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/writeback.h>
12 #include <linux/blkdev.h>
13 #include <linux/falloc.h>
14 #include <linux/types.h>
15 #include <linux/compat.h>
16 #include <linux/uaccess.h>
17 #include <linux/mount.h>
18 #include <linux/pagevec.h>
19 #include <linux/uio.h>
20 #include <linux/uuid.h>
21 #include <linux/file.h>
22 #include <linux/nls.h>
23 #include <linux/sched/signal.h>
24 #include <linux/fileattr.h>
25 #include <linux/fadvise.h>
26 #include <linux/iomap.h>
27
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "acl.h"
33 #include "gc.h"
34 #include "iostat.h"
35 #include <trace/events/f2fs.h>
36 #include <uapi/linux/f2fs.h>
37
f2fs_zero_post_eof_page(struct inode * inode,loff_t new_size)38 static void f2fs_zero_post_eof_page(struct inode *inode, loff_t new_size)
39 {
40 loff_t old_size = i_size_read(inode);
41
42 if (old_size >= new_size)
43 return;
44
45 /* zero or drop pages only in range of [old_size, new_size] */
46 truncate_pagecache(inode, old_size);
47 }
48
f2fs_filemap_fault(struct vm_fault * vmf)49 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
50 {
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 vm_flags_t flags = vmf->vma->vm_flags;
53 vm_fault_t ret;
54
55 ret = filemap_fault(vmf);
56 if (ret & VM_FAULT_LOCKED)
57 f2fs_update_iostat(F2FS_I_SB(inode), inode,
58 APP_MAPPED_READ_IO, F2FS_BLKSIZE);
59
60 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret);
61
62 return ret;
63 }
64
f2fs_vm_page_mkwrite(struct vm_fault * vmf)65 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
66 {
67 struct folio *folio = page_folio(vmf->page);
68 struct inode *inode = file_inode(vmf->vma->vm_file);
69 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
70 struct dnode_of_data dn;
71 bool need_alloc = !f2fs_is_pinned_file(inode);
72 int err = 0;
73 vm_fault_t ret;
74
75 if (unlikely(IS_IMMUTABLE(inode)))
76 return VM_FAULT_SIGBUS;
77
78 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
79 err = -EIO;
80 goto out;
81 }
82
83 if (unlikely(f2fs_cp_error(sbi))) {
84 err = -EIO;
85 goto out;
86 }
87
88 if (!f2fs_is_checkpoint_ready(sbi)) {
89 err = -ENOSPC;
90 goto out;
91 }
92
93 err = f2fs_convert_inline_inode(inode);
94 if (err)
95 goto out;
96
97 #ifdef CONFIG_F2FS_FS_COMPRESSION
98 if (f2fs_compressed_file(inode)) {
99 int ret = f2fs_is_compressed_cluster(inode, folio->index);
100
101 if (ret < 0) {
102 err = ret;
103 goto out;
104 } else if (ret) {
105 need_alloc = false;
106 }
107 }
108 #endif
109 /* should do out of any locked page */
110 if (need_alloc)
111 f2fs_balance_fs(sbi, true);
112
113 sb_start_pagefault(inode->i_sb);
114
115 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
116
117 filemap_invalidate_lock(inode->i_mapping);
118 f2fs_zero_post_eof_page(inode, (folio->index + 1) << PAGE_SHIFT);
119 filemap_invalidate_unlock(inode->i_mapping);
120
121 file_update_time(vmf->vma->vm_file);
122 filemap_invalidate_lock_shared(inode->i_mapping);
123
124 folio_lock(folio);
125 if (unlikely(folio->mapping != inode->i_mapping ||
126 folio_pos(folio) > i_size_read(inode) ||
127 !folio_test_uptodate(folio))) {
128 folio_unlock(folio);
129 err = -EFAULT;
130 goto out_sem;
131 }
132
133 set_new_dnode(&dn, inode, NULL, NULL, 0);
134 if (need_alloc) {
135 /* block allocation */
136 err = f2fs_get_block_locked(&dn, folio->index);
137 } else {
138 err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE);
139 f2fs_put_dnode(&dn);
140 if (f2fs_is_pinned_file(inode) &&
141 !__is_valid_data_blkaddr(dn.data_blkaddr))
142 err = -EIO;
143 }
144
145 if (err) {
146 folio_unlock(folio);
147 goto out_sem;
148 }
149
150 f2fs_folio_wait_writeback(folio, DATA, false, true);
151
152 /* wait for GCed page writeback via META_MAPPING */
153 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
154
155 /*
156 * check to see if the page is mapped already (no holes)
157 */
158 if (folio_test_mappedtodisk(folio))
159 goto out_sem;
160
161 /* page is wholly or partially inside EOF */
162 if (((loff_t)(folio->index + 1) << PAGE_SHIFT) >
163 i_size_read(inode)) {
164 loff_t offset;
165
166 offset = i_size_read(inode) & ~PAGE_MASK;
167 folio_zero_segment(folio, offset, folio_size(folio));
168 }
169 folio_mark_dirty(folio);
170
171 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
172 f2fs_update_time(sbi, REQ_TIME);
173
174 out_sem:
175 filemap_invalidate_unlock_shared(inode->i_mapping);
176
177 sb_end_pagefault(inode->i_sb);
178 out:
179 ret = vmf_fs_error(err);
180
181 trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret);
182 return ret;
183 }
184
185 static const struct vm_operations_struct f2fs_file_vm_ops = {
186 .fault = f2fs_filemap_fault,
187 .map_pages = filemap_map_pages,
188 .page_mkwrite = f2fs_vm_page_mkwrite,
189 };
190
get_parent_ino(struct inode * inode,nid_t * pino)191 static int get_parent_ino(struct inode *inode, nid_t *pino)
192 {
193 struct dentry *dentry;
194
195 /*
196 * Make sure to get the non-deleted alias. The alias associated with
197 * the open file descriptor being fsync()'ed may be deleted already.
198 */
199 dentry = d_find_alias(inode);
200 if (!dentry)
201 return 0;
202
203 *pino = d_parent_ino(dentry);
204 dput(dentry);
205 return 1;
206 }
207
need_do_checkpoint(struct inode * inode)208 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
209 {
210 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
211 enum cp_reason_type cp_reason = CP_NO_NEEDED;
212
213 if (!S_ISREG(inode->i_mode))
214 cp_reason = CP_NON_REGULAR;
215 else if (f2fs_compressed_file(inode))
216 cp_reason = CP_COMPRESSED;
217 else if (inode->i_nlink != 1)
218 cp_reason = CP_HARDLINK;
219 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
220 cp_reason = CP_SB_NEED_CP;
221 else if (file_wrong_pino(inode))
222 cp_reason = CP_WRONG_PINO;
223 else if (!f2fs_space_for_roll_forward(sbi))
224 cp_reason = CP_NO_SPC_ROLL;
225 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
226 cp_reason = CP_NODE_NEED_CP;
227 else if (test_opt(sbi, FASTBOOT))
228 cp_reason = CP_FASTBOOT_MODE;
229 else if (F2FS_OPTION(sbi).active_logs == 2)
230 cp_reason = CP_SPEC_LOG_NUM;
231 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
232 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
233 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
234 TRANS_DIR_INO))
235 cp_reason = CP_RECOVER_DIR;
236 else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
237 XATTR_DIR_INO))
238 cp_reason = CP_XATTR_DIR;
239
240 return cp_reason;
241 }
242
need_inode_page_update(struct f2fs_sb_info * sbi,nid_t ino)243 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
244 {
245 struct folio *i = filemap_get_folio(NODE_MAPPING(sbi), ino);
246 bool ret = false;
247 /* But we need to avoid that there are some inode updates */
248 if ((!IS_ERR(i) && folio_test_dirty(i)) ||
249 f2fs_need_inode_block_update(sbi, ino))
250 ret = true;
251 f2fs_folio_put(i, false);
252 return ret;
253 }
254
try_to_fix_pino(struct inode * inode)255 static void try_to_fix_pino(struct inode *inode)
256 {
257 struct f2fs_inode_info *fi = F2FS_I(inode);
258 nid_t pino;
259
260 f2fs_down_write(&fi->i_sem);
261 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
262 get_parent_ino(inode, &pino)) {
263 f2fs_i_pino_write(inode, pino);
264 file_got_pino(inode);
265 }
266 f2fs_up_write(&fi->i_sem);
267 }
268
f2fs_do_sync_file(struct file * file,loff_t start,loff_t end,int datasync,bool atomic)269 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
270 int datasync, bool atomic)
271 {
272 struct inode *inode = file->f_mapping->host;
273 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
274 nid_t ino = inode->i_ino;
275 int ret = 0;
276 enum cp_reason_type cp_reason = 0;
277 struct writeback_control wbc = {
278 .sync_mode = WB_SYNC_ALL,
279 .nr_to_write = LONG_MAX,
280 };
281 unsigned int seq_id = 0;
282
283 if (unlikely(f2fs_readonly(inode->i_sb)))
284 return 0;
285
286 trace_f2fs_sync_file_enter(inode);
287
288 if (S_ISDIR(inode->i_mode))
289 goto go_write;
290
291 /* if fdatasync is triggered, let's do in-place-update */
292 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
293 set_inode_flag(inode, FI_NEED_IPU);
294 ret = file_write_and_wait_range(file, start, end);
295 clear_inode_flag(inode, FI_NEED_IPU);
296
297 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
298 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
299 return ret;
300 }
301
302 /* if the inode is dirty, let's recover all the time */
303 if (!f2fs_skip_inode_update(inode, datasync)) {
304 f2fs_write_inode(inode, NULL);
305 goto go_write;
306 }
307
308 /*
309 * if there is no written data, don't waste time to write recovery info.
310 */
311 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
312 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
313
314 /* it may call write_inode just prior to fsync */
315 if (need_inode_page_update(sbi, ino))
316 goto go_write;
317
318 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
319 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
320 goto flush_out;
321 goto out;
322 } else {
323 /*
324 * for OPU case, during fsync(), node can be persisted before
325 * data when lower device doesn't support write barrier, result
326 * in data corruption after SPO.
327 * So for strict fsync mode, force to use atomic write semantics
328 * to keep write order in between data/node and last node to
329 * avoid potential data corruption.
330 */
331 if (F2FS_OPTION(sbi).fsync_mode ==
332 FSYNC_MODE_STRICT && !atomic)
333 atomic = true;
334 }
335 go_write:
336 /*
337 * Both of fdatasync() and fsync() are able to be recovered from
338 * sudden-power-off.
339 */
340 f2fs_down_read(&F2FS_I(inode)->i_sem);
341 cp_reason = need_do_checkpoint(inode);
342 f2fs_up_read(&F2FS_I(inode)->i_sem);
343
344 if (cp_reason) {
345 /* all the dirty node pages should be flushed for POR */
346 ret = f2fs_sync_fs(inode->i_sb, 1);
347
348 /*
349 * We've secured consistency through sync_fs. Following pino
350 * will be used only for fsynced inodes after checkpoint.
351 */
352 try_to_fix_pino(inode);
353 clear_inode_flag(inode, FI_APPEND_WRITE);
354 clear_inode_flag(inode, FI_UPDATE_WRITE);
355 goto out;
356 }
357 sync_nodes:
358 atomic_inc(&sbi->wb_sync_req[NODE]);
359 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
360 atomic_dec(&sbi->wb_sync_req[NODE]);
361 if (ret)
362 goto out;
363
364 /* if cp_error was enabled, we should avoid infinite loop */
365 if (unlikely(f2fs_cp_error(sbi))) {
366 ret = -EIO;
367 goto out;
368 }
369
370 if (f2fs_need_inode_block_update(sbi, ino)) {
371 f2fs_mark_inode_dirty_sync(inode, true);
372 f2fs_write_inode(inode, NULL);
373 goto sync_nodes;
374 }
375
376 /*
377 * If it's atomic_write, it's just fine to keep write ordering. So
378 * here we don't need to wait for node write completion, since we use
379 * node chain which serializes node blocks. If one of node writes are
380 * reordered, we can see simply broken chain, resulting in stopping
381 * roll-forward recovery. It means we'll recover all or none node blocks
382 * given fsync mark.
383 */
384 if (!atomic) {
385 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
386 if (ret)
387 goto out;
388 }
389
390 /* once recovery info is written, don't need to tack this */
391 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
392 clear_inode_flag(inode, FI_APPEND_WRITE);
393 flush_out:
394 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
395 ret = f2fs_issue_flush(sbi, inode->i_ino);
396 if (!ret) {
397 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
398 clear_inode_flag(inode, FI_UPDATE_WRITE);
399 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
400 }
401 f2fs_update_time(sbi, REQ_TIME);
402 out:
403 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
404 return ret;
405 }
406
f2fs_sync_file(struct file * file,loff_t start,loff_t end,int datasync)407 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
408 {
409 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
410 return -EIO;
411 return f2fs_do_sync_file(file, start, end, datasync, false);
412 }
413
__found_offset(struct address_space * mapping,struct dnode_of_data * dn,pgoff_t index,int whence)414 static bool __found_offset(struct address_space *mapping,
415 struct dnode_of_data *dn, pgoff_t index, int whence)
416 {
417 block_t blkaddr = f2fs_data_blkaddr(dn);
418 struct inode *inode = mapping->host;
419 bool compressed_cluster = false;
420
421 if (f2fs_compressed_file(inode)) {
422 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio,
423 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size));
424
425 compressed_cluster = first_blkaddr == COMPRESS_ADDR;
426 }
427
428 switch (whence) {
429 case SEEK_DATA:
430 if (__is_valid_data_blkaddr(blkaddr))
431 return true;
432 if (blkaddr == NEW_ADDR &&
433 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
434 return true;
435 if (compressed_cluster)
436 return true;
437 break;
438 case SEEK_HOLE:
439 if (compressed_cluster)
440 return false;
441 if (blkaddr == NULL_ADDR)
442 return true;
443 break;
444 }
445 return false;
446 }
447
f2fs_seek_block(struct file * file,loff_t offset,int whence)448 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
449 {
450 struct inode *inode = file->f_mapping->host;
451 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
452 struct dnode_of_data dn;
453 pgoff_t pgofs, end_offset;
454 loff_t data_ofs = offset;
455 loff_t isize;
456 int err = 0;
457
458 inode_lock_shared(inode);
459
460 isize = i_size_read(inode);
461 if (offset >= isize)
462 goto fail;
463
464 /* handle inline data case */
465 if (f2fs_has_inline_data(inode)) {
466 if (whence == SEEK_HOLE) {
467 data_ofs = isize;
468 goto found;
469 } else if (whence == SEEK_DATA) {
470 data_ofs = offset;
471 goto found;
472 }
473 }
474
475 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
476
477 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
478 set_new_dnode(&dn, inode, NULL, NULL, 0);
479 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
480 if (err && err != -ENOENT) {
481 goto fail;
482 } else if (err == -ENOENT) {
483 /* direct node does not exists */
484 if (whence == SEEK_DATA) {
485 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
486 continue;
487 } else {
488 goto found;
489 }
490 }
491
492 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
493
494 /* find data/hole in dnode block */
495 for (; dn.ofs_in_node < end_offset;
496 dn.ofs_in_node++, pgofs++,
497 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
498 block_t blkaddr;
499
500 blkaddr = f2fs_data_blkaddr(&dn);
501
502 if (__is_valid_data_blkaddr(blkaddr) &&
503 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
504 blkaddr, DATA_GENERIC_ENHANCE)) {
505 f2fs_put_dnode(&dn);
506 goto fail;
507 }
508
509 if (__found_offset(file->f_mapping, &dn,
510 pgofs, whence)) {
511 f2fs_put_dnode(&dn);
512 goto found;
513 }
514 }
515 f2fs_put_dnode(&dn);
516 }
517
518 if (whence == SEEK_DATA)
519 goto fail;
520 found:
521 if (whence == SEEK_HOLE && data_ofs > isize)
522 data_ofs = isize;
523 inode_unlock_shared(inode);
524 return vfs_setpos(file, data_ofs, maxbytes);
525 fail:
526 inode_unlock_shared(inode);
527 return -ENXIO;
528 }
529
f2fs_llseek(struct file * file,loff_t offset,int whence)530 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
531 {
532 struct inode *inode = file->f_mapping->host;
533 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
534
535 switch (whence) {
536 case SEEK_SET:
537 case SEEK_CUR:
538 case SEEK_END:
539 return generic_file_llseek_size(file, offset, whence,
540 maxbytes, i_size_read(inode));
541 case SEEK_DATA:
542 case SEEK_HOLE:
543 if (offset < 0)
544 return -ENXIO;
545 return f2fs_seek_block(file, offset, whence);
546 }
547
548 return -EINVAL;
549 }
550
f2fs_file_mmap(struct file * file,struct vm_area_struct * vma)551 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
552 {
553 struct inode *inode = file_inode(file);
554
555 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
556 return -EIO;
557
558 if (!f2fs_is_compress_backend_ready(inode))
559 return -EOPNOTSUPP;
560
561 file_accessed(file);
562 vma->vm_ops = &f2fs_file_vm_ops;
563
564 f2fs_down_read(&F2FS_I(inode)->i_sem);
565 set_inode_flag(inode, FI_MMAP_FILE);
566 f2fs_up_read(&F2FS_I(inode)->i_sem);
567
568 return 0;
569 }
570
finish_preallocate_blocks(struct inode * inode)571 static int finish_preallocate_blocks(struct inode *inode)
572 {
573 int ret = 0;
574 bool opened;
575
576 f2fs_down_read(&F2FS_I(inode)->i_sem);
577 opened = is_inode_flag_set(inode, FI_OPENED_FILE);
578 f2fs_up_read(&F2FS_I(inode)->i_sem);
579 if (opened)
580 return 0;
581
582 inode_lock(inode);
583 if (is_inode_flag_set(inode, FI_OPENED_FILE))
584 goto out_unlock;
585
586 if (!file_should_truncate(inode))
587 goto out_update;
588
589 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
590 filemap_invalidate_lock(inode->i_mapping);
591
592 truncate_setsize(inode, i_size_read(inode));
593 ret = f2fs_truncate(inode);
594
595 filemap_invalidate_unlock(inode->i_mapping);
596 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
597 if (ret)
598 goto out_unlock;
599
600 file_dont_truncate(inode);
601 out_update:
602 f2fs_down_write(&F2FS_I(inode)->i_sem);
603 set_inode_flag(inode, FI_OPENED_FILE);
604 f2fs_up_write(&F2FS_I(inode)->i_sem);
605 out_unlock:
606 inode_unlock(inode);
607 return ret;
608 }
609
f2fs_file_open(struct inode * inode,struct file * filp)610 static int f2fs_file_open(struct inode *inode, struct file *filp)
611 {
612 int err = fscrypt_file_open(inode, filp);
613
614 if (err)
615 return err;
616
617 if (!f2fs_is_compress_backend_ready(inode))
618 return -EOPNOTSUPP;
619
620 err = fsverity_file_open(inode, filp);
621 if (err)
622 return err;
623
624 filp->f_mode |= FMODE_NOWAIT;
625 filp->f_mode |= FMODE_CAN_ODIRECT;
626
627 err = dquot_file_open(inode, filp);
628 if (err)
629 return err;
630
631 return finish_preallocate_blocks(inode);
632 }
633
f2fs_truncate_data_blocks_range(struct dnode_of_data * dn,int count)634 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
635 {
636 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
637 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
638 __le32 *addr;
639 bool compressed_cluster = false;
640 int cluster_index = 0, valid_blocks = 0;
641 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
642 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
643 block_t blkstart;
644 int blklen = 0;
645
646 addr = get_dnode_addr(dn->inode, dn->node_folio) + ofs;
647 blkstart = le32_to_cpu(*addr);
648
649 /* Assumption: truncation starts with cluster */
650 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
651 block_t blkaddr = le32_to_cpu(*addr);
652
653 if (f2fs_compressed_file(dn->inode) &&
654 !(cluster_index & (cluster_size - 1))) {
655 if (compressed_cluster)
656 f2fs_i_compr_blocks_update(dn->inode,
657 valid_blocks, false);
658 compressed_cluster = (blkaddr == COMPRESS_ADDR);
659 valid_blocks = 0;
660 }
661
662 if (blkaddr == NULL_ADDR)
663 goto next;
664
665 f2fs_set_data_blkaddr(dn, NULL_ADDR);
666
667 if (__is_valid_data_blkaddr(blkaddr)) {
668 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE))
669 goto next;
670 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr,
671 DATA_GENERIC_ENHANCE))
672 goto next;
673 if (compressed_cluster)
674 valid_blocks++;
675 }
676
677 if (blkstart + blklen == blkaddr) {
678 blklen++;
679 } else {
680 f2fs_invalidate_blocks(sbi, blkstart, blklen);
681 blkstart = blkaddr;
682 blklen = 1;
683 }
684
685 if (!released || blkaddr != COMPRESS_ADDR)
686 nr_free++;
687
688 continue;
689
690 next:
691 if (blklen)
692 f2fs_invalidate_blocks(sbi, blkstart, blklen);
693
694 blkstart = le32_to_cpu(*(addr + 1));
695 blklen = 0;
696 }
697
698 if (blklen)
699 f2fs_invalidate_blocks(sbi, blkstart, blklen);
700
701 if (compressed_cluster)
702 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
703
704 if (nr_free) {
705 pgoff_t fofs;
706 /*
707 * once we invalidate valid blkaddr in range [ofs, ofs + count],
708 * we will invalidate all blkaddr in the whole range.
