projects / linux-2.6-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
f2fs: fix to dirty inode for i_mode recovery
[linux-2.6-block.git] / fs / f2fs / file.c
... / ...
CommitLineData
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
24#include "f2fs.h"
25#include "node.h"
26#include "segment.h"
27#include "xattr.h"
28#include "acl.h"
29#include "gc.h"
30#include "trace.h"
31#include <trace/events/f2fs.h>
32
33static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
34{
35 struct inode *inode = file_inode(vmf->vma->vm_file);
36 vm_fault_t ret;
37
38 down_read(&F2FS_I(inode)->i_mmap_sem);
39 ret = filemap_fault(vmf);
40 up_read(&F2FS_I(inode)->i_mmap_sem);
41
42 return ret;
43}
44
45static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
46{
47 struct page *page = vmf->page;
48 struct inode *inode = file_inode(vmf->vma->vm_file);
49 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
50 struct dnode_of_data dn = { .node_changed = false };
51 int err;
52
53 if (unlikely(f2fs_cp_error(sbi))) {
54 err = -EIO;
55 goto err;
56 }
57
58 sb_start_pagefault(inode->i_sb);
59
60 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
61
62 file_update_time(vmf->vma->vm_file);
63 down_read(&F2FS_I(inode)->i_mmap_sem);
64 lock_page(page);
65 if (unlikely(page->mapping != inode->i_mapping ||
66 page_offset(page) > i_size_read(inode) ||
67 !PageUptodate(page))) {
68 unlock_page(page);
69 err = -EFAULT;
70 goto out_sem;
71 }
72
73 /* block allocation */
74 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
75 set_new_dnode(&dn, inode, NULL, NULL, 0);
76 err = f2fs_get_block(&dn, page->index);
77 f2fs_put_dnode(&dn);
78 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
79 if (err) {
80 unlock_page(page);
81 goto out_sem;
82 }
83
84 /* fill the page */
85 f2fs_wait_on_page_writeback(page, DATA, false, true);
86
87 /* wait for GCed page writeback via META_MAPPING */
88 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
89
90 /*
91 * check to see if the page is mapped already (no holes)
92 */
93 if (PageMappedToDisk(page))
94 goto out_sem;
95
96 /* page is wholly or partially inside EOF */
97 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
98 i_size_read(inode)) {
99 loff_t offset;
100
101 offset = i_size_read(inode) & ~PAGE_MASK;
102 zero_user_segment(page, offset, PAGE_SIZE);
103 }
104 set_page_dirty(page);
105 if (!PageUptodate(page))
106 SetPageUptodate(page);
107
108 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
109 f2fs_update_time(sbi, REQ_TIME);
110
111 trace_f2fs_vm_page_mkwrite(page, DATA);
112out_sem:
113 up_read(&F2FS_I(inode)->i_mmap_sem);
114
115 f2fs_balance_fs(sbi, dn.node_changed);
116
117 sb_end_pagefault(inode->i_sb);
118err:
119 return block_page_mkwrite_return(err);
120}
121
122static const struct vm_operations_struct f2fs_file_vm_ops = {
123 .fault = f2fs_filemap_fault,
124 .map_pages = filemap_map_pages,
125 .page_mkwrite = f2fs_vm_page_mkwrite,
126};
127
128static int get_parent_ino(struct inode *inode, nid_t *pino)
129{
130 struct dentry *dentry;
131
132 inode = igrab(inode);
133 dentry = d_find_any_alias(inode);
134 iput(inode);
135 if (!dentry)
136 return 0;
137
138 *pino = parent_ino(dentry);
139 dput(dentry);
140 return 1;
141}
142
143static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
144{
145 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
146 enum cp_reason_type cp_reason = CP_NO_NEEDED;
147
148 if (!S_ISREG(inode->i_mode))
149 cp_reason = CP_NON_REGULAR;
150 else if (inode->i_nlink != 1)
151 cp_reason = CP_HARDLINK;
152 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
153 cp_reason = CP_SB_NEED_CP;
154 else if (file_wrong_pino(inode))
155 cp_reason = CP_WRONG_PINO;
156 else if (!f2fs_space_for_roll_forward(sbi))
157 cp_reason = CP_NO_SPC_ROLL;
158 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
159 cp_reason = CP_NODE_NEED_CP;
160 else if (test_opt(sbi, FASTBOOT))
161 cp_reason = CP_FASTBOOT_MODE;
162 else if (F2FS_OPTION(sbi).active_logs == 2)
163 cp_reason = CP_SPEC_LOG_NUM;
164 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
