]> git.ipfire.org Git - thirdparty/linux.git/blob - fs/f2fs/file.c
projects / thirdparty / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
orangefs: use attach/detach_page_private
[thirdparty/linux.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
25 #include "f2fs.h"
26 #include "node.h"
27 #include "segment.h"
28 #include "xattr.h"
29 #include "acl.h"
30 #include "gc.h"
31 #include "trace.h"
32 #include <trace/events/f2fs.h>
33
34 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
35 {
36 struct inode *inode = file_inode(vmf->vma->vm_file);
37 vm_fault_t ret;
38
39 down_read(&F2FS_I(inode)->i_mmap_sem);
40 ret = filemap_fault(vmf);
41 up_read(&F2FS_I(inode)->i_mmap_sem);
42
43 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
44
45 return ret;
46 }
47
48 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 {
50 struct page *page = vmf->page;
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 struct dnode_of_data dn;
54 bool need_alloc = true;
55 int err = 0;
56
57 if (unlikely(f2fs_cp_error(sbi))) {
58 err = -EIO;
59 goto err;
60 }
61
62 if (!f2fs_is_checkpoint_ready(sbi)) {
63 err = -ENOSPC;
64 goto err;
65 }
66
67 #ifdef CONFIG_F2FS_FS_COMPRESSION
68 if (f2fs_compressed_file(inode)) {
69 int ret = f2fs_is_compressed_cluster(inode, page->index);
70
71 if (ret < 0) {
72 err = ret;
73 goto err;
74 } else if (ret) {
75 if (ret < F2FS_I(inode)->i_cluster_size) {
76 err = -EAGAIN;
77 goto err;
78 }
79 need_alloc = false;
80 }
81 }
82 #endif
83 /* should do out of any locked page */
84 if (need_alloc)
85 f2fs_balance_fs(sbi, true);
86
87 sb_start_pagefault(inode->i_sb);
88
89 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
90
91 file_update_time(vmf->vma->vm_file);
92 down_read(&F2FS_I(inode)->i_mmap_sem);
93 lock_page(page);
94 if (unlikely(page->mapping != inode->i_mapping ||
95 page_offset(page) > i_size_read(inode) ||
96 !PageUptodate(page))) {
97 unlock_page(page);
98 err = -EFAULT;
99 goto out_sem;
100 }
101
102 if (need_alloc) {
103 /* block allocation */
104 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
105 set_new_dnode(&dn, inode, NULL, NULL, 0);
106 err = f2fs_get_block(&dn, page->index);
107 f2fs_put_dnode(&dn);
108 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
109 }
110
111 #ifdef CONFIG_F2FS_FS_COMPRESSION
112 if (!need_alloc) {
113 set_new_dnode(&dn, inode, NULL, NULL, 0);
114 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
115 f2fs_put_dnode(&dn);
116 }
117 #endif
118 if (err) {
119 unlock_page(page);
120 goto out_sem;
121 }
122
123 f2fs_wait_on_page_writeback(page, DATA, false, true);
124
125 /* wait for GCed page writeback via META_MAPPING */
126 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
127
128 /*
129 * check to see if the page is mapped already (no holes)
130 */
131 if (PageMappedToDisk(page))
132 goto out_sem;
133
134 /* page is wholly or partially inside EOF */
135 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
136 i_size_read(inode)) {
137 loff_t offset;
138
139 offset = i_size_read(inode) & ~PAGE_MASK;
140 zero_user_segment(page, offset, PAGE_SIZE);
141 }
142 set_page_dirty(page);
143 if (!PageUptodate(page))
144 SetPageUptodate(page);
145
146 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
147 f2fs_update_time(sbi, REQ_TIME);
148
149 trace_f2fs_vm_page_mkwrite(page, DATA);
150 out_sem:
151 up_read(&F2FS_I(inode)->i_mmap_sem);
152
153 sb_end_pagefault(inode->i_sb);
154 err:
155 return block_page_mkwrite_return(err);
156 }
157
158 static const struct vm_operations_struct f2fs_file_vm_ops = {
159 .fault = f2fs_filemap_fault,
160 .map_pages = filemap_map_pages,
161 .page_mkwrite = f2fs_vm_page_mkwrite,
162 };
163
164 static int get_parent_ino(struct inode *inode, nid_t *pino)
165 {
166 struct dentry *dentry;
167
168 inode = igrab(inode);
169 dentry = d_find_any_alias(inode);
170 iput(inode);
171 if (!dentry)
172 return 0;
173
174 *pino = parent_ino(dentry);
175 dput(dentry);
176 return 1;
177 }
178
179 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
180 {
181 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
182 enum cp_reason_type cp_reason = CP_NO_NEEDED;
183
184 if (!S_ISREG(inode->i_mode))
185 cp_reason = CP_NON_REGULAR;
186 else if (f2fs_compressed_file(inode))
187 cp_reason = CP_COMPRESSED;
188 else if (inode->i_nlink != 1)
189 cp_reason = CP_HARDLINK;
190 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
191 cp_reason = CP_SB_NEED_CP;
192 else if (file_wrong_pino(inode))
193 cp_reason = CP_WRONG_PINO;
194 else if (!f2fs_space_for_roll_forward(sbi))
195 cp_reason = CP_NO_SPC_ROLL;
196 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
197 cp_reason = CP_NODE_NEED_CP;
198 else if (test_opt(sbi, FASTBOOT))
199 cp_reason = CP_FASTBOOT_MODE;
200 else if (F2FS_OPTION(sbi).active_logs == 2)
201 cp_reason = CP_SPEC_LOG_NUM;
202 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
203 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
204 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
205 TRANS_DIR_INO))
206 cp_reason = CP_RECOVER_DIR;
207
208 return cp_reason;
209 }
210
211 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
212 {
213 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
214 bool ret = false;
215 /* But we need to avoid that there are some inode updates */
216 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
217 ret = true;
218 f2fs_put_page(i, 0);
219 return ret;
220 }
221
222 static void try_to_fix_pino(struct inode *inode)
223 {
224 struct f2fs_inode_info *fi = F2FS_I(inode);
225 nid_t pino;
226
227 down_write(&fi->i_sem);
228 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
229 get_parent_ino(inode, &pino)) {
230 f2fs_i_pino_write(inode, pino);
231 file_got_pino(inode);
232 }
233 up_write(&fi->i_sem);
234 }
235
236 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
237 int datasync, bool atomic)
238 {
239 struct inode *inode = file->f_mapping->host;
240 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
241 nid_t ino = inode->i_ino;
242 int ret = 0;
243 enum cp_reason_type cp_reason = 0;
244 struct writeback_control wbc = {
245 .sync_mode = WB_SYNC_ALL,
246 .nr_to_write = LONG_MAX,
247 .for_reclaim = 0,
248 };
249 unsigned int seq_id = 0;
250
251 if (unlikely(f2fs_readonly(inode->i_sb) ||
252 is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
253 return 0;
254
255 trace_f2fs_sync_file_enter(inode);
256
257 if (S_ISDIR(inode->i_mode))
258 goto go_write;
259
260 /* if fdatasync is triggered, let's do in-place-update */
261 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
262 set_inode_flag(inode, FI_NEED_IPU);
263 ret = file_write_and_wait_range(file, start, end);
264 clear_inode_flag(inode, FI_NEED_IPU);
265
266 if (ret) {
267 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
268 return ret;
269 }
270
271 /* if the inode is dirty, let's recover all the time */
272 if (!f2fs_skip_inode_update(inode, datasync)) {
273 f2fs_write_inode(inode, NULL);
274 goto go_write;
275 }
276
277 /*
278 * if there is no written data, don't waste time to write recovery info.
279 */
280 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
281 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
282
283 /* it may call write_inode just prior to fsync */
284 if (need_inode_page_update(sbi, ino))
285 goto go_write;
286
287 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
288 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
289 goto flush_out;
290 goto out;
291 }
292 go_write:
293 /*
294 * Both of fdatasync() and fsync() are able to be recovered from
295 * sudden-power-off.
296 */
297 down_read(&F2FS_I(inode)->i_sem);
298 cp_reason = need_do_checkpoint(inode);
299 up_read(&F2FS_I(inode)->i_sem);
300
301 if (cp_reason) {
302 /* all the dirty node pages should be flushed for POR */
303 ret = f2fs_sync_fs(inode->i_sb, 1);
304
305 /*
306 * We've secured consistency through sync_fs. Following pino
307 * will be used only for fsynced inodes after checkpoint.
308 */
309 try_to_fix_pino(inode);
310 clear_inode_flag(inode, FI_APPEND_WRITE);
311 clear_inode_flag(inode, FI_UPDATE_WRITE);
312 goto out;
313 }
314 sync_nodes:
315 atomic_inc(&sbi->wb_sync_req[NODE]);
316 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
317 atomic_dec(&sbi->wb_sync_req[NODE]);
318 if (ret)
319 goto out;
320
321 /* if cp_error was enabled, we should avoid infinite loop */
322 if (unlikely(f2fs_cp_error(sbi))) {
323 ret = -EIO;
324 goto out;
325 }
326
327 if (f2fs_need_inode_block_update(sbi, ino)) {
328 f2fs_mark_inode_dirty_sync(inode, true);
329 f2fs_write_inode(inode, NULL);
330 goto sync_nodes;
331 }
332
333 /*
334 * If it's atomic_write, it's just fine to keep write ordering. So
335 * here we don't need to wait for node write completion, since we use
336 * node chain which serializes node blocks. If one of node writes are
337 * reordered, we can see simply broken chain, resulting in stopping
338 * roll-forward recovery. It means we'll recover all or none node blocks
339 * given fsync mark.
