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), APP_MAPPED_READ_IO,
47 F2FS_BLKSIZE);
48
49 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
50
51 return ret;
52 }
53
f2fs_vm_page_mkwrite(struct vm_fault * vmf)54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
55 {
56 struct page *page = vmf->page;
57 struct inode *inode = file_inode(vmf->vma->vm_file);
58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
59 struct dnode_of_data dn;
60 bool need_alloc = true;
61 int err = 0;
62
63 if (unlikely(IS_IMMUTABLE(inode)))
64 return VM_FAULT_SIGBUS;
65
66 if (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, 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_getattr(struct user_namespace * mnt_userns,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)811 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
812 struct kstat *stat, u32 request_mask, unsigned int query_flags)
813 {
814 struct inode *inode = d_inode(path->dentry);
815 struct f2fs_inode_info *fi = F2FS_I(inode);
816 struct f2fs_inode *ri = NULL;
817 unsigned int flags;
818
819 if (f2fs_has_extra_attr(inode) &&
820 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
821 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
822 stat->result_mask |= STATX_BTIME;
823 stat->btime.tv_sec = fi->i_crtime.tv_sec;
824 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
825 }
826
827 flags = fi->i_flags;
828 if (flags & F2FS_COMPR_FL)
829 stat->attributes |= STATX_ATTR_COMPRESSED;
830 if (flags & F2FS_APPEND_FL)
831 stat->attributes |= STATX_ATTR_APPEND;
832 if (IS_ENCRYPTED(inode))
833 stat->attributes |= STATX_ATTR_ENCRYPTED;
834 if (flags & F2FS_IMMUTABLE_FL)
835 stat->attributes |= STATX_ATTR_IMMUTABLE;
836 if (flags & F2FS_NODUMP_FL)
837 stat->attributes |= STATX_ATTR_NODUMP;
838 if (IS_VERITY(inode))
839 stat->attributes |= STATX_ATTR_VERITY;
840
841 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
842 STATX_ATTR_APPEND |
843 STATX_ATTR_ENCRYPTED |
844 STATX_ATTR_IMMUTABLE |
845 STATX_ATTR_NODUMP |
846 STATX_ATTR_VERITY);
847
848 generic_fillattr(mnt_userns, inode, stat);
849
850 /* we need to show initial sectors used for inline_data/dentries */
851 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
852 f2fs_has_inline_dentry(inode))
853 stat->blocks += (stat->size + 511) >> 9;
854
855 return 0;
856 }
857
858 #ifdef CONFIG_F2FS_FS_POSIX_ACL
__setattr_copy(struct user_namespace * mnt_userns,struct inode * inode,const struct iattr * attr)859 static void __setattr_copy(struct user_namespace *mnt_userns,
860 struct inode *inode, const struct iattr *attr)
861 {
862 unsigned int ia_valid = attr->ia_valid;
863
864 if (ia_valid & ATTR_UID)
865 inode->i_uid = attr->ia_uid;
866 if (ia_valid & ATTR_GID)
867 inode->i_gid = attr->ia_gid;
868 if (ia_valid & ATTR_ATIME)
869 inode->i_atime = attr->ia_atime;
870 if (ia_valid & ATTR_MTIME)
871 inode->i_mtime = attr->ia_mtime;
872 if (ia_valid & ATTR_CTIME)
873 inode->i_ctime = attr->ia_ctime;
874 if (ia_valid & ATTR_MODE) {
875 umode_t mode = attr->ia_mode;
876 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
877
878 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
879 mode &= ~S_ISGID;
880 set_acl_inode(inode, mode);
881 }
882 }
883 #else
884 #define __setattr_copy setattr_copy
885 #endif
886
f2fs_setattr(struct user_namespace * mnt_userns,struct dentry * dentry,struct iattr * attr)887 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
888 struct iattr *attr)
889 {
890 struct inode *inode = d_inode(dentry);
891 int err;
892
893 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
894 return -EIO;
895
896 if (unlikely(IS_IMMUTABLE(inode)))
897 return -EPERM;
898
899 if (unlikely(IS_APPEND(inode) &&
900 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
901 ATTR_GID | ATTR_TIMES_SET))))
902 return -EPERM;
903
904 if ((attr->ia_valid & ATTR_SIZE) &&
905 !f2fs_is_compress_backend_ready(inode))
906 return -EOPNOTSUPP;
907
908 err = setattr_prepare(mnt_userns, dentry, attr);
909 if (err)
910 return err;
911
912 err = fscrypt_prepare_setattr(dentry, attr);
913 if (err)
914 return err;
915
916 err = fsverity_prepare_setattr(dentry, attr);
917 if (err)
918 return err;
919
920 if (is_quota_modification(inode, attr)) {
921 err = f2fs_dquot_initialize(inode);
922 if (err)
923 return err;
924 }
925 if ((attr->ia_valid & ATTR_UID &&
926 !uid_eq(attr->ia_uid, inode->i_uid)) ||
927 (attr->ia_valid & ATTR_GID &&
928 !gid_eq(attr->ia_gid, inode->i_gid))) {
929 f2fs_lock_op(F2FS_I_SB(inode));
930 err = dquot_transfer(inode, attr);
931 if (err) {
932 set_sbi_flag(F2FS_I_SB(inode),
933 SBI_QUOTA_NEED_REPAIR);
934 f2fs_unlock_op(F2FS_I_SB(inode));
935 return err;
936 }
937 /*
938 * update uid/gid under lock_op(), so that dquot and inode can
939 * be updated atomically.
940 */
941 if (attr->ia_valid & ATTR_UID)
942 inode->i_uid = attr->ia_uid;
943 if (attr->ia_valid & ATTR_GID)
944 inode->i_gid = attr->ia_gid;
945 f2fs_mark_inode_dirty_sync(inode, true);
946 f2fs_unlock_op(F2FS_I_SB(inode));
947 }
948
949 if (attr->ia_valid & ATTR_SIZE) {
950 loff_t old_size = i_size_read(inode);
951
952 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
953 /*
954 * should convert inline inode before i_size_write to
955 * keep smaller than inline_data size with inline flag.
956 */
957 err = f2fs_convert_inline_inode(inode);
958 if (err)
959 return err;
960 }
961
962 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
963 filemap_invalidate_lock(inode->i_mapping);
964
965 truncate_setsize(inode, attr->ia_size);
966
967 if (attr->ia_size <= old_size)
968 err = f2fs_truncate(inode);
969 /*
970 * do not trim all blocks after i_size if target size is
971 * larger than i_size.
972 */
973 filemap_invalidate_unlock(inode->i_mapping);
974 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
975 if (err)
976 return err;
977
978 spin_lock(&F2FS_I(inode)->i_size_lock);
979 inode->i_mtime = inode->i_ctime = current_time(inode);
980 F2FS_I(inode)->last_disk_size = i_size_read(inode);
981 spin_unlock(&F2FS_I(inode)->i_size_lock);
982 }
983
984 __setattr_copy(mnt_userns, inode, attr);
985
986 if (attr->ia_valid & ATTR_MODE) {
987 err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
988
989 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
990 if (!err)
991 inode->i_mode = F2FS_I(inode)->i_acl_mode;
992 clear_inode_flag(inode, FI_ACL_MODE);
993 }
994 }
995
996 /* file size may changed here */
997 f2fs_mark_inode_dirty_sync(inode, true);
998
999 /* inode change will produce dirty node pages flushed by checkpoint */
1000 f2fs_balance_fs(F2FS_I_SB(inode), true);
1001
1002 return err;
1003 }
1004
1005 const struct inode_operations f2fs_file_inode_operations = {
1006 .getattr = f2fs_getattr,
1007 .setattr = f2fs_setattr,
1008 .get_acl = f2fs_get_acl,
1009 .set_acl = f2fs_set_acl,
1010 .listxattr = f2fs_listxattr,
1011 .fiemap = f2fs_fiemap,
1012 .fileattr_get = f2fs_fileattr_get,
1013 .fileattr_set = f2fs_fileattr_set,
1014 };
1015
fill_zero(struct inode * inode,pgoff_t index,loff_t start,loff_t len)1016 static int fill_zero(struct inode *inode, pgoff_t index,
1017 loff_t start, loff_t len)
1018 {
1019 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1020 struct page *page;
1021
1022 if (!len)
1023 return 0;
1024
1025 f2fs_balance_fs(sbi, true);
1026
1027 f2fs_lock_op(sbi);
1028 page = f2fs_get_new_data_page(inode, NULL, index, false);
1029 f2fs_unlock_op(sbi);
1030
1031 if (IS_ERR(page))
1032 return PTR_ERR(page);
1033
1034 f2fs_wait_on_page_writeback(page, DATA, true, true);
1035 zero_user(page, start, len);
1036 set_page_dirty(page);
1037 f2fs_put_page(page, 1);
1038 return 0;
1039 }
1040
f2fs_truncate_hole(struct inode * inode,pgoff_t pg_start,pgoff_t pg_end)1041 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1042 {
1043 int err;
1044
1045 while (pg_start < pg_end) {
1046 struct dnode_of_data dn;
1047 pgoff_t end_offset, count;
1048
1049 set_new_dnode(&dn, inode, NULL, NULL, 0);
1050 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1051 if (err) {
1052 if (err == -ENOENT) {
1053 pg_start = f2fs_get_next_page_offset(&dn,
1054 pg_start);
1055 continue;
1056 }
1057 return err;
1058 }
1059
1060 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1061 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1062
1063 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1064
1065 f2fs_truncate_data_blocks_range(&dn, count);
1066 f2fs_put_dnode(&dn);
1067
1068 pg_start += count;
1069 }
1070 return 0;
1071 }
1072
punch_hole(struct inode * inode,loff_t offset,loff_t len)1073 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1074 {
1075 pgoff_t pg_start, pg_end;
1076 loff_t off_start, off_end;
1077 int ret;
1078
1079 ret = f2fs_convert_inline_inode(inode);
1080 if (ret)
1081 return ret;
1082
1083 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1084 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1085
1086 off_start = offset & (PAGE_SIZE - 1);
1087 off_end = (offset + len) & (PAGE_SIZE - 1);
1088
1089 if (pg_start == pg_end) {
1090 ret = fill_zero(inode, pg_start, off_start,
1091 off_end - off_start);
1092 if (ret)
1093 return ret;
1094 } else {
1095 if (off_start) {
1096 ret = fill_zero(inode, pg_start++, off_start,
1097 PAGE_SIZE - off_start);
1098 if (ret)
1099 return ret;
1100 }
1101 if (off_end) {
1102 ret = fill_zero(inode, pg_end, 0, off_end);
1103 if (ret)
1104 return ret;
1105 }
1106
1107 if (pg_start < pg_end) {
1108 loff_t blk_start, blk_end;
1109 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1110
1111 f2fs_balance_fs(sbi, true);
1112
1113 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1114 blk_end = (loff_t)pg_end << PAGE_SHIFT;
1115
1116 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1117 filemap_invalidate_lock(inode->i_mapping);
1118
1119 truncate_pagecache_range(inode, blk_start, blk_end - 1);
1120
1121 f2fs_lock_op(sbi);
1122 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1123 f2fs_unlock_op(sbi);
1124
1125 filemap_invalidate_unlock(inode->i_mapping);
1126 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1127 }
1128 }
1129
1130 return ret;
1131 }
1132
__read_out_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,pgoff_t len)1133 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1134 int *do_replace, pgoff_t off, pgoff_t len)
1135 {
1136 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1137 struct dnode_of_data dn;
1138 int ret, done, i;
1139
1140 next_dnode:
1141 set_new_dnode(&dn, inode, NULL, NULL, 0);
1142 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1143 if (ret && ret != -ENOENT) {
1144 return ret;
1145 } else if (ret == -ENOENT) {
1146 if (dn.max_level == 0)
1147 return -ENOENT;
1148 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1149 dn.ofs_in_node, len);
1150 blkaddr += done;
1151 do_replace += done;
1152 goto next;
1153 }
1154
1155 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1156 dn.ofs_in_node, len);
1157 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1158 *blkaddr = f2fs_data_blkaddr(&dn);
1159
1160 if (__is_valid_data_blkaddr(*blkaddr) &&
1161 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1162 DATA_GENERIC_ENHANCE)) {
1163 f2fs_put_dnode(&dn);
1164 return -EFSCORRUPTED;
1165 }
1166
1167 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1168
1169 if (f2fs_lfs_mode(sbi)) {
1170 f2fs_put_dnode(&dn);
1171 return -EOPNOTSUPP;
1172 }
1173
1174 /* do not invalidate this block address */
1175 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1176 *do_replace = 1;
1177 }
1178 }
1179 f2fs_put_dnode(&dn);
1180 next:
1181 len -= done;
1182 off += done;
1183 if (len)
1184 goto next_dnode;
1185 return 0;
1186 }
1187
__roll_back_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,int len)1188 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1189 int *do_replace, pgoff_t off, int len)
1190 {
1191 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1192 struct dnode_of_data dn;
1193 int ret, i;
1194
1195 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1196 if (*do_replace == 0)
1197 continue;
1198
1199 set_new_dnode(&dn, inode, NULL, NULL, 0);
1200 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1201 if (ret) {
1202 dec_valid_block_count(sbi, inode, 1);
1203 f2fs_invalidate_blocks(sbi, *blkaddr);
1204 } else {
1205 f2fs_update_data_blkaddr(&dn, *blkaddr);
1206 }
1207 f2fs_put_dnode(&dn);
1208 }
1209 return 0;
1210 }
1211
__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)1212 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1213 block_t *blkaddr, int *do_replace,
1214 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1215 {
1216 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1217 pgoff_t i = 0;
1218 int ret;
1219
1220 while (i < len) {
1221 if (blkaddr[i] == NULL_ADDR && !full) {
1222 i++;
1223 continue;
1224 }
1225
1226 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1227 struct dnode_of_data dn;
1228 struct node_info ni;
1229 size_t new_size;
1230 pgoff_t ilen;
1231
1232 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1233 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1234 if (ret)
1235 return ret;
1236
1237 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1238 if (ret) {
1239 f2fs_put_dnode(&dn);
1240 return ret;
1241 }
1242
1243 ilen = min((pgoff_t)
1244 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1245 dn.ofs_in_node, len - i);
1246 do {
1247 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1248 f2fs_truncate_data_blocks_range(&dn, 1);
1249
1250 if (do_replace[i]) {
1251 f2fs_i_blocks_write(src_inode,
1252 1, false, false);
1253 f2fs_i_blocks_write(dst_inode,
1254 1, true, false);
1255 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1256 blkaddr[i], ni.version, true, false);
1257
1258 do_replace[i] = 0;
1259 }
1260 dn.ofs_in_node++;
1261 i++;
1262 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1263 if (dst_inode->i_size < new_size)
1264 f2fs_i_size_write(dst_inode, new_size);
1265 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1266
1267 f2fs_put_dnode(&dn);
1268 } else {
1269 struct page *psrc, *pdst;