709 */
710 fofs = f2fs_start_bidx_of_node(ofs_of_node(&dn->node_folio->page),
711 dn->inode) + ofs;
712 f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
713 f2fs_update_age_extent_cache_range(dn, fofs, len);
714 dec_valid_block_count(sbi, dn->inode, nr_free);
715 }
716 dn->ofs_in_node = ofs;
717
718 f2fs_update_time(sbi, REQ_TIME);
719 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
720 dn->ofs_in_node, nr_free);
721 }
722
truncate_partial_data_page(struct inode * inode,u64 from,bool cache_only)723 static int truncate_partial_data_page(struct inode *inode, u64 from,
724 bool cache_only)
725 {
726 loff_t offset = from & (PAGE_SIZE - 1);
727 pgoff_t index = from >> PAGE_SHIFT;
728 struct address_space *mapping = inode->i_mapping;
729 struct folio *folio;
730
731 if (!offset && !cache_only)
732 return 0;
733
734 if (cache_only) {
735 folio = filemap_lock_folio(mapping, index);
736 if (IS_ERR(folio))
737 return 0;
738 if (folio_test_uptodate(folio))
739 goto truncate_out;
740 f2fs_folio_put(folio, true);
741 return 0;
742 }
743
744 folio = f2fs_get_lock_data_folio(inode, index, true);
745 if (IS_ERR(folio))
746 return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio);
747 truncate_out:
748 f2fs_folio_wait_writeback(folio, DATA, true, true);
749 folio_zero_segment(folio, offset, folio_size(folio));
750
751 /* An encrypted inode should have a key and truncate the last page. */
752 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
753 if (!cache_only)
754 folio_mark_dirty(folio);
755 f2fs_folio_put(folio, true);
756 return 0;
757 }
758
f2fs_do_truncate_blocks(struct inode * inode,u64 from,bool lock)759 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
760 {
761 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
762 struct dnode_of_data dn;
763 pgoff_t free_from;
764 int count = 0, err = 0;
765 struct folio *ifolio;
766 bool truncate_page = false;
767
768 trace_f2fs_truncate_blocks_enter(inode, from);
769
770 if (IS_DEVICE_ALIASING(inode) && from) {
771 err = -EINVAL;
772 goto out_err;
773 }
774
775 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
776
777 if (free_from >= max_file_blocks(inode))
778 goto free_partial;
779
780 if (lock)
781 f2fs_lock_op(sbi);
782
783 ifolio = f2fs_get_inode_folio(sbi, inode->i_ino);
784 if (IS_ERR(ifolio)) {
785 err = PTR_ERR(ifolio);
786 goto out;
787 }
788
789 if (IS_DEVICE_ALIASING(inode)) {
790 struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
791 struct extent_info ei = et->largest;
792
793 f2fs_invalidate_blocks(sbi, ei.blk, ei.len);
794
795 dec_valid_block_count(sbi, inode, ei.len);
796 f2fs_update_time(sbi, REQ_TIME);
797
798 f2fs_folio_put(ifolio, true);
799 goto out;
800 }
801
802 if (f2fs_has_inline_data(inode)) {
803 f2fs_truncate_inline_inode(inode, ifolio, from);
804 f2fs_folio_put(ifolio, true);
805 truncate_page = true;
806 goto out;
807 }
808
809 set_new_dnode(&dn, inode, ifolio, NULL, 0);
810 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
811 if (err) {
812 if (err == -ENOENT)
813 goto free_next;
814 goto out;
815 }
816
817 count = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
818
819 count -= dn.ofs_in_node;
820 f2fs_bug_on(sbi, count < 0);
821
822 if (dn.ofs_in_node || IS_INODE(&dn.node_folio->page)) {
823 f2fs_truncate_data_blocks_range(&dn, count);
824 free_from += count;
825 }
826
827 f2fs_put_dnode(&dn);
828 free_next:
829 err = f2fs_truncate_inode_blocks(inode, free_from);
830 out:
831 if (lock)
832 f2fs_unlock_op(sbi);
833 free_partial:
834 /* lastly zero out the first data page */
835 if (!err)
836 err = truncate_partial_data_page(inode, from, truncate_page);
837 out_err:
838 trace_f2fs_truncate_blocks_exit(inode, err);
839 return err;
840 }
841
f2fs_truncate_blocks(struct inode * inode,u64 from,bool lock)842 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
843 {
844 u64 free_from = from;
845 int err;
846
847 #ifdef CONFIG_F2FS_FS_COMPRESSION
848 /*
849 * for compressed file, only support cluster size
850 * aligned truncation.
851 */
852 if (f2fs_compressed_file(inode))
853 free_from = round_up(from,
854 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
855 #endif
856
857 err = f2fs_do_truncate_blocks(inode, free_from, lock);
858 if (err)
859 return err;
860
861 #ifdef CONFIG_F2FS_FS_COMPRESSION
862 /*
863 * For compressed file, after release compress blocks, don't allow write
864 * direct, but we should allow write direct after truncate to zero.
865 */
866 if (f2fs_compressed_file(inode) && !free_from
867 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
868 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
869
870 if (from != free_from) {
871 err = f2fs_truncate_partial_cluster(inode, from, lock);
872 if (err)
873 return err;
874 }
875 #endif
876
877 return 0;
878 }
879
f2fs_truncate(struct inode * inode)880 int f2fs_truncate(struct inode *inode)
881 {
882 int err;
883
884 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
885 return -EIO;
886
887 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
888 S_ISLNK(inode->i_mode)))
889 return 0;
890
891 trace_f2fs_truncate(inode);
892
893 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
894 return -EIO;
895
896 err = f2fs_dquot_initialize(inode);
897 if (err)
898 return err;
899
900 /* we should check inline_data size */
901 if (!f2fs_may_inline_data(inode)) {
902 err = f2fs_convert_inline_inode(inode);
903 if (err)
904 return err;
905 }
906
907 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
908 if (err)
909 return err;
910
911 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
912 f2fs_mark_inode_dirty_sync(inode, false);
913 return 0;
914 }
915
f2fs_force_buffered_io(struct inode * inode,int rw)916 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
917 {
918 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
919
920 if (!fscrypt_dio_supported(inode))
921 return true;
922 if (fsverity_active(inode))
923 return true;
924 if (f2fs_compressed_file(inode))
925 return true;
926 /*
927 * only force direct read to use buffered IO, for direct write,
928 * it expects inline data conversion before committing IO.
929 */
930 if (f2fs_has_inline_data(inode) && rw == READ)
931 return true;
932
933 /* disallow direct IO if any of devices has unaligned blksize */
934 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
935 return true;
936 /*
937 * for blkzoned device, fallback direct IO to buffered IO, so
938 * all IOs can be serialized by log-structured write.
939 */
940 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) &&
941 !f2fs_is_pinned_file(inode))
942 return true;
943 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
944 return true;
945
946 return false;
947 }
948
f2fs_getattr(struct mnt_idmap * idmap,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)949 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path,
950 struct kstat *stat, u32 request_mask, unsigned int query_flags)
951 {
952 struct inode *inode = d_inode(path->dentry);
953 struct f2fs_inode_info *fi = F2FS_I(inode);
954 struct f2fs_inode *ri = NULL;
955 unsigned int flags;
956
957 if (f2fs_has_extra_attr(inode) &&
958 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
959 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
960 stat->result_mask |= STATX_BTIME;
961 stat->btime.tv_sec = fi->i_crtime.tv_sec;
962 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
963 }
964
965 /*
966 * Return the DIO alignment restrictions if requested. We only return
967 * this information when requested, since on encrypted files it might
968 * take a fair bit of work to get if the file wasn't opened recently.
969 *
970 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN
971 * cannot represent that, so in that case we report no DIO support.
972 */
973 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
974 unsigned int bsize = i_blocksize(inode);
975
976 stat->result_mask |= STATX_DIOALIGN;
977 if (!f2fs_force_buffered_io(inode, WRITE)) {
978 stat->dio_mem_align = bsize;
979 stat->dio_offset_align = bsize;
980 }
981 }
982
983 flags = fi->i_flags;
984 if (flags & F2FS_COMPR_FL)
985 stat->attributes |= STATX_ATTR_COMPRESSED;
986 if (flags & F2FS_APPEND_FL)
987 stat->attributes |= STATX_ATTR_APPEND;
988 if (IS_ENCRYPTED(inode))
989 stat->attributes |= STATX_ATTR_ENCRYPTED;
990 if (flags & F2FS_IMMUTABLE_FL)
991 stat->attributes |= STATX_ATTR_IMMUTABLE;
992 if (flags & F2FS_NODUMP_FL)
993 stat->attributes |= STATX_ATTR_NODUMP;
994 if (IS_VERITY(inode))
995 stat->attributes |= STATX_ATTR_VERITY;
996
997 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
998 STATX_ATTR_APPEND |
999 STATX_ATTR_ENCRYPTED |
1000 STATX_ATTR_IMMUTABLE |
1001 STATX_ATTR_NODUMP |
1002 STATX_ATTR_VERITY);
1003
1004 generic_fillattr(idmap, request_mask, inode, stat);
1005
1006 /* we need to show initial sectors used for inline_data/dentries */
1007 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
1008 f2fs_has_inline_dentry(inode))
1009 stat->blocks += (stat->size + 511) >> 9;
1010
1011 return 0;
1012 }
1013
1014 #ifdef CONFIG_F2FS_FS_POSIX_ACL
__setattr_copy(struct mnt_idmap * idmap,struct inode * inode,const struct iattr * attr)1015 static void __setattr_copy(struct mnt_idmap *idmap,
1016 struct inode *inode, const struct iattr *attr)
1017 {
1018 unsigned int ia_valid = attr->ia_valid;
1019
1020 i_uid_update(idmap, attr, inode);
1021 i_gid_update(idmap, attr, inode);
1022 if (ia_valid & ATTR_ATIME)
1023 inode_set_atime_to_ts(inode, attr->ia_atime);
1024 if (ia_valid & ATTR_MTIME)
1025 inode_set_mtime_to_ts(inode, attr->ia_mtime);
1026 if (ia_valid & ATTR_CTIME)
1027 inode_set_ctime_to_ts(inode, attr->ia_ctime);
1028 if (ia_valid & ATTR_MODE) {
1029 umode_t mode = attr->ia_mode;
1030
1031 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode)))
1032 mode &= ~S_ISGID;
1033 set_acl_inode(inode, mode);
1034 }
1035 }
1036 #else
1037 #define __setattr_copy setattr_copy
1038 #endif
1039
f2fs_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * attr)1040 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
1041 struct iattr *attr)
1042 {
1043 struct inode *inode = d_inode(dentry);
1044 struct f2fs_inode_info *fi = F2FS_I(inode);
1045 int err;
1046
1047 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1048 return -EIO;
1049
1050 if (unlikely(IS_IMMUTABLE(inode)))
1051 return -EPERM;
1052
1053 if (unlikely(IS_APPEND(inode) &&
1054 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
1055 ATTR_GID | ATTR_TIMES_SET))))
1056 return -EPERM;
1057
1058 if ((attr->ia_valid & ATTR_SIZE)) {
1059 if (!f2fs_is_compress_backend_ready(inode) ||
1060 IS_DEVICE_ALIASING(inode))
1061 return -EOPNOTSUPP;
1062 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) &&
1063 !IS_ALIGNED(attr->ia_size,
1064 F2FS_BLK_TO_BYTES(fi->i_cluster_size)))
1065 return -EINVAL;
1066 }
1067
1068 err = setattr_prepare(idmap, dentry, attr);
1069 if (err)
1070 return err;
1071
1072 err = fscrypt_prepare_setattr(dentry, attr);
1073 if (err)
1074 return err;
1075
1076 err = fsverity_prepare_setattr(dentry, attr);
1077 if (err)
1078 return err;
1079
1080 if (is_quota_modification(idmap, inode, attr)) {
1081 err = f2fs_dquot_initialize(inode);
1082 if (err)
1083 return err;
1084 }
1085 if (i_uid_needs_update(idmap, attr, inode) ||
1086 i_gid_needs_update(idmap, attr, inode)) {
1087 f2fs_lock_op(F2FS_I_SB(inode));
1088 err = dquot_transfer(idmap, inode, attr);
1089 if (err) {
1090 set_sbi_flag(F2FS_I_SB(inode),
1091 SBI_QUOTA_NEED_REPAIR);
1092 f2fs_unlock_op(F2FS_I_SB(inode));
1093 return err;
1094 }
1095 /*
1096 * update uid/gid under lock_op(), so that dquot and inode can
1097 * be updated atomically.
1098 */
1099 i_uid_update(idmap, attr, inode);
1100 i_gid_update(idmap, attr, inode);
1101 f2fs_mark_inode_dirty_sync(inode, true);
1102 f2fs_unlock_op(F2FS_I_SB(inode));
1103 }
1104
1105 if (attr->ia_valid & ATTR_SIZE) {
1106 loff_t old_size = i_size_read(inode);
1107
1108 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
1109 /*
1110 * should convert inline inode before i_size_write to
1111 * keep smaller than inline_data size with inline flag.
1112 */
1113 err = f2fs_convert_inline_inode(inode);
1114 if (err)
1115 return err;
1116 }
1117
1118 /*
1119 * wait for inflight dio, blocks should be removed after
1120 * IO completion.
1121 */
1122 if (attr->ia_size < old_size)
1123 inode_dio_wait(inode);
1124
1125 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
1126 filemap_invalidate_lock(inode->i_mapping);
1127
1128 if (attr->ia_size > old_size)
1129 f2fs_zero_post_eof_page(inode, attr->ia_size);
1130 truncate_setsize(inode, attr->ia_size);
1131
1132 if (attr->ia_size <= old_size)
1133 err = f2fs_truncate(inode);
1134 /*
1135 * do not trim all blocks after i_size if target size is
1136 * larger than i_size.
1137 */
1138 filemap_invalidate_unlock(inode->i_mapping);
1139 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
1140 if (err)
1141 return err;
1142
1143 spin_lock(&fi->i_size_lock);
1144 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1145 fi->last_disk_size = i_size_read(inode);
1146 spin_unlock(&fi->i_size_lock);
1147 }
1148
1149 __setattr_copy(idmap, inode, attr);
1150
1151 if (attr->ia_valid & ATTR_MODE) {
1152 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
1153
1154 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1155 if (!err)
1156 inode->i_mode = fi->i_acl_mode;
1157 clear_inode_flag(inode, FI_ACL_MODE);
1158 }
1159 }
1160
1161 /* file size may changed here */
1162 f2fs_mark_inode_dirty_sync(inode, true);
1163
1164 /* inode change will produce dirty node pages flushed by checkpoint */
1165 f2fs_balance_fs(F2FS_I_SB(inode), true);
1166
1167 return err;
1168 }
1169
1170 const struct inode_operations f2fs_file_inode_operations = {
1171 .getattr = f2fs_getattr,
1172 .setattr = f2fs_setattr,
1173 .get_inode_acl = f2fs_get_acl,
1174 .set_acl = f2fs_set_acl,
1175 .listxattr = f2fs_listxattr,
1176 .fiemap = f2fs_fiemap,
1177 .fileattr_get = f2fs_fileattr_get,
1178 .fileattr_set = f2fs_fileattr_set,
1179 };
1180
fill_zero(struct inode * inode,pgoff_t index,loff_t start,loff_t len)1181 static int fill_zero(struct inode *inode, pgoff_t index,
1182 loff_t start, loff_t len)
1183 {
1184 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1185 struct folio *folio;
1186
1187 if (!len)
1188 return 0;
1189
1190 f2fs_balance_fs(sbi, true);
1191
1192 f2fs_lock_op(sbi);
1193 folio = f2fs_get_new_data_folio(inode, NULL, index, false);
1194 f2fs_unlock_op(sbi);
1195
1196 if (IS_ERR(folio))
1197 return PTR_ERR(folio);
1198
1199 f2fs_folio_wait_writeback(folio, DATA, true, true);
1200 folio_zero_range(folio, start, len);
1201 folio_mark_dirty(folio);
1202 f2fs_folio_put(folio, true);
1203 return 0;
1204 }
1205
f2fs_truncate_hole(struct inode * inode,pgoff_t pg_start,pgoff_t pg_end)1206 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1207 {
1208 int err;
1209
1210 while (pg_start < pg_end) {
1211 struct dnode_of_data dn;
1212 pgoff_t end_offset, count;
1213
1214 set_new_dnode(&dn, inode, NULL, NULL, 0);
1215 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1216 if (err) {
1217 if (err == -ENOENT) {
1218 pg_start = f2fs_get_next_page_offset(&dn,
1219 pg_start);
1220 continue;
1221 }
1222 return err;
1223 }
1224
1225 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
1226 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1227
1228 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1229
1230 f2fs_truncate_data_blocks_range(&dn, count);
1231 f2fs_put_dnode(&dn);
1232
1233 pg_start += count;
1234 }
1235 return 0;
1236 }
1237
f2fs_punch_hole(struct inode * inode,loff_t offset,loff_t len)1238 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
1239 {
1240 pgoff_t pg_start, pg_end;
1241 loff_t off_start, off_end;
1242 int ret;
1243
1244 ret = f2fs_convert_inline_inode(inode);
1245 if (ret)
1246 return ret;
1247
1248 filemap_invalidate_lock(inode->i_mapping);
1249 f2fs_zero_post_eof_page(inode, offset + len);
1250 filemap_invalidate_unlock(inode->i_mapping);
1251
1252 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1253 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1254
1255 off_start = offset & (PAGE_SIZE - 1);
1256 off_end = (offset + len) & (PAGE_SIZE - 1);
1257
1258 if (pg_start == pg_end) {
1259 ret = fill_zero(inode, pg_start, off_start,
1260 off_end - off_start);
1261 if (ret)
1262 return ret;
1263 } else {
1264 if (off_start) {
1265 ret = fill_zero(inode, pg_start++, off_start,
1266 PAGE_SIZE - off_start);
1267 if (ret)
1268 return ret;
1269 }
1270 if (off_end) {
1271 ret = fill_zero(inode, pg_end, 0, off_end);
1272 if (ret)
1273 return ret;
1274 }
1275
1276 if (pg_start < pg_end) {
1277 loff_t blk_start, blk_end;
1278 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1279
1280 f2fs_balance_fs(sbi, true);
1281
1282 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1283 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1284
1285 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1286 filemap_invalidate_lock(inode->i_mapping);
1287
1288 truncate_pagecache_range(inode, blk_start, blk_end - 1);
1289
1290 f2fs_lock_op(sbi);
1291 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1292 f2fs_unlock_op(sbi);
1293
1294 filemap_invalidate_unlock(inode->i_mapping);
1295 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1296 }
1297 }
1298
1299 return ret;
1300 }
1301
__read_out_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,pgoff_t len)1302 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1303 int *do_replace, pgoff_t off, pgoff_t len)
1304 {
1305 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1306 struct dnode_of_data dn;
1307 int ret, done, i;
1308
1309 next_dnode:
1310 set_new_dnode(&dn, inode, NULL, NULL, 0);
1311 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1312 if (ret && ret != -ENOENT) {
1313 return ret;
1314 } else if (ret == -ENOENT) {
1315 if (dn.max_level == 0)
1316 return -ENOENT;
1317 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1318 dn.ofs_in_node, len);
1319 blkaddr += done;
1320 do_replace += done;
1321 goto next;
1322 }
1323
1324 done = min((pgoff_t)ADDRS_PER_PAGE(&dn.node_folio->page, inode) -
1325 dn.ofs_in_node, len);
1326 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1327 *blkaddr = f2fs_data_blkaddr(&dn);
1328
1329 if (__is_valid_data_blkaddr(*blkaddr) &&
1330 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1331 DATA_GENERIC_ENHANCE)) {
1332 f2fs_put_dnode(&dn);
1333 return -EFSCORRUPTED;
1334 }
1335
1336 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1337
1338 if (f2fs_lfs_mode(sbi)) {
1339 f2fs_put_dnode(&dn);
1340 return -EOPNOTSUPP;
1341 }
1342
1343 /* do not invalidate this block address */
1344 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1345 *do_replace = 1;
1346 }
1347 }
1348 f2fs_put_dnode(&dn);
1349 next:
1350 len -= done;
1351 off += done;
1352 if (len)
1353 goto next_dnode;
1354 return 0;
1355 }
1356
__roll_back_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,int len)1357 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1358 int *do_replace, pgoff_t off, int len)
1359 {
1360 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1361 struct dnode_of_data dn;
1362 int ret, i;
1363
1364 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1365 if (*do_replace == 0)
1366 continue;
1367
1368 set_new_dnode(&dn, inode, NULL, NULL, 0);
1369 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1370 if (ret) {
1371 dec_valid_block_count(sbi, inode, 1);
1372 f2fs_invalidate_blocks(sbi, *blkaddr, 1);
1373 } else {
1374 f2fs_update_data_blkaddr(&dn, *blkaddr);
1375 }
