_file_blocks)
583 goto free_partial;
584
585 if (lock)
586 f2fs_lock_op(sbi);
587
588 ipage = get_node_page(sbi, inode->i_ino);
589 if (IS_ERR(ipage)) {
590 err = PTR_ERR(ipage);
591 goto out;
592 }
593
594 if (f2fs_has_inline_data(inode)) {
595 truncate_inline_inode(inode, ipage, from);
596 f2fs_put_page(ipage, 1);
597 truncate_page = true;
598 goto out;
599 }
600
601 set_new_dnode(&dn, inode, ipage, NULL, 0);
602 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
603 if (err) {
604 if (err == -ENOENT)
605 goto free_next;
606 goto out;
607 }
608
609 count = ADDRS_PER_PAGE(dn.node_page, inode);
610
611 count -= dn.ofs_in_node;
612 f2fs_bug_on(sbi, count < 0);
613
614 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
615 truncate_data_blocks_range(&dn, count);
616 free_from += count;
617 }
618
619 f2fs_put_dnode(&dn);
620 free_next:
621 err = truncate_inode_blocks(inode, free_from);
622 out:
623 if (lock)
624 f2fs_unlock_op(sbi);
625 free_partial:
626 /* lastly zero out the first data page */
627 if (!err)
628 err = truncate_partial_data_page(inode, from, truncate_page);
629
630 trace_f2fs_truncate_blocks_exit(inode, err);
631 return err;
632 }
633
634 int f2fs_truncate(struct inode *inode)
635 {
636 int err;
637
638 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
639 return -EIO;
640
641 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
642 S_ISLNK(inode->i_mode)))
643 return 0;
644
645 trace_f2fs_truncate(inode);
646
647 #ifdef CONFIG_F2FS_FAULT_INJECTION
648 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
649 f2fs_show_injection_info(FAULT_TRUNCATE);
650 return -EIO;
651 }
652 #endif
653 /* we should check inline_data size */
654 if (!f2fs_may_inline_data(inode)) {
655 err = f2fs_convert_inline_inode(inode);
656 if (err)
657 return err;
658 }
659
660 err = truncate_blocks(inode, i_size_read(inode), true);
661 if (err)
662 return err;
663
664 inode->i_mtime = inode->i_ctime = current_time(inode);
665 f2fs_mark_inode_dirty_sync(inode, false);
666 return 0;
667 }
668
669 int f2fs_getattr(const struct path *path, struct kstat *stat,
670 u32 request_mask, unsigned int query_flags)
671 {
672 struct inode *inode = d_inode(path->dentry);
673 struct f2fs_inode_info *fi = F2FS_I(inode);
674 struct f2fs_inode *ri;
675 unsigned int flags;
676
677 if (f2fs_has_extra_attr(inode) &&
678 f2fs_sb_has_inode_crtime(inode->i_sb) &&
679 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
680 stat->result_mask |= STATX_BTIME;
681 stat->btime.tv_sec = fi->i_crtime.tv_sec;
682 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
683 }
684
685 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
686 if (flags & FS_APPEND_FL)
687 stat->attributes |= STATX_ATTR_APPEND;
688 if (flags & FS_COMPR_FL)
689 stat->attributes |= STATX_ATTR_COMPRESSED;
690 if (f2fs_encrypted_inode(inode))
691 stat->attributes |= STATX_ATTR_ENCRYPTED;
692 if (flags & FS_IMMUTABLE_FL)
693 stat->attributes |= STATX_ATTR_IMMUTABLE;
694 if (flags & FS_NODUMP_FL)
695 stat->attributes |= STATX_ATTR_NODUMP;
696
697 stat->attributes_mask |= (STATX_ATTR_APPEND |
698 STATX_ATTR_COMPRESSED |
699 STATX_ATTR_ENCRYPTED |
700 STATX_ATTR_IMMUTABLE |
701 STATX_ATTR_NODUMP);
702
703 generic_fillattr(inode, stat);
704
705 /* we need to show initial sectors used for inline_data/dentries */
706 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
707 f2fs_has_inline_dentry(inode))
708 stat->blocks += (stat->size + 511) >> 9;
709
710 return 0;
711 }
712
713 #ifdef CONFIG_F2FS_FS_POSIX_ACL
714 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
715 {
716 unsigned int ia_valid = attr->ia_valid;
717
718 if (ia_valid & ATTR_UID)
719 inode->i_uid = attr->ia_uid;
720 if (ia_valid & ATTR_GID)
721 inode->i_gid = attr->ia_gid;
722 if (ia_valid & ATTR_ATIME)
723 inode->i_atime = timespec_trunc(attr->ia_atime,
724 inode->i_sb->s_time_gran);
725 if (ia_valid & ATTR_MTIME)
726 inode->i_mtime = timespec_trunc(attr->ia_mtime,
727 inode->i_sb->s_time_gran);
728 if (ia_valid & ATTR_CTIME)
729 inode->i_ctime = timespec_trunc(attr->ia_ctime,
730 inode->i_sb->s_time_gran);
731 if (ia_valid & ATTR_MODE) {
732 umode_t mode = attr->ia_mode;
733
734 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
735 mode &= ~S_ISGID;
736 set_acl_inode(inode, mode);
737 }
738 }
739 #else
740 #define __setattr_copy setattr_copy
741 #endif
742
743 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
744 {
745 struct inode *inode = d_inode(dentry);
746 int err;
747 bool size_changed = false;
748
749 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
750 return -EIO;
751
752 err = setattr_prepare(dentry, attr);
753 if (err)
754 return err;
755
756 err = fscrypt_prepare_setattr(dentry, attr);
757 if (err)
758 return err;
759
760 if (is_quota_modification(inode, attr)) {
761 err = dquot_initialize(inode);
762 if (err)
763 return err;
764 }
765 if ((attr->ia_valid & ATTR_UID &&
766 !uid_eq(attr->ia_uid, inode->i_uid)) ||
767 (attr->ia_valid & ATTR_GID &&
768 !gid_eq(attr->ia_gid, inode->i_gid))) {
769 err = dquot_transfer(inode, attr);
770 if (err)
771 return err;
772 }
773
774 if (attr->ia_valid & ATTR_SIZE) {
775 if (attr->ia_size <= i_size_read(inode)) {
776 down_write(&F2FS_I(inode)->i_mmap_sem);
777 truncate_setsize(inode, attr->ia_size);
778 err = f2fs_truncate(inode);
779 up_write(&F2FS_I(inode)->i_mmap_sem);
780 if (err)
781 return err;
782 } else {
783 /*
784 * do not trim all blocks after i_size if target size is
785 * larger than i_size.
