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 #include <linux/fadvise.h>
27 #include <linux/iomap.h>
28
29 #include "f2fs.h"
30 #include "node.h"
31 #include "segment.h"
32 #include "xattr.h"
33 #include "acl.h"
34 #include "gc.h"
35 #include "iostat.h"
36 #include <trace/events/f2fs.h>
37 #include <uapi/linux/f2fs.h>
38
39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
40 {
41 struct inode *inode = file_inode(vmf->vma->vm_file);
42 vm_flags_t flags = vmf->vma->vm_flags;
43 vm_fault_t ret;
44
45 ret = filemap_fault(vmf);
46 if (ret & VM_FAULT_LOCKED)
47 f2fs_update_iostat(F2FS_I_SB(inode), inode,
48 APP_MAPPED_READ_IO, F2FS_BLKSIZE);
49
50 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret);
51
52 return ret;
53 }
54
55 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
56 {
57 struct page *page = vmf->page;
58 struct inode *inode = file_inode(vmf->vma->vm_file);
59 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
60 struct dnode_of_data dn;
61 bool need_alloc = !f2fs_is_pinned_file(inode);
62 int err = 0;
63 vm_fault_t ret;
64
65 if (unlikely(IS_IMMUTABLE(inode)))
66 return VM_FAULT_SIGBUS;
67
68 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
69 err = -EIO;
70 goto out;
71 }
72
73 if (unlikely(f2fs_cp_error(sbi))) {
74 err = -EIO;
75 goto out;
76 }
77
78 if (!f2fs_is_checkpoint_ready(sbi)) {
79 err = -ENOSPC;
80 goto out;
81 }
82
83 err = f2fs_convert_inline_inode(inode);
84 if (err)
85 goto out;
86
87 #ifdef CONFIG_F2FS_FS_COMPRESSION
88 if (f2fs_compressed_file(inode)) {
89 int ret = f2fs_is_compressed_cluster(inode, page->index);
90
91 if (ret < 0) {
92 err = ret;
93 goto out;
94 } else if (ret) {
95 need_alloc = false;
96 }
97 }
98 #endif
99 /* should do out of any locked page */
100 if (need_alloc)
101 f2fs_balance_fs(sbi, true);
102
103 sb_start_pagefault(inode->i_sb);
104
105 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
106
107 file_update_time(vmf->vma->vm_file);
108 filemap_invalidate_lock_shared(inode->i_mapping);
109 lock_page(page);
110 if (unlikely(page->mapping != inode->i_mapping ||
111 page_offset(page) > i_size_read(inode) ||
112 !PageUptodate(page))) {
113 unlock_page(page);
114 err = -EFAULT;
115 goto out_sem;
116 }
117
118 set_new_dnode(&dn, inode, NULL, NULL, 0);
119 if (need_alloc) {
120 /* block allocation */
121 err = f2fs_get_block_locked(&dn, page->index);
122 } else {
123 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
124 f2fs_put_dnode(&dn);
125 if (f2fs_is_pinned_file(inode) &&
126 !__is_valid_data_blkaddr(dn.data_blkaddr))
127 err = -EIO;
128 }
129
130 if (err) {
131 unlock_page(page);
132 goto out_sem;
133 }
134
135 f2fs_wait_on_page_writeback(page, DATA, false, true);
136
137 /* wait for GCed page writeback via META_MAPPING */
138 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
139
140 /*
141 * check to see if the page is mapped already (no holes)
142 */
143 if (PageMappedToDisk(page))
144 goto out_sem;
145
146 /* page is wholly or partially inside EOF */
147 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
148 i_size_read(inode)) {
149 loff_t offset;
150
151 offset = i_size_read(inode) & ~PAGE_MASK;
152 zero_user_segment(page, offset, PAGE_SIZE);
153 }
154 set_page_dirty(page);
155
156 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
157 f2fs_update_time(sbi, REQ_TIME);
158
159 out_sem:
160 filemap_invalidate_unlock_shared(inode->i_mapping);
161
162 sb_end_pagefault(inode->i_sb);
163 out:
164 ret = vmf_fs_error(err);
165
166 trace_f2fs_vm_page_mkwrite(inode, page->index, vmf->vma->vm_flags, ret);
167 return ret;
168 }
169
170 static const struct vm_operations_struct f2fs_file_vm_ops = {
171 .fault = f2fs_filemap_fault,
172 .map_pages = filemap_map_pages,
173 .page_mkwrite = f2fs_vm_page_mkwrite,
174 };
175
176 static int get_parent_ino(struct inode *inode, nid_t *pino)
177 {
178 struct dentry *dentry;
179
180 /*
181 * Make sure to get the non-deleted alias. The alias associated with
182 * the open file descriptor being fsync()'ed may be deleted already.
183 */
184 dentry = d_find_alias(inode);
185 if (!dentry)
186 return 0;
187
188 *pino = d_parent_ino(dentry);
189 dput(dentry);
190 return 1;
191 }
192
193 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
194 {
195 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
196 enum cp_reason_type cp_reason = CP_NO_NEEDED;
197
198 if (!S_ISREG(inode->i_mode))
199 cp_reason = CP_NON_REGULAR;
200 else if (f2fs_compressed_file(inode))
201 cp_reason = CP_COMPRESSED;
202 else if (inode->i_nlink != 1)
203 cp_reason = CP_HARDLINK;
204 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
205 cp_reason = CP_SB_NEED_CP;
206 else if (file_wrong_pino(inode))
207 cp_reason = CP_WRONG_PINO;
208 else if (!f2fs_space_for_roll_forward(sbi))
209 cp_reason = CP_NO_SPC_ROLL;
210 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
211 cp_reason = CP_NODE_NEED_CP;
212 else if (test_opt(sbi, FASTBOOT))
213 cp_reason = CP_FASTBOOT_MODE;
214 else if (F2FS_OPTION(sbi).active_logs == 2)
215 cp_reason = CP_SPEC_LOG_NUM;
216 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
217 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
218 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
219 TRANS_DIR_INO))
220 cp_reason = CP_RECOVER_DIR;
221 else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
222 XATTR_DIR_INO))
223 cp_reason = CP_XATTR_DIR;
224
225 return cp_reason;
226 }
227
228 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
229 {
230 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
231 bool ret = false;
232 /* But we need to avoid that there are some inode updates */
233 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
234 ret = true;
235 f2fs_put_page(i, 0);
236 return ret;
237 }
238
239 static void try_to_fix_pino(struct inode *inode)
240 {
241 struct f2fs_inode_info *fi = F2FS_I(inode);
242 nid_t pino;
243
244 f2fs_down_write(&fi->i_sem);
245 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
246 get_parent_ino(inode, &pino)) {
247 f2fs_i_pino_write(inode, pino);
248 file_got_pino(inode);
249 }
250 f2fs_up_write(&fi->i_sem);
251 }
252
253 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
254 int datasync, bool atomic)
255 {
256 struct inode *inode = file->f_mapping->host;
257 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
258 nid_t ino = inode->i_ino;
259 int ret = 0;
260 enum cp_reason_type cp_reason = 0;
261 struct writeback_control wbc = {
262 .sync_mode = WB_SYNC_ALL,
263 .nr_to_write = LONG_MAX,
264 .for_reclaim = 0,
265 };
266 unsigned int seq_id = 0;
267
268 if (unlikely(f2fs_readonly(inode->i_sb)))
269 return 0;
270
271 trace_f2fs_sync_file_enter(inode);
272
273 if (S_ISDIR(inode->i_mode))
274 goto go_write;
275
276 /* if fdatasync is triggered, let's do in-place-update */
277 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
278 set_inode_flag(inode, FI_NEED_IPU);
279 ret = file_write_and_wait_range(file, start, end);
280 clear_inode_flag(inode, FI_NEED_IPU);
281
282 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
283 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
284 return ret;
285 }
286
287 /* if the inode is dirty, let's recover all the time */
288 if (!f2fs_skip_inode_update(inode, datasync)) {
289 f2fs_write_inode(inode, NULL);
290 goto go_write;
291 }
292
293 /*
294 * if there is no written data, don't waste time to write recovery info.
295 */
296 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
297 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
298
299 /* it may call write_inode just prior to fsync */
300 if (need_inode_page_update(sbi, ino))
301 goto go_write;
302
303 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
304 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
305 goto flush_out;
306 goto out;
307 } else {
308 /*
309 * for OPU case, during fsync(), node can be persisted before
310 * data when lower device doesn't support write barrier, result
311 * in data corruption after SPO.
312 * So for strict fsync mode, force to use atomic write semantics
313 * to keep write order in between data/node and last node to
314 * avoid potential data corruption.
315 */
316 if (F2FS_OPTION(sbi).fsync_mode ==
317 FSYNC_MODE_STRICT && !atomic)
318 atomic = true;
319 }
320 go_write:
321 /*
322 * Both of fdatasync() and fsync() are able to be recovered from
323 * sudden-power-off.
324 */
325 f2fs_down_read(&F2FS_I(inode)->i_sem);
326 cp_reason = need_do_checkpoint(inode);
327 f2fs_up_read(&F2FS_I(inode)->i_sem);
328
329 if (cp_reason) {
330 /* all the dirty node pages should be flushed for POR */
331 ret = f2fs_sync_fs(inode->i_sb, 1);
332
333 /*
334 * We've secured consistency through sync_fs. Following pino
335 * will be used only for fsynced inodes after checkpoint.
336 */
337 try_to_fix_pino(inode);
338 clear_inode_flag(inode, FI_APPEND_WRITE);
339 clear_inode_flag(inode, FI_UPDATE_WRITE);
340 goto out;
341 }
342 sync_nodes:
343 atomic_inc(&sbi->wb_sync_req[NODE]);
344 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
345 atomic_dec(&sbi->wb_sync_req[NODE]);
346 if (ret)
347 goto out;
348
349 /* if cp_error was enabled, we should avoid infinite loop */
350 if (unlikely(f2fs_cp_error(sbi))) {
351 ret = -EIO;
352 goto out;
353 }
354
355 if (f2fs_need_inode_block_update(sbi, ino)) {
356 f2fs_mark_inode_dirty_sync(inode, true);
357 f2fs_write_inode(inode, NULL);
358 goto sync_nodes;
359 }
360
361 /*
362 * If it's atomic_write, it's just fine to keep write ordering. So
363 * here we don't need to wait for node write completion, since we use
364 * node chain which serializes node blocks. If one of node writes are
365 * reordered, we can see simply broken chain, resulting in stopping
366 * roll-forward recovery. It means we'll recover all or none node blocks
367 * given fsync mark.
368 */
369 if (!atomic) {
370 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
371 if (ret)
372 goto out;
373 }
374
375 /* once recovery info is written, don't need to tack this */
376 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
377 clear_inode_flag(inode, FI_APPEND_WRITE);
378 flush_out:
379 if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) ||
380 (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
381 ret = f2fs_issue_flush(sbi, inode->i_ino);
382 if (!ret) {
383 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
384 clear_inode_flag(inode, FI_UPDATE_WRITE);
385 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
386 }
387 f2fs_update_time(sbi, REQ_TIME);
388 out:
389 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
390 return ret;
391 }
392
393 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
394 {
395 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
396 return -EIO;
397 return f2fs_do_sync_file(file, start, end, datasync, false);
398 }
399
400 static bool __found_offset(struct address_space *mapping,
401 struct dnode_of_data *dn, pgoff_t index, int whence)
402 {
403 block_t blkaddr = f2fs_data_blkaddr(dn);
404 struct inode *inode = mapping->host;
405 bool compressed_cluster = false;
406
407 if (f2fs_compressed_file(inode)) {
408 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_page,
409 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size));
410
411 compressed_cluster = first_blkaddr == COMPRESS_ADDR;
412 }
413
414 switch (whence) {
415 case SEEK_DATA:
416 if (__is_valid_data_blkaddr(blkaddr))
417 return true;
418 if (blkaddr == NEW_ADDR &&
419 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
420 return true;
421 if (compressed_cluster)
422 return true;
423 break;
424 case SEEK_HOLE:
425 if (compressed_cluster)
426 return false;
427 if (blkaddr == NULL_ADDR)
428 return true;
429 break;
430 }
431 return false;
432 }
433
434 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
435 {
436 struct inode *inode = file->f_mapping->host;
437 loff_t maxbytes = inode->i_sb->s_maxbytes;
438 struct dnode_of_data dn;
439 pgoff_t pgofs, end_offset;
440 loff_t data_ofs = offset;
441 loff_t isize;
442 int err = 0;
443
444 inode_lock_shared(inode);
445
446 isize = i_size_read(inode);
447 if (offset >= isize)
448 goto fail;
449
450 /* handle inline data case */
451 if (f2fs_has_inline_data(inode)) {
452 if (whence == SEEK_HOLE) {
453 data_ofs = isize;
454 goto found;
455 } else if (whence == SEEK_DATA) {
456 data_ofs = offset;
457 goto found;
458 }
459 }
460
461 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
462
463 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
464 set_new_dnode(&dn, inode, NULL, NULL, 0);
465 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
466 if (err && err != -ENOENT) {
467 goto fail;
468 } else if (err == -ENOENT) {
469 /* direct node does not exists */
470 if (whence == SEEK_DATA) {
471 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
472 continue;
473 } else {
474 goto found;
475 }
476 }
477
478 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
479
480 /* find data/hole in dnode block */
481 for (; dn.ofs_in_node < end_offset;
482 dn.ofs_in_node++, pgofs++,
483 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
484 block_t blkaddr;
485
486 blkaddr = f2fs_data_blkaddr(&dn);
487
488 if (__is_valid_data_blkaddr(blkaddr) &&
489 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
490 blkaddr, DATA_GENERIC_ENHANCE)) {
491 f2fs_put_dnode(&dn);
492 goto fail;
493 }
494
495 if (__found_offset(file->f_mapping, &dn,
496 pgofs, whence)) {
497 f2fs_put_dnode(&dn);
498 goto found;
499 }
500 }
501 f2fs_put_dnode(&dn);
502 }
503
504 if (whence == SEEK_DATA)
505 goto fail;
506 found:
507 if (whence == SEEK_HOLE && data_ofs > isize)
508 data_ofs = isize;
509 inode_unlock_shared(inode);
510 return vfs_setpos(file, data_ofs, maxbytes);
511 fail:
512 inode_unlock_shared(inode);
513 return -ENXIO;
514 }
515
516 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
517 {
518 struct inode *inode = file->f_mapping->host;
519 loff_t maxbytes = inode->i_sb->s_maxbytes;
520
521 if (f2fs_compressed_file(inode))
522 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
523
524 switch (whence) {
525 case SEEK_SET:
526 case SEEK_CUR:
527 case SEEK_END:
528 return generic_file_llseek_size(file, offset, whence,
529 maxbytes, i_size_read(inode));
530 case SEEK_DATA:
531 case SEEK_HOLE:
532 if (offset < 0)
533 return -ENXIO;
534 return f2fs_seek_block(file, offset, whence);
535 }
536
537 return -EINVAL;
538 }
539
540 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
541 {
542 struct inode *inode = file_inode(file);
543
544 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
545 return -EIO;
546
547 if (!f2fs_is_compress_backend_ready(inode))
548 return -EOPNOTSUPP;
549
550 file_accessed(file);
551 vma->vm_ops = &f2fs_file_vm_ops;
552
553 f2fs_down_read(&F2FS_I(inode)->i_sem);
554 set_inode_flag(inode, FI_MMAP_FILE);
555 f2fs_up_read(&F2FS_I(inode)->i_sem);
556
557 return 0;
558 }
559
560 static int finish_preallocate_blocks(struct inode *inode)
561 {
562 int ret;
563
564 inode_lock(inode);
565 if (is_inode_flag_set(inode, FI_OPENED_FILE)) {
566 inode_unlock(inode);
567 return 0;
568 }
569
570 if (!file_should_truncate(inode)) {
571 set_inode_flag(inode, FI_OPENED_FILE);
572 inode_unlock(inode);
573 return 0;
574 }
575
576 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
577 filemap_invalidate_lock(inode->i_mapping);
578
579 truncate_setsize(inode, i_size_read(inode));
580 ret = f2fs_truncate(inode);
581
582 filemap_invalidate_unlock(inode->i_mapping);
583 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
584
585 if (!ret)
586 set_inode_flag(inode, FI_OPENED_FILE);
587
588 inode_unlock(inode);
589 if (ret)
590 return ret;
591
592 file_dont_truncate(inode);
593 return 0;
594 }
595
596 static int f2fs_file_open(struct inode *inode, struct file *filp)
597 {
598 int err = fscrypt_file_open(inode, filp);
599
600 if (err)
601 return err;
602
603 if (!f2fs_is_compress_backend_ready(inode))
604 return -EOPNOTSUPP;
605
606 err = fsverity_file_open(inode, filp);
607 if (err)
608 return err;
609
610 filp->f_mode |= FMODE_NOWAIT;
611 filp->f_mode |= FMODE_CAN_ODIRECT;
612
613 err = dquot_file_open(inode, filp);
614 if (err)
615 return err;
616
617 return finish_preallocate_blocks(inode);
618 }
619
620 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
621 {
622 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
623 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
624 __le32 *addr;
625 bool compressed_cluster = false;
626 int cluster_index = 0, valid_blocks = 0;
627 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
628 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
629
630 addr = get_dnode_addr(dn->inode, dn->node_page) + ofs;
631
632 /* Assumption: truncation starts with cluster */
633 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
634 block_t blkaddr = le32_to_cpu(*addr);
635
636 if (f2fs_compressed_file(dn->inode) &&
637 !(cluster_index & (cluster_size - 1))) {
638 if (compressed_cluster)
639 f2fs_i_compr_blocks_update(dn->inode,
640 valid_blocks, false);
641 compressed_cluster = (blkaddr == COMPRESS_ADDR);
642 valid_blocks = 0;
643 }
644
645 if (blkaddr == NULL_ADDR)
646 continue;
647
648 f2fs_set_data_blkaddr(dn, NULL_ADDR);
649
650 if (__is_valid_data_blkaddr(blkaddr)) {
651 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE))
652 continue;
653 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr,
654 DATA_GENERIC_ENHANCE))
655 continue;
656 if (compressed_cluster)
657 valid_blocks++;
658 }
659
660 f2fs_invalidate_blocks(sbi, blkaddr);
661
662 if (!released || blkaddr != COMPRESS_ADDR)
663 nr_free++;
664 }
665
666 if (compressed_cluster)
667 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
668
669 if (nr_free) {
670 pgoff_t fofs;
671 /*
672 * once we invalidate valid blkaddr in range [ofs, ofs + count],
673 * we will invalidate all blkaddr in the whole range.
