projects / GitHub / MotorolaMobilityLLC / kernel-slsi.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge 4.14.73 into android-4.14-p
[GitHub/MotorolaMobilityLLC/kernel-slsi.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 page writeback via META_MAPPING */
114 if (f2fs_post_read_required(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 (F2FS_OPTION(sbi).active_logs == 2)
167 cp_reason = CP_SPEC_LOG_NUM;
168 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
169 need_dentry_mark(sbi, inode->i_ino) &&
170 exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
171 cp_reason = CP_RECOVER_DIR;
172
173 return cp_reason;
174 }
175
176 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
177 {
178 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
179 bool ret = false;
180 /* But we need to avoid that there are some inode updates */
181 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
182 ret = true;
183 f2fs_put_page(i, 0);
184 return ret;
185 }
186
187 static void try_to_fix_pino(struct inode *inode)
188 {
189 struct f2fs_inode_info *fi = F2FS_I(inode);
190 nid_t pino;
191
192 down_write(&fi->i_sem);
193 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
194 get_parent_ino(inode, &pino)) {
195 f2fs_i_pino_write(inode, pino);
196 file_got_pino(inode);
197 }
198 up_write(&fi->i_sem);
199 }
200
201 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
202 int datasync, bool atomic)
203 {
204 struct inode *inode = file->f_mapping->host;
205 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
206 nid_t ino = inode->i_ino;
207 int ret = 0;
208 enum cp_reason_type cp_reason = 0;
209 struct writeback_control wbc = {
210 .sync_mode = WB_SYNC_ALL,
211 .nr_to_write = LONG_MAX,
212 .for_reclaim = 0,
213 };
214
215 if (unlikely(f2fs_readonly(inode->i_sb)))
216 return 0;
217
218 trace_f2fs_sync_file_enter(inode);
219
220 /* if fdatasync is triggered, let's do in-place-update */
221 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
222 set_inode_flag(inode, FI_NEED_IPU);
223 ret = file_write_and_wait_range(file, start, end);
224 clear_inode_flag(inode, FI_NEED_IPU);
225
226 if (ret) {
227 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
228 return ret;
229 }
230
231 /* if the inode is dirty, let's recover all the time */
232 if (!f2fs_skip_inode_update(inode, datasync)) {
233 f2fs_write_inode(inode, NULL);
234 goto go_write;
235 }
236
237 /*
238 * if there is no written data, don't waste time to write recovery info.
239 */
240 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
241 !exist_written_data(sbi, ino, APPEND_INO)) {
242
243 /* it may call write_inode just prior to fsync */
244 if (need_inode_page_update(sbi, ino))
245 goto go_write;
246
247 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
248 exist_written_data(sbi, ino, UPDATE_INO))
249 goto flush_out;
250 goto out;
251 }
252 go_write:
253 /*
254 * Both of fdatasync() and fsync() are able to be recovered from
255 * sudden-power-off.
256 */
257 down_read(&F2FS_I(inode)->i_sem);
258 cp_reason = need_do_checkpoint(inode);
259 up_read(&F2FS_I(inode)->i_sem);
260
261 if (cp_reason) {
262 /* all the dirty node pages should be flushed for POR */
263 ret = f2fs_sync_fs(inode->i_sb, 1);
264
265 /*
266 * We've secured consistency through sync_fs. Following pino
267 * will be used only for fsynced inodes after checkpoint.
268 */
269 try_to_fix_pino(inode);
270 clear_inode_flag(inode, FI_APPEND_WRITE);
271 clear_inode_flag(inode, FI_UPDATE_WRITE);
272 goto out;
273 }
274 sync_nodes:
275 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
276 if (ret)
277 goto out;
278
279 /* if cp_error was enabled, we should avoid infinite loop */
280 if (unlikely(f2fs_cp_error(sbi))) {
281 ret = -EIO;
282 goto out;
283 }
284
285 if (need_inode_block_update(sbi, ino)) {
286 f2fs_mark_inode_dirty_sync(inode, true);
287 f2fs_write_inode(inode, NULL);
288 goto sync_nodes;
289 }
290
291 /*
292 * If it's atomic_write, it's just fine to keep write ordering. So
293 * here we don't need to wait for node write completion, since we use
294 * node chain which serializes node blocks. If one of node writes are
295 * reordered, we can see simply broken chain, resulting in stopping
296 * roll-forward recovery. It means we'll recover all or none node blocks
297 * given fsync mark.
298 */
299 if (!atomic) {
300 ret = wait_on_node_pages_writeback(sbi, ino);
301 if (ret)
302 goto out;
303 }
304
305 /* once recovery info is written, don't need to tack this */
306 remove_ino_entry(sbi, ino, APPEND_INO);
307 clear_inode_flag(inode, FI_APPEND_WRITE);
308 flush_out:
309 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
310 ret = f2fs_issue_flush(sbi, inode->i_ino);
311 if (!ret) {
312 remove_ino_entry(sbi, ino, UPDATE_INO);
313 clear_inode_flag(inode, FI_UPDATE_WRITE);
314 remove_ino_entry(sbi, ino, FLUSH_INO);
315 }
316 f2fs_update_time(sbi, REQ_TIME);
317 out:
318 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
319 f2fs_trace_ios(NULL, 1);
320 return ret;
321 }
322
323 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
324 {
325 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
326 return -EIO;
327 return f2fs_do_sync_file(file, start, end, datasync, false);
328 }
329
330 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
331 pgoff_t pgofs, int whence)
332 {
333 struct pagevec pvec;
334 int nr_pages;
335
336 if (whence != SEEK_DATA)
337 return 0;
338
339 /* find first dirty page index */
340 pagevec_init(&pvec, 0);
341 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
342 PAGECACHE_TAG_DIRTY, 1);
343 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
344 pagevec_release(&pvec);
345 return pgofs;
346 }
347
348 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
349 int whence)
350 {
351 switch (whence) {
352 case SEEK_DATA:
353 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
354 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
355 return true;
356 break;
357 case SEEK_HOLE:
358 if (blkaddr == NULL_ADDR)
359 return true;
360 break;
361 }
362 return false;
363 }
364
365 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
366 {
367 struct inode *inode = file->f_mapping->host;
368 loff_t maxbytes = inode->i_sb->s_maxbytes;
369 struct dnode_of_data dn;
370 pgoff_t pgofs, end_offset, dirty;
371 loff_t data_ofs = offset;
372 loff_t isize;
373 int err = 0;
374
375 inode_lock(inode);
376
377 isize = i_size_read(inode);
378 if (offset >= isize)
379 goto fail;
380
381 /* handle inline data case */
382 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
383 if (whence == SEEK_HOLE)
384 data_ofs = isize;
385 goto found;
386 }
387
388 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
389
390 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
391
392 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
393 set_new_dnode(&dn, inode, NULL, NULL, 0);
394 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
395 if (err && err != -ENOENT) {
396 goto fail;
397 } else if (err == -ENOENT) {
398 /* direct node does not exists */
399 if (whence == SEEK_DATA) {
400 pgofs = get_next_page_offset(&dn, pgofs);
401 continue;
402 } else {
403 goto found;
404 }
405 }
406
407 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
408
409 /* find data/hole in dnode block */
410 for (; dn.ofs_in_node < end_offset;
411 dn.ofs_in_node++, pgofs++,
412 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
413 block_t blkaddr;
414 blkaddr = datablock_addr(dn.inode,
415 dn.node_page, dn.ofs_in_node);
416
417 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
418 f2fs_put_dnode(&dn);
419 goto found;
420 }
421 }
422 f2fs_put_dnode(&dn);
423 }
424
425 if (whence == SEEK_DATA)
426 goto fail;
427 found:
428 if (whence == SEEK_HOLE && data_ofs > isize)
429 data_ofs = isize;
430 inode_unlock(inode);
431 return vfs_setpos(file, data_ofs, maxbytes);
432 fail:
433 inode_unlock(inode);
434 return -ENXIO;
435 }
436
437 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
438 {
439 struct inode *inode = file->f_mapping->host;
440 loff_t maxbytes = inode->i_sb->s_maxbytes;
441
442 switch (whence) {
443 case SEEK_SET:
444 case SEEK_CUR:
445 case SEEK_END:
446 return generic_file_llseek_size(file, offset, whence,
447 maxbytes, i_size_read(inode));
448 case SEEK_DATA:
449 case SEEK_HOLE:
450 if (offset < 0)
451 return -ENXIO;
452 return f2fs_seek_block(file, offset, whence);
453 }
454
455 return -EINVAL;
456 }
457
458 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
459 {
460 struct inode *inode = file_inode(file);
461 int err;
462
463 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
464 return -EIO;
465
466 /* we don't need to use inline_data strictly */
467 err = f2fs_convert_inline_inode(inode);
468 if (err)
469 return err;
470
471 file_accessed(file);
472 vma->vm_ops = &f2fs_file_vm_ops;
473 return 0;
474 }
475
476 static int f2fs_file_open(struct inode *inode, struct file *filp)
477 {
478 int err = fscrypt_file_open(inode, filp);
479
480 if (err)
481 return err;
482
483 filp->f_mode |= FMODE_NOWAIT;
484
485 return dquot_file_open(inode, filp);
486 }
487
488 void truncate_data_blocks_range(struct dnode_of_data *dn, int count)
489 {
490 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
491 struct f2fs_node *raw_node;
492 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
493 __le32 *addr;
494 int base = 0;
495
496 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
497 base = get_extra_isize(dn->inode);
498
499 raw_node = F2FS_NODE(dn->node_page);
500 addr = blkaddr_in_node(raw_node) + base + ofs;
501
502 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
503 block_t blkaddr = le32_to_cpu(*addr);
504 if (blkaddr == NULL_ADDR)
505 continue;
506
507 dn->data_blkaddr = NULL_ADDR;
508 set_data_blkaddr(dn);
509 invalidate_blocks(sbi, blkaddr);
510 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
511 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
512 nr_free++;
513 }
514
515 if (nr_free) {
516 pgoff_t fofs;
517 /*
518 * once we invalidate valid blkaddr in range [ofs, ofs + count],
519 * we will invalidate all blkaddr in the whole range.