340 */
341 if (!atomic) {
342 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
343 if (ret)
344 goto out;
345 }
346
347 /* once recovery info is written, don't need to tack this */
348 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
349 clear_inode_flag(inode, FI_APPEND_WRITE);
350 flush_out:
351 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
352 ret = f2fs_issue_flush(sbi, inode->i_ino);
353 if (!ret) {
354 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
355 clear_inode_flag(inode, FI_UPDATE_WRITE);
356 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
357 }
358 f2fs_update_time(sbi, REQ_TIME);
359 out:
360 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
361 f2fs_trace_ios(NULL, 1);
362 return ret;
363 }
364
365 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
366 {
367 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
368 return -EIO;
369 return f2fs_do_sync_file(file, start, end, datasync, false);
370 }
371
372 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
373 pgoff_t pgofs, int whence)
374 {
375 struct page *page;
376 int nr_pages;
377
378 if (whence != SEEK_DATA)
379 return 0;
380
381 /* find first dirty page index */
382 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
383 1, &page);
384 if (!nr_pages)
385 return ULONG_MAX;
386 pgofs = page->index;
387 put_page(page);
388 return pgofs;
389 }
390
391 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
392 pgoff_t dirty, pgoff_t pgofs, int whence)
393 {
394 switch (whence) {
395 case SEEK_DATA:
396 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
397 __is_valid_data_blkaddr(blkaddr))
398 return true;
399 break;
400 case SEEK_HOLE:
401 if (blkaddr == NULL_ADDR)
402 return true;
403 break;
404 }
405 return false;
406 }
407
408 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
409 {
410 struct inode *inode = file->f_mapping->host;
411 loff_t maxbytes = inode->i_sb->s_maxbytes;
412 struct dnode_of_data dn;
413 pgoff_t pgofs, end_offset, dirty;
414 loff_t data_ofs = offset;
415 loff_t isize;
416 int err = 0;
417
418 inode_lock(inode);
419
420 isize = i_size_read(inode);
421 if (offset >= isize)
422 goto fail;
423
424 /* handle inline data case */
425 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
426 if (whence == SEEK_HOLE)
427 data_ofs = isize;
428 goto found;
429 }
430
431 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
432
433 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
434
435 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
436 set_new_dnode(&dn, inode, NULL, NULL, 0);
437 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
438 if (err && err != -ENOENT) {
439 goto fail;
440 } else if (err == -ENOENT) {
441 /* direct node does not exists */
442 if (whence == SEEK_DATA) {
443 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
444 continue;
445 } else {
446 goto found;
447 }
448 }
449
450 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
451
452 /* find data/hole in dnode block */
453 for (; dn.ofs_in_node < end_offset;
454 dn.ofs_in_node++, pgofs++,
455 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
456 block_t blkaddr;
457
458 blkaddr = f2fs_data_blkaddr(&dn);
459
460 if (__is_valid_data_blkaddr(blkaddr) &&
461 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
462 blkaddr, DATA_GENERIC_ENHANCE)) {
463 f2fs_put_dnode(&dn);
464 goto fail;
465 }
466
467 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
468 pgofs, whence)) {
469 f2fs_put_dnode(&dn);
470 goto found;
471 }
472 }
473 f2fs_put_dnode(&dn);
474 }
475
476 if (whence == SEEK_DATA)
477 goto fail;
478 found:
479 if (whence == SEEK_HOLE && data_ofs > isize)
480 data_ofs = isize;
481 inode_unlock(inode);
482 return vfs_setpos(file, data_ofs, maxbytes);
483 fail:
484 inode_unlock(inode);
485 return -ENXIO;
486 }
487
488 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
489 {
490 struct inode *inode = file->f_mapping->host;
491 loff_t maxbytes = inode->i_sb->s_maxbytes;
492
493 switch (whence) {
494 case SEEK_SET:
495 case SEEK_CUR:
496 case SEEK_END:
497 return generic_file_llseek_size(file, offset, whence,
498 maxbytes, i_size_read(inode));
499 case SEEK_DATA:
500 case SEEK_HOLE:
501 if (offset < 0)
502 return -ENXIO;
503 return f2fs_seek_block(file, offset, whence);
504 }
505
506 return -EINVAL;
507 }
508
509 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
510 {
511 struct inode *inode = file_inode(file);
512 int err;
513
514 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
515 return -EIO;
516
517 if (!f2fs_is_compress_backend_ready(inode))
518 return -EOPNOTSUPP;
519
520 /* we don't need to use inline_data strictly */
521 err = f2fs_convert_inline_inode(inode);
522 if (err)
523 return err;
524
525 file_accessed(file);
526 vma->vm_ops = &f2fs_file_vm_ops;
527 set_inode_flag(inode, FI_MMAP_FILE);
528 return 0;
529 }
530
531 static int f2fs_file_open(struct inode *inode, struct file *filp)
532 {
533 int err = fscrypt_file_open(inode, filp);
534
535 if (err)
536 return err;
537
538 if (!f2fs_is_compress_backend_ready(inode))
539 return -EOPNOTSUPP;
540
541 err = fsverity_file_open(inode, filp);
542 if (err)
543 return err;
544
545 filp->f_mode |= FMODE_NOWAIT;
546
547 return dquot_file_open(inode, filp);
548 }
549
550 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
551 {
552 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
553 struct f2fs_node *raw_node;
554 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
555 __le32 *addr;
556 int base = 0;
557 bool compressed_cluster = false;
558 int cluster_index = 0, valid_blocks = 0;
559 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
560
561 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
562 base = get_extra_isize(dn->inode);
563
564 raw_node = F2FS_NODE(dn->node_page);
565 addr = blkaddr_in_node(raw_node) + base + ofs;
566
567 /* Assumption: truncateion starts with cluster */
568 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
569 block_t blkaddr = le32_to_cpu(*addr);
570
571 if (f2fs_compressed_file(dn->inode) &&
572 !(cluster_index & (cluster_size - 1))) {
573 if (compressed_cluster)
574 f2fs_i_compr_blocks_update(dn->inode,
575 valid_blocks, false);
576 compressed_cluster = (blkaddr == COMPRESS_ADDR);
577 valid_blocks = 0;
578 }
579
580 if (blkaddr == NULL_ADDR)
581 continue;
582
583 dn->data_blkaddr = NULL_ADDR;
584 f2fs_set_data_blkaddr(dn);
585
586 if (__is_valid_data_blkaddr(blkaddr)) {
587 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
588 DATA_GENERIC_ENHANCE))
589 continue;
590 if (compressed_cluster)
591 valid_blocks++;
592 }
593
594 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
595 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
596
597 f2fs_invalidate_blocks(sbi, blkaddr);
598 nr_free++;
599 }
600
601 if (compressed_cluster)
602 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
603
604 if (nr_free) {
605 pgoff_t fofs;
606 /*
607 * once we invalidate valid blkaddr in range [ofs, ofs + count],
608 * we will invalidate all blkaddr in the whole range.
609 */
610 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
611 dn->inode) + ofs;
612 f2fs_update_extent_cache_range(dn, fofs, 0, len);
613 dec_valid_block_count(sbi, dn->inode, nr_free);
614 }
615 dn->ofs_in_node = ofs;
616
617 f2fs_update_time(sbi, REQ_TIME);
618 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
619 dn->ofs_in_node, nr_free);
620 }
621
622 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
623 {
624 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
625 }
626
627 static int truncate_partial_data_page(struct inode *inode, u64 from,
628 bool cache_only)
629 {
630 loff_t offset = from & (PAGE_SIZE - 1);
631 pgoff_t index = from >> PAGE_SHIFT;
632 struct address_space *mapping = inode->i_mapping;
633 struct page *page;
634
635 if (!offset && !cache_only)
636 return 0;
637
638 if (cache_only) {
639 page = find_lock_page(mapping, index);
640 if (page && PageUptodate(page))
641 goto truncate_out;
642 f2fs_put_page(page, 1);
643 return 0;
644 }
645
646 if (f2fs_compressed_file(inode))
647 return 0;
648
649 page = f2fs_get_lock_data_page(inode, index, true);
650 if (IS_ERR(page))
651 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
652 truncate_out:
653 f2fs_wait_on_page_writeback(page, DATA, true, true);
654 zero_user(page, offset, PAGE_SIZE - offset);
655
656 /* An encrypted inode should have a key and truncate the last page. */
657 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
658 if (!cache_only)
659 set_page_dirty(page);
660 f2fs_put_page(page, 1);
661 return 0;
662 }
663
664 static int do_truncate_blocks(struct inode *inode, u64 from, bool lock)
665 {
666 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
667 struct dnode_of_data dn;
668 pgoff_t free_from;
669 int count = 0, err = 0;
670 struct page *ipage;
671 bool truncate_page = false;
672
673 trace_f2fs_truncate_blocks_enter(inode, from);
674
675 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
676
677 if (free_from >= sbi->max_file_blocks)
678 goto free_partial;
679
680 if (lock)
681 f2fs_lock_op(sbi);
682
683 ipage = f2fs_get_node_page(sbi, inode->i_ino);
684 if (IS_ERR(ipage)) {
685 err = PTR_ERR(ipage);
686 goto out;
687 }
688
689 if (f2fs_has_inline_data(inode)) {
690 f2fs_truncate_inline_inode(inode, ipage, from);
691 f2fs_put_page(ipage, 1);
692 truncate_page = true;
693 goto out;
694 }
695
696 set_new_dnode(&dn, inode, ipage, NULL, 0);
697 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
698 if (err) {
699 if (err == -ENOENT)
700 goto free_next;
701 goto out;
702 }
703
704 count = ADDRS_PER_PAGE(dn.node_page, inode);
705
706 count -= dn.ofs_in_node;
707 f2fs_bug_on(sbi, count < 0);
708
709 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
710 f2fs_truncate_data_blocks_range(&dn, count);
711 free_from += count;
712 }
713
714 f2fs_put_dnode(&dn);
715 free_next:
716 err = f2fs_truncate_inode_blocks(inode, free_from);
717 out:
718 if (lock)
719 f2fs_unlock_op(sbi);
720 free_partial:
721 /* lastly zero out the first data page */
722 if (!err)
723 err = truncate_partial_data_page(inode, from, truncate_page);
724
725 trace_f2fs_truncate_blocks_exit(inode, err);
726 return err;
727 }
728
729 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
730 {
731 u64 free_from = from;
732
733 /*
734 * for compressed file, only support cluster size
735 * aligned truncation.
736 */
737 if (f2fs_compressed_file(inode)) {
738 size_t cluster_shift = PAGE_SHIFT +
739 F2FS_I(inode)->i_log_cluster_size;
740 size_t cluster_mask = (1 << cluster_shift) - 1;
741
742 free_from = from >> cluster_shift;
743 if (from & cluster_mask)
744 free_from++;
745 free_from <<= cluster_shift;
746 }
747
748 return do_truncate_blocks(inode, free_from, lock);
749 }
750
751 int f2fs_truncate(struct inode *inode)
752 {
753 int err;
754
755 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
756 return -EIO;
757
758 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
759 S_ISLNK(inode->i_mode)))
760 return 0;
761
762 trace_f2fs_truncate(inode);
763
764 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
765 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
766 return -EIO;
767 }
768
769 /* we should check inline_data size */
770 if (!f2fs_may_inline_data(inode)) {
771 err = f2fs_convert_inline_inode(inode);
772 if (err)
773 return err;
774 }
775
776 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
777 if (err)
778 return err;
779
780 inode->i_mtime = inode->i_ctime = current_time(inode);
781 f2fs_mark_inode_dirty_sync(inode, false);
782 return 0;
783 }
784
785 int f2fs_getattr(const struct path *path, struct kstat *stat,
786 u32 request_mask, unsigned int query_flags)
787 {
788 struct inode *inode = d_inode(path->dentry);
789 struct f2fs_inode_info *fi = F2FS_I(inode);
790 struct f2fs_inode *ri;
791 unsigned int flags;
792
793 if (f2fs_has_extra_attr(inode) &&
794 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
795 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
796 stat->result_mask |= STATX_BTIME;
797 stat->btime.tv_sec = fi->i_crtime.tv_sec;
798 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
799 }
800
801 flags = fi->i_flags;
802 if (flags & F2FS_COMPR_FL)
803 stat->attributes |= STATX_ATTR_COMPRESSED;
804 if (flags & F2FS_APPEND_FL)
805 stat->attributes |= STATX_ATTR_APPEND;
806 if (IS_ENCRYPTED(inode))
807 stat->attributes |= STATX_ATTR_ENCRYPTED;
808 if (flags & F2FS_IMMUTABLE_FL)
809 stat->attributes |= STATX_ATTR_IMMUTABLE;
810 if (flags & F2FS_NODUMP_FL)
811 stat->attributes |= STATX_ATTR_NODUMP;
812 if (IS_VERITY(inode))
813 stat->attributes |= STATX_ATTR_VERITY;
814
815 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
816 STATX_ATTR_APPEND |
817 STATX_ATTR_ENCRYPTED |
818 STATX_ATTR_IMMUTABLE |
819 STATX_ATTR_NODUMP |
820 STATX_ATTR_VERITY);
821
822 generic_fillattr(inode, stat);
823
824 /* we need to show initial sectors used for inline_data/dentries */
825 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
826 f2fs_has_inline_dentry(inode))
827 stat->blocks += (stat->size + 511) >> 9;
828
829 return 0;
830 }
831
832 #ifdef CONFIG_F2FS_FS_POSIX_ACL
833 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
834 {
835 unsigned int ia_valid = attr->ia_valid;
836
837 if (ia_valid & ATTR_UID)
838 inode->i_uid = attr->ia_uid;
839 if (ia_valid & ATTR_GID)
840 inode->i_gid = attr->ia_gid;
841 if (ia_valid & ATTR_ATIME)
842 inode->i_atime = attr->ia_atime;
843 if (ia_valid & ATTR_MTIME)
844 inode->i_mtime = attr->ia_mtime;
845 if (ia_valid & ATTR_CTIME)
846 inode->i_ctime = attr->ia_ctime;
847 if (ia_valid & ATTR_MODE) {
848 umode_t mode = attr->ia_mode;
849
850 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
851 mode &= ~S_ISGID;
852 set_acl_inode(inode, mode);
853 }
854 }
855 #else
856 #define __setattr_copy setattr_copy
857 #endif
858
859 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
860 {
861 struct inode *inode = d_inode(dentry);
862 int err;
863
864 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
865 return -EIO;
866
867 if ((attr->ia_valid & ATTR_SIZE) &&
868 !f2fs_is_compress_backend_ready(inode))
869 return -EOPNOTSUPP;
870
871 err = setattr_prepare(dentry, attr);
872 if (err)
873 return err;
874
875 err = fscrypt_prepare_setattr(dentry, attr);
876 if (err)
877 return err;
878
879 err = fsverity_prepare_setattr(dentry, attr);
880 if (err)
881 return err;
882
883 if (is_quota_modification(inode, attr)) {
884 err = dquot_initialize(inode);
885 if (err)
886 return err;
887 }
888 if ((attr->ia_valid & ATTR_UID &&
889 !uid_eq(attr->ia_uid, inode->i_uid)) ||
890 (attr->ia_valid & ATTR_GID &&
891 !gid_eq(attr->ia_gid, inode->i_gid))) {
892 f2fs_lock_op(F2FS_I_SB(inode));
893 err = dquot_transfer(inode, attr);
894 if (err) {
895 set_sbi_flag(F2FS_I_SB(inode),
896 SBI_QUOTA_NEED_REPAIR);
897 f2fs_unlock_op(F2FS_I_SB(inode));
898 return err;
899 }
900 /*
901 * update uid/gid under lock_op(), so that dquot and inode can
902 * be updated atomically.
903 */
904 if (attr->ia_valid & ATTR_UID)
905 inode->i_uid = attr->ia_uid;
906 if (attr->ia_valid & ATTR_GID)
907 inode->i_gid = attr->ia_gid;
908 f2fs_mark_inode_dirty_sync(inode, true);
909 f2fs_unlock_op(F2FS_I_SB(inode));
910 }
911
912 if (attr->ia_valid & ATTR_SIZE) {
913 loff_t old_size = i_size_read(inode);
914
915 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
916 /*
917 * should convert inline inode before i_size_write to
918 * keep smaller than inline_data size with inline flag.
