]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - fs/f2fs/file.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
UBUNTU: Ubuntu-5.15.0-39.42
[mirror_ubuntu-jammy-kernel.git] / fs / f2fs / file.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/file.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/buffer_head.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/types.h>
16 #include <linux/compat.h>
17 #include <linux/uaccess.h>
18 #include <linux/mount.h>
19 #include <linux/pagevec.h>
20 #include <linux/uio.h>
21 #include <linux/uuid.h>
22 #include <linux/file.h>
23 #include <linux/nls.h>
24 #include <linux/sched/signal.h>
25 #include <linux/fileattr.h>
26 #include <linux/fadvise.h>
27
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
38 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
39 {
40 struct inode *inode = file_inode(vmf->vma->vm_file);
41 vm_fault_t ret;
42
43 ret = filemap_fault(vmf);
44 if (!ret)
45 f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
46 F2FS_BLKSIZE);
47
48 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
49
50 return ret;
51 }
52
53 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
54 {
55 struct page *page = vmf->page;
56 struct inode *inode = file_inode(vmf->vma->vm_file);
57 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
58 struct dnode_of_data dn;
59 bool need_alloc = true;
60 int err = 0;
61
62 if (unlikely(IS_IMMUTABLE(inode)))
63 return VM_FAULT_SIGBUS;
64
65 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
66 return VM_FAULT_SIGBUS;
67
68 if (unlikely(f2fs_cp_error(sbi))) {
69 err = -EIO;
70 goto err;
71 }
72
73 if (!f2fs_is_checkpoint_ready(sbi)) {
74 err = -ENOSPC;
75 goto err;
76 }
77
78 err = f2fs_convert_inline_inode(inode);
79 if (err)
80 goto err;
81
82 #ifdef CONFIG_F2FS_FS_COMPRESSION
83 if (f2fs_compressed_file(inode)) {
84 int ret = f2fs_is_compressed_cluster(inode, page->index);
85
86 if (ret < 0) {
87 err = ret;
88 goto err;
89 } else if (ret) {
90 need_alloc = false;
91 }
92 }
93 #endif
94 /* should do out of any locked page */
95 if (need_alloc)
96 f2fs_balance_fs(sbi, true);
97
98 sb_start_pagefault(inode->i_sb);
99
100 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
101
102 file_update_time(vmf->vma->vm_file);
103 filemap_invalidate_lock_shared(inode->i_mapping);
104 lock_page(page);
105 if (unlikely(page->mapping != inode->i_mapping ||
106 page_offset(page) > i_size_read(inode) ||
107 !PageUptodate(page))) {
108 unlock_page(page);
109 err = -EFAULT;
110 goto out_sem;
111 }
112
113 if (need_alloc) {
114 /* block allocation */
115 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
116 set_new_dnode(&dn, inode, NULL, NULL, 0);
117 err = f2fs_get_block(&dn, page->index);
118 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
119 }
120
121 #ifdef CONFIG_F2FS_FS_COMPRESSION
122 if (!need_alloc) {
123 set_new_dnode(&dn, inode, NULL, NULL, 0);
124 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
125 f2fs_put_dnode(&dn);
126 }
127 #endif
128 if (err) {
129 unlock_page(page);
130 goto out_sem;
131 }
132
133 f2fs_wait_on_page_writeback(page, DATA, false, true);
134
135 /* wait for GCed page writeback via META_MAPPING */
136 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
137
138 /*
139 * check to see if the page is mapped already (no holes)
140 */
141 if (PageMappedToDisk(page))
142 goto out_sem;
143
144 /* page is wholly or partially inside EOF */
145 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
146 i_size_read(inode)) {
147 loff_t offset;
148
149 offset = i_size_read(inode) & ~PAGE_MASK;
150 zero_user_segment(page, offset, PAGE_SIZE);
151 }
152 set_page_dirty(page);
153 if (!PageUptodate(page))
154 SetPageUptodate(page);
155
156 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
157 f2fs_update_time(sbi, REQ_TIME);
158
159 trace_f2fs_vm_page_mkwrite(page, DATA);
160 out_sem:
161 filemap_invalidate_unlock_shared(inode->i_mapping);
162
163 sb_end_pagefault(inode->i_sb);
164 err:
165 return block_page_mkwrite_return(err);
166 }
167
168 static const struct vm_operations_struct f2fs_file_vm_ops = {
169 .fault = f2fs_filemap_fault,
170 .map_pages = filemap_map_pages,
171 .page_mkwrite = f2fs_vm_page_mkwrite,
172 };
173
174 static int get_parent_ino(struct inode *inode, nid_t *pino)
175 {
176 struct dentry *dentry;
177
178 /*
179 * Make sure to get the non-deleted alias. The alias associated with
180 * the open file descriptor being fsync()'ed may be deleted already.
181 */
182 dentry = d_find_alias(inode);
183 if (!dentry)
184 return 0;
185
186 *pino = parent_ino(dentry);
187 dput(dentry);
188 return 1;
189 }
190
191 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
192 {
193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
194 enum cp_reason_type cp_reason = CP_NO_NEEDED;
195
196 if (!S_ISREG(inode->i_mode))
197 cp_reason = CP_NON_REGULAR;
198 else if (f2fs_compressed_file(inode))
199 cp_reason = CP_COMPRESSED;
200 else if (inode->i_nlink != 1)
201 cp_reason = CP_HARDLINK;
202 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
203 cp_reason = CP_SB_NEED_CP;
204 else if (file_wrong_pino(inode))
205 cp_reason = CP_WRONG_PINO;
206 else if (!f2fs_space_for_roll_forward(sbi))
207 cp_reason = CP_NO_SPC_ROLL;
208 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
209 cp_reason = CP_NODE_NEED_CP;
210 else if (test_opt(sbi, FASTBOOT))
211 cp_reason = CP_FASTBOOT_MODE;
212 else if (F2FS_OPTION(sbi).active_logs == 2)
213 cp_reason = CP_SPEC_LOG_NUM;
214 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
215 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
216 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
217 TRANS_DIR_INO))
218 cp_reason = CP_RECOVER_DIR;
219
220 return cp_reason;
221 }
222
223 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
224 {
225 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
226 bool ret = false;
227 /* But we need to avoid that there are some inode updates */
228 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
229 ret = true;
230 f2fs_put_page(i, 0);
231 return ret;
232 }
233
234 static void try_to_fix_pino(struct inode *inode)
235 {
236 struct f2fs_inode_info *fi = F2FS_I(inode);
237 nid_t pino;
238
239 down_write(&fi->i_sem);
240 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
241 get_parent_ino(inode, &pino)) {
242 f2fs_i_pino_write(inode, pino);
243 file_got_pino(inode);
244 }
245 up_write(&fi->i_sem);
246 }
247
248 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
249 int datasync, bool atomic)
250 {
251 struct inode *inode = file->f_mapping->host;
252 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
253 nid_t ino = inode->i_ino;
254 int ret = 0;
255 enum cp_reason_type cp_reason = 0;
256 struct writeback_control wbc = {
257 .sync_mode = WB_SYNC_ALL,
258 .nr_to_write = LONG_MAX,
259 .for_reclaim = 0,
260 };
261 unsigned int seq_id = 0;
262
263 if (unlikely(f2fs_readonly(inode->i_sb)))
264 return 0;
265
266 trace_f2fs_sync_file_enter(inode);
267
268 if (S_ISDIR(inode->i_mode))
269 goto go_write;
270
271 /* if fdatasync is triggered, let's do in-place-update */
272 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
273 set_inode_flag(inode, FI_NEED_IPU);
274 ret = file_write_and_wait_range(file, start, end);
275 clear_inode_flag(inode, FI_NEED_IPU);
276
277 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
278 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
279 return ret;
280 }
281
282 /* if the inode is dirty, let's recover all the time */
283 if (!f2fs_skip_inode_update(inode, datasync)) {
284 f2fs_write_inode(inode, NULL);
285 goto go_write;
286 }
287
288 /*
289 * if there is no written data, don't waste time to write recovery info.
290 */
291 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
292 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
293
294 /* it may call write_inode just prior to fsync */
295 if (need_inode_page_update(sbi, ino))
296 goto go_write;
297
298 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
299 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
300 goto flush_out;
301 goto out;
302 } else {
303 /*
304 * for OPU case, during fsync(), node can be persisted before
305 * data when lower device doesn't support write barrier, result
306 * in data corruption after SPO.
307 * So for strict fsync mode, force to use atomic write sematics
308 * to keep write order in between data/node and last node to
309 * avoid potential data corruption.
310 */
311 if (F2FS_OPTION(sbi).fsync_mode ==
312 FSYNC_MODE_STRICT && !atomic)
313 atomic = true;
314 }
315 go_write:
316 /*
317 * Both of fdatasync() and fsync() are able to be recovered from
318 * sudden-power-off.
319 */
320 down_read(&F2FS_I(inode)->i_sem);
321 cp_reason = need_do_checkpoint(inode);
322 up_read(&F2FS_I(inode)->i_sem);
323
324 if (cp_reason) {
325 /* all the dirty node pages should be flushed for POR */
326 ret = f2fs_sync_fs(inode->i_sb, 1);
327
328 /*
329 * We've secured consistency through sync_fs. Following pino
330 * will be used only for fsynced inodes after checkpoint.
331 */
332 try_to_fix_pino(inode);
333 clear_inode_flag(inode, FI_APPEND_WRITE);
334 clear_inode_flag(inode, FI_UPDATE_WRITE);
335 goto out;
336 }
337 sync_nodes:
338 atomic_inc(&sbi->wb_sync_req[NODE]);
339 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
340 atomic_dec(&sbi->wb_sync_req[NODE]);
341 if (ret)
342 goto out;
343
344 /* if cp_error was enabled, we should avoid infinite loop */
345 if (unlikely(f2fs_cp_error(sbi))) {
346 ret = -EIO;
347 goto out;
348 }
349
350 if (f2fs_need_inode_block_update(sbi, ino)) {
351 f2fs_mark_inode_dirty_sync(inode, true);
352 f2fs_write_inode(inode, NULL);
353 goto sync_nodes;
354 }
355
356 /*
357 * If it's atomic_write, it's just fine to keep write ordering. So
358 * here we don't need to wait for node write completion, since we use
359 * node chain which serializes node blocks. If one of node writes are
360 * reordered, we can see simply broken chain, resulting in stopping
361 * roll-forward recovery. It means we'll recover all or none node blocks
362 * given fsync mark.
363 */
364 if (!atomic) {
365 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
366 if (ret)
367 goto out;
368 }
369
370 /* once recovery info is written, don't need to tack this */
371 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
372 clear_inode_flag(inode, FI_APPEND_WRITE);
373 flush_out:
374 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
375 ret = f2fs_issue_flush(sbi, inode->i_ino);
376 if (!ret) {
377 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
378 clear_inode_flag(inode, FI_UPDATE_WRITE);
379 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
380 }
381 f2fs_update_time(sbi, REQ_TIME);
382 out:
383 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
384 return ret;
385 }
386
387 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
388 {
389 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
390 return -EIO;
391 return f2fs_do_sync_file(file, start, end, datasync, false);
392 }
393
394 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
395 pgoff_t index, int whence)
396 {
397 switch (whence) {
398 case SEEK_DATA:
399 if (__is_valid_data_blkaddr(blkaddr))
400 return true;
401 if (blkaddr == NEW_ADDR &&
402 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
403 return true;
404 break;
405 case SEEK_HOLE:
406 if (blkaddr == NULL_ADDR)
407 return true;
408 break;
409 }
410 return false;
411 }
412
413 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
414 {
415 struct inode *inode = file->f_mapping->host;
416 loff_t maxbytes = inode->i_sb->s_maxbytes;
417 struct dnode_of_data dn;
418 pgoff_t pgofs, end_offset;
419 loff_t data_ofs = offset;
420 loff_t isize;
421 int err = 0;
422
423 inode_lock(inode);
424
425 isize = i_size_read(inode);
426 if (offset >= isize)
427 goto fail;
428
429 /* handle inline data case */
430 if (f2fs_has_inline_data(inode)) {
431 if (whence == SEEK_HOLE) {
432 data_ofs = isize;
433 goto found;
434 } else if (whence == SEEK_DATA) {
435 data_ofs = offset;
436 goto found;
437 }
438 }
439
440 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
441
442 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
443 set_new_dnode(&dn, inode, NULL, NULL, 0);
444 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
445 if (err && err != -ENOENT) {
446 goto fail;
447 } else if (err == -ENOENT) {
448 /* direct node does not exists */
449 if (whence == SEEK_DATA) {
450 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
451 continue;
452 } else {
453 goto found;
454 }
455 }
456
457 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
458
459 /* find data/hole in dnode block */
460 for (; dn.ofs_in_node < end_offset;
461 dn.ofs_in_node++, pgofs++,
462 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
463 block_t blkaddr;
464
465 blkaddr = f2fs_data_blkaddr(&dn);
466
467 if (__is_valid_data_blkaddr(blkaddr) &&
468 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
469 blkaddr, DATA_GENERIC_ENHANCE)) {
470 f2fs_put_dnode(&dn);
471 goto fail;
472 }
473
474 if (__found_offset(file->f_mapping, blkaddr,
475 pgofs, whence)) {
476 f2fs_put_dnode(&dn);
477 goto found;
478 }
479 }
480 f2fs_put_dnode(&dn);
481 }
482
483 if (whence == SEEK_DATA)
484 goto fail;
485 found:
486 if (whence == SEEK_HOLE && data_ofs > isize)
487 data_ofs = isize;
488 inode_unlock(inode);
489 return vfs_setpos(file, data_ofs, maxbytes);
490 fail:
491 inode_unlock(inode);
492 return -ENXIO;
493 }
494
495 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
496 {
497 struct inode *inode = file->f_mapping->host;
498 loff_t maxbytes = inode->i_sb->s_maxbytes;
499
500 if (f2fs_compressed_file(inode))
501 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
502
503 switch (whence) {
504 case SEEK_SET:
505 case SEEK_CUR:
506 case SEEK_END:
507 return generic_file_llseek_size(file, offset, whence,
508 maxbytes, i_size_read(inode));
509 case SEEK_DATA:
510 case SEEK_HOLE:
511 if (offset < 0)
512 return -ENXIO;
513 return f2fs_seek_block(file, offset, whence);
514 }
515
516 return -EINVAL;
517 }
518
519 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
520 {
521 struct inode *inode = file_inode(file);
522
523 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
524 return -EIO;
525
526 if (!f2fs_is_compress_backend_ready(inode))
527 return -EOPNOTSUPP;
528
529 file_accessed(file);
530 vma->vm_ops = &f2fs_file_vm_ops;
531 set_inode_flag(inode, FI_MMAP_FILE);
532 return 0;
533 }
534
535 static int f2fs_file_open(struct inode *inode, struct file *filp)
536 {
537 int err = fscrypt_file_open(inode, filp);
538
539 if (err)
540 return err;
541
542 if (!f2fs_is_compress_backend_ready(inode))
543 return -EOPNOTSUPP;
544
545 err = fsverity_file_open(inode, filp);
546 if (err)
547 return err;
548
549 filp->f_mode |= FMODE_NOWAIT;
550
551 return dquot_file_open(inode, filp);
552 }
553
554 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
555 {
556 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
557 struct f2fs_node *raw_node;
558 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
559 __le32 *addr;
560 int base = 0;
561 bool compressed_cluster = false;
562 int cluster_index = 0, valid_blocks = 0;
563 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
564 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
565
566 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
567 base = get_extra_isize(dn->inode);
568
569 raw_node = F2FS_NODE(dn->node_page);
570 addr = blkaddr_in_node(raw_node) + base + ofs;
571
572 /* Assumption: truncateion starts with cluster */
573 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
574 block_t blkaddr = le32_to_cpu(*addr);
575
576 if (f2fs_compressed_file(dn->inode) &&
577 !(cluster_index & (cluster_size - 1))) {
578 if (compressed_cluster)
579 f2fs_i_compr_blocks_update(dn->inode,
580 valid_blocks, false);
581 compressed_cluster = (blkaddr == COMPRESS_ADDR);
582 valid_blocks = 0;
583 }
584
585 if (blkaddr == NULL_ADDR)
586 continue;
587
588 dn->data_blkaddr = NULL_ADDR;
589 f2fs_set_data_blkaddr(dn);
590
591 if (__is_valid_data_blkaddr(blkaddr)) {
592 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
593 DATA_GENERIC_ENHANCE))
594 continue;
595 if (compressed_cluster)
596 valid_blocks++;
597 }
598
599 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
600 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
601
602 f2fs_invalidate_blocks(sbi, blkaddr);
603
604 if (!released || blkaddr != COMPRESS_ADDR)
605 nr_free++;
606 }
607
608 if (compressed_cluster)
609 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
610
611 if (nr_free) {
612 pgoff_t fofs;
613 /*
614 * once we invalidate valid blkaddr in range [ofs, ofs + count],
615 * we will invalidate all blkaddr in the whole range.