520 */
521 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
522 dn->inode) + ofs;
523 f2fs_update_extent_cache_range(dn, fofs, 0, len);
524 dec_valid_block_count(sbi, dn->inode, nr_free);
525 }
526 dn->ofs_in_node = ofs;
527
528 f2fs_update_time(sbi, REQ_TIME);
529 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
530 dn->ofs_in_node, nr_free);
531 }
532
533 void truncate_data_blocks(struct dnode_of_data *dn)
534 {
535 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
536 }
537
538 static int truncate_partial_data_page(struct inode *inode, u64 from,
539 bool cache_only)
540 {
541 unsigned offset = from & (PAGE_SIZE - 1);
542 pgoff_t index = from >> PAGE_SHIFT;
543 struct address_space *mapping = inode->i_mapping;
544 struct page *page;
545
546 if (!offset && !cache_only)
547 return 0;
548
549 if (cache_only) {
550 page = find_lock_page(mapping, index);
551 if (page && PageUptodate(page))
552 goto truncate_out;
553 f2fs_put_page(page, 1);
554 return 0;
555 }
556
557 page = get_lock_data_page(inode, index, true);
558 if (IS_ERR(page))
559 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
560 truncate_out:
561 f2fs_wait_on_page_writeback(page, DATA, true);
562 zero_user(page, offset, PAGE_SIZE - offset);
563
564 /* An encrypted inode should have a key and truncate the last page. */
565 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
566 if (!cache_only)
567 set_page_dirty(page);
568 f2fs_put_page(page, 1);
569 return 0;
570 }
571
572 int truncate_blocks(struct inode *inode, u64 from, bool lock)
573 {
574 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
575 struct dnode_of_data dn;
576 pgoff_t free_from;
577 int count = 0, err = 0;
578 struct page *ipage;
579 bool truncate_page = false;
580
581 trace_f2fs_truncate_blocks_enter(inode, from);
582
583 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
584
585 if (free_from >= sbi->max_file_blocks)
586 goto free_partial;
587
588 if (lock)
589 f2fs_lock_op(sbi);
590
591 ipage = get_node_page(sbi, inode->i_ino);
592 if (IS_ERR(ipage)) {
593 err = PTR_ERR(ipage);
594 goto out;
595 }
596
597 if (f2fs_has_inline_data(inode)) {
598 truncate_inline_inode(inode, ipage, from);
599 f2fs_put_page(ipage, 1);
600 truncate_page = true;
601 goto out;
602 }
603
604 set_new_dnode(&dn, inode, ipage, NULL, 0);
605 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
606 if (err) {
607 if (err == -ENOENT)
608 goto free_next;
609 goto out;
610 }
611
612 count = ADDRS_PER_PAGE(dn.node_page, inode);
613
614 count -= dn.ofs_in_node;
615 f2fs_bug_on(sbi, count < 0);
616
617 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
618 truncate_data_blocks_range(&dn, count);
619 free_from += count;
620 }
621
622 f2fs_put_dnode(&dn);
623 free_next:
624 err = truncate_inode_blocks(inode, free_from);
625 out:
626 if (lock)
627 f2fs_unlock_op(sbi);
628 free_partial:
629 /* lastly zero out the first data page */
630 if (!err)
631 err = truncate_partial_data_page(inode, from, truncate_page);
632
633 trace_f2fs_truncate_blocks_exit(inode, err);
634 return err;
635 }
636
637 int f2fs_truncate(struct inode *inode)
638 {
639 int err;
640
641 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
642 return -EIO;
643
644 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
645 S_ISLNK(inode->i_mode)))
646 return 0;
647
648 trace_f2fs_truncate(inode);
649
650 #ifdef CONFIG_F2FS_FAULT_INJECTION
651 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
652 f2fs_show_injection_info(FAULT_TRUNCATE);
653 return -EIO;
654 }
655 #endif
656 /* we should check inline_data size */
657 if (!f2fs_may_inline_data(inode)) {
658 err = f2fs_convert_inline_inode(inode);
659 if (err)
660 return err;
661 }
662
663 err = truncate_blocks(inode, i_size_read(inode), true);
664 if (err)
665 return err;
666
667 inode->i_mtime = inode->i_ctime = current_time(inode);
668 f2fs_mark_inode_dirty_sync(inode, false);
669 return 0;
670 }
671
672 int f2fs_getattr(const struct path *path, struct kstat *stat,
673 u32 request_mask, unsigned int query_flags)
674 {
675 struct inode *inode = d_inode(path->dentry);
676 struct f2fs_inode_info *fi = F2FS_I(inode);
677 struct f2fs_inode *ri;
678 unsigned int flags;
679
680 if (f2fs_has_extra_attr(inode) &&
681 f2fs_sb_has_inode_crtime(inode->i_sb) &&
682 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
683 stat->result_mask |= STATX_BTIME;
684 stat->btime.tv_sec = fi->i_crtime.tv_sec;
685 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
686 }
687
688 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
689 if (flags & FS_APPEND_FL)
690 stat->attributes |= STATX_ATTR_APPEND;
691 if (flags & FS_COMPR_FL)
692 stat->attributes |= STATX_ATTR_COMPRESSED;
693 if (f2fs_encrypted_inode(inode))
694 stat->attributes |= STATX_ATTR_ENCRYPTED;
695 if (flags & FS_IMMUTABLE_FL)
696 stat->attributes |= STATX_ATTR_IMMUTABLE;
697 if (flags & FS_NODUMP_FL)
698 stat->attributes |= STATX_ATTR_NODUMP;
699
700 stat->attributes_mask |= (STATX_ATTR_APPEND |
701 STATX_ATTR_COMPRESSED |
702 STATX_ATTR_ENCRYPTED |
703 STATX_ATTR_IMMUTABLE |
704 STATX_ATTR_NODUMP);
705
706 generic_fillattr(inode, stat);
707
708 /* we need to show initial sectors used for inline_data/dentries */
709 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
710 f2fs_has_inline_dentry(inode))
711 stat->blocks += (stat->size + 511) >> 9;
712
713 return 0;
714 }
715
716 #ifdef CONFIG_F2FS_FS_POSIX_ACL
717 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
718 {
719 unsigned int ia_valid = attr->ia_valid;
720
721 if (ia_valid & ATTR_UID)
722 inode->i_uid = attr->ia_uid;
723 if (ia_valid & ATTR_GID)
724 inode->i_gid = attr->ia_gid;
725 if (ia_valid & ATTR_ATIME)
726 inode->i_atime = timespec_trunc(attr->ia_atime,
727 inode->i_sb->s_time_gran);
728 if (ia_valid & ATTR_MTIME)
729 inode->i_mtime = timespec_trunc(attr->ia_mtime,
730 inode->i_sb->s_time_gran);
731 if (ia_valid & ATTR_CTIME)
732 inode->i_ctime = timespec_trunc(attr->ia_ctime,
733 inode->i_sb->s_time_gran);
734 if (ia_valid & ATTR_MODE) {
735 umode_t mode = attr->ia_mode;
736
737 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
738 mode &= ~S_ISGID;
739 set_acl_inode(inode, mode);
740 }
741 }
742 #else
743 #define __setattr_copy setattr_copy
744 #endif
745
746 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
747 {
748 struct inode *inode = d_inode(dentry);
749 int err;
750 bool size_changed = false;
751
752 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
753 return -EIO;
754
755 err = setattr_prepare(dentry, attr);
756 if (err)
757 return err;
758
759 err = fscrypt_prepare_setattr(dentry, attr);
760 if (err)
761 return err;
762
763 if (is_quota_modification(inode, attr)) {
764 err = dquot_initialize(inode);
765 if (err)
766 return err;
767 }
768 if ((attr->ia_valid & ATTR_UID &&
769 !uid_eq(attr->ia_uid, inode->i_uid)) ||
770 (attr->ia_valid & ATTR_GID &&
771 !gid_eq(attr->ia_gid, inode->i_gid))) {
772 err = dquot_transfer(inode, attr);
773 if (err)
774 return err;
775 }
776
777 if (attr->ia_valid & ATTR_SIZE) {
778 if (attr->ia_size <= i_size_read(inode)) {
779 down_write(&F2FS_I(inode)->i_mmap_sem);
780 truncate_setsize(inode, attr->ia_size);
781 err = f2fs_truncate(inode);
782 up_write(&F2FS_I(inode)->i_mmap_sem);
783 if (err)
784 return err;
785 } else {
786 /*
787 * do not trim all blocks after i_size if target size is
788 * larger than i_size.
