Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * fs/f2fs/file.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/stat.h>
11#include <linux/buffer_head.h>
12#include <linux/writeback.h>
13#include <linux/blkdev.h>
14#include <linux/falloc.h>
15#include <linux/types.h>
16#include <linux/compat.h>
17#include <linux/uaccess.h>
18#include <linux/mount.h>
19#include <linux/pagevec.h>
20#include <linux/uio.h>
21#include <linux/uuid.h>
22#include <linux/file.h>
23#include <linux/nls.h>
24#include <linux/sched/signal.h>
25#include <linux/fileattr.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/events/f2fs.h>
34#include <uapi/linux/f2fs.h>
35
36static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
37{
38 struct inode *inode = file_inode(vmf->vma->vm_file);
39 vm_fault_t ret;
40
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 ret = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
44
45 if (!ret)
46 f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
47 F2FS_BLKSIZE);
48
49 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
50
51 return ret;
52}
53
54static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
55{
56 struct page *page = vmf->page;
57 struct inode *inode = file_inode(vmf->vma->vm_file);
58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
59 struct dnode_of_data dn;
60 bool need_alloc = true;
61 int err = 0;
62
63 if (unlikely(IS_IMMUTABLE(inode)))
64 return VM_FAULT_SIGBUS;
65
66 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
67 return VM_FAULT_SIGBUS;
68
69 if (unlikely(f2fs_cp_error(sbi))) {
70 err = -EIO;
71 goto err;
72 }
73
74 if (!f2fs_is_checkpoint_ready(sbi)) {
75 err = -ENOSPC;
76 goto err;
77 }
78
79 err = f2fs_convert_inline_inode(inode);
80 if (err)
81 goto err;
82
83#ifdef CONFIG_F2FS_FS_COMPRESSION
84 if (f2fs_compressed_file(inode)) {
85 int ret = f2fs_is_compressed_cluster(inode, page->index);
86
87 if (ret < 0) {
88 err = ret;
89 goto err;
90 } else if (ret) {
91 need_alloc = false;
92 }
93 }
94#endif
95 /* should do out of any locked page */
96 if (need_alloc)
97 f2fs_balance_fs(sbi, true);
98
99 sb_start_pagefault(inode->i_sb);
100
101 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
102
103 file_update_time(vmf->vma->vm_file);
104 down_read(&F2FS_I(inode)->i_mmap_sem);
105 lock_page(page);
106 if (unlikely(page->mapping != inode->i_mapping ||
107 page_offset(page) > i_size_read(inode) ||
108 !PageUptodate(page))) {
109 unlock_page(page);
110 err = -EFAULT;
111 goto out_sem;
112 }
113
114 if (need_alloc) {
115 /* block allocation */
116 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
117 set_new_dnode(&dn, inode, NULL, NULL, 0);
118 err = f2fs_get_block(&dn, page->index);
119 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
120 }
121
122#ifdef CONFIG_F2FS_FS_COMPRESSION
123 if (!need_alloc) {
124 set_new_dnode(&dn, inode, NULL, NULL, 0);
125 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
126 f2fs_put_dnode(&dn);
127 }
128#endif
129 if (err) {
130 unlock_page(page);
131 goto out_sem;
132 }
133
134 f2fs_wait_on_page_writeback(page, DATA, false, true);
135
136 /* wait for GCed page writeback via META_MAPPING */
137 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
138
139 /*
140 * check to see if the page is mapped already (no holes)
141 */
142 if (PageMappedToDisk(page))
143 goto out_sem;
144
145 /* page is wholly or partially inside EOF */
146 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
147 i_size_read(inode)) {
148 loff_t offset;
149
150 offset = i_size_read(inode) & ~PAGE_MASK;
151 zero_user_segment(page, offset, PAGE_SIZE);
152 }
153 set_page_dirty(page);
154 if (!PageUptodate(page))
155 SetPageUptodate(page);
156
157 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
158 f2fs_update_time(sbi, REQ_TIME);
159
160 trace_f2fs_vm_page_mkwrite(page, DATA);
161out_sem:
162 up_read(&F2FS_I(inode)->i_mmap_sem);
163
164 sb_end_pagefault(inode->i_sb);
165err:
166 return block_page_mkwrite_return(err);
167}
168
169static const struct vm_operations_struct f2fs_file_vm_ops = {
170 .fault = f2fs_filemap_fault,
171 .map_pages = filemap_map_pages,
172 .page_mkwrite = f2fs_vm_page_mkwrite,
173};
174
175static int get_parent_ino(struct inode *inode, nid_t *pino)
176{
177 struct dentry *dentry;
178
179 /*
180 * Make sure to get the non-deleted alias. The alias associated with
181 * the open file descriptor being fsync()'ed may be deleted already.
182 */
183 dentry = d_find_alias(inode);
184 if (!dentry)
185 return 0;
186
187 *pino = parent_ino(dentry);
188 dput(dentry);
189 return 1;
190}
191
192static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
193{
194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
195 enum cp_reason_type cp_reason = CP_NO_NEEDED;
196
197 if (!S_ISREG(inode->i_mode))
198 cp_reason = CP_NON_REGULAR;
199 else if (f2fs_compressed_file(inode))
200 cp_reason = CP_COMPRESSED;
201 else if (inode->i_nlink != 1)
202 cp_reason = CP_HARDLINK;
203 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
204 cp_reason = CP_SB_NEED_CP;
205 else if (file_wrong_pino(inode))
206 cp_reason = CP_WRONG_PINO;
207 else if (!f2fs_space_for_roll_forward(sbi))
208 cp_reason = CP_NO_SPC_ROLL;
209 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
210 cp_reason = CP_NODE_NEED_CP;
211 else if (test_opt(sbi, FASTBOOT))
212 cp_reason = CP_FASTBOOT_MODE;
213 else if (F2FS_OPTION(sbi).active_logs == 2)
214 cp_reason = CP_SPEC_LOG_NUM;
215 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
216 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
217 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
218 TRANS_DIR_INO))
219 cp_reason = CP_RECOVER_DIR;
220
221 return cp_reason;
222}
223
224static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
225{
226 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
227 bool ret = false;
228 /* But we need to avoid that there are some inode updates */
229 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
230 ret = true;
231 f2fs_put_page(i, 0);
232 return ret;
233}
234
235static void try_to_fix_pino(struct inode *inode)
236{
237 struct f2fs_inode_info *fi = F2FS_I(inode);
238 nid_t pino;
239
240 down_write(&fi->i_sem);
241 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
242 get_parent_ino(inode, &pino)) {
243 f2fs_i_pino_write(inode, pino);
244 file_got_pino(inode);
245 }
246 up_write(&fi->i_sem);
247}
248
249static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
250 int datasync, bool atomic)
251{
252 struct inode *inode = file->f_mapping->host;
253 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
254 nid_t ino = inode->i_ino;
255 int ret = 0;
256 enum cp_reason_type cp_reason = 0;
257 struct writeback_control wbc = {
258 .sync_mode = WB_SYNC_ALL,
259 .nr_to_write = LONG_MAX,
260 .for_reclaim = 0,
261 };
262 unsigned int seq_id = 0;
263
264 if (unlikely(f2fs_readonly(inode->i_sb)))
265 return 0;
266
267 trace_f2fs_sync_file_enter(inode);
268
269 if (S_ISDIR(inode->i_mode))
270 goto go_write;
271
272 /* if fdatasync is triggered, let's do in-place-update */
273 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
274 set_inode_flag(inode, FI_NEED_IPU);
275 ret = file_write_and_wait_range(file, start, end);
276 clear_inode_flag(inode, FI_NEED_IPU);
277
278 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
279 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
280 return ret;
281 }
282
283 /* if the inode is dirty, let's recover all the time */
284 if (!f2fs_skip_inode_update(inode, datasync)) {
285 f2fs_write_inode(inode, NULL);
286 goto go_write;
287 }
288
289 /*
290 * if there is no written data, don't waste time to write recovery info.
291 */
292 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
293 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
294
295 /* it may call write_inode just prior to fsync */
296 if (need_inode_page_update(sbi, ino))
297 goto go_write;
298
299 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
300 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
301 goto flush_out;
302 goto out;
303 }
304go_write:
305 /*
306 * Both of fdatasync() and fsync() are able to be recovered from
307 * sudden-power-off.
308 */
309 down_read(&F2FS_I(inode)->i_sem);
310 cp_reason = need_do_checkpoint(inode);
311 up_read(&F2FS_I(inode)->i_sem);
312
313 if (cp_reason) {
314 /* all the dirty node pages should be flushed for POR */
315 ret = f2fs_sync_fs(inode->i_sb, 1);
316
317 /*
318 * We've secured consistency through sync_fs. Following pino
319 * will be used only for fsynced inodes after checkpoint.
320 */
321 try_to_fix_pino(inode);
322 clear_inode_flag(inode, FI_APPEND_WRITE);
323 clear_inode_flag(inode, FI_UPDATE_WRITE);
324 goto out;
325 }
326sync_nodes:
327 atomic_inc(&sbi->wb_sync_req[NODE]);
328 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
329 atomic_dec(&sbi->wb_sync_req[NODE]);
330 if (ret)
331 goto out;
332
333 /* if cp_error was enabled, we should avoid infinite loop */
334 if (unlikely(f2fs_cp_error(sbi))) {
335 ret = -EIO;
336 goto out;
337 }
338
339 if (f2fs_need_inode_block_update(sbi, ino)) {
340 f2fs_mark_inode_dirty_sync(inode, true);
341 f2fs_write_inode(inode, NULL);
342 goto sync_nodes;
343 }
344
345 /*
346 * If it's atomic_write, it's just fine to keep write ordering. So
347 * here we don't need to wait for node write completion, since we use
348 * node chain which serializes node blocks. If one of node writes are
349 * reordered, we can see simply broken chain, resulting in stopping
350 * roll-forward recovery. It means we'll recover all or none node blocks
351 * given fsync mark.
352 */
353 if (!atomic) {
354 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
355 if (ret)
356 goto out;
357 }
358
359 /* once recovery info is written, don't need to tack this */
360 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
361 clear_inode_flag(inode, FI_APPEND_WRITE);
362flush_out:
363 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
364 ret = f2fs_issue_flush(sbi, inode->i_ino);
365 if (!ret) {
366 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
367 clear_inode_flag(inode, FI_UPDATE_WRITE);
368 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
369 }
370 f2fs_update_time(sbi, REQ_TIME);
371out:
372 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
373 return ret;
374}
375
376int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
377{
378 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
379 return -EIO;
380 return f2fs_do_sync_file(file, start, end, datasync, false);
381}
382
383static bool __found_offset(struct address_space *mapping, block_t blkaddr,
384 pgoff_t index, int whence)
385{
386 switch (whence) {
387 case SEEK_DATA:
388 if (__is_valid_data_blkaddr(blkaddr))
389 return true;
390 if (blkaddr == NEW_ADDR &&
391 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
392 return true;
393 break;
394 case SEEK_HOLE:
395 if (blkaddr == NULL_ADDR)
396 return true;
397 break;
398 }
399 return false;
400}
401
402static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
403{
404 struct inode *inode = file->f_mapping->host;
405 loff_t maxbytes = inode->i_sb->s_maxbytes;
406 struct dnode_of_data dn;
407 pgoff_t pgofs, end_offset;
408 loff_t data_ofs = offset;
409 loff_t isize;
410 int err = 0;
411
412 inode_lock(inode);
413
414 isize = i_size_read(inode);
415 if (offset >= isize)
416 goto fail;
417
418 /* handle inline data case */
419 if (f2fs_has_inline_data(inode)) {
420 if (whence == SEEK_HOLE) {
421 data_ofs = isize;
422 goto found;
423 } else if (whence == SEEK_DATA) {
424 data_ofs = offset;
425 goto found;
426 }
427 }
428
429 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
430
431 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
432 set_new_dnode(&dn, inode, NULL, NULL, 0);
433 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
434 if (err && err != -ENOENT) {
435 goto fail;
436 } else if (err == -ENOENT) {
437 /* direct node does not exists */
438 if (whence == SEEK_DATA) {
439 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
440 continue;
441 } else {
442 goto found;
443 }
444 }
445
446 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
447
448 /* find data/hole in dnode block */
449 for (; dn.ofs_in_node < end_offset;
450 dn.ofs_in_node++, pgofs++,
451 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
452 block_t blkaddr;
453
454 blkaddr = f2fs_data_blkaddr(&dn);
455
456 if (__is_valid_data_blkaddr(blkaddr) &&
457 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
458 blkaddr, DATA_GENERIC_ENHANCE)) {
459 f2fs_put_dnode(&dn);
460 goto fail;
461 }
462
463 if (__found_offset(file->f_mapping, blkaddr,
464 pgofs, whence)) {
465 f2fs_put_dnode(&dn);
466 goto found;
467 }
468 }
469 f2fs_put_dnode(&dn);
470 }
471
472 if (whence == SEEK_DATA)
473 goto fail;
474found:
475 if (whence == SEEK_HOLE && data_ofs > isize)
476 data_ofs = isize;
477 inode_unlock(inode);
478 return vfs_setpos(file, data_ofs, maxbytes);
479fail:
480 inode_unlock(inode);
481 return -ENXIO;
482}
483
484static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
485{
486 struct inode *inode = file->f_mapping->host;
487 loff_t maxbytes = inode->i_sb->s_maxbytes;
488
489 if (f2fs_compressed_file(inode))
490 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
491
492 switch (whence) {
493 case SEEK_SET:
494 case SEEK_CUR:
495 case SEEK_END:
496 return generic_file_llseek_size(file, offset, whence,
497 maxbytes, i_size_read(inode));
498 case SEEK_DATA:
499 case SEEK_HOLE:
500 if (offset < 0)
501 return -ENXIO;
502 return f2fs_seek_block(file, offset, whence);
503 }
504
505 return -EINVAL;
506}
507
508static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
509{
510 struct inode *inode = file_inode(file);
511
512 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
513 return -EIO;
514
515 if (!f2fs_is_compress_backend_ready(inode))
516 return -EOPNOTSUPP;
517
518 file_accessed(file);
519 vma->vm_ops = &f2fs_file_vm_ops;
520 set_inode_flag(inode, FI_MMAP_FILE);
521 return 0;
522}
523
524static int f2fs_file_open(struct inode *inode, struct file *filp)
525{
526 int err = fscrypt_file_open(inode, filp);
527
528 if (err)
529 return err;
530
531 if (!f2fs_is_compress_backend_ready(inode))
532 return -EOPNOTSUPP;
533
534 err = fsverity_file_open(inode, filp);
535 if (err)
536 return err;
537
538 filp->f_mode |= FMODE_NOWAIT;
539
540 return dquot_file_open(inode, filp);
541}
542
543void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
544{
545 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
546 struct f2fs_node *raw_node;
547 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
548 __le32 *addr;
549 int base = 0;
550 bool compressed_cluster = false;
551 int cluster_index = 0, valid_blocks = 0;
552 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
553 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
554
555 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
556 base = get_extra_isize(dn->inode);
557
558 raw_node = F2FS_NODE(dn->node_page);
559 addr = blkaddr_in_node(raw_node) + base + ofs;
560
561 /* Assumption: truncateion starts with cluster */
562 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
563 block_t blkaddr = le32_to_cpu(*addr);
564
565 if (f2fs_compressed_file(dn->inode) &&
566 !(cluster_index & (cluster_size - 1))) {
567 if (compressed_cluster)
568 f2fs_i_compr_blocks_update(dn->inode,
569 valid_blocks, false);
570 compressed_cluster = (blkaddr == COMPRESS_ADDR);
571 valid_blocks = 0;
572 }
573
574 if (blkaddr == NULL_ADDR)
575 continue;
576
577 dn->data_blkaddr = NULL_ADDR;
578 f2fs_set_data_blkaddr(dn);
579
580 if (__is_valid_data_blkaddr(blkaddr)) {
581 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
582 DATA_GENERIC_ENHANCE))
583 continue;
584 if (compressed_cluster)
585 valid_blocks++;
586 }
587
588 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
589 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
590
591 f2fs_invalidate_blocks(sbi, blkaddr);
592
593 if (!released || blkaddr != COMPRESS_ADDR)
594 nr_free++;
595 }
596
597 if (compressed_cluster)
598 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
599
600 if (nr_free) {
601 pgoff_t fofs;
602 /*
603 * once we invalidate valid blkaddr in range [ofs, ofs + count],
604 * we will invalidate all blkaddr in the whole range.
605 */
606 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
607 dn->inode) + ofs;
608 f2fs_update_extent_cache_range(dn, fofs, 0, len);
609 dec_valid_block_count(sbi, dn->inode, nr_free);
610 }
611 dn->ofs_in_node = ofs;
612
613 f2fs_update_time(sbi, REQ_TIME);
614 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
615 dn->ofs_in_node, nr_free);
616}
617
618void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
619{
620 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
621}
622
623static int truncate_partial_data_page(struct inode *inode, u64 from,
624 bool cache_only)
625{
626 loff_t offset = from & (PAGE_SIZE - 1);
627 pgoff_t index = from >> PAGE_SHIFT;
628 struct address_space *mapping = inode->i_mapping;
629 struct page *page;
630
631 if (!offset && !cache_only)
632 return 0;
633
634 if (cache_only) {
635 page = find_lock_page(mapping, index);
636 if (page && PageUptodate(page))
637 goto truncate_out;
638 f2fs_put_page(page, 1);
639 return 0;
640 }
641
642 page = f2fs_get_lock_data_page(inode, index, true);
643 if (IS_ERR(page))
644 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
645truncate_out:
646 f2fs_wait_on_page_writeback(page, DATA, true, true);
647 zero_user(page, offset, PAGE_SIZE - offset);
648
649 /* An encrypted inode should have a key and truncate the last page. */
650 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
651 if (!cache_only)
652 set_page_dirty(page);
653 f2fs_put_page(page, 1);
654 return 0;
655}
656
657int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
658{
659 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
660 struct dnode_of_data dn;
661 pgoff_t free_from;
662 int count = 0, err = 0;
663 struct page *ipage;
664 bool truncate_page = false;
665
666 trace_f2fs_truncate_blocks_enter(inode, from);
667
668 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
669
670 if (free_from >= max_file_blocks(inode))
671 goto free_partial;
672
673 if (lock)
674 f2fs_lock_op(sbi);
675
676 ipage = f2fs_get_node_page(sbi, inode->i_ino);
677 if (IS_ERR(ipage)) {
678 err = PTR_ERR(ipage);
679 goto out;
680 }
681
682 if (f2fs_has_inline_data(inode)) {
683 f2fs_truncate_inline_inode(inode, ipage, from);
684 f2fs_put_page(ipage, 1);
685 truncate_page = true;
686 goto out;
687 }
688
689 set_new_dnode(&dn, inode, ipage, NULL, 0);
690 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
691 if (err) {
692 if (err == -ENOENT)
693 goto free_next;
694 goto out;
695 }
696
697 count = ADDRS_PER_PAGE(dn.node_page, inode);
698
699 count -= dn.ofs_in_node;
700 f2fs_bug_on(sbi, count < 0);
701
702 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
703 f2fs_truncate_data_blocks_range(&dn, count);
704 free_from += count;
705 }
706
707 f2fs_put_dnode(&dn);
708free_next:
709 err = f2fs_truncate_inode_blocks(inode, free_from);
710out:
711 if (lock)
712 f2fs_unlock_op(sbi);
713free_partial:
714 /* lastly zero out the first data page */
715 if (!err)
716 err = truncate_partial_data_page(inode, from, truncate_page);
717
718 trace_f2fs_truncate_blocks_exit(inode, err);
719 return err;
720}
721
722int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
723{
724 u64 free_from = from;
725 int err;
726
727#ifdef CONFIG_F2FS_FS_COMPRESSION
728 /*
729 * for compressed file, only support cluster size
730 * aligned truncation.