1270
1271 psrc = f2fs_get_lock_data_page(src_inode,
1272 src + i, true);
1273 if (IS_ERR(psrc))
1274 return PTR_ERR(psrc);
1275 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1276 true);
1277 if (IS_ERR(pdst)) {
1278 f2fs_put_page(psrc, 1);
1279 return PTR_ERR(pdst);
1280 }
1281 f2fs_copy_page(psrc, pdst);
1282 set_page_dirty(pdst);
1283 f2fs_put_page(pdst, 1);
1284 f2fs_put_page(psrc, 1);
1285
1286 ret = f2fs_truncate_hole(src_inode,
1287 src + i, src + i + 1);
1288 if (ret)
1289 return ret;
1290 i++;
1291 }
1292 }
1293 return 0;
1294 }
1295
__exchange_data_block(struct inode * src_inode,struct inode * dst_inode,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1296 static int __exchange_data_block(struct inode *src_inode,
1297 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1298 pgoff_t len, bool full)
1299 {
1300 block_t *src_blkaddr;
1301 int *do_replace;
1302 pgoff_t olen;
1303 int ret;
1304
1305 while (len) {
1306 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1307
1308 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1309 array_size(olen, sizeof(block_t)),
1310 GFP_NOFS);
1311 if (!src_blkaddr)
1312 return -ENOMEM;
1313
1314 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1315 array_size(olen, sizeof(int)),
1316 GFP_NOFS);
1317 if (!do_replace) {
1318 kvfree(src_blkaddr);
1319 return -ENOMEM;
1320 }
1321
1322 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1323 do_replace, src, olen);
1324 if (ret)
1325 goto roll_back;
1326
1327 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1328 do_replace, src, dst, olen, full);
1329 if (ret)
1330 goto roll_back;
1331
1332 src += olen;
1333 dst += olen;
1334 len -= olen;
1335
1336 kvfree(src_blkaddr);
1337 kvfree(do_replace);
1338 }
1339 return 0;
1340
1341 roll_back:
1342 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1343 kvfree(src_blkaddr);
1344 kvfree(do_replace);
1345 return ret;
1346 }
1347
f2fs_do_collapse(struct inode * inode,loff_t offset,loff_t len)1348 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1349 {
1350 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1351 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1352 pgoff_t start = offset >> PAGE_SHIFT;
1353 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1354 int ret;
1355
1356 f2fs_balance_fs(sbi, true);
1357
1358 /* avoid gc operation during block exchange */
1359 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1360 filemap_invalidate_lock(inode->i_mapping);
1361
1362 f2fs_lock_op(sbi);
1363 f2fs_drop_extent_tree(inode);
1364 truncate_pagecache(inode, offset);
1365 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1366 f2fs_unlock_op(sbi);
1367
1368 filemap_invalidate_unlock(inode->i_mapping);
1369 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1370 return ret;
1371 }
1372
f2fs_collapse_range(struct inode * inode,loff_t offset,loff_t len)1373 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1374 {
1375 loff_t new_size;
1376 int ret;
1377
1378 if (offset + len >= i_size_read(inode))
1379 return -EINVAL;
1380
1381 /* collapse range should be aligned to block size of f2fs. */
1382 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1383 return -EINVAL;
1384
1385 ret = f2fs_convert_inline_inode(inode);
1386 if (ret)
1387 return ret;
1388
1389 /* write out all dirty pages from offset */
1390 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1391 if (ret)
1392 return ret;
1393
1394 ret = f2fs_do_collapse(inode, offset, len);
1395 if (ret)
1396 return ret;
1397
1398 /* write out all moved pages, if possible */
1399 filemap_invalidate_lock(inode->i_mapping);
1400 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1401 truncate_pagecache(inode, offset);
1402
1403 new_size = i_size_read(inode) - len;
1404 ret = f2fs_truncate_blocks(inode, new_size, true);
1405 filemap_invalidate_unlock(inode->i_mapping);
1406 if (!ret)
1407 f2fs_i_size_write(inode, new_size);
1408 return ret;
1409 }
1410
f2fs_do_zero_range(struct dnode_of_data * dn,pgoff_t start,pgoff_t end)1411 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1412 pgoff_t end)
1413 {
1414 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1415 pgoff_t index = start;
1416 unsigned int ofs_in_node = dn->ofs_in_node;
1417 blkcnt_t count = 0;
1418 int ret;
1419
1420 for (; index < end; index++, dn->ofs_in_node++) {
1421 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1422 count++;
1423 }
1424
1425 dn->ofs_in_node = ofs_in_node;
1426 ret = f2fs_reserve_new_blocks(dn, count);
1427 if (ret)
1428 return ret;
1429
1430 dn->ofs_in_node = ofs_in_node;
1431 for (index = start; index < end; index++, dn->ofs_in_node++) {
1432 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1433 /*
1434 * f2fs_reserve_new_blocks will not guarantee entire block
1435 * allocation.
1436 */
1437 if (dn->data_blkaddr == NULL_ADDR) {
1438 ret = -ENOSPC;
1439 break;
1440 }
1441
1442 if (dn->data_blkaddr == NEW_ADDR)
1443 continue;
1444
1445 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1446 DATA_GENERIC_ENHANCE)) {
1447 ret = -EFSCORRUPTED;
1448 break;
1449 }
1450
1451 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1452 dn->data_blkaddr = NEW_ADDR;
1453 f2fs_set_data_blkaddr(dn);
1454 }
1455
1456 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1457
1458 return ret;
1459 }
1460
f2fs_zero_range(struct inode * inode,loff_t offset,loff_t len,int mode)1461 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1462 int mode)
1463 {
1464 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1465 struct address_space *mapping = inode->i_mapping;
1466 pgoff_t index, pg_start, pg_end;
1467 loff_t new_size = i_size_read(inode);
1468 loff_t off_start, off_end;
1469 int ret = 0;
1470
1471 ret = inode_newsize_ok(inode, (len + offset));
1472 if (ret)
1473 return ret;
1474
1475 ret = f2fs_convert_inline_inode(inode);
1476 if (ret)
1477 return ret;
1478
1479 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1480 if (ret)
1481 return ret;
1482
1483 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1484 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1485
1486 off_start = offset & (PAGE_SIZE - 1);
1487 off_end = (offset + len) & (PAGE_SIZE - 1);
1488
1489 if (pg_start == pg_end) {
1490 ret = fill_zero(inode, pg_start, off_start,
1491 off_end - off_start);
1492 if (ret)
1493 return ret;
1494
1495 new_size = max_t(loff_t, new_size, offset + len);
1496 } else {
1497 if (off_start) {
1498 ret = fill_zero(inode, pg_start++, off_start,
1499 PAGE_SIZE - off_start);
1500 if (ret)
1501 return ret;
1502
1503 new_size = max_t(loff_t, new_size,
1504 (loff_t)pg_start << PAGE_SHIFT);
1505 }
1506
1507 for (index = pg_start; index < pg_end;) {
1508 struct dnode_of_data dn;
1509 unsigned int end_offset;
1510 pgoff_t end;
1511
1512 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1513 filemap_invalidate_lock(mapping);
1514
1515 truncate_pagecache_range(inode,
1516 (loff_t)index << PAGE_SHIFT,
1517 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1518
1519 f2fs_lock_op(sbi);
1520
1521 set_new_dnode(&dn, inode, NULL, NULL, 0);
1522 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1523 if (ret) {
1524 f2fs_unlock_op(sbi);
1525 filemap_invalidate_unlock(mapping);
1526 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1527 goto out;
1528 }
1529
1530 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1531 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1532
1533 ret = f2fs_do_zero_range(&dn, index, end);
1534 f2fs_put_dnode(&dn);
1535
1536 f2fs_unlock_op(sbi);
1537 filemap_invalidate_unlock(mapping);
1538 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1539
1540 f2fs_balance_fs(sbi, dn.node_changed);
1541
1542 if (ret)
1543 goto out;
1544
1545 index = end;
1546 new_size = max_t(loff_t, new_size,
1547 (loff_t)index << PAGE_SHIFT);
1548 }
1549
1550 if (off_end) {
1551 ret = fill_zero(inode, pg_end, 0, off_end);
1552 if (ret)
1553 goto out;
1554
1555 new_size = max_t(loff_t, new_size, offset + len);
1556 }
1557 }
1558
1559 out:
1560 if (new_size > i_size_read(inode)) {
1561 if (mode & FALLOC_FL_KEEP_SIZE)
1562 file_set_keep_isize(inode);
1563 else
1564 f2fs_i_size_write(inode, new_size);
1565 }
1566 return ret;
1567 }
1568
f2fs_insert_range(struct inode * inode,loff_t offset,loff_t len)1569 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1570 {
1571 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1572 struct address_space *mapping = inode->i_mapping;
1573 pgoff_t nr, pg_start, pg_end, delta, idx;
1574 loff_t new_size;
1575 int ret = 0;
1576
1577 new_size = i_size_read(inode) + len;
1578 ret = inode_newsize_ok(inode, new_size);
1579 if (ret)
1580 return ret;
1581
1582 if (offset >= i_size_read(inode))
1583 return -EINVAL;
1584
1585 /* insert range should be aligned to block size of f2fs. */
1586 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1587 return -EINVAL;
1588
1589 ret = f2fs_convert_inline_inode(inode);
1590 if (ret)
1591 return ret;
1592
1593 f2fs_balance_fs(sbi, true);
1594
1595 filemap_invalidate_lock(mapping);
1596 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1597 filemap_invalidate_unlock(mapping);
1598 if (ret)
1599 return ret;
1600
1601 /* write out all dirty pages from offset */
1602 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1603 if (ret)
1604 return ret;
1605
1606 pg_start = offset >> PAGE_SHIFT;
1607 pg_end = (offset + len) >> PAGE_SHIFT;
1608 delta = pg_end - pg_start;
1609 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1610
1611 /* avoid gc operation during block exchange */
1612 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1613 filemap_invalidate_lock(mapping);
1614 truncate_pagecache(inode, offset);
1615
1616 while (!ret && idx > pg_start) {
1617 nr = idx - pg_start;
1618 if (nr > delta)
1619 nr = delta;
1620 idx -= nr;
1621
1622 f2fs_lock_op(sbi);
1623 f2fs_drop_extent_tree(inode);
1624
1625 ret = __exchange_data_block(inode, inode, idx,
1626 idx + delta, nr, false);
1627 f2fs_unlock_op(sbi);
1628 }
1629 filemap_invalidate_unlock(mapping);
1630 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1631
1632 /* write out all moved pages, if possible */
1633 filemap_invalidate_lock(mapping);
1634 filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1635 truncate_pagecache(inode, offset);
1636 filemap_invalidate_unlock(mapping);
1637
1638 if (!ret)
1639 f2fs_i_size_write(inode, new_size);
1640 return ret;
1641 }
1642
expand_inode_data(struct inode * inode,loff_t offset,loff_t len,int mode)1643 static int expand_inode_data(struct inode *inode, loff_t offset,
1644 loff_t len, int mode)
1645 {
1646 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1647 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1648 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1649 .m_may_create = true };
1650 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1651 .init_gc_type = FG_GC,
1652 .should_migrate_blocks = false,
1653 .err_gc_skipped = true,
1654 .nr_free_secs = 0 };
1655 pgoff_t pg_start, pg_end;
1656 loff_t new_size = i_size_read(inode);
1657 loff_t off_end;
1658 block_t expanded = 0;
1659 int err;
1660
1661 err = inode_newsize_ok(inode, (len + offset));
1662 if (err)
1663 return err;
1664
1665 err = f2fs_convert_inline_inode(inode);
1666 if (err)
1667 return err;
1668
1669 f2fs_balance_fs(sbi, true);
1670
1671 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1672 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1673 off_end = (offset + len) & (PAGE_SIZE - 1);
1674
1675 map.m_lblk = pg_start;
1676 map.m_len = pg_end - pg_start;
1677 if (off_end)
1678 map.m_len++;
1679
1680 if (!map.m_len)
1681 return 0;
1682
1683 if (f2fs_is_pinned_file(inode)) {
1684 block_t sec_blks = BLKS_PER_SEC(sbi);
1685 block_t sec_len = roundup(map.m_len, sec_blks);
1686
1687 map.m_len = sec_blks;
1688 next_alloc:
1689 if (has_not_enough_free_secs(sbi, 0,
1690 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1691 f2fs_down_write(&sbi->gc_lock);
1692 err = f2fs_gc(sbi, &gc_control);
1693 if (err && err != -ENODATA)
1694 goto out_err;
1695 }
1696
1697 f2fs_down_write(&sbi->pin_sem);
1698
1699 f2fs_lock_op(sbi);
1700 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1701 f2fs_unlock_op(sbi);
1702
1703 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1704 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1705 file_dont_truncate(inode);
1706
1707 f2fs_up_write(&sbi->pin_sem);
1708
1709 expanded += map.m_len;
1710 sec_len -= map.m_len;
1711 map.m_lblk += map.m_len;
1712 if (!err && sec_len)
1713 goto next_alloc;
1714
1715 map.m_len = expanded;
1716 } else {
1717 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1718 expanded = map.m_len;
1719 }
1720 out_err:
1721 if (err) {
1722 pgoff_t last_off;
1723
1724 if (!expanded)
1725 return err;
1726
1727 last_off = pg_start + expanded - 1;
1728
1729 /* update new size to the failed position */
1730 new_size = (last_off == pg_end) ? offset + len :
1731 (loff_t)(last_off + 1) << PAGE_SHIFT;
1732 } else {
1733 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1734 }
1735
1736 if (new_size > i_size_read(inode)) {
1737 if (mode & FALLOC_FL_KEEP_SIZE)
1738 file_set_keep_isize(inode);
1739 else
1740 f2fs_i_size_write(inode, new_size);
1741 }
1742
1743 return err;
1744 }
1745
f2fs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)1746 static long f2fs_fallocate(struct file *file, int mode,
1747 loff_t offset, loff_t len)
1748 {
1749 struct inode *inode = file_inode(file);
1750 long ret = 0;
1751
1752 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1753 return -EIO;
1754 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1755 return -ENOSPC;
1756 if (!f2fs_is_compress_backend_ready(inode))
1757 return -EOPNOTSUPP;
1758
1759 /* f2fs only support ->fallocate for regular file */
1760 if (!S_ISREG(inode->i_mode))
1761 return -EINVAL;
1762
1763 if (IS_ENCRYPTED(inode) &&
1764 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1765 return -EOPNOTSUPP;
1766
1767 /*
1768 * Pinned file should not support partial trucation since the block
1769 * can be used by applications.