919 */
920 err = f2fs_convert_inline_inode(inode);
921 if (err)
922 return err;
923 }
924
925 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
926 down_write(&F2FS_I(inode)->i_mmap_sem);
927
928 truncate_setsize(inode, attr->ia_size);
929
930 if (attr->ia_size <= old_size)
931 err = f2fs_truncate(inode);
932 /*
933 * do not trim all blocks after i_size if target size is
934 * larger than i_size.
935 */
936 up_write(&F2FS_I(inode)->i_mmap_sem);
937 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
938 if (err)
939 return err;
940
941 spin_lock(&F2FS_I(inode)->i_size_lock);
942 inode->i_mtime = inode->i_ctime = current_time(inode);
943 F2FS_I(inode)->last_disk_size = i_size_read(inode);
944 spin_unlock(&F2FS_I(inode)->i_size_lock);
945 }
946
947 __setattr_copy(inode, attr);
948
949 if (attr->ia_valid & ATTR_MODE) {
950 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
951 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
952 inode->i_mode = F2FS_I(inode)->i_acl_mode;
953 clear_inode_flag(inode, FI_ACL_MODE);
954 }
955 }
956
957 /* file size may changed here */
958 f2fs_mark_inode_dirty_sync(inode, true);
959
960 /* inode change will produce dirty node pages flushed by checkpoint */
961 f2fs_balance_fs(F2FS_I_SB(inode), true);
962
963 return err;
964 }
965
966 const struct inode_operations f2fs_file_inode_operations = {
967 .getattr = f2fs_getattr,
968 .setattr = f2fs_setattr,
969 .get_acl = f2fs_get_acl,
970 .set_acl = f2fs_set_acl,
971 #ifdef CONFIG_F2FS_FS_XATTR
972 .listxattr = f2fs_listxattr,
973 #endif
974 .fiemap = f2fs_fiemap,
975 };
976
977 static int fill_zero(struct inode *inode, pgoff_t index,
978 loff_t start, loff_t len)
979 {
980 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
981 struct page *page;
982
983 if (!len)
984 return 0;
985
986 f2fs_balance_fs(sbi, true);
987
988 f2fs_lock_op(sbi);
989 page = f2fs_get_new_data_page(inode, NULL, index, false);
990 f2fs_unlock_op(sbi);
991
992 if (IS_ERR(page))
993 return PTR_ERR(page);
994
995 f2fs_wait_on_page_writeback(page, DATA, true, true);
996 zero_user(page, start, len);
997 set_page_dirty(page);
998 f2fs_put_page(page, 1);
999 return 0;
1000 }
1001
1002 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1003 {
1004 int err;
1005
1006 while (pg_start < pg_end) {
1007 struct dnode_of_data dn;
1008 pgoff_t end_offset, count;
1009
1010 set_new_dnode(&dn, inode, NULL, NULL, 0);
1011 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1012 if (err) {
1013 if (err == -ENOENT) {
1014 pg_start = f2fs_get_next_page_offset(&dn,
1015 pg_start);
1016 continue;
1017 }
1018 return err;
1019 }
1020
1021 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1022 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1023
1024 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1025
1026 f2fs_truncate_data_blocks_range(&dn, count);
1027 f2fs_put_dnode(&dn);
1028
1029 pg_start += count;
1030 }
1031 return 0;
1032 }
1033
1034 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1035 {
1036 pgoff_t pg_start, pg_end;
1037 loff_t off_start, off_end;
1038 int ret;
1039
1040 ret = f2fs_convert_inline_inode(inode);
1041 if (ret)
1042 return ret;
1043
1044 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1045 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1046
1047 off_start = offset & (PAGE_SIZE - 1);
1048 off_end = (offset + len) & (PAGE_SIZE - 1);
1049
1050 if (pg_start == pg_end) {
1051 ret = fill_zero(inode, pg_start, off_start,
1052 off_end - off_start);
1053 if (ret)
1054 return ret;
1055 } else {
1056 if (off_start) {
1057 ret = fill_zero(inode, pg_start++, off_start,
1058 PAGE_SIZE - off_start);
1059 if (ret)
1060 return ret;
1061 }
1062 if (off_end) {
1063 ret = fill_zero(inode, pg_end, 0, off_end);
1064 if (ret)
1065 return ret;
1066 }
1067
1068 if (pg_start < pg_end) {
1069 struct address_space *mapping = inode->i_mapping;
1070 loff_t blk_start, blk_end;
1071 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1072
1073 f2fs_balance_fs(sbi, true);
1074
1075 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1076 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1077
1078 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1079 down_write(&F2FS_I(inode)->i_mmap_sem);
1080
1081 truncate_inode_pages_range(mapping, blk_start,
1082 blk_end - 1);
1083
1084 f2fs_lock_op(sbi);
1085 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1086 f2fs_unlock_op(sbi);
1087
1088 up_write(&F2FS_I(inode)->i_mmap_sem);
1089 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1090 }
1091 }
1092
1093 return ret;
1094 }
1095
1096 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1097 int *do_replace, pgoff_t off, pgoff_t len)
1098 {
1099 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1100 struct dnode_of_data dn;
1101 int ret, done, i;
1102
1103 next_dnode:
1104 set_new_dnode(&dn, inode, NULL, NULL, 0);
1105 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1106 if (ret && ret != -ENOENT) {
1107 return ret;
1108 } else if (ret == -ENOENT) {
1109 if (dn.max_level == 0)
1110 return -ENOENT;
1111 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1112 dn.ofs_in_node, len);
1113 blkaddr += done;
1114 do_replace += done;
1115 goto next;
1116 }
1117
1118 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1119 dn.ofs_in_node, len);
1120 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1121 *blkaddr = f2fs_data_blkaddr(&dn);
1122
1123 if (__is_valid_data_blkaddr(*blkaddr) &&
1124 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1125 DATA_GENERIC_ENHANCE)) {
1126 f2fs_put_dnode(&dn);
1127 return -EFSCORRUPTED;
1128 }
1129
1130 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1131
1132 if (f2fs_lfs_mode(sbi)) {
1133 f2fs_put_dnode(&dn);
1134 return -EOPNOTSUPP;
1135 }
1136
1137 /* do not invalidate this block address */
1138 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1139 *do_replace = 1;
1140 }
1141 }
1142 f2fs_put_dnode(&dn);
1143 next:
1144 len -= done;
1145 off += done;
1146 if (len)
1147 goto next_dnode;
1148 return 0;
1149 }
1150
1151 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1152 int *do_replace, pgoff_t off, int len)
1153 {
1154 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1155 struct dnode_of_data dn;
1156 int ret, i;
1157
1158 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1159 if (*do_replace == 0)
1160 continue;
1161
1162 set_new_dnode(&dn, inode, NULL, NULL, 0);
1163 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1164 if (ret) {
1165 dec_valid_block_count(sbi, inode, 1);
1166 f2fs_invalidate_blocks(sbi, *blkaddr);
1167 } else {
1168 f2fs_update_data_blkaddr(&dn, *blkaddr);
1169 }
1170 f2fs_put_dnode(&dn);
1171 }
1172 return 0;
1173 }
1174
1175 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1176 block_t *blkaddr, int *do_replace,
1177 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1178 {
1179 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1180 pgoff_t i = 0;
1181 int ret;
1182
1183 while (i < len) {
1184 if (blkaddr[i] == NULL_ADDR && !full) {
1185 i++;
1186 continue;
1187 }
1188
1189 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1190 struct dnode_of_data dn;
1191 struct node_info ni;
1192 size_t new_size;
1193 pgoff_t ilen;
1194
1195 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1196 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1197 if (ret)
1198 return ret;
1199
1200 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1201 if (ret) {
1202 f2fs_put_dnode(&dn);
1203 return ret;
1204 }
1205
1206 ilen = min((pgoff_t)
1207 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1208 dn.ofs_in_node, len - i);
1209 do {
1210 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1211 f2fs_truncate_data_blocks_range(&dn, 1);
1212
1213 if (do_replace[i]) {
1214 f2fs_i_blocks_write(src_inode,
1215 1, false, false);
1216 f2fs_i_blocks_write(dst_inode,
1217 1, true, false);
1218 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1219 blkaddr[i], ni.version, true, false);
1220
1221 do_replace[i] = 0;
1222 }
1223 dn.ofs_in_node++;
1224 i++;
1225 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1226 if (dst_inode->i_size < new_size)
1227 f2fs_i_size_write(dst_inode, new_size);
1228 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1229
1230 f2fs_put_dnode(&dn);
1231 } else {
1232 struct page *psrc, *pdst;
1233
1234 psrc = f2fs_get_lock_data_page(src_inode,
1235 src + i, true);
1236 if (IS_ERR(psrc))
1237 return PTR_ERR(psrc);
1238 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1239 true);
1240 if (IS_ERR(pdst)) {
1241 f2fs_put_page(psrc, 1);
1242 return PTR_ERR(pdst);
1243 }
1244 f2fs_copy_page(psrc, pdst);
1245 set_page_dirty(pdst);
1246 f2fs_put_page(pdst, 1);
1247 f2fs_put_page(psrc, 1);
1248
1249 ret = f2fs_truncate_hole(src_inode,
1250 src + i, src + i + 1);
1251 if (ret)
1252 return ret;
1253 i++;
1254 }
1255 }
1256 return 0;
1257 }
1258
1259 static int __exchange_data_block(struct inode *src_inode,
1260 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1261 pgoff_t len, bool full)
1262 {
1263 block_t *src_blkaddr;
1264 int *do_replace;
1265 pgoff_t olen;
1266 int ret;
1267
1268 while (len) {
1269 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1270
1271 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1272 array_size(olen, sizeof(block_t)),
1273 GFP_NOFS);
1274 if (!src_blkaddr)
1275 return -ENOMEM;
1276
1277 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1278 array_size(olen, sizeof(int)),
1279 GFP_NOFS);
1280 if (!do_replace) {
1281 kvfree(src_blkaddr);
1282 return -ENOMEM;
1283 }
1284
1285 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1286 do_replace, src, olen);
1287 if (ret)
1288 goto roll_back;
1289
1290 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1291 do_replace, src, dst, olen, full);
1292 if (ret)
1293 goto roll_back;
1294
1295 src += olen;
1296 dst += olen;
1297 len -= olen;
1298
1299 kvfree(src_blkaddr);
1300 kvfree(do_replace);
1301 }
1302 return 0;
1303
1304 roll_back:
1305 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1306 kvfree(src_blkaddr);
1307 kvfree(do_replace);
1308 return ret;
1309 }
1310
1311 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1312 {
1313 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1314 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1315 pgoff_t start = offset >> PAGE_SHIFT;
1316 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1317 int ret;
1318
1319 f2fs_balance_fs(sbi, true);
1320
1321 /* avoid gc operation during block exchange */
1322 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1323 down_write(&F2FS_I(inode)->i_mmap_sem);
1324
1325 f2fs_lock_op(sbi);
1326 f2fs_drop_extent_tree(inode);
1327 truncate_pagecache(inode, offset);
1328 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1329 f2fs_unlock_op(sbi);
1330
1331 up_write(&F2FS_I(inode)->i_mmap_sem);
1332 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1333 return ret;
1334 }
1335
1336 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1337 {
1338 loff_t new_size;
1339 int ret;
1340
1341 if (offset + len >= i_size_read(inode))
1342 return -EINVAL;
1343
1344 /* collapse range should be aligned to block size of f2fs. */
1345 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1346 return -EINVAL;
1347
1348 ret = f2fs_convert_inline_inode(inode);
1349 if (ret)
1350 return ret;
1351
1352 /* write out all dirty pages from offset */
1353 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1354 if (ret)
1355 return ret;
1356
1357 ret = f2fs_do_collapse(inode, offset, len);
1358 if (ret)
1359 return ret;
1360
1361 /* write out all moved pages, if possible */
1362 down_write(&F2FS_I(inode)->i_mmap_sem);
1363 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1364 truncate_pagecache(inode, offset);
1365
1366 new_size = i_size_read(inode) - len;
1367 truncate_pagecache(inode, new_size);
1368
1369 ret = f2fs_truncate_blocks(inode, new_size, true);
1370 up_write(&F2FS_I(inode)->i_mmap_sem);
1371 if (!ret)
1372 f2fs_i_size_write(inode, new_size);
1373 return ret;
1374 }
1375
1376 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1377 pgoff_t end)
1378 {
1379 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1380 pgoff_t index = start;
1381 unsigned int ofs_in_node = dn->ofs_in_node;
1382 blkcnt_t count = 0;
1383 int ret;
1384
1385 for (; index < end; index++, dn->ofs_in_node++) {
1386 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1387 count++;
1388 }
1389
1390 dn->ofs_in_node = ofs_in_node;
1391 ret = f2fs_reserve_new_blocks(dn, count);
1392 if (ret)
1393 return ret;
1394
1395 dn->ofs_in_node = ofs_in_node;
1396 for (index = start; index < end; index++, dn->ofs_in_node++) {
1397 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1398 /*
1399 * f2fs_reserve_new_blocks will not guarantee entire block
1400 * allocation.
