1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/file.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/buffer_head.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/types.h>
16 #include <linux/compat.h>
17 #include <linux/uaccess.h>
18 #include <linux/mount.h>
19 #include <linux/pagevec.h>
20 #include <linux/uio.h>
21 #include <linux/uuid.h>
22 #include <linux/file.h>
23 #include <linux/nls.h>
24 #include <linux/sched/signal.h>
25 #include <linux/fileattr.h>
26 #include <linux/fadvise.h>
27 #include <linux/iomap.h>
28
29 #include "f2fs.h"
30 #include "node.h"
31 #include "segment.h"
32 #include "xattr.h"
33 #include "acl.h"
34 #include "gc.h"
35 #include "iostat.h"
36 #include <trace/events/f2fs.h>
37 #include <uapi/linux/f2fs.h>
38
f2fs_filemap_fault(struct vm_fault * vmf)39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
40 {
41 struct inode *inode = file_inode(vmf->vma->vm_file);
42 vm_fault_t ret;
43
44 ret = filemap_fault(vmf);
45 if (!ret)
46 f2fs_update_iostat(F2FS_I_SB(inode), inode,
47 APP_MAPPED_READ_IO, 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 (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
67 return VM_FAULT_SIGBUS;
68
69 if (unlikely(f2fs_cp_error(sbi))) {
70 err = -EIO;
71 goto err;
72 }
73
74 if (!f2fs_is_checkpoint_ready(sbi)) {
75 err = -ENOSPC;
76 goto err;
77 }
78
79 err = f2fs_convert_inline_inode(inode);
80 if (err)
81 goto err;
82
83 #ifdef CONFIG_F2FS_FS_COMPRESSION
84 if (f2fs_compressed_file(inode)) {
85 int ret = f2fs_is_compressed_cluster(inode, page->index);
86
87 if (ret < 0) {
88 err = ret;
89 goto err;
90 } else if (ret) {
91 need_alloc = false;
92 }
93 }
94 #endif
95 /* should do out of any locked page */
96 if (need_alloc)
97 f2fs_balance_fs(sbi, true);
98
99 sb_start_pagefault(inode->i_sb);
100
101 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
102
103 file_update_time(vmf->vma->vm_file);
104 filemap_invalidate_lock_shared(inode->i_mapping);
105 lock_page(page);
106 if (unlikely(page->mapping != inode->i_mapping ||
107 page_offset(page) > i_size_read(inode) ||
108 !PageUptodate(page))) {
109 unlock_page(page);
110 err = -EFAULT;
111 goto out_sem;
112 }
113
114 if (need_alloc) {
115 /* block allocation */
116 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
117 set_new_dnode(&dn, inode, NULL, NULL, 0);
118 err = f2fs_get_block(&dn, page->index);
119 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
120 }
121
122 #ifdef CONFIG_F2FS_FS_COMPRESSION
123 if (!need_alloc) {
124 set_new_dnode(&dn, inode, NULL, NULL, 0);
125 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
126 f2fs_put_dnode(&dn);
127 }
128 #endif
129 if (err) {
130 unlock_page(page);
131 goto out_sem;
132 }
133
134 f2fs_wait_on_page_writeback(page, DATA, false, true);
135
136 /* wait for GCed page writeback via META_MAPPING */
137 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
138
139 /*
140 * check to see if the page is mapped already (no holes)
141 */
142 if (PageMappedToDisk(page))
143 goto out_sem;
144
145 /* page is wholly or partially inside EOF */
146 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
147 i_size_read(inode)) {
148 loff_t offset;
149
150 offset = i_size_read(inode) & ~PAGE_MASK;
151 zero_user_segment(page, offset, PAGE_SIZE);
152 }
153 set_page_dirty(page);
154 if (!PageUptodate(page))
155 SetPageUptodate(page);
156
157 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
158 f2fs_update_time(sbi, REQ_TIME);
159
160 trace_f2fs_vm_page_mkwrite(page, DATA);
161 out_sem:
162 filemap_invalidate_unlock_shared(inode->i_mapping);
163
164 sb_end_pagefault(inode->i_sb);
165 err:
166 return block_page_mkwrite_return(err);
167 }
168
169 static const struct vm_operations_struct f2fs_file_vm_ops = {
170 .fault = f2fs_filemap_fault,
171 .map_pages = filemap_map_pages,
172 .page_mkwrite = f2fs_vm_page_mkwrite,
173 };
174
get_parent_ino(struct inode * inode,nid_t * pino)175 static int get_parent_ino(struct inode *inode, nid_t *pino)
176 {
177 struct dentry *dentry;
178
179 /*
180 * Make sure to get the non-deleted alias. The alias associated with
181 * the open file descriptor being fsync()'ed may be deleted already.
182 */
183 dentry = d_find_alias(inode);
184 if (!dentry)
185 return 0;
186
187 *pino = parent_ino(dentry);
188 dput(dentry);
189 return 1;
190 }
191
need_do_checkpoint(struct inode * inode)192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
193 {
194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
195 enum cp_reason_type cp_reason = CP_NO_NEEDED;
196
197 if (!S_ISREG(inode->i_mode))
198 cp_reason = CP_NON_REGULAR;
199 else if (f2fs_compressed_file(inode))
200 cp_reason = CP_COMPRESSED;
201 else if (inode->i_nlink != 1)
202 cp_reason = CP_HARDLINK;
203 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
204 cp_reason = CP_SB_NEED_CP;
205 else if (file_wrong_pino(inode))
206 cp_reason = CP_WRONG_PINO;
207 else if (!f2fs_space_for_roll_forward(sbi))
208 cp_reason = CP_NO_SPC_ROLL;
209 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
210 cp_reason = CP_NODE_NEED_CP;
211 else if (test_opt(sbi, FASTBOOT))
212 cp_reason = CP_FASTBOOT_MODE;
213 else if (F2FS_OPTION(sbi).active_logs == 2)
214 cp_reason = CP_SPEC_LOG_NUM;
215 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
216 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
217 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
218 TRANS_DIR_INO))
219 cp_reason = CP_RECOVER_DIR;
220
221 return cp_reason;
222 }
223
need_inode_page_update(struct f2fs_sb_info * sbi,nid_t ino)224 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
225 {
226 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
227 bool ret = false;
228 /* But we need to avoid that there are some inode updates */
229 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
230 ret = true;
231 f2fs_put_page(i, 0);
232 return ret;
233 }
234
try_to_fix_pino(struct inode * inode)235 static void try_to_fix_pino(struct inode *inode)
236 {
237 struct f2fs_inode_info *fi = F2FS_I(inode);
238 nid_t pino;
239
240 f2fs_down_write(&fi->i_sem);
241 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
242 get_parent_ino(inode, &pino)) {
243 f2fs_i_pino_write(inode, pino);
244 file_got_pino(inode);
245 }
246 f2fs_up_write(&fi->i_sem);
247 }
248
f2fs_do_sync_file(struct file * file,loff_t start,loff_t end,int datasync,bool atomic)249 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
250 int datasync, bool atomic)
251 {
252 struct inode *inode = file->f_mapping->host;
253 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
254 nid_t ino = inode->i_ino;
255 int ret = 0;
256 enum cp_reason_type cp_reason = 0;
257 struct writeback_control wbc = {
258 .sync_mode = WB_SYNC_ALL,
259 .nr_to_write = LONG_MAX,
260 .for_reclaim = 0,
261 };
262 unsigned int seq_id = 0;
263
264 if (unlikely(f2fs_readonly(inode->i_sb)))
265 return 0;
266
267 trace_f2fs_sync_file_enter(inode);
268
269 if (S_ISDIR(inode->i_mode))
270 goto go_write;
271
272 /* if fdatasync is triggered, let's do in-place-update */
273 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
274 set_inode_flag(inode, FI_NEED_IPU);
275 ret = file_write_and_wait_range(file, start, end);
276 clear_inode_flag(inode, FI_NEED_IPU);
277
278 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
279 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
280 return ret;
281 }
282
283 /* if the inode is dirty, let's recover all the time */
284 if (!f2fs_skip_inode_update(inode, datasync)) {
285 f2fs_write_inode(inode, NULL);
286 goto go_write;
287 }
288
289 /*
290 * if there is no written data, don't waste time to write recovery info.
291 */
292 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
293 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
294
295 /* it may call write_inode just prior to fsync */
296 if (need_inode_page_update(sbi, ino))
297 goto go_write;
298
299 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
300 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
301 goto flush_out;
302 goto out;
303 } else {
304 /*
305 * for OPU case, during fsync(), node can be persisted before
306 * data when lower device doesn't support write barrier, result
307 * in data corruption after SPO.
308 * So for strict fsync mode, force to use atomic write sematics
309 * to keep write order in between data/node and last node to
310 * avoid potential data corruption.
311 */
312 if (F2FS_OPTION(sbi).fsync_mode ==
313 FSYNC_MODE_STRICT && !atomic)
314 atomic = true;
315 }
316 go_write:
317 /*
318 * Both of fdatasync() and fsync() are able to be recovered from
319 * sudden-power-off.
320 */
321 f2fs_down_read(&F2FS_I(inode)->i_sem);
322 cp_reason = need_do_checkpoint(inode);
323 f2fs_up_read(&F2FS_I(inode)->i_sem);
324
325 if (cp_reason) {
326 /* all the dirty node pages should be flushed for POR */
327 ret = f2fs_sync_fs(inode->i_sb, 1);
328
329 /*
330 * We've secured consistency through sync_fs. Following pino
331 * will be used only for fsynced inodes after checkpoint.
332 */
333 try_to_fix_pino(inode);
334 clear_inode_flag(inode, FI_APPEND_WRITE);
335 clear_inode_flag(inode, FI_UPDATE_WRITE);
336 goto out;
337 }
338 sync_nodes:
339 atomic_inc(&sbi->wb_sync_req[NODE]);
340 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
341 atomic_dec(&sbi->wb_sync_req[NODE]);
342 if (ret)
343 goto out;
344
345 /* if cp_error was enabled, we should avoid infinite loop */
346 if (unlikely(f2fs_cp_error(sbi))) {
347 ret = -EIO;
348 goto out;
349 }
350
351 if (f2fs_need_inode_block_update(sbi, ino)) {
352 f2fs_mark_inode_dirty_sync(inode, true);
353 f2fs_write_inode(inode, NULL);
354 goto sync_nodes;
355 }
356
357 /*
358 * If it's atomic_write, it's just fine to keep write ordering. So
359 * here we don't need to wait for node write completion, since we use
360 * node chain which serializes node blocks. If one of node writes are
361 * reordered, we can see simply broken chain, resulting in stopping
362 * roll-forward recovery. It means we'll recover all or none node blocks
363 * given fsync mark.
364 */
365 if (!atomic) {
366 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
367 if (ret)
368 goto out;
369 }
370
371 /* once recovery info is written, don't need to tack this */
372 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
373 clear_inode_flag(inode, FI_APPEND_WRITE);
374 flush_out:
375 if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) ||
376 (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
377 ret = f2fs_issue_flush(sbi, inode->i_ino);
378 if (!ret) {
379 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
380 clear_inode_flag(inode, FI_UPDATE_WRITE);
381 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
382 }
383 f2fs_update_time(sbi, REQ_TIME);
384 out:
385 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
386 return ret;
387 }
388
f2fs_sync_file(struct file * file,loff_t start,loff_t end,int datasync)389 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
390 {
391 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
392 return -EIO;
393 return f2fs_do_sync_file(file, start, end, datasync, false);
394 }
395
__found_offset(struct address_space * mapping,block_t blkaddr,pgoff_t index,int whence)396 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
397 pgoff_t index, int whence)
398 {
399 switch (whence) {
400 case SEEK_DATA:
401 if (__is_valid_data_blkaddr(blkaddr))
402 return true;
403 if (blkaddr == NEW_ADDR &&
404 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
405 return true;
406 break;
407 case SEEK_HOLE:
408 if (blkaddr == NULL_ADDR)
409 return true;
410 break;
411 }
412 return false;
413 }
414
f2fs_seek_block(struct file * file,loff_t offset,int whence)415 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
416 {
417 struct inode *inode = file->f_mapping->host;
418 loff_t maxbytes = inode->i_sb->s_maxbytes;
419 struct dnode_of_data dn;
420 pgoff_t pgofs, end_offset;
421 loff_t data_ofs = offset;
422 loff_t isize;
423 int err = 0;
424
425 inode_lock(inode);
426
427 isize = i_size_read(inode);
428 if (offset >= isize)
429 goto fail;
430
431 /* handle inline data case */
432 if (f2fs_has_inline_data(inode)) {
433 if (whence == SEEK_HOLE) {
434 data_ofs = isize;
435 goto found;
436 } else if (whence == SEEK_DATA) {
437 data_ofs = offset;
438 goto found;
439 }
440 }
441
442 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
443
444 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
445 set_new_dnode(&dn, inode, NULL, NULL, 0);
446 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
447 if (err && err != -ENOENT) {
448 goto fail;
449 } else if (err == -ENOENT) {
450 /* direct node does not exists */
451 if (whence == SEEK_DATA) {
452 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
453 continue;
454 } else {
455 goto found;
456 }
457 }
458
459 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
460
461 /* find data/hole in dnode block */
462 for (; dn.ofs_in_node < end_offset;
463 dn.ofs_in_node++, pgofs++,
464 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
465 block_t blkaddr;
466
467 blkaddr = f2fs_data_blkaddr(&dn);
468
469 if (__is_valid_data_blkaddr(blkaddr) &&
470 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
471 blkaddr, DATA_GENERIC_ENHANCE)) {
472 f2fs_put_dnode(&dn);
473 goto fail;
474 }
475
476 if (__found_offset(file->f_mapping, blkaddr,
477 pgofs, whence)) {
478 f2fs_put_dnode(&dn);
479 goto found;
480 }
481 }
482 f2fs_put_dnode(&dn);
483 }
484
485 if (whence == SEEK_DATA)
486 goto fail;
487 found:
488 if (whence == SEEK_HOLE && data_ofs > isize)
489 data_ofs = isize;
490 inode_unlock(inode);
491 return vfs_setpos(file, data_ofs, maxbytes);
492 fail:
493 inode_unlock(inode);
494 return -ENXIO;
495 }
496
f2fs_llseek(struct file * file,loff_t offset,int whence)497 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
498 {
499 struct inode *inode = file->f_mapping->host;
500 loff_t maxbytes = inode->i_sb->s_maxbytes;
501
502 if (f2fs_compressed_file(inode))
503 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
504
505 switch (whence) {
506 case SEEK_SET:
507 case SEEK_CUR:
508 case SEEK_END:
509 return generic_file_llseek_size(file, offset, whence,
510 maxbytes, i_size_read(inode));
511 case SEEK_DATA:
512 case SEEK_HOLE:
513 if (offset < 0)
514 return -ENXIO;
515 return f2fs_seek_block(file, offset, whence);
516 }
517
518 return -EINVAL;
519 }
520
f2fs_file_mmap(struct file * file,struct vm_area_struct * vma)521 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
522 {
523 struct inode *inode = file_inode(file);
524
525 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
526 return -EIO;
527
528 if (!f2fs_is_compress_backend_ready(inode))
529 return -EOPNOTSUPP;
530
531 file_accessed(file);
532 vma->vm_ops = &f2fs_file_vm_ops;
533 set_inode_flag(inode, FI_MMAP_FILE);
534 return 0;
535 }
536
f2fs_file_open(struct inode * inode,struct file * filp)537 static int f2fs_file_open(struct inode *inode, struct file *filp)
538 {
539 int err = fscrypt_file_open(inode, filp);
540
541 if (err)
542 return err;
543
544 if (!f2fs_is_compress_backend_ready(inode))
545 return -EOPNOTSUPP;
546
547 err = fsverity_file_open(inode, filp);
548 if (err)
549 return err;
550
551 filp->f_mode |= FMODE_NOWAIT;
552
553 return dquot_file_open(inode, filp);
554 }
555
f2fs_truncate_data_blocks_range(struct dnode_of_data * dn,int count)556 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
557 {
558 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
559 struct f2fs_node *raw_node;
560 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
561 __le32 *addr;
562 int base = 0;
563 bool compressed_cluster = false;
564 int cluster_index = 0, valid_blocks = 0;
565 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
566 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
567
568 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
569 base = get_extra_isize(dn->inode);
570
571 raw_node = F2FS_NODE(dn->node_page);
572 addr = blkaddr_in_node(raw_node) + base + ofs;
573
574 /* Assumption: truncateion starts with cluster */
575 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
576 block_t blkaddr = le32_to_cpu(*addr);
577
578 if (f2fs_compressed_file(dn->inode) &&
579 !(cluster_index & (cluster_size - 1))) {
580 if (compressed_cluster)
581 f2fs_i_compr_blocks_update(dn->inode,
582 valid_blocks, false);
583 compressed_cluster = (blkaddr == COMPRESS_ADDR);
584 valid_blocks = 0;
585 }
586
587 if (blkaddr == NULL_ADDR)
588 continue;
589
590 dn->data_blkaddr = NULL_ADDR;
591 f2fs_set_data_blkaddr(dn);
592
593 if (__is_valid_data_blkaddr(blkaddr)) {
594 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
595 DATA_GENERIC_ENHANCE))
596 continue;
597 if (compressed_cluster)
598 valid_blocks++;
599 }
600
601 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
602 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
603
604 f2fs_invalidate_blocks(sbi, blkaddr);
605
606 if (!released || blkaddr != COMPRESS_ADDR)
607 nr_free++;
608 }
609
610 if (compressed_cluster)
611 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
612
613 if (nr_free) {
614 pgoff_t fofs;
615 /*
616 * once we invalidate valid blkaddr in range [ofs, ofs + count],
617 * we will invalidate all blkaddr in the whole range.