674 */
675 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
676 dn->inode) + ofs;
677 f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
678 f2fs_update_age_extent_cache_range(dn, fofs, len);
679 dec_valid_block_count(sbi, dn->inode, nr_free);
680 }
681 dn->ofs_in_node = ofs;
682
683 f2fs_update_time(sbi, REQ_TIME);
684 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
685 dn->ofs_in_node, nr_free);
686 }
687
688 static int truncate_partial_data_page(struct inode *inode, u64 from,
689 bool cache_only)
690 {
691 loff_t offset = from & (PAGE_SIZE - 1);
692 pgoff_t index = from >> PAGE_SHIFT;
693 struct address_space *mapping = inode->i_mapping;
694 struct page *page;
695
696 if (!offset && !cache_only)
697 return 0;
698
699 if (cache_only) {
700 page = find_lock_page(mapping, index);
701 if (page && PageUptodate(page))
702 goto truncate_out;
703 f2fs_put_page(page, 1);
704 return 0;
705 }
706
707 page = f2fs_get_lock_data_page(inode, index, true);
708 if (IS_ERR(page))
709 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
710 truncate_out:
711 f2fs_wait_on_page_writeback(page, DATA, true, true);
712 zero_user(page, offset, PAGE_SIZE - offset);
713
714 /* An encrypted inode should have a key and truncate the last page. */
715 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
716 if (!cache_only)
717 set_page_dirty(page);
718 f2fs_put_page(page, 1);
719 return 0;
720 }
721
722 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
723 {
724 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
725 struct dnode_of_data dn;
726 pgoff_t free_from;
727 int count = 0, err = 0;
728 struct page *ipage;
729 bool truncate_page = false;
730
731 trace_f2fs_truncate_blocks_enter(inode, from);
732
733 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
734
735 if (free_from >= max_file_blocks(inode))
736 goto free_partial;
737
738 if (lock)
739 f2fs_lock_op(sbi);
740
741 ipage = f2fs_get_node_page(sbi, inode->i_ino);
742 if (IS_ERR(ipage)) {
743 err = PTR_ERR(ipage);
744 goto out;
745 }
746
747 if (f2fs_has_inline_data(inode)) {
748 f2fs_truncate_inline_inode(inode, ipage, from);
749 f2fs_put_page(ipage, 1);
750 truncate_page = true;
751 goto out;
752 }
753
754 set_new_dnode(&dn, inode, ipage, NULL, 0);
755 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
756 if (err) {
757 if (err == -ENOENT)
758 goto free_next;
759 goto out;
760 }
761
762 count = ADDRS_PER_PAGE(dn.node_page, inode);
763
764 count -= dn.ofs_in_node;
765 f2fs_bug_on(sbi, count < 0);
766
767 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
768 f2fs_truncate_data_blocks_range(&dn, count);
769 free_from += count;
770 }
771
772 f2fs_put_dnode(&dn);
773 free_next:
774 err = f2fs_truncate_inode_blocks(inode, free_from);
775 out:
776 if (lock)
777 f2fs_unlock_op(sbi);
778 free_partial:
779 /* lastly zero out the first data page */
780 if (!err)
781 err = truncate_partial_data_page(inode, from, truncate_page);
782
783 trace_f2fs_truncate_blocks_exit(inode, err);
784 return err;
785 }
786
787 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
788 {
789 u64 free_from = from;
790 int err;
791
792 #ifdef CONFIG_F2FS_FS_COMPRESSION
793 /*
794 * for compressed file, only support cluster size
795 * aligned truncation.
796 */
797 if (f2fs_compressed_file(inode))
798 free_from = round_up(from,
799 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
800 #endif
801
802 err = f2fs_do_truncate_blocks(inode, free_from, lock);
803 if (err)
804 return err;
805
806 #ifdef CONFIG_F2FS_FS_COMPRESSION
807 /*
808 * For compressed file, after release compress blocks, don't allow write
809 * direct, but we should allow write direct after truncate to zero.
810 */
811 if (f2fs_compressed_file(inode) && !free_from
812 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
813 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
814
815 if (from != free_from) {
816 err = f2fs_truncate_partial_cluster(inode, from, lock);
817 if (err)
818 return err;
819 }
820 #endif
821
822 return 0;
823 }
824
825 int f2fs_truncate(struct inode *inode)
826 {
827 int err;
828
829 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
830 return -EIO;
831
832 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
833 S_ISLNK(inode->i_mode)))
834 return 0;
835
836 trace_f2fs_truncate(inode);
837
838 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
839 return -EIO;
840
841 err = f2fs_dquot_initialize(inode);
842 if (err)
843 return err;
844
845 /* we should check inline_data size */
846 if (!f2fs_may_inline_data(inode)) {
847 err = f2fs_convert_inline_inode(inode);
848 if (err)
849 return err;
850 }
851
852 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
853 if (err)
854 return err;
855
856 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
857 f2fs_mark_inode_dirty_sync(inode, false);
858 return 0;
859 }
860
861 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
862 {
863 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
864
865 if (!fscrypt_dio_supported(inode))
866 return true;
867 if (fsverity_active(inode))
868 return true;
869 if (f2fs_compressed_file(inode))
870 return true;
871 if (f2fs_has_inline_data(inode))
872 return true;
873
874 /* disallow direct IO if any of devices has unaligned blksize */
875 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
876 return true;
877 /*
878 * for blkzoned device, fallback direct IO to buffered IO, so
879 * all IOs can be serialized by log-structured write.
880 */
881 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) &&
882 !f2fs_is_pinned_file(inode))
883 return true;
884 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
885 return true;
886
887 return false;
888 }
889
890 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path,
891 struct kstat *stat, u32 request_mask, unsigned int query_flags)
892 {
893 struct inode *inode = d_inode(path->dentry);
894 struct f2fs_inode_info *fi = F2FS_I(inode);
895 struct f2fs_inode *ri = NULL;
896 unsigned int flags;
897
898 if (f2fs_has_extra_attr(inode) &&
899 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
900 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
901 stat->result_mask |= STATX_BTIME;
902 stat->btime.tv_sec = fi->i_crtime.tv_sec;
903 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
904 }
905
906 /*
907 * Return the DIO alignment restrictions if requested. We only return
908 * this information when requested, since on encrypted files it might
909 * take a fair bit of work to get if the file wasn't opened recently.
910 *
911 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN
912 * cannot represent that, so in that case we report no DIO support.
913 */
914 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
915 unsigned int bsize = i_blocksize(inode);
916
917 stat->result_mask |= STATX_DIOALIGN;
918 if (!f2fs_force_buffered_io(inode, WRITE)) {
919 stat->dio_mem_align = bsize;
920 stat->dio_offset_align = bsize;
921 }
922 }
923
924 flags = fi->i_flags;
925 if (flags & F2FS_COMPR_FL)
926 stat->attributes |= STATX_ATTR_COMPRESSED;
927 if (flags & F2FS_APPEND_FL)
928 stat->attributes |= STATX_ATTR_APPEND;
929 if (IS_ENCRYPTED(inode))
930 stat->attributes |= STATX_ATTR_ENCRYPTED;
931 if (flags & F2FS_IMMUTABLE_FL)
932 stat->attributes |= STATX_ATTR_IMMUTABLE;
933 if (flags & F2FS_NODUMP_FL)
934 stat->attributes |= STATX_ATTR_NODUMP;
935 if (IS_VERITY(inode))
936 stat->attributes |= STATX_ATTR_VERITY;
937
938 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
939 STATX_ATTR_APPEND |
940 STATX_ATTR_ENCRYPTED |
941 STATX_ATTR_IMMUTABLE |
942 STATX_ATTR_NODUMP |
943 STATX_ATTR_VERITY);
944
945 generic_fillattr(idmap, request_mask, inode, stat);
946
947 /* we need to show initial sectors used for inline_data/dentries */
948 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
949 f2fs_has_inline_dentry(inode))
950 stat->blocks += (stat->size + 511) >> 9;
951
952 return 0;
953 }
954
955 #ifdef CONFIG_F2FS_FS_POSIX_ACL
956 static void __setattr_copy(struct mnt_idmap *idmap,
957 struct inode *inode, const struct iattr *attr)
958 {
959 unsigned int ia_valid = attr->ia_valid;
960
961 i_uid_update(idmap, attr, inode);
962 i_gid_update(idmap, attr, inode);
963 if (ia_valid & ATTR_ATIME)
964 inode_set_atime_to_ts(inode, attr->ia_atime);
965 if (ia_valid & ATTR_MTIME)
966 inode_set_mtime_to_ts(inode, attr->ia_mtime);
967 if (ia_valid & ATTR_CTIME)
968 inode_set_ctime_to_ts(inode, attr->ia_ctime);
969 if (ia_valid & ATTR_MODE) {
970 umode_t mode = attr->ia_mode;
971
972 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode)))
973 mode &= ~S_ISGID;
974 set_acl_inode(inode, mode);
975 }
976 }
977 #else
978 #define __setattr_copy setattr_copy
979 #endif
980
981 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
982 struct iattr *attr)
983 {
984 struct inode *inode = d_inode(dentry);
985 struct f2fs_inode_info *fi = F2FS_I(inode);
986 int err;
987
988 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
989 return -EIO;
990
991 if (unlikely(IS_IMMUTABLE(inode)))
992 return -EPERM;
993
994 if (unlikely(IS_APPEND(inode) &&
995 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
996 ATTR_GID | ATTR_TIMES_SET))))
997 return -EPERM;
998
999 if ((attr->ia_valid & ATTR_SIZE)) {
1000 if (!f2fs_is_compress_backend_ready(inode))
1001 return -EOPNOTSUPP;
1002 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) &&
1003 !IS_ALIGNED(attr->ia_size,
1004 F2FS_BLK_TO_BYTES(fi->i_cluster_size)))
1005 return -EINVAL;
1006 }
1007
1008 err = setattr_prepare(idmap, dentry, attr);
1009 if (err)
1010 return err;
1011
1012 err = fscrypt_prepare_setattr(dentry, attr);
1013 if (err)
1014 return err;
1015
1016 err = fsverity_prepare_setattr(dentry, attr);
1017 if (err)
1018 return err;
1019
1020 if (is_quota_modification(idmap, inode, attr)) {
1021 err = f2fs_dquot_initialize(inode);
1022 if (err)
1023 return err;
1024 }
1025 if (i_uid_needs_update(idmap, attr, inode) ||
1026 i_gid_needs_update(idmap, attr, inode)) {
1027 f2fs_lock_op(F2FS_I_SB(inode));
1028 err = dquot_transfer(idmap, inode, attr);
1029 if (err) {
1030 set_sbi_flag(F2FS_I_SB(inode),
1031 SBI_QUOTA_NEED_REPAIR);
1032 f2fs_unlock_op(F2FS_I_SB(inode));
1033 return err;
1034 }
1035 /*
1036 * update uid/gid under lock_op(), so that dquot and inode can
1037 * be updated atomically.
1038 */
1039 i_uid_update(idmap, attr, inode);
1040 i_gid_update(idmap, attr, inode);
1041 f2fs_mark_inode_dirty_sync(inode, true);
1042 f2fs_unlock_op(F2FS_I_SB(inode));
1043 }
1044
1045 if (attr->ia_valid & ATTR_SIZE) {
1046 loff_t old_size = i_size_read(inode);
1047
1048 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
1049 /*
1050 * should convert inline inode before i_size_write to
1051 * keep smaller than inline_data size with inline flag.
1052 */
1053 err = f2fs_convert_inline_inode(inode);
1054 if (err)
1055 return err;
1056 }
1057
1058 /*
1059 * wait for inflight dio, blocks should be removed after
1060 * IO completion.
1061 */
1062 if (attr->ia_size < old_size)
1063 inode_dio_wait(inode);
1064
1065 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
1066 filemap_invalidate_lock(inode->i_mapping);
1067
1068 truncate_setsize(inode, attr->ia_size);
1069
1070 if (attr->ia_size <= old_size)
1071 err = f2fs_truncate(inode);
1072 /*
1073 * do not trim all blocks after i_size if target size is
1074 * larger than i_size.
1075 */
1076 filemap_invalidate_unlock(inode->i_mapping);
1077 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
1078 if (err)
1079 return err;
1080
1081 spin_lock(&fi->i_size_lock);
1082 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1083 fi->last_disk_size = i_size_read(inode);
1084 spin_unlock(&fi->i_size_lock);
1085 }
1086
1087 __setattr_copy(idmap, inode, attr);
1088
1089 if (attr->ia_valid & ATTR_MODE) {
1090 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
1091
1092 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1093 if (!err)
1094 inode->i_mode = fi->i_acl_mode;
1095 clear_inode_flag(inode, FI_ACL_MODE);
1096 }
1097 }
1098
1099 /* file size may changed here */
1100 f2fs_mark_inode_dirty_sync(inode, true);
1101
1102 /* inode change will produce dirty node pages flushed by checkpoint */
1103 f2fs_balance_fs(F2FS_I_SB(inode), true);
1104
1105 return err;
1106 }
1107
1108 const struct inode_operations f2fs_file_inode_operations = {
1109 .getattr = f2fs_getattr,
1110 .setattr = f2fs_setattr,
1111 .get_inode_acl = f2fs_get_acl,
1112 .set_acl = f2fs_set_acl,
1113 .listxattr = f2fs_listxattr,
1114 .fiemap = f2fs_fiemap,
1115 .fileattr_get = f2fs_fileattr_get,
1116 .fileattr_set = f2fs_fileattr_set,
1117 };
1118
1119 static int fill_zero(struct inode *inode, pgoff_t index,
1120 loff_t start, loff_t len)
1121 {
1122 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1123 struct page *page;
1124
1125 if (!len)
1126 return 0;
1127
1128 f2fs_balance_fs(sbi, true);
1129
1130 f2fs_lock_op(sbi);
1131 page = f2fs_get_new_data_page(inode, NULL, index, false);
1132 f2fs_unlock_op(sbi);
1133
1134 if (IS_ERR(page))
1135 return PTR_ERR(page);
1136
1137 f2fs_wait_on_page_writeback(page, DATA, true, true);
1138 zero_user(page, start, len);
1139 set_page_dirty(page);
1140 f2fs_put_page(page, 1);
1141 return 0;
1142 }
1143
1144 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1145 {
1146 int err;
1147
1148 while (pg_start < pg_end) {
1149 struct dnode_of_data dn;
1150 pgoff_t end_offset, count;
1151
1152 set_new_dnode(&dn, inode, NULL, NULL, 0);
1153 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1154 if (err) {
1155 if (err == -ENOENT) {
1156 pg_start = f2fs_get_next_page_offset(&dn,
1157 pg_start);
1158 continue;
1159 }
1160 return err;
1161 }
1162
1163 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1164 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1165
1166 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1167
1168 f2fs_truncate_data_blocks_range(&dn, count);
1169 f2fs_put_dnode(&dn);
1170
1171 pg_start += count;
1172 }
1173 return 0;
1174 }
1175
1176 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
1177 {
1178 pgoff_t pg_start, pg_end;
1179 loff_t off_start, off_end;
1180 int ret;
1181
1182 ret = f2fs_convert_inline_inode(inode);
1183 if (ret)
1184 return ret;
1185
1186 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1187 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1188
1189 off_start = offset & (PAGE_SIZE - 1);
1190 off_end = (offset + len) & (PAGE_SIZE - 1);
1191
1192 if (pg_start == pg_end) {
1193 ret = fill_zero(inode, pg_start, off_start,
1194 off_end - off_start);
1195 if (ret)
1196 return ret;
1197 } else {
1198 if (off_start) {
1199 ret = fill_zero(inode, pg_start++, off_start,
1200 PAGE_SIZE - off_start);
1201 if (ret)
1202 return ret;
1203 }
1204 if (off_end) {
1205 ret = fill_zero(inode, pg_end, 0, off_end);
1206 if (ret)
1207 return ret;
1208 }
1209
1210 if (pg_start < pg_end) {
1211 loff_t blk_start, blk_end;
1212 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1213
1214 f2fs_balance_fs(sbi, true);
1215
1216 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1217 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1218
1219 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1220 filemap_invalidate_lock(inode->i_mapping);
1221
1222 truncate_pagecache_range(inode, blk_start, blk_end - 1);
1223
1224 f2fs_lock_op(sbi);
1225 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1226 f2fs_unlock_op(sbi);
1227
1228 filemap_invalidate_unlock(inode->i_mapping);
1229 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1230 }
1231 }
1232
1233 return ret;
1234 }
1235
1236 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1237 int *do_replace, pgoff_t off, pgoff_t len)
1238 {
1239 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1240 struct dnode_of_data dn;
1241 int ret, done, i;
1242
1243 next_dnode:
1244 set_new_dnode(&dn, inode, NULL, NULL, 0);
1245 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1246 if (ret && ret != -ENOENT) {
1247 return ret;
1248 } else if (ret == -ENOENT) {
1249 if (dn.max_level == 0)
1250 return -ENOENT;
1251 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1252 dn.ofs_in_node, len);
1253 blkaddr += done;
1254 do_replace += done;
1255 goto next;
1256 }
1257
1258 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1259 dn.ofs_in_node, len);
1260 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1261 *blkaddr = f2fs_data_blkaddr(&dn);
1262
1263 if (__is_valid_data_blkaddr(*blkaddr) &&
1264 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1265 DATA_GENERIC_ENHANCE)) {
1266 f2fs_put_dnode(&dn);
1267 return -EFSCORRUPTED;
1268 }
1269
1270 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1271
1272 if (f2fs_lfs_mode(sbi)) {
1273 f2fs_put_dnode(&dn);
1274 return -EOPNOTSUPP;
1275 }
1276
1277 /* do not invalidate this block address */
1278 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1279 *do_replace = 1;
1280 }
1281 }
1282 f2fs_put_dnode(&dn);
1283 next:
1284 len -= done;
1285 off += done;
1286 if (len)
1287 goto next_dnode;
1288 return 0;
1289 }
1290
1291 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1292 int *do_replace, pgoff_t off, int len)
1293 {
1294 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1295 struct dnode_of_data dn;
1296 int ret, i;
1297
1298 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1299 if (*do_replace == 0)
1300 continue;
1301
1302 set_new_dnode(&dn, inode, NULL, NULL, 0);
1303 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1304 if (ret) {
1305 dec_valid_block_count(sbi, inode, 1);
1306 f2fs_invalidate_blocks(sbi, *blkaddr);
1307 } else {
1308 f2fs_update_data_blkaddr(&dn, *blkaddr);
1309 }
1310 f2fs_put_dnode(&dn);
1311 }
1312 return 0;
1313 }
1314