789 */
790 down_write(&F2FS_I(inode)->i_mmap_sem);
791 truncate_setsize(inode, attr->ia_size);
792 up_write(&F2FS_I(inode)->i_mmap_sem);
793
794 /* should convert inline inode here */
795 if (!f2fs_may_inline_data(inode)) {
796 err = f2fs_convert_inline_inode(inode);
797 if (err)
798 return err;
799 }
800 inode->i_mtime = inode->i_ctime = current_time(inode);
801 }
802
803 down_write(&F2FS_I(inode)->i_sem);
804 F2FS_I(inode)->last_disk_size = i_size_read(inode);
805 up_write(&F2FS_I(inode)->i_sem);
806
807 size_changed = true;
808 }
809
810 __setattr_copy(inode, attr);
811
812 if (attr->ia_valid & ATTR_MODE) {
813 err = posix_acl_chmod(inode, get_inode_mode(inode));
814 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
815 inode->i_mode = F2FS_I(inode)->i_acl_mode;
816 clear_inode_flag(inode, FI_ACL_MODE);
817 }
818 }
819
820 /* file size may changed here */
821 f2fs_mark_inode_dirty_sync(inode, size_changed);
822
823 /* inode change will produce dirty node pages flushed by checkpoint */
824 f2fs_balance_fs(F2FS_I_SB(inode), true);
825
826 return err;
827 }
828
829 const struct inode_operations f2fs_file_inode_operations = {
830 .getattr = f2fs_getattr,
831 .setattr = f2fs_setattr,
832 .get_acl = f2fs_get_acl,
833 .set_acl = f2fs_set_acl,
834 #ifdef CONFIG_F2FS_FS_XATTR
835 .listxattr = f2fs_listxattr,
836 #endif
837 .fiemap = f2fs_fiemap,
838 };
839
840 static int fill_zero(struct inode *inode, pgoff_t index,
841 loff_t start, loff_t len)
842 {
843 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
844 struct page *page;
845
846 if (!len)
847 return 0;
848
849 f2fs_balance_fs(sbi, true);
850
851 f2fs_lock_op(sbi);
852 page = get_new_data_page(inode, NULL, index, false);
853 f2fs_unlock_op(sbi);
854
855 if (IS_ERR(page))
856 return PTR_ERR(page);
857
858 f2fs_wait_on_page_writeback(page, DATA, true);
859 zero_user(page, start, len);
860 set_page_dirty(page);
861 f2fs_put_page(page, 1);
862 return 0;
863 }
864
865 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
866 {
867 int err;
868
869 while (pg_start < pg_end) {
870 struct dnode_of_data dn;
871 pgoff_t end_offset, count;
872
873 set_new_dnode(&dn, inode, NULL, NULL, 0);
874 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
875 if (err) {
876 if (err == -ENOENT) {
877 pg_start = get_next_page_offset(&dn, pg_start);
878 continue;
879 }
880 return err;
881 }
882
883 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
884 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
885
886 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
887
888 truncate_data_blocks_range(&dn, count);
889 f2fs_put_dnode(&dn);
890
891 pg_start += count;
892 }
893 return 0;
894 }
895
896 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
897 {
898 pgoff_t pg_start, pg_end;
899 loff_t off_start, off_end;
900 int ret;
901
902 ret = f2fs_convert_inline_inode(inode);
903 if (ret)
904 return ret;
905
906 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
907 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
908
909 off_start = offset & (PAGE_SIZE - 1);
910 off_end = (offset + len) & (PAGE_SIZE - 1);
911
912 if (pg_start == pg_end) {
913 ret = fill_zero(inode, pg_start, off_start,
914 off_end - off_start);
915 if (ret)
916 return ret;
917 } else {
918 if (off_start) {
919 ret = fill_zero(inode, pg_start++, off_start,
920 PAGE_SIZE - off_start);
921 if (ret)
922 return ret;
923 }
924 if (off_end) {
925 ret = fill_zero(inode, pg_end, 0, off_end);
926 if (ret)
927 return ret;
928 }
929
930 if (pg_start < pg_end) {
931 struct address_space *mapping = inode->i_mapping;
932 loff_t blk_start, blk_end;
933 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
934
935 f2fs_balance_fs(sbi, true);
936
937 blk_start = (loff_t)pg_start << PAGE_SHIFT;
938 blk_end = (loff_t)pg_end << PAGE_SHIFT;
939 down_write(&F2FS_I(inode)->i_mmap_sem);
940 truncate_inode_pages_range(mapping, blk_start,
941 blk_end - 1);
942
943 f2fs_lock_op(sbi);
944 ret = truncate_hole(inode, pg_start, pg_end);
945 f2fs_unlock_op(sbi);
946 up_write(&F2FS_I(inode)->i_mmap_sem);
947 }
948 }
949
950 return ret;
951 }
952
953 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
954 int *do_replace, pgoff_t off, pgoff_t len)
955 {
956 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
957 struct dnode_of_data dn;
958 int ret, done, i;
959
960 next_dnode:
961 set_new_dnode(&dn, inode, NULL, NULL, 0);
962 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
963 if (ret && ret != -ENOENT) {
964 return ret;
965 } else if (ret == -ENOENT) {
966 if (dn.max_level == 0)
967 return -ENOENT;
968 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
969 blkaddr += done;
970 do_replace += done;
971 goto next;
972 }
973
974 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
975 dn.ofs_in_node, len);
976 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
977 *blkaddr = datablock_addr(dn.inode,
978 dn.node_page, dn.ofs_in_node);
979 if (!is_checkpointed_data(sbi, *blkaddr)) {
980
981 if (test_opt(sbi, LFS)) {
982 f2fs_put_dnode(&dn);
983 return -ENOTSUPP;
984 }
985
986 /* do not invalidate this block address */
987 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
988 *do_replace = 1;
989 }
990 }
991 f2fs_put_dnode(&dn);
992 next:
993 len -= done;
994 off += done;
995 if (len)
996 goto next_dnode;
997 return 0;
998 }
999
1000 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1001 int *do_replace, pgoff_t off, int len)
1002 {
1003 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1004 struct dnode_of_data dn;
1005 int ret, i;
1006
1007 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1008 if (*do_replace == 0)
1009 continue;
1010
1011 set_new_dnode(&dn, inode, NULL, NULL, 0);
1012 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1013 if (ret) {
1014 dec_valid_block_count(sbi, inode, 1);
1015 invalidate_blocks(sbi, *blkaddr);
1016 } else {
1017 f2fs_update_data_blkaddr(&dn, *blkaddr);
1018 }
1019 f2fs_put_dnode(&dn);
1020 }
1021 return 0;
1022 }
1023
1024 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1025 block_t *blkaddr, int *do_replace,
1026 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1027 {
1028 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1029 pgoff_t i = 0;
1030 int ret;
1031
1032 while (i < len) {
1033 if (blkaddr[i] == NULL_ADDR && !full) {
1034 i++;
1035 continue;
1036 }
1037
1038 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1039 struct dnode_of_data dn;
1040 struct node_info ni;
1041 size_t new_size;
1042 pgoff_t ilen;
1043
1044 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1045 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1046 if (ret)
1047 return ret;
1048
1049 get_node_info(sbi, dn.nid, &ni);
1050 ilen = min((pgoff_t)
1051 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1052 dn.ofs_in_node, len - i);
1053 do {
1054 dn.data_blkaddr = datablock_addr(dn.inode,
1055 dn.node_page, dn.ofs_in_node);
1056 truncate_data_blocks_range(&dn, 1);
1057
1058 if (do_replace[i]) {
1059 f2fs_i_blocks_write(src_inode,
1060 1, false, false);
1061 f2fs_i_blocks_write(dst_inode,
1062 1, true, false);
1063 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1064 blkaddr[i], ni.version, true, false);
1065
1066 do_replace[i] = 0;
1067 }
1068 dn.ofs_in_node++;
1069 i++;
1070 new_size = (dst + i) << PAGE_SHIFT;
1071 if (dst_inode->i_size < new_size)
1072 f2fs_i_size_write(dst_inode, new_size);
1073 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1074
1075 f2fs_put_dnode(&dn);
1076 } else {
1077 struct page *psrc, *pdst;
1078
1079 psrc = get_lock_data_page(src_inode, src + i, true);
1080 if (IS_ERR(psrc))
1081 return PTR_ERR(psrc);
1082 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1083 true);
1084 if (IS_ERR(pdst)) {
1085 f2fs_put_page(psrc, 1);
1086 return PTR_ERR(pdst);
1087 }
1088 f2fs_copy_page(psrc, pdst);
1089 set_page_dirty(pdst);
1090 f2fs_put_page(pdst, 1);
1091 f2fs_put_page(psrc, 1);
1092
1093 ret = truncate_hole(src_inode, src + i, src + i + 1);
1094 if (ret)
1095 return ret;
1096 i++;
1097 }
1098 }
1099 return 0;
1100 }
1101
1102 static int __exchange_data_block(struct inode *src_inode,
1103 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1104 pgoff_t len, bool full)
1105 {
1106 block_t *src_blkaddr;
1107 int *do_replace;
1108 pgoff_t olen;
1109 int ret;
1110
1111 while (len) {
1112 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1113
1114 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1115 sizeof(block_t) * olen, GFP_KERNEL);
1116 if (!src_blkaddr)
1117 return -ENOMEM;
1118
1119 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1120 sizeof(int) * olen, GFP_KERNEL);
1121 if (!do_replace) {
1122 kvfree(src_blkaddr);
1123 return -ENOMEM;
1124 }
1125
1126 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1127 do_replace, src, olen);
1128 if (ret)
1129 goto roll_back;
1130
1131 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1132 do_replace, src, dst, olen, full);
1133 if (ret)
1134 goto roll_back;
1135
1136 src += olen;
1137 dst += olen;
1138 len -= olen;
1139
1140 kvfree(src_blkaddr);
1141 kvfree(do_replace);
1142 }
1143 return 0;
1144
1145 roll_back:
1146 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1147 kvfree(src_blkaddr);
1148 kvfree(do_replace);
1149 return ret;
1150 }
1151
1152 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1153 {
1154 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1155 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1156 int ret;
1157
1158 f2fs_balance_fs(sbi, true);
1159 f2fs_lock_op(sbi);
1160
1161 f2fs_drop_extent_tree(inode);
1162
1163 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1164 f2fs_unlock_op(sbi);
1165 return ret;
1166 }
1167
1168 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1169 {
1170 pgoff_t pg_start, pg_end;
1171 loff_t new_size;
1172 int ret;
1173
1174 if (offset + len >= i_size_read(inode))
1175 return -EINVAL;
1176
1177 /* collapse range should be aligned to block size of f2fs. */
1178 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1179 return -EINVAL;
1180
1181 ret = f2fs_convert_inline_inode(inode);
1182 if (ret)
1183 return ret;
1184
1185 pg_start = offset >> PAGE_SHIFT;
1186 pg_end = (offset + len) >> PAGE_SHIFT;
1187
1188 /* avoid gc operation during block exchange */
1189 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1190
1191 down_write(&F2FS_I(inode)->i_mmap_sem);
1192 /* write out all dirty pages from offset */
1193 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1194 if (ret)
1195 goto out_unlock;
1196
1197 truncate_pagecache(inode, offset);
1198
1199 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1200 if (ret)
1201 goto out_unlock;
1202
1203 /* write out all moved pages, if possible */
1204 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1205 truncate_pagecache(inode, offset);
1206
1207 new_size = i_size_read(inode) - len;
1208 truncate_pagecache(inode, new_size);
1209
1210 ret = truncate_blocks(inode, new_size, true);
1211 if (!ret)
1212 f2fs_i_size_write(inode, new_size);
1213 out_unlock:
1214 up_write(&F2FS_I(inode)->i_mmap_sem);
1215 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1216 return ret;
1217 }
1218
1219 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1220 pgoff_t end)
1221 {
1222 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1223 pgoff_t index = start;
1224 unsigned int ofs_in_node = dn->ofs_in_node;
1225 blkcnt_t count = 0;
1226 int ret;
1227
1228 for (; index < end; index++, dn->ofs_in_node++) {
1229 if (datablock_addr(dn->inode, dn->node_page,
1230 dn->ofs_in_node) == NULL_ADDR)
1231 count++;
1232 }
1233
1234 dn->ofs_in_node = ofs_in_node;
1235 ret = reserve_new_blocks(dn, count);
1236 if (ret)
1237 return ret;
1238
1239 dn->ofs_in_node = ofs_in_node;
1240 for (index = start; index < end; index++, dn->ofs_in_node++) {
1241 dn->data_blkaddr = datablock_addr(dn->inode,
1242 dn->node_page, dn->ofs_in_node);
1243 /*
1244 * reserve_new_blocks will not guarantee entire block
1245 * allocation.