786 */
787 down_write(&F2FS_I(inode)->i_mmap_sem);
788 truncate_setsize(inode, attr->ia_size);
789 up_write(&F2FS_I(inode)->i_mmap_sem);
790
791 /* should convert inline inode here */
792 if (!f2fs_may_inline_data(inode)) {
793 err = f2fs_convert_inline_inode(inode);
794 if (err)
795 return err;
796 }
797 inode->i_mtime = inode->i_ctime = current_time(inode);
798 }
799
800 down_write(&F2FS_I(inode)->i_sem);
801 F2FS_I(inode)->last_disk_size = i_size_read(inode);
802 up_write(&F2FS_I(inode)->i_sem);
803
804 size_changed = true;
805 }
806
807 __setattr_copy(inode, attr);
808
809 if (attr->ia_valid & ATTR_MODE) {
810 err = posix_acl_chmod(inode, get_inode_mode(inode));
811 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
812 inode->i_mode = F2FS_I(inode)->i_acl_mode;
813 clear_inode_flag(inode, FI_ACL_MODE);
814 }
815 }
816
817 /* file size may changed here */
818 f2fs_mark_inode_dirty_sync(inode, size_changed);
819
820 /* inode change will produce dirty node pages flushed by checkpoint */
821 f2fs_balance_fs(F2FS_I_SB(inode), true);
822
823 return err;
824 }
825
826 const struct inode_operations f2fs_file_inode_operations = {
827 .getattr = f2fs_getattr,
828 .setattr = f2fs_setattr,
829 .get_acl = f2fs_get_acl,
830 .set_acl = f2fs_set_acl,
831 #ifdef CONFIG_F2FS_FS_XATTR
832 .listxattr = f2fs_listxattr,
833 #endif
834 .fiemap = f2fs_fiemap,
835 };
836
837 static int fill_zero(struct inode *inode, pgoff_t index,
838 loff_t start, loff_t len)
839 {
840 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
841 struct page *page;
842
843 if (!len)
844 return 0;
845
846 f2fs_balance_fs(sbi, true);
847
848 f2fs_lock_op(sbi);
849 page = get_new_data_page(inode, NULL, index, false);
850 f2fs_unlock_op(sbi);
851
852 if (IS_ERR(page))
853 return PTR_ERR(page);
854
855 f2fs_wait_on_page_writeback(page, DATA, true);
856 zero_user(page, start, len);
857 set_page_dirty(page);
858 f2fs_put_page(page, 1);
859 return 0;
860 }
861
862 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
863 {
864 int err;
865
866 while (pg_start < pg_end) {
867 struct dnode_of_data dn;
868 pgoff_t end_offset, count;
869
870 set_new_dnode(&dn, inode, NULL, NULL, 0);
871 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
872 if (err) {
873 if (err == -ENOENT) {
874 pg_start = get_next_page_offset(&dn, pg_start);
875 continue;
876 }
877 return err;
878 }
879
880 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
881 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
882
883 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
884
885 truncate_data_blocks_range(&dn, count);
886 f2fs_put_dnode(&dn);
887
888 pg_start += count;
889 }
890 return 0;
891 }
892
893 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
894 {
895 pgoff_t pg_start, pg_end;
896 loff_t off_start, off_end;
897 int ret;
898
899 ret = f2fs_convert_inline_inode(inode);
900 if (ret)
901 return ret;
902
903 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
904 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
905
906 off_start = offset & (PAGE_SIZE - 1);
907 off_end = (offset + len) & (PAGE_SIZE - 1);
908
909 if (pg_start == pg_end) {
910 ret = fill_zero(inode, pg_start, off_start,
911 off_end - off_start);
912 if (ret)
913 return ret;
914 } else {
915 if (off_start) {
916 ret = fill_zero(inode, pg_start++, off_start,
917 PAGE_SIZE - off_start);
918 if (ret)
919 return ret;
920 }
921 if (off_end) {
922 ret = fill_zero(inode, pg_end, 0, off_end);
923 if (ret)
924 return ret;
925 }
926
927 if (pg_start < pg_end) {
928 struct address_space *mapping = inode->i_mapping;
929 loff_t blk_start, blk_end;
930 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
931
932 f2fs_balance_fs(sbi, true);
933
934 blk_start = (loff_t)pg_start << PAGE_SHIFT;
935 blk_end = (loff_t)pg_end << PAGE_SHIFT;
936 down_write(&F2FS_I(inode)->i_mmap_sem);
937 truncate_inode_pages_range(mapping, blk_start,
938 blk_end - 1);
939
940 f2fs_lock_op(sbi);
941 ret = truncate_hole(inode, pg_start, pg_end);
942 f2fs_unlock_op(sbi);
943 up_write(&F2FS_I(inode)->i_mmap_sem);
944 }
945 }
946
947 return ret;
948 }
949
950 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
951 int *do_replace, pgoff_t off, pgoff_t len)
952 {
953 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
954 struct dnode_of_data dn;
955 int ret, done, i;
956
957 next_dnode:
958 set_new_dnode(&dn, inode, NULL, NULL, 0);
959 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
960 if (ret && ret != -ENOENT) {
961 return ret;
962 } else if (ret == -ENOENT) {
963 if (dn.max_level == 0)
964 return -ENOENT;
965 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
966 blkaddr += done;
967 do_replace += done;
968 goto next;
969 }
970
971 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
972 dn.ofs_in_node, len);
973 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
974 *blkaddr = datablock_addr(dn.inode,
975 dn.node_page, dn.ofs_in_node);
976 if (!is_checkpointed_data(sbi, *blkaddr)) {
977
978 if (test_opt(sbi, LFS)) {
979 f2fs_put_dnode(&dn);
980 return -ENOTSUPP;
981 }
982
983 /* do not invalidate this block address */
984 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
985 *do_replace = 1;
986 }
987 }
988 f2fs_put_dnode(&dn);
989 next:
990 len -= done;
991 off += done;
992 if (len)
993 goto next_dnode;
994 return 0;
995 }
996
997 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
998 int *do_replace, pgoff_t off, int len)
999 {
1000 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1001 struct dnode_of_data dn;
1002 int ret, i;
1003
1004 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1005 if (*do_replace == 0)
1006 continue;
1007
1008 set_new_dnode(&dn, inode, NULL, NULL, 0);
1009 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1010 if (ret) {
1011 dec_valid_block_count(sbi, inode, 1);
1012 invalidate_blocks(sbi, *blkaddr);
1013 } else {
1014 f2fs_update_data_blkaddr(&dn, *blkaddr);
1015 }
1016 f2fs_put_dnode(&dn);
1017 }
1018 return 0;
1019 }
1020
1021 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1022 block_t *blkaddr, int *do_replace,
1023 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1024 {
1025 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1026 pgoff_t i = 0;
1027 int ret;
1028
1029 while (i < len) {
1030 if (blkaddr[i] == NULL_ADDR && !full) {
1031 i++;
1032 continue;
1033 }
1034
1035 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1036 struct dnode_of_data dn;
1037 struct node_info ni;
1038 size_t new_size;
1039 pgoff_t ilen;
1040
1041 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1042 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1043 if (ret)
1044 return ret;
1045
1046 get_node_info(sbi, dn.nid, &ni);
1047 ilen = min((pgoff_t)
1048 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1049 dn.ofs_in_node, len - i);
1050 do {
1051 dn.data_blkaddr = datablock_addr(dn.inode,
1052 dn.node_page, dn.ofs_in_node);
1053 truncate_data_blocks_range(&dn, 1);
1054
1055 if (do_replace[i]) {
1056 f2fs_i_blocks_write(src_inode,
1057 1, false, false);
1058 f2fs_i_blocks_write(dst_inode,
1059 1, true, false);
1060 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1061 blkaddr[i], ni.version, true, false);
1062
1063 do_replace[i] = 0;
1064 }
1065 dn.ofs_in_node++;
1066 i++;
1067 new_size = (dst + i) << PAGE_SHIFT;
1068 if (dst_inode->i_size < new_size)
1069 f2fs_i_size_write(dst_inode, new_size);
1070 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1071
1072 f2fs_put_dnode(&dn);
1073 } else {
1074 struct page *psrc, *pdst;
1075
1076 psrc = get_lock_data_page(src_inode, src + i, true);
1077 if (IS_ERR(psrc))
1078 return PTR_ERR(psrc);
1079 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1080 true);
1081 if (IS_ERR(pdst)) {
1082 f2fs_put_page(psrc, 1);
1083 return PTR_ERR(pdst);
1084 }
1085 f2fs_copy_page(psrc, pdst);
1086 set_page_dirty(pdst);
1087 f2fs_put_page(pdst, 1);
1088 f2fs_put_page(psrc, 1);
1089
1090 ret = truncate_hole(src_inode, src + i, src + i + 1);
1091 if (ret)
1092 return ret;
1093 i++;
1094 }
1095 }
1096 return 0;
1097 }
1098
1099 static int __exchange_data_block(struct inode *src_inode,
1100 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1101 pgoff_t len, bool full)
1102 {
1103 block_t *src_blkaddr;
1104 int *do_replace;
1105 pgoff_t olen;
1106 int ret;
1107
1108 while (len) {
1109 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1110
1111 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1112 sizeof(block_t) * olen, GFP_KERNEL);
1113 if (!src_blkaddr)
1114 return -ENOMEM;
1115
1116 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1117 sizeof(int) * olen, GFP_KERNEL);
1118 if (!do_replace) {
1119 kvfree(src_blkaddr);
1120 return -ENOMEM;
1121 }
1122
1123 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1124 do_replace, src, olen);
1125 if (ret)
1126 goto roll_back;
1127
1128 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1129 do_replace, src, dst, olen, full);
1130 if (ret)
1131 goto roll_back;
1132
1133 src += olen;
1134 dst += olen;
1135 len -= olen;
1136
1137 kvfree(src_blkaddr);
1138 kvfree(do_replace);
1139 }
1140 return 0;
1141
1142 roll_back:
1143 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1144 kvfree(src_blkaddr);
1145 kvfree(do_replace);
1146 return ret;
1147 }
1148
1149 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1150 {
1151 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1152 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1153 int ret;
1154
1155 f2fs_balance_fs(sbi, true);
1156 f2fs_lock_op(sbi);
1157
1158 f2fs_drop_extent_tree(inode);
1159
1160 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1161 f2fs_unlock_op(sbi);
1162 return ret;
1163 }
1164
1165 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1166 {
1167 pgoff_t pg_start, pg_end;
1168 loff_t new_size;
1169 int ret;
1170
1171 if (offset + len >= i_size_read(inode))
1172 return -EINVAL;
1173
1174 /* collapse range should be aligned to block size of f2fs. */
1175 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1176 return -EINVAL;
1177
1178 ret = f2fs_convert_inline_inode(inode);
1179 if (ret)
1180 return ret;
1181
1182 pg_start = offset >> PAGE_SHIFT;
1183 pg_end = (offset + len) >> PAGE_SHIFT;
1184
1185 /* avoid gc operation during block exchange */
1186 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1187
1188 down_write(&F2FS_I(inode)->i_mmap_sem);
1189 /* write out all dirty pages from offset */
1190 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1191 if (ret)
1192 goto out_unlock;
1193
1194 truncate_pagecache(inode, offset);
1195
1196 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1197 if (ret)
1198 goto out_unlock;
1199
1200 /* write out all moved pages, if possible */
1201 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1202 truncate_pagecache(inode, offset);
1203
1204 new_size = i_size_read(inode) - len;
1205 truncate_pagecache(inode, new_size);
1206
1207 ret = truncate_blocks(inode, new_size, true);
1208 if (!ret)
1209 f2fs_i_size_write(inode, new_size);
1210 out_unlock:
1211 up_write(&F2FS_I(inode)->i_mmap_sem);
1212 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1213 return ret;
1214 }
1215
1216 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1217 pgoff_t end)
1218 {
1219 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1220 pgoff_t index = start;
1221 unsigned int ofs_in_node = dn->ofs_in_node;
1222 blkcnt_t count = 0;
1223 int ret;
1224
1225 for (; index < end; index++, dn->ofs_in_node++) {
1226 if (datablock_addr(dn->inode, dn->node_page,
1227 dn->ofs_in_node) == NULL_ADDR)
1228 count++;
1229 }
1230
1231 dn->ofs_in_node = ofs_in_node;
1232 ret = reserve_new_blocks(dn, count);
1233 if (ret)
1234 return ret;
1235
1236 dn->ofs_in_node = ofs_in_node;
1237 for (index = start; index < end; index++, dn->ofs_in_node++) {
1238 dn->data_blkaddr = datablock_addr(dn->inode,
1239 dn->node_page, dn->ofs_in_node);
1240 /*
1241 * reserve_new_blocks will not guarantee entire block
1242 * allocation.