618 */
619 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
620 dn->inode) + ofs;
621 f2fs_update_extent_cache_range(dn, fofs, 0, len);
622 dec_valid_block_count(sbi, dn->inode, nr_free);
623 }
624 dn->ofs_in_node = ofs;
625
626 f2fs_update_time(sbi, REQ_TIME);
627 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
628 dn->ofs_in_node, nr_free);
629 }
630
f2fs_truncate_data_blocks(struct dnode_of_data * dn)631 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
632 {
633 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
634 }
635
truncate_partial_data_page(struct inode * inode,u64 from,bool cache_only)636 static int truncate_partial_data_page(struct inode *inode, u64 from,
637 bool cache_only)
638 {
639 loff_t offset = from & (PAGE_SIZE - 1);
640 pgoff_t index = from >> PAGE_SHIFT;
641 struct address_space *mapping = inode->i_mapping;
642 struct page *page;
643
644 if (!offset && !cache_only)
645 return 0;
646
647 if (cache_only) {
648 page = find_lock_page(mapping, index);
649 if (page && PageUptodate(page))
650 goto truncate_out;
651 f2fs_put_page(page, 1);
652 return 0;
653 }
654
655 page = f2fs_get_lock_data_page(inode, index, true);
656 if (IS_ERR(page))
657 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
658 truncate_out:
659 f2fs_wait_on_page_writeback(page, DATA, true, true);
660 zero_user(page, offset, PAGE_SIZE - offset);
661
662 /* An encrypted inode should have a key and truncate the last page. */
663 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
664 if (!cache_only)
665 set_page_dirty(page);
666 f2fs_put_page(page, 1);
667 return 0;
668 }
669
f2fs_do_truncate_blocks(struct inode * inode,u64 from,bool lock)670 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
671 {
672 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
673 struct dnode_of_data dn;
674 pgoff_t free_from;
675 int count = 0, err = 0;
676 struct page *ipage;
677 bool truncate_page = false;
678
679 trace_f2fs_truncate_blocks_enter(inode, from);
680
681 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
682
683 if (free_from >= max_file_blocks(inode))
684 goto free_partial;
685
686 if (lock)
687 f2fs_lock_op(sbi);
688
689 ipage = f2fs_get_node_page(sbi, inode->i_ino);
690 if (IS_ERR(ipage)) {
691 err = PTR_ERR(ipage);
692 goto out;
693 }
694
695 if (f2fs_has_inline_data(inode)) {
696 f2fs_truncate_inline_inode(inode, ipage, from);
697 f2fs_put_page(ipage, 1);
698 truncate_page = true;
699 goto out;
700 }
701
702 set_new_dnode(&dn, inode, ipage, NULL, 0);
703 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
704 if (err) {
705 if (err == -ENOENT)
706 goto free_next;
707 goto out;
708 }
709
710 count = ADDRS_PER_PAGE(dn.node_page, inode);
711
712 count -= dn.ofs_in_node;
713 f2fs_bug_on(sbi, count < 0);
714
715 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
716 f2fs_truncate_data_blocks_range(&dn, count);
717 free_from += count;
718 }
719
720 f2fs_put_dnode(&dn);
721 free_next:
722 err = f2fs_truncate_inode_blocks(inode, free_from);
723 out:
724 if (lock)
725 f2fs_unlock_op(sbi);
726 free_partial:
727 /* lastly zero out the first data page */
728 if (!err)
729 err = truncate_partial_data_page(inode, from, truncate_page);
730
731 trace_f2fs_truncate_blocks_exit(inode, err);
732 return err;
733 }
734
f2fs_truncate_blocks(struct inode * inode,u64 from,bool lock)735 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
736 {
737 u64 free_from = from;
738 int err;
739
740 #ifdef CONFIG_F2FS_FS_COMPRESSION
741 /*
742 * for compressed file, only support cluster size
743 * aligned truncation.
744 */
745 if (f2fs_compressed_file(inode))
746 free_from = round_up(from,
747 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
748 #endif
749
750 err = f2fs_do_truncate_blocks(inode, free_from, lock);
751 if (err)
752 return err;
753
754 #ifdef CONFIG_F2FS_FS_COMPRESSION
755 /*
756 * For compressed file, after release compress blocks, don't allow write
757 * direct, but we should allow write direct after truncate to zero.
758 */
759 if (f2fs_compressed_file(inode) && !free_from
760 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
761 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
762
763 if (from != free_from) {
764 err = f2fs_truncate_partial_cluster(inode, from, lock);
765 if (err)
766 return err;
767 }
768 #endif
769
770 return 0;
771 }
772
f2fs_truncate(struct inode * inode)773 int f2fs_truncate(struct inode *inode)
774 {
775 int err;
776
777 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
778 return -EIO;
779
780 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
781 S_ISLNK(inode->i_mode)))
782 return 0;
783
784 trace_f2fs_truncate(inode);
785
786 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
787 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
788 return -EIO;
789 }
790
791 err = f2fs_dquot_initialize(inode);
792 if (err)
793 return err;
794
795 /* we should check inline_data size */
796 if (!f2fs_may_inline_data(inode)) {
797 err = f2fs_convert_inline_inode(inode);
798 if (err)
799 return err;
800 }
801
802 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
803 if (err)
804 return err;
805
806 inode->i_mtime = inode->i_ctime = current_time(inode);
807 f2fs_mark_inode_dirty_sync(inode, false);
808 return 0;
809 }
810
f2fs_force_buffered_io(struct inode * inode,int rw)811 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
812 {
813 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
814
815 if (!fscrypt_dio_supported(inode))
816 return true;
817 if (fsverity_active(inode))
818 return true;
819 if (f2fs_compressed_file(inode))
820 return true;
821
822 /* disallow direct IO if any of devices has unaligned blksize */
823 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
824 return true;
825 /*
826 * for blkzoned device, fallback direct IO to buffered IO, so
827 * all IOs can be serialized by log-structured write.
828 */
829 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE))
830 return true;
831 if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi))
832 return true;
833 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
834 return true;
835
836 return false;
837 }
838
f2fs_getattr(struct user_namespace * mnt_userns,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)839 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
840 struct kstat *stat, u32 request_mask, unsigned int query_flags)
841 {
842 struct inode *inode = d_inode(path->dentry);
843 struct f2fs_inode_info *fi = F2FS_I(inode);
844 struct f2fs_inode *ri = NULL;
845 unsigned int flags;
846
847 if (f2fs_has_extra_attr(inode) &&
848 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
849 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
850 stat->result_mask |= STATX_BTIME;
851 stat->btime.tv_sec = fi->i_crtime.tv_sec;
852 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
853 }
854
855 /*
856 * Return the DIO alignment restrictions if requested. We only return
857 * this information when requested, since on encrypted files it might
858 * take a fair bit of work to get if the file wasn't opened recently.
859 *
860 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN
861 * cannot represent that, so in that case we report no DIO support.
862 */
863 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
864 unsigned int bsize = i_blocksize(inode);
865
866 stat->result_mask |= STATX_DIOALIGN;
867 if (!f2fs_force_buffered_io(inode, WRITE)) {
868 stat->dio_mem_align = bsize;
869 stat->dio_offset_align = bsize;
870 }
871 }
872
873 flags = fi->i_flags;
874 if (flags & F2FS_COMPR_FL)
875 stat->attributes |= STATX_ATTR_COMPRESSED;
876 if (flags & F2FS_APPEND_FL)
877 stat->attributes |= STATX_ATTR_APPEND;
878 if (IS_ENCRYPTED(inode))
879 stat->attributes |= STATX_ATTR_ENCRYPTED;
880 if (flags & F2FS_IMMUTABLE_FL)
881 stat->attributes |= STATX_ATTR_IMMUTABLE;
882 if (flags & F2FS_NODUMP_FL)
883 stat->attributes |= STATX_ATTR_NODUMP;
884 if (IS_VERITY(inode))
885 stat->attributes |= STATX_ATTR_VERITY;
886
887 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
888 STATX_ATTR_APPEND |
889 STATX_ATTR_ENCRYPTED |
890 STATX_ATTR_IMMUTABLE |
891 STATX_ATTR_NODUMP |
892 STATX_ATTR_VERITY);
893
894 generic_fillattr(mnt_userns, inode, stat);
895
896 /* we need to show initial sectors used for inline_data/dentries */
897 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
898 f2fs_has_inline_dentry(inode))
899 stat->blocks += (stat->size + 511) >> 9;
900
901 return 0;
902 }
903
904 #ifdef CONFIG_F2FS_FS_POSIX_ACL
__setattr_copy(struct user_namespace * mnt_userns,struct inode * inode,const struct iattr * attr)905 static void __setattr_copy(struct user_namespace *mnt_userns,
906 struct inode *inode, const struct iattr *attr)
907 {
908 unsigned int ia_valid = attr->ia_valid;
909
910 i_uid_update(mnt_userns, attr, inode);
911 i_gid_update(mnt_userns, attr, inode);
912 if (ia_valid & ATTR_ATIME)
913 inode->i_atime = attr->ia_atime;
914 if (ia_valid & ATTR_MTIME)
915 inode->i_mtime = attr->ia_mtime;
916 if (ia_valid & ATTR_CTIME)
917 inode->i_ctime = attr->ia_ctime;
918 if (ia_valid & ATTR_MODE) {
919 umode_t mode = attr->ia_mode;
920 vfsgid_t vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
921
922 if (!vfsgid_in_group_p(vfsgid) &&
923 !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
924 mode &= ~S_ISGID;
925 set_acl_inode(inode, mode);
926 }
927 }
928 #else
929 #define __setattr_copy setattr_copy
930 #endif
931
f2fs_setattr(struct user_namespace * mnt_userns,struct dentry * dentry,struct iattr * attr)932 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
933 struct iattr *attr)
934 {
935 struct inode *inode = d_inode(dentry);
936 int err;
937
938 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
939 return -EIO;
940
941 if (unlikely(IS_IMMUTABLE(inode)))
942 return -EPERM;
943
944 if (unlikely(IS_APPEND(inode) &&
945 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
946 ATTR_GID | ATTR_TIMES_SET))))
947 return -EPERM;
948
949 if ((attr->ia_valid & ATTR_SIZE) &&
950 !f2fs_is_compress_backend_ready(inode))
951 return -EOPNOTSUPP;
952
953 err = setattr_prepare(mnt_userns, dentry, attr);
954 if (err)
955 return err;
956
957 err = fscrypt_prepare_setattr(dentry, attr);
958 if (err)
959 return err;
960
961 err = fsverity_prepare_setattr(dentry, attr);
962 if (err)
963 return err;
964
965 if (is_quota_modification(mnt_userns, inode, attr)) {
966 err = f2fs_dquot_initialize(inode);
967 if (err)
968 return err;
969 }
970 if (i_uid_needs_update(mnt_userns, attr, inode) ||
971 i_gid_needs_update(mnt_userns, attr, inode)) {
972 f2fs_lock_op(F2FS_I_SB(inode));
973 err = dquot_transfer(mnt_userns, inode, attr);
974 if (err) {
975 set_sbi_flag(F2FS_I_SB(inode),
976 SBI_QUOTA_NEED_REPAIR);
977 f2fs_unlock_op(F2FS_I_SB(inode));
978 return err;
979 }
980 /*
981 * update uid/gid under lock_op(), so that dquot and inode can
982 * be updated atomically.
983 */
984 i_uid_update(mnt_userns, attr, inode);
985 i_gid_update(mnt_userns, attr, inode);
986 f2fs_mark_inode_dirty_sync(inode, true);
987 f2fs_unlock_op(F2FS_I_SB(inode));
988 }
989
990 if (attr->ia_valid & ATTR_SIZE) {
991 loff_t old_size = i_size_read(inode);
992
993 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
994 /*
995 * should convert inline inode before i_size_write to
996 * keep smaller than inline_data size with inline flag.
997 */
998 err = f2fs_convert_inline_inode(inode);
999 if (err)
1000 return err;
1001 }
1002
1003 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1004 filemap_invalidate_lock(inode->i_mapping);
1005
1006 truncate_setsize(inode, attr->ia_size);
1007
1008 if (attr->ia_size <= old_size)
1009 err = f2fs_truncate(inode);
1010 /*
1011 * do not trim all blocks after i_size if target size is
1012 * larger than i_size.
1013 */
1014 filemap_invalidate_unlock(inode->i_mapping);
1015 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1016 if (err)
1017 return err;
1018
1019 spin_lock(&F2FS_I(inode)->i_size_lock);
1020 inode->i_mtime = inode->i_ctime = current_time(inode);
1021 F2FS_I(inode)->last_disk_size = i_size_read(inode);
1022 spin_unlock(&F2FS_I(inode)->i_size_lock);
1023 }
1024
1025 __setattr_copy(mnt_userns, inode, attr);
1026
1027 if (attr->ia_valid & ATTR_MODE) {
1028 err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
1029
1030 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1031 if (!err)
1032 inode->i_mode = F2FS_I(inode)->i_acl_mode;
1033 clear_inode_flag(inode, FI_ACL_MODE);
1034 }
1035 }
1036
1037 /* file size may changed here */
1038 f2fs_mark_inode_dirty_sync(inode, true);
1039
1040 /* inode change will produce dirty node pages flushed by checkpoint */
1041 f2fs_balance_fs(F2FS_I_SB(inode), true);
1042
1043 return err;
1044 }
1045
1046 const struct inode_operations f2fs_file_inode_operations = {
1047 .getattr = f2fs_getattr,
1048 .setattr = f2fs_setattr,
1049 .get_acl = f2fs_get_acl,
1050 .set_acl = f2fs_set_acl,
1051 .listxattr = f2fs_listxattr,
1052 .fiemap = f2fs_fiemap,
1053 .fileattr_get = f2fs_fileattr_get,
1054 .fileattr_set = f2fs_fileattr_set,
1055 };
1056
fill_zero(struct inode * inode,pgoff_t index,loff_t start,loff_t len)1057 static int fill_zero(struct inode *inode, pgoff_t index,
1058 loff_t start, loff_t len)
1059 {
1060 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1061 struct page *page;
1062
1063 if (!len)
1064 return 0;
1065
1066 f2fs_balance_fs(sbi, true);
1067
1068 f2fs_lock_op(sbi);
1069 page = f2fs_get_new_data_page(inode, NULL, index, false);
1070 f2fs_unlock_op(sbi);
1071
1072 if (IS_ERR(page))
1073 return PTR_ERR(page);
1074
1075 f2fs_wait_on_page_writeback(page, DATA, true, true);
1076 zero_user(page, start, len);
1077 set_page_dirty(page);
1078 f2fs_put_page(page, 1);
1079 return 0;
1080 }
1081
f2fs_truncate_hole(struct inode * inode,pgoff_t pg_start,pgoff_t pg_end)1082 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1083 {
1084 int err;
1085
1086 while (pg_start < pg_end) {
1087 struct dnode_of_data dn;
1088 pgoff_t end_offset, count;
1089
1090 set_new_dnode(&dn, inode, NULL, NULL, 0);
1091 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1092 if (err) {
1093 if (err == -ENOENT) {
1094 pg_start = f2fs_get_next_page_offset(&dn,
1095 pg_start);
1096 continue;
1097 }
1098 return err;
1099 }
1100
1101 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1102 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1103
1104 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1105
1106 f2fs_truncate_data_blocks_range(&dn, count);
1107 f2fs_put_dnode(&dn);
1108
1109 pg_start += count;
1110 }
1111 return 0;
1112 }
1113
punch_hole(struct inode * inode,loff_t offset,loff_t len)1114 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1115 {
1116 pgoff_t pg_start, pg_end;
1117 loff_t off_start, off_end;
1118 int ret;
1119
1120 ret = f2fs_convert_inline_inode(inode);
1121 if (ret)
1122 return ret;
1123
1124 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1125 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1126
1127 off_start = offset & (PAGE_SIZE - 1);
1128 off_end = (offset + len) & (PAGE_SIZE - 1);
1129
1130 if (pg_start == pg_end) {
1131 ret = fill_zero(inode, pg_start, off_start,
1132 off_end - off_start);
1133 if (ret)
1134 return ret;
1135 } else {
1136 if (off_start) {
1137 ret = fill_zero(inode, pg_start++, off_start,
1138 PAGE_SIZE - off_start);
1139 if (ret)
1140 return ret;
1141 }
1142 if (off_end) {
1143 ret = fill_zero(inode, pg_end, 0, off_end);
1144 if (ret)
1145 return ret;
1146 }
1147
1148 if (pg_start < pg_end) {
1149 loff_t blk_start, blk_end;
1150 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1151
1152 f2fs_balance_fs(sbi, true);
1153
1154 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1155 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1156
1157 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1158 filemap_invalidate_lock(inode->i_mapping);
1159
1160 truncate_pagecache_range(inode, blk_start, blk_end - 1);
1161
1162 f2fs_lock_op(sbi);
1163 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1164 f2fs_unlock_op(sbi);
1165
1166 filemap_invalidate_unlock(inode->i_mapping);
1167 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1168 }
1169 }
1170
1171 return ret;
1172 }
1173
__read_out_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,pgoff_t len)1174 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1175 int *do_replace, pgoff_t off, pgoff_t len)
1176 {
1177 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1178 struct dnode_of_data dn;
1179 int ret, done, i;
1180
1181 next_dnode:
1182 set_new_dnode(&dn, inode, NULL, NULL, 0);
1183 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1184 if (ret && ret != -ENOENT) {
1185 return ret;
1186 } else if (ret == -ENOENT) {
1187 if (dn.max_level == 0)
1188 return -ENOENT;
1189 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1190 dn.ofs_in_node, len);
1191 blkaddr += done;
1192 do_replace += done;
1193 goto next;
1194 }
1195
1196 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1197 dn.ofs_in_node, len);
1198 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1199 *blkaddr = f2fs_data_blkaddr(&dn);
1200
1201 if (__is_valid_data_blkaddr(*blkaddr) &&
1202 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1203 DATA_GENERIC_ENHANCE)) {
1204 f2fs_put_dnode(&dn);
1205 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1206 return -EFSCORRUPTED;
1207 }
1208
1209 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1210
1211 if (f2fs_lfs_mode(sbi)) {
1212 f2fs_put_dnode(&dn);
1213 return -EOPNOTSUPP;
1214 }
1215
1216 /* do not invalidate this block address */
1217 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1218 *do_replace = 1;
1219 }
1220 }
1221 f2fs_put_dnode(&dn);
1222 next:
1223 len -= done;
1224 off += done;
1225 if (len)
1226 goto next_dnode;
1227 return 0;
1228 }
1229
__roll_back_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,int len)1230 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1231 int *do_replace, pgoff_t off, int len)
1232 {