1376 f2fs_put_dnode(&dn);
1377 }
1378 return 0;
1379 }
1380
__clone_blkaddrs(struct inode * src_inode,struct inode * dst_inode,block_t * blkaddr,int * do_replace,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1381 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1382 block_t *blkaddr, int *do_replace,
1383 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1384 {
1385 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1386 pgoff_t i = 0;
1387 int ret;
1388
1389 while (i < len) {
1390 if (blkaddr[i] == NULL_ADDR && !full) {
1391 i++;
1392 continue;
1393 }
1394
1395 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1396 struct dnode_of_data dn;
1397 struct node_info ni;
1398 size_t new_size;
1399 pgoff_t ilen;
1400
1401 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1402 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1403 if (ret)
1404 return ret;
1405
1406 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1407 if (ret) {
1408 f2fs_put_dnode(&dn);
1409 return ret;
1410 }
1411
1412 ilen = min((pgoff_t)
1413 ADDRS_PER_PAGE(&dn.node_folio->page, dst_inode) -
1414 dn.ofs_in_node, len - i);
1415 do {
1416 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1417 f2fs_truncate_data_blocks_range(&dn, 1);
1418
1419 if (do_replace[i]) {
1420 f2fs_i_blocks_write(src_inode,
1421 1, false, false);
1422 f2fs_i_blocks_write(dst_inode,
1423 1, true, false);
1424 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1425 blkaddr[i], ni.version, true, false);
1426
1427 do_replace[i] = 0;
1428 }
1429 dn.ofs_in_node++;
1430 i++;
1431 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1432 if (dst_inode->i_size < new_size)
1433 f2fs_i_size_write(dst_inode, new_size);
1434 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1435
1436 f2fs_put_dnode(&dn);
1437 } else {
1438 struct folio *fsrc, *fdst;
1439
1440 fsrc = f2fs_get_lock_data_folio(src_inode,
1441 src + i, true);
1442 if (IS_ERR(fsrc))
1443 return PTR_ERR(fsrc);
1444 fdst = f2fs_get_new_data_folio(dst_inode, NULL, dst + i,
1445 true);
1446 if (IS_ERR(fdst)) {
1447 f2fs_folio_put(fsrc, true);
1448 return PTR_ERR(fdst);
1449 }
1450
1451 f2fs_folio_wait_writeback(fdst, DATA, true, true);
1452
1453 memcpy_folio(fdst, 0, fsrc, 0, PAGE_SIZE);
1454 folio_mark_dirty(fdst);
1455 set_page_private_gcing(&fdst->page);
1456 f2fs_folio_put(fdst, true);
1457 f2fs_folio_put(fsrc, true);
1458
1459 ret = f2fs_truncate_hole(src_inode,
1460 src + i, src + i + 1);
1461 if (ret)
1462 return ret;
1463 i++;
1464 }
1465 }
1466 return 0;
1467 }
1468
__exchange_data_block(struct inode * src_inode,struct inode * dst_inode,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1469 static int __exchange_data_block(struct inode *src_inode,
1470 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1471 pgoff_t len, bool full)
1472 {
1473 block_t *src_blkaddr;
1474 int *do_replace;
1475 pgoff_t olen;
1476 int ret;
1477
1478 while (len) {
1479 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1480
1481 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1482 array_size(olen, sizeof(block_t)),
1483 GFP_NOFS);
1484 if (!src_blkaddr)
1485 return -ENOMEM;
1486
1487 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1488 array_size(olen, sizeof(int)),
1489 GFP_NOFS);
1490 if (!do_replace) {
1491 kvfree(src_blkaddr);
1492 return -ENOMEM;
1493 }
1494
1495 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1496 do_replace, src, olen);
1497 if (ret)
1498 goto roll_back;
1499
1500 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1501 do_replace, src, dst, olen, full);
1502 if (ret)
1503 goto roll_back;
1504
1505 src += olen;
1506 dst += olen;
1507 len -= olen;
1508
1509 kvfree(src_blkaddr);
1510 kvfree(do_replace);
1511 }
1512 return 0;
1513
1514 roll_back:
1515 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1516 kvfree(src_blkaddr);
1517 kvfree(do_replace);
1518 return ret;
1519 }
1520
f2fs_do_collapse(struct inode * inode,loff_t offset,loff_t len)1521 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1522 {
1523 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1524 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1525 pgoff_t start = offset >> PAGE_SHIFT;
1526 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1527 int ret;
1528
1529 f2fs_balance_fs(sbi, true);
1530
1531 /* avoid gc operation during block exchange */
1532 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1533 filemap_invalidate_lock(inode->i_mapping);
1534
1535 f2fs_zero_post_eof_page(inode, offset + len);
1536
1537 f2fs_lock_op(sbi);
1538 f2fs_drop_extent_tree(inode);
1539 truncate_pagecache(inode, offset);
1540 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1541 f2fs_unlock_op(sbi);
1542
1543 filemap_invalidate_unlock(inode->i_mapping);
1544 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1545 return ret;
1546 }
1547
f2fs_collapse_range(struct inode * inode,loff_t offset,loff_t len)1548 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1549 {
1550 loff_t new_size;
1551 int ret;
1552
1553 if (offset + len >= i_size_read(inode))
1554 return -EINVAL;
1555
1556 /* collapse range should be aligned to block size of f2fs. */
1557 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1558 return -EINVAL;
1559
1560 ret = f2fs_convert_inline_inode(inode);
1561 if (ret)
1562 return ret;
1563
1564 /* write out all dirty pages from offset */
1565 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1566 if (ret)
1567 return ret;
1568
1569 ret = f2fs_do_collapse(inode, offset, len);
1570 if (ret)
1571 return ret;
1572
1573 /* write out all moved pages, if possible */
1574 filemap_invalidate_lock(inode->i_mapping);
1575 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1576 truncate_pagecache(inode, offset);
1577
1578 new_size = i_size_read(inode) - len;
1579 ret = f2fs_truncate_blocks(inode, new_size, true);
1580 filemap_invalidate_unlock(inode->i_mapping);
1581 if (!ret)
1582 f2fs_i_size_write(inode, new_size);
1583 return ret;
1584 }
1585
f2fs_do_zero_range(struct dnode_of_data * dn,pgoff_t start,pgoff_t end)1586 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1587 pgoff_t end)
1588 {
1589 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1590 pgoff_t index = start;
1591 unsigned int ofs_in_node = dn->ofs_in_node;
1592 blkcnt_t count = 0;
1593 int ret;
1594
1595 for (; index < end; index++, dn->ofs_in_node++) {
1596 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1597 count++;
1598 }
1599
1600 dn->ofs_in_node = ofs_in_node;
1601 ret = f2fs_reserve_new_blocks(dn, count);
1602 if (ret)
1603 return ret;
1604
1605 dn->ofs_in_node = ofs_in_node;
1606 for (index = start; index < end; index++, dn->ofs_in_node++) {
1607 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1608 /*
1609 * f2fs_reserve_new_blocks will not guarantee entire block
1610 * allocation.
1611 */
1612 if (dn->data_blkaddr == NULL_ADDR) {
1613 ret = -ENOSPC;
1614 break;
1615 }
1616
1617 if (dn->data_blkaddr == NEW_ADDR)
1618 continue;
1619
1620 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1621 DATA_GENERIC_ENHANCE)) {
1622 ret = -EFSCORRUPTED;
1623 break;
1624 }
1625
1626 f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1);
1627 f2fs_set_data_blkaddr(dn, NEW_ADDR);
1628 }
1629
1630 f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
1631 f2fs_update_age_extent_cache_range(dn, start, index - start);
1632
1633 return ret;
1634 }
1635
f2fs_zero_range(struct inode * inode,loff_t offset,loff_t len,int mode)1636 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1637 int mode)
1638 {
1639 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1640 struct address_space *mapping = inode->i_mapping;
1641 pgoff_t index, pg_start, pg_end;
1642 loff_t new_size = i_size_read(inode);
1643 loff_t off_start, off_end;
1644 int ret = 0;
1645
1646 ret = inode_newsize_ok(inode, (len + offset));
1647 if (ret)
1648 return ret;
1649
1650 ret = f2fs_convert_inline_inode(inode);
1651 if (ret)
1652 return ret;
1653
1654 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1655 if (ret)
1656 return ret;
1657
1658 filemap_invalidate_lock(mapping);
1659 f2fs_zero_post_eof_page(inode, offset + len);
1660 filemap_invalidate_unlock(mapping);
1661
1662 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1663 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1664
1665 off_start = offset & (PAGE_SIZE - 1);
1666 off_end = (offset + len) & (PAGE_SIZE - 1);
1667
1668 if (pg_start == pg_end) {
1669 ret = fill_zero(inode, pg_start, off_start,
1670 off_end - off_start);
1671 if (ret)
1672 return ret;
1673
1674 new_size = max_t(loff_t, new_size, offset + len);
1675 } else {
1676 if (off_start) {
1677 ret = fill_zero(inode, pg_start++, off_start,
1678 PAGE_SIZE - off_start);
1679 if (ret)
1680 return ret;
1681
1682 new_size = max_t(loff_t, new_size,
1683 (loff_t)pg_start << PAGE_SHIFT);
1684 }
1685
1686 for (index = pg_start; index < pg_end;) {
1687 struct dnode_of_data dn;
1688 unsigned int end_offset;
1689 pgoff_t end;
1690
1691 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1692 filemap_invalidate_lock(mapping);
1693
1694 truncate_pagecache_range(inode,
1695 (loff_t)index << PAGE_SHIFT,
1696 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1697
1698 f2fs_lock_op(sbi);
1699
1700 set_new_dnode(&dn, inode, NULL, NULL, 0);
1701 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1702 if (ret) {
1703 f2fs_unlock_op(sbi);
1704 filemap_invalidate_unlock(mapping);
1705 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1706 goto out;
1707 }
1708
1709 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
1710 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1711
1712 ret = f2fs_do_zero_range(&dn, index, end);
1713 f2fs_put_dnode(&dn);
1714
1715 f2fs_unlock_op(sbi);
1716 filemap_invalidate_unlock(mapping);
1717 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1718
1719 f2fs_balance_fs(sbi, dn.node_changed);
1720
1721 if (ret)
1722 goto out;
1723
1724 index = end;
1725 new_size = max_t(loff_t, new_size,
1726 (loff_t)index << PAGE_SHIFT);
1727 }
1728
1729 if (off_end) {
1730 ret = fill_zero(inode, pg_end, 0, off_end);
1731 if (ret)
1732 goto out;
1733
1734 new_size = max_t(loff_t, new_size, offset + len);
1735 }
1736 }
1737
1738 out:
1739 if (new_size > i_size_read(inode)) {
1740 if (mode & FALLOC_FL_KEEP_SIZE)
1741 file_set_keep_isize(inode);
1742 else
1743 f2fs_i_size_write(inode, new_size);
1744 }
1745 return ret;
1746 }
1747
f2fs_insert_range(struct inode * inode,loff_t offset,loff_t len)1748 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1749 {
1750 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1751 struct address_space *mapping = inode->i_mapping;
1752 pgoff_t nr, pg_start, pg_end, delta, idx;
1753 loff_t new_size;
1754 int ret = 0;
1755
1756 new_size = i_size_read(inode) + len;
1757 ret = inode_newsize_ok(inode, new_size);
1758 if (ret)
1759 return ret;
1760
1761 if (offset >= i_size_read(inode))
1762 return -EINVAL;
1763
1764 /* insert range should be aligned to block size of f2fs. */
1765 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1766 return -EINVAL;
1767
1768 ret = f2fs_convert_inline_inode(inode);
1769 if (ret)
1770 return ret;
1771
1772 f2fs_balance_fs(sbi, true);
1773
1774 filemap_invalidate_lock(mapping);
1775 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1776 filemap_invalidate_unlock(mapping);
1777 if (ret)
1778 return ret;
1779
1780 /* write out all dirty pages from offset */
1781 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1782 if (ret)
1783 return ret;
1784
1785 pg_start = offset >> PAGE_SHIFT;
1786 pg_end = (offset + len) >> PAGE_SHIFT;
1787 delta = pg_end - pg_start;
1788 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1789
1790 /* avoid gc operation during block exchange */
1791 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1792 filemap_invalidate_lock(mapping);
1793
1794 f2fs_zero_post_eof_page(inode, offset + len);
1795 truncate_pagecache(inode, offset);
1796
1797 while (!ret && idx > pg_start) {
1798 nr = idx - pg_start;
1799 if (nr > delta)
1800 nr = delta;
1801 idx -= nr;
1802
1803 f2fs_lock_op(sbi);
1804 f2fs_drop_extent_tree(inode);
1805
1806 ret = __exchange_data_block(inode, inode, idx,
1807 idx + delta, nr, false);
1808 f2fs_unlock_op(sbi);
1809 }
1810 filemap_invalidate_unlock(mapping);
1811 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1812 if (ret)
1813 return ret;
1814
1815 /* write out all moved pages, if possible */
1816 filemap_invalidate_lock(mapping);
1817 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1818 truncate_pagecache(inode, offset);
1819 filemap_invalidate_unlock(mapping);
1820
1821 if (!ret)
1822 f2fs_i_size_write(inode, new_size);
1823 return ret;
1824 }
1825
f2fs_expand_inode_data(struct inode * inode,loff_t offset,loff_t len,int mode)1826 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
1827 loff_t len, int mode)
1828 {
1829 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1830 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1831 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1832 .m_may_create = true };
1833 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1834 .init_gc_type = FG_GC,
1835 .should_migrate_blocks = false,
1836 .err_gc_skipped = true,
1837 .nr_free_secs = 0 };
1838 pgoff_t pg_start, pg_end;
1839 loff_t new_size;
1840 loff_t off_end;
1841 block_t expanded = 0;
1842 int err;
1843
1844 err = inode_newsize_ok(inode, (len + offset));
1845 if (err)
1846 return err;
1847
1848 err = f2fs_convert_inline_inode(inode);
1849 if (err)
1850 return err;
1851
1852 filemap_invalidate_lock(inode->i_mapping);
1853 f2fs_zero_post_eof_page(inode, offset + len);
1854 filemap_invalidate_unlock(inode->i_mapping);
1855
1856 f2fs_balance_fs(sbi, true);
1857
1858 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1859 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1860 off_end = (offset + len) & (PAGE_SIZE - 1);
1861
1862 map.m_lblk = pg_start;
1863 map.m_len = pg_end - pg_start;
1864 if (off_end)
1865 map.m_len++;
1866
1867 if (!map.m_len)
1868 return 0;
1869
1870 if (f2fs_is_pinned_file(inode)) {
1871 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1872 block_t sec_len = roundup(map.m_len, sec_blks);
1873
1874 map.m_len = sec_blks;
1875 next_alloc:
1876 f2fs_down_write(&sbi->pin_sem);
1877
1878 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1879 if (has_not_enough_free_secs(sbi, 0, 0)) {
1880 f2fs_up_write(&sbi->pin_sem);
1881 err = -ENOSPC;
1882 f2fs_warn_ratelimited(sbi,
1883 "ino:%lu, start:%lu, end:%lu, need to trigger GC to "
1884 "reclaim enough free segment when checkpoint is enabled",
1885 inode->i_ino, pg_start, pg_end);
1886 goto out_err;
1887 }
1888 }
1889
1890 if (has_not_enough_free_secs(sbi, 0, f2fs_sb_has_blkzoned(sbi) ?
1891 ZONED_PIN_SEC_REQUIRED_COUNT :
1892 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1893 f2fs_down_write(&sbi->gc_lock);
1894 stat_inc_gc_call_count(sbi, FOREGROUND);
1895 err = f2fs_gc(sbi, &gc_control);
1896 if (err && err != -ENODATA) {
1897 f2fs_up_write(&sbi->pin_sem);
1898 goto out_err;
1899 }
1900 }
1901
1902 err = f2fs_allocate_pinning_section(sbi);
1903 if (err) {
1904 f2fs_up_write(&sbi->pin_sem);
1905 goto out_err;
1906 }
1907
1908 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1909 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
1910 file_dont_truncate(inode);
1911
1912 f2fs_up_write(&sbi->pin_sem);
1913
1914 expanded += map.m_len;
1915 sec_len -= map.m_len;
1916 map.m_lblk += map.m_len;
1917 if (!err && sec_len)
1918 goto next_alloc;
1919
1920 map.m_len = expanded;
1921 } else {
1922 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
1923 expanded = map.m_len;
1924 }
1925 out_err:
1926 if (err) {
1927 pgoff_t last_off;
1928
1929 if (!expanded)
1930 return err;
1931
1932 last_off = pg_start + expanded - 1;
1933
1934 /* update new size to the failed position */
1935 new_size = (last_off == pg_end) ? offset + len :
1936 (loff_t)(last_off + 1) << PAGE_SHIFT;
1937 } else {
1938 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1939 }
1940
1941 if (new_size > i_size_read(inode)) {
1942 if (mode & FALLOC_FL_KEEP_SIZE)
1943 file_set_keep_isize(inode);
1944 else
1945 f2fs_i_size_write(inode, new_size);
1946 }
1947
1948 return err;
1949 }
1950
f2fs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)1951 static long f2fs_fallocate(struct file *file, int mode,
1952 loff_t offset, loff_t len)
1953 {
1954 struct inode *inode = file_inode(file);
1955 long ret = 0;
1956
1957 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1958 return -EIO;
1959 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1960 return -ENOSPC;
1961 if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode))
1962 return -EOPNOTSUPP;
1963
1964 /* f2fs only support ->fallocate for regular file */
1965 if (!S_ISREG(inode->i_mode))
1966 return -EINVAL;
1967
1968 if (IS_ENCRYPTED(inode) &&
1969 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1970 return -EOPNOTSUPP;
1971
1972 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1973 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1974 FALLOC_FL_INSERT_RANGE))
1975 return -EOPNOTSUPP;
1976
1977 inode_lock(inode);
1978
1979 /*
1980 * Pinned file should not support partial truncation since the block
1981 * can be used by applications.
1982 */
1983 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1984 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1985 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) {
1986 ret = -EOPNOTSUPP;
1987 goto out;
1988 }
1989
1990 ret = file_modified(file);
1991 if (ret)
1992 goto out;
1993
1994 /*
1995 * wait for inflight dio, blocks should be removed after IO
1996 * completion.
1997 */
1998 inode_dio_wait(inode);
1999
2000 if (mode & FALLOC_FL_PUNCH_HOLE) {
2001 if (offset >= inode->i_size)
2002 goto out;
2003
2004 ret = f2fs_punch_hole(inode, offset, len);
2005 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
2006 ret = f2fs_collapse_range(inode, offset, len);
2007 } else if (mode & FALLOC_FL_ZERO_RANGE) {
2008 ret = f2fs_zero_range(inode, offset, len, mode);
2009 } else if (mode & FALLOC_FL_INSERT_RANGE) {
2010 ret = f2fs_insert_range(inode, offset, len);
2011 } else {
2012 ret = f2fs_expand_inode_data(inode, offset, len, mode);
2013 }
2014
2015 if (!ret) {
2016 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2017 f2fs_mark_inode_dirty_sync(inode, false);
2018 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2019 }
2020
2021 out:
2022 inode_unlock(inode);
2023
2024 trace_f2fs_fallocate(inode, mode, offset, len, ret);
2025 return ret;
2026 }
2027
f2fs_release_file(struct inode * inode,struct file * filp)2028 static int f2fs_release_file(struct inode *inode, struct file *filp)
2029 {
2030 /*
2031 * f2fs_release_file is called at every close calls. So we should
2032 * not drop any inmemory pages by close called by other process.
2033 */
2034 if (!(filp->f_mode & FMODE_WRITE) ||
2035 atomic_read(&inode->i_writecount) != 1)
2036 return 0;
2037
2038 inode_lock(inode);
2039 f2fs_abort_atomic_write(inode, true);
2040 inode_unlock(inode);
2041
2042 return 0;
2043 }
2044
f2fs_file_flush(struct file * file,fl_owner_t id)2045 static int f2fs_file_flush(struct file *file, fl_owner_t id)
2046 {
2047 struct inode *inode = file_inode(file);
2048
2049 /*
2050 * If the process doing a transaction is crashed, we should do
2051 * roll-back. Otherwise, other reader/write can see corrupted database
2052 * until all the writers close its file. Since this should be done
2053 * before dropping file lock, it needs to do in ->flush.