1401 */
1402 if (dn->data_blkaddr == NULL_ADDR) {
1403 ret = -ENOSPC;
1404 break;
1405 }
1406 if (dn->data_blkaddr != NEW_ADDR) {
1407 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1408 dn->data_blkaddr = NEW_ADDR;
1409 f2fs_set_data_blkaddr(dn);
1410 }
1411 }
1412
1413 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1414
1415 return ret;
1416 }
1417
1418 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1419 int mode)
1420 {
1421 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1422 struct address_space *mapping = inode->i_mapping;
1423 pgoff_t index, pg_start, pg_end;
1424 loff_t new_size = i_size_read(inode);
1425 loff_t off_start, off_end;
1426 int ret = 0;
1427
1428 ret = inode_newsize_ok(inode, (len + offset));
1429 if (ret)
1430 return ret;
1431
1432 ret = f2fs_convert_inline_inode(inode);
1433 if (ret)
1434 return ret;
1435
1436 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1437 if (ret)
1438 return ret;
1439
1440 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1441 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1442
1443 off_start = offset & (PAGE_SIZE - 1);
1444 off_end = (offset + len) & (PAGE_SIZE - 1);
1445
1446 if (pg_start == pg_end) {
1447 ret = fill_zero(inode, pg_start, off_start,
1448 off_end - off_start);
1449 if (ret)
1450 return ret;
1451
1452 new_size = max_t(loff_t, new_size, offset + len);
1453 } else {
1454 if (off_start) {
1455 ret = fill_zero(inode, pg_start++, off_start,
1456 PAGE_SIZE - off_start);
1457 if (ret)
1458 return ret;
1459
1460 new_size = max_t(loff_t, new_size,
1461 (loff_t)pg_start << PAGE_SHIFT);
1462 }
1463
1464 for (index = pg_start; index < pg_end;) {
1465 struct dnode_of_data dn;
1466 unsigned int end_offset;
1467 pgoff_t end;
1468
1469 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1470 down_write(&F2FS_I(inode)->i_mmap_sem);
1471
1472 truncate_pagecache_range(inode,
1473 (loff_t)index << PAGE_SHIFT,
1474 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1475
1476 f2fs_lock_op(sbi);
1477
1478 set_new_dnode(&dn, inode, NULL, NULL, 0);
1479 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1480 if (ret) {
1481 f2fs_unlock_op(sbi);
1482 up_write(&F2FS_I(inode)->i_mmap_sem);
1483 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1484 goto out;
1485 }
1486
1487 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1488 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1489
1490 ret = f2fs_do_zero_range(&dn, index, end);
1491 f2fs_put_dnode(&dn);
1492
1493 f2fs_unlock_op(sbi);
1494 up_write(&F2FS_I(inode)->i_mmap_sem);
1495 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1496
1497 f2fs_balance_fs(sbi, dn.node_changed);
1498
1499 if (ret)
1500 goto out;
1501
1502 index = end;
1503 new_size = max_t(loff_t, new_size,
1504 (loff_t)index << PAGE_SHIFT);
1505 }
1506
1507 if (off_end) {
1508 ret = fill_zero(inode, pg_end, 0, off_end);
1509 if (ret)
1510 goto out;
1511
1512 new_size = max_t(loff_t, new_size, offset + len);
1513 }
1514 }
1515
1516 out:
1517 if (new_size > i_size_read(inode)) {
1518 if (mode & FALLOC_FL_KEEP_SIZE)
1519 file_set_keep_isize(inode);
1520 else
1521 f2fs_i_size_write(inode, new_size);
1522 }
1523 return ret;
1524 }
1525
1526 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1527 {
1528 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1529 pgoff_t nr, pg_start, pg_end, delta, idx;
1530 loff_t new_size;
1531 int ret = 0;
1532
1533 new_size = i_size_read(inode) + len;
1534 ret = inode_newsize_ok(inode, new_size);
1535 if (ret)
1536 return ret;
1537
1538 if (offset >= i_size_read(inode))
1539 return -EINVAL;
1540
1541 /* insert range should be aligned to block size of f2fs. */
1542 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1543 return -EINVAL;
1544
1545 ret = f2fs_convert_inline_inode(inode);
1546 if (ret)
1547 return ret;
1548
1549 f2fs_balance_fs(sbi, true);
1550
1551 down_write(&F2FS_I(inode)->i_mmap_sem);
1552 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1553 up_write(&F2FS_I(inode)->i_mmap_sem);
1554 if (ret)
1555 return ret;
1556
1557 /* write out all dirty pages from offset */
1558 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1559 if (ret)
1560 return ret;
1561
1562 pg_start = offset >> PAGE_SHIFT;
1563 pg_end = (offset + len) >> PAGE_SHIFT;
1564 delta = pg_end - pg_start;
1565 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1566
1567 /* avoid gc operation during block exchange */
1568 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1569 down_write(&F2FS_I(inode)->i_mmap_sem);
1570 truncate_pagecache(inode, offset);
1571
1572 while (!ret && idx > pg_start) {
1573 nr = idx - pg_start;
1574 if (nr > delta)
1575 nr = delta;
1576 idx -= nr;
1577
1578 f2fs_lock_op(sbi);
1579 f2fs_drop_extent_tree(inode);
1580
1581 ret = __exchange_data_block(inode, inode, idx,
1582 idx + delta, nr, false);
1583 f2fs_unlock_op(sbi);
1584 }
1585 up_write(&F2FS_I(inode)->i_mmap_sem);
1586 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1587
1588 /* write out all moved pages, if possible */
1589 down_write(&F2FS_I(inode)->i_mmap_sem);
1590 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1591 truncate_pagecache(inode, offset);
1592 up_write(&F2FS_I(inode)->i_mmap_sem);
1593
1594 if (!ret)
1595 f2fs_i_size_write(inode, new_size);
1596 return ret;
1597 }
1598
1599 static int expand_inode_data(struct inode *inode, loff_t offset,
1600 loff_t len, int mode)
1601 {
1602 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1603 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1604 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1605 .m_may_create = true };
1606 pgoff_t pg_end;
1607 loff_t new_size = i_size_read(inode);
1608 loff_t off_end;
1609 int err;
1610
1611 err = inode_newsize_ok(inode, (len + offset));
1612 if (err)
1613 return err;
1614
1615 err = f2fs_convert_inline_inode(inode);
1616 if (err)
1617 return err;
1618
1619 f2fs_balance_fs(sbi, true);
1620
1621 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1622 off_end = (offset + len) & (PAGE_SIZE - 1);
1623
1624 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1625 map.m_len = pg_end - map.m_lblk;
1626 if (off_end)
1627 map.m_len++;
1628
1629 if (!map.m_len)
1630 return 0;
1631
1632 if (f2fs_is_pinned_file(inode)) {
1633 block_t len = (map.m_len >> sbi->log_blocks_per_seg) <<
1634 sbi->log_blocks_per_seg;
1635 block_t done = 0;
1636
1637 if (map.m_len % sbi->blocks_per_seg)
1638 len += sbi->blocks_per_seg;
1639
1640 map.m_len = sbi->blocks_per_seg;
1641 next_alloc:
1642 if (has_not_enough_free_secs(sbi, 0,
1643 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1644 down_write(&sbi->gc_lock);
1645 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1646 if (err && err != -ENODATA && err != -EAGAIN)
1647 goto out_err;
1648 }
1649
1650 down_write(&sbi->pin_sem);
1651 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1652 f2fs_allocate_new_segments(sbi, CURSEG_COLD_DATA);
1653 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1654 up_write(&sbi->pin_sem);
1655
1656 done += map.m_len;
1657 len -= map.m_len;
1658 map.m_lblk += map.m_len;
1659 if (!err && len)
1660 goto next_alloc;
1661
1662 map.m_len = done;
1663 } else {
1664 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1665 }
1666 out_err:
1667 if (err) {
1668 pgoff_t last_off;
1669
1670 if (!map.m_len)
1671 return err;
1672
1673 last_off = map.m_lblk + map.m_len - 1;
1674
1675 /* update new size to the failed position */
1676 new_size = (last_off == pg_end) ? offset + len :
1677 (loff_t)(last_off + 1) << PAGE_SHIFT;
1678 } else {
1679 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1680 }
1681
1682 if (new_size > i_size_read(inode)) {
1683 if (mode & FALLOC_FL_KEEP_SIZE)
1684 file_set_keep_isize(inode);
1685 else
1686 f2fs_i_size_write(inode, new_size);
1687 }
1688
1689 return err;
1690 }
1691
1692 static long f2fs_fallocate(struct file *file, int mode,
1693 loff_t offset, loff_t len)
1694 {
1695 struct inode *inode = file_inode(file);
1696 long ret = 0;
1697
1698 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1699 return -EIO;
1700 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1701 return -ENOSPC;
1702 if (!f2fs_is_compress_backend_ready(inode))
1703 return -EOPNOTSUPP;
1704
1705 /* f2fs only support ->fallocate for regular file */
1706 if (!S_ISREG(inode->i_mode))
1707 return -EINVAL;
1708
1709 if (IS_ENCRYPTED(inode) &&
1710 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1711 return -EOPNOTSUPP;
1712
1713 if (f2fs_compressed_file(inode) &&
1714 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1715 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1716 return -EOPNOTSUPP;
1717
1718 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1719 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1720 FALLOC_FL_INSERT_RANGE))
1721 return -EOPNOTSUPP;
1722
1723 inode_lock(inode);
1724
1725 if (mode & FALLOC_FL_PUNCH_HOLE) {
1726 if (offset >= inode->i_size)
1727 goto out;
1728
1729 ret = punch_hole(inode, offset, len);
1730 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1731 ret = f2fs_collapse_range(inode, offset, len);
1732 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1733 ret = f2fs_zero_range(inode, offset, len, mode);
1734 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1735 ret = f2fs_insert_range(inode, offset, len);
1736 } else {
1737 ret = expand_inode_data(inode, offset, len, mode);
1738 }
1739
1740 if (!ret) {
1741 inode->i_mtime = inode->i_ctime = current_time(inode);
1742 f2fs_mark_inode_dirty_sync(inode, false);
1743 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1744 }
1745
1746 out:
1747 inode_unlock(inode);
1748
1749 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1750 return ret;
1751 }
1752
1753 static int f2fs_release_file(struct inode *inode, struct file *filp)
1754 {
1755 /*
1756 * f2fs_relase_file is called at every close calls. So we should
1757 * not drop any inmemory pages by close called by other process.
1758 */
1759 if (!(filp->f_mode & FMODE_WRITE) ||
1760 atomic_read(&inode->i_writecount) != 1)
1761 return 0;
1762
1763 /* some remained atomic pages should discarded */
1764 if (f2fs_is_atomic_file(inode))
1765 f2fs_drop_inmem_pages(inode);
1766 if (f2fs_is_volatile_file(inode)) {
1767 set_inode_flag(inode, FI_DROP_CACHE);
1768 filemap_fdatawrite(inode->i_mapping);
1769 clear_inode_flag(inode, FI_DROP_CACHE);
1770 clear_inode_flag(inode, FI_VOLATILE_FILE);
1771 stat_dec_volatile_write(inode);
1772 }
1773 return 0;
1774 }
1775
1776 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1777 {
1778 struct inode *inode = file_inode(file);
1779
1780 /*
1781 * If the process doing a transaction is crashed, we should do
1782 * roll-back. Otherwise, other reader/write can see corrupted database
1783 * until all the writers close its file. Since this should be done
1784 * before dropping file lock, it needs to do in ->flush.