1770 */
1771 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1772 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1773 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1774 return -EOPNOTSUPP;
1775
1776 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1777 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1778 FALLOC_FL_INSERT_RANGE))
1779 return -EOPNOTSUPP;
1780
1781 inode_lock(inode);
1782
1783 ret = file_modified(file);
1784 if (ret)
1785 goto out;
1786
1787 if (mode & FALLOC_FL_PUNCH_HOLE) {
1788 if (offset >= inode->i_size)
1789 goto out;
1790
1791 ret = punch_hole(inode, offset, len);
1792 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1793 ret = f2fs_collapse_range(inode, offset, len);
1794 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1795 ret = f2fs_zero_range(inode, offset, len, mode);
1796 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1797 ret = f2fs_insert_range(inode, offset, len);
1798 } else {
1799 ret = expand_inode_data(inode, offset, len, mode);
1800 }
1801
1802 if (!ret) {
1803 inode->i_mtime = inode->i_ctime = current_time(inode);
1804 f2fs_mark_inode_dirty_sync(inode, false);
1805 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1806 }
1807
1808 out:
1809 inode_unlock(inode);
1810
1811 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1812 return ret;
1813 }
1814
f2fs_release_file(struct inode * inode,struct file * filp)1815 static int f2fs_release_file(struct inode *inode, struct file *filp)
1816 {
1817 /*
1818 * f2fs_relase_file is called at every close calls. So we should
1819 * not drop any inmemory pages by close called by other process.
1820 */
1821 if (!(filp->f_mode & FMODE_WRITE) ||
1822 atomic_read(&inode->i_writecount) != 1)
1823 return 0;
1824
1825 if (f2fs_is_atomic_file(inode))
1826 f2fs_abort_atomic_write(inode, true);
1827 return 0;
1828 }
1829
f2fs_file_flush(struct file * file,fl_owner_t id)1830 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1831 {
1832 struct inode *inode = file_inode(file);
1833
1834 /*
1835 * If the process doing a transaction is crashed, we should do
1836 * roll-back. Otherwise, other reader/write can see corrupted database
1837 * until all the writers close its file. Since this should be done
1838 * before dropping file lock, it needs to do in ->flush.
1839 */
1840 if (f2fs_is_atomic_file(inode) &&
1841 F2FS_I(inode)->atomic_write_task == current)
1842 f2fs_abort_atomic_write(inode, true);
1843 return 0;
1844 }
1845
f2fs_setflags_common(struct inode * inode,u32 iflags,u32 mask)1846 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1847 {
1848 struct f2fs_inode_info *fi = F2FS_I(inode);
1849 u32 masked_flags = fi->i_flags & mask;
1850
1851 /* mask can be shrunk by flags_valid selector */
1852 iflags &= mask;
1853
1854 /* Is it quota file? Do not allow user to mess with it */
1855 if (IS_NOQUOTA(inode))
1856 return -EPERM;
1857
1858 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1859 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1860 return -EOPNOTSUPP;
1861 if (!f2fs_empty_dir(inode))
1862 return -ENOTEMPTY;
1863 }
1864
1865 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1866 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1867 return -EOPNOTSUPP;
1868 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1869 return -EINVAL;
1870 }
1871
1872 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1873 if (masked_flags & F2FS_COMPR_FL) {
1874 if (!f2fs_disable_compressed_file(inode))
1875 return -EINVAL;
1876 } else {
1877 if (!f2fs_may_compress(inode))
1878 return -EINVAL;
1879 if (S_ISREG(inode->i_mode) && inode->i_size)
1880 return -EINVAL;
1881
1882 set_compress_context(inode);
1883 }
1884 }
1885
1886 fi->i_flags = iflags | (fi->i_flags & ~mask);
1887 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1888 (fi->i_flags & F2FS_NOCOMP_FL));
1889
1890 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1891 set_inode_flag(inode, FI_PROJ_INHERIT);
1892 else
1893 clear_inode_flag(inode, FI_PROJ_INHERIT);
1894
1895 inode->i_ctime = current_time(inode);
1896 f2fs_set_inode_flags(inode);
1897 f2fs_mark_inode_dirty_sync(inode, true);
1898 return 0;
1899 }
1900
1901 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1902
1903 /*
1904 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1905 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1906 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1907 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1908 *
1909 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1910 * FS_IOC_FSSETXATTR is done by the VFS.
1911 */
1912
1913 static const struct {
1914 u32 iflag;
1915 u32 fsflag;
1916 } f2fs_fsflags_map[] = {
1917 { F2FS_COMPR_FL, FS_COMPR_FL },
1918 { F2FS_SYNC_FL, FS_SYNC_FL },
1919 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1920 { F2FS_APPEND_FL, FS_APPEND_FL },
1921 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1922 { F2FS_NOATIME_FL, FS_NOATIME_FL },
1923 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
1924 { F2FS_INDEX_FL, FS_INDEX_FL },
1925 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1926 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
1927 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1928 };
1929
1930 #define F2FS_GETTABLE_FS_FL ( \
1931 FS_COMPR_FL | \
1932 FS_SYNC_FL | \
1933 FS_IMMUTABLE_FL | \
1934 FS_APPEND_FL | \
1935 FS_NODUMP_FL | \
1936 FS_NOATIME_FL | \
1937 FS_NOCOMP_FL | \
1938 FS_INDEX_FL | \
1939 FS_DIRSYNC_FL | \
1940 FS_PROJINHERIT_FL | \
1941 FS_ENCRYPT_FL | \
1942 FS_INLINE_DATA_FL | \
1943 FS_NOCOW_FL | \
1944 FS_VERITY_FL | \
1945 FS_CASEFOLD_FL)
1946
1947 #define F2FS_SETTABLE_FS_FL ( \
1948 FS_COMPR_FL | \
1949 FS_SYNC_FL | \
1950 FS_IMMUTABLE_FL | \
1951 FS_APPEND_FL | \
1952 FS_NODUMP_FL | \
1953 FS_NOATIME_FL | \
1954 FS_NOCOMP_FL | \
1955 FS_DIRSYNC_FL | \
1956 FS_PROJINHERIT_FL | \
1957 FS_CASEFOLD_FL)
1958
1959 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
f2fs_iflags_to_fsflags(u32 iflags)1960 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1961 {
1962 u32 fsflags = 0;
1963 int i;
1964
1965 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1966 if (iflags & f2fs_fsflags_map[i].iflag)
1967 fsflags |= f2fs_fsflags_map[i].fsflag;
1968
1969 return fsflags;
1970 }
1971
1972 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
f2fs_fsflags_to_iflags(u32 fsflags)1973 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1974 {
1975 u32 iflags = 0;
1976 int i;
1977
1978 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1979 if (fsflags & f2fs_fsflags_map[i].fsflag)
1980 iflags |= f2fs_fsflags_map[i].iflag;
1981
1982 return iflags;
1983 }
1984
f2fs_ioc_getversion(struct file * filp,unsigned long arg)1985 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1986 {
1987 struct inode *inode = file_inode(filp);
1988
1989 return put_user(inode->i_generation, (int __user *)arg);
1990 }
1991
f2fs_ioc_start_atomic_write(struct file * filp)1992 static int f2fs_ioc_start_atomic_write(struct file *filp)
1993 {
1994 struct inode *inode = file_inode(filp);
1995 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
1996 struct f2fs_inode_info *fi = F2FS_I(inode);
1997 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1998 struct inode *pinode;
1999 int ret;
2000
2001 if (!inode_owner_or_capable(mnt_userns, inode))
2002 return -EACCES;
2003
2004 if (!S_ISREG(inode->i_mode))
2005 return -EINVAL;
2006
2007 if (filp->f_flags & O_DIRECT)
2008 return -EINVAL;
2009
2010 ret = mnt_want_write_file(filp);
2011 if (ret)
2012 return ret;
2013
2014 inode_lock(inode);
2015
2016 if (!f2fs_disable_compressed_file(inode)) {
2017 ret = -EINVAL;
2018 goto out;
2019 }
2020
2021 if (f2fs_is_atomic_file(inode))
2022 goto out;
2023
2024 ret = f2fs_convert_inline_inode(inode);
2025 if (ret)
2026 goto out;
2027
2028 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2029
2030 /*
2031 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2032 * f2fs_is_atomic_file.
2033 */
2034 if (get_dirty_pages(inode))
2035 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2036 inode->i_ino, get_dirty_pages(inode));
2037 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2038 if (ret) {
2039 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2040 goto out;
2041 }
2042
2043 /* Create a COW inode for atomic write */
2044 pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2045 if (IS_ERR(pinode)) {
2046 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2047 ret = PTR_ERR(pinode);
2048 goto out;
2049 }
2050
2051 ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2052 iput(pinode);
2053 if (ret) {
2054 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2055 goto out;
2056 }
2057 f2fs_i_size_write(fi->cow_inode, i_size_read(inode));
2058
2059 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2060 sbi->atomic_files++;
2061 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2062
2063 set_inode_flag(inode, FI_ATOMIC_FILE);
2064 set_inode_flag(fi->cow_inode, FI_COW_FILE);
2065 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2066 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2067
2068 f2fs_update_time(sbi, REQ_TIME);
2069 fi->atomic_write_task = current;
2070 stat_update_max_atomic_write(inode);
2071 out:
2072 inode_unlock(inode);
2073 mnt_drop_write_file(filp);
2074 return ret;
2075 }
2076
f2fs_ioc_commit_atomic_write(struct file * filp)2077 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2078 {
2079 struct inode *inode = file_inode(filp);
2080 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2081 int ret;
2082
2083 if (!inode_owner_or_capable(mnt_userns, inode))
2084 return -EACCES;
2085
2086 ret = mnt_want_write_file(filp);
2087 if (ret)
2088 return ret;
2089
2090 f2fs_balance_fs(F2FS_I_SB(inode), true);
2091
2092 inode_lock(inode);
2093
2094 if (f2fs_is_atomic_file(inode)) {
2095 ret = f2fs_commit_atomic_write(inode);
2096 if (ret)
2097 goto unlock_out;
2098
2099 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2100 if (!ret)
2101 f2fs_abort_atomic_write(inode, false);
2102 } else {
2103 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2104 }
2105 unlock_out:
2106 inode_unlock(inode);
2107 mnt_drop_write_file(filp);
2108 return ret;
2109 }
2110
f2fs_ioc_abort_atomic_write(struct file * filp)2111 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2112 {
2113 struct inode *inode = file_inode(filp);
2114 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2115 int ret;
2116
2117 if (!inode_owner_or_capable(mnt_userns, inode))
2118 return -EACCES;
2119
2120 ret = mnt_want_write_file(filp);
2121 if (ret)
2122 return ret;
2123
2124 inode_lock(inode);
2125
2126 if (f2fs_is_atomic_file(inode))
2127 f2fs_abort_atomic_write(inode, true);
2128
2129 inode_unlock(inode);
2130
2131 mnt_drop_write_file(filp);
2132 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2133 return ret;
2134 }
2135
f2fs_ioc_shutdown(struct file * filp,unsigned long arg)2136 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2137 {
2138 struct inode *inode = file_inode(filp);
2139 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2140 struct super_block *sb = sbi->sb;
2141 __u32 in;
2142 int ret = 0;
2143
2144 if (!capable(CAP_SYS_ADMIN))
2145 return -EPERM;
2146
2147 if (get_user(in, (__u32 __user *)arg))
2148 return -EFAULT;
2149
2150 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2151 ret = mnt_want_write_file(filp);
2152 if (ret) {
2153 if (ret == -EROFS) {
2154 ret = 0;
2155 f2fs_stop_checkpoint(sbi, false);
2156 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2157 trace_f2fs_shutdown(sbi, in, ret);
2158 }
2159 return ret;
2160 }
2161 }
2162
2163 switch (in) {
2164 case F2FS_GOING_DOWN_FULLSYNC:
2165 ret = freeze_bdev(sb->s_bdev);
2166 if (ret)
2167 goto out;
2168 f2fs_stop_checkpoint(sbi, false);
2169 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2170 thaw_bdev(sb->s_bdev);
2171 break;
2172 case F2FS_GOING_DOWN_METASYNC:
2173 /* do checkpoint only */
2174 ret = f2fs_sync_fs(sb, 1);
2175 if (ret)
2176 goto out;
2177 f2fs_stop_checkpoint(sbi, false);
2178 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2179 break;
2180 case F2FS_GOING_DOWN_NOSYNC:
2181 f2fs_stop_checkpoint(sbi, false);
2182 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2183 break;
2184 case F2FS_GOING_DOWN_METAFLUSH:
2185 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2186 f2fs_stop_checkpoint(sbi, false);
2187 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2188 break;
2189 case F2FS_GOING_DOWN_NEED_FSCK:
2190 set_sbi_flag(sbi, SBI_NEED_FSCK);
2191 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2192 set_sbi_flag(sbi, SBI_IS_DIRTY);
2193 /* do checkpoint only */
2194 ret = f2fs_sync_fs(sb, 1);
2195 goto out;
2196 default:
2197 ret = -EINVAL;
2198 goto out;
2199 }
2200
2201 f2fs_stop_gc_thread(sbi);
2202 f2fs_stop_discard_thread(sbi);
2203
2204 f2fs_drop_discard_cmd(sbi);
2205 clear_opt(sbi, DISCARD);
2206
2207 f2fs_update_time(sbi, REQ_TIME);
2208 out:
2209 if (in != F2FS_GOING_DOWN_FULLSYNC)
2210 mnt_drop_write_file(filp);
2211
2212 trace_f2fs_shutdown(sbi, in, ret);
2213
2214 return ret;
2215 }
2216
f2fs_ioc_fitrim(struct file * filp,unsigned long arg)2217 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2218 {
2219 struct inode *inode = file_inode(filp);
2220 struct super_block *sb = inode->i_sb;
2221 struct fstrim_range range;
2222 int ret;
2223
2224 if (!capable(CAP_SYS_ADMIN))
2225 return -EPERM;
2226
2227 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2228 return -EOPNOTSUPP;
2229
2230 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2231 sizeof(range)))
2232 return -EFAULT;
2233
2234 ret = mnt_want_write_file(filp);
2235 if (ret)
2236 return ret;
2237
2238 range.minlen = max((unsigned int)range.minlen,
2239 bdev_discard_granularity(sb->s_bdev));
2240 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2241 mnt_drop_write_file(filp);
2242 if (ret < 0)
2243 return ret;
2244
2245 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2246 sizeof(range)))
2247 return -EFAULT;
2248 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2249 return 0;
2250 }
2251
uuid_is_nonzero(__u8 u[16])2252 static bool uuid_is_nonzero(__u8 u[16])
2253 {
2254 int i;
2255
2256 for (i = 0; i < 16; i++)
2257 if (u[i])
2258 return true;
2259 return false;
2260 }
2261
f2fs_ioc_set_encryption_policy(struct file * filp,unsigned long arg)2262 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2263 {
2264 struct inode *inode = file_inode(filp);
2265
2266 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2267 return -EOPNOTSUPP;
2268
2269 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2270
2271 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2272 }
2273
f2fs_ioc_get_encryption_policy(struct file * filp,unsigned long arg)2274 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2275 {
2276 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2277 return -EOPNOTSUPP;
2278 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2279 }