1243 */
1244 if (dn->data_blkaddr == NULL_ADDR) {
1245 ret = -ENOSPC;
1246 break;
1247 }
1248 if (dn->data_blkaddr != NEW_ADDR) {
1249 invalidate_blocks(sbi, dn->data_blkaddr);
1250 dn->data_blkaddr = NEW_ADDR;
1251 set_data_blkaddr(dn);
1252 }
1253 }
1254
1255 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1256
1257 return ret;
1258 }
1259
1260 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1261 int mode)
1262 {
1263 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1264 struct address_space *mapping = inode->i_mapping;
1265 pgoff_t index, pg_start, pg_end;
1266 loff_t new_size = i_size_read(inode);
1267 loff_t off_start, off_end;
1268 int ret = 0;
1269
1270 ret = inode_newsize_ok(inode, (len + offset));
1271 if (ret)
1272 return ret;
1273
1274 ret = f2fs_convert_inline_inode(inode);
1275 if (ret)
1276 return ret;
1277
1278 down_write(&F2FS_I(inode)->i_mmap_sem);
1279 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1280 if (ret)
1281 goto out_sem;
1282
1283 truncate_pagecache_range(inode, offset, offset + len - 1);
1284
1285 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1286 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1287
1288 off_start = offset & (PAGE_SIZE - 1);
1289 off_end = (offset + len) & (PAGE_SIZE - 1);
1290
1291 if (pg_start == pg_end) {
1292 ret = fill_zero(inode, pg_start, off_start,
1293 off_end - off_start);
1294 if (ret)
1295 goto out_sem;
1296
1297 new_size = max_t(loff_t, new_size, offset + len);
1298 } else {
1299 if (off_start) {
1300 ret = fill_zero(inode, pg_start++, off_start,
1301 PAGE_SIZE - off_start);
1302 if (ret)
1303 goto out_sem;
1304
1305 new_size = max_t(loff_t, new_size,
1306 (loff_t)pg_start << PAGE_SHIFT);
1307 }
1308
1309 for (index = pg_start; index < pg_end;) {
1310 struct dnode_of_data dn;
1311 unsigned int end_offset;
1312 pgoff_t end;
1313
1314 f2fs_lock_op(sbi);
1315
1316 set_new_dnode(&dn, inode, NULL, NULL, 0);
1317 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1318 if (ret) {
1319 f2fs_unlock_op(sbi);
1320 goto out;
1321 }
1322
1323 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1324 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1325
1326 ret = f2fs_do_zero_range(&dn, index, end);
1327 f2fs_put_dnode(&dn);
1328 f2fs_unlock_op(sbi);
1329
1330 f2fs_balance_fs(sbi, dn.node_changed);
1331
1332 if (ret)
1333 goto out;
1334
1335 index = end;
1336 new_size = max_t(loff_t, new_size,
1337 (loff_t)index << PAGE_SHIFT);
1338 }
1339
1340 if (off_end) {
1341 ret = fill_zero(inode, pg_end, 0, off_end);
1342 if (ret)
1343 goto out;
1344
1345 new_size = max_t(loff_t, new_size, offset + len);
1346 }
1347 }
1348
1349 out:
1350 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1351 f2fs_i_size_write(inode, new_size);
1352 out_sem:
1353 up_write(&F2FS_I(inode)->i_mmap_sem);
1354
1355 return ret;
1356 }
1357
1358 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1359 {
1360 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1361 pgoff_t nr, pg_start, pg_end, delta, idx;
1362 loff_t new_size;
1363 int ret = 0;
1364
1365 new_size = i_size_read(inode) + len;
1366 ret = inode_newsize_ok(inode, new_size);
1367 if (ret)
1368 return ret;
1369
1370 if (offset >= i_size_read(inode))
1371 return -EINVAL;
1372
1373 /* insert range should be aligned to block size of f2fs. */
1374 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1375 return -EINVAL;
1376
1377 ret = f2fs_convert_inline_inode(inode);
1378 if (ret)
1379 return ret;
1380
1381 f2fs_balance_fs(sbi, true);
1382
1383 /* avoid gc operation during block exchange */
1384 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1385
1386 down_write(&F2FS_I(inode)->i_mmap_sem);
1387 ret = truncate_blocks(inode, i_size_read(inode), true);
1388 if (ret)
1389 goto out;
1390
1391 /* write out all dirty pages from offset */
1392 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1393 if (ret)
1394 goto out;
1395
1396 truncate_pagecache(inode, offset);
1397
1398 pg_start = offset >> PAGE_SHIFT;
1399 pg_end = (offset + len) >> PAGE_SHIFT;
1400 delta = pg_end - pg_start;
1401 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1402
1403 while (!ret && idx > pg_start) {
1404 nr = idx - pg_start;
1405 if (nr > delta)
1406 nr = delta;
1407 idx -= nr;
1408
1409 f2fs_lock_op(sbi);
1410 f2fs_drop_extent_tree(inode);
1411
1412 ret = __exchange_data_block(inode, inode, idx,
1413 idx + delta, nr, false);
1414 f2fs_unlock_op(sbi);
1415 }
1416
1417 /* write out all moved pages, if possible */
1418 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1419 truncate_pagecache(inode, offset);
1420
1421 if (!ret)
1422 f2fs_i_size_write(inode, new_size);
1423 out:
1424 up_write(&F2FS_I(inode)->i_mmap_sem);
1425 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1426 return ret;
1427 }
1428
1429 static int expand_inode_data(struct inode *inode, loff_t offset,
1430 loff_t len, int mode)
1431 {
1432 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1433 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1434 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1435 pgoff_t pg_end;
1436 loff_t new_size = i_size_read(inode);
1437 loff_t off_end;
1438 int err;
1439
1440 err = inode_newsize_ok(inode, (len + offset));
1441 if (err)
1442 return err;
1443
1444 err = f2fs_convert_inline_inode(inode);
1445 if (err)
1446 return err;
1447
1448 f2fs_balance_fs(sbi, true);
1449
1450 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1451 off_end = (offset + len) & (PAGE_SIZE - 1);
1452
1453 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1454 map.m_len = pg_end - map.m_lblk;
1455 if (off_end)
1456 map.m_len++;
1457
1458 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1459 if (err) {
1460 pgoff_t last_off;
1461
1462 if (!map.m_len)
1463 return err;
1464
1465 last_off = map.m_lblk + map.m_len - 1;
1466
1467 /* update new size to the failed position */
1468 new_size = (last_off == pg_end) ? offset + len:
1469 (loff_t)(last_off + 1) << PAGE_SHIFT;
1470 } else {
1471 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1472 }
1473
1474 if (new_size > i_size_read(inode)) {
1475 if (mode & FALLOC_FL_KEEP_SIZE)
1476 file_set_keep_isize(inode);
1477 else
1478 f2fs_i_size_write(inode, new_size);
1479 }
1480
1481 return err;
1482 }
1483
1484 static long f2fs_fallocate(struct file *file, int mode,
1485 loff_t offset, loff_t len)
1486 {
1487 struct inode *inode = file_inode(file);
1488 long ret = 0;
1489
1490 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1491 return -EIO;
1492
1493 /* f2fs only support ->fallocate for regular file */
1494 if (!S_ISREG(inode->i_mode))
1495 return -EINVAL;
1496
1497 if (f2fs_encrypted_inode(inode) &&
1498 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1499 return -EOPNOTSUPP;
1500
1501 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1502 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1503 FALLOC_FL_INSERT_RANGE))
1504 return -EOPNOTSUPP;
1505
1506 inode_lock(inode);
1507
1508 if (mode & FALLOC_FL_PUNCH_HOLE) {
1509 if (offset >= inode->i_size)
1510 goto out;
1511
1512 ret = punch_hole(inode, offset, len);
1513 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1514 ret = f2fs_collapse_range(inode, offset, len);
1515 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1516 ret = f2fs_zero_range(inode, offset, len, mode);
1517 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1518 ret = f2fs_insert_range(inode, offset, len);
1519 } else {
1520 ret = expand_inode_data(inode, offset, len, mode);
1521 }
1522
1523 if (!ret) {
1524 inode->i_mtime = inode->i_ctime = current_time(inode);
1525 f2fs_mark_inode_dirty_sync(inode, false);
1526 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1527 }
1528
1529 out:
1530 inode_unlock(inode);
1531
1532 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1533 return ret;
1534 }
1535
1536 static int f2fs_release_file(struct inode *inode, struct file *filp)
1537 {
1538 /*
1539 * f2fs_relase_file is called at every close calls. So we should
1540 * not drop any inmemory pages by close called by other process.