1785 */
1786 if (f2fs_is_atomic_file(inode) &&
1787 F2FS_I(inode)->inmem_task == current)
1788 f2fs_drop_inmem_pages(inode);
1789 return 0;
1790 }
1791
1792 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1793 {
1794 struct f2fs_inode_info *fi = F2FS_I(inode);
1795 u32 masked_flags = fi->i_flags & mask;
1796
1797 f2fs_bug_on(F2FS_I_SB(inode), (iflags & ~mask));
1798
1799 /* Is it quota file? Do not allow user to mess with it */
1800 if (IS_NOQUOTA(inode))
1801 return -EPERM;
1802
1803 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1804 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1805 return -EOPNOTSUPP;
1806 if (!f2fs_empty_dir(inode))
1807 return -ENOTEMPTY;
1808 }
1809
1810 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1811 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1812 return -EOPNOTSUPP;
1813 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1814 return -EINVAL;
1815 }
1816
1817 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1818 if (masked_flags & F2FS_COMPR_FL) {
1819 if (f2fs_disable_compressed_file(inode))
1820 return -EINVAL;
1821 }
1822 if (iflags & F2FS_NOCOMP_FL)
1823 return -EINVAL;
1824 if (iflags & F2FS_COMPR_FL) {
1825 if (!f2fs_may_compress(inode))
1826 return -EINVAL;
1827
1828 set_compress_context(inode);
1829 }
1830 }
1831 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1832 if (masked_flags & F2FS_COMPR_FL)
1833 return -EINVAL;
1834 }
1835
1836 fi->i_flags = iflags | (fi->i_flags & ~mask);
1837 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1838 (fi->i_flags & F2FS_NOCOMP_FL));
1839
1840 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1841 set_inode_flag(inode, FI_PROJ_INHERIT);
1842 else
1843 clear_inode_flag(inode, FI_PROJ_INHERIT);
1844
1845 inode->i_ctime = current_time(inode);
1846 f2fs_set_inode_flags(inode);
1847 f2fs_mark_inode_dirty_sync(inode, true);
1848 return 0;
1849 }
1850
1851 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1852
1853 /*
1854 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1855 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1856 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1857 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1858 */
1859
1860 static const struct {
1861 u32 iflag;
1862 u32 fsflag;
1863 } f2fs_fsflags_map[] = {
1864 { F2FS_COMPR_FL, FS_COMPR_FL },
1865 { F2FS_SYNC_FL, FS_SYNC_FL },
1866 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1867 { F2FS_APPEND_FL, FS_APPEND_FL },
1868 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1869 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1870 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
1871 { F2FS_INDEX_FL, FS_INDEX_FL },
1872 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1873 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1874 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1875 };
1876
1877 #define F2FS_GETTABLE_FS_FL ( \
1878 FS_COMPR_FL | \
1879 FS_SYNC_FL | \
1880 FS_IMMUTABLE_FL | \
1881 FS_APPEND_FL | \
1882 FS_NODUMP_FL | \
1883 FS_NOATIME_FL | \
1884 FS_NOCOMP_FL | \
1885 FS_INDEX_FL | \
1886 FS_DIRSYNC_FL | \
1887 FS_PROJINHERIT_FL | \
1888 FS_ENCRYPT_FL | \
1889 FS_INLINE_DATA_FL | \
1890 FS_NOCOW_FL | \
1891 FS_VERITY_FL | \
1892 FS_CASEFOLD_FL)
1893
1894 #define F2FS_SETTABLE_FS_FL ( \
1895 FS_COMPR_FL | \
1896 FS_SYNC_FL | \
1897 FS_IMMUTABLE_FL | \
1898 FS_APPEND_FL | \
1899 FS_NODUMP_FL | \
1900 FS_NOATIME_FL | \
1901 FS_NOCOMP_FL | \
1902 FS_DIRSYNC_FL | \
1903 FS_PROJINHERIT_FL | \
1904 FS_CASEFOLD_FL)
1905
1906 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1907 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1908 {
1909 u32 fsflags = 0;
1910 int i;
1911
1912 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1913 if (iflags & f2fs_fsflags_map[i].iflag)
1914 fsflags |= f2fs_fsflags_map[i].fsflag;
1915
1916 return fsflags;
1917 }
1918
1919 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1920 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1921 {
1922 u32 iflags = 0;
1923 int i;
1924
1925 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1926 if (fsflags & f2fs_fsflags_map[i].fsflag)
1927 iflags |= f2fs_fsflags_map[i].iflag;
1928
1929 return iflags;
1930 }
1931
1932 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1933 {
1934 struct inode *inode = file_inode(filp);
1935 struct f2fs_inode_info *fi = F2FS_I(inode);
1936 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1937
1938 if (IS_ENCRYPTED(inode))
1939 fsflags |= FS_ENCRYPT_FL;
1940 if (IS_VERITY(inode))
1941 fsflags |= FS_VERITY_FL;
1942 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1943 fsflags |= FS_INLINE_DATA_FL;
1944 if (is_inode_flag_set(inode, FI_PIN_FILE))
1945 fsflags |= FS_NOCOW_FL;
1946
1947 fsflags &= F2FS_GETTABLE_FS_FL;
1948
1949 return put_user(fsflags, (int __user *)arg);
1950 }
1951
1952 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1953 {
1954 struct inode *inode = file_inode(filp);
1955 struct f2fs_inode_info *fi = F2FS_I(inode);
1956 u32 fsflags, old_fsflags;
1957 u32 iflags;
1958 int ret;
1959
1960 if (!inode_owner_or_capable(inode))
1961 return -EACCES;
1962
1963 if (get_user(fsflags, (int __user *)arg))
1964 return -EFAULT;
1965
1966 if (fsflags & ~F2FS_GETTABLE_FS_FL)
1967 return -EOPNOTSUPP;
1968 fsflags &= F2FS_SETTABLE_FS_FL;
1969
1970 iflags = f2fs_fsflags_to_iflags(fsflags);
1971 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1972 return -EOPNOTSUPP;
1973
1974 ret = mnt_want_write_file(filp);
1975 if (ret)
1976 return ret;
1977
1978 inode_lock(inode);
1979
1980 old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1981 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1982 if (ret)
1983 goto out;
1984
1985 ret = f2fs_setflags_common(inode, iflags,
1986 f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1987 out:
1988 inode_unlock(inode);
1989 mnt_drop_write_file(filp);
1990 return ret;
1991 }
1992
1993 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1994 {
1995 struct inode *inode = file_inode(filp);
1996
1997 return put_user(inode->i_generation, (int __user *)arg);
1998 }
1999
2000 static int f2fs_ioc_start_atomic_write(struct file *filp)
2001 {
2002 struct inode *inode = file_inode(filp);
2003 struct f2fs_inode_info *fi = F2FS_I(inode);
2004 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2005 int ret;
2006
2007 if (!inode_owner_or_capable(inode))
2008 return -EACCES;
2009
2010 if (!S_ISREG(inode->i_mode))
2011 return -EINVAL;
2012
2013 if (filp->f_flags & O_DIRECT)
2014 return -EINVAL;
2015
2016 ret = mnt_want_write_file(filp);
2017 if (ret)
2018 return ret;
2019
2020 inode_lock(inode);
2021
2022 f2fs_disable_compressed_file(inode);
2023
2024 if (f2fs_is_atomic_file(inode)) {
2025 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2026 ret = -EINVAL;
2027 goto out;
2028 }
2029
2030 ret = f2fs_convert_inline_inode(inode);
2031 if (ret)
2032 goto out;
2033
2034 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2035
2036 /*
2037 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2038 * f2fs_is_atomic_file.
2039 */
2040 if (get_dirty_pages(inode))
2041 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2042 inode->i_ino, get_dirty_pages(inode));
2043 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2044 if (ret) {
2045 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2046 goto out;
2047 }
2048
2049 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2050 if (list_empty(&fi->inmem_ilist))
2051 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2052 sbi->atomic_files++;
2053 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2054
2055 /* add inode in inmem_list first and set atomic_file */
2056 set_inode_flag(inode, FI_ATOMIC_FILE);
2057 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2058 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2059
2060 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2061 F2FS_I(inode)->inmem_task = current;
2062 stat_update_max_atomic_write(inode);
2063 out:
2064 inode_unlock(inode);
2065 mnt_drop_write_file(filp);
2066 return ret;
2067 }
2068
2069 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2070 {
2071 struct inode *inode = file_inode(filp);
2072 int ret;
2073
2074 if (!inode_owner_or_capable(inode))
2075 return -EACCES;
2076
2077 ret = mnt_want_write_file(filp);
2078 if (ret)
2079 return ret;
2080
2081 f2fs_balance_fs(F2FS_I_SB(inode), true);
2082
2083 inode_lock(inode);
2084
2085 if (f2fs_is_volatile_file(inode)) {
2086 ret = -EINVAL;
2087 goto err_out;
2088 }
2089
2090 if (f2fs_is_atomic_file(inode)) {
2091 ret = f2fs_commit_inmem_pages(inode);
2092 if (ret)
2093 goto err_out;
2094
2095 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2096 if (!ret)
2097 f2fs_drop_inmem_pages(inode);
2098 } else {
2099 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2100 }
2101 err_out:
2102 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2103 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2104 ret = -EINVAL;
2105 }
2106 inode_unlock(inode);
2107 mnt_drop_write_file(filp);
2108 return ret;
2109 }
2110
2111 static int f2fs_ioc_start_volatile_write(struct file *filp)
2112 {
2113 struct inode *inode = file_inode(filp);
2114 int ret;
2115
2116 if (!inode_owner_or_capable(inode))
2117 return -EACCES;
2118
2119 if (!S_ISREG(inode->i_mode))
2120 return -EINVAL;
2121
2122 ret = mnt_want_write_file(filp);
2123 if (ret)
2124 return ret;
2125
2126 inode_lock(inode);
2127
2128 if (f2fs_is_volatile_file(inode))
2129 goto out;
2130
2131 ret = f2fs_convert_inline_inode(inode);
2132 if (ret)
2133 goto out;
2134
2135 stat_inc_volatile_write(inode);
2136 stat_update_max_volatile_write(inode);
2137
2138 set_inode_flag(inode, FI_VOLATILE_FILE);
2139 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2140 out:
2141 inode_unlock(inode);
2142 mnt_drop_write_file(filp);
2143 return ret;
2144 }
2145
2146 static int f2fs_ioc_release_volatile_write(struct file *filp)
2147 {
2148 struct inode *inode = file_inode(filp);
2149 int ret;
2150
2151 if (!inode_owner_or_capable(inode))
2152 return -EACCES;
2153
2154 ret = mnt_want_write_file(filp);
2155 if (ret)
2156 return ret;
2157
2158 inode_lock(inode);
2159
2160 if (!f2fs_is_volatile_file(inode))
2161 goto out;
2162
2163 if (!f2fs_is_first_block_written(inode)) {
2164 ret = truncate_partial_data_page(inode, 0, true);
2165 goto out;
2166 }
2167
2168 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2169 out:
2170 inode_unlock(inode);
2171 mnt_drop_write_file(filp);
2172 return ret;
2173 }
2174
2175 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2176 {
2177 struct inode *inode = file_inode(filp);
2178 int ret;
2179
2180 if (!inode_owner_or_capable(inode))
2181 return -EACCES;
2182
2183 ret = mnt_want_write_file(filp);
2184 if (ret)
2185 return ret;
2186
2187 inode_lock(inode);
2188
2189 if (f2fs_is_atomic_file(inode))
2190 f2fs_drop_inmem_pages(inode);
2191 if (f2fs_is_volatile_file(inode)) {
2192 clear_inode_flag(inode, FI_VOLATILE_FILE);
2193 stat_dec_volatile_write(inode);
2194 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2195 }
2196
2197 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2198
2199 inode_unlock(inode);
2200
2201 mnt_drop_write_file(filp);
2202 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2203 return ret;
2204 }
2205
2206 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2207 {
2208 struct inode *inode = file_inode(filp);
2209 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2210 struct super_block *sb = sbi->sb;
2211 __u32 in;
2212 int ret = 0;
2213
2214 if (!capable(CAP_SYS_ADMIN))
2215 return -EPERM;
2216
2217 if (get_user(in, (__u32 __user *)arg))
2218 return -EFAULT;
2219
2220 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2221 ret = mnt_want_write_file(filp);
2222 if (ret)
2223 return ret;
2224 }
2225
2226 switch (in) {
2227 case F2FS_GOING_DOWN_FULLSYNC:
2228 sb = freeze_bdev(sb->s_bdev);
2229 if (IS_ERR(sb)) {
2230 ret = PTR_ERR(sb);
2231 goto out;
2232 }
2233 if (sb) {
2234 f2fs_stop_checkpoint(sbi, false);
2235 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2236 thaw_bdev(sb->s_bdev, sb);
2237 }
2238 break;
2239 case F2FS_GOING_DOWN_METASYNC:
2240 /* do checkpoint only */
2241 ret = f2fs_sync_fs(sb, 1);
2242 if (ret)
2243 goto out;
2244 f2fs_stop_checkpoint(sbi, false);
2245 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2246 break;
2247 case F2FS_GOING_DOWN_NOSYNC:
2248 f2fs_stop_checkpoint(sbi, false);
2249 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2250 break;
2251 case F2FS_GOING_DOWN_METAFLUSH:
2252 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2253 f2fs_stop_checkpoint(sbi, false);