616 */
617 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
618 dn->inode) + ofs;
619 f2fs_update_extent_cache_range(dn, fofs, 0, len);
620 dec_valid_block_count(sbi, dn->inode, nr_free);
621 }
622 dn->ofs_in_node = ofs;
623
624 f2fs_update_time(sbi, REQ_TIME);
625 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
626 dn->ofs_in_node, nr_free);
627 }
628
629 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
630 {
631 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
632 }
633
634 static int truncate_partial_data_page(struct inode *inode, u64 from,
635 bool cache_only)
636 {
637 loff_t offset = from & (PAGE_SIZE - 1);
638 pgoff_t index = from >> PAGE_SHIFT;
639 struct address_space *mapping = inode->i_mapping;
640 struct page *page;
641
642 if (!offset && !cache_only)
643 return 0;
644
645 if (cache_only) {
646 page = find_lock_page(mapping, index);
647 if (page && PageUptodate(page))
648 goto truncate_out;
649 f2fs_put_page(page, 1);
650 return 0;
651 }
652
653 page = f2fs_get_lock_data_page(inode, index, true);
654 if (IS_ERR(page))
655 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
656 truncate_out:
657 f2fs_wait_on_page_writeback(page, DATA, true, true);
658 zero_user(page, offset, PAGE_SIZE - offset);
659
660 /* An encrypted inode should have a key and truncate the last page. */
661 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
662 if (!cache_only)
663 set_page_dirty(page);
664 f2fs_put_page(page, 1);
665 return 0;
666 }
667
668 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
669 {
670 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
671 struct dnode_of_data dn;
672 pgoff_t free_from;
673 int count = 0, err = 0;
674 struct page *ipage;
675 bool truncate_page = false;
676
677 trace_f2fs_truncate_blocks_enter(inode, from);
678
679 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
680
681 if (free_from >= max_file_blocks(inode))
682 goto free_partial;
683
684 if (lock)
685 f2fs_lock_op(sbi);
686
687 ipage = f2fs_get_node_page(sbi, inode->i_ino);
688 if (IS_ERR(ipage)) {
689 err = PTR_ERR(ipage);
690 goto out;
691 }
692
693 if (f2fs_has_inline_data(inode)) {
694 f2fs_truncate_inline_inode(inode, ipage, from);
695 f2fs_put_page(ipage, 1);
696 truncate_page = true;
697 goto out;
698 }
699
700 set_new_dnode(&dn, inode, ipage, NULL, 0);
701 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
702 if (err) {
703 if (err == -ENOENT)
704 goto free_next;
705 goto out;
706 }
707
708 count = ADDRS_PER_PAGE(dn.node_page, inode);
709
710 count -= dn.ofs_in_node;
711 f2fs_bug_on(sbi, count < 0);
712
713 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
714 f2fs_truncate_data_blocks_range(&dn, count);
715 free_from += count;
716 }
717
718 f2fs_put_dnode(&dn);
719 free_next:
720 err = f2fs_truncate_inode_blocks(inode, free_from);
721 out:
722 if (lock)
723 f2fs_unlock_op(sbi);
724 free_partial:
725 /* lastly zero out the first data page */
726 if (!err)
727 err = truncate_partial_data_page(inode, from, truncate_page);
728
729 trace_f2fs_truncate_blocks_exit(inode, err);
730 return err;
731 }
732
733 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
734 {
735 u64 free_from = from;
736 int err;
737
738 #ifdef CONFIG_F2FS_FS_COMPRESSION
739 /*
740 * for compressed file, only support cluster size
741 * aligned truncation.
742 */
743 if (f2fs_compressed_file(inode))
744 free_from = round_up(from,
745 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
746 #endif
747
748 err = f2fs_do_truncate_blocks(inode, free_from, lock);
749 if (err)
750 return err;
751
752 #ifdef CONFIG_F2FS_FS_COMPRESSION
753 /*
754 * For compressed file, after release compress blocks, don't allow write
755 * direct, but we should allow write direct after truncate to zero.
756 */
757 if (f2fs_compressed_file(inode) && !free_from
758 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
759 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
760
761 if (from != free_from) {
762 err = f2fs_truncate_partial_cluster(inode, from, lock);
763 if (err)
764 return err;
765 }
766 #endif
767
768 return 0;
769 }
770
771 int f2fs_truncate(struct inode *inode)
772 {
773 int err;
774
775 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
776 return -EIO;
777
778 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
779 S_ISLNK(inode->i_mode)))
780 return 0;
781
782 trace_f2fs_truncate(inode);
783
784 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
785 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
786 return -EIO;
787 }
788
789 err = dquot_initialize(inode);
790 if (err)
791 return err;
792
793 /* we should check inline_data size */
794 if (!f2fs_may_inline_data(inode)) {
795 err = f2fs_convert_inline_inode(inode);
796 if (err)
797 return err;
798 }
799
800 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
801 if (err)
802 return err;
803
804 inode->i_mtime = inode->i_ctime = current_time(inode);
805 f2fs_mark_inode_dirty_sync(inode, false);
806 return 0;
807 }
808
809 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
810 struct kstat *stat, u32 request_mask, unsigned int query_flags)
811 {
812 struct inode *inode = d_inode(path->dentry);
813 struct f2fs_inode_info *fi = F2FS_I(inode);
814 struct f2fs_inode *ri;
815 unsigned int flags;
816
817 if (f2fs_has_extra_attr(inode) &&
818 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
819 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
820 stat->result_mask |= STATX_BTIME;
821 stat->btime.tv_sec = fi->i_crtime.tv_sec;
822 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
823 }
824
825 flags = fi->i_flags;
826 if (flags & F2FS_COMPR_FL)
827 stat->attributes |= STATX_ATTR_COMPRESSED;
828 if (flags & F2FS_APPEND_FL)
829 stat->attributes |= STATX_ATTR_APPEND;
830 if (IS_ENCRYPTED(inode))
831 stat->attributes |= STATX_ATTR_ENCRYPTED;
832 if (flags & F2FS_IMMUTABLE_FL)
833 stat->attributes |= STATX_ATTR_IMMUTABLE;
834 if (flags & F2FS_NODUMP_FL)
835 stat->attributes |= STATX_ATTR_NODUMP;
836 if (IS_VERITY(inode))
837 stat->attributes |= STATX_ATTR_VERITY;
838
839 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
840 STATX_ATTR_APPEND |
841 STATX_ATTR_ENCRYPTED |
842 STATX_ATTR_IMMUTABLE |
843 STATX_ATTR_NODUMP |
844 STATX_ATTR_VERITY);
845
846 generic_fillattr(&init_user_ns, inode, stat);
847
848 /* we need to show initial sectors used for inline_data/dentries */
849 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
850 f2fs_has_inline_dentry(inode))
851 stat->blocks += (stat->size + 511) >> 9;
852
853 return 0;
854 }
855
856 #ifdef CONFIG_F2FS_FS_POSIX_ACL
857 static void __setattr_copy(struct user_namespace *mnt_userns,
858 struct inode *inode, const struct iattr *attr)
859 {
860 unsigned int ia_valid = attr->ia_valid;
861
862 if (ia_valid & ATTR_UID)
863 inode->i_uid = attr->ia_uid;
864 if (ia_valid & ATTR_GID)
865 inode->i_gid = attr->ia_gid;
866 if (ia_valid & ATTR_ATIME)
867 inode->i_atime = attr->ia_atime;
868 if (ia_valid & ATTR_MTIME)
869 inode->i_mtime = attr->ia_mtime;
870 if (ia_valid & ATTR_CTIME)
871 inode->i_ctime = attr->ia_ctime;
872 if (ia_valid & ATTR_MODE) {
873 umode_t mode = attr->ia_mode;
874 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
875
876 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
877 mode &= ~S_ISGID;
878 set_acl_inode(inode, mode);
879 }
880 }
881 #else
882 #define __setattr_copy setattr_copy
883 #endif
884
885 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
886 struct iattr *attr)
887 {
888 struct inode *inode = d_inode(dentry);
889 int err;
890
891 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
892 return -EIO;
893
894 if (unlikely(IS_IMMUTABLE(inode)))
895 return -EPERM;
896
897 if (unlikely(IS_APPEND(inode) &&
898 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
899 ATTR_GID | ATTR_TIMES_SET))))
900 return -EPERM;
901
902 if ((attr->ia_valid & ATTR_SIZE) &&
903 !f2fs_is_compress_backend_ready(inode))
904 return -EOPNOTSUPP;
905
906 err = setattr_prepare(&init_user_ns, dentry, attr);
907 if (err)
908 return err;
909
910 err = fscrypt_prepare_setattr(dentry, attr);
911 if (err)
912 return err;
913
914 err = fsverity_prepare_setattr(dentry, attr);
915 if (err)
916 return err;
917
918 if (is_quota_modification(inode, attr)) {
919 err = dquot_initialize(inode);
920 if (err)
921 return err;
922 }
923 if ((attr->ia_valid & ATTR_UID &&
924 !uid_eq(attr->ia_uid, inode->i_uid)) ||
925 (attr->ia_valid & ATTR_GID &&
926 !gid_eq(attr->ia_gid, inode->i_gid))) {
927 f2fs_lock_op(F2FS_I_SB(inode));
928 err = dquot_transfer(inode, attr);
929 if (err) {
930 set_sbi_flag(F2FS_I_SB(inode),
931 SBI_QUOTA_NEED_REPAIR);
932 f2fs_unlock_op(F2FS_I_SB(inode));
933 return err;
934 }
935 /*
936 * update uid/gid under lock_op(), so that dquot and inode can
937 * be updated atomically.
938 */
939 if (attr->ia_valid & ATTR_UID)
940 inode->i_uid = attr->ia_uid;
941 if (attr->ia_valid & ATTR_GID)
942 inode->i_gid = attr->ia_gid;
943 f2fs_mark_inode_dirty_sync(inode, true);
944 f2fs_unlock_op(F2FS_I_SB(inode));
945 }
946
947 if (attr->ia_valid & ATTR_SIZE) {
948 loff_t old_size = i_size_read(inode);
949
950 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
951 /*
952 * should convert inline inode before i_size_write to
953 * keep smaller than inline_data size with inline flag.
954 */
955 err = f2fs_convert_inline_inode(inode);
956 if (err)
957 return err;
958 }
959
960 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
961 filemap_invalidate_lock(inode->i_mapping);
962
963 truncate_setsize(inode, attr->ia_size);
964
965 if (attr->ia_size <= old_size)
966 err = f2fs_truncate(inode);
967 /*
968 * do not trim all blocks after i_size if target size is
969 * larger than i_size.