731 */
732 if (f2fs_compressed_file(inode))
733 free_from = round_up(from,
734 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
735#endif
736
737 err = f2fs_do_truncate_blocks(inode, free_from, lock);
738 if (err)
739 return err;
740
741#ifdef CONFIG_F2FS_FS_COMPRESSION
742 if (from != free_from) {
743 err = f2fs_truncate_partial_cluster(inode, from, lock);
744 if (err)
745 return err;
746 }
747#endif
748
749 return 0;
750}
751
752int f2fs_truncate(struct inode *inode)
753{
754 int err;
755
756 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
757 return -EIO;
758
759 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
760 S_ISLNK(inode->i_mode)))
761 return 0;
762
763 trace_f2fs_truncate(inode);
764
765 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
766 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
767 return -EIO;
768 }
769
770 err = dquot_initialize(inode);
771 if (err)
772 return err;
773
774 /* we should check inline_data size */
775 if (!f2fs_may_inline_data(inode)) {
776 err = f2fs_convert_inline_inode(inode);
777 if (err)
778 return err;
779 }
780
781 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
782 if (err)
783 return err;
784
785 inode->i_mtime = inode->i_ctime = current_time(inode);
786 f2fs_mark_inode_dirty_sync(inode, false);
787 return 0;
788}
789
790int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
791 struct kstat *stat, u32 request_mask, unsigned int query_flags)
792{
793 struct inode *inode = d_inode(path->dentry);
794 struct f2fs_inode_info *fi = F2FS_I(inode);
795 struct f2fs_inode *ri;
796 unsigned int flags;
797
798 if (f2fs_has_extra_attr(inode) &&
799 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
800 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
801 stat->result_mask |= STATX_BTIME;
802 stat->btime.tv_sec = fi->i_crtime.tv_sec;
803 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
804 }
805
806 flags = fi->i_flags;
807 if (flags & F2FS_COMPR_FL)
808 stat->attributes |= STATX_ATTR_COMPRESSED;
809 if (flags & F2FS_APPEND_FL)
810 stat->attributes |= STATX_ATTR_APPEND;
811 if (IS_ENCRYPTED(inode))
812 stat->attributes |= STATX_ATTR_ENCRYPTED;
813 if (flags & F2FS_IMMUTABLE_FL)
814 stat->attributes |= STATX_ATTR_IMMUTABLE;
815 if (flags & F2FS_NODUMP_FL)
816 stat->attributes |= STATX_ATTR_NODUMP;
817 if (IS_VERITY(inode))
818 stat->attributes |= STATX_ATTR_VERITY;
819
820 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
821 STATX_ATTR_APPEND |
822 STATX_ATTR_ENCRYPTED |
823 STATX_ATTR_IMMUTABLE |
824 STATX_ATTR_NODUMP |
825 STATX_ATTR_VERITY);
826
827 generic_fillattr(&init_user_ns, inode, stat);
828
829 /* we need to show initial sectors used for inline_data/dentries */
830 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
831 f2fs_has_inline_dentry(inode))
832 stat->blocks += (stat->size + 511) >> 9;
833
834 return 0;
835}
836
837#ifdef CONFIG_F2FS_FS_POSIX_ACL
838static void __setattr_copy(struct user_namespace *mnt_userns,
839 struct inode *inode, const struct iattr *attr)
840{
841 unsigned int ia_valid = attr->ia_valid;
842
843 if (ia_valid & ATTR_UID)
844 inode->i_uid = attr->ia_uid;
845 if (ia_valid & ATTR_GID)
846 inode->i_gid = attr->ia_gid;
847 if (ia_valid & ATTR_ATIME)
848 inode->i_atime = attr->ia_atime;
849 if (ia_valid & ATTR_MTIME)
850 inode->i_mtime = attr->ia_mtime;
851 if (ia_valid & ATTR_CTIME)
852 inode->i_ctime = attr->ia_ctime;
853 if (ia_valid & ATTR_MODE) {
854 umode_t mode = attr->ia_mode;
855 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
856
857 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
858 mode &= ~S_ISGID;
859 set_acl_inode(inode, mode);
860 }
861}
862#else
863#define __setattr_copy setattr_copy
864#endif
865
866int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
867 struct iattr *attr)
868{
869 struct inode *inode = d_inode(dentry);
870 int err;
871
872 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
873 return -EIO;
874
875 if (unlikely(IS_IMMUTABLE(inode)))
876 return -EPERM;
877
878 if (unlikely(IS_APPEND(inode) &&
879 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
880 ATTR_GID | ATTR_TIMES_SET))))
881 return -EPERM;
882
883 if ((attr->ia_valid & ATTR_SIZE) &&
884 !f2fs_is_compress_backend_ready(inode))
885 return -EOPNOTSUPP;
886
887 err = setattr_prepare(&init_user_ns, dentry, attr);
888 if (err)
889 return err;
890
891 err = fscrypt_prepare_setattr(dentry, attr);
892 if (err)
893 return err;
894
895 err = fsverity_prepare_setattr(dentry, attr);
896 if (err)
897 return err;
898
899 if (is_quota_modification(inode, attr)) {
900 err = dquot_initialize(inode);
901 if (err)
902 return err;
903 }
904 if ((attr->ia_valid & ATTR_UID &&
905 !uid_eq(attr->ia_uid, inode->i_uid)) ||
906 (attr->ia_valid & ATTR_GID &&
907 !gid_eq(attr->ia_gid, inode->i_gid))) {
908 f2fs_lock_op(F2FS_I_SB(inode));
909 err = dquot_transfer(inode, attr);
910 if (err) {
911 set_sbi_flag(F2FS_I_SB(inode),
912 SBI_QUOTA_NEED_REPAIR);
913 f2fs_unlock_op(F2FS_I_SB(inode));
914 return err;
915 }
916 /*
917 * update uid/gid under lock_op(), so that dquot and inode can
918 * be updated atomically.
919 */
920 if (attr->ia_valid & ATTR_UID)
921 inode->i_uid = attr->ia_uid;
922 if (attr->ia_valid & ATTR_GID)
923 inode->i_gid = attr->ia_gid;
924 f2fs_mark_inode_dirty_sync(inode, true);
925 f2fs_unlock_op(F2FS_I_SB(inode));
926 }
927
928 if (attr->ia_valid & ATTR_SIZE) {
929 loff_t old_size = i_size_read(inode);
930
931 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
932 /*
933 * should convert inline inode before i_size_write to
934 * keep smaller than inline_data size with inline flag.
935 */
936 err = f2fs_convert_inline_inode(inode);
937 if (err)
938 return err;
939 }
940
941 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
942 down_write(&F2FS_I(inode)->i_mmap_sem);
943
944 truncate_setsize(inode, attr->ia_size);
945
946 if (attr->ia_size <= old_size)
947 err = f2fs_truncate(inode);
948 /*
949 * do not trim all blocks after i_size if target size is
950 * larger than i_size.
951 */
952 up_write(&F2FS_I(inode)->i_mmap_sem);
953 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
954 if (err)
955 return err;
956
957 spin_lock(&F2FS_I(inode)->i_size_lock);
958 inode->i_mtime = inode->i_ctime = current_time(inode);
959 F2FS_I(inode)->last_disk_size = i_size_read(inode);
960 spin_unlock(&F2FS_I(inode)->i_size_lock);
961 }
962
963 __setattr_copy(&init_user_ns, inode, attr);
964
965 if (attr->ia_valid & ATTR_MODE) {
966 err = posix_acl_chmod(&init_user_ns, inode, f2fs_get_inode_mode(inode));
967
968 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
969 if (!err)
970 inode->i_mode = F2FS_I(inode)->i_acl_mode;
971 clear_inode_flag(inode, FI_ACL_MODE);
972 }
973 }
974
975 /* file size may changed here */
976 f2fs_mark_inode_dirty_sync(inode, true);
977
978 /* inode change will produce dirty node pages flushed by checkpoint */
979 f2fs_balance_fs(F2FS_I_SB(inode), true);
980
981 return err;
982}
983
984const struct inode_operations f2fs_file_inode_operations = {
985 .getattr = f2fs_getattr,
986 .setattr = f2fs_setattr,
987 .get_acl = f2fs_get_acl,
988 .set_acl = f2fs_set_acl,
989 .listxattr = f2fs_listxattr,
990 .fiemap = f2fs_fiemap,
991 .fileattr_get = f2fs_fileattr_get,
992 .fileattr_set = f2fs_fileattr_set,
993};
994
995static int fill_zero(struct inode *inode, pgoff_t index,
996 loff_t start, loff_t len)
997{
998 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
999 struct page *page;
1000
1001 if (!len)
1002 return 0;
1003
1004 f2fs_balance_fs(sbi, true);
1005
1006 f2fs_lock_op(sbi);
1007 page = f2fs_get_new_data_page(inode, NULL, index, false);
1008 f2fs_unlock_op(sbi);
1009
1010 if (IS_ERR(page))
1011 return PTR_ERR(page);
1012
1013 f2fs_wait_on_page_writeback(page, DATA, true, true);
1014 zero_user(page, start, len);
1015 set_page_dirty(page);
1016 f2fs_put_page(page, 1);
1017 return 0;
1018}
1019
1020int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1021{
1022 int err;
1023
1024 while (pg_start < pg_end) {
1025 struct dnode_of_data dn;
1026 pgoff_t end_offset, count;
1027
1028 set_new_dnode(&dn, inode, NULL, NULL, 0);
1029 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1030 if (err) {
1031 if (err == -ENOENT) {
1032 pg_start = f2fs_get_next_page_offset(&dn,
1033 pg_start);
1034 continue;
1035 }
1036 return err;
1037 }
1038
1039 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1040 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1041
1042 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1043
1044 f2fs_truncate_data_blocks_range(&dn, count);
1045 f2fs_put_dnode(&dn);
1046
1047 pg_start += count;
1048 }
1049 return 0;
1050}
1051
1052static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1053{
1054 pgoff_t pg_start, pg_end;
1055 loff_t off_start, off_end;
1056 int ret;
1057
1058 ret = f2fs_convert_inline_inode(inode);
1059 if (ret)
1060 return ret;
1061
1062 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1063 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1064
1065 off_start = offset & (PAGE_SIZE - 1);
1066 off_end = (offset + len) & (PAGE_SIZE - 1);
1067
1068 if (pg_start == pg_end) {
1069 ret = fill_zero(inode, pg_start, off_start,
1070 off_end - off_start);
1071 if (ret)
1072 return ret;
1073 } else {
1074 if (off_start) {
1075 ret = fill_zero(inode, pg_start++, off_start,
1076 PAGE_SIZE - off_start);
1077 if (ret)
1078 return ret;
1079 }
1080 if (off_end) {
1081 ret = fill_zero(inode, pg_end, 0, off_end);
1082 if (ret)
1083 return ret;
1084 }
1085
1086 if (pg_start < pg_end) {
1087 loff_t blk_start, blk_end;
1088 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1089
1090 f2fs_balance_fs(sbi, true);
1091
1092 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1093 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1094
1095 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1096 down_write(&F2FS_I(inode)->i_mmap_sem);
1097
1098 truncate_pagecache_range(inode, blk_start, blk_end - 1);
1099
1100 f2fs_lock_op(sbi);
1101 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1102 f2fs_unlock_op(sbi);
1103
1104 up_write(&F2FS_I(inode)->i_mmap_sem);
1105 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1106 }
1107 }
1108
1109 return ret;
1110}
1111
1112static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1113 int *do_replace, pgoff_t off, pgoff_t len)
1114{
1115 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1116 struct dnode_of_data dn;
1117 int ret, done, i;
1118
1119next_dnode:
1120 set_new_dnode(&dn, inode, NULL, NULL, 0);
1121 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1122 if (ret && ret != -ENOENT) {
1123 return ret;
1124 } else if (ret == -ENOENT) {
1125 if (dn.max_level == 0)
1126 return -ENOENT;
1127 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1128 dn.ofs_in_node, len);
1129 blkaddr += done;
1130 do_replace += done;
1131 goto next;
1132 }
1133
1134 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1135 dn.ofs_in_node, len);
1136 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1137 *blkaddr = f2fs_data_blkaddr(&dn);
1138
1139 if (__is_valid_data_blkaddr(*blkaddr) &&
1140 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1141 DATA_GENERIC_ENHANCE)) {
1142 f2fs_put_dnode(&dn);
1143 return -EFSCORRUPTED;
1144 }
1145
1146 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1147
1148 if (f2fs_lfs_mode(sbi)) {
1149 f2fs_put_dnode(&dn);
1150 return -EOPNOTSUPP;
1151 }
1152
1153 /* do not invalidate this block address */
1154 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1155 *do_replace = 1;
1156 }
1157 }
1158 f2fs_put_dnode(&dn);
1159next:
1160 len -= done;
1161 off += done;
1162 if (len)
1163 goto next_dnode;
1164 return 0;
1165}
1166
1167static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1168 int *do_replace, pgoff_t off, int len)
1169{
1170 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1171 struct dnode_of_data dn;
1172 int ret, i;
1173
1174 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1175 if (*do_replace == 0)
1176 continue;
1177
1178 set_new_dnode(&dn, inode, NULL, NULL, 0);
1179 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1180 if (ret) {
1181 dec_valid_block_count(sbi, inode, 1);
1182 f2fs_invalidate_blocks(sbi, *blkaddr);
1183 } else {
1184 f2fs_update_data_blkaddr(&dn, *blkaddr);
1185 }
1186 f2fs_put_dnode(&dn);
1187 }
1188 return 0;
1189}
1190
1191static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1192 block_t *blkaddr, int *do_replace,
1193 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1194{
1195 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1196 pgoff_t i = 0;
1197 int ret;
1198
1199 while (i < len) {
1200 if (blkaddr[i] == NULL_ADDR && !full) {
1201 i++;
1202 continue;
1203 }
1204
1205 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1206 struct dnode_of_data dn;
1207 struct node_info ni;
1208 size_t new_size;
1209 pgoff_t ilen;
1210
1211 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1212 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1213 if (ret)
1214 return ret;
1215
1216 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1217 if (ret) {
1218 f2fs_put_dnode(&dn);
1219 return ret;
1220 }
1221
1222 ilen = min((pgoff_t)
1223 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1224 dn.ofs_in_node, len - i);
1225 do {
1226 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1227 f2fs_truncate_data_blocks_range(&dn, 1);
1228
1229 if (do_replace[i]) {
1230 f2fs_i_blocks_write(src_inode,
1231 1, false, false);
1232 f2fs_i_blocks_write(dst_inode,
1233 1, true, false);
1234 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1235 blkaddr[i], ni.version, true, false);
1236
1237 do_replace[i] = 0;
1238 }
1239 dn.ofs_in_node++;
1240 i++;
1241 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1242 if (dst_inode->i_size < new_size)
1243 f2fs_i_size_write(dst_inode, new_size);
1244 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1245
1246 f2fs_put_dnode(&dn);
1247 } else {
1248 struct page *psrc, *pdst;
1249
1250 psrc = f2fs_get_lock_data_page(src_inode,
1251 src + i, true);
1252 if (IS_ERR(psrc))
1253 return PTR_ERR(psrc);
1254 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1255 true);
1256 if (IS_ERR(pdst)) {
1257 f2fs_put_page(psrc, 1);
1258 return PTR_ERR(pdst);
1259 }
1260 f2fs_copy_page(psrc, pdst);
1261 set_page_dirty(pdst);
1262 f2fs_put_page(pdst, 1);
1263 f2fs_put_page(psrc, 1);
1264
1265 ret = f2fs_truncate_hole(src_inode,
1266 src + i, src + i + 1);
1267 if (ret)
1268 return ret;
1269 i++;
1270 }
1271 }
1272 return 0;
1273}
1274
1275static int __exchange_data_block(struct inode *src_inode,
1276 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1277 pgoff_t len, bool full)
1278{
1279 block_t *src_blkaddr;
1280 int *do_replace;
1281 pgoff_t olen;
1282 int ret;
1283
1284 while (len) {
1285 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1286
1287 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1288 array_size(olen, sizeof(block_t)),
1289 GFP_NOFS);
1290 if (!src_blkaddr)
1291 return -ENOMEM;
1292
1293 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1294 array_size(olen, sizeof(int)),
1295 GFP_NOFS);
1296 if (!do_replace) {
1297 kvfree(src_blkaddr);
1298 return -ENOMEM;
1299 }
1300
1301 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1302 do_replace, src, olen);
1303 if (ret)
1304 goto roll_back;
1305
1306 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1307 do_replace, src, dst, olen, full);
1308 if (ret)
1309 goto roll_back;
1310
1311 src += olen;
1312 dst += olen;
1313 len -= olen;
1314
1315 kvfree(src_blkaddr);
1316 kvfree(do_replace);
1317 }
1318 return 0;
1319
1320roll_back:
1321 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1322 kvfree(src_blkaddr);
1323 kvfree(do_replace);
1324 return ret;
1325}
1326
1327static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1328{
1329 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1330 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1331 pgoff_t start = offset >> PAGE_SHIFT;
1332 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1333 int ret;
1334
1335 f2fs_balance_fs(sbi, true);
1336
1337 /* avoid gc operation during block exchange */
1338 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1339 down_write(&F2FS_I(inode)->i_mmap_sem);
1340
1341 f2fs_lock_op(sbi);
1342 f2fs_drop_extent_tree(inode);
1343 truncate_pagecache(inode, offset);
1344 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1345 f2fs_unlock_op(sbi);
1346
1347 up_write(&F2FS_I(inode)->i_mmap_sem);
1348 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1349 return ret;
1350}
1351
1352static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1353{
1354 loff_t new_size;
1355 int ret;
1356
1357 if (offset + len >= i_size_read(inode))
1358 return -EINVAL;
1359
1360 /* collapse range should be aligned to block size of f2fs. */
1361 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1362 return -EINVAL;
1363
1364 ret = f2fs_convert_inline_inode(inode);
1365 if (ret)
1366 return ret;
1367
1368 /* write out all dirty pages from offset */
1369 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1370 if (ret)
1371 return ret;
1372
1373 ret = f2fs_do_collapse(inode, offset, len);
1374 if (ret)
1375 return ret;
1376
1377 /* write out all moved pages, if possible */
1378 down_write(&F2FS_I(inode)->i_mmap_sem);
1379 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1380 truncate_pagecache(inode, offset);
1381
1382 new_size = i_size_read(inode) - len;
1383 ret = f2fs_truncate_blocks(inode, new_size, true);
1384 up_write(&F2FS_I(inode)->i_mmap_sem);
1385 if (!ret)
1386 f2fs_i_size_write(inode, new_size);
1387 return ret;
1388}
1389
1390static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1391 pgoff_t end)
1392{
1393 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1394 pgoff_t index = start;
1395 unsigned int ofs_in_node = dn->ofs_in_node;
1396 blkcnt_t count = 0;
1397 int ret;
1398
1399 for (; index < end; index++, dn->ofs_in_node++) {
1400 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1401 count++;
1402 }
1403
1404 dn->ofs_in_node = ofs_in_node;
1405 ret = f2fs_reserve_new_blocks(dn, count);
1406 if (ret)
1407 return ret;
1408
1409 dn->ofs_in_node = ofs_in_node;
1410 for (index = start; index < end; index++, dn->ofs_in_node++) {
1411 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1412 /*
1413 * f2fs_reserve_new_blocks will not guarantee entire block
1414 * allocation.
1415 */
1416 if (dn->data_blkaddr == NULL_ADDR) {
1417 ret = -ENOSPC;
1418 break;
1419 }
1420 if (dn->data_blkaddr != NEW_ADDR) {
1421 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1422 dn->data_blkaddr = NEW_ADDR;
1423 f2fs_set_data_blkaddr(dn);
1424 }
1425 }
1426
1427 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1428
1429 return ret;
1430}
1431
1432static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1433 int mode)
1434{
1435 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1436 struct address_space *mapping = inode->i_mapping;
1437 pgoff_t index, pg_start, pg_end;
1438 loff_t new_size = i_size_read(inode);
1439 loff_t off_start, off_end;
1440 int ret = 0;
1441
1442 ret = inode_newsize_ok(inode, (len + offset));
1443 if (ret)
1444 return ret;
1445
1446 ret = f2fs_convert_inline_inode(inode);
1447 if (ret)
1448 return ret;
1449
1450 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1451 if (ret)
1452 return ret;
1453
1454 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1455 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1456
1457 off_start = offset & (PAGE_SIZE - 1);
1458 off_end = (offset + len) & (PAGE_SIZE - 1);
1459
1460 if (pg_start == pg_end) {
1461 ret = fill_zero(inode, pg_start, off_start,
1462 off_end - off_start);
1463 if (ret)
1464 return ret;
1465
1466 new_size = max_t(loff_t, new_size, offset + len);
1467 } else {
1468 if (off_start) {
1469 ret = fill_zero(inode, pg_start++, off_start,
1470 PAGE_SIZE - off_start);
1471 if (ret)
1472 return ret;
1473
1474 new_size = max_t(loff_t, new_size,
1475 (loff_t)pg_start << PAGE_SHIFT);
1476 }
1477
1478 for (index = pg_start; index < pg_end;) {
1479 struct dnode_of_data dn;
1480 unsigned int end_offset;
1481 pgoff_t end;
1482
1483 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1484 down_write(&F2FS_I(inode)->i_mmap_sem);
1485
1486 truncate_pagecache_range(inode,
1487 (loff_t)index << PAGE_SHIFT,
1488 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1489
1490 f2fs_lock_op(sbi);
1491
1492 set_new_dnode(&dn, inode, NULL, NULL, 0);
1493 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1494 if (ret) {
1495 f2fs_unlock_op(sbi);
1496 up_write(&F2FS_I(inode)->i_mmap_sem);
1497 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1498 goto out;
1499 }
1500
1501 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1502 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1503
1504 ret = f2fs_do_zero_range(&dn, index, end);
1505 f2fs_put_dnode(&dn);
1506
1507 f2fs_unlock_op(sbi);
1508 up_write(&F2FS_I(inode)->i_mmap_sem);
1509 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1510
1511 f2fs_balance_fs(sbi, dn.node_changed);
1512
1513 if (ret)
1514 goto out;
1515
1516 index = end;
1517 new_size = max_t(loff_t, new_size,
1518 (loff_t)index << PAGE_SHIFT);
1519 }
1520
1521 if (off_end) {
1522 ret = fill_zero(inode, pg_end, 0, off_end);
1523 if (ret)
1524 goto out;
1525
1526 new_size = max_t(loff_t, new_size, offset + len);
1527 }
1528 }
1529
1530out:
1531 if (new_size > i_size_read(inode)) {
1532 if (mode & FALLOC_FL_KEEP_SIZE)
1533 file_set_keep_isize(inode);
1534 else
1535 f2fs_i_size_write(inode, new_size);
1536 }
1537 return ret;
1538}
1539
1540static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1541{
1542 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1543 pgoff_t nr, pg_start, pg_end, delta, idx;
1544 loff_t new_size;
1545 int ret = 0;
1546
1547 new_size = i_size_read(inode) + len;
1548 ret = inode_newsize_ok(inode, new_size);
1549 if (ret)
1550 return ret;
1551
1552 if (offset >= i_size_read(inode))
1553 return -EINVAL;
1554
1555 /* insert range should be aligned to block size of f2fs. */
1556 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1557 return -EINVAL;
1558
1559 ret = f2fs_convert_inline_inode(inode);
1560 if (ret)
1561 return ret;
1562
1563 f2fs_balance_fs(sbi, true);
1564
1565 down_write(&F2FS_I(inode)->i_mmap_sem);
1566 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1567 up_write(&F2FS_I(inode)->i_mmap_sem);
1568 if (ret)
1569 return ret;
1570
1571 /* write out all dirty pages from offset */
1572 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1573 if (ret)
1574 return ret;
1575
1576 pg_start = offset >> PAGE_SHIFT;
1577 pg_end = (offset + len) >> PAGE_SHIFT;
1578 delta = pg_end - pg_start;
1579 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1580
1581 /* avoid gc operation during block exchange */
1582 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1583 down_write(&F2FS_I(inode)->i_mmap_sem);
1584 truncate_pagecache(inode, offset);
1585
1586 while (!ret && idx > pg_start) {
1587 nr = idx - pg_start;
1588 if (nr > delta)
1589 nr = delta;
1590 idx -= nr;
1591
1592 f2fs_lock_op(sbi);
1593 f2fs_drop_extent_tree(inode);
1594
1595 ret = __exchange_data_block(inode, inode, idx,
1596 idx + delta, nr, false);
1597 f2fs_unlock_op(sbi);
1598 }
1599 up_write(&F2FS_I(inode)->i_mmap_sem);
1600 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1601
1602 /* write out all moved pages, if possible */
1603 down_write(&F2FS_I(inode)->i_mmap_sem);
1604 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1605 truncate_pagecache(inode, offset);
1606 up_write(&F2FS_I(inode)->i_mmap_sem);
1607
1608 if (!ret)
1609 f2fs_i_size_write(inode, new_size);
1610 return ret;
1611}
1612
1613static int expand_inode_data(struct inode *inode, loff_t offset,
1614 loff_t len, int mode)
1615{
1616 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1617 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1618 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1619 .m_may_create = true };
1620 pgoff_t pg_start, pg_end;
1621 loff_t new_size = i_size_read(inode);
1622 loff_t off_end;
1623 block_t expanded = 0;
1624 int err;
1625
1626 err = inode_newsize_ok(inode, (len + offset));
1627 if (err)
1628 return err;
1629
1630 err = f2fs_convert_inline_inode(inode);
1631 if (err)
1632 return err;
1633
1634 f2fs_balance_fs(sbi, true);
1635
1636 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1637 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1638 off_end = (offset + len) & (PAGE_SIZE - 1);
1639
1640 map.m_lblk = pg_start;
1641 map.m_len = pg_end - pg_start;
1642 if (off_end)
1643 map.m_len++;
1644
1645 if (!map.m_len)
1646 return 0;
1647
1648 if (f2fs_is_pinned_file(inode)) {
1649 block_t sec_blks = BLKS_PER_SEC(sbi);
1650 block_t sec_len = roundup(map.m_len, sec_blks);
1651
1652 map.m_len = sec_blks;
1653next_alloc:
1654 if (has_not_enough_free_secs(sbi, 0,
1655 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1656 down_write(&sbi->gc_lock);
1657 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1658 if (err && err != -ENODATA && err != -EAGAIN)
1659 goto out_err;
1660 }
1661
1662 down_write(&sbi->pin_sem);
1663
1664 f2fs_lock_op(sbi);
1665 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1666 f2fs_unlock_op(sbi);
1667
1668 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1669 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1670
1671 up_write(&sbi->pin_sem);
1672
1673 expanded += map.m_len;
1674 sec_len -= map.m_len;
1675 map.m_lblk += map.m_len;
1676 if (!err && sec_len)
1677 goto next_alloc;
1678
1679 map.m_len = expanded;
1680 } else {
1681 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1682 expanded = map.m_len;
1683 }
1684out_err:
1685 if (err) {
1686 pgoff_t last_off;
1687
1688 if (!expanded)
1689 return err;
1690
1691 last_off = pg_start + expanded - 1;
1692
1693 /* update new size to the failed position */
1694 new_size = (last_off == pg_end) ? offset + len :
1695 (loff_t)(last_off + 1) << PAGE_SHIFT;
1696 } else {
1697 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1698 }
1699
1700 if (new_size > i_size_read(inode)) {
1701 if (mode & FALLOC_FL_KEEP_SIZE)
1702 file_set_keep_isize(inode);
1703 else
1704 f2fs_i_size_write(inode, new_size);
1705 }
1706
1707 return err;
1708}
1709
1710static long f2fs_fallocate(struct file *file, int mode,
1711 loff_t offset, loff_t len)
1712{
1713 struct inode *inode = file_inode(file);
1714 long ret = 0;
1715
1716 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1717 return -EIO;
1718 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1719 return -ENOSPC;
1720 if (!f2fs_is_compress_backend_ready(inode))
1721 return -EOPNOTSUPP;
1722
1723 /* f2fs only support ->fallocate for regular file */
1724 if (!S_ISREG(inode->i_mode))
1725 return -EINVAL;
1726
1727 if (IS_ENCRYPTED(inode) &&
1728 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1729 return -EOPNOTSUPP;
1730
1731 if (f2fs_compressed_file(inode) &&
1732 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1733 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1734 return -EOPNOTSUPP;
1735
1736 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1737 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1738 FALLOC_FL_INSERT_RANGE))
1739 return -EOPNOTSUPP;
1740
1741 inode_lock(inode);
1742
1743 if (mode & FALLOC_FL_PUNCH_HOLE) {
1744 if (offset >= inode->i_size)
1745 goto out;
1746
1747 ret = punch_hole(inode, offset, len);
1748 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1749 ret = f2fs_collapse_range(inode, offset, len);
1750 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1751 ret = f2fs_zero_range(inode, offset, len, mode);
1752 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1753 ret = f2fs_insert_range(inode, offset, len);
1754 } else {
1755 ret = expand_inode_data(inode, offset, len, mode);
1756 }
1757
1758 if (!ret) {
1759 inode->i_mtime = inode->i_ctime = current_time(inode);
1760 f2fs_mark_inode_dirty_sync(inode, false);
1761 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1762 }
1763
1764out:
1765 inode_unlock(inode);
1766
1767 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1768 return ret;
1769}
1770
1771static int f2fs_release_file(struct inode *inode, struct file *filp)
1772{
1773 /*
1774 * f2fs_relase_file is called at every close calls. So we should
1775 * not drop any inmemory pages by close called by other process.