2254 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2255 break;
2256 case F2FS_GOING_DOWN_NEED_FSCK:
2257 set_sbi_flag(sbi, SBI_NEED_FSCK);
2258 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2259 set_sbi_flag(sbi, SBI_IS_DIRTY);
2260 /* do checkpoint only */
2261 ret = f2fs_sync_fs(sb, 1);
2262 goto out;
2263 default:
2264 ret = -EINVAL;
2265 goto out;
2266 }
2267
2268 f2fs_stop_gc_thread(sbi);
2269 f2fs_stop_discard_thread(sbi);
2270
2271 f2fs_drop_discard_cmd(sbi);
2272 clear_opt(sbi, DISCARD);
2273
2274 f2fs_update_time(sbi, REQ_TIME);
2275 out:
2276 if (in != F2FS_GOING_DOWN_FULLSYNC)
2277 mnt_drop_write_file(filp);
2278
2279 trace_f2fs_shutdown(sbi, in, ret);
2280
2281 return ret;
2282 }
2283
2284 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2285 {
2286 struct inode *inode = file_inode(filp);
2287 struct super_block *sb = inode->i_sb;
2288 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2289 struct fstrim_range range;
2290 int ret;
2291
2292 if (!capable(CAP_SYS_ADMIN))
2293 return -EPERM;
2294
2295 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2296 return -EOPNOTSUPP;
2297
2298 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2299 sizeof(range)))
2300 return -EFAULT;
2301
2302 ret = mnt_want_write_file(filp);
2303 if (ret)
2304 return ret;
2305
2306 range.minlen = max((unsigned int)range.minlen,
2307 q->limits.discard_granularity);
2308 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2309 mnt_drop_write_file(filp);
2310 if (ret < 0)
2311 return ret;
2312
2313 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2314 sizeof(range)))
2315 return -EFAULT;
2316 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2317 return 0;
2318 }
2319
2320 static bool uuid_is_nonzero(__u8 u[16])
2321 {
2322 int i;
2323
2324 for (i = 0; i < 16; i++)
2325 if (u[i])
2326 return true;
2327 return false;
2328 }
2329
2330 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2331 {
2332 struct inode *inode = file_inode(filp);
2333
2334 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2335 return -EOPNOTSUPP;
2336
2337 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2338
2339 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2340 }
2341
2342 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2343 {
2344 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2345 return -EOPNOTSUPP;
2346 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2347 }
2348
2349 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2350 {
2351 struct inode *inode = file_inode(filp);
2352 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2353 int err;
2354
2355 if (!f2fs_sb_has_encrypt(sbi))
2356 return -EOPNOTSUPP;
2357
2358 err = mnt_want_write_file(filp);
2359 if (err)
2360 return err;
2361
2362 down_write(&sbi->sb_lock);
2363
2364 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2365 goto got_it;
2366
2367 /* update superblock with uuid */
2368 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2369
2370 err = f2fs_commit_super(sbi, false);
2371 if (err) {
2372 /* undo new data */
2373 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2374 goto out_err;
2375 }
2376 got_it:
2377 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2378 16))
2379 err = -EFAULT;
2380 out_err:
2381 up_write(&sbi->sb_lock);
2382 mnt_drop_write_file(filp);
2383 return err;
2384 }
2385
2386 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2387 unsigned long arg)
2388 {
2389 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2390 return -EOPNOTSUPP;
2391
2392 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2393 }
2394
2395 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2396 {
2397 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2398 return -EOPNOTSUPP;
2399
2400 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2401 }
2402
2403 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2404 {
2405 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2406 return -EOPNOTSUPP;
2407
2408 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2409 }
2410
2411 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2412 unsigned long arg)
2413 {
2414 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2415 return -EOPNOTSUPP;
2416
2417 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2418 }
2419
2420 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2421 unsigned long arg)
2422 {
2423 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2424 return -EOPNOTSUPP;
2425
2426 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2427 }
2428
2429 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2430 {
2431 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2432 return -EOPNOTSUPP;
2433
2434 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2435 }
2436
2437 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2438 {
2439 struct inode *inode = file_inode(filp);
2440 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2441 __u32 sync;
2442 int ret;
2443
2444 if (!capable(CAP_SYS_ADMIN))
2445 return -EPERM;
2446
2447 if (get_user(sync, (__u32 __user *)arg))
2448 return -EFAULT;
2449
2450 if (f2fs_readonly(sbi->sb))
2451 return -EROFS;
2452
2453 ret = mnt_want_write_file(filp);
2454 if (ret)
2455 return ret;
2456
2457 if (!sync) {
2458 if (!down_write_trylock(&sbi->gc_lock)) {
2459 ret = -EBUSY;
2460 goto out;
2461 }
2462 } else {
2463 down_write(&sbi->gc_lock);
2464 }
2465
2466 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2467 out:
2468 mnt_drop_write_file(filp);
2469 return ret;
2470 }
2471
2472 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2473 {
2474 struct inode *inode = file_inode(filp);
2475 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2476 struct f2fs_gc_range range;
2477 u64 end;
2478 int ret;
2479
2480 if (!capable(CAP_SYS_ADMIN))
2481 return -EPERM;
2482
2483 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2484 sizeof(range)))
2485 return -EFAULT;
2486
2487 if (f2fs_readonly(sbi->sb))
2488 return -EROFS;
2489
2490 end = range.start + range.len;
2491 if (end < range.start || range.start < MAIN_BLKADDR(sbi) ||
2492 end >= MAX_BLKADDR(sbi))
2493 return -EINVAL;
2494
2495 ret = mnt_want_write_file(filp);
2496 if (ret)
2497 return ret;
2498
2499 do_more:
2500 if (!range.sync) {
2501 if (!down_write_trylock(&sbi->gc_lock)) {
2502 ret = -EBUSY;
2503 goto out;
2504 }
2505 } else {
2506 down_write(&sbi->gc_lock);
2507 }
2508
2509 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2510 range.start += BLKS_PER_SEC(sbi);
2511 if (range.start <= end)
2512 goto do_more;
2513 out:
2514 mnt_drop_write_file(filp);
2515 return ret;
2516 }
2517
2518 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2519 {
2520 struct inode *inode = file_inode(filp);
2521 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2522 int ret;
2523
2524 if (!capable(CAP_SYS_ADMIN))
2525 return -EPERM;
2526
2527 if (f2fs_readonly(sbi->sb))
2528 return -EROFS;
2529
2530 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2531 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2532 return -EINVAL;
2533 }
2534
2535 ret = mnt_want_write_file(filp);
2536 if (ret)
2537 return ret;
2538
2539 ret = f2fs_sync_fs(sbi->sb, 1);
2540
2541 mnt_drop_write_file(filp);
2542 return ret;
2543 }
2544
2545 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2546 struct file *filp,
2547 struct f2fs_defragment *range)
2548 {
2549 struct inode *inode = file_inode(filp);
2550 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2551 .m_seg_type = NO_CHECK_TYPE ,
2552 .m_may_create = false };
2553 struct extent_info ei = {0, 0, 0};
2554 pgoff_t pg_start, pg_end, next_pgofs;
2555 unsigned int blk_per_seg = sbi->blocks_per_seg;
2556 unsigned int total = 0, sec_num;
2557 block_t blk_end = 0;
2558 bool fragmented = false;
2559 int err;
2560
2561 /* if in-place-update policy is enabled, don't waste time here */
2562 if (f2fs_should_update_inplace(inode, NULL))
2563 return -EINVAL;
2564
2565 pg_start = range->start >> PAGE_SHIFT;
2566 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2567
2568 f2fs_balance_fs(sbi, true);
2569
2570 inode_lock(inode);
2571
2572 /* writeback all dirty pages in the range */
2573 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2574 range->start + range->len - 1);
2575 if (err)
2576 goto out;
2577
2578 /*
2579 * lookup mapping info in extent cache, skip defragmenting if physical
2580 * block addresses are continuous.
2581 */
2582 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2583 if (ei.fofs + ei.len >= pg_end)
2584 goto out;
2585 }
2586
2587 map.m_lblk = pg_start;
2588 map.m_next_pgofs = &next_pgofs;
2589
2590 /*
2591 * lookup mapping info in dnode page cache, skip defragmenting if all
2592 * physical block addresses are continuous even if there are hole(s)
2593 * in logical blocks.
2594 */
2595 while (map.m_lblk < pg_end) {
2596 map.m_len = pg_end - map.m_lblk;
2597 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2598 if (err)
2599 goto out;
2600
2601 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2602 map.m_lblk = next_pgofs;
2603 continue;
2604 }
2605
2606 if (blk_end && blk_end != map.m_pblk)
2607 fragmented = true;
2608
2609 /* record total count of block that we're going to move */
2610 total += map.m_len;
2611
2612 blk_end = map.m_pblk + map.m_len;
2613
2614 map.m_lblk += map.m_len;
2615 }
2616
2617 if (!fragmented) {
2618 total = 0;
2619 goto out;
2620 }
2621
2622 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2623
2624 /*
2625 * make sure there are enough free section for LFS allocation, this can
2626 * avoid defragment running in SSR mode when free section are allocated
2627 * intensively
2628 */
2629 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2630 err = -EAGAIN;
2631 goto out;
2632 }
2633
2634 map.m_lblk = pg_start;
2635 map.m_len = pg_end - pg_start;
2636 total = 0;
2637
2638 while (map.m_lblk < pg_end) {
2639 pgoff_t idx;
2640 int cnt = 0;
2641
2642 do_map:
2643 map.m_len = pg_end - map.m_lblk;
2644 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2645 if (err)
2646 goto clear_out;
2647
2648 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2649 map.m_lblk = next_pgofs;
2650 goto check;
2651 }
2652
2653 set_inode_flag(inode, FI_DO_DEFRAG);
2654
2655 idx = map.m_lblk;
2656 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2657 struct page *page;
2658
2659 page = f2fs_get_lock_data_page(inode, idx, true);
2660 if (IS_ERR(page)) {
2661 err = PTR_ERR(page);
2662 goto clear_out;
2663 }
2664
2665 set_page_dirty(page);
2666 f2fs_put_page(page, 1);
2667
2668 idx++;
2669 cnt++;
2670 total++;
2671 }
2672
2673 map.m_lblk = idx;
2674 check:
2675 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2676 goto do_map;
2677
2678 clear_inode_flag(inode, FI_DO_DEFRAG);
2679
2680 err = filemap_fdatawrite(inode->i_mapping);
2681 if (err)
2682 goto out;
2683 }
2684 clear_out:
2685 clear_inode_flag(inode, FI_DO_DEFRAG);
2686 out:
2687 inode_unlock(inode);
2688 if (!err)
2689 range->len = (u64)total << PAGE_SHIFT;
2690 return err;
2691 }
2692
2693 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2694 {
2695 struct inode *inode = file_inode(filp);
2696 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2697 struct f2fs_defragment range;
2698 int err;
2699
2700 if (!capable(CAP_SYS_ADMIN))
2701 return -EPERM;
2702
2703 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2704 return -EINVAL;
2705
2706 if (f2fs_readonly(sbi->sb))
2707 return -EROFS;
2708
2709 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2710 sizeof(range)))
2711 return -EFAULT;
2712
2713 /* verify alignment of offset & size */
2714 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2715 return -EINVAL;
2716
2717 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2718 sbi->max_file_blocks))
2719 return -EINVAL;
2720
2721 err = mnt_want_write_file(filp);
2722 if (err)
2723 return err;
2724
2725 err = f2fs_defragment_range(sbi, filp, &range);
2726 mnt_drop_write_file(filp);
2727
2728 f2fs_update_time(sbi, REQ_TIME);
2729 if (err < 0)
2730 return err;
2731
2732 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2733 sizeof(range)))
2734 return -EFAULT;
2735
2736 return 0;
2737 }
2738
2739 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2740 struct file *file_out, loff_t pos_out, size_t len)
2741 {
2742 struct inode *src = file_inode(file_in);
2743 struct inode *dst = file_inode(file_out);
2744 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2745 size_t olen = len, dst_max_i_size = 0;
2746 size_t dst_osize;
2747 int ret;
2748