970 */
971 filemap_invalidate_unlock(inode->i_mapping);
972 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
973 if (err)
974 return err;
975
976 spin_lock(&F2FS_I(inode)->i_size_lock);
977 inode->i_mtime = inode->i_ctime = current_time(inode);
978 F2FS_I(inode)->last_disk_size = i_size_read(inode);
979 spin_unlock(&F2FS_I(inode)->i_size_lock);
980 }
981
982 __setattr_copy(&init_user_ns, inode, attr);
983
984 if (attr->ia_valid & ATTR_MODE) {
985 err = posix_acl_chmod(&init_user_ns, inode, f2fs_get_inode_mode(inode));
986
987 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
988 if (!err)
989 inode->i_mode = F2FS_I(inode)->i_acl_mode;
990 clear_inode_flag(inode, FI_ACL_MODE);
991 }
992 }
993
994 /* file size may changed here */
995 f2fs_mark_inode_dirty_sync(inode, true);
996
997 /* inode change will produce dirty node pages flushed by checkpoint */
998 f2fs_balance_fs(F2FS_I_SB(inode), true);
999
1000 return err;
1001 }
1002
1003 const struct inode_operations f2fs_file_inode_operations = {
1004 .getattr = f2fs_getattr,
1005 .setattr = f2fs_setattr,
1006 .get_acl = f2fs_get_acl,
1007 .set_acl = f2fs_set_acl,
1008 .listxattr = f2fs_listxattr,
1009 .fiemap = f2fs_fiemap,
1010 .fileattr_get = f2fs_fileattr_get,
1011 .fileattr_set = f2fs_fileattr_set,
1012 };
1013
1014 static int fill_zero(struct inode *inode, pgoff_t index,
1015 loff_t start, loff_t len)
1016 {
1017 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1018 struct page *page;
1019
1020 if (!len)
1021 return 0;
1022
1023 f2fs_balance_fs(sbi, true);
1024
1025 f2fs_lock_op(sbi);
1026 page = f2fs_get_new_data_page(inode, NULL, index, false);
1027 f2fs_unlock_op(sbi);
1028
1029 if (IS_ERR(page))
1030 return PTR_ERR(page);
1031
1032 f2fs_wait_on_page_writeback(page, DATA, true, true);
1033 zero_user(page, start, len);
1034 set_page_dirty(page);
1035 f2fs_put_page(page, 1);
1036 return 0;
1037 }
1038
1039 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1040 {
1041 int err;
1042
1043 while (pg_start < pg_end) {
1044 struct dnode_of_data dn;
1045 pgoff_t end_offset, count;
1046
1047 set_new_dnode(&dn, inode, NULL, NULL, 0);
1048 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1049 if (err) {
1050 if (err == -ENOENT) {
1051 pg_start = f2fs_get_next_page_offset(&dn,
1052 pg_start);
1053 continue;
1054 }
1055 return err;
1056 }
1057
1058 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1059 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1060
1061 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1062
1063 f2fs_truncate_data_blocks_range(&dn, count);
1064 f2fs_put_dnode(&dn);
1065
1066 pg_start += count;
1067 }
1068 return 0;
1069 }
1070
1071 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1072 {
1073 pgoff_t pg_start, pg_end;
1074 loff_t off_start, off_end;
1075 int ret;
1076
1077 ret = f2fs_convert_inline_inode(inode);
1078 if (ret)
1079 return ret;
1080
1081 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1082 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1083
1084 off_start = offset & (PAGE_SIZE - 1);
1085 off_end = (offset + len) & (PAGE_SIZE - 1);
1086
1087 if (pg_start == pg_end) {
1088 ret = fill_zero(inode, pg_start, off_start,
1089 off_end - off_start);
1090 if (ret)
1091 return ret;
1092 } else {
1093 if (off_start) {
1094 ret = fill_zero(inode, pg_start++, off_start,
1095 PAGE_SIZE - off_start);
1096 if (ret)
1097 return ret;
1098 }
1099 if (off_end) {
1100 ret = fill_zero(inode, pg_end, 0, off_end);
1101 if (ret)
1102 return ret;
1103 }
1104
1105 if (pg_start < pg_end) {
1106 loff_t blk_start, blk_end;
1107 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1108
1109 f2fs_balance_fs(sbi, true);
1110
1111 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1112 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1113
1114 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1115 filemap_invalidate_lock(inode->i_mapping);
1116
1117 truncate_pagecache_range(inode, blk_start, blk_end - 1);
1118
1119 f2fs_lock_op(sbi);
1120 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1121 f2fs_unlock_op(sbi);
1122
1123 filemap_invalidate_unlock(inode->i_mapping);
1124 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1125 }
1126 }
1127
1128 return ret;
1129 }
1130
1131 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1132 int *do_replace, pgoff_t off, pgoff_t len)
1133 {
1134 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1135 struct dnode_of_data dn;
1136 int ret, done, i;
1137
1138 next_dnode:
1139 set_new_dnode(&dn, inode, NULL, NULL, 0);
1140 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1141 if (ret && ret != -ENOENT) {
1142 return ret;
1143 } else if (ret == -ENOENT) {
1144 if (dn.max_level == 0)
1145 return -ENOENT;
1146 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1147 dn.ofs_in_node, len);
1148 blkaddr += done;
1149 do_replace += done;
1150 goto next;
1151 }
1152
1153 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1154 dn.ofs_in_node, len);
1155 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1156 *blkaddr = f2fs_data_blkaddr(&dn);
1157
1158 if (__is_valid_data_blkaddr(*blkaddr) &&
1159 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1160 DATA_GENERIC_ENHANCE)) {
1161 f2fs_put_dnode(&dn);
1162 return -EFSCORRUPTED;
1163 }
1164
1165 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1166
1167 if (f2fs_lfs_mode(sbi)) {
1168 f2fs_put_dnode(&dn);
1169 return -EOPNOTSUPP;
1170 }
1171
1172 /* do not invalidate this block address */
1173 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1174 *do_replace = 1;
1175 }
1176 }
1177 f2fs_put_dnode(&dn);
1178 next:
1179 len -= done;
1180 off += done;
1181 if (len)
1182 goto next_dnode;
1183 return 0;
1184 }
1185
1186 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1187 int *do_replace, pgoff_t off, int len)
1188 {
1189 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1190 struct dnode_of_data dn;
1191 int ret, i;
1192
1193 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1194 if (*do_replace == 0)
1195 continue;
1196
1197 set_new_dnode(&dn, inode, NULL, NULL, 0);
1198 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1199 if (ret) {
1200 dec_valid_block_count(sbi, inode, 1);
1201 f2fs_invalidate_blocks(sbi, *blkaddr);
1202 } else {
1203 f2fs_update_data_blkaddr(&dn, *blkaddr);
1204 }
1205 f2fs_put_dnode(&dn);
1206 }
1207 return 0;
1208 }
1209
1210 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1211 block_t *blkaddr, int *do_replace,
1212 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1213 {
1214 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1215 pgoff_t i = 0;
1216 int ret;
1217
1218 while (i < len) {
1219 if (blkaddr[i] == NULL_ADDR && !full) {
1220 i++;
1221 continue;
1222 }
1223
1224 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1225 struct dnode_of_data dn;
1226 struct node_info ni;
1227 size_t new_size;
1228 pgoff_t ilen;
1229
1230 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1231 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1232 if (ret)
1233 return ret;
1234
1235 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1236 if (ret) {
1237 f2fs_put_dnode(&dn);
1238 return ret;
1239 }
1240
1241 ilen = min((pgoff_t)
1242 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1243 dn.ofs_in_node, len - i);
1244 do {
1245 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1246 f2fs_truncate_data_blocks_range(&dn, 1);
1247
1248 if (do_replace[i]) {
1249 f2fs_i_blocks_write(src_inode,
1250 1, false, false);
1251 f2fs_i_blocks_write(dst_inode,
1252 1, true, false);
1253 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1254 blkaddr[i], ni.version, true, false);
1255
1256 do_replace[i] = 0;
1257 }
1258 dn.ofs_in_node++;
1259 i++;
1260 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1261 if (dst_inode->i_size < new_size)
1262 f2fs_i_size_write(dst_inode, new_size);
1263 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1264
1265 f2fs_put_dnode(&dn);
1266 } else {
1267 struct page *psrc, *pdst;
1268
1269 psrc = f2fs_get_lock_data_page(src_inode,
1270 src + i, true);
1271 if (IS_ERR(psrc))
1272 return PTR_ERR(psrc);
1273 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1274 true);
1275 if (IS_ERR(pdst)) {
1276 f2fs_put_page(psrc, 1);
1277 return PTR_ERR(pdst);
1278 }
1279 f2fs_copy_page(psrc, pdst);
1280 set_page_dirty(pdst);
1281 f2fs_put_page(pdst, 1);
1282 f2fs_put_page(psrc, 1);
1283
1284 ret = f2fs_truncate_hole(src_inode,
1285 src + i, src + i + 1);
1286 if (ret)
1287 return ret;
1288 i++;
1289 }
1290 }
1291 return 0;
1292 }
1293
1294 static int __exchange_data_block(struct inode *src_inode,
1295 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1296 pgoff_t len, bool full)
1297 {
1298 block_t *src_blkaddr;
1299 int *do_replace;
1300 pgoff_t olen;
1301 int ret;
1302
1303 while (len) {
1304 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1305
1306 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1307 array_size(olen, sizeof(block_t)),
1308 GFP_NOFS);
1309 if (!src_blkaddr)
1310 return -ENOMEM;
1311
1312 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1313 array_size(olen, sizeof(int)),
1314 GFP_NOFS);
1315 if (!do_replace) {
1316 kvfree(src_blkaddr);
1317 return -ENOMEM;
1318 }
1319
1320 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1321 do_replace, src, olen);
1322 if (ret)
1323 goto roll_back;
1324
1325 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1326 do_replace, src, dst, olen, full);
1327 if (ret)
1328 goto roll_back;
1329
1330 src += olen;
1331 dst += olen;
1332 len -= olen;
1333
1334 kvfree(src_blkaddr);
1335 kvfree(do_replace);
1336 }
1337 return 0;
1338
1339 roll_back:
1340 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1341 kvfree(src_blkaddr);
1342 kvfree(do_replace);
1343 return ret;
1344 }
1345
1346 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1347 {
1348 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1349 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1350 pgoff_t start = offset >> PAGE_SHIFT;
1351 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1352 int ret;
1353
1354 f2fs_balance_fs(sbi, true);
1355
1356 /* avoid gc operation during block exchange */
1357 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1358 filemap_invalidate_lock(inode->i_mapping);
1359
1360 f2fs_lock_op(sbi);
1361 f2fs_drop_extent_tree(inode);
1362 truncate_pagecache(inode, offset);
1363 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1364 f2fs_unlock_op(sbi);
1365
1366 filemap_invalidate_unlock(inode->i_mapping);
1367 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1368 return ret;
1369 }
1370
1371 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1372 {
1373 loff_t new_size;
1374 int ret;
1375
1376 if (offset + len >= i_size_read(inode))
1377 return -EINVAL;
1378
1379 /* collapse range should be aligned to block size of f2fs. */
1380 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1381 return -EINVAL;
1382
1383 ret = f2fs_convert_inline_inode(inode);
1384 if (ret)
1385 return ret;
1386
1387 /* write out all dirty pages from offset */
1388 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1389 if (ret)
1390 return ret;
1391
1392 ret = f2fs_do_collapse(inode, offset, len);
1393 if (ret)
1394 return ret;
1395
1396 /* write out all moved pages, if possible */
1397 filemap_invalidate_lock(inode->i_mapping);
1398 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1399 truncate_pagecache(inode, offset);
1400
1401 new_size = i_size_read(inode) - len;
1402 ret = f2fs_truncate_blocks(inode, new_size, true);
1403 filemap_invalidate_unlock(inode->i_mapping);
1404 if (!ret)
1405 f2fs_i_size_write(inode, new_size);
1406 return ret;
1407 }
1408
1409 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1410 pgoff_t end)
1411 {
1412 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1413 pgoff_t index = start;
1414 unsigned int ofs_in_node = dn->ofs_in_node;
1415 blkcnt_t count = 0;
1416 int ret;
1417
1418 for (; index < end; index++, dn->ofs_in_node++) {
1419 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1420 count++;
1421 }
1422
1423 dn->ofs_in_node = ofs_in_node;
1424 ret = f2fs_reserve_new_blocks(dn, count);
1425 if (ret)
1426 return ret;
1427
1428 dn->ofs_in_node = ofs_in_node;
1429 for (index = start; index < end; index++, dn->ofs_in_node++) {
1430 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1431 /*
1432 * f2fs_reserve_new_blocks will not guarantee entire block
1433 * allocation.
1434 */
1435 if (dn->data_blkaddr == NULL_ADDR) {
1436 ret = -ENOSPC;
1437 break;
1438 }
1439 if (dn->data_blkaddr != NEW_ADDR) {
1440 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1441 dn->data_blkaddr = NEW_ADDR;
1442 f2fs_set_data_blkaddr(dn);
1443 }
1444 }
1445
1446 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1447
1448 return ret;
1449 }
1450
1451 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1452 int mode)
1453 {
1454 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1455 struct address_space *mapping = inode->i_mapping;
1456 pgoff_t index, pg_start, pg_end;
1457 loff_t new_size = i_size_read(inode);
1458 loff_t off_start, off_end;
1459 int ret = 0;
1460
1461 ret = inode_newsize_ok(inode, (len + offset));
1462 if (ret)
1463 return ret;
1464
1465 ret = f2fs_convert_inline_inode(inode);
1466 if (ret)
1467 return ret;
1468
1469 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1470 if (ret)
1471 return ret;
1472
1473 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1474 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1475
1476 off_start = offset & (PAGE_SIZE - 1);
1477 off_end = (offset + len) & (PAGE_SIZE - 1);
1478
1479 if (pg_start == pg_end) {
1480 ret = fill_zero(inode, pg_start, off_start,
1481 off_end - off_start);
1482 if (ret)
1483 return ret;
1484
1485 new_size = max_t(loff_t, new_size, offset + len);
1486 } else {
1487 if (off_start) {
1488 ret = fill_zero(inode, pg_start++, off_start,
1489 PAGE_SIZE - off_start);
1490 if (ret)
1491 return ret;
1492
1493 new_size = max_t(loff_t, new_size,
1494 (loff_t)pg_start << PAGE_SHIFT);
1495 }
1496
1497 for (index = pg_start; index < pg_end;) {
1498 struct dnode_of_data dn;
1499 unsigned int end_offset;
1500 pgoff_t end;
1501
1502 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1503 filemap_invalidate_lock(mapping);
1504
1505 truncate_pagecache_range(inode,
1506 (loff_t)index << PAGE_SHIFT,
1507 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1508
1509 f2fs_lock_op(sbi);
1510
1511 set_new_dnode(&dn, inode, NULL, NULL, 0);
1512 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1513 if (ret) {
1514 f2fs_unlock_op(sbi);
1515 filemap_invalidate_unlock(mapping);
1516 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1517 goto out;
1518 }
1519
1520 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1521 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1522
1523 ret = f2fs_do_zero_range(&dn, index, end);
1524 f2fs_put_dnode(&dn);
1525
1526 f2fs_unlock_op(sbi);
1527 filemap_invalidate_unlock(mapping);
1528 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1529
1530 f2fs_balance_fs(sbi, dn.node_changed);
1531
1532 if (ret)
1533 goto out;
1534
1535 index = end;
1536 new_size = max_t(loff_t, new_size,
1537 (loff_t)index << PAGE_SHIFT);
1538 }
1539
1540 if (off_end) {
1541 ret = fill_zero(inode, pg_end, 0, off_end);
1542 if (ret)
1543 goto out;
1544
1545 new_size = max_t(loff_t, new_size, offset + len);
1546 }
1547 }
1548
1549 out:
1550 if (new_size > i_size_read(inode)) {
1551 if (mode & FALLOC_FL_KEEP_SIZE)
1552 file_set_keep_isize(inode);
1553 else
1554 f2fs_i_size_write(inode, new_size);
1555 }
1556 return ret;
1557 }
1558
1559 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1560 {
1561 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1562 struct address_space *mapping = inode->i_mapping;
1563 pgoff_t nr, pg_start, pg_end, delta, idx;
1564 loff_t new_size;
1565 int ret = 0;
1566
1567 new_size = i_size_read(inode) + len;
1568 ret = inode_newsize_ok(inode, new_size);
1569 if (ret)
1570 return ret;
1571
1572 if (offset >= i_size_read(inode))
1573 return -EINVAL;
1574
1575 /* insert range should be aligned to block size of f2fs. */
1576 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1577 return -EINVAL;
1578
1579 ret = f2fs_convert_inline_inode(inode);
1580 if (ret)
1581 return ret;
1582
1583 f2fs_balance_fs(sbi, true);
1584
1585 filemap_invalidate_lock(mapping);
1586 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1587 filemap_invalidate_unlock(mapping);
1588 if (ret)
1589 return ret;
1590
1591 /* write out all dirty pages from offset */
1592 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1593 if (ret)
1594 return ret;
1595
1596 pg_start = offset >> PAGE_SHIFT;
1597 pg_end = (offset + len) >> PAGE_SHIFT;
1598 delta = pg_end - pg_start;
1599 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1600
1601 /* avoid gc operation during block exchange */
1602 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1603 filemap_invalidate_lock(mapping);
1604 truncate_pagecache(inode, offset);
1605
1606 while (!ret && idx > pg_start) {
1607 nr = idx - pg_start;
1608 if (nr > delta)
1609 nr = delta;
1610 idx -= nr;
1611
1612 f2fs_lock_op(sbi);
1613 f2fs_drop_extent_tree(inode);
1614
1615 ret = __exchange_data_block(inode, inode, idx,
1616 idx + delta, nr, false);
1617 f2fs_unlock_op(sbi);
1618 }
1619 filemap_invalidate_unlock(mapping);
1620 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1621
1622 /* write out all moved pages, if possible */
1623 filemap_invalidate_lock(mapping);
1624 filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1625 truncate_pagecache(inode, offset);
1626 filemap_invalidate_unlock(mapping);
1627
1628 if (!ret)
1629 f2fs_i_size_write(inode, new_size);
1630 return ret;
1631 }
1632
1633 static int expand_inode_data(struct inode *inode, loff_t offset,
1634 loff_t len, int mode)
1635 {
1636 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1637 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1638 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1639 .m_may_create = true };
1640 pgoff_t pg_start, pg_end;
1641 loff_t new_size = i_size_read(inode);
1642 loff_t off_end;
1643 block_t expanded = 0;
1644 int err;
1645
1646 err = inode_newsize_ok(inode, (len + offset));
1647 if (err)
1648 return err;
1649
1650 err = f2fs_convert_inline_inode(inode);
1651 if (err)
1652 return err;
1653
1654 f2fs_balance_fs(sbi, true);
1655
1656 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1657 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1658 off_end = (offset + len) & (PAGE_SIZE - 1);
1659
1660 map.m_lblk = pg_start;
1661 map.m_len = pg_end - pg_start;
1662 if (off_end)
1663 map.m_len++;
1664
1665 if (!map.m_len)
1666 return 0;
1667
1668 if (f2fs_is_pinned_file(inode)) {
1669 block_t sec_blks = BLKS_PER_SEC(sbi);
1670 block_t sec_len = roundup(map.m_len, sec_blks);
1671
1672 map.m_len = sec_blks;
1673 next_alloc:
1674 if (has_not_enough_free_secs(sbi, 0,
1675 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1676 down_write(&sbi->gc_lock);
1677 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1678 if (err && err != -ENODATA && err != -EAGAIN)
1679 goto out_err;
1680 }
1681
1682 down_write(&sbi->pin_sem);
1683
1684 f2fs_lock_op(sbi);
1685 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1686 f2fs_unlock_op(sbi);
1687
1688 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1689 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1690
1691 up_write(&sbi->pin_sem);
1692
1693 expanded += map.m_len;
1694 sec_len -= map.m_len;
1695 map.m_lblk += map.m_len;
1696 if (!err && sec_len)
1697 goto next_alloc;
1698
1699 map.m_len = expanded;
1700 } else {
1701 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1702 expanded = map.m_len;
1703 }
1704 out_err:
1705 if (err) {
1706 pgoff_t last_off;
1707
1708 if (!expanded)
1709 return err;
1710
1711 last_off = pg_start + expanded - 1;
1712
1713 /* update new size to the failed position */
1714 new_size = (last_off == pg_end) ? offset + len :
1715 (loff_t)(last_off + 1) << PAGE_SHIFT;
1716 } else {
1717 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1718 }
1719
1720 if (new_size > i_size_read(inode)) {
1721 if (mode & FALLOC_FL_KEEP_SIZE)
1722 file_set_keep_isize(inode);
1723 else
1724 f2fs_i_size_write(inode, new_size);
1725 }
1726
1727 return err;
1728 }
1729
1730 static long f2fs_fallocate(struct file *file, int mode,
1731 loff_t offset, loff_t len)
1732 {
1733 struct inode *inode = file_inode(file);
1734 long ret = 0;
1735
1736 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1737 return -EIO;
1738 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1739 return -ENOSPC;
1740 if (!f2fs_is_compress_backend_ready(inode))
1741 return -EOPNOTSUPP;
1742
1743 /* f2fs only support ->fallocate for regular file */
1744 if (!S_ISREG(inode->i_mode))
1745 return -EINVAL;
1746
1747 if (IS_ENCRYPTED(inode) &&
1748 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1749 return -EOPNOTSUPP;
1750
1751 if (f2fs_compressed_file(inode) &&
1752 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1753 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1754 return -EOPNOTSUPP;
1755
1756 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1757 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1758 FALLOC_FL_INSERT_RANGE))
1759 return -EOPNOTSUPP;
1760
1761 inode_lock(inode);
1762
1763 if (mode & FALLOC_FL_PUNCH_HOLE) {
1764 if (offset >= inode->i_size)
1765 goto out;
1766
1767 ret = punch_hole(inode, offset, len);
1768 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1769 ret = f2fs_collapse_range(inode, offset, len);
1770 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1771 ret = f2fs_zero_range(inode, offset, len, mode);
1772 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1773 ret = f2fs_insert_range(inode, offset, len);
1774 } else {
1775 ret = expand_inode_data(inode, offset, len, mode);
1776 }
1777
1778 if (!ret) {
1779 inode->i_mtime = inode->i_ctime = current_time(inode);
1780 f2fs_mark_inode_dirty_sync(inode, false);
1781 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1782 }
1783
1784 out:
1785 inode_unlock(inode);
1786
1787 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1788 return ret;
1789 }
1790
1791 static int f2fs_release_file(struct inode *inode, struct file *filp)
1792 {
1793 /*
1794 * f2fs_relase_file is called at every close calls. So we should
1795 * not drop any inmemory pages by close called by other process.
