1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/file.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/buffer_head.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/types.h>
16 #include <linux/compat.h>
17 #include <linux/uaccess.h>
18 #include <linux/mount.h>
19 #include <linux/pagevec.h>
20 #include <linux/uio.h>
21 #include <linux/uuid.h>
22 #include <linux/file.h>
23 #include <linux/nls.h>
24 #include <linux/sched/signal.h>
25
26 #include "f2fs.h"
27 #include "node.h"
28 #include "segment.h"
29 #include "xattr.h"
30 #include "acl.h"
31 #include "gc.h"
32 #include "trace.h"
33 #include <trace/events/f2fs.h>
34 #include <uapi/linux/f2fs.h>
35
f2fs_filemap_fault(struct vm_fault * vmf)36 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
37 {
38 struct inode *inode = file_inode(vmf->vma->vm_file);
39 vm_fault_t ret;
40
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 ret = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
44
45 if (!ret)
46 f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
47 F2FS_BLKSIZE);
48
49 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
50
51 return ret;
52 }
53
f2fs_vm_page_mkwrite(struct vm_fault * vmf)54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
55 {
56 struct page *page = vmf->page;
57 struct inode *inode = file_inode(vmf->vma->vm_file);
58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
59 struct dnode_of_data dn;
60 bool need_alloc = true;
61 int err = 0;
62
63 if (unlikely(IS_IMMUTABLE(inode)))
64 return VM_FAULT_SIGBUS;
65
66 if (unlikely(f2fs_cp_error(sbi))) {
67 err = -EIO;
68 goto err;
69 }
70
71 if (!f2fs_is_checkpoint_ready(sbi)) {
72 err = -ENOSPC;
73 goto err;
74 }
75
76 #ifdef CONFIG_F2FS_FS_COMPRESSION
77 if (f2fs_compressed_file(inode)) {
78 int ret = f2fs_is_compressed_cluster(inode, page->index);
79
80 if (ret < 0) {
81 err = ret;
82 goto err;
83 } else if (ret) {
84 if (ret < F2FS_I(inode)->i_cluster_size) {
85 err = -EAGAIN;
86 goto err;
87 }
88 need_alloc = false;
89 }
90 }
91 #endif
92 /* should do out of any locked page */
93 if (need_alloc)
94 f2fs_balance_fs(sbi, true);
95
96 sb_start_pagefault(inode->i_sb);
97
98 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
99
100 file_update_time(vmf->vma->vm_file);
101 down_read(&F2FS_I(inode)->i_mmap_sem);
102 lock_page(page);
103 if (unlikely(page->mapping != inode->i_mapping ||
104 page_offset(page) > i_size_read(inode) ||
105 !PageUptodate(page))) {
106 unlock_page(page);
107 err = -EFAULT;
108 goto out_sem;
109 }
110
111 if (need_alloc) {
112 /* block allocation */
113 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
114 set_new_dnode(&dn, inode, NULL, NULL, 0);
115 err = f2fs_get_block(&dn, page->index);
116 f2fs_put_dnode(&dn);
117 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
118 }
119
120 #ifdef CONFIG_F2FS_FS_COMPRESSION
121 if (!need_alloc) {
122 set_new_dnode(&dn, inode, NULL, NULL, 0);
123 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
124 f2fs_put_dnode(&dn);
125 }
126 #endif
127 if (err) {
128 unlock_page(page);
129 goto out_sem;
130 }
131
132 f2fs_wait_on_page_writeback(page, DATA, false, true);
133
134 /* wait for GCed page writeback via META_MAPPING */
135 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
136
137 /*
138 * check to see if the page is mapped already (no holes)
139 */
140 if (PageMappedToDisk(page))
141 goto out_sem;
142
143 /* page is wholly or partially inside EOF */
144 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
145 i_size_read(inode)) {
146 loff_t offset;
147
148 offset = i_size_read(inode) & ~PAGE_MASK;
149 zero_user_segment(page, offset, PAGE_SIZE);
150 }
151 set_page_dirty(page);
152 if (!PageUptodate(page))
153 SetPageUptodate(page);
154
155 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
156 f2fs_update_time(sbi, REQ_TIME);
157
158 trace_f2fs_vm_page_mkwrite(page, DATA);
159 out_sem:
160 up_read(&F2FS_I(inode)->i_mmap_sem);
161
162 sb_end_pagefault(inode->i_sb);
163 err:
164 return block_page_mkwrite_return(err);
165 }
166
167 static const struct vm_operations_struct f2fs_file_vm_ops = {
168 .fault = f2fs_filemap_fault,
169 .map_pages = filemap_map_pages,
170 .page_mkwrite = f2fs_vm_page_mkwrite,
171 };
172
get_parent_ino(struct inode * inode,nid_t * pino)173 static int get_parent_ino(struct inode *inode, nid_t *pino)
174 {
175 struct dentry *dentry;
176
177 /*
178 * Make sure to get the non-deleted alias. The alias associated with
179 * the open file descriptor being fsync()'ed may be deleted already.
180 */
181 dentry = d_find_alias(inode);
182 if (!dentry)
183 return 0;
184
185 *pino = parent_ino(dentry);
186 dput(dentry);
187 return 1;
188 }
189
need_do_checkpoint(struct inode * inode)190 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
191 {
192 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
193 enum cp_reason_type cp_reason = CP_NO_NEEDED;
194
195 if (!S_ISREG(inode->i_mode))
196 cp_reason = CP_NON_REGULAR;
197 else if (f2fs_compressed_file(inode))
198 cp_reason = CP_COMPRESSED;
199 else if (inode->i_nlink != 1)
200 cp_reason = CP_HARDLINK;
201 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
202 cp_reason = CP_SB_NEED_CP;
203 else if (file_wrong_pino(inode))
204 cp_reason = CP_WRONG_PINO;
205 else if (!f2fs_space_for_roll_forward(sbi))
206 cp_reason = CP_NO_SPC_ROLL;
207 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
208 cp_reason = CP_NODE_NEED_CP;
209 else if (test_opt(sbi, FASTBOOT))
210 cp_reason = CP_FASTBOOT_MODE;
211 else if (F2FS_OPTION(sbi).active_logs == 2)
212 cp_reason = CP_SPEC_LOG_NUM;
213 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
214 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
215 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
216 TRANS_DIR_INO))
217 cp_reason = CP_RECOVER_DIR;
218
219 return cp_reason;
220 }
221
need_inode_page_update(struct f2fs_sb_info * sbi,nid_t ino)222 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
223 {
224 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
225 bool ret = false;
226 /* But we need to avoid that there are some inode updates */
227 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
228 ret = true;
229 f2fs_put_page(i, 0);
230 return ret;
231 }
232
try_to_fix_pino(struct inode * inode)233 static void try_to_fix_pino(struct inode *inode)
234 {
235 struct f2fs_inode_info *fi = F2FS_I(inode);
236 nid_t pino;
237
238 down_write(&fi->i_sem);
239 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
240 get_parent_ino(inode, &pino)) {
241 f2fs_i_pino_write(inode, pino);
242 file_got_pino(inode);
243 }
244 up_write(&fi->i_sem);
245 }
246
f2fs_do_sync_file(struct file * file,loff_t start,loff_t end,int datasync,bool atomic)247 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
248 int datasync, bool atomic)
249 {
250 struct inode *inode = file->f_mapping->host;
251 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
252 nid_t ino = inode->i_ino;
253 int ret = 0;
254 enum cp_reason_type cp_reason = 0;
255 struct writeback_control wbc = {
256 .sync_mode = WB_SYNC_ALL,
257 .nr_to_write = LONG_MAX,
258 .for_reclaim = 0,
259 };
260 unsigned int seq_id = 0;
261
262 if (unlikely(f2fs_readonly(inode->i_sb)))
263 return 0;
264
265 trace_f2fs_sync_file_enter(inode);
266
267 if (S_ISDIR(inode->i_mode))
268 goto go_write;
269
270 /* if fdatasync is triggered, let's do in-place-update */
271 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
272 set_inode_flag(inode, FI_NEED_IPU);
273 ret = file_write_and_wait_range(file, start, end);
274 clear_inode_flag(inode, FI_NEED_IPU);
275
276 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
277 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
278 return ret;
279 }
280
281 /* if the inode is dirty, let's recover all the time */
282 if (!f2fs_skip_inode_update(inode, datasync)) {
283 f2fs_write_inode(inode, NULL);
284 goto go_write;
285 }
286
287 /*
288 * if there is no written data, don't waste time to write recovery info.
289 */
290 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
291 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
292
293 /* it may call write_inode just prior to fsync */
294 if (need_inode_page_update(sbi, ino))
295 goto go_write;
296
297 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
298 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
299 goto flush_out;
300 goto out;
301 }
302 go_write:
303 /*
304 * Both of fdatasync() and fsync() are able to be recovered from
305 * sudden-power-off.
306 */
307 down_read(&F2FS_I(inode)->i_sem);
308 cp_reason = need_do_checkpoint(inode);
309 up_read(&F2FS_I(inode)->i_sem);
310
311 if (cp_reason) {
312 /* all the dirty node pages should be flushed for POR */
313 ret = f2fs_sync_fs(inode->i_sb, 1);
314
315 /*
316 * We've secured consistency through sync_fs. Following pino
317 * will be used only for fsynced inodes after checkpoint.
318 */
319 try_to_fix_pino(inode);
320 clear_inode_flag(inode, FI_APPEND_WRITE);
321 clear_inode_flag(inode, FI_UPDATE_WRITE);
322 goto out;
323 }
324 sync_nodes:
325 atomic_inc(&sbi->wb_sync_req[NODE]);
326 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
327 atomic_dec(&sbi->wb_sync_req[NODE]);
328 if (ret)
329 goto out;
330
331 /* if cp_error was enabled, we should avoid infinite loop */
332 if (unlikely(f2fs_cp_error(sbi))) {
333 ret = -EIO;
334 goto out;
335 }
336
337 if (f2fs_need_inode_block_update(sbi, ino)) {
338 f2fs_mark_inode_dirty_sync(inode, true);
339 f2fs_write_inode(inode, NULL);
340 goto sync_nodes;
341 }
342
343 /*
344 * If it's atomic_write, it's just fine to keep write ordering. So
345 * here we don't need to wait for node write completion, since we use
346 * node chain which serializes node blocks. If one of node writes are
347 * reordered, we can see simply broken chain, resulting in stopping
348 * roll-forward recovery. It means we'll recover all or none node blocks
349 * given fsync mark.
350 */
351 if (!atomic) {
352 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
353 if (ret)
354 goto out;
355 }
356
357 /* once recovery info is written, don't need to tack this */
358 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
359 clear_inode_flag(inode, FI_APPEND_WRITE);
360 flush_out:
361 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
362 ret = f2fs_issue_flush(sbi, inode->i_ino);
363 if (!ret) {
364 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
365 clear_inode_flag(inode, FI_UPDATE_WRITE);
366 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
367 }
368 f2fs_update_time(sbi, REQ_TIME);
369 out:
370 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
371 f2fs_trace_ios(NULL, 1);
372 return ret;
373 }
374
f2fs_sync_file(struct file * file,loff_t start,loff_t end,int datasync)375 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
376 {
377 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
378 return -EIO;
379 return f2fs_do_sync_file(file, start, end, datasync, false);
380 }
381
__found_offset(struct address_space * mapping,block_t blkaddr,pgoff_t index,int whence)382 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
383 pgoff_t index, int whence)
384 {
385 switch (whence) {
386 case SEEK_DATA:
387 if (__is_valid_data_blkaddr(blkaddr))
388 return true;
389 if (blkaddr == NEW_ADDR &&
390 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
391 return true;
392 break;
393 case SEEK_HOLE:
394 if (blkaddr == NULL_ADDR)
395 return true;
396 break;
397 }
398 return false;
399 }
400
f2fs_seek_block(struct file * file,loff_t offset,int whence)401 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
402 {
403 struct inode *inode = file->f_mapping->host;
404 loff_t maxbytes = inode->i_sb->s_maxbytes;
405 struct dnode_of_data dn;
406 pgoff_t pgofs, end_offset;
407 loff_t data_ofs = offset;
408 loff_t isize;
409 int err = 0;
410
411 inode_lock(inode);
412
413 isize = i_size_read(inode);
414 if (offset >= isize)
415 goto fail;
416
417 /* handle inline data case */
418 if (f2fs_has_inline_data(inode)) {
419 if (whence == SEEK_HOLE) {
420 data_ofs = isize;
421 goto found;
422 } else if (whence == SEEK_DATA) {
423 data_ofs = offset;
424 goto found;
425 }
426 }
427
428 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
429
430 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
431 set_new_dnode(&dn, inode, NULL, NULL, 0);
432 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
433 if (err && err != -ENOENT) {
434 goto fail;
435 } else if (err == -ENOENT) {
436 /* direct node does not exists */
437 if (whence == SEEK_DATA) {
438 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
439 continue;
440 } else {
441 goto found;
442 }
443 }
444
445 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
446
447 /* find data/hole in dnode block */
448 for (; dn.ofs_in_node < end_offset;
449 dn.ofs_in_node++, pgofs++,
450 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
451 block_t blkaddr;
452
453 blkaddr = f2fs_data_blkaddr(&dn);
454
455 if (__is_valid_data_blkaddr(blkaddr) &&
456 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
457 blkaddr, DATA_GENERIC_ENHANCE)) {
458 f2fs_put_dnode(&dn);
459 goto fail;
460 }
461
462 if (__found_offset(file->f_mapping, blkaddr,
463 pgofs, whence)) {
464 f2fs_put_dnode(&dn);
465 goto found;
466 }
467 }
468 f2fs_put_dnode(&dn);
469 }
470
471 if (whence == SEEK_DATA)
472 goto fail;
473 found:
474 if (whence == SEEK_HOLE && data_ofs > isize)
475 data_ofs = isize;
476 inode_unlock(inode);
477 return vfs_setpos(file, data_ofs, maxbytes);
478 fail:
479 inode_unlock(inode);
480 return -ENXIO;
481 }
482
f2fs_llseek(struct file * file,loff_t offset,int whence)483 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
484 {
485 struct inode *inode = file->f_mapping->host;
486 loff_t maxbytes = inode->i_sb->s_maxbytes;
487
488 switch (whence) {
489 case SEEK_SET:
490 case SEEK_CUR:
491 case SEEK_END:
492 return generic_file_llseek_size(file, offset, whence,
493 maxbytes, i_size_read(inode));
494 case SEEK_DATA:
495 case SEEK_HOLE:
496 if (offset < 0)
497 return -ENXIO;
498 return f2fs_seek_block(file, offset, whence);
499 }
500
501 return -EINVAL;
502 }
503
f2fs_file_mmap(struct file * file,struct vm_area_struct * vma)504 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
505 {
506 struct inode *inode = file_inode(file);
507 int err;
508
509 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
510 return -EIO;
511
512 if (!f2fs_is_compress_backend_ready(inode))
513 return -EOPNOTSUPP;
514
515 /* we don't need to use inline_data strictly */
516 err = f2fs_convert_inline_inode(inode);
517 if (err)
518 return err;
519
520 file_accessed(file);
521 vma->vm_ops = &f2fs_file_vm_ops;
522 set_inode_flag(inode, FI_MMAP_FILE);
523 return 0;
524 }
525
f2fs_file_open(struct inode * inode,struct file * filp)526 static int f2fs_file_open(struct inode *inode, struct file *filp)
527 {
528 int err = fscrypt_file_open(inode, filp);
529
530 if (err)
531 return err;
532
533 if (!f2fs_is_compress_backend_ready(inode))
534 return -EOPNOTSUPP;
535
536 err = fsverity_file_open(inode, filp);
537 if (err)
538 return err;
539
540 filp->f_mode |= FMODE_NOWAIT;
541
542 return dquot_file_open(inode, filp);
543 }
544
f2fs_truncate_data_blocks_range(struct dnode_of_data * dn,int count)545 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
546 {
547 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
548 struct f2fs_node *raw_node;
549 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
550 __le32 *addr;
551 int base = 0;
552 bool compressed_cluster = false;
553 int cluster_index = 0, valid_blocks = 0;
554 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
555 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
556
557 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
558 base = get_extra_isize(dn->inode);
559
560 raw_node = F2FS_NODE(dn->node_page);
561 addr = blkaddr_in_node(raw_node) + base + ofs;
562
563 /* Assumption: truncateion starts with cluster */
564 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
565 block_t blkaddr = le32_to_cpu(*addr);
566
567 if (f2fs_compressed_file(dn->inode) &&
568 !(cluster_index & (cluster_size - 1))) {
569 if (compressed_cluster)
570 f2fs_i_compr_blocks_update(dn->inode,
571 valid_blocks, false);
572 compressed_cluster = (blkaddr == COMPRESS_ADDR);
573 valid_blocks = 0;
574 }
575
576 if (blkaddr == NULL_ADDR)
577 continue;
578
579 dn->data_blkaddr = NULL_ADDR;
580 f2fs_set_data_blkaddr(dn);
581
582 if (__is_valid_data_blkaddr(blkaddr)) {
583 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
584 DATA_GENERIC_ENHANCE))
585 continue;
586 if (compressed_cluster)
587 valid_blocks++;
588 }
589
590 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
591 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
592
593 f2fs_invalidate_blocks(sbi, blkaddr);
594
595 if (!released || blkaddr != COMPRESS_ADDR)
596 nr_free++;
597 }
598
599 if (compressed_cluster)
600 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
601
602 if (nr_free) {
603 pgoff_t fofs;
604 /*
605 * once we invalidate valid blkaddr in range [ofs, ofs + count],
606 * we will invalidate all blkaddr in the whole range.