1315 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1316 block_t *blkaddr, int *do_replace,
1317 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1318 {
1319 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1320 pgoff_t i = 0;
1321 int ret;
1322
1323 while (i < len) {
1324 if (blkaddr[i] == NULL_ADDR && !full) {
1325 i++;
1326 continue;
1327 }
1328
1329 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1330 struct dnode_of_data dn;
1331 struct node_info ni;
1332 size_t new_size;
1333 pgoff_t ilen;
1334
1335 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1336 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1337 if (ret)
1338 return ret;
1339
1340 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1341 if (ret) {
1342 f2fs_put_dnode(&dn);
1343 return ret;
1344 }
1345
1346 ilen = min((pgoff_t)
1347 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1348 dn.ofs_in_node, len - i);
1349 do {
1350 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1351 f2fs_truncate_data_blocks_range(&dn, 1);
1352
1353 if (do_replace[i]) {
1354 f2fs_i_blocks_write(src_inode,
1355 1, false, false);
1356 f2fs_i_blocks_write(dst_inode,
1357 1, true, false);
1358 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1359 blkaddr[i], ni.version, true, false);
1360
1361 do_replace[i] = 0;
1362 }
1363 dn.ofs_in_node++;
1364 i++;
1365 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1366 if (dst_inode->i_size < new_size)
1367 f2fs_i_size_write(dst_inode, new_size);
1368 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1369
1370 f2fs_put_dnode(&dn);
1371 } else {
1372 struct page *psrc, *pdst;
1373
1374 psrc = f2fs_get_lock_data_page(src_inode,
1375 src + i, true);
1376 if (IS_ERR(psrc))
1377 return PTR_ERR(psrc);
1378 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1379 true);
1380 if (IS_ERR(pdst)) {
1381 f2fs_put_page(psrc, 1);
1382 return PTR_ERR(pdst);
1383 }
1384
1385 f2fs_wait_on_page_writeback(pdst, DATA, true, true);
1386
1387 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1388 set_page_dirty(pdst);
1389 set_page_private_gcing(pdst);
1390 f2fs_put_page(pdst, 1);
1391 f2fs_put_page(psrc, 1);
1392
1393 ret = f2fs_truncate_hole(src_inode,
1394 src + i, src + i + 1);
1395 if (ret)
1396 return ret;
1397 i++;
1398 }
1399 }
1400 return 0;
1401 }
1402
1403 static int __exchange_data_block(struct inode *src_inode,
1404 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1405 pgoff_t len, bool full)
1406 {
1407 block_t *src_blkaddr;
1408 int *do_replace;
1409 pgoff_t olen;
1410 int ret;
1411
1412 while (len) {
1413 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1414
1415 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1416 array_size(olen, sizeof(block_t)),
1417 GFP_NOFS);
1418 if (!src_blkaddr)
1419 return -ENOMEM;
1420
1421 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1422 array_size(olen, sizeof(int)),
1423 GFP_NOFS);
1424 if (!do_replace) {
1425 kvfree(src_blkaddr);
1426 return -ENOMEM;
1427 }
1428
1429 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1430 do_replace, src, olen);
1431 if (ret)
1432 goto roll_back;
1433
1434 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1435 do_replace, src, dst, olen, full);
1436 if (ret)
1437 goto roll_back;
1438
1439 src += olen;
1440 dst += olen;
1441 len -= olen;
1442
1443 kvfree(src_blkaddr);
1444 kvfree(do_replace);
1445 }
1446 return 0;
1447
1448 roll_back:
1449 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1450 kvfree(src_blkaddr);
1451 kvfree(do_replace);
1452 return ret;
1453 }
1454
1455 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1456 {
1457 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1458 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1459 pgoff_t start = offset >> PAGE_SHIFT;
1460 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1461 int ret;
1462
1463 f2fs_balance_fs(sbi, true);
1464
1465 /* avoid gc operation during block exchange */
1466 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1467 filemap_invalidate_lock(inode->i_mapping);
1468
1469 f2fs_lock_op(sbi);
1470 f2fs_drop_extent_tree(inode);
1471 truncate_pagecache(inode, offset);
1472 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1473 f2fs_unlock_op(sbi);
1474
1475 filemap_invalidate_unlock(inode->i_mapping);
1476 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1477 return ret;
1478 }
1479
1480 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1481 {
1482 loff_t new_size;
1483 int ret;
1484
1485 if (offset + len >= i_size_read(inode))
1486 return -EINVAL;
1487
1488 /* collapse range should be aligned to block size of f2fs. */
1489 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1490 return -EINVAL;
1491
1492 ret = f2fs_convert_inline_inode(inode);
1493 if (ret)
1494 return ret;
1495
1496 /* write out all dirty pages from offset */
1497 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1498 if (ret)
1499 return ret;
1500
1501 ret = f2fs_do_collapse(inode, offset, len);
1502 if (ret)
1503 return ret;
1504
1505 /* write out all moved pages, if possible */
1506 filemap_invalidate_lock(inode->i_mapping);
1507 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1508 truncate_pagecache(inode, offset);
1509
1510 new_size = i_size_read(inode) - len;
1511 ret = f2fs_truncate_blocks(inode, new_size, true);
1512 filemap_invalidate_unlock(inode->i_mapping);
1513 if (!ret)
1514 f2fs_i_size_write(inode, new_size);
1515 return ret;
1516 }
1517
1518 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1519 pgoff_t end)
1520 {
1521 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1522 pgoff_t index = start;
1523 unsigned int ofs_in_node = dn->ofs_in_node;
1524 blkcnt_t count = 0;
1525 int ret;
1526
1527 for (; index < end; index++, dn->ofs_in_node++) {
1528 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1529 count++;
1530 }
1531
1532 dn->ofs_in_node = ofs_in_node;
1533 ret = f2fs_reserve_new_blocks(dn, count);
1534 if (ret)
1535 return ret;
1536
1537 dn->ofs_in_node = ofs_in_node;
1538 for (index = start; index < end; index++, dn->ofs_in_node++) {
1539 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1540 /*
1541 * f2fs_reserve_new_blocks will not guarantee entire block
1542 * allocation.
1543 */
1544 if (dn->data_blkaddr == NULL_ADDR) {
1545 ret = -ENOSPC;
1546 break;
1547 }
1548
1549 if (dn->data_blkaddr == NEW_ADDR)
1550 continue;
1551
1552 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1553 DATA_GENERIC_ENHANCE)) {
1554 ret = -EFSCORRUPTED;
1555 break;
1556 }
1557
1558 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1559 f2fs_set_data_blkaddr(dn, NEW_ADDR);
1560 }
1561
1562 f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
1563 f2fs_update_age_extent_cache_range(dn, start, index - start);
1564
1565 return ret;
1566 }
1567
1568 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1569 int mode)
1570 {
1571 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1572 struct address_space *mapping = inode->i_mapping;
1573 pgoff_t index, pg_start, pg_end;
1574 loff_t new_size = i_size_read(inode);
1575 loff_t off_start, off_end;
1576 int ret = 0;
1577
1578 ret = inode_newsize_ok(inode, (len + offset));
1579 if (ret)
1580 return ret;
1581
1582 ret = f2fs_convert_inline_inode(inode);
1583 if (ret)
1584 return ret;
1585
1586 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1587 if (ret)
1588 return ret;
1589
1590 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1591 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1592
1593 off_start = offset & (PAGE_SIZE - 1);
1594 off_end = (offset + len) & (PAGE_SIZE - 1);
1595
1596 if (pg_start == pg_end) {
1597 ret = fill_zero(inode, pg_start, off_start,
1598 off_end - off_start);
1599 if (ret)
1600 return ret;
1601
1602 new_size = max_t(loff_t, new_size, offset + len);
1603 } else {
1604 if (off_start) {
1605 ret = fill_zero(inode, pg_start++, off_start,
1606 PAGE_SIZE - off_start);
1607 if (ret)
1608 return ret;
1609
1610 new_size = max_t(loff_t, new_size,
1611 (loff_t)pg_start << PAGE_SHIFT);
1612 }
1613
1614 for (index = pg_start; index < pg_end;) {
1615 struct dnode_of_data dn;
1616 unsigned int end_offset;
1617 pgoff_t end;
1618
1619 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1620 filemap_invalidate_lock(mapping);
1621
1622 truncate_pagecache_range(inode,
1623 (loff_t)index << PAGE_SHIFT,
1624 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1625
1626 f2fs_lock_op(sbi);
1627
1628 set_new_dnode(&dn, inode, NULL, NULL, 0);
1629 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1630 if (ret) {
1631 f2fs_unlock_op(sbi);
1632 filemap_invalidate_unlock(mapping);
1633 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1634 goto out;
1635 }
1636
1637 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1638 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1639
1640 ret = f2fs_do_zero_range(&dn, index, end);
1641 f2fs_put_dnode(&dn);
1642
1643 f2fs_unlock_op(sbi);
1644 filemap_invalidate_unlock(mapping);
1645 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1646
1647 f2fs_balance_fs(sbi, dn.node_changed);
1648
1649 if (ret)
1650 goto out;
1651
1652 index = end;
1653 new_size = max_t(loff_t, new_size,
1654 (loff_t)index << PAGE_SHIFT);
1655 }
1656
1657 if (off_end) {
1658 ret = fill_zero(inode, pg_end, 0, off_end);
1659 if (ret)
1660 goto out;
1661
1662 new_size = max_t(loff_t, new_size, offset + len);
1663 }
1664 }
1665
1666 out:
1667 if (new_size > i_size_read(inode)) {
1668 if (mode & FALLOC_FL_KEEP_SIZE)
1669 file_set_keep_isize(inode);
1670 else
1671 f2fs_i_size_write(inode, new_size);
1672 }
1673 return ret;
1674 }
1675
1676 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1677 {
1678 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1679 struct address_space *mapping = inode->i_mapping;
1680 pgoff_t nr, pg_start, pg_end, delta, idx;
1681 loff_t new_size;
1682 int ret = 0;
1683
1684 new_size = i_size_read(inode) + len;
1685 ret = inode_newsize_ok(inode, new_size);
1686 if (ret)
1687 return ret;
1688
1689 if (offset >= i_size_read(inode))
1690 return -EINVAL;
1691
1692 /* insert range should be aligned to block size of f2fs. */
1693 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1694 return -EINVAL;
1695
1696 ret = f2fs_convert_inline_inode(inode);
1697 if (ret)
1698 return ret;
1699
1700 f2fs_balance_fs(sbi, true);
1701
1702 filemap_invalidate_lock(mapping);
1703 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1704 filemap_invalidate_unlock(mapping);
1705 if (ret)
1706 return ret;
1707
1708 /* write out all dirty pages from offset */
1709 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1710 if (ret)
1711 return ret;
1712
1713 pg_start = offset >> PAGE_SHIFT;
1714 pg_end = (offset + len) >> PAGE_SHIFT;
1715 delta = pg_end - pg_start;
1716 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1717
1718 /* avoid gc operation during block exchange */
1719 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1720 filemap_invalidate_lock(mapping);
1721 truncate_pagecache(inode, offset);
1722
1723 while (!ret && idx > pg_start) {
1724 nr = idx - pg_start;
1725 if (nr > delta)
1726 nr = delta;
1727 idx -= nr;
1728
1729 f2fs_lock_op(sbi);
1730 f2fs_drop_extent_tree(inode);
1731
1732 ret = __exchange_data_block(inode, inode, idx,
1733 idx + delta, nr, false);
1734 f2fs_unlock_op(sbi);
1735 }
1736 filemap_invalidate_unlock(mapping);
1737 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1738 if (ret)
1739 return ret;
1740
1741 /* write out all moved pages, if possible */
1742 filemap_invalidate_lock(mapping);
1743 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1744 truncate_pagecache(inode, offset);
1745 filemap_invalidate_unlock(mapping);
1746
1747 if (!ret)
1748 f2fs_i_size_write(inode, new_size);
1749 return ret;
1750 }
1751
1752 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
1753 loff_t len, int mode)
1754 {
1755 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1756 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1757 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1758 .m_may_create = true };
1759 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1760 .init_gc_type = FG_GC,
1761 .should_migrate_blocks = false,
1762 .err_gc_skipped = true,
1763 .nr_free_secs = 0 };
1764 pgoff_t pg_start, pg_end;
1765 loff_t new_size;
1766 loff_t off_end;
1767 block_t expanded = 0;
1768 int err;
1769
1770 err = inode_newsize_ok(inode, (len + offset));
1771 if (err)
1772 return err;
1773
1774 err = f2fs_convert_inline_inode(inode);
1775 if (err)
1776 return err;
1777
1778 f2fs_balance_fs(sbi, true);
1779
1780 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1781 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1782 off_end = (offset + len) & (PAGE_SIZE - 1);
1783
1784 map.m_lblk = pg_start;
1785 map.m_len = pg_end - pg_start;
1786 if (off_end)
1787 map.m_len++;
1788
1789 if (!map.m_len)
1790 return 0;
1791
1792 if (f2fs_is_pinned_file(inode)) {
1793 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1794 block_t sec_len = roundup(map.m_len, sec_blks);
1795
1796 map.m_len = sec_blks;
1797 next_alloc:
1798 if (has_not_enough_free_secs(sbi, 0,
1799 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1800 f2fs_down_write(&sbi->gc_lock);
1801 stat_inc_gc_call_count(sbi, FOREGROUND);
1802 err = f2fs_gc(sbi, &gc_control);
1803 if (err && err != -ENODATA)
1804 goto out_err;
1805 }
1806
1807 f2fs_down_write(&sbi->pin_sem);
1808
1809 err = f2fs_allocate_pinning_section(sbi);
1810 if (err) {
1811 f2fs_up_write(&sbi->pin_sem);
1812 goto out_err;
1813 }
1814
1815 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1816 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
1817 file_dont_truncate(inode);
1818
1819 f2fs_up_write(&sbi->pin_sem);
1820
1821 expanded += map.m_len;
1822 sec_len -= map.m_len;
1823 map.m_lblk += map.m_len;
1824 if (!err && sec_len)
1825 goto next_alloc;
1826
1827 map.m_len = expanded;
1828 } else {
1829 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
1830 expanded = map.m_len;
1831 }
1832 out_err:
1833 if (err) {
1834 pgoff_t last_off;
1835
1836 if (!expanded)
1837 return err;
1838
1839 last_off = pg_start + expanded - 1;
1840
1841 /* update new size to the failed position */
1842 new_size = (last_off == pg_end) ? offset + len :
1843 (loff_t)(last_off + 1) << PAGE_SHIFT;
1844 } else {
1845 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1846 }
1847
1848 if (new_size > i_size_read(inode)) {
1849 if (mode & FALLOC_FL_KEEP_SIZE)
1850 file_set_keep_isize(inode);
1851 else
1852 f2fs_i_size_write(inode, new_size);
1853 }
1854
1855 return err;
1856 }
1857
1858 static long f2fs_fallocate(struct file *file, int mode,
1859 loff_t offset, loff_t len)
1860 {
1861 struct inode *inode = file_inode(file);
1862 long ret = 0;
1863
1864 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1865 return -EIO;
1866 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1867 return -ENOSPC;
1868 if (!f2fs_is_compress_backend_ready(inode))
1869 return -EOPNOTSUPP;
1870
1871 /* f2fs only support ->fallocate for regular file */
1872 if (!S_ISREG(inode->i_mode))
1873 return -EINVAL;
1874
1875 if (IS_ENCRYPTED(inode) &&
1876 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1877 return -EOPNOTSUPP;
1878
1879 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1880 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1881 FALLOC_FL_INSERT_RANGE))
1882 return -EOPNOTSUPP;
1883
1884 inode_lock(inode);
1885
1886 /*
1887 * Pinned file should not support partial truncation since the block
1888 * can be used by applications.
1889 */
1890 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1891 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1892 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) {
1893 ret = -EOPNOTSUPP;
1894 goto out;
1895 }
1896
1897 ret = file_modified(file);
1898 if (ret)
1899 goto out;
1900
1901 /*
1902 * wait for inflight dio, blocks should be removed after IO
1903 * completion.
1904 */
1905 inode_dio_wait(inode);
1906
1907 if (mode & FALLOC_FL_PUNCH_HOLE) {
1908 if (offset >= inode->i_size)
1909 goto out;
1910
1911 ret = f2fs_punch_hole(inode, offset, len);
1912 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1913 ret = f2fs_collapse_range(inode, offset, len);
1914 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1915 ret = f2fs_zero_range(inode, offset, len, mode);
1916 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1917 ret = f2fs_insert_range(inode, offset, len);
1918 } else {
1919 ret = f2fs_expand_inode_data(inode, offset, len, mode);
1920 }
1921
1922 if (!ret) {
1923 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1924 f2fs_mark_inode_dirty_sync(inode, false);
1925 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1926 }
1927
1928 out:
1929 inode_unlock(inode);
1930
1931 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1932 return ret;
1933 }
1934
1935 static int f2fs_release_file(struct inode *inode, struct file *filp)
1936 {
1937 /*
1938 * f2fs_release_file is called at every close calls. So we should
1939 * not drop any inmemory pages by close called by other process.
1940 */
1941 if (!(filp->f_mode & FMODE_WRITE) ||
1942 atomic_read(&inode->i_writecount) != 1)
1943 return 0;
1944
1945 inode_lock(inode);
1946 f2fs_abort_atomic_write(inode, true);
1947 inode_unlock(inode);
1948
1949 return 0;
1950 }
1951
1952 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1953 {
1954 struct inode *inode = file_inode(file);
1955
1956 /*
1957 * If the process doing a transaction is crashed, we should do
1958 * roll-back. Otherwise, other reader/write can see corrupted database
1959 * until all the writers close its file. Since this should be done
1960 * before dropping file lock, it needs to do in ->flush.