1246 */
1247 if (dn->data_blkaddr == NULL_ADDR) {
1248 ret = -ENOSPC;
1249 break;
1250 }
1251 if (dn->data_blkaddr != NEW_ADDR) {
1252 invalidate_blocks(sbi, dn->data_blkaddr);
1253 dn->data_blkaddr = NEW_ADDR;
1254 set_data_blkaddr(dn);
1255 }
1256 }
1257
1258 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1259
1260 return ret;
1261 }
1262
1263 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1264 int mode)
1265 {
1266 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1267 struct address_space *mapping = inode->i_mapping;
1268 pgoff_t index, pg_start, pg_end;
1269 loff_t new_size = i_size_read(inode);
1270 loff_t off_start, off_end;
1271 int ret = 0;
1272
1273 ret = inode_newsize_ok(inode, (len + offset));
1274 if (ret)
1275 return ret;
1276
1277 ret = f2fs_convert_inline_inode(inode);
1278 if (ret)
1279 return ret;
1280
1281 down_write(&F2FS_I(inode)->i_mmap_sem);
1282 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1283 if (ret)
1284 goto out_sem;
1285
1286 truncate_pagecache_range(inode, offset, offset + len - 1);
1287
1288 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1289 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1290
1291 off_start = offset & (PAGE_SIZE - 1);
1292 off_end = (offset + len) & (PAGE_SIZE - 1);
1293
1294 if (pg_start == pg_end) {
1295 ret = fill_zero(inode, pg_start, off_start,
1296 off_end - off_start);
1297 if (ret)
1298 goto out_sem;
1299
1300 new_size = max_t(loff_t, new_size, offset + len);
1301 } else {
1302 if (off_start) {
1303 ret = fill_zero(inode, pg_start++, off_start,
1304 PAGE_SIZE - off_start);
1305 if (ret)
1306 goto out_sem;
1307
1308 new_size = max_t(loff_t, new_size,
1309 (loff_t)pg_start << PAGE_SHIFT);
1310 }
1311
1312 for (index = pg_start; index < pg_end;) {
1313 struct dnode_of_data dn;
1314 unsigned int end_offset;
1315 pgoff_t end;
1316
1317 f2fs_lock_op(sbi);
1318
1319 set_new_dnode(&dn, inode, NULL, NULL, 0);
1320 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1321 if (ret) {
1322 f2fs_unlock_op(sbi);
1323 goto out;
1324 }
1325
1326 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1327 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1328
1329 ret = f2fs_do_zero_range(&dn, index, end);
1330 f2fs_put_dnode(&dn);
1331 f2fs_unlock_op(sbi);
1332
1333 f2fs_balance_fs(sbi, dn.node_changed);
1334
1335 if (ret)
1336 goto out;
1337
1338 index = end;
1339 new_size = max_t(loff_t, new_size,
1340 (loff_t)index << PAGE_SHIFT);
1341 }
1342
1343 if (off_end) {
1344 ret = fill_zero(inode, pg_end, 0, off_end);
1345 if (ret)
1346 goto out;
1347
1348 new_size = max_t(loff_t, new_size, offset + len);
1349 }
1350 }
1351
1352 out:
1353 if (new_size > i_size_read(inode)) {
1354 if (mode & FALLOC_FL_KEEP_SIZE)
1355 file_set_keep_isize(inode);
1356 else
1357 f2fs_i_size_write(inode, new_size);
1358 }
1359 out_sem:
1360 up_write(&F2FS_I(inode)->i_mmap_sem);
1361
1362 return ret;
1363 }
1364
1365 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1366 {
1367 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1368 pgoff_t nr, pg_start, pg_end, delta, idx;
1369 loff_t new_size;
1370 int ret = 0;
1371
1372 new_size = i_size_read(inode) + len;
1373 ret = inode_newsize_ok(inode, new_size);
1374 if (ret)
1375 return ret;
1376
1377 if (offset >= i_size_read(inode))
1378 return -EINVAL;
1379
1380 /* insert range should be aligned to block size of f2fs. */
1381 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1382 return -EINVAL;
1383
1384 ret = f2fs_convert_inline_inode(inode);
1385 if (ret)
1386 return ret;
1387
1388 f2fs_balance_fs(sbi, true);
1389
1390 /* avoid gc operation during block exchange */
1391 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1392
1393 down_write(&F2FS_I(inode)->i_mmap_sem);
1394 ret = truncate_blocks(inode, i_size_read(inode), true);
1395 if (ret)
1396 goto out;
1397
1398 /* write out all dirty pages from offset */
1399 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1400 if (ret)
1401 goto out;
1402
1403 truncate_pagecache(inode, offset);
1404
1405 pg_start = offset >> PAGE_SHIFT;
1406 pg_end = (offset + len) >> PAGE_SHIFT;
1407 delta = pg_end - pg_start;
1408 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1409
1410 while (!ret && idx > pg_start) {
1411 nr = idx - pg_start;
1412 if (nr > delta)
1413 nr = delta;
1414 idx -= nr;
1415
1416 f2fs_lock_op(sbi);
1417 f2fs_drop_extent_tree(inode);
1418
1419 ret = __exchange_data_block(inode, inode, idx,
1420 idx + delta, nr, false);
1421 f2fs_unlock_op(sbi);
1422 }
1423
1424 /* write out all moved pages, if possible */
1425 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1426 truncate_pagecache(inode, offset);
1427
1428 if (!ret)
1429 f2fs_i_size_write(inode, new_size);
1430 out:
1431 up_write(&F2FS_I(inode)->i_mmap_sem);
1432 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1433 return ret;
1434 }
1435
1436 static int expand_inode_data(struct inode *inode, loff_t offset,
1437 loff_t len, int mode)
1438 {
1439 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1440 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1441 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1442 pgoff_t pg_end;
1443 loff_t new_size = i_size_read(inode);
1444 loff_t off_end;
1445 int err;
1446
1447 err = inode_newsize_ok(inode, (len + offset));
1448 if (err)
1449 return err;
1450
1451 err = f2fs_convert_inline_inode(inode);
1452 if (err)
1453 return err;
1454
1455 f2fs_balance_fs(sbi, true);
1456
1457 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1458 off_end = (offset + len) & (PAGE_SIZE - 1);
1459
1460 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1461 map.m_len = pg_end - map.m_lblk;
1462 if (off_end)
1463 map.m_len++;
1464
1465 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1466 if (err) {
1467 pgoff_t last_off;
1468
1469 if (!map.m_len)
1470 return err;
1471
1472 last_off = map.m_lblk + map.m_len - 1;
1473
1474 /* update new size to the failed position */
1475 new_size = (last_off == pg_end) ? offset + len:
1476 (loff_t)(last_off + 1) << PAGE_SHIFT;
1477 } else {
1478 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1479 }
1480
1481 if (new_size > i_size_read(inode)) {
1482 if (mode & FALLOC_FL_KEEP_SIZE)
1483 file_set_keep_isize(inode);
1484 else
1485 f2fs_i_size_write(inode, new_size);
1486 }
1487
1488 return err;
1489 }
1490
1491 static long f2fs_fallocate(struct file *file, int mode,
1492 loff_t offset, loff_t len)
1493 {
1494 struct inode *inode = file_inode(file);
1495 long ret = 0;
1496
1497 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1498 return -EIO;
1499
1500 /* f2fs only support ->fallocate for regular file */
1501 if (!S_ISREG(inode->i_mode))
1502 return -EINVAL;
1503
1504 if (f2fs_encrypted_inode(inode) &&
1505 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1506 return -EOPNOTSUPP;
1507
1508 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1509 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1510 FALLOC_FL_INSERT_RANGE))
1511 return -EOPNOTSUPP;
1512
1513 inode_lock(inode);
1514
1515 if (mode & FALLOC_FL_PUNCH_HOLE) {
1516 if (offset >= inode->i_size)
1517 goto out;
1518
1519 ret = punch_hole(inode, offset, len);
1520 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1521 ret = f2fs_collapse_range(inode, offset, len);
1522 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1523 ret = f2fs_zero_range(inode, offset, len, mode);
1524 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1525 ret = f2fs_insert_range(inode, offset, len);
1526 } else {
1527 ret = expand_inode_data(inode, offset, len, mode);
1528 }
1529
1530 if (!ret) {
1531 inode->i_mtime = inode->i_ctime = current_time(inode);
1532 f2fs_mark_inode_dirty_sync(inode, false);
1533 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1534 }
1535
1536 out:
1537 inode_unlock(inode);
1538
1539 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1540 return ret;
1541 }
1542
1543 static int f2fs_release_file(struct inode *inode, struct file *filp)
1544 {
1545 /*
1546 * f2fs_relase_file is called at every close calls. So we should
1547 * not drop any inmemory pages by close called by other process.
1548 */
1549 if (!(filp->f_mode & FMODE_WRITE) ||
1550 atomic_read(&inode->i_writecount) != 1)
1551 return 0;
1552
1553 /* some remained atomic pages should discarded */
1554 if (f2fs_is_atomic_file(inode))
1555 drop_inmem_pages(inode);
1556 if (f2fs_is_volatile_file(inode)) {
1557 clear_inode_flag(inode, FI_VOLATILE_FILE);
1558 stat_dec_volatile_write(inode);
1559 set_inode_flag(inode, FI_DROP_CACHE);
1560 filemap_fdatawrite(inode->i_mapping);
1561 clear_inode_flag(inode, FI_DROP_CACHE);
1562 }
1563 return 0;
1564 }
1565
1566 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1567 {
1568 struct inode *inode = file_inode(file);
1569
1570 /*
1571 * If the process doing a transaction is crashed, we should do
1572 * roll-back. Otherwise, other reader/write can see corrupted database
1573 * until all the writers close its file. Since this should be done
1574 * before dropping file lock, it needs to do in ->flush.