607 */
608 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
609 dn->inode) + ofs;
610 f2fs_update_extent_cache_range(dn, fofs, 0, len);
611 dec_valid_block_count(sbi, dn->inode, nr_free);
612 }
613 dn->ofs_in_node = ofs;
614
615 f2fs_update_time(sbi, REQ_TIME);
616 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
617 dn->ofs_in_node, nr_free);
618 }
619
f2fs_truncate_data_blocks(struct dnode_of_data * dn)620 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
621 {
622 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
623 }
624
truncate_partial_data_page(struct inode * inode,u64 from,bool cache_only)625 static int truncate_partial_data_page(struct inode *inode, u64 from,
626 bool cache_only)
627 {
628 loff_t offset = from & (PAGE_SIZE - 1);
629 pgoff_t index = from >> PAGE_SHIFT;
630 struct address_space *mapping = inode->i_mapping;
631 struct page *page;
632
633 if (!offset && !cache_only)
634 return 0;
635
636 if (cache_only) {
637 page = find_lock_page(mapping, index);
638 if (page && PageUptodate(page))
639 goto truncate_out;
640 f2fs_put_page(page, 1);
641 return 0;
642 }
643
644 page = f2fs_get_lock_data_page(inode, index, true);
645 if (IS_ERR(page))
646 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
647 truncate_out:
648 f2fs_wait_on_page_writeback(page, DATA, true, true);
649 zero_user(page, offset, PAGE_SIZE - offset);
650
651 /* An encrypted inode should have a key and truncate the last page. */
652 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
653 if (!cache_only)
654 set_page_dirty(page);
655 f2fs_put_page(page, 1);
656 return 0;
657 }
658
f2fs_do_truncate_blocks(struct inode * inode,u64 from,bool lock)659 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
660 {
661 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
662 struct dnode_of_data dn;
663 pgoff_t free_from;
664 int count = 0, err = 0;
665 struct page *ipage;
666 bool truncate_page = false;
667
668 trace_f2fs_truncate_blocks_enter(inode, from);
669
670 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
671
672 if (free_from >= sbi->max_file_blocks)
673 goto free_partial;
674
675 if (lock)
676 f2fs_lock_op(sbi);
677
678 ipage = f2fs_get_node_page(sbi, inode->i_ino);
679 if (IS_ERR(ipage)) {
680 err = PTR_ERR(ipage);
681 goto out;
682 }
683
684 if (f2fs_has_inline_data(inode)) {
685 f2fs_truncate_inline_inode(inode, ipage, from);
686 f2fs_put_page(ipage, 1);
687 truncate_page = true;
688 goto out;
689 }
690
691 set_new_dnode(&dn, inode, ipage, NULL, 0);
692 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
693 if (err) {
694 if (err == -ENOENT)
695 goto free_next;
696 goto out;
697 }
698
699 count = ADDRS_PER_PAGE(dn.node_page, inode);
700
701 count -= dn.ofs_in_node;
702 f2fs_bug_on(sbi, count < 0);
703
704 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
705 f2fs_truncate_data_blocks_range(&dn, count);
706 free_from += count;
707 }
708
709 f2fs_put_dnode(&dn);
710 free_next:
711 err = f2fs_truncate_inode_blocks(inode, free_from);
712 out:
713 if (lock)
714 f2fs_unlock_op(sbi);
715 free_partial:
716 /* lastly zero out the first data page */
717 if (!err)
718 err = truncate_partial_data_page(inode, from, truncate_page);
719
720 trace_f2fs_truncate_blocks_exit(inode, err);
721 return err;
722 }
723
f2fs_truncate_blocks(struct inode * inode,u64 from,bool lock)724 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
725 {
726 u64 free_from = from;
727 int err;
728
729 #ifdef CONFIG_F2FS_FS_COMPRESSION
730 /*
731 * for compressed file, only support cluster size
732 * aligned truncation.
733 */
734 if (f2fs_compressed_file(inode))
735 free_from = round_up(from,
736 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
737 #endif
738
739 err = f2fs_do_truncate_blocks(inode, free_from, lock);
740 if (err)
741 return err;
742
743 #ifdef CONFIG_F2FS_FS_COMPRESSION
744 if (from != free_from) {
745 err = f2fs_truncate_partial_cluster(inode, from, lock);
746 if (err)
747 return err;
748 }
749 #endif
750
751 return 0;
752 }
753
f2fs_truncate(struct inode * inode)754 int f2fs_truncate(struct inode *inode)
755 {
756 int err;
757
758 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
759 return -EIO;
760
761 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
762 S_ISLNK(inode->i_mode)))
763 return 0;
764
765 trace_f2fs_truncate(inode);
766
767 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
768 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
769 return -EIO;
770 }
771
772 err = dquot_initialize(inode);
773 if (err)
774 return err;
775
776 /* we should check inline_data size */
777 if (!f2fs_may_inline_data(inode)) {
778 err = f2fs_convert_inline_inode(inode);
779 if (err)
780 return err;
781 }
782
783 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
784 if (err)
785 return err;
786
787 inode->i_mtime = inode->i_ctime = current_time(inode);
788 f2fs_mark_inode_dirty_sync(inode, false);
789 return 0;
790 }
791
f2fs_getattr(const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)792 int f2fs_getattr(const struct path *path, struct kstat *stat,
793 u32 request_mask, unsigned int query_flags)
794 {
795 struct inode *inode = d_inode(path->dentry);
796 struct f2fs_inode_info *fi = F2FS_I(inode);
797 struct f2fs_inode *ri;
798 unsigned int flags;
799
800 if (f2fs_has_extra_attr(inode) &&
801 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
802 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
803 stat->result_mask |= STATX_BTIME;
804 stat->btime.tv_sec = fi->i_crtime.tv_sec;
805 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
806 }
807
808 flags = fi->i_flags;
809 if (flags & F2FS_COMPR_FL)
810 stat->attributes |= STATX_ATTR_COMPRESSED;
811 if (flags & F2FS_APPEND_FL)
812 stat->attributes |= STATX_ATTR_APPEND;
813 if (IS_ENCRYPTED(inode))
814 stat->attributes |= STATX_ATTR_ENCRYPTED;
815 if (flags & F2FS_IMMUTABLE_FL)
816 stat->attributes |= STATX_ATTR_IMMUTABLE;
817 if (flags & F2FS_NODUMP_FL)
818 stat->attributes |= STATX_ATTR_NODUMP;
819 if (IS_VERITY(inode))
820 stat->attributes |= STATX_ATTR_VERITY;
821
822 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
823 STATX_ATTR_APPEND |
824 STATX_ATTR_ENCRYPTED |
825 STATX_ATTR_IMMUTABLE |
826 STATX_ATTR_NODUMP |
827 STATX_ATTR_VERITY);
828
829 generic_fillattr(inode, stat);
830
831 /* we need to show initial sectors used for inline_data/dentries */
832 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
833 f2fs_has_inline_dentry(inode))
834 stat->blocks += (stat->size + 511) >> 9;
835
836 return 0;
837 }
838
839 #ifdef CONFIG_F2FS_FS_POSIX_ACL
__setattr_copy(struct inode * inode,const struct iattr * attr)840 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
841 {
842 unsigned int ia_valid = attr->ia_valid;
843
844 if (ia_valid & ATTR_UID)
845 inode->i_uid = attr->ia_uid;
846 if (ia_valid & ATTR_GID)
847 inode->i_gid = attr->ia_gid;
848 if (ia_valid & ATTR_ATIME)
849 inode->i_atime = attr->ia_atime;
850 if (ia_valid & ATTR_MTIME)
851 inode->i_mtime = attr->ia_mtime;
852 if (ia_valid & ATTR_CTIME)
853 inode->i_ctime = attr->ia_ctime;
854 if (ia_valid & ATTR_MODE) {
855 umode_t mode = attr->ia_mode;
856
857 if (!in_group_p(inode->i_gid) &&
858 !capable_wrt_inode_uidgid(inode, CAP_FSETID))
859 mode &= ~S_ISGID;
860 set_acl_inode(inode, mode);
861 }
862 }
863 #else
864 #define __setattr_copy setattr_copy
865 #endif
866
f2fs_setattr(struct dentry * dentry,struct iattr * attr)867 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
868 {
869 struct inode *inode = d_inode(dentry);
870 int err;
871
872 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
873 return -EIO;
874
875 if (unlikely(IS_IMMUTABLE(inode)))
876 return -EPERM;
877
878 if (unlikely(IS_APPEND(inode) &&
879 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
880 ATTR_GID | ATTR_TIMES_SET))))
881 return -EPERM;
882
883 if ((attr->ia_valid & ATTR_SIZE) &&
884 !f2fs_is_compress_backend_ready(inode))
885 return -EOPNOTSUPP;
886
887 err = setattr_prepare(dentry, attr);
888 if (err)
889 return err;
890
891 err = fscrypt_prepare_setattr(dentry, attr);
892 if (err)
893 return err;
894
895 err = fsverity_prepare_setattr(dentry, attr);
896 if (err)
897 return err;
898
899 if (is_quota_modification(inode, attr)) {
900 err = dquot_initialize(inode);
901 if (err)
902 return err;
903 }
904 if ((attr->ia_valid & ATTR_UID &&
905 !uid_eq(attr->ia_uid, inode->i_uid)) ||
906 (attr->ia_valid & ATTR_GID &&
907 !gid_eq(attr->ia_gid, inode->i_gid))) {
908 f2fs_lock_op(F2FS_I_SB(inode));
909 err = dquot_transfer(inode, attr);
910 if (err) {
911 set_sbi_flag(F2FS_I_SB(inode),
912 SBI_QUOTA_NEED_REPAIR);
913 f2fs_unlock_op(F2FS_I_SB(inode));
914 return err;
915 }
916 /*
917 * update uid/gid under lock_op(), so that dquot and inode can
918 * be updated atomically.
919 */
920 if (attr->ia_valid & ATTR_UID)
921 inode->i_uid = attr->ia_uid;
922 if (attr->ia_valid & ATTR_GID)
923 inode->i_gid = attr->ia_gid;
924 f2fs_mark_inode_dirty_sync(inode, true);
925 f2fs_unlock_op(F2FS_I_SB(inode));
926 }
927
928 if (attr->ia_valid & ATTR_SIZE) {
929 loff_t old_size = i_size_read(inode);
930
931 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
932 /*
933 * should convert inline inode before i_size_write to
934 * keep smaller than inline_data size with inline flag.
935 */
936 err = f2fs_convert_inline_inode(inode);
937 if (err)
938 return err;
939 }
940
941 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
942 down_write(&F2FS_I(inode)->i_mmap_sem);
943
944 truncate_setsize(inode, attr->ia_size);
945
946 if (attr->ia_size <= old_size)
947 err = f2fs_truncate(inode);
948 /*
949 * do not trim all blocks after i_size if target size is
950 * larger than i_size.