1541 */
1542 if (!(filp->f_mode & FMODE_WRITE) ||
1543 atomic_read(&inode->i_writecount) != 1)
1544 return 0;
1545
1546 /* some remained atomic pages should discarded */
1547 if (f2fs_is_atomic_file(inode))
1548 drop_inmem_pages(inode);
1549 if (f2fs_is_volatile_file(inode)) {
1550 clear_inode_flag(inode, FI_VOLATILE_FILE);
1551 stat_dec_volatile_write(inode);
1552 set_inode_flag(inode, FI_DROP_CACHE);
1553 filemap_fdatawrite(inode->i_mapping);
1554 clear_inode_flag(inode, FI_DROP_CACHE);
1555 }
1556 return 0;
1557 }
1558
1559 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1560 {
1561 struct inode *inode = file_inode(file);
1562
1563 /*
1564 * If the process doing a transaction is crashed, we should do
1565 * roll-back. Otherwise, other reader/write can see corrupted database
1566 * until all the writers close its file. Since this should be done
1567 * before dropping file lock, it needs to do in ->flush.
1568 */
1569 if (f2fs_is_atomic_file(inode) &&
1570 F2FS_I(inode)->inmem_task == current)
1571 drop_inmem_pages(inode);
1572 return 0;
1573 }
1574
1575 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1576 {
1577 struct inode *inode = file_inode(filp);
1578 struct f2fs_inode_info *fi = F2FS_I(inode);
1579 unsigned int flags = fi->i_flags &
1580 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1581 return put_user(flags, (int __user *)arg);
1582 }
1583
1584 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1585 {
1586 struct f2fs_inode_info *fi = F2FS_I(inode);
1587 unsigned int oldflags;
1588
1589 /* Is it quota file? Do not allow user to mess with it */
1590 if (IS_NOQUOTA(inode))
1591 return -EPERM;
1592
1593 flags = f2fs_mask_flags(inode->i_mode, flags);
1594
1595 oldflags = fi->i_flags;
1596
1597 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1598 if (!capable(CAP_LINUX_IMMUTABLE))
1599 return -EPERM;
1600
1601 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1602 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1603 fi->i_flags = flags;
1604
1605 if (fi->i_flags & FS_PROJINHERIT_FL)
1606 set_inode_flag(inode, FI_PROJ_INHERIT);
1607 else
1608 clear_inode_flag(inode, FI_PROJ_INHERIT);
1609
1610 inode->i_ctime = current_time(inode);
1611 f2fs_set_inode_flags(inode);
1612 f2fs_mark_inode_dirty_sync(inode, false);
1613 return 0;
1614 }
1615
1616 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1617 {
1618 struct inode *inode = file_inode(filp);
1619 unsigned int flags;
1620 int ret;
1621
1622 if (!inode_owner_or_capable(inode))
1623 return -EACCES;
1624
1625 if (get_user(flags, (int __user *)arg))
1626 return -EFAULT;
1627
1628 ret = mnt_want_write_file(filp);
1629 if (ret)
1630 return ret;
1631
1632 inode_lock(inode);
1633
1634 ret = __f2fs_ioc_setflags(inode, flags);
1635
1636 inode_unlock(inode);
1637 mnt_drop_write_file(filp);
1638 return ret;
1639 }
1640
1641 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1642 {
1643 struct inode *inode = file_inode(filp);
1644
1645 return put_user(inode->i_generation, (int __user *)arg);
1646 }
1647
1648 static int f2fs_ioc_start_atomic_write(struct file *filp)
1649 {
1650 struct inode *inode = file_inode(filp);
1651 int ret;
1652
1653 if (!inode_owner_or_capable(inode))
1654 return -EACCES;
1655
1656 if (!S_ISREG(inode->i_mode))
1657 return -EINVAL;
1658
1659 ret = mnt_want_write_file(filp);
1660 if (ret)
1661 return ret;
1662
1663 inode_lock(inode);
1664
1665 if (f2fs_is_atomic_file(inode))
1666 goto out;
1667
1668 ret = f2fs_convert_inline_inode(inode);
1669 if (ret)
1670 goto out;
1671
1672 set_inode_flag(inode, FI_ATOMIC_FILE);
1673 set_inode_flag(inode, FI_HOT_DATA);
1674 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1675
1676 if (!get_dirty_pages(inode))
1677 goto inc_stat;
1678
1679 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1680 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1681 inode->i_ino, get_dirty_pages(inode));
1682 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1683 if (ret) {
1684 clear_inode_flag(inode, FI_ATOMIC_FILE);
1685 clear_inode_flag(inode, FI_HOT_DATA);
1686 goto out;
1687 }
1688
1689 inc_stat:
1690 F2FS_I(inode)->inmem_task = current;
1691 stat_inc_atomic_write(inode);
1692 stat_update_max_atomic_write(inode);
1693 out:
1694 inode_unlock(inode);
1695 mnt_drop_write_file(filp);
1696 return ret;
1697 }
1698
1699 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1700 {
1701 struct inode *inode = file_inode(filp);
1702 int ret;
1703
1704 if (!inode_owner_or_capable(inode))
1705 return -EACCES;
1706
1707 ret = mnt_want_write_file(filp);
1708 if (ret)
1709 return ret;
1710
1711 inode_lock(inode);
1712
1713 if (f2fs_is_volatile_file(inode))
1714 goto err_out;
1715
1716 if (f2fs_is_atomic_file(inode)) {
1717 ret = commit_inmem_pages(inode);
1718 if (ret)
1719 goto err_out;
1720
1721 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1722 if (!ret) {
1723 clear_inode_flag(inode, FI_ATOMIC_FILE);
1724 clear_inode_flag(inode, FI_HOT_DATA);
1725 stat_dec_atomic_write(inode);
1726 }
1727 } else {
1728 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1729 }
1730 err_out:
1731 inode_unlock(inode);
1732 mnt_drop_write_file(filp);
1733 return ret;
1734 }
1735
1736 static int f2fs_ioc_start_volatile_write(struct file *filp)
1737 {
1738 struct inode *inode = file_inode(filp);
1739 int ret;
1740
1741 if (!inode_owner_or_capable(inode))
1742 return -EACCES;
1743
1744 if (!S_ISREG(inode->i_mode))
1745 return -EINVAL;
1746
1747 ret = mnt_want_write_file(filp);
1748 if (ret)
1749 return ret;
1750
1751 inode_lock(inode);
1752
1753 if (f2fs_is_volatile_file(inode))
1754 goto out;
1755
1756 ret = f2fs_convert_inline_inode(inode);
1757 if (ret)
1758 goto out;
1759
1760 stat_inc_volatile_write(inode);
1761 stat_update_max_volatile_write(inode);
1762
1763 set_inode_flag(inode, FI_VOLATILE_FILE);
1764 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1765 out:
1766 inode_unlock(inode);
1767 mnt_drop_write_file(filp);
1768 return ret;
1769 }
1770
1771 static int f2fs_ioc_release_volatile_write(struct file *filp)
1772 {
1773 struct inode *inode = file_inode(filp);
1774 int ret;
1775
1776 if (!inode_owner_or_capable(inode))
1777 return -EACCES;
1778
1779 ret = mnt_want_write_file(filp);
1780 if (ret)
1781 return ret;
1782
1783 inode_lock(inode);
1784
1785 if (!f2fs_is_volatile_file(inode))
1786 goto out;
1787
1788 if (!f2fs_is_first_block_written(inode)) {
1789 ret = truncate_partial_data_page(inode, 0, true);
1790 goto out;
1791 }
1792
1793 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1794 out:
1795 inode_unlock(inode);
1796 mnt_drop_write_file(filp);
1797 return ret;
1798 }
1799
1800 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1801 {
1802 struct inode *inode = file_inode(filp);
1803 int ret;
1804
1805 if (!inode_owner_or_capable(inode))
1806 return -EACCES;
1807
1808 ret = mnt_want_write_file(filp);
1809 if (ret)
1810 return ret;
1811
1812 inode_lock(inode);
1813
1814 if (f2fs_is_atomic_file(inode))
1815 drop_inmem_pages(inode);
1816 if (f2fs_is_volatile_file(inode)) {
1817 clear_inode_flag(inode, FI_VOLATILE_FILE);
1818 stat_dec_volatile_write(inode);
1819 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1820 }
1821
1822 inode_unlock(inode);
1823
1824 mnt_drop_write_file(filp);
1825 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1826 return ret;
1827 }
1828