1961 */
1962 if (F2FS_I(inode)->atomic_write_task == current &&
1963 (current->flags & PF_EXITING)) {
1964 inode_lock(inode);
1965 f2fs_abort_atomic_write(inode, true);
1966 inode_unlock(inode);
1967 }
1968
1969 return 0;
1970 }
1971
1972 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1973 {
1974 struct f2fs_inode_info *fi = F2FS_I(inode);
1975 u32 masked_flags = fi->i_flags & mask;
1976
1977 /* mask can be shrunk by flags_valid selector */
1978 iflags &= mask;
1979
1980 /* Is it quota file? Do not allow user to mess with it */
1981 if (IS_NOQUOTA(inode))
1982 return -EPERM;
1983
1984 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1985 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1986 return -EOPNOTSUPP;
1987 if (!f2fs_empty_dir(inode))
1988 return -ENOTEMPTY;
1989 }
1990
1991 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1992 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1993 return -EOPNOTSUPP;
1994 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1995 return -EINVAL;
1996 }
1997
1998 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1999 if (masked_flags & F2FS_COMPR_FL) {
2000 if (!f2fs_disable_compressed_file(inode))
2001 return -EINVAL;
2002 } else {
2003 /* try to convert inline_data to support compression */
2004 int err = f2fs_convert_inline_inode(inode);
2005 if (err)
2006 return err;
2007
2008 f2fs_down_write(&fi->i_sem);
2009 if (!f2fs_may_compress(inode) ||
2010 (S_ISREG(inode->i_mode) &&
2011 F2FS_HAS_BLOCKS(inode))) {
2012 f2fs_up_write(&fi->i_sem);
2013 return -EINVAL;
2014 }
2015 err = set_compress_context(inode);
2016 f2fs_up_write(&fi->i_sem);
2017
2018 if (err)
2019 return err;
2020 }
2021 }
2022
2023 fi->i_flags = iflags | (fi->i_flags & ~mask);
2024 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
2025 (fi->i_flags & F2FS_NOCOMP_FL));
2026
2027 if (fi->i_flags & F2FS_PROJINHERIT_FL)
2028 set_inode_flag(inode, FI_PROJ_INHERIT);
2029 else
2030 clear_inode_flag(inode, FI_PROJ_INHERIT);
2031
2032 inode_set_ctime_current(inode);
2033 f2fs_set_inode_flags(inode);
2034 f2fs_mark_inode_dirty_sync(inode, true);
2035 return 0;
2036 }
2037
2038 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
2039
2040 /*
2041 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
2042 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
2043 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
2044 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
2045 *
2046 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
2047 * FS_IOC_FSSETXATTR is done by the VFS.
2048 */
2049
2050 static const struct {
2051 u32 iflag;
2052 u32 fsflag;
2053 } f2fs_fsflags_map[] = {
2054 { F2FS_COMPR_FL, FS_COMPR_FL },
2055 { F2FS_SYNC_FL, FS_SYNC_FL },
2056 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
2057 { F2FS_APPEND_FL, FS_APPEND_FL },
2058 { F2FS_NODUMP_FL, FS_NODUMP_FL },
2059 { F2FS_NOATIME_FL, FS_NOATIME_FL },
2060 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
2061 { F2FS_INDEX_FL, FS_INDEX_FL },
2062 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
2063 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
2064 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
2065 };
2066
2067 #define F2FS_GETTABLE_FS_FL ( \
2068 FS_COMPR_FL | \
2069 FS_SYNC_FL | \
2070 FS_IMMUTABLE_FL | \
2071 FS_APPEND_FL | \
2072 FS_NODUMP_FL | \
2073 FS_NOATIME_FL | \
2074 FS_NOCOMP_FL | \
2075 FS_INDEX_FL | \
2076 FS_DIRSYNC_FL | \
2077 FS_PROJINHERIT_FL | \
2078 FS_ENCRYPT_FL | \
2079 FS_INLINE_DATA_FL | \
2080 FS_NOCOW_FL | \
2081 FS_VERITY_FL | \
2082 FS_CASEFOLD_FL)
2083
2084 #define F2FS_SETTABLE_FS_FL ( \
2085 FS_COMPR_FL | \
2086 FS_SYNC_FL | \
2087 FS_IMMUTABLE_FL | \
2088 FS_APPEND_FL | \
2089 FS_NODUMP_FL | \
2090 FS_NOATIME_FL | \
2091 FS_NOCOMP_FL | \
2092 FS_DIRSYNC_FL | \
2093 FS_PROJINHERIT_FL | \
2094 FS_CASEFOLD_FL)
2095
2096 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2097 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2098 {
2099 u32 fsflags = 0;
2100 int i;
2101
2102 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2103 if (iflags & f2fs_fsflags_map[i].iflag)
2104 fsflags |= f2fs_fsflags_map[i].fsflag;
2105
2106 return fsflags;
2107 }
2108
2109 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2110 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2111 {
2112 u32 iflags = 0;
2113 int i;
2114
2115 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2116 if (fsflags & f2fs_fsflags_map[i].fsflag)
2117 iflags |= f2fs_fsflags_map[i].iflag;
2118
2119 return iflags;
2120 }
2121
2122 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2123 {
2124 struct inode *inode = file_inode(filp);
2125
2126 return put_user(inode->i_generation, (int __user *)arg);
2127 }
2128
2129 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2130 {
2131 struct inode *inode = file_inode(filp);
2132 struct mnt_idmap *idmap = file_mnt_idmap(filp);
2133 struct f2fs_inode_info *fi = F2FS_I(inode);
2134 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2135 loff_t isize;
2136 int ret;
2137
2138 if (!(filp->f_mode & FMODE_WRITE))
2139 return -EBADF;
2140
2141 if (!inode_owner_or_capable(idmap, inode))
2142 return -EACCES;
2143
2144 if (!S_ISREG(inode->i_mode))
2145 return -EINVAL;
2146
2147 if (filp->f_flags & O_DIRECT)
2148 return -EINVAL;
2149
2150 ret = mnt_want_write_file(filp);
2151 if (ret)
2152 return ret;
2153
2154 inode_lock(inode);
2155
2156 if (!f2fs_disable_compressed_file(inode) ||
2157 f2fs_is_pinned_file(inode)) {
2158 ret = -EINVAL;
2159 goto out;
2160 }
2161
2162 if (f2fs_is_atomic_file(inode))
2163 goto out;
2164
2165 ret = f2fs_convert_inline_inode(inode);
2166 if (ret)
2167 goto out;
2168
2169 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2170
2171 /*
2172 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2173 * f2fs_is_atomic_file.
2174 */
2175 if (get_dirty_pages(inode))
2176 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2177 inode->i_ino, get_dirty_pages(inode));
2178 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2179 if (ret) {
2180 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2181 goto out;
2182 }
2183
2184 /* Check if the inode already has a COW inode */
2185 if (fi->cow_inode == NULL) {
2186 /* Create a COW inode for atomic write */
2187 struct dentry *dentry = file_dentry(filp);
2188 struct inode *dir = d_inode(dentry->d_parent);
2189
2190 ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode);
2191 if (ret) {
2192 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2193 goto out;
2194 }
2195
2196 set_inode_flag(fi->cow_inode, FI_COW_FILE);
2197 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2198
2199 /* Set the COW inode's atomic_inode to the atomic inode */
2200 F2FS_I(fi->cow_inode)->atomic_inode = inode;
2201 } else {
2202 /* Reuse the already created COW inode */
2203 f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode));
2204
2205 invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1);
2206
2207 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2208 if (ret) {
2209 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2210 goto out;
2211 }
2212 }
2213
2214 f2fs_write_inode(inode, NULL);
2215
2216 stat_inc_atomic_inode(inode);
2217
2218 set_inode_flag(inode, FI_ATOMIC_FILE);
2219
2220 isize = i_size_read(inode);
2221 fi->original_i_size = isize;
2222 if (truncate) {
2223 set_inode_flag(inode, FI_ATOMIC_REPLACE);
2224 truncate_inode_pages_final(inode->i_mapping);
2225 f2fs_i_size_write(inode, 0);
2226 isize = 0;
2227 }
2228 f2fs_i_size_write(fi->cow_inode, isize);
2229
2230 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2231
2232 f2fs_update_time(sbi, REQ_TIME);
2233 fi->atomic_write_task = current;
2234 stat_update_max_atomic_write(inode);
2235 fi->atomic_write_cnt = 0;
2236 out:
2237 inode_unlock(inode);
2238 mnt_drop_write_file(filp);
2239 return ret;
2240 }
2241
2242 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2243 {
2244 struct inode *inode = file_inode(filp);
2245 struct mnt_idmap *idmap = file_mnt_idmap(filp);
2246 int ret;
2247
2248 if (!(filp->f_mode & FMODE_WRITE))
2249 return -EBADF;
2250
2251 if (!inode_owner_or_capable(idmap, inode))
2252 return -EACCES;
2253
2254 ret = mnt_want_write_file(filp);
2255 if (ret)
2256 return ret;
2257
2258 f2fs_balance_fs(F2FS_I_SB(inode), true);
2259
2260 inode_lock(inode);
2261
2262 if (f2fs_is_atomic_file(inode)) {
2263 ret = f2fs_commit_atomic_write(inode);
2264 if (!ret)
2265 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2266
2267 f2fs_abort_atomic_write(inode, ret);
2268 } else {
2269 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2270 }
2271
2272 inode_unlock(inode);
2273 mnt_drop_write_file(filp);
2274 return ret;
2275 }
2276
2277 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2278 {
2279 struct inode *inode = file_inode(filp);
2280 struct mnt_idmap *idmap = file_mnt_idmap(filp);
2281 int ret;
2282
2283 if (!(filp->f_mode & FMODE_WRITE))
2284 return -EBADF;
2285
2286 if (!inode_owner_or_capable(idmap, inode))
2287 return -EACCES;
2288
2289 ret = mnt_want_write_file(filp);
2290 if (ret)
2291 return ret;
2292
2293 inode_lock(inode);
2294
2295 f2fs_abort_atomic_write(inode, true);
2296
2297 inode_unlock(inode);
2298
2299 mnt_drop_write_file(filp);
2300 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2301 return ret;
2302 }
2303
2304 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
2305 bool readonly, bool need_lock)
2306 {
2307 struct super_block *sb = sbi->sb;
2308 int ret = 0;
2309
2310 switch (flag) {
2311 case F2FS_GOING_DOWN_FULLSYNC:
2312 ret = bdev_freeze(sb->s_bdev);
2313 if (ret)
2314 goto out;
2315 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2316 bdev_thaw(sb->s_bdev);
2317 break;
2318 case F2FS_GOING_DOWN_METASYNC:
2319 /* do checkpoint only */
2320 ret = f2fs_sync_fs(sb, 1);
2321 if (ret) {
2322 if (ret == -EIO)
2323 ret = 0;
2324 goto out;
2325 }
2326 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2327 break;
2328 case F2FS_GOING_DOWN_NOSYNC:
2329 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2330 break;
2331 case F2FS_GOING_DOWN_METAFLUSH:
2332 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2333 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2334 break;
2335 case F2FS_GOING_DOWN_NEED_FSCK:
2336 set_sbi_flag(sbi, SBI_NEED_FSCK);
2337 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2338 set_sbi_flag(sbi, SBI_IS_DIRTY);
2339 /* do checkpoint only */
2340 ret = f2fs_sync_fs(sb, 1);
2341 if (ret == -EIO)
2342 ret = 0;
2343 goto out;
2344 default:
2345 ret = -EINVAL;
2346 goto out;
2347 }
2348
2349 if (readonly)
2350 goto out;
2351
2352 /* grab sb->s_umount to avoid racing w/ remount() */
2353 if (need_lock)
2354 down_read(&sbi->sb->s_umount);
2355
2356 f2fs_stop_gc_thread(sbi);
2357 f2fs_stop_discard_thread(sbi);
2358
2359 f2fs_drop_discard_cmd(sbi);
2360 clear_opt(sbi, DISCARD);
2361
2362 if (need_lock)
2363 up_read(&sbi->sb->s_umount);
2364
2365 f2fs_update_time(sbi, REQ_TIME);
2366 out:
2367
2368 trace_f2fs_shutdown(sbi, flag, ret);
2369
2370 return ret;
2371 }
2372
2373 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2374 {
2375 struct inode *inode = file_inode(filp);
2376 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2377 __u32 in;
2378 int ret;
2379 bool need_drop = false, readonly = false;
2380
2381 if (!capable(CAP_SYS_ADMIN))
2382 return -EPERM;
2383
2384 if (get_user(in, (__u32 __user *)arg))
2385 return -EFAULT;
2386
2387 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2388 ret = mnt_want_write_file(filp);
2389 if (ret) {
2390 if (ret != -EROFS)
2391 return ret;
2392
2393 /* fallback to nosync shutdown for readonly fs */
2394 in = F2FS_GOING_DOWN_NOSYNC;
2395 readonly = true;
2396 } else {
2397 need_drop = true;
2398 }
2399 }
2400
2401 ret = f2fs_do_shutdown(sbi, in, readonly, true);
2402
2403 if (need_drop)
2404 mnt_drop_write_file(filp);
2405
2406 return ret;
2407 }
2408
2409 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2410 {
2411 struct inode *inode = file_inode(filp);
2412 struct super_block *sb = inode->i_sb;
2413 struct fstrim_range range;
2414 int ret;
2415
2416 if (!capable(CAP_SYS_ADMIN))
2417 return -EPERM;
2418
2419 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2420 return -EOPNOTSUPP;
2421
2422 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2423 sizeof(range)))
2424 return -EFAULT;
2425
2426 ret = mnt_want_write_file(filp);
2427 if (ret)
2428 return ret;
2429
2430 range.minlen = max((unsigned int)range.minlen,
2431 bdev_discard_granularity(sb->s_bdev));
2432 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2433 mnt_drop_write_file(filp);
2434 if (ret < 0)
2435 return ret;
2436
2437 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2438 sizeof(range)))
2439 return -EFAULT;
2440 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2441 return 0;
2442 }
2443
2444 static bool uuid_is_nonzero(__u8 u[16])
2445 {
2446 int i;
2447
2448 for (i = 0; i < 16; i++)
2449 if (u[i])
2450 return true;
2451 return false;
2452 }
2453
2454 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2455 {
2456 struct inode *inode = file_inode(filp);
2457 int ret;
2458
2459 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2460 return -EOPNOTSUPP;
2461
2462 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2463 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2464 return ret;
2465 }
2466
2467 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2468 {
2469 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2470 return -EOPNOTSUPP;
2471 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2472 }
2473
2474 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2475 {
2476 struct inode *inode = file_inode(filp);
2477 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2478 u8 encrypt_pw_salt[16];
2479 int err;
2480
2481 if (!f2fs_sb_has_encrypt(sbi))
2482 return -EOPNOTSUPP;
2483
2484 err = mnt_want_write_file(filp);
2485 if (err)
2486 return err;
2487
2488 f2fs_down_write(&sbi->sb_lock);
2489
2490 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2491 goto got_it;
2492
2493 /* update superblock with uuid */
2494 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2495
2496 err = f2fs_commit_super(sbi, false);
2497 if (err) {
2498 /* undo new data */
2499 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2500 goto out_err;
2501 }
2502 got_it:
2503 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2504 out_err:
2505 f2fs_up_write(&sbi->sb_lock);
2506 mnt_drop_write_file(filp);
2507
2508 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2509 err = -EFAULT;
2510
2511 return err;
2512 }
2513
2514 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2515 unsigned long arg)
2516 {
2517 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2518 return -EOPNOTSUPP;
2519
2520 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2521 }
2522
2523 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2524 {
2525 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2526 return -EOPNOTSUPP;
2527
2528 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2529 }
2530
2531 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2532 {
2533 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2534 return -EOPNOTSUPP;
2535
2536 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2537 }
2538
2539 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2540 unsigned long arg)
2541 {
2542 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2543 return -EOPNOTSUPP;
2544
2545 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2546 }
2547
2548 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2549 unsigned long arg)
2550 {
2551 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2552 return -EOPNOTSUPP;
2553
2554 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2555 }
2556
2557 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2558 {
2559 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2560 return -EOPNOTSUPP;
2561
2562 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2563 }
2564
2565 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2566 {
2567 struct inode *inode = file_inode(filp);
2568 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2569 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2570 .no_bg_gc = false,
2571 .should_migrate_blocks = false,
2572 .nr_free_secs = 0 };
2573 __u32 sync;
2574 int ret;
2575
2576 if (!capable(CAP_SYS_ADMIN))
2577 return -EPERM;
2578
2579 if (get_user(sync, (__u32 __user *)arg))
2580 return -EFAULT;
2581
2582 if (f2fs_readonly(sbi->sb))
2583 return -EROFS;
2584
2585 ret = mnt_want_write_file(filp);
2586 if (ret)
2587 return ret;
2588
2589 if (!sync) {
2590 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2591 ret = -EBUSY;
2592 goto out;
2593 }
2594 } else {
2595 f2fs_down_write(&sbi->gc_lock);
2596 }
2597
2598 gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2599 gc_control.err_gc_skipped = sync;
2600 stat_inc_gc_call_count(sbi, FOREGROUND);
2601 ret = f2fs_gc(sbi, &gc_control);
2602 out:
2603 mnt_drop_write_file(filp);
2604 return ret;
2605 }
2606
2607 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2608 {
2609 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2610 struct f2fs_gc_control gc_control = {
2611 .init_gc_type = range->sync ? FG_GC : BG_GC,
2612 .no_bg_gc = false,
2613 .should_migrate_blocks = false,
2614 .err_gc_skipped = range->sync,
2615 .nr_free_secs = 0 };
2616 u64 end;
2617 int ret;
2618
2619 if (!capable(CAP_SYS_ADMIN))
2620 return -EPERM;
2621 if (f2fs_readonly(sbi->sb))
2622 return -EROFS;
2623
2624 end = range->start + range->len;
2625 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2626 end >= MAX_BLKADDR(sbi))
2627 return -EINVAL;
2628
2629 ret = mnt_want_write_file(filp);
2630 if (ret)
2631 return ret;
2632
2633 do_more:
2634 if (!range->sync) {
2635 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2636 ret = -EBUSY;
2637 goto out;
2638 }
2639 } else {
2640 f2fs_down_write(&sbi->gc_lock);
2641 }
2642
2643 gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2644 stat_inc_gc_call_count(sbi, FOREGROUND);
2645 ret = f2fs_gc(sbi, &gc_control);
2646 if (ret) {
2647 if (ret == -EBUSY)
2648 ret = -EAGAIN;
2649 goto out;
2650 }
2651 range->start += CAP_BLKS_PER_SEC(sbi);
2652 if (range->start <= end)
2653 goto do_more;
2654 out:
2655 mnt_drop_write_file(filp);
2656 return ret;
2657 }
2658
2659 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2660 {
2661 struct f2fs_gc_range range;
2662
2663 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2664 sizeof(range)))
2665 return -EFAULT;
2666 return __f2fs_ioc_gc_range(filp, &range);
2667 }
2668
2669 static int f2fs_ioc_write_checkpoint(struct file *filp)
2670 {
2671 struct inode *inode = file_inode(filp);
2672 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2673 int ret;
2674
2675 if (!capable(CAP_SYS_ADMIN))
2676 return -EPERM;
2677
2678 if (f2fs_readonly(sbi->sb))
2679 return -EROFS;
2680
2681 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2682 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2683 return -EINVAL;
2684 }
2685
2686 ret = mnt_want_write_file(filp);
2687 if (ret)
2688 return ret;
2689
2690 ret = f2fs_sync_fs(sbi->sb, 1);
2691
2692 mnt_drop_write_file(filp);
2693 return ret;
2694 }
2695
2696 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2697 struct file *filp,
2698 struct f2fs_defragment *range)
2699 {
2700 struct inode *inode = file_inode(filp);
2701 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2702 .m_seg_type = NO_CHECK_TYPE,
2703 .m_may_create = false };
2704 struct extent_info ei = {};
2705 pgoff_t pg_start, pg_end, next_pgofs;
2706 unsigned int total = 0, sec_num;
2707 block_t blk_end = 0;
2708 bool fragmented = false;
2709 int err;
2710
2711 f2fs_balance_fs(sbi, true);
2712
2713 inode_lock(inode);
2714 pg_start = range->start >> PAGE_SHIFT;
2715 pg_end = min_t(pgoff_t,
2716 (range->start + range->len) >> PAGE_SHIFT,
2717 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));
2718
2719 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) ||
2720 f2fs_is_atomic_file(inode)) {
2721 err = -EINVAL;
2722 goto unlock_out;
2723 }
2724
2725 /* if in-place-update policy is enabled, don't waste time here */
2726 set_inode_flag(inode, FI_OPU_WRITE);
2727 if (f2fs_should_update_inplace(inode, NULL)) {
2728 err = -EINVAL;
2729 goto out;
2730 }
2731
2732 /* writeback all dirty pages in the range */
2733 err = filemap_write_and_wait_range(inode->i_mapping,
2734 pg_start << PAGE_SHIFT,
2735 (pg_end << PAGE_SHIFT) - 1);
2736 if (err)
2737 goto out;
2738
2739 /*
2740 * lookup mapping info in extent cache, skip defragmenting if physical
2741 * block addresses are continuous.