1796 */
1797 if (!(filp->f_mode & FMODE_WRITE) ||
1798 atomic_read(&inode->i_writecount) != 1)
1799 return 0;
1800
1801 /* some remained atomic pages should discarded */
1802 if (f2fs_is_atomic_file(inode))
1803 f2fs_drop_inmem_pages(inode);
1804 if (f2fs_is_volatile_file(inode)) {
1805 set_inode_flag(inode, FI_DROP_CACHE);
1806 filemap_fdatawrite(inode->i_mapping);
1807 clear_inode_flag(inode, FI_DROP_CACHE);
1808 clear_inode_flag(inode, FI_VOLATILE_FILE);
1809 stat_dec_volatile_write(inode);
1810 }
1811 return 0;
1812 }
1813
1814 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1815 {
1816 struct inode *inode = file_inode(file);
1817
1818 /*
1819 * If the process doing a transaction is crashed, we should do
1820 * roll-back. Otherwise, other reader/write can see corrupted database
1821 * until all the writers close its file. Since this should be done
1822 * before dropping file lock, it needs to do in ->flush.
1823 */
1824 if (f2fs_is_atomic_file(inode) &&
1825 F2FS_I(inode)->inmem_task == current)
1826 f2fs_drop_inmem_pages(inode);
1827 return 0;
1828 }
1829
1830 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1831 {
1832 struct f2fs_inode_info *fi = F2FS_I(inode);
1833 u32 masked_flags = fi->i_flags & mask;
1834
1835 /* mask can be shrunk by flags_valid selector */
1836 iflags &= mask;
1837
1838 /* Is it quota file? Do not allow user to mess with it */
1839 if (IS_NOQUOTA(inode))
1840 return -EPERM;
1841
1842 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1843 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1844 return -EOPNOTSUPP;
1845 if (!f2fs_empty_dir(inode))
1846 return -ENOTEMPTY;
1847 }
1848
1849 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1850 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1851 return -EOPNOTSUPP;
1852 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1853 return -EINVAL;
1854 }
1855
1856 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1857 if (masked_flags & F2FS_COMPR_FL) {
1858 if (!f2fs_disable_compressed_file(inode))
1859 return -EINVAL;
1860 }
1861 if (iflags & F2FS_NOCOMP_FL)
1862 return -EINVAL;
1863 if (iflags & F2FS_COMPR_FL) {
1864 if (!f2fs_may_compress(inode))
1865 return -EINVAL;
1866 if (S_ISREG(inode->i_mode) && inode->i_size)
1867 return -EINVAL;
1868
1869 set_compress_context(inode);
1870 }
1871 }
1872 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1873 if (masked_flags & F2FS_COMPR_FL)
1874 return -EINVAL;
1875 }
1876
1877 fi->i_flags = iflags | (fi->i_flags & ~mask);
1878 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1879 (fi->i_flags & F2FS_NOCOMP_FL));
1880
1881 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1882 set_inode_flag(inode, FI_PROJ_INHERIT);
1883 else
1884 clear_inode_flag(inode, FI_PROJ_INHERIT);
1885
1886 inode->i_ctime = current_time(inode);
1887 f2fs_set_inode_flags(inode);
1888 f2fs_mark_inode_dirty_sync(inode, true);
1889 return 0;
1890 }
1891
1892 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1893
1894 /*
1895 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1896 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1897 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1898 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1899 *
1900 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1901 * FS_IOC_FSSETXATTR is done by the VFS.
1902 */
1903
1904 static const struct {
1905 u32 iflag;
1906 u32 fsflag;
1907 } f2fs_fsflags_map[] = {
1908 { F2FS_COMPR_FL, FS_COMPR_FL },
1909 { F2FS_SYNC_FL, FS_SYNC_FL },
1910 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1911 { F2FS_APPEND_FL, FS_APPEND_FL },
1912 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1913 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1914 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
1915 { F2FS_INDEX_FL, FS_INDEX_FL },
1916 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1917 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1918 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1919 };
1920
1921 #define F2FS_GETTABLE_FS_FL ( \
1922 FS_COMPR_FL | \
1923 FS_SYNC_FL | \
1924 FS_IMMUTABLE_FL | \
1925 FS_APPEND_FL | \
1926 FS_NODUMP_FL | \
1927 FS_NOATIME_FL | \
1928 FS_NOCOMP_FL | \
1929 FS_INDEX_FL | \
1930 FS_DIRSYNC_FL | \
1931 FS_PROJINHERIT_FL | \
1932 FS_ENCRYPT_FL | \
1933 FS_INLINE_DATA_FL | \
1934 FS_NOCOW_FL | \
1935 FS_VERITY_FL | \
1936 FS_CASEFOLD_FL)
1937
1938 #define F2FS_SETTABLE_FS_FL ( \
1939 FS_COMPR_FL | \
1940 FS_SYNC_FL | \
1941 FS_IMMUTABLE_FL | \
1942 FS_APPEND_FL | \
1943 FS_NODUMP_FL | \
1944 FS_NOATIME_FL | \
1945 FS_NOCOMP_FL | \
1946 FS_DIRSYNC_FL | \
1947 FS_PROJINHERIT_FL | \
1948 FS_CASEFOLD_FL)
1949
1950 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1951 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1952 {
1953 u32 fsflags = 0;
1954 int i;
1955
1956 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1957 if (iflags & f2fs_fsflags_map[i].iflag)
1958 fsflags |= f2fs_fsflags_map[i].fsflag;
1959
1960 return fsflags;
1961 }
1962
1963 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1964 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1965 {
1966 u32 iflags = 0;
1967 int i;
1968
1969 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1970 if (fsflags & f2fs_fsflags_map[i].fsflag)
1971 iflags |= f2fs_fsflags_map[i].iflag;
1972
1973 return iflags;
1974 }
1975
1976 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1977 {
1978 struct inode *inode = file_inode(filp);
1979
1980 return put_user(inode->i_generation, (int __user *)arg);
1981 }
1982
1983 static int f2fs_ioc_start_atomic_write(struct file *filp)
1984 {
1985 struct inode *inode = file_inode(filp);
1986 struct f2fs_inode_info *fi = F2FS_I(inode);
1987 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1988 int ret;
1989
1990 if (!inode_owner_or_capable(&init_user_ns, inode))
1991 return -EACCES;
1992
1993 if (!S_ISREG(inode->i_mode))
1994 return -EINVAL;
1995
1996 if (filp->f_flags & O_DIRECT)
1997 return -EINVAL;
1998
1999 ret = mnt_want_write_file(filp);
2000 if (ret)
2001 return ret;
2002
2003 inode_lock(inode);
2004
2005 if (!f2fs_disable_compressed_file(inode)) {
2006 ret = -EINVAL;
2007 goto out;
2008 }
2009
2010 if (f2fs_is_atomic_file(inode)) {
2011 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2012 ret = -EINVAL;
2013 goto out;
2014 }
2015
2016 ret = f2fs_convert_inline_inode(inode);
2017 if (ret)
2018 goto out;
2019
2020 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2021
2022 /*
2023 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2024 * f2fs_is_atomic_file.
2025 */
2026 if (get_dirty_pages(inode))
2027 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2028 inode->i_ino, get_dirty_pages(inode));
2029 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2030 if (ret) {
2031 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2032 goto out;
2033 }
2034
2035 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2036 if (list_empty(&fi->inmem_ilist))
2037 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2038 sbi->atomic_files++;
2039 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2040
2041 /* add inode in inmem_list first and set atomic_file */
2042 set_inode_flag(inode, FI_ATOMIC_FILE);
2043 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2044 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2045
2046 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2047 F2FS_I(inode)->inmem_task = current;
2048 stat_update_max_atomic_write(inode);
2049 out:
2050 inode_unlock(inode);
2051 mnt_drop_write_file(filp);
2052 return ret;
2053 }
2054
2055 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2056 {
2057 struct inode *inode = file_inode(filp);
2058 int ret;
2059
2060 if (!inode_owner_or_capable(&init_user_ns, inode))
2061 return -EACCES;
2062
2063 ret = mnt_want_write_file(filp);
2064 if (ret)
2065 return ret;
2066
2067 f2fs_balance_fs(F2FS_I_SB(inode), true);
2068
2069 inode_lock(inode);
2070
2071 if (f2fs_is_volatile_file(inode)) {
2072 ret = -EINVAL;
2073 goto err_out;
2074 }
2075
2076 if (f2fs_is_atomic_file(inode)) {
2077 ret = f2fs_commit_inmem_pages(inode);
2078 if (ret)
2079 goto err_out;
2080
2081 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2082 if (!ret)
2083 f2fs_drop_inmem_pages(inode);
2084 } else {
2085 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2086 }
2087 err_out:
2088 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2089 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2090 ret = -EINVAL;
2091 }
2092 inode_unlock(inode);
2093 mnt_drop_write_file(filp);
2094 return ret;
2095 }
2096
2097 static int f2fs_ioc_start_volatile_write(struct file *filp)
2098 {
2099 struct inode *inode = file_inode(filp);
2100 int ret;
2101
2102 if (!inode_owner_or_capable(&init_user_ns, inode))
2103 return -EACCES;
2104
2105 if (!S_ISREG(inode->i_mode))
2106 return -EINVAL;
2107
2108 ret = mnt_want_write_file(filp);
2109 if (ret)
2110 return ret;
2111
2112 inode_lock(inode);
2113
2114 if (f2fs_is_volatile_file(inode))
2115 goto out;
2116
2117 ret = f2fs_convert_inline_inode(inode);
2118 if (ret)
2119 goto out;
2120
2121 stat_inc_volatile_write(inode);
2122 stat_update_max_volatile_write(inode);
2123
2124 set_inode_flag(inode, FI_VOLATILE_FILE);
2125 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2126 out:
2127 inode_unlock(inode);
2128 mnt_drop_write_file(filp);
2129 return ret;
2130 }
2131
2132 static int f2fs_ioc_release_volatile_write(struct file *filp)
2133 {
2134 struct inode *inode = file_inode(filp);
2135 int ret;
2136
2137 if (!inode_owner_or_capable(&init_user_ns, inode))
2138 return -EACCES;
2139
2140 ret = mnt_want_write_file(filp);
2141 if (ret)
2142 return ret;
2143
2144 inode_lock(inode);
2145
2146 if (!f2fs_is_volatile_file(inode))
2147 goto out;
2148
2149 if (!f2fs_is_first_block_written(inode)) {
2150 ret = truncate_partial_data_page(inode, 0, true);
2151 goto out;
2152 }
2153
2154 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2155 out:
2156 inode_unlock(inode);
2157 mnt_drop_write_file(filp);
2158 return ret;
2159 }
2160
2161 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2162 {
2163 struct inode *inode = file_inode(filp);
2164 int ret;
2165
2166 if (!inode_owner_or_capable(&init_user_ns, inode))
2167 return -EACCES;
2168
2169 ret = mnt_want_write_file(filp);
2170 if (ret)
2171 return ret;
2172
2173 inode_lock(inode);
2174
2175 if (f2fs_is_atomic_file(inode))
2176 f2fs_drop_inmem_pages(inode);
2177 if (f2fs_is_volatile_file(inode)) {
2178 clear_inode_flag(inode, FI_VOLATILE_FILE);
2179 stat_dec_volatile_write(inode);
2180 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2181 }
2182
2183 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2184
2185 inode_unlock(inode);
2186
2187 mnt_drop_write_file(filp);
2188 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2189 return ret;
2190 }
2191
2192 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2193 {
2194 struct inode *inode = file_inode(filp);
2195 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2196 struct super_block *sb = sbi->sb;
2197 __u32 in;
2198 int ret = 0;
2199
2200 if (!capable(CAP_SYS_ADMIN))
2201 return -EPERM;
2202
2203 if (get_user(in, (__u32 __user *)arg))
2204 return -EFAULT;
2205
2206 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2207 ret = mnt_want_write_file(filp);
2208 if (ret) {
2209 if (ret == -EROFS) {
2210 ret = 0;
2211 f2fs_stop_checkpoint(sbi, false);
2212 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2213 trace_f2fs_shutdown(sbi, in, ret);
2214 }
2215 return ret;
2216 }
2217 }
2218
2219 switch (in) {
2220 case F2FS_GOING_DOWN_FULLSYNC:
2221 ret = freeze_bdev(sb->s_bdev);
2222 if (ret)
2223 goto out;
2224 f2fs_stop_checkpoint(sbi, false);
2225 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2226 thaw_bdev(sb->s_bdev);
2227 break;
2228 case F2FS_GOING_DOWN_METASYNC:
2229 /* do checkpoint only */
2230 ret = f2fs_sync_fs(sb, 1);
2231 if (ret)
2232 goto out;
2233 f2fs_stop_checkpoint(sbi, false);
2234 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2235 break;
2236 case F2FS_GOING_DOWN_NOSYNC:
2237 f2fs_stop_checkpoint(sbi, false);
2238 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2239 break;
2240 case F2FS_GOING_DOWN_METAFLUSH:
2241 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2242 f2fs_stop_checkpoint(sbi, false);
2243 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2244 break;
2245 case F2FS_GOING_DOWN_NEED_FSCK:
2246 set_sbi_flag(sbi, SBI_NEED_FSCK);
2247 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2248 set_sbi_flag(sbi, SBI_IS_DIRTY);
2249 /* do checkpoint only */
2250 ret = f2fs_sync_fs(sb, 1);
2251 goto out;
2252 default:
2253 ret = -EINVAL;
2254 goto out;
2255 }
2256
2257 f2fs_stop_gc_thread(sbi);
2258 f2fs_stop_discard_thread(sbi);
2259
2260 f2fs_drop_discard_cmd(sbi);
2261 clear_opt(sbi, DISCARD);
2262
2263 f2fs_update_time(sbi, REQ_TIME);
2264 out:
2265 if (in != F2FS_GOING_DOWN_FULLSYNC)
2266 mnt_drop_write_file(filp);
2267
2268 trace_f2fs_shutdown(sbi, in, ret);
2269
2270 return ret;
2271 }
2272
2273 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2274 {
2275 struct inode *inode = file_inode(filp);
2276 struct super_block *sb = inode->i_sb;
2277 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2278 struct fstrim_range range;
2279 int ret;
2280
2281 if (!capable(CAP_SYS_ADMIN))
2282 return -EPERM;
2283
2284 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2285 return -EOPNOTSUPP;
2286
2287 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2288 sizeof(range)))
2289 return -EFAULT;
2290
2291 ret = mnt_want_write_file(filp);
2292 if (ret)
2293 return ret;
2294
2295 range.minlen = max((unsigned int)range.minlen,
2296 q->limits.discard_granularity);
2297 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2298 mnt_drop_write_file(filp);
2299 if (ret < 0)
2300 return ret;
2301
2302 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2303 sizeof(range)))
2304 return -EFAULT;
2305 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2306 return 0;
2307 }
2308
2309 static bool uuid_is_nonzero(__u8 u[16])
2310 {
2311 int i;
2312
2313 for (i = 0; i < 16; i++)
2314 if (u[i])
2315 return true;
2316 return false;
2317 }
2318
2319 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2320 {
2321 struct inode *inode = file_inode(filp);
2322
2323 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2324 return -EOPNOTSUPP;
2325
2326 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2327
2328 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2329 }
2330
2331 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2332 {
2333 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2334 return -EOPNOTSUPP;