1829 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1830 {
1831 struct inode *inode = file_inode(filp);
1832 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1833 struct super_block *sb = sbi->sb;
1834 __u32 in;
1835 int ret;
1836
1837 if (!capable(CAP_SYS_ADMIN))
1838 return -EPERM;
1839
1840 if (get_user(in, (__u32 __user *)arg))
1841 return -EFAULT;
1842
1843 ret = mnt_want_write_file(filp);
1844 if (ret)
1845 return ret;
1846
1847 switch (in) {
1848 case F2FS_GOING_DOWN_FULLSYNC:
1849 sb = freeze_bdev(sb->s_bdev);
1850 if (IS_ERR(sb)) {
1851 ret = PTR_ERR(sb);
1852 goto out;
1853 }
1854 if (sb) {
1855 f2fs_stop_checkpoint(sbi, false);
1856 thaw_bdev(sb->s_bdev, sb);
1857 }
1858 break;
1859 case F2FS_GOING_DOWN_METASYNC:
1860 /* do checkpoint only */
1861 ret = f2fs_sync_fs(sb, 1);
1862 if (ret)
1863 goto out;
1864 f2fs_stop_checkpoint(sbi, false);
1865 break;
1866 case F2FS_GOING_DOWN_NOSYNC:
1867 f2fs_stop_checkpoint(sbi, false);
1868 break;
1869 case F2FS_GOING_DOWN_METAFLUSH:
1870 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1871 f2fs_stop_checkpoint(sbi, false);
1872 break;
1873 default:
1874 ret = -EINVAL;
1875 goto out;
1876 }
1877
1878 stop_gc_thread(sbi);
1879 stop_discard_thread(sbi);
1880
1881 drop_discard_cmd(sbi);
1882 clear_opt(sbi, DISCARD);
1883
1884 f2fs_update_time(sbi, REQ_TIME);
1885 out:
1886 mnt_drop_write_file(filp);
1887 return ret;
1888 }
1889
1890 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1891 {
1892 struct inode *inode = file_inode(filp);
1893 struct super_block *sb = inode->i_sb;
1894 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1895 struct fstrim_range range;
1896 int ret;
1897
1898 if (!capable(CAP_SYS_ADMIN))
1899 return -EPERM;
1900
1901 if (!blk_queue_discard(q))
1902 return -EOPNOTSUPP;
1903
1904 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1905 sizeof(range)))
1906 return -EFAULT;
1907
1908 ret = mnt_want_write_file(filp);
1909 if (ret)
1910 return ret;
1911
1912 range.minlen = max((unsigned int)range.minlen,
1913 q->limits.discard_granularity);
1914 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1915 mnt_drop_write_file(filp);
1916 if (ret < 0)
1917 return ret;
1918
1919 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1920 sizeof(range)))
1921 return -EFAULT;
1922 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1923 return 0;
1924 }
1925
1926 static bool uuid_is_nonzero(__u8 u[16])
1927 {
1928 int i;
1929
1930 for (i = 0; i < 16; i++)
1931 if (u[i])
1932 return true;
1933 return false;
1934 }
1935
1936 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1937 {
1938 struct inode *inode = file_inode(filp);
1939
1940 if (!f2fs_sb_has_crypto(inode->i_sb))
1941 return -EOPNOTSUPP;
1942
1943 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1944
1945 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1946 }
1947
1948 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1949 {
1950 if (!f2fs_sb_has_crypto(file_inode(filp)->i_sb))
1951 return -EOPNOTSUPP;
1952 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1953 }
1954
1955 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1956 {
1957 struct inode *inode = file_inode(filp);
1958 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1959 int err;
1960
1961 if (!f2fs_sb_has_crypto(inode->i_sb))
1962 return -EOPNOTSUPP;
1963
1964 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1965 goto got_it;
1966
1967 err = mnt_want_write_file(filp);
1968 if (err)
1969 return err;
1970
1971 /* update superblock with uuid */
1972 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1973
1974 err = f2fs_commit_super(sbi, false);
1975 if (err) {
1976 /* undo new data */
1977 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1978 mnt_drop_write_file(filp);
1979 return err;
1980 }
1981 mnt_drop_write_file(filp);
1982 got_it:
1983 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1984 16))
1985 return -EFAULT;
1986 return 0;
1987 }
1988
1989 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1990 {
1991 struct inode *inode = file_inode(filp);
1992 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1993 __u32 sync;
1994 int ret;
1995
1996 if (!capable(CAP_SYS_ADMIN))
1997 return -EPERM;
1998
1999 if (get_user(sync, (__u32 __user *)arg))
2000 return -EFAULT;
2001
2002 if (f2fs_readonly(sbi->sb))
2003 return -EROFS;
2004
2005 ret = mnt_want_write_file(filp);
2006 if (ret)
2007 return ret;
2008
2009 if (!sync) {
2010 if (!mutex_trylock(&sbi->gc_mutex)) {
2011 ret = -EBUSY;
2012 goto out;
2013 }
2014 } else {
2015 mutex_lock(&sbi->gc_mutex);
2016 }
2017
2018 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2019 out:
2020 mnt_drop_write_file(filp);
2021 return ret;
2022 }
2023
2024 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2025 {
2026 struct inode *inode = file_inode(filp);
2027 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2028 struct f2fs_gc_range range;
2029 u64 end;
2030 int ret;
2031
2032 if (!capable(CAP_SYS_ADMIN))
2033 return -EPERM;
2034
2035 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2036 sizeof(range)))
2037 return -EFAULT;
2038
2039 if (f2fs_readonly(sbi->sb))
2040 return -EROFS;
2041
2042 ret = mnt_want_write_file(filp);
2043 if (ret)
2044 return ret;
2045
2046 end = range.start + range.len;
2047 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi))
2048 return -EINVAL;
2049 do_more:
2050 if (!range.sync) {
2051 if (!mutex_trylock(&sbi->gc_mutex)) {
2052 ret = -EBUSY;
2053 goto out;
2054 }
2055 } else {
2056 mutex_lock(&sbi->gc_mutex);
2057 }
2058
2059 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2060 range.start += sbi->blocks_per_seg;
2061 if (range.start <= end)
2062 goto do_more;
2063 out:
2064 mnt_drop_write_file(filp);
2065 return ret;
2066 }
2067
2068 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2069 {
2070 struct inode *inode = file_inode(filp);
2071 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2072 int ret;
2073
2074 if (!capable(CAP_SYS_ADMIN))
2075 return -EPERM;
2076
2077 if (f2fs_readonly(sbi->sb))
2078 return -EROFS;
2079
2080 ret = mnt_want_write_file(filp);
2081 if (ret)
2082 return ret;
2083
2084 ret = f2fs_sync_fs(sbi->sb, 1);
2085
2086 mnt_drop_write_file(filp);
2087 return ret;
2088 }
2089
2090 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2091 struct file *filp,
2092 struct f2fs_defragment *range)
2093 {
2094 struct inode *inode = file_inode(filp);
2095 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2096 .m_seg_type = NO_CHECK_TYPE };
2097 struct extent_info ei = {0,0,0};
2098 pgoff_t pg_start, pg_end, next_pgofs;
2099 unsigned int blk_per_seg = sbi->blocks_per_seg;
2100 unsigned int total = 0, sec_num;
2101 block_t blk_end = 0;
2102 bool fragmented = false;
2103 int err;
2104
2105 /* if in-place-update policy is enabled, don't waste time here */
2106 if (should_update_inplace(inode, NULL))
2107 return -EINVAL;
2108
2109 pg_start = range->start >> PAGE_SHIFT;
2110 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2111
2112 f2fs_balance_fs(sbi, true);
2113
2114 inode_lock(inode);
2115
2116 /* writeback all dirty pages in the range */
2117 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2118 range->start + range->len - 1);
2119 if (err)
2120 goto out;
2121
2122 /*
2123 * lookup mapping info in extent cache, skip defragmenting if physical
2124 * block addresses are continuous.