2742 */
2743 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2744 if ((pgoff_t)ei.fofs + ei.len >= pg_end)
2745 goto out;
2746 }
2747
2748 map.m_lblk = pg_start;
2749 map.m_next_pgofs = &next_pgofs;
2750
2751 /*
2752 * lookup mapping info in dnode page cache, skip defragmenting if all
2753 * physical block addresses are continuous even if there are hole(s)
2754 * in logical blocks.
2755 */
2756 while (map.m_lblk < pg_end) {
2757 map.m_len = pg_end - map.m_lblk;
2758 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2759 if (err)
2760 goto out;
2761
2762 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2763 map.m_lblk = next_pgofs;
2764 continue;
2765 }
2766
2767 if (blk_end && blk_end != map.m_pblk)
2768 fragmented = true;
2769
2770 /* record total count of block that we're going to move */
2771 total += map.m_len;
2772
2773 blk_end = map.m_pblk + map.m_len;
2774
2775 map.m_lblk += map.m_len;
2776 }
2777
2778 if (!fragmented) {
2779 total = 0;
2780 goto out;
2781 }
2782
2783 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2784
2785 /*
2786 * make sure there are enough free section for LFS allocation, this can
2787 * avoid defragment running in SSR mode when free section are allocated
2788 * intensively
2789 */
2790 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2791 err = -EAGAIN;
2792 goto out;
2793 }
2794
2795 map.m_lblk = pg_start;
2796 map.m_len = pg_end - pg_start;
2797 total = 0;
2798
2799 while (map.m_lblk < pg_end) {
2800 pgoff_t idx;
2801 int cnt = 0;
2802
2803 do_map:
2804 map.m_len = pg_end - map.m_lblk;
2805 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2806 if (err)
2807 goto clear_out;
2808
2809 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2810 map.m_lblk = next_pgofs;
2811 goto check;
2812 }
2813
2814 set_inode_flag(inode, FI_SKIP_WRITES);
2815
2816 idx = map.m_lblk;
2817 while (idx < map.m_lblk + map.m_len &&
2818 cnt < BLKS_PER_SEG(sbi)) {
2819 struct page *page;
2820
2821 page = f2fs_get_lock_data_page(inode, idx, true);
2822 if (IS_ERR(page)) {
2823 err = PTR_ERR(page);
2824 goto clear_out;
2825 }
2826
2827 f2fs_wait_on_page_writeback(page, DATA, true, true);
2828
2829 set_page_dirty(page);
2830 set_page_private_gcing(page);
2831 f2fs_put_page(page, 1);
2832
2833 idx++;
2834 cnt++;
2835 total++;
2836 }
2837
2838 map.m_lblk = idx;
2839 check:
2840 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi))
2841 goto do_map;
2842
2843 clear_inode_flag(inode, FI_SKIP_WRITES);
2844
2845 err = filemap_fdatawrite(inode->i_mapping);
2846 if (err)
2847 goto out;
2848 }
2849 clear_out:
2850 clear_inode_flag(inode, FI_SKIP_WRITES);
2851 out:
2852 clear_inode_flag(inode, FI_OPU_WRITE);
2853 unlock_out:
2854 inode_unlock(inode);
2855 if (!err)
2856 range->len = (u64)total << PAGE_SHIFT;
2857 return err;
2858 }
2859
2860 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2861 {
2862 struct inode *inode = file_inode(filp);
2863 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2864 struct f2fs_defragment range;
2865 int err;
2866
2867 if (!capable(CAP_SYS_ADMIN))
2868 return -EPERM;
2869
2870 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2871 return -EINVAL;
2872
2873 if (f2fs_readonly(sbi->sb))
2874 return -EROFS;
2875
2876 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2877 sizeof(range)))
2878 return -EFAULT;
2879
2880 /* verify alignment of offset & size */
2881 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2882 return -EINVAL;
2883
2884 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2885 max_file_blocks(inode)))
2886 return -EINVAL;
2887
2888 err = mnt_want_write_file(filp);
2889 if (err)
2890 return err;
2891
2892 err = f2fs_defragment_range(sbi, filp, &range);
2893 mnt_drop_write_file(filp);
2894
2895 if (range.len)
2896 f2fs_update_time(sbi, REQ_TIME);
2897 if (err < 0)
2898 return err;
2899
2900 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2901 sizeof(range)))
2902 return -EFAULT;
2903
2904 return 0;
2905 }
2906
2907 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2908 struct file *file_out, loff_t pos_out, size_t len)
2909 {
2910 struct inode *src = file_inode(file_in);
2911 struct inode *dst = file_inode(file_out);
2912 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2913 size_t olen = len, dst_max_i_size = 0;
2914 size_t dst_osize;
2915 int ret;
2916
2917 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2918 src->i_sb != dst->i_sb)
2919 return -EXDEV;
2920
2921 if (unlikely(f2fs_readonly(src->i_sb)))
2922 return -EROFS;
2923
2924 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2925 return -EINVAL;
2926
2927 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2928 return -EOPNOTSUPP;
2929
2930 if (pos_out < 0 || pos_in < 0)
2931 return -EINVAL;
2932
2933 if (src == dst) {
2934 if (pos_in == pos_out)
2935 return 0;
2936 if (pos_out > pos_in && pos_out < pos_in + len)
2937 return -EINVAL;
2938 }
2939
2940 inode_lock(src);
2941 if (src != dst) {
2942 ret = -EBUSY;
2943 if (!inode_trylock(dst))
2944 goto out;
2945 }
2946
2947 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) ||
2948 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) {
2949 ret = -EOPNOTSUPP;
2950 goto out_unlock;
2951 }
2952
2953 if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) {
2954 ret = -EINVAL;
2955 goto out_unlock;
2956 }
2957
2958 ret = -EINVAL;
2959 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2960 goto out_unlock;
2961 if (len == 0)
2962 olen = len = src->i_size - pos_in;
2963 if (pos_in + len == src->i_size)
2964 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2965 if (len == 0) {
2966 ret = 0;
2967 goto out_unlock;
2968 }
2969
2970 dst_osize = dst->i_size;
2971 if (pos_out + olen > dst->i_size)
2972 dst_max_i_size = pos_out + olen;
2973
2974 /* verify the end result is block aligned */
2975 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2976 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2977 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2978 goto out_unlock;
2979
2980 ret = f2fs_convert_inline_inode(src);
2981 if (ret)
2982 goto out_unlock;
2983
2984 ret = f2fs_convert_inline_inode(dst);
2985 if (ret)
2986 goto out_unlock;
2987
2988 /* write out all dirty pages from offset */
2989 ret = filemap_write_and_wait_range(src->i_mapping,
2990 pos_in, pos_in + len);
2991 if (ret)
2992 goto out_unlock;
2993
2994 ret = filemap_write_and_wait_range(dst->i_mapping,
2995 pos_out, pos_out + len);
2996 if (ret)
2997 goto out_unlock;
2998
2999 f2fs_balance_fs(sbi, true);
3000
3001 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3002 if (src != dst) {
3003 ret = -EBUSY;
3004 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
3005 goto out_src;
3006 }
3007
3008 f2fs_lock_op(sbi);
3009 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
3010 pos_out >> F2FS_BLKSIZE_BITS,
3011 len >> F2FS_BLKSIZE_BITS, false);
3012
3013 if (!ret) {
3014 if (dst_max_i_size)
3015 f2fs_i_size_write(dst, dst_max_i_size);
3016 else if (dst_osize != dst->i_size)
3017 f2fs_i_size_write(dst, dst_osize);
3018 }
3019 f2fs_unlock_op(sbi);
3020
3021 if (src != dst)
3022 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
3023 out_src:
3024 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3025 if (ret)
3026 goto out_unlock;
3027
3028 inode_set_mtime_to_ts(src, inode_set_ctime_current(src));
3029 f2fs_mark_inode_dirty_sync(src, false);
3030 if (src != dst) {
3031 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst));
3032 f2fs_mark_inode_dirty_sync(dst, false);
3033 }
3034 f2fs_update_time(sbi, REQ_TIME);
3035
3036 out_unlock:
3037 if (src != dst)
3038 inode_unlock(dst);
3039 out:
3040 inode_unlock(src);
3041 return ret;
3042 }
3043
3044 static int __f2fs_ioc_move_range(struct file *filp,
3045 struct f2fs_move_range *range)
3046 {
3047 struct fd dst;
3048 int err;
3049
3050 if (!(filp->f_mode & FMODE_READ) ||
3051 !(filp->f_mode & FMODE_WRITE))
3052 return -EBADF;
3053
3054 dst = fdget(range->dst_fd);
3055 if (!dst.file)
3056 return -EBADF;
3057
3058 if (!(dst.file->f_mode & FMODE_WRITE)) {
3059 err = -EBADF;
3060 goto err_out;
3061 }
3062
3063 err = mnt_want_write_file(filp);
3064 if (err)
3065 goto err_out;
3066
3067 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
3068 range->pos_out, range->len);
3069
3070 mnt_drop_write_file(filp);
3071 err_out:
3072 fdput(dst);
3073 return err;
3074 }
3075
3076 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
3077 {
3078 struct f2fs_move_range range;
3079
3080 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
3081 sizeof(range)))
3082 return -EFAULT;
3083 return __f2fs_ioc_move_range(filp, &range);
3084 }
3085
3086 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
3087 {
3088 struct inode *inode = file_inode(filp);
3089 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3090 struct sit_info *sm = SIT_I(sbi);
3091 unsigned int start_segno = 0, end_segno = 0;
3092 unsigned int dev_start_segno = 0, dev_end_segno = 0;
3093 struct f2fs_flush_device range;
3094 struct f2fs_gc_control gc_control = {
3095 .init_gc_type = FG_GC,
3096 .should_migrate_blocks = true,
3097 .err_gc_skipped = true,
3098 .nr_free_secs = 0 };
3099 int ret;
3100
3101 if (!capable(CAP_SYS_ADMIN))
3102 return -EPERM;
3103
3104 if (f2fs_readonly(sbi->sb))
3105 return -EROFS;
3106
3107 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3108 return -EINVAL;
3109
3110 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
3111 sizeof(range)))
3112 return -EFAULT;
3113
3114 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
3115 __is_large_section(sbi)) {
3116 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1",
3117 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi));
3118 return -EINVAL;
3119 }
3120
3121 ret = mnt_want_write_file(filp);
3122 if (ret)
3123 return ret;
3124
3125 if (range.dev_num != 0)
3126 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
3127 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
3128
3129 start_segno = sm->last_victim[FLUSH_DEVICE];
3130 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
3131 start_segno = dev_start_segno;
3132 end_segno = min(start_segno + range.segments, dev_end_segno);
3133
3134 while (start_segno < end_segno) {
3135 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
3136 ret = -EBUSY;
3137 goto out;
3138 }
3139 sm->last_victim[GC_CB] = end_segno + 1;
3140 sm->last_victim[GC_GREEDY] = end_segno + 1;
3141 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
3142
3143 gc_control.victim_segno = start_segno;
3144 stat_inc_gc_call_count(sbi, FOREGROUND);
3145 ret = f2fs_gc(sbi, &gc_control);
3146 if (ret == -EAGAIN)
3147 ret = 0;
3148 else if (ret < 0)
3149 break;
3150 start_segno++;
3151 }
3152 out:
3153 mnt_drop_write_file(filp);
3154 return ret;
3155 }
3156
3157 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3158 {
3159 struct inode *inode = file_inode(filp);
3160 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3161
3162 /* Must validate to set it with SQLite behavior in Android. */
3163 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3164
3165 return put_user(sb_feature, (u32 __user *)arg);
3166 }
3167
3168 #ifdef CONFIG_QUOTA
3169 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3170 {
3171 struct dquot *transfer_to[MAXQUOTAS] = {};
3172 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3173 struct super_block *sb = sbi->sb;
3174 int err;
3175
3176 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3177 if (IS_ERR(transfer_to[PRJQUOTA]))
3178 return PTR_ERR(transfer_to[PRJQUOTA]);
3179
3180 err = __dquot_transfer(inode, transfer_to);
3181 if (err)
3182 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3183 dqput(transfer_to[PRJQUOTA]);
3184 return err;
3185 }
3186
3187 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3188 {
3189 struct f2fs_inode_info *fi = F2FS_I(inode);
3190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3191 struct f2fs_inode *ri = NULL;
3192 kprojid_t kprojid;
3193 int err;
3194
3195 if (!f2fs_sb_has_project_quota(sbi)) {
3196 if (projid != F2FS_DEF_PROJID)
3197 return -EOPNOTSUPP;
3198 else
3199 return 0;
3200 }
3201
3202 if (!f2fs_has_extra_attr(inode))
3203 return -EOPNOTSUPP;
3204
3205 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3206
3207 if (projid_eq(kprojid, fi->i_projid))
3208 return 0;
3209
3210 err = -EPERM;
3211 /* Is it quota file? Do not allow user to mess with it */
3212 if (IS_NOQUOTA(inode))
3213 return err;
3214
3215 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3216 return -EOVERFLOW;
3217
3218 err = f2fs_dquot_initialize(inode);
3219 if (err)
3220 return err;
3221
3222 f2fs_lock_op(sbi);
3223 err = f2fs_transfer_project_quota(inode, kprojid);
3224 if (err)
3225 goto out_unlock;
3226
3227 fi->i_projid = kprojid;
3228 inode_set_ctime_current(inode);
3229 f2fs_mark_inode_dirty_sync(inode, true);
3230 out_unlock:
3231 f2fs_unlock_op(sbi);
3232 return err;
3233 }
3234 #else
3235 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3236 {
3237 return 0;
3238 }
3239
3240 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3241 {
3242 if (projid != F2FS_DEF_PROJID)
3243 return -EOPNOTSUPP;
3244 return 0;
3245 }
3246 #endif
3247
3248 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3249 {
3250 struct inode *inode = d_inode(dentry);
3251 struct f2fs_inode_info *fi = F2FS_I(inode);
3252 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3253
3254 if (IS_ENCRYPTED(inode))
3255 fsflags |= FS_ENCRYPT_FL;
3256 if (IS_VERITY(inode))
3257 fsflags |= FS_VERITY_FL;
3258 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3259 fsflags |= FS_INLINE_DATA_FL;
3260 if (is_inode_flag_set(inode, FI_PIN_FILE))
3261 fsflags |= FS_NOCOW_FL;
3262
3263 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3264
3265 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3266 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3267
3268 return 0;
3269 }
3270
3271 int f2fs_fileattr_set(struct mnt_idmap *idmap,
3272 struct dentry *dentry, struct fileattr *fa)
3273 {
3274 struct inode *inode = d_inode(dentry);
3275 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3276 u32 iflags;
3277 int err;
3278
3279 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3280 return -EIO;
3281 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3282 return -ENOSPC;
3283 if (fsflags & ~F2FS_GETTABLE_FS_FL)
3284 return -EOPNOTSUPP;
3285 fsflags &= F2FS_SETTABLE_FS_FL;
3286 if (!fa->flags_valid)
3287 mask &= FS_COMMON_FL;
3288
3289 iflags = f2fs_fsflags_to_iflags(fsflags);
3290 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3291 return -EOPNOTSUPP;
3292
3293 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3294 if (!err)
3295 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3296
3297 return err;
3298 }
3299
3300 int f2fs_pin_file_control(struct inode *inode, bool inc)
3301 {
3302 struct f2fs_inode_info *fi = F2FS_I(inode);
3303 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3304
3305 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) {
3306 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3307 __func__, inode->i_ino, fi->i_gc_failures);
3308 clear_inode_flag(inode, FI_PIN_FILE);
3309 return -EAGAIN;
3310 }
3311
3312 /* Use i_gc_failures for normal file as a risk signal. */
3313 if (inc)
3314 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
3315
3316 return 0;
3317 }
3318
3319 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3320 {
3321 struct inode *inode = file_inode(filp);
3322 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3323 __u32 pin;
3324 int ret = 0;
3325
3326 if (get_user(pin, (__u32 __user *)arg))
3327 return -EFAULT;
3328
3329 if (!S_ISREG(inode->i_mode))
3330 return -EINVAL;
3331
3332 if (f2fs_readonly(sbi->sb))
3333 return -EROFS;
3334
3335 ret = mnt_want_write_file(filp);
3336 if (ret)
3337 return ret;
3338
3339 inode_lock(inode);
3340
3341 if (f2fs_is_atomic_file(inode)) {
3342 ret = -EINVAL;
3343 goto out;
3344 }
3345
3346 if (!pin) {
3347 clear_inode_flag(inode, FI_PIN_FILE);
3348 f2fs_i_gc_failures_write(inode, 0);
3349 goto done;
3350 } else if (f2fs_is_pinned_file(inode)) {
3351 goto done;
3352 }
3353
3354 if (F2FS_HAS_BLOCKS(inode)) {
3355 ret = -EFBIG;
3356 goto out;
3357 }
3358
3359 /* Let's allow file pinning on zoned device. */
3360 if (!f2fs_sb_has_blkzoned(sbi) &&
3361 f2fs_should_update_outplace(inode, NULL)) {
3362 ret = -EINVAL;
3363 goto out;
3364 }
3365
3366 if (f2fs_pin_file_control(inode, false)) {