2054 */
2055 if (F2FS_I(inode)->atomic_write_task == current &&
2056 (current->flags & PF_EXITING)) {
2057 inode_lock(inode);
2058 f2fs_abort_atomic_write(inode, true);
2059 inode_unlock(inode);
2060 }
2061
2062 return 0;
2063 }
2064
f2fs_setflags_common(struct inode * inode,u32 iflags,u32 mask)2065 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
2066 {
2067 struct f2fs_inode_info *fi = F2FS_I(inode);
2068 u32 masked_flags = fi->i_flags & mask;
2069
2070 /* mask can be shrunk by flags_valid selector */
2071 iflags &= mask;
2072
2073 /* Is it quota file? Do not allow user to mess with it */
2074 if (IS_NOQUOTA(inode))
2075 return -EPERM;
2076
2077 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
2078 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
2079 return -EOPNOTSUPP;
2080 if (!f2fs_empty_dir(inode))
2081 return -ENOTEMPTY;
2082 }
2083
2084 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
2085 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
2086 return -EOPNOTSUPP;
2087 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
2088 return -EINVAL;
2089 }
2090
2091 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
2092 if (masked_flags & F2FS_COMPR_FL) {
2093 if (!f2fs_disable_compressed_file(inode))
2094 return -EINVAL;
2095 } else {
2096 /* try to convert inline_data to support compression */
2097 int err = f2fs_convert_inline_inode(inode);
2098 if (err)
2099 return err;
2100
2101 f2fs_down_write(&fi->i_sem);
2102 if (!f2fs_may_compress(inode) ||
2103 (S_ISREG(inode->i_mode) &&
2104 F2FS_HAS_BLOCKS(inode))) {
2105 f2fs_up_write(&fi->i_sem);
2106 return -EINVAL;
2107 }
2108 err = set_compress_context(inode);
2109 f2fs_up_write(&fi->i_sem);
2110
2111 if (err)
2112 return err;
2113 }
2114 }
2115
2116 fi->i_flags = iflags | (fi->i_flags & ~mask);
2117 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
2118 (fi->i_flags & F2FS_NOCOMP_FL));
2119
2120 if (fi->i_flags & F2FS_PROJINHERIT_FL)
2121 set_inode_flag(inode, FI_PROJ_INHERIT);
2122 else
2123 clear_inode_flag(inode, FI_PROJ_INHERIT);
2124
2125 inode_set_ctime_current(inode);
2126 f2fs_set_inode_flags(inode);
2127 f2fs_mark_inode_dirty_sync(inode, true);
2128 return 0;
2129 }
2130
2131 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
2132
2133 /*
2134 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
2135 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
2136 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
2137 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
2138 *
2139 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
2140 * FS_IOC_FSSETXATTR is done by the VFS.
2141 */
2142
2143 static const struct {
2144 u32 iflag;
2145 u32 fsflag;
2146 } f2fs_fsflags_map[] = {
2147 { F2FS_COMPR_FL, FS_COMPR_FL },
2148 { F2FS_SYNC_FL, FS_SYNC_FL },
2149 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
2150 { F2FS_APPEND_FL, FS_APPEND_FL },
2151 { F2FS_NODUMP_FL, FS_NODUMP_FL },
2152 { F2FS_NOATIME_FL, FS_NOATIME_FL },
2153 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
2154 { F2FS_INDEX_FL, FS_INDEX_FL },
2155 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
2156 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
2157 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
2158 };
2159
2160 #define F2FS_GETTABLE_FS_FL ( \
2161 FS_COMPR_FL | \
2162 FS_SYNC_FL | \
2163 FS_IMMUTABLE_FL | \
2164 FS_APPEND_FL | \
2165 FS_NODUMP_FL | \
2166 FS_NOATIME_FL | \
2167 FS_NOCOMP_FL | \
2168 FS_INDEX_FL | \
2169 FS_DIRSYNC_FL | \
2170 FS_PROJINHERIT_FL | \
2171 FS_ENCRYPT_FL | \
2172 FS_INLINE_DATA_FL | \
2173 FS_NOCOW_FL | \
2174 FS_VERITY_FL | \
2175 FS_CASEFOLD_FL)
2176
2177 #define F2FS_SETTABLE_FS_FL ( \
2178 FS_COMPR_FL | \
2179 FS_SYNC_FL | \
2180 FS_IMMUTABLE_FL | \
2181 FS_APPEND_FL | \
2182 FS_NODUMP_FL | \
2183 FS_NOATIME_FL | \
2184 FS_NOCOMP_FL | \
2185 FS_DIRSYNC_FL | \
2186 FS_PROJINHERIT_FL | \
2187 FS_CASEFOLD_FL)
2188
2189 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
f2fs_iflags_to_fsflags(u32 iflags)2190 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2191 {
2192 u32 fsflags = 0;
2193 int i;
2194
2195 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2196 if (iflags & f2fs_fsflags_map[i].iflag)
2197 fsflags |= f2fs_fsflags_map[i].fsflag;
2198
2199 return fsflags;
2200 }
2201
2202 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
f2fs_fsflags_to_iflags(u32 fsflags)2203 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2204 {
2205 u32 iflags = 0;
2206 int i;
2207
2208 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2209 if (fsflags & f2fs_fsflags_map[i].fsflag)
2210 iflags |= f2fs_fsflags_map[i].iflag;
2211
2212 return iflags;
2213 }
2214
f2fs_ioc_getversion(struct file * filp,unsigned long arg)2215 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2216 {
2217 struct inode *inode = file_inode(filp);
2218
2219 return put_user(inode->i_generation, (int __user *)arg);
2220 }
2221
f2fs_ioc_start_atomic_write(struct file * filp,bool truncate)2222 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2223 {
2224 struct inode *inode = file_inode(filp);
2225 struct mnt_idmap *idmap = file_mnt_idmap(filp);
2226 struct f2fs_inode_info *fi = F2FS_I(inode);
2227 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2228 loff_t isize;
2229 int ret;
2230
2231 if (!(filp->f_mode & FMODE_WRITE))
2232 return -EBADF;
2233
2234 if (!inode_owner_or_capable(idmap, inode))
2235 return -EACCES;
2236
2237 if (!S_ISREG(inode->i_mode))
2238 return -EINVAL;
2239
2240 if (filp->f_flags & O_DIRECT)
2241 return -EINVAL;
2242
2243 ret = mnt_want_write_file(filp);
2244 if (ret)
2245 return ret;
2246
2247 inode_lock(inode);
2248
2249 if (!f2fs_disable_compressed_file(inode) ||
2250 f2fs_is_pinned_file(inode)) {
2251 ret = -EINVAL;
2252 goto out;
2253 }
2254
2255 if (f2fs_is_atomic_file(inode))
2256 goto out;
2257
2258 ret = f2fs_convert_inline_inode(inode);
2259 if (ret)
2260 goto out;
2261
2262 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2263 f2fs_down_write(&fi->i_gc_rwsem[READ]);
2264
2265 /*
2266 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2267 * f2fs_is_atomic_file.
2268 */
2269 if (get_dirty_pages(inode))
2270 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2271 inode->i_ino, get_dirty_pages(inode));
2272 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2273 if (ret)
2274 goto out_unlock;
2275
2276 /* Check if the inode already has a COW inode */
2277 if (fi->cow_inode == NULL) {
2278 /* Create a COW inode for atomic write */
2279 struct dentry *dentry = file_dentry(filp);
2280 struct inode *dir = d_inode(dentry->d_parent);
2281
2282 ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode);
2283 if (ret)
2284 goto out_unlock;
2285
2286 set_inode_flag(fi->cow_inode, FI_COW_FILE);
2287 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2288
2289 /* Set the COW inode's atomic_inode to the atomic inode */
2290 F2FS_I(fi->cow_inode)->atomic_inode = inode;
2291 } else {
2292 /* Reuse the already created COW inode */
2293 f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode));
2294
2295 invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1);
2296
2297 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2298 if (ret)
2299 goto out_unlock;
2300 }
2301
2302 f2fs_write_inode(inode, NULL);
2303
2304 stat_inc_atomic_inode(inode);
2305
2306 set_inode_flag(inode, FI_ATOMIC_FILE);
2307
2308 isize = i_size_read(inode);
2309 fi->original_i_size = isize;
2310 if (truncate) {
2311 set_inode_flag(inode, FI_ATOMIC_REPLACE);
2312 truncate_inode_pages_final(inode->i_mapping);
2313 f2fs_i_size_write(inode, 0);
2314 isize = 0;
2315 }
2316 f2fs_i_size_write(fi->cow_inode, isize);
2317
2318 out_unlock:
2319 f2fs_up_write(&fi->i_gc_rwsem[READ]);
2320 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2321 if (ret)
2322 goto out;
2323
2324 f2fs_update_time(sbi, REQ_TIME);
2325 fi->atomic_write_task = current;
2326 stat_update_max_atomic_write(inode);
2327 fi->atomic_write_cnt = 0;
2328 out:
2329 inode_unlock(inode);
2330 mnt_drop_write_file(filp);
2331 return ret;
2332 }
2333
f2fs_ioc_commit_atomic_write(struct file * filp)2334 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2335 {
2336 struct inode *inode = file_inode(filp);
2337 struct mnt_idmap *idmap = file_mnt_idmap(filp);
2338 int ret;
2339
2340 if (!(filp->f_mode & FMODE_WRITE))
2341 return -EBADF;
2342
2343 if (!inode_owner_or_capable(idmap, inode))
2344 return -EACCES;
2345
2346 ret = mnt_want_write_file(filp);
2347 if (ret)
2348 return ret;
2349
2350 f2fs_balance_fs(F2FS_I_SB(inode), true);
2351
2352 inode_lock(inode);
2353
2354 if (f2fs_is_atomic_file(inode)) {
2355 ret = f2fs_commit_atomic_write(inode);
2356 if (!ret)
2357 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2358
2359 f2fs_abort_atomic_write(inode, ret);
2360 } else {
2361 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2362 }
2363
2364 inode_unlock(inode);
2365 mnt_drop_write_file(filp);
2366 return ret;
2367 }
2368
f2fs_ioc_abort_atomic_write(struct file * filp)2369 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2370 {
2371 struct inode *inode = file_inode(filp);
2372 struct mnt_idmap *idmap = file_mnt_idmap(filp);
2373 int ret;
2374
2375 if (!(filp->f_mode & FMODE_WRITE))
2376 return -EBADF;
2377
2378 if (!inode_owner_or_capable(idmap, inode))
2379 return -EACCES;
2380
2381 ret = mnt_want_write_file(filp);
2382 if (ret)
2383 return ret;
2384
2385 inode_lock(inode);
2386
2387 f2fs_abort_atomic_write(inode, true);
2388
2389 inode_unlock(inode);
2390
2391 mnt_drop_write_file(filp);
2392 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2393 return ret;
2394 }
2395
f2fs_do_shutdown(struct f2fs_sb_info * sbi,unsigned int flag,bool readonly,bool need_lock)2396 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
2397 bool readonly, bool need_lock)
2398 {
2399 struct super_block *sb = sbi->sb;
2400 int ret = 0;
2401
2402 switch (flag) {
2403 case F2FS_GOING_DOWN_FULLSYNC:
2404 ret = bdev_freeze(sb->s_bdev);
2405 if (ret)
2406 goto out;
2407 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2408 bdev_thaw(sb->s_bdev);
2409 break;
2410 case F2FS_GOING_DOWN_METASYNC:
2411 /* do checkpoint only */
2412 ret = f2fs_sync_fs(sb, 1);
2413 if (ret) {
2414 if (ret == -EIO)
2415 ret = 0;
2416 goto out;
2417 }
2418 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2419 break;
2420 case F2FS_GOING_DOWN_NOSYNC:
2421 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2422 break;
2423 case F2FS_GOING_DOWN_METAFLUSH:
2424 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2425 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2426 break;
2427 case F2FS_GOING_DOWN_NEED_FSCK:
2428 set_sbi_flag(sbi, SBI_NEED_FSCK);
2429 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2430 set_sbi_flag(sbi, SBI_IS_DIRTY);
2431 /* do checkpoint only */
2432 ret = f2fs_sync_fs(sb, 1);
2433 if (ret == -EIO)
2434 ret = 0;
2435 goto out;
2436 default:
2437 ret = -EINVAL;
2438 goto out;
2439 }
2440
2441 if (readonly)
2442 goto out;
2443
2444 /*
2445 * grab sb->s_umount to avoid racing w/ remount() and other shutdown
2446 * paths.
2447 */
2448 if (need_lock)
2449 down_write(&sbi->sb->s_umount);
2450
2451 f2fs_stop_gc_thread(sbi);
2452 f2fs_stop_discard_thread(sbi);
2453
2454 f2fs_drop_discard_cmd(sbi);
2455 clear_opt(sbi, DISCARD);
2456
2457 if (need_lock)
2458 up_write(&sbi->sb->s_umount);
2459
2460 f2fs_update_time(sbi, REQ_TIME);
2461 out:
2462
2463 trace_f2fs_shutdown(sbi, flag, ret);
2464
2465 return ret;
2466 }
2467
f2fs_ioc_shutdown(struct file * filp,unsigned long arg)2468 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2469 {
2470 struct inode *inode = file_inode(filp);
2471 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2472 __u32 in;
2473 int ret;
2474 bool need_drop = false, readonly = false;
2475
2476 if (!capable(CAP_SYS_ADMIN))
2477 return -EPERM;
2478
2479 if (get_user(in, (__u32 __user *)arg))
2480 return -EFAULT;
2481
2482 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2483 ret = mnt_want_write_file(filp);
2484 if (ret) {
2485 if (ret != -EROFS)
2486 return ret;
2487
2488 /* fallback to nosync shutdown for readonly fs */
2489 in = F2FS_GOING_DOWN_NOSYNC;
2490 readonly = true;
2491 } else {
2492 need_drop = true;
2493 }
2494 }
2495
2496 ret = f2fs_do_shutdown(sbi, in, readonly, true);
2497
2498 if (need_drop)
2499 mnt_drop_write_file(filp);
2500
2501 return ret;
2502 }
2503
f2fs_keep_noreuse_range(struct inode * inode,loff_t offset,loff_t len)2504 static int f2fs_keep_noreuse_range(struct inode *inode,
2505 loff_t offset, loff_t len)
2506 {
2507 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2508 u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
2509 u64 start, end;
2510 int ret = 0;
2511
2512 if (!S_ISREG(inode->i_mode))
2513 return 0;
2514
2515 if (offset >= max_bytes || len > max_bytes ||
2516 (offset + len) > max_bytes)
2517 return 0;
2518
2519 start = offset >> PAGE_SHIFT;
2520 end = DIV_ROUND_UP(offset + len, PAGE_SIZE);
2521
2522 inode_lock(inode);
2523 if (f2fs_is_atomic_file(inode)) {
2524 inode_unlock(inode);
2525 return 0;
2526 }
2527
2528 spin_lock(&sbi->inode_lock[DONATE_INODE]);
2529 /* let's remove the range, if len = 0 */
2530 if (!len) {
2531 if (!list_empty(&F2FS_I(inode)->gdonate_list)) {
2532 list_del_init(&F2FS_I(inode)->gdonate_list);
2533 sbi->donate_files--;
2534 if (is_inode_flag_set(inode, FI_DONATE_FINISHED))
2535 ret = -EALREADY;
2536 else
2537 set_inode_flag(inode, FI_DONATE_FINISHED);
2538 } else
2539 ret = -ENOENT;
2540 } else {
2541 if (list_empty(&F2FS_I(inode)->gdonate_list)) {
2542 list_add_tail(&F2FS_I(inode)->gdonate_list,
2543 &sbi->inode_list[DONATE_INODE]);
2544 sbi->donate_files++;
2545 } else {
2546 list_move_tail(&F2FS_I(inode)->gdonate_list,
2547 &sbi->inode_list[DONATE_INODE]);
2548 }
2549 F2FS_I(inode)->donate_start = start;
2550 F2FS_I(inode)->donate_end = end - 1;
2551 clear_inode_flag(inode, FI_DONATE_FINISHED);
2552 }
2553 spin_unlock(&sbi->inode_lock[DONATE_INODE]);
2554 inode_unlock(inode);
2555
2556 return ret;
2557 }
2558
f2fs_ioc_fitrim(struct file * filp,unsigned long arg)2559 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2560 {
2561 struct inode *inode = file_inode(filp);
2562 struct super_block *sb = inode->i_sb;
2563 struct fstrim_range range;
2564 int ret;
2565
2566 if (!capable(CAP_SYS_ADMIN))
2567 return -EPERM;
2568
2569 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2570 return -EOPNOTSUPP;
2571
2572 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2573 sizeof(range)))
2574 return -EFAULT;
2575
2576 ret = mnt_want_write_file(filp);
2577 if (ret)
2578 return ret;
2579
2580 range.minlen = max((unsigned int)range.minlen,
2581 bdev_discard_granularity(sb->s_bdev));
2582 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2583 mnt_drop_write_file(filp);
2584 if (ret < 0)
2585 return ret;
2586
2587 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2588 sizeof(range)))
2589 return -EFAULT;
2590 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2591 return 0;
2592 }
2593
uuid_is_nonzero(__u8 u[16])2594 static bool uuid_is_nonzero(__u8 u[16])
2595 {
2596 int i;
2597
2598 for (i = 0; i < 16; i++)
2599 if (u[i])
2600 return true;
2601 return false;
2602 }
2603
f2fs_ioc_set_encryption_policy(struct file * filp,unsigned long arg)2604 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2605 {
2606 struct inode *inode = file_inode(filp);
2607 int ret;
2608
2609 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2610 return -EOPNOTSUPP;
2611
2612 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2613 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2614 return ret;
2615 }
2616
f2fs_ioc_get_encryption_policy(struct file * filp,unsigned long arg)2617 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2618 {
2619 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2620 return -EOPNOTSUPP;
2621 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2622 }
2623
f2fs_ioc_get_encryption_pwsalt(struct file * filp,unsigned long arg)2624 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2625 {
2626 struct inode *inode = file_inode(filp);
2627 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2628 u8 encrypt_pw_salt[16];
2629 int err;
2630
2631 if (!f2fs_sb_has_encrypt(sbi))
2632 return -EOPNOTSUPP;
2633
2634 err = mnt_want_write_file(filp);
2635 if (err)
2636 return err;
2637
2638 f2fs_down_write(&sbi->sb_lock);
2639
2640 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2641 goto got_it;
2642
2643 /* update superblock with uuid */
2644 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2645
2646 err = f2fs_commit_super(sbi, false);
2647 if (err) {
2648 /* undo new data */
2649 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2650 goto out_err;
2651 }
2652 got_it:
2653 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2654 out_err:
2655 f2fs_up_write(&sbi->sb_lock);
2656 mnt_drop_write_file(filp);
2657
2658 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2659 err = -EFAULT;
2660
2661 return err;
2662 }
2663
f2fs_ioc_get_encryption_policy_ex(struct file * filp,unsigned long arg)2664 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2665 unsigned long arg)
2666 {
2667 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2668 return -EOPNOTSUPP;
2669
2670 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2671 }
2672
f2fs_ioc_add_encryption_key(struct file * filp,unsigned long arg)2673 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2674 {
2675 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2676 return -EOPNOTSUPP;
2677
2678 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2679 }
2680
f2fs_ioc_remove_encryption_key(struct file * filp,unsigned long arg)2681 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2682 {
2683 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2684 return -EOPNOTSUPP;
2685
2686 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2687 }
2688
f2fs_ioc_remove_encryption_key_all_users(struct file * filp,unsigned long arg)2689 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2690 unsigned long arg)
2691 {
2692 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2693 return -EOPNOTSUPP;
2694
2695 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2696 }
2697
f2fs_ioc_get_encryption_key_status(struct file * filp,unsigned long arg)2698 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2699 unsigned long arg)
2700 {
2701 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2702 return -EOPNOTSUPP;
2703
2704 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2705 }
2706
f2fs_ioc_get_encryption_nonce(struct file * filp,unsigned long arg)2707 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2708 {
2709 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2710 return -EOPNOTSUPP;
2711
2712 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2713 }
2714
f2fs_ioc_gc(struct file * filp,unsigned long arg)2715 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2716 {
2717 struct inode *inode = file_inode(filp);
2718 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2719 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2720 .no_bg_gc = false,
2721 .should_migrate_blocks = false,
2722 .nr_free_secs = 0 };
2723 __u32 sync;
2724 int ret;
2725
2726 if (!capable(CAP_SYS_ADMIN))
2727 return -EPERM;
2728
2729 if (get_user(sync, (__u32 __user *)arg))
2730 return -EFAULT;
2731
2732 if (f2fs_readonly(sbi->sb))
2733 return -EROFS;
2734
2735 ret = mnt_want_write_file(filp);
2736 if (ret)
2737 return ret;
2738
2739 if (!sync) {
2740 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2741 ret = -EBUSY;
2742 goto out;
2743 }
2744 } else {
2745 f2fs_down_write(&sbi->gc_lock);
2746 }
2747
2748 gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2749 gc_control.err_gc_skipped = sync;
2750 stat_inc_gc_call_count(sbi, FOREGROUND);
2751 ret = f2fs_gc(sbi, &gc_control);
2752 out:
2753 mnt_drop_write_file(filp);
2754 return ret;
2755 }
2756
__f2fs_ioc_gc_range(struct file * filp,struct f2fs_gc_range * range)2757 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2758 {
2759 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2760 struct f2fs_gc_control gc_control = {
2761 .init_gc_type = range->sync ? FG_GC : BG_GC,
2762 .no_bg_gc = false,
2763 .should_migrate_blocks = false,
2764 .err_gc_skipped = range->sync,
2765 .nr_free_secs = 0 };
2766 u64 end;
2767 int ret;
2768
2769 if (!capable(CAP_SYS_ADMIN))
2770 return -EPERM;
2771 if (f2fs_readonly(sbi->sb))
2772 return -EROFS;
2773
2774 end = range->start + range->len;
2775 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2776 end >= MAX_BLKADDR(sbi))
2777 return -EINVAL;
2778
2779 ret = mnt_want_write_file(filp);
2780 if (ret)
2781 return ret;
2782
2783 do_more:
2784 if (!range->sync) {
2785 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2786 ret = -EBUSY;
2787 goto out;
2788 }
2789 } else {
2790 f2fs_down_write(&sbi->gc_lock);
2791 }
2792
2793 gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2794 stat_inc_gc_call_count(sbi, FOREGROUND);
2795 ret = f2fs_gc(sbi, &gc_control);
2796 if (ret) {
2797 if (ret == -EBUSY)
2798 ret = -EAGAIN;
2799 goto out;
2800 }
2801 range->start += CAP_BLKS_PER_SEC(sbi);
2802 if (range->start <= end)
2803 goto do_more;
2804 out:
2805 mnt_drop_write_file(filp);
2806 return ret;
2807 }
2808
f2fs_ioc_gc_range(struct file * filp,unsigned long arg)2809 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2810 {
2811 struct f2fs_gc_range range;
2812
2813 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2814 sizeof(range)))
2815 return -EFAULT;
2816 return __f2fs_ioc_gc_range(filp, &range);
2817 }
2818
f2fs_ioc_write_checkpoint(struct file * filp)2819 static int f2fs_ioc_write_checkpoint(struct file *filp)
2820 {
2821 struct inode *inode = file_inode(filp);
2822 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2823 int ret;
2824
2825 if (!capable(CAP_SYS_ADMIN))
2826 return -EPERM;
2827
2828 if (f2fs_readonly(sbi->sb))
2829 return -EROFS;
2830
2831 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2832 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2833 return -EINVAL;
2834 }
2835
2836 ret = mnt_want_write_file(filp);
2837 if (ret)
2838 return ret;
2839
2840 ret = f2fs_sync_fs(sbi->sb, 1);
2841
2842 mnt_drop_write_file(filp);
2843 return ret;
2844 }
2845
f2fs_defragment_range(struct f2fs_sb_info * sbi,struct file * filp,struct f2fs_defragment * range)2846 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2847 struct file *filp,
2848 struct f2fs_defragment *range)
2849 {
2850 struct inode *inode = file_inode(filp);
2851 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2852 .m_seg_type = NO_CHECK_TYPE,
2853 .m_may_create = false };
2854 struct extent_info ei = {};
2855 pgoff_t pg_start, pg_end, next_pgofs;
2856 unsigned int total = 0, sec_num;
2857 block_t blk_end = 0;
2858 bool fragmented = false;
2859 int err;
2860
2861 f2fs_balance_fs(sbi, true);
2862
2863 inode_lock(inode);
2864 pg_start = range->start >> PAGE_SHIFT;
2865 pg_end = min_t(pgoff_t,
2866 (range->start + range->len) >> PAGE_SHIFT,
2867 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));
2868
2869 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) ||
2870 f2fs_is_atomic_file(inode)) {
2871 err = -EINVAL;
2872 goto unlock_out;
2873 }
2874
2875 /* if in-place-update policy is enabled, don't waste time here */
2876 set_inode_flag(inode, FI_OPU_WRITE);
2877 if (f2fs_should_update_inplace(inode, NULL)) {
2878 err = -EINVAL;
2879 goto out;
2880 }
2881
2882 /* writeback all dirty pages in the range */
2883 err = filemap_write_and_wait_range(inode->i_mapping,
2884 pg_start << PAGE_SHIFT,
2885 (pg_end << PAGE_SHIFT) - 1);
2886 if (err)
2887 goto out;
2888
2889 /*
2890 * lookup mapping info in extent cache, skip defragmenting if physical
2891 * block addresses are continuous.