165 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
166 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
167 TRANS_DIR_INO))
168 cp_reason = CP_RECOVER_DIR;
169
170 return cp_reason;
171}
172
173static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
174{
175 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
176 bool ret = false;
177 /* But we need to avoid that there are some inode updates */
178 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
179 ret = true;
180 f2fs_put_page(i, 0);
181 return ret;
182}
183
184static void try_to_fix_pino(struct inode *inode)
185{
186 struct f2fs_inode_info *fi = F2FS_I(inode);
187 nid_t pino;
188
189 down_write(&fi->i_sem);
190 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
191 get_parent_ino(inode, &pino)) {
192 f2fs_i_pino_write(inode, pino);
193 file_got_pino(inode);
194 }
195 up_write(&fi->i_sem);
196}
197
198static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
199 int datasync, bool atomic)
200{
201 struct inode *inode = file->f_mapping->host;
202 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
203 nid_t ino = inode->i_ino;
204 int ret = 0;
205 enum cp_reason_type cp_reason = 0;
206 struct writeback_control wbc = {
207 .sync_mode = WB_SYNC_ALL,
208 .nr_to_write = LONG_MAX,
209 .for_reclaim = 0,
210 };
211 unsigned int seq_id = 0;
212
213 if (unlikely(f2fs_readonly(inode->i_sb) ||
214 is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
215 return 0;
216
217 trace_f2fs_sync_file_enter(inode);
218
219 if (S_ISDIR(inode->i_mode))
220 goto go_write;
221
222 /* if fdatasync is triggered, let's do in-place-update */
223 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
224 set_inode_flag(inode, FI_NEED_IPU);
225 ret = file_write_and_wait_range(file, start, end);
226 clear_inode_flag(inode, FI_NEED_IPU);
227
228 if (ret) {
229 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
230 return ret;
231 }
232
233 /* if the inode is dirty, let's recover all the time */
234 if (!f2fs_skip_inode_update(inode, datasync)) {
235 f2fs_write_inode(inode, NULL);
236 goto go_write;
237 }
238
239 /*
240 * if there is no written data, don't waste time to write recovery info.
241 */
242 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
243 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
244
245 /* it may call write_inode just prior to fsync */
246 if (need_inode_page_update(sbi, ino))
247 goto go_write;
248
249 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
250 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
251 goto flush_out;
252 goto out;
253 }
254go_write:
255 /*
256 * Both of fdatasync() and fsync() are able to be recovered from
257 * sudden-power-off.
258 */
259 down_read(&F2FS_I(inode)->i_sem);
260 cp_reason = need_do_checkpoint(inode);
261 up_read(&F2FS_I(inode)->i_sem);
262
263 if (cp_reason) {
264 /* all the dirty node pages should be flushed for POR */
265 ret = f2fs_sync_fs(inode->i_sb, 1);
266
267 /*
268 * We've secured consistency through sync_fs. Following pino
269 * will be used only for fsynced inodes after checkpoint.
270 */
271 try_to_fix_pino(inode);
272 clear_inode_flag(inode, FI_APPEND_WRITE);
273 clear_inode_flag(inode, FI_UPDATE_WRITE);
274 goto out;
275 }
276sync_nodes:
277 atomic_inc(&sbi->wb_sync_req[NODE]);
278 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
279 atomic_dec(&sbi->wb_sync_req[NODE]);
280 if (ret)
281 goto out;
282
283 /* if cp_error was enabled, we should avoid infinite loop */
284 if (unlikely(f2fs_cp_error(sbi))) {
285 ret = -EIO;
286 goto out;
287 }
288
289 if (f2fs_need_inode_block_update(sbi, ino)) {
290 f2fs_mark_inode_dirty_sync(inode, true);
291 f2fs_write_inode(inode, NULL);
292 goto sync_nodes;
293 }
294
295 /*
296 * If it's atomic_write, it's just fine to keep write ordering. So
297 * here we don't need to wait for node write completion, since we use
298 * node chain which serializes node blocks. If one of node writes are
299 * reordered, we can see simply broken chain, resulting in stopping
300 * roll-forward recovery. It means we'll recover all or none node blocks
301 * given fsync mark.