951 */
952 up_write(&F2FS_I(inode)->i_mmap_sem);
953 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
954 if (err)
955 return err;
956
957 spin_lock(&F2FS_I(inode)->i_size_lock);
958 inode->i_mtime = inode->i_ctime = current_time(inode);
959 F2FS_I(inode)->last_disk_size = i_size_read(inode);
960 spin_unlock(&F2FS_I(inode)->i_size_lock);
961 }
962
963 __setattr_copy(inode, attr);
964
965 if (attr->ia_valid & ATTR_MODE) {
966 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
967 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
968 inode->i_mode = F2FS_I(inode)->i_acl_mode;
969 clear_inode_flag(inode, FI_ACL_MODE);
970 }
971 }
972
973 /* file size may changed here */
974 f2fs_mark_inode_dirty_sync(inode, true);
975
976 /* inode change will produce dirty node pages flushed by checkpoint */
977 f2fs_balance_fs(F2FS_I_SB(inode), true);
978
979 return err;
980 }
981
982 const struct inode_operations f2fs_file_inode_operations = {
983 .getattr = f2fs_getattr,
984 .setattr = f2fs_setattr,
985 .get_acl = f2fs_get_acl,
986 .set_acl = f2fs_set_acl,
987 .listxattr = f2fs_listxattr,
988 .fiemap = f2fs_fiemap,
989 };
990
fill_zero(struct inode * inode,pgoff_t index,loff_t start,loff_t len)991 static int fill_zero(struct inode *inode, pgoff_t index,
992 loff_t start, loff_t len)
993 {
994 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
995 struct page *page;
996
997 if (!len)
998 return 0;
999
1000 f2fs_balance_fs(sbi, true);
1001
1002 f2fs_lock_op(sbi);
1003 page = f2fs_get_new_data_page(inode, NULL, index, false);
1004 f2fs_unlock_op(sbi);
1005
1006 if (IS_ERR(page))
1007 return PTR_ERR(page);
1008
1009 f2fs_wait_on_page_writeback(page, DATA, true, true);
1010 zero_user(page, start, len);
1011 set_page_dirty(page);
1012 f2fs_put_page(page, 1);
1013 return 0;
1014 }
1015
f2fs_truncate_hole(struct inode * inode,pgoff_t pg_start,pgoff_t pg_end)1016 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1017 {
1018 int err;
1019
1020 while (pg_start < pg_end) {
1021 struct dnode_of_data dn;
1022 pgoff_t end_offset, count;
1023
1024 set_new_dnode(&dn, inode, NULL, NULL, 0);
1025 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1026 if (err) {
1027 if (err == -ENOENT) {
1028 pg_start = f2fs_get_next_page_offset(&dn,
1029 pg_start);
1030 continue;
1031 }
1032 return err;
1033 }
1034
1035 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1036 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1037
1038 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1039
1040 f2fs_truncate_data_blocks_range(&dn, count);
1041 f2fs_put_dnode(&dn);
1042
1043 pg_start += count;
1044 }
1045 return 0;
1046 }
1047
punch_hole(struct inode * inode,loff_t offset,loff_t len)1048 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1049 {
1050 pgoff_t pg_start, pg_end;
1051 loff_t off_start, off_end;
1052 int ret;
1053
1054 ret = f2fs_convert_inline_inode(inode);
1055 if (ret)
1056 return ret;
1057
1058 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1059 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1060
1061 off_start = offset & (PAGE_SIZE - 1);
1062 off_end = (offset + len) & (PAGE_SIZE - 1);
1063
1064 if (pg_start == pg_end) {
1065 ret = fill_zero(inode, pg_start, off_start,
1066 off_end - off_start);
1067 if (ret)
1068 return ret;
1069 } else {
1070 if (off_start) {
1071 ret = fill_zero(inode, pg_start++, off_start,
1072 PAGE_SIZE - off_start);
1073 if (ret)
1074 return ret;
1075 }
1076 if (off_end) {
1077 ret = fill_zero(inode, pg_end, 0, off_end);
1078 if (ret)
1079 return ret;
1080 }
1081
1082 if (pg_start < pg_end) {
1083 loff_t blk_start, blk_end;
1084 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1085
1086 f2fs_balance_fs(sbi, true);
1087
1088 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1089 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1090
1091 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1092 down_write(&F2FS_I(inode)->i_mmap_sem);
1093
1094 truncate_pagecache_range(inode, blk_start, blk_end - 1);
1095
1096 f2fs_lock_op(sbi);
1097 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1098 f2fs_unlock_op(sbi);
1099
1100 up_write(&F2FS_I(inode)->i_mmap_sem);
1101 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1102 }
1103 }
1104
1105 return ret;
1106 }
1107
__read_out_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,pgoff_t len)1108 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1109 int *do_replace, pgoff_t off, pgoff_t len)
1110 {
1111 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1112 struct dnode_of_data dn;
1113 int ret, done, i;
1114
1115 next_dnode:
1116 set_new_dnode(&dn, inode, NULL, NULL, 0);
1117 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1118 if (ret && ret != -ENOENT) {
1119 return ret;
1120 } else if (ret == -ENOENT) {
1121 if (dn.max_level == 0)
1122 return -ENOENT;
1123 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1124 dn.ofs_in_node, len);
1125 blkaddr += done;
1126 do_replace += done;
1127 goto next;
1128 }
1129
1130 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1131 dn.ofs_in_node, len);
1132 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1133 *blkaddr = f2fs_data_blkaddr(&dn);
1134
1135 if (__is_valid_data_blkaddr(*blkaddr) &&
1136 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1137 DATA_GENERIC_ENHANCE)) {
1138 f2fs_put_dnode(&dn);
1139 return -EFSCORRUPTED;
1140 }
1141
1142 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1143
1144 if (f2fs_lfs_mode(sbi)) {
1145 f2fs_put_dnode(&dn);
1146 return -EOPNOTSUPP;
1147 }
1148
1149 /* do not invalidate this block address */
1150 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1151 *do_replace = 1;
1152 }
1153 }
1154 f2fs_put_dnode(&dn);
1155 next:
1156 len -= done;
1157 off += done;
1158 if (len)
1159 goto next_dnode;
1160 return 0;
1161 }
1162
__roll_back_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,int len)1163 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1164 int *do_replace, pgoff_t off, int len)
1165 {
1166 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1167 struct dnode_of_data dn;
1168 int ret, i;
1169
1170 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1171 if (*do_replace == 0)
1172 continue;
1173
1174 set_new_dnode(&dn, inode, NULL, NULL, 0);
1175 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1176 if (ret) {
1177 dec_valid_block_count(sbi, inode, 1);
1178 f2fs_invalidate_blocks(sbi, *blkaddr);
1179 } else {
1180 f2fs_update_data_blkaddr(&dn, *blkaddr);
1181 }
1182 f2fs_put_dnode(&dn);
1183 }
1184 return 0;
1185 }
1186
__clone_blkaddrs(struct inode * src_inode,struct inode * dst_inode,block_t * blkaddr,int * do_replace,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1187 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1188 block_t *blkaddr, int *do_replace,
1189 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1190 {
1191 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1192 pgoff_t i = 0;
1193 int ret;
1194
1195 while (i < len) {
1196 if (blkaddr[i] == NULL_ADDR && !full) {
1197 i++;
1198 continue;
1199 }
1200
1201 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1202 struct dnode_of_data dn;
1203 struct node_info ni;
1204 size_t new_size;
1205 pgoff_t ilen;
1206
1207 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1208 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1209 if (ret)
1210 return ret;
1211
1212 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1213 if (ret) {
1214 f2fs_put_dnode(&dn);
1215 return ret;
1216 }
1217
1218 ilen = min((pgoff_t)
1219 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1220 dn.ofs_in_node, len - i);
1221 do {
1222 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1223 f2fs_truncate_data_blocks_range(&dn, 1);
1224
1225 if (do_replace[i]) {
1226 f2fs_i_blocks_write(src_inode,
1227 1, false, false);
1228 f2fs_i_blocks_write(dst_inode,
1229 1, true, false);
1230 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1231 blkaddr[i], ni.version, true, false);
1232
1233 do_replace[i] = 0;
1234 }
1235 dn.ofs_in_node++;
1236 i++;
1237 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1238 if (dst_inode->i_size < new_size)
1239 f2fs_i_size_write(dst_inode, new_size);
1240 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1241
1242 f2fs_put_dnode(&dn);
1243 } else {
1244 struct page *psrc, *pdst;
1245
1246 psrc = f2fs_get_lock_data_page(src_inode,
1247 src + i, true);
1248 if (IS_ERR(psrc))
1249 return PTR_ERR(psrc);
1250 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1251 true);
1252 if (IS_ERR(pdst)) {
1253 f2fs_put_page(psrc, 1);
1254 return PTR_ERR(pdst);
1255 }
1256 f2fs_copy_page(psrc, pdst);
1257 set_page_dirty(pdst);
1258 f2fs_put_page(pdst, 1);
1259 f2fs_put_page(psrc, 1);
1260
1261 ret = f2fs_truncate_hole(src_inode,
1262 src + i, src + i + 1);
1263 if (ret)
1264 return ret;
1265 i++;
1266 }
1267 }
1268 return 0;
1269 }
1270
__exchange_data_block(struct inode * src_inode,struct inode * dst_inode,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1271 static int __exchange_data_block(struct inode *src_inode,
1272 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1273 pgoff_t len, bool full)
1274 {
1275 block_t *src_blkaddr;
1276 int *do_replace;
1277 pgoff_t olen;
1278 int ret;
1279
1280 while (len) {
1281 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1282
1283 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1284 array_size(olen, sizeof(block_t)),
1285 GFP_NOFS);
1286 if (!src_blkaddr)
1287 return -ENOMEM;
1288
1289 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1290 array_size(olen, sizeof(int)),
1291 GFP_NOFS);
1292 if (!do_replace) {
1293 kvfree(src_blkaddr);
1294 return -ENOMEM;
1295 }
1296
1297 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1298 do_replace, src, olen);
1299 if (ret)
1300 goto roll_back;
1301
1302 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1303 do_replace, src, dst, olen, full);
1304 if (ret)
1305 goto roll_back;
1306
1307 src += olen;
1308 dst += olen;
1309 len -= olen;
1310
1311 kvfree(src_blkaddr);
1312 kvfree(do_replace);
1313 }
1314 return 0;
1315
1316 roll_back:
1317 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1318 kvfree(src_blkaddr);
1319 kvfree(do_replace);
1320 return ret;
1321 }
1322
f2fs_do_collapse(struct inode * inode,loff_t offset,loff_t len)1323 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1324 {
1325 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1326 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1327 pgoff_t start = offset >> PAGE_SHIFT;
1328 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1329 int ret;
1330
1331 f2fs_balance_fs(sbi, true);
1332
1333 /* avoid gc operation during block exchange */
1334 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1335 down_write(&F2FS_I(inode)->i_mmap_sem);
1336
1337 f2fs_lock_op(sbi);
1338 f2fs_drop_extent_tree(inode);
1339 truncate_pagecache(inode, offset);
1340 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1341 f2fs_unlock_op(sbi);
1342
1343 up_write(&F2FS_I(inode)->i_mmap_sem);
1344 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1345 return ret;
1346 }
1347
f2fs_collapse_range(struct inode * inode,loff_t offset,loff_t len)1348 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1349 {
1350 loff_t new_size;
1351 int ret;
1352
1353 if (offset + len >= i_size_read(inode))
1354 return -EINVAL;
1355
1356 /* collapse range should be aligned to block size of f2fs. */
1357 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1358 return -EINVAL;
1359
1360 ret = f2fs_convert_inline_inode(inode);
1361 if (ret)
1362 return ret;
1363
1364 /* write out all dirty pages from offset */
1365 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1366 if (ret)
1367 return ret;
1368
1369 ret = f2fs_do_collapse(inode, offset, len);
1370 if (ret)
1371 return ret;
1372
1373 /* write out all moved pages, if possible */
1374 down_write(&F2FS_I(inode)->i_mmap_sem);
1375 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1376 truncate_pagecache(inode, offset);
1377
1378 new_size = i_size_read(inode) - len;
1379 ret = f2fs_truncate_blocks(inode, new_size, true);
1380 up_write(&F2FS_I(inode)->i_mmap_sem);
1381 if (!ret)
1382 f2fs_i_size_write(inode, new_size);
1383 return ret;
1384 }
1385
f2fs_do_zero_range(struct dnode_of_data * dn,pgoff_t start,pgoff_t end)1386 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1387 pgoff_t end)
1388 {
1389 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1390 pgoff_t index = start;
1391 unsigned int ofs_in_node = dn->ofs_in_node;
1392 blkcnt_t count = 0;
1393 int ret;
1394
1395 for (; index < end; index++, dn->ofs_in_node++) {
1396 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1397 count++;
1398 }
1399
1400 dn->ofs_in_node = ofs_in_node;
1401 ret = f2fs_reserve_new_blocks(dn, count);
1402 if (ret)
1403 return ret;
1404
1405 dn->ofs_in_node = ofs_in_node;
1406 for (index = start; index < end; index++, dn->ofs_in_node++) {
1407 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1408 /*
1409 * f2fs_reserve_new_blocks will not guarantee entire block
1410 * allocation.
1411 */
1412 if (dn->data_blkaddr == NULL_ADDR) {
1413 ret = -ENOSPC;
1414 break;
1415 }
1416 if (dn->data_blkaddr != NEW_ADDR) {
1417 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1418 dn->data_blkaddr = NEW_ADDR;
1419 f2fs_set_data_blkaddr(dn);
1420 }
1421 }
1422
1423 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1424
1425 return ret;
1426 }
1427
f2fs_zero_range(struct inode * inode,loff_t offset,loff_t len,int mode)1428 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1429 int mode)
1430 {
1431 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1432 struct address_space *mapping = inode->i_mapping;
1433 pgoff_t index, pg_start, pg_end;
1434 loff_t new_size = i_size_read(inode);
1435 loff_t off_start, off_end;
1436 int ret = 0;
1437
1438 ret = inode_newsize_ok(inode, (len + offset));
1439 if (ret)
1440 return ret;
1441
1442 ret = f2fs_convert_inline_inode(inode);
1443 if (ret)
1444 return ret;
1445
1446 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1447 if (ret)
1448 return ret;
1449
1450 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1451 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1452
1453 off_start = offset & (PAGE_SIZE - 1);
1454 off_end = (offset + len) & (PAGE_SIZE - 1);
1455
1456 if (pg_start == pg_end) {
1457 ret = fill_zero(inode, pg_start, off_start,
1458 off_end - off_start);
1459 if (ret)
1460 return ret;
1461
1462 new_size = max_t(loff_t, new_size, offset + len);
1463 } else {
1464 if (off_start) {
1465 ret = fill_zero(inode, pg_start++, off_start,
1466 PAGE_SIZE - off_start);
1467 if (ret)
1468 return ret;
1469
1470 new_size = max_t(loff_t, new_size,
1471 (loff_t)pg_start << PAGE_SHIFT);
1472 }
1473
1474 for (index = pg_start; index < pg_end;) {
1475 struct dnode_of_data dn;
1476 unsigned int end_offset;
1477 pgoff_t end;
1478
1479 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1480 down_write(&F2FS_I(inode)->i_mmap_sem);
1481
1482 truncate_pagecache_range(inode,
1483 (loff_t)index << PAGE_SHIFT,
1484 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1485
1486 f2fs_lock_op(sbi);
1487
1488 set_new_dnode(&dn, inode, NULL, NULL, 0);
1489 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1490 if (ret) {
1491 f2fs_unlock_op(sbi);
1492 up_write(&F2FS_I(inode)->i_mmap_sem);
1493 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1494 goto out;
1495 }
1496
1497 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1498 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1499
1500 ret = f2fs_do_zero_range(&dn, index, end);
1501 f2fs_put_dnode(&dn);
1502
1503 f2fs_unlock_op(sbi);
1504 up_write(&F2FS_I(inode)->i_mmap_sem);
1505 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1506
1507 f2fs_balance_fs(sbi, dn.node_changed);
1508
1509 if (ret)
1510 goto out;
1511
1512 index = end;
1513 new_size = max_t(loff_t, new_size,
1514 (loff_t)index << PAGE_SHIFT);
1515 }
1516
1517 if (off_end) {
1518 ret = fill_zero(inode, pg_end, 0, off_end);
1519 if (ret)
1520 goto out;
1521
1522 new_size = max_t(loff_t, new_size, offset + len);
1523 }
1524 }
1525
1526 out:
1527 if (new_size > i_size_read(inode)) {
1528 if (mode & FALLOC_FL_KEEP_SIZE)
1529 file_set_keep_isize(inode);
1530 else
1531 f2fs_i_size_write(inode, new_size);
1532 }
1533 return ret;
1534 }
1535
f2fs_insert_range(struct inode * inode,loff_t offset,loff_t len)1536 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1537 {
1538 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1539 pgoff_t nr, pg_start, pg_end, delta, idx;
1540 loff_t new_size;
1541 int ret = 0;
1542
1543 new_size = i_size_read(inode) + len;
1544 ret = inode_newsize_ok(inode, new_size);
1545 if (ret)
1546 return ret;
1547
1548 if (offset >= i_size_read(inode))
1549 return -EINVAL;
1550
1551 /* insert range should be aligned to block size of f2fs. */
1552 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1553 return -EINVAL;
1554
1555 ret = f2fs_convert_inline_inode(inode);
1556 if (ret)
1557 return ret;
1558
1559 f2fs_balance_fs(sbi, true);
1560
1561 down_write(&F2FS_I(inode)->i_mmap_sem);
1562 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1563 up_write(&F2FS_I(inode)->i_mmap_sem);
1564 if (ret)
1565 return ret;
1566
1567 /* write out all dirty pages from offset */
1568 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1569 if (ret)
1570 return ret;
1571
1572 pg_start = offset >> PAGE_SHIFT;
1573 pg_end = (offset + len) >> PAGE_SHIFT;
1574 delta = pg_end - pg_start;
1575 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1576
1577 /* avoid gc operation during block exchange */
1578 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1579 down_write(&F2FS_I(inode)->i_mmap_sem);
1580 truncate_pagecache(inode, offset);
1581
1582 while (!ret && idx > pg_start) {
1583 nr = idx - pg_start;
1584 if (nr > delta)
1585 nr = delta;
1586 idx -= nr;
1587
1588 f2fs_lock_op(sbi);
1589 f2fs_drop_extent_tree(inode);
1590
1591 ret = __exchange_data_block(inode, inode, idx,
1592 idx + delta, nr, false);
1593 f2fs_unlock_op(sbi);
1594 }
1595 up_write(&F2FS_I(inode)->i_mmap_sem);
1596 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1597
1598 /* write out all moved pages, if possible */
1599 down_write(&F2FS_I(inode)->i_mmap_sem);
1600 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1601 truncate_pagecache(inode, offset);
1602 up_write(&F2FS_I(inode)->i_mmap_sem);
1603
1604 if (!ret)
1605 f2fs_i_size_write(inode, new_size);
1606 return ret;
1607 }
1608
expand_inode_data(struct inode * inode,loff_t offset,loff_t len,int mode)1609 static int expand_inode_data(struct inode *inode, loff_t offset,
1610 loff_t len, int mode)
1611 {
1612 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1613 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1614 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1615 .m_may_create = true };
1616 pgoff_t pg_start, pg_end;
1617 loff_t new_size = i_size_read(inode);
1618 loff_t off_end;
1619 block_t expanded = 0;
1620 int err;
1621
1622 err = inode_newsize_ok(inode, (len + offset));
1623 if (err)
1624 return err;
1625
1626 err = f2fs_convert_inline_inode(inode);
1627 if (err)
1628 return err;
1629
1630 f2fs_balance_fs(sbi, true);
1631
1632 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1633 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1634 off_end = (offset + len) & (PAGE_SIZE - 1);
1635
1636 map.m_lblk = pg_start;
1637 map.m_len = pg_end - pg_start;
1638 if (off_end)
1639 map.m_len++;
1640
1641 if (!map.m_len)
1642 return 0;
1643
1644 if (f2fs_is_pinned_file(inode)) {
1645 block_t sec_blks = BLKS_PER_SEC(sbi);
1646 block_t sec_len = roundup(map.m_len, sec_blks);
1647
1648 map.m_len = sec_blks;
1649 next_alloc:
1650 if (has_not_enough_free_secs(sbi, 0,
1651 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1652 down_write(&sbi->gc_lock);
1653 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1654 if (err && err != -ENODATA && err != -EAGAIN)
1655 goto out_err;
1656 }
1657
1658 down_write(&sbi->pin_sem);
1659
1660 f2fs_lock_op(sbi);
1661 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED);
1662 f2fs_unlock_op(sbi);
1663
1664 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1665 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1666
1667 up_write(&sbi->pin_sem);
1668
1669 expanded += map.m_len;
1670 sec_len -= map.m_len;
1671 map.m_lblk += map.m_len;
1672 if (!err && sec_len)
1673 goto next_alloc;
1674
1675 map.m_len = expanded;
1676 } else {
1677 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1678 expanded = map.m_len;
1679 }
1680 out_err:
1681 if (err) {
1682 pgoff_t last_off;
1683
1684 if (!expanded)
1685 return err;
1686
1687 last_off = pg_start + expanded - 1;
1688
1689 /* update new size to the failed position */
1690 new_size = (last_off == pg_end) ? offset + len :
1691 (loff_t)(last_off + 1) << PAGE_SHIFT;
1692 } else {
1693 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1694 }
1695
1696 if (new_size > i_size_read(inode)) {
1697 if (mode & FALLOC_FL_KEEP_SIZE)
1698 file_set_keep_isize(inode);
1699 else
1700 f2fs_i_size_write(inode, new_size);
1701 }
1702
1703 return err;
1704 }
1705
f2fs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)1706 static long f2fs_fallocate(struct file *file, int mode,
1707 loff_t offset, loff_t len)
1708 {
1709 struct inode *inode = file_inode(file);
1710 long ret = 0;
1711
1712 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1713 return -EIO;
1714 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1715 return -ENOSPC;
1716 if (!f2fs_is_compress_backend_ready(inode))
1717 return -EOPNOTSUPP;
1718
1719 /* f2fs only support ->fallocate for regular file */
1720 if (!S_ISREG(inode->i_mode))
1721 return -EINVAL;
1722
1723 if (IS_ENCRYPTED(inode) &&
1724 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1725 return -EOPNOTSUPP;
1726
1727 if (f2fs_compressed_file(inode) &&
1728 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1729 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1730 return -EOPNOTSUPP;
1731
1732 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1733 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1734 FALLOC_FL_INSERT_RANGE))
1735 return -EOPNOTSUPP;
1736
1737 inode_lock(inode);
1738
1739 if (mode & FALLOC_FL_PUNCH_HOLE) {
1740 if (offset >= inode->i_size)
1741 goto out;
1742
1743 ret = punch_hole(inode, offset, len);
1744 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1745 ret = f2fs_collapse_range(inode, offset, len);
1746 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1747 ret = f2fs_zero_range(inode, offset, len, mode);
1748 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1749 ret = f2fs_insert_range(inode, offset, len);
1750 } else {
1751 ret = expand_inode_data(inode, offset, len, mode);
1752 }
1753
1754 if (!ret) {
1755 inode->i_mtime = inode->i_ctime = current_time(inode);
1756 f2fs_mark_inode_dirty_sync(inode, false);
1757 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1758 }
1759
1760 out:
1761 inode_unlock(inode);
1762
1763 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1764 return ret;
1765 }
1766
f2fs_release_file(struct inode * inode,struct file * filp)1767 static int f2fs_release_file(struct inode *inode, struct file *filp)
1768 {
1769 /*
1770 * f2fs_relase_file is called at every close calls. So we should
1771 * not drop any inmemory pages by close called by other process.