2892 */
2893 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2894 if ((pgoff_t)ei.fofs + ei.len >= pg_end)
2895 goto out;
2896 }
2897
2898 map.m_lblk = pg_start;
2899 map.m_next_pgofs = &next_pgofs;
2900
2901 /*
2902 * lookup mapping info in dnode page cache, skip defragmenting if all
2903 * physical block addresses are continuous even if there are hole(s)
2904 * in logical blocks.
2905 */
2906 while (map.m_lblk < pg_end) {
2907 map.m_len = pg_end - map.m_lblk;
2908 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2909 if (err)
2910 goto out;
2911
2912 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2913 map.m_lblk = next_pgofs;
2914 continue;
2915 }
2916
2917 if (blk_end && blk_end != map.m_pblk)
2918 fragmented = true;
2919
2920 /* record total count of block that we're going to move */
2921 total += map.m_len;
2922
2923 blk_end = map.m_pblk + map.m_len;
2924
2925 map.m_lblk += map.m_len;
2926 }
2927
2928 if (!fragmented) {
2929 total = 0;
2930 goto out;
2931 }
2932
2933 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2934
2935 /*
2936 * make sure there are enough free section for LFS allocation, this can
2937 * avoid defragment running in SSR mode when free section are allocated
2938 * intensively
2939 */
2940 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2941 err = -EAGAIN;
2942 goto out;
2943 }
2944
2945 map.m_lblk = pg_start;
2946 map.m_len = pg_end - pg_start;
2947 total = 0;
2948
2949 while (map.m_lblk < pg_end) {
2950 pgoff_t idx;
2951 int cnt = 0;
2952
2953 do_map:
2954 map.m_len = pg_end - map.m_lblk;
2955 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2956 if (err)
2957 goto clear_out;
2958
2959 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2960 map.m_lblk = next_pgofs;
2961 goto check;
2962 }
2963
2964 set_inode_flag(inode, FI_SKIP_WRITES);
2965
2966 idx = map.m_lblk;
2967 while (idx < map.m_lblk + map.m_len &&
2968 cnt < BLKS_PER_SEG(sbi)) {
2969 struct folio *folio;
2970
2971 folio = f2fs_get_lock_data_folio(inode, idx, true);
2972 if (IS_ERR(folio)) {
2973 err = PTR_ERR(folio);
2974 goto clear_out;
2975 }
2976
2977 f2fs_folio_wait_writeback(folio, DATA, true, true);
2978
2979 folio_mark_dirty(folio);
2980 set_page_private_gcing(&folio->page);
2981 f2fs_folio_put(folio, true);
2982
2983 idx++;
2984 cnt++;
2985 total++;
2986 }
2987
2988 map.m_lblk = idx;
2989 check:
2990 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi))
2991 goto do_map;
2992
2993 clear_inode_flag(inode, FI_SKIP_WRITES);
2994
2995 err = filemap_fdatawrite(inode->i_mapping);
2996 if (err)
2997 goto out;
2998 }
2999 clear_out:
3000 clear_inode_flag(inode, FI_SKIP_WRITES);
3001 out:
3002 clear_inode_flag(inode, FI_OPU_WRITE);
3003 unlock_out:
3004 inode_unlock(inode);
3005 if (!err)
3006 range->len = (u64)total << PAGE_SHIFT;
3007 return err;
3008 }
3009
f2fs_ioc_defragment(struct file * filp,unsigned long arg)3010 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
3011 {
3012 struct inode *inode = file_inode(filp);
3013 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3014 struct f2fs_defragment range;
3015 int err;
3016
3017 if (!capable(CAP_SYS_ADMIN))
3018 return -EPERM;
3019
3020 if (!S_ISREG(inode->i_mode))
3021 return -EINVAL;
3022
3023 if (f2fs_readonly(sbi->sb))
3024 return -EROFS;
3025
3026 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
3027 sizeof(range)))
3028 return -EFAULT;
3029
3030 /* verify alignment of offset & size */
3031 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
3032 return -EINVAL;
3033
3034 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
3035 max_file_blocks(inode)))
3036 return -EINVAL;
3037
3038 err = mnt_want_write_file(filp);
3039 if (err)
3040 return err;
3041
3042 err = f2fs_defragment_range(sbi, filp, &range);
3043 mnt_drop_write_file(filp);
3044
3045 if (range.len)
3046 f2fs_update_time(sbi, REQ_TIME);
3047 if (err < 0)
3048 return err;
3049
3050 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
3051 sizeof(range)))
3052 return -EFAULT;
3053
3054 return 0;
3055 }
3056
f2fs_move_file_range(struct file * file_in,loff_t pos_in,struct file * file_out,loff_t pos_out,size_t len)3057 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
3058 struct file *file_out, loff_t pos_out, size_t len)
3059 {
3060 struct inode *src = file_inode(file_in);
3061 struct inode *dst = file_inode(file_out);
3062 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
3063 size_t olen = len, dst_max_i_size = 0;
3064 size_t dst_osize;
3065 int ret;
3066
3067 if (file_in->f_path.mnt != file_out->f_path.mnt ||
3068 src->i_sb != dst->i_sb)
3069 return -EXDEV;
3070
3071 if (unlikely(f2fs_readonly(src->i_sb)))
3072 return -EROFS;
3073
3074 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
3075 return -EINVAL;
3076
3077 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
3078 return -EOPNOTSUPP;
3079
3080 if (pos_out < 0 || pos_in < 0)
3081 return -EINVAL;
3082
3083 if (src == dst) {
3084 if (pos_in == pos_out)
3085 return 0;
3086 if (pos_out > pos_in && pos_out < pos_in + len)
3087 return -EINVAL;
3088 }
3089
3090 inode_lock(src);
3091 if (src != dst) {
3092 ret = -EBUSY;
3093 if (!inode_trylock(dst))
3094 goto out;
3095 }
3096
3097 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) ||
3098 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) {
3099 ret = -EOPNOTSUPP;
3100 goto out_unlock;
3101 }
3102
3103 if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) {
3104 ret = -EINVAL;
3105 goto out_unlock;
3106 }
3107
3108 ret = -EINVAL;
3109 if (pos_in + len > src->i_size || pos_in + len < pos_in)
3110 goto out_unlock;
3111 if (len == 0)
3112 olen = len = src->i_size - pos_in;
3113 if (pos_in + len == src->i_size)
3114 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
3115 if (len == 0) {
3116 ret = 0;
3117 goto out_unlock;
3118 }
3119
3120 dst_osize = dst->i_size;
3121 if (pos_out + olen > dst->i_size)
3122 dst_max_i_size = pos_out + olen;
3123
3124 /* verify the end result is block aligned */
3125 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
3126 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
3127 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
3128 goto out_unlock;
3129
3130 ret = f2fs_convert_inline_inode(src);
3131 if (ret)
3132 goto out_unlock;
3133
3134 ret = f2fs_convert_inline_inode(dst);
3135 if (ret)
3136 goto out_unlock;
3137
3138 /* write out all dirty pages from offset */
3139 ret = filemap_write_and_wait_range(src->i_mapping,
3140 pos_in, pos_in + len);
3141 if (ret)
3142 goto out_unlock;
3143
3144 ret = filemap_write_and_wait_range(dst->i_mapping,
3145 pos_out, pos_out + len);
3146 if (ret)
3147 goto out_unlock;
3148
3149 f2fs_balance_fs(sbi, true);
3150
3151 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3152 if (src != dst) {
3153 ret = -EBUSY;
3154 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
3155 goto out_src;
3156 }
3157
3158 f2fs_lock_op(sbi);
3159 ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in),
3160 F2FS_BYTES_TO_BLK(pos_out),
3161 F2FS_BYTES_TO_BLK(len), false);
3162
3163 if (!ret) {
3164 if (dst_max_i_size)
3165 f2fs_i_size_write(dst, dst_max_i_size);
3166 else if (dst_osize != dst->i_size)
3167 f2fs_i_size_write(dst, dst_osize);
3168 }
3169 f2fs_unlock_op(sbi);
3170
3171 if (src != dst)
3172 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
3173 out_src:
3174 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
3175 if (ret)
3176 goto out_unlock;
3177
3178 inode_set_mtime_to_ts(src, inode_set_ctime_current(src));
3179 f2fs_mark_inode_dirty_sync(src, false);
3180 if (src != dst) {
3181 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst));
3182 f2fs_mark_inode_dirty_sync(dst, false);
3183 }
3184 f2fs_update_time(sbi, REQ_TIME);
3185
3186 out_unlock:
3187 if (src != dst)
3188 inode_unlock(dst);
3189 out:
3190 inode_unlock(src);
3191 return ret;
3192 }
3193
__f2fs_ioc_move_range(struct file * filp,struct f2fs_move_range * range)3194 static int __f2fs_ioc_move_range(struct file *filp,
3195 struct f2fs_move_range *range)
3196 {
3197 int err;
3198
3199 if (!(filp->f_mode & FMODE_READ) ||
3200 !(filp->f_mode & FMODE_WRITE))
3201 return -EBADF;
3202
3203 CLASS(fd, dst)(range->dst_fd);
3204 if (fd_empty(dst))
3205 return -EBADF;
3206
3207 if (!(fd_file(dst)->f_mode & FMODE_WRITE))
3208 return -EBADF;
3209
3210 err = mnt_want_write_file(filp);
3211 if (err)
3212 return err;
3213
3214 err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst),
3215 range->pos_out, range->len);
3216
3217 mnt_drop_write_file(filp);
3218 return err;
3219 }
3220
f2fs_ioc_move_range(struct file * filp,unsigned long arg)3221 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
3222 {
3223 struct f2fs_move_range range;
3224
3225 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
3226 sizeof(range)))
3227 return -EFAULT;
3228 return __f2fs_ioc_move_range(filp, &range);
3229 }
3230
f2fs_ioc_flush_device(struct file * filp,unsigned long arg)3231 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
3232 {
3233 struct inode *inode = file_inode(filp);
3234 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3235 struct sit_info *sm = SIT_I(sbi);
3236 unsigned int start_segno = 0, end_segno = 0;
3237 unsigned int dev_start_segno = 0, dev_end_segno = 0;
3238 struct f2fs_flush_device range;
3239 struct f2fs_gc_control gc_control = {
3240 .init_gc_type = FG_GC,
3241 .should_migrate_blocks = true,
3242 .err_gc_skipped = true,
3243 .nr_free_secs = 0 };
3244 int ret;
3245
3246 if (!capable(CAP_SYS_ADMIN))
3247 return -EPERM;
3248
3249 if (f2fs_readonly(sbi->sb))
3250 return -EROFS;
3251
3252 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3253 return -EINVAL;
3254
3255 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
3256 sizeof(range)))
3257 return -EFAULT;
3258
3259 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
3260 __is_large_section(sbi)) {
3261 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1",
3262 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi));
3263 return -EINVAL;
3264 }
3265
3266 ret = mnt_want_write_file(filp);
3267 if (ret)
3268 return ret;
3269
3270 if (range.dev_num != 0)
3271 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
3272 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
3273
3274 start_segno = sm->last_victim[FLUSH_DEVICE];
3275 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
3276 start_segno = dev_start_segno;
3277 end_segno = min(start_segno + range.segments, dev_end_segno);
3278
3279 while (start_segno < end_segno) {
3280 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
3281 ret = -EBUSY;
3282 goto out;
3283 }
3284 sm->last_victim[GC_CB] = end_segno + 1;
3285 sm->last_victim[GC_GREEDY] = end_segno + 1;
3286 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
3287
3288 gc_control.victim_segno = start_segno;
3289 stat_inc_gc_call_count(sbi, FOREGROUND);
3290 ret = f2fs_gc(sbi, &gc_control);
3291 if (ret == -EAGAIN)
3292 ret = 0;
3293 else if (ret < 0)
3294 break;
3295 start_segno++;
3296 }
3297 out:
3298 mnt_drop_write_file(filp);
3299 return ret;
3300 }
3301
f2fs_ioc_get_features(struct file * filp,unsigned long arg)3302 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3303 {
3304 struct inode *inode = file_inode(filp);
3305 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3306
3307 /* Must validate to set it with SQLite behavior in Android. */
3308 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3309
3310 return put_user(sb_feature, (u32 __user *)arg);
3311 }
3312
3313 #ifdef CONFIG_QUOTA
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3314 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3315 {
3316 struct dquot *transfer_to[MAXQUOTAS] = {};
3317 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3318 struct super_block *sb = sbi->sb;
3319 int err;
3320
3321 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3322 if (IS_ERR(transfer_to[PRJQUOTA]))
3323 return PTR_ERR(transfer_to[PRJQUOTA]);
3324
3325 err = __dquot_transfer(inode, transfer_to);
3326 if (err)
3327 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3328 dqput(transfer_to[PRJQUOTA]);
3329 return err;
3330 }
3331
f2fs_ioc_setproject(struct inode * inode,__u32 projid)3332 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3333 {
3334 struct f2fs_inode_info *fi = F2FS_I(inode);
3335 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3336 struct f2fs_inode *ri = NULL;
3337 kprojid_t kprojid;
3338 int err;
3339
3340 if (!f2fs_sb_has_project_quota(sbi)) {
3341 if (projid != F2FS_DEF_PROJID)
3342 return -EOPNOTSUPP;
3343 else
3344 return 0;
3345 }
3346
3347 if (!f2fs_has_extra_attr(inode))
3348 return -EOPNOTSUPP;
3349
3350 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3351
3352 if (projid_eq(kprojid, fi->i_projid))
3353 return 0;
3354
3355 err = -EPERM;
3356 /* Is it quota file? Do not allow user to mess with it */
3357 if (IS_NOQUOTA(inode))
3358 return err;
3359
3360 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3361 return -EOVERFLOW;
3362
3363 err = f2fs_dquot_initialize(inode);
3364 if (err)
3365 return err;
3366
3367 f2fs_lock_op(sbi);
3368 err = f2fs_transfer_project_quota(inode, kprojid);
3369 if (err)
3370 goto out_unlock;
3371
3372 fi->i_projid = kprojid;
3373 inode_set_ctime_current(inode);
3374 f2fs_mark_inode_dirty_sync(inode, true);
3375 out_unlock:
3376 f2fs_unlock_op(sbi);
3377 return err;
3378 }
3379 #else
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3380 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3381 {
3382 return 0;
3383 }
3384
f2fs_ioc_setproject(struct inode * inode,__u32 projid)3385 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3386 {
3387 if (projid != F2FS_DEF_PROJID)
3388 return -EOPNOTSUPP;
3389 return 0;
3390 }
3391 #endif
3392
f2fs_fileattr_get(struct dentry * dentry,struct fileattr * fa)3393 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3394 {
3395 struct inode *inode = d_inode(dentry);
3396 struct f2fs_inode_info *fi = F2FS_I(inode);
3397 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3398
3399 if (IS_ENCRYPTED(inode))
3400 fsflags |= FS_ENCRYPT_FL;
3401 if (IS_VERITY(inode))
3402 fsflags |= FS_VERITY_FL;
3403 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3404 fsflags |= FS_INLINE_DATA_FL;
3405 if (is_inode_flag_set(inode, FI_PIN_FILE))
3406 fsflags |= FS_NOCOW_FL;
3407
3408 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3409
3410 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3411 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3412
3413 return 0;
3414 }
3415
f2fs_fileattr_set(struct mnt_idmap * idmap,struct dentry * dentry,struct fileattr * fa)3416 int f2fs_fileattr_set(struct mnt_idmap *idmap,
3417 struct dentry *dentry, struct fileattr *fa)
3418 {
3419 struct inode *inode = d_inode(dentry);
3420 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3421 u32 iflags;
3422 int err;
3423
3424 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3425 return -EIO;
3426 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3427 return -ENOSPC;
3428 if (fsflags & ~F2FS_GETTABLE_FS_FL)
3429 return -EOPNOTSUPP;
3430 fsflags &= F2FS_SETTABLE_FS_FL;
3431 if (!fa->flags_valid)
3432 mask &= FS_COMMON_FL;
3433
3434 iflags = f2fs_fsflags_to_iflags(fsflags);
3435 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3436 return -EOPNOTSUPP;
3437
3438 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3439 if (!err)
3440 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3441
3442 return err;
3443 }
3444
f2fs_pin_file_control(struct inode * inode,bool inc)3445 int f2fs_pin_file_control(struct inode *inode, bool inc)
3446 {
3447 struct f2fs_inode_info *fi = F2FS_I(inode);
3448 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3449
3450 if (IS_DEVICE_ALIASING(inode))
3451 return -EINVAL;
3452
3453 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) {
3454 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3455 __func__, inode->i_ino, fi->i_gc_failures);
3456 clear_inode_flag(inode, FI_PIN_FILE);
3457 return -EAGAIN;
3458 }
3459
3460 /* Use i_gc_failures for normal file as a risk signal. */
3461 if (inc)
3462 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
3463
3464 return 0;
3465 }
3466
f2fs_ioc_set_pin_file(struct file * filp,unsigned long arg)3467 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3468 {
3469 struct inode *inode = file_inode(filp);
3470 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3471 __u32 pin;
3472 int ret = 0;
3473
3474 if (get_user(pin, (__u32 __user *)arg))
3475 return -EFAULT;
3476
3477 if (!S_ISREG(inode->i_mode))
3478 return -EINVAL;
3479
3480 if (f2fs_readonly(sbi->sb))
3481 return -EROFS;
3482
3483 if (!pin && IS_DEVICE_ALIASING(inode))
3484 return -EOPNOTSUPP;
3485
3486 ret = mnt_want_write_file(filp);
3487 if (ret)
3488 return ret;
3489
3490 inode_lock(inode);
3491
3492 if (f2fs_is_atomic_file(inode)) {
3493 ret = -EINVAL;
3494 goto out;
3495 }
3496
3497 if (!pin) {
3498 clear_inode_flag(inode, FI_PIN_FILE);
3499 f2fs_i_gc_failures_write(inode, 0);
3500 goto done;
3501 } else if (f2fs_is_pinned_file(inode)) {
3502 goto done;
3503 }
3504
3505 if (F2FS_HAS_BLOCKS(inode)) {
3506 ret = -EFBIG;
3507 goto out;
3508 }
3509
3510 /* Let's allow file pinning on zoned device. */
3511 if (!f2fs_sb_has_blkzoned(sbi) &&
3512 f2fs_should_update_outplace(inode, NULL)) {
3513 ret = -EINVAL;
3514 goto out;
3515 }
3516
3517 if (f2fs_pin_file_control(inode, false)) {
3518 ret = -EAGAIN;
3519 goto out;
3520 }
3521
3522 ret = f2fs_convert_inline_inode(inode);
3523 if (ret)
3524 goto out;
3525
3526 if (!f2fs_disable_compressed_file(inode)) {
3527 ret = -EOPNOTSUPP;
3528 goto out;
3529 }
3530
3531 set_inode_flag(inode, FI_PIN_FILE);
3532 ret = F2FS_I(inode)->i_gc_failures;
3533 done:
3534 f2fs_update_time(sbi, REQ_TIME);
3535 out:
3536 inode_unlock(inode);
3537 mnt_drop_write_file(filp);
3538 return ret;
3539 }
3540