2335 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2336 }
2337
2338 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2339 {
2340 struct inode *inode = file_inode(filp);
2341 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2342 int err;
2343
2344 if (!f2fs_sb_has_encrypt(sbi))
2345 return -EOPNOTSUPP;
2346
2347 err = mnt_want_write_file(filp);
2348 if (err)
2349 return err;
2350
2351 down_write(&sbi->sb_lock);
2352
2353 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2354 goto got_it;
2355
2356 /* update superblock with uuid */
2357 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2358
2359 err = f2fs_commit_super(sbi, false);
2360 if (err) {
2361 /* undo new data */
2362 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2363 goto out_err;
2364 }
2365 got_it:
2366 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2367 16))
2368 err = -EFAULT;
2369 out_err:
2370 up_write(&sbi->sb_lock);
2371 mnt_drop_write_file(filp);
2372 return err;
2373 }
2374
2375 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2376 unsigned long arg)
2377 {
2378 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2379 return -EOPNOTSUPP;
2380
2381 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2382 }
2383
2384 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2385 {
2386 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2387 return -EOPNOTSUPP;
2388
2389 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2390 }
2391
2392 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2393 {
2394 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2395 return -EOPNOTSUPP;
2396
2397 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2398 }
2399
2400 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2401 unsigned long arg)
2402 {
2403 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2404 return -EOPNOTSUPP;
2405
2406 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2407 }
2408
2409 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2410 unsigned long arg)
2411 {
2412 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2413 return -EOPNOTSUPP;
2414
2415 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2416 }
2417
2418 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2419 {
2420 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2421 return -EOPNOTSUPP;
2422
2423 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2424 }
2425
2426 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2427 {
2428 struct inode *inode = file_inode(filp);
2429 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2430 __u32 sync;
2431 int ret;
2432
2433 if (!capable(CAP_SYS_ADMIN))
2434 return -EPERM;
2435
2436 if (get_user(sync, (__u32 __user *)arg))
2437 return -EFAULT;
2438
2439 if (f2fs_readonly(sbi->sb))
2440 return -EROFS;
2441
2442 ret = mnt_want_write_file(filp);
2443 if (ret)
2444 return ret;
2445
2446 if (!sync) {
2447 if (!down_write_trylock(&sbi->gc_lock)) {
2448 ret = -EBUSY;
2449 goto out;
2450 }
2451 } else {
2452 down_write(&sbi->gc_lock);
2453 }
2454
2455 ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO);
2456 out:
2457 mnt_drop_write_file(filp);
2458 return ret;
2459 }
2460
2461 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2462 {
2463 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2464 u64 end;
2465 int ret;
2466
2467 if (!capable(CAP_SYS_ADMIN))
2468 return -EPERM;
2469 if (f2fs_readonly(sbi->sb))
2470 return -EROFS;
2471
2472 end = range->start + range->len;
2473 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2474 end >= MAX_BLKADDR(sbi))
2475 return -EINVAL;
2476
2477 ret = mnt_want_write_file(filp);
2478 if (ret)
2479 return ret;
2480
2481 do_more:
2482 if (!range->sync) {
2483 if (!down_write_trylock(&sbi->gc_lock)) {
2484 ret = -EBUSY;
2485 goto out;
2486 }
2487 } else {
2488 down_write(&sbi->gc_lock);
2489 }
2490
2491 ret = f2fs_gc(sbi, range->sync, true, false,
2492 GET_SEGNO(sbi, range->start));
2493 if (ret) {
2494 if (ret == -EBUSY)
2495 ret = -EAGAIN;
2496 goto out;
2497 }
2498 range->start += BLKS_PER_SEC(sbi);
2499 if (range->start <= end)
2500 goto do_more;
2501 out:
2502 mnt_drop_write_file(filp);
2503 return ret;
2504 }
2505
2506 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2507 {
2508 struct f2fs_gc_range range;
2509
2510 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2511 sizeof(range)))
2512 return -EFAULT;
2513 return __f2fs_ioc_gc_range(filp, &range);
2514 }
2515
2516 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2517 {
2518 struct inode *inode = file_inode(filp);
2519 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2520 int ret;
2521
2522 if (!capable(CAP_SYS_ADMIN))
2523 return -EPERM;
2524
2525 if (f2fs_readonly(sbi->sb))
2526 return -EROFS;
2527
2528 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2529 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2530 return -EINVAL;
2531 }
2532
2533 ret = mnt_want_write_file(filp);
2534 if (ret)
2535 return ret;
2536
2537 ret = f2fs_sync_fs(sbi->sb, 1);
2538
2539 mnt_drop_write_file(filp);
2540 return ret;
2541 }
2542
2543 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2544 struct file *filp,
2545 struct f2fs_defragment *range)
2546 {
2547 struct inode *inode = file_inode(filp);
2548 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2549 .m_seg_type = NO_CHECK_TYPE,
2550 .m_may_create = false };
2551 struct extent_info ei = {0, 0, 0};
2552 pgoff_t pg_start, pg_end, next_pgofs;
2553 unsigned int blk_per_seg = sbi->blocks_per_seg;
2554 unsigned int total = 0, sec_num;
2555 block_t blk_end = 0;
2556 bool fragmented = false;
2557 int err;
2558
2559 /* if in-place-update policy is enabled, don't waste time here */
2560 if (f2fs_should_update_inplace(inode, NULL))
2561 return -EINVAL;
2562
2563 pg_start = range->start >> PAGE_SHIFT;
2564 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2565
2566 f2fs_balance_fs(sbi, true);
2567
2568 inode_lock(inode);
2569
2570 /* writeback all dirty pages in the range */
2571 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2572 range->start + range->len - 1);
2573 if (err)
2574 goto out;
2575
2576 /*
2577 * lookup mapping info in extent cache, skip defragmenting if physical
2578 * block addresses are continuous.
2579 */
2580 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2581 if (ei.fofs + ei.len >= pg_end)
2582 goto out;
2583 }
2584
2585 map.m_lblk = pg_start;
2586 map.m_next_pgofs = &next_pgofs;
2587
2588 /*
2589 * lookup mapping info in dnode page cache, skip defragmenting if all
2590 * physical block addresses are continuous even if there are hole(s)
2591 * in logical blocks.
2592 */
2593 while (map.m_lblk < pg_end) {
2594 map.m_len = pg_end - map.m_lblk;
2595 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2596 if (err)
2597 goto out;
2598
2599 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2600 map.m_lblk = next_pgofs;
2601 continue;
2602 }
2603
2604 if (blk_end && blk_end != map.m_pblk)
2605 fragmented = true;
2606
2607 /* record total count of block that we're going to move */
2608 total += map.m_len;
2609
2610 blk_end = map.m_pblk + map.m_len;
2611
2612 map.m_lblk += map.m_len;
2613 }
2614
2615 if (!fragmented) {
2616 total = 0;
2617 goto out;
2618 }
2619
2620 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2621
2622 /*
2623 * make sure there are enough free section for LFS allocation, this can
2624 * avoid defragment running in SSR mode when free section are allocated
2625 * intensively
2626 */
2627 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2628 err = -EAGAIN;
2629 goto out;
2630 }
2631
2632 map.m_lblk = pg_start;
2633 map.m_len = pg_end - pg_start;
2634 total = 0;
2635
2636 while (map.m_lblk < pg_end) {
2637 pgoff_t idx;
2638 int cnt = 0;
2639
2640 do_map:
2641 map.m_len = pg_end - map.m_lblk;
2642 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2643 if (err)
2644 goto clear_out;
2645
2646 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2647 map.m_lblk = next_pgofs;
2648 goto check;
2649 }
2650
2651 set_inode_flag(inode, FI_DO_DEFRAG);
2652
2653 idx = map.m_lblk;
2654 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2655 struct page *page;
2656
2657 page = f2fs_get_lock_data_page(inode, idx, true);
2658 if (IS_ERR(page)) {
2659 err = PTR_ERR(page);
2660 goto clear_out;
2661 }
2662
2663 set_page_dirty(page);
2664 f2fs_put_page(page, 1);
2665
2666 idx++;
2667 cnt++;
2668 total++;
2669 }
2670
2671 map.m_lblk = idx;
2672 check:
2673 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2674 goto do_map;
2675
2676 clear_inode_flag(inode, FI_DO_DEFRAG);
2677
2678 err = filemap_fdatawrite(inode->i_mapping);
2679 if (err)
2680 goto out;
2681 }
2682 clear_out:
2683 clear_inode_flag(inode, FI_DO_DEFRAG);
2684 out:
2685 inode_unlock(inode);
2686 if (!err)
2687 range->len = (u64)total << PAGE_SHIFT;
2688 return err;
2689 }
2690
2691 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2692 {
2693 struct inode *inode = file_inode(filp);
2694 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2695 struct f2fs_defragment range;
2696 int err;
2697
2698 if (!capable(CAP_SYS_ADMIN))
2699 return -EPERM;
2700
2701 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2702 return -EINVAL;
2703
2704 if (f2fs_readonly(sbi->sb))
2705 return -EROFS;
2706
2707 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2708 sizeof(range)))
2709 return -EFAULT;
2710
2711 /* verify alignment of offset & size */
2712 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2713 return -EINVAL;
2714
2715 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2716 max_file_blocks(inode)))
2717 return -EINVAL;
2718
2719 err = mnt_want_write_file(filp);
2720 if (err)
2721 return err;
2722
2723 err = f2fs_defragment_range(sbi, filp, &range);
2724 mnt_drop_write_file(filp);
2725
2726 f2fs_update_time(sbi, REQ_TIME);
2727 if (err < 0)
2728 return err;
2729
2730 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2731 sizeof(range)))
2732 return -EFAULT;
2733
2734 return 0;
2735 }
2736
2737 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2738 struct file *file_out, loff_t pos_out, size_t len)
2739 {
2740 struct inode *src = file_inode(file_in);
2741 struct inode *dst = file_inode(file_out);
2742 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2743 size_t olen = len, dst_max_i_size = 0;
2744 size_t dst_osize;
2745 int ret;
2746
2747 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2748 src->i_sb != dst->i_sb)
2749 return -EXDEV;
2750
2751 if (unlikely(f2fs_readonly(src->i_sb)))
2752 return -EROFS;
2753
2754 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2755 return -EINVAL;
2756
2757 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2758 return -EOPNOTSUPP;
2759
2760 if (pos_out < 0 || pos_in < 0)
2761 return -EINVAL;
2762
2763 if (src == dst) {
2764 if (pos_in == pos_out)
2765 return 0;
2766 if (pos_out > pos_in && pos_out < pos_in + len)
2767 return -EINVAL;
2768 }
2769
2770 inode_lock(src);
2771 if (src != dst) {
2772 ret = -EBUSY;
2773 if (!inode_trylock(dst))
2774 goto out;
2775 }
2776
2777 ret = -EINVAL;
2778 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2779 goto out_unlock;
2780 if (len == 0)
2781 olen = len = src->i_size - pos_in;
2782 if (pos_in + len == src->i_size)
2783 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2784 if (len == 0) {
2785 ret = 0;
2786 goto out_unlock;
2787 }
2788
2789 dst_osize = dst->i_size;
2790 if (pos_out + olen > dst->i_size)
2791 dst_max_i_size = pos_out + olen;
2792
2793 /* verify the end result is block aligned */
2794 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2795 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2796 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2797 goto out_unlock;
2798
2799 ret = f2fs_convert_inline_inode(src);
2800 if (ret)
2801 goto out_unlock;
2802
2803 ret = f2fs_convert_inline_inode(dst);
2804 if (ret)
2805 goto out_unlock;
2806
2807 /* write out all dirty pages from offset */
2808 ret = filemap_write_and_wait_range(src->i_mapping,
2809 pos_in, pos_in + len);
2810 if (ret)
2811 goto out_unlock;
2812
2813 ret = filemap_write_and_wait_range(dst->i_mapping,
2814 pos_out, pos_out + len);
2815 if (ret)
2816 goto out_unlock;
2817
2818 f2fs_balance_fs(sbi, true);
2819
2820 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2821 if (src != dst) {
2822 ret = -EBUSY;
2823 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2824 goto out_src;
2825 }
2826
2827 f2fs_lock_op(sbi);
2828 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2829 pos_out >> F2FS_BLKSIZE_BITS,
2830 len >> F2FS_BLKSIZE_BITS, false);
2831
2832 if (!ret) {
2833 if (dst_max_i_size)
2834 f2fs_i_size_write(dst, dst_max_i_size);
2835 else if (dst_osize != dst->i_size)
2836 f2fs_i_size_write(dst, dst_osize);
2837 }
2838 f2fs_unlock_op(sbi);
2839
2840 if (src != dst)
2841 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2842 out_src:
2843 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2844 out_unlock:
2845 if (src != dst)
2846 inode_unlock(dst);
2847 out:
2848 inode_unlock(src);
2849 return ret;
2850 }
2851
2852 static int __f2fs_ioc_move_range(struct file *filp,
2853 struct f2fs_move_range *range)
2854 {
2855 struct fd dst;
2856 int err;
2857
2858 if (!(filp->f_mode & FMODE_READ) ||
2859 !(filp->f_mode & FMODE_WRITE))
2860 return -EBADF;
2861
2862 dst = fdget(range->dst_fd);
2863 if (!dst.file)
2864 return -EBADF;
2865
2866 if (!(dst.file->f_mode & FMODE_WRITE)) {
2867 err = -EBADF;
2868 goto err_out;
2869 }
2870
2871 err = mnt_want_write_file(filp);
2872 if (err)
2873 goto err_out;
2874
2875 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2876 range->pos_out, range->len);
2877
2878 mnt_drop_write_file(filp);
2879 err_out:
2880 fdput(dst);
2881 return err;
2882 }
2883
2884 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2885 {
2886 struct f2fs_move_range range;
2887
2888 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2889 sizeof(range)))
2890 return -EFAULT;
2891 return __f2fs_ioc_move_range(filp, &range);
2892 }
2893
2894 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2895 {
2896 struct inode *inode = file_inode(filp);
2897 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2898 struct sit_info *sm = SIT_I(sbi);
2899 unsigned int start_segno = 0, end_segno = 0;
2900 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2901 struct f2fs_flush_device range;
2902 int ret;
2903
2904 if (!capable(CAP_SYS_ADMIN))
2905 return -EPERM;
2906
2907 if (f2fs_readonly(sbi->sb))
2908 return -EROFS;
2909
2910 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2911 return -EINVAL;
2912
2913 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2914 sizeof(range)))
2915 return -EFAULT;
2916