1772 */
1773 if (!(filp->f_mode & FMODE_WRITE) ||
1774 atomic_read(&inode->i_writecount) != 1)
1775 return 0;
1776
1777 /* some remained atomic pages should discarded */
1778 if (f2fs_is_atomic_file(inode))
1779 f2fs_drop_inmem_pages(inode);
1780 if (f2fs_is_volatile_file(inode)) {
1781 set_inode_flag(inode, FI_DROP_CACHE);
1782 filemap_fdatawrite(inode->i_mapping);
1783 clear_inode_flag(inode, FI_DROP_CACHE);
1784 clear_inode_flag(inode, FI_VOLATILE_FILE);
1785 stat_dec_volatile_write(inode);
1786 }
1787 return 0;
1788 }
1789
f2fs_file_flush(struct file * file,fl_owner_t id)1790 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1791 {
1792 struct inode *inode = file_inode(file);
1793
1794 /*
1795 * If the process doing a transaction is crashed, we should do
1796 * roll-back. Otherwise, other reader/write can see corrupted database
1797 * until all the writers close its file. Since this should be done
1798 * before dropping file lock, it needs to do in ->flush.
1799 */
1800 if (f2fs_is_atomic_file(inode) &&
1801 F2FS_I(inode)->inmem_task == current)
1802 f2fs_drop_inmem_pages(inode);
1803 return 0;
1804 }
1805
f2fs_setflags_common(struct inode * inode,u32 iflags,u32 mask)1806 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1807 {
1808 struct f2fs_inode_info *fi = F2FS_I(inode);
1809 u32 masked_flags = fi->i_flags & mask;
1810
1811 f2fs_bug_on(F2FS_I_SB(inode), (iflags & ~mask));
1812
1813 /* Is it quota file? Do not allow user to mess with it */
1814 if (IS_NOQUOTA(inode))
1815 return -EPERM;
1816
1817 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1818 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1819 return -EOPNOTSUPP;
1820 if (!f2fs_empty_dir(inode))
1821 return -ENOTEMPTY;
1822 }
1823
1824 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1825 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1826 return -EOPNOTSUPP;
1827 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1828 return -EINVAL;
1829 }
1830
1831 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1832 if (masked_flags & F2FS_COMPR_FL) {
1833 if (!f2fs_disable_compressed_file(inode))
1834 return -EINVAL;
1835 }
1836 if (iflags & F2FS_NOCOMP_FL)
1837 return -EINVAL;
1838 if (iflags & F2FS_COMPR_FL) {
1839 if (!f2fs_may_compress(inode))
1840 return -EINVAL;
1841 if (S_ISREG(inode->i_mode) && inode->i_size)
1842 return -EINVAL;
1843
1844 set_compress_context(inode);
1845 }
1846 }
1847 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1848 if (masked_flags & F2FS_COMPR_FL)
1849 return -EINVAL;
1850 }
1851
1852 fi->i_flags = iflags | (fi->i_flags & ~mask);
1853 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1854 (fi->i_flags & F2FS_NOCOMP_FL));
1855
1856 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1857 set_inode_flag(inode, FI_PROJ_INHERIT);
1858 else
1859 clear_inode_flag(inode, FI_PROJ_INHERIT);
1860
1861 inode->i_ctime = current_time(inode);
1862 f2fs_set_inode_flags(inode);
1863 f2fs_mark_inode_dirty_sync(inode, true);
1864 return 0;
1865 }
1866
1867 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1868
1869 /*
1870 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1871 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1872 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1873 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1874 */
1875
1876 static const struct {
1877 u32 iflag;
1878 u32 fsflag;
1879 } f2fs_fsflags_map[] = {
1880 { F2FS_COMPR_FL, FS_COMPR_FL },
1881 { F2FS_SYNC_FL, FS_SYNC_FL },
1882 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1883 { F2FS_APPEND_FL, FS_APPEND_FL },
1884 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1885 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1886 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
1887 { F2FS_INDEX_FL, FS_INDEX_FL },
1888 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1889 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1890 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1891 };
1892
1893 #define F2FS_GETTABLE_FS_FL ( \
1894 FS_COMPR_FL | \
1895 FS_SYNC_FL | \
1896 FS_IMMUTABLE_FL | \
1897 FS_APPEND_FL | \
1898 FS_NODUMP_FL | \
1899 FS_NOATIME_FL | \
1900 FS_NOCOMP_FL | \
1901 FS_INDEX_FL | \
1902 FS_DIRSYNC_FL | \
1903 FS_PROJINHERIT_FL | \
1904 FS_ENCRYPT_FL | \
1905 FS_INLINE_DATA_FL | \
1906 FS_NOCOW_FL | \
1907 FS_VERITY_FL | \
1908 FS_CASEFOLD_FL)
1909
1910 #define F2FS_SETTABLE_FS_FL ( \
1911 FS_COMPR_FL | \
1912 FS_SYNC_FL | \
1913 FS_IMMUTABLE_FL | \
1914 FS_APPEND_FL | \
1915 FS_NODUMP_FL | \
1916 FS_NOATIME_FL | \
1917 FS_NOCOMP_FL | \
1918 FS_DIRSYNC_FL | \
1919 FS_PROJINHERIT_FL | \
1920 FS_CASEFOLD_FL)
1921
1922 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
f2fs_iflags_to_fsflags(u32 iflags)1923 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1924 {
1925 u32 fsflags = 0;
1926 int i;
1927
1928 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1929 if (iflags & f2fs_fsflags_map[i].iflag)
1930 fsflags |= f2fs_fsflags_map[i].fsflag;
1931
1932 return fsflags;
1933 }
1934
1935 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
f2fs_fsflags_to_iflags(u32 fsflags)1936 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1937 {
1938 u32 iflags = 0;
1939 int i;
1940
1941 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1942 if (fsflags & f2fs_fsflags_map[i].fsflag)
1943 iflags |= f2fs_fsflags_map[i].iflag;
1944
1945 return iflags;
1946 }
1947
f2fs_ioc_getflags(struct file * filp,unsigned long arg)1948 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1949 {
1950 struct inode *inode = file_inode(filp);
1951 struct f2fs_inode_info *fi = F2FS_I(inode);
1952 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1953
1954 if (IS_ENCRYPTED(inode))
1955 fsflags |= FS_ENCRYPT_FL;
1956 if (IS_VERITY(inode))
1957 fsflags |= FS_VERITY_FL;
1958 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1959 fsflags |= FS_INLINE_DATA_FL;
1960 if (is_inode_flag_set(inode, FI_PIN_FILE))
1961 fsflags |= FS_NOCOW_FL;
1962
1963 fsflags &= F2FS_GETTABLE_FS_FL;
1964
1965 return put_user(fsflags, (int __user *)arg);
1966 }
1967
f2fs_ioc_setflags(struct file * filp,unsigned long arg)1968 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1969 {
1970 struct inode *inode = file_inode(filp);
1971 struct f2fs_inode_info *fi = F2FS_I(inode);
1972 u32 fsflags, old_fsflags;
1973 u32 iflags;
1974 int ret;
1975
1976 if (!inode_owner_or_capable(inode))
1977 return -EACCES;
1978
1979 if (get_user(fsflags, (int __user *)arg))
1980 return -EFAULT;
1981
1982 if (fsflags & ~F2FS_GETTABLE_FS_FL)
1983 return -EOPNOTSUPP;
1984 fsflags &= F2FS_SETTABLE_FS_FL;
1985
1986 iflags = f2fs_fsflags_to_iflags(fsflags);
1987 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1988 return -EOPNOTSUPP;
1989
1990 ret = mnt_want_write_file(filp);
1991 if (ret)
1992 return ret;
1993
1994 inode_lock(inode);
1995
1996 old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1997 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1998 if (ret)
1999 goto out;
2000
2001 ret = f2fs_setflags_common(inode, iflags,
2002 f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
2003 out:
2004 inode_unlock(inode);
2005 mnt_drop_write_file(filp);
2006 return ret;
2007 }
2008
f2fs_ioc_getversion(struct file * filp,unsigned long arg)2009 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2010 {
2011 struct inode *inode = file_inode(filp);
2012
2013 return put_user(inode->i_generation, (int __user *)arg);
2014 }
2015
f2fs_ioc_start_atomic_write(struct file * filp)2016 static int f2fs_ioc_start_atomic_write(struct file *filp)
2017 {
2018 struct inode *inode = file_inode(filp);
2019 struct f2fs_inode_info *fi = F2FS_I(inode);
2020 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2021 int ret;
2022
2023 if (!inode_owner_or_capable(inode))
2024 return -EACCES;
2025
2026 if (!S_ISREG(inode->i_mode))
2027 return -EINVAL;
2028
2029 if (filp->f_flags & O_DIRECT)
2030 return -EINVAL;
2031
2032 ret = mnt_want_write_file(filp);
2033 if (ret)
2034 return ret;
2035
2036 inode_lock(inode);
2037
2038 f2fs_disable_compressed_file(inode);
2039
2040 if (f2fs_is_atomic_file(inode)) {
2041 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2042 ret = -EINVAL;
2043 goto out;
2044 }
2045
2046 ret = f2fs_convert_inline_inode(inode);
2047 if (ret)
2048 goto out;
2049
2050 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2051
2052 /*
2053 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2054 * f2fs_is_atomic_file.