f2fs_ioc_get_pin_file(struct file * filp,unsigned long arg)3541 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3542 {
3543 struct inode *inode = file_inode(filp);
3544 __u32 pin = 0;
3545
3546 if (is_inode_flag_set(inode, FI_PIN_FILE))
3547 pin = F2FS_I(inode)->i_gc_failures;
3548 return put_user(pin, (u32 __user *)arg);
3549 }
3550
f2fs_ioc_get_dev_alias_file(struct file * filp,unsigned long arg)3551 static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg)
3552 {
3553 return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0,
3554 (u32 __user *)arg);
3555 }
3556
f2fs_ioc_io_prio(struct file * filp,unsigned long arg)3557 static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg)
3558 {
3559 struct inode *inode = file_inode(filp);
3560 __u32 level;
3561
3562 if (get_user(level, (__u32 __user *)arg))
3563 return -EFAULT;
3564
3565 if (!S_ISREG(inode->i_mode) || level >= F2FS_IOPRIO_MAX)
3566 return -EINVAL;
3567
3568 inode_lock(inode);
3569 F2FS_I(inode)->ioprio_hint = level;
3570 inode_unlock(inode);
3571 return 0;
3572 }
3573
f2fs_precache_extents(struct inode * inode)3574 int f2fs_precache_extents(struct inode *inode)
3575 {
3576 struct f2fs_inode_info *fi = F2FS_I(inode);
3577 struct f2fs_map_blocks map;
3578 pgoff_t m_next_extent;
3579 loff_t end;
3580 int err;
3581
3582 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3583 return -EOPNOTSUPP;
3584
3585 map.m_lblk = 0;
3586 map.m_pblk = 0;
3587 map.m_next_pgofs = NULL;
3588 map.m_next_extent = &m_next_extent;
3589 map.m_seg_type = NO_CHECK_TYPE;
3590 map.m_may_create = false;
3591 end = F2FS_BLK_ALIGN(i_size_read(inode));
3592
3593 while (map.m_lblk < end) {
3594 map.m_len = end - map.m_lblk;
3595
3596 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3597 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
3598 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3599 if (err || !map.m_len)
3600 return err;
3601
3602 map.m_lblk = m_next_extent;
3603 }
3604
3605 return 0;
3606 }
3607
f2fs_ioc_precache_extents(struct file * filp)3608 static int f2fs_ioc_precache_extents(struct file *filp)
3609 {
3610 return f2fs_precache_extents(file_inode(filp));
3611 }
3612
f2fs_ioc_resize_fs(struct file * filp,unsigned long arg)3613 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3614 {
3615 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3616 __u64 block_count;
3617
3618 if (!capable(CAP_SYS_ADMIN))
3619 return -EPERM;
3620
3621 if (f2fs_readonly(sbi->sb))
3622 return -EROFS;
3623
3624 if (copy_from_user(&block_count, (void __user *)arg,
3625 sizeof(block_count)))
3626 return -EFAULT;
3627
3628 return f2fs_resize_fs(filp, block_count);
3629 }
3630
f2fs_ioc_enable_verity(struct file * filp,unsigned long arg)3631 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3632 {
3633 struct inode *inode = file_inode(filp);
3634
3635 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3636
3637 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3638 f2fs_warn(F2FS_I_SB(inode),
3639 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3640 inode->i_ino);
3641 return -EOPNOTSUPP;
3642 }
3643
3644 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3645 }
3646
f2fs_ioc_measure_verity(struct file * filp,unsigned long arg)3647 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3648 {
3649 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3650 return -EOPNOTSUPP;
3651
3652 return fsverity_ioctl_measure(filp, (void __user *)arg);
3653 }
3654
f2fs_ioc_read_verity_metadata(struct file * filp,unsigned long arg)3655 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3656 {
3657 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3658 return -EOPNOTSUPP;
3659
3660 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3661 }
3662
f2fs_ioc_getfslabel(struct file * filp,unsigned long arg)3663 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3664 {
3665 struct inode *inode = file_inode(filp);
3666 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3667 char *vbuf;
3668 int count;
3669 int err = 0;
3670
3671 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3672 if (!vbuf)
3673 return -ENOMEM;
3674
3675 f2fs_down_read(&sbi->sb_lock);
3676 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3677 ARRAY_SIZE(sbi->raw_super->volume_name),
3678 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3679 f2fs_up_read(&sbi->sb_lock);
3680
3681 if (copy_to_user((char __user *)arg, vbuf,
3682 min(FSLABEL_MAX, count)))
3683 err = -EFAULT;
3684
3685 kfree(vbuf);
3686 return err;
3687 }
3688
f2fs_ioc_setfslabel(struct file * filp,unsigned long arg)3689 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3690 {
3691 struct inode *inode = file_inode(filp);
3692 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3693 char *vbuf;
3694 int err = 0;
3695
3696 if (!capable(CAP_SYS_ADMIN))
3697 return -EPERM;
3698
3699 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3700 if (IS_ERR(vbuf))
3701 return PTR_ERR(vbuf);
3702
3703 err = mnt_want_write_file(filp);
3704 if (err)
3705 goto out;
3706
3707 f2fs_down_write(&sbi->sb_lock);
3708
3709 memset(sbi->raw_super->volume_name, 0,
3710 sizeof(sbi->raw_super->volume_name));
3711 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3712 sbi->raw_super->volume_name,
3713 ARRAY_SIZE(sbi->raw_super->volume_name));
3714
3715 err = f2fs_commit_super(sbi, false);
3716
3717 f2fs_up_write(&sbi->sb_lock);
3718
3719 mnt_drop_write_file(filp);
3720 out:
3721 kfree(vbuf);
3722 return err;
3723 }
3724
f2fs_get_compress_blocks(struct inode * inode,__u64 * blocks)3725 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks)
3726 {
3727 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3728 return -EOPNOTSUPP;
3729
3730 if (!f2fs_compressed_file(inode))
3731 return -EINVAL;
3732
3733 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3734
3735 return 0;
3736 }
3737
f2fs_ioc_get_compress_blocks(struct file * filp,unsigned long arg)3738 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg)
3739 {
3740 struct inode *inode = file_inode(filp);
3741 __u64 blocks;
3742 int ret;
3743
3744 ret = f2fs_get_compress_blocks(inode, &blocks);
3745 if (ret < 0)
3746 return ret;
3747
3748 return put_user(blocks, (u64 __user *)arg);
3749 }
3750
release_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3751 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3752 {
3753 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3754 unsigned int released_blocks = 0;
3755 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3756 block_t blkaddr;
3757 int i;
3758
3759 for (i = 0; i < count; i++) {
3760 blkaddr = data_blkaddr(dn->inode, dn->node_folio,
3761 dn->ofs_in_node + i);
3762
3763 if (!__is_valid_data_blkaddr(blkaddr))
3764 continue;
3765 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3766 DATA_GENERIC_ENHANCE)))
3767 return -EFSCORRUPTED;
3768 }
3769
3770 while (count) {
3771 int compr_blocks = 0;
3772
3773 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3774 blkaddr = f2fs_data_blkaddr(dn);
3775
3776 if (i == 0) {
3777 if (blkaddr == COMPRESS_ADDR)
3778 continue;
3779 dn->ofs_in_node += cluster_size;
3780 goto next;
3781 }
3782
3783 if (__is_valid_data_blkaddr(blkaddr))
3784 compr_blocks++;
3785
3786 if (blkaddr != NEW_ADDR)
3787 continue;
3788
3789 f2fs_set_data_blkaddr(dn, NULL_ADDR);
3790 }
3791
3792 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3793 dec_valid_block_count(sbi, dn->inode,
3794 cluster_size - compr_blocks);
3795
3796 released_blocks += cluster_size - compr_blocks;
3797 next:
3798 count -= cluster_size;
3799 }
3800
3801 return released_blocks;
3802 }
3803
f2fs_release_compress_blocks(struct file * filp,unsigned long arg)3804 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3805 {
3806 struct inode *inode = file_inode(filp);
3807 struct f2fs_inode_info *fi = F2FS_I(inode);
3808 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3809 pgoff_t page_idx = 0, last_idx;
3810 unsigned int released_blocks = 0;
3811 int ret;
3812 int writecount;
3813
3814 if (!f2fs_sb_has_compression(sbi))
3815 return -EOPNOTSUPP;
3816
3817 if (f2fs_readonly(sbi->sb))
3818 return -EROFS;
3819
3820 ret = mnt_want_write_file(filp);
3821 if (ret)
3822 return ret;
3823
3824 f2fs_balance_fs(sbi, true);
3825
3826 inode_lock(inode);
3827
3828 writecount = atomic_read(&inode->i_writecount);
3829 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3830 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3831 ret = -EBUSY;
3832 goto out;
3833 }
3834
3835 if (!f2fs_compressed_file(inode) ||
3836 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3837 ret = -EINVAL;
3838 goto out;
3839 }
3840
3841 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3842 if (ret)
3843 goto out;
3844
3845 if (!atomic_read(&fi->i_compr_blocks)) {
3846 ret = -EPERM;
3847 goto out;
3848 }
3849
3850 set_inode_flag(inode, FI_COMPRESS_RELEASED);
3851 inode_set_ctime_current(inode);
3852 f2fs_mark_inode_dirty_sync(inode, true);
3853
3854 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3855 filemap_invalidate_lock(inode->i_mapping);
3856
3857 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3858
3859 while (page_idx < last_idx) {
3860 struct dnode_of_data dn;
3861 pgoff_t end_offset, count;
3862
3863 f2fs_lock_op(sbi);
3864
3865 set_new_dnode(&dn, inode, NULL, NULL, 0);
3866 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3867 if (ret) {
3868 f2fs_unlock_op(sbi);
3869 if (ret == -ENOENT) {
3870 page_idx = f2fs_get_next_page_offset(&dn,
3871 page_idx);
3872 ret = 0;
3873 continue;
3874 }
3875 break;
3876 }
3877
3878 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
3879 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3880 count = round_up(count, fi->i_cluster_size);
3881
3882 ret = release_compress_blocks(&dn, count);
3883
3884 f2fs_put_dnode(&dn);
3885
3886 f2fs_unlock_op(sbi);
3887
3888 if (ret < 0)
3889 break;
3890
3891 page_idx += count;
3892 released_blocks += ret;
3893 }
3894
3895 filemap_invalidate_unlock(inode->i_mapping);
3896 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3897 out:
3898 if (released_blocks)
3899 f2fs_update_time(sbi, REQ_TIME);
3900 inode_unlock(inode);
3901
3902 mnt_drop_write_file(filp);
3903
3904 if (ret >= 0) {
3905 ret = put_user(released_blocks, (u64 __user *)arg);
3906 } else if (released_blocks &&
3907 atomic_read(&fi->i_compr_blocks)) {
3908 set_sbi_flag(sbi, SBI_NEED_FSCK);
3909 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3910 "iblocks=%llu, released=%u, compr_blocks=%u, "
3911 "run fsck to fix.",
3912 __func__, inode->i_ino, inode->i_blocks,
3913 released_blocks,
3914 atomic_read(&fi->i_compr_blocks));
3915 }
3916
3917 return ret;
3918 }
3919
reserve_compress_blocks(struct dnode_of_data * dn,pgoff_t count,unsigned int * reserved_blocks)3920 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count,
3921 unsigned int *reserved_blocks)
3922 {
3923 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3924 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3925 block_t blkaddr;
3926 int i;
3927
3928 for (i = 0; i < count; i++) {
3929 blkaddr = data_blkaddr(dn->inode, dn->node_folio,
3930 dn->ofs_in_node + i);
3931
3932 if (!__is_valid_data_blkaddr(blkaddr))
3933 continue;
3934 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3935 DATA_GENERIC_ENHANCE)))
3936 return -EFSCORRUPTED;
3937 }
3938
3939 while (count) {
3940 int compr_blocks = 0;
3941 blkcnt_t reserved = 0;
3942 blkcnt_t to_reserved;
3943 int ret;
3944
3945 for (i = 0; i < cluster_size; i++) {
3946 blkaddr = data_blkaddr(dn->inode, dn->node_folio,
3947 dn->ofs_in_node + i);
3948
3949 if (i == 0) {
3950 if (blkaddr != COMPRESS_ADDR) {
3951 dn->ofs_in_node += cluster_size;
3952 goto next;
3953 }
3954 continue;
3955 }
3956
3957 /*
3958 * compressed cluster was not released due to it
3959 * fails in release_compress_blocks(), so NEW_ADDR
3960 * is a possible case.
3961 */
3962 if (blkaddr == NEW_ADDR) {
3963 reserved++;
3964 continue;
3965 }
3966 if (__is_valid_data_blkaddr(blkaddr)) {
3967 compr_blocks++;
3968 continue;
3969 }
3970 }
3971
3972 to_reserved = cluster_size - compr_blocks - reserved;
3973
3974 /* for the case all blocks in cluster were reserved */
3975 if (reserved && to_reserved == 1) {
3976 dn->ofs_in_node += cluster_size;
3977 goto next;
3978 }
3979
3980 ret = inc_valid_block_count(sbi, dn->inode,
3981 &to_reserved, false);
3982 if (unlikely(ret))
3983 return ret;
3984
3985 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3986 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
3987 f2fs_set_data_blkaddr(dn, NEW_ADDR);
3988 }
3989
3990 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3991
3992 *reserved_blocks += to_reserved;
3993 next:
3994 count -= cluster_size;
3995 }
3996
3997 return 0;
3998 }
3999
f2fs_reserve_compress_blocks(struct file * filp,unsigned long arg)4000 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
4001 {
4002 struct inode *inode = file_inode(filp);
4003 struct f2fs_inode_info *fi = F2FS_I(inode);
4004 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4005 pgoff_t page_idx = 0, last_idx;
4006 unsigned int reserved_blocks = 0;
4007 int ret;
4008
4009 if (!f2fs_sb_has_compression(sbi))
4010 return -EOPNOTSUPP;
4011
4012 if (f2fs_readonly(sbi->sb))
4013 return -EROFS;
4014
4015 ret = mnt_want_write_file(filp);
4016 if (ret)
4017 return ret;
4018
4019 f2fs_balance_fs(sbi, true);
4020
4021 inode_lock(inode);
4022
4023 if (!f2fs_compressed_file(inode) ||
4024 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4025 ret = -EINVAL;
4026 goto unlock_inode;
4027 }
4028
4029 if (atomic_read(&fi->i_compr_blocks))
4030 goto unlock_inode;
4031
4032 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
4033 filemap_invalidate_lock(inode->i_mapping);
4034
4035 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4036
4037 while (page_idx < last_idx) {
4038 struct dnode_of_data dn;
4039 pgoff_t end_offset, count;
4040
4041 f2fs_lock_op(sbi);
4042
4043 set_new_dnode(&dn, inode, NULL, NULL, 0);
4044 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
4045 if (ret) {
4046 f2fs_unlock_op(sbi);
4047 if (ret == -ENOENT) {
4048 page_idx = f2fs_get_next_page_offset(&dn,
4049 page_idx);
4050 ret = 0;
4051 continue;
4052 }
4053 break;
4054 }
4055
4056 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
4057 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
4058 count = round_up(count, fi->i_cluster_size);
4059
4060 ret = reserve_compress_blocks(&dn, count, &reserved_blocks);
4061
4062 f2fs_put_dnode(&dn);
4063
4064 f2fs_unlock_op(sbi);
4065
4066 if (ret < 0)
4067 break;
4068
4069 page_idx += count;
4070 }
4071
4072 filemap_invalidate_unlock(inode->i_mapping);
4073 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
4074
4075 if (!ret) {
4076 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
4077 inode_set_ctime_current(inode);
4078 f2fs_mark_inode_dirty_sync(inode, true);
4079 }
4080 unlock_inode:
4081 if (reserved_blocks)
4082 f2fs_update_time(sbi, REQ_TIME);
4083 inode_unlock(inode);
4084 mnt_drop_write_file(filp);
4085
4086 if (!ret) {
4087 ret = put_user(reserved_blocks, (u64 __user *)arg);
4088 } else if (reserved_blocks &&
4089 atomic_read(&fi->i_compr_blocks)) {
4090 set_sbi_flag(sbi, SBI_NEED_FSCK);
4091 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx "
4092 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
4093 "run fsck to fix.",
4094 __func__, inode->i_ino, inode->i_blocks,
4095 reserved_blocks,
4096 atomic_read(&fi->i_compr_blocks));
4097 }
4098
4099 return ret;
4100 }
4101
f2fs_secure_erase(struct block_device * bdev,struct inode * inode,pgoff_t off,block_t block,block_t len,u32 flags)4102 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
4103 pgoff_t off, block_t block, block_t len, u32 flags)
4104 {
4105 sector_t sector = SECTOR_FROM_BLOCK(block);
4106 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
4107 int ret = 0;
4108
4109 if (flags & F2FS_TRIM_FILE_DISCARD) {
4110 if (bdev_max_secure_erase_sectors(bdev))
4111 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
4112 GFP_NOFS);
4113 else
4114 ret = blkdev_issue_discard(bdev, sector, nr_sects,
4115 GFP_NOFS);
4116 }
4117
4118 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
4119 if (IS_ENCRYPTED(inode))
4120 ret = fscrypt_zeroout_range(inode, off, block, len);
4121 else
4122 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
4123 GFP_NOFS, 0);
4124 }
4125
4126 return ret;
4127 }
4128
f2fs_sec_trim_file(struct file * filp,unsigned long arg)4129 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
4130 {
4131 struct inode *inode = file_inode(filp);
4132 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4133 struct address_space *mapping = inode->i_mapping;
4134 struct block_device *prev_bdev = NULL;
4135 struct f2fs_sectrim_range range;
4136 pgoff_t index, pg_end, prev_index = 0;
4137 block_t prev_block = 0, len = 0;
4138 loff_t end_addr;
4139 bool to_end = false;
4140 int ret = 0;
4141
4142 if (!(filp->f_mode & FMODE_WRITE))
4143 return -EBADF;
4144
4145 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
4146 sizeof(range)))
4147 return -EFAULT;
4148
4149 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
4150 !S_ISREG(inode->i_mode))
4151 return -EINVAL;
4152
4153 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
4154 !f2fs_hw_support_discard(sbi)) ||
4155 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
4156 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
4157 return -EOPNOTSUPP;
4158
4159 ret = mnt_want_write_file(filp);
4160 if (ret)
4161 return ret;
4162 inode_lock(inode);
4163
4164 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
4165 range.start >= inode->i_size) {
4166 ret = -EINVAL;
4167 goto err;
4168 }
4169
4170 if (range.len == 0)
4171 goto err;
4172
4173 if (inode->i_size - range.start > range.len) {
4174 end_addr = range.start + range.len;
4175 } else {
4176 end_addr = range.len == (u64)-1 ?