1233 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1234 struct dnode_of_data dn;
1235 int ret, i;
1236
1237 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1238 if (*do_replace == 0)
1239 continue;
1240
1241 set_new_dnode(&dn, inode, NULL, NULL, 0);
1242 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1243 if (ret) {
1244 dec_valid_block_count(sbi, inode, 1);
1245 f2fs_invalidate_blocks(sbi, *blkaddr);
1246 } else {
1247 f2fs_update_data_blkaddr(&dn, *blkaddr);
1248 }
1249 f2fs_put_dnode(&dn);
1250 }
1251 return 0;
1252 }
1253
__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)1254 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1255 block_t *blkaddr, int *do_replace,
1256 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1257 {
1258 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1259 pgoff_t i = 0;
1260 int ret;
1261
1262 while (i < len) {
1263 if (blkaddr[i] == NULL_ADDR && !full) {
1264 i++;
1265 continue;
1266 }
1267
1268 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1269 struct dnode_of_data dn;
1270 struct node_info ni;
1271 size_t new_size;
1272 pgoff_t ilen;
1273
1274 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1275 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1276 if (ret)
1277 return ret;
1278
1279 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1280 if (ret) {
1281 f2fs_put_dnode(&dn);
1282 return ret;
1283 }
1284
1285 ilen = min((pgoff_t)
1286 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1287 dn.ofs_in_node, len - i);
1288 do {
1289 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1290 f2fs_truncate_data_blocks_range(&dn, 1);
1291
1292 if (do_replace[i]) {
1293 f2fs_i_blocks_write(src_inode,
1294 1, false, false);
1295 f2fs_i_blocks_write(dst_inode,
1296 1, true, false);
1297 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1298 blkaddr[i], ni.version, true, false);
1299
1300 do_replace[i] = 0;
1301 }
1302 dn.ofs_in_node++;
1303 i++;
1304 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1305 if (dst_inode->i_size < new_size)
1306 f2fs_i_size_write(dst_inode, new_size);
1307 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1308
1309 f2fs_put_dnode(&dn);
1310 } else {
1311 struct page *psrc, *pdst;
1312
1313 psrc = f2fs_get_lock_data_page(src_inode,
1314 src + i, true);
1315 if (IS_ERR(psrc))
1316 return PTR_ERR(psrc);
1317 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1318 true);
1319 if (IS_ERR(pdst)) {
1320 f2fs_put_page(psrc, 1);
1321 return PTR_ERR(pdst);
1322 }
1323 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1324 set_page_dirty(pdst);
1325 f2fs_put_page(pdst, 1);
1326 f2fs_put_page(psrc, 1);
1327
1328 ret = f2fs_truncate_hole(src_inode,
1329 src + i, src + i + 1);
1330 if (ret)
1331 return ret;
1332 i++;
1333 }
1334 }
1335 return 0;
1336 }
1337
__exchange_data_block(struct inode * src_inode,struct inode * dst_inode,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1338 static int __exchange_data_block(struct inode *src_inode,
1339 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1340 pgoff_t len, bool full)
1341 {
1342 block_t *src_blkaddr;
1343 int *do_replace;
1344 pgoff_t olen;
1345 int ret;
1346
1347 while (len) {
1348 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1349
1350 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1351 array_size(olen, sizeof(block_t)),
1352 GFP_NOFS);
1353 if (!src_blkaddr)
1354 return -ENOMEM;
1355
1356 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1357 array_size(olen, sizeof(int)),
1358 GFP_NOFS);
1359 if (!do_replace) {
1360 kvfree(src_blkaddr);
1361 return -ENOMEM;
1362 }
1363
1364 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1365 do_replace, src, olen);
1366 if (ret)
1367 goto roll_back;
1368
1369 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1370 do_replace, src, dst, olen, full);
1371 if (ret)
1372 goto roll_back;
1373
1374 src += olen;
1375 dst += olen;
1376 len -= olen;
1377
1378 kvfree(src_blkaddr);
1379 kvfree(do_replace);
1380 }
1381 return 0;
1382
1383 roll_back:
1384 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1385 kvfree(src_blkaddr);
1386 kvfree(do_replace);
1387 return ret;
1388 }
1389
f2fs_do_collapse(struct inode * inode,loff_t offset,loff_t len)1390 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1391 {
1392 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1393 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1394 pgoff_t start = offset >> PAGE_SHIFT;
1395 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1396 int ret;
1397
1398 f2fs_balance_fs(sbi, true);
1399
1400 /* avoid gc operation during block exchange */
1401 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1402 filemap_invalidate_lock(inode->i_mapping);
1403
1404 f2fs_lock_op(sbi);
1405 f2fs_drop_extent_tree(inode);
1406 truncate_pagecache(inode, offset);
1407 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1408 f2fs_unlock_op(sbi);
1409
1410 filemap_invalidate_unlock(inode->i_mapping);
1411 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1412 return ret;
1413 }
1414
f2fs_collapse_range(struct inode * inode,loff_t offset,loff_t len)1415 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1416 {
1417 loff_t new_size;
1418 int ret;
1419
1420 if (offset + len >= i_size_read(inode))
1421 return -EINVAL;
1422
1423 /* collapse range should be aligned to block size of f2fs. */
1424 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1425 return -EINVAL;
1426
1427 ret = f2fs_convert_inline_inode(inode);
1428 if (ret)
1429 return ret;
1430
1431 /* write out all dirty pages from offset */
1432 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1433 if (ret)
1434 return ret;
1435
1436 ret = f2fs_do_collapse(inode, offset, len);
1437 if (ret)
1438 return ret;
1439
1440 /* write out all moved pages, if possible */
1441 filemap_invalidate_lock(inode->i_mapping);
1442 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1443 truncate_pagecache(inode, offset);
1444
1445 new_size = i_size_read(inode) - len;
1446 ret = f2fs_truncate_blocks(inode, new_size, true);
1447 filemap_invalidate_unlock(inode->i_mapping);
1448 if (!ret)
1449 f2fs_i_size_write(inode, new_size);
1450 return ret;
1451 }
1452
f2fs_do_zero_range(struct dnode_of_data * dn,pgoff_t start,pgoff_t end)1453 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1454 pgoff_t end)
1455 {
1456 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1457 pgoff_t index = start;
1458 unsigned int ofs_in_node = dn->ofs_in_node;
1459 blkcnt_t count = 0;
1460 int ret;
1461
1462 for (; index < end; index++, dn->ofs_in_node++) {
1463 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1464 count++;
1465 }
1466
1467 dn->ofs_in_node = ofs_in_node;
1468 ret = f2fs_reserve_new_blocks(dn, count);
1469 if (ret)
1470 return ret;
1471
1472 dn->ofs_in_node = ofs_in_node;
1473 for (index = start; index < end; index++, dn->ofs_in_node++) {
1474 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1475 /*
1476 * f2fs_reserve_new_blocks will not guarantee entire block
1477 * allocation.
1478 */
1479 if (dn->data_blkaddr == NULL_ADDR) {
1480 ret = -ENOSPC;
1481 break;
1482 }
1483
1484 if (dn->data_blkaddr == NEW_ADDR)
1485 continue;
1486
1487 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1488 DATA_GENERIC_ENHANCE)) {
1489 ret = -EFSCORRUPTED;
1490 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1491 break;
1492 }
1493
1494 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1495 dn->data_blkaddr = NEW_ADDR;
1496 f2fs_set_data_blkaddr(dn);
1497 }
1498
1499 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1500
1501 return ret;
1502 }
1503
f2fs_zero_range(struct inode * inode,loff_t offset,loff_t len,int mode)1504 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1505 int mode)
1506 {
1507 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1508 struct address_space *mapping = inode->i_mapping;
1509 pgoff_t index, pg_start, pg_end;
1510 loff_t new_size = i_size_read(inode);
1511 loff_t off_start, off_end;
1512 int ret = 0;
1513
1514 ret = inode_newsize_ok(inode, (len + offset));
1515 if (ret)
1516 return ret;
1517
1518 ret = f2fs_convert_inline_inode(inode);
1519 if (ret)
1520 return ret;
1521
1522 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1523 if (ret)
1524 return ret;
1525
1526 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1527 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1528
1529 off_start = offset & (PAGE_SIZE - 1);
1530 off_end = (offset + len) & (PAGE_SIZE - 1);
1531
1532 if (pg_start == pg_end) {
1533 ret = fill_zero(inode, pg_start, off_start,
1534 off_end - off_start);
1535 if (ret)
1536 return ret;
1537
1538 new_size = max_t(loff_t, new_size, offset + len);
1539 } else {
1540 if (off_start) {
1541 ret = fill_zero(inode, pg_start++, off_start,
1542 PAGE_SIZE - off_start);
1543 if (ret)
1544 return ret;
1545
1546 new_size = max_t(loff_t, new_size,
1547 (loff_t)pg_start << PAGE_SHIFT);
1548 }
1549
1550 for (index = pg_start; index < pg_end;) {
1551 struct dnode_of_data dn;
1552 unsigned int end_offset;
1553 pgoff_t end;
1554
1555 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1556 filemap_invalidate_lock(mapping);
1557
1558 truncate_pagecache_range(inode,
1559 (loff_t)index << PAGE_SHIFT,
1560 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1561
1562 f2fs_lock_op(sbi);
1563
1564 set_new_dnode(&dn, inode, NULL, NULL, 0);
1565 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1566 if (ret) {
1567 f2fs_unlock_op(sbi);
1568 filemap_invalidate_unlock(mapping);
1569 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1570 goto out;
1571 }
1572
1573 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1574 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1575
1576 ret = f2fs_do_zero_range(&dn, index, end);
1577 f2fs_put_dnode(&dn);
1578
1579 f2fs_unlock_op(sbi);
1580 filemap_invalidate_unlock(mapping);
1581 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1582
1583 f2fs_balance_fs(sbi, dn.node_changed);
1584
1585 if (ret)
1586 goto out;
1587
1588 index = end;
1589 new_size = max_t(loff_t, new_size,
1590 (loff_t)index << PAGE_SHIFT);
1591 }
1592
1593 if (off_end) {
1594 ret = fill_zero(inode, pg_end, 0, off_end);
1595 if (ret)
1596 goto out;
1597
1598 new_size = max_t(loff_t, new_size, offset + len);
1599 }
1600 }
1601
1602 out:
1603 if (new_size > i_size_read(inode)) {
1604 if (mode & FALLOC_FL_KEEP_SIZE)
1605 file_set_keep_isize(inode);
1606 else
1607 f2fs_i_size_write(inode, new_size);
1608 }
1609 return ret;
1610 }
1611
f2fs_insert_range(struct inode * inode,loff_t offset,loff_t len)1612 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1613 {
1614 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1615 struct address_space *mapping = inode->i_mapping;
1616 pgoff_t nr, pg_start, pg_end, delta, idx;
1617 loff_t new_size;
1618 int ret = 0;
1619
1620 new_size = i_size_read(inode) + len;
1621 ret = inode_newsize_ok(inode, new_size);
1622 if (ret)
1623 return ret;
1624
1625 if (offset >= i_size_read(inode))
1626 return -EINVAL;
1627
1628 /* insert range should be aligned to block size of f2fs. */
1629 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1630 return -EINVAL;
1631
1632 ret = f2fs_convert_inline_inode(inode);
1633 if (ret)
1634 return ret;
1635
1636 f2fs_balance_fs(sbi, true);
1637
1638 filemap_invalidate_lock(mapping);
1639 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1640 filemap_invalidate_unlock(mapping);
1641 if (ret)
1642 return ret;
1643
1644 /* write out all dirty pages from offset */
1645 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1646 if (ret)
1647 return ret;
1648
1649 pg_start = offset >> PAGE_SHIFT;
1650 pg_end = (offset + len) >> PAGE_SHIFT;
1651 delta = pg_end - pg_start;
1652 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1653
1654 /* avoid gc operation during block exchange */
1655 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1656 filemap_invalidate_lock(mapping);
1657 truncate_pagecache(inode, offset);
1658
1659 while (!ret && idx > pg_start) {
1660 nr = idx - pg_start;
1661 if (nr > delta)
1662 nr = delta;
1663 idx -= nr;
1664
1665 f2fs_lock_op(sbi);
1666 f2fs_drop_extent_tree(inode);
1667
1668 ret = __exchange_data_block(inode, inode, idx,
1669 idx + delta, nr, false);
1670 f2fs_unlock_op(sbi);
1671 }
1672 filemap_invalidate_unlock(mapping);
1673 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1674
1675 /* write out all moved pages, if possible */
1676 filemap_invalidate_lock(mapping);
1677 filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1678 truncate_pagecache(inode, offset);
1679 filemap_invalidate_unlock(mapping);
1680
1681 if (!ret)
1682 f2fs_i_size_write(inode, new_size);
1683 return ret;
1684 }
1685
expand_inode_data(struct inode * inode,loff_t offset,loff_t len,int mode)1686 static int expand_inode_data(struct inode *inode, loff_t offset,
1687 loff_t len, int mode)
1688 {
1689 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1690 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1691 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1692 .m_may_create = true };
1693 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1694 .init_gc_type = FG_GC,
1695 .should_migrate_blocks = false,
1696 .err_gc_skipped = true,
1697 .nr_free_secs = 0 };
1698 pgoff_t pg_start, pg_end;
1699 loff_t new_size = i_size_read(inode);
1700 loff_t off_end;
1701 block_t expanded = 0;
1702 int err;
1703
1704 err = inode_newsize_ok(inode, (len + offset));
1705 if (err)
1706 return err;
1707
1708 err = f2fs_convert_inline_inode(inode);
1709 if (err)
1710 return err;
1711
1712 f2fs_balance_fs(sbi, true);
1713
1714 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1715 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1716 off_end = (offset + len) & (PAGE_SIZE - 1);
1717
1718 map.m_lblk = pg_start;
1719 map.m_len = pg_end - pg_start;
1720 if (off_end)
1721 map.m_len++;
1722
1723 if (!map.m_len)
1724 return 0;
1725
1726 if (f2fs_is_pinned_file(inode)) {
1727 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1728 block_t sec_len = roundup(map.m_len, sec_blks);
1729
1730 map.m_len = sec_blks;
1731 next_alloc:
1732 if (has_not_enough_free_secs(sbi, 0,
1733 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1734 f2fs_down_write(&sbi->gc_lock);
1735 err = f2fs_gc(sbi, &gc_control);
1736 if (err && err != -ENODATA)
1737 goto out_err;
1738 }
1739
1740 f2fs_down_write(&sbi->pin_sem);
1741
1742 f2fs_lock_op(sbi);
1743 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1744 f2fs_unlock_op(sbi);
1745
1746 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1747 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1748 file_dont_truncate(inode);
1749
1750 f2fs_up_write(&sbi->pin_sem);
1751
1752 expanded += map.m_len;
1753 sec_len -= map.m_len;
1754 map.m_lblk += map.m_len;
1755 if (!err && sec_len)
1756 goto next_alloc;
1757
1758 map.m_len = expanded;
1759 } else {
1760 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1761 expanded = map.m_len;
1762 }
1763 out_err:
1764 if (err) {
1765 pgoff_t last_off;
1766
1767 if (!expanded)
1768 return err;
1769
1770 last_off = pg_start + expanded - 1;
1771
1772 /* update new size to the failed position */
1773 new_size = (last_off == pg_end) ? offset + len :
1774 (loff_t)(last_off + 1) << PAGE_SHIFT;
1775 } else {
1776 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1777 }
1778
1779 if (new_size > i_size_read(inode)) {
1780 if (mode & FALLOC_FL_KEEP_SIZE)
1781 file_set_keep_isize(inode);
1782 else
1783 f2fs_i_size_write(inode, new_size);
1784 }
1785
1786 return err;
1787 }
1788
f2fs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)1789 static long f2fs_fallocate(struct file *file, int mode,
1790 loff_t offset, loff_t len)
1791 {
1792 struct inode *inode = file_inode(file);
1793 long ret = 0;
1794
1795 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1796 return -EIO;
1797 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1798 return -ENOSPC;
1799 if (!f2fs_is_compress_backend_ready(inode))
1800 return -EOPNOTSUPP;
1801
1802 /* f2fs only support ->fallocate for regular file */
1803 if (!S_ISREG(inode->i_mode))
1804 return -EINVAL;
1805
1806 if (IS_ENCRYPTED(inode) &&
1807 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1808 return -EOPNOTSUPP;
1809
1810 /*
1811 * Pinned file should not support partial trucation since the block
1812 * can be used by applications.
1813 */
1814 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1815 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1816 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1817 return -EOPNOTSUPP;
1818
1819 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1820 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1821 FALLOC_FL_INSERT_RANGE))
1822 return -EOPNOTSUPP;
1823
1824 inode_lock(inode);
1825
1826 ret = file_modified(file);
1827 if (ret)
1828 goto out;
1829
1830 if (mode & FALLOC_FL_PUNCH_HOLE) {
1831 if (offset >= inode->i_size)
1832 goto out;
1833
1834 ret = punch_hole(inode, offset, len);
1835 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1836 ret = f2fs_collapse_range(inode, offset, len);
1837 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1838 ret = f2fs_zero_range(inode, offset, len, mode);
1839 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1840 ret = f2fs_insert_range(inode, offset, len);
1841 } else {
1842 ret = expand_inode_data(inode, offset, len, mode);
1843 }
1844
1845 if (!ret) {
1846 inode->i_mtime = inode->i_ctime = current_time(inode);
1847 f2fs_mark_inode_dirty_sync(inode, false);
1848 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1849 }
1850
1851 out:
1852 inode_unlock(inode);
1853
1854 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1855 return ret;
1856 }
1857
f2fs_release_file(struct inode * inode,struct file * filp)1858 static int f2fs_release_file(struct inode *inode, struct file *filp)
1859 {
1860 /*
1861 * f2fs_relase_file is called at every close calls. So we should
1862 * not drop any inmemory pages by close called by other process.