1776 */
1777 if (!(filp->f_mode & FMODE_WRITE) ||
1778 atomic_read(&inode->i_writecount) != 1)
1779 return 0;
1780
1781 /* some remained atomic pages should discarded */
1782 if (f2fs_is_atomic_file(inode))
1783 f2fs_drop_inmem_pages(inode);
1784 if (f2fs_is_volatile_file(inode)) {
1785 set_inode_flag(inode, FI_DROP_CACHE);
1786 filemap_fdatawrite(inode->i_mapping);
1787 clear_inode_flag(inode, FI_DROP_CACHE);
1788 clear_inode_flag(inode, FI_VOLATILE_FILE);
1789 stat_dec_volatile_write(inode);
1790 }
1791 return 0;
1792}
1793
1794static int f2fs_file_flush(struct file *file, fl_owner_t id)
1795{
1796 struct inode *inode = file_inode(file);
1797
1798 /*
1799 * If the process doing a transaction is crashed, we should do
1800 * roll-back. Otherwise, other reader/write can see corrupted database
1801 * until all the writers close its file. Since this should be done
1802 * before dropping file lock, it needs to do in ->flush.
1803 */
1804 if (f2fs_is_atomic_file(inode) &&
1805 F2FS_I(inode)->inmem_task == current)
1806 f2fs_drop_inmem_pages(inode);
1807 return 0;
1808}
1809
1810static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1811{
1812 struct f2fs_inode_info *fi = F2FS_I(inode);
1813 u32 masked_flags = fi->i_flags & mask;
1814
1815 /* mask can be shrunk by flags_valid selector */
1816 iflags &= mask;
1817
1818 /* Is it quota file? Do not allow user to mess with it */
1819 if (IS_NOQUOTA(inode))
1820 return -EPERM;
1821
1822 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1823 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1824 return -EOPNOTSUPP;
1825 if (!f2fs_empty_dir(inode))
1826 return -ENOTEMPTY;
1827 }
1828
1829 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1830 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1831 return -EOPNOTSUPP;
1832 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1833 return -EINVAL;
1834 }
1835
1836 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1837 if (masked_flags & F2FS_COMPR_FL) {
1838 if (!f2fs_disable_compressed_file(inode))
1839 return -EINVAL;
1840 }
1841 if (iflags & F2FS_NOCOMP_FL)
1842 return -EINVAL;
1843 if (iflags & F2FS_COMPR_FL) {
1844 if (!f2fs_may_compress(inode))
1845 return -EINVAL;
1846 if (S_ISREG(inode->i_mode) && inode->i_size)
1847 return -EINVAL;
1848
1849 set_compress_context(inode);
1850 }
1851 }
1852 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1853 if (masked_flags & F2FS_COMPR_FL)
1854 return -EINVAL;
1855 }
1856
1857 fi->i_flags = iflags | (fi->i_flags & ~mask);
1858 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1859 (fi->i_flags & F2FS_NOCOMP_FL));
1860
1861 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1862 set_inode_flag(inode, FI_PROJ_INHERIT);
1863 else
1864 clear_inode_flag(inode, FI_PROJ_INHERIT);
1865
1866 inode->i_ctime = current_time(inode);
1867 f2fs_set_inode_flags(inode);
1868 f2fs_mark_inode_dirty_sync(inode, true);
1869 return 0;
1870}
1871
1872/* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1873
1874/*
1875 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1876 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1877 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1878 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1879 *
1880 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1881 * FS_IOC_FSSETXATTR is done by the VFS.
1882 */
1883
1884static const struct {
1885 u32 iflag;
1886 u32 fsflag;
1887} f2fs_fsflags_map[] = {
1888 { F2FS_COMPR_FL, FS_COMPR_FL },
1889 { F2FS_SYNC_FL, FS_SYNC_FL },
1890 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1891 { F2FS_APPEND_FL, FS_APPEND_FL },
1892 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1893 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1894 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
1895 { F2FS_INDEX_FL, FS_INDEX_FL },
1896 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1897 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1898 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1899};
1900
1901#define F2FS_GETTABLE_FS_FL ( \
1902 FS_COMPR_FL | \
1903 FS_SYNC_FL | \
1904 FS_IMMUTABLE_FL | \
1905 FS_APPEND_FL | \
1906 FS_NODUMP_FL | \
1907 FS_NOATIME_FL | \
1908 FS_NOCOMP_FL | \
1909 FS_INDEX_FL | \
1910 FS_DIRSYNC_FL | \
1911 FS_PROJINHERIT_FL | \
1912 FS_ENCRYPT_FL | \
1913 FS_INLINE_DATA_FL | \
1914 FS_NOCOW_FL | \
1915 FS_VERITY_FL | \
1916 FS_CASEFOLD_FL)
1917
1918#define F2FS_SETTABLE_FS_FL ( \
1919 FS_COMPR_FL | \
1920 FS_SYNC_FL | \
1921 FS_IMMUTABLE_FL | \
1922 FS_APPEND_FL | \
1923 FS_NODUMP_FL | \
1924 FS_NOATIME_FL | \
1925 FS_NOCOMP_FL | \
1926 FS_DIRSYNC_FL | \
1927 FS_PROJINHERIT_FL | \
1928 FS_CASEFOLD_FL)
1929
1930/* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1931static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1932{
1933 u32 fsflags = 0;
1934 int i;
1935
1936 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1937 if (iflags & f2fs_fsflags_map[i].iflag)
1938 fsflags |= f2fs_fsflags_map[i].fsflag;
1939
1940 return fsflags;
1941}
1942
1943/* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1944static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1945{
1946 u32 iflags = 0;
1947 int i;
1948
1949 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1950 if (fsflags & f2fs_fsflags_map[i].fsflag)
1951 iflags |= f2fs_fsflags_map[i].iflag;
1952
1953 return iflags;
1954}
1955
1956static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1957{
1958 struct inode *inode = file_inode(filp);
1959
1960 return put_user(inode->i_generation, (int __user *)arg);
1961}
1962
1963static int f2fs_ioc_start_atomic_write(struct file *filp)
1964{
1965 struct inode *inode = file_inode(filp);
1966 struct f2fs_inode_info *fi = F2FS_I(inode);
1967 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1968 int ret;
1969
1970 if (!inode_owner_or_capable(&init_user_ns, inode))
1971 return -EACCES;
1972
1973 if (!S_ISREG(inode->i_mode))
1974 return -EINVAL;
1975
1976 if (filp->f_flags & O_DIRECT)
1977 return -EINVAL;
1978
1979 ret = mnt_want_write_file(filp);
1980 if (ret)
1981 return ret;
1982
1983 inode_lock(inode);
1984
1985 f2fs_disable_compressed_file(inode);
1986
1987 if (f2fs_is_atomic_file(inode)) {
1988 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1989 ret = -EINVAL;
1990 goto out;
1991 }
1992
1993 ret = f2fs_convert_inline_inode(inode);
1994 if (ret)
1995 goto out;
1996
1997 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1998
1999 /*
2000 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2001 * f2fs_is_atomic_file.
2002 */
2003 if (get_dirty_pages(inode))
2004 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2005 inode->i_ino, get_dirty_pages(inode));
2006 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2007 if (ret) {
2008 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2009 goto out;
2010 }
2011
2012 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2013 if (list_empty(&fi->inmem_ilist))
2014 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2015 sbi->atomic_files++;
2016 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2017
2018 /* add inode in inmem_list first and set atomic_file */
2019 set_inode_flag(inode, FI_ATOMIC_FILE);
2020 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2021 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2022
2023 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2024 F2FS_I(inode)->inmem_task = current;
2025 stat_update_max_atomic_write(inode);
2026out:
2027 inode_unlock(inode);
2028 mnt_drop_write_file(filp);
2029 return ret;
2030}
2031
2032static int f2fs_ioc_commit_atomic_write(struct file *filp)
2033{
2034 struct inode *inode = file_inode(filp);
2035 int ret;
2036
2037 if (!inode_owner_or_capable(&init_user_ns, inode))
2038 return -EACCES;
2039
2040 ret = mnt_want_write_file(filp);
2041 if (ret)
2042 return ret;
2043
2044 f2fs_balance_fs(F2FS_I_SB(inode), true);
2045
2046 inode_lock(inode);
2047
2048 if (f2fs_is_volatile_file(inode)) {
2049 ret = -EINVAL;
2050 goto err_out;
2051 }
2052
2053 if (f2fs_is_atomic_file(inode)) {
2054 ret = f2fs_commit_inmem_pages(inode);
2055 if (ret)
2056 goto err_out;
2057
2058 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2059 if (!ret)
2060 f2fs_drop_inmem_pages(inode);
2061 } else {
2062 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2063 }
2064err_out:
2065 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2066 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2067 ret = -EINVAL;
2068 }
2069 inode_unlock(inode);
2070 mnt_drop_write_file(filp);
2071 return ret;
2072}
2073
2074static int f2fs_ioc_start_volatile_write(struct file *filp)
2075{
2076 struct inode *inode = file_inode(filp);
2077 int ret;
2078
2079 if (!inode_owner_or_capable(&init_user_ns, inode))
2080 return -EACCES;
2081
2082 if (!S_ISREG(inode->i_mode))
2083 return -EINVAL;
2084
2085 ret = mnt_want_write_file(filp);
2086 if (ret)
2087 return ret;
2088
2089 inode_lock(inode);
2090
2091 if (f2fs_is_volatile_file(inode))
2092 goto out;
2093
2094 ret = f2fs_convert_inline_inode(inode);
2095 if (ret)
2096 goto out;
2097
2098 stat_inc_volatile_write(inode);
2099 stat_update_max_volatile_write(inode);
2100
2101 set_inode_flag(inode, FI_VOLATILE_FILE);
2102 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2103out:
2104 inode_unlock(inode);
2105 mnt_drop_write_file(filp);
2106 return ret;
2107}
2108
2109static int f2fs_ioc_release_volatile_write(struct file *filp)
2110{
2111 struct inode *inode = file_inode(filp);
2112 int ret;
2113
2114 if (!inode_owner_or_capable(&init_user_ns, inode))
2115 return -EACCES;
2116
2117 ret = mnt_want_write_file(filp);
2118 if (ret)
2119 return ret;
2120
2121 inode_lock(inode);
2122
2123 if (!f2fs_is_volatile_file(inode))
2124 goto out;
2125
2126 if (!f2fs_is_first_block_written(inode)) {
2127 ret = truncate_partial_data_page(inode, 0, true);
2128 goto out;
2129 }
2130
2131 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2132out:
2133 inode_unlock(inode);
2134 mnt_drop_write_file(filp);
2135 return ret;
2136}
2137
2138static int f2fs_ioc_abort_volatile_write(struct file *filp)
2139{
2140 struct inode *inode = file_inode(filp);
2141 int ret;
2142
2143 if (!inode_owner_or_capable(&init_user_ns, inode))
2144 return -EACCES;
2145
2146 ret = mnt_want_write_file(filp);
2147 if (ret)
2148 return ret;
2149
2150 inode_lock(inode);
2151
2152 if (f2fs_is_atomic_file(inode))
2153 f2fs_drop_inmem_pages(inode);
2154 if (f2fs_is_volatile_file(inode)) {
2155 clear_inode_flag(inode, FI_VOLATILE_FILE);
2156 stat_dec_volatile_write(inode);
2157 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2158 }
2159
2160 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2161
2162 inode_unlock(inode);
2163
2164 mnt_drop_write_file(filp);
2165 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2166 return ret;
2167}
2168
2169static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2170{
2171 struct inode *inode = file_inode(filp);
2172 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2173 struct super_block *sb = sbi->sb;
2174 __u32 in;
2175 int ret = 0;
2176
2177 if (!capable(CAP_SYS_ADMIN))
2178 return -EPERM;
2179
2180 if (get_user(in, (__u32 __user *)arg))
2181 return -EFAULT;
2182
2183 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2184 ret = mnt_want_write_file(filp);
2185 if (ret) {
2186 if (ret == -EROFS) {
2187 ret = 0;
2188 f2fs_stop_checkpoint(sbi, false);
2189 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2190 trace_f2fs_shutdown(sbi, in, ret);
2191 }
2192 return ret;
2193 }
2194 }
2195
2196 switch (in) {
2197 case F2FS_GOING_DOWN_FULLSYNC:
2198 ret = freeze_bdev(sb->s_bdev);
2199 if (ret)
2200 goto out;
2201 f2fs_stop_checkpoint(sbi, false);
2202 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2203 thaw_bdev(sb->s_bdev);
2204 break;
2205 case F2FS_GOING_DOWN_METASYNC:
2206 /* do checkpoint only */
2207 ret = f2fs_sync_fs(sb, 1);
2208 if (ret)
2209 goto out;
2210 f2fs_stop_checkpoint(sbi, false);
2211 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2212 break;
2213 case F2FS_GOING_DOWN_NOSYNC:
2214 f2fs_stop_checkpoint(sbi, false);
2215 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2216 break;
2217 case F2FS_GOING_DOWN_METAFLUSH:
2218 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2219 f2fs_stop_checkpoint(sbi, false);
2220 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2221 break;
2222 case F2FS_GOING_DOWN_NEED_FSCK:
2223 set_sbi_flag(sbi, SBI_NEED_FSCK);
2224 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2225 set_sbi_flag(sbi, SBI_IS_DIRTY);
2226 /* do checkpoint only */
2227 ret = f2fs_sync_fs(sb, 1);
2228 goto out;
2229 default:
2230 ret = -EINVAL;
2231 goto out;
2232 }
2233
2234 f2fs_stop_gc_thread(sbi);
2235 f2fs_stop_discard_thread(sbi);
2236
2237 f2fs_drop_discard_cmd(sbi);
2238 clear_opt(sbi, DISCARD);
2239
2240 f2fs_update_time(sbi, REQ_TIME);
2241out:
2242 if (in != F2FS_GOING_DOWN_FULLSYNC)
2243 mnt_drop_write_file(filp);
2244
2245 trace_f2fs_shutdown(sbi, in, ret);
2246
2247 return ret;
2248}
2249
2250static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2251{
2252 struct inode *inode = file_inode(filp);
2253 struct super_block *sb = inode->i_sb;
2254 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2255 struct fstrim_range range;
2256 int ret;
2257
2258 if (!capable(CAP_SYS_ADMIN))
2259 return -EPERM;
2260
2261 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2262 return -EOPNOTSUPP;
2263
2264 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2265 sizeof(range)))
2266 return -EFAULT;
2267
2268 ret = mnt_want_write_file(filp);
2269 if (ret)
2270 return ret;
2271
2272 range.minlen = max((unsigned int)range.minlen,
2273 q->limits.discard_granularity);
2274 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2275 mnt_drop_write_file(filp);
2276 if (ret < 0)
2277 return ret;
2278
2279 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2280 sizeof(range)))
2281 return -EFAULT;
2282 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2283 return 0;
2284}
2285
2286static bool uuid_is_nonzero(__u8 u[16])
2287{
2288 int i;
2289
2290 for (i = 0; i < 16; i++)
2291 if (u[i])
2292 return true;
2293 return false;
2294}
2295
2296static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2297{
2298 struct inode *inode = file_inode(filp);
2299
2300 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2301 return -EOPNOTSUPP;
2302
2303 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2304
2305 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2306}
2307
2308static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2309{
2310 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2311 return -EOPNOTSUPP;
2312 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2313}
2314
2315static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2316{
2317 struct inode *inode = file_inode(filp);
2318 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2319 int err;
2320
2321 if (!f2fs_sb_has_encrypt(sbi))
2322 return -EOPNOTSUPP;
2323
2324 err = mnt_want_write_file(filp);
2325 if (err)
2326 return err;
2327
2328 down_write(&sbi->sb_lock);
2329
2330 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2331 goto got_it;
2332
2333 /* update superblock with uuid */
2334 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2335
2336 err = f2fs_commit_super(sbi, false);
2337 if (err) {
2338 /* undo new data */
2339 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2340 goto out_err;
2341 }
2342got_it:
2343 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2344 16))
2345 err = -EFAULT;
2346out_err:
2347 up_write(&sbi->sb_lock);
2348 mnt_drop_write_file(filp);
2349 return err;
2350}
2351
2352static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2353 unsigned long arg)
2354{
2355 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2356 return -EOPNOTSUPP;
2357
2358 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2359}
2360
2361static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2362{
2363 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2364 return -EOPNOTSUPP;
2365
2366 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2367}
2368
2369static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2370{
2371 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2372 return -EOPNOTSUPP;
2373
2374 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2375}
2376
2377static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2378 unsigned long arg)
2379{
2380 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2381 return -EOPNOTSUPP;
2382
2383 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2384}
2385
2386static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2387 unsigned long arg)
2388{
2389 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2390 return -EOPNOTSUPP;
2391
2392 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2393}
2394
2395static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2396{
2397 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2398 return -EOPNOTSUPP;
2399
2400 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2401}
2402
2403static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2404{
2405 struct inode *inode = file_inode(filp);
2406 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2407 __u32 sync;
2408 int ret;
2409
2410 if (!capable(CAP_SYS_ADMIN))
2411 return -EPERM;
2412
2413 if (get_user(sync, (__u32 __user *)arg))
2414 return -EFAULT;
2415
2416 if (f2fs_readonly(sbi->sb))
2417 return -EROFS;
2418
2419 ret = mnt_want_write_file(filp);
2420 if (ret)
2421 return ret;
2422
2423 if (!sync) {
2424 if (!down_write_trylock(&sbi->gc_lock)) {
2425 ret = -EBUSY;
2426 goto out;
2427 }
2428 } else {
2429 down_write(&sbi->gc_lock);
2430 }
2431
2432 ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO);
2433out:
2434 mnt_drop_write_file(filp);
2435 return ret;
2436}
2437
2438static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2439{
2440 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2441 u64 end;
2442 int ret;
2443
2444 if (!capable(CAP_SYS_ADMIN))
2445 return -EPERM;
2446 if (f2fs_readonly(sbi->sb))
2447 return -EROFS;
2448
2449 end = range->start + range->len;
2450 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2451 end >= MAX_BLKADDR(sbi))
2452 return -EINVAL;
2453
2454 ret = mnt_want_write_file(filp);
2455 if (ret)
2456 return ret;
2457
2458do_more:
2459 if (!range->sync) {
2460 if (!down_write_trylock(&sbi->gc_lock)) {
2461 ret = -EBUSY;
2462 goto out;
2463 }
2464 } else {
2465 down_write(&sbi->gc_lock);
2466 }
2467
2468 ret = f2fs_gc(sbi, range->sync, true, false,
2469 GET_SEGNO(sbi, range->start));
2470 if (ret) {
2471 if (ret == -EBUSY)
2472 ret = -EAGAIN;
2473 goto out;
2474 }
2475 range->start += BLKS_PER_SEC(sbi);
2476 if (range->start <= end)
2477 goto do_more;
2478out:
2479 mnt_drop_write_file(filp);
2480 return ret;
2481}
2482
2483static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2484{
2485 struct f2fs_gc_range range;
2486
2487 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2488 sizeof(range)))
2489 return -EFAULT;
2490 return __f2fs_ioc_gc_range(filp, &range);
2491}
2492
2493static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2494{
2495 struct inode *inode = file_inode(filp);
2496 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2497 int ret;
2498
2499 if (!capable(CAP_SYS_ADMIN))
2500 return -EPERM;
2501
2502 if (f2fs_readonly(sbi->sb))
2503 return -EROFS;
2504
2505 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2506 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2507 return -EINVAL;
2508 }
2509
2510 ret = mnt_want_write_file(filp);
2511 if (ret)
2512 return ret;
2513
2514 ret = f2fs_sync_fs(sbi->sb, 1);
2515
2516 mnt_drop_write_file(filp);
2517 return ret;
2518}
2519
2520static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2521 struct file *filp,
2522 struct f2fs_defragment *range)
2523{
2524 struct inode *inode = file_inode(filp);
2525 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2526 .m_seg_type = NO_CHECK_TYPE,
2527 .m_may_create = false };
2528 struct extent_info ei = {0, 0, 0};
2529 pgoff_t pg_start, pg_end, next_pgofs;
2530 unsigned int blk_per_seg = sbi->blocks_per_seg;
2531 unsigned int total = 0, sec_num;
2532 block_t blk_end = 0;
2533 bool fragmented = false;
2534 int err;
2535
2536 /* if in-place-update policy is enabled, don't waste time here */
2537 if (f2fs_should_update_inplace(inode, NULL))
2538 return -EINVAL;
2539
2540 pg_start = range->start >> PAGE_SHIFT;
2541 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2542
2543 f2fs_balance_fs(sbi, true);
2544
2545 inode_lock(inode);
2546
2547 /* writeback all dirty pages in the range */
2548 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2549 range->start + range->len - 1);
2550 if (err)
2551 goto out;
2552
2553 /*
2554 * lookup mapping info in extent cache, skip defragmenting if physical
2555 * block addresses are continuous.