4177 sbi->sb->s_maxbytes : inode->i_size;
4178 to_end = true;
4179 }
4180
4181 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
4182 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
4183 ret = -EINVAL;
4184 goto err;
4185 }
4186
4187 index = F2FS_BYTES_TO_BLK(range.start);
4188 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
4189
4190 ret = f2fs_convert_inline_inode(inode);
4191 if (ret)
4192 goto err;
4193
4194 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4195 filemap_invalidate_lock(mapping);
4196
4197 ret = filemap_write_and_wait_range(mapping, range.start,
4198 to_end ? LLONG_MAX : end_addr - 1);
4199 if (ret)
4200 goto out;
4201
4202 truncate_inode_pages_range(mapping, range.start,
4203 to_end ? -1 : end_addr - 1);
4204
4205 while (index < pg_end) {
4206 struct dnode_of_data dn;
4207 pgoff_t end_offset, count;
4208 int i;
4209
4210 set_new_dnode(&dn, inode, NULL, NULL, 0);
4211 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4212 if (ret) {
4213 if (ret == -ENOENT) {
4214 index = f2fs_get_next_page_offset(&dn, index);
4215 continue;
4216 }
4217 goto out;
4218 }
4219
4220 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
4221 count = min(end_offset - dn.ofs_in_node, pg_end - index);
4222 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
4223 struct block_device *cur_bdev;
4224 block_t blkaddr = f2fs_data_blkaddr(&dn);
4225
4226 if (!__is_valid_data_blkaddr(blkaddr))
4227 continue;
4228
4229 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
4230 DATA_GENERIC_ENHANCE)) {
4231 ret = -EFSCORRUPTED;
4232 f2fs_put_dnode(&dn);
4233 goto out;
4234 }
4235
4236 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
4237 if (f2fs_is_multi_device(sbi)) {
4238 int di = f2fs_target_device_index(sbi, blkaddr);
4239
4240 blkaddr -= FDEV(di).start_blk;
4241 }
4242
4243 if (len) {
4244 if (prev_bdev == cur_bdev &&
4245 index == prev_index + len &&
4246 blkaddr == prev_block + len) {
4247 len++;
4248 } else {
4249 ret = f2fs_secure_erase(prev_bdev,
4250 inode, prev_index, prev_block,
4251 len, range.flags);
4252 if (ret) {
4253 f2fs_put_dnode(&dn);
4254 goto out;
4255 }
4256
4257 len = 0;
4258 }
4259 }
4260
4261 if (!len) {
4262 prev_bdev = cur_bdev;
4263 prev_index = index;
4264 prev_block = blkaddr;
4265 len = 1;
4266 }
4267 }
4268
4269 f2fs_put_dnode(&dn);
4270
4271 if (fatal_signal_pending(current)) {
4272 ret = -EINTR;
4273 goto out;
4274 }
4275 cond_resched();
4276 }
4277
4278 if (len)
4279 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
4280 prev_block, len, range.flags);
4281 f2fs_update_time(sbi, REQ_TIME);
4282 out:
4283 filemap_invalidate_unlock(mapping);
4284 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4285 err:
4286 inode_unlock(inode);
4287 mnt_drop_write_file(filp);
4288
4289 return ret;
4290 }
4291
f2fs_ioc_get_compress_option(struct file * filp,unsigned long arg)4292 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
4293 {
4294 struct inode *inode = file_inode(filp);
4295 struct f2fs_comp_option option;
4296
4297 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4298 return -EOPNOTSUPP;
4299
4300 inode_lock_shared(inode);
4301
4302 if (!f2fs_compressed_file(inode)) {
4303 inode_unlock_shared(inode);
4304 return -ENODATA;
4305 }
4306
4307 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
4308 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
4309
4310 inode_unlock_shared(inode);
4311
4312 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
4313 sizeof(option)))
4314 return -EFAULT;
4315
4316 return 0;
4317 }
4318
f2fs_ioc_set_compress_option(struct file * filp,unsigned long arg)4319 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4320 {
4321 struct inode *inode = file_inode(filp);
4322 struct f2fs_inode_info *fi = F2FS_I(inode);
4323 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4324 struct f2fs_comp_option option;
4325 int ret = 0;
4326
4327 if (!f2fs_sb_has_compression(sbi))
4328 return -EOPNOTSUPP;
4329
4330 if (!(filp->f_mode & FMODE_WRITE))
4331 return -EBADF;
4332
4333 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4334 sizeof(option)))
4335 return -EFAULT;
4336
4337 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
4338 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
4339 option.algorithm >= COMPRESS_MAX)
4340 return -EINVAL;
4341
4342 ret = mnt_want_write_file(filp);
4343 if (ret)
4344 return ret;
4345 inode_lock(inode);
4346
4347 f2fs_down_write(&F2FS_I(inode)->i_sem);
4348 if (!f2fs_compressed_file(inode)) {
4349 ret = -EINVAL;
4350 goto out;
4351 }
4352
4353 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4354 ret = -EBUSY;
4355 goto out;
4356 }
4357
4358 if (F2FS_HAS_BLOCKS(inode)) {
4359 ret = -EFBIG;
4360 goto out;
4361 }
4362
4363 fi->i_compress_algorithm = option.algorithm;
4364 fi->i_log_cluster_size = option.log_cluster_size;
4365 fi->i_cluster_size = BIT(option.log_cluster_size);
4366 /* Set default level */
4367 if (fi->i_compress_algorithm == COMPRESS_ZSTD)
4368 fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
4369 else
4370 fi->i_compress_level = 0;
4371 /* Adjust mount option level */
4372 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm &&
4373 F2FS_OPTION(sbi).compress_level)
4374 fi->i_compress_level = F2FS_OPTION(sbi).compress_level;
4375 f2fs_mark_inode_dirty_sync(inode, true);
4376
4377 if (!f2fs_is_compress_backend_ready(inode))
4378 f2fs_warn(sbi, "compression algorithm is successfully set, "
4379 "but current kernel doesn't support this algorithm.");
4380 out:
4381 f2fs_up_write(&fi->i_sem);
4382 inode_unlock(inode);
4383 mnt_drop_write_file(filp);
4384
4385 return ret;
4386 }
4387
redirty_blocks(struct inode * inode,pgoff_t page_idx,int len)4388 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4389 {
4390 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
4391 struct address_space *mapping = inode->i_mapping;
4392 struct folio *folio;
4393 pgoff_t redirty_idx = page_idx;
4394 int page_len = 0, ret = 0;
4395
4396 page_cache_ra_unbounded(&ractl, len, 0);
4397
4398 do {
4399 folio = read_cache_folio(mapping, page_idx, NULL, NULL);
4400 if (IS_ERR(folio)) {
4401 ret = PTR_ERR(folio);
4402 break;
4403 }
4404 page_len += folio_nr_pages(folio) - (page_idx - folio->index);
4405 page_idx = folio_next_index(folio);
4406 } while (page_len < len);
4407
4408 do {
4409 folio = filemap_lock_folio(mapping, redirty_idx);
4410
4411 /* It will never fail, when folio has pinned above */
4412 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(folio));
4413
4414 f2fs_folio_wait_writeback(folio, DATA, true, true);
4415
4416 folio_mark_dirty(folio);
4417 set_page_private_gcing(&folio->page);
4418 redirty_idx = folio_next_index(folio);
4419 folio_unlock(folio);
4420 folio_put_refs(folio, 2);
4421 } while (redirty_idx < page_idx);
4422
4423 return ret;
4424 }
4425
f2fs_ioc_decompress_file(struct file * filp)4426 static int f2fs_ioc_decompress_file(struct file *filp)
4427 {
4428 struct inode *inode = file_inode(filp);
4429 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4430 struct f2fs_inode_info *fi = F2FS_I(inode);
4431 pgoff_t page_idx = 0, last_idx, cluster_idx;
4432 int ret;
4433
4434 if (!f2fs_sb_has_compression(sbi) ||
4435 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4436 return -EOPNOTSUPP;
4437
4438 if (!(filp->f_mode & FMODE_WRITE))
4439 return -EBADF;
4440
4441 f2fs_balance_fs(sbi, true);
4442
4443 ret = mnt_want_write_file(filp);
4444 if (ret)
4445 return ret;
4446 inode_lock(inode);
4447
4448 if (!f2fs_is_compress_backend_ready(inode)) {
4449 ret = -EOPNOTSUPP;
4450 goto out;
4451 }
4452
4453 if (!f2fs_compressed_file(inode) ||
4454 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4455 ret = -EINVAL;
4456 goto out;
4457 }
4458
4459 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4460 if (ret)
4461 goto out;
4462
4463 if (!atomic_read(&fi->i_compr_blocks))
4464 goto out;
4465
4466 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4467 last_idx >>= fi->i_log_cluster_size;
4468
4469 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4470 page_idx = cluster_idx << fi->i_log_cluster_size;
4471
4472 if (!f2fs_is_compressed_cluster(inode, page_idx))
4473 continue;
4474
4475 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4476 if (ret < 0)
4477 break;
4478
4479 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4480 ret = filemap_fdatawrite(inode->i_mapping);
4481 if (ret < 0)
4482 break;
4483 }
4484
4485 cond_resched();
4486 if (fatal_signal_pending(current)) {
4487 ret = -EINTR;
4488 break;
4489 }
4490 }
4491
4492 if (!ret)
4493 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4494 LLONG_MAX);
4495
4496 if (ret)
4497 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4498 __func__, ret);
4499 f2fs_update_time(sbi, REQ_TIME);
4500 out:
4501 inode_unlock(inode);
4502 mnt_drop_write_file(filp);
4503
4504 return ret;
4505 }
4506
f2fs_ioc_compress_file(struct file * filp)4507 static int f2fs_ioc_compress_file(struct file *filp)
4508 {
4509 struct inode *inode = file_inode(filp);
4510 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4511 struct f2fs_inode_info *fi = F2FS_I(inode);
4512 pgoff_t page_idx = 0, last_idx, cluster_idx;
4513 int ret;
4514
4515 if (!f2fs_sb_has_compression(sbi) ||
4516 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4517 return -EOPNOTSUPP;
4518
4519 if (!(filp->f_mode & FMODE_WRITE))
4520 return -EBADF;
4521
4522 f2fs_balance_fs(sbi, true);
4523
4524 ret = mnt_want_write_file(filp);
4525 if (ret)
4526 return ret;
4527 inode_lock(inode);
4528
4529 if (!f2fs_is_compress_backend_ready(inode)) {
4530 ret = -EOPNOTSUPP;
4531 goto out;
4532 }
4533
4534 if (!f2fs_compressed_file(inode) ||
4535 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4536 ret = -EINVAL;
4537 goto out;
4538 }
4539
4540 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4541 if (ret)
4542 goto out;
4543
4544 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4545
4546 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4547 last_idx >>= fi->i_log_cluster_size;
4548
4549 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4550 page_idx = cluster_idx << fi->i_log_cluster_size;
4551
4552 if (f2fs_is_sparse_cluster(inode, page_idx))
4553 continue;
4554
4555 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
4556 if (ret < 0)
4557 break;
4558
4559 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
4560 ret = filemap_fdatawrite(inode->i_mapping);
4561 if (ret < 0)
4562 break;
4563 }
4564
4565 cond_resched();
4566 if (fatal_signal_pending(current)) {
4567 ret = -EINTR;
4568 break;
4569 }
4570 }
4571
4572 if (!ret)
4573 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4574 LLONG_MAX);
4575
4576 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4577
4578 if (ret)
4579 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4580 __func__, ret);
4581 f2fs_update_time(sbi, REQ_TIME);
4582 out:
4583 inode_unlock(inode);
4584 mnt_drop_write_file(filp);
4585
4586 return ret;
4587 }
4588
__f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4589 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4590 {
4591 switch (cmd) {
4592 case FS_IOC_GETVERSION:
4593 return f2fs_ioc_getversion(filp, arg);
4594 case F2FS_IOC_START_ATOMIC_WRITE:
4595 return f2fs_ioc_start_atomic_write(filp, false);
4596 case F2FS_IOC_START_ATOMIC_REPLACE:
4597 return f2fs_ioc_start_atomic_write(filp, true);
4598 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4599 return f2fs_ioc_commit_atomic_write(filp);
4600 case F2FS_IOC_ABORT_ATOMIC_WRITE:
4601 return f2fs_ioc_abort_atomic_write(filp);
4602 case F2FS_IOC_START_VOLATILE_WRITE:
4603 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4604 return -EOPNOTSUPP;
4605 case F2FS_IOC_SHUTDOWN:
4606 return f2fs_ioc_shutdown(filp, arg);
4607 case FITRIM:
4608 return f2fs_ioc_fitrim(filp, arg);
4609 case FS_IOC_SET_ENCRYPTION_POLICY:
4610 return f2fs_ioc_set_encryption_policy(filp, arg);
4611 case FS_IOC_GET_ENCRYPTION_POLICY:
4612 return f2fs_ioc_get_encryption_policy(filp, arg);
4613 case FS_IOC_GET_ENCRYPTION_PWSALT:
4614 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4615 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4616 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4617 case FS_IOC_ADD_ENCRYPTION_KEY:
4618 return f2fs_ioc_add_encryption_key(filp, arg);
4619 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4620 return f2fs_ioc_remove_encryption_key(filp, arg);
4621 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4622 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4623 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4624 return f2fs_ioc_get_encryption_key_status(filp, arg);
4625 case FS_IOC_GET_ENCRYPTION_NONCE:
4626 return f2fs_ioc_get_encryption_nonce(filp, arg);
4627 case F2FS_IOC_GARBAGE_COLLECT:
4628 return f2fs_ioc_gc(filp, arg);
4629 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4630 return f2fs_ioc_gc_range(filp, arg);
4631 case F2FS_IOC_WRITE_CHECKPOINT:
4632 return f2fs_ioc_write_checkpoint(filp);
4633 case F2FS_IOC_DEFRAGMENT:
4634 return f2fs_ioc_defragment(filp, arg);
4635 case F2FS_IOC_MOVE_RANGE:
4636 return f2fs_ioc_move_range(filp, arg);
4637 case F2FS_IOC_FLUSH_DEVICE:
4638 return f2fs_ioc_flush_device(filp, arg);
4639 case F2FS_IOC_GET_FEATURES:
4640 return f2fs_ioc_get_features(filp, arg);
4641 case F2FS_IOC_GET_PIN_FILE:
4642 return f2fs_ioc_get_pin_file(filp, arg);
4643 case F2FS_IOC_SET_PIN_FILE:
4644 return f2fs_ioc_set_pin_file(filp, arg);
4645 case F2FS_IOC_PRECACHE_EXTENTS:
4646 return f2fs_ioc_precache_extents(filp);
4647 case F2FS_IOC_RESIZE_FS:
4648 return f2fs_ioc_resize_fs(filp, arg);
4649 case FS_IOC_ENABLE_VERITY:
4650 return f2fs_ioc_enable_verity(filp, arg);
4651 case FS_IOC_MEASURE_VERITY:
4652 return f2fs_ioc_measure_verity(filp, arg);
4653 case FS_IOC_READ_VERITY_METADATA:
4654 return f2fs_ioc_read_verity_metadata(filp, arg);
4655 case FS_IOC_GETFSLABEL:
4656 return f2fs_ioc_getfslabel(filp, arg);
4657 case FS_IOC_SETFSLABEL:
4658 return f2fs_ioc_setfslabel(filp, arg);
4659 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4660 return f2fs_ioc_get_compress_blocks(filp, arg);
4661 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4662 return f2fs_release_compress_blocks(filp, arg);
4663 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4664 return f2fs_reserve_compress_blocks(filp, arg);
4665 case F2FS_IOC_SEC_TRIM_FILE:
4666 return f2fs_sec_trim_file(filp, arg);
4667 case F2FS_IOC_GET_COMPRESS_OPTION:
4668 return f2fs_ioc_get_compress_option(filp, arg);
4669 case F2FS_IOC_SET_COMPRESS_OPTION:
4670 return f2fs_ioc_set_compress_option(filp, arg);
4671 case F2FS_IOC_DECOMPRESS_FILE:
4672 return f2fs_ioc_decompress_file(filp);
4673 case F2FS_IOC_COMPRESS_FILE:
4674 return f2fs_ioc_compress_file(filp);
4675 case F2FS_IOC_GET_DEV_ALIAS_FILE:
4676 return f2fs_ioc_get_dev_alias_file(filp, arg);
4677 case F2FS_IOC_IO_PRIO:
4678 return f2fs_ioc_io_prio(filp, arg);
4679 default:
4680 return -ENOTTY;
4681 }
4682 }
4683
f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4684 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4685 {
4686 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4687 return -EIO;
4688 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4689 return -ENOSPC;
4690
4691 return __f2fs_ioctl(filp, cmd, arg);
4692 }
4693
4694 /*
4695 * Return %true if the given read or write request should use direct I/O, or
4696 * %false if it should use buffered I/O.
4697 */
f2fs_should_use_dio(struct inode * inode,struct kiocb * iocb,struct iov_iter * iter)4698 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4699 struct iov_iter *iter)
4700 {
4701 unsigned int align;
4702
4703 if (!(iocb->ki_flags & IOCB_DIRECT))
4704 return false;
4705
4706 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4707 return false;
4708
4709 /*
4710 * Direct I/O not aligned to the disk's logical_block_size will be
4711 * attempted, but will fail with -EINVAL.
4712 *
4713 * f2fs additionally requires that direct I/O be aligned to the
4714 * filesystem block size, which is often a stricter requirement.
4715 * However, f2fs traditionally falls back to buffered I/O on requests
4716 * that are logical_block_size-aligned but not fs-block aligned.
4717 *
4718 * The below logic implements this behavior.
4719 */
4720 align = iocb->ki_pos | iov_iter_alignment(iter);
4721 if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4722 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4723 return false;
4724
4725 return true;
4726 }
4727
f2fs_dio_read_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4728 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4729 unsigned int flags)
4730 {
4731 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4732
4733 dec_page_count(sbi, F2FS_DIO_READ);
4734 if (error)
4735 return error;
4736 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4737 return 0;
4738 }
4739
4740 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4741 .end_io = f2fs_dio_read_end_io,
4742 };
4743
f2fs_dio_read_iter(struct kiocb * iocb,struct iov_iter * to)4744 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4745 {
4746 struct file *file = iocb->ki_filp;
4747 struct inode *inode = file_inode(file);
4748 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4749 struct f2fs_inode_info *fi = F2FS_I(inode);
4750 const loff_t pos = iocb->ki_pos;
4751 const size_t count = iov_iter_count(to);
4752 struct iomap_dio *dio;
4753 ssize_t ret;
4754
4755 if (count == 0)
4756 return 0; /* skip atime update */
4757
4758 trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4759
4760 if (iocb->ki_flags & IOCB_NOWAIT) {
4761 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4762 ret = -EAGAIN;
4763 goto out;
4764 }
4765 } else {
4766 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4767 }
4768
4769 /* dio is not compatible w/ atomic file */
4770 if (f2fs_is_atomic_file(inode)) {
4771 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4772 ret = -EOPNOTSUPP;
4773 goto out;
4774 }
4775
4776 /*
4777 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4778 * the higher-level function iomap_dio_rw() in order to ensure that the
4779 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4780 */
4781 inc_page_count(sbi, F2FS_DIO_READ);
4782 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4783 &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4784 if (IS_ERR_OR_NULL(dio)) {
4785 ret = PTR_ERR_OR_ZERO(dio);
4786 if (ret != -EIOCBQUEUED)
4787 dec_page_count(sbi, F2FS_DIO_READ);
4788 } else {
4789 ret = iomap_dio_complete(dio);
4790 }
4791
4792 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4793
4794 file_accessed(file);
4795 out:
4796 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4797 return ret;
4798 }
4799
f2fs_trace_rw_file_path(struct file * file,loff_t pos,size_t count,int rw)4800 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count,
4801 int rw)
4802 {
4803 struct inode *inode = file_inode(file);
4804 char *buf, *path;
4805
4806 buf = f2fs_getname(F2FS_I_SB(inode));
4807 if (!buf)
4808 return;
4809 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX);
4810 if (IS_ERR(path))
4811 goto free_buf;
4812 if (rw == WRITE)
4813 trace_f2fs_datawrite_start(inode, pos, count,
4814 current->pid, path, current->comm);
4815 else
4816 trace_f2fs_dataread_start(inode, pos, count,
4817 current->pid, path, current->comm);
4818 free_buf:
4819 f2fs_putname(buf);
4820 }
4821
f2fs_file_read_iter(struct kiocb * iocb,struct iov_iter * to)4822 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4823 {
4824 struct inode *inode = file_inode(iocb->ki_filp);
4825 const loff_t pos = iocb->ki_pos;
4826 ssize_t ret;
4827
4828 if (!f2fs_is_compress_backend_ready(inode))
4829 return -EOPNOTSUPP;
4830
4831 if (trace_f2fs_dataread_start_enabled())
4832 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
4833 iov_iter_count(to), READ);
4834
4835 /* In LFS mode, if there is inflight dio, wait for its completion */
4836 if (f2fs_lfs_mode(F2FS_I_SB(inode)) &&
4837 get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE))
4838 inode_dio_wait(inode);
4839
4840 if (f2fs_should_use_dio(inode, iocb, to)) {
4841 ret = f2fs_dio_read_iter(iocb, to);
4842 } else {
4843 ret = filemap_read(iocb, to, 0);
4844 if (ret > 0)
4845 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4846 APP_BUFFERED_READ_IO, ret);
4847 }
4848 if (trace_f2fs_dataread_end_enabled())
4849 trace_f2fs_dataread_end(inode, pos, ret);
4850 return ret;
4851 }
4852
f2fs_file_splice_read(struct file * in,loff_t * ppos,struct pipe_inode_info * pipe,size_t len,unsigned int flags)4853 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos,
4854 struct pipe_inode_info *pipe,
4855 size_t len, unsigned int flags)
4856 {
4857 struct inode *inode = file_inode(in);
4858 const loff_t pos = *ppos;
4859 ssize_t ret;
4860
4861 if (!f2fs_is_compress_backend_ready(inode))
4862 return -EOPNOTSUPP;
4863
4864 if (trace_f2fs_dataread_start_enabled())
4865 f2fs_trace_rw_file_path(in, pos, len, READ);
4866
4867 ret = filemap_splice_read(in, ppos, pipe, len, flags);
4868 if (ret > 0)
4869 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4870 APP_BUFFERED_READ_IO, ret);
4871
4872 if (trace_f2fs_dataread_end_enabled())
4873 trace_f2fs_dataread_end(inode, pos, ret);
4874 return ret;
4875 }
4876
f2fs_write_checks(struct kiocb * iocb,struct iov_iter * from)4877 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4878 {
4879 struct file *file = iocb->ki_filp;
4880 struct inode *inode = file_inode(file);
4881 ssize_t count;
4882 int err;
4883
4884 if (IS_IMMUTABLE(inode))
4885 return -EPERM;
4886
4887 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4888 return -EPERM;
4889
4890 count = generic_write_checks(iocb, from);
4891 if (count <= 0)
4892 return count;
4893
4894 err = file_modified(file);
4895 if (err)
4896 return err;
4897
4898 filemap_invalidate_lock(inode->i_mapping);
4899 f2fs_zero_post_eof_page(inode, iocb->ki_pos + iov_iter_count(from));
4900 filemap_invalidate_unlock(inode->i_mapping);
4901 return count;
4902 }
4903
4904 /*
4905 * Preallocate blocks for a write request, if it is possible and helpful to do
4906 * so. Returns a positive number if blocks may have been preallocated, 0 if no
4907 * blocks were preallocated, or a negative errno value if something went
4908 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4909 * requested blocks (not just some of them) have been allocated.