2280
f2fs_ioc_get_encryption_pwsalt(struct file * filp,unsigned long arg)2281 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2282 {
2283 struct inode *inode = file_inode(filp);
2284 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2285 int err;
2286
2287 if (!f2fs_sb_has_encrypt(sbi))
2288 return -EOPNOTSUPP;
2289
2290 err = mnt_want_write_file(filp);
2291 if (err)
2292 return err;
2293
2294 f2fs_down_write(&sbi->sb_lock);
2295
2296 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2297 goto got_it;
2298
2299 /* update superblock with uuid */
2300 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2301
2302 err = f2fs_commit_super(sbi, false);
2303 if (err) {
2304 /* undo new data */
2305 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2306 goto out_err;
2307 }
2308 got_it:
2309 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2310 16))
2311 err = -EFAULT;
2312 out_err:
2313 f2fs_up_write(&sbi->sb_lock);
2314 mnt_drop_write_file(filp);
2315 return err;
2316 }
2317
f2fs_ioc_get_encryption_policy_ex(struct file * filp,unsigned long arg)2318 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2319 unsigned long arg)
2320 {
2321 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2322 return -EOPNOTSUPP;
2323
2324 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2325 }
2326
f2fs_ioc_add_encryption_key(struct file * filp,unsigned long arg)2327 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2328 {
2329 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2330 return -EOPNOTSUPP;
2331
2332 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2333 }
2334
f2fs_ioc_remove_encryption_key(struct file * filp,unsigned long arg)2335 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2336 {
2337 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2338 return -EOPNOTSUPP;
2339
2340 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2341 }
2342
f2fs_ioc_remove_encryption_key_all_users(struct file * filp,unsigned long arg)2343 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2344 unsigned long arg)
2345 {
2346 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2347 return -EOPNOTSUPP;
2348
2349 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2350 }
2351
f2fs_ioc_get_encryption_key_status(struct file * filp,unsigned long arg)2352 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2353 unsigned long arg)
2354 {
2355 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2356 return -EOPNOTSUPP;
2357
2358 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2359 }
2360
f2fs_ioc_get_encryption_nonce(struct file * filp,unsigned long arg)2361 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2362 {
2363 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2364 return -EOPNOTSUPP;
2365
2366 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2367 }
2368
f2fs_ioc_gc(struct file * filp,unsigned long arg)2369 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2370 {
2371 struct inode *inode = file_inode(filp);
2372 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2373 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2374 .no_bg_gc = false,
2375 .should_migrate_blocks = false,
2376 .nr_free_secs = 0 };
2377 __u32 sync;
2378 int ret;
2379
2380 if (!capable(CAP_SYS_ADMIN))
2381 return -EPERM;
2382
2383 if (get_user(sync, (__u32 __user *)arg))
2384 return -EFAULT;
2385
2386 if (f2fs_readonly(sbi->sb))
2387 return -EROFS;
2388
2389 ret = mnt_want_write_file(filp);
2390 if (ret)
2391 return ret;
2392
2393 if (!sync) {
2394 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2395 ret = -EBUSY;
2396 goto out;
2397 }
2398 } else {
2399 f2fs_down_write(&sbi->gc_lock);
2400 }
2401
2402 gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2403 gc_control.err_gc_skipped = sync;
2404 ret = f2fs_gc(sbi, &gc_control);
2405 out:
2406 mnt_drop_write_file(filp);
2407 return ret;
2408 }
2409
__f2fs_ioc_gc_range(struct file * filp,struct f2fs_gc_range * range)2410 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2411 {
2412 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2413 struct f2fs_gc_control gc_control = {
2414 .init_gc_type = range->sync ? FG_GC : BG_GC,
2415 .no_bg_gc = false,
2416 .should_migrate_blocks = false,
2417 .err_gc_skipped = range->sync,
2418 .nr_free_secs = 0 };
2419 u64 end;
2420 int ret;
2421
2422 if (!capable(CAP_SYS_ADMIN))
2423 return -EPERM;
2424 if (f2fs_readonly(sbi->sb))
2425 return -EROFS;
2426
2427 end = range->start + range->len;
2428 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2429 end >= MAX_BLKADDR(sbi))
2430 return -EINVAL;
2431
2432 ret = mnt_want_write_file(filp);
2433 if (ret)
2434 return ret;
2435
2436 do_more:
2437 if (!range->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.victim_segno = GET_SEGNO(sbi, range->start);
2447 ret = f2fs_gc(sbi, &gc_control);
2448 if (ret) {
2449 if (ret == -EBUSY)
2450 ret = -EAGAIN;
2451 goto out;
2452 }
2453 range->start += BLKS_PER_SEC(sbi);
2454 if (range->start <= end)
2455 goto do_more;
2456 out:
2457 mnt_drop_write_file(filp);
2458 return ret;
2459 }
2460
f2fs_ioc_gc_range(struct file * filp,unsigned long arg)2461 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2462 {
2463 struct f2fs_gc_range range;
2464
2465 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2466 sizeof(range)))
2467 return -EFAULT;
2468 return __f2fs_ioc_gc_range(filp, &range);
2469 }
2470
f2fs_ioc_write_checkpoint(struct file * filp,unsigned long arg)2471 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2472 {
2473 struct inode *inode = file_inode(filp);
2474 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2475 int ret;
2476
2477 if (!capable(CAP_SYS_ADMIN))
2478 return -EPERM;
2479
2480 if (f2fs_readonly(sbi->sb))
2481 return -EROFS;
2482
2483 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2484 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2485 return -EINVAL;
2486 }
2487
2488 ret = mnt_want_write_file(filp);
2489 if (ret)
2490 return ret;
2491
2492 ret = f2fs_sync_fs(sbi->sb, 1);
2493
2494 mnt_drop_write_file(filp);
2495 return ret;
2496 }
2497
f2fs_defragment_range(struct f2fs_sb_info * sbi,struct file * filp,struct f2fs_defragment * range)2498 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2499 struct file *filp,
2500 struct f2fs_defragment *range)
2501 {
2502 struct inode *inode = file_inode(filp);
2503 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2504 .m_seg_type = NO_CHECK_TYPE,
2505 .m_may_create = false };
2506 struct extent_info ei = {0, 0, 0};
2507 pgoff_t pg_start, pg_end, next_pgofs;
2508 unsigned int blk_per_seg = sbi->blocks_per_seg;
2509 unsigned int total = 0, sec_num;
2510 block_t blk_end = 0;
2511 bool fragmented = false;
2512 int err;
2513
2514 pg_start = range->start >> PAGE_SHIFT;
2515 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2516
2517 f2fs_balance_fs(sbi, true);
2518
2519 inode_lock(inode);
2520
2521 /* if in-place-update policy is enabled, don't waste time here */
2522 set_inode_flag(inode, FI_OPU_WRITE);
2523 if (f2fs_should_update_inplace(inode, NULL)) {
2524 err = -EINVAL;
2525 goto out;
2526 }
2527
2528 /* writeback all dirty pages in the range */
2529 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2530 range->start + range->len - 1);
2531 if (err)
2532 goto out;
2533
2534 /*
2535 * lookup mapping info in extent cache, skip defragmenting if physical
2536 * block addresses are continuous.
2537 */
2538 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2539 if (ei.fofs + ei.len >= pg_end)
2540 goto out;
2541 }
2542
2543 map.m_lblk = pg_start;
2544 map.m_next_pgofs = &next_pgofs;
2545
2546 /*
2547 * lookup mapping info in dnode page cache, skip defragmenting if all
2548 * physical block addresses are continuous even if there are hole(s)
2549 * in logical blocks.
2550 */
2551 while (map.m_lblk < pg_end) {
2552 map.m_len = pg_end - map.m_lblk;
2553 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2554 if (err)
2555 goto out;
2556
2557 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2558 map.m_lblk = next_pgofs;
2559 continue;
2560 }
2561
2562 if (blk_end && blk_end != map.m_pblk)
2563 fragmented = true;
2564
2565 /* record total count of block that we're going to move */
2566 total += map.m_len;
2567
2568 blk_end = map.m_pblk + map.m_len;
2569
2570 map.m_lblk += map.m_len;
2571 }
2572
2573 if (!fragmented) {
2574 total = 0;
2575 goto out;
2576 }
2577
2578 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2579
2580 /*
2581 * make sure there are enough free section for LFS allocation, this can
2582 * avoid defragment running in SSR mode when free section are allocated
2583 * intensively
2584 */
2585 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2586 err = -EAGAIN;
2587 goto out;
2588 }
2589
2590 map.m_lblk = pg_start;
2591 map.m_len = pg_end - pg_start;
2592 total = 0;
2593
2594 while (map.m_lblk < pg_end) {
2595 pgoff_t idx;
2596 int cnt = 0;
2597
2598 do_map:
2599 map.m_len = pg_end - map.m_lblk;
2600 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2601 if (err)
2602 goto clear_out;
2603
2604 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2605 map.m_lblk = next_pgofs;
2606 goto check;
2607 }
2608
2609 set_inode_flag(inode, FI_SKIP_WRITES);
2610
2611 idx = map.m_lblk;
2612 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2613 struct page *page;
2614
2615 page = f2fs_get_lock_data_page(inode, idx, true);
2616 if (IS_ERR(page)) {
2617 err = PTR_ERR(page);
2618 goto clear_out;
2619 }
2620
2621 set_page_dirty(page);
2622 set_page_private_gcing(page);
2623 f2fs_put_page(page, 1);
2624
2625 idx++;
2626 cnt++;
2627 total++;
2628 }
2629
2630 map.m_lblk = idx;
2631 check:
2632 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2633 goto do_map;
2634
2635 clear_inode_flag(inode, FI_SKIP_WRITES);
2636
2637 err = filemap_fdatawrite(inode->i_mapping);
2638 if (err)
2639 goto out;
2640 }
2641 clear_out:
2642 clear_inode_flag(inode, FI_SKIP_WRITES);
2643 out:
2644 clear_inode_flag(inode, FI_OPU_WRITE);
2645 inode_unlock(inode);
2646 if (!err)
2647 range->len = (u64)total << PAGE_SHIFT;
2648 return err;
2649 }
2650
f2fs_ioc_defragment(struct file * filp,unsigned long arg)2651 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2652 {
2653 struct inode *inode = file_inode(filp);
2654 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2655 struct f2fs_defragment range;
2656 int err;
2657
2658 if (!capable(CAP_SYS_ADMIN))
2659 return -EPERM;
2660
2661 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2662 return -EINVAL;
2663
2664 if (f2fs_readonly(sbi->sb))
2665 return -EROFS;
2666
2667 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2668 sizeof(range)))
2669 return -EFAULT;
2670
2671 /* verify alignment of offset & size */
2672 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2673 return -EINVAL;
2674
2675 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2676 max_file_blocks(inode)))
2677 return -EINVAL;
2678
2679 err = mnt_want_write_file(filp);
2680 if (err)
2681 return err;
2682
2683 err = f2fs_defragment_range(sbi, filp, &range);
2684 mnt_drop_write_file(filp);
2685
2686 f2fs_update_time(sbi, REQ_TIME);
2687 if (err < 0)
2688 return err;
2689
2690 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2691 sizeof(range)))
2692 return -EFAULT;
2693
2694 return 0;
2695 }
2696
f2fs_move_file_range(struct file * file_in,loff_t pos_in,struct file * file_out,loff_t pos_out,size_t len)2697 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2698 struct file *file_out, loff_t pos_out, size_t len)
2699 {
2700 struct inode *src = file_inode(file_in);
2701 struct inode *dst = file_inode(file_out);
2702 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2703 size_t olen = len, dst_max_i_size = 0;
2704 size_t dst_osize;
2705 int ret;
2706
2707 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2708 src->i_sb != dst->i_sb)
2709 return -EXDEV;
2710
2711 if (unlikely(f2fs_readonly(src->i_sb)))
2712 return -EROFS;
2713
2714 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2715 return -EINVAL;
2716
2717 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2718 return -EOPNOTSUPP;
2719
2720 if (pos_out < 0 || pos_in < 0)
2721 return -EINVAL;
2722
2723 if (src == dst) {
2724 if (pos_in == pos_out)
2725 return 0;
2726 if (pos_out > pos_in && pos_out < pos_in + len)
2727 return -EINVAL;
2728 }
2729
2730 inode_lock(src);
2731 if (src != dst) {
2732 ret = -EBUSY;
2733 if (!inode_trylock(dst))
2734 goto out;
2735 }
2736
2737 ret = -EINVAL;
2738 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2739 goto out_unlock;
2740 if (len == 0)
2741 olen = len = src->i_size - pos_in;
2742 if (pos_in + len == src->i_size)
2743 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2744 if (len == 0) {
2745 ret = 0;
2746 goto out_unlock;
2747 }
2748
2749 dst_osize = dst->i_size;
2750 if (pos_out + olen > dst->i_size)
2751 dst_max_i_size = pos_out + olen;
2752
2753 /* verify the end result is block aligned */
2754 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2755 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2756 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2757 goto out_unlock;
2758
2759 ret = f2fs_convert_inline_inode(src);
2760 if (ret)
2761 goto out_unlock;
2762
2763 ret = f2fs_convert_inline_inode(dst);
2764 if (ret)
2765 goto out_unlock;
2766
2767 /* write out all dirty pages from offset */
2768 ret = filemap_write_and_wait_range(src->i_mapping,
2769 pos_in, pos_in + len);
2770 if (ret)
2771 goto out_unlock;
2772
2773 ret = filemap_write_and_wait_range(dst->i_mapping,
2774 pos_out, pos_out + len);
2775 if (ret)
2776 goto out_unlock;
2777
2778 f2fs_balance_fs(sbi, true);
2779
2780 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2781 if (src != dst) {
2782 ret = -EBUSY;
2783 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2784 goto out_src;
2785 }
2786
2787 f2fs_lock_op(sbi);
2788 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2789 pos_out >> F2FS_BLKSIZE_BITS,
2790 len >> F2FS_BLKSIZE_BITS, false);
2791
2792 if (!ret) {
2793 if (dst_max_i_size)
2794 f2fs_i_size_write(dst, dst_max_i_size);
2795 else if (dst_osize != dst->i_size)
2796 f2fs_i_size_write(dst, dst_osize);
2797 }
2798 f2fs_unlock_op(sbi);
2799
2800 if (src != dst)
2801 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2802 out_src:
2803 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2804 out_unlock:
2805 if (src != dst)
2806 inode_unlock(dst);
2807 out:
2808 inode_unlock(src);
2809 return ret;
2810 }
2811
__f2fs_ioc_move_range(struct file * filp,struct f2fs_move_range * range)2812 static int __f2fs_ioc_move_range(struct file *filp,
2813 struct f2fs_move_range *range)
2814 {
2815 struct fd dst;
2816 int err;
2817
2818 if (!(filp->f_mode & FMODE_READ) ||
2819 !(filp->f_mode & FMODE_WRITE))
2820 return -EBADF;
2821
2822 dst = fdget(range->dst_fd);
2823 if (!dst.file)
2824 return -EBADF;
2825
2826 if (!(dst.file->f_mode & FMODE_WRITE)) {
2827 err = -EBADF;
2828 goto err_out;
2829 }
2830