2556 */
2557 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2558 if (ei.fofs + ei.len >= pg_end)
2559 goto out;
2560 }
2561
2562 map.m_lblk = pg_start;
2563 map.m_next_pgofs = &next_pgofs;
2564
2565 /*
2566 * lookup mapping info in dnode page cache, skip defragmenting if all
2567 * physical block addresses are continuous even if there are hole(s)
2568 * in logical blocks.
2569 */
2570 while (map.m_lblk < pg_end) {
2571 map.m_len = pg_end - map.m_lblk;
2572 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2573 if (err)
2574 goto out;
2575
2576 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2577 map.m_lblk = next_pgofs;
2578 continue;
2579 }
2580
2581 if (blk_end && blk_end != map.m_pblk)
2582 fragmented = true;
2583
2584 /* record total count of block that we're going to move */
2585 total += map.m_len;
2586
2587 blk_end = map.m_pblk + map.m_len;
2588
2589 map.m_lblk += map.m_len;
2590 }
2591
2592 if (!fragmented) {
2593 total = 0;
2594 goto out;
2595 }
2596
2597 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2598
2599 /*
2600 * make sure there are enough free section for LFS allocation, this can
2601 * avoid defragment running in SSR mode when free section are allocated
2602 * intensively
2603 */
2604 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2605 err = -EAGAIN;
2606 goto out;
2607 }
2608
2609 map.m_lblk = pg_start;
2610 map.m_len = pg_end - pg_start;
2611 total = 0;
2612
2613 while (map.m_lblk < pg_end) {
2614 pgoff_t idx;
2615 int cnt = 0;
2616
2617do_map:
2618 map.m_len = pg_end - map.m_lblk;
2619 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2620 if (err)
2621 goto clear_out;
2622
2623 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2624 map.m_lblk = next_pgofs;
2625 goto check;
2626 }
2627
2628 set_inode_flag(inode, FI_DO_DEFRAG);
2629
2630 idx = map.m_lblk;
2631 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2632 struct page *page;
2633
2634 page = f2fs_get_lock_data_page(inode, idx, true);
2635 if (IS_ERR(page)) {
2636 err = PTR_ERR(page);
2637 goto clear_out;
2638 }
2639
2640 set_page_dirty(page);
2641 f2fs_put_page(page, 1);
2642
2643 idx++;
2644 cnt++;
2645 total++;
2646 }
2647
2648 map.m_lblk = idx;
2649check:
2650 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2651 goto do_map;
2652
2653 clear_inode_flag(inode, FI_DO_DEFRAG);
2654
2655 err = filemap_fdatawrite(inode->i_mapping);
2656 if (err)
2657 goto out;
2658 }
2659clear_out:
2660 clear_inode_flag(inode, FI_DO_DEFRAG);
2661out:
2662 inode_unlock(inode);
2663 if (!err)
2664 range->len = (u64)total << PAGE_SHIFT;
2665 return err;
2666}
2667
2668static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2669{
2670 struct inode *inode = file_inode(filp);
2671 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2672 struct f2fs_defragment range;
2673 int err;
2674
2675 if (!capable(CAP_SYS_ADMIN))
2676 return -EPERM;
2677
2678 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2679 return -EINVAL;
2680
2681 if (f2fs_readonly(sbi->sb))
2682 return -EROFS;
2683
2684 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2685 sizeof(range)))
2686 return -EFAULT;
2687
2688 /* verify alignment of offset & size */
2689 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2690 return -EINVAL;
2691
2692 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2693 max_file_blocks(inode)))
2694 return -EINVAL;
2695
2696 err = mnt_want_write_file(filp);
2697 if (err)
2698 return err;
2699
2700 err = f2fs_defragment_range(sbi, filp, &range);
2701 mnt_drop_write_file(filp);
2702
2703 f2fs_update_time(sbi, REQ_TIME);
2704 if (err < 0)
2705 return err;
2706
2707 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2708 sizeof(range)))
2709 return -EFAULT;
2710
2711 return 0;
2712}
2713
2714static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2715 struct file *file_out, loff_t pos_out, size_t len)
2716{
2717 struct inode *src = file_inode(file_in);
2718 struct inode *dst = file_inode(file_out);
2719 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2720 size_t olen = len, dst_max_i_size = 0;
2721 size_t dst_osize;
2722 int ret;
2723
2724 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2725 src->i_sb != dst->i_sb)
2726 return -EXDEV;
2727
2728 if (unlikely(f2fs_readonly(src->i_sb)))
2729 return -EROFS;
2730
2731 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2732 return -EINVAL;
2733
2734 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2735 return -EOPNOTSUPP;
2736
2737 if (pos_out < 0 || pos_in < 0)
2738 return -EINVAL;
2739
2740 if (src == dst) {
2741 if (pos_in == pos_out)
2742 return 0;
2743 if (pos_out > pos_in && pos_out < pos_in + len)
2744 return -EINVAL;
2745 }
2746
2747 inode_lock(src);
2748 if (src != dst) {
2749 ret = -EBUSY;
2750 if (!inode_trylock(dst))
2751 goto out;
2752 }
2753
2754 ret = -EINVAL;
2755 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2756 goto out_unlock;
2757 if (len == 0)
2758 olen = len = src->i_size - pos_in;
2759 if (pos_in + len == src->i_size)
2760 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2761 if (len == 0) {
2762 ret = 0;
2763 goto out_unlock;
2764 }
2765
2766 dst_osize = dst->i_size;
2767 if (pos_out + olen > dst->i_size)
2768 dst_max_i_size = pos_out + olen;
2769
2770 /* verify the end result is block aligned */
2771 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2772 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2773 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2774 goto out_unlock;
2775
2776 ret = f2fs_convert_inline_inode(src);
2777 if (ret)
2778 goto out_unlock;
2779
2780 ret = f2fs_convert_inline_inode(dst);
2781 if (ret)
2782 goto out_unlock;
2783
2784 /* write out all dirty pages from offset */
2785 ret = filemap_write_and_wait_range(src->i_mapping,
2786 pos_in, pos_in + len);
2787 if (ret)
2788 goto out_unlock;
2789
2790 ret = filemap_write_and_wait_range(dst->i_mapping,
2791 pos_out, pos_out + len);
2792 if (ret)
2793 goto out_unlock;
2794
2795 f2fs_balance_fs(sbi, true);
2796
2797 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2798 if (src != dst) {
2799 ret = -EBUSY;
2800 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2801 goto out_src;
2802 }
2803
2804 f2fs_lock_op(sbi);
2805 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2806 pos_out >> F2FS_BLKSIZE_BITS,
2807 len >> F2FS_BLKSIZE_BITS, false);
2808
2809 if (!ret) {
2810 if (dst_max_i_size)
2811 f2fs_i_size_write(dst, dst_max_i_size);
2812 else if (dst_osize != dst->i_size)
2813 f2fs_i_size_write(dst, dst_osize);
2814 }
2815 f2fs_unlock_op(sbi);
2816
2817 if (src != dst)
2818 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2819out_src:
2820 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2821out_unlock:
2822 if (src != dst)
2823 inode_unlock(dst);
2824out:
2825 inode_unlock(src);
2826 return ret;
2827}
2828
2829static int __f2fs_ioc_move_range(struct file *filp,
2830 struct f2fs_move_range *range)
2831{
2832 struct fd dst;
2833 int err;
2834
2835 if (!(filp->f_mode & FMODE_READ) ||
2836 !(filp->f_mode & FMODE_WRITE))
2837 return -EBADF;
2838
2839 dst = fdget(range->dst_fd);
2840 if (!dst.file)
2841 return -EBADF;
2842
2843 if (!(dst.file->f_mode & FMODE_WRITE)) {
2844 err = -EBADF;
2845 goto err_out;
2846 }
2847
2848 err = mnt_want_write_file(filp);
2849 if (err)
2850 goto err_out;
2851
2852 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2853 range->pos_out, range->len);
2854
2855 mnt_drop_write_file(filp);
2856err_out:
2857 fdput(dst);
2858 return err;
2859}
2860
2861static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2862{
2863 struct f2fs_move_range range;
2864
2865 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2866 sizeof(range)))
2867 return -EFAULT;
2868 return __f2fs_ioc_move_range(filp, &range);
2869}
2870
2871static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2872{
2873 struct inode *inode = file_inode(filp);
2874 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2875 struct sit_info *sm = SIT_I(sbi);
2876 unsigned int start_segno = 0, end_segno = 0;
2877 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2878 struct f2fs_flush_device range;
2879 int ret;
2880
2881 if (!capable(CAP_SYS_ADMIN))
2882 return -EPERM;
2883
2884 if (f2fs_readonly(sbi->sb))
2885 return -EROFS;
2886
2887 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2888 return -EINVAL;
2889
2890 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2891 sizeof(range)))
2892 return -EFAULT;
2893
2894 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2895 __is_large_section(sbi)) {
2896 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2897 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2898 return -EINVAL;
2899 }
2900
2901 ret = mnt_want_write_file(filp);
2902 if (ret)
2903 return ret;
2904
2905 if (range.dev_num != 0)
2906 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2907 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2908
2909 start_segno = sm->last_victim[FLUSH_DEVICE];
2910 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2911 start_segno = dev_start_segno;
2912 end_segno = min(start_segno + range.segments, dev_end_segno);
2913
2914 while (start_segno < end_segno) {
2915 if (!down_write_trylock(&sbi->gc_lock)) {
2916 ret = -EBUSY;
2917 goto out;
2918 }
2919 sm->last_victim[GC_CB] = end_segno + 1;
2920 sm->last_victim[GC_GREEDY] = end_segno + 1;
2921 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2922 ret = f2fs_gc(sbi, true, true, true, start_segno);
2923 if (ret == -EAGAIN)
2924 ret = 0;
2925 else if (ret < 0)
2926 break;
2927 start_segno++;
2928 }
2929out:
2930 mnt_drop_write_file(filp);
2931 return ret;
2932}
2933
2934static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2935{
2936 struct inode *inode = file_inode(filp);
2937 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2938
2939 /* Must validate to set it with SQLite behavior in Android. */
2940 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2941
2942 return put_user(sb_feature, (u32 __user *)arg);
2943}
2944
2945#ifdef CONFIG_QUOTA
2946int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2947{
2948 struct dquot *transfer_to[MAXQUOTAS] = {};
2949 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2950 struct super_block *sb = sbi->sb;
2951 int err = 0;
2952
2953 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2954 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2955 err = __dquot_transfer(inode, transfer_to);
2956 if (err)
2957 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2958 dqput(transfer_to[PRJQUOTA]);
2959 }
2960 return err;
2961}
2962
2963static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2964{
2965 struct f2fs_inode_info *fi = F2FS_I(inode);
2966 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2967 struct page *ipage;
2968 kprojid_t kprojid;
2969 int err;
2970
2971 if (!f2fs_sb_has_project_quota(sbi)) {
2972 if (projid != F2FS_DEF_PROJID)
2973 return -EOPNOTSUPP;
2974 else
2975 return 0;
2976 }
2977
2978 if (!f2fs_has_extra_attr(inode))
2979 return -EOPNOTSUPP;
2980
2981 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2982
2983 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2984 return 0;
2985
2986 err = -EPERM;
2987 /* Is it quota file? Do not allow user to mess with it */
2988 if (IS_NOQUOTA(inode))
2989 return err;
2990
2991 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2992 if (IS_ERR(ipage))
2993 return PTR_ERR(ipage);
2994
2995 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2996 i_projid)) {
2997 err = -EOVERFLOW;
2998 f2fs_put_page(ipage, 1);
2999 return err;
3000 }
3001 f2fs_put_page(ipage, 1);
3002
3003 err = dquot_initialize(inode);
3004 if (err)
3005 return err;
3006
3007 f2fs_lock_op(sbi);
3008 err = f2fs_transfer_project_quota(inode, kprojid);
3009 if (err)
3010 goto out_unlock;
3011
3012 F2FS_I(inode)->i_projid = kprojid;
3013 inode->i_ctime = current_time(inode);
3014 f2fs_mark_inode_dirty_sync(inode, true);
3015out_unlock:
3016 f2fs_unlock_op(sbi);
3017 return err;
3018}
3019#else
3020int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3021{
3022 return 0;
3023}
3024
3025static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3026{
3027 if (projid != F2FS_DEF_PROJID)
3028 return -EOPNOTSUPP;
3029 return 0;
3030}
3031#endif
3032
3033int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3034{
3035 struct inode *inode = d_inode(dentry);
3036 struct f2fs_inode_info *fi = F2FS_I(inode);
3037 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3038
3039 if (IS_ENCRYPTED(inode))
3040 fsflags |= FS_ENCRYPT_FL;
3041 if (IS_VERITY(inode))
3042 fsflags |= FS_VERITY_FL;
3043 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3044 fsflags |= FS_INLINE_DATA_FL;
3045 if (is_inode_flag_set(inode, FI_PIN_FILE))
3046 fsflags |= FS_NOCOW_FL;
3047
3048 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3049
3050 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3051 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3052
3053 return 0;
3054}
3055
3056int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3057 struct dentry *dentry, struct fileattr *fa)
3058{
3059 struct inode *inode = d_inode(dentry);
3060 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3061 u32 iflags;
3062 int err;
3063
3064 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3065 return -EIO;
3066 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3067 return -ENOSPC;
3068 if (fsflags & ~F2FS_GETTABLE_FS_FL)
3069 return -EOPNOTSUPP;
3070 fsflags &= F2FS_SETTABLE_FS_FL;
3071 if (!fa->flags_valid)
3072 mask &= FS_COMMON_FL;
3073
3074 iflags = f2fs_fsflags_to_iflags(fsflags);
3075 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3076 return -EOPNOTSUPP;
3077
3078 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3079 if (!err)
3080 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3081
3082 return err;
3083}
3084
3085int f2fs_pin_file_control(struct inode *inode, bool inc)
3086{
3087 struct f2fs_inode_info *fi = F2FS_I(inode);
3088 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3089
3090 /* Use i_gc_failures for normal file as a risk signal. */
3091 if (inc)
3092 f2fs_i_gc_failures_write(inode,
3093 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3094
3095 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3096 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3097 __func__, inode->i_ino,
3098 fi->i_gc_failures[GC_FAILURE_PIN]);
3099 clear_inode_flag(inode, FI_PIN_FILE);
3100 return -EAGAIN;
3101 }
3102 return 0;
3103}
3104
3105static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3106{
3107 struct inode *inode = file_inode(filp);
3108 __u32 pin;
3109 int ret = 0;
3110
3111 if (get_user(pin, (__u32 __user *)arg))
3112 return -EFAULT;
3113
3114 if (!S_ISREG(inode->i_mode))
3115 return -EINVAL;
3116
3117 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3118 return -EROFS;
3119
3120 ret = mnt_want_write_file(filp);
3121 if (ret)
3122 return ret;
3123
3124 inode_lock(inode);
3125
3126 if (f2fs_should_update_outplace(inode, NULL)) {
3127 ret = -EINVAL;
3128 goto out;
3129 }
3130
3131 if (!pin) {
3132 clear_inode_flag(inode, FI_PIN_FILE);
3133 f2fs_i_gc_failures_write(inode, 0);
3134 goto done;
3135 }
3136
3137 if (f2fs_pin_file_control(inode, false)) {
3138 ret = -EAGAIN;
3139 goto out;
3140 }
3141
3142 ret = f2fs_convert_inline_inode(inode);
3143 if (ret)
3144 goto out;
3145
3146 if (!f2fs_disable_compressed_file(inode)) {
3147 ret = -EOPNOTSUPP;
3148 goto out;
3149 }
3150
3151 set_inode_flag(inode, FI_PIN_FILE);
3152 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3153done:
3154 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3155out:
3156 inode_unlock(inode);
3157 mnt_drop_write_file(filp);
3158 return ret;
3159}
3160
3161static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3162{
3163 struct inode *inode = file_inode(filp);
3164 __u32 pin = 0;
3165
3166 if (is_inode_flag_set(inode, FI_PIN_FILE))
3167 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3168 return put_user(pin, (u32 __user *)arg);
3169}
3170
3171int f2fs_precache_extents(struct inode *inode)
3172{
3173 struct f2fs_inode_info *fi = F2FS_I(inode);
3174 struct f2fs_map_blocks map;
3175 pgoff_t m_next_extent;
3176 loff_t end;
3177 int err;
3178
3179 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3180 return -EOPNOTSUPP;
3181
3182 map.m_lblk = 0;
3183 map.m_next_pgofs = NULL;
3184 map.m_next_extent = &m_next_extent;
3185 map.m_seg_type = NO_CHECK_TYPE;
3186 map.m_may_create = false;
3187 end = max_file_blocks(inode);
3188
3189 while (map.m_lblk < end) {
3190 map.m_len = end - map.m_lblk;
3191
3192 down_write(&fi->i_gc_rwsem[WRITE]);
3193 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3194 up_write(&fi->i_gc_rwsem[WRITE]);
3195 if (err)
3196 return err;
3197
3198 map.m_lblk = m_next_extent;
3199 }
3200
3201 return 0;
3202}
3203
3204static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3205{
3206 return f2fs_precache_extents(file_inode(filp));
3207}
3208
3209static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3210{
3211 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3212 __u64 block_count;
3213
3214 if (!capable(CAP_SYS_ADMIN))
3215 return -EPERM;
3216
3217 if (f2fs_readonly(sbi->sb))
3218 return -EROFS;
3219
3220 if (copy_from_user(&block_count, (void __user *)arg,
3221 sizeof(block_count)))
3222 return -EFAULT;
3223
3224 return f2fs_resize_fs(sbi, block_count);
3225}
3226
3227static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3228{
3229 struct inode *inode = file_inode(filp);
3230
3231 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3232
3233 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3234 f2fs_warn(F2FS_I_SB(inode),
3235 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3236 inode->i_ino);
3237 return -EOPNOTSUPP;
3238 }
3239
3240 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3241}
3242
3243static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3244{
3245 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3246 return -EOPNOTSUPP;
3247
3248 return fsverity_ioctl_measure(filp, (void __user *)arg);
3249}
3250
3251static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3252{
3253 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3254 return -EOPNOTSUPP;
3255
3256 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3257}
3258
3259static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3260{
3261 struct inode *inode = file_inode(filp);
3262 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3263 char *vbuf;
3264 int count;
3265 int err = 0;
3266
3267 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3268 if (!vbuf)
3269 return -ENOMEM;
3270
3271 down_read(&sbi->sb_lock);
3272 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3273 ARRAY_SIZE(sbi->raw_super->volume_name),
3274 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3275 up_read(&sbi->sb_lock);
3276
3277 if (copy_to_user((char __user *)arg, vbuf,
3278 min(FSLABEL_MAX, count)))
3279 err = -EFAULT;
3280
3281 kfree(vbuf);
3282 return err;
3283}
3284
3285static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3286{
3287 struct inode *inode = file_inode(filp);
3288 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3289 char *vbuf;
3290 int err = 0;
3291
3292 if (!capable(CAP_SYS_ADMIN))
3293 return -EPERM;
3294
3295 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3296 if (IS_ERR(vbuf))
3297 return PTR_ERR(vbuf);
3298
3299 err = mnt_want_write_file(filp);
3300 if (err)
3301 goto out;
3302
3303 down_write(&sbi->sb_lock);
3304
3305 memset(sbi->raw_super->volume_name, 0,
3306 sizeof(sbi->raw_super->volume_name));
3307 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3308 sbi->raw_super->volume_name,
3309 ARRAY_SIZE(sbi->raw_super->volume_name));
3310
3311 err = f2fs_commit_super(sbi, false);
3312
3313 up_write(&sbi->sb_lock);
3314
3315 mnt_drop_write_file(filp);
3316out:
3317 kfree(vbuf);
3318 return err;
3319}
3320
3321static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3322{
3323 struct inode *inode = file_inode(filp);
3324 __u64 blocks;
3325
3326 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3327 return -EOPNOTSUPP;
3328
3329 if (!f2fs_compressed_file(inode))
3330 return -EINVAL;
3331
3332 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3333 return put_user(blocks, (u64 __user *)arg);
3334}
3335
3336static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3337{
3338 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3339 unsigned int released_blocks = 0;
3340 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3341 block_t blkaddr;
3342 int i;
3343
3344 for (i = 0; i < count; i++) {
3345 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3346 dn->ofs_in_node + i);
3347
3348 if (!__is_valid_data_blkaddr(blkaddr))
3349 continue;
3350 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3351 DATA_GENERIC_ENHANCE)))
3352 return -EFSCORRUPTED;
3353 }
3354
3355 while (count) {
3356 int compr_blocks = 0;
3357
3358 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3359 blkaddr = f2fs_data_blkaddr(dn);
3360
3361 if (i == 0) {
3362 if (blkaddr == COMPRESS_ADDR)
3363 continue;
3364 dn->ofs_in_node += cluster_size;
3365 goto next;
3366 }
3367
3368 if (__is_valid_data_blkaddr(blkaddr))
3369 compr_blocks++;
3370
3371 if (blkaddr != NEW_ADDR)
3372 continue;
3373
3374 dn->data_blkaddr = NULL_ADDR;
3375 f2fs_set_data_blkaddr(dn);
3376 }
3377
3378 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3379 dec_valid_block_count(sbi, dn->inode,
3380 cluster_size - compr_blocks);
3381
3382 released_blocks += cluster_size - compr_blocks;
3383next:
3384 count -= cluster_size;
3385 }
3386
3387 return released_blocks;
3388}
3389
3390static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3391{
3392 struct inode *inode = file_inode(filp);
3393 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3394 pgoff_t page_idx = 0, last_idx;
3395 unsigned int released_blocks = 0;
3396 int ret;
3397 int writecount;
3398
3399 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3400 return -EOPNOTSUPP;
3401
3402 if (!f2fs_compressed_file(inode))
3403 return -EINVAL;
3404
3405 if (f2fs_readonly(sbi->sb))
3406 return -EROFS;
3407
3408 ret = mnt_want_write_file(filp);
3409 if (ret)
3410 return ret;
3411
3412 f2fs_balance_fs(F2FS_I_SB(inode), true);
3413
3414 inode_lock(inode);
3415
3416 writecount = atomic_read(&inode->i_writecount);
3417 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3418 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3419 ret = -EBUSY;
3420 goto out;
3421 }
3422
3423 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3424 ret = -EINVAL;
3425 goto out;
3426 }
3427
3428 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3429 if (ret)
3430 goto out;
3431
3432 set_inode_flag(inode, FI_COMPRESS_RELEASED);
3433 inode->i_ctime = current_time(inode);
3434 f2fs_mark_inode_dirty_sync(inode, true);
3435
3436 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3437 goto out;
3438
3439 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3440 down_write(&F2FS_I(inode)->i_mmap_sem);
3441