3367 ret = -EAGAIN;
3368 goto out;
3369 }
3370
3371 ret = f2fs_convert_inline_inode(inode);
3372 if (ret)
3373 goto out;
3374
3375 if (!f2fs_disable_compressed_file(inode)) {
3376 ret = -EOPNOTSUPP;
3377 goto out;
3378 }
3379
3380 set_inode_flag(inode, FI_PIN_FILE);
3381 ret = F2FS_I(inode)->i_gc_failures;
3382 done:
3383 f2fs_update_time(sbi, REQ_TIME);
3384 out:
3385 inode_unlock(inode);
3386 mnt_drop_write_file(filp);
3387 return ret;
3388 }
3389
3390 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3391 {
3392 struct inode *inode = file_inode(filp);
3393 __u32 pin = 0;
3394
3395 if (is_inode_flag_set(inode, FI_PIN_FILE))
3396 pin = F2FS_I(inode)->i_gc_failures;
3397 return put_user(pin, (u32 __user *)arg);
3398 }
3399
3400 int f2fs_precache_extents(struct inode *inode)
3401 {
3402 struct f2fs_inode_info *fi = F2FS_I(inode);
3403 struct f2fs_map_blocks map;
3404 pgoff_t m_next_extent;
3405 loff_t end;
3406 int err;
3407
3408 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3409 return -EOPNOTSUPP;
3410
3411 map.m_lblk = 0;
3412 map.m_pblk = 0;
3413 map.m_next_pgofs = NULL;
3414 map.m_next_extent = &m_next_extent;
3415 map.m_seg_type = NO_CHECK_TYPE;
3416 map.m_may_create = false;
3417 end = F2FS_BLK_ALIGN(i_size_read(inode));
3418
3419 while (map.m_lblk < end) {
3420 map.m_len = end - map.m_lblk;
3421
3422 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3423 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
3424 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3425 if (err || !map.m_len)
3426 return err;
3427
3428 map.m_lblk = m_next_extent;
3429 }
3430
3431 return 0;
3432 }
3433
3434 static int f2fs_ioc_precache_extents(struct file *filp)
3435 {
3436 return f2fs_precache_extents(file_inode(filp));
3437 }
3438
3439 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3440 {
3441 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3442 __u64 block_count;
3443
3444 if (!capable(CAP_SYS_ADMIN))
3445 return -EPERM;
3446
3447 if (f2fs_readonly(sbi->sb))
3448 return -EROFS;
3449
3450 if (copy_from_user(&block_count, (void __user *)arg,
3451 sizeof(block_count)))
3452 return -EFAULT;
3453
3454 return f2fs_resize_fs(filp, block_count);
3455 }
3456
3457 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3458 {
3459 struct inode *inode = file_inode(filp);
3460
3461 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3462
3463 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3464 f2fs_warn(F2FS_I_SB(inode),
3465 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3466 inode->i_ino);
3467 return -EOPNOTSUPP;
3468 }
3469
3470 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3471 }
3472
3473 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3474 {
3475 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3476 return -EOPNOTSUPP;
3477
3478 return fsverity_ioctl_measure(filp, (void __user *)arg);
3479 }
3480
3481 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3482 {
3483 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3484 return -EOPNOTSUPP;
3485
3486 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3487 }
3488
3489 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3490 {
3491 struct inode *inode = file_inode(filp);
3492 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3493 char *vbuf;
3494 int count;
3495 int err = 0;
3496
3497 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3498 if (!vbuf)
3499 return -ENOMEM;
3500
3501 f2fs_down_read(&sbi->sb_lock);
3502 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3503 ARRAY_SIZE(sbi->raw_super->volume_name),
3504 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3505 f2fs_up_read(&sbi->sb_lock);
3506
3507 if (copy_to_user((char __user *)arg, vbuf,
3508 min(FSLABEL_MAX, count)))
3509 err = -EFAULT;
3510
3511 kfree(vbuf);
3512 return err;
3513 }
3514
3515 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3516 {
3517 struct inode *inode = file_inode(filp);
3518 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3519 char *vbuf;
3520 int err = 0;
3521
3522 if (!capable(CAP_SYS_ADMIN))
3523 return -EPERM;
3524
3525 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3526 if (IS_ERR(vbuf))
3527 return PTR_ERR(vbuf);
3528
3529 err = mnt_want_write_file(filp);
3530 if (err)
3531 goto out;
3532
3533 f2fs_down_write(&sbi->sb_lock);
3534
3535 memset(sbi->raw_super->volume_name, 0,
3536 sizeof(sbi->raw_super->volume_name));
3537 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3538 sbi->raw_super->volume_name,
3539 ARRAY_SIZE(sbi->raw_super->volume_name));
3540
3541 err = f2fs_commit_super(sbi, false);
3542
3543 f2fs_up_write(&sbi->sb_lock);
3544
3545 mnt_drop_write_file(filp);
3546 out:
3547 kfree(vbuf);
3548 return err;
3549 }
3550
3551 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks)
3552 {
3553 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3554 return -EOPNOTSUPP;
3555
3556 if (!f2fs_compressed_file(inode))
3557 return -EINVAL;
3558
3559 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3560
3561 return 0;
3562 }
3563
3564 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg)
3565 {
3566 struct inode *inode = file_inode(filp);
3567 __u64 blocks;
3568 int ret;
3569
3570 ret = f2fs_get_compress_blocks(inode, &blocks);
3571 if (ret < 0)
3572 return ret;
3573
3574 return put_user(blocks, (u64 __user *)arg);
3575 }
3576
3577 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3578 {
3579 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3580 unsigned int released_blocks = 0;
3581 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3582 block_t blkaddr;
3583 int i;
3584
3585 for (i = 0; i < count; i++) {
3586 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3587 dn->ofs_in_node + i);
3588
3589 if (!__is_valid_data_blkaddr(blkaddr))
3590 continue;
3591 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3592 DATA_GENERIC_ENHANCE)))
3593 return -EFSCORRUPTED;
3594 }
3595
3596 while (count) {
3597 int compr_blocks = 0;
3598
3599 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3600 blkaddr = f2fs_data_blkaddr(dn);
3601
3602 if (i == 0) {
3603 if (blkaddr == COMPRESS_ADDR)
3604 continue;
3605 dn->ofs_in_node += cluster_size;
3606 goto next;
3607 }
3608
3609 if (__is_valid_data_blkaddr(blkaddr))
3610 compr_blocks++;
3611
3612 if (blkaddr != NEW_ADDR)
3613 continue;
3614
3615 f2fs_set_data_blkaddr(dn, NULL_ADDR);
3616 }
3617
3618 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3619 dec_valid_block_count(sbi, dn->inode,
3620 cluster_size - compr_blocks);
3621
3622 released_blocks += cluster_size - compr_blocks;
3623 next:
3624 count -= cluster_size;
3625 }
3626
3627 return released_blocks;
3628 }
3629
3630 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3631 {
3632 struct inode *inode = file_inode(filp);
3633 struct f2fs_inode_info *fi = F2FS_I(inode);
3634 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3635 pgoff_t page_idx = 0, last_idx;
3636 unsigned int released_blocks = 0;
3637 int ret;
3638 int writecount;
3639
3640 if (!f2fs_sb_has_compression(sbi))
3641 return -EOPNOTSUPP;
3642
3643 if (f2fs_readonly(sbi->sb))
3644 return -EROFS;
3645
3646 ret = mnt_want_write_file(filp);
3647 if (ret)
3648 return ret;
3649
3650 f2fs_balance_fs(sbi, true);
3651
3652 inode_lock(inode);
3653
3654 writecount = atomic_read(&inode->i_writecount);
3655 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3656 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3657 ret = -EBUSY;
3658 goto out;
3659 }
3660
3661 if (!f2fs_compressed_file(inode) ||
3662 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3663 ret = -EINVAL;
3664 goto out;
3665 }
3666
3667 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3668 if (ret)
3669 goto out;
3670
3671 if (!atomic_read(&fi->i_compr_blocks)) {
3672 ret = -EPERM;
3673 goto out;
3674 }
3675
3676 set_inode_flag(inode, FI_COMPRESS_RELEASED);
3677 inode_set_ctime_current(inode);
3678 f2fs_mark_inode_dirty_sync(inode, true);
3679
3680 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3681 filemap_invalidate_lock(inode->i_mapping);
3682
3683 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3684
3685 while (page_idx < last_idx) {
3686 struct dnode_of_data dn;
3687 pgoff_t end_offset, count;
3688
3689 f2fs_lock_op(sbi);
3690
3691 set_new_dnode(&dn, inode, NULL, NULL, 0);
3692 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3693 if (ret) {
3694 f2fs_unlock_op(sbi);
3695 if (ret == -ENOENT) {
3696 page_idx = f2fs_get_next_page_offset(&dn,
3697 page_idx);
3698 ret = 0;
3699 continue;
3700 }
3701 break;
3702 }
3703
3704 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3705 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3706 count = round_up(count, fi->i_cluster_size);
3707
3708 ret = release_compress_blocks(&dn, count);
3709
3710 f2fs_put_dnode(&dn);
3711
3712 f2fs_unlock_op(sbi);
3713
3714 if (ret < 0)
3715 break;
3716
3717 page_idx += count;
3718 released_blocks += ret;
3719 }
3720
3721 filemap_invalidate_unlock(inode->i_mapping);
3722 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3723 out:
3724 if (released_blocks)
3725 f2fs_update_time(sbi, REQ_TIME);
3726 inode_unlock(inode);
3727
3728 mnt_drop_write_file(filp);
3729
3730 if (ret >= 0) {
3731 ret = put_user(released_blocks, (u64 __user *)arg);
3732 } else if (released_blocks &&
3733 atomic_read(&fi->i_compr_blocks)) {
3734 set_sbi_flag(sbi, SBI_NEED_FSCK);
3735 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3736 "iblocks=%llu, released=%u, compr_blocks=%u, "
3737 "run fsck to fix.",
3738 __func__, inode->i_ino, inode->i_blocks,
3739 released_blocks,
3740 atomic_read(&fi->i_compr_blocks));
3741 }
3742
3743 return ret;
3744 }
3745
3746 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count,
3747 unsigned int *reserved_blocks)
3748 {
3749 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3750 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3751 block_t blkaddr;
3752 int i;
3753
3754 for (i = 0; i < count; i++) {
3755 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3756 dn->ofs_in_node + i);
3757
3758 if (!__is_valid_data_blkaddr(blkaddr))
3759 continue;
3760 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3761 DATA_GENERIC_ENHANCE)))
3762 return -EFSCORRUPTED;
3763 }
3764
3765 while (count) {
3766 int compr_blocks = 0;
3767 blkcnt_t reserved = 0;
3768 blkcnt_t to_reserved;
3769 int ret;
3770
3771 for (i = 0; i < cluster_size; i++) {
3772 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3773 dn->ofs_in_node + i);
3774
3775 if (i == 0) {
3776 if (blkaddr != COMPRESS_ADDR) {
3777 dn->ofs_in_node += cluster_size;
3778 goto next;
3779 }
3780 continue;
3781 }
3782
3783 /*
3784 * compressed cluster was not released due to it
3785 * fails in release_compress_blocks(), so NEW_ADDR
3786 * is a possible case.
3787 */
3788 if (blkaddr == NEW_ADDR) {
3789 reserved++;
3790 continue;
3791 }
3792 if (__is_valid_data_blkaddr(blkaddr)) {
3793 compr_blocks++;
3794 continue;
3795 }
3796 }
3797
3798 to_reserved = cluster_size - compr_blocks - reserved;
3799
3800 /* for the case all blocks in cluster were reserved */
3801 if (to_reserved == 1) {
3802 dn->ofs_in_node += cluster_size;
3803 goto next;
3804 }
3805
3806 ret = inc_valid_block_count(sbi, dn->inode,
3807 &to_reserved, false);
3808 if (unlikely(ret))
3809 return ret;
3810
3811 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3812 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
3813 f2fs_set_data_blkaddr(dn, NEW_ADDR);
3814 }
3815
3816 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3817
3818 *reserved_blocks += to_reserved;
3819 next:
3820 count -= cluster_size;
3821 }
3822
3823 return 0;
3824 }
3825
3826 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3827 {
3828 struct inode *inode = file_inode(filp);
3829 struct f2fs_inode_info *fi = F2FS_I(inode);
3830 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3831 pgoff_t page_idx = 0, last_idx;
3832 unsigned int reserved_blocks = 0;
3833 int ret;
3834
3835 if (!f2fs_sb_has_compression(sbi))
3836 return -EOPNOTSUPP;
3837
3838 if (f2fs_readonly(sbi->sb))
3839 return -EROFS;
3840
3841 ret = mnt_want_write_file(filp);
3842 if (ret)
3843 return ret;
3844
3845 f2fs_balance_fs(sbi, true);
3846
3847 inode_lock(inode);
3848
3849 if (!f2fs_compressed_file(inode) ||
3850 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3851 ret = -EINVAL;
3852 goto unlock_inode;
3853 }
3854
3855 if (atomic_read(&fi->i_compr_blocks))
3856 goto unlock_inode;
3857
3858 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3859 filemap_invalidate_lock(inode->i_mapping);
3860
3861 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3862
3863 while (page_idx < last_idx) {
3864 struct dnode_of_data dn;
3865 pgoff_t end_offset, count;
3866
3867 f2fs_lock_op(sbi);
3868
3869 set_new_dnode(&dn, inode, NULL, NULL, 0);
3870 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3871 if (ret) {
3872 f2fs_unlock_op(sbi);
3873 if (ret == -ENOENT) {
3874 page_idx = f2fs_get_next_page_offset(&dn,
3875 page_idx);
3876 ret = 0;
3877 continue;
3878 }
3879 break;
3880 }
3881
3882 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3883 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3884 count = round_up(count, fi->i_cluster_size);
3885
3886 ret = reserve_compress_blocks(&dn, count, &reserved_blocks);
3887
3888 f2fs_put_dnode(&dn);
3889
3890 f2fs_unlock_op(sbi);
3891
3892 if (ret < 0)
3893 break;
3894
3895 page_idx += count;
3896 }
3897
3898 filemap_invalidate_unlock(inode->i_mapping);
3899 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3900
3901 if (!ret) {
3902 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3903 inode_set_ctime_current(inode);
3904 f2fs_mark_inode_dirty_sync(inode, true);
3905 }
3906 unlock_inode:
3907 if (reserved_blocks)
3908 f2fs_update_time(sbi, REQ_TIME);
3909 inode_unlock(inode);
3910 mnt_drop_write_file(filp);
3911
3912 if (!ret) {
3913 ret = put_user(reserved_blocks, (u64 __user *)arg);
3914 } else if (reserved_blocks &&
3915 atomic_read(&fi->i_compr_blocks)) {
3916 set_sbi_flag(sbi, SBI_NEED_FSCK);
3917 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx "
3918 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3919 "run fsck to fix.",
3920 __func__, inode->i_ino, inode->i_blocks,
3921 reserved_blocks,
3922 atomic_read(&fi->i_compr_blocks));
3923 }
3924
3925 return ret;
3926 }
3927
3928 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3929 pgoff_t off, block_t block, block_t len, u32 flags)
3930 {
3931 sector_t sector = SECTOR_FROM_BLOCK(block);
3932 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3933 int ret = 0;
3934
3935 if (flags & F2FS_TRIM_FILE_DISCARD) {
3936 if (bdev_max_secure_erase_sectors(bdev))
3937 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3938 GFP_NOFS);
3939 else
3940 ret = blkdev_issue_discard(bdev, sector, nr_sects,
3941 GFP_NOFS);
3942 }
3943
3944 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3945 if (IS_ENCRYPTED(inode))
3946 ret = fscrypt_zeroout_range(inode, off, block, len);
3947 else
3948 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3949 GFP_NOFS, 0);
3950 }
3951
3952 return ret;
3953 }
3954
3955 static int f2fs_sec_trim_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 address_space *mapping = inode->i_mapping;
3960 struct block_device *prev_bdev = NULL;
3961 struct f2fs_sectrim_range range;
3962 pgoff_t index, pg_end, prev_index = 0;
3963 block_t prev_block = 0, len = 0;
3964 loff_t end_addr;
3965 bool to_end = false;
3966 int ret = 0;
3967
3968 if (!(filp->f_mode & FMODE_WRITE))
3969 return -EBADF;
3970
3971 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3972 sizeof(range)))
3973 return -EFAULT;
3974
3975 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3976 !S_ISREG(inode->i_mode))
3977 return -EINVAL;
3978
3979 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3980 !f2fs_hw_support_discard(sbi)) ||
3981 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3982 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3983 return -EOPNOTSUPP;
3984
3985 ret = mnt_want_write_file(filp);
3986 if (ret)
3987 return ret;
3988 inode_lock(inode);
3989
3990 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3991 range.start >= inode->i_size) {
3992 ret = -EINVAL;
3993 goto err;
3994 }
3995
3996 if (range.len == 0)
3997 goto err;
3998
3999 if (inode->i_size - range.start > range.len) {
4000 end_addr = range.start + range.len;
4001 } else {
4002 end_addr = range.len == (u64)-1 ?