2831 err = mnt_want_write_file(filp);
2832 if (err)
2833 goto err_out;
2834
2835 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2836 range->pos_out, range->len);
2837
2838 mnt_drop_write_file(filp);
2839 err_out:
2840 fdput(dst);
2841 return err;
2842 }
2843
f2fs_ioc_move_range(struct file * filp,unsigned long arg)2844 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2845 {
2846 struct f2fs_move_range range;
2847
2848 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2849 sizeof(range)))
2850 return -EFAULT;
2851 return __f2fs_ioc_move_range(filp, &range);
2852 }
2853
f2fs_ioc_flush_device(struct file * filp,unsigned long arg)2854 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2855 {
2856 struct inode *inode = file_inode(filp);
2857 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2858 struct sit_info *sm = SIT_I(sbi);
2859 unsigned int start_segno = 0, end_segno = 0;
2860 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2861 struct f2fs_flush_device range;
2862 struct f2fs_gc_control gc_control = {
2863 .init_gc_type = FG_GC,
2864 .should_migrate_blocks = true,
2865 .err_gc_skipped = true,
2866 .nr_free_secs = 0 };
2867 int ret;
2868
2869 if (!capable(CAP_SYS_ADMIN))
2870 return -EPERM;
2871
2872 if (f2fs_readonly(sbi->sb))
2873 return -EROFS;
2874
2875 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2876 return -EINVAL;
2877
2878 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2879 sizeof(range)))
2880 return -EFAULT;
2881
2882 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2883 __is_large_section(sbi)) {
2884 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2885 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2886 return -EINVAL;
2887 }
2888
2889 ret = mnt_want_write_file(filp);
2890 if (ret)
2891 return ret;
2892
2893 if (range.dev_num != 0)
2894 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2895 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2896
2897 start_segno = sm->last_victim[FLUSH_DEVICE];
2898 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2899 start_segno = dev_start_segno;
2900 end_segno = min(start_segno + range.segments, dev_end_segno);
2901
2902 while (start_segno < end_segno) {
2903 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2904 ret = -EBUSY;
2905 goto out;
2906 }
2907 sm->last_victim[GC_CB] = end_segno + 1;
2908 sm->last_victim[GC_GREEDY] = end_segno + 1;
2909 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2910
2911 gc_control.victim_segno = start_segno;
2912 ret = f2fs_gc(sbi, &gc_control);
2913 if (ret == -EAGAIN)
2914 ret = 0;
2915 else if (ret < 0)
2916 break;
2917 start_segno++;
2918 }
2919 out:
2920 mnt_drop_write_file(filp);
2921 return ret;
2922 }
2923
f2fs_ioc_get_features(struct file * filp,unsigned long arg)2924 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2925 {
2926 struct inode *inode = file_inode(filp);
2927 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2928
2929 /* Must validate to set it with SQLite behavior in Android. */
2930 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2931
2932 return put_user(sb_feature, (u32 __user *)arg);
2933 }
2934
2935 #ifdef CONFIG_QUOTA
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)2936 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2937 {
2938 struct dquot *transfer_to[MAXQUOTAS] = {};
2939 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2940 struct super_block *sb = sbi->sb;
2941 int err = 0;
2942
2943 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2944 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2945 err = __dquot_transfer(inode, transfer_to);
2946 if (err)
2947 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2948 dqput(transfer_to[PRJQUOTA]);
2949 }
2950 return err;
2951 }
2952
f2fs_ioc_setproject(struct inode * inode,__u32 projid)2953 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2954 {
2955 struct f2fs_inode_info *fi = F2FS_I(inode);
2956 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2957 struct f2fs_inode *ri = NULL;
2958 kprojid_t kprojid;
2959 int err;
2960
2961 if (!f2fs_sb_has_project_quota(sbi)) {
2962 if (projid != F2FS_DEF_PROJID)
2963 return -EOPNOTSUPP;
2964 else
2965 return 0;
2966 }
2967
2968 if (!f2fs_has_extra_attr(inode))
2969 return -EOPNOTSUPP;
2970
2971 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2972
2973 if (projid_eq(kprojid, fi->i_projid))
2974 return 0;
2975
2976 err = -EPERM;
2977 /* Is it quota file? Do not allow user to mess with it */
2978 if (IS_NOQUOTA(inode))
2979 return err;
2980
2981 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
2982 return -EOVERFLOW;
2983
2984 err = f2fs_dquot_initialize(inode);
2985 if (err)
2986 return err;
2987
2988 f2fs_lock_op(sbi);
2989 err = f2fs_transfer_project_quota(inode, kprojid);
2990 if (err)
2991 goto out_unlock;
2992
2993 fi->i_projid = kprojid;
2994 inode->i_ctime = current_time(inode);
2995 f2fs_mark_inode_dirty_sync(inode, true);
2996 out_unlock:
2997 f2fs_unlock_op(sbi);
2998 return err;
2999 }
3000 #else
f2fs_transfer_project_quota(struct inode * inode,kprojid_t kprojid)3001 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3002 {
3003 return 0;
3004 }
3005
f2fs_ioc_setproject(struct inode * inode,__u32 projid)3006 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3007 {
3008 if (projid != F2FS_DEF_PROJID)
3009 return -EOPNOTSUPP;
3010 return 0;
3011 }
3012 #endif
3013
f2fs_fileattr_get(struct dentry * dentry,struct fileattr * fa)3014 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3015 {
3016 struct inode *inode = d_inode(dentry);
3017 struct f2fs_inode_info *fi = F2FS_I(inode);
3018 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3019
3020 if (IS_ENCRYPTED(inode))
3021 fsflags |= FS_ENCRYPT_FL;
3022 if (IS_VERITY(inode))
3023 fsflags |= FS_VERITY_FL;
3024 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3025 fsflags |= FS_INLINE_DATA_FL;
3026 if (is_inode_flag_set(inode, FI_PIN_FILE))
3027 fsflags |= FS_NOCOW_FL;
3028
3029 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3030
3031 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3032 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3033
3034 return 0;
3035 }
3036
f2fs_fileattr_set(struct user_namespace * mnt_userns,struct dentry * dentry,struct fileattr * fa)3037 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3038 struct dentry *dentry, struct fileattr *fa)
3039 {
3040 struct inode *inode = d_inode(dentry);
3041 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3042 u32 iflags;
3043 int err;
3044
3045 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3046 return -EIO;
3047 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3048 return -ENOSPC;
3049 if (fsflags & ~F2FS_GETTABLE_FS_FL)
3050 return -EOPNOTSUPP;
3051 fsflags &= F2FS_SETTABLE_FS_FL;
3052 if (!fa->flags_valid)
3053 mask &= FS_COMMON_FL;
3054
3055 iflags = f2fs_fsflags_to_iflags(fsflags);
3056 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3057 return -EOPNOTSUPP;
3058
3059 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3060 if (!err)
3061 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3062
3063 return err;
3064 }
3065
f2fs_pin_file_control(struct inode * inode,bool inc)3066 int f2fs_pin_file_control(struct inode *inode, bool inc)
3067 {
3068 struct f2fs_inode_info *fi = F2FS_I(inode);
3069 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3070
3071 /* Use i_gc_failures for normal file as a risk signal. */
3072 if (inc)
3073 f2fs_i_gc_failures_write(inode,
3074 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3075
3076 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3077 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3078 __func__, inode->i_ino,
3079 fi->i_gc_failures[GC_FAILURE_PIN]);
3080 clear_inode_flag(inode, FI_PIN_FILE);
3081 return -EAGAIN;
3082 }
3083 return 0;
3084 }
3085
f2fs_ioc_set_pin_file(struct file * filp,unsigned long arg)3086 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3087 {
3088 struct inode *inode = file_inode(filp);
3089 __u32 pin;
3090 int ret = 0;
3091
3092 if (get_user(pin, (__u32 __user *)arg))
3093 return -EFAULT;
3094
3095 if (!S_ISREG(inode->i_mode))
3096 return -EINVAL;
3097
3098 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3099 return -EROFS;
3100
3101 ret = mnt_want_write_file(filp);
3102 if (ret)
3103 return ret;
3104
3105 inode_lock(inode);
3106
3107 if (!pin) {
3108 clear_inode_flag(inode, FI_PIN_FILE);
3109 f2fs_i_gc_failures_write(inode, 0);
3110 goto done;
3111 }
3112
3113 if (f2fs_should_update_outplace(inode, NULL)) {
3114 ret = -EINVAL;
3115 goto out;
3116 }
3117
3118 if (f2fs_pin_file_control(inode, false)) {
3119 ret = -EAGAIN;
3120 goto out;
3121 }
3122
3123 ret = f2fs_convert_inline_inode(inode);
3124 if (ret)
3125 goto out;
3126
3127 if (!f2fs_disable_compressed_file(inode)) {
3128 ret = -EOPNOTSUPP;
3129 goto out;
3130 }
3131
3132 set_inode_flag(inode, FI_PIN_FILE);
3133 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3134 done:
3135 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3136 out:
3137 inode_unlock(inode);
3138 mnt_drop_write_file(filp);
3139 return ret;
3140 }
3141
f2fs_ioc_get_pin_file(struct file * filp,unsigned long arg)3142 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3143 {
3144 struct inode *inode = file_inode(filp);
3145 __u32 pin = 0;
3146
3147 if (is_inode_flag_set(inode, FI_PIN_FILE))
3148 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3149 return put_user(pin, (u32 __user *)arg);
3150 }
3151
f2fs_precache_extents(struct inode * inode)3152 int f2fs_precache_extents(struct inode *inode)
3153 {
3154 struct f2fs_inode_info *fi = F2FS_I(inode);
3155 struct f2fs_map_blocks map;
3156 pgoff_t m_next_extent;
3157 loff_t end;
3158 int err;
3159
3160 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3161 return -EOPNOTSUPP;
3162
3163 map.m_lblk = 0;
3164 map.m_next_pgofs = NULL;
3165 map.m_next_extent = &m_next_extent;
3166 map.m_seg_type = NO_CHECK_TYPE;
3167 map.m_may_create = false;
3168 end = max_file_blocks(inode);
3169
3170 while (map.m_lblk < end) {
3171 map.m_len = end - map.m_lblk;
3172
3173 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3174 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3175 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3176 if (err)
3177 return err;
3178
3179 map.m_lblk = m_next_extent;
3180 }
3181
3182 return 0;
3183 }
3184
f2fs_ioc_precache_extents(struct file * filp,unsigned long arg)3185 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3186 {
3187 return f2fs_precache_extents(file_inode(filp));
3188 }
3189
f2fs_ioc_resize_fs(struct file * filp,unsigned long arg)3190 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3191 {
3192 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3193 __u64 block_count;
3194
3195 if (!capable(CAP_SYS_ADMIN))
3196 return -EPERM;
3197
3198 if (f2fs_readonly(sbi->sb))
3199 return -EROFS;
3200
3201 if (copy_from_user(&block_count, (void __user *)arg,
3202 sizeof(block_count)))
3203 return -EFAULT;
3204
3205 return f2fs_resize_fs(sbi, block_count);
3206 }
3207
f2fs_ioc_enable_verity(struct file * filp,unsigned long arg)3208 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3209 {
3210 struct inode *inode = file_inode(filp);
3211
3212 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3213
3214 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3215 f2fs_warn(F2FS_I_SB(inode),
3216 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3217 inode->i_ino);
3218 return -EOPNOTSUPP;
3219 }
3220
3221 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3222 }
3223
f2fs_ioc_measure_verity(struct file * filp,unsigned long arg)3224 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3225 {
3226 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3227 return -EOPNOTSUPP;
3228
3229 return fsverity_ioctl_measure(filp, (void __user *)arg);
3230 }
3231
f2fs_ioc_read_verity_metadata(struct file * filp,unsigned long arg)3232 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3233 {
3234 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3235 return -EOPNOTSUPP;
3236
3237 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3238 }
3239
f2fs_ioc_getfslabel(struct file * filp,unsigned long arg)3240 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3241 {
3242 struct inode *inode = file_inode(filp);
3243 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3244 char *vbuf;
3245 int count;
3246 int err = 0;
3247
3248 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3249 if (!vbuf)
3250 return -ENOMEM;
3251
3252 f2fs_down_read(&sbi->sb_lock);
3253 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3254 ARRAY_SIZE(sbi->raw_super->volume_name),
3255 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3256 f2fs_up_read(&sbi->sb_lock);
3257
3258 if (copy_to_user((char __user *)arg, vbuf,
3259 min(FSLABEL_MAX, count)))
3260 err = -EFAULT;
3261
3262 kfree(vbuf);
3263 return err;
3264 }
3265
f2fs_ioc_setfslabel(struct file * filp,unsigned long arg)3266 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3267 {
3268 struct inode *inode = file_inode(filp);
3269 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3270 char *vbuf;
3271 int err = 0;
3272
3273 if (!capable(CAP_SYS_ADMIN))
3274 return -EPERM;
3275
3276 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3277 if (IS_ERR(vbuf))
3278 return PTR_ERR(vbuf);
3279
3280 err = mnt_want_write_file(filp);
3281 if (err)
3282 goto out;
3283
3284 f2fs_down_write(&sbi->sb_lock);
3285
3286 memset(sbi->raw_super->volume_name, 0,
3287 sizeof(sbi->raw_super->volume_name));
3288 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3289 sbi->raw_super->volume_name,
3290 ARRAY_SIZE(sbi->raw_super->volume_name));
3291
3292 err = f2fs_commit_super(sbi, false);
3293
3294 f2fs_up_write(&sbi->sb_lock);
3295
3296 mnt_drop_write_file(filp);
3297 out:
3298 kfree(vbuf);
3299 return err;
3300 }
3301
f2fs_get_compress_blocks(struct file * filp,unsigned long arg)3302 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3303 {
3304 struct inode *inode = file_inode(filp);
3305 __u64 blocks;
3306
3307 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3308 return -EOPNOTSUPP;
3309
3310 if (!f2fs_compressed_file(inode))
3311 return -EINVAL;
3312
3313 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3314 return put_user(blocks, (u64 __user *)arg);
3315 }
3316
release_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3317 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3318 {
3319 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3320 unsigned int released_blocks = 0;
3321 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3322 block_t blkaddr;
3323 int i;
3324
3325 for (i = 0; i < count; i++) {
3326 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3327 dn->ofs_in_node + i);
3328
3329 if (!__is_valid_data_blkaddr(blkaddr))
3330 continue;
3331 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3332 DATA_GENERIC_ENHANCE)))
3333 return -EFSCORRUPTED;
3334 }
3335