3442 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3443
3444 while (page_idx < last_idx) {
3445 struct dnode_of_data dn;
3446 pgoff_t end_offset, count;
3447
3448 set_new_dnode(&dn, inode, NULL, NULL, 0);
3449 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3450 if (ret) {
3451 if (ret == -ENOENT) {
3452 page_idx = f2fs_get_next_page_offset(&dn,
3453 page_idx);
3454 ret = 0;
3455 continue;
3456 }
3457 break;
3458 }
3459
3460 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3461 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3462 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3463
3464 ret = release_compress_blocks(&dn, count);
3465
3466 f2fs_put_dnode(&dn);
3467
3468 if (ret < 0)
3469 break;
3470
3471 page_idx += count;
3472 released_blocks += ret;
3473 }
3474
3475 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3476 up_write(&F2FS_I(inode)->i_mmap_sem);
3477out:
3478 inode_unlock(inode);
3479
3480 mnt_drop_write_file(filp);
3481
3482 if (ret >= 0) {
3483 ret = put_user(released_blocks, (u64 __user *)arg);
3484 } else if (released_blocks &&
3485 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3486 set_sbi_flag(sbi, SBI_NEED_FSCK);
3487 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3488 "iblocks=%llu, released=%u, compr_blocks=%u, "
3489 "run fsck to fix.",
3490 __func__, inode->i_ino, inode->i_blocks,
3491 released_blocks,
3492 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3493 }
3494
3495 return ret;
3496}
3497
3498static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3499{
3500 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3501 unsigned int reserved_blocks = 0;
3502 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3503 block_t blkaddr;
3504 int i;
3505
3506 for (i = 0; i < count; i++) {
3507 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3508 dn->ofs_in_node + i);
3509
3510 if (!__is_valid_data_blkaddr(blkaddr))
3511 continue;
3512 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3513 DATA_GENERIC_ENHANCE)))
3514 return -EFSCORRUPTED;
3515 }
3516
3517 while (count) {
3518 int compr_blocks = 0;
3519 blkcnt_t reserved;
3520 int ret;
3521
3522 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3523 blkaddr = f2fs_data_blkaddr(dn);
3524
3525 if (i == 0) {
3526 if (blkaddr == COMPRESS_ADDR)
3527 continue;
3528 dn->ofs_in_node += cluster_size;
3529 goto next;
3530 }
3531
3532 if (__is_valid_data_blkaddr(blkaddr)) {
3533 compr_blocks++;
3534 continue;
3535 }
3536
3537 dn->data_blkaddr = NEW_ADDR;
3538 f2fs_set_data_blkaddr(dn);
3539 }
3540
3541 reserved = cluster_size - compr_blocks;
3542 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3543 if (ret)
3544 return ret;
3545
3546 if (reserved != cluster_size - compr_blocks)
3547 return -ENOSPC;
3548
3549 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3550
3551 reserved_blocks += reserved;
3552next:
3553 count -= cluster_size;
3554 }
3555
3556 return reserved_blocks;
3557}
3558
3559static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3560{
3561 struct inode *inode = file_inode(filp);
3562 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3563 pgoff_t page_idx = 0, last_idx;
3564 unsigned int reserved_blocks = 0;
3565 int ret;
3566
3567 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3568 return -EOPNOTSUPP;
3569
3570 if (!f2fs_compressed_file(inode))
3571 return -EINVAL;
3572
3573 if (f2fs_readonly(sbi->sb))
3574 return -EROFS;
3575
3576 ret = mnt_want_write_file(filp);
3577 if (ret)
3578 return ret;
3579
3580 if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3581 goto out;
3582
3583 f2fs_balance_fs(F2FS_I_SB(inode), true);
3584
3585 inode_lock(inode);
3586
3587 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3588 ret = -EINVAL;
3589 goto unlock_inode;
3590 }
3591
3592 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3593 down_write(&F2FS_I(inode)->i_mmap_sem);
3594
3595 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3596
3597 while (page_idx < last_idx) {
3598 struct dnode_of_data dn;
3599 pgoff_t end_offset, count;
3600
3601 set_new_dnode(&dn, inode, NULL, NULL, 0);
3602 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3603 if (ret) {
3604 if (ret == -ENOENT) {
3605 page_idx = f2fs_get_next_page_offset(&dn,
3606 page_idx);
3607 ret = 0;
3608 continue;
3609 }
3610 break;
3611 }
3612
3613 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3614 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3615 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3616
3617 ret = reserve_compress_blocks(&dn, count);
3618
3619 f2fs_put_dnode(&dn);
3620
3621 if (ret < 0)
3622 break;
3623
3624 page_idx += count;
3625 reserved_blocks += ret;
3626 }
3627
3628 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3629 up_write(&F2FS_I(inode)->i_mmap_sem);
3630
3631 if (ret >= 0) {
3632 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3633 inode->i_ctime = current_time(inode);
3634 f2fs_mark_inode_dirty_sync(inode, true);
3635 }
3636unlock_inode:
3637 inode_unlock(inode);
3638out:
3639 mnt_drop_write_file(filp);
3640
3641 if (ret >= 0) {
3642 ret = put_user(reserved_blocks, (u64 __user *)arg);
3643 } else if (reserved_blocks &&
3644 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3645 set_sbi_flag(sbi, SBI_NEED_FSCK);
3646 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3647 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3648 "run fsck to fix.",
3649 __func__, inode->i_ino, inode->i_blocks,
3650 reserved_blocks,
3651 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3652 }
3653
3654 return ret;
3655}
3656
3657static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3658 pgoff_t off, block_t block, block_t len, u32 flags)
3659{
3660 struct request_queue *q = bdev_get_queue(bdev);
3661 sector_t sector = SECTOR_FROM_BLOCK(block);
3662 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3663 int ret = 0;
3664
3665 if (!q)
3666 return -ENXIO;
3667
3668 if (flags & F2FS_TRIM_FILE_DISCARD)
3669 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3670 blk_queue_secure_erase(q) ?
3671 BLKDEV_DISCARD_SECURE : 0);
3672
3673 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3674 if (IS_ENCRYPTED(inode))
3675 ret = fscrypt_zeroout_range(inode, off, block, len);
3676 else
3677 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3678 GFP_NOFS, 0);
3679 }
3680
3681 return ret;
3682}
3683
3684static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3685{
3686 struct inode *inode = file_inode(filp);
3687 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3688 struct address_space *mapping = inode->i_mapping;
3689 struct block_device *prev_bdev = NULL;
3690 struct f2fs_sectrim_range range;
3691 pgoff_t index, pg_end, prev_index = 0;
3692 block_t prev_block = 0, len = 0;
3693 loff_t end_addr;
3694 bool to_end = false;
3695 int ret = 0;
3696
3697 if (!(filp->f_mode & FMODE_WRITE))
3698 return -EBADF;
3699
3700 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3701 sizeof(range)))
3702 return -EFAULT;
3703
3704 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3705 !S_ISREG(inode->i_mode))
3706 return -EINVAL;
3707
3708 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3709 !f2fs_hw_support_discard(sbi)) ||
3710 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3711 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3712 return -EOPNOTSUPP;
3713
3714 file_start_write(filp);
3715 inode_lock(inode);
3716
3717 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3718 range.start >= inode->i_size) {
3719 ret = -EINVAL;
3720 goto err;
3721 }
3722
3723 if (range.len == 0)
3724 goto err;
3725
3726 if (inode->i_size - range.start > range.len) {
3727 end_addr = range.start + range.len;
3728 } else {
3729 end_addr = range.len == (u64)-1 ?
3730 sbi->sb->s_maxbytes : inode->i_size;
3731 to_end = true;
3732 }
3733
3734 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3735 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3736 ret = -EINVAL;
3737 goto err;
3738 }
3739
3740 index = F2FS_BYTES_TO_BLK(range.start);
3741 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3742
3743 ret = f2fs_convert_inline_inode(inode);
3744 if (ret)
3745 goto err;
3746
3747 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3748 down_write(&F2FS_I(inode)->i_mmap_sem);
3749
3750 ret = filemap_write_and_wait_range(mapping, range.start,
3751 to_end ? LLONG_MAX : end_addr - 1);
3752 if (ret)
3753 goto out;
3754
3755 truncate_inode_pages_range(mapping, range.start,
3756 to_end ? -1 : end_addr - 1);
3757
3758 while (index < pg_end) {
3759 struct dnode_of_data dn;
3760 pgoff_t end_offset, count;
3761 int i;
3762
3763 set_new_dnode(&dn, inode, NULL, NULL, 0);
3764 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3765 if (ret) {
3766 if (ret == -ENOENT) {
3767 index = f2fs_get_next_page_offset(&dn, index);
3768 continue;
3769 }
3770 goto out;
3771 }
3772
3773 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3774 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3775 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3776 struct block_device *cur_bdev;
3777 block_t blkaddr = f2fs_data_blkaddr(&dn);
3778
3779 if (!__is_valid_data_blkaddr(blkaddr))
3780 continue;
3781
3782 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3783 DATA_GENERIC_ENHANCE)) {
3784 ret = -EFSCORRUPTED;
3785 f2fs_put_dnode(&dn);
3786 goto out;
3787 }
3788
3789 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3790 if (f2fs_is_multi_device(sbi)) {
3791 int di = f2fs_target_device_index(sbi, blkaddr);
3792
3793 blkaddr -= FDEV(di).start_blk;
3794 }
3795
3796 if (len) {
3797 if (prev_bdev == cur_bdev &&
3798 index == prev_index + len &&
3799 blkaddr == prev_block + len) {
3800 len++;
3801 } else {
3802 ret = f2fs_secure_erase(prev_bdev,
3803 inode, prev_index, prev_block,
3804 len, range.flags);
3805 if (ret) {
3806 f2fs_put_dnode(&dn);
3807 goto out;
3808 }
3809
3810 len = 0;
3811 }
3812 }
3813
3814 if (!len) {
3815 prev_bdev = cur_bdev;
3816 prev_index = index;
3817 prev_block = blkaddr;
3818 len = 1;
3819 }
3820 }
3821
3822 f2fs_put_dnode(&dn);
3823
3824 if (fatal_signal_pending(current)) {
3825 ret = -EINTR;
3826 goto out;
3827 }
3828 cond_resched();
3829 }
3830
3831 if (len)
3832 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3833 prev_block, len, range.flags);
3834out:
3835 up_write(&F2FS_I(inode)->i_mmap_sem);
3836 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3837err:
3838 inode_unlock(inode);
3839 file_end_write(filp);
3840
3841 return ret;
3842}
3843
3844static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3845{
3846 struct inode *inode = file_inode(filp);
3847 struct f2fs_comp_option option;
3848
3849 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3850 return -EOPNOTSUPP;
3851
3852 inode_lock_shared(inode);
3853
3854 if (!f2fs_compressed_file(inode)) {
3855 inode_unlock_shared(inode);
3856 return -ENODATA;
3857 }
3858
3859 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3860 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3861
3862 inode_unlock_shared(inode);
3863
3864 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3865 sizeof(option)))
3866 return -EFAULT;
3867
3868 return 0;
3869}
3870
3871static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3872{
3873 struct inode *inode = file_inode(filp);
3874 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3875 struct f2fs_comp_option option;
3876 int ret = 0;
3877
3878 if (!f2fs_sb_has_compression(sbi))
3879 return -EOPNOTSUPP;
3880
3881 if (!(filp->f_mode & FMODE_WRITE))
3882 return -EBADF;
3883
3884 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3885 sizeof(option)))
3886 return -EFAULT;
3887
3888 if (!f2fs_compressed_file(inode) ||
3889 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3890 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3891 option.algorithm >= COMPRESS_MAX)
3892 return -EINVAL;
3893
3894 file_start_write(filp);
3895 inode_lock(inode);
3896
3897 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3898 ret = -EBUSY;
3899 goto out;
3900 }
3901
3902 if (inode->i_size != 0) {
3903 ret = -EFBIG;
3904 goto out;
3905 }
3906
3907 F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3908 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3909 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3910 f2fs_mark_inode_dirty_sync(inode, true);
3911
3912 if (!f2fs_is_compress_backend_ready(inode))
3913 f2fs_warn(sbi, "compression algorithm is successfully set, "
3914 "but current kernel doesn't support this algorithm.");
3915out:
3916 inode_unlock(inode);
3917 file_end_write(filp);
3918
3919 return ret;
3920}
3921
3922static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3923{
3924 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3925 struct address_space *mapping = inode->i_mapping;
3926 struct page *page;
3927 pgoff_t redirty_idx = page_idx;
3928 int i, page_len = 0, ret = 0;
3929
3930 page_cache_ra_unbounded(&ractl, len, 0);
3931
3932 for (i = 0; i < len; i++, page_idx++) {
3933 page = read_cache_page(mapping, page_idx, NULL, NULL);
3934 if (IS_ERR(page)) {
3935 ret = PTR_ERR(page);
3936 break;
3937 }
3938 page_len++;
3939 }
3940
3941 for (i = 0; i < page_len; i++, redirty_idx++) {
3942 page = find_lock_page(mapping, redirty_idx);
3943 if (!page) {
3944 ret = -ENOMEM;
3945 break;
3946 }
3947 set_page_dirty(page);
3948 f2fs_put_page(page, 1);
3949 f2fs_put_page(page, 0);
3950 }
3951
3952 return ret;
3953}
3954
3955static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3956{
3957 struct inode *inode = file_inode(filp);
3958 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3959 struct f2fs_inode_info *fi = F2FS_I(inode);
3960 pgoff_t page_idx = 0, last_idx;
3961 unsigned int blk_per_seg = sbi->blocks_per_seg;
3962 int cluster_size = F2FS_I(inode)->i_cluster_size;
3963 int count, ret;
3964
3965 if (!f2fs_sb_has_compression(sbi) ||
3966 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3967 return -EOPNOTSUPP;
3968
3969 if (!(filp->f_mode & FMODE_WRITE))
3970 return -EBADF;
3971
3972 if (!f2fs_compressed_file(inode))
3973 return -EINVAL;
3974
3975 f2fs_balance_fs(F2FS_I_SB(inode), true);
3976
3977 file_start_write(filp);
3978 inode_lock(inode);
3979
3980 if (!f2fs_is_compress_backend_ready(inode)) {
3981 ret = -EOPNOTSUPP;
3982 goto out;
3983 }
3984
3985 if (f2fs_is_mmap_file(inode)) {
3986 ret = -EBUSY;
3987 goto out;
3988 }
3989
3990 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3991 if (ret)
3992 goto out;
3993
3994 if (!atomic_read(&fi->i_compr_blocks))
3995 goto out;
3996
3997 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3998
3999 count = last_idx - page_idx;
4000 while (count) {
4001 int len = min(cluster_size, count);
4002
4003 ret = redirty_blocks(inode, page_idx, len);
4004 if (ret < 0)
4005 break;
4006
4007 if (get_dirty_pages(inode) >= blk_per_seg)
4008 filemap_fdatawrite(inode->i_mapping);
4009
4010 count -= len;
4011 page_idx += len;
4012 }
4013
4014 if (!ret)
4015 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4016 LLONG_MAX);
4017
4018 if (ret)
4019 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4020 __func__, ret);
4021out:
4022 inode_unlock(inode);
4023 file_end_write(filp);
4024
4025 return ret;
4026}
4027
4028static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4029{
4030 struct inode *inode = file_inode(filp);
4031 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4032 pgoff_t page_idx = 0, last_idx;
4033 unsigned int blk_per_seg = sbi->blocks_per_seg;
4034 int cluster_size = F2FS_I(inode)->i_cluster_size;
4035 int count, ret;
4036
4037 if (!f2fs_sb_has_compression(sbi) ||
4038 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4039 return -EOPNOTSUPP;
4040
4041 if (!(filp->f_mode & FMODE_WRITE))
4042 return -EBADF;
4043
4044 if (!f2fs_compressed_file(inode))
4045 return -EINVAL;
4046
4047 f2fs_balance_fs(F2FS_I_SB(inode), true);
4048
4049 file_start_write(filp);
4050 inode_lock(inode);
4051
4052 if (!f2fs_is_compress_backend_ready(inode)) {
4053 ret = -EOPNOTSUPP;
4054 goto out;
4055 }
4056
4057 if (f2fs_is_mmap_file(inode)) {
4058 ret = -EBUSY;
4059 goto out;
4060 }
4061
4062 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4063 if (ret)
4064 goto out;
4065
4066 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4067
4068 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4069
4070 count = last_idx - page_idx;
4071 while (count) {
4072 int len = min(cluster_size, count);
4073
4074 ret = redirty_blocks(inode, page_idx, len);
4075 if (ret < 0)
4076 break;
4077
4078 if (get_dirty_pages(inode) >= blk_per_seg)
4079 filemap_fdatawrite(inode->i_mapping);
4080
4081 count -= len;
4082 page_idx += len;
4083 }
4084
4085 if (!ret)
4086 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4087 LLONG_MAX);
4088
4089 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4090
4091 if (ret)
4092 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4093 __func__, ret);
4094out:
4095 inode_unlock(inode);
4096 file_end_write(filp);
4097
4098 return ret;
4099}
4100
4101static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4102{
4103 switch (cmd) {
4104 case FS_IOC_GETVERSION:
4105 return f2fs_ioc_getversion(filp, arg);
4106 case F2FS_IOC_START_ATOMIC_WRITE:
4107 return f2fs_ioc_start_atomic_write(filp);
4108 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4109 return f2fs_ioc_commit_atomic_write(filp);
4110 case F2FS_IOC_START_VOLATILE_WRITE:
4111 return f2fs_ioc_start_volatile_write(filp);
4112 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4113 return f2fs_ioc_release_volatile_write(filp);
4114 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4115 return f2fs_ioc_abort_volatile_write(filp);
4116 case F2FS_IOC_SHUTDOWN:
4117 return f2fs_ioc_shutdown(filp, arg);
4118 case FITRIM:
4119 return f2fs_ioc_fitrim(filp, arg);
4120 case FS_IOC_SET_ENCRYPTION_POLICY:
4121 return f2fs_ioc_set_encryption_policy(filp, arg);
4122 case FS_IOC_GET_ENCRYPTION_POLICY:
4123 return f2fs_ioc_get_encryption_policy(filp, arg);
4124 case FS_IOC_GET_ENCRYPTION_PWSALT:
4125 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4126 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4127 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4128 case FS_IOC_ADD_ENCRYPTION_KEY:
4129 return f2fs_ioc_add_encryption_key(filp, arg);
4130 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4131 return f2fs_ioc_remove_encryption_key(filp, arg);
4132 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4133 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4134 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4135 return f2fs_ioc_get_encryption_key_status(filp, arg);
4136 case FS_IOC_GET_ENCRYPTION_NONCE:
4137 return f2fs_ioc_get_encryption_nonce(filp, arg);
4138 case F2FS_IOC_GARBAGE_COLLECT:
4139 return f2fs_ioc_gc(filp, arg);
4140 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4141 return f2fs_ioc_gc_range(filp, arg);
4142 case F2FS_IOC_WRITE_CHECKPOINT:
4143 return f2fs_ioc_write_checkpoint(filp, arg);
4144 case F2FS_IOC_DEFRAGMENT:
4145 return f2fs_ioc_defragment(filp, arg);
4146 case F2FS_IOC_MOVE_RANGE:
4147 return f2fs_ioc_move_range(filp, arg);
4148 case F2FS_IOC_FLUSH_DEVICE:
4149 return f2fs_ioc_flush_device(filp, arg);
4150 case F2FS_IOC_GET_FEATURES:
4151 return f2fs_ioc_get_features(filp, arg);
4152 case F2FS_IOC_GET_PIN_FILE:
4153 return f2fs_ioc_get_pin_file(filp, arg);
4154 case F2FS_IOC_SET_PIN_FILE:
4155 return f2fs_ioc_set_pin_file(filp, arg);
4156 case F2FS_IOC_PRECACHE_EXTENTS:
4157 return f2fs_ioc_precache_extents(filp, arg);
4158 case F2FS_IOC_RESIZE_FS:
4159 return f2fs_ioc_resize_fs(filp, arg);
4160 case FS_IOC_ENABLE_VERITY:
4161 return f2fs_ioc_enable_verity(filp, arg);
4162 case FS_IOC_MEASURE_VERITY:
4163 return f2fs_ioc_measure_verity(filp, arg);
4164 case FS_IOC_READ_VERITY_METADATA:
4165 return f2fs_ioc_read_verity_metadata(filp, arg);
4166 case FS_IOC_GETFSLABEL:
4167 return f2fs_ioc_getfslabel(filp, arg);
4168 case FS_IOC_SETFSLABEL:
4169 return f2fs_ioc_setfslabel(filp, arg);
4170 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4171 return f2fs_get_compress_blocks(filp, arg);
4172 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4173 return f2fs_release_compress_blocks(filp, arg);
4174 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4175 return f2fs_reserve_compress_blocks(filp, arg);
4176 case F2FS_IOC_SEC_TRIM_FILE:
4177 return f2fs_sec_trim_file(filp, arg);
4178 case F2FS_IOC_GET_COMPRESS_OPTION:
4179 return f2fs_ioc_get_compress_option(filp, arg);
4180 case F2FS_IOC_SET_COMPRESS_OPTION:
4181 return f2fs_ioc_set_compress_option(filp, arg);
4182 case F2FS_IOC_DECOMPRESS_FILE:
4183 return f2fs_ioc_decompress_file(filp, arg);
4184 case F2FS_IOC_COMPRESS_FILE:
4185 return f2fs_ioc_compress_file(filp, arg);
4186 default:
4187 return -ENOTTY;
4188 }
4189}
4190
4191long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4192{
4193 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4194 return -EIO;
4195 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4196 return -ENOSPC;
4197
4198 return __f2fs_ioctl(filp, cmd, arg);
4199}
4200
4201static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
4202{
4203 struct file *file = iocb->ki_filp;
4204 struct inode *inode = file_inode(file);
4205 int ret;
4206
4207 if (!f2fs_is_compress_backend_ready(inode))
4208 return -EOPNOTSUPP;
4209
4210 ret = generic_file_read_iter(iocb, iter);
4211
4212 if (ret > 0)
4213 f2fs_update_iostat(F2FS_I_SB(inode), APP_READ_IO, ret);
4214
4215 return ret;
4216}
4217
4218static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4219{
4220 struct file *file = iocb->ki_filp;
4221 struct inode *inode = file_inode(file);
4222 ssize_t ret;
4223
4224 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4225 ret = -EIO;
4226 goto out;
4227 }
4228
4229 if (!f2fs_is_compress_backend_ready(inode)) {
4230 ret = -EOPNOTSUPP;
4231 goto out;
4232 }
4233
4234 if (iocb->ki_flags & IOCB_NOWAIT) {
4235 if (!inode_trylock(inode)) {
4236 ret = -EAGAIN;
4237 goto out;
4238 }
4239 } else {
4240 inode_lock(inode);
4241 }
4242
4243 if (unlikely(IS_IMMUTABLE(inode))) {
4244 ret = -EPERM;
4245 goto unlock;
4246 }
4247
4248 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4249 ret = -EPERM;
4250 goto unlock;
4251 }
4252
4253 ret = generic_write_checks(iocb, from);
4254 if (ret > 0) {
4255 bool preallocated = false;
4256 size_t target_size = 0;
4257 int err;
4258
4259 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
4260 set_inode_flag(inode, FI_NO_PREALLOC);
4261
4262 if ((iocb->ki_flags & IOCB_NOWAIT)) {
4263 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
4264 iov_iter_count(from)) ||
4265 f2fs_has_inline_data(inode) ||
4266 f2fs_force_buffered_io(inode, iocb, from)) {
4267 clear_inode_flag(inode, FI_NO_PREALLOC);
4268 inode_unlock(inode);
4269 ret = -EAGAIN;
4270 goto out;
4271 }
4272 goto write;
4273 }
4274
4275 if (is_inode_flag_set(inode, FI_NO_PREALLOC))
4276 goto write;
4277
4278 if (iocb->ki_flags & IOCB_DIRECT) {
4279 /*
4280 * Convert inline data for Direct I/O before entering
4281 * f2fs_direct_IO().