2917 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2918 __is_large_section(sbi)) {
2919 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2920 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2921 return -EINVAL;
2922 }
2923
2924 ret = mnt_want_write_file(filp);
2925 if (ret)
2926 return ret;
2927
2928 if (range.dev_num != 0)
2929 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2930 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2931
2932 start_segno = sm->last_victim[FLUSH_DEVICE];
2933 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2934 start_segno = dev_start_segno;
2935 end_segno = min(start_segno + range.segments, dev_end_segno);
2936
2937 while (start_segno < end_segno) {
2938 if (!down_write_trylock(&sbi->gc_lock)) {
2939 ret = -EBUSY;
2940 goto out;
2941 }
2942 sm->last_victim[GC_CB] = end_segno + 1;
2943 sm->last_victim[GC_GREEDY] = end_segno + 1;
2944 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2945 ret = f2fs_gc(sbi, true, true, true, start_segno);
2946 if (ret == -EAGAIN)
2947 ret = 0;
2948 else if (ret < 0)
2949 break;
2950 start_segno++;
2951 }
2952 out:
2953 mnt_drop_write_file(filp);
2954 return ret;
2955 }
2956
2957 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2958 {
2959 struct inode *inode = file_inode(filp);
2960 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2961
2962 /* Must validate to set it with SQLite behavior in Android. */
2963 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2964
2965 return put_user(sb_feature, (u32 __user *)arg);
2966 }
2967
2968 #ifdef CONFIG_QUOTA
2969 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2970 {
2971 struct dquot *transfer_to[MAXQUOTAS] = {};
2972 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2973 struct super_block *sb = sbi->sb;
2974 int err = 0;
2975
2976 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2977 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2978 err = __dquot_transfer(inode, transfer_to);
2979 if (err)
2980 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2981 dqput(transfer_to[PRJQUOTA]);
2982 }
2983 return err;
2984 }
2985
2986 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2987 {
2988 struct f2fs_inode_info *fi = F2FS_I(inode);
2989 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2990 struct page *ipage;
2991 kprojid_t kprojid;
2992 int err;
2993
2994 if (!f2fs_sb_has_project_quota(sbi)) {
2995 if (projid != F2FS_DEF_PROJID)
2996 return -EOPNOTSUPP;
2997 else
2998 return 0;
2999 }
3000
3001 if (!f2fs_has_extra_attr(inode))
3002 return -EOPNOTSUPP;
3003
3004 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3005
3006 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
3007 return 0;
3008
3009 err = -EPERM;
3010 /* Is it quota file? Do not allow user to mess with it */
3011 if (IS_NOQUOTA(inode))
3012 return err;
3013
3014 ipage = f2fs_get_node_page(sbi, inode->i_ino);
3015 if (IS_ERR(ipage))
3016 return PTR_ERR(ipage);
3017
3018 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
3019 i_projid)) {
3020 err = -EOVERFLOW;
3021 f2fs_put_page(ipage, 1);
3022 return err;
3023 }
3024 f2fs_put_page(ipage, 1);
3025
3026 err = dquot_initialize(inode);
3027 if (err)
3028 return err;
3029
3030 f2fs_lock_op(sbi);
3031 err = f2fs_transfer_project_quota(inode, kprojid);
3032 if (err)
3033 goto out_unlock;
3034
3035 F2FS_I(inode)->i_projid = kprojid;
3036 inode->i_ctime = current_time(inode);
3037 f2fs_mark_inode_dirty_sync(inode, true);
3038 out_unlock:
3039 f2fs_unlock_op(sbi);
3040 return err;
3041 }
3042 #else
3043 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3044 {
3045 return 0;
3046 }
3047
3048 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3049 {
3050 if (projid != F2FS_DEF_PROJID)
3051 return -EOPNOTSUPP;
3052 return 0;
3053 }
3054 #endif
3055
3056 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3057 {
3058 struct inode *inode = d_inode(dentry);
3059 struct f2fs_inode_info *fi = F2FS_I(inode);
3060 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3061
3062 if (IS_ENCRYPTED(inode))
3063 fsflags |= FS_ENCRYPT_FL;
3064 if (IS_VERITY(inode))
3065 fsflags |= FS_VERITY_FL;
3066 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3067 fsflags |= FS_INLINE_DATA_FL;
3068 if (is_inode_flag_set(inode, FI_PIN_FILE))
3069 fsflags |= FS_NOCOW_FL;
3070
3071 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3072
3073 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3074 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3075
3076 return 0;
3077 }
3078
3079 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3080 struct dentry *dentry, struct fileattr *fa)
3081 {
3082 struct inode *inode = d_inode(dentry);
3083 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3084 u32 iflags;
3085 int err;
3086
3087 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3088 return -EIO;
3089 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3090 return -ENOSPC;
3091 if (fsflags & ~F2FS_GETTABLE_FS_FL)
3092 return -EOPNOTSUPP;
3093 fsflags &= F2FS_SETTABLE_FS_FL;
3094 if (!fa->flags_valid)
3095 mask &= FS_COMMON_FL;
3096
3097 iflags = f2fs_fsflags_to_iflags(fsflags);
3098 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3099 return -EOPNOTSUPP;
3100
3101 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3102 if (!err)
3103 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3104
3105 return err;
3106 }
3107
3108 int f2fs_pin_file_control(struct inode *inode, bool inc)
3109 {
3110 struct f2fs_inode_info *fi = F2FS_I(inode);
3111 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3112
3113 /* Use i_gc_failures for normal file as a risk signal. */
3114 if (inc)
3115 f2fs_i_gc_failures_write(inode,
3116 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3117
3118 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3119 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3120 __func__, inode->i_ino,
3121 fi->i_gc_failures[GC_FAILURE_PIN]);
3122 clear_inode_flag(inode, FI_PIN_FILE);
3123 return -EAGAIN;
3124 }
3125 return 0;
3126 }
3127
3128 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3129 {
3130 struct inode *inode = file_inode(filp);
3131 __u32 pin;
3132 int ret = 0;
3133
3134 if (get_user(pin, (__u32 __user *)arg))
3135 return -EFAULT;
3136
3137 if (!S_ISREG(inode->i_mode))
3138 return -EINVAL;
3139
3140 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3141 return -EROFS;
3142
3143 ret = mnt_want_write_file(filp);
3144 if (ret)
3145 return ret;
3146
3147 inode_lock(inode);
3148
3149 if (!pin) {
3150 clear_inode_flag(inode, FI_PIN_FILE);
3151 f2fs_i_gc_failures_write(inode, 0);
3152 goto done;
3153 }
3154
3155 if (f2fs_should_update_outplace(inode, NULL)) {
3156 ret = -EINVAL;
3157 goto out;
3158 }
3159
3160 if (f2fs_pin_file_control(inode, false)) {
3161 ret = -EAGAIN;
3162 goto out;
3163 }
3164
3165 ret = f2fs_convert_inline_inode(inode);
3166 if (ret)
3167 goto out;
3168
3169 if (!f2fs_disable_compressed_file(inode)) {
3170 ret = -EOPNOTSUPP;
3171 goto out;
3172 }
3173
3174 set_inode_flag(inode, FI_PIN_FILE);
3175 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3176 done:
3177 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3178 out:
3179 inode_unlock(inode);
3180 mnt_drop_write_file(filp);
3181 return ret;
3182 }
3183
3184 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3185 {
3186 struct inode *inode = file_inode(filp);
3187 __u32 pin = 0;
3188
3189 if (is_inode_flag_set(inode, FI_PIN_FILE))
3190 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3191 return put_user(pin, (u32 __user *)arg);
3192 }
3193
3194 int f2fs_precache_extents(struct inode *inode)
3195 {
3196 struct f2fs_inode_info *fi = F2FS_I(inode);
3197 struct f2fs_map_blocks map;
3198 pgoff_t m_next_extent;
3199 loff_t end;
3200 int err;
3201
3202 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3203 return -EOPNOTSUPP;
3204
3205 map.m_lblk = 0;
3206 map.m_next_pgofs = NULL;
3207 map.m_next_extent = &m_next_extent;
3208 map.m_seg_type = NO_CHECK_TYPE;
3209 map.m_may_create = false;
3210 end = max_file_blocks(inode);
3211
3212 while (map.m_lblk < end) {
3213 map.m_len = end - map.m_lblk;
3214
3215 down_write(&fi->i_gc_rwsem[WRITE]);
3216 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3217 up_write(&fi->i_gc_rwsem[WRITE]);
3218 if (err)
3219 return err;
3220
3221 map.m_lblk = m_next_extent;
3222 }
3223
3224 return 0;
3225 }
3226
3227 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3228 {
3229 return f2fs_precache_extents(file_inode(filp));
3230 }
3231
3232 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3233 {
3234 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3235 __u64 block_count;
3236
3237 if (!capable(CAP_SYS_ADMIN))
3238 return -EPERM;
3239
3240 if (f2fs_readonly(sbi->sb))
3241 return -EROFS;
3242
3243 if (copy_from_user(&block_count, (void __user *)arg,
3244 sizeof(block_count)))
3245 return -EFAULT;
3246
3247 return f2fs_resize_fs(sbi, block_count);
3248 }
3249
3250 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3251 {
3252 struct inode *inode = file_inode(filp);
3253
3254 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3255
3256 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3257 f2fs_warn(F2FS_I_SB(inode),
3258 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3259 inode->i_ino);
3260 return -EOPNOTSUPP;
3261 }
3262
3263 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3264 }
3265
3266 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3267 {
3268 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3269 return -EOPNOTSUPP;
3270
3271 return fsverity_ioctl_measure(filp, (void __user *)arg);
3272 }
3273
3274 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3275 {
3276 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3277 return -EOPNOTSUPP;
3278
3279 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3280 }
3281
3282 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3283 {
3284 struct inode *inode = file_inode(filp);
3285 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3286 char *vbuf;
3287 int count;
3288 int err = 0;
3289
3290 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3291 if (!vbuf)
3292 return -ENOMEM;
3293
3294 down_read(&sbi->sb_lock);
3295 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3296 ARRAY_SIZE(sbi->raw_super->volume_name),
3297 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3298 up_read(&sbi->sb_lock);
3299
3300 if (copy_to_user((char __user *)arg, vbuf,
3301 min(FSLABEL_MAX, count)))
3302 err = -EFAULT;
3303
3304 kfree(vbuf);
3305 return err;
3306 }
3307
3308 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3309 {
3310 struct inode *inode = file_inode(filp);
3311 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3312 char *vbuf;
3313 int err = 0;
3314
3315 if (!capable(CAP_SYS_ADMIN))
3316 return -EPERM;
3317
3318 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3319 if (IS_ERR(vbuf))
3320 return PTR_ERR(vbuf);
3321
3322 err = mnt_want_write_file(filp);
3323 if (err)
3324 goto out;
3325
3326 down_write(&sbi->sb_lock);
3327
3328 memset(sbi->raw_super->volume_name, 0,
3329 sizeof(sbi->raw_super->volume_name));
3330 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3331 sbi->raw_super->volume_name,
3332 ARRAY_SIZE(sbi->raw_super->volume_name));
3333
3334 err = f2fs_commit_super(sbi, false);
3335
3336 up_write(&sbi->sb_lock);
3337
3338 mnt_drop_write_file(filp);
3339 out:
3340 kfree(vbuf);
3341 return err;
3342 }
3343
3344 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3345 {
3346 struct inode *inode = file_inode(filp);
3347 __u64 blocks;
3348
3349 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3350 return -EOPNOTSUPP;
3351
3352 if (!f2fs_compressed_file(inode))
3353 return -EINVAL;
3354
3355 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3356 return put_user(blocks, (u64 __user *)arg);
3357 }
3358
3359 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3360 {
3361 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3362 unsigned int released_blocks = 0;
3363 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3364 block_t blkaddr;
3365 int i;
3366
3367 for (i = 0; i < count; i++) {
3368 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3369 dn->ofs_in_node + i);
3370
3371 if (!__is_valid_data_blkaddr(blkaddr))
3372 continue;
3373 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3374 DATA_GENERIC_ENHANCE)))
3375 return -EFSCORRUPTED;
3376 }
3377
3378 while (count) {
3379 int compr_blocks = 0;
3380
3381 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3382 blkaddr = f2fs_data_blkaddr(dn);
3383
3384 if (i == 0) {
3385 if (blkaddr == COMPRESS_ADDR)
3386 continue;
3387 dn->ofs_in_node += cluster_size;
3388 goto next;
3389 }
3390
3391 if (__is_valid_data_blkaddr(blkaddr))
3392 compr_blocks++;
3393
3394 if (blkaddr != NEW_ADDR)
3395 continue;
3396
3397 dn->data_blkaddr = NULL_ADDR;
3398 f2fs_set_data_blkaddr(dn);
3399 }
3400
3401 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3402 dec_valid_block_count(sbi, dn->inode,
3403 cluster_size - compr_blocks);
3404
3405 released_blocks += cluster_size - compr_blocks;
3406 next:
3407 count -= cluster_size;
3408 }
3409
3410 return released_blocks;
3411 }
3412
3413 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3414 {
3415 struct inode *inode = file_inode(filp);
3416 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3417 pgoff_t page_idx = 0, last_idx;
3418 unsigned int released_blocks = 0;
3419 int ret;
3420 int writecount;
3421
3422 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3423 return -EOPNOTSUPP;
3424
3425 if (!f2fs_compressed_file(inode))
3426 return -EINVAL;
3427
3428 if (f2fs_readonly(sbi->sb))
3429 return -EROFS;
3430
3431 ret = mnt_want_write_file(filp);
3432 if (ret)
3433 return ret;
3434
3435 f2fs_balance_fs(F2FS_I_SB(inode), true);
3436
3437 inode_lock(inode);
3438
3439 writecount = atomic_read(&inode->i_writecount);
3440 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3441 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3442 ret = -EBUSY;
3443 goto out;
3444 }
3445
3446 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3447 ret = -EINVAL;
3448 goto out;
3449 }
3450
3451 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3452 if (ret)
3453 goto out;
3454
3455 set_inode_flag(inode, FI_COMPRESS_RELEASED);
3456 inode->i_ctime = current_time(inode);
3457 f2fs_mark_inode_dirty_sync(inode, true);
3458
3459 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3460 goto out;
3461
3462 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3463 filemap_invalidate_lock(inode->i_mapping);
3464
3465 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3466
3467 while (page_idx < last_idx) {
3468 struct dnode_of_data dn;