1575 */
1576 if (f2fs_is_atomic_file(inode) &&
1577 F2FS_I(inode)->inmem_task == current)
1578 drop_inmem_pages(inode);
1579 return 0;
1580 }
1581
1582 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1583 {
1584 struct inode *inode = file_inode(filp);
1585 struct f2fs_inode_info *fi = F2FS_I(inode);
1586 unsigned int flags = fi->i_flags &
1587 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1588 return put_user(flags, (int __user *)arg);
1589 }
1590
1591 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1592 {
1593 struct f2fs_inode_info *fi = F2FS_I(inode);
1594 unsigned int oldflags;
1595
1596 /* Is it quota file? Do not allow user to mess with it */
1597 if (IS_NOQUOTA(inode))
1598 return -EPERM;
1599
1600 flags = f2fs_mask_flags(inode->i_mode, flags);
1601
1602 oldflags = fi->i_flags;
1603
1604 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1605 if (!capable(CAP_LINUX_IMMUTABLE))
1606 return -EPERM;
1607
1608 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1609 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1610 fi->i_flags = flags;
1611
1612 if (fi->i_flags & FS_PROJINHERIT_FL)
1613 set_inode_flag(inode, FI_PROJ_INHERIT);
1614 else
1615 clear_inode_flag(inode, FI_PROJ_INHERIT);
1616
1617 inode->i_ctime = current_time(inode);
1618 f2fs_set_inode_flags(inode);
1619 f2fs_mark_inode_dirty_sync(inode, false);
1620 return 0;
1621 }
1622
1623 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1624 {
1625 struct inode *inode = file_inode(filp);
1626 unsigned int flags;
1627 int ret;
1628
1629 if (!inode_owner_or_capable(inode))
1630 return -EACCES;
1631
1632 if (get_user(flags, (int __user *)arg))
1633 return -EFAULT;
1634
1635 ret = mnt_want_write_file(filp);
1636 if (ret)
1637 return ret;
1638
1639 inode_lock(inode);
1640
1641 ret = __f2fs_ioc_setflags(inode, flags);
1642
1643 inode_unlock(inode);
1644 mnt_drop_write_file(filp);
1645 return ret;
1646 }
1647
1648 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1649 {
1650 struct inode *inode = file_inode(filp);
1651
1652 return put_user(inode->i_generation, (int __user *)arg);
1653 }
1654
1655 static int f2fs_ioc_start_atomic_write(struct file *filp)
1656 {
1657 struct inode *inode = file_inode(filp);
1658 int ret;
1659
1660 if (!inode_owner_or_capable(inode))
1661 return -EACCES;
1662
1663 if (!S_ISREG(inode->i_mode))
1664 return -EINVAL;
1665
1666 ret = mnt_want_write_file(filp);
1667 if (ret)
1668 return ret;
1669
1670 inode_lock(inode);
1671
1672 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1673
1674 if (f2fs_is_atomic_file(inode))
1675 goto out;
1676
1677 ret = f2fs_convert_inline_inode(inode);
1678 if (ret)
1679 goto out;
1680
1681 set_inode_flag(inode, FI_ATOMIC_FILE);
1682 set_inode_flag(inode, FI_HOT_DATA);
1683 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1684
1685 if (!get_dirty_pages(inode))
1686 goto inc_stat;
1687
1688 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1689 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1690 inode->i_ino, get_dirty_pages(inode));
1691 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1692 if (ret) {
1693 clear_inode_flag(inode, FI_ATOMIC_FILE);
1694 clear_inode_flag(inode, FI_HOT_DATA);
1695 goto out;
1696 }
1697
1698 inc_stat:
1699 F2FS_I(inode)->inmem_task = current;
1700 stat_inc_atomic_write(inode);
1701 stat_update_max_atomic_write(inode);
1702 out:
1703 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1704 inode_unlock(inode);
1705 mnt_drop_write_file(filp);
1706 return ret;
1707 }
1708
1709 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1710 {
1711 struct inode *inode = file_inode(filp);
1712 int ret;
1713
1714 if (!inode_owner_or_capable(inode))
1715 return -EACCES;
1716
1717 ret = mnt_want_write_file(filp);
1718 if (ret)
1719 return ret;
1720
1721 inode_lock(inode);
1722
1723 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1724
1725 if (f2fs_is_volatile_file(inode))
1726 goto err_out;
1727
1728 if (f2fs_is_atomic_file(inode)) {
1729 ret = commit_inmem_pages(inode);
1730 if (ret)
1731 goto err_out;
1732
1733 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1734 if (!ret) {
1735 clear_inode_flag(inode, FI_ATOMIC_FILE);
1736 clear_inode_flag(inode, FI_HOT_DATA);
1737 stat_dec_atomic_write(inode);
1738 }
1739 } else {
1740 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1741 }
1742 err_out:
1743 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1744 inode_unlock(inode);
1745 mnt_drop_write_file(filp);
1746 return ret;
1747 }
1748
1749 static int f2fs_ioc_start_volatile_write(struct file *filp)
1750 {
1751 struct inode *inode = file_inode(filp);
1752 int ret;
1753
1754 if (!inode_owner_or_capable(inode))
1755 return -EACCES;
1756
1757 if (!S_ISREG(inode->i_mode))
1758 return -EINVAL;
1759
1760 ret = mnt_want_write_file(filp);
1761 if (ret)
1762 return ret;
1763
1764 inode_lock(inode);
1765
1766 if (f2fs_is_volatile_file(inode))
1767 goto out;
1768
1769 ret = f2fs_convert_inline_inode(inode);
1770 if (ret)
1771 goto out;
1772
1773 stat_inc_volatile_write(inode);
1774 stat_update_max_volatile_write(inode);
1775
1776 set_inode_flag(inode, FI_VOLATILE_FILE);
1777 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1778 out:
1779 inode_unlock(inode);
1780 mnt_drop_write_file(filp);
1781 return ret;
1782 }
1783
1784 static int f2fs_ioc_release_volatile_write(struct file *filp)
1785 {
1786 struct inode *inode = file_inode(filp);
1787 int ret;
1788
1789 if (!inode_owner_or_capable(inode))
1790 return -EACCES;
1791
1792 ret = mnt_want_write_file(filp);
1793 if (ret)
1794 return ret;
1795
1796 inode_lock(inode);
1797
1798 if (!f2fs_is_volatile_file(inode))
1799 goto out;
1800
1801 if (!f2fs_is_first_block_written(inode)) {
1802 ret = truncate_partial_data_page(inode, 0, true);
1803 goto out;
1804 }
1805
1806 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1807 out:
1808 inode_unlock(inode);
1809 mnt_drop_write_file(filp);
1810 return ret;
1811 }
1812
1813 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1814 {
1815 struct inode *inode = file_inode(filp);
1816 int ret;
1817
1818 if (!inode_owner_or_capable(inode))
1819 return -EACCES;
1820
1821 ret = mnt_want_write_file(filp);
1822 if (ret)
1823 return ret;
1824
1825 inode_lock(inode);
1826
1827 if (f2fs_is_atomic_file(inode))
1828 drop_inmem_pages(inode);
1829 if (f2fs_is_volatile_file(inode)) {
1830 clear_inode_flag(inode, FI_VOLATILE_FILE);
1831 stat_dec_volatile_write(inode);
1832 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1833 }
1834
1835 inode_unlock(inode);
1836
1837 mnt_drop_write_file(filp);
1838 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1839 return ret;
1840 }
1841
1842 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1843 {
1844 struct inode *inode = file_inode(filp);
1845 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1846 struct super_block *sb = sbi->sb;
1847 __u32 in;
1848 int ret = 0;
1849
1850 if (!capable(CAP_SYS_ADMIN))
1851 return -EPERM;
1852
1853 if (get_user(in, (__u32 __user *)arg))
1854 return -EFAULT;
1855
1856 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1857 ret = mnt_want_write_file(filp);
1858 if (ret)
1859 return ret;
1860 }
1861
1862 switch (in) {
1863 case F2FS_GOING_DOWN_FULLSYNC:
1864 sb = freeze_bdev(sb->s_bdev);
1865 if (IS_ERR(sb)) {
1866 ret = PTR_ERR(sb);
1867 goto out;
1868 }
1869 if (sb) {
1870 f2fs_stop_checkpoint(sbi, false);
1871 thaw_bdev(sb->s_bdev, sb);
1872 }
1873 break;
1874 case F2FS_GOING_DOWN_METASYNC:
1875 /* do checkpoint only */
1876 ret = f2fs_sync_fs(sb, 1);
1877 if (ret)
1878 goto out;
1879 f2fs_stop_checkpoint(sbi, false);
1880 break;
1881 case F2FS_GOING_DOWN_NOSYNC:
1882 f2fs_stop_checkpoint(sbi, false);
1883 break;
1884 case F2FS_GOING_DOWN_METAFLUSH:
1885 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1886 f2fs_stop_checkpoint(sbi, false);
1887 break;
1888 default:
1889 ret = -EINVAL;
1890 goto out;
1891 }
1892
1893 stop_gc_thread(sbi);
1894 stop_discard_thread(sbi);
1895
1896 drop_discard_cmd(sbi);
1897 clear_opt(sbi, DISCARD);
1898
1899 f2fs_update_time(sbi, REQ_TIME);
1900 out:
1901 if (in != F2FS_GOING_DOWN_FULLSYNC)
1902 mnt_drop_write_file(filp);
1903 return ret;
1904 }
1905
1906 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1907 {
1908 struct inode *inode = file_inode(filp);
1909 struct super_block *sb = inode->i_sb;
1910 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1911 struct fstrim_range range;
1912 int ret;
1913
1914 if (!capable(CAP_SYS_ADMIN))
1915 return -EPERM;
1916
1917 if (!blk_queue_discard(q))
1918 return -EOPNOTSUPP;
1919
1920 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1921 sizeof(range)))
1922 return -EFAULT;
1923
1924 ret = mnt_want_write_file(filp);
1925 if (ret)
1926 return ret;
1927
1928 range.minlen = max((unsigned int)range.minlen,
1929 q->limits.discard_granularity);
1930 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1931 mnt_drop_write_file(filp);
1932 if (ret < 0)
1933 return ret;
1934
1935 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1936 sizeof(range)))
1937 return -EFAULT;
1938 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1939 return 0;
1940 }
1941
1942 static bool uuid_is_nonzero(__u8 u[16])
1943 {
1944 int i;
1945
1946 for (i = 0; i < 16; i++)
1947 if (u[i])
1948 return true;
1949 return false;
1950 }
1951
1952 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1953 {
1954 struct inode *inode = file_inode(filp);
1955
1956 if (!f2fs_sb_has_encrypt(inode->i_sb))
1957 return -EOPNOTSUPP;
1958
1959 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1960
1961 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1962 }