1863 */
1864 if (!(filp->f_mode & FMODE_WRITE) ||
1865 atomic_read(&inode->i_writecount) != 1)
1866 return 0;
1867
1868 f2fs_abort_atomic_write(inode, true);
1869 return 0;
1870 }
1871
f2fs_file_flush(struct file * file,fl_owner_t id)1872 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1873 {
1874 struct inode *inode = file_inode(file);
1875
1876 /*
1877 * If the process doing a transaction is crashed, we should do
1878 * roll-back. Otherwise, other reader/write can see corrupted database
1879 * until all the writers close its file. Since this should be done
1880 * before dropping file lock, it needs to do in ->flush.
1881 */
1882 if (F2FS_I(inode)->atomic_write_task == current)
1883 f2fs_abort_atomic_write(inode, true);
1884 return 0;
1885 }
1886
f2fs_setflags_common(struct inode * inode,u32 iflags,u32 mask)1887 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1888 {
1889 struct f2fs_inode_info *fi = F2FS_I(inode);
1890 u32 masked_flags = fi->i_flags & mask;
1891
1892 /* mask can be shrunk by flags_valid selector */
1893 iflags &= mask;
1894
1895 /* Is it quota file? Do not allow user to mess with it */
1896 if (IS_NOQUOTA(inode))
1897 return -EPERM;
1898
1899 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1900 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1901 return -EOPNOTSUPP;
1902 if (!f2fs_empty_dir(inode))
1903 return -ENOTEMPTY;
1904 }
1905
1906 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1907 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1908 return -EOPNOTSUPP;
1909 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1910 return -EINVAL;
1911 }
1912
1913 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1914 if (masked_flags & F2FS_COMPR_FL) {
1915 if (!f2fs_disable_compressed_file(inode))
1916 return -EINVAL;
1917 } else {
1918 /* try to convert inline_data to support compression */
1919 int err = f2fs_convert_inline_inode(inode);
1920 if (err)
1921 return err;
1922 if (!f2fs_may_compress(inode))
1923 return -EINVAL;
1924 if (S_ISREG(inode->i_mode) && F2FS_HAS_BLOCKS(inode))
1925 return -EINVAL;
1926 if (set_compress_context(inode))
1927 return -EOPNOTSUPP;
1928 }
1929 }
1930
1931 fi->i_flags = iflags | (fi->i_flags & ~mask);
1932 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1933 (fi->i_flags & F2FS_NOCOMP_FL));
1934
1935 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1936 set_inode_flag(inode, FI_PROJ_INHERIT);
1937 else
1938 clear_inode_flag(inode, FI_PROJ_INHERIT);
1939
1940 inode->i_ctime = current_time(inode);
1941 f2fs_set_inode_flags(inode);
1942 f2fs_mark_inode_dirty_sync(inode, true);
1943 return 0;
1944 }
1945
1946 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1947
1948 /*
1949 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1950 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1951 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1952 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1953 *
1954 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1955 * FS_IOC_FSSETXATTR is done by the VFS.
1956 */
1957
1958 static const struct {
1959 u32 iflag;
1960 u32 fsflag;
1961 } f2fs_fsflags_map[] = {
1962 { F2FS_COMPR_FL, FS_COMPR_FL },
1963 { F2FS_SYNC_FL, FS_SYNC_FL },
1964 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1965 { F2FS_APPEND_FL, FS_APPEND_FL },
1966 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1967 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1968 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
1969 { F2FS_INDEX_FL, FS_INDEX_FL },
1970 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1971 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1972 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1973 };
1974
1975 #define F2FS_GETTABLE_FS_FL ( \
1976 FS_COMPR_FL | \
1977 FS_SYNC_FL | \
1978 FS_IMMUTABLE_FL | \
1979 FS_APPEND_FL | \
1980 FS_NODUMP_FL | \
1981 FS_NOATIME_FL | \
1982 FS_NOCOMP_FL | \
1983 FS_INDEX_FL | \
1984 FS_DIRSYNC_FL | \
1985 FS_PROJINHERIT_FL | \
1986 FS_ENCRYPT_FL | \
1987 FS_INLINE_DATA_FL | \
1988 FS_NOCOW_FL | \
1989 FS_VERITY_FL | \
1990 FS_CASEFOLD_FL)
1991
1992 #define F2FS_SETTABLE_FS_FL ( \
1993 FS_COMPR_FL | \
1994 FS_SYNC_FL | \
1995 FS_IMMUTABLE_FL | \
1996 FS_APPEND_FL | \
1997 FS_NODUMP_FL | \
1998 FS_NOATIME_FL | \
1999 FS_NOCOMP_FL | \
2000 FS_DIRSYNC_FL | \
2001 FS_PROJINHERIT_FL | \
2002 FS_CASEFOLD_FL)
2003
2004 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
f2fs_iflags_to_fsflags(u32 iflags)2005 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2006 {
2007 u32 fsflags = 0;
2008 int i;
2009
2010 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2011 if (iflags & f2fs_fsflags_map[i].iflag)
2012 fsflags |= f2fs_fsflags_map[i].fsflag;
2013
2014 return fsflags;
2015 }
2016
2017 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
f2fs_fsflags_to_iflags(u32 fsflags)2018 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2019 {
2020 u32 iflags = 0;
2021 int i;
2022
2023 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2024 if (fsflags & f2fs_fsflags_map[i].fsflag)
2025 iflags |= f2fs_fsflags_map[i].iflag;
2026
2027 return iflags;
2028 }
2029
f2fs_ioc_getversion(struct file * filp,unsigned long arg)2030 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2031 {
2032 struct inode *inode = file_inode(filp);
2033
2034 return put_user(inode->i_generation, (int __user *)arg);
2035 }
2036
f2fs_ioc_start_atomic_write(struct file * filp)2037 static int f2fs_ioc_start_atomic_write(struct file *filp)
2038 {
2039 struct inode *inode = file_inode(filp);
2040 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2041 struct f2fs_inode_info *fi = F2FS_I(inode);
2042 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2043 struct inode *pinode;
2044 int ret;
2045
2046 if (!inode_owner_or_capable(mnt_userns, inode))
2047 return -EACCES;
2048
2049 if (!S_ISREG(inode->i_mode))
2050 return -EINVAL;
2051
2052 if (filp->f_flags & O_DIRECT)
2053 return -EINVAL;
2054
2055 ret = mnt_want_write_file(filp);
2056 if (ret)
2057 return ret;
2058
2059 inode_lock(inode);
2060
2061 if (!f2fs_disable_compressed_file(inode)) {
2062 ret = -EINVAL;
2063 goto out;
2064 }
2065
2066 if (f2fs_is_atomic_file(inode))
2067 goto out;
2068
2069 ret = f2fs_convert_inline_inode(inode);
2070 if (ret)
2071 goto out;
2072
2073 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2074
2075 /*
2076 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2077 * f2fs_is_atomic_file.
2078 */
2079 if (get_dirty_pages(inode))
2080 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2081 inode->i_ino, get_dirty_pages(inode));
2082 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2083 if (ret) {
2084 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2085 goto out;
2086 }
2087
2088 /* Create a COW inode for atomic write */
2089 pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2090 if (IS_ERR(pinode)) {
2091 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2092 ret = PTR_ERR(pinode);
2093 goto out;
2094 }
2095
2096 ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2097 iput(pinode);
2098 if (ret) {
2099 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2100 goto out;
2101 }
2102 f2fs_i_size_write(fi->cow_inode, i_size_read(inode));
2103
2104 stat_inc_atomic_inode(inode);
2105
2106 set_inode_flag(inode, FI_ATOMIC_FILE);
2107 set_inode_flag(fi->cow_inode, FI_COW_FILE);
2108 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2109 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2110
2111 f2fs_update_time(sbi, REQ_TIME);
2112 fi->atomic_write_task = current;
2113 stat_update_max_atomic_write(inode);
2114 fi->atomic_write_cnt = 0;
2115 out:
2116 inode_unlock(inode);
2117 mnt_drop_write_file(filp);
2118 return ret;
2119 }
2120
f2fs_ioc_commit_atomic_write(struct file * filp)2121 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2122 {
2123 struct inode *inode = file_inode(filp);
2124 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2125 int ret;
2126
2127 if (!inode_owner_or_capable(mnt_userns, inode))
2128 return -EACCES;
2129
2130 ret = mnt_want_write_file(filp);
2131 if (ret)
2132 return ret;
2133
2134 f2fs_balance_fs(F2FS_I_SB(inode), true);
2135
2136 inode_lock(inode);
2137
2138 if (f2fs_is_atomic_file(inode)) {
2139 ret = f2fs_commit_atomic_write(inode);
2140 if (ret)
2141 goto unlock_out;
2142
2143 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2144 if (!ret)
2145 f2fs_abort_atomic_write(inode, false);
2146 } else {
2147 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2148 }
2149 unlock_out:
2150 inode_unlock(inode);
2151 mnt_drop_write_file(filp);
2152 return ret;
2153 }
2154
f2fs_ioc_abort_atomic_write(struct file * filp)2155 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2156 {
2157 struct inode *inode = file_inode(filp);
2158 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2159 int ret;
2160
2161 if (!inode_owner_or_capable(mnt_userns, inode))
2162 return -EACCES;
2163
2164 ret = mnt_want_write_file(filp);
2165 if (ret)
2166 return ret;
2167
2168 inode_lock(inode);
2169
2170 f2fs_abort_atomic_write(inode, true);
2171
2172 inode_unlock(inode);
2173
2174 mnt_drop_write_file(filp);
2175 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2176 return ret;
2177 }
2178
f2fs_ioc_shutdown(struct file * filp,unsigned long arg)2179 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2180 {
2181 struct inode *inode = file_inode(filp);
2182 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2183 struct super_block *sb = sbi->sb;
2184 __u32 in;
2185 int ret = 0;
2186
2187 if (!capable(CAP_SYS_ADMIN))
2188 return -EPERM;
2189
2190 if (get_user(in, (__u32 __user *)arg))
2191 return -EFAULT;
2192
2193 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2194 ret = mnt_want_write_file(filp);
2195 if (ret) {
2196 if (ret == -EROFS) {
2197 ret = 0;
2198 f2fs_stop_checkpoint(sbi, false,
2199 STOP_CP_REASON_SHUTDOWN);
2200 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2201 trace_f2fs_shutdown(sbi, in, ret);
2202 }
2203 return ret;
2204 }
2205 }
2206
2207 switch (in) {
2208 case F2FS_GOING_DOWN_FULLSYNC:
2209 ret = freeze_bdev(sb->s_bdev);
2210 if (ret)
2211 goto out;
2212 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2213 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2214 thaw_bdev(sb->s_bdev);
2215 break;
2216 case F2FS_GOING_DOWN_METASYNC:
2217 /* do checkpoint only */
2218 ret = f2fs_sync_fs(sb, 1);
2219 if (ret)
2220 goto out;
2221 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2222 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2223 break;
2224 case F2FS_GOING_DOWN_NOSYNC:
2225 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2226 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2227 break;
2228 case F2FS_GOING_DOWN_METAFLUSH:
2229 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2230 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2231 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2232 break;
2233 case F2FS_GOING_DOWN_NEED_FSCK:
2234 set_sbi_flag(sbi, SBI_NEED_FSCK);
2235 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2236 set_sbi_flag(sbi, SBI_IS_DIRTY);
2237 /* do checkpoint only */
2238 ret = f2fs_sync_fs(sb, 1);
2239 goto out;
2240 default:
2241 ret = -EINVAL;
2242 goto out;
2243 }
2244
2245 f2fs_stop_gc_thread(sbi);
2246 f2fs_stop_discard_thread(sbi);
2247
2248 f2fs_drop_discard_cmd(sbi);
2249 clear_opt(sbi, DISCARD);
2250
2251 f2fs_update_time(sbi, REQ_TIME);
2252 out:
2253 if (in != F2FS_GOING_DOWN_FULLSYNC)
2254 mnt_drop_write_file(filp);
2255
2256 trace_f2fs_shutdown(sbi, in, ret);
2257
2258 return ret;
2259 }
2260
f2fs_ioc_fitrim(struct file * filp,unsigned long arg)2261 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2262 {
2263 struct inode *inode = file_inode(filp);
2264 struct super_block *sb = inode->i_sb;
2265 struct fstrim_range range;
2266 int ret;
2267
2268 if (!capable(CAP_SYS_ADMIN))
2269 return -EPERM;
2270
2271 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2272 return -EOPNOTSUPP;
2273
2274 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2275 sizeof(range)))
2276 return -EFAULT;
2277
2278 ret = mnt_want_write_file(filp);
2279 if (ret)
2280 return ret;
2281
2282 range.minlen = max((unsigned int)range.minlen,
2283 bdev_discard_granularity(sb->s_bdev));
2284 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2285 mnt_drop_write_file(filp);
2286 if (ret < 0)
2287 return ret;
2288
2289 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2290 sizeof(range)))
2291 return -EFAULT;
2292 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2293 return 0;
2294 }
2295
uuid_is_nonzero(__u8 u[16])2296 static bool uuid_is_nonzero(__u8 u[16])
2297 {
2298 int i;
2299
2300 for (i = 0; i < 16; i++)
2301 if (u[i])
2302 return true;
2303 return false;
2304 }
2305
f2fs_ioc_set_encryption_policy(struct file * filp,unsigned long arg)2306 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2307 {
2308 struct inode *inode = file_inode(filp);
2309
2310 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2311 return -EOPNOTSUPP;
2312
2313 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2314
2315 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2316 }
2317
f2fs_ioc_get_encryption_policy(struct file * filp,unsigned long arg)2318 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2319 {
2320 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2321 return -EOPNOTSUPP;
2322 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2323 }
2324
f2fs_ioc_get_encryption_pwsalt(struct file * filp,unsigned long arg)2325 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2326 {
2327 struct inode *inode = file_inode(filp);
2328 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2329 int err;
2330
2331 if (!f2fs_sb_has_encrypt(sbi))
2332 return -EOPNOTSUPP;
2333
2334 err = mnt_want_write_file(filp);
2335 if (err)
2336 return err;
2337
2338 f2fs_down_write(&sbi->sb_lock);
2339
2340 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2341 goto got_it;
2342
2343 /* update superblock with uuid */
2344 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2345
2346 err = f2fs_commit_super(sbi, false);
2347 if (err) {
2348 /* undo new data */
2349 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2350 goto out_err;
2351 }
2352 got_it:
2353 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2354 16))
2355 err = -EFAULT;
2356 out_err:
2357 f2fs_up_write(&sbi->sb_lock);
2358 mnt_drop_write_file(filp);
2359 return err;
2360 }
2361
f2fs_ioc_get_encryption_policy_ex(struct file * filp,unsigned long arg)2362 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2363 unsigned long arg)
2364 {
2365 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2366 return -EOPNOTSUPP;
2367
2368 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2369 }
2370
f2fs_ioc_add_encryption_key(struct file * filp,unsigned long arg)2371 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2372 {
2373 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2374 return -EOPNOTSUPP;
2375
2376 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2377 }
2378
f2fs_ioc_remove_encryption_key(struct file * filp,unsigned long arg)2379 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2380 {
2381 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2382 return -EOPNOTSUPP;
2383
2384 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2385 }
2386
f2fs_ioc_remove_encryption_key_all_users(struct file * filp,unsigned long arg)2387 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2388 unsigned long arg)
2389 {
2390 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2391 return -EOPNOTSUPP;
2392
2393 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2394 }
2395
f2fs_ioc_get_encryption_key_status(struct file * filp,unsigned long arg)2396 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2397 unsigned long arg)
2398 {
2399 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2400 return -EOPNOTSUPP;
2401
2402 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2403 }
2404
f2fs_ioc_get_encryption_nonce(struct file * filp,unsigned long arg)2405 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2406 {
2407 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2408 return -EOPNOTSUPP;
2409
2410 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2411 }
2412
f2fs_ioc_gc(struct file * filp,unsigned long arg)2413 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2414 {
2415 struct inode *inode = file_inode(filp);
2416 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2417 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2418 .no_bg_gc = false,
2419 .should_migrate_blocks = false,
2420 .nr_free_secs = 0 };
2421 __u32 sync;
2422 int ret;
2423
2424 if (!capable(CAP_SYS_ADMIN))
2425 return -EPERM;
2426
2427 if (get_user(sync, (__u32 __user *)arg))
2428 return -EFAULT;
2429
2430 if (f2fs_readonly(sbi->sb))
2431 return -EROFS;
2432
2433 ret = mnt_want_write_file(filp);
2434 if (ret)
2435 return ret;
2436
2437 if (!sync) {
2438 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2439 ret = -EBUSY;
2440 goto out;
2441 }
2442 } else {
2443 f2fs_down_write(&sbi->gc_lock);
2444 }
2445
2446 gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2447 gc_control.err_gc_skipped = sync;
2448 ret = f2fs_gc(sbi, &gc_control);
2449 out:
2450 mnt_drop_write_file(filp);
2451 return ret;
2452 }
2453
__f2fs_ioc_gc_range(struct file * filp,struct f2fs_gc_range * range)2454 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2455 {
2456 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2457 struct f2fs_gc_control gc_control = {
2458 .init_gc_type = range->sync ? FG_GC : BG_GC,
2459 .no_bg_gc = false,
2460 .should_migrate_blocks = false,
2461 .err_gc_skipped = range->sync,
2462 .nr_free_secs = 0 };
2463 u64 end;
2464 int ret;
2465
2466 if (!capable(CAP_SYS_ADMIN))
2467 return -EPERM;
2468 if (f2fs_readonly(sbi->sb))
2469 return -EROFS;
2470
2471 end = range->start + range->len;
2472 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2473 end >= MAX_BLKADDR(sbi))
2474 return -EINVAL;
2475
2476 ret = mnt_want_write_file(filp);
2477 if (ret)
2478 return ret;
2479
2480 do_more:
2481 if (!range->sync) {
2482 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2483 ret = -EBUSY;
2484 goto out;
2485 }
2486 } else {
2487 f2fs_down_write(&sbi->gc_lock);
2488 }
2489
2490 gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2491 ret = f2fs_gc(sbi, &gc_control);
2492 if (ret) {
2493 if (ret == -EBUSY)
2494 ret = -EAGAIN;
2495 goto out;
2496 }
2497 range->start += CAP_BLKS_PER_SEC(sbi);
2498 if (range->start <= end)
2499 goto do_more;
2500 out:
2501 mnt_drop_write_file(filp);
2502 return ret;
2503 }
2504
f2fs_ioc_gc_range(struct file * filp,unsigned long arg)2505 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2506 {
2507 struct f2fs_gc_range range;
2508
2509 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2510 sizeof(range)))
2511 return -EFAULT;
2512 return __f2fs_ioc_gc_range(filp, &range);
2513 }
2514
f2fs_ioc_write_checkpoint(struct file * filp,unsigned long arg)2515 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2516 {
2517 struct inode *inode = file_inode(filp);
2518 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2519 int ret;
2520
2521 if (!capable(CAP_SYS_ADMIN))
2522 return -EPERM;
2523
2524 if (f2fs_readonly(sbi->sb))
2525 return -EROFS;
2526
2527 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2528 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2529 return -EINVAL;
2530 }
2531
2532 ret = mnt_want_write_file(filp);
2533 if (ret)
2534 return ret;
2535
2536 ret = f2fs_sync_fs(sbi->sb, 1);
2537
2538 mnt_drop_write_file(filp);
2539 return ret;
2540 }
2541
f2fs_defragment_range(struct f2fs_sb_info * sbi,struct file * filp,struct f2fs_defragment * range)2542 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2543 struct file *filp,
2544 struct f2fs_defragment *range)
2545 {
2546 struct inode *inode = file_inode(filp);
2547 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2548 .m_seg_type = NO_CHECK_TYPE,
2549 .m_may_create = false };
2550 struct extent_info ei = {0, 0, 0};
2551 pgoff_t pg_start, pg_end, next_pgofs;
2552 unsigned int blk_per_seg = sbi->blocks_per_seg;
2553 unsigned int total = 0, sec_num;
2554 block_t blk_end = 0;
2555 bool fragmented = false;
2556 int err;
2557
2558 pg_start = range->start >> PAGE_SHIFT;
2559 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2560
2561 f2fs_balance_fs(sbi, true);
2562
2563 inode_lock(inode);
2564
2565 /* if in-place-update policy is enabled, don't waste time here */
2566 set_inode_flag(inode, FI_OPU_WRITE);
2567 if (f2fs_should_update_inplace(inode, NULL)) {
2568 err = -EINVAL;
2569 goto out;
2570 }
2571
2572 /* writeback all dirty pages in the range */
2573 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2574 range->start + range->len - 1);
2575 if (err)
2576 goto out;
2577
2578 /*
2579 * lookup mapping info in extent cache, skip defragmenting if physical
2580 * block addresses are continuous.