3469 pgoff_t end_offset, count;
3470
3471 set_new_dnode(&dn, inode, NULL, NULL, 0);
3472 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3473 if (ret) {
3474 if (ret == -ENOENT) {
3475 page_idx = f2fs_get_next_page_offset(&dn,
3476 page_idx);
3477 ret = 0;
3478 continue;
3479 }
3480 break;
3481 }
3482
3483 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3484 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3485 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3486
3487 ret = release_compress_blocks(&dn, count);
3488
3489 f2fs_put_dnode(&dn);
3490
3491 if (ret < 0)
3492 break;
3493
3494 page_idx += count;
3495 released_blocks += ret;
3496 }
3497
3498 filemap_invalidate_unlock(inode->i_mapping);
3499 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3500 out:
3501 inode_unlock(inode);
3502
3503 mnt_drop_write_file(filp);
3504
3505 if (ret >= 0) {
3506 ret = put_user(released_blocks, (u64 __user *)arg);
3507 } else if (released_blocks &&
3508 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3509 set_sbi_flag(sbi, SBI_NEED_FSCK);
3510 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3511 "iblocks=%llu, released=%u, compr_blocks=%u, "
3512 "run fsck to fix.",
3513 __func__, inode->i_ino, inode->i_blocks,
3514 released_blocks,
3515 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3516 }
3517
3518 return ret;
3519 }
3520
3521 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3522 {
3523 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3524 unsigned int reserved_blocks = 0;
3525 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3526 block_t blkaddr;
3527 int i;
3528
3529 for (i = 0; i < count; i++) {
3530 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3531 dn->ofs_in_node + i);
3532
3533 if (!__is_valid_data_blkaddr(blkaddr))
3534 continue;
3535 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3536 DATA_GENERIC_ENHANCE)))
3537 return -EFSCORRUPTED;
3538 }
3539
3540 while (count) {
3541 int compr_blocks = 0;
3542 blkcnt_t reserved;
3543 int ret;
3544
3545 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3546 blkaddr = f2fs_data_blkaddr(dn);
3547
3548 if (i == 0) {
3549 if (blkaddr == COMPRESS_ADDR)
3550 continue;
3551 dn->ofs_in_node += cluster_size;
3552 goto next;
3553 }
3554
3555 if (__is_valid_data_blkaddr(blkaddr)) {
3556 compr_blocks++;
3557 continue;
3558 }
3559
3560 dn->data_blkaddr = NEW_ADDR;
3561 f2fs_set_data_blkaddr(dn);
3562 }
3563
3564 reserved = cluster_size - compr_blocks;
3565 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3566 if (ret)
3567 return ret;
3568
3569 if (reserved != cluster_size - compr_blocks)
3570 return -ENOSPC;
3571
3572 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3573
3574 reserved_blocks += reserved;
3575 next:
3576 count -= cluster_size;
3577 }
3578
3579 return reserved_blocks;
3580 }
3581
3582 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3583 {
3584 struct inode *inode = file_inode(filp);
3585 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3586 pgoff_t page_idx = 0, last_idx;
3587 unsigned int reserved_blocks = 0;
3588 int ret;
3589
3590 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3591 return -EOPNOTSUPP;
3592
3593 if (!f2fs_compressed_file(inode))
3594 return -EINVAL;
3595
3596 if (f2fs_readonly(sbi->sb))
3597 return -EROFS;
3598
3599 ret = mnt_want_write_file(filp);
3600 if (ret)
3601 return ret;
3602
3603 if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3604 goto out;
3605
3606 f2fs_balance_fs(F2FS_I_SB(inode), true);
3607
3608 inode_lock(inode);
3609
3610 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3611 ret = -EINVAL;
3612 goto unlock_inode;
3613 }
3614
3615 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3616 filemap_invalidate_lock(inode->i_mapping);
3617
3618 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3619
3620 while (page_idx < last_idx) {
3621 struct dnode_of_data dn;
3622 pgoff_t end_offset, count;
3623
3624 set_new_dnode(&dn, inode, NULL, NULL, 0);
3625 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3626 if (ret) {
3627 if (ret == -ENOENT) {
3628 page_idx = f2fs_get_next_page_offset(&dn,
3629 page_idx);
3630 ret = 0;
3631 continue;
3632 }
3633 break;
3634 }
3635
3636 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3637 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3638 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3639
3640 ret = reserve_compress_blocks(&dn, count);
3641
3642 f2fs_put_dnode(&dn);
3643
3644 if (ret < 0)
3645 break;
3646
3647 page_idx += count;
3648 reserved_blocks += ret;
3649 }
3650
3651 filemap_invalidate_unlock(inode->i_mapping);
3652 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3653
3654 if (ret >= 0) {
3655 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3656 inode->i_ctime = current_time(inode);
3657 f2fs_mark_inode_dirty_sync(inode, true);
3658 }
3659 unlock_inode:
3660 inode_unlock(inode);
3661 out:
3662 mnt_drop_write_file(filp);
3663
3664 if (ret >= 0) {
3665 ret = put_user(reserved_blocks, (u64 __user *)arg);
3666 } else if (reserved_blocks &&
3667 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3668 set_sbi_flag(sbi, SBI_NEED_FSCK);
3669 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3670 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3671 "run fsck to fix.",
3672 __func__, inode->i_ino, inode->i_blocks,
3673 reserved_blocks,
3674 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3675 }
3676
3677 return ret;
3678 }
3679
3680 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3681 pgoff_t off, block_t block, block_t len, u32 flags)
3682 {
3683 struct request_queue *q = bdev_get_queue(bdev);
3684 sector_t sector = SECTOR_FROM_BLOCK(block);
3685 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3686 int ret = 0;
3687
3688 if (!q)
3689 return -ENXIO;
3690
3691 if (flags & F2FS_TRIM_FILE_DISCARD)
3692 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3693 blk_queue_secure_erase(q) ?
3694 BLKDEV_DISCARD_SECURE : 0);
3695
3696 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3697 if (IS_ENCRYPTED(inode))
3698 ret = fscrypt_zeroout_range(inode, off, block, len);
3699 else
3700 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3701 GFP_NOFS, 0);
3702 }
3703
3704 return ret;
3705 }
3706
3707 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3708 {
3709 struct inode *inode = file_inode(filp);
3710 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3711 struct address_space *mapping = inode->i_mapping;
3712 struct block_device *prev_bdev = NULL;
3713 struct f2fs_sectrim_range range;
3714 pgoff_t index, pg_end, prev_index = 0;
3715 block_t prev_block = 0, len = 0;
3716 loff_t end_addr;
3717 bool to_end = false;
3718 int ret = 0;
3719
3720 if (!(filp->f_mode & FMODE_WRITE))
3721 return -EBADF;
3722
3723 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3724 sizeof(range)))
3725 return -EFAULT;
3726
3727 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3728 !S_ISREG(inode->i_mode))
3729 return -EINVAL;
3730
3731 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3732 !f2fs_hw_support_discard(sbi)) ||
3733 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3734 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3735 return -EOPNOTSUPP;
3736
3737 file_start_write(filp);
3738 inode_lock(inode);
3739
3740 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3741 range.start >= inode->i_size) {
3742 ret = -EINVAL;
3743 goto err;
3744 }
3745
3746 if (range.len == 0)
3747 goto err;
3748
3749 if (inode->i_size - range.start > range.len) {
3750 end_addr = range.start + range.len;
3751 } else {
3752 end_addr = range.len == (u64)-1 ?
3753 sbi->sb->s_maxbytes : inode->i_size;
3754 to_end = true;
3755 }
3756
3757 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3758 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3759 ret = -EINVAL;
3760 goto err;
3761 }
3762
3763 index = F2FS_BYTES_TO_BLK(range.start);
3764 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3765
3766 ret = f2fs_convert_inline_inode(inode);
3767 if (ret)
3768 goto err;
3769
3770 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3771 filemap_invalidate_lock(mapping);
3772
3773 ret = filemap_write_and_wait_range(mapping, range.start,
3774 to_end ? LLONG_MAX : end_addr - 1);
3775 if (ret)
3776 goto out;
3777
3778 truncate_inode_pages_range(mapping, range.start,
3779 to_end ? -1 : end_addr - 1);
3780
3781 while (index < pg_end) {
3782 struct dnode_of_data dn;
3783 pgoff_t end_offset, count;
3784 int i;
3785
3786 set_new_dnode(&dn, inode, NULL, NULL, 0);
3787 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3788 if (ret) {
3789 if (ret == -ENOENT) {
3790 index = f2fs_get_next_page_offset(&dn, index);
3791 continue;
3792 }
3793 goto out;
3794 }
3795
3796 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3797 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3798 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3799 struct block_device *cur_bdev;
3800 block_t blkaddr = f2fs_data_blkaddr(&dn);
3801
3802 if (!__is_valid_data_blkaddr(blkaddr))
3803 continue;
3804
3805 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3806 DATA_GENERIC_ENHANCE)) {
3807 ret = -EFSCORRUPTED;
3808 f2fs_put_dnode(&dn);
3809 goto out;
3810 }
3811
3812 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3813 if (f2fs_is_multi_device(sbi)) {
3814 int di = f2fs_target_device_index(sbi, blkaddr);
3815
3816 blkaddr -= FDEV(di).start_blk;
3817 }
3818
3819 if (len) {
3820 if (prev_bdev == cur_bdev &&
3821 index == prev_index + len &&
3822 blkaddr == prev_block + len) {
3823 len++;
3824 } else {
3825 ret = f2fs_secure_erase(prev_bdev,
3826 inode, prev_index, prev_block,
3827 len, range.flags);
3828 if (ret) {
3829 f2fs_put_dnode(&dn);
3830 goto out;
3831 }
3832
3833 len = 0;
3834 }
3835 }
3836
3837 if (!len) {
3838 prev_bdev = cur_bdev;
3839 prev_index = index;
3840 prev_block = blkaddr;
3841 len = 1;
3842 }
3843 }
3844
3845 f2fs_put_dnode(&dn);
3846
3847 if (fatal_signal_pending(current)) {
3848 ret = -EINTR;
3849 goto out;
3850 }
3851 cond_resched();
3852 }
3853
3854 if (len)
3855 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3856 prev_block, len, range.flags);
3857 out:
3858 filemap_invalidate_unlock(mapping);
3859 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3860 err:
3861 inode_unlock(inode);
3862 file_end_write(filp);
3863
3864 return ret;
3865 }
3866
3867 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3868 {
3869 struct inode *inode = file_inode(filp);
3870 struct f2fs_comp_option option;
3871
3872 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3873 return -EOPNOTSUPP;
3874
3875 inode_lock_shared(inode);
3876
3877 if (!f2fs_compressed_file(inode)) {
3878 inode_unlock_shared(inode);
3879 return -ENODATA;
3880 }
3881
3882 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3883 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3884
3885 inode_unlock_shared(inode);
3886
3887 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3888 sizeof(option)))
3889 return -EFAULT;
3890
3891 return 0;
3892 }
3893
3894 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3895 {
3896 struct inode *inode = file_inode(filp);
3897 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3898 struct f2fs_comp_option option;
3899 int ret = 0;
3900
3901 if (!f2fs_sb_has_compression(sbi))
3902 return -EOPNOTSUPP;
3903
3904 if (!(filp->f_mode & FMODE_WRITE))
3905 return -EBADF;
3906
3907 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3908 sizeof(option)))
3909 return -EFAULT;
3910
3911 if (!f2fs_compressed_file(inode) ||
3912 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3913 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3914 option.algorithm >= COMPRESS_MAX)
3915 return -EINVAL;
3916
3917 file_start_write(filp);
3918 inode_lock(inode);
3919
3920 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3921 ret = -EBUSY;
3922 goto out;
3923 }
3924
3925 if (inode->i_size != 0) {
3926 ret = -EFBIG;
3927 goto out;
3928 }
3929
3930 F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3931 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3932 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3933 f2fs_mark_inode_dirty_sync(inode, true);
3934
3935 if (!f2fs_is_compress_backend_ready(inode))
3936 f2fs_warn(sbi, "compression algorithm is successfully set, "
3937 "but current kernel doesn't support this algorithm.");
3938 out:
3939 inode_unlock(inode);
3940 file_end_write(filp);
3941
3942 return ret;
3943 }
3944
3945 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3946 {
3947 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3948 struct address_space *mapping = inode->i_mapping;
3949 struct page *page;
3950 pgoff_t redirty_idx = page_idx;
3951 int i, page_len = 0, ret = 0;
3952
3953 page_cache_ra_unbounded(&ractl, len, 0);
3954
3955 for (i = 0; i < len; i++, page_idx++) {
3956 page = read_cache_page(mapping, page_idx, NULL, NULL);
3957 if (IS_ERR(page)) {
3958 ret = PTR_ERR(page);
3959 break;
3960 }
3961 page_len++;
3962 }
3963
3964 for (i = 0; i < page_len; i++, redirty_idx++) {
3965 page = find_lock_page(mapping, redirty_idx);
3966 if (!page) {
3967 ret = -ENOMEM;
3968 break;
3969 }
3970 set_page_dirty(page);
3971 f2fs_put_page(page, 1);
3972 f2fs_put_page(page, 0);
3973 }
3974
3975 return ret;
3976 }
3977
3978 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3979 {
3980 struct inode *inode = file_inode(filp);
3981 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3982 struct f2fs_inode_info *fi = F2FS_I(inode);
3983 pgoff_t page_idx = 0, last_idx;
3984 unsigned int blk_per_seg = sbi->blocks_per_seg;
3985 int cluster_size = F2FS_I(inode)->i_cluster_size;
3986 int count, ret;
3987
3988 if (!f2fs_sb_has_compression(sbi) ||
3989 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3990 return -EOPNOTSUPP;
3991
3992 if (!(filp->f_mode & FMODE_WRITE))
3993 return -EBADF;
3994
3995 if (!f2fs_compressed_file(inode))
3996 return -EINVAL;
3997
3998 f2fs_balance_fs(F2FS_I_SB(inode), true);
3999
4000 file_start_write(filp);
4001 inode_lock(inode);
4002
4003 if (!f2fs_is_compress_backend_ready(inode)) {
4004 ret = -EOPNOTSUPP;
4005 goto out;
4006 }
4007
4008 if (f2fs_is_mmap_file(inode)) {
4009 ret = -EBUSY;
4010 goto out;
4011 }
4012
4013 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4014 if (ret)
4015 goto out;
4016
4017 if (!atomic_read(&fi->i_compr_blocks))
4018 goto out;
4019
4020 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4021
4022 count = last_idx - page_idx;
4023 while (count) {
4024 int len = min(cluster_size, count);
4025
4026 ret = redirty_blocks(inode, page_idx, len);
4027 if (ret < 0)
4028 break;
4029