2749 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2750 src->i_sb != dst->i_sb)
2751 return -EXDEV;
2752
2753 if (unlikely(f2fs_readonly(src->i_sb)))
2754 return -EROFS;
2755
2756 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2757 return -EINVAL;
2758
2759 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2760 return -EOPNOTSUPP;
2761
2762 if (src == dst) {
2763 if (pos_in == pos_out)
2764 return 0;
2765 if (pos_out > pos_in && pos_out < pos_in + len)
2766 return -EINVAL;
2767 }
2768
2769 inode_lock(src);
2770 if (src != dst) {
2771 ret = -EBUSY;
2772 if (!inode_trylock(dst))
2773 goto out;
2774 }
2775
2776 ret = -EINVAL;
2777 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2778 goto out_unlock;
2779 if (len == 0)
2780 olen = len = src->i_size - pos_in;
2781 if (pos_in + len == src->i_size)
2782 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2783 if (len == 0) {
2784 ret = 0;
2785 goto out_unlock;
2786 }
2787
2788 dst_osize = dst->i_size;
2789 if (pos_out + olen > dst->i_size)
2790 dst_max_i_size = pos_out + olen;
2791
2792 /* verify the end result is block aligned */
2793 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2794 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2795 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2796 goto out_unlock;
2797
2798 ret = f2fs_convert_inline_inode(src);
2799 if (ret)
2800 goto out_unlock;
2801
2802 ret = f2fs_convert_inline_inode(dst);
2803 if (ret)
2804 goto out_unlock;
2805
2806 /* write out all dirty pages from offset */
2807 ret = filemap_write_and_wait_range(src->i_mapping,
2808 pos_in, pos_in + len);
2809 if (ret)
2810 goto out_unlock;
2811
2812 ret = filemap_write_and_wait_range(dst->i_mapping,
2813 pos_out, pos_out + len);
2814 if (ret)
2815 goto out_unlock;
2816
2817 f2fs_balance_fs(sbi, true);
2818
2819 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2820 if (src != dst) {
2821 ret = -EBUSY;
2822 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2823 goto out_src;
2824 }
2825
2826 f2fs_lock_op(sbi);
2827 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2828 pos_out >> F2FS_BLKSIZE_BITS,
2829 len >> F2FS_BLKSIZE_BITS, false);
2830
2831 if (!ret) {
2832 if (dst_max_i_size)
2833 f2fs_i_size_write(dst, dst_max_i_size);
2834 else if (dst_osize != dst->i_size)
2835 f2fs_i_size_write(dst, dst_osize);
2836 }
2837 f2fs_unlock_op(sbi);
2838
2839 if (src != dst)
2840 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2841 out_src:
2842 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2843 out_unlock:
2844 if (src != dst)
2845 inode_unlock(dst);
2846 out:
2847 inode_unlock(src);
2848 return ret;
2849 }
2850
2851 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2852 {
2853 struct f2fs_move_range range;
2854 struct fd dst;
2855 int err;
2856
2857 if (!(filp->f_mode & FMODE_READ) ||
2858 !(filp->f_mode & FMODE_WRITE))
2859 return -EBADF;
2860
2861 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2862 sizeof(range)))
2863 return -EFAULT;
2864
2865 dst = fdget(range.dst_fd);
2866 if (!dst.file)
2867 return -EBADF;
2868
2869 if (!(dst.file->f_mode & FMODE_WRITE)) {
2870 err = -EBADF;
2871 goto err_out;
2872 }
2873
2874 err = mnt_want_write_file(filp);
2875 if (err)
2876 goto err_out;
2877
2878 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2879 range.pos_out, range.len);
2880
2881 mnt_drop_write_file(filp);
2882 if (err)
2883 goto err_out;
2884
2885 if (copy_to_user((struct f2fs_move_range __user *)arg,
2886 &range, sizeof(range)))
2887 err = -EFAULT;
2888 err_out:
2889 fdput(dst);
2890 return err;
2891 }
2892
2893 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2894 {
2895 struct inode *inode = file_inode(filp);
2896 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2897 struct sit_info *sm = SIT_I(sbi);
2898 unsigned int start_segno = 0, end_segno = 0;
2899 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2900 struct f2fs_flush_device range;
2901 int ret;
2902
2903 if (!capable(CAP_SYS_ADMIN))
2904 return -EPERM;
2905
2906 if (f2fs_readonly(sbi->sb))
2907 return -EROFS;
2908
2909 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2910 return -EINVAL;
2911
2912 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2913 sizeof(range)))
2914 return -EFAULT;
2915
2916 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2917 __is_large_section(sbi)) {
2918 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2919 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2920 return -EINVAL;
2921 }
2922
2923 ret = mnt_want_write_file(filp);
2924 if (ret)
2925 return ret;
2926
2927 if (range.dev_num != 0)
2928 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2929 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2930
2931 start_segno = sm->last_victim[FLUSH_DEVICE];
2932 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2933 start_segno = dev_start_segno;
2934 end_segno = min(start_segno + range.segments, dev_end_segno);
2935
2936 while (start_segno < end_segno) {
2937 if (!down_write_trylock(&sbi->gc_lock)) {
2938 ret = -EBUSY;
2939 goto out;
2940 }
2941 sm->last_victim[GC_CB] = end_segno + 1;
2942 sm->last_victim[GC_GREEDY] = end_segno + 1;
2943 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2944 ret = f2fs_gc(sbi, true, true, start_segno);
2945 if (ret == -EAGAIN)
2946 ret = 0;
2947 else if (ret < 0)
2948 break;
2949 start_segno++;
2950 }
2951 out:
2952 mnt_drop_write_file(filp);
2953 return ret;
2954 }
2955
2956 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2957 {
2958 struct inode *inode = file_inode(filp);
2959 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2960
2961 /* Must validate to set it with SQLite behavior in Android. */
2962 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2963
2964 return put_user(sb_feature, (u32 __user *)arg);
2965 }
2966
2967 #ifdef CONFIG_QUOTA
2968 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2969 {
2970 struct dquot *transfer_to[MAXQUOTAS] = {};
2971 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2972 struct super_block *sb = sbi->sb;
2973 int err = 0;
2974
2975 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2976 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2977 err = __dquot_transfer(inode, transfer_to);
2978 if (err)
2979 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2980 dqput(transfer_to[PRJQUOTA]);
2981 }
2982 return err;
2983 }
2984
2985 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2986 {
2987 struct inode *inode = file_inode(filp);
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 file *filp, __u32 projid)
3049 {
3050 if (projid != F2FS_DEF_PROJID)
3051 return -EOPNOTSUPP;
3052 return 0;
3053 }
3054 #endif
3055
3056 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
3057
3058 /*
3059 * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
3060 * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
3061 * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
3062 */
3063
3064 static const struct {
3065 u32 iflag;
3066 u32 xflag;
3067 } f2fs_xflags_map[] = {
3068 { F2FS_SYNC_FL, FS_XFLAG_SYNC },
3069 { F2FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE },
3070 { F2FS_APPEND_FL, FS_XFLAG_APPEND },
3071 { F2FS_NODUMP_FL, FS_XFLAG_NODUMP },
3072 { F2FS_NOATIME_FL, FS_XFLAG_NOATIME },
3073 { F2FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT },
3074 };
3075
3076 #define F2FS_SUPPORTED_XFLAGS ( \
3077 FS_XFLAG_SYNC | \
3078 FS_XFLAG_IMMUTABLE | \
3079 FS_XFLAG_APPEND | \
3080 FS_XFLAG_NODUMP | \
3081 FS_XFLAG_NOATIME | \
3082 FS_XFLAG_PROJINHERIT)
3083
3084 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
3085 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
3086 {
3087 u32 xflags = 0;
3088 int i;
3089
3090 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3091 if (iflags & f2fs_xflags_map[i].iflag)
3092 xflags |= f2fs_xflags_map[i].xflag;
3093
3094 return xflags;
3095 }
3096
3097 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
3098 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
3099 {
3100 u32 iflags = 0;
3101 int i;
3102
3103 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3104 if (xflags & f2fs_xflags_map[i].xflag)
3105 iflags |= f2fs_xflags_map[i].iflag;
3106
3107 return iflags;
3108 }
3109
3110 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
3111 {
3112 struct f2fs_inode_info *fi = F2FS_I(inode);
3113
3114 simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
3115
3116 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3117 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3118 }
3119
3120 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
3121 {
3122 struct inode *inode = file_inode(filp);
3123 struct fsxattr fa;
3124
3125 f2fs_fill_fsxattr(inode, &fa);
3126
3127 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
3128 return -EFAULT;
3129 return 0;
3130 }
3131
3132 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
3133 {
3134 struct inode *inode = file_inode(filp);
3135 struct fsxattr fa, old_fa;
3136 u32 iflags;
3137 int err;
3138
3139 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
3140 return -EFAULT;
3141
3142 /* Make sure caller has proper permission */
3143 if (!inode_owner_or_capable(inode))
3144 return -EACCES;
3145
3146 if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
3147 return -EOPNOTSUPP;
3148
3149 iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
3150 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3151 return -EOPNOTSUPP;
3152
3153 err = mnt_want_write_file(filp);
3154 if (err)
3155 return err;
3156
3157 inode_lock(inode);
3158
3159 f2fs_fill_fsxattr(inode, &old_fa);
3160 err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
3161 if (err)
3162 goto out;
3163
3164 err = f2fs_setflags_common(inode, iflags,
3165 f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
3166 if (err)
3167 goto out;
3168
3169 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3170 out:
3171 inode_unlock(inode);
3172 mnt_drop_write_file(filp);
3173 return err;
3174 }
3175
3176 int f2fs_pin_file_control(struct inode *inode, bool inc)
3177 {
3178 struct f2fs_inode_info *fi = F2FS_I(inode);
3179 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3180
3181 /* Use i_gc_failures for normal file as a risk signal. */
3182 if (inc)
3183 f2fs_i_gc_failures_write(inode,
3184 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3185
3186 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3187 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3188 __func__, inode->i_ino,
3189 fi->i_gc_failures[GC_FAILURE_PIN]);
3190 clear_inode_flag(inode, FI_PIN_FILE);
3191 return -EAGAIN;
3192 }
3193 return 0;
3194 }
3195
3196 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3197 {
3198 struct inode *inode = file_inode(filp);
3199 __u32 pin;
3200 int ret = 0;
3201
3202 if (get_user(pin, (__u32 __user *)arg))
3203 return -EFAULT;
3204
3205 if (!S_ISREG(inode->i_mode))
3206 return -EINVAL;
3207
3208 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3209 return -EROFS;
3210
3211 ret = mnt_want_write_file(filp);
3212 if (ret)
3213 return ret;
3214
3215 inode_lock(inode);
3216
3217 if (f2fs_should_update_outplace(inode, NULL)) {
3218 ret = -EINVAL;
3219 goto out;
3220 }
3221
3222 if (!pin) {
3223 clear_inode_flag(inode, FI_PIN_FILE);
3224 f2fs_i_gc_failures_write(inode, 0);
3225 goto done;
3226 }
3227
3228 if (f2fs_pin_file_control(inode, false)) {
3229 ret = -EAGAIN;
3230 goto out;
3231 }
3232
3233 ret = f2fs_convert_inline_inode(inode);
3234 if (ret)
3235 goto out;
3236
3237 if (f2fs_disable_compressed_file(inode)) {
3238 ret = -EOPNOTSUPP;
3239 goto out;
3240 }
3241
3242 set_inode_flag(inode, FI_PIN_FILE);
3243 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3244 done:
3245 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3246 out:
3247 inode_unlock(inode);
3248 mnt_drop_write_file(filp);
3249 return ret;
3250 }
3251
3252 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3253 {
3254 struct inode *inode = file_inode(filp);
3255 __u32 pin = 0;
3256
3257 if (is_inode_flag_set(inode, FI_PIN_FILE))
3258 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3259 return put_user(pin, (u32 __user *)arg);
3260 }
3261
3262 int f2fs_precache_extents(struct inode *inode)
3263 {
3264 struct f2fs_inode_info *fi = F2FS_I(inode);
3265 struct f2fs_map_blocks map;
3266 pgoff_t m_next_extent;
3267 loff_t end;
3268 int err;
3269
3270 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3271 return -EOPNOTSUPP;
3272
3273 map.m_lblk = 0;
3274 map.m_next_pgofs = NULL;
3275 map.m_next_extent = &m_next_extent;
3276 map.m_seg_type = NO_CHECK_TYPE;
3277 map.m_may_create = false;
3278 end = F2FS_I_SB(inode)->max_file_blocks;