1963
1964 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1965 {
1966 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
1967 return -EOPNOTSUPP;
1968 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1969 }
1970
1971 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1972 {
1973 struct inode *inode = file_inode(filp);
1974 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1975 int err;
1976
1977 if (!f2fs_sb_has_encrypt(inode->i_sb))
1978 return -EOPNOTSUPP;
1979
1980 err = mnt_want_write_file(filp);
1981 if (err)
1982 return err;
1983
1984 down_write(&sbi->sb_lock);
1985
1986 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1987 goto got_it;
1988
1989 /* update superblock with uuid */
1990 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1991
1992 err = f2fs_commit_super(sbi, false);
1993 if (err) {
1994 /* undo new data */
1995 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1996 goto out_err;
1997 }
1998 got_it:
1999 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2000 16))
2001 err = -EFAULT;
2002 out_err:
2003 up_write(&sbi->sb_lock);
2004 mnt_drop_write_file(filp);
2005 return err;
2006 }
2007
2008 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2009 {
2010 struct inode *inode = file_inode(filp);
2011 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2012 __u32 sync;
2013 int ret;
2014
2015 if (!capable(CAP_SYS_ADMIN))
2016 return -EPERM;
2017
2018 if (get_user(sync, (__u32 __user *)arg))
2019 return -EFAULT;
2020
2021 if (f2fs_readonly(sbi->sb))
2022 return -EROFS;
2023
2024 ret = mnt_want_write_file(filp);
2025 if (ret)
2026 return ret;
2027
2028 if (!sync) {
2029 if (!mutex_trylock(&sbi->gc_mutex)) {
2030 ret = -EBUSY;
2031 goto out;
2032 }
2033 } else {
2034 mutex_lock(&sbi->gc_mutex);
2035 }
2036
2037 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2038 out:
2039 mnt_drop_write_file(filp);
2040 return ret;
2041 }
2042
2043 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2044 {
2045 struct inode *inode = file_inode(filp);
2046 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2047 struct f2fs_gc_range range;
2048 u64 end;
2049 int ret;
2050
2051 if (!capable(CAP_SYS_ADMIN))
2052 return -EPERM;
2053
2054 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2055 sizeof(range)))
2056 return -EFAULT;
2057
2058 if (f2fs_readonly(sbi->sb))
2059 return -EROFS;
2060
2061 ret = mnt_want_write_file(filp);
2062 if (ret)
2063 return ret;
2064
2065 end = range.start + range.len;
2066 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2067 ret = -EINVAL;
2068 goto out;
2069 }
2070 do_more:
2071 if (!range.sync) {
2072 if (!mutex_trylock(&sbi->gc_mutex)) {
2073 ret = -EBUSY;
2074 goto out;
2075 }
2076 } else {
2077 mutex_lock(&sbi->gc_mutex);
2078 }
2079
2080 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2081 range.start += sbi->blocks_per_seg;
2082 if (range.start <= end)
2083 goto do_more;
2084 out:
2085 mnt_drop_write_file(filp);
2086 return ret;
2087 }
2088
2089 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2090 {
2091 struct inode *inode = file_inode(filp);
2092 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2093 int ret;
2094
2095 if (!capable(CAP_SYS_ADMIN))
2096 return -EPERM;
2097
2098 if (f2fs_readonly(sbi->sb))
2099 return -EROFS;
2100
2101 ret = mnt_want_write_file(filp);
2102 if (ret)
2103 return ret;
2104
2105 ret = f2fs_sync_fs(sbi->sb, 1);
2106
2107 mnt_drop_write_file(filp);
2108 return ret;
2109 }
2110
2111 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2112 struct file *filp,
2113 struct f2fs_defragment *range)
2114 {
2115 struct inode *inode = file_inode(filp);
2116 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2117 .m_seg_type = NO_CHECK_TYPE };
2118 struct extent_info ei = {0,0,0};
2119 pgoff_t pg_start, pg_end, next_pgofs;
2120 unsigned int blk_per_seg = sbi->blocks_per_seg;
2121 unsigned int total = 0, sec_num;
2122 block_t blk_end = 0;
2123 bool fragmented = false;
2124 int err;
2125
2126 /* if in-place-update policy is enabled, don't waste time here */
2127 if (should_update_inplace(inode, NULL))
2128 return -EINVAL;
2129
2130 pg_start = range->start >> PAGE_SHIFT;
2131 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2132
2133 f2fs_balance_fs(sbi, true);
2134
2135 inode_lock(inode);
2136
2137 /* writeback all dirty pages in the range */
2138 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2139 range->start + range->len - 1);
2140 if (err)
2141 goto out;
2142
2143 /*
2144 * lookup mapping info in extent cache, skip defragmenting if physical
2145 * block addresses are continuous.
2146 */
2147 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2148 if (ei.fofs + ei.len >= pg_end)
2149 goto out;
2150 }
2151
2152 map.m_lblk = pg_start;
2153 map.m_next_pgofs = &next_pgofs;
2154
2155 /*
2156 * lookup mapping info in dnode page cache, skip defragmenting if all
2157 * physical block addresses are continuous even if there are hole(s)
2158 * in logical blocks.
2159 */
2160 while (map.m_lblk < pg_end) {
2161 map.m_len = pg_end - map.m_lblk;
2162 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2163 if (err)
2164 goto out;
2165
2166 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2167 map.m_lblk = next_pgofs;
2168 continue;
2169 }
2170
2171 if (blk_end && blk_end != map.m_pblk)
2172 fragmented = true;
2173
2174 /* record total count of block that we're going to move */
2175 total += map.m_len;
2176
2177 blk_end = map.m_pblk + map.m_len;
2178
2179 map.m_lblk += map.m_len;
2180 }
2181
2182 if (!fragmented)
2183 goto out;
2184
2185 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2186
2187 /*
2188 * make sure there are enough free section for LFS allocation, this can
2189 * avoid defragment running in SSR mode when free section are allocated
2190 * intensively
2191 */
2192 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2193 err = -EAGAIN;
2194 goto out;
2195 }
2196
2197 map.m_lblk = pg_start;
2198 map.m_len = pg_end - pg_start;
2199 total = 0;
2200
2201 while (map.m_lblk < pg_end) {
2202 pgoff_t idx;
2203 int cnt = 0;
2204
2205 do_map:
2206 map.m_len = pg_end - map.m_lblk;
2207 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2208 if (err)
2209 goto clear_out;
2210
2211 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2212 map.m_lblk = next_pgofs;
2213 continue;
2214 }
2215
2216 set_inode_flag(inode, FI_DO_DEFRAG);
2217
2218 idx = map.m_lblk;
2219 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2220 struct page *page;
2221
2222 page = get_lock_data_page(inode, idx, true);
2223 if (IS_ERR(page)) {
2224 err = PTR_ERR(page);
2225 goto clear_out;
2226 }
2227
2228 set_page_dirty(page);
2229 f2fs_put_page(page, 1);
2230
2231 idx++;
2232 cnt++;
2233 total++;
2234 }
2235
2236 map.m_lblk = idx;
2237
2238 if (idx < pg_end && cnt < blk_per_seg)
2239 goto do_map;
2240
2241 clear_inode_flag(inode, FI_DO_DEFRAG);
2242
2243 err = filemap_fdatawrite(inode->i_mapping);
2244 if (err)
2245 goto out;
2246 }
2247 clear_out:
2248 clear_inode_flag(inode, FI_DO_DEFRAG);
2249 out:
2250 inode_unlock(inode);
2251 if (!err)
2252 range->len = (u64)total << PAGE_SHIFT;
2253 return err;
2254 }
2255
2256 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2257 {
2258 struct inode *inode = file_inode(filp);
2259 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2260 struct f2fs_defragment range;
2261 int err;
2262
2263 if (!capable(CAP_SYS_ADMIN))
2264 return -EPERM;
2265
2266 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2267 return -EINVAL;
2268
2269 if (f2fs_readonly(sbi->sb))
2270 return -EROFS;
2271
2272 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2273 sizeof(range)))
2274 return -EFAULT;
2275
2276 /* verify alignment of offset & size */
2277 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2278 return -EINVAL;
2279
2280 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2281 sbi->max_file_blocks))
2282 return -EINVAL;
2283
2284 err = mnt_want_write_file(filp);
2285 if (err)
2286 return err;
2287
2288 err = f2fs_defragment_range(sbi, filp, &range);
2289 mnt_drop_write_file(filp);
2290
2291 f2fs_update_time(sbi, REQ_TIME);
2292 if (err < 0)
2293 return err;
2294
2295 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2296 sizeof(range)))
2297 return -EFAULT;
2298
2299 return 0;
2300 }
2301
2302 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2303 struct file *file_out, loff_t pos_out, size_t len)
2304 {
2305 struct inode *src = file_inode(file_in);
2306 struct inode *dst = file_inode(file_out);
2307 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2308 size_t olen = len, dst_max_i_size = 0;
2309 size_t dst_osize;
2310 int ret;
2311
2312 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2313 src->i_sb != dst->i_sb)
2314 return -EXDEV;
2315
2316 if (unlikely(f2fs_readonly(src->i_sb)))
2317 return -EROFS;
2318
2319 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2320 return -EINVAL;
2321
2322 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2323 return -EOPNOTSUPP;
2324
2325 if (src == dst) {
2326 if (pos_in == pos_out)
2327 return 0;
2328 if (pos_out > pos_in && pos_out < pos_in + len)
2329 return -EINVAL;
2330 }
2331
2332 inode_lock(src);
2333 down_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2334 if (src != dst) {
2335 ret = -EBUSY;
2336 if (!inode_trylock(dst))
2337 goto out;
2338 if (!down_write_trylock(&F2FS_I(dst)->dio_rwsem[WRITE])) {
2339 inode_unlock(dst);
2340 goto out;
2341 }
2342 }
2343
2344 ret = -EINVAL;
2345 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2346 goto out_unlock;
2347 if (len == 0)
2348 olen = len = src->i_size - pos_in;