302 */
303 if (!atomic) {
304 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
305 if (ret)
306 goto out;
307 }
308
309 /* once recovery info is written, don't need to tack this */
310 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
311 clear_inode_flag(inode, FI_APPEND_WRITE);
312flush_out:
313 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
314 ret = f2fs_issue_flush(sbi, inode->i_ino);
315 if (!ret) {
316 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
317 clear_inode_flag(inode, FI_UPDATE_WRITE);
318 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
319 }
320 f2fs_update_time(sbi, REQ_TIME);
321out:
322 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
323 f2fs_trace_ios(NULL, 1);
324 return ret;
325}
326
327int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
328{
329 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
330 return -EIO;
331 return f2fs_do_sync_file(file, start, end, datasync, false);
332}
333
334static pgoff_t __get_first_dirty_index(struct address_space *mapping,
335 pgoff_t pgofs, int whence)
336{
337 struct page *page;
338 int nr_pages;
339
340 if (whence != SEEK_DATA)
341 return 0;
342
343 /* find first dirty page index */
344 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
345 1, &page);
346 if (!nr_pages)
347 return ULONG_MAX;
348 pgofs = page->index;
349 put_page(page);
350 return pgofs;
351}
352
353static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
354 pgoff_t dirty, pgoff_t pgofs, int whence)
355{
356 switch (whence) {
357 case SEEK_DATA:
358 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
359 is_valid_data_blkaddr(sbi, blkaddr))
360 return true;
361 break;
362 case SEEK_HOLE:
363 if (blkaddr == NULL_ADDR)
364 return true;
365 break;
366 }
367 return false;
368}
369
370static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
371{
372 struct inode *inode = file->f_mapping->host;
373 loff_t maxbytes = inode->i_sb->s_maxbytes;
374 struct dnode_of_data dn;
375 pgoff_t pgofs, end_offset, dirty;
376 loff_t data_ofs = offset;
377 loff_t isize;
378 int err = 0;
379
380 inode_lock(inode);
381
382 isize = i_size_read(inode);
383 if (offset >= isize)
384 goto fail;
385
386 /* handle inline data case */
387 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
388 if (whence == SEEK_HOLE)
389 data_ofs = isize;
390 goto found;
391 }
392
393 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
394
395 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
396
397 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
398 set_new_dnode(&dn, inode, NULL, NULL, 0);
399 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
400 if (err && err != -ENOENT) {
401 goto fail;
402 } else if (err == -ENOENT) {
403 /* direct node does not exists */
404 if (whence == SEEK_DATA) {
405 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
406 continue;
407 } else {
408 goto found;
409 }
410 }
411
412 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
413
414 /* find data/hole in dnode block */
415 for (; dn.ofs_in_node < end_offset;
416 dn.ofs_in_node++, pgofs++,
417 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
418 block_t blkaddr;
419
420 blkaddr = datablock_addr(dn.inode,
421 dn.node_page, dn.ofs_in_node);
422
423 if (__is_valid_data_blkaddr(blkaddr) &&
424 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
425 blkaddr, DATA_GENERIC)) {
426 f2fs_put_dnode(&dn);
427 goto fail;
428 }
429
430 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
431 pgofs, whence)) {
432 f2fs_put_dnode(&dn);
433 goto found;
434 }
435 }
436 f2fs_put_dnode(&dn);
437 }
438
439 if (whence == SEEK_DATA)
440 goto fail;
441found:
442 if (whence == SEEK_HOLE && data_ofs > isize)
443 data_ofs = isize;
444 inode_unlock(inode);
445 return vfs_setpos(file, data_ofs, maxbytes);
446fail:
447 inode_unlock(inode);
448 return -ENXIO;
449}
450
451static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
452{
453 struct inode *inode = file->f_mapping->host;
454 loff_t maxbytes = inode->i_sb->s_maxbytes;
455
456 switch (whence) {
457 case SEEK_SET:
458 case SEEK_CUR:
459 case SEEK_END:
460 return generic_file_llseek_size(file, offset, whence,
461 maxbytes, i_size_read(inode));
462 case SEEK_DATA:
463 