2581 */
2582 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2583 if (ei.fofs + ei.len >= pg_end)
2584 goto out;
2585 }
2586
2587 map.m_lblk = pg_start;
2588 map.m_next_pgofs = &next_pgofs;
2589
2590 /*
2591 * lookup mapping info in dnode page cache, skip defragmenting if all
2592 * physical block addresses are continuous even if there are hole(s)
2593 * in logical blocks.
2594 */
2595 while (map.m_lblk < pg_end) {
2596 map.m_len = pg_end - map.m_lblk;
2597 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2598 if (err)
2599 goto out;
2600
2601 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2602 map.m_lblk = next_pgofs;
2603 continue;
2604 }
2605
2606 if (blk_end && blk_end != map.m_pblk)
2607 fragmented = true;
2608
2609 /* record total count of block that we're going to move */
2610 total += map.m_len;
2611
2612 blk_end = map.m_pblk + map.m_len;
2613
2614 map.m_lblk += map.m_len;
2615 }
2616
2617 if (!fragmented) {
2618 total = 0;
2619 goto out;
2620 }
2621
2622 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2623
2624 /*
2625 * make sure there are enough free section for LFS allocation, this can
2626 * avoid defragment running in SSR mode when free section are allocated
2627 * intensively
2628 */
2629 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2630 err = -EAGAIN;
2631 goto out;
2632 }
2633
2634 map.m_lblk = pg_start;
2635 map.m_len = pg_end - pg_start;
2636 total = 0;
2637
2638 while (map.m_lblk < pg_end) {
2639 pgoff_t idx;
2640 int cnt = 0;
2641
2642 do_map:
2643 map.m_len = pg_end - map.m_lblk;
2644 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2645 if (err)
2646 goto clear_out;
2647
2648 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2649 map.m_lblk = next_pgofs;
2650 goto check;
2651 }
2652
2653 set_inode_flag(inode, FI_SKIP_WRITES);
2654
2655 idx = map.m_lblk;
2656 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2657 struct page *page;
2658
2659 page = f2fs_get_lock_data_page(inode, idx, true);
2660 if (IS_ERR(page)) {
2661 err = PTR_ERR(page);
2662 goto clear_out;
2663 }
2664
2665 set_page_dirty(page);
2666 set_page_private_gcing(page);
2667 f2fs_put_page(page, 1);
2668
2669 idx++;
2670 cnt++;
2671 total++;
2672 }
2673
2674 map.m_lblk = idx;
2675 check:
2676 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2677 goto do_map;
2678
2679 clear_inode_flag(inode, FI_SKIP_WRITES);
2680
2681 err = filemap_fdatawrite(inode->i_mapping);
2682 if (err)
2683 goto out;
2684 }
2685 clear_out:
2686 clear_inode_flag(inode, FI_SKIP_WRITES);
2687 out:
2688 clear_inode_flag(inode, FI_OPU_WRITE);
2689 inode_unlock(inode);
2690 if (!err)
2691 range->len = (u64)total << PAGE_SHIFT;
2692 return err;
2693 }
2694
f2fs_ioc_defragment(struct file * filp,unsigned long arg)2695 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2696 {
2697 struct inode *inode = file_inode(filp);
2698 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2699 struct f2fs_defragment range;
2700 int err;
2701
2702 if (!capable(CAP_SYS_ADMIN))
2703 return -EPERM;
2704
2705 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2706 return -EINVAL;
2707
2708 if (f2fs_readonly(sbi->sb))
2709 return -EROFS;
2710
2711 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2712 sizeof(range)))
2713 return -EFAULT;
2714
2715 /* verify alignment of offset & size */
2716 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2717 return -EINVAL;
2718
2719 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2720 max_file_blocks(inode)))
2721 return -EINVAL;
2722
2723 err = mnt_want_write_file(filp);
2724 if (err)
2725 return err;
2726
2727 err = f2fs_defragment_range(sbi, filp, &range);
2728 mnt_drop_write_file(filp);
2729
2730 f2fs_update_time(sbi, REQ_TIME);
2731 if (err < 0)
2732 return err;
2733
2734 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2735 sizeof(range)))
2736 return -EFAULT;
2737
2738 return 0;
2739 }
2740
f2fs_move_file_range(struct file * file_in,loff_t pos_in,struct file * file_out,loff_t pos_out,size_t len)2741 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2742 struct file *file_out, loff_t pos_out, size_t len)
2743 {
2744 struct inode *src = file_inode(file_in);
2745 struct inode *dst = file_inode(file_out);
2746 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2747 size_t olen = len, dst_max_i_size = 0;
2748 size_t dst_osize;
2749 int ret;
2750
2751 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2752 src->i_sb != dst->i_sb)
2753 return -EXDEV;
2754
2755 if (unlikely(f2fs_readonly(src->i_sb)))
2756 return -EROFS;
2757
2758 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2759 return -EINVAL;
2760
2761 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2762 return -EOPNOTSUPP;
2763
2764 if (pos_out < 0 || pos_in < 0)
2765 return -EINVAL;
2766
2767 if (src == dst) {
2768 if (pos_in == pos_out)
2769 return 0;
2770 if (pos_out > pos_in && pos_out < pos_in + len)
2771 return -EINVAL;
2772 }
2773
2774 inode_lock(src);
2775 if (src != dst) {
2776 ret = -EBUSY;
2777 if (!inode_trylock(dst))
2778 goto out;
2779 }
2780
2781 ret = -EINVAL;
2782 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2783 goto out_unlock;
2784 if (len == 0)
2785 olen = len = src->i_size - pos_in;
2786 if (pos_in + len == src->i_size)
2787 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2788 if (len == 0) {
2789 ret = 0;
2790 goto out_unlock;
2791 }
2792
2793 dst_osize = dst->i_size;
2794 if (pos_out + olen > dst->i_size)
2795 dst_max_i_size = pos_out + olen;
2796
2797 /* verify the end result is block aligned */
2798 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2799 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2800 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2801 goto out_unlock;
2802
2803 ret = f2fs_convert_inline_inode(src);
2804 if (ret)
2805 goto out_unlock;
2806
2807 ret = f2fs_convert_inline_inode(dst);
2808 if (ret)
2809 goto out_unlock;
2810
2811 /* write out all dirty pages from offset */
2812 ret = filemap_write_and_wait_range(src->i_mapping,
2813 pos_in, pos_in + len);
2814 if (ret)
2815 goto out_unlock;
2816
2817 ret = filemap_write_and_wait_range(dst->i_mapping,
2818 pos_out, pos_out + len);
2819 if (ret)
2820 goto out_unlock;
2821
2822 f2fs_balance_fs(sbi, true);
2823
2824 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2825 if (src != dst) {
2826 ret = -EBUSY;
2827 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2828 goto out_src;
2829 }
2830
2831 f2fs_lock_op(sbi);
2832 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2833 pos_out >> F2FS_BLKSIZE_BITS,
2834 len >> F2FS_BLKSIZE_BITS, false);
2835
2836 if (!ret) {
2837 if (dst_max_i_size)
2838 f2fs_i_size_write(dst, dst_max_i_size);
2839 else if (dst_osize != dst->i_size)
2840 f2fs_i_size_write(dst, dst_osize);
2841 }
2842 f2fs_unlock_op(sbi);
2843
2844 if (src != dst)
2845 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2846 out_src:
2847 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2848 out_unlock:
2849 if (src != dst)
2850 inode_unlock(dst);
2851 out:
2852 inode_unlock(src);
2853 return ret;
2854 }
2855
__f2fs_ioc_move_range(struct file * filp,struct f2fs_move_range * range)2856 static int __f2fs_ioc_move_range(struct file *filp,
2857 struct f2fs_move_range *range)
2858 {
2859 struct fd dst;
2860 int err;
2861
2862 if (!(filp->f_mode & FMODE_READ) ||
2863 !(filp->f_mode & FMODE_WRITE))
2864 return -EBADF;
2865
2866 dst = fdget(range->dst_fd);
2867 if (!dst.file)
2868 return -EBADF;
2869
2870 if (!(dst.file->f_mode & FMODE_WRITE)) {
2871 err = -EBADF;
2872 goto err_out;
2873 }
2874
2875 err = mnt_want_write_file(filp);
2876 if (err)
2877 goto err_out;
2878
2879 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2880 range->pos_out, range->len);
2881
2882 mnt_drop_write_file(filp);
2883 err_out:
2884 fdput(dst);
2885 return err;
2886 }
2887
f2fs_ioc_move_range(struct file * filp,unsigned long arg)2888 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2889 {
2890 struct f2fs_move_range range;
2891
2892 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2893 sizeof(range)))
2894 return -EFAULT;
2895 return __f2fs_ioc_move_range(filp, &range);
2896 }
2897
f2fs_ioc_flush_device(struct file * filp,unsigned long arg)2898 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2899 {
2900 struct inode *inode = file_inode(filp);
2901 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2902 struct sit_info *sm = SIT_I(sbi);
2903 unsigned int start_segno = 0, end_segno = 0;
2904 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2905 struct f2fs_flush_device range;
2906 struct f2fs_gc_control gc_control = {
2907 .init_gc_type = FG_GC,
2908 .should_migrate_blocks = true,
2909 .err_gc_skipped = true,
2910 .nr_free_secs = 0 };
2911 int ret;
2912
2913 if (!capable(CAP_SYS_ADMIN))
2914 return -EPERM;
2915
2916 if (f2fs_readonly(sbi->sb))
2917 return -EROFS;
2918
2919 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2920 return -EINVAL;
2921
2922 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2923 sizeof(range)))
2924 return -EFAULT;
2925
2926 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2927 __is_large_section(sbi)) {
2928 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2929 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2930 return -EINVAL;
2931 }
2932
2933 ret = mnt_want_write_file(filp);
2934 if (ret)
2935 return ret;
2936
2937 if (range.dev_num != 0)
2938 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2939 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2940
2941 start_segno = sm->last_victim[FLUSH_DEVICE];
2942 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2943 start_segno = dev_start_segno;
2944 end_segno = min(start_segno + range.segments, dev_end_segno);
2945
2946 while (start_segno < end_segno) {
2947 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2948 ret = -EBUSY;
2949 goto out;
2950 }
2951 sm->last_victim[GC_CB] = end_segno + 1;
2952 sm->last_victim[GC_GREEDY] = end_segno + 1;
2953 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2954
2955 gc_control.victim_segno = start_segno;
2956 ret = f2fs_gc(sbi, &gc_control);
2957 if (ret == -EAGAIN)
2958 ret = 0;
2959 else if (ret < 0)
2960 break;
2961 start_segno++;
2962 }
2963 out:
2964 mnt_drop_write_file(filp);
2965 return ret;
2966 }
2967
f2fs_ioc_get_features(struct file * filp,unsigned long arg)2968 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2969 {
2970 struct inode *inode = file_inode(filp);
2971 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2972
2973 /* Must validate to set it with SQLite behavior in Android. */
2974 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2975
2976 return put_user(sb_feature, (u32 __user *)arg);
2977 }
2978
2979 #ifdef CONFIG_QUOTA
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)2980 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2981 {
2982 struct dquot *transfer_to[MAXQUOTAS] = {};
2983 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2984 struct super_block *sb = sbi->sb;
2985 int err = 0;
2986
2987 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2988 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2989 err = __dquot_transfer(inode, transfer_to);
2990 if (err)
2991 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2992 dqput(transfer_to[PRJQUOTA]);
2993 }
2994 return err;
2995 }
2996
f2fs_ioc_setproject(struct inode * inode,__u32 projid)2997 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2998 {
2999 struct f2fs_inode_info *fi = F2FS_I(inode);
3000 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3001 struct f2fs_inode *ri = NULL;
3002 kprojid_t kprojid;
3003 int err;
3004
3005 if (!f2fs_sb_has_project_quota(sbi)) {
3006 if (projid != F2FS_DEF_PROJID)
3007 return -EOPNOTSUPP;
3008 else
3009 return 0;
3010 }
3011
3012 if (!f2fs_has_extra_attr(inode))
3013 return -EOPNOTSUPP;
3014
3015 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3016
3017 if (projid_eq(kprojid, fi->i_projid))
3018 return 0;
3019
3020 err = -EPERM;
3021 /* Is it quota file? Do not allow user to mess with it */
3022 if (IS_NOQUOTA(inode))
3023 return err;
3024
3025 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3026 return -EOVERFLOW;
3027
3028 err = f2fs_dquot_initialize(inode);
3029 if (err)
3030 return err;
3031
3032 f2fs_lock_op(sbi);
3033 err = f2fs_transfer_project_quota(inode, kprojid);
3034 if (err)
3035 goto out_unlock;
3036
3037 fi->i_projid = kprojid;
3038 inode->i_ctime = current_time(inode);
3039 f2fs_mark_inode_dirty_sync(inode, true);
3040 out_unlock:
3041 f2fs_unlock_op(sbi);
3042 return err;
3043 }
3044 #else
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3045 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3046 {
3047 return 0;
3048 }
3049
f2fs_ioc_setproject(struct inode * inode,__u32 projid)3050 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3051 {
3052 if (projid != F2FS_DEF_PROJID)
3053 return -EOPNOTSUPP;
3054 return 0;
3055 }
3056 #endif
3057
f2fs_fileattr_get(struct dentry * dentry,struct fileattr * fa)3058 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3059 {
3060 struct inode *inode = d_inode(dentry);
3061 struct f2fs_inode_info *fi = F2FS_I(inode);
3062 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3063
3064 if (IS_ENCRYPTED(inode))
3065 fsflags |= FS_ENCRYPT_FL;
3066 if (IS_VERITY(inode))
3067 fsflags |= FS_VERITY_FL;
3068 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3069 fsflags |= FS_INLINE_DATA_FL;
3070 if (is_inode_flag_set(inode, FI_PIN_FILE))
3071 fsflags |= FS_NOCOW_FL;
3072
3073 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3074
3075 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3076 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3077
3078 return 0;
3079 }
3080
f2fs_fileattr_set(struct user_namespace * mnt_userns,struct dentry * dentry,struct fileattr * fa)3081 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3082 struct dentry *dentry, struct fileattr *fa)
3083 {
3084 struct inode *inode = d_inode(dentry);
3085 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3086 u32 iflags;
3087 int err;
3088
3089 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3090 return -EIO;
3091 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3092 return -ENOSPC;
3093 if (fsflags & ~F2FS_GETTABLE_FS_FL)
3094 return -EOPNOTSUPP;
3095 fsflags &= F2FS_SETTABLE_FS_FL;
3096 if (!fa->flags_valid)
3097 mask &= FS_COMMON_FL;
3098
3099 iflags = f2fs_fsflags_to_iflags(fsflags);
3100 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3101 return -EOPNOTSUPP;
3102
3103 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3104 if (!err)
3105 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3106
3107 return err;
3108 }
3109
f2fs_pin_file_control(struct inode * inode,bool inc)3110 int f2fs_pin_file_control(struct inode *inode, bool inc)
3111 {
3112 struct f2fs_inode_info *fi = F2FS_I(inode);
3113 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3114
3115 /* Use i_gc_failures for normal file as a risk signal. */
3116 if (inc)
3117 f2fs_i_gc_failures_write(inode,
3118 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3119
3120 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3121 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3122 __func__, inode->i_ino,
3123 fi->i_gc_failures[GC_FAILURE_PIN]);
3124 clear_inode_flag(inode, FI_PIN_FILE);
3125 return -EAGAIN;
3126 }
3127 return 0;
3128 }
3129
f2fs_ioc_set_pin_file(struct file * filp,unsigned long arg)3130 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3131 {
3132 struct inode *inode = file_inode(filp);
3133 __u32 pin;
3134 int ret = 0;
3135
3136 if (get_user(pin, (__u32 __user *)arg))
3137 return -EFAULT;
3138
3139 if (!S_ISREG(inode->i_mode))
3140 return -EINVAL;
3141
3142 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3143 return -EROFS;
3144
3145 ret = mnt_want_write_file(filp);
3146 if (ret)
3147 return ret;
3148
3149 inode_lock(inode);
3150
3151 if (!pin) {
3152 clear_inode_flag(inode, FI_PIN_FILE);
3153 f2fs_i_gc_failures_write(inode, 0);
3154 goto done;
3155 }
3156
3157 if (f2fs_should_update_outplace(inode, NULL)) {
3158 ret = -EINVAL;
3159 goto out;
3160 }
3161
3162 if (f2fs_pin_file_control(inode, false)) {
3163 ret = -EAGAIN;
3164 goto out;
3165 }
3166
3167 ret = f2fs_convert_inline_inode(inode);
3168 if (ret)
3169 goto out;
3170
3171 if (!f2fs_disable_compressed_file(inode)) {
3172 ret = -EOPNOTSUPP;
3173 goto out;
3174 }
3175
3176 set_inode_flag(inode, FI_PIN_FILE);