4282 */
4283 err = f2fs_convert_inline_inode(inode);
4284 if (err)
4285 goto out_err;
4286 /*
4287 * If force_buffere_io() is true, we have to allocate
4288 * blocks all the time, since f2fs_direct_IO will fall
4289 * back to buffered IO.
4290 */
4291 if (!f2fs_force_buffered_io(inode, iocb, from) &&
4292 allow_outplace_dio(inode, iocb, from))
4293 goto write;
4294 }
4295 preallocated = true;
4296 target_size = iocb->ki_pos + iov_iter_count(from);
4297
4298 err = f2fs_preallocate_blocks(iocb, from);
4299 if (err) {
4300out_err:
4301 clear_inode_flag(inode, FI_NO_PREALLOC);
4302 inode_unlock(inode);
4303 ret = err;
4304 goto out;
4305 }
4306write:
4307 ret = __generic_file_write_iter(iocb, from);
4308 clear_inode_flag(inode, FI_NO_PREALLOC);
4309
4310 /* if we couldn't write data, we should deallocate blocks. */
4311 if (preallocated && i_size_read(inode) < target_size) {
4312 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4313 down_write(&F2FS_I(inode)->i_mmap_sem);
4314 f2fs_truncate(inode);
4315 up_write(&F2FS_I(inode)->i_mmap_sem);
4316 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4317 }
4318
4319 if (ret > 0)
4320 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
4321 }
4322unlock:
4323 inode_unlock(inode);
4324out:
4325 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
4326 iov_iter_count(from), ret);
4327 if (ret > 0)
4328 ret = generic_write_sync(iocb, ret);
4329 return ret;
4330}
4331
4332#ifdef CONFIG_COMPAT
4333struct compat_f2fs_gc_range {
4334 u32 sync;
4335 compat_u64 start;
4336 compat_u64 len;
4337};
4338#define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
4339 struct compat_f2fs_gc_range)
4340
4341static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4342{
4343 struct compat_f2fs_gc_range __user *urange;
4344 struct f2fs_gc_range range;
4345 int err;
4346
4347 urange = compat_ptr(arg);
4348 err = get_user(range.sync, &urange->sync);
4349 err |= get_user(range.start, &urange->start);
4350 err |= get_user(range.len, &urange->len);
4351 if (err)
4352 return -EFAULT;
4353
4354 return __f2fs_ioc_gc_range(file, &range);
4355}
4356
4357struct compat_f2fs_move_range {
4358 u32 dst_fd;
4359 compat_u64 pos_in;
4360 compat_u64 pos_out;
4361 compat_u64 len;
4362};
4363#define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
4364 struct compat_f2fs_move_range)
4365
4366static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4367{
4368 struct compat_f2fs_move_range __user *urange;
4369 struct f2fs_move_range range;
4370 int err;
4371
4372 urange = compat_ptr(arg);
4373 err = get_user(range.dst_fd, &urange->dst_fd);
4374 err |= get_user(range.pos_in, &urange->pos_in);
4375 err |= get_user(range.pos_out, &urange->pos_out);
4376 err |= get_user(range.len, &urange->len);
4377 if (err)
4378 return -EFAULT;
4379
4380 return __f2fs_ioc_move_range(file, &range);
4381}
4382
4383long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4384{
4385 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4386 return -EIO;
4387 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4388 return -ENOSPC;
4389
4390 switch (cmd) {
4391 case FS_IOC32_GETVERSION:
4392 cmd = FS_IOC_GETVERSION;
4393 break;
4394 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4395 return f2fs_compat_ioc_gc_range(file, arg);
4396 case F2FS_IOC32_MOVE_RANGE:
4397 return f2fs_compat_ioc_move_range(file, arg);
4398 case F2FS_IOC_START_ATOMIC_WRITE:
4399 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4400 case F2FS_IOC_START_VOLATILE_WRITE:
4401 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4402 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4403 case F2FS_IOC_SHUTDOWN:
4404 case FITRIM:
4405 case FS_IOC_SET_ENCRYPTION_POLICY:
4406 case FS_IOC_GET_ENCRYPTION_PWSALT:
4407 case FS_IOC_GET_ENCRYPTION_POLICY:
4408 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4409 case FS_IOC_ADD_ENCRYPTION_KEY:
4410 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4411 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4412 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4413 case FS_IOC_GET_ENCRYPTION_NONCE:
4414 case F2FS_IOC_GARBAGE_COLLECT:
4415 case F2FS_IOC_WRITE_CHECKPOINT:
4416 case F2FS_IOC_DEFRAGMENT:
4417 case F2FS_IOC_FLUSH_DEVICE:
4418 case F2FS_IOC_GET_FEATURES:
4419 case F2FS_IOC_GET_PIN_FILE:
4420 case F2FS_IOC_SET_PIN_FILE:
4421 case F2FS_IOC_PRECACHE_EXTENTS:
4422 case F2FS_IOC_RESIZE_FS:
4423 case FS_IOC_ENABLE_VERITY:
4424 case FS_IOC_MEASURE_VERITY:
4425 case FS_IOC_READ_VERITY_METADATA:
4426 case FS_IOC_GETFSLABEL:
4427 case FS_IOC_SETFSLABEL:
4428 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4429 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4430 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4431 case F2FS_IOC_SEC_TRIM_FILE:
4432 case F2FS_IOC_GET_COMPRESS_OPTION:
4433 case F2FS_IOC_SET_COMPRESS_OPTION:
4434 case F2FS_IOC_DECOMPRESS_FILE:
4435 case F2FS_IOC_COMPRESS_FILE:
4436 break;
4437 default:
4438 return -ENOIOCTLCMD;
4439 }
4440 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4441}
4442#endif
4443
4444const struct file_operations f2fs_file_operations = {
4445 .llseek = f2fs_llseek,
4446 .read_iter = f2fs_file_read_iter,
4447 .write_iter = f2fs_file_write_iter,
4448 .open = f2fs_file_open,
4449 .release = f2fs_release_file,
4450 .mmap = f2fs_file_mmap,
4451 .flush = f2fs_file_flush,
4452 .fsync = f2fs_sync_file,
4453 .fallocate = f2fs_fallocate,
4454 .unlocked_ioctl = f2fs_ioctl,
4455#ifdef CONFIG_COMPAT
4456 .compat_ioctl = f2fs_compat_ioctl,
4457#endif
4458 .splice_read = generic_file_splice_read,
4459 .splice_write = iter_file_splice_write,
4460};
1/*
2 * fs/f2fs/file.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11#include <linux/fs.h>
12#include <linux/f2fs_fs.h>
13#include <linux/stat.h>
14#include <linux/buffer_head.h>
15#include <linux/writeback.h>
16#include <linux/blkdev.h>
17#include <linux/falloc.h>
18#include <linux/types.h>
19#include <linux/compat.h>
20#include <linux/uaccess.h>
21#include <linux/mount.h>
22#include <linux/pagevec.h>
23#include <linux/random.h>
24
25#include "f2fs.h"
26#include "node.h"
27#include "segment.h"
28#include "xattr.h"
29#include "acl.h"
30#include "gc.h"
31#include "trace.h"
32#include <trace/events/f2fs.h>
33
34static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
35 struct vm_fault *vmf)
36{
37 struct page *page = vmf->page;
38 struct inode *inode = file_inode(vma->vm_file);
39 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
40 struct dnode_of_data dn;
41 int err;
42
43 sb_start_pagefault(inode->i_sb);
44
45 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
46
47 /* block allocation */
48 f2fs_lock_op(sbi);
49 set_new_dnode(&dn, inode, NULL, NULL, 0);
50 err = f2fs_reserve_block(&dn, page->index);
51 if (err) {
52 f2fs_unlock_op(sbi);
53 goto out;
54 }
55 f2fs_put_dnode(&dn);
56 f2fs_unlock_op(sbi);
57
58 f2fs_balance_fs(sbi, dn.node_changed);
59
60 file_update_time(vma->vm_file);
61 lock_page(page);
62 if (unlikely(page->mapping != inode->i_mapping ||
63 page_offset(page) > i_size_read(inode) ||
64 !PageUptodate(page))) {
65 unlock_page(page);
66 err = -EFAULT;
67 goto out;
68 }
69
70 /*
71 * check to see if the page is mapped already (no holes)
72 */
73 if (PageMappedToDisk(page))
74 goto mapped;
75
76 /* page is wholly or partially inside EOF */
77 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
78 i_size_read(inode)) {
79 unsigned offset;
80 offset = i_size_read(inode) & ~PAGE_MASK;
81 zero_user_segment(page, offset, PAGE_SIZE);
82 }
83 set_page_dirty(page);
84 SetPageUptodate(page);
85
86 trace_f2fs_vm_page_mkwrite(page, DATA);
87mapped:
88 /* fill the page */
89 f2fs_wait_on_page_writeback(page, DATA, false);
90
91 /* wait for GCed encrypted page writeback */
92 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
93 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
94
95 /* if gced page is attached, don't write to cold segment */
96 clear_cold_data(page);
97out:
98 sb_end_pagefault(inode->i_sb);
99 f2fs_update_time(sbi, REQ_TIME);
100 return block_page_mkwrite_return(err);
101}
102
103static const struct vm_operations_struct f2fs_file_vm_ops = {
104 .fault = filemap_fault,
105 .map_pages = filemap_map_pages,
106 .page_mkwrite = f2fs_vm_page_mkwrite,
107};
108
109static int get_parent_ino(struct inode *inode, nid_t *pino)
110{
111 struct dentry *dentry;
112
113 inode = igrab(inode);
114 dentry = d_find_any_alias(inode);
115 iput(inode);
116 if (!dentry)
117 return 0;
118
119 if (update_dent_inode(inode, inode, &dentry->d_name)) {
120 dput(dentry);
121 return 0;
122 }
123
124 *pino = parent_ino(dentry);
125 dput(dentry);
126 return 1;
127}
128
129static inline bool need_do_checkpoint(struct inode *inode)
130{
131 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
132 bool need_cp = false;
133
134 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
135 need_cp = true;
136 else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
137 need_cp = true;
138 else if (file_wrong_pino(inode))
139 need_cp = true;
140 else if (!space_for_roll_forward(sbi))
141 need_cp = true;
142 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
143 need_cp = true;
144 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
145 need_cp = true;
146 else if (test_opt(sbi, FASTBOOT))
147 need_cp = true;
148 else if (sbi->active_logs == 2)
149 need_cp = true;
150
151 return need_cp;
152}
153
154static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
155{
156 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
157 bool ret = false;
158 /* But we need to avoid that there are some inode updates */
159 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
160 ret = true;
161 f2fs_put_page(i, 0);
162 return ret;
163}
164
165static void try_to_fix_pino(struct inode *inode)
166{
167 struct f2fs_inode_info *fi = F2FS_I(inode);
168 nid_t pino;
169
170 down_write(&fi->i_sem);
171 fi->xattr_ver = 0;
172 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
173 get_parent_ino(inode, &pino)) {
174 fi->i_pino = pino;
175 file_got_pino(inode);
176 up_write(&fi->i_sem);
177
178 mark_inode_dirty_sync(inode);
179 f2fs_write_inode(inode, NULL);
180 } else {
181 up_write(&fi->i_sem);
182 }
183}
184
185int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
186{
187 struct inode *inode = file->f_mapping->host;
188 struct f2fs_inode_info *fi = F2FS_I(inode);
189 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
190 nid_t ino = inode->i_ino;
191 int ret = 0;
192 bool need_cp = false;
193 struct writeback_control wbc = {
194 .sync_mode = WB_SYNC_ALL,
195 .nr_to_write = LONG_MAX,
196 .for_reclaim = 0,
197 };
198
199 if (unlikely(f2fs_readonly(inode->i_sb)))
200 return 0;
201
202 trace_f2fs_sync_file_enter(inode);
203
204 /* if fdatasync is triggered, let's do in-place-update */
205 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
206 set_inode_flag(fi, FI_NEED_IPU);
207 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
208 clear_inode_flag(fi, FI_NEED_IPU);
209
210 if (ret) {
211 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
212 return ret;
213 }
214
215 /* if the inode is dirty, let's recover all the time */
216 if (!datasync) {
217 f2fs_write_inode(inode, NULL);
218 goto go_write;
219 }
220
221 /*
222 * if there is no written data, don't waste time to write recovery info.
223 */
224 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
225 !exist_written_data(sbi, ino, APPEND_INO)) {
226
227 /* it may call write_inode just prior to fsync */
228 if (need_inode_page_update(sbi, ino))
229 goto go_write;
230
231 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
232 exist_written_data(sbi, ino, UPDATE_INO))
233 goto flush_out;
234 goto out;
235 }
236go_write:
237 /*
238 * Both of fdatasync() and fsync() are able to be recovered from
239 * sudden-power-off.
240 */
241 down_read(&fi->i_sem);
242 need_cp = need_do_checkpoint(inode);
243 up_read(&fi->i_sem);
244
245 if (need_cp) {
246 /* all the dirty node pages should be flushed for POR */
247 ret = f2fs_sync_fs(inode->i_sb, 1);
248
249 /*
250 * We've secured consistency through sync_fs. Following pino
251 * will be used only for fsynced inodes after checkpoint.
252 */
253 try_to_fix_pino(inode);
254 clear_inode_flag(fi, FI_APPEND_WRITE);
255 clear_inode_flag(fi, FI_UPDATE_WRITE);
256 goto out;
257 }
258sync_nodes:
259 sync_node_pages(sbi, ino, &wbc);
260
261 /* if cp_error was enabled, we should avoid infinite loop */
262 if (unlikely(f2fs_cp_error(sbi))) {
263 ret = -EIO;
264 goto out;
265 }
266
267 if (need_inode_block_update(sbi, ino)) {
268 mark_inode_dirty_sync(inode);
269 f2fs_write_inode(inode, NULL);
270 goto sync_nodes;
271 }
272
273 ret = wait_on_node_pages_writeback(sbi, ino);
274 if (ret)
275 goto out;
276
277 /* once recovery info is written, don't need to tack this */
278 remove_ino_entry(sbi, ino, APPEND_INO);
279 clear_inode_flag(fi, FI_APPEND_WRITE);
280flush_out:
281 remove_ino_entry(sbi, ino, UPDATE_INO);
282 clear_inode_flag(fi, FI_UPDATE_WRITE);
283 ret = f2fs_issue_flush(sbi);
284 f2fs_update_time(sbi, REQ_TIME);
285out:
286 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
287 f2fs_trace_ios(NULL, 1);
288 return ret;
289}
290
291static pgoff_t __get_first_dirty_index(struct address_space *mapping,
292 pgoff_t pgofs, int whence)
293{
294 struct pagevec pvec;
295 int nr_pages;
296
297 if (whence != SEEK_DATA)
298 return 0;
299
300 /* find first dirty page index */
301 pagevec_init(&pvec, 0);
302 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
303 PAGECACHE_TAG_DIRTY, 1);
304 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
305 pagevec_release(&pvec);
306 return pgofs;
307}
308
309static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
310 int whence)
311{
312 switch (whence) {
313 case SEEK_DATA:
314 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
315 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
316 return true;
317 break;
318 case SEEK_HOLE:
319 if (blkaddr == NULL_ADDR)
320 return true;
321 break;
322 }
323 return false;
324}
325
326static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
327{
328 struct inode *inode = file->f_mapping->host;
329 loff_t maxbytes = inode->i_sb->s_maxbytes;
330 struct dnode_of_data dn;
331 pgoff_t pgofs, end_offset, dirty;
332 loff_t data_ofs = offset;
333 loff_t isize;
334 int err = 0;
335
336 inode_lock(inode);
337
338 isize = i_size_read(inode);
339 if (offset >= isize)
340 goto fail;
341
342 /* handle inline data case */
343 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
344 if (whence == SEEK_HOLE)
345 data_ofs = isize;
346 goto found;
347 }
348
349 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
350
351 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
352
353 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
354 set_new_dnode(&dn, inode, NULL, NULL, 0);
355 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
356 if (err && err != -ENOENT) {
357 goto fail;
358 } else if (err == -ENOENT) {
359 /* direct node does not exists */
360 if (whence == SEEK_DATA) {
361 pgofs = get_next_page_offset(&dn, pgofs);
362 continue;
363 } else {
364 goto found;
365 }
366 }
367
368 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
369
370 /* find data/hole in dnode block */
371 for (; dn.ofs_in_node < end_offset;
372 dn.ofs_in_node++, pgofs++,
373 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
374 block_t blkaddr;
375 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
376
377 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
378 f2fs_put_dnode(&dn);
379 goto found;
380 }
381 }
382 f2fs_put_dnode(&dn);
383 }
384
385 if (whence == SEEK_DATA)
386 goto fail;
387found:
388 if (whence == SEEK_HOLE && data_ofs > isize)
389 data_ofs = isize;
390 inode_unlock(inode);
391 return vfs_setpos(file, data_ofs, maxbytes);
392fail:
393 inode_unlock(inode);
394 return -ENXIO;
395}
396
397static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
398{
399 struct inode *inode = file->f_mapping->host;
400 loff_t maxbytes = inode->i_sb->s_maxbytes;
401
402 switch (whence) {
403 case SEEK_SET:
404 case SEEK_CUR:
405 case SEEK_END:
406 return generic_file_llseek_size(file, offset, whence,
407 maxbytes, i_size_read(inode));
408 case SEEK_DATA:
409 case SEEK_HOLE:
410 if (offset < 0)
411 return -ENXIO;
412 return f2fs_seek_block(file, offset, whence);
413 }
414
415 return -EINVAL;
416}
417
418static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
419{
420 struct inode *inode = file_inode(file);
421 int err;
422
423 if (f2fs_encrypted_inode(inode)) {
424 err = fscrypt_get_encryption_info(inode);
425 if (err)
426 return 0;
427 if (!f2fs_encrypted_inode(inode))
428 return -ENOKEY;
429 }
430
431 /* we don't need to use inline_data strictly */
432 err = f2fs_convert_inline_inode(inode);
433 if (err)
434 return err;
435
436 file_accessed(file);
437 vma->vm_ops = &f2fs_file_vm_ops;
438 return 0;
439}
440
441static int f2fs_file_open(struct inode *inode, struct file *filp)
442{
443 int ret = generic_file_open(inode, filp);
444 struct dentry *dir;
445
446 if (!ret && f2fs_encrypted_inode(inode)) {
447 ret = fscrypt_get_encryption_info(inode);
448 if (ret)
449 return -EACCES;
450 if (!fscrypt_has_encryption_key(inode))
451 return -ENOKEY;
452 }
453 dir = dget_parent(file_dentry(filp));
454 if (f2fs_encrypted_inode(d_inode(dir)) &&
455 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
456 dput(dir);
457 return -EPERM;
458 }
459 dput(dir);
460 return ret;
461}
462
463int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
464{
465 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
466 struct f2fs_node *raw_node;
467 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
468 __le32 *addr;
469
470 raw_node = F2FS_NODE(dn->node_page);
471 addr = blkaddr_in_node(raw_node) + ofs;
472
473 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
474 block_t blkaddr = le32_to_cpu(*addr);
475 if (blkaddr == NULL_ADDR)
476 continue;
477
478 dn->data_blkaddr = NULL_ADDR;
479 set_data_blkaddr(dn);
480 invalidate_blocks(sbi, blkaddr);
481 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
482 clear_inode_flag(F2FS_I(dn->inode),
483 FI_FIRST_BLOCK_WRITTEN);
484 nr_free++;
485 }
486
487 if (nr_free) {
488 pgoff_t fofs;
489 /*
490 * once we invalidate valid blkaddr in range [ofs, ofs + count],
491 * we will invalidate all blkaddr in the whole range.