2349 if (pos_in + len == src->i_size)
2350 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2351 if (len == 0) {
2352 ret = 0;
2353 goto out_unlock;
2354 }
2355
2356 dst_osize = dst->i_size;
2357 if (pos_out + olen > dst->i_size)
2358 dst_max_i_size = pos_out + olen;
2359
2360 /* verify the end result is block aligned */
2361 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2362 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2363 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2364 goto out_unlock;
2365
2366 ret = f2fs_convert_inline_inode(src);
2367 if (ret)
2368 goto out_unlock;
2369
2370 ret = f2fs_convert_inline_inode(dst);
2371 if (ret)
2372 goto out_unlock;
2373
2374 /* write out all dirty pages from offset */
2375 ret = filemap_write_and_wait_range(src->i_mapping,
2376 pos_in, pos_in + len);
2377 if (ret)
2378 goto out_unlock;
2379
2380 ret = filemap_write_and_wait_range(dst->i_mapping,
2381 pos_out, pos_out + len);
2382 if (ret)
2383 goto out_unlock;
2384
2385 f2fs_balance_fs(sbi, true);
2386 f2fs_lock_op(sbi);
2387 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2388 pos_out >> F2FS_BLKSIZE_BITS,
2389 len >> F2FS_BLKSIZE_BITS, false);
2390
2391 if (!ret) {
2392 if (dst_max_i_size)
2393 f2fs_i_size_write(dst, dst_max_i_size);
2394 else if (dst_osize != dst->i_size)
2395 f2fs_i_size_write(dst, dst_osize);
2396 }
2397 f2fs_unlock_op(sbi);
2398 out_unlock:
2399 if (src != dst) {
2400 up_write(&F2FS_I(dst)->dio_rwsem[WRITE]);
2401 inode_unlock(dst);
2402 }
2403 out:
2404 up_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2405 inode_unlock(src);
2406 return ret;
2407 }
2408
2409 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2410 {
2411 struct f2fs_move_range range;
2412 struct fd dst;
2413 int err;
2414
2415 if (!(filp->f_mode & FMODE_READ) ||
2416 !(filp->f_mode & FMODE_WRITE))
2417 return -EBADF;
2418
2419 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2420 sizeof(range)))
2421 return -EFAULT;
2422
2423 dst = fdget(range.dst_fd);
2424 if (!dst.file)
2425 return -EBADF;
2426
2427 if (!(dst.file->f_mode & FMODE_WRITE)) {
2428 err = -EBADF;
2429 goto err_out;
2430 }
2431
2432 err = mnt_want_write_file(filp);
2433 if (err)
2434 goto err_out;
2435
2436 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2437 range.pos_out, range.len);
2438
2439 mnt_drop_write_file(filp);
2440 if (err)
2441 goto err_out;
2442
2443 if (copy_to_user((struct f2fs_move_range __user *)arg,
2444 &range, sizeof(range)))
2445 err = -EFAULT;
2446 err_out:
2447 fdput(dst);
2448 return err;
2449 }
2450
2451 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2452 {
2453 struct inode *inode = file_inode(filp);
2454 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2455 struct sit_info *sm = SIT_I(sbi);
2456 unsigned int start_segno = 0, end_segno = 0;
2457 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2458 struct f2fs_flush_device range;
2459 int ret;
2460
2461 if (!capable(CAP_SYS_ADMIN))
2462 return -EPERM;
2463
2464 if (f2fs_readonly(sbi->sb))
2465 return -EROFS;
2466
2467 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2468 sizeof(range)))
2469 return -EFAULT;
2470
2471 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2472 sbi->segs_per_sec != 1) {
2473 f2fs_msg(sbi->sb, KERN_WARNING,
2474 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2475 range.dev_num, sbi->s_ndevs,
2476 sbi->segs_per_sec);
2477 return -EINVAL;
2478 }
2479
2480 ret = mnt_want_write_file(filp);
2481 if (ret)
2482 return ret;
2483
2484 if (range.dev_num != 0)
2485 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2486 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2487
2488 start_segno = sm->last_victim[FLUSH_DEVICE];
2489 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2490 start_segno = dev_start_segno;
2491 end_segno = min(start_segno + range.segments, dev_end_segno);
2492
2493 while (start_segno < end_segno) {
2494 if (!mutex_trylock(&sbi->gc_mutex)) {
2495 ret = -EBUSY;
2496 goto out;
2497 }
2498 sm->last_victim[GC_CB] = end_segno + 1;
2499 sm->last_victim[GC_GREEDY] = end_segno + 1;
2500 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2501 ret = f2fs_gc(sbi, true, true, start_segno);
2502 if (ret == -EAGAIN)
2503 ret = 0;
2504 else if (ret < 0)
2505 break;
2506 start_segno++;
2507 }
2508 out:
2509 mnt_drop_write_file(filp);
2510 return ret;
2511 }
2512
2513 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2514 {
2515 struct inode *inode = file_inode(filp);
2516 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2517
2518 /* Must validate to set it with SQLite behavior in Android. */
2519 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2520
2521 return put_user(sb_feature, (u32 __user *)arg);
2522 }
2523
2524 #ifdef CONFIG_QUOTA
2525 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2526 {
2527 struct inode *inode = file_inode(filp);
2528 struct f2fs_inode_info *fi = F2FS_I(inode);
2529 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2530 struct super_block *sb = sbi->sb;
2531 struct dquot *transfer_to[MAXQUOTAS] = {};
2532 struct page *ipage;
2533 kprojid_t kprojid;
2534 int err;
2535
2536 if (!f2fs_sb_has_project_quota(sb)) {
2537 if (projid != F2FS_DEF_PROJID)
2538 return -EOPNOTSUPP;
2539 else
2540 return 0;
2541 }
2542
2543 if (!f2fs_has_extra_attr(inode))
2544 return -EOPNOTSUPP;
2545
2546 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2547
2548 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2549 return 0;
2550
2551 err = mnt_want_write_file(filp);
2552 if (err)
2553 return err;
2554
2555 err = -EPERM;
2556 inode_lock(inode);
2557
2558 /* Is it quota file? Do not allow user to mess with it */
2559 if (IS_NOQUOTA(inode))
2560 goto out_unlock;
2561
2562 ipage = get_node_page(sbi, inode->i_ino);
2563 if (IS_ERR(ipage)) {
2564 err = PTR_ERR(ipage);
2565 goto out_unlock;
2566 }
2567
2568 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2569 i_projid)) {
2570 err = -EOVERFLOW;
2571 f2fs_put_page(ipage, 1);
2572 goto out_unlock;
2573 }
2574 f2fs_put_page(ipage, 1);
2575
2576 err = dquot_initialize(inode);
2577 if (err)
2578 goto out_unlock;
2579
2580 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2581 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2582 err = __dquot_transfer(inode, transfer_to);
2583 dqput(transfer_to[PRJQUOTA]);
2584 if (err)
2585 goto out_dirty;
2586 }
2587
2588 F2FS_I(inode)->i_projid = kprojid;
2589 inode->i_ctime = current_time(inode);
2590 out_dirty:
2591 f2fs_mark_inode_dirty_sync(inode, true);
2592 out_unlock:
2593 inode_unlock(inode);
2594 mnt_drop_write_file(filp);
2595 return err;
2596 }
2597 #else
2598 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2599 {
2600 if (projid != F2FS_DEF_PROJID)
2601 return -EOPNOTSUPP;
2602 return 0;
2603 }
2604 #endif
2605
2606 /* Transfer internal flags to xflags */
2607 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2608 {
2609 __u32 xflags = 0;
2610
2611 if (iflags & FS_SYNC_FL)
2612 xflags |= FS_XFLAG_SYNC;
2613 if (iflags & FS_IMMUTABLE_FL)
2614 xflags |= FS_XFLAG_IMMUTABLE;
2615 if (iflags & FS_APPEND_FL)
2616 xflags |= FS_XFLAG_APPEND;
2617 if (iflags & FS_NODUMP_FL)
2618 xflags |= FS_XFLAG_NODUMP;
2619 if (iflags & FS_NOATIME_FL)
2620 xflags |= FS_XFLAG_NOATIME;
2621 if (iflags & FS_PROJINHERIT_FL)
2622 xflags |= FS_XFLAG_PROJINHERIT;
2623 return xflags;
2624 }
2625
2626 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2627 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2628 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2629
2630 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2631 #define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2632 FS_IMMUTABLE_FL | \
2633 FS_APPEND_FL | \
2634 FS_NODUMP_FL | \
2635 FS_NOATIME_FL | \
2636 FS_PROJINHERIT_FL)
2637
2638 /* Transfer xflags flags to internal */
2639 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2640 {
2641 unsigned long iflags = 0;
2642
2643 if (xflags & FS_XFLAG_SYNC)
2644 iflags |= FS_SYNC_FL;
2645 if (xflags & FS_XFLAG_IMMUTABLE)
2646 iflags |= FS_IMMUTABLE_FL;
2647 if (xflags & FS_XFLAG_APPEND)
2648 iflags |= FS_APPEND_FL;
2649 if (xflags & FS_XFLAG_NODUMP)
2650 iflags |= FS_NODUMP_FL;
2651 if (xflags & FS_XFLAG_NOATIME)
2652 iflags |= FS_NOATIME_FL;
2653 if (xflags & FS_XFLAG_PROJINHERIT)
2654 iflags |= FS_PROJINHERIT_FL;
2655
2656 return iflags;
2657 }
2658
2659 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2660 {
2661 struct inode *inode = file_inode(filp);
2662 struct f2fs_inode_info *fi = F2FS_I(inode);
2663 struct fsxattr fa;
2664
2665 memset(&fa, 0, sizeof(struct fsxattr));
2666 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2667 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2668
2669 if (f2fs_sb_has_project_quota(inode->i_sb))
2670 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2671 fi->i_projid);
2672
2673 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2674 return -EFAULT;
2675 return 0;
2676 }
2677
2678 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2679 {
2680 struct inode *inode = file_inode(filp);
2681 struct f2fs_inode_info *fi = F2FS_I(inode);
2682 struct fsxattr fa;
2683 unsigned int flags;
2684 int err;
2685
2686 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2687 return -EFAULT;
2688
2689 /* Make sure caller has proper permission */
2690 if (!inode_owner_or_capable(inode))
2691 return -EACCES;
2692
2693 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2694 return -EOPNOTSUPP;
2695
2696 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2697 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2698 return -EOPNOTSUPP;