4910 */
f2fs_preallocate_blocks(struct kiocb * iocb,struct iov_iter * iter,bool dio)4911 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4912 bool dio)
4913 {
4914 struct inode *inode = file_inode(iocb->ki_filp);
4915 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4916 const loff_t pos = iocb->ki_pos;
4917 const size_t count = iov_iter_count(iter);
4918 struct f2fs_map_blocks map = {};
4919 int flag;
4920 int ret;
4921
4922 /* If it will be an out-of-place direct write, don't bother. */
4923 if (dio && f2fs_lfs_mode(sbi))
4924 return 0;
4925 /*
4926 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4927 * buffered IO, if DIO meets any holes.
4928 */
4929 if (dio && i_size_read(inode) &&
4930 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4931 return 0;
4932
4933 /* No-wait I/O can't allocate blocks. */
4934 if (iocb->ki_flags & IOCB_NOWAIT)
4935 return 0;
4936
4937 /* If it will be a short write, don't bother. */
4938 if (fault_in_iov_iter_readable(iter, count))
4939 return 0;
4940
4941 if (f2fs_has_inline_data(inode)) {
4942 /* If the data will fit inline, don't bother. */
4943 if (pos + count <= MAX_INLINE_DATA(inode))
4944 return 0;
4945 ret = f2fs_convert_inline_inode(inode);
4946 if (ret)
4947 return ret;
4948 }
4949
4950 /* Do not preallocate blocks that will be written partially in 4KB. */
4951 map.m_lblk = F2FS_BLK_ALIGN(pos);
4952 map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4953 if (map.m_len > map.m_lblk)
4954 map.m_len -= map.m_lblk;
4955 else
4956 return 0;
4957
4958 if (!IS_DEVICE_ALIASING(inode))
4959 map.m_may_create = true;
4960 if (dio) {
4961 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi,
4962 inode->i_write_hint);
4963 flag = F2FS_GET_BLOCK_PRE_DIO;
4964 } else {
4965 map.m_seg_type = NO_CHECK_TYPE;
4966 flag = F2FS_GET_BLOCK_PRE_AIO;
4967 }
4968
4969 ret = f2fs_map_blocks(inode, &map, flag);
4970 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4971 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4972 return ret;
4973 if (ret == 0)
4974 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4975 return map.m_len;
4976 }
4977
f2fs_buffered_write_iter(struct kiocb * iocb,struct iov_iter * from)4978 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4979 struct iov_iter *from)
4980 {
4981 struct file *file = iocb->ki_filp;
4982 struct inode *inode = file_inode(file);
4983 ssize_t ret;
4984
4985 if (iocb->ki_flags & IOCB_NOWAIT)
4986 return -EOPNOTSUPP;
4987
4988 ret = generic_perform_write(iocb, from);
4989
4990 if (ret > 0) {
4991 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4992 APP_BUFFERED_IO, ret);
4993 }
4994 return ret;
4995 }
4996
f2fs_dio_write_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4997 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4998 unsigned int flags)
4999 {
5000 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
5001
5002 dec_page_count(sbi, F2FS_DIO_WRITE);
5003 if (error)
5004 return error;
5005 f2fs_update_time(sbi, REQ_TIME);
5006 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
5007 return 0;
5008 }
5009
f2fs_dio_write_submit_io(const struct iomap_iter * iter,struct bio * bio,loff_t file_offset)5010 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter,
5011 struct bio *bio, loff_t file_offset)
5012 {
5013 struct inode *inode = iter->inode;
5014 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5015 enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint);
5016 enum temp_type temp = f2fs_get_segment_temp(sbi, type);
5017
5018 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp);
5019 submit_bio(bio);
5020 }
5021
5022 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
5023 .end_io = f2fs_dio_write_end_io,
5024 .submit_io = f2fs_dio_write_submit_io,
5025 };
5026
f2fs_flush_buffered_write(struct address_space * mapping,loff_t start_pos,loff_t end_pos)5027 static void f2fs_flush_buffered_write(struct address_space *mapping,
5028 loff_t start_pos, loff_t end_pos)
5029 {
5030 int ret;
5031
5032 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos);
5033 if (ret < 0)
5034 return;
5035 invalidate_mapping_pages(mapping,
5036 start_pos >> PAGE_SHIFT,
5037 end_pos >> PAGE_SHIFT);
5038 }
5039
f2fs_dio_write_iter(struct kiocb * iocb,struct iov_iter * from,bool * may_need_sync)5040 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
5041 bool *may_need_sync)
5042 {
5043 struct file *file = iocb->ki_filp;
5044 struct inode *inode = file_inode(file);
5045 struct f2fs_inode_info *fi = F2FS_I(inode);
5046 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5047 const bool do_opu = f2fs_lfs_mode(sbi);
5048 const loff_t pos = iocb->ki_pos;
5049 const ssize_t count = iov_iter_count(from);
5050 unsigned int dio_flags;
5051 struct iomap_dio *dio;
5052 ssize_t ret;
5053
5054 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
5055
5056 if (iocb->ki_flags & IOCB_NOWAIT) {
5057 /* f2fs_convert_inline_inode() and block allocation can block */
5058 if (f2fs_has_inline_data(inode) ||
5059 !f2fs_overwrite_io(inode, pos, count)) {
5060 ret = -EAGAIN;
5061 goto out;
5062 }
5063
5064 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
5065 ret = -EAGAIN;
5066 goto out;
5067 }
5068 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
5069 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
5070 ret = -EAGAIN;
5071 goto out;
5072 }
5073 } else {
5074 ret = f2fs_convert_inline_inode(inode);
5075 if (ret)
5076 goto out;
5077
5078 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
5079 if (do_opu)
5080 f2fs_down_read(&fi->i_gc_rwsem[READ]);
5081 }
5082
5083 /*
5084 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
5085 * the higher-level function iomap_dio_rw() in order to ensure that the
5086 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
5087 */
5088 inc_page_count(sbi, F2FS_DIO_WRITE);
5089 dio_flags = 0;
5090 if (pos + count > inode->i_size)
5091 dio_flags |= IOMAP_DIO_FORCE_WAIT;
5092 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
5093 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
5094 if (IS_ERR_OR_NULL(dio)) {
5095 ret = PTR_ERR_OR_ZERO(dio);
5096 if (ret == -ENOTBLK)
5097 ret = 0;
5098 if (ret != -EIOCBQUEUED)
5099 dec_page_count(sbi, F2FS_DIO_WRITE);
5100 } else {
5101 ret = iomap_dio_complete(dio);
5102 }
5103
5104 if (do_opu)
5105 f2fs_up_read(&fi->i_gc_rwsem[READ]);
5106 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
5107
5108 if (ret < 0)
5109 goto out;
5110 if (pos + ret > inode->i_size)
5111 f2fs_i_size_write(inode, pos + ret);
5112 if (!do_opu)
5113 set_inode_flag(inode, FI_UPDATE_WRITE);
5114
5115 if (iov_iter_count(from)) {
5116 ssize_t ret2;
5117 loff_t bufio_start_pos = iocb->ki_pos;
5118
5119 /*
5120 * The direct write was partial, so we need to fall back to a
5121 * buffered write for the remainder.
5122 */
5123
5124 ret2 = f2fs_buffered_write_iter(iocb, from);
5125 if (iov_iter_count(from))
5126 f2fs_write_failed(inode, iocb->ki_pos);
5127 if (ret2 < 0)
5128 goto out;
5129
5130 /*
5131 * Ensure that the pagecache pages are written to disk and
5132 * invalidated to preserve the expected O_DIRECT semantics.
5133 */
5134 if (ret2 > 0) {
5135 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
5136
5137 ret += ret2;
5138
5139 f2fs_flush_buffered_write(file->f_mapping,
5140 bufio_start_pos,
5141 bufio_end_pos);
5142 }
5143 } else {
5144 /* iomap_dio_rw() already handled the generic_write_sync(). */
5145 *may_need_sync = false;
5146 }
5147 out:
5148 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
5149 return ret;
5150 }
5151
f2fs_file_write_iter(struct kiocb * iocb,struct iov_iter * from)5152 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
5153 {
5154 struct inode *inode = file_inode(iocb->ki_filp);
5155 const loff_t orig_pos = iocb->ki_pos;
5156 const size_t orig_count = iov_iter_count(from);
5157 loff_t target_size;
5158 bool dio;
5159 bool may_need_sync = true;
5160 int preallocated;
5161 const loff_t pos = iocb->ki_pos;
5162 const ssize_t count = iov_iter_count(from);
5163 ssize_t ret;
5164
5165 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
5166 ret = -EIO;
5167 goto out;
5168 }
5169
5170 if (!f2fs_is_compress_backend_ready(inode)) {
5171 ret = -EOPNOTSUPP;
5172 goto out;
5173 }
5174
5175 if (iocb->ki_flags & IOCB_NOWAIT) {
5176 if (!inode_trylock(inode)) {
5177 ret = -EAGAIN;
5178 goto out;
5179 }
5180 } else {
5181 inode_lock(inode);
5182 }
5183
5184 if (f2fs_is_pinned_file(inode) &&
5185 !f2fs_overwrite_io(inode, pos, count)) {
5186 ret = -EIO;
5187 goto out_unlock;
5188 }
5189
5190 ret = f2fs_write_checks(iocb, from);
5191 if (ret <= 0)
5192 goto out_unlock;
5193
5194 /* Determine whether we will do a direct write or a buffered write. */
5195 dio = f2fs_should_use_dio(inode, iocb, from);
5196
5197 /* dio is not compatible w/ atomic write */
5198 if (dio && f2fs_is_atomic_file(inode)) {
5199 ret = -EOPNOTSUPP;
5200 goto out_unlock;
5201 }
5202
5203 /* Possibly preallocate the blocks for the write. */
5204 target_size = iocb->ki_pos + iov_iter_count(from);
5205 preallocated = f2fs_preallocate_blocks(iocb, from, dio);
5206 if (preallocated < 0) {
5207 ret = preallocated;
5208 } else {
5209 if (trace_f2fs_datawrite_start_enabled())
5210 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
5211 orig_count, WRITE);
5212
5213 /* Do the actual write. */
5214 ret = dio ?
5215 f2fs_dio_write_iter(iocb, from, &may_need_sync) :
5216 f2fs_buffered_write_iter(iocb, from);
5217
5218 if (trace_f2fs_datawrite_end_enabled())
5219 trace_f2fs_datawrite_end(inode, orig_pos, ret);
5220 }
5221
5222 /* Don't leave any preallocated blocks around past i_size. */
5223 if (preallocated && i_size_read(inode) < target_size) {
5224 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5225 filemap_invalidate_lock(inode->i_mapping);
5226 if (!f2fs_truncate(inode))
5227 file_dont_truncate(inode);
5228 filemap_invalidate_unlock(inode->i_mapping);
5229 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
5230 } else {
5231 file_dont_truncate(inode);
5232 }
5233
5234 clear_inode_flag(inode, FI_PREALLOCATED_ALL);
5235 out_unlock:
5236 inode_unlock(inode);
5237 out:
5238 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
5239
5240 if (ret > 0 && may_need_sync)
5241 ret = generic_write_sync(iocb, ret);
5242
5243 /* If buffered IO was forced, flush and drop the data from
5244 * the page cache to preserve O_DIRECT semantics
5245 */
5246 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
5247 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
5248 orig_pos,
5249 orig_pos + ret - 1);
5250
5251 return ret;
5252 }
5253
f2fs_file_fadvise(struct file * filp,loff_t offset,loff_t len,int advice)5254 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
5255 int advice)
5256 {
5257 struct address_space *mapping;
5258 struct backing_dev_info *bdi;
5259 struct inode *inode = file_inode(filp);
5260 int err;
5261
5262 if (advice == POSIX_FADV_SEQUENTIAL) {
5263 if (S_ISFIFO(inode->i_mode))
5264 return -ESPIPE;
5265
5266 mapping = filp->f_mapping;
5267 if (!mapping || len < 0)
5268 return -EINVAL;
5269
5270 bdi = inode_to_bdi(mapping->host);
5271 filp->f_ra.ra_pages = bdi->ra_pages *
5272 F2FS_I_SB(inode)->seq_file_ra_mul;
5273 spin_lock(&filp->f_lock);
5274 filp->f_mode &= ~FMODE_RANDOM;
5275 spin_unlock(&filp->f_lock);
5276 return 0;
5277 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) {
5278 /* Load extent cache at the first readahead. */
5279 f2fs_precache_extents(inode);
5280 }
5281
5282 err = generic_fadvise(filp, offset, len, advice);
5283 if (err)
5284 return err;
5285
5286 if (advice == POSIX_FADV_DONTNEED &&
5287 (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
5288 f2fs_compressed_file(inode)))
5289 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
5290 else if (advice == POSIX_FADV_NOREUSE)
5291 err = f2fs_keep_noreuse_range(inode, offset, len);
5292 return err;
5293 }
5294
5295 #ifdef CONFIG_COMPAT
5296 struct compat_f2fs_gc_range {
5297 u32 sync;
5298 compat_u64 start;
5299 compat_u64 len;
5300 };
5301 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
5302 struct compat_f2fs_gc_range)
5303
f2fs_compat_ioc_gc_range(struct file * file,unsigned long arg)5304 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
5305 {
5306 struct compat_f2fs_gc_range __user *urange;
5307 struct f2fs_gc_range range;
5308 int err;
5309
5310 urange = compat_ptr(arg);
5311 err = get_user(range.sync, &urange->sync);
5312 err |= get_user(range.start, &urange->start);
5313 err |= get_user(range.len, &urange->len);
5314 if (err)
5315 return -EFAULT;
5316
5317 return __f2fs_ioc_gc_range(file, &range);
5318 }
5319
5320 struct compat_f2fs_move_range {
5321 u32 dst_fd;
5322 compat_u64 pos_in;
5323 compat_u64 pos_out;
5324 compat_u64 len;
5325 };
5326 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
5327 struct compat_f2fs_move_range)
5328
f2fs_compat_ioc_move_range(struct file * file,unsigned long arg)5329 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
5330 {
5331 struct compat_f2fs_move_range __user *urange;
5332 struct f2fs_move_range range;
5333 int err;
5334
5335 urange = compat_ptr(arg);
5336 err = get_user(range.dst_fd, &urange->dst_fd);
5337 err |= get_user(range.pos_in, &urange->pos_in);
5338 err |= get_user(range.pos_out, &urange->pos_out);
5339 err |= get_user(range.len, &urange->len);
5340 if (err)
5341 return -EFAULT;
5342
5343 return __f2fs_ioc_move_range(file, &range);
5344 }
5345
f2fs_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)5346 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5347 {
5348 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
5349 return -EIO;
5350 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
5351 return -ENOSPC;
5352
5353 switch (cmd) {
5354 case FS_IOC32_GETVERSION:
5355 cmd = FS_IOC_GETVERSION;
5356 break;
5357 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
5358 return f2fs_compat_ioc_gc_range(file, arg);
5359 case F2FS_IOC32_MOVE_RANGE:
5360 return f2fs_compat_ioc_move_range(file, arg);
5361 case F2FS_IOC_START_ATOMIC_WRITE:
5362 case F2FS_IOC_START_ATOMIC_REPLACE:
5363 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
5364 case F2FS_IOC_START_VOLATILE_WRITE:
5365 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
5366 case F2FS_IOC_ABORT_ATOMIC_WRITE:
5367 case F2FS_IOC_SHUTDOWN:
5368 case FITRIM:
5369 case FS_IOC_SET_ENCRYPTION_POLICY:
5370 case FS_IOC_GET_ENCRYPTION_PWSALT:
5371 case FS_IOC_GET_ENCRYPTION_POLICY:
5372 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
5373 case FS_IOC_ADD_ENCRYPTION_KEY:
5374 case FS_IOC_REMOVE_ENCRYPTION_KEY:
5375 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
5376 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
5377 case FS_IOC_GET_ENCRYPTION_NONCE:
5378 case F2FS_IOC_GARBAGE_COLLECT:
5379 case F2FS_IOC_WRITE_CHECKPOINT:
5380 case F2FS_IOC_DEFRAGMENT:
5381 case F2FS_IOC_FLUSH_DEVICE:
5382 case F2FS_IOC_GET_FEATURES:
5383 case F2FS_IOC_GET_PIN_FILE:
5384 case F2FS_IOC_SET_PIN_FILE:
5385 case F2FS_IOC_PRECACHE_EXTENTS:
5386 case F2FS_IOC_RESIZE_FS:
5387 case FS_IOC_ENABLE_VERITY:
5388 case FS_IOC_MEASURE_VERITY:
5389 case FS_IOC_READ_VERITY_METADATA:
5390 case FS_IOC_GETFSLABEL:
5391 case FS_IOC_SETFSLABEL:
5392 case F2FS_IOC_GET_COMPRESS_BLOCKS:
5393 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
5394 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
5395 case F2FS_IOC_SEC_TRIM_FILE:
5396 case F2FS_IOC_GET_COMPRESS_OPTION:
5397 case F2FS_IOC_SET_COMPRESS_OPTION:
5398 case F2FS_IOC_DECOMPRESS_FILE:
5399 case F2FS_IOC_COMPRESS_FILE:
5400 case F2FS_IOC_GET_DEV_ALIAS_FILE:
5401 case F2FS_IOC_IO_PRIO:
5402 break;
5403 default:
5404 return -ENOIOCTLCMD;
5405 }
5406 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5407 }
5408 #endif
5409
5410 const struct file_operations f2fs_file_operations = {
5411 .llseek = f2fs_llseek,
5412 .read_iter = f2fs_file_read_iter,
5413 .write_iter = f2fs_file_write_iter,
5414 .iopoll = iocb_bio_iopoll,
5415 .open = f2fs_file_open,
5416 .release = f2fs_release_file,
5417 .mmap = f2fs_file_mmap,
5418 .flush = f2fs_file_flush,
5419 .fsync = f2fs_sync_file,
5420 .fallocate = f2fs_fallocate,
5421 .unlocked_ioctl = f2fs_ioctl,
5422 #ifdef CONFIG_COMPAT
5423 .compat_ioctl = f2fs_compat_ioctl,
5424 #endif
5425 .splice_read = f2fs_file_splice_read,
5426 .splice_write = iter_file_splice_write,
5427 .fadvise = f2fs_file_fadvise,
5428 .fop_flags = FOP_BUFFER_RASYNC,
5429 };
5430