4003 sbi->sb->s_maxbytes : inode->i_size;
4004 to_end = true;
4005 }
4006
4007 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
4008 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
4009 ret = -EINVAL;
4010 goto err;
4011 }
4012
4013 index = F2FS_BYTES_TO_BLK(range.start);
4014 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
4015
4016 ret = f2fs_convert_inline_inode(inode);
4017 if (ret)
4018 goto err;
4019
4020 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4021 filemap_invalidate_lock(mapping);
4022
4023 ret = filemap_write_and_wait_range(mapping, range.start,
4024 to_end ? LLONG_MAX : end_addr - 1);
4025 if (ret)
4026 goto out;
4027
4028 truncate_inode_pages_range(mapping, range.start,
4029 to_end ? -1 : end_addr - 1);
4030
4031 while (index < pg_end) {
4032 struct dnode_of_data dn;
4033 pgoff_t end_offset, count;
4034 int i;
4035
4036 set_new_dnode(&dn, inode, NULL, NULL, 0);
4037 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4038 if (ret) {
4039 if (ret == -ENOENT) {
4040 index = f2fs_get_next_page_offset(&dn, index);
4041 continue;
4042 }
4043 goto out;
4044 }
4045
4046 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
4047 count = min(end_offset - dn.ofs_in_node, pg_end - index);
4048 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
4049 struct block_device *cur_bdev;
4050 block_t blkaddr = f2fs_data_blkaddr(&dn);
4051
4052 if (!__is_valid_data_blkaddr(blkaddr))
4053 continue;
4054
4055 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
4056 DATA_GENERIC_ENHANCE)) {
4057 ret = -EFSCORRUPTED;
4058 f2fs_put_dnode(&dn);
4059 goto out;
4060 }
4061
4062 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
4063 if (f2fs_is_multi_device(sbi)) {
4064 int di = f2fs_target_device_index(sbi, blkaddr);
4065
4066 blkaddr -= FDEV(di).start_blk;
4067 }
4068
4069 if (len) {
4070 if (prev_bdev == cur_bdev &&
4071 index == prev_index + len &&
4072 blkaddr == prev_block + len) {
4073 len++;
4074 } else {
4075 ret = f2fs_secure_erase(prev_bdev,
4076 inode, prev_index, prev_block,
4077 len, range.flags);
4078 if (ret) {
4079 f2fs_put_dnode(&dn);
4080 goto out;
4081 }
4082
4083 len = 0;
4084 }
4085 }
4086
4087 if (!len) {
4088 prev_bdev = cur_bdev;
4089 prev_index = index;
4090 prev_block = blkaddr;
4091 len = 1;
4092 }
4093 }
4094
4095 f2fs_put_dnode(&dn);
4096
4097 if (fatal_signal_pending(current)) {
4098 ret = -EINTR;
4099 goto out;
4100 }
4101 cond_resched();
4102 }
4103
4104 if (len)
4105 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
4106 prev_block, len, range.flags);
4107 f2fs_update_time(sbi, REQ_TIME);
4108 out:
4109 filemap_invalidate_unlock(mapping);
4110 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4111 err:
4112 inode_unlock(inode);
4113 mnt_drop_write_file(filp);
4114
4115 return ret;
4116 }
4117
4118 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
4119 {
4120 struct inode *inode = file_inode(filp);
4121 struct f2fs_comp_option option;
4122
4123 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4124 return -EOPNOTSUPP;
4125
4126 inode_lock_shared(inode);
4127
4128 if (!f2fs_compressed_file(inode)) {
4129 inode_unlock_shared(inode);
4130 return -ENODATA;
4131 }
4132
4133 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
4134 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
4135
4136 inode_unlock_shared(inode);
4137
4138 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
4139 sizeof(option)))
4140 return -EFAULT;
4141
4142 return 0;
4143 }
4144
4145 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4146 {
4147 struct inode *inode = file_inode(filp);
4148 struct f2fs_inode_info *fi = F2FS_I(inode);
4149 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4150 struct f2fs_comp_option option;
4151 int ret = 0;
4152
4153 if (!f2fs_sb_has_compression(sbi))
4154 return -EOPNOTSUPP;
4155
4156 if (!(filp->f_mode & FMODE_WRITE))
4157 return -EBADF;
4158
4159 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4160 sizeof(option)))
4161 return -EFAULT;
4162
4163 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
4164 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
4165 option.algorithm >= COMPRESS_MAX)
4166 return -EINVAL;
4167
4168 ret = mnt_want_write_file(filp);
4169 if (ret)
4170 return ret;
4171 inode_lock(inode);
4172
4173 f2fs_down_write(&F2FS_I(inode)->i_sem);
4174 if (!f2fs_compressed_file(inode)) {
4175 ret = -EINVAL;
4176 goto out;
4177 }
4178
4179 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4180 ret = -EBUSY;
4181 goto out;
4182 }
4183
4184 if (F2FS_HAS_BLOCKS(inode)) {
4185 ret = -EFBIG;
4186 goto out;
4187 }
4188
4189 fi->i_compress_algorithm = option.algorithm;
4190 fi->i_log_cluster_size = option.log_cluster_size;
4191 fi->i_cluster_size = BIT(option.log_cluster_size);
4192 /* Set default level */
4193 if (fi->i_compress_algorithm == COMPRESS_ZSTD)
4194 fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
4195 else
4196 fi->i_compress_level = 0;
4197 /* Adjust mount option level */
4198 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm &&
4199 F2FS_OPTION(sbi).compress_level)
4200 fi->i_compress_level = F2FS_OPTION(sbi).compress_level;
4201 f2fs_mark_inode_dirty_sync(inode, true);
4202
4203 if (!f2fs_is_compress_backend_ready(inode))
4204 f2fs_warn(sbi, "compression algorithm is successfully set, "
4205 "but current kernel doesn't support this algorithm.");
4206 out:
4207 f2fs_up_write(&fi->i_sem);
4208 inode_unlock(inode);
4209 mnt_drop_write_file(filp);
4210
4211 return ret;
4212 }
4213
4214 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4215 {
4216 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
4217 struct address_space *mapping = inode->i_mapping;
4218 struct page *page;
4219 pgoff_t redirty_idx = page_idx;
4220 int i, page_len = 0, ret = 0;
4221
4222 page_cache_ra_unbounded(&ractl, len, 0);
4223
4224 for (i = 0; i < len; i++, page_idx++) {
4225 page = read_cache_page(mapping, page_idx, NULL, NULL);
4226 if (IS_ERR(page)) {
4227 ret = PTR_ERR(page);
4228 break;
4229 }
4230 page_len++;
4231 }
4232
4233 for (i = 0; i < page_len; i++, redirty_idx++) {
4234 page = find_lock_page(mapping, redirty_idx);
4235
4236 /* It will never fail, when page has pinned above */
4237 f2fs_bug_on(F2FS_I_SB(inode), !page);
4238
4239 f2fs_wait_on_page_writeback(page, DATA, true, true);
4240
4241 set_page_dirty(page);
4242 set_page_private_gcing(page);
4243 f2fs_put_page(page, 1);
4244 f2fs_put_page(page, 0);
4245 }
4246
4247 return ret;
4248 }
4249
4250 static int f2fs_ioc_decompress_file(struct file *filp)
4251 {
4252 struct inode *inode = file_inode(filp);
4253 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4254 struct f2fs_inode_info *fi = F2FS_I(inode);
4255 pgoff_t page_idx = 0, last_idx, cluster_idx;
4256 int ret;
4257
4258 if (!f2fs_sb_has_compression(sbi) ||
4259 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4260 return -EOPNOTSUPP;
4261
4262 if (!(filp->f_mode & FMODE_WRITE))
4263 return -EBADF;
4264
4265 f2fs_balance_fs(sbi, true);
4266
4267 ret = mnt_want_write_file(filp);
4268 if (ret)
4269 return ret;
4270 inode_lock(inode);
4271
4272 if (!f2fs_is_compress_backend_ready(inode)) {
4273 ret = -EOPNOTSUPP;
4274 goto out;
4275 }
4276
4277 if (!f2fs_compressed_file(inode) ||
4278 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4279 ret = -EINVAL;
4280 goto out;
4281 }
4282
4283 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4284 if (ret)
4285 goto out;
4286
4287 if (!atomic_read(&fi->i_compr_blocks))
4288 goto out;
4289
4290 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4291 last_idx >>= fi->i_log_cluster_size;
4292
4293 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4294 page_idx = cluster_idx << fi->i_log_cluster_size;
4295
4296 if (!f2fs_is_compressed_cluster(inode, page_idx))
4297 continue;
4298
4299 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4300 if (ret < 0)
4301 break;
4302
4303 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4304 ret = filemap_fdatawrite(inode->i_mapping);
4305 if (ret < 0)
4306 break;
4307 }
4308
4309 cond_resched();
4310 if (fatal_signal_pending(current)) {
4311 ret = -EINTR;
4312 break;
4313 }
4314 }
4315
4316 if (!ret)
4317 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4318 LLONG_MAX);
4319
4320 if (ret)
4321 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4322 __func__, ret);
4323 f2fs_update_time(sbi, REQ_TIME);
4324 out:
4325 inode_unlock(inode);
4326 mnt_drop_write_file(filp);
4327
4328 return ret;
4329 }
4330
4331 static int f2fs_ioc_compress_file(struct file *filp)
4332 {
4333 struct inode *inode = file_inode(filp);
4334 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4335 struct f2fs_inode_info *fi = F2FS_I(inode);
4336 pgoff_t page_idx = 0, last_idx, cluster_idx;
4337 int ret;
4338
4339 if (!f2fs_sb_has_compression(sbi) ||
4340 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4341 return -EOPNOTSUPP;
4342
4343 if (!(filp->f_mode & FMODE_WRITE))
4344 return -EBADF;
4345
4346 f2fs_balance_fs(sbi, true);
4347
4348 ret = mnt_want_write_file(filp);
4349 if (ret)
4350 return ret;
4351 inode_lock(inode);
4352
4353 if (!f2fs_is_compress_backend_ready(inode)) {
4354 ret = -EOPNOTSUPP;
4355 goto out;
4356 }
4357
4358 if (!f2fs_compressed_file(inode) ||
4359 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4360 ret = -EINVAL;
4361 goto out;
4362 }
4363
4364 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4365 if (ret)
4366 goto out;
4367
4368 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4369
4370 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4371 last_idx >>= fi->i_log_cluster_size;
4372
4373 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4374 page_idx = cluster_idx << fi->i_log_cluster_size;
4375
4376 if (f2fs_is_sparse_cluster(inode, page_idx))
4377 continue;
4378
4379 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4380 if (ret < 0)
4381 break;
4382
4383 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4384 ret = filemap_fdatawrite(inode->i_mapping);
4385 if (ret < 0)
4386 break;
4387 }
4388
4389 cond_resched();
4390 if (fatal_signal_pending(current)) {
4391 ret = -EINTR;
4392 break;
4393 }
4394 }
4395
4396 if (!ret)
4397 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4398 LLONG_MAX);
4399
4400 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4401
4402 if (ret)
4403 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4404 __func__, ret);
4405 f2fs_update_time(sbi, REQ_TIME);
4406 out:
4407 inode_unlock(inode);
4408 mnt_drop_write_file(filp);
4409
4410 return ret;
4411 }
4412
4413 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4414 {
4415 switch (cmd) {
4416 case FS_IOC_GETVERSION:
4417 return f2fs_ioc_getversion(filp, arg);
4418 case F2FS_IOC_START_ATOMIC_WRITE:
4419 return f2fs_ioc_start_atomic_write(filp, false);
4420 case F2FS_IOC_START_ATOMIC_REPLACE:
4421 return f2fs_ioc_start_atomic_write(filp, true);
4422 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4423 return f2fs_ioc_commit_atomic_write(filp);
4424 case F2FS_IOC_ABORT_ATOMIC_WRITE:
4425 return f2fs_ioc_abort_atomic_write(filp);
4426 case F2FS_IOC_START_VOLATILE_WRITE:
4427 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4428 return -EOPNOTSUPP;
4429 case F2FS_IOC_SHUTDOWN:
4430 return f2fs_ioc_shutdown(filp, arg);
4431 case FITRIM:
4432 return f2fs_ioc_fitrim(filp, arg);
4433 case FS_IOC_SET_ENCRYPTION_POLICY:
4434 return f2fs_ioc_set_encryption_policy(filp, arg);
4435 case FS_IOC_GET_ENCRYPTION_POLICY:
4436 return f2fs_ioc_get_encryption_policy(filp, arg);
4437 case FS_IOC_GET_ENCRYPTION_PWSALT:
4438 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4439 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4440 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4441 case FS_IOC_ADD_ENCRYPTION_KEY:
4442 return f2fs_ioc_add_encryption_key(filp, arg);
4443 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4444 return f2fs_ioc_remove_encryption_key(filp, arg);
4445 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4446 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4447 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4448 return f2fs_ioc_get_encryption_key_status(filp, arg);
4449 case FS_IOC_GET_ENCRYPTION_NONCE:
4450 return f2fs_ioc_get_encryption_nonce(filp, arg);
4451 case F2FS_IOC_GARBAGE_COLLECT:
4452 return f2fs_ioc_gc(filp, arg);
4453 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4454 return f2fs_ioc_gc_range(filp, arg);
4455 case F2FS_IOC_WRITE_CHECKPOINT:
4456 return f2fs_ioc_write_checkpoint(filp);
4457 case F2FS_IOC_DEFRAGMENT:
4458 return f2fs_ioc_defragment(filp, arg);
4459 case F2FS_IOC_MOVE_RANGE:
4460 return f2fs_ioc_move_range(filp, arg);
4461 case F2FS_IOC_FLUSH_DEVICE:
4462 return f2fs_ioc_flush_device(filp, arg);
4463 case F2FS_IOC_GET_FEATURES:
4464 return f2fs_ioc_get_features(filp, arg);
4465 case F2FS_IOC_GET_PIN_FILE:
4466 return f2fs_ioc_get_pin_file(filp, arg);
4467 case F2FS_IOC_SET_PIN_FILE:
4468 return f2fs_ioc_set_pin_file(filp, arg);
4469 case F2FS_IOC_PRECACHE_EXTENTS:
4470 return f2fs_ioc_precache_extents(filp);
4471 case F2FS_IOC_RESIZE_FS:
4472 return f2fs_ioc_resize_fs(filp, arg);
4473 case FS_IOC_ENABLE_VERITY:
4474 return f2fs_ioc_enable_verity(filp, arg);
4475 case FS_IOC_MEASURE_VERITY:
4476 return f2fs_ioc_measure_verity(filp, arg);
4477 case FS_IOC_READ_VERITY_METADATA:
4478 return f2fs_ioc_read_verity_metadata(filp, arg);
4479 case FS_IOC_GETFSLABEL:
4480 return f2fs_ioc_getfslabel(filp, arg);
4481 case FS_IOC_SETFSLABEL:
4482 return f2fs_ioc_setfslabel(filp, arg);
4483 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4484 return f2fs_ioc_get_compress_blocks(filp, arg);
4485 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4486 return f2fs_release_compress_blocks(filp, arg);
4487 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4488 return f2fs_reserve_compress_blocks(filp, arg);
4489 case F2FS_IOC_SEC_TRIM_FILE:
4490 return f2fs_sec_trim_file(filp, arg);
4491 case F2FS_IOC_GET_COMPRESS_OPTION:
4492 return f2fs_ioc_get_compress_option(filp, arg);
4493 case F2FS_IOC_SET_COMPRESS_OPTION:
4494 return f2fs_ioc_set_compress_option(filp, arg);
4495 case F2FS_IOC_DECOMPRESS_FILE:
4496 return f2fs_ioc_decompress_file(filp);
4497 case F2FS_IOC_COMPRESS_FILE:
4498 return f2fs_ioc_compress_file(filp);
4499 default:
4500 return -ENOTTY;
4501 }
4502 }
4503
4504 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4505 {
4506 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4507 return -EIO;
4508 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4509 return -ENOSPC;
4510
4511 return __f2fs_ioctl(filp, cmd, arg);
4512 }
4513
4514 /*
4515 * Return %true if the given read or write request should use direct I/O, or
4516 * %false if it should use buffered I/O.
4517 */
4518 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4519 struct iov_iter *iter)
4520 {
4521 unsigned int align;
4522
4523 if (!(iocb->ki_flags & IOCB_DIRECT))
4524 return false;
4525
4526 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4527 return false;
4528
4529 /*
4530 * Direct I/O not aligned to the disk's logical_block_size will be
4531 * attempted, but will fail with -EINVAL.
4532 *
4533 * f2fs additionally requires that direct I/O be aligned to the
4534 * filesystem block size, which is often a stricter requirement.
4535 * However, f2fs traditionally falls back to buffered I/O on requests
4536 * that are logical_block_size-aligned but not fs-block aligned.
4537 *
4538 * The below logic implements this behavior.
4539 */
4540 align = iocb->ki_pos | iov_iter_alignment(iter);
4541 if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4542 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4543 return false;
4544
4545 return true;
4546 }
4547
4548 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4549 unsigned int flags)
4550 {
4551 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4552
4553 dec_page_count(sbi, F2FS_DIO_READ);
4554 if (error)
4555 return error;
4556 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4557 return 0;
4558 }
4559
4560 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4561 .end_io = f2fs_dio_read_end_io,
4562 };
4563
4564 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4565 {
4566 struct file *file = iocb->ki_filp;
4567 struct inode *inode = file_inode(file);
4568 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4569 struct f2fs_inode_info *fi = F2FS_I(inode);
4570 const loff_t pos = iocb->ki_pos;
4571 const size_t count = iov_iter_count(to);
4572 struct iomap_dio *dio;
4573 ssize_t ret;
4574
4575 if (count == 0)
4576 return 0; /* skip atime update */
4577
4578 trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4579
4580 if (iocb->ki_flags & IOCB_NOWAIT) {
4581 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4582 ret = -EAGAIN;
4583 goto out;
4584 }
4585 } else {
4586 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4587 }
4588
4589 /*
4590 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4591 * the higher-level function iomap_dio_rw() in order to ensure that the
4592 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4593 */
4594 inc_page_count(sbi, F2FS_DIO_READ);
4595 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4596 &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4597 if (IS_ERR_OR_NULL(dio)) {
4598 ret = PTR_ERR_OR_ZERO(dio);
4599 if (ret != -EIOCBQUEUED)
4600 dec_page_count(sbi, F2FS_DIO_READ);
4601 } else {
4602 ret = iomap_dio_complete(dio);
4603 }
4604
4605 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4606
4607 file_accessed(file);
4608 out:
4609 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4610 return ret;
4611 }
4612
4613 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count,
4614 int rw)
4615 {
4616 struct inode *inode = file_inode(file);
4617 char *buf, *path;
4618
4619 buf = f2fs_getname(F2FS_I_SB(inode));
4620 if (!buf)
4621 return;
4622 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX);
4623 if (IS_ERR(path))
4624 goto free_buf;
4625 if (rw == WRITE)
4626 trace_f2fs_datawrite_start(inode, pos, count,
4627 current->pid, path, current->comm);
4628 else
4629 trace_f2fs_dataread_start(inode, pos, count,
4630 current->pid, path, current->comm);
4631 free_buf:
4632 f2fs_putname(buf);
4633 }
4634
4635 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4636 {
4637 struct inode *inode = file_inode(iocb->ki_filp);
4638 const loff_t pos = iocb->ki_pos;
4639 ssize_t ret;
4640
4641 if (!f2fs_is_compress_backend_ready(inode))
4642 return -EOPNOTSUPP;
4643
4644 if (trace_f2fs_dataread_start_enabled())
4645 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
4646 iov_iter_count(to), READ);
4647
4648 /* In LFS mode, if there is inflight dio, wait for its completion */
4649 if (f2fs_lfs_mode(F2FS_I_SB(inode)))
4650 inode_dio_wait(inode);
4651
4652 if (f2fs_should_use_dio(inode, iocb, to)) {
4653 ret = f2fs_dio_read_iter(iocb, to);
4654 } else {
4655 ret = filemap_read(iocb, to, 0);
4656 if (ret > 0)
4657 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4658 APP_BUFFERED_READ_IO, ret);
4659 }
4660 if (trace_f2fs_dataread_end_enabled())
4661 trace_f2fs_dataread_end(inode, pos, ret);
4662 return ret;
4663 }
4664
4665 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos,
4666 struct pipe_inode_info *pipe,
4667 size_t len, unsigned int flags)
4668 {
4669 struct inode *inode = file_inode(in);
4670 const loff_t pos = *ppos;
4671 ssize_t ret;
4672
4673 if (!f2fs_is_compress_backend_ready(inode))
4674 return -EOPNOTSUPP;
4675
4676 if (trace_f2fs_dataread_start_enabled())
4677 f2fs_trace_rw_file_path(in, pos, len, READ);
4678
4679 ret = filemap_splice_read(in, ppos, pipe, len, flags);
4680 if (ret > 0)
4681 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4682 APP_BUFFERED_READ_IO, ret);
4683
4684 if (trace_f2fs_dataread_end_enabled())
4685 trace_f2fs_dataread_end(inode, pos, ret);
4686 return ret;
4687 }
4688
4689 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4690 {
4691 struct file *file = iocb->ki_filp;
4692 struct inode *inode = file_inode(file);
4693 ssize_t count;
4694 int err;
4695
4696 if (IS_IMMUTABLE(inode))
4697 return -EPERM;
4698
4699 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4700 return -EPERM;
4701
4702 count = generic_write_checks(iocb, from);
4703 if (count <= 0)
4704 return count;
4705
4706 err = file_modified(file);
4707 if (err)
4708 return err;
4709 return count;
4710 }
4711
4712 /*
4713 * Preallocate blocks for a write request, if it is possible and helpful to do
4714 * so. Returns a positive number if blocks may have been preallocated, 0 if no
4715 * blocks were preallocated, or a negative errno value if something went
4716 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4717 * requested blocks (not just some of them) have been allocated.