4030 if (get_dirty_pages(inode) >= blk_per_seg)
4031 filemap_fdatawrite(inode->i_mapping);
4032
4033 count -= len;
4034 page_idx += len;
4035 }
4036
4037 if (!ret)
4038 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4039 LLONG_MAX);
4040
4041 if (ret)
4042 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4043 __func__, ret);
4044 out:
4045 inode_unlock(inode);
4046 file_end_write(filp);
4047
4048 return ret;
4049 }
4050
4051 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4052 {
4053 struct inode *inode = file_inode(filp);
4054 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4055 pgoff_t page_idx = 0, last_idx;
4056 unsigned int blk_per_seg = sbi->blocks_per_seg;
4057 int cluster_size = F2FS_I(inode)->i_cluster_size;
4058 int count, ret;
4059
4060 if (!f2fs_sb_has_compression(sbi) ||
4061 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4062 return -EOPNOTSUPP;
4063
4064 if (!(filp->f_mode & FMODE_WRITE))
4065 return -EBADF;
4066
4067 if (!f2fs_compressed_file(inode))
4068 return -EINVAL;
4069
4070 f2fs_balance_fs(F2FS_I_SB(inode), true);
4071
4072 file_start_write(filp);
4073 inode_lock(inode);
4074
4075 if (!f2fs_is_compress_backend_ready(inode)) {
4076 ret = -EOPNOTSUPP;
4077 goto out;
4078 }
4079
4080 if (f2fs_is_mmap_file(inode)) {
4081 ret = -EBUSY;
4082 goto out;
4083 }
4084
4085 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4086 if (ret)
4087 goto out;
4088
4089 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4090
4091 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4092
4093 count = last_idx - page_idx;
4094 while (count) {
4095 int len = min(cluster_size, count);
4096
4097 ret = redirty_blocks(inode, page_idx, len);
4098 if (ret < 0)
4099 break;
4100
4101 if (get_dirty_pages(inode) >= blk_per_seg)
4102 filemap_fdatawrite(inode->i_mapping);
4103
4104 count -= len;
4105 page_idx += len;
4106 }
4107
4108 if (!ret)
4109 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4110 LLONG_MAX);
4111
4112 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4113
4114 if (ret)
4115 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4116 __func__, ret);
4117 out:
4118 inode_unlock(inode);
4119 file_end_write(filp);
4120
4121 return ret;
4122 }
4123
4124 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4125 {
4126 switch (cmd) {
4127 case FS_IOC_GETVERSION:
4128 return f2fs_ioc_getversion(filp, arg);
4129 case F2FS_IOC_START_ATOMIC_WRITE:
4130 return f2fs_ioc_start_atomic_write(filp);
4131 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4132 return f2fs_ioc_commit_atomic_write(filp);
4133 case F2FS_IOC_START_VOLATILE_WRITE:
4134 return f2fs_ioc_start_volatile_write(filp);
4135 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4136 return f2fs_ioc_release_volatile_write(filp);
4137 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4138 return f2fs_ioc_abort_volatile_write(filp);
4139 case F2FS_IOC_SHUTDOWN:
4140 return f2fs_ioc_shutdown(filp, arg);
4141 case FITRIM:
4142 return f2fs_ioc_fitrim(filp, arg);
4143 case FS_IOC_SET_ENCRYPTION_POLICY:
4144 return f2fs_ioc_set_encryption_policy(filp, arg);
4145 case FS_IOC_GET_ENCRYPTION_POLICY:
4146 return f2fs_ioc_get_encryption_policy(filp, arg);
4147 case FS_IOC_GET_ENCRYPTION_PWSALT:
4148 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4149 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4150 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4151 case FS_IOC_ADD_ENCRYPTION_KEY:
4152 return f2fs_ioc_add_encryption_key(filp, arg);
4153 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4154 return f2fs_ioc_remove_encryption_key(filp, arg);
4155 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4156 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4157 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4158 return f2fs_ioc_get_encryption_key_status(filp, arg);
4159 case FS_IOC_GET_ENCRYPTION_NONCE:
4160 return f2fs_ioc_get_encryption_nonce(filp, arg);
4161 case F2FS_IOC_GARBAGE_COLLECT:
4162 return f2fs_ioc_gc(filp, arg);
4163 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4164 return f2fs_ioc_gc_range(filp, arg);
4165 case F2FS_IOC_WRITE_CHECKPOINT:
4166 return f2fs_ioc_write_checkpoint(filp, arg);
4167 case F2FS_IOC_DEFRAGMENT:
4168 return f2fs_ioc_defragment(filp, arg);
4169 case F2FS_IOC_MOVE_RANGE:
4170 return f2fs_ioc_move_range(filp, arg);
4171 case F2FS_IOC_FLUSH_DEVICE:
4172 return f2fs_ioc_flush_device(filp, arg);
4173 case F2FS_IOC_GET_FEATURES:
4174 return f2fs_ioc_get_features(filp, arg);
4175 case F2FS_IOC_GET_PIN_FILE:
4176 return f2fs_ioc_get_pin_file(filp, arg);
4177 case F2FS_IOC_SET_PIN_FILE:
4178 return f2fs_ioc_set_pin_file(filp, arg);
4179 case F2FS_IOC_PRECACHE_EXTENTS:
4180 return f2fs_ioc_precache_extents(filp, arg);
4181 case F2FS_IOC_RESIZE_FS:
4182 return f2fs_ioc_resize_fs(filp, arg);
4183 case FS_IOC_ENABLE_VERITY:
4184 return f2fs_ioc_enable_verity(filp, arg);
4185 case FS_IOC_MEASURE_VERITY:
4186 return f2fs_ioc_measure_verity(filp, arg);
4187 case FS_IOC_READ_VERITY_METADATA:
4188 return f2fs_ioc_read_verity_metadata(filp, arg);
4189 case FS_IOC_GETFSLABEL:
4190 return f2fs_ioc_getfslabel(filp, arg);
4191 case FS_IOC_SETFSLABEL:
4192 return f2fs_ioc_setfslabel(filp, arg);
4193 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4194 return f2fs_get_compress_blocks(filp, arg);
4195 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4196 return f2fs_release_compress_blocks(filp, arg);
4197 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4198 return f2fs_reserve_compress_blocks(filp, arg);
4199 case F2FS_IOC_SEC_TRIM_FILE:
4200 return f2fs_sec_trim_file(filp, arg);
4201 case F2FS_IOC_GET_COMPRESS_OPTION:
4202 return f2fs_ioc_get_compress_option(filp, arg);
4203 case F2FS_IOC_SET_COMPRESS_OPTION:
4204 return f2fs_ioc_set_compress_option(filp, arg);
4205 case F2FS_IOC_DECOMPRESS_FILE:
4206 return f2fs_ioc_decompress_file(filp, arg);
4207 case F2FS_IOC_COMPRESS_FILE:
4208 return f2fs_ioc_compress_file(filp, arg);
4209 default:
4210 return -ENOTTY;
4211 }
4212 }
4213
4214 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4215 {
4216 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4217 return -EIO;
4218 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4219 return -ENOSPC;
4220
4221 return __f2fs_ioctl(filp, cmd, arg);
4222 }
4223
4224 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
4225 {
4226 struct file *file = iocb->ki_filp;
4227 struct inode *inode = file_inode(file);
4228 int ret;
4229
4230 if (!f2fs_is_compress_backend_ready(inode))
4231 return -EOPNOTSUPP;
4232
4233 ret = generic_file_read_iter(iocb, iter);
4234
4235 if (ret > 0)
4236 f2fs_update_iostat(F2FS_I_SB(inode), APP_READ_IO, ret);
4237
4238 return ret;
4239 }
4240
4241 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4242 {
4243 struct file *file = iocb->ki_filp;
4244 struct inode *inode = file_inode(file);
4245 ssize_t ret;
4246
4247 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4248 ret = -EIO;
4249 goto out;
4250 }
4251
4252 if (!f2fs_is_compress_backend_ready(inode)) {
4253 ret = -EOPNOTSUPP;
4254 goto out;
4255 }
4256
4257 if (iocb->ki_flags & IOCB_NOWAIT) {
4258 if (!inode_trylock(inode)) {
4259 ret = -EAGAIN;
4260 goto out;
4261 }
4262 } else {
4263 inode_lock(inode);
4264 }
4265
4266 if (unlikely(IS_IMMUTABLE(inode))) {
4267 ret = -EPERM;
4268 goto unlock;
4269 }
4270
4271 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4272 ret = -EPERM;
4273 goto unlock;
4274 }
4275
4276 ret = generic_write_checks(iocb, from);
4277 if (ret > 0) {
4278 bool preallocated = false;
4279 size_t target_size = 0;
4280 int err;
4281
4282 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
4283 set_inode_flag(inode, FI_NO_PREALLOC);
4284
4285 if ((iocb->ki_flags & IOCB_NOWAIT)) {
4286 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
4287 iov_iter_count(from)) ||
4288 f2fs_has_inline_data(inode) ||
4289 f2fs_force_buffered_io(inode, iocb, from)) {
4290 clear_inode_flag(inode, FI_NO_PREALLOC);
4291 inode_unlock(inode);
4292 ret = -EAGAIN;
4293 goto out;
4294 }
4295 goto write;
4296 }
4297
4298 if (is_inode_flag_set(inode, FI_NO_PREALLOC))
4299 goto write;
4300
4301 if (iocb->ki_flags & IOCB_DIRECT) {
4302 /*
4303 * Convert inline data for Direct I/O before entering
4304 * f2fs_direct_IO().
4305 */
4306 err = f2fs_convert_inline_inode(inode);
4307 if (err)
4308 goto out_err;
4309 /*
4310 * If force_buffere_io() is true, we have to allocate
4311 * blocks all the time, since f2fs_direct_IO will fall
4312 * back to buffered IO.
4313 */
4314 if (!f2fs_force_buffered_io(inode, iocb, from) &&
4315 f2fs_lfs_mode(F2FS_I_SB(inode)))
4316 goto write;
4317 }
4318 preallocated = true;
4319 target_size = iocb->ki_pos + iov_iter_count(from);
4320
4321 err = f2fs_preallocate_blocks(iocb, from);
4322 if (err) {
4323 out_err:
4324 clear_inode_flag(inode, FI_NO_PREALLOC);
4325 inode_unlock(inode);
4326 ret = err;
4327 goto out;
4328 }
4329 write:
4330 ret = __generic_file_write_iter(iocb, from);
4331 clear_inode_flag(inode, FI_NO_PREALLOC);
4332
4333 /* if we couldn't write data, we should deallocate blocks. */
4334 if (preallocated && i_size_read(inode) < target_size) {
4335 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4336 filemap_invalidate_lock(inode->i_mapping);
4337 f2fs_truncate(inode);
4338 filemap_invalidate_unlock(inode->i_mapping);
4339 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4340 }
4341
4342 if (ret > 0)
4343 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
4344 }
4345 unlock:
4346 inode_unlock(inode);
4347 out:
4348 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
4349 iov_iter_count(from), ret);
4350 if (ret > 0)
4351 ret = generic_write_sync(iocb, ret);
4352 return ret;
4353 }
4354
4355 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4356 int advice)
4357 {
4358 struct inode *inode;
4359 struct address_space *mapping;
4360 struct backing_dev_info *bdi;
4361
4362 if (advice == POSIX_FADV_SEQUENTIAL) {
4363 inode = file_inode(filp);
4364 if (S_ISFIFO(inode->i_mode))
4365 return -ESPIPE;
4366
4367 mapping = filp->f_mapping;
4368 if (!mapping || len < 0)
4369 return -EINVAL;
4370
4371 bdi = inode_to_bdi(mapping->host);
4372 filp->f_ra.ra_pages = bdi->ra_pages *
4373 F2FS_I_SB(inode)->seq_file_ra_mul;
4374 spin_lock(&filp->f_lock);
4375 filp->f_mode &= ~FMODE_RANDOM;
4376 spin_unlock(&filp->f_lock);
4377 return 0;
4378 }
4379
4380 return generic_fadvise(filp, offset, len, advice);
4381 }
4382
4383 #ifdef CONFIG_COMPAT
4384 struct compat_f2fs_gc_range {
4385 u32 sync;
4386 compat_u64 start;
4387 compat_u64 len;
4388 };
4389 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
4390 struct compat_f2fs_gc_range)
4391
4392 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4393 {
4394 struct compat_f2fs_gc_range __user *urange;
4395 struct f2fs_gc_range range;
4396 int err;
4397
4398 urange = compat_ptr(arg);
4399 err = get_user(range.sync, &urange->sync);
4400 err |= get_user(range.start, &urange->start);
4401 err |= get_user(range.len, &urange->len);
4402 if (err)
4403 return -EFAULT;
4404
4405 return __f2fs_ioc_gc_range(file, &range);
4406 }
4407
4408 struct compat_f2fs_move_range {
4409 u32 dst_fd;
4410 compat_u64 pos_in;
4411 compat_u64 pos_out;
4412 compat_u64 len;
4413 };
4414 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
4415 struct compat_f2fs_move_range)
4416
4417 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4418 {
4419 struct compat_f2fs_move_range __user *urange;
4420 struct f2fs_move_range range;
4421 int err;
4422
4423 urange = compat_ptr(arg);
4424 err = get_user(range.dst_fd, &urange->dst_fd);
4425 err |= get_user(range.pos_in, &urange->pos_in);
4426 err |= get_user(range.pos_out, &urange->pos_out);
4427 err |= get_user(range.len, &urange->len);
4428 if (err)
4429 return -EFAULT;
4430
4431 return __f2fs_ioc_move_range(file, &range);
4432 }
4433
4434 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4435 {
4436 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4437 return -EIO;
4438 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4439 return -ENOSPC;
4440
4441 switch (cmd) {
4442 case FS_IOC32_GETVERSION:
4443 cmd = FS_IOC_GETVERSION;
4444 break;
4445 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4446 return f2fs_compat_ioc_gc_range(file, arg);
4447 case F2FS_IOC32_MOVE_RANGE:
4448 return f2fs_compat_ioc_move_range(file, arg);
4449 case F2FS_IOC_START_ATOMIC_WRITE:
4450 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4451 case F2FS_IOC_START_VOLATILE_WRITE:
4452 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4453 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4454 case F2FS_IOC_SHUTDOWN:
4455 case FITRIM:
4456 case FS_IOC_SET_ENCRYPTION_POLICY:
4457 case FS_IOC_GET_ENCRYPTION_PWSALT:
4458 case FS_IOC_GET_ENCRYPTION_POLICY:
4459 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4460 case FS_IOC_ADD_ENCRYPTION_KEY:
4461 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4462 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4463 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4464 case FS_IOC_GET_ENCRYPTION_NONCE:
4465 case F2FS_IOC_GARBAGE_COLLECT:
4466 case F2FS_IOC_WRITE_CHECKPOINT:
4467 case F2FS_IOC_DEFRAGMENT:
4468 case F2FS_IOC_FLUSH_DEVICE:
4469 case F2FS_IOC_GET_FEATURES:
4470 case F2FS_IOC_GET_PIN_FILE:
4471 case F2FS_IOC_SET_PIN_FILE:
4472 case F2FS_IOC_PRECACHE_EXTENTS:
4473 case F2FS_IOC_RESIZE_FS:
4474 case FS_IOC_ENABLE_VERITY:
4475 case FS_IOC_MEASURE_VERITY:
4476 case FS_IOC_READ_VERITY_METADATA:
4477 case FS_IOC_GETFSLABEL:
4478 case FS_IOC_SETFSLABEL:
4479 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4480 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4481 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4482 case F2FS_IOC_SEC_TRIM_FILE:
4483 case F2FS_IOC_GET_COMPRESS_OPTION:
4484 case F2FS_IOC_SET_COMPRESS_OPTION:
4485 case F2FS_IOC_DECOMPRESS_FILE:
4486 case F2FS_IOC_COMPRESS_FILE:
4487 break;
4488 default:
4489 return -ENOIOCTLCMD;
4490 }
4491 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4492 }
4493 #endif
4494
4495 const struct file_operations f2fs_file_operations = {
4496 .llseek = f2fs_llseek,
4497 .read_iter = f2fs_file_read_iter,
4498 .write_iter = f2fs_file_write_iter,
4499 .open = f2fs_file_open,
4500 .release = f2fs_release_file,
4501 .mmap = f2fs_file_mmap,
4502 .flush = f2fs_file_flush,
4503 .fsync = f2fs_sync_file,
4504 .fallocate = f2fs_fallocate,
4505 .unlocked_ioctl = f2fs_ioctl,
4506 #ifdef CONFIG_COMPAT
4507 .compat_ioctl = f2fs_compat_ioctl,
4508 #endif
4509 .splice_read = generic_file_splice_read,
4510 .splice_write = iter_file_splice_write,
4511 .fadvise = f2fs_file_fadvise,
4512 };
Ubuntu Jammy Linux kernel mirror