3336 while (count) {
3337 int compr_blocks = 0;
3338
3339 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3340 blkaddr = f2fs_data_blkaddr(dn);
3341
3342 if (i == 0) {
3343 if (blkaddr == COMPRESS_ADDR)
3344 continue;
3345 dn->ofs_in_node += cluster_size;
3346 goto next;
3347 }
3348
3349 if (__is_valid_data_blkaddr(blkaddr))
3350 compr_blocks++;
3351
3352 if (blkaddr != NEW_ADDR)
3353 continue;
3354
3355 dn->data_blkaddr = NULL_ADDR;
3356 f2fs_set_data_blkaddr(dn);
3357 }
3358
3359 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3360 dec_valid_block_count(sbi, dn->inode,
3361 cluster_size - compr_blocks);
3362
3363 released_blocks += cluster_size - compr_blocks;
3364 next:
3365 count -= cluster_size;
3366 }
3367
3368 return released_blocks;
3369 }
3370
f2fs_release_compress_blocks(struct file * filp,unsigned long arg)3371 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3372 {
3373 struct inode *inode = file_inode(filp);
3374 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3375 pgoff_t page_idx = 0, last_idx;
3376 unsigned int released_blocks = 0;
3377 int ret;
3378 int writecount;
3379
3380 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3381 return -EOPNOTSUPP;
3382
3383 if (!f2fs_compressed_file(inode))
3384 return -EINVAL;
3385
3386 if (f2fs_readonly(sbi->sb))
3387 return -EROFS;
3388
3389 ret = mnt_want_write_file(filp);
3390 if (ret)
3391 return ret;
3392
3393 f2fs_balance_fs(F2FS_I_SB(inode), true);
3394
3395 inode_lock(inode);
3396
3397 writecount = atomic_read(&inode->i_writecount);
3398 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3399 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3400 ret = -EBUSY;
3401 goto out;
3402 }
3403
3404 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3405 ret = -EINVAL;
3406 goto out;
3407 }
3408
3409 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3410 if (ret)
3411 goto out;
3412
3413 set_inode_flag(inode, FI_COMPRESS_RELEASED);
3414 inode->i_ctime = current_time(inode);
3415 f2fs_mark_inode_dirty_sync(inode, true);
3416
3417 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3418 goto out;
3419
3420 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3421 filemap_invalidate_lock(inode->i_mapping);
3422
3423 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3424
3425 while (page_idx < last_idx) {
3426 struct dnode_of_data dn;
3427 pgoff_t end_offset, count;
3428
3429 set_new_dnode(&dn, inode, NULL, NULL, 0);
3430 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3431 if (ret) {
3432 if (ret == -ENOENT) {
3433 page_idx = f2fs_get_next_page_offset(&dn,
3434 page_idx);
3435 ret = 0;
3436 continue;
3437 }
3438 break;
3439 }
3440
3441 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3442 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3443 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3444
3445 ret = release_compress_blocks(&dn, count);
3446
3447 f2fs_put_dnode(&dn);
3448
3449 if (ret < 0)
3450 break;
3451
3452 page_idx += count;
3453 released_blocks += ret;
3454 }
3455
3456 filemap_invalidate_unlock(inode->i_mapping);
3457 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3458 out:
3459 inode_unlock(inode);
3460
3461 mnt_drop_write_file(filp);
3462
3463 if (ret >= 0) {
3464 ret = put_user(released_blocks, (u64 __user *)arg);
3465 } else if (released_blocks &&
3466 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3467 set_sbi_flag(sbi, SBI_NEED_FSCK);
3468 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3469 "iblocks=%llu, released=%u, compr_blocks=%u, "
3470 "run fsck to fix.",
3471 __func__, inode->i_ino, inode->i_blocks,
3472 released_blocks,
3473 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3474 }
3475
3476 return ret;
3477 }
3478
reserve_compress_blocks(struct dnode_of_data * dn,pgoff_t count)3479 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3480 {
3481 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3482 unsigned int reserved_blocks = 0;
3483 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3484 block_t blkaddr;
3485 int i;
3486
3487 for (i = 0; i < count; i++) {
3488 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3489 dn->ofs_in_node + i);
3490
3491 if (!__is_valid_data_blkaddr(blkaddr))
3492 continue;
3493 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3494 DATA_GENERIC_ENHANCE)))
3495 return -EFSCORRUPTED;
3496 }
3497
3498 while (count) {
3499 int compr_blocks = 0;
3500 blkcnt_t reserved;
3501 int ret;
3502
3503 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3504 blkaddr = f2fs_data_blkaddr(dn);
3505
3506 if (i == 0) {
3507 if (blkaddr == COMPRESS_ADDR)
3508 continue;
3509 dn->ofs_in_node += cluster_size;
3510 goto next;
3511 }
3512
3513 if (__is_valid_data_blkaddr(blkaddr)) {
3514 compr_blocks++;
3515 continue;
3516 }
3517
3518 dn->data_blkaddr = NEW_ADDR;
3519 f2fs_set_data_blkaddr(dn);
3520 }
3521
3522 reserved = cluster_size - compr_blocks;
3523 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3524 if (ret)
3525 return ret;
3526
3527 if (reserved != cluster_size - compr_blocks)
3528 return -ENOSPC;
3529
3530 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3531
3532 reserved_blocks += reserved;
3533 next:
3534 count -= cluster_size;
3535 }
3536
3537 return reserved_blocks;
3538 }
3539
f2fs_reserve_compress_blocks(struct file * filp,unsigned long arg)3540 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3541 {
3542 struct inode *inode = file_inode(filp);
3543 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3544 pgoff_t page_idx = 0, last_idx;
3545 unsigned int reserved_blocks = 0;
3546 int ret;
3547
3548 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3549 return -EOPNOTSUPP;
3550
3551 if (!f2fs_compressed_file(inode))
3552 return -EINVAL;
3553
3554 if (f2fs_readonly(sbi->sb))
3555 return -EROFS;
3556
3557 ret = mnt_want_write_file(filp);
3558 if (ret)
3559 return ret;
3560
3561 if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3562 goto out;
3563
3564 f2fs_balance_fs(F2FS_I_SB(inode), true);
3565
3566 inode_lock(inode);
3567
3568 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3569 ret = -EINVAL;
3570 goto unlock_inode;
3571 }
3572
3573 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3574 filemap_invalidate_lock(inode->i_mapping);
3575
3576 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3577
3578 while (page_idx < last_idx) {
3579 struct dnode_of_data dn;
3580 pgoff_t end_offset, count;
3581
3582 set_new_dnode(&dn, inode, NULL, NULL, 0);
3583 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3584 if (ret) {
3585 if (ret == -ENOENT) {
3586 page_idx = f2fs_get_next_page_offset(&dn,
3587 page_idx);
3588 ret = 0;
3589 continue;
3590 }
3591 break;
3592 }
3593
3594 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3595 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3596 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3597
3598 ret = reserve_compress_blocks(&dn, count);
3599
3600 f2fs_put_dnode(&dn);
3601
3602 if (ret < 0)
3603 break;
3604
3605 page_idx += count;
3606 reserved_blocks += ret;
3607 }
3608
3609 filemap_invalidate_unlock(inode->i_mapping);
3610 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3611
3612 if (ret >= 0) {
3613 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3614 inode->i_ctime = current_time(inode);
3615 f2fs_mark_inode_dirty_sync(inode, true);
3616 }
3617 unlock_inode:
3618 inode_unlock(inode);
3619 out:
3620 mnt_drop_write_file(filp);
3621
3622 if (ret >= 0) {
3623 ret = put_user(reserved_blocks, (u64 __user *)arg);
3624 } else if (reserved_blocks &&
3625 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3626 set_sbi_flag(sbi, SBI_NEED_FSCK);
3627 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3628 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3629 "run fsck to fix.",
3630 __func__, inode->i_ino, inode->i_blocks,
3631 reserved_blocks,
3632 atomic_read(&F2FS_I(inode)->i_compr_blocks));
3633 }
3634
3635 return ret;
3636 }
3637
f2fs_secure_erase(struct block_device * bdev,struct inode * inode,pgoff_t off,block_t block,block_t len,u32 flags)3638 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3639 pgoff_t off, block_t block, block_t len, u32 flags)
3640 {
3641 sector_t sector = SECTOR_FROM_BLOCK(block);
3642 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3643 int ret = 0;
3644
3645 if (flags & F2FS_TRIM_FILE_DISCARD) {
3646 if (bdev_max_secure_erase_sectors(bdev))
3647 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3648 GFP_NOFS);
3649 else
3650 ret = blkdev_issue_discard(bdev, sector, nr_sects,
3651 GFP_NOFS);
3652 }
3653
3654 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3655 if (IS_ENCRYPTED(inode))
3656 ret = fscrypt_zeroout_range(inode, off, block, len);
3657 else
3658 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3659 GFP_NOFS, 0);
3660 }
3661
3662 return ret;
3663 }
3664
f2fs_sec_trim_file(struct file * filp,unsigned long arg)3665 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3666 {
3667 struct inode *inode = file_inode(filp);
3668 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3669 struct address_space *mapping = inode->i_mapping;
3670 struct block_device *prev_bdev = NULL;
3671 struct f2fs_sectrim_range range;
3672 pgoff_t index, pg_end, prev_index = 0;
3673 block_t prev_block = 0, len = 0;
3674 loff_t end_addr;
3675 bool to_end = false;
3676 int ret = 0;
3677
3678 if (!(filp->f_mode & FMODE_WRITE))
3679 return -EBADF;
3680
3681 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3682 sizeof(range)))
3683 return -EFAULT;
3684
3685 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3686 !S_ISREG(inode->i_mode))
3687 return -EINVAL;
3688
3689 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3690 !f2fs_hw_support_discard(sbi)) ||
3691 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3692 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3693 return -EOPNOTSUPP;
3694
3695 file_start_write(filp);
3696 inode_lock(inode);
3697
3698 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3699 range.start >= inode->i_size) {
3700 ret = -EINVAL;
3701 goto err;
3702 }
3703
3704 if (range.len == 0)
3705 goto err;
3706
3707 if (inode->i_size - range.start > range.len) {
3708 end_addr = range.start + range.len;
3709 } else {
3710 end_addr = range.len == (u64)-1 ?
3711 sbi->sb->s_maxbytes : inode->i_size;
3712 to_end = true;
3713 }
3714
3715 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3716 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3717 ret = -EINVAL;
3718 goto err;
3719 }
3720
3721 index = F2FS_BYTES_TO_BLK(range.start);
3722 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3723
3724 ret = f2fs_convert_inline_inode(inode);
3725 if (ret)
3726 goto err;
3727
3728 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3729 filemap_invalidate_lock(mapping);
3730
3731 ret = filemap_write_and_wait_range(mapping, range.start,
3732 to_end ? LLONG_MAX : end_addr - 1);
3733 if (ret)
3734 goto out;
3735
3736 truncate_inode_pages_range(mapping, range.start,
3737 to_end ? -1 : end_addr - 1);
3738
3739 while (index < pg_end) {
3740 struct dnode_of_data dn;
3741 pgoff_t end_offset, count;
3742 int i;
3743
3744 set_new_dnode(&dn, inode, NULL, NULL, 0);
3745 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3746 if (ret) {
3747 if (ret == -ENOENT) {
3748 index = f2fs_get_next_page_offset(&dn, index);
3749 continue;
3750 }
3751 goto out;
3752 }
3753
3754 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3755 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3756 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3757 struct block_device *cur_bdev;
3758 block_t blkaddr = f2fs_data_blkaddr(&dn);
3759
3760 if (!__is_valid_data_blkaddr(blkaddr))
3761 continue;
3762
3763 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3764 DATA_GENERIC_ENHANCE)) {
3765 ret = -EFSCORRUPTED;
3766 f2fs_put_dnode(&dn);
3767 goto out;
3768 }
3769
3770 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3771 if (f2fs_is_multi_device(sbi)) {
3772 int di = f2fs_target_device_index(sbi, blkaddr);
3773
3774 blkaddr -= FDEV(di).start_blk;
3775 }
3776
3777 if (len) {
3778 if (prev_bdev == cur_bdev &&
3779 index == prev_index + len &&
3780 blkaddr == prev_block + len) {
3781 len++;
3782 } else {
3783 ret = f2fs_secure_erase(prev_bdev,
3784 inode, prev_index, prev_block,
3785 len, range.flags);
3786 if (ret) {
3787 f2fs_put_dnode(&dn);
3788 goto out;
3789 }
3790
3791 len = 0;
3792 }
3793 }
3794
3795 if (!len) {
3796 prev_bdev = cur_bdev;
3797 prev_index = index;
3798 prev_block = blkaddr;
3799 len = 1;
3800 }
3801 }
3802
3803 f2fs_put_dnode(&dn);
3804
3805 if (fatal_signal_pending(current)) {
3806 ret = -EINTR;
3807 goto out;
3808 }
3809 cond_resched();
3810 }
3811
3812 if (len)
3813 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3814 prev_block, len, range.flags);
3815 out:
3816 filemap_invalidate_unlock(mapping);
3817 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3818 err:
3819 inode_unlock(inode);
3820 file_end_write(filp);
3821
3822 return ret;
3823 }
3824
f2fs_ioc_get_compress_option(struct file * filp,unsigned long arg)3825 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3826 {
3827 struct inode *inode = file_inode(filp);
3828 struct f2fs_comp_option option;
3829
3830 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3831 return -EOPNOTSUPP;
3832
3833 inode_lock_shared(inode);
3834
3835 if (!f2fs_compressed_file(inode)) {
3836 inode_unlock_shared(inode);
3837 return -ENODATA;
3838 }
3839
3840 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3841 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3842
3843 inode_unlock_shared(inode);
3844
3845 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3846 sizeof(option)))
3847 return -EFAULT;
3848
3849 return 0;
3850 }
3851
f2fs_ioc_set_compress_option(struct file * filp,unsigned long arg)3852 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3853 {
3854 struct inode *inode = file_inode(filp);
3855 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3856 struct f2fs_comp_option option;
3857 int ret = 0;
3858
3859 if (!f2fs_sb_has_compression(sbi))
3860 return -EOPNOTSUPP;
3861
3862 if (!(filp->f_mode & FMODE_WRITE))
3863 return -EBADF;
3864
3865 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3866 sizeof(option)))
3867 return -EFAULT;
3868
3869 if (!f2fs_compressed_file(inode) ||
3870 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3871 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3872 option.algorithm >= COMPRESS_MAX)
3873 return -EINVAL;
3874
3875 file_start_write(filp);
3876 inode_lock(inode);
3877
3878 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3879 ret = -EBUSY;
3880 goto out;
3881 }
3882
3883 if (inode->i_size != 0) {
3884 ret = -EFBIG;
3885 goto out;
3886 }
3887
3888 F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3889 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3890 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3891 f2fs_mark_inode_dirty_sync(inode, true);