3177 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3178 done:
3179 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3180 out:
3181 inode_unlock(inode);
3182 mnt_drop_write_file(filp);
3183 return ret;
3184 }
3185
f2fs_ioc_get_pin_file(struct file * filp,unsigned long arg)3186 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3187 {
3188 struct inode *inode = file_inode(filp);
3189 __u32 pin = 0;
3190
3191 if (is_inode_flag_set(inode, FI_PIN_FILE))
3192 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3193 return put_user(pin, (u32 __user *)arg);
3194 }
3195
f2fs_precache_extents(struct inode * inode)3196 int f2fs_precache_extents(struct inode *inode)
3197 {
3198 struct f2fs_inode_info *fi = F2FS_I(inode);
3199 struct f2fs_map_blocks map;
3200 pgoff_t m_next_extent;
3201 loff_t end;
3202 int err;
3203
3204 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3205 return -EOPNOTSUPP;
3206
3207 map.m_lblk = 0;
3208 map.m_next_pgofs = NULL;
3209 map.m_next_extent = &m_next_extent;
3210 map.m_seg_type = NO_CHECK_TYPE;
3211 map.m_may_create = false;
3212 end = max_file_blocks(inode);
3213
3214 while (map.m_lblk < end) {
3215 map.m_len = end - map.m_lblk;
3216
3217 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3218 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3219 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3220 if (err)
3221 return err;
3222
3223 map.m_lblk = m_next_extent;
3224 }
3225
3226 return 0;
3227 }
3228
f2fs_ioc_precache_extents(struct file * filp,unsigned long arg)3229 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3230 {
3231 return f2fs_precache_extents(file_inode(filp));
3232 }
3233
f2fs_ioc_resize_fs(struct file * filp,unsigned long arg)3234 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3235 {
3236 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3237 __u64 block_count;
3238
3239 if (!capable(CAP_SYS_ADMIN))
3240 return -EPERM;
3241
3242 if (f2fs_readonly(sbi->sb))
3243 return -EROFS;
3244
3245 if (copy_from_user(&block_count, (void __user *)arg,
3246 sizeof(block_count)))
3247 return -EFAULT;
3248
3249 return f2fs_resize_fs(sbi, block_count);
3250 }
3251
f2fs_ioc_enable_verity(struct file * filp,unsigned long arg)3252 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3253 {
3254 struct inode *inode = file_inode(filp);
3255
3256 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3257
3258 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3259 f2fs_warn(F2FS_I_SB(inode),
3260 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3261 inode->i_ino);
3262 return -EOPNOTSUPP;
3263 }
3264
3265 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3266 }
3267
f2fs_ioc_measure_verity(struct file * filp,unsigned long arg)3268 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3269 {
3270 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3271 return -EOPNOTSUPP;
3272
3273 return fsverity_ioctl_measure(filp, (void __user *)arg);
3274 }
3275
f2fs_ioc_read_verity_metadata(struct file * filp,unsigned long arg)3276 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3277 {
3278 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3279 return -EOPNOTSUPP;
3280
3281 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3282 }
3283
f2fs_ioc_getfslabel(struct file * filp,unsigned long arg)3284 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3285 {
3286 struct inode *inode = file_inode(filp);
3287 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3288 char *vbuf;
3289 int count;
3290 int err = 0;
3291
3292 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3293 if (!vbuf)
3294 return -ENOMEM;
3295
3296 f2fs_down_read(&sbi->sb_lock);
3297 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3298 ARRAY_SIZE(sbi->raw_super->volume_name),
3299 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3300 f2fs_up_read(&sbi->sb_lock);
3301
3302 if (copy_to_user((char __user *)arg, vbuf,
3303 min(FSLABEL_MAX, count)))
3304 err = -EFAULT;
3305
3306 kfree(vbuf);
3307 return err;
3308 }
3309
f2fs_ioc_setfslabel(struct file * filp,unsigned long arg)3310 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3311 {
3312 struct inode *inode = file_inode(filp);
3313 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3314 char *vbuf;
3315 int err = 0;
3316
3317 if (!capable(CAP_SYS_ADMIN))
3318 return -EPERM;
3319
3320 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3321 if (IS_ERR(vbuf))
3322 return PTR_ERR(vbuf);
3323
3324 err = mnt_want_write_file(filp);
3325 if (err)
3326 goto out;
3327
3328 f2fs_down_write(&sbi->sb_lock);
3329
3330 memset(sbi->raw_super->volume_name, 0,
3331 sizeof(sbi->raw_super->volume_name));
3332 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3333 sbi->raw_super->volume_name,
3334 ARRAY_SIZE(sbi->raw_super->volume_name));
3335
3336 err = f2fs_commit_super(sbi, false);
3337
3338 f2fs_up_write(&sbi->sb_lock);
3339
3340 mnt_drop_write_file(filp);
3341 out:
3342 kfree(vbuf);
3343 return err;
3344 }
3345
f2fs_get_compress_blocks(struct file * filp,unsigned long arg)3346 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3347 {
3348 struct inode *inode = file_inode(filp);
3349 __u64 blocks;
3350
3351 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3352 return -EOPNOTSUPP;
3353
3354 if (!f2fs_compressed_file(inode))
3355 return -EINVAL;
3356
3357 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3358 return put_user(blocks, (u64 __user *)arg);
3359 }
3360
release_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3361 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3362 {
3363 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3364 unsigned int released_blocks = 0;
3365 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3366 block_t blkaddr;
3367 int i;
3368
3369 for (i = 0; i < count; i++) {
3370 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3371 dn->ofs_in_node + i);
3372
3373 if (!__is_valid_data_blkaddr(blkaddr))
3374 continue;
3375 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3376 DATA_GENERIC_ENHANCE))) {
3377 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3378 return -EFSCORRUPTED;
3379 }
3380 }
3381
3382 while (count) {
3383 int compr_blocks = 0;
3384
3385 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3386 blkaddr = f2fs_data_blkaddr(dn);
3387
3388 if (i == 0) {
3389 if (blkaddr == COMPRESS_ADDR)
3390 continue;
3391 dn->ofs_in_node += cluster_size;
3392 goto next;
3393 }
3394
3395 if (__is_valid_data_blkaddr(blkaddr))
3396 compr_blocks++;
3397
3398 if (blkaddr != NEW_ADDR)
3399 continue;
3400
3401 dn->data_blkaddr = NULL_ADDR;
3402 f2fs_set_data_blkaddr(dn);
3403 }
3404
3405 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3406 dec_valid_block_count(sbi, dn->inode,
3407 cluster_size - compr_blocks);
3408
3409 released_blocks += cluster_size - compr_blocks;
3410 next:
3411 count -= cluster_size;
3412 }
3413
3414 return released_blocks;
3415 }
3416
f2fs_release_compress_blocks(struct file * filp,unsigned long arg)3417 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3418 {
3419 struct inode *inode = file_inode(filp);
3420 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3421 pgoff_t page_idx = 0, last_idx;
3422 unsigned int released_blocks = 0;
3423 int ret;
3424 int writecount;
3425
3426 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3427 return -EOPNOTSUPP;
3428
3429 if (!f2fs_compressed_file(inode))
3430 return -EINVAL;
3431
3432 if (f2fs_readonly(sbi->sb))
3433 return -EROFS;
3434
3435 ret = mnt_want_write_file(filp);
3436 if (ret)
3437 return ret;
3438
3439 f2fs_balance_fs(F2FS_I_SB(inode), true);
3440
3441 inode_lock(inode);
3442
3443 writecount = atomic_read(&inode->i_writecount);
3444 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3445 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3446 ret = -EBUSY;
3447 goto out;
3448 }
3449
3450 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3451 ret = -EINVAL;
3452 goto out;
3453 }
3454
3455 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3456 if (ret)
3457 goto out;
3458
3459 set_inode_flag(inode, FI_COMPRESS_RELEASED);
3460 inode->i_ctime = current_time(inode);
3461 f2fs_mark_inode_dirty_sync(inode, true);
3462
3463 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3464 goto out;
3465
3466 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3467 filemap_invalidate_lock(inode->i_mapping);
3468
3469 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3470
3471 while (page_idx < last_idx) {
3472 struct dnode_of_data dn;
3473 pgoff_t end_offset, count;
3474
3475 set_new_dnode(&dn, inode, NULL, NULL, 0);
3476 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3477 if (ret) {
3478 if (ret == -ENOENT) {
3479 page_idx = f2fs_get_next_page_offset(&dn,
3480 page_idx);
3481 ret = 0;
3482 continue;
3483 }
3484 break;
3485 }
3486
3487 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3488 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3489 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3490
3491 ret = release_compress_blocks(&dn, count);
3492
3493 f2fs_put_dnode(&dn);
3494
3495 if (ret < 0)
3496 break;
3497
3498 page_idx += count;
3499 released_blocks += ret;
3500 }
3501
3502 filemap_invalidate_unlock(inode->i_mapping);
3503 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3504 out:
3505 inode_unlock(inode);
3506
3507 mnt_drop_write_file(filp);
3508
3509 if (ret >= 0) {
3510 ret = put_user(released_blocks, (u64 __user *)arg);
3511 } else if (released_blocks &&
3512 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3513 set_sbi_flag(sbi, SBI_NEED_FSCK);
3514 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3515 "iblocks=%llu, released=%u, compr_blocks=%u, "
3516 "run fsck to fix.",
3517 __func__, inode->i_ino, inode->i_blocks,
3518 released_blocks,
3519 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3520 }
3521
3522 return ret;
3523 }
3524
reserve_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3525 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3526 {
3527 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3528 unsigned int reserved_blocks = 0;
3529 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3530 block_t blkaddr;
3531 int i;
3532
3533 for (i = 0; i < count; i++) {
3534 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3535 dn->ofs_in_node + i);
3536
3537 if (!__is_valid_data_blkaddr(blkaddr))
3538 continue;
3539 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3540 DATA_GENERIC_ENHANCE))) {
3541 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3542 return -EFSCORRUPTED;
3543 }
3544 }
3545
3546 while (count) {
3547 int compr_blocks = 0;
3548 blkcnt_t reserved;
3549 int ret;
3550
3551 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3552 blkaddr = f2fs_data_blkaddr(dn);
3553
3554 if (i == 0) {
3555 if (blkaddr == COMPRESS_ADDR)
3556 continue;
3557 dn->ofs_in_node += cluster_size;
3558 goto next;
3559 }
3560
3561 if (__is_valid_data_blkaddr(blkaddr)) {
3562 compr_blocks++;
3563 continue;
3564 }
3565
3566 dn->data_blkaddr = NEW_ADDR;
3567 f2fs_set_data_blkaddr(dn);
3568 }
3569
3570 reserved = cluster_size - compr_blocks;
3571 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3572 if (ret)
3573 return ret;
3574
3575 if (reserved != cluster_size - compr_blocks)
3576 return -ENOSPC;
3577
3578 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3579
3580 reserved_blocks += reserved;
3581 next:
3582 count -= cluster_size;
3583 }
3584
3585 return reserved_blocks;
3586 }
3587
f2fs_reserve_compress_blocks(struct file * filp,unsigned long arg)3588 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3589 {
3590 struct inode *inode = file_inode(filp);
3591 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3592 pgoff_t page_idx = 0, last_idx;
3593 unsigned int reserved_blocks = 0;
3594 int ret;
3595
3596 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3597 return -EOPNOTSUPP;
3598
3599 if (!f2fs_compressed_file(inode))
3600 return -EINVAL;
3601
3602 if (f2fs_readonly(sbi->sb))
3603 return -EROFS;
3604
3605 ret = mnt_want_write_file(filp);
3606 if (ret)
3607 return ret;
3608
3609 if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3610 goto out;
3611
3612 f2fs_balance_fs(F2FS_I_SB(inode), true);
3613
3614 inode_lock(inode);
3615
3616 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3617 ret = -EINVAL;
3618 goto unlock_inode;
3619 }
3620
3621 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3622 filemap_invalidate_lock(inode->i_mapping);
3623
3624 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3625
3626 while (page_idx < last_idx) {
3627 struct dnode_of_data dn;
3628 pgoff_t end_offset, count;
3629
3630 set_new_dnode(&dn, inode, NULL, NULL, 0);
3631 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3632 if (ret) {
3633 if (ret == -ENOENT) {
3634 page_idx = f2fs_get_next_page_offset(&dn,
3635 page_idx);
3636 ret = 0;
3637 continue;
3638 }
3639 break;
3640 }
3641
3642 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3643 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3644 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3645
3646 ret = reserve_compress_blocks(&dn, count);
3647
3648 f2fs_put_dnode(&dn);
3649
3650 if (ret < 0)
3651 break;
3652
3653 page_idx += count;
3654 reserved_blocks += ret;
3655 }
3656
3657 filemap_invalidate_unlock(inode->i_mapping);
3658 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3659
3660 if (ret >= 0) {
3661 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3662 inode->i_ctime = current_time(inode);
3663 f2fs_mark_inode_dirty_sync(inode, true);
3664 }
3665 unlock_inode:
3666 inode_unlock(inode);
3667 out:
3668 mnt_drop_write_file(filp);
3669
3670 if (ret >= 0) {
3671 ret = put_user(reserved_blocks, (u64 __user *)arg);
3672 } else if (reserved_blocks &&
3673 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3674 set_sbi_flag(sbi, SBI_NEED_FSCK);
3675 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3676 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3677 "run fsck to fix.",
3678 __func__, inode->i_ino, inode->i_blocks,
3679 reserved_blocks,
3680 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3681 }
3682
3683 return ret;
3684 }
3685
f2fs_secure_erase(struct block_device * bdev,struct inode * inode,pgoff_t off,block_t block,block_t len,u32 flags)3686 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3687 pgoff_t off, block_t block, block_t len, u32 flags)
3688 {
3689 sector_t sector = SECTOR_FROM_BLOCK(block);
3690 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3691 int ret = 0;
3692
3693 if (flags & F2FS_TRIM_FILE_DISCARD) {
3694 if (bdev_max_secure_erase_sectors(bdev))
3695 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3696 GFP_NOFS);
3697 else
3698 ret = blkdev_issue_discard(bdev, sector, nr_sects,
3699 GFP_NOFS);
3700 }
3701
3702 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3703 if (IS_ENCRYPTED(inode))
3704 ret = fscrypt_zeroout_range(inode, off, block, len);
3705 else
3706 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3707 GFP_NOFS, 0);
3708 }
3709
3710 return ret;
3711 }
3712
f2fs_sec_trim_file(struct file * filp,unsigned long arg)3713 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3714 {
3715 struct inode *inode = file_inode(filp);
3716 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3717 struct address_space *mapping = inode->i_mapping;
3718 struct block_device *prev_bdev = NULL;
3719 struct f2fs_sectrim_range range;
3720 pgoff_t index, pg_end, prev_index = 0;
3721 block_t prev_block = 0, len = 0;
3722 loff_t end_addr;
3723 bool to_end = false;
3724 int ret = 0;
3725
3726 if (!(filp->f_mode & FMODE_WRITE))
3727 return -EBADF;
3728
3729 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3730 sizeof(range)))
3731 return -EFAULT;
3732
3733 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3734 !S_ISREG(inode->i_mode))
3735 return -EINVAL;
3736
3737 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3738 !f2fs_hw_support_discard(sbi)) ||
3739 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3740 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3741 return -EOPNOTSUPP;
3742
3743 file_start_write(filp);
3744 inode_lock(inode);
3745
3746 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3747 range.start >= inode->i_size) {
3748 ret = -EINVAL;
3749 goto err;
3750 }
3751
3752 if (range.len == 0)
3753 goto err;
3754
3755 if (inode->i_size - range.start > range.len) {
3756 end_addr = range.start + range.len;
3757 } else {
3758 end_addr = range.len == (u64)-1 ?