2055 */
2056 if (get_dirty_pages(inode))
2057 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2058 inode->i_ino, get_dirty_pages(inode));
2059 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2060 if (ret) {
2061 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2062 goto out;
2063 }
2064
2065 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2066 if (list_empty(&fi->inmem_ilist))
2067 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2068 sbi->atomic_files++;
2069 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2070
2071 /* add inode in inmem_list first and set atomic_file */
2072 set_inode_flag(inode, FI_ATOMIC_FILE);
2073 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2074 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2075
2076 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2077 F2FS_I(inode)->inmem_task = current;
2078 stat_update_max_atomic_write(inode);
2079 out:
2080 inode_unlock(inode);
2081 mnt_drop_write_file(filp);
2082 return ret;
2083 }
2084
f2fs_ioc_commit_atomic_write(struct file * filp)2085 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2086 {
2087 struct inode *inode = file_inode(filp);
2088 int ret;
2089
2090 if (!inode_owner_or_capable(inode))
2091 return -EACCES;
2092
2093 ret = mnt_want_write_file(filp);
2094 if (ret)
2095 return ret;
2096
2097 f2fs_balance_fs(F2FS_I_SB(inode), true);
2098
2099 inode_lock(inode);
2100
2101 if (f2fs_is_volatile_file(inode)) {
2102 ret = -EINVAL;
2103 goto err_out;
2104 }
2105
2106 if (f2fs_is_atomic_file(inode)) {
2107 ret = f2fs_commit_inmem_pages(inode);
2108 if (ret)
2109 goto err_out;
2110
2111 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2112 if (!ret)
2113 f2fs_drop_inmem_pages(inode);
2114 } else {
2115 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2116 }
2117 err_out:
2118 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2119 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2120 ret = -EINVAL;
2121 }
2122 inode_unlock(inode);
2123 mnt_drop_write_file(filp);
2124 return ret;
2125 }
2126
f2fs_ioc_start_volatile_write(struct file * filp)2127 static int f2fs_ioc_start_volatile_write(struct file *filp)
2128 {
2129 struct inode *inode = file_inode(filp);
2130 int ret;
2131
2132 if (!inode_owner_or_capable(inode))
2133 return -EACCES;
2134
2135 if (!S_ISREG(inode->i_mode))
2136 return -EINVAL;
2137
2138 ret = mnt_want_write_file(filp);
2139 if (ret)
2140 return ret;
2141
2142 inode_lock(inode);
2143
2144 if (f2fs_is_volatile_file(inode))
2145 goto out;
2146
2147 ret = f2fs_convert_inline_inode(inode);
2148 if (ret)
2149 goto out;
2150
2151 stat_inc_volatile_write(inode);
2152 stat_update_max_volatile_write(inode);
2153
2154 set_inode_flag(inode, FI_VOLATILE_FILE);
2155 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2156 out:
2157 inode_unlock(inode);
2158 mnt_drop_write_file(filp);
2159 return ret;
2160 }
2161
f2fs_ioc_release_volatile_write(struct file * filp)2162 static int f2fs_ioc_release_volatile_write(struct file *filp)
2163 {
2164 struct inode *inode = file_inode(filp);
2165 int ret;
2166
2167 if (!inode_owner_or_capable(inode))
2168 return -EACCES;
2169
2170 ret = mnt_want_write_file(filp);
2171 if (ret)
2172 return ret;
2173
2174 inode_lock(inode);
2175
2176 if (!f2fs_is_volatile_file(inode))
2177 goto out;
2178
2179 if (!f2fs_is_first_block_written(inode)) {
2180 ret = truncate_partial_data_page(inode, 0, true);
2181 goto out;
2182 }
2183
2184 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2185 out:
2186 inode_unlock(inode);
2187 mnt_drop_write_file(filp);
2188 return ret;
2189 }
2190
f2fs_ioc_abort_volatile_write(struct file * filp)2191 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2192 {
2193 struct inode *inode = file_inode(filp);
2194 int ret;
2195
2196 if (!inode_owner_or_capable(inode))
2197 return -EACCES;
2198
2199 ret = mnt_want_write_file(filp);
2200 if (ret)
2201 return ret;
2202
2203 inode_lock(inode);
2204
2205 if (f2fs_is_atomic_file(inode))
2206 f2fs_drop_inmem_pages(inode);
2207 if (f2fs_is_volatile_file(inode)) {
2208 clear_inode_flag(inode, FI_VOLATILE_FILE);
2209 stat_dec_volatile_write(inode);
2210 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2211 }
2212
2213 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2214
2215 inode_unlock(inode);
2216
2217 mnt_drop_write_file(filp);
2218 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2219 return ret;
2220 }
2221
f2fs_ioc_shutdown(struct file * filp,unsigned long arg)2222 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2223 {
2224 struct inode *inode = file_inode(filp);
2225 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2226 struct super_block *sb = sbi->sb;
2227 __u32 in;
2228 int ret = 0;
2229
2230 if (!capable(CAP_SYS_ADMIN))
2231 return -EPERM;
2232
2233 if (get_user(in, (__u32 __user *)arg))
2234 return -EFAULT;
2235
2236 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2237 ret = mnt_want_write_file(filp);
2238 if (ret) {
2239 if (ret == -EROFS) {
2240 ret = 0;
2241 f2fs_stop_checkpoint(sbi, false);
2242 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2243 trace_f2fs_shutdown(sbi, in, ret);
2244 }
2245 return ret;
2246 }
2247 }
2248
2249 switch (in) {
2250 case F2FS_GOING_DOWN_FULLSYNC:
2251 sb = freeze_bdev(sb->s_bdev);
2252 if (IS_ERR(sb)) {
2253 ret = PTR_ERR(sb);
2254 goto out;
2255 }
2256 if (sb) {
2257 f2fs_stop_checkpoint(sbi, false);
2258 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2259 thaw_bdev(sb->s_bdev, sb);
2260 }
2261 break;
2262 case F2FS_GOING_DOWN_METASYNC:
2263 /* do checkpoint only */
2264 ret = f2fs_sync_fs(sb, 1);
2265 if (ret)
2266 goto out;
2267 f2fs_stop_checkpoint(sbi, false);
2268 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2269 break;
2270 case F2FS_GOING_DOWN_NOSYNC:
2271 f2fs_stop_checkpoint(sbi, false);
2272 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2273 break;
2274 case F2FS_GOING_DOWN_METAFLUSH:
2275 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2276 f2fs_stop_checkpoint(sbi, false);
2277 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2278 break;
2279 case F2FS_GOING_DOWN_NEED_FSCK:
2280 set_sbi_flag(sbi, SBI_NEED_FSCK);
2281 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2282 set_sbi_flag(sbi, SBI_IS_DIRTY);
2283 /* do checkpoint only */
2284 ret = f2fs_sync_fs(sb, 1);
2285 goto out;
2286 default:
2287 ret = -EINVAL;
2288 goto out;
2289 }
2290
2291 f2fs_stop_gc_thread(sbi);
2292 f2fs_stop_discard_thread(sbi);
2293
2294 f2fs_drop_discard_cmd(sbi);
2295 clear_opt(sbi, DISCARD);
2296
2297 f2fs_update_time(sbi, REQ_TIME);
2298 out:
2299 if (in != F2FS_GOING_DOWN_FULLSYNC)
2300 mnt_drop_write_file(filp);
2301
2302 trace_f2fs_shutdown(sbi, in, ret);
2303
2304 return ret;
2305 }
2306
f2fs_ioc_fitrim(struct file * filp,unsigned long arg)2307 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2308 {
2309 struct inode *inode = file_inode(filp);
2310 struct super_block *sb = inode->i_sb;
2311 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2312 struct fstrim_range range;
2313 int ret;
2314
2315 if (!capable(CAP_SYS_ADMIN))
2316 return -EPERM;
2317
2318 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2319 return -EOPNOTSUPP;
2320
2321 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2322 sizeof(range)))
2323 return -EFAULT;
2324
2325 ret = mnt_want_write_file(filp);
2326 if (ret)
2327 return ret;
2328
2329 range.minlen = max((unsigned int)range.minlen,
2330 q->limits.discard_granularity);
2331 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2332 mnt_drop_write_file(filp);
2333 if (ret < 0)
2334 return ret;
2335
2336 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2337 sizeof(range)))
2338 return -EFAULT;
2339 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2340 return 0;
2341 }
2342
uuid_is_nonzero(__u8 u[16])2343 static bool uuid_is_nonzero(__u8 u[16])
2344 {
2345 int i;
2346
2347 for (i = 0; i < 16; i++)
2348 if (u[i])
2349 return true;
2350 return false;
2351 }
2352
f2fs_ioc_set_encryption_policy(struct file * filp,unsigned long arg)2353 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2354 {
2355 struct inode *inode = file_inode(filp);
2356
2357 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2358 return -EOPNOTSUPP;
2359
2360 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2361
2362 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2363 }
2364
f2fs_ioc_get_encryption_policy(struct file * filp,unsigned long arg)2365 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2366 {
2367 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2368 return -EOPNOTSUPP;
2369 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2370 }
2371
f2fs_ioc_get_encryption_pwsalt(struct file * filp,unsigned long arg)2372 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2373 {
2374 struct inode *inode = file_inode(filp);
2375 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2376 int err;
2377
2378 if (!f2fs_sb_has_encrypt(sbi))
2379 return -EOPNOTSUPP;
2380
2381 err = mnt_want_write_file(filp);
2382 if (err)
2383 return err;
2384
2385 down_write(&sbi->sb_lock);
2386
2387 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2388 goto got_it;
2389
2390 /* update superblock with uuid */
2391 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2392
2393 err = f2fs_commit_super(sbi, false);
2394 if (err) {
2395 /* undo new data */
2396 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2397 goto out_err;
2398 }
2399 got_it:
2400 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2401 16))
2402 err = -EFAULT;
2403 out_err:
2404 up_write(&sbi->sb_lock);
2405 mnt_drop_write_file(filp);
2406 return err;
2407 }
2408
f2fs_ioc_get_encryption_policy_ex(struct file * filp,unsigned long arg)2409 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2410 unsigned long arg)
2411 {
2412 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2413 return -EOPNOTSUPP;
2414
2415 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2416 }
2417
f2fs_ioc_add_encryption_key(struct file * filp,unsigned long arg)2418 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2419 {
2420 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2421 return -EOPNOTSUPP;
2422
2423 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2424 }
2425
f2fs_ioc_remove_encryption_key(struct file * filp,unsigned long arg)2426 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2427 {
2428 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2429 return -EOPNOTSUPP;
2430
2431 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2432 }
2433
f2fs_ioc_remove_encryption_key_all_users(struct file * filp,unsigned long arg)2434 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2435 unsigned long arg)
2436 {
2437 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2438 return -EOPNOTSUPP;
2439
2440 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2441 }
2442
f2fs_ioc_get_encryption_key_status(struct file * filp,unsigned long arg)2443 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2444 unsigned long arg)
2445 {
2446 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2447 return -EOPNOTSUPP;
2448
2449 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2450 }
2451
f2fs_ioc_get_encryption_nonce(struct file * filp,unsigned long arg)2452 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2453 {
2454 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2455 return -EOPNOTSUPP;
2456
2457 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2458 }
2459
f2fs_ioc_gc(struct file * filp,unsigned long arg)2460 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2461 {
2462 struct inode *inode = file_inode(filp);
2463 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2464 __u32 sync;
2465 int ret;
2466
2467 if (!capable(CAP_SYS_ADMIN))
2468 return -EPERM;
2469
2470 if (get_user(sync, (__u32 __user *)arg))
2471 return -EFAULT;
2472
2473 if (f2fs_readonly(sbi->sb))
2474 return -EROFS;
2475
2476 ret = mnt_want_write_file(filp);
2477 if (ret)
2478 return ret;
2479
2480 if (!sync) {
2481 if (!down_write_trylock(&sbi->gc_lock)) {
2482 ret = -EBUSY;
2483 goto out;
2484 }
2485 } else {
2486 down_write(&sbi->gc_lock);
2487 }
2488
2489 ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO);
2490 out:
2491 mnt_drop_write_file(filp);
2492 return ret;
2493 }
2494
__f2fs_ioc_gc_range(struct file * filp,struct f2fs_gc_range * range)2495 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2496 {
2497 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2498 u64 end;
2499 int ret;
2500
2501 if (!capable(CAP_SYS_ADMIN))
2502 return -EPERM;
2503 if (f2fs_readonly(sbi->sb))
2504 return -EROFS;
2505
2506 end = range->start + range->len;
2507 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2508 end >= MAX_BLKADDR(sbi))
2509 return -EINVAL;
2510
2511 ret = mnt_want_write_file(filp);
2512 if (ret)
2513 return ret;
2514
2515 do_more:
2516 if (!range->sync) {
2517 if (!down_write_trylock(&sbi->gc_lock)) {
2518 ret = -EBUSY;
2519 goto out;
2520 }
2521 } else {
2522 down_write(&sbi->gc_lock);
2523 }
2524
2525 ret = f2fs_gc(sbi, range->sync, true, false,
2526 GET_SEGNO(sbi, range->start));
2527 if (ret) {
2528 if (ret == -EBUSY)
2529 ret = -EAGAIN;
2530 goto out;
2531 }
2532 range->start += BLKS_PER_SEC(sbi);
2533 if (range->start <= end)
2534 goto do_more;
2535 out:
2536 mnt_drop_write_file(filp);
2537 return ret;
2538 }
2539
f2fs_ioc_gc_range(struct file * filp,unsigned long arg)2540 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2541 {
2542 struct f2fs_gc_range range;
2543
2544 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2545 sizeof(range)))
2546 return -EFAULT;
2547 return __f2fs_ioc_gc_range(filp, &range);
2548 }
2549
f2fs_ioc_write_checkpoint(struct file * filp,unsigned long arg)2550 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2551 {
2552 struct inode *inode = file_inode(filp);
2553 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2554 int ret;
2555
2556 if (!capable(CAP_SYS_ADMIN))
2557 return -EPERM;
2558
2559 if (f2fs_readonly(sbi->sb))
2560 return -EROFS;
2561
2562 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2563 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2564 return -EINVAL;
2565 }
2566
2567 ret = mnt_want_write_file(filp);
2568 if (ret)
2569 return ret;
2570
2571 ret = f2fs_sync_fs(sbi->sb, 1);
2572
2573 mnt_drop_write_file(filp);
2574 return ret;
2575 }
2576
f2fs_defragment_range(struct f2fs_sb_info * sbi,struct file * filp,struct f2fs_defragment * range)2577 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2578 struct file *filp,
2579 struct f2fs_defragment *range)
2580 {
2581 struct inode *inode = file_inode(filp);
2582 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2583 .m_seg_type = NO_CHECK_TYPE ,
2584 .m_may_create = false };
2585 struct extent_info ei = {0, 0, 0};
2586 pgoff_t pg_start, pg_end, next_pgofs;
2587 unsigned int blk_per_seg = sbi->blocks_per_seg;
2588 unsigned int total = 0, sec_num;
2589 block_t blk_end = 0;
2590 bool fragmented = false;
2591 int err;
2592
2593 /* if in-place-update policy is enabled, don't waste time here */
2594 if (f2fs_should_update_inplace(inode, NULL))
2595 return -EINVAL;
2596
2597 pg_start = range->start >> PAGE_SHIFT;
2598 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2599
2600 f2fs_balance_fs(sbi, true);
2601
2602 inode_lock(inode);
2603
2604 /* writeback all dirty pages in the range */
2605 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2606 range->start + range->len - 1);
2607 if (err)
2608 goto out;
2609
2610 /*
2611 * lookup mapping info in extent cache, skip defragmenting if physical
2612 * block addresses are continuous.
2613 */
2614 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2615 if (ei.fofs + ei.len >= pg_end)
2616 goto out;
2617 }
2618
2619 map.m_lblk = pg_start;
2620 map.m_next_pgofs = &next_pgofs;
2621
2622 /*
2623 * lookup mapping info in dnode page cache, skip defragmenting if all
2624 * physical block addresses are continuous even if there are hole(s)
2625 * in logical blocks.