2125 */
2126 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2127 if (ei.fofs + ei.len >= pg_end)
2128 goto out;
2129 }
2130
2131 map.m_lblk = pg_start;
2132 map.m_next_pgofs = &next_pgofs;
2133
2134 /*
2135 * lookup mapping info in dnode page cache, skip defragmenting if all
2136 * physical block addresses are continuous even if there are hole(s)
2137 * in logical blocks.
2138 */
2139 while (map.m_lblk < pg_end) {
2140 map.m_len = pg_end - map.m_lblk;
2141 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2142 if (err)
2143 goto out;
2144
2145 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2146 map.m_lblk = next_pgofs;
2147 continue;
2148 }
2149
2150 if (blk_end && blk_end != map.m_pblk)
2151 fragmented = true;
2152
2153 /* record total count of block that we're going to move */
2154 total += map.m_len;
2155
2156 blk_end = map.m_pblk + map.m_len;
2157
2158 map.m_lblk += map.m_len;
2159 }
2160
2161 if (!fragmented)
2162 goto out;
2163
2164 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2165
2166 /*
2167 * make sure there are enough free section for LFS allocation, this can
2168 * avoid defragment running in SSR mode when free section are allocated
2169 * intensively
2170 */
2171 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2172 err = -EAGAIN;
2173 goto out;
2174 }
2175
2176 map.m_lblk = pg_start;
2177 map.m_len = pg_end - pg_start;
2178 total = 0;
2179
2180 while (map.m_lblk < pg_end) {
2181 pgoff_t idx;
2182 int cnt = 0;
2183
2184 do_map:
2185 map.m_len = pg_end - map.m_lblk;
2186 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2187 if (err)
2188 goto clear_out;
2189
2190 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2191 map.m_lblk = next_pgofs;
2192 continue;
2193 }
2194
2195 set_inode_flag(inode, FI_DO_DEFRAG);
2196
2197 idx = map.m_lblk;
2198 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2199 struct page *page;
2200
2201 page = get_lock_data_page(inode, idx, true);
2202 if (IS_ERR(page)) {
2203 err = PTR_ERR(page);
2204 goto clear_out;
2205 }
2206
2207 set_page_dirty(page);
2208 f2fs_put_page(page, 1);
2209
2210 idx++;
2211 cnt++;
2212 total++;
2213 }
2214
2215 map.m_lblk = idx;
2216
2217 if (idx < pg_end && cnt < blk_per_seg)
2218 goto do_map;
2219
2220 clear_inode_flag(inode, FI_DO_DEFRAG);
2221
2222 err = filemap_fdatawrite(inode->i_mapping);
2223 if (err)
2224 goto out;
2225 }
2226 clear_out:
2227 clear_inode_flag(inode, FI_DO_DEFRAG);
2228 out:
2229 inode_unlock(inode);
2230 if (!err)
2231 range->len = (u64)total << PAGE_SHIFT;
2232 return err;
2233 }
2234
2235 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2236 {
2237 struct inode *inode = file_inode(filp);
2238 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2239 struct f2fs_defragment range;
2240 int err;
2241
2242 if (!capable(CAP_SYS_ADMIN))
2243 return -EPERM;
2244
2245 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2246 return -EINVAL;
2247
2248 if (f2fs_readonly(sbi->sb))
2249 return -EROFS;
2250
2251 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2252 sizeof(range)))
2253 return -EFAULT;
2254
2255 /* verify alignment of offset & size */
2256 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2257 return -EINVAL;
2258
2259 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2260 sbi->max_file_blocks))
2261 return -EINVAL;
2262
2263 err = mnt_want_write_file(filp);
2264 if (err)
2265 return err;
2266
2267 err = f2fs_defragment_range(sbi, filp, &range);
2268 mnt_drop_write_file(filp);
2269
2270 f2fs_update_time(sbi, REQ_TIME);
2271 if (err < 0)
2272 return err;
2273
2274 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2275 sizeof(range)))
2276 return -EFAULT;
2277
2278 return 0;
2279 }
2280
2281 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2282 struct file *file_out, loff_t pos_out, size_t len)
2283 {
2284 struct inode *src = file_inode(file_in);
2285 struct inode *dst = file_inode(file_out);
2286 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2287 size_t olen = len, dst_max_i_size = 0;
2288 size_t dst_osize;
2289 int ret;
2290
2291 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2292 src->i_sb != dst->i_sb)
2293 return -EXDEV;
2294
2295 if (unlikely(f2fs_readonly(src->i_sb)))
2296 return -EROFS;
2297
2298 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2299 return -EINVAL;
2300
2301 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2302 return -EOPNOTSUPP;
2303
2304 if (src == dst) {
2305 if (pos_in == pos_out)
2306 return 0;
2307 if (pos_out > pos_in && pos_out < pos_in + len)
2308 return -EINVAL;
2309 }
2310
2311 inode_lock(src);
2312 down_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2313 if (src != dst) {
2314 ret = -EBUSY;
2315 if (!inode_trylock(dst))
2316 goto out;
2317 if (!down_write_trylock(&F2FS_I(dst)->dio_rwsem[WRITE])) {
2318 inode_unlock(dst);
2319 goto out;
2320 }
2321 }
2322
2323 ret = -EINVAL;
2324 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2325 goto out_unlock;
2326 if (len == 0)
2327 olen = len = src->i_size - pos_in;
2328 if (pos_in + len == src->i_size)
2329 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2330 if (len == 0) {
2331 ret = 0;
2332 goto out_unlock;
2333 }
2334
2335 dst_osize = dst->i_size;
2336 if (pos_out + olen > dst->i_size)
2337 dst_max_i_size = pos_out + olen;
2338
2339 /* verify the end result is block aligned */
2340 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2341 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2342 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2343 goto out_unlock;
2344
2345 ret = f2fs_convert_inline_inode(src);
2346 if (ret)
2347 goto out_unlock;
2348
2349 ret = f2fs_convert_inline_inode(dst);
2350 if (ret)
2351 goto out_unlock;
2352
2353 /* write out all dirty pages from offset */
2354 ret = filemap_write_and_wait_range(src->i_mapping,
2355 pos_in, pos_in + len);
2356 if (ret)
2357 goto out_unlock;
2358
2359 ret = filemap_write_and_wait_range(dst->i_mapping,
2360 pos_out, pos_out + len);
2361 if (ret)
2362 goto out_unlock;
2363
2364 f2fs_balance_fs(sbi, true);
2365 f2fs_lock_op(sbi);
2366 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2367 pos_out >> F2FS_BLKSIZE_BITS,
2368 len >> F2FS_BLKSIZE_BITS, false);
2369
2370 if (!ret) {
2371 if (dst_max_i_size)
2372 f2fs_i_size_write(dst, dst_max_i_size);
2373 else if (dst_osize != dst->i_size)
2374 f2fs_i_size_write(dst, dst_osize);
2375 }
2376 f2fs_unlock_op(sbi);
2377 out_unlock:
2378 if (src != dst) {
2379 up_write(&F2FS_I(dst)->dio_rwsem[WRITE]);
2380 inode_unlock(dst);
2381 }
2382 out:
2383 up_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2384 inode_unlock(src);
2385 return ret;
2386 }
2387
2388 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2389 {
2390 struct f2fs_move_range range;
2391 struct fd dst;
2392 int err;
2393
2394 if (!(filp->f_mode & FMODE_READ) ||
2395 !(filp->f_mode & FMODE_WRITE))
2396 return -EBADF;
2397
2398 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2399 sizeof(range)))
2400 return -EFAULT;
2401
2402 dst = fdget(range.dst_fd);
2403 if (!dst.file)
2404 return -EBADF;
2405
2406 if (!(dst.file->f_mode & FMODE_WRITE)) {
2407 err = -EBADF;
2408 goto err_out;
2409 }
2410
2411 err = mnt_want_write_file(filp);
2412 if (err)
2413 goto err_out;
2414
2415 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2416 range.pos_out, range.len);
2417
2418 mnt_drop_write_file(filp);
2419 if (err)
2420 goto err_out;
2421
2422 if (copy_to_user((struct f2fs_move_range __user *)arg,
2423 &range, sizeof(range)))
2424 err = -EFAULT;
2425 err_out:
2426 fdput(dst);
2427 return err;
2428 }
2429
2430 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2431 {
2432 struct inode *inode = file_inode(filp);
2433 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2434 struct sit_info *sm = SIT_I(sbi);