2699
2700 err = mnt_want_write_file(filp);
2701 if (err)
2702 return err;
2703
2704 inode_lock(inode);
2705 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2706 (flags & F2FS_FL_XFLAG_VISIBLE);
2707 err = __f2fs_ioc_setflags(inode, flags);
2708 inode_unlock(inode);
2709 mnt_drop_write_file(filp);
2710 if (err)
2711 return err;
2712
2713 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2714 if (err)
2715 return err;
2716
2717 return 0;
2718 }
2719
2720 int f2fs_pin_file_control(struct inode *inode, bool inc)
2721 {
2722 struct f2fs_inode_info *fi = F2FS_I(inode);
2723 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2724
2725 /* Use i_gc_failures for normal file as a risk signal. */
2726 if (inc)
2727 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
2728
2729 if (fi->i_gc_failures > sbi->gc_pin_file_threshold) {
2730 f2fs_msg(sbi->sb, KERN_WARNING,
2731 "%s: Enable GC = ino %lx after %x GC trials\n",
2732 __func__, inode->i_ino, fi->i_gc_failures);
2733 clear_inode_flag(inode, FI_PIN_FILE);
2734 return -EAGAIN;
2735 }
2736 return 0;
2737 }
2738
2739 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2740 {
2741 struct inode *inode = file_inode(filp);
2742 __u32 pin;
2743 int ret = 0;
2744
2745 if (!inode_owner_or_capable(inode))
2746 return -EACCES;
2747
2748 if (get_user(pin, (__u32 __user *)arg))
2749 return -EFAULT;
2750
2751 if (!S_ISREG(inode->i_mode))
2752 return -EINVAL;
2753
2754 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2755 return -EROFS;
2756
2757 ret = mnt_want_write_file(filp);
2758 if (ret)
2759 return ret;
2760
2761 inode_lock(inode);
2762
2763 if (should_update_outplace(inode, NULL)) {
2764 ret = -EINVAL;
2765 goto out;
2766 }
2767
2768 if (!pin) {
2769 clear_inode_flag(inode, FI_PIN_FILE);
2770 F2FS_I(inode)->i_gc_failures = 1;
2771 goto done;
2772 }
2773
2774 if (f2fs_pin_file_control(inode, false)) {
2775 ret = -EAGAIN;
2776 goto out;
2777 }
2778 ret = f2fs_convert_inline_inode(inode);
2779 if (ret)
2780 goto out;
2781
2782 set_inode_flag(inode, FI_PIN_FILE);
2783 ret = F2FS_I(inode)->i_gc_failures;
2784 done:
2785 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2786 out:
2787 inode_unlock(inode);
2788 mnt_drop_write_file(filp);
2789 return ret;
2790 }
2791
2792 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2793 {
2794 struct inode *inode = file_inode(filp);
2795 __u32 pin = 0;
2796
2797 if (is_inode_flag_set(inode, FI_PIN_FILE))
2798 pin = F2FS_I(inode)->i_gc_failures;
2799 return put_user(pin, (u32 __user *)arg);
2800 }
2801
2802 int f2fs_precache_extents(struct inode *inode)
2803 {
2804 struct f2fs_inode_info *fi = F2FS_I(inode);
2805 struct f2fs_map_blocks map;
2806 pgoff_t m_next_extent;
2807 loff_t end;
2808 int err;
2809
2810 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2811 return -EOPNOTSUPP;
2812
2813 map.m_lblk = 0;
2814 map.m_next_pgofs = NULL;
2815 map.m_next_extent = &m_next_extent;
2816 map.m_seg_type = NO_CHECK_TYPE;
2817 end = F2FS_I_SB(inode)->max_file_blocks;
2818
2819 while (map.m_lblk < end) {
2820 map.m_len = end - map.m_lblk;
2821
2822 down_write(&fi->dio_rwsem[WRITE]);
2823 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2824 up_write(&fi->dio_rwsem[WRITE]);
2825 if (err)
2826 return err;
2827
2828 map.m_lblk = m_next_extent;
2829 }
2830
2831 return err;
2832 }
2833
2834 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2835 {
2836 return f2fs_precache_extents(file_inode(filp));
2837 }
2838
2839 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2840 {
2841 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2842 return -EIO;
2843
2844 switch (cmd) {
2845 case F2FS_IOC_GETFLAGS:
2846 return f2fs_ioc_getflags(filp, arg);
2847 case F2FS_IOC_SETFLAGS:
2848 return f2fs_ioc_setflags(filp, arg);
2849 case F2FS_IOC_GETVERSION:
2850 return f2fs_ioc_getversion(filp, arg);
2851 case F2FS_IOC_START_ATOMIC_WRITE:
2852 return f2fs_ioc_start_atomic_write(filp);
2853 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2854 return f2fs_ioc_commit_atomic_write(filp);
2855 case F2FS_IOC_START_VOLATILE_WRITE:
2856 return f2fs_ioc_start_volatile_write(filp);
2857 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2858 return f2fs_ioc_release_volatile_write(filp);
2859 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2860 return f2fs_ioc_abort_volatile_write(filp);
2861 case F2FS_IOC_SHUTDOWN:
2862 return f2fs_ioc_shutdown(filp, arg);
2863 case FITRIM:
2864 return f2fs_ioc_fitrim(filp, arg);
2865 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2866 return f2fs_ioc_set_encryption_policy(filp, arg);
2867 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2868 return f2fs_ioc_get_encryption_policy(filp, arg);
2869 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2870 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2871 case F2FS_IOC_GARBAGE_COLLECT:
2872 return f2fs_ioc_gc(filp, arg);
2873 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2874 return f2fs_ioc_gc_range(filp, arg);
2875 case F2FS_IOC_WRITE_CHECKPOINT:
2876 return f2fs_ioc_write_checkpoint(filp, arg);
2877 case F2FS_IOC_DEFRAGMENT:
2878 return f2fs_ioc_defragment(filp, arg);
2879 case F2FS_IOC_MOVE_RANGE:
2880 return f2fs_ioc_move_range(filp, arg);
2881 case F2FS_IOC_FLUSH_DEVICE:
2882 return f2fs_ioc_flush_device(filp, arg);
2883 case F2FS_IOC_GET_FEATURES:
2884 return f2fs_ioc_get_features(filp, arg);
2885 case F2FS_IOC_FSGETXATTR:
2886 return f2fs_ioc_fsgetxattr(filp, arg);
2887 case F2FS_IOC_FSSETXATTR:
2888 return f2fs_ioc_fssetxattr(filp, arg);
2889 case F2FS_IOC_GET_PIN_FILE:
2890 return f2fs_ioc_get_pin_file(filp, arg);
2891 case F2FS_IOC_SET_PIN_FILE:
2892 return f2fs_ioc_set_pin_file(filp, arg);
2893 case F2FS_IOC_PRECACHE_EXTENTS:
2894 return f2fs_ioc_precache_extents(filp, arg);
2895 default:
2896 return -ENOTTY;
2897 }
2898 }
2899
2900 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2901 {
2902 struct file *file = iocb->ki_filp;
2903 struct inode *inode = file_inode(file);
2904 struct blk_plug plug;
2905 ssize_t ret;
2906
2907 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2908 return -EIO;
2909
2910 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
2911 return -EINVAL;
2912
2913 if (!inode_trylock(inode)) {
2914 if (iocb->ki_flags & IOCB_NOWAIT)
2915 return -EAGAIN;
2916 inode_lock(inode);
2917 }
2918
2919 ret = generic_write_checks(iocb, from);
2920 if (ret > 0) {
2921 bool preallocated = false;
2922 size_t target_size = 0;
2923 int err;
2924
2925 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2926 set_inode_flag(inode, FI_NO_PREALLOC);
2927
2928 if ((iocb->ki_flags & IOCB_NOWAIT) &&
2929 (iocb->ki_flags & IOCB_DIRECT)) {
2930 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
2931 iov_iter_count(from)) ||
2932 f2fs_has_inline_data(inode) ||
2933 f2fs_force_buffered_io(inode, WRITE)) {
2934 inode_unlock(inode);
2935 return -EAGAIN;
2936 }
2937
2938 } else {
2939 preallocated = true;
2940 target_size = iocb->ki_pos + iov_iter_count(from);
2941
2942 err = f2fs_preallocate_blocks(iocb, from);
2943 if (err) {
2944 clear_inode_flag(inode, FI_NO_PREALLOC);
2945 inode_unlock(inode);
2946 return err;
2947 }
2948 }
2949 blk_start_plug(&plug);
2950 ret = __generic_file_write_iter(iocb, from);
2951 blk_finish_plug(&plug);
2952 clear_inode_flag(inode, FI_NO_PREALLOC);
2953
2954 /* if we couldn't write data, we should deallocate blocks. */
2955 if (preallocated && i_size_read(inode) < target_size)
2956 f2fs_truncate(inode);
2957
2958 if (ret > 0)
2959 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2960 }
2961 inode_unlock(inode);
2962
2963 if (ret > 0)
2964 ret = generic_write_sync(iocb, ret);
2965 return ret;
2966 }
2967
2968 #ifdef CONFIG_COMPAT
2969 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2970 {
2971 switch (cmd) {
2972 case F2FS_IOC32_GETFLAGS:
2973 cmd = F2FS_IOC_GETFLAGS;
2974 break;
2975 case F2FS_IOC32_SETFLAGS:
2976 cmd = F2FS_IOC_SETFLAGS;
2977 break;
2978 case F2FS_IOC32_GETVERSION:
2979 cmd = F2FS_IOC_GETVERSION;
2980 break;
2981 case F2FS_IOC_START_ATOMIC_WRITE:
2982 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2983 case F2FS_IOC_START_VOLATILE_WRITE:
2984 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2985 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2986 case F2FS_IOC_SHUTDOWN:
2987 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2988 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2989 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2990 case F2FS_IOC_GARBAGE_COLLECT:
2991 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2992 case F2FS_IOC_WRITE_CHECKPOINT:
2993 case F2FS_IOC_DEFRAGMENT:
2994 case F2FS_IOC_MOVE_RANGE:
2995 case F2FS_IOC_FLUSH_DEVICE:
2996 case F2FS_IOC_GET_FEATURES:
2997 case F2FS_IOC_FSGETXATTR:
2998 case F2FS_IOC_FSSETXATTR:
2999 case F2FS_IOC_GET_PIN_FILE:
3000 case F2FS_IOC_SET_PIN_FILE:
3001 case F2FS_IOC_PRECACHE_EXTENTS:
3002 break;
3003 default:
3004 return -ENOIOCTLCMD;
3005 }
3006 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3007 }
3008 #endif
3009
3010 const struct file_operations f2fs_file_operations = {
3011 .llseek = f2fs_llseek,
3012 .read_iter = generic_file_read_iter,
3013 .write_iter = f2fs_file_write_iter,
3014 .open = f2fs_file_open,
3015 .release = f2fs_release_file,
3016 .mmap = f2fs_file_mmap,
3017 .flush = f2fs_file_flush,
3018 .fsync = f2fs_sync_file,
3019 .fallocate = f2fs_fallocate,
3020 .unlocked_ioctl = f2fs_ioctl,
3021 #ifdef CONFIG_COMPAT
3022 .compat_ioctl = f2fs_compat_ioctl,
3023 #endif
3024 .splice_read = generic_file_splice_read,
3025 .splice_write = iter_file_splice_write,
3026 };
Motorola kernel-slsi