3759 sbi->sb->s_maxbytes : inode->i_size;
3760 to_end = true;
3761 }
3762
3763 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3764 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3765 ret = -EINVAL;
3766 goto err;
3767 }
3768
3769 index = F2FS_BYTES_TO_BLK(range.start);
3770 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3771
3772 ret = f2fs_convert_inline_inode(inode);
3773 if (ret)
3774 goto err;
3775
3776 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3777 filemap_invalidate_lock(mapping);
3778
3779 ret = filemap_write_and_wait_range(mapping, range.start,
3780 to_end ? LLONG_MAX : end_addr - 1);
3781 if (ret)
3782 goto out;
3783
3784 truncate_inode_pages_range(mapping, range.start,
3785 to_end ? -1 : end_addr - 1);
3786
3787 while (index < pg_end) {
3788 struct dnode_of_data dn;
3789 pgoff_t end_offset, count;
3790 int i;
3791
3792 set_new_dnode(&dn, inode, NULL, NULL, 0);
3793 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3794 if (ret) {
3795 if (ret == -ENOENT) {
3796 index = f2fs_get_next_page_offset(&dn, index);
3797 continue;
3798 }
3799 goto out;
3800 }
3801
3802 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3803 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3804 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3805 struct block_device *cur_bdev;
3806 block_t blkaddr = f2fs_data_blkaddr(&dn);
3807
3808 if (!__is_valid_data_blkaddr(blkaddr))
3809 continue;
3810
3811 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3812 DATA_GENERIC_ENHANCE)) {
3813 ret = -EFSCORRUPTED;
3814 f2fs_put_dnode(&dn);
3815 f2fs_handle_error(sbi,
3816 ERROR_INVALID_BLKADDR);
3817 goto out;
3818 }
3819
3820 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3821 if (f2fs_is_multi_device(sbi)) {
3822 int di = f2fs_target_device_index(sbi, blkaddr);
3823
3824 blkaddr -= FDEV(di).start_blk;
3825 }
3826
3827 if (len) {
3828 if (prev_bdev == cur_bdev &&
3829 index == prev_index + len &&
3830 blkaddr == prev_block + len) {
3831 len++;
3832 } else {
3833 ret = f2fs_secure_erase(prev_bdev,
3834 inode, prev_index, prev_block,
3835 len, range.flags);
3836 if (ret) {
3837 f2fs_put_dnode(&dn);
3838 goto out;
3839 }
3840
3841 len = 0;
3842 }
3843 }
3844
3845 if (!len) {
3846 prev_bdev = cur_bdev;
3847 prev_index = index;
3848 prev_block = blkaddr;
3849 len = 1;
3850 }
3851 }
3852
3853 f2fs_put_dnode(&dn);
3854
3855 if (fatal_signal_pending(current)) {
3856 ret = -EINTR;
3857 goto out;
3858 }
3859 cond_resched();
3860 }
3861
3862 if (len)
3863 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3864 prev_block, len, range.flags);
3865 out:
3866 filemap_invalidate_unlock(mapping);
3867 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3868 err:
3869 inode_unlock(inode);
3870 file_end_write(filp);
3871
3872 return ret;
3873 }
3874
f2fs_ioc_get_compress_option(struct file * filp,unsigned long arg)3875 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3876 {
3877 struct inode *inode = file_inode(filp);
3878 struct f2fs_comp_option option;
3879
3880 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3881 return -EOPNOTSUPP;
3882
3883 inode_lock_shared(inode);
3884
3885 if (!f2fs_compressed_file(inode)) {
3886 inode_unlock_shared(inode);
3887 return -ENODATA;
3888 }
3889
3890 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3891 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3892
3893 inode_unlock_shared(inode);
3894
3895 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3896 sizeof(option)))
3897 return -EFAULT;
3898
3899 return 0;
3900 }
3901
f2fs_ioc_set_compress_option(struct file * filp,unsigned long arg)3902 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3903 {
3904 struct inode *inode = file_inode(filp);
3905 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3906 struct f2fs_comp_option option;
3907 int ret = 0;
3908
3909 if (!f2fs_sb_has_compression(sbi))
3910 return -EOPNOTSUPP;
3911
3912 if (!(filp->f_mode & FMODE_WRITE))
3913 return -EBADF;
3914
3915 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3916 sizeof(option)))
3917 return -EFAULT;
3918
3919 if (!f2fs_compressed_file(inode) ||
3920 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3921 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3922 option.algorithm >= COMPRESS_MAX)
3923 return -EINVAL;
3924
3925 file_start_write(filp);
3926 inode_lock(inode);
3927
3928 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3929 ret = -EBUSY;
3930 goto out;
3931 }
3932
3933 if (inode->i_size != 0) {
3934 ret = -EFBIG;
3935 goto out;
3936 }
3937
3938 F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3939 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3940 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3941 f2fs_mark_inode_dirty_sync(inode, true);
3942
3943 if (!f2fs_is_compress_backend_ready(inode))
3944 f2fs_warn(sbi, "compression algorithm is successfully set, "
3945 "but current kernel doesn't support this algorithm.");
3946 out:
3947 inode_unlock(inode);
3948 file_end_write(filp);
3949
3950 return ret;
3951 }
3952
redirty_blocks(struct inode * inode,pgoff_t page_idx,int len)3953 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3954 {
3955 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3956 struct address_space *mapping = inode->i_mapping;
3957 struct page *page;
3958 pgoff_t redirty_idx = page_idx;
3959 int i, page_len = 0, ret = 0;
3960
3961 page_cache_ra_unbounded(&ractl, len, 0);
3962
3963 for (i = 0; i < len; i++, page_idx++) {
3964 page = read_cache_page(mapping, page_idx, NULL, NULL);
3965 if (IS_ERR(page)) {
3966 ret = PTR_ERR(page);
3967 break;
3968 }
3969 page_len++;
3970 }
3971
3972 for (i = 0; i < page_len; i++, redirty_idx++) {
3973 page = find_lock_page(mapping, redirty_idx);
3974
3975 /* It will never fail, when page has pinned above */
3976 f2fs_bug_on(F2FS_I_SB(inode), !page);
3977
3978 set_page_dirty(page);
3979 f2fs_put_page(page, 1);
3980 f2fs_put_page(page, 0);
3981 }
3982
3983 return ret;
3984 }
3985
f2fs_ioc_decompress_file(struct file * filp,unsigned long arg)3986 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3987 {
3988 struct inode *inode = file_inode(filp);
3989 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3990 struct f2fs_inode_info *fi = F2FS_I(inode);
3991 pgoff_t page_idx = 0, last_idx;
3992 unsigned int blk_per_seg = sbi->blocks_per_seg;
3993 int cluster_size = fi->i_cluster_size;
3994 int count, ret;
3995
3996 if (!f2fs_sb_has_compression(sbi) ||
3997 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3998 return -EOPNOTSUPP;
3999
4000 if (!(filp->f_mode & FMODE_WRITE))
4001 return -EBADF;
4002
4003 if (!f2fs_compressed_file(inode))
4004 return -EINVAL;
4005
4006 f2fs_balance_fs(F2FS_I_SB(inode), true);
4007
4008 file_start_write(filp);
4009 inode_lock(inode);
4010
4011 if (!f2fs_is_compress_backend_ready(inode)) {
4012 ret = -EOPNOTSUPP;
4013 goto out;
4014 }
4015
4016 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4017 ret = -EINVAL;
4018 goto out;
4019 }
4020
4021 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4022 if (ret)
4023 goto out;
4024
4025 if (!atomic_read(&fi->i_compr_blocks))
4026 goto out;
4027
4028 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4029
4030 count = last_idx - page_idx;
4031 while (count) {
4032 int len = min(cluster_size, count);
4033
4034 ret = redirty_blocks(inode, page_idx, len);
4035 if (ret < 0)
4036 break;
4037
4038 if (get_dirty_pages(inode) >= blk_per_seg)
4039 filemap_fdatawrite(inode->i_mapping);
4040
4041 count -= len;
4042 page_idx += len;
4043 }
4044
4045 if (!ret)
4046 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4047 LLONG_MAX);
4048
4049 if (ret)
4050 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4051 __func__, ret);
4052 out:
4053 inode_unlock(inode);
4054 file_end_write(filp);
4055
4056 return ret;
4057 }
4058
f2fs_ioc_compress_file(struct file * filp,unsigned long arg)4059 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4060 {
4061 struct inode *inode = file_inode(filp);
4062 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4063 pgoff_t page_idx = 0, last_idx;
4064 unsigned int blk_per_seg = sbi->blocks_per_seg;
4065 int cluster_size = F2FS_I(inode)->i_cluster_size;
4066 int count, ret;
4067
4068 if (!f2fs_sb_has_compression(sbi) ||
4069 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4070 return -EOPNOTSUPP;
4071
4072 if (!(filp->f_mode & FMODE_WRITE))
4073 return -EBADF;
4074
4075 if (!f2fs_compressed_file(inode))
4076 return -EINVAL;
4077
4078 f2fs_balance_fs(F2FS_I_SB(inode), true);
4079
4080 file_start_write(filp);
4081 inode_lock(inode);
4082
4083 if (!f2fs_is_compress_backend_ready(inode)) {
4084 ret = -EOPNOTSUPP;
4085 goto out;
4086 }
4087
4088 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4089 ret = -EINVAL;
4090 goto out;
4091 }
4092
4093 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4094 if (ret)
4095 goto out;
4096
4097 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4098
4099 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4100
4101 count = last_idx - page_idx;
4102 while (count) {
4103 int len = min(cluster_size, count);
4104
4105 ret = redirty_blocks(inode, page_idx, len);
4106 if (ret < 0)
4107 break;
4108
4109 if (get_dirty_pages(inode) >= blk_per_seg)
4110 filemap_fdatawrite(inode->i_mapping);
4111
4112 count -= len;
4113 page_idx += len;
4114 }
4115
4116 if (!ret)
4117 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4118 LLONG_MAX);
4119
4120 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4121
4122 if (ret)
4123 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4124 __func__, ret);
4125 out:
4126 inode_unlock(inode);
4127 file_end_write(filp);
4128
4129 return ret;
4130 }
4131
__f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4132 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4133 {
4134 switch (cmd) {
4135 case FS_IOC_GETVERSION:
4136 return f2fs_ioc_getversion(filp, arg);
4137 case F2FS_IOC_START_ATOMIC_WRITE:
4138 return f2fs_ioc_start_atomic_write(filp);
4139 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4140 return f2fs_ioc_commit_atomic_write(filp);
4141 case F2FS_IOC_ABORT_ATOMIC_WRITE:
4142 return f2fs_ioc_abort_atomic_write(filp);
4143 case F2FS_IOC_START_VOLATILE_WRITE:
4144 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4145 return -EOPNOTSUPP;
4146 case F2FS_IOC_SHUTDOWN:
4147 return f2fs_ioc_shutdown(filp, arg);
4148 case FITRIM:
4149 return f2fs_ioc_fitrim(filp, arg);
4150 case FS_IOC_SET_ENCRYPTION_POLICY:
4151 return f2fs_ioc_set_encryption_policy(filp, arg);
4152 case FS_IOC_GET_ENCRYPTION_POLICY:
4153 return f2fs_ioc_get_encryption_policy(filp, arg);
4154 case FS_IOC_GET_ENCRYPTION_PWSALT:
4155 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4156 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4157 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4158 case FS_IOC_ADD_ENCRYPTION_KEY:
4159 return f2fs_ioc_add_encryption_key(filp, arg);
4160 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4161 return f2fs_ioc_remove_encryption_key(filp, arg);
4162 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4163 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4164 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4165 return f2fs_ioc_get_encryption_key_status(filp, arg);
4166 case FS_IOC_GET_ENCRYPTION_NONCE:
4167 return f2fs_ioc_get_encryption_nonce(filp, arg);
4168 case F2FS_IOC_GARBAGE_COLLECT:
4169 return f2fs_ioc_gc(filp, arg);
4170 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4171 return f2fs_ioc_gc_range(filp, arg);
4172 case F2FS_IOC_WRITE_CHECKPOINT:
4173 return f2fs_ioc_write_checkpoint(filp, arg);
4174 case F2FS_IOC_DEFRAGMENT:
4175 return f2fs_ioc_defragment(filp, arg);
4176 case F2FS_IOC_MOVE_RANGE:
4177 return f2fs_ioc_move_range(filp, arg);
4178 case F2FS_IOC_FLUSH_DEVICE:
4179 return f2fs_ioc_flush_device(filp, arg);
4180 case F2FS_IOC_GET_FEATURES:
4181 return f2fs_ioc_get_features(filp, arg);
4182 case F2FS_IOC_GET_PIN_FILE:
4183 return f2fs_ioc_get_pin_file(filp, arg);
4184 case F2FS_IOC_SET_PIN_FILE:
4185 return f2fs_ioc_set_pin_file(filp, arg);
4186 case F2FS_IOC_PRECACHE_EXTENTS:
4187 return f2fs_ioc_precache_extents(filp, arg);
4188 case F2FS_IOC_RESIZE_FS:
4189 return f2fs_ioc_resize_fs(filp, arg);
4190 case FS_IOC_ENABLE_VERITY:
4191 return f2fs_ioc_enable_verity(filp, arg);
4192 case FS_IOC_MEASURE_VERITY:
4193 return f2fs_ioc_measure_verity(filp, arg);
4194 case FS_IOC_READ_VERITY_METADATA:
4195 return f2fs_ioc_read_verity_metadata(filp, arg);
4196 case FS_IOC_GETFSLABEL:
4197 return f2fs_ioc_getfslabel(filp, arg);
4198 case FS_IOC_SETFSLABEL:
4199 return f2fs_ioc_setfslabel(filp, arg);
4200 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4201 return f2fs_get_compress_blocks(filp, arg);
4202 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4203 return f2fs_release_compress_blocks(filp, arg);
4204 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4205 return f2fs_reserve_compress_blocks(filp, arg);
4206 case F2FS_IOC_SEC_TRIM_FILE:
4207 return f2fs_sec_trim_file(filp, arg);
4208 case F2FS_IOC_GET_COMPRESS_OPTION:
4209 return f2fs_ioc_get_compress_option(filp, arg);
4210 case F2FS_IOC_SET_COMPRESS_OPTION:
4211 return f2fs_ioc_set_compress_option(filp, arg);
4212 case F2FS_IOC_DECOMPRESS_FILE:
4213 return f2fs_ioc_decompress_file(filp, arg);
4214 case F2FS_IOC_COMPRESS_FILE:
4215 return f2fs_ioc_compress_file(filp, arg);
4216 default:
4217 return -ENOTTY;
4218 }
4219 }
4220
f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4221 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4222 {
4223 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4224 return -EIO;
4225 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4226 return -ENOSPC;
4227
4228 return __f2fs_ioctl(filp, cmd, arg);
4229 }
4230
4231 /*
4232 * Return %true if the given read or write request should use direct I/O, or
4233 * %false if it should use buffered I/O.
4234 */
f2fs_should_use_dio(struct inode * inode,struct kiocb * iocb,struct iov_iter * iter)4235 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4236 struct iov_iter *iter)
4237 {
4238 unsigned int align;
4239
4240 if (!(iocb->ki_flags & IOCB_DIRECT))
4241 return false;
4242
4243 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4244 return false;
4245
4246 /*
4247 * Direct I/O not aligned to the disk's logical_block_size will be
4248 * attempted, but will fail with -EINVAL.
4249 *
4250 * f2fs additionally requires that direct I/O be aligned to the
4251 * filesystem block size, which is often a stricter requirement.
4252 * However, f2fs traditionally falls back to buffered I/O on requests
4253 * that are logical_block_size-aligned but not fs-block aligned.
4254 *
4255 * The below logic implements this behavior.
4256 */
4257 align = iocb->ki_pos | iov_iter_alignment(iter);
4258 if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4259 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4260 return false;
4261
4262 return true;
4263 }
4264
f2fs_dio_read_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4265 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4266 unsigned int flags)
4267 {
4268 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4269
4270 dec_page_count(sbi, F2FS_DIO_READ);
4271 if (error)
4272 return error;
4273 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4274 return 0;
4275 }
4276
4277 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4278 .end_io = f2fs_dio_read_end_io,
4279 };
4280
f2fs_dio_read_iter(struct kiocb * iocb,struct iov_iter * to)4281 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4282 {
4283 struct file *file = iocb->ki_filp;
4284 struct inode *inode = file_inode(file);
4285 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4286 struct f2fs_inode_info *fi = F2FS_I(inode);
4287 const loff_t pos = iocb->ki_pos;
4288 const size_t count = iov_iter_count(to);
4289 struct iomap_dio *dio;
4290 ssize_t ret;
4291
4292 if (count == 0)
4293 return 0; /* skip atime update */
4294
4295 trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4296
4297 if (iocb->ki_flags & IOCB_NOWAIT) {
4298 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4299 ret = -EAGAIN;
4300 goto out;
4301 }
4302 } else {
4303 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4304 }
4305
4306 /*
4307 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4308 * the higher-level function iomap_dio_rw() in order to ensure that the
4309 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4310 */
4311 inc_page_count(sbi, F2FS_DIO_READ);
4312 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4313 &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4314 if (IS_ERR_OR_NULL(dio)) {
4315 ret = PTR_ERR_OR_ZERO(dio);
4316 if (ret != -EIOCBQUEUED)
4317 dec_page_count(sbi, F2FS_DIO_READ);
4318 } else {
4319 ret = iomap_dio_complete(dio);
4320 }
4321
4322 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4323
4324 file_accessed(file);
4325 out:
4326 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4327 return ret;
4328 }
4329
f2fs_file_read_iter(struct kiocb * iocb,struct iov_iter * to)4330 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4331 {
4332 struct inode *inode = file_inode(iocb->ki_filp);
4333 const loff_t pos = iocb->ki_pos;
4334 ssize_t ret;
4335
4336 if (!f2fs_is_compress_backend_ready(inode))
4337 return -EOPNOTSUPP;
4338
4339 if (trace_f2fs_dataread_start_enabled()) {
4340 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4341 char *path;
4342
4343 if (!p)
4344 goto skip_read_trace;
4345
4346 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4347 if (IS_ERR(path)) {
4348 kfree(p);
4349 goto skip_read_trace;
4350 }
4351
4352 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4353 current->pid, path, current->comm);
4354 kfree(p);
4355 }
4356 skip_read_trace:
4357 if (f2fs_should_use_dio(inode, iocb, to)) {
4358 ret = f2fs_dio_read_iter(iocb, to);
4359 } else {
4360 ret = filemap_read(iocb, to, 0);
4361 if (ret > 0)
4362 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4363 APP_BUFFERED_READ_IO, ret);
4364 }
4365 if (trace_f2fs_dataread_end_enabled())
4366 trace_f2fs_dataread_end(inode, pos, ret);
4367 return ret;
4368 }
4369
f2fs_write_checks(struct kiocb * iocb,struct iov_iter * from)4370 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4371 {
4372 struct file *file = iocb->ki_filp;
4373 struct inode *inode = file_inode(file);
4374 ssize_t count;
4375 int err;
4376
4377 if (IS_IMMUTABLE(inode))
4378 return -EPERM;
4379
4380 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4381 return -EPERM;
4382
4383 count = generic_write_checks(iocb, from);
4384 if (count <= 0)
4385 return count;
4386
4387 err = file_modified(file);
4388 if (err)
4389 return err;
4390 return count;
4391 }
4392
4393 /*
4394 * Preallocate blocks for a write request, if it is possible and helpful to do
4395 * so. Returns a positive number if blocks may have been preallocated, 0 if no
4396 * blocks were preallocated, or a negative errno value if something went
4397 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4398 * requested blocks (not just some of them) have been allocated.