492 */
493 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
494 dn->inode) + ofs;
495 f2fs_update_extent_cache_range(dn, fofs, 0, len);
496 dec_valid_block_count(sbi, dn->inode, nr_free);
497 sync_inode_page(dn);
498 }
499 dn->ofs_in_node = ofs;
500
501 f2fs_update_time(sbi, REQ_TIME);
502 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
503 dn->ofs_in_node, nr_free);
504 return nr_free;
505}
506
507void truncate_data_blocks(struct dnode_of_data *dn)
508{
509 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
510}
511
512static int truncate_partial_data_page(struct inode *inode, u64 from,
513 bool cache_only)
514{
515 unsigned offset = from & (PAGE_SIZE - 1);
516 pgoff_t index = from >> PAGE_SHIFT;
517 struct address_space *mapping = inode->i_mapping;
518 struct page *page;
519
520 if (!offset && !cache_only)
521 return 0;
522
523 if (cache_only) {
524 page = f2fs_grab_cache_page(mapping, index, false);
525 if (page && PageUptodate(page))
526 goto truncate_out;
527 f2fs_put_page(page, 1);
528 return 0;
529 }
530
531 page = get_lock_data_page(inode, index, true);
532 if (IS_ERR(page))
533 return 0;
534truncate_out:
535 f2fs_wait_on_page_writeback(page, DATA, true);
536 zero_user(page, offset, PAGE_SIZE - offset);
537 if (!cache_only || !f2fs_encrypted_inode(inode) ||
538 !S_ISREG(inode->i_mode))
539 set_page_dirty(page);
540 f2fs_put_page(page, 1);
541 return 0;
542}
543
544int truncate_blocks(struct inode *inode, u64 from, bool lock)
545{
546 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
547 unsigned int blocksize = inode->i_sb->s_blocksize;
548 struct dnode_of_data dn;
549 pgoff_t free_from;
550 int count = 0, err = 0;
551 struct page *ipage;
552 bool truncate_page = false;
553
554 trace_f2fs_truncate_blocks_enter(inode, from);
555
556 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
557
558 if (lock)
559 f2fs_lock_op(sbi);
560
561 ipage = get_node_page(sbi, inode->i_ino);
562 if (IS_ERR(ipage)) {
563 err = PTR_ERR(ipage);
564 goto out;
565 }
566
567 if (f2fs_has_inline_data(inode)) {
568 if (truncate_inline_inode(ipage, from))
569 set_page_dirty(ipage);
570 f2fs_put_page(ipage, 1);
571 truncate_page = true;
572 goto out;
573 }
574
575 set_new_dnode(&dn, inode, ipage, NULL, 0);
576 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
577 if (err) {
578 if (err == -ENOENT)
579 goto free_next;
580 goto out;
581 }
582
583 count = ADDRS_PER_PAGE(dn.node_page, inode);
584
585 count -= dn.ofs_in_node;
586 f2fs_bug_on(sbi, count < 0);
587
588 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
589 truncate_data_blocks_range(&dn, count);
590 free_from += count;
591 }
592
593 f2fs_put_dnode(&dn);
594free_next:
595 err = truncate_inode_blocks(inode, free_from);
596out:
597 if (lock)
598 f2fs_unlock_op(sbi);
599
600 /* lastly zero out the first data page */
601 if (!err)
602 err = truncate_partial_data_page(inode, from, truncate_page);
603
604 trace_f2fs_truncate_blocks_exit(inode, err);
605 return err;
606}
607
608int f2fs_truncate(struct inode *inode, bool lock)
609{
610 int err;
611
612 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
613 S_ISLNK(inode->i_mode)))
614 return 0;
615
616 trace_f2fs_truncate(inode);
617
618 /* we should check inline_data size */
619 if (!f2fs_may_inline_data(inode)) {
620 err = f2fs_convert_inline_inode(inode);
621 if (err)
622 return err;
623 }
624
625 err = truncate_blocks(inode, i_size_read(inode), lock);
626 if (err)
627 return err;
628
629 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
630 mark_inode_dirty(inode);
631 return 0;
632}
633
634int f2fs_getattr(struct vfsmount *mnt,
635 struct dentry *dentry, struct kstat *stat)
636{
637 struct inode *inode = d_inode(dentry);
638 generic_fillattr(inode, stat);
639 stat->blocks <<= 3;
640 return 0;
641}
642
643#ifdef CONFIG_F2FS_FS_POSIX_ACL
644static void __setattr_copy(struct inode *inode, const struct iattr *attr)
645{
646 struct f2fs_inode_info *fi = F2FS_I(inode);
647 unsigned int ia_valid = attr->ia_valid;
648
649 if (ia_valid & ATTR_UID)
650 inode->i_uid = attr->ia_uid;
651 if (ia_valid & ATTR_GID)
652 inode->i_gid = attr->ia_gid;
653 if (ia_valid & ATTR_ATIME)
654 inode->i_atime = timespec_trunc(attr->ia_atime,
655 inode->i_sb->s_time_gran);
656 if (ia_valid & ATTR_MTIME)
657 inode->i_mtime = timespec_trunc(attr->ia_mtime,
658 inode->i_sb->s_time_gran);
659 if (ia_valid & ATTR_CTIME)
660 inode->i_ctime = timespec_trunc(attr->ia_ctime,
661 inode->i_sb->s_time_gran);
662 if (ia_valid & ATTR_MODE) {
663 umode_t mode = attr->ia_mode;
664
665 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
666 mode &= ~S_ISGID;
667 set_acl_inode(fi, mode);
668 }
669}
670#else
671#define __setattr_copy setattr_copy
672#endif
673
674int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
675{
676 struct inode *inode = d_inode(dentry);
677 struct f2fs_inode_info *fi = F2FS_I(inode);
678 int err;
679
680 err = inode_change_ok(inode, attr);
681 if (err)
682 return err;
683
684 if (attr->ia_valid & ATTR_SIZE) {
685 if (f2fs_encrypted_inode(inode) &&
686 fscrypt_get_encryption_info(inode))
687 return -EACCES;
688
689 if (attr->ia_size <= i_size_read(inode)) {
690 truncate_setsize(inode, attr->ia_size);
691 err = f2fs_truncate(inode, true);
692 if (err)
693 return err;
694 f2fs_balance_fs(F2FS_I_SB(inode), true);
695 } else {
696 /*
697 * do not trim all blocks after i_size if target size is
698 * larger than i_size.
699 */
700 truncate_setsize(inode, attr->ia_size);
701
702 /* should convert inline inode here */
703 if (!f2fs_may_inline_data(inode)) {
704 err = f2fs_convert_inline_inode(inode);
705 if (err)
706 return err;
707 }
708 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
709 }
710 }
711
712 __setattr_copy(inode, attr);
713
714 if (attr->ia_valid & ATTR_MODE) {
715 err = posix_acl_chmod(inode, get_inode_mode(inode));
716 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
717 inode->i_mode = fi->i_acl_mode;
718 clear_inode_flag(fi, FI_ACL_MODE);
719 }
720 }
721
722 mark_inode_dirty(inode);
723 return err;
724}
725
726const struct inode_operations f2fs_file_inode_operations = {
727 .getattr = f2fs_getattr,
728 .setattr = f2fs_setattr,
729 .get_acl = f2fs_get_acl,
730 .set_acl = f2fs_set_acl,
731#ifdef CONFIG_F2FS_FS_XATTR
732 .setxattr = generic_setxattr,
733 .getxattr = generic_getxattr,
734 .listxattr = f2fs_listxattr,
735 .removexattr = generic_removexattr,
736#endif
737 .fiemap = f2fs_fiemap,
738};
739
740static int fill_zero(struct inode *inode, pgoff_t index,
741 loff_t start, loff_t len)
742{
743 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
744 struct page *page;
745
746 if (!len)
747 return 0;
748
749 f2fs_balance_fs(sbi, true);
750
751 f2fs_lock_op(sbi);
752 page = get_new_data_page(inode, NULL, index, false);
753 f2fs_unlock_op(sbi);
754
755 if (IS_ERR(page))
756 return PTR_ERR(page);
757
758 f2fs_wait_on_page_writeback(page, DATA, true);
759 zero_user(page, start, len);
760 set_page_dirty(page);
761 f2fs_put_page(page, 1);
762 return 0;
763}
764
765int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
766{
767 int err;
768
769 while (pg_start < pg_end) {
770 struct dnode_of_data dn;
771 pgoff_t end_offset, count;
772
773 set_new_dnode(&dn, inode, NULL, NULL, 0);
774 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
775 if (err) {
776 if (err == -ENOENT) {
777 pg_start++;
778 continue;
779 }
780 return err;
781 }
782
783 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
784 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
785
786 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
787
788 truncate_data_blocks_range(&dn, count);
789 f2fs_put_dnode(&dn);
790
791 pg_start += count;
792 }
793 return 0;
794}
795
796static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
797{
798 pgoff_t pg_start, pg_end;
799 loff_t off_start, off_end;
800 int ret;
801
802 ret = f2fs_convert_inline_inode(inode);
803 if (ret)
804 return ret;
805
806 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
807 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
808
809 off_start = offset & (PAGE_SIZE - 1);
810 off_end = (offset + len) & (PAGE_SIZE - 1);
811
812 if (pg_start == pg_end) {
813 ret = fill_zero(inode, pg_start, off_start,
814 off_end - off_start);
815 if (ret)
816 return ret;
817 } else {
818 if (off_start) {
819 ret = fill_zero(inode, pg_start++, off_start,
820 PAGE_SIZE - off_start);
821 if (ret)
822 return ret;
823 }
824 if (off_end) {
825 ret = fill_zero(inode, pg_end, 0, off_end);
826 if (ret)
827 return ret;
828 }
829
830 if (pg_start < pg_end) {
831 struct address_space *mapping = inode->i_mapping;
832 loff_t blk_start, blk_end;
833 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
834
835 f2fs_balance_fs(sbi, true);
836
837 blk_start = (loff_t)pg_start << PAGE_SHIFT;
838 blk_end = (loff_t)pg_end << PAGE_SHIFT;
839 truncate_inode_pages_range(mapping, blk_start,
840 blk_end - 1);
841
842 f2fs_lock_op(sbi);
843 ret = truncate_hole(inode, pg_start, pg_end);
844 f2fs_unlock_op(sbi);
845 }
846 }
847
848 return ret;
849}
850
851static int __exchange_data_block(struct inode *inode, pgoff_t src,
852 pgoff_t dst, bool full)
853{
854 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
855 struct dnode_of_data dn;
856 block_t new_addr;
857 bool do_replace = false;
858 int ret;
859
860 set_new_dnode(&dn, inode, NULL, NULL, 0);
861 ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
862 if (ret && ret != -ENOENT) {
863 return ret;
864 } else if (ret == -ENOENT) {
865 new_addr = NULL_ADDR;
866 } else {
867 new_addr = dn.data_blkaddr;
868 if (!is_checkpointed_data(sbi, new_addr)) {
869 /* do not invalidate this block address */
870 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
871 do_replace = true;
872 }
873 f2fs_put_dnode(&dn);
874 }
875
876 if (new_addr == NULL_ADDR)
877 return full ? truncate_hole(inode, dst, dst + 1) : 0;
878
879 if (do_replace) {
880 struct page *ipage = get_node_page(sbi, inode->i_ino);
881 struct node_info ni;
882
883 if (IS_ERR(ipage)) {
884 ret = PTR_ERR(ipage);
885 goto err_out;
886 }
887
888 set_new_dnode(&dn, inode, ipage, NULL, 0);
889 ret = f2fs_reserve_block(&dn, dst);
890 if (ret)
891 goto err_out;
892
893 truncate_data_blocks_range(&dn, 1);
894
895 get_node_info(sbi, dn.nid, &ni);
896 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
897 ni.version, true, false);
898 f2fs_put_dnode(&dn);
899 } else {
900 struct page *psrc, *pdst;
901
902 psrc = get_lock_data_page(inode, src, true);
903 if (IS_ERR(psrc))
904 return PTR_ERR(psrc);
905 pdst = get_new_data_page(inode, NULL, dst, true);
906 if (IS_ERR(pdst)) {
907 f2fs_put_page(psrc, 1);
908 return PTR_ERR(pdst);
909 }
910 f2fs_copy_page(psrc, pdst);
911 set_page_dirty(pdst);
912 f2fs_put_page(pdst, 1);
913 f2fs_put_page(psrc, 1);
914
915 return truncate_hole(inode, src, src + 1);
916 }
917 return 0;
918
919err_out:
920 if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
921 f2fs_update_data_blkaddr(&dn, new_addr);
922 f2fs_put_dnode(&dn);
923 }
924 return ret;
925}
926
927static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
928{
929 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
930 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
931 int ret = 0;
932
933 for (; end < nrpages; start++, end++) {
934 f2fs_balance_fs(sbi, true);
935 f2fs_lock_op(sbi);
936 ret = __exchange_data_block(inode, end, start, true);
937 f2fs_unlock_op(sbi);
938 if (ret)
939 break;
940 }
941 return ret;
942}
943
944static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
945{
946 pgoff_t pg_start, pg_end;
947 loff_t new_size;
948 int ret;
949
950 if (offset + len >= i_size_read(inode))
951 return -EINVAL;
952
953 /* collapse range should be aligned to block size of f2fs. */
954 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
955 return -EINVAL;
956
957 ret = f2fs_convert_inline_inode(inode);
958 if (ret)
959 return ret;
960
961 pg_start = offset >> PAGE_SHIFT;
962 pg_end = (offset + len) >> PAGE_SHIFT;
963
964 /* write out all dirty pages from offset */
965 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
966 if (ret)
967 return ret;
968
969 truncate_pagecache(inode, offset);
970
971 ret = f2fs_do_collapse(inode, pg_start, pg_end);
972 if (ret)
973 return ret;
974
975 /* write out all moved pages, if possible */
976 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
977 truncate_pagecache(inode, offset);
978
979 new_size = i_size_read(inode) - len;
980 truncate_pagecache(inode, new_size);
981
982 ret = truncate_blocks(inode, new_size, true);
983 if (!ret)
984 i_size_write(inode, new_size);
985
986 return ret;
987}
988
989static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
990 int mode)
991{
992 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
993 struct address_space *mapping = inode->i_mapping;
994 pgoff_t index, pg_start, pg_end;
995 loff_t new_size = i_size_read(inode);
996 loff_t off_start, off_end;
997 int ret = 0;
998
999 ret = inode_newsize_ok(inode, (len + offset));
1000 if (ret)
1001 return ret;
1002
1003 ret = f2fs_convert_inline_inode(inode);
1004 if (ret)
1005 return ret;
1006
1007 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1008 if (ret)
1009 return ret;
1010
1011 truncate_pagecache_range(inode, offset, offset + len - 1);
1012
1013 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1014 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1015
1016 off_start = offset & (PAGE_SIZE - 1);
1017 off_end = (offset + len) & (PAGE_SIZE - 1);
1018
1019 if (pg_start == pg_end) {
1020 ret = fill_zero(inode, pg_start, off_start,
1021 off_end - off_start);
1022 if (ret)
1023 return ret;
1024
1025 if (offset + len > new_size)
1026 new_size = offset + len;
1027 new_size = max_t(loff_t, new_size, offset + len);
1028 } else {
1029 if (off_start) {
1030 ret = fill_zero(inode, pg_start++, off_start,
1031 PAGE_SIZE - off_start);
1032 if (ret)
1033 return ret;
1034
1035 new_size = max_t(loff_t, new_size,
1036 (loff_t)pg_start << PAGE_SHIFT);
1037 }
1038
1039 for (index = pg_start; index < pg_end; index++) {
1040 struct dnode_of_data dn;
1041 struct page *ipage;
1042
1043 f2fs_lock_op(sbi);
1044
1045 ipage = get_node_page(sbi, inode->i_ino);
1046 if (IS_ERR(ipage)) {
1047 ret = PTR_ERR(ipage);
1048 f2fs_unlock_op(sbi);
1049 goto out;
1050 }
1051
1052 set_new_dnode(&dn, inode, ipage, NULL, 0);
1053 ret = f2fs_reserve_block(&dn, index);
1054 if (ret) {
1055 f2fs_unlock_op(sbi);
1056 goto out;
1057 }
1058
1059 if (dn.data_blkaddr != NEW_ADDR) {
1060 invalidate_blocks(sbi, dn.data_blkaddr);
1061 f2fs_update_data_blkaddr(&dn, NEW_ADDR);
1062 }
1063 f2fs_put_dnode(&dn);
1064 f2fs_unlock_op(sbi);
1065
1066 new_size = max_t(loff_t, new_size,
1067 (loff_t)(index + 1) << PAGE_SHIFT);
1068 }
1069
1070 if (off_end) {
1071 ret = fill_zero(inode, pg_end, 0, off_end);
1072 if (ret)
1073 goto out;
1074
1075 new_size = max_t(loff_t, new_size, offset + len);
1076 }
1077 }
1078
1079out:
1080 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
1081 i_size_write(inode, new_size);
1082 mark_inode_dirty(inode);
1083 update_inode_page(inode);
1084 }
1085
1086 return ret;
1087}
1088
1089static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1090{
1091 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1092 pgoff_t pg_start, pg_end, delta, nrpages, idx;
1093 loff_t new_size;
1094 int ret = 0;
1095
1096 new_size = i_size_read(inode) + len;
1097 if (new_size > inode->i_sb->s_maxbytes)
1098 return -EFBIG;
1099
1100 if (offset >= i_size_read(inode))
1101 return -EINVAL;
1102
1103 /* insert range should be aligned to block size of f2fs. */
1104 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1105 return -EINVAL;
1106
1107 ret = f2fs_convert_inline_inode(inode);
1108 if (ret)
1109 return ret;
1110
1111 f2fs_balance_fs(sbi, true);
1112
1113 ret = truncate_blocks(inode, i_size_read(inode), true);
1114 if (ret)
1115 return ret;
1116
1117 /* write out all dirty pages from offset */
1118 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1119 if (ret)
1120 return ret;
1121
1122 truncate_pagecache(inode, offset);
1123
1124 pg_start = offset >> PAGE_SHIFT;
1125 pg_end = (offset + len) >> PAGE_SHIFT;
1126 delta = pg_end - pg_start;
1127 nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1128
1129 for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
1130 f2fs_lock_op(sbi);
1131 ret = __exchange_data_block(inode, idx, idx + delta, false);
1132 f2fs_unlock_op(sbi);
1133 if (ret)
1134 break;
1135 }
1136
1137 /* write out all moved pages, if possible */
1138 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1139 truncate_pagecache(inode, offset);
1140
1141 if (!ret)
1142 i_size_write(inode, new_size);
1143 return ret;
1144}
1145
1146static int expand_inode_data(struct inode *inode, loff_t offset,
1147 loff_t len, int mode)
1148{
1149 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1150 pgoff_t index, pg_start, pg_end;
1151 loff_t new_size = i_size_read(inode);
1152 loff_t off_start, off_end;
1153 int ret = 0;
1154
1155 ret = inode_newsize_ok(inode, (len + offset));
1156 if (ret)
1157 return ret;
1158
1159 ret = f2fs_convert_inline_inode(inode);
1160 if (ret)
1161 return ret;
1162
1163 f2fs_balance_fs(sbi, true);
1164
1165 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1166 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1167
1168 off_start = offset & (PAGE_SIZE - 1);
1169 off_end = (offset + len) & (PAGE_SIZE - 1);
1170
1171 f2fs_lock_op(sbi);
1172
1173 for (index = pg_start; index <= pg_end; index++) {
1174 struct dnode_of_data dn;
1175
1176 if (index == pg_end && !off_end)
1177 goto noalloc;
1178
1179 set_new_dnode(&dn, inode, NULL, NULL, 0);
1180 ret = f2fs_reserve_block(&dn, index);
1181 if (ret)
1182 break;
1183noalloc:
1184 if (pg_start == pg_end)
1185 new_size = offset + len;
1186 else if (index == pg_start && off_start)
1187 new_size = (loff_t)(index + 1) << PAGE_SHIFT;
1188 else if (index == pg_end)
1189 new_size = ((loff_t)index << PAGE_SHIFT) +
1190 off_end;
1191 else
1192 new_size += PAGE_SIZE;
1193 }
1194
1195 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
1196 i_size_read(inode) < new_size) {
1197 i_size_write(inode, new_size);
1198 mark_inode_dirty(inode);
1199 update_inode_page(inode);
1200 }
1201 f2fs_unlock_op(sbi);
1202
1203 return ret;
1204}
1205
1206static long f2fs_fallocate(struct file *file, int mode,
1207 loff_t offset, loff_t len)
1208{
1209 struct inode *inode = file_inode(file);
1210 long ret = 0;
1211
1212 /* f2fs only support ->fallocate for regular file */
1213 if (!S_ISREG(inode->i_mode))
1214 return -EINVAL;
1215
1216 if (f2fs_encrypted_inode(inode) &&
1217 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1218 return -EOPNOTSUPP;
1219
1220 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1221 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1222 FALLOC_FL_INSERT_RANGE))
1223 return -EOPNOTSUPP;
1224
1225 inode_lock(inode);
1226
1227 if (mode & FALLOC_FL_PUNCH_HOLE) {
1228 if (offset >= inode->i_size)
1229 goto out;
1230
1231 ret = punch_hole(inode, offset, len);