2626 */
2627 while (map.m_lblk < pg_end) {
2628 map.m_len = pg_end - map.m_lblk;
2629 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2630 if (err)
2631 goto out;
2632
2633 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2634 map.m_lblk = next_pgofs;
2635 continue;
2636 }
2637
2638 if (blk_end && blk_end != map.m_pblk)
2639 fragmented = true;
2640
2641 /* record total count of block that we're going to move */
2642 total += map.m_len;
2643
2644 blk_end = map.m_pblk + map.m_len;
2645
2646 map.m_lblk += map.m_len;
2647 }
2648
2649 if (!fragmented) {
2650 total = 0;
2651 goto out;
2652 }
2653
2654 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2655
2656 /*
2657 * make sure there are enough free section for LFS allocation, this can
2658 * avoid defragment running in SSR mode when free section are allocated
2659 * intensively
2660 */
2661 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2662 err = -EAGAIN;
2663 goto out;
2664 }
2665
2666 map.m_lblk = pg_start;
2667 map.m_len = pg_end - pg_start;
2668 total = 0;
2669
2670 while (map.m_lblk < pg_end) {
2671 pgoff_t idx;
2672 int cnt = 0;
2673
2674 do_map:
2675 map.m_len = pg_end - map.m_lblk;
2676 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2677 if (err)
2678 goto clear_out;
2679
2680 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2681 map.m_lblk = next_pgofs;
2682 goto check;
2683 }
2684
2685 set_inode_flag(inode, FI_DO_DEFRAG);
2686
2687 idx = map.m_lblk;
2688 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2689 struct page *page;
2690
2691 page = f2fs_get_lock_data_page(inode, idx, true);
2692 if (IS_ERR(page)) {
2693 err = PTR_ERR(page);
2694 goto clear_out;
2695 }
2696
2697 set_page_dirty(page);
2698 f2fs_put_page(page, 1);
2699
2700 idx++;
2701 cnt++;
2702 total++;
2703 }
2704
2705 map.m_lblk = idx;
2706 check:
2707 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2708 goto do_map;
2709
2710 clear_inode_flag(inode, FI_DO_DEFRAG);
2711
2712 err = filemap_fdatawrite(inode->i_mapping);
2713 if (err)
2714 goto out;
2715 }
2716 clear_out:
2717 clear_inode_flag(inode, FI_DO_DEFRAG);
2718 out:
2719 inode_unlock(inode);
2720 if (!err)
2721 range->len = (u64)total << PAGE_SHIFT;
2722 return err;
2723 }
2724
f2fs_ioc_defragment(struct file * filp,unsigned long arg)2725 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2726 {
2727 struct inode *inode = file_inode(filp);
2728 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2729 struct f2fs_defragment range;
2730 int err;
2731
2732 if (!capable(CAP_SYS_ADMIN))
2733 return -EPERM;
2734
2735 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2736 return -EINVAL;
2737
2738 if (f2fs_readonly(sbi->sb))
2739 return -EROFS;
2740
2741 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2742 sizeof(range)))
2743 return -EFAULT;
2744
2745 /* verify alignment of offset & size */
2746 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2747 return -EINVAL;
2748
2749 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2750 sbi->max_file_blocks))
2751 return -EINVAL;
2752
2753 err = mnt_want_write_file(filp);
2754 if (err)
2755 return err;
2756
2757 err = f2fs_defragment_range(sbi, filp, &range);
2758 mnt_drop_write_file(filp);
2759
2760 f2fs_update_time(sbi, REQ_TIME);
2761 if (err < 0)
2762 return err;
2763
2764 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2765 sizeof(range)))
2766 return -EFAULT;
2767
2768 return 0;
2769 }
2770
f2fs_move_file_range(struct file * file_in,loff_t pos_in,struct file * file_out,loff_t pos_out,size_t len)2771 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2772 struct file *file_out, loff_t pos_out, size_t len)
2773 {
2774 struct inode *src = file_inode(file_in);
2775 struct inode *dst = file_inode(file_out);
2776 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2777 size_t olen = len, dst_max_i_size = 0;
2778 size_t dst_osize;
2779 int ret;
2780
2781 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2782 src->i_sb != dst->i_sb)
2783 return -EXDEV;
2784
2785 if (unlikely(f2fs_readonly(src->i_sb)))
2786 return -EROFS;
2787
2788 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2789 return -EINVAL;
2790
2791 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2792 return -EOPNOTSUPP;
2793
2794 if (pos_out < 0 || pos_in < 0)
2795 return -EINVAL;
2796
2797 if (src == dst) {
2798 if (pos_in == pos_out)
2799 return 0;
2800 if (pos_out > pos_in && pos_out < pos_in + len)
2801 return -EINVAL;
2802 }
2803
2804 inode_lock(src);
2805 if (src != dst) {
2806 ret = -EBUSY;
2807 if (!inode_trylock(dst))
2808 goto out;
2809 }
2810
2811 ret = -EINVAL;
2812 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2813 goto out_unlock;
2814 if (len == 0)
2815 olen = len = src->i_size - pos_in;
2816 if (pos_in + len == src->i_size)
2817 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2818 if (len == 0) {
2819 ret = 0;
2820 goto out_unlock;
2821 }
2822
2823 dst_osize = dst->i_size;
2824 if (pos_out + olen > dst->i_size)
2825 dst_max_i_size = pos_out + olen;
2826
2827 /* verify the end result is block aligned */
2828 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2829 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2830 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2831 goto out_unlock;
2832
2833 ret = f2fs_convert_inline_inode(src);
2834 if (ret)
2835 goto out_unlock;
2836
2837 ret = f2fs_convert_inline_inode(dst);
2838 if (ret)
2839 goto out_unlock;
2840
2841 /* write out all dirty pages from offset */
2842 ret = filemap_write_and_wait_range(src->i_mapping,
2843 pos_in, pos_in + len);
2844 if (ret)
2845 goto out_unlock;
2846
2847 ret = filemap_write_and_wait_range(dst->i_mapping,
2848 pos_out, pos_out + len);
2849 if (ret)
2850 goto out_unlock;
2851
2852 f2fs_balance_fs(sbi, true);
2853
2854 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2855 if (src != dst) {
2856 ret = -EBUSY;
2857 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2858 goto out_src;
2859 }
2860
2861 f2fs_lock_op(sbi);
2862 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2863 pos_out >> F2FS_BLKSIZE_BITS,
2864 len >> F2FS_BLKSIZE_BITS, false);
2865
2866 if (!ret) {
2867 if (dst_max_i_size)
2868 f2fs_i_size_write(dst, dst_max_i_size);
2869 else if (dst_osize != dst->i_size)
2870 f2fs_i_size_write(dst, dst_osize);
2871 }
2872 f2fs_unlock_op(sbi);
2873
2874 if (src != dst)
2875 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2876 out_src:
2877 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2878 out_unlock:
2879 if (src != dst)
2880 inode_unlock(dst);
2881 out:
2882 inode_unlock(src);
2883 return ret;
2884 }
2885
__f2fs_ioc_move_range(struct file * filp,struct f2fs_move_range * range)2886 static int __f2fs_ioc_move_range(struct file *filp,
2887 struct f2fs_move_range *range)
2888 {
2889 struct fd dst;
2890 int err;
2891
2892 if (!(filp->f_mode & FMODE_READ) ||
2893 !(filp->f_mode & FMODE_WRITE))
2894 return -EBADF;
2895
2896 dst = fdget(range->dst_fd);
2897 if (!dst.file)
2898 return -EBADF;
2899
2900 if (!(dst.file->f_mode & FMODE_WRITE)) {
2901 err = -EBADF;
2902 goto err_out;
2903 }
2904
2905 err = mnt_want_write_file(filp);
2906 if (err)
2907 goto err_out;
2908
2909 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2910 range->pos_out, range->len);
2911
2912 mnt_drop_write_file(filp);
2913 err_out:
2914 fdput(dst);
2915 return err;
2916 }
2917
f2fs_ioc_move_range(struct file * filp,unsigned long arg)2918 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2919 {
2920 struct f2fs_move_range range;
2921
2922 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2923 sizeof(range)))
2924 return -EFAULT;
2925 return __f2fs_ioc_move_range(filp, &range);
2926 }
2927
f2fs_ioc_flush_device(struct file * filp,unsigned long arg)2928 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2929 {
2930 struct inode *inode = file_inode(filp);
2931 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2932 struct sit_info *sm = SIT_I(sbi);
2933 unsigned int start_segno = 0, end_segno = 0;
2934 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2935 struct f2fs_flush_device range;
2936 int ret;
2937
2938 if (!capable(CAP_SYS_ADMIN))
2939 return -EPERM;
2940
2941 if (f2fs_readonly(sbi->sb))
2942 return -EROFS;
2943
2944 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2945 return -EINVAL;
2946
2947 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2948 sizeof(range)))
2949 return -EFAULT;
2950
2951 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2952 __is_large_section(sbi)) {
2953 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2954 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2955 return -EINVAL;
2956 }
2957
2958 ret = mnt_want_write_file(filp);
2959 if (ret)
2960 return ret;
2961
2962 if (range.dev_num != 0)
2963 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2964 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2965
2966 start_segno = sm->last_victim[FLUSH_DEVICE];
2967 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2968 start_segno = dev_start_segno;
2969 end_segno = min(start_segno + range.segments, dev_end_segno);
2970
2971 while (start_segno < end_segno) {
2972 if (!down_write_trylock(&sbi->gc_lock)) {
2973 ret = -EBUSY;
2974 goto out;
2975 }
2976 sm->last_victim[GC_CB] = end_segno + 1;
2977 sm->last_victim[GC_GREEDY] = end_segno + 1;
2978 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2979 ret = f2fs_gc(sbi, true, true, true, start_segno);
2980 if (ret == -EAGAIN)
2981 ret = 0;
2982 else if (ret < 0)
2983 break;
2984 start_segno++;
2985 }
2986 out:
2987 mnt_drop_write_file(filp);
2988 return ret;
2989 }
2990
f2fs_ioc_get_features(struct file * filp,unsigned long arg)2991 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2992 {
2993 struct inode *inode = file_inode(filp);
2994 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2995
2996 /* Must validate to set it with SQLite behavior in Android. */
2997 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2998
2999 return put_user(sb_feature, (u32 __user *)arg);
3000 }
3001
3002 #ifdef CONFIG_QUOTA
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3003 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3004 {
3005 struct dquot *transfer_to[MAXQUOTAS] = {};
3006 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3007 struct super_block *sb = sbi->sb;
3008 int err = 0;
3009
3010 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3011 if (!IS_ERR(transfer_to[PRJQUOTA])) {
3012 err = __dquot_transfer(inode, transfer_to);
3013 if (err)
3014 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3015 dqput(transfer_to[PRJQUOTA]);
3016 }
3017 return err;
3018 }
3019
f2fs_ioc_setproject(struct file * filp,__u32 projid)3020 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3021 {
3022 struct inode *inode = file_inode(filp);
3023 struct f2fs_inode_info *fi = F2FS_I(inode);
3024 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3025 struct page *ipage;
3026 kprojid_t kprojid;
3027 int err;
3028
3029 if (!f2fs_sb_has_project_quota(sbi)) {
3030 if (projid != F2FS_DEF_PROJID)
3031 return -EOPNOTSUPP;
3032 else
3033 return 0;
3034 }
3035
3036 if (!f2fs_has_extra_attr(inode))
3037 return -EOPNOTSUPP;
3038
3039 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3040
3041 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
3042 return 0;
3043
3044 err = -EPERM;
3045 /* Is it quota file? Do not allow user to mess with it */
3046 if (IS_NOQUOTA(inode))
3047 return err;
3048
3049 ipage = f2fs_get_node_page(sbi, inode->i_ino);
3050 if (IS_ERR(ipage))
3051 return PTR_ERR(ipage);
3052
3053 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
3054 i_projid)) {
3055 err = -EOVERFLOW;
3056 f2fs_put_page(ipage, 1);
3057 return err;
3058 }
3059 f2fs_put_page(ipage, 1);
3060
3061 err = dquot_initialize(inode);
3062 if (err)
3063 return err;
3064
3065 f2fs_lock_op(sbi);
3066 err = f2fs_transfer_project_quota(inode, kprojid);
3067 if (err)
3068 goto out_unlock;
3069
3070 F2FS_I(inode)->i_projid = kprojid;
3071 inode->i_ctime = current_time(inode);
3072 f2fs_mark_inode_dirty_sync(inode, true);
3073 out_unlock:
3074 f2fs_unlock_op(sbi);
3075 return err;
3076 }
3077 #else
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3078 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3079 {
3080 return 0;
3081 }
3082
f2fs_ioc_setproject(struct file * filp,__u32 projid)3083 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3084 {
3085 if (projid != F2FS_DEF_PROJID)
3086 return -EOPNOTSUPP;
3087 return 0;
3088 }
3089 #endif
3090
3091 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
3092
3093 /*
3094 * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
3095 * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
3096 * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
3097 */
3098
3099 static const struct {
3100 u32 iflag;
3101 u32 xflag;
3102 } f2fs_xflags_map[] = {
3103 { F2FS_SYNC_FL, FS_XFLAG_SYNC },
3104 { F2FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE },
3105 { F2FS_APPEND_FL, FS_XFLAG_APPEND },
3106 { F2FS_NODUMP_FL, FS_XFLAG_NODUMP },
3107 { F2FS_NOATIME_FL, FS_XFLAG_NOATIME },
3108 { F2FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT },
3109 };
3110
3111 #define F2FS_SUPPORTED_XFLAGS ( \
3112 FS_XFLAG_SYNC | \
3113 FS_XFLAG_IMMUTABLE | \
3114 FS_XFLAG_APPEND | \
3115 FS_XFLAG_NODUMP | \
3116 FS_XFLAG_NOATIME | \
3117 FS_XFLAG_PROJINHERIT)
3118
3119 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
f2fs_iflags_to_xflags(u32 iflags)3120 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
3121 {
3122 u32 xflags = 0;
3123 int i;
3124
3125 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3126 if (iflags & f2fs_xflags_map[i].iflag)
3127 xflags |= f2fs_xflags_map[i].xflag;
3128
3129 return xflags;
3130 }
3131
3132 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
f2fs_xflags_to_iflags(u32 xflags)3133 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
3134 {
3135 u32 iflags = 0;
3136 int i;
3137
3138 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3139 if (xflags & f2fs_xflags_map[i].xflag)
3140 iflags |= f2fs_xflags_map[i].iflag;
3141
3142 return iflags;
3143 }
3144
f2fs_fill_fsxattr(struct inode * inode,struct fsxattr * fa)3145 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
3146 {
3147 struct f2fs_inode_info *fi = F2FS_I(inode);
3148
3149 simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
3150
3151 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3152 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3153 }
3154
f2fs_ioc_fsgetxattr(struct file * filp,unsigned long arg)3155 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
3156 {
3157 struct inode *inode = file_inode(filp);
3158 struct fsxattr fa;
3159
3160 f2fs_fill_fsxattr(inode, &fa);
3161
3162 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
3163 return -EFAULT;
3164 return 0;
3165 }
3166
f2fs_ioc_fssetxattr(struct file * filp,unsigned long arg)3167 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
3168 {
3169 struct inode *inode = file_inode(filp);
3170 struct fsxattr fa, old_fa;
3171 u32 iflags;
3172 int err;
3173
3174 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
3175 return -EFAULT;
3176
3177 /* Make sure caller has proper permission */
3178 if (!inode_owner_or_capable(inode))
3179 return -EACCES;
3180
3181 if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
3182 return -EOPNOTSUPP;
3183
3184 iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
3185 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3186 return -EOPNOTSUPP;
3187
3188 err = mnt_want_write_file(filp);
3189 if (err)
3190 return err;
3191
3192 inode_lock(inode);
3193
3194 f2fs_fill_fsxattr(inode, &old_fa);
3195 err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
3196 if (err)
3197 goto out;
3198
3199 err = f2fs_setflags_common(inode, iflags,
3200 f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
3201 if (err)
3202 goto out;
3203
3204 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3205 out:
3206 inode_unlock(inode);
3207 mnt_drop_write_file(filp);
3208 return err;
3209 }
3210
f2fs_pin_file_control(struct inode * inode,bool inc)3211 int f2fs_pin_file_control(struct inode *inode, bool inc)
3212 {
3213 struct f2fs_inode_info *fi = F2FS_I(inode);
3214 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3215
3216 /* Use i_gc_failures for normal file as a risk signal. */
3217 if (inc)
3218 f2fs_i_gc_failures_write(inode,
3219 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3220
3221 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3222 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3223 __func__, inode->i_ino,
3224 fi->i_gc_failures[GC_FAILURE_PIN]);
3225 clear_inode_flag(inode, FI_PIN_FILE);
3226 return -EAGAIN;
3227 }
3228 return 0;
3229 }
3230
f2fs_ioc_set_pin_file(struct file * filp,unsigned long arg)3231 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3232 {
3233 struct inode *inode = file_inode(filp);