4718 */
4719 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4720 bool dio)
4721 {
4722 struct inode *inode = file_inode(iocb->ki_filp);
4723 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4724 const loff_t pos = iocb->ki_pos;
4725 const size_t count = iov_iter_count(iter);
4726 struct f2fs_map_blocks map = {};
4727 int flag;
4728 int ret;
4729
4730 /* If it will be an out-of-place direct write, don't bother. */
4731 if (dio && f2fs_lfs_mode(sbi))
4732 return 0;
4733 /*
4734 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4735 * buffered IO, if DIO meets any holes.
4736 */
4737 if (dio && i_size_read(inode) &&
4738 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4739 return 0;
4740
4741 /* No-wait I/O can't allocate blocks. */
4742 if (iocb->ki_flags & IOCB_NOWAIT)
4743 return 0;
4744
4745 /* If it will be a short write, don't bother. */
4746 if (fault_in_iov_iter_readable(iter, count))
4747 return 0;
4748
4749 if (f2fs_has_inline_data(inode)) {
4750 /* If the data will fit inline, don't bother. */
4751 if (pos + count <= MAX_INLINE_DATA(inode))
4752 return 0;
4753 ret = f2fs_convert_inline_inode(inode);
4754 if (ret)
4755 return ret;
4756 }
4757
4758 /* Do not preallocate blocks that will be written partially in 4KB. */
4759 map.m_lblk = F2FS_BLK_ALIGN(pos);
4760 map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4761 if (map.m_len > map.m_lblk)
4762 map.m_len -= map.m_lblk;
4763 else
4764 return 0;
4765
4766 map.m_may_create = true;
4767 if (dio) {
4768 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi,
4769 inode->i_write_hint);
4770 flag = F2FS_GET_BLOCK_PRE_DIO;
4771 } else {
4772 map.m_seg_type = NO_CHECK_TYPE;
4773 flag = F2FS_GET_BLOCK_PRE_AIO;
4774 }
4775
4776 ret = f2fs_map_blocks(inode, &map, flag);
4777 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4778 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4779 return ret;
4780 if (ret == 0)
4781 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4782 return map.m_len;
4783 }
4784
4785 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4786 struct iov_iter *from)
4787 {
4788 struct file *file = iocb->ki_filp;
4789 struct inode *inode = file_inode(file);
4790 ssize_t ret;
4791
4792 if (iocb->ki_flags & IOCB_NOWAIT)
4793 return -EOPNOTSUPP;
4794
4795 ret = generic_perform_write(iocb, from);
4796
4797 if (ret > 0) {
4798 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4799 APP_BUFFERED_IO, ret);
4800 }
4801 return ret;
4802 }
4803
4804 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4805 unsigned int flags)
4806 {
4807 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4808
4809 dec_page_count(sbi, F2FS_DIO_WRITE);
4810 if (error)
4811 return error;
4812 f2fs_update_time(sbi, REQ_TIME);
4813 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4814 return 0;
4815 }
4816
4817 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter,
4818 struct bio *bio, loff_t file_offset)
4819 {
4820 struct inode *inode = iter->inode;
4821 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4822 int seg_type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint);
4823 enum temp_type temp = f2fs_get_segment_temp(seg_type);
4824
4825 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp);
4826 submit_bio(bio);
4827 }
4828
4829 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4830 .end_io = f2fs_dio_write_end_io,
4831 .submit_io = f2fs_dio_write_submit_io,
4832 };
4833
4834 static void f2fs_flush_buffered_write(struct address_space *mapping,
4835 loff_t start_pos, loff_t end_pos)
4836 {
4837 int ret;
4838
4839 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos);
4840 if (ret < 0)
4841 return;
4842 invalidate_mapping_pages(mapping,
4843 start_pos >> PAGE_SHIFT,
4844 end_pos >> PAGE_SHIFT);
4845 }
4846
4847 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4848 bool *may_need_sync)
4849 {
4850 struct file *file = iocb->ki_filp;
4851 struct inode *inode = file_inode(file);
4852 struct f2fs_inode_info *fi = F2FS_I(inode);
4853 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4854 const bool do_opu = f2fs_lfs_mode(sbi);
4855 const loff_t pos = iocb->ki_pos;
4856 const ssize_t count = iov_iter_count(from);
4857 unsigned int dio_flags;
4858 struct iomap_dio *dio;
4859 ssize_t ret;
4860
4861 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4862
4863 if (iocb->ki_flags & IOCB_NOWAIT) {
4864 /* f2fs_convert_inline_inode() and block allocation can block */
4865 if (f2fs_has_inline_data(inode) ||
4866 !f2fs_overwrite_io(inode, pos, count)) {
4867 ret = -EAGAIN;
4868 goto out;
4869 }
4870
4871 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4872 ret = -EAGAIN;
4873 goto out;
4874 }
4875 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4876 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4877 ret = -EAGAIN;
4878 goto out;
4879 }
4880 } else {
4881 ret = f2fs_convert_inline_inode(inode);
4882 if (ret)
4883 goto out;
4884
4885 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4886 if (do_opu)
4887 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4888 }
4889
4890 /*
4891 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4892 * the higher-level function iomap_dio_rw() in order to ensure that the
4893 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4894 */
4895 inc_page_count(sbi, F2FS_DIO_WRITE);
4896 dio_flags = 0;
4897 if (pos + count > inode->i_size)
4898 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4899 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4900 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4901 if (IS_ERR_OR_NULL(dio)) {
4902 ret = PTR_ERR_OR_ZERO(dio);
4903 if (ret == -ENOTBLK)
4904 ret = 0;
4905 if (ret != -EIOCBQUEUED)
4906 dec_page_count(sbi, F2FS_DIO_WRITE);
4907 } else {
4908 ret = iomap_dio_complete(dio);
4909 }
4910
4911 if (do_opu)
4912 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4913 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4914
4915 if (ret < 0)
4916 goto out;
4917 if (pos + ret > inode->i_size)
4918 f2fs_i_size_write(inode, pos + ret);
4919 if (!do_opu)
4920 set_inode_flag(inode, FI_UPDATE_WRITE);
4921
4922 if (iov_iter_count(from)) {
4923 ssize_t ret2;
4924 loff_t bufio_start_pos = iocb->ki_pos;
4925
4926 /*
4927 * The direct write was partial, so we need to fall back to a
4928 * buffered write for the remainder.
4929 */
4930
4931 ret2 = f2fs_buffered_write_iter(iocb, from);
4932 if (iov_iter_count(from))
4933 f2fs_write_failed(inode, iocb->ki_pos);
4934 if (ret2 < 0)
4935 goto out;
4936
4937 /*
4938 * Ensure that the pagecache pages are written to disk and
4939 * invalidated to preserve the expected O_DIRECT semantics.
4940 */
4941 if (ret2 > 0) {
4942 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4943
4944 ret += ret2;
4945
4946 f2fs_flush_buffered_write(file->f_mapping,
4947 bufio_start_pos,
4948 bufio_end_pos);
4949 }
4950 } else {
4951 /* iomap_dio_rw() already handled the generic_write_sync(). */
4952 *may_need_sync = false;
4953 }
4954 out:
4955 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4956 return ret;
4957 }
4958
4959 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4960 {
4961 struct inode *inode = file_inode(iocb->ki_filp);
4962 const loff_t orig_pos = iocb->ki_pos;
4963 const size_t orig_count = iov_iter_count(from);
4964 loff_t target_size;
4965 bool dio;
4966 bool may_need_sync = true;
4967 int preallocated;
4968 const loff_t pos = iocb->ki_pos;
4969 const ssize_t count = iov_iter_count(from);
4970 ssize_t ret;
4971
4972 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4973 ret = -EIO;
4974 goto out;
4975 }
4976
4977 if (!f2fs_is_compress_backend_ready(inode)) {
4978 ret = -EOPNOTSUPP;
4979 goto out;
4980 }
4981
4982 if (iocb->ki_flags & IOCB_NOWAIT) {
4983 if (!inode_trylock(inode)) {
4984 ret = -EAGAIN;
4985 goto out;
4986 }
4987 } else {
4988 inode_lock(inode);
4989 }
4990
4991 if (f2fs_is_pinned_file(inode) &&
4992 !f2fs_overwrite_io(inode, pos, count)) {
4993 ret = -EIO;
4994 goto out_unlock;
4995 }
4996
4997 ret = f2fs_write_checks(iocb, from);
4998 if (ret <= 0)
4999 goto out_unlock;
5000
5001 /* Determine whether we will do a direct write or a buffered write. */
5002 dio = f2fs_should_use_dio(inode, iocb, from);
5003
5004 /* Possibly preallocate the blocks for the write. */
5005 target_size = iocb->ki_pos + iov_iter_count(from);
5006 preallocated = f2fs_preallocate_blocks(iocb, from, dio);
5007 if (preallocated < 0) {
5008 ret = preallocated;
5009 } else {
5010 if (trace_f2fs_datawrite_start_enabled())
5011 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
5012 orig_count, WRITE);
5013
5014 /* Do the actual write. */
5015 ret = dio ?
5016 f2fs_dio_write_iter(iocb, from, &may_need_sync) :
5017 f2fs_buffered_write_iter(iocb, from);
5018
5019 if (trace_f2fs_datawrite_end_enabled())
5020 trace_f2fs_datawrite_end(inode, orig_pos, ret);
5021 }
5022
5023 /* Don't leave any preallocated blocks around past i_size. */
5024 if (preallocated && i_size_read(inode) < target_size) {
5025 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5026 filemap_invalidate_lock(inode->i_mapping);
5027 if (!f2fs_truncate(inode))
5028 file_dont_truncate(inode);
5029 filemap_invalidate_unlock(inode->i_mapping);
5030 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5031 } else {
5032 file_dont_truncate(inode);
5033 }
5034
5035 clear_inode_flag(inode, FI_PREALLOCATED_ALL);
5036 out_unlock:
5037 inode_unlock(inode);
5038 out:
5039 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
5040
5041 if (ret > 0 && may_need_sync)
5042 ret = generic_write_sync(iocb, ret);
5043
5044 /* If buffered IO was forced, flush and drop the data from
5045 * the page cache to preserve O_DIRECT semantics
5046 */
5047 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
5048 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
5049 orig_pos,
5050 orig_pos + ret - 1);
5051
5052 return ret;
5053 }
5054
5055 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
5056 int advice)
5057 {
5058 struct address_space *mapping;
5059 struct backing_dev_info *bdi;
5060 struct inode *inode = file_inode(filp);
5061 int err;
5062
5063 if (advice == POSIX_FADV_SEQUENTIAL) {
5064 if (S_ISFIFO(inode->i_mode))
5065 return -ESPIPE;
5066
5067 mapping = filp->f_mapping;
5068 if (!mapping || len < 0)
5069 return -EINVAL;
5070
5071 bdi = inode_to_bdi(mapping->host);
5072 filp->f_ra.ra_pages = bdi->ra_pages *
5073 F2FS_I_SB(inode)->seq_file_ra_mul;
5074 spin_lock(&filp->f_lock);
5075 filp->f_mode &= ~FMODE_RANDOM;
5076 spin_unlock(&filp->f_lock);
5077 return 0;
5078 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) {
5079 /* Load extent cache at the first readahead. */
5080 f2fs_precache_extents(inode);
5081 }
5082
5083 err = generic_fadvise(filp, offset, len, advice);
5084 if (!err && advice == POSIX_FADV_DONTNEED &&
5085 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
5086 f2fs_compressed_file(inode))
5087 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
5088
5089 return err;
5090 }
5091
5092 #ifdef CONFIG_COMPAT
5093 struct compat_f2fs_gc_range {
5094 u32 sync;
5095 compat_u64 start;
5096 compat_u64 len;
5097 };
5098 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
5099 struct compat_f2fs_gc_range)
5100
5101 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
5102 {
5103 struct compat_f2fs_gc_range __user *urange;
5104 struct f2fs_gc_range range;
5105 int err;
5106
5107 urange = compat_ptr(arg);
5108 err = get_user(range.sync, &urange->sync);
5109 err |= get_user(range.start, &urange->start);
5110 err |= get_user(range.len, &urange->len);
5111 if (err)
5112 return -EFAULT;
5113
5114 return __f2fs_ioc_gc_range(file, &range);
5115 }
5116
5117 struct compat_f2fs_move_range {
5118 u32 dst_fd;
5119 compat_u64 pos_in;
5120 compat_u64 pos_out;
5121 compat_u64 len;
5122 };
5123 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
5124 struct compat_f2fs_move_range)
5125
5126 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
5127 {
5128 struct compat_f2fs_move_range __user *urange;
5129 struct f2fs_move_range range;
5130 int err;
5131
5132 urange = compat_ptr(arg);
5133 err = get_user(range.dst_fd, &urange->dst_fd);
5134 err |= get_user(range.pos_in, &urange->pos_in);
5135 err |= get_user(range.pos_out, &urange->pos_out);
5136 err |= get_user(range.len, &urange->len);
5137 if (err)
5138 return -EFAULT;
5139
5140 return __f2fs_ioc_move_range(file, &range);
5141 }
5142
5143 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5144 {
5145 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
5146 return -EIO;
5147 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
5148 return -ENOSPC;
5149
5150 switch (cmd) {
5151 case FS_IOC32_GETVERSION:
5152 cmd = FS_IOC_GETVERSION;
5153 break;
5154 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
5155 return f2fs_compat_ioc_gc_range(file, arg);
5156 case F2FS_IOC32_MOVE_RANGE:
5157 return f2fs_compat_ioc_move_range(file, arg);
5158 case F2FS_IOC_START_ATOMIC_WRITE:
5159 case F2FS_IOC_START_ATOMIC_REPLACE:
5160 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
5161 case F2FS_IOC_START_VOLATILE_WRITE:
5162 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
5163 case F2FS_IOC_ABORT_ATOMIC_WRITE:
5164 case F2FS_IOC_SHUTDOWN:
5165 case FITRIM:
5166 case FS_IOC_SET_ENCRYPTION_POLICY:
5167 case FS_IOC_GET_ENCRYPTION_PWSALT:
5168 case FS_IOC_GET_ENCRYPTION_POLICY:
5169 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
5170 case FS_IOC_ADD_ENCRYPTION_KEY:
5171 case FS_IOC_REMOVE_ENCRYPTION_KEY:
5172 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
5173 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
5174 case FS_IOC_GET_ENCRYPTION_NONCE:
5175 case F2FS_IOC_GARBAGE_COLLECT:
5176 case F2FS_IOC_WRITE_CHECKPOINT:
5177 case F2FS_IOC_DEFRAGMENT:
5178 case F2FS_IOC_FLUSH_DEVICE:
5179 case F2FS_IOC_GET_FEATURES:
5180 case F2FS_IOC_GET_PIN_FILE:
5181 case F2FS_IOC_SET_PIN_FILE:
5182 case F2FS_IOC_PRECACHE_EXTENTS:
5183 case F2FS_IOC_RESIZE_FS:
5184 case FS_IOC_ENABLE_VERITY:
5185 case FS_IOC_MEASURE_VERITY:
5186 case FS_IOC_READ_VERITY_METADATA:
5187 case FS_IOC_GETFSLABEL:
5188 case FS_IOC_SETFSLABEL:
5189 case F2FS_IOC_GET_COMPRESS_BLOCKS:
5190 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
5191 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
5192 case F2FS_IOC_SEC_TRIM_FILE:
5193 case F2FS_IOC_GET_COMPRESS_OPTION:
5194 case F2FS_IOC_SET_COMPRESS_OPTION:
5195 case F2FS_IOC_DECOMPRESS_FILE:
5196 case F2FS_IOC_COMPRESS_FILE:
5197 break;
5198 default:
5199 return -ENOIOCTLCMD;
5200 }
5201 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5202 }
5203 #endif
5204
5205 const struct file_operations f2fs_file_operations = {
5206 .llseek = f2fs_llseek,
5207 .read_iter = f2fs_file_read_iter,
5208 .write_iter = f2fs_file_write_iter,
5209 .iopoll = iocb_bio_iopoll,
5210 .open = f2fs_file_open,
5211 .release = f2fs_release_file,
5212 .mmap = f2fs_file_mmap,
5213 .flush = f2fs_file_flush,
5214 .fsync = f2fs_sync_file,
5215 .fallocate = f2fs_fallocate,
5216 .unlocked_ioctl = f2fs_ioctl,
5217 #ifdef CONFIG_COMPAT
5218 .compat_ioctl = f2fs_compat_ioctl,
5219 #endif
5220 .splice_read = f2fs_file_splice_read,
5221 .splice_write = iter_file_splice_write,
5222 .fadvise = f2fs_file_fadvise,
5223 .fop_flags = FOP_BUFFER_RASYNC,
5224 };
5225
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.