3892
3893 if (!f2fs_is_compress_backend_ready(inode))
3894 f2fs_warn(sbi, "compression algorithm is successfully set, "
3895 "but current kernel doesn't support this algorithm.");
3896 out:
3897 inode_unlock(inode);
3898 file_end_write(filp);
3899
3900 return ret;
3901 }
3902
redirty_blocks(struct inode * inode,pgoff_t page_idx,int len)3903 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3904 {
3905 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3906 struct address_space *mapping = inode->i_mapping;
3907 struct page *page;
3908 pgoff_t redirty_idx = page_idx;
3909 int i, page_len = 0, ret = 0;
3910
3911 page_cache_ra_unbounded(&ractl, len, 0);
3912
3913 for (i = 0; i < len; i++, page_idx++) {
3914 page = read_cache_page(mapping, page_idx, NULL, NULL);
3915 if (IS_ERR(page)) {
3916 ret = PTR_ERR(page);
3917 break;
3918 }
3919 page_len++;
3920 }
3921
3922 for (i = 0; i < page_len; i++, redirty_idx++) {
3923 page = find_lock_page(mapping, redirty_idx);
3924 if (!page) {
3925 ret = -ENOMEM;
3926 break;
3927 }
3928 set_page_dirty(page);
3929 f2fs_put_page(page, 1);
3930 f2fs_put_page(page, 0);
3931 }
3932
3933 return ret;
3934 }
3935
f2fs_ioc_decompress_file(struct file * filp,unsigned long arg)3936 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3937 {
3938 struct inode *inode = file_inode(filp);
3939 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3940 struct f2fs_inode_info *fi = F2FS_I(inode);
3941 pgoff_t page_idx = 0, last_idx;
3942 unsigned int blk_per_seg = sbi->blocks_per_seg;
3943 int cluster_size = fi->i_cluster_size;
3944 int count, ret;
3945
3946 if (!f2fs_sb_has_compression(sbi) ||
3947 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3948 return -EOPNOTSUPP;
3949
3950 if (!(filp->f_mode & FMODE_WRITE))
3951 return -EBADF;
3952
3953 if (!f2fs_compressed_file(inode))
3954 return -EINVAL;
3955
3956 f2fs_balance_fs(F2FS_I_SB(inode), true);
3957
3958 file_start_write(filp);
3959 inode_lock(inode);
3960
3961 if (!f2fs_is_compress_backend_ready(inode)) {
3962 ret = -EOPNOTSUPP;
3963 goto out;
3964 }
3965
3966 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3967 ret = -EINVAL;
3968 goto out;
3969 }
3970
3971 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3972 if (ret)
3973 goto out;
3974
3975 if (!atomic_read(&fi->i_compr_blocks))
3976 goto out;
3977
3978 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3979
3980 count = last_idx - page_idx;
3981 while (count) {
3982 int len = min(cluster_size, count);
3983
3984 ret = redirty_blocks(inode, page_idx, len);
3985 if (ret < 0)
3986 break;
3987
3988 if (get_dirty_pages(inode) >= blk_per_seg)
3989 filemap_fdatawrite(inode->i_mapping);
3990
3991 count -= len;
3992 page_idx += len;
3993 }
3994
3995 if (!ret)
3996 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
3997 LLONG_MAX);
3998
3999 if (ret)
4000 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4001 __func__, ret);
4002 out:
4003 inode_unlock(inode);
4004 file_end_write(filp);
4005
4006 return ret;
4007 }
4008
f2fs_ioc_compress_file(struct file * filp,unsigned long arg)4009 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4010 {
4011 struct inode *inode = file_inode(filp);
4012 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4013 pgoff_t page_idx = 0, last_idx;
4014 unsigned int blk_per_seg = sbi->blocks_per_seg;
4015 int cluster_size = F2FS_I(inode)->i_cluster_size;
4016 int count, ret;
4017
4018 if (!f2fs_sb_has_compression(sbi) ||
4019 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4020 return -EOPNOTSUPP;
4021
4022 if (!(filp->f_mode & FMODE_WRITE))
4023 return -EBADF;
4024
4025 if (!f2fs_compressed_file(inode))
4026 return -EINVAL;
4027
4028 f2fs_balance_fs(F2FS_I_SB(inode), true);
4029
4030 file_start_write(filp);
4031 inode_lock(inode);
4032
4033 if (!f2fs_is_compress_backend_ready(inode)) {
4034 ret = -EOPNOTSUPP;
4035 goto out;
4036 }
4037
4038 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4039 ret = -EINVAL;
4040 goto out;
4041 }
4042
4043 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4044 if (ret)
4045 goto out;
4046
4047 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4048
4049 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4050
4051 count = last_idx - page_idx;
4052 while (count) {
4053 int len = min(cluster_size, count);
4054
4055 ret = redirty_blocks(inode, page_idx, len);
4056 if (ret < 0)
4057 break;
4058
4059 if (get_dirty_pages(inode) >= blk_per_seg)
4060 filemap_fdatawrite(inode->i_mapping);
4061
4062 count -= len;
4063 page_idx += len;
4064 }
4065
4066 if (!ret)
4067 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4068 LLONG_MAX);
4069
4070 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4071
4072 if (ret)
4073 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4074 __func__, ret);
4075 out:
4076 inode_unlock(inode);
4077 file_end_write(filp);
4078
4079 return ret;
4080 }
4081
__f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4082 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4083 {
4084 switch (cmd) {
4085 case FS_IOC_GETVERSION:
4086 return f2fs_ioc_getversion(filp, arg);
4087 case F2FS_IOC_START_ATOMIC_WRITE:
4088 return f2fs_ioc_start_atomic_write(filp);
4089 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4090 return f2fs_ioc_commit_atomic_write(filp);
4091 case F2FS_IOC_ABORT_ATOMIC_WRITE:
4092 return f2fs_ioc_abort_atomic_write(filp);
4093 case F2FS_IOC_START_VOLATILE_WRITE:
4094 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4095 return -EOPNOTSUPP;
4096 case F2FS_IOC_SHUTDOWN:
4097 return f2fs_ioc_shutdown(filp, arg);
4098 case FITRIM:
4099 return f2fs_ioc_fitrim(filp, arg);
4100 case FS_IOC_SET_ENCRYPTION_POLICY:
4101 return f2fs_ioc_set_encryption_policy(filp, arg);
4102 case FS_IOC_GET_ENCRYPTION_POLICY:
4103 return f2fs_ioc_get_encryption_policy(filp, arg);
4104 case FS_IOC_GET_ENCRYPTION_PWSALT:
4105 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4106 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4107 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4108 case FS_IOC_ADD_ENCRYPTION_KEY:
4109 return f2fs_ioc_add_encryption_key(filp, arg);
4110 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4111 return f2fs_ioc_remove_encryption_key(filp, arg);
4112 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4113 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4114 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4115 return f2fs_ioc_get_encryption_key_status(filp, arg);
4116 case FS_IOC_GET_ENCRYPTION_NONCE:
4117 return f2fs_ioc_get_encryption_nonce(filp, arg);
4118 case F2FS_IOC_GARBAGE_COLLECT:
4119 return f2fs_ioc_gc(filp, arg);
4120 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4121 return f2fs_ioc_gc_range(filp, arg);
4122 case F2FS_IOC_WRITE_CHECKPOINT:
4123 return f2fs_ioc_write_checkpoint(filp, arg);
4124 case F2FS_IOC_DEFRAGMENT:
4125 return f2fs_ioc_defragment(filp, arg);
4126 case F2FS_IOC_MOVE_RANGE:
4127 return f2fs_ioc_move_range(filp, arg);
4128 case F2FS_IOC_FLUSH_DEVICE:
4129 return f2fs_ioc_flush_device(filp, arg);
4130 case F2FS_IOC_GET_FEATURES:
4131 return f2fs_ioc_get_features(filp, arg);
4132 case F2FS_IOC_GET_PIN_FILE:
4133 return f2fs_ioc_get_pin_file(filp, arg);
4134 case F2FS_IOC_SET_PIN_FILE:
4135 return f2fs_ioc_set_pin_file(filp, arg);
4136 case F2FS_IOC_PRECACHE_EXTENTS:
4137 return f2fs_ioc_precache_extents(filp, arg);
4138 case F2FS_IOC_RESIZE_FS:
4139 return f2fs_ioc_resize_fs(filp, arg);
4140 case FS_IOC_ENABLE_VERITY:
4141 return f2fs_ioc_enable_verity(filp, arg);
4142 case FS_IOC_MEASURE_VERITY:
4143 return f2fs_ioc_measure_verity(filp, arg);
4144 case FS_IOC_READ_VERITY_METADATA:
4145 return f2fs_ioc_read_verity_metadata(filp, arg);
4146 case FS_IOC_GETFSLABEL:
4147 return f2fs_ioc_getfslabel(filp, arg);
4148 case FS_IOC_SETFSLABEL:
4149 return f2fs_ioc_setfslabel(filp, arg);
4150 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4151 return f2fs_get_compress_blocks(filp, arg);
4152 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4153 return f2fs_release_compress_blocks(filp, arg);
4154 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4155 return f2fs_reserve_compress_blocks(filp, arg);
4156 case F2FS_IOC_SEC_TRIM_FILE:
4157 return f2fs_sec_trim_file(filp, arg);
4158 case F2FS_IOC_GET_COMPRESS_OPTION:
4159 return f2fs_ioc_get_compress_option(filp, arg);
4160 case F2FS_IOC_SET_COMPRESS_OPTION:
4161 return f2fs_ioc_set_compress_option(filp, arg);
4162 case F2FS_IOC_DECOMPRESS_FILE:
4163 return f2fs_ioc_decompress_file(filp, arg);
4164 case F2FS_IOC_COMPRESS_FILE:
4165 return f2fs_ioc_compress_file(filp, arg);
4166 default:
4167 return -ENOTTY;
4168 }
4169 }
4170
f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)4171 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4172 {
4173 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4174 return -EIO;
4175 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4176 return -ENOSPC;
4177
4178 return __f2fs_ioctl(filp, cmd, arg);
4179 }
4180
4181 /*
4182 * Return %true if the given read or write request should use direct I/O, or
4183 * %false if it should use buffered I/O.
4184 */
f2fs_should_use_dio(struct inode * inode,struct kiocb * iocb,struct iov_iter * iter)4185 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4186 struct iov_iter *iter)
4187 {
4188 unsigned int align;
4189
4190 if (!(iocb->ki_flags & IOCB_DIRECT))
4191 return false;
4192
4193 if (f2fs_force_buffered_io(inode, iocb, iter))
4194 return false;
4195
4196 /*
4197 * Direct I/O not aligned to the disk's logical_block_size will be
4198 * attempted, but will fail with -EINVAL.
4199 *
4200 * f2fs additionally requires that direct I/O be aligned to the
4201 * filesystem block size, which is often a stricter requirement.
4202 * However, f2fs traditionally falls back to buffered I/O on requests
4203 * that are logical_block_size-aligned but not fs-block aligned.
4204 *
4205 * The below logic implements this behavior.
4206 */
4207 align = iocb->ki_pos | iov_iter_alignment(iter);
4208 if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4209 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4210 return false;
4211
4212 return true;
4213 }
4214
f2fs_dio_read_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4215 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4216 unsigned int flags)
4217 {
4218 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4219
4220 dec_page_count(sbi, F2FS_DIO_READ);
4221 if (error)
4222 return error;
4223 f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size);
4224 return 0;
4225 }
4226
4227 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4228 .end_io = f2fs_dio_read_end_io,
4229 };
4230
f2fs_dio_read_iter(struct kiocb * iocb,struct iov_iter * to)4231 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4232 {
4233 struct file *file = iocb->ki_filp;
4234 struct inode *inode = file_inode(file);
4235 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4236 struct f2fs_inode_info *fi = F2FS_I(inode);
4237 const loff_t pos = iocb->ki_pos;
4238 const size_t count = iov_iter_count(to);
4239 struct iomap_dio *dio;
4240 ssize_t ret;
4241
4242 if (count == 0)
4243 return 0; /* skip atime update */
4244
4245 trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4246
4247 if (iocb->ki_flags & IOCB_NOWAIT) {
4248 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4249 ret = -EAGAIN;
4250 goto out;
4251 }
4252 } else {
4253 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4254 }
4255
4256 /*
4257 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4258 * the higher-level function iomap_dio_rw() in order to ensure that the
4259 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4260 */
4261 inc_page_count(sbi, F2FS_DIO_READ);
4262 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4263 &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4264 if (IS_ERR_OR_NULL(dio)) {
4265 ret = PTR_ERR_OR_ZERO(dio);
4266 if (ret != -EIOCBQUEUED)
4267 dec_page_count(sbi, F2FS_DIO_READ);
4268 } else {
4269 ret = iomap_dio_complete(dio);
4270 }
4271
4272 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4273
4274 file_accessed(file);
4275 out:
4276 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4277 return ret;
4278 }
4279
f2fs_file_read_iter(struct kiocb * iocb,struct iov_iter * to)4280 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4281 {
4282 struct inode *inode = file_inode(iocb->ki_filp);
4283 const loff_t pos = iocb->ki_pos;
4284 ssize_t ret;
4285
4286 if (!f2fs_is_compress_backend_ready(inode))
4287 return -EOPNOTSUPP;
4288
4289 if (trace_f2fs_dataread_start_enabled()) {
4290 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4291 char *path;
4292
4293 if (!p)
4294 goto skip_read_trace;
4295
4296 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4297 if (IS_ERR(path)) {
4298 kfree(p);
4299 goto skip_read_trace;
4300 }
4301
4302 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4303 current->pid, path, current->comm);
4304 kfree(p);
4305 }
4306 skip_read_trace:
4307 if (f2fs_should_use_dio(inode, iocb, to)) {
4308 ret = f2fs_dio_read_iter(iocb, to);
4309 } else {
4310 ret = filemap_read(iocb, to, 0);
4311 if (ret > 0)
4312 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret);
4313 }
4314 if (trace_f2fs_dataread_end_enabled())
4315 trace_f2fs_dataread_end(inode, pos, ret);
4316 return ret;
4317 }
4318
f2fs_write_checks(struct kiocb * iocb,struct iov_iter * from)4319 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4320 {
4321 struct file *file = iocb->ki_filp;
4322 struct inode *inode = file_inode(file);
4323 ssize_t count;
4324 int err;
4325
4326 if (IS_IMMUTABLE(inode))
4327 return -EPERM;
4328
4329 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4330 return -EPERM;
4331
4332 count = generic_write_checks(iocb, from);
4333 if (count <= 0)
4334 return count;
4335
4336 err = file_modified(file);
4337 if (err)
4338 return err;
4339 return count;
4340 }
4341
4342 /*
4343 * Preallocate blocks for a write request, if it is possible and helpful to do
4344 * so. Returns a positive number if blocks may have been preallocated, 0 if no
4345 * blocks were preallocated, or a negative errno value if something went
4346 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4347 * requested blocks (not just some of them) have been allocated.
4348 */
f2fs_preallocate_blocks(struct kiocb * iocb,struct iov_iter * iter,bool dio)4349 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4350 bool dio)
4351 {
4352 struct inode *inode = file_inode(iocb->ki_filp);
4353 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4354 const loff_t pos = iocb->ki_pos;
4355 const size_t count = iov_iter_count(iter);
4356 struct f2fs_map_blocks map = {};
4357 int flag;
4358 int ret;
4359
4360 /* If it will be an out-of-place direct write, don't bother. */
4361 if (dio && f2fs_lfs_mode(sbi))
4362 return 0;
4363 /*
4364 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4365 * buffered IO, if DIO meets any holes.