3234 __u32 pin;
3235 int ret = 0;
3236
3237 if (get_user(pin, (__u32 __user *)arg))
3238 return -EFAULT;
3239
3240 if (!S_ISREG(inode->i_mode))
3241 return -EINVAL;
3242
3243 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3244 return -EROFS;
3245
3246 ret = mnt_want_write_file(filp);
3247 if (ret)
3248 return ret;
3249
3250 inode_lock(inode);
3251
3252 if (f2fs_should_update_outplace(inode, NULL)) {
3253 ret = -EINVAL;
3254 goto out;
3255 }
3256
3257 if (!pin) {
3258 clear_inode_flag(inode, FI_PIN_FILE);
3259 f2fs_i_gc_failures_write(inode, 0);
3260 goto done;
3261 }
3262
3263 if (f2fs_pin_file_control(inode, false)) {
3264 ret = -EAGAIN;
3265 goto out;
3266 }
3267
3268 ret = f2fs_convert_inline_inode(inode);
3269 if (ret)
3270 goto out;
3271
3272 if (!f2fs_disable_compressed_file(inode)) {
3273 ret = -EOPNOTSUPP;
3274 goto out;
3275 }
3276
3277 set_inode_flag(inode, FI_PIN_FILE);
3278 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3279 done:
3280 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3281 out:
3282 inode_unlock(inode);
3283 mnt_drop_write_file(filp);
3284 return ret;
3285 }
3286
f2fs_ioc_get_pin_file(struct file * filp,unsigned long arg)3287 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3288 {
3289 struct inode *inode = file_inode(filp);
3290 __u32 pin = 0;
3291
3292 if (is_inode_flag_set(inode, FI_PIN_FILE))
3293 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3294 return put_user(pin, (u32 __user *)arg);
3295 }
3296
f2fs_precache_extents(struct inode * inode)3297 int f2fs_precache_extents(struct inode *inode)
3298 {
3299 struct f2fs_inode_info *fi = F2FS_I(inode);
3300 struct f2fs_map_blocks map;
3301 pgoff_t m_next_extent;
3302 loff_t end;
3303 int err;
3304
3305 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3306 return -EOPNOTSUPP;
3307
3308 map.m_lblk = 0;
3309 map.m_next_pgofs = NULL;
3310 map.m_next_extent = &m_next_extent;
3311 map.m_seg_type = NO_CHECK_TYPE;
3312 map.m_may_create = false;
3313 end = F2FS_I_SB(inode)->max_file_blocks;
3314
3315 while (map.m_lblk < end) {
3316 map.m_len = end - map.m_lblk;
3317
3318 down_write(&fi->i_gc_rwsem[WRITE]);
3319 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3320 up_write(&fi->i_gc_rwsem[WRITE]);
3321 if (err)
3322 return err;
3323
3324 map.m_lblk = m_next_extent;
3325 }
3326
3327 return err;
3328 }
3329
f2fs_ioc_precache_extents(struct file * filp,unsigned long arg)3330 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3331 {
3332 return f2fs_precache_extents(file_inode(filp));
3333 }
3334
f2fs_ioc_resize_fs(struct file * filp,unsigned long arg)3335 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3336 {
3337 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3338 __u64 block_count;
3339
3340 if (!capable(CAP_SYS_ADMIN))
3341 return -EPERM;
3342
3343 if (f2fs_readonly(sbi->sb))
3344 return -EROFS;
3345
3346 if (copy_from_user(&block_count, (void __user *)arg,
3347 sizeof(block_count)))
3348 return -EFAULT;
3349
3350 return f2fs_resize_fs(sbi, block_count);
3351 }
3352
f2fs_ioc_enable_verity(struct file * filp,unsigned long arg)3353 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3354 {
3355 struct inode *inode = file_inode(filp);
3356
3357 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3358
3359 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3360 f2fs_warn(F2FS_I_SB(inode),
3361 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3362 inode->i_ino);
3363 return -EOPNOTSUPP;
3364 }
3365
3366 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3367 }
3368
f2fs_ioc_measure_verity(struct file * filp,unsigned long arg)3369 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3370 {
3371 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3372 return -EOPNOTSUPP;
3373
3374 return fsverity_ioctl_measure(filp, (void __user *)arg);
3375 }
3376
f2fs_ioc_getfslabel(struct file * filp,unsigned long arg)3377 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3378 {
3379 struct inode *inode = file_inode(filp);
3380 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3381 char *vbuf;
3382 int count;
3383 int err = 0;
3384
3385 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3386 if (!vbuf)
3387 return -ENOMEM;
3388
3389 down_read(&sbi->sb_lock);
3390 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3391 ARRAY_SIZE(sbi->raw_super->volume_name),
3392 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3393 up_read(&sbi->sb_lock);
3394
3395 if (copy_to_user((char __user *)arg, vbuf,
3396 min(FSLABEL_MAX, count)))
3397 err = -EFAULT;
3398
3399 kfree(vbuf);
3400 return err;
3401 }
3402
f2fs_ioc_setfslabel(struct file * filp,unsigned long arg)3403 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3404 {
3405 struct inode *inode = file_inode(filp);
3406 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3407 char *vbuf;
3408 int err = 0;
3409
3410 if (!capable(CAP_SYS_ADMIN))
3411 return -EPERM;
3412
3413 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3414 if (IS_ERR(vbuf))
3415 return PTR_ERR(vbuf);
3416
3417 err = mnt_want_write_file(filp);
3418 if (err)
3419 goto out;
3420
3421 down_write(&sbi->sb_lock);
3422
3423 memset(sbi->raw_super->volume_name, 0,
3424 sizeof(sbi->raw_super->volume_name));
3425 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3426 sbi->raw_super->volume_name,
3427 ARRAY_SIZE(sbi->raw_super->volume_name));
3428
3429 err = f2fs_commit_super(sbi, false);
3430
3431 up_write(&sbi->sb_lock);
3432
3433 mnt_drop_write_file(filp);
3434 out:
3435 kfree(vbuf);
3436 return err;
3437 }
3438
f2fs_get_compress_blocks(struct file * filp,unsigned long arg)3439 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3440 {
3441 struct inode *inode = file_inode(filp);
3442 __u64 blocks;
3443
3444 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3445 return -EOPNOTSUPP;
3446
3447 if (!f2fs_compressed_file(inode))
3448 return -EINVAL;
3449
3450 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3451 return put_user(blocks, (u64 __user *)arg);
3452 }
3453
release_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3454 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3455 {
3456 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3457 unsigned int released_blocks = 0;
3458 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3459 block_t blkaddr;
3460 int i;
3461
3462 for (i = 0; i < count; i++) {
3463 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3464 dn->ofs_in_node + i);
3465
3466 if (!__is_valid_data_blkaddr(blkaddr))
3467 continue;
3468 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3469 DATA_GENERIC_ENHANCE)))
3470 return -EFSCORRUPTED;
3471 }
3472
3473 while (count) {
3474 int compr_blocks = 0;
3475
3476 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3477 blkaddr = f2fs_data_blkaddr(dn);
3478
3479 if (i == 0) {
3480 if (blkaddr == COMPRESS_ADDR)
3481 continue;
3482 dn->ofs_in_node += cluster_size;
3483 goto next;
3484 }
3485
3486 if (__is_valid_data_blkaddr(blkaddr))
3487 compr_blocks++;
3488
3489 if (blkaddr != NEW_ADDR)
3490 continue;
3491
3492 dn->data_blkaddr = NULL_ADDR;
3493 f2fs_set_data_blkaddr(dn);
3494 }
3495
3496 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3497 dec_valid_block_count(sbi, dn->inode,
3498 cluster_size - compr_blocks);
3499
3500 released_blocks += cluster_size - compr_blocks;
3501 next:
3502 count -= cluster_size;
3503 }
3504
3505 return released_blocks;
3506 }
3507
f2fs_release_compress_blocks(struct file * filp,unsigned long arg)3508 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3509 {
3510 struct inode *inode = file_inode(filp);
3511 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3512 pgoff_t page_idx = 0, last_idx;
3513 unsigned int released_blocks = 0;
3514 int ret;
3515 int writecount;
3516
3517 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3518 return -EOPNOTSUPP;
3519
3520 if (!f2fs_compressed_file(inode))
3521 return -EINVAL;
3522
3523 if (f2fs_readonly(sbi->sb))
3524 return -EROFS;
3525
3526 ret = mnt_want_write_file(filp);
3527 if (ret)
3528 return ret;
3529
3530 f2fs_balance_fs(F2FS_I_SB(inode), true);
3531
3532 inode_lock(inode);
3533
3534 writecount = atomic_read(&inode->i_writecount);
3535 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3536 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3537 ret = -EBUSY;
3538 goto out;
3539 }
3540
3541 if (IS_IMMUTABLE(inode)) {
3542 ret = -EINVAL;
3543 goto out;
3544 }
3545
3546 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3547 if (ret)
3548 goto out;
3549
3550 F2FS_I(inode)->i_flags |= F2FS_IMMUTABLE_FL;
3551 f2fs_set_inode_flags(inode);
3552 inode->i_ctime = current_time(inode);
3553 f2fs_mark_inode_dirty_sync(inode, true);
3554
3555 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3556 goto out;
3557
3558 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3559 down_write(&F2FS_I(inode)->i_mmap_sem);
3560
3561 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3562
3563 while (page_idx < last_idx) {
3564 struct dnode_of_data dn;
3565 pgoff_t end_offset, count;
3566
3567 set_new_dnode(&dn, inode, NULL, NULL, 0);
3568 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3569 if (ret) {
3570 if (ret == -ENOENT) {
3571 page_idx = f2fs_get_next_page_offset(&dn,
3572 page_idx);
3573 ret = 0;
3574 continue;
3575 }
3576 break;
3577 }
3578
3579 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3580 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3581 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3582
3583 ret = release_compress_blocks(&dn, count);
3584
3585 f2fs_put_dnode(&dn);
3586
3587 if (ret < 0)
3588 break;
3589
3590 page_idx += count;
3591 released_blocks += ret;
3592 }
3593
3594 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3595 up_write(&F2FS_I(inode)->i_mmap_sem);
3596 out:
3597 inode_unlock(inode);
3598
3599 mnt_drop_write_file(filp);
3600
3601 if (ret >= 0) {
3602 ret = put_user(released_blocks, (u64 __user *)arg);
3603 } else if (released_blocks &&
3604 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3605 set_sbi_flag(sbi, SBI_NEED_FSCK);
3606 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3607 "iblocks=%llu, released=%u, compr_blocks=%u, "
3608 "run fsck to fix.",
3609 __func__, inode->i_ino, inode->i_blocks,
3610 released_blocks,
3611 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3612 }
3613
3614 return ret;
3615 }
3616
reserve_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3617 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3618 {
3619 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3620 unsigned int reserved_blocks = 0;
3621 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3622 block_t blkaddr;
3623 int i;
3624
3625 for (i = 0; i < count; i++) {
3626 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3627 dn->ofs_in_node + i);
3628
3629 if (!__is_valid_data_blkaddr(blkaddr))
3630 continue;
3631 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3632 DATA_GENERIC_ENHANCE)))
3633 return -EFSCORRUPTED;
3634 }
3635
3636 while (count) {
3637 int compr_blocks = 0;
3638 blkcnt_t reserved;
3639 int ret;
3640
3641 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3642 blkaddr = f2fs_data_blkaddr(dn);
3643
3644 if (i == 0) {
3645 if (blkaddr == COMPRESS_ADDR)
3646 continue;
3647 dn->ofs_in_node += cluster_size;
3648 goto next;
3649 }
3650
3651 if (__is_valid_data_blkaddr(blkaddr)) {
3652 compr_blocks++;
3653 continue;
3654 }
3655
3656 dn->data_blkaddr = NEW_ADDR;
3657 f2fs_set_data_blkaddr(dn);
3658 }
3659
3660 reserved = cluster_size - compr_blocks;
3661 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3662 if (ret)
3663 return ret;
3664
3665 if (reserved != cluster_size - compr_blocks)
3666 return -ENOSPC;
3667
3668 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3669
3670 reserved_blocks += reserved;
3671 next:
3672 count -= cluster_size;
3673 }
3674
3675 return reserved_blocks;
3676 }
3677
f2fs_reserve_compress_blocks(struct file * filp,unsigned long arg)3678 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3679 {
3680 struct inode *inode = file_inode(filp);
3681 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3682 pgoff_t page_idx = 0, last_idx;
3683 unsigned int reserved_blocks = 0;
3684 int ret;
3685
3686 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3687 return -EOPNOTSUPP;
3688
3689 if (!f2fs_compressed_file(inode))
3690 return -EINVAL;
3691
3692 if (f2fs_readonly(sbi->sb))
3693 return -EROFS;
3694
3695 ret = mnt_want_write_file(filp);
3696 if (ret)
3697 return ret;
3698
3699 if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3700 goto out;
3701
3702 f2fs_balance_fs(F2FS_I_SB(inode), true);
3703
3704 inode_lock(inode);
3705
3706 if (!IS_IMMUTABLE(inode)) {
3707 ret = -EINVAL;
3708 goto unlock_inode;
3709 }
3710
3711 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3712 down_write(&F2FS_I(inode)->i_mmap_sem);
3713
3714 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3715
3716 while (page_idx < last_idx) {
3717 struct dnode_of_data dn;
3718 pgoff_t end_offset, count;
3719
3720 set_new_dnode(&dn, inode, NULL, NULL, 0);
3721 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3722 if (ret) {
3723 if (ret == -ENOENT) {
3724 page_idx = f2fs_get_next_page_offset(&dn,
3725 page_idx);
3726 ret = 0;
3727 continue;
3728 }
3729 break;
3730 }
3731
3732 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3733 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3734 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3735
3736 ret = reserve_compress_blocks(&dn, count);
3737
3738 f2fs_put_dnode(&dn);
3739
3740 if (ret < 0)
3741 break;
3742
3743 page_idx += count;
3744 reserved_blocks += ret;
3745 }
3746
3747 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3748 up_write(&F2FS_I(inode)->i_mmap_sem);
3749
3750 if (ret >= 0) {
3751 F2FS_I(inode)->i_flags &= ~F2FS_IMMUTABLE_FL;
3752 f2fs_set_inode_flags(inode);
3753 inode->i_ctime = current_time(inode);
3754 f2fs_mark_inode_dirty_sync(inode, true);
3755 }
3756 unlock_inode:
3757 inode_unlock(inode);
3758 out:
3759 mnt_drop_write_file(filp);
3760
3761 if (ret >= 0) {
3762 ret = put_user(reserved_blocks, (u64 __user *)arg);
3763 } else if (reserved_blocks &&
3764 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3765 set_sbi_flag(sbi, SBI_NEED_FSCK);
3766 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3767 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3768 "run fsck to fix.",
3769 __func__, inode->i_ino, inode->i_blocks,
3770 reserved_blocks,
3771 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3772 }
3773
3774 return ret;
3775 }
3776
f2fs_secure_erase(struct block_device * bdev,struct inode * inode,pgoff_t off,block_t block,block_t len,u32 flags)3777 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3778 pgoff_t off, block_t block, block_t len, u32 flags)
3779 {
3780 struct request_queue *q = bdev_get_queue(bdev);
3781 sector_t sector = SECTOR_FROM_BLOCK(block);
3782 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3783 int ret = 0;
3784
3785 if (!q)
3786 return -ENXIO;
3787
3788 if (flags & F2FS_TRIM_FILE_DISCARD)
3789 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3790 blk_queue_secure_erase(q) ?
3791 BLKDEV_DISCARD_SECURE : 0);
3792
3793 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3794 if (IS_ENCRYPTED(inode))
3795 ret = fscrypt_zeroout_range(inode, off, block, len);
3796 else
3797 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3798 GFP_NOFS, 0);
3799 }
3800
3801 return ret;
3802 }
3803
f2fs_sec_trim_file(struct file * filp,unsigned long arg)3804 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3805 {
3806 struct inode *inode = file_inode(filp);
3807 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3808 struct address_space *mapping = inode->i_mapping;
3809 struct block_device *prev_bdev = NULL;
3810 struct f2fs_sectrim_range range;
3811 pgoff_t index, pg_end, prev_index = 0;
3812 block_t prev_block = 0, len = 0;
3813 loff_t end_addr;
3814 bool to_end = false;
3815 int ret = 0;
3816
3817 if (!(filp->f_mode & FMODE_WRITE))
3818 return -EBADF;
3819
3820 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3821 sizeof(range)))
3822 return -EFAULT;
3823
3824 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3825 !S_ISREG(inode->i_mode))
3826 return -EINVAL;
3827
3828 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3829 !f2fs_hw_support_discard(sbi)) ||
3830 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3831 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3832 return -EOPNOTSUPP;
3833
3834 file_start_write(filp);
3835 inode_lock(inode);
3836
3837 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3838 range.start >= inode->i_size) {
3839 ret = -EINVAL;
3840 goto err;
3841 }
3842
3843 if (range.len == 0)
3844 goto err;
3845
3846 if (inode->i_size - range.start > range.len) {
3847 end_addr = range.start + range.len;
3848 } else {
3849 end_addr = range.len == (u64)-1 ?