2435 unsigned int start_segno = 0, end_segno = 0;
2436 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2437 struct f2fs_flush_device range;
2438 int ret;
2439
2440 if (!capable(CAP_SYS_ADMIN))
2441 return -EPERM;
2442
2443 if (f2fs_readonly(sbi->sb))
2444 return -EROFS;
2445
2446 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2447 sizeof(range)))
2448 return -EFAULT;
2449
2450 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2451 sbi->segs_per_sec != 1) {
2452 f2fs_msg(sbi->sb, KERN_WARNING,
2453 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2454 range.dev_num, sbi->s_ndevs,
2455 sbi->segs_per_sec);
2456 return -EINVAL;
2457 }
2458
2459 ret = mnt_want_write_file(filp);
2460 if (ret)
2461 return ret;
2462
2463 if (range.dev_num != 0)
2464 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2465 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2466
2467 start_segno = sm->last_victim[FLUSH_DEVICE];
2468 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2469 start_segno = dev_start_segno;
2470 end_segno = min(start_segno + range.segments, dev_end_segno);
2471
2472 while (start_segno < end_segno) {
2473 if (!mutex_trylock(&sbi->gc_mutex)) {
2474 ret = -EBUSY;
2475 goto out;
2476 }
2477 sm->last_victim[GC_CB] = end_segno + 1;
2478 sm->last_victim[GC_GREEDY] = end_segno + 1;
2479 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2480 ret = f2fs_gc(sbi, true, true, start_segno);
2481 if (ret == -EAGAIN)
2482 ret = 0;
2483 else if (ret < 0)
2484 break;
2485 start_segno++;
2486 }
2487 out:
2488 mnt_drop_write_file(filp);
2489 return ret;
2490 }
2491
2492 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2493 {
2494 struct inode *inode = file_inode(filp);
2495 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2496
2497 /* Must validate to set it with SQLite behavior in Android. */
2498 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2499
2500 return put_user(sb_feature, (u32 __user *)arg);
2501 }
2502
2503 #ifdef CONFIG_QUOTA
2504 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2505 {
2506 struct inode *inode = file_inode(filp);
2507 struct f2fs_inode_info *fi = F2FS_I(inode);
2508 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2509 struct super_block *sb = sbi->sb;
2510 struct dquot *transfer_to[MAXQUOTAS] = {};
2511 struct page *ipage;
2512 kprojid_t kprojid;
2513 int err;
2514
2515 if (!f2fs_sb_has_project_quota(sb)) {
2516 if (projid != F2FS_DEF_PROJID)
2517 return -EOPNOTSUPP;
2518 else
2519 return 0;
2520 }
2521
2522 if (!f2fs_has_extra_attr(inode))
2523 return -EOPNOTSUPP;
2524
2525 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2526
2527 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2528 return 0;
2529
2530 err = mnt_want_write_file(filp);
2531 if (err)
2532 return err;
2533
2534 err = -EPERM;
2535 inode_lock(inode);
2536
2537 /* Is it quota file? Do not allow user to mess with it */
2538 if (IS_NOQUOTA(inode))
2539 goto out_unlock;
2540
2541 ipage = get_node_page(sbi, inode->i_ino);
2542 if (IS_ERR(ipage)) {
2543 err = PTR_ERR(ipage);
2544 goto out_unlock;
2545 }
2546
2547 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2548 i_projid)) {
2549 err = -EOVERFLOW;
2550 f2fs_put_page(ipage, 1);
2551 goto out_unlock;
2552 }
2553 f2fs_put_page(ipage, 1);
2554
2555 dquot_initialize(inode);
2556
2557 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2558 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2559 err = __dquot_transfer(inode, transfer_to);
2560 dqput(transfer_to[PRJQUOTA]);
2561 if (err)
2562 goto out_dirty;
2563 }
2564
2565 F2FS_I(inode)->i_projid = kprojid;
2566 inode->i_ctime = current_time(inode);
2567 out_dirty:
2568 f2fs_mark_inode_dirty_sync(inode, true);
2569 out_unlock:
2570 inode_unlock(inode);
2571 mnt_drop_write_file(filp);
2572 return err;
2573 }
2574 #else
2575 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2576 {
2577 if (projid != F2FS_DEF_PROJID)
2578 return -EOPNOTSUPP;
2579 return 0;
2580 }
2581 #endif
2582
2583 /* Transfer internal flags to xflags */
2584 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2585 {
2586 __u32 xflags = 0;
2587
2588 if (iflags & FS_SYNC_FL)
2589 xflags |= FS_XFLAG_SYNC;
2590 if (iflags & FS_IMMUTABLE_FL)
2591 xflags |= FS_XFLAG_IMMUTABLE;
2592 if (iflags & FS_APPEND_FL)
2593 xflags |= FS_XFLAG_APPEND;
2594 if (iflags & FS_NODUMP_FL)
2595 xflags |= FS_XFLAG_NODUMP;
2596 if (iflags & FS_NOATIME_FL)
2597 xflags |= FS_XFLAG_NOATIME;
2598 if (iflags & FS_PROJINHERIT_FL)
2599 xflags |= FS_XFLAG_PROJINHERIT;
2600 return xflags;
2601 }
2602
2603 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2604 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2605 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2606
2607 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2608 #define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2609 FS_IMMUTABLE_FL | \
2610 FS_APPEND_FL | \
2611 FS_NODUMP_FL | \
2612 FS_NOATIME_FL | \
2613 FS_PROJINHERIT_FL)
2614
2615 /* Transfer xflags flags to internal */
2616 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2617 {
2618 unsigned long iflags = 0;
2619
2620 if (xflags & FS_XFLAG_SYNC)
2621 iflags |= FS_SYNC_FL;
2622 if (xflags & FS_XFLAG_IMMUTABLE)
2623 iflags |= FS_IMMUTABLE_FL;
2624 if (xflags & FS_XFLAG_APPEND)
2625 iflags |= FS_APPEND_FL;
2626 if (xflags & FS_XFLAG_NODUMP)
2627 iflags |= FS_NODUMP_FL;
2628 if (xflags & FS_XFLAG_NOATIME)
2629 iflags |= FS_NOATIME_FL;
2630 if (xflags & FS_XFLAG_PROJINHERIT)
2631 iflags |= FS_PROJINHERIT_FL;
2632
2633 return iflags;
2634 }
2635
2636 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2637 {
2638 struct inode *inode = file_inode(filp);
2639 struct f2fs_inode_info *fi = F2FS_I(inode);
2640 struct fsxattr fa;
2641
2642 memset(&fa, 0, sizeof(struct fsxattr));
2643 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2644 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2645
2646 if (f2fs_sb_has_project_quota(inode->i_sb))
2647 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2648 fi->i_projid);
2649
2650 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2651 return -EFAULT;
2652 return 0;
2653 }
2654
2655 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2656 {
2657 struct inode *inode = file_inode(filp);
2658 struct f2fs_inode_info *fi = F2FS_I(inode);
2659 struct fsxattr fa;
2660 unsigned int flags;
2661 int err;
2662
2663 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2664 return -EFAULT;
2665
2666 /* Make sure caller has proper permission */
2667 if (!inode_owner_or_capable(inode))
2668 return -EACCES;
2669
2670 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2671 return -EOPNOTSUPP;
2672
2673 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2674 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2675 return -EOPNOTSUPP;
2676
2677 err = mnt_want_write_file(filp);
2678 if (err)
2679 return err;
2680
2681 inode_lock(inode);
2682 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2683 (flags & F2FS_FL_XFLAG_VISIBLE);
2684 err = __f2fs_ioc_setflags(inode, flags);
2685 inode_unlock(inode);
2686 mnt_drop_write_file(filp);
2687 if (err)
2688 return err;
2689
2690 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2691 if (err)
2692 return err;
2693
2694 return 0;
2695 }
2696
2697 int f2fs_pin_file_control(struct inode *inode, bool inc)
2698 {
2699 struct f2fs_inode_info *fi = F2FS_I(inode);
2700 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2701
2702 /* Use i_gc_failures for normal file as a risk signal. */
2703 if (inc)
2704 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
2705
2706 if (fi->i_gc_failures > sbi->gc_pin_file_threshold) {
2707 f2fs_msg(sbi->sb, KERN_WARNING,
2708 "%s: Enable GC = ino %lx after %x GC trials\n",
2709 __func__, inode->i_ino, fi->i_gc_failures);