4366 */
4367 if (dio && i_size_read(inode) &&
4368 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4369 return 0;
4370
4371 /* No-wait I/O can't allocate blocks. */
4372 if (iocb->ki_flags & IOCB_NOWAIT)
4373 return 0;
4374
4375 /* If it will be a short write, don't bother. */
4376 if (fault_in_iov_iter_readable(iter, count))
4377 return 0;
4378
4379 if (f2fs_has_inline_data(inode)) {
4380 /* If the data will fit inline, don't bother. */
4381 if (pos + count <= MAX_INLINE_DATA(inode))
4382 return 0;
4383 ret = f2fs_convert_inline_inode(inode);
4384 if (ret)
4385 return ret;
4386 }
4387
4388 /* Do not preallocate blocks that will be written partially in 4KB. */
4389 map.m_lblk = F2FS_BLK_ALIGN(pos);
4390 map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4391 if (map.m_len > map.m_lblk)
4392 map.m_len -= map.m_lblk;
4393 else
4394 map.m_len = 0;
4395 map.m_may_create = true;
4396 if (dio) {
4397 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4398 flag = F2FS_GET_BLOCK_PRE_DIO;
4399 } else {
4400 map.m_seg_type = NO_CHECK_TYPE;
4401 flag = F2FS_GET_BLOCK_PRE_AIO;
4402 }
4403
4404 ret = f2fs_map_blocks(inode, &map, 1, flag);
4405 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4406 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4407 return ret;
4408 if (ret == 0)
4409 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4410 return map.m_len;
4411 }
4412
f2fs_buffered_write_iter(struct kiocb * iocb,struct iov_iter * from)4413 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4414 struct iov_iter *from)
4415 {
4416 struct file *file = iocb->ki_filp;
4417 struct inode *inode = file_inode(file);
4418 ssize_t ret;
4419
4420 if (iocb->ki_flags & IOCB_NOWAIT)
4421 return -EOPNOTSUPP;
4422
4423 current->backing_dev_info = inode_to_bdi(inode);
4424 ret = generic_perform_write(iocb, from);
4425 current->backing_dev_info = NULL;
4426
4427 if (ret > 0) {
4428 iocb->ki_pos += ret;
4429 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret);
4430 }
4431 return ret;
4432 }
4433
f2fs_dio_write_end_io(struct kiocb * iocb,ssize_t size,int error,unsigned int flags)4434 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4435 unsigned int flags)
4436 {
4437 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4438
4439 dec_page_count(sbi, F2FS_DIO_WRITE);
4440 if (error)
4441 return error;
4442 f2fs_update_iostat(sbi, APP_DIRECT_IO, size);
4443 return 0;
4444 }
4445
4446 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4447 .end_io = f2fs_dio_write_end_io,
4448 };
4449
f2fs_dio_write_iter(struct kiocb * iocb,struct iov_iter * from,bool * may_need_sync)4450 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4451 bool *may_need_sync)
4452 {
4453 struct file *file = iocb->ki_filp;
4454 struct inode *inode = file_inode(file);
4455 struct f2fs_inode_info *fi = F2FS_I(inode);
4456 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4457 const bool do_opu = f2fs_lfs_mode(sbi);
4458 const loff_t pos = iocb->ki_pos;
4459 const ssize_t count = iov_iter_count(from);
4460 unsigned int dio_flags;
4461 struct iomap_dio *dio;
4462 ssize_t ret;
4463
4464 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4465
4466 if (iocb->ki_flags & IOCB_NOWAIT) {
4467 /* f2fs_convert_inline_inode() and block allocation can block */
4468 if (f2fs_has_inline_data(inode) ||
4469 !f2fs_overwrite_io(inode, pos, count)) {
4470 ret = -EAGAIN;
4471 goto out;
4472 }
4473
4474 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4475 ret = -EAGAIN;
4476 goto out;
4477 }
4478 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4479 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4480 ret = -EAGAIN;
4481 goto out;
4482 }
4483 } else {
4484 ret = f2fs_convert_inline_inode(inode);
4485 if (ret)
4486 goto out;
4487
4488 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4489 if (do_opu)
4490 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4491 }
4492
4493 /*
4494 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4495 * the higher-level function iomap_dio_rw() in order to ensure that the
4496 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4497 */
4498 inc_page_count(sbi, F2FS_DIO_WRITE);
4499 dio_flags = 0;
4500 if (pos + count > inode->i_size)
4501 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4502 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4503 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4504 if (IS_ERR_OR_NULL(dio)) {
4505 ret = PTR_ERR_OR_ZERO(dio);
4506 if (ret == -ENOTBLK)
4507 ret = 0;
4508 if (ret != -EIOCBQUEUED)
4509 dec_page_count(sbi, F2FS_DIO_WRITE);
4510 } else {
4511 ret = iomap_dio_complete(dio);
4512 }
4513
4514 if (do_opu)
4515 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4516 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4517
4518 if (ret < 0)
4519 goto out;
4520 if (pos + ret > inode->i_size)
4521 f2fs_i_size_write(inode, pos + ret);
4522 if (!do_opu)
4523 set_inode_flag(inode, FI_UPDATE_WRITE);
4524
4525 if (iov_iter_count(from)) {
4526 ssize_t ret2;
4527 loff_t bufio_start_pos = iocb->ki_pos;
4528
4529 /*
4530 * The direct write was partial, so we need to fall back to a
4531 * buffered write for the remainder.
4532 */
4533
4534 ret2 = f2fs_buffered_write_iter(iocb, from);
4535 if (iov_iter_count(from))
4536 f2fs_write_failed(inode, iocb->ki_pos);
4537 if (ret2 < 0)
4538 goto out;
4539
4540 /*
4541 * Ensure that the pagecache pages are written to disk and
4542 * invalidated to preserve the expected O_DIRECT semantics.
4543 */
4544 if (ret2 > 0) {
4545 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4546
4547 ret += ret2;
4548
4549 ret2 = filemap_write_and_wait_range(file->f_mapping,
4550 bufio_start_pos,
4551 bufio_end_pos);
4552 if (ret2 < 0)
4553 goto out;
4554 invalidate_mapping_pages(file->f_mapping,
4555 bufio_start_pos >> PAGE_SHIFT,
4556 bufio_end_pos >> PAGE_SHIFT);
4557 }
4558 } else {
4559 /* iomap_dio_rw() already handled the generic_write_sync(). */
4560 *may_need_sync = false;
4561 }
4562 out:
4563 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4564 return ret;
4565 }
4566
f2fs_file_write_iter(struct kiocb * iocb,struct iov_iter * from)4567 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4568 {
4569 struct inode *inode = file_inode(iocb->ki_filp);
4570 const loff_t orig_pos = iocb->ki_pos;
4571 const size_t orig_count = iov_iter_count(from);
4572 loff_t target_size;
4573 bool dio;
4574 bool may_need_sync = true;
4575 int preallocated;
4576 ssize_t ret;
4577
4578 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4579 ret = -EIO;
4580 goto out;
4581 }
4582
4583 if (!f2fs_is_compress_backend_ready(inode)) {
4584 ret = -EOPNOTSUPP;
4585 goto out;
4586 }
4587
4588 if (iocb->ki_flags & IOCB_NOWAIT) {
4589 if (!inode_trylock(inode)) {
4590 ret = -EAGAIN;
4591 goto out;
4592 }
4593 } else {
4594 inode_lock(inode);
4595 }
4596
4597 ret = f2fs_write_checks(iocb, from);
4598 if (ret <= 0)
4599 goto out_unlock;
4600
4601 /* Determine whether we will do a direct write or a buffered write. */
4602 dio = f2fs_should_use_dio(inode, iocb, from);
4603
4604 /* Possibly preallocate the blocks for the write. */
4605 target_size = iocb->ki_pos + iov_iter_count(from);
4606 preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4607 if (preallocated < 0) {
4608 ret = preallocated;
4609 } else {
4610 if (trace_f2fs_datawrite_start_enabled()) {
4611 char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4612 PATH_MAX, GFP_KERNEL);
4613 char *path;
4614
4615 if (!p)
4616 goto skip_write_trace;
4617 path = dentry_path_raw(file_dentry(iocb->ki_filp),
4618 p, PATH_MAX);
4619 if (IS_ERR(path)) {
4620 kfree(p);
4621 goto skip_write_trace;
4622 }
4623 trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4624 current->pid, path, current->comm);
4625 kfree(p);
4626 }
4627 skip_write_trace:
4628 /* Do the actual write. */
4629 ret = dio ?
4630 f2fs_dio_write_iter(iocb, from, &may_need_sync):
4631 f2fs_buffered_write_iter(iocb, from);
4632
4633 if (trace_f2fs_datawrite_end_enabled())
4634 trace_f2fs_datawrite_end(inode, orig_pos, ret);
4635 }
4636
4637 /* Don't leave any preallocated blocks around past i_size. */
4638 if (preallocated && i_size_read(inode) < target_size) {
4639 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4640 filemap_invalidate_lock(inode->i_mapping);
4641 if (!f2fs_truncate(inode))
4642 file_dont_truncate(inode);
4643 filemap_invalidate_unlock(inode->i_mapping);
4644 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4645 } else {
4646 file_dont_truncate(inode);
4647 }
4648
4649 clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4650 out_unlock:
4651 inode_unlock(inode);
4652 out:
4653 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4654 if (ret > 0 && may_need_sync)
4655 ret = generic_write_sync(iocb, ret);
4656 return ret;
4657 }
4658
f2fs_file_fadvise(struct file * filp,loff_t offset,loff_t len,int advice)4659 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4660 int advice)
4661 {
4662 struct address_space *mapping;
4663 struct backing_dev_info *bdi;
4664 struct inode *inode = file_inode(filp);
4665 int err;
4666
4667 if (advice == POSIX_FADV_SEQUENTIAL) {
4668 if (S_ISFIFO(inode->i_mode))
4669 return -ESPIPE;
4670
4671 mapping = filp->f_mapping;
4672 if (!mapping || len < 0)
4673 return -EINVAL;
4674
4675 bdi = inode_to_bdi(mapping->host);
4676 filp->f_ra.ra_pages = bdi->ra_pages *
4677 F2FS_I_SB(inode)->seq_file_ra_mul;
4678 spin_lock(&filp->f_lock);
4679 filp->f_mode &= ~FMODE_RANDOM;
4680 spin_unlock(&filp->f_lock);
4681 return 0;
4682 }
4683
4684 err = generic_fadvise(filp, offset, len, advice);
4685 if (!err && advice == POSIX_FADV_DONTNEED &&
4686 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4687 f2fs_compressed_file(inode))
4688 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4689
4690 return err;
4691 }
4692
4693 #ifdef CONFIG_COMPAT
4694 struct compat_f2fs_gc_range {
4695 u32 sync;
4696 compat_u64 start;
4697 compat_u64 len;
4698 };
4699 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
4700 struct compat_f2fs_gc_range)
4701
f2fs_compat_ioc_gc_range(struct file * file,unsigned long arg)4702 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4703 {
4704 struct compat_f2fs_gc_range __user *urange;
4705 struct f2fs_gc_range range;
4706 int err;
4707
4708 urange = compat_ptr(arg);
4709 err = get_user(range.sync, &urange->sync);
4710 err |= get_user(range.start, &urange->start);
4711 err |= get_user(range.len, &urange->len);
4712 if (err)
4713 return -EFAULT;
4714
4715 return __f2fs_ioc_gc_range(file, &range);
4716 }
4717
4718 struct compat_f2fs_move_range {
4719 u32 dst_fd;
4720 compat_u64 pos_in;
4721 compat_u64 pos_out;
4722 compat_u64 len;
4723 };
4724 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
4725 struct compat_f2fs_move_range)
4726
f2fs_compat_ioc_move_range(struct file * file,unsigned long arg)4727 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4728 {
4729 struct compat_f2fs_move_range __user *urange;
4730 struct f2fs_move_range range;
4731 int err;
4732
4733 urange = compat_ptr(arg);
4734 err = get_user(range.dst_fd, &urange->dst_fd);
4735 err |= get_user(range.pos_in, &urange->pos_in);
4736 err |= get_user(range.pos_out, &urange->pos_out);
4737 err |= get_user(range.len, &urange->len);
4738 if (err)
4739 return -EFAULT;
4740
4741 return __f2fs_ioc_move_range(file, &range);
4742 }
4743
f2fs_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)4744 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4745 {
4746 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4747 return -EIO;
4748 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4749 return -ENOSPC;
4750
4751 switch (cmd) {
4752 case FS_IOC32_GETVERSION:
4753 cmd = FS_IOC_GETVERSION;
4754 break;
4755 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4756 return f2fs_compat_ioc_gc_range(file, arg);
4757 case F2FS_IOC32_MOVE_RANGE:
4758 return f2fs_compat_ioc_move_range(file, arg);
4759 case F2FS_IOC_START_ATOMIC_WRITE:
4760 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4761 case F2FS_IOC_START_VOLATILE_WRITE:
4762 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4763 case F2FS_IOC_ABORT_ATOMIC_WRITE:
4764 case F2FS_IOC_SHUTDOWN:
4765 case FITRIM:
4766 case FS_IOC_SET_ENCRYPTION_POLICY:
4767 case FS_IOC_GET_ENCRYPTION_PWSALT:
4768 case FS_IOC_GET_ENCRYPTION_POLICY:
4769 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4770 case FS_IOC_ADD_ENCRYPTION_KEY:
4771 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4772 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4773 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4774 case FS_IOC_GET_ENCRYPTION_NONCE:
4775 case F2FS_IOC_GARBAGE_COLLECT:
4776 case F2FS_IOC_WRITE_CHECKPOINT:
4777 case F2FS_IOC_DEFRAGMENT:
4778 case F2FS_IOC_FLUSH_DEVICE:
4779 case F2FS_IOC_GET_FEATURES:
4780 case F2FS_IOC_GET_PIN_FILE:
4781 case F2FS_IOC_SET_PIN_FILE:
4782 case F2FS_IOC_PRECACHE_EXTENTS:
4783 case F2FS_IOC_RESIZE_FS:
4784 case FS_IOC_ENABLE_VERITY:
4785 case FS_IOC_MEASURE_VERITY:
4786 case FS_IOC_READ_VERITY_METADATA:
4787 case FS_IOC_GETFSLABEL:
4788 case FS_IOC_SETFSLABEL:
4789 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4790 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4791 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4792 case F2FS_IOC_SEC_TRIM_FILE:
4793 case F2FS_IOC_GET_COMPRESS_OPTION:
4794 case F2FS_IOC_SET_COMPRESS_OPTION:
4795 case F2FS_IOC_DECOMPRESS_FILE:
4796 case F2FS_IOC_COMPRESS_FILE:
4797 break;
4798 default:
4799 return -ENOIOCTLCMD;
4800 }
4801 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4802 }
4803 #endif
4804
4805 const struct file_operations f2fs_file_operations = {
4806 .llseek = f2fs_llseek,
4807 .read_iter = f2fs_file_read_iter,
4808 .write_iter = f2fs_file_write_iter,
4809 .open = f2fs_file_open,
4810 .release = f2fs_release_file,
4811 .mmap = f2fs_file_mmap,
4812 .flush = f2fs_file_flush,
4813 .fsync = f2fs_sync_file,
4814 .fallocate = f2fs_fallocate,
4815 .unlocked_ioctl = f2fs_ioctl,
4816 #ifdef CONFIG_COMPAT
4817 .compat_ioctl = f2fs_compat_ioctl,
4818 #endif
4819 .splice_read = generic_file_splice_read,
4820 .splice_write = iter_file_splice_write,
4821 .fadvise = f2fs_file_fadvise,
4822 };
4823