4399 */
f2fs_preallocate_blocks(struct kiocb * iocb,struct iov_iter * iter,bool dio)4400 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4401 bool dio)
4402 {
4403 struct inode *inode = file_inode(iocb->ki_filp);
4404 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4405 const loff_t pos = iocb->ki_pos;
4406 const size_t count = iov_iter_count(iter);
4407 struct f2fs_map_blocks map = {};
4408 int flag;
4409 int ret;
4410
4411 /* If it will be an out-of-place direct write, don't bother. */
4412 if (dio && f2fs_lfs_mode(sbi))
4413 return 0;
4414 /*
4415 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4416 * buffered IO, if DIO meets any holes.
4417 */
4418 if (dio && i_size_read(inode) &&
4419 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4420 return 0;
4421
4422 /* No-wait I/O can't allocate blocks. */
4423 if (iocb->ki_flags & IOCB_NOWAIT)
4424 return 0;
4425
4426 /* If it will be a short write, don't bother. */
4427 if (fault_in_iov_iter_readable(iter, count))
4428 return 0;
4429
4430 if (f2fs_has_inline_data(inode)) {
4431 /* If the data will fit inline, don't bother. */
4432 if (pos + count <= MAX_INLINE_DATA(inode))
4433 return 0;
4434 ret = f2fs_convert_inline_inode(inode);
4435 if (ret)
4436 return ret;
4437 }
4438
4439 /* Do not preallocate blocks that will be written partially in 4KB. */
4440 map.m_lblk = F2FS_BLK_ALIGN(pos);
4441 map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4442 if (map.m_len > map.m_lblk)
4443 map.m_len -= map.m_lblk;
4444 else
4445 map.m_len = 0;
4446 map.m_may_create = true;
4447 if (dio) {
4448 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4449 flag = F2FS_GET_BLOCK_PRE_DIO;
4450 } else {
4451 map.m_seg_type = NO_CHECK_TYPE;
4452 flag = F2FS_GET_BLOCK_PRE_AIO;
4453 }
4454
4455 ret = f2fs_map_blocks(inode, &map, 1, flag);
4456 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4457 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4458 return ret;
4459 if (ret == 0)
4460 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4461 return map.m_len;
4462 }
4463
f2fs_buffered_write_iter(struct kiocb * iocb,struct iov_iter * from)4464 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4465 struct iov_iter *from)
4466 {
4467 struct file *file = iocb->ki_filp;
4468 struct inode *inode = file_inode(file);
4469 ssize_t ret;
4470
4471 if (iocb->ki_flags & IOCB_NOWAIT)
4472 return -EOPNOTSUPP;
4473
4474 current->backing_dev_info = inode_to_bdi(inode);
4475 ret = generic_perform_write(iocb, from);
4476 current->backing_dev_info = NULL;
4477
4478 if (ret > 0) {
4479 iocb->ki_pos += ret;
4480 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4481 APP_BUFFERED_IO, ret);
4482 }
4483 return ret;
4484 }
4485
f2fs_dio_write_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4486 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4487 unsigned int flags)
4488 {
4489 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4490
4491 dec_page_count(sbi, F2FS_DIO_WRITE);
4492 if (error)
4493 return error;
4494 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4495 return 0;
4496 }
4497
4498 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4499 .end_io = f2fs_dio_write_end_io,
4500 };
4501
f2fs_dio_write_iter(struct kiocb * iocb,struct iov_iter * from,bool * may_need_sync)4502 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4503 bool *may_need_sync)
4504 {
4505 struct file *file = iocb->ki_filp;
4506 struct inode *inode = file_inode(file);
4507 struct f2fs_inode_info *fi = F2FS_I(inode);
4508 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4509 const bool do_opu = f2fs_lfs_mode(sbi);
4510 const loff_t pos = iocb->ki_pos;
4511 const ssize_t count = iov_iter_count(from);
4512 unsigned int dio_flags;
4513 struct iomap_dio *dio;
4514 ssize_t ret;
4515
4516 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4517
4518 if (iocb->ki_flags & IOCB_NOWAIT) {
4519 /* f2fs_convert_inline_inode() and block allocation can block */
4520 if (f2fs_has_inline_data(inode) ||
4521 !f2fs_overwrite_io(inode, pos, count)) {
4522 ret = -EAGAIN;
4523 goto out;
4524 }
4525
4526 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4527 ret = -EAGAIN;
4528 goto out;
4529 }
4530 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4531 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4532 ret = -EAGAIN;
4533 goto out;
4534 }
4535 } else {
4536 ret = f2fs_convert_inline_inode(inode);
4537 if (ret)
4538 goto out;
4539
4540 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4541 if (do_opu)
4542 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4543 }
4544
4545 /*
4546 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4547 * the higher-level function iomap_dio_rw() in order to ensure that the
4548 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4549 */
4550 inc_page_count(sbi, F2FS_DIO_WRITE);
4551 dio_flags = 0;
4552 if (pos + count > inode->i_size)
4553 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4554 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4555 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4556 if (IS_ERR_OR_NULL(dio)) {
4557 ret = PTR_ERR_OR_ZERO(dio);
4558 if (ret == -ENOTBLK)
4559 ret = 0;
4560 if (ret != -EIOCBQUEUED)
4561 dec_page_count(sbi, F2FS_DIO_WRITE);
4562 } else {
4563 ret = iomap_dio_complete(dio);
4564 }
4565
4566 if (do_opu)
4567 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4568 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4569
4570 if (ret < 0)
4571 goto out;
4572 if (pos + ret > inode->i_size)
4573 f2fs_i_size_write(inode, pos + ret);
4574 if (!do_opu)
4575 set_inode_flag(inode, FI_UPDATE_WRITE);
4576
4577 if (iov_iter_count(from)) {
4578 ssize_t ret2;
4579 loff_t bufio_start_pos = iocb->ki_pos;
4580
4581 /*
4582 * The direct write was partial, so we need to fall back to a
4583 * buffered write for the remainder.
4584 */
4585
4586 ret2 = f2fs_buffered_write_iter(iocb, from);
4587 if (iov_iter_count(from))
4588 f2fs_write_failed(inode, iocb->ki_pos);
4589 if (ret2 < 0)
4590 goto out;
4591
4592 /*
4593 * Ensure that the pagecache pages are written to disk and
4594 * invalidated to preserve the expected O_DIRECT semantics.
4595 */
4596 if (ret2 > 0) {
4597 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4598
4599 ret += ret2;
4600
4601 ret2 = filemap_write_and_wait_range(file->f_mapping,
4602 bufio_start_pos,
4603 bufio_end_pos);
4604 if (ret2 < 0)
4605 goto out;
4606 invalidate_mapping_pages(file->f_mapping,
4607 bufio_start_pos >> PAGE_SHIFT,
4608 bufio_end_pos >> PAGE_SHIFT);
4609 }
4610 } else {
4611 /* iomap_dio_rw() already handled the generic_write_sync(). */
4612 *may_need_sync = false;
4613 }
4614 out:
4615 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4616 return ret;
4617 }
4618
f2fs_file_write_iter(struct kiocb * iocb,struct iov_iter * from)4619 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4620 {
4621 struct inode *inode = file_inode(iocb->ki_filp);
4622 const loff_t orig_pos = iocb->ki_pos;
4623 const size_t orig_count = iov_iter_count(from);
4624 loff_t target_size;
4625 bool dio;
4626 bool may_need_sync = true;
4627 int preallocated;
4628 ssize_t ret;
4629
4630 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4631 ret = -EIO;
4632 goto out;
4633 }
4634
4635 if (!f2fs_is_compress_backend_ready(inode)) {
4636 ret = -EOPNOTSUPP;
4637 goto out;
4638 }
4639
4640 if (iocb->ki_flags & IOCB_NOWAIT) {
4641 if (!inode_trylock(inode)) {
4642 ret = -EAGAIN;
4643 goto out;
4644 }
4645 } else {
4646 inode_lock(inode);
4647 }
4648
4649 ret = f2fs_write_checks(iocb, from);
4650 if (ret <= 0)
4651 goto out_unlock;
4652
4653 /* Determine whether we will do a direct write or a buffered write. */
4654 dio = f2fs_should_use_dio(inode, iocb, from);
4655
4656 /* Possibly preallocate the blocks for the write. */
4657 target_size = iocb->ki_pos + iov_iter_count(from);
4658 preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4659 if (preallocated < 0) {
4660 ret = preallocated;
4661 } else {
4662 if (trace_f2fs_datawrite_start_enabled()) {
4663 char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4664 PATH_MAX, GFP_KERNEL);
4665 char *path;
4666
4667 if (!p)
4668 goto skip_write_trace;
4669 path = dentry_path_raw(file_dentry(iocb->ki_filp),
4670 p, PATH_MAX);
4671 if (IS_ERR(path)) {
4672 kfree(p);
4673 goto skip_write_trace;
4674 }
4675 trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4676 current->pid, path, current->comm);
4677 kfree(p);
4678 }
4679 skip_write_trace:
4680 /* Do the actual write. */
4681 ret = dio ?
4682 f2fs_dio_write_iter(iocb, from, &may_need_sync) :
4683 f2fs_buffered_write_iter(iocb, from);
4684
4685 if (trace_f2fs_datawrite_end_enabled())
4686 trace_f2fs_datawrite_end(inode, orig_pos, ret);
4687 }
4688
4689 /* Don't leave any preallocated blocks around past i_size. */
4690 if (preallocated && i_size_read(inode) < target_size) {
4691 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4692 filemap_invalidate_lock(inode->i_mapping);
4693 if (!f2fs_truncate(inode))
4694 file_dont_truncate(inode);
4695 filemap_invalidate_unlock(inode->i_mapping);
4696 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4697 } else {
4698 file_dont_truncate(inode);
4699 }
4700
4701 clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4702 out_unlock:
4703 inode_unlock(inode);
4704 out:
4705 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4706 if (ret > 0 && may_need_sync)
4707 ret = generic_write_sync(iocb, ret);
4708 return ret;
4709 }
4710
f2fs_file_fadvise(struct file * filp,loff_t offset,loff_t len,int advice)4711 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4712 int advice)
4713 {
4714 struct address_space *mapping;
4715 struct backing_dev_info *bdi;
4716 struct inode *inode = file_inode(filp);
4717 int err;
4718
4719 if (advice == POSIX_FADV_SEQUENTIAL) {
4720 if (S_ISFIFO(inode->i_mode))
4721 return -ESPIPE;
4722
4723 mapping = filp->f_mapping;
4724 if (!mapping || len < 0)
4725 return -EINVAL;
4726
4727 bdi = inode_to_bdi(mapping->host);
4728 filp->f_ra.ra_pages = bdi->ra_pages *
4729 F2FS_I_SB(inode)->seq_file_ra_mul;
4730 spin_lock(&filp->f_lock);
4731 filp->f_mode &= ~FMODE_RANDOM;
4732 spin_unlock(&filp->f_lock);
4733 return 0;
4734 }
4735
4736 err = generic_fadvise(filp, offset, len, advice);
4737 if (!err && advice == POSIX_FADV_DONTNEED &&
4738 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4739 f2fs_compressed_file(inode))
4740 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4741
4742 return err;
4743 }
4744
4745 #ifdef CONFIG_COMPAT
4746 struct compat_f2fs_gc_range {
4747 u32 sync;
4748 compat_u64 start;
4749 compat_u64 len;
4750 };
4751 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
4752 struct compat_f2fs_gc_range)
4753
f2fs_compat_ioc_gc_range(struct file * file,unsigned long arg)4754 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4755 {
4756 struct compat_f2fs_gc_range __user *urange;
4757 struct f2fs_gc_range range;
4758 int err;
4759
4760 urange = compat_ptr(arg);
4761 err = get_user(range.sync, &urange->sync);
4762 err |= get_user(range.start, &urange->start);
4763 err |= get_user(range.len, &urange->len);
4764 if (err)
4765 return -EFAULT;
4766
4767 return __f2fs_ioc_gc_range(file, &range);
4768 }
4769
4770 struct compat_f2fs_move_range {
4771 u32 dst_fd;
4772 compat_u64 pos_in;
4773 compat_u64 pos_out;
4774 compat_u64 len;
4775 };
4776 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
4777 struct compat_f2fs_move_range)
4778
f2fs_compat_ioc_move_range(struct file * file,unsigned long arg)4779 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4780 {
4781 struct compat_f2fs_move_range __user *urange;
4782 struct f2fs_move_range range;
4783 int err;
4784
4785 urange = compat_ptr(arg);
4786 err = get_user(range.dst_fd, &urange->dst_fd);
4787 err |= get_user(range.pos_in, &urange->pos_in);
4788 err |= get_user(range.pos_out, &urange->pos_out);
4789 err |= get_user(range.len, &urange->len);
4790 if (err)
4791 return -EFAULT;
4792
4793 return __f2fs_ioc_move_range(file, &range);
4794 }
4795
f2fs_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)4796 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4797 {
4798 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4799 return -EIO;
4800 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4801 return -ENOSPC;
4802
4803 switch (cmd) {
4804 case FS_IOC32_GETVERSION:
4805 cmd = FS_IOC_GETVERSION;
4806 break;
4807 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4808 return f2fs_compat_ioc_gc_range(file, arg);
4809 case F2FS_IOC32_MOVE_RANGE:
4810 return f2fs_compat_ioc_move_range(file, arg);
4811 case F2FS_IOC_START_ATOMIC_WRITE:
4812 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4813 case F2FS_IOC_START_VOLATILE_WRITE:
4814 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4815 case F2FS_IOC_ABORT_ATOMIC_WRITE:
4816 case F2FS_IOC_SHUTDOWN:
4817 case FITRIM:
4818 case FS_IOC_SET_ENCRYPTION_POLICY:
4819 case FS_IOC_GET_ENCRYPTION_PWSALT:
4820 case FS_IOC_GET_ENCRYPTION_POLICY:
4821 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4822 case FS_IOC_ADD_ENCRYPTION_KEY:
4823 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4824 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4825 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4826 case FS_IOC_GET_ENCRYPTION_NONCE:
4827 case F2FS_IOC_GARBAGE_COLLECT:
4828 case F2FS_IOC_WRITE_CHECKPOINT:
4829 case F2FS_IOC_DEFRAGMENT:
4830 case F2FS_IOC_FLUSH_DEVICE:
4831 case F2FS_IOC_GET_FEATURES:
4832 case F2FS_IOC_GET_PIN_FILE:
4833 case F2FS_IOC_SET_PIN_FILE:
4834 case F2FS_IOC_PRECACHE_EXTENTS:
4835 case F2FS_IOC_RESIZE_FS:
4836 case FS_IOC_ENABLE_VERITY:
4837 case FS_IOC_MEASURE_VERITY:
4838 case FS_IOC_READ_VERITY_METADATA:
4839 case FS_IOC_GETFSLABEL:
4840 case FS_IOC_SETFSLABEL:
4841 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4842 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4843 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4844 case F2FS_IOC_SEC_TRIM_FILE:
4845 case F2FS_IOC_GET_COMPRESS_OPTION:
4846 case F2FS_IOC_SET_COMPRESS_OPTION:
4847 case F2FS_IOC_DECOMPRESS_FILE:
4848 case F2FS_IOC_COMPRESS_FILE:
4849 break;
4850 default:
4851 return -ENOIOCTLCMD;
4852 }
4853 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4854 }
4855 #endif
4856
4857 const struct file_operations f2fs_file_operations = {
4858 .llseek = f2fs_llseek,
4859 .read_iter = f2fs_file_read_iter,
4860 .write_iter = f2fs_file_write_iter,
4861 .open = f2fs_file_open,
4862 .release = f2fs_release_file,
4863 .mmap = f2fs_file_mmap,
4864 .flush = f2fs_file_flush,
4865 .fsync = f2fs_sync_file,
4866 .fallocate = f2fs_fallocate,
4867 .unlocked_ioctl = f2fs_ioctl,
4868 #ifdef CONFIG_COMPAT
4869 .compat_ioctl = f2fs_compat_ioctl,
4870 #endif
4871 .splice_read = generic_file_splice_read,
4872 .splice_write = iter_file_splice_write,
4873 .fadvise = f2fs_file_fadvise,
4874 };
4875