2710 clear_inode_flag(inode, FI_PIN_FILE);
2711 return -EAGAIN;
2712 }
2713 return 0;
2714 }
2715
2716 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2717 {
2718 struct inode *inode = file_inode(filp);
2719 __u32 pin;
2720 int ret = 0;
2721
2722 if (!inode_owner_or_capable(inode))
2723 return -EACCES;
2724
2725 if (get_user(pin, (__u32 __user *)arg))
2726 return -EFAULT;
2727
2728 if (!S_ISREG(inode->i_mode))
2729 return -EINVAL;
2730
2731 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2732 return -EROFS;
2733
2734 ret = mnt_want_write_file(filp);
2735 if (ret)
2736 return ret;
2737
2738 inode_lock(inode);
2739
2740 if (should_update_outplace(inode, NULL)) {
2741 ret = -EINVAL;
2742 goto out;
2743 }
2744
2745 if (!pin) {
2746 clear_inode_flag(inode, FI_PIN_FILE);
2747 F2FS_I(inode)->i_gc_failures = 1;
2748 goto done;
2749 }
2750
2751 if (f2fs_pin_file_control(inode, false)) {
2752 ret = -EAGAIN;
2753 goto out;
2754 }
2755 ret = f2fs_convert_inline_inode(inode);
2756 if (ret)
2757 goto out;
2758
2759 set_inode_flag(inode, FI_PIN_FILE);
2760 ret = F2FS_I(inode)->i_gc_failures;
2761 done:
2762 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2763 out:
2764 inode_unlock(inode);
2765 mnt_drop_write_file(filp);
2766 return ret;
2767 }
2768
2769 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2770 {
2771 struct inode *inode = file_inode(filp);
2772 __u32 pin = 0;
2773
2774 if (is_inode_flag_set(inode, FI_PIN_FILE))
2775 pin = F2FS_I(inode)->i_gc_failures;
2776 return put_user(pin, (u32 __user *)arg);
2777 }
2778
2779 int f2fs_precache_extents(struct inode *inode)
2780 {
2781 struct f2fs_inode_info *fi = F2FS_I(inode);
2782 struct f2fs_map_blocks map;
2783 pgoff_t m_next_extent;
2784 loff_t end;
2785 int err;
2786
2787 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2788 return -EOPNOTSUPP;
2789
2790 map.m_lblk = 0;
2791 map.m_next_pgofs = NULL;
2792 map.m_next_extent = &m_next_extent;
2793 map.m_seg_type = NO_CHECK_TYPE;
2794 end = F2FS_I_SB(inode)->max_file_blocks;
2795
2796 while (map.m_lblk < end) {
2797 map.m_len = end - map.m_lblk;
2798
2799 down_write(&fi->dio_rwsem[WRITE]);
2800 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2801 up_write(&fi->dio_rwsem[WRITE]);
2802 if (err)
2803 return err;
2804
2805 map.m_lblk = m_next_extent;
2806 }
2807
2808 return err;
2809 }
2810
2811 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2812 {
2813 return f2fs_precache_extents(file_inode(filp));
2814 }
2815
2816 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2817 {
2818 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2819 return -EIO;
2820
2821 switch (cmd) {
2822 case F2FS_IOC_GETFLAGS:
2823 return f2fs_ioc_getflags(filp, arg);
2824 case F2FS_IOC_SETFLAGS:
2825 return f2fs_ioc_setflags(filp, arg);
2826 case F2FS_IOC_GETVERSION:
2827 return f2fs_ioc_getversion(filp, arg);
2828 case F2FS_IOC_START_ATOMIC_WRITE:
2829 return f2fs_ioc_start_atomic_write(filp);
2830 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2831 return f2fs_ioc_commit_atomic_write(filp);
2832 case F2FS_IOC_START_VOLATILE_WRITE:
2833 return f2fs_ioc_start_volatile_write(filp);
2834 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2835 return f2fs_ioc_release_volatile_write(filp);
2836 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2837 return f2fs_ioc_abort_volatile_write(filp);
2838 case F2FS_IOC_SHUTDOWN:
2839 return f2fs_ioc_shutdown(filp, arg);
2840 case FITRIM:
2841 return f2fs_ioc_fitrim(filp, arg);
2842 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2843 return f2fs_ioc_set_encryption_policy(filp, arg);
2844 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2845 return f2fs_ioc_get_encryption_policy(filp, arg);
2846 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2847 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2848 case F2FS_IOC_GARBAGE_COLLECT:
2849 return f2fs_ioc_gc(filp, arg);
2850 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2851 return f2fs_ioc_gc_range(filp, arg);
2852 case F2FS_IOC_WRITE_CHECKPOINT:
2853 return f2fs_ioc_write_checkpoint(filp, arg);
2854 case F2FS_IOC_DEFRAGMENT:
2855 return f2fs_ioc_defragment(filp, arg);
2856 case F2FS_IOC_MOVE_RANGE:
2857 return f2fs_ioc_move_range(filp, arg);
2858 case F2FS_IOC_FLUSH_DEVICE:
2859 return f2fs_ioc_flush_device(filp, arg);
2860 case F2FS_IOC_GET_FEATURES:
2861 return f2fs_ioc_get_features(filp, arg);
2862 case F2FS_IOC_FSGETXATTR:
2863 return f2fs_ioc_fsgetxattr(filp, arg);
2864 case F2FS_IOC_FSSETXATTR:
2865 return f2fs_ioc_fssetxattr(filp, arg);
2866 case F2FS_IOC_GET_PIN_FILE:
2867 return f2fs_ioc_get_pin_file(filp, arg);
2868 case F2FS_IOC_SET_PIN_FILE:
2869 return f2fs_ioc_set_pin_file(filp, arg);
2870 case F2FS_IOC_PRECACHE_EXTENTS:
2871 return f2fs_ioc_precache_extents(filp, arg);
2872 default:
2873 return -ENOTTY;
2874 }
2875 }
2876
2877 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2878 {
2879 struct file *file = iocb->ki_filp;
2880 struct inode *inode = file_inode(file);
2881 struct blk_plug plug;
2882 ssize_t ret;
2883
2884 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2885 return -EIO;
2886
2887 inode_lock(inode);
2888 ret = generic_write_checks(iocb, from);
2889 if (ret > 0) {
2890 int err;
2891
2892 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2893 set_inode_flag(inode, FI_NO_PREALLOC);
2894
2895 err = f2fs_preallocate_blocks(iocb, from);
2896 if (err) {
2897 clear_inode_flag(inode, FI_NO_PREALLOC);
2898 inode_unlock(inode);
2899 return err;
2900 }
2901 blk_start_plug(&plug);
2902 ret = __generic_file_write_iter(iocb, from);
2903 blk_finish_plug(&plug);
2904 clear_inode_flag(inode, FI_NO_PREALLOC);
2905
2906 if (ret > 0)
2907 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2908 }
2909 inode_unlock(inode);
2910
2911 if (ret > 0)
2912 ret = generic_write_sync(iocb, ret);
2913 return ret;
2914 }
2915
2916 #ifdef CONFIG_COMPAT
2917 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2918 {
2919 switch (cmd) {
2920 case F2FS_IOC32_GETFLAGS:
2921 cmd = F2FS_IOC_GETFLAGS;
2922 break;
2923 case F2FS_IOC32_SETFLAGS:
2924 cmd = F2FS_IOC_SETFLAGS;
2925 break;
2926 case F2FS_IOC32_GETVERSION:
2927 cmd = F2FS_IOC_GETVERSION;
2928 break;
2929 case F2FS_IOC_START_ATOMIC_WRITE:
2930 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2931 case F2FS_IOC_START_VOLATILE_WRITE:
2932 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2933 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2934 case F2FS_IOC_SHUTDOWN:
2935 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2936 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2937 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2938 case F2FS_IOC_GARBAGE_COLLECT:
2939 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2940 case F2FS_IOC_WRITE_CHECKPOINT:
2941 case F2FS_IOC_DEFRAGMENT:
2942 case F2FS_IOC_MOVE_RANGE:
2943 case F2FS_IOC_FLUSH_DEVICE:
2944 case F2FS_IOC_GET_FEATURES:
2945 case F2FS_IOC_FSGETXATTR:
2946 case F2FS_IOC_FSSETXATTR:
2947 case F2FS_IOC_GET_PIN_FILE:
2948 case F2FS_IOC_SET_PIN_FILE:
2949 case F2FS_IOC_PRECACHE_EXTENTS:
2950 break;
2951 default:
2952 return -ENOIOCTLCMD;
2953 }
2954 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2955 }
2956 #endif
2957
2958 const struct file_operations f2fs_file_operations = {
2959 .llseek = f2fs_llseek,
2960 .read_iter = generic_file_read_iter,
2961 .write_iter = f2fs_file_write_iter,
2962 .open = f2fs_file_open,
2963 .release = f2fs_release_file,
2964 .mmap = f2fs_file_mmap,
2965 .flush = f2fs_file_flush,
2966 .fsync = f2fs_sync_file,
2967 .fallocate = f2fs_fallocate,
2968 .unlocked_ioctl = f2fs_ioctl,
2969 #ifdef CONFIG_COMPAT
2970 .compat_ioctl = f2fs_compat_ioctl,
2971 #endif
2972 .splice_read = generic_file_splice_read,
2973 .splice_write = iter_file_splice_write,
2974 };