3279
3280 while (map.m_lblk < end) {
3281 map.m_len = end - map.m_lblk;
3282
3283 down_write(&fi->i_gc_rwsem[WRITE]);
3284 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3285 up_write(&fi->i_gc_rwsem[WRITE]);
3286 if (err)
3287 return err;
3288
3289 map.m_lblk = m_next_extent;
3290 }
3291
3292 return err;
3293 }
3294
3295 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3296 {
3297 return f2fs_precache_extents(file_inode(filp));
3298 }
3299
3300 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3301 {
3302 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3303 __u64 block_count;
3304 int ret;
3305
3306 if (!capable(CAP_SYS_ADMIN))
3307 return -EPERM;
3308
3309 if (f2fs_readonly(sbi->sb))
3310 return -EROFS;
3311
3312 if (copy_from_user(&block_count, (void __user *)arg,
3313 sizeof(block_count)))
3314 return -EFAULT;
3315
3316 ret = f2fs_resize_fs(sbi, block_count);
3317
3318 return ret;
3319 }
3320
3321 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3322 {
3323 struct inode *inode = file_inode(filp);
3324
3325 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3326
3327 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3328 f2fs_warn(F2FS_I_SB(inode),
3329 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3330 inode->i_ino);
3331 return -EOPNOTSUPP;
3332 }
3333
3334 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3335 }
3336
3337 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3338 {
3339 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3340 return -EOPNOTSUPP;
3341
3342 return fsverity_ioctl_measure(filp, (void __user *)arg);
3343 }
3344
3345 static int f2fs_get_volume_name(struct file *filp, unsigned long arg)
3346 {
3347 struct inode *inode = file_inode(filp);
3348 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3349 char *vbuf;
3350 int count;
3351 int err = 0;
3352
3353 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3354 if (!vbuf)
3355 return -ENOMEM;
3356
3357 down_read(&sbi->sb_lock);
3358 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3359 ARRAY_SIZE(sbi->raw_super->volume_name),
3360 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3361 up_read(&sbi->sb_lock);
3362
3363 if (copy_to_user((char __user *)arg, vbuf,
3364 min(FSLABEL_MAX, count)))
3365 err = -EFAULT;
3366
3367 kvfree(vbuf);
3368 return err;
3369 }
3370
3371 static int f2fs_set_volume_name(struct file *filp, unsigned long arg)
3372 {
3373 struct inode *inode = file_inode(filp);
3374 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3375 char *vbuf;
3376 int err = 0;
3377
3378 if (!capable(CAP_SYS_ADMIN))
3379 return -EPERM;
3380
3381 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3382 if (IS_ERR(vbuf))
3383 return PTR_ERR(vbuf);
3384
3385 err = mnt_want_write_file(filp);
3386 if (err)
3387 goto out;
3388
3389 down_write(&sbi->sb_lock);
3390
3391 memset(sbi->raw_super->volume_name, 0,
3392 sizeof(sbi->raw_super->volume_name));
3393 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3394 sbi->raw_super->volume_name,
3395 ARRAY_SIZE(sbi->raw_super->volume_name));
3396
3397 err = f2fs_commit_super(sbi, false);
3398
3399 up_write(&sbi->sb_lock);
3400
3401 mnt_drop_write_file(filp);
3402 out:
3403 kfree(vbuf);
3404 return err;
3405 }
3406
3407 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3408 {
3409 struct inode *inode = file_inode(filp);
3410 __u64 blocks;
3411
3412 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3413 return -EOPNOTSUPP;
3414
3415 if (!f2fs_compressed_file(inode))
3416 return -EINVAL;
3417
3418 blocks = F2FS_I(inode)->i_compr_blocks;
3419 return put_user(blocks, (u64 __user *)arg);
3420 }
3421
3422 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3423 {
3424 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3425 return -EIO;
3426 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
3427 return -ENOSPC;
3428
3429 switch (cmd) {
3430 case F2FS_IOC_GETFLAGS:
3431 return f2fs_ioc_getflags(filp, arg);
3432 case F2FS_IOC_SETFLAGS:
3433 return f2fs_ioc_setflags(filp, arg);
3434 case F2FS_IOC_GETVERSION:
3435 return f2fs_ioc_getversion(filp, arg);
3436 case F2FS_IOC_START_ATOMIC_WRITE:
3437 return f2fs_ioc_start_atomic_write(filp);
3438 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3439 return f2fs_ioc_commit_atomic_write(filp);
3440 case F2FS_IOC_START_VOLATILE_WRITE:
3441 return f2fs_ioc_start_volatile_write(filp);
3442 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3443 return f2fs_ioc_release_volatile_write(filp);
3444 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3445 return f2fs_ioc_abort_volatile_write(filp);
3446 case F2FS_IOC_SHUTDOWN:
3447 return f2fs_ioc_shutdown(filp, arg);
3448 case FITRIM:
3449 return f2fs_ioc_fitrim(filp, arg);
3450 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3451 return f2fs_ioc_set_encryption_policy(filp, arg);
3452 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3453 return f2fs_ioc_get_encryption_policy(filp, arg);
3454 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3455 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3456 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3457 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3458 case FS_IOC_ADD_ENCRYPTION_KEY:
3459 return f2fs_ioc_add_encryption_key(filp, arg);
3460 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3461 return f2fs_ioc_remove_encryption_key(filp, arg);
3462 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3463 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3464 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3465 return f2fs_ioc_get_encryption_key_status(filp, arg);
3466 case FS_IOC_GET_ENCRYPTION_NONCE:
3467 return f2fs_ioc_get_encryption_nonce(filp, arg);
3468 case F2FS_IOC_GARBAGE_COLLECT:
3469 return f2fs_ioc_gc(filp, arg);
3470 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3471 return f2fs_ioc_gc_range(filp, arg);
3472 case F2FS_IOC_WRITE_CHECKPOINT:
3473 return f2fs_ioc_write_checkpoint(filp, arg);
3474 case F2FS_IOC_DEFRAGMENT:
3475 return f2fs_ioc_defragment(filp, arg);
3476 case F2FS_IOC_MOVE_RANGE:
3477 return f2fs_ioc_move_range(filp, arg);
3478 case F2FS_IOC_FLUSH_DEVICE:
3479 return f2fs_ioc_flush_device(filp, arg);
3480 case F2FS_IOC_GET_FEATURES:
3481 return f2fs_ioc_get_features(filp, arg);
3482 case F2FS_IOC_FSGETXATTR:
3483 return f2fs_ioc_fsgetxattr(filp, arg);
3484 case F2FS_IOC_FSSETXATTR:
3485 return f2fs_ioc_fssetxattr(filp, arg);
3486 case F2FS_IOC_GET_PIN_FILE:
3487 return f2fs_ioc_get_pin_file(filp, arg);
3488 case F2FS_IOC_SET_PIN_FILE:
3489 return f2fs_ioc_set_pin_file(filp, arg);
3490 case F2FS_IOC_PRECACHE_EXTENTS:
3491 return f2fs_ioc_precache_extents(filp, arg);
3492 case F2FS_IOC_RESIZE_FS:
3493 return f2fs_ioc_resize_fs(filp, arg);
3494 case FS_IOC_ENABLE_VERITY:
3495 return f2fs_ioc_enable_verity(filp, arg);
3496 case FS_IOC_MEASURE_VERITY:
3497 return f2fs_ioc_measure_verity(filp, arg);
3498 case F2FS_IOC_GET_VOLUME_NAME:
3499 return f2fs_get_volume_name(filp, arg);
3500 case F2FS_IOC_SET_VOLUME_NAME:
3501 return f2fs_set_volume_name(filp, arg);
3502 case F2FS_IOC_GET_COMPRESS_BLOCKS:
3503 return f2fs_get_compress_blocks(filp, arg);
3504 default:
3505 return -ENOTTY;
3506 }
3507 }
3508
3509 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
3510 {
3511 struct file *file = iocb->ki_filp;
3512 struct inode *inode = file_inode(file);
3513
3514 if (!f2fs_is_compress_backend_ready(inode))
3515 return -EOPNOTSUPP;
3516
3517 return generic_file_read_iter(iocb, iter);
3518 }
3519
3520 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3521 {
3522 struct file *file = iocb->ki_filp;
3523 struct inode *inode = file_inode(file);
3524 ssize_t ret;
3525
3526 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3527 ret = -EIO;
3528 goto out;
3529 }
3530
3531 if (!f2fs_is_compress_backend_ready(inode)) {
3532 ret = -EOPNOTSUPP;
3533 goto out;
3534 }
3535
3536 if (iocb->ki_flags & IOCB_NOWAIT) {
3537 if (!inode_trylock(inode)) {
3538 ret = -EAGAIN;
3539 goto out;
3540 }
3541 } else {
3542 inode_lock(inode);
3543 }
3544
3545 ret = generic_write_checks(iocb, from);
3546 if (ret > 0) {
3547 bool preallocated = false;
3548 size_t target_size = 0;
3549 int err;
3550
3551 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3552 set_inode_flag(inode, FI_NO_PREALLOC);
3553
3554 if ((iocb->ki_flags & IOCB_NOWAIT)) {
3555 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3556 iov_iter_count(from)) ||
3557 f2fs_has_inline_data(inode) ||
3558 f2fs_force_buffered_io(inode, iocb, from)) {
3559 clear_inode_flag(inode, FI_NO_PREALLOC);
3560 inode_unlock(inode);
3561 ret = -EAGAIN;
3562 goto out;
3563 }
3564 goto write;
3565 }
3566
3567 if (is_inode_flag_set(inode, FI_NO_PREALLOC))
3568 goto write;
3569
3570 if (iocb->ki_flags & IOCB_DIRECT) {
3571 /*
3572 * Convert inline data for Direct I/O before entering
3573 * f2fs_direct_IO().
3574 */
3575 err = f2fs_convert_inline_inode(inode);
3576 if (err)
3577 goto out_err;
3578 /*
3579 * If force_buffere_io() is true, we have to allocate
3580 * blocks all the time, since f2fs_direct_IO will fall
3581 * back to buffered IO.
3582 */
3583 if (!f2fs_force_buffered_io(inode, iocb, from) &&
3584 allow_outplace_dio(inode, iocb, from))
3585 goto write;
3586 }
3587 preallocated = true;
3588 target_size = iocb->ki_pos + iov_iter_count(from);
3589
3590 err = f2fs_preallocate_blocks(iocb, from);
3591 if (err) {
3592 out_err:
3593 clear_inode_flag(inode, FI_NO_PREALLOC);
3594 inode_unlock(inode);
3595 ret = err;
3596 goto out;
3597 }
3598 write:
3599 ret = __generic_file_write_iter(iocb, from);
3600 clear_inode_flag(inode, FI_NO_PREALLOC);
3601
3602 /* if we couldn't write data, we should deallocate blocks. */
3603 if (preallocated && i_size_read(inode) < target_size)
3604 f2fs_truncate(inode);
3605
3606 if (ret > 0)
3607 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3608 }
3609 inode_unlock(inode);
3610 out:
3611 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3612 iov_iter_count(from), ret);
3613 if (ret > 0)
3614 ret = generic_write_sync(iocb, ret);
3615 return ret;
3616 }
3617
3618 #ifdef CONFIG_COMPAT
3619 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3620 {
3621 switch (cmd) {
3622 case F2FS_IOC32_GETFLAGS:
3623 cmd = F2FS_IOC_GETFLAGS;
3624 break;
3625 case F2FS_IOC32_SETFLAGS:
3626 cmd = F2FS_IOC_SETFLAGS;
3627 break;
3628 case F2FS_IOC32_GETVERSION:
3629 cmd = F2FS_IOC_GETVERSION;
3630 break;
3631 case F2FS_IOC_START_ATOMIC_WRITE:
3632 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3633 case F2FS_IOC_START_VOLATILE_WRITE:
3634 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3635 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3636 case F2FS_IOC_SHUTDOWN:
3637 case FITRIM:
3638 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3639 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3640 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3641 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3642 case FS_IOC_ADD_ENCRYPTION_KEY:
3643 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3644 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3645 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3646 case FS_IOC_GET_ENCRYPTION_NONCE:
3647 case F2FS_IOC_GARBAGE_COLLECT:
3648 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3649 case F2FS_IOC_WRITE_CHECKPOINT:
3650 case F2FS_IOC_DEFRAGMENT:
3651 case F2FS_IOC_MOVE_RANGE:
3652 case F2FS_IOC_FLUSH_DEVICE:
3653 case F2FS_IOC_GET_FEATURES:
3654 case F2FS_IOC_FSGETXATTR:
3655 case F2FS_IOC_FSSETXATTR:
3656 case F2FS_IOC_GET_PIN_FILE:
3657 case F2FS_IOC_SET_PIN_FILE:
3658 case F2FS_IOC_PRECACHE_EXTENTS:
3659 case F2FS_IOC_RESIZE_FS:
3660 case FS_IOC_ENABLE_VERITY:
3661 case FS_IOC_MEASURE_VERITY:
3662 case F2FS_IOC_GET_VOLUME_NAME:
3663 case F2FS_IOC_SET_VOLUME_NAME:
3664 case F2FS_IOC_GET_COMPRESS_BLOCKS:
3665 break;
3666 default:
3667 return -ENOIOCTLCMD;
3668 }
3669 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3670 }
3671 #endif
3672
3673 const struct file_operations f2fs_file_operations = {
3674 .llseek = f2fs_llseek,
3675 .read_iter = f2fs_file_read_iter,
3676 .write_iter = f2fs_file_write_iter,
3677 .open = f2fs_file_open,
3678 .release = f2fs_release_file,
3679 .mmap = f2fs_file_mmap,
3680 .flush = f2fs_file_flush,
3681 .fsync = f2fs_sync_file,
3682 .fallocate = f2fs_fallocate,
3683 .unlocked_ioctl = f2fs_ioctl,
3684 #ifdef CONFIG_COMPAT
3685 .compat_ioctl = f2fs_compat_ioctl,
3686 #endif
3687 .splice_read = generic_file_splice_read,
3688 .splice_write = iter_file_splice_write,
3689 };
A mirror of Linus' kernel repository