3850 sbi->sb->s_maxbytes : inode->i_size;
3851 to_end = true;
3852 }
3853
3854 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3855 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3856 ret = -EINVAL;
3857 goto err;
3858 }
3859
3860 index = F2FS_BYTES_TO_BLK(range.start);
3861 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3862
3863 ret = f2fs_convert_inline_inode(inode);
3864 if (ret)
3865 goto err;
3866
3867 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3868 down_write(&F2FS_I(inode)->i_mmap_sem);
3869
3870 ret = filemap_write_and_wait_range(mapping, range.start,
3871 to_end ? LLONG_MAX : end_addr - 1);
3872 if (ret)
3873 goto out;
3874
3875 truncate_inode_pages_range(mapping, range.start,
3876 to_end ? -1 : end_addr - 1);
3877
3878 while (index < pg_end) {
3879 struct dnode_of_data dn;
3880 pgoff_t end_offset, count;
3881 int i;
3882
3883 set_new_dnode(&dn, inode, NULL, NULL, 0);
3884 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3885 if (ret) {
3886 if (ret == -ENOENT) {
3887 index = f2fs_get_next_page_offset(&dn, index);
3888 continue;
3889 }
3890 goto out;
3891 }
3892
3893 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3894 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3895 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3896 struct block_device *cur_bdev;
3897 block_t blkaddr = f2fs_data_blkaddr(&dn);
3898
3899 if (!__is_valid_data_blkaddr(blkaddr))
3900 continue;
3901
3902 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3903 DATA_GENERIC_ENHANCE)) {
3904 ret = -EFSCORRUPTED;
3905 f2fs_put_dnode(&dn);
3906 goto out;
3907 }
3908
3909 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3910 if (f2fs_is_multi_device(sbi)) {
3911 int di = f2fs_target_device_index(sbi, blkaddr);
3912
3913 blkaddr -= FDEV(di).start_blk;
3914 }
3915
3916 if (len) {
3917 if (prev_bdev == cur_bdev &&
3918 index == prev_index + len &&
3919 blkaddr == prev_block + len) {
3920 len++;
3921 } else {
3922 ret = f2fs_secure_erase(prev_bdev,
3923 inode, prev_index, prev_block,
3924 len, range.flags);
3925 if (ret) {
3926 f2fs_put_dnode(&dn);
3927 goto out;
3928 }
3929
3930 len = 0;
3931 }
3932 }
3933
3934 if (!len) {
3935 prev_bdev = cur_bdev;
3936 prev_index = index;
3937 prev_block = blkaddr;
3938 len = 1;
3939 }
3940 }
3941
3942 f2fs_put_dnode(&dn);
3943
3944 if (fatal_signal_pending(current)) {
3945 ret = -EINTR;
3946 goto out;
3947 }
3948 cond_resched();
3949 }
3950
3951 if (len)
3952 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3953 prev_block, len, range.flags);
3954 out:
3955 up_write(&F2FS_I(inode)->i_mmap_sem);
3956 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3957 err:
3958 inode_unlock(inode);
3959 file_end_write(filp);
3960
3961 return ret;
3962 }
3963
__f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)3964 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3965 {
3966 switch (cmd) {
3967 case FS_IOC_GETFLAGS:
3968 return f2fs_ioc_getflags(filp, arg);
3969 case FS_IOC_SETFLAGS:
3970 return f2fs_ioc_setflags(filp, arg);
3971 case FS_IOC_GETVERSION:
3972 return f2fs_ioc_getversion(filp, arg);
3973 case F2FS_IOC_START_ATOMIC_WRITE:
3974 return f2fs_ioc_start_atomic_write(filp);
3975 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3976 return f2fs_ioc_commit_atomic_write(filp);
3977 case F2FS_IOC_START_VOLATILE_WRITE:
3978 return f2fs_ioc_start_volatile_write(filp);
3979 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3980 return f2fs_ioc_release_volatile_write(filp);
3981 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3982 return f2fs_ioc_abort_volatile_write(filp);
3983 case F2FS_IOC_SHUTDOWN:
3984 return f2fs_ioc_shutdown(filp, arg);
3985 case FITRIM:
3986 return f2fs_ioc_fitrim(filp, arg);
3987 case FS_IOC_SET_ENCRYPTION_POLICY:
3988 return f2fs_ioc_set_encryption_policy(filp, arg);
3989 case FS_IOC_GET_ENCRYPTION_POLICY:
3990 return f2fs_ioc_get_encryption_policy(filp, arg);
3991 case FS_IOC_GET_ENCRYPTION_PWSALT:
3992 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3993 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3994 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3995 case FS_IOC_ADD_ENCRYPTION_KEY:
3996 return f2fs_ioc_add_encryption_key(filp, arg);
3997 case FS_IOC_REMOVE_ENCRYPTION_KEY:
3998 return f2fs_ioc_remove_encryption_key(filp, arg);
3999 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4000 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4001 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4002 return f2fs_ioc_get_encryption_key_status(filp, arg);
4003 case FS_IOC_GET_ENCRYPTION_NONCE:
4004 return f2fs_ioc_get_encryption_nonce(filp, arg);
4005 case F2FS_IOC_GARBAGE_COLLECT:
4006 return f2fs_ioc_gc(filp, arg);
4007 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4008 return f2fs_ioc_gc_range(filp, arg);
4009 case F2FS_IOC_WRITE_CHECKPOINT:
4010 return f2fs_ioc_write_checkpoint(filp, arg);
4011 case F2FS_IOC_DEFRAGMENT:
4012 return f2fs_ioc_defragment(filp, arg);
4013 case F2FS_IOC_MOVE_RANGE:
4014 return f2fs_ioc_move_range(filp, arg);
4015 case F2FS_IOC_FLUSH_DEVICE:
4016 return f2fs_ioc_flush_device(filp, arg);
4017 case F2FS_IOC_GET_FEATURES:
4018 return f2fs_ioc_get_features(filp, arg);
4019 case FS_IOC_FSGETXATTR:
4020 return f2fs_ioc_fsgetxattr(filp, arg);
4021 case FS_IOC_FSSETXATTR:
4022 return f2fs_ioc_fssetxattr(filp, arg);
4023 case F2FS_IOC_GET_PIN_FILE:
4024 return f2fs_ioc_get_pin_file(filp, arg);
4025 case F2FS_IOC_SET_PIN_FILE:
4026 return f2fs_ioc_set_pin_file(filp, arg);
4027 case F2FS_IOC_PRECACHE_EXTENTS:
4028 return f2fs_ioc_precache_extents(filp, arg);
4029 case F2FS_IOC_RESIZE_FS:
4030 return f2fs_ioc_resize_fs(filp, arg);
4031 case FS_IOC_ENABLE_VERITY:
4032 return f2fs_ioc_enable_verity(filp, arg);
4033 case FS_IOC_MEASURE_VERITY:
4034 return f2fs_ioc_measure_verity(filp, arg);
4035 case FS_IOC_GETFSLABEL:
4036 return f2fs_ioc_getfslabel(filp, arg);
4037 case FS_IOC_SETFSLABEL:
4038 return f2fs_ioc_setfslabel(filp, arg);
4039 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4040 return f2fs_get_compress_blocks(filp, arg);
4041 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4042 return f2fs_release_compress_blocks(filp, arg);
4043 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4044 return f2fs_reserve_compress_blocks(filp, arg);
4045 case F2FS_IOC_SEC_TRIM_FILE:
4046 return f2fs_sec_trim_file(filp, arg);
4047 default:
4048 return -ENOTTY;
4049 }
4050 }
4051
f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4052 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4053 {
4054 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4055 return -EIO;
4056 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4057 return -ENOSPC;
4058
4059 return __f2fs_ioctl(filp, cmd, arg);
4060 }
4061
f2fs_file_read_iter(struct kiocb * iocb,struct iov_iter * iter)4062 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
4063 {
4064 struct file *file = iocb->ki_filp;
4065 struct inode *inode = file_inode(file);
4066 int ret;
4067
4068 if (!f2fs_is_compress_backend_ready(inode))
4069 return -EOPNOTSUPP;
4070
4071 ret = generic_file_read_iter(iocb, iter);
4072
4073 if (ret > 0)
4074 f2fs_update_iostat(F2FS_I_SB(inode), APP_READ_IO, ret);
4075
4076 return ret;
4077 }
4078
f2fs_file_write_iter(struct kiocb * iocb,struct iov_iter * from)4079 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4080 {
4081 struct file *file = iocb->ki_filp;
4082 struct inode *inode = file_inode(file);
4083 ssize_t ret;
4084
4085 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4086 ret = -EIO;
4087 goto out;
4088 }
4089
4090 if (!f2fs_is_compress_backend_ready(inode)) {
4091 ret = -EOPNOTSUPP;
4092 goto out;
4093 }
4094
4095 if (iocb->ki_flags & IOCB_NOWAIT) {
4096 if (!inode_trylock(inode)) {
4097 ret = -EAGAIN;
4098 goto out;
4099 }
4100 } else {
4101 inode_lock(inode);
4102 }
4103
4104 if (unlikely(IS_IMMUTABLE(inode))) {
4105 ret = -EPERM;
4106 goto unlock;
4107 }
4108
4109 ret = generic_write_checks(iocb, from);
4110 if (ret > 0) {
4111 bool preallocated = false;
4112 size_t target_size = 0;
4113 int err;
4114
4115 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
4116 set_inode_flag(inode, FI_NO_PREALLOC);
4117
4118 if ((iocb->ki_flags & IOCB_NOWAIT)) {
4119 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
4120 iov_iter_count(from)) ||
4121 f2fs_has_inline_data(inode) ||
4122 f2fs_force_buffered_io(inode, iocb, from)) {
4123 clear_inode_flag(inode, FI_NO_PREALLOC);
4124 inode_unlock(inode);
4125 ret = -EAGAIN;
4126 goto out;
4127 }
4128 goto write;
4129 }
4130
4131 if (is_inode_flag_set(inode, FI_NO_PREALLOC))
4132 goto write;
4133
4134 if (iocb->ki_flags & IOCB_DIRECT) {
4135 /*
4136 * Convert inline data for Direct I/O before entering
4137 * f2fs_direct_IO().
4138 */
4139 err = f2fs_convert_inline_inode(inode);
4140 if (err)
4141 goto out_err;
4142 /*
4143 * If force_buffere_io() is true, we have to allocate
4144 * blocks all the time, since f2fs_direct_IO will fall
4145 * back to buffered IO.
4146 */
4147 if (!f2fs_force_buffered_io(inode, iocb, from) &&
4148 allow_outplace_dio(inode, iocb, from))
4149 goto write;
4150 }
4151 preallocated = true;
4152 target_size = iocb->ki_pos + iov_iter_count(from);
4153
4154 err = f2fs_preallocate_blocks(iocb, from);
4155 if (err) {
4156 out_err:
4157 clear_inode_flag(inode, FI_NO_PREALLOC);
4158 inode_unlock(inode);
4159 ret = err;
4160 goto out;
4161 }
4162 write:
4163 ret = __generic_file_write_iter(iocb, from);
4164 clear_inode_flag(inode, FI_NO_PREALLOC);
4165
4166 /* if we couldn't write data, we should deallocate blocks. */
4167 if (preallocated && i_size_read(inode) < target_size)
4168 f2fs_truncate(inode);
4169
4170 if (ret > 0)
4171 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
4172 }
4173 unlock:
4174 inode_unlock(inode);
4175 out:
4176 trace_f2fs_file_write_iter(inode, iocb->ki_pos,
4177 iov_iter_count(from), ret);
4178 if (ret > 0)
4179 ret = generic_write_sync(iocb, ret);
4180 return ret;
4181 }
4182
4183 #ifdef CONFIG_COMPAT
4184 struct compat_f2fs_gc_range {
4185 u32 sync;
4186 compat_u64 start;
4187 compat_u64 len;
4188 };
4189 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
4190 struct compat_f2fs_gc_range)
4191
f2fs_compat_ioc_gc_range(struct file * file,unsigned long arg)4192 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4193 {
4194 struct compat_f2fs_gc_range __user *urange;
4195 struct f2fs_gc_range range;
4196 int err;
4197
4198 urange = compat_ptr(arg);
4199 err = get_user(range.sync, &urange->sync);
4200 err |= get_user(range.start, &urange->start);
4201 err |= get_user(range.len, &urange->len);
4202 if (err)
4203 return -EFAULT;
4204
4205 return __f2fs_ioc_gc_range(file, &range);
4206 }
4207
4208 struct compat_f2fs_move_range {
4209 u32 dst_fd;
4210 compat_u64 pos_in;
4211 compat_u64 pos_out;
4212 compat_u64 len;
4213 };
4214 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
4215 struct compat_f2fs_move_range)
4216
f2fs_compat_ioc_move_range(struct file * file,unsigned long arg)4217 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4218 {
4219 struct compat_f2fs_move_range __user *urange;
4220 struct f2fs_move_range range;
4221 int err;
4222
4223 urange = compat_ptr(arg);
4224 err = get_user(range.dst_fd, &urange->dst_fd);
4225 err |= get_user(range.pos_in, &urange->pos_in);
4226 err |= get_user(range.pos_out, &urange->pos_out);
4227 err |= get_user(range.len, &urange->len);
4228 if (err)
4229 return -EFAULT;
4230
4231 return __f2fs_ioc_move_range(file, &range);
4232 }
4233
f2fs_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)4234 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4235 {
4236 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4237 return -EIO;
4238 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4239 return -ENOSPC;
4240
4241 switch (cmd) {
4242 case FS_IOC32_GETFLAGS:
4243 cmd = FS_IOC_GETFLAGS;
4244 break;
4245 case FS_IOC32_SETFLAGS:
4246 cmd = FS_IOC_SETFLAGS;
4247 break;
4248 case FS_IOC32_GETVERSION:
4249 cmd = FS_IOC_GETVERSION;
4250 break;
4251 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4252 return f2fs_compat_ioc_gc_range(file, arg);
4253 case F2FS_IOC32_MOVE_RANGE:
4254 return f2fs_compat_ioc_move_range(file, arg);
4255 case F2FS_IOC_START_ATOMIC_WRITE:
4256 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4257 case F2FS_IOC_START_VOLATILE_WRITE:
4258 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4259 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4260 case F2FS_IOC_SHUTDOWN:
4261 case FITRIM:
4262 case FS_IOC_SET_ENCRYPTION_POLICY:
4263 case FS_IOC_GET_ENCRYPTION_PWSALT:
4264 case FS_IOC_GET_ENCRYPTION_POLICY:
4265 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4266 case FS_IOC_ADD_ENCRYPTION_KEY:
4267 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4268 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4269 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4270 case FS_IOC_GET_ENCRYPTION_NONCE:
4271 case F2FS_IOC_GARBAGE_COLLECT:
4272 case F2FS_IOC_WRITE_CHECKPOINT:
4273 case F2FS_IOC_DEFRAGMENT:
4274 case F2FS_IOC_FLUSH_DEVICE:
4275 case F2FS_IOC_GET_FEATURES:
4276 case FS_IOC_FSGETXATTR:
4277 case FS_IOC_FSSETXATTR:
4278 case F2FS_IOC_GET_PIN_FILE:
4279 case F2FS_IOC_SET_PIN_FILE:
4280 case F2FS_IOC_PRECACHE_EXTENTS:
4281 case F2FS_IOC_RESIZE_FS:
4282 case FS_IOC_ENABLE_VERITY:
4283 case FS_IOC_MEASURE_VERITY:
4284 case FS_IOC_GETFSLABEL:
4285 case FS_IOC_SETFSLABEL:
4286 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4287 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4288 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4289 case F2FS_IOC_SEC_TRIM_FILE:
4290 break;
4291 default:
4292 return -ENOIOCTLCMD;
4293 }
4294 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4295 }
4296 #endif
4297
4298 const struct file_operations f2fs_file_operations = {
4299 .llseek = f2fs_llseek,
4300 .read_iter = f2fs_file_read_iter,
4301 .write_iter = f2fs_file_write_iter,
4302 .open = f2fs_file_open,
4303 .release = f2fs_release_file,
4304 .mmap = f2fs_file_mmap,
4305 .flush = f2fs_file_flush,
4306 .fsync = f2fs_sync_file,
4307 .fallocate = f2fs_fallocate,
4308 .unlocked_ioctl = f2fs_ioctl,
4309 #ifdef CONFIG_COMPAT
4310 .compat_ioctl = f2fs_compat_ioctl,
4311 #endif
4312 .splice_read = generic_file_splice_read,
4313 .splice_write = iter_file_splice_write,
4314 };
4315