1232 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1233 ret = f2fs_collapse_range(inode, offset, len);
1234 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1235 ret = f2fs_zero_range(inode, offset, len, mode);
1236 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1237 ret = f2fs_insert_range(inode, offset, len);
1238 } else {
1239 ret = expand_inode_data(inode, offset, len, mode);
1240 }
1241
1242 if (!ret) {
1243 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1244 mark_inode_dirty(inode);
1245 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1246 }
1247
1248out:
1249 inode_unlock(inode);
1250
1251 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1252 return ret;
1253}
1254
1255static int f2fs_release_file(struct inode *inode, struct file *filp)
1256{
1257 /* some remained atomic pages should discarded */
1258 if (f2fs_is_atomic_file(inode))
1259 drop_inmem_pages(inode);
1260 if (f2fs_is_volatile_file(inode)) {
1261 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1262 filemap_fdatawrite(inode->i_mapping);
1263 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
1264 }
1265 return 0;
1266}
1267
1268#define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1269#define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1270
1271static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1272{
1273 if (S_ISDIR(mode))
1274 return flags;
1275 else if (S_ISREG(mode))
1276 return flags & F2FS_REG_FLMASK;
1277 else
1278 return flags & F2FS_OTHER_FLMASK;
1279}
1280
1281static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1282{
1283 struct inode *inode = file_inode(filp);
1284 struct f2fs_inode_info *fi = F2FS_I(inode);
1285 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1286 return put_user(flags, (int __user *)arg);
1287}
1288
1289static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1290{
1291 struct inode *inode = file_inode(filp);
1292 struct f2fs_inode_info *fi = F2FS_I(inode);
1293 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1294 unsigned int oldflags;
1295 int ret;
1296
1297 ret = mnt_want_write_file(filp);
1298 if (ret)
1299 return ret;
1300
1301 if (!inode_owner_or_capable(inode)) {
1302 ret = -EACCES;
1303 goto out;
1304 }
1305
1306 if (get_user(flags, (int __user *)arg)) {
1307 ret = -EFAULT;
1308 goto out;
1309 }
1310
1311 flags = f2fs_mask_flags(inode->i_mode, flags);
1312
1313 inode_lock(inode);
1314
1315 oldflags = fi->i_flags;
1316
1317 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1318 if (!capable(CAP_LINUX_IMMUTABLE)) {
1319 inode_unlock(inode);
1320 ret = -EPERM;
1321 goto out;
1322 }
1323 }
1324
1325 flags = flags & FS_FL_USER_MODIFIABLE;
1326 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1327 fi->i_flags = flags;
1328 inode_unlock(inode);
1329
1330 f2fs_set_inode_flags(inode);
1331 inode->i_ctime = CURRENT_TIME;
1332 mark_inode_dirty(inode);
1333out:
1334 mnt_drop_write_file(filp);
1335 return ret;
1336}
1337
1338static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1339{
1340 struct inode *inode = file_inode(filp);
1341
1342 return put_user(inode->i_generation, (int __user *)arg);
1343}
1344
1345static int f2fs_ioc_start_atomic_write(struct file *filp)
1346{
1347 struct inode *inode = file_inode(filp);
1348 int ret;
1349
1350 if (!inode_owner_or_capable(inode))
1351 return -EACCES;
1352
1353 if (f2fs_is_atomic_file(inode))
1354 return 0;
1355
1356 ret = f2fs_convert_inline_inode(inode);
1357 if (ret)
1358 return ret;
1359
1360 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1361 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1362
1363 return 0;
1364}
1365
1366static int f2fs_ioc_commit_atomic_write(struct file *filp)
1367{
1368 struct inode *inode = file_inode(filp);
1369 int ret;
1370
1371 if (!inode_owner_or_capable(inode))
1372 return -EACCES;
1373
1374 if (f2fs_is_volatile_file(inode))
1375 return 0;
1376
1377 ret = mnt_want_write_file(filp);
1378 if (ret)
1379 return ret;
1380
1381 if (f2fs_is_atomic_file(inode)) {
1382 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1383 ret = commit_inmem_pages(inode);
1384 if (ret) {
1385 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1386 goto err_out;
1387 }
1388 }
1389
1390 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1391err_out:
1392 mnt_drop_write_file(filp);
1393 return ret;
1394}
1395
1396static int f2fs_ioc_start_volatile_write(struct file *filp)
1397{
1398 struct inode *inode = file_inode(filp);
1399 int ret;
1400
1401 if (!inode_owner_or_capable(inode))
1402 return -EACCES;
1403
1404 if (f2fs_is_volatile_file(inode))
1405 return 0;
1406
1407 ret = f2fs_convert_inline_inode(inode);
1408 if (ret)
1409 return ret;
1410
1411 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1412 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1413 return 0;
1414}
1415
1416static int f2fs_ioc_release_volatile_write(struct file *filp)
1417{
1418 struct inode *inode = file_inode(filp);
1419
1420 if (!inode_owner_or_capable(inode))
1421 return -EACCES;
1422
1423 if (!f2fs_is_volatile_file(inode))
1424 return 0;
1425
1426 if (!f2fs_is_first_block_written(inode))
1427 return truncate_partial_data_page(inode, 0, true);
1428
1429 return punch_hole(inode, 0, F2FS_BLKSIZE);
1430}
1431
1432static int f2fs_ioc_abort_volatile_write(struct file *filp)
1433{
1434 struct inode *inode = file_inode(filp);
1435 int ret;
1436
1437 if (!inode_owner_or_capable(inode))
1438 return -EACCES;
1439
1440 ret = mnt_want_write_file(filp);
1441 if (ret)
1442 return ret;
1443
1444 if (f2fs_is_atomic_file(inode)) {
1445 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1446 drop_inmem_pages(inode);
1447 }
1448 if (f2fs_is_volatile_file(inode)) {
1449 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1450 ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
1451 }
1452
1453 mnt_drop_write_file(filp);
1454 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1455 return ret;
1456}
1457
1458static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1459{
1460 struct inode *inode = file_inode(filp);
1461 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1462 struct super_block *sb = sbi->sb;
1463 __u32 in;
1464
1465 if (!capable(CAP_SYS_ADMIN))
1466 return -EPERM;
1467
1468 if (get_user(in, (__u32 __user *)arg))
1469 return -EFAULT;
1470
1471 switch (in) {
1472 case F2FS_GOING_DOWN_FULLSYNC:
1473 sb = freeze_bdev(sb->s_bdev);
1474 if (sb && !IS_ERR(sb)) {
1475 f2fs_stop_checkpoint(sbi);
1476 thaw_bdev(sb->s_bdev, sb);
1477 }
1478 break;
1479 case F2FS_GOING_DOWN_METASYNC:
1480 /* do checkpoint only */
1481 f2fs_sync_fs(sb, 1);
1482 f2fs_stop_checkpoint(sbi);
1483 break;
1484 case F2FS_GOING_DOWN_NOSYNC:
1485 f2fs_stop_checkpoint(sbi);
1486 break;
1487 case F2FS_GOING_DOWN_METAFLUSH:
1488 sync_meta_pages(sbi, META, LONG_MAX);
1489 f2fs_stop_checkpoint(sbi);
1490 break;
1491 default:
1492 return -EINVAL;
1493 }
1494 f2fs_update_time(sbi, REQ_TIME);
1495 return 0;
1496}
1497
1498static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1499{
1500 struct inode *inode = file_inode(filp);
1501 struct super_block *sb = inode->i_sb;
1502 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1503 struct fstrim_range range;
1504 int ret;
1505
1506 if (!capable(CAP_SYS_ADMIN))
1507 return -EPERM;
1508
1509 if (!blk_queue_discard(q))
1510 return -EOPNOTSUPP;
1511
1512 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1513 sizeof(range)))
1514 return -EFAULT;
1515
1516 range.minlen = max((unsigned int)range.minlen,
1517 q->limits.discard_granularity);
1518 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1519 if (ret < 0)
1520 return ret;
1521
1522 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1523 sizeof(range)))
1524 return -EFAULT;
1525 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1526 return 0;
1527}
1528
1529static bool uuid_is_nonzero(__u8 u[16])
1530{
1531 int i;
1532
1533 for (i = 0; i < 16; i++)
1534 if (u[i])
1535 return true;
1536 return false;
1537}
1538
1539static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1540{
1541 struct fscrypt_policy policy;
1542 struct inode *inode = file_inode(filp);
1543
1544 if (copy_from_user(&policy, (struct fscrypt_policy __user *)arg,
1545 sizeof(policy)))
1546 return -EFAULT;
1547
1548 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1549 return fscrypt_process_policy(inode, &policy);
1550}
1551
1552static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1553{
1554 struct fscrypt_policy policy;
1555 struct inode *inode = file_inode(filp);
1556 int err;
1557
1558 err = fscrypt_get_policy(inode, &policy);
1559 if (err)
1560 return err;
1561
1562 if (copy_to_user((struct fscrypt_policy __user *)arg, &policy, sizeof(policy)))
1563 return -EFAULT;
1564 return 0;
1565}
1566
1567static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1568{
1569 struct inode *inode = file_inode(filp);
1570 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1571 int err;
1572
1573 if (!f2fs_sb_has_crypto(inode->i_sb))
1574 return -EOPNOTSUPP;
1575
1576 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1577 goto got_it;
1578
1579 err = mnt_want_write_file(filp);
1580 if (err)
1581 return err;
1582
1583 /* update superblock with uuid */
1584 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1585
1586 err = f2fs_commit_super(sbi, false);
1587 if (err) {
1588 /* undo new data */
1589 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1590 mnt_drop_write_file(filp);
1591 return err;
1592 }
1593 mnt_drop_write_file(filp);
1594got_it:
1595 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1596 16))
1597 return -EFAULT;
1598 return 0;
1599}
1600
1601static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1602{
1603 struct inode *inode = file_inode(filp);
1604 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1605 __u32 sync;
1606
1607 if (!capable(CAP_SYS_ADMIN))
1608 return -EPERM;
1609
1610 if (get_user(sync, (__u32 __user *)arg))
1611 return -EFAULT;
1612
1613 if (f2fs_readonly(sbi->sb))
1614 return -EROFS;
1615
1616 if (!sync) {
1617 if (!mutex_trylock(&sbi->gc_mutex))
1618 return -EBUSY;
1619 } else {
1620 mutex_lock(&sbi->gc_mutex);
1621 }
1622
1623 return f2fs_gc(sbi, sync);
1624}
1625
1626static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1627{
1628 struct inode *inode = file_inode(filp);
1629 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1630
1631 if (!capable(CAP_SYS_ADMIN))
1632 return -EPERM;
1633
1634 if (f2fs_readonly(sbi->sb))
1635 return -EROFS;
1636
1637 return f2fs_sync_fs(sbi->sb, 1);
1638}
1639
1640static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1641 struct file *filp,
1642 struct f2fs_defragment *range)
1643{
1644 struct inode *inode = file_inode(filp);
1645 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1646 struct extent_info ei;
1647 pgoff_t pg_start, pg_end;
1648 unsigned int blk_per_seg = sbi->blocks_per_seg;
1649 unsigned int total = 0, sec_num;
1650 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1651 block_t blk_end = 0;
1652 bool fragmented = false;
1653 int err;
1654
1655 /* if in-place-update policy is enabled, don't waste time here */
1656 if (need_inplace_update(inode))
1657 return -EINVAL;
1658
1659 pg_start = range->start >> PAGE_SHIFT;
1660 pg_end = (range->start + range->len) >> PAGE_SHIFT;
1661
1662 f2fs_balance_fs(sbi, true);
1663
1664 inode_lock(inode);
1665
1666 /* writeback all dirty pages in the range */
1667 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1668 range->start + range->len - 1);
1669 if (err)
1670 goto out;
1671
1672 /*
1673 * lookup mapping info in extent cache, skip defragmenting if physical
1674 * block addresses are continuous.
1675 */
1676 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1677 if (ei.fofs + ei.len >= pg_end)
1678 goto out;
1679 }
1680
1681 map.m_lblk = pg_start;
1682
1683 /*
1684 * lookup mapping info in dnode page cache, skip defragmenting if all
1685 * physical block addresses are continuous even if there are hole(s)
1686 * in logical blocks.
1687 */
1688 while (map.m_lblk < pg_end) {
1689 map.m_len = pg_end - map.m_lblk;
1690 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1691 if (err)
1692 goto out;
1693
1694 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1695 map.m_lblk++;
1696 continue;
1697 }
1698
1699 if (blk_end && blk_end != map.m_pblk) {
1700 fragmented = true;
1701 break;
1702 }
1703 blk_end = map.m_pblk + map.m_len;
1704
1705 map.m_lblk += map.m_len;
1706 }
1707
1708 if (!fragmented)
1709 goto out;
1710
1711 map.m_lblk = pg_start;
1712 map.m_len = pg_end - pg_start;
1713
1714 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1715
1716 /*
1717 * make sure there are enough free section for LFS allocation, this can
1718 * avoid defragment running in SSR mode when free section are allocated
1719 * intensively
1720 */
1721 if (has_not_enough_free_secs(sbi, sec_num)) {
1722 err = -EAGAIN;
1723 goto out;
1724 }
1725
1726 while (map.m_lblk < pg_end) {
1727 pgoff_t idx;
1728 int cnt = 0;
1729
1730do_map:
1731 map.m_len = pg_end - map.m_lblk;
1732 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1733 if (err)
1734 goto clear_out;
1735
1736 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1737 map.m_lblk++;
1738 continue;
1739 }
1740
1741 set_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1742
1743 idx = map.m_lblk;
1744 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1745 struct page *page;
1746
1747 page = get_lock_data_page(inode, idx, true);
1748 if (IS_ERR(page)) {
1749 err = PTR_ERR(page);
1750 goto clear_out;
1751 }
1752
1753 set_page_dirty(page);
1754 f2fs_put_page(page, 1);
1755
1756 idx++;
1757 cnt++;
1758 total++;
1759 }
1760
1761 map.m_lblk = idx;
1762
1763 if (idx < pg_end && cnt < blk_per_seg)
1764 goto do_map;
1765
1766 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1767
1768 err = filemap_fdatawrite(inode->i_mapping);
1769 if (err)
1770 goto out;
1771 }
1772clear_out:
1773 clear_inode_flag(F2FS_I(inode), FI_DO_DEFRAG);
1774out:
1775 inode_unlock(inode);
1776 if (!err)
1777 range->len = (u64)total << PAGE_SHIFT;
1778 return err;
1779}
1780
1781static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
1782{
1783 struct inode *inode = file_inode(filp);
1784 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1785 struct f2fs_defragment range;
1786 int err;
1787
1788 if (!capable(CAP_SYS_ADMIN))
1789 return -EPERM;
1790
1791 if (!S_ISREG(inode->i_mode))
1792 return -EINVAL;
1793
1794 err = mnt_want_write_file(filp);
1795 if (err)
1796 return err;
1797
1798 if (f2fs_readonly(sbi->sb)) {
1799 err = -EROFS;
1800 goto out;
1801 }
1802
1803 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
1804 sizeof(range))) {
1805 err = -EFAULT;
1806 goto out;
1807 }
1808
1809 /* verify alignment of offset & size */
1810 if (range.start & (F2FS_BLKSIZE - 1) ||
1811 range.len & (F2FS_BLKSIZE - 1)) {
1812 err = -EINVAL;
1813 goto out;
1814 }
1815
1816 err = f2fs_defragment_range(sbi, filp, &range);
1817 f2fs_update_time(sbi, REQ_TIME);
1818 if (err < 0)
1819 goto out;
1820
1821 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
1822 sizeof(range)))
1823 err = -EFAULT;
1824out:
1825 mnt_drop_write_file(filp);
1826 return err;
1827}
1828
1829long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1830{
1831 switch (cmd) {
1832 case F2FS_IOC_GETFLAGS:
1833 return f2fs_ioc_getflags(filp, arg);
1834 case F2FS_IOC_SETFLAGS:
1835 return f2fs_ioc_setflags(filp, arg);
1836 case F2FS_IOC_GETVERSION:
1837 return f2fs_ioc_getversion(filp, arg);
1838 case F2FS_IOC_START_ATOMIC_WRITE:
1839 return f2fs_ioc_start_atomic_write(filp);
1840 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1841 return f2fs_ioc_commit_atomic_write(filp);
1842 case F2FS_IOC_START_VOLATILE_WRITE:
1843 return f2fs_ioc_start_volatile_write(filp);
1844 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1845 return f2fs_ioc_release_volatile_write(filp);
1846 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1847 return f2fs_ioc_abort_volatile_write(filp);
1848 case F2FS_IOC_SHUTDOWN:
1849 return f2fs_ioc_shutdown(filp, arg);
1850 case FITRIM:
1851 return f2fs_ioc_fitrim(filp, arg);
1852 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1853 return f2fs_ioc_set_encryption_policy(filp, arg);
1854 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1855 return f2fs_ioc_get_encryption_policy(filp, arg);
1856 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1857 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
1858 case F2FS_IOC_GARBAGE_COLLECT:
1859 return f2fs_ioc_gc(filp, arg);
1860 case F2FS_IOC_WRITE_CHECKPOINT:
1861 return f2fs_ioc_write_checkpoint(filp, arg);
1862 case F2FS_IOC_DEFRAGMENT:
1863 return f2fs_ioc_defragment(filp, arg);
1864 default:
1865 return -ENOTTY;
1866 }
1867}
1868
1869static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1870{
1871 struct file *file = iocb->ki_filp;
1872 struct inode *inode = file_inode(file);
1873 ssize_t ret;
1874
1875 if (f2fs_encrypted_inode(inode) &&
1876 !fscrypt_has_encryption_key(inode) &&
1877 fscrypt_get_encryption_info(inode))
1878 return -EACCES;
1879
1880 inode_lock(inode);
1881 ret = generic_write_checks(iocb, from);
1882 if (ret > 0) {
1883 ret = f2fs_preallocate_blocks(iocb, from);
1884 if (!ret)
1885 ret = __generic_file_write_iter(iocb, from);
1886 }
1887 inode_unlock(inode);
1888
1889 if (ret > 0) {
1890 ssize_t err;
1891
1892 err = generic_write_sync(file, iocb->ki_pos - ret, ret);
1893 if (err < 0)
1894 ret = err;
1895 }
1896 return ret;
1897}
1898
1899#ifdef CONFIG_COMPAT
1900long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1901{
1902 switch (cmd) {
1903 case F2FS_IOC32_GETFLAGS:
1904 cmd = F2FS_IOC_GETFLAGS;
1905 break;
1906 case F2FS_IOC32_SETFLAGS:
1907 cmd = F2FS_IOC_SETFLAGS;
1908 break;
1909 case F2FS_IOC32_GETVERSION:
1910 cmd = F2FS_IOC_GETVERSION;
1911 break;
1912 case F2FS_IOC_START_ATOMIC_WRITE:
1913 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1914 case F2FS_IOC_START_VOLATILE_WRITE:
1915 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1916 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1917 case F2FS_IOC_SHUTDOWN:
1918 case F2FS_IOC_SET_ENCRYPTION_POLICY:
1919 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
1920 case F2FS_IOC_GET_ENCRYPTION_POLICY:
1921 case F2FS_IOC_GARBAGE_COLLECT:
1922 case F2FS_IOC_WRITE_CHECKPOINT:
1923 case F2FS_IOC_DEFRAGMENT:
1924 break;
1925 default:
1926 return -ENOIOCTLCMD;
1927 }
1928 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1929}
1930#endif
1931
1932const struct file_operations f2fs_file_operations = {
1933 .llseek = f2fs_llseek,
1934 .read_iter = generic_file_read_iter,
1935 .write_iter = f2fs_file_write_iter,
1936 .open = f2fs_file_open,
1937 .release = f2fs_release_file,
1938 .mmap = f2fs_file_mmap,
1939 .fsync = f2fs_sync_file,
1940 .fallocate = f2fs_fallocate,
1941 .unlocked_ioctl = f2fs_ioctl,
1942#ifdef CONFIG_COMPAT
1943 .compat_ioctl = f2fs_compat_ioctl,
1944#endif
1945 .splice_read = generic_file_splice_read,
1946 .splice_write = iter_file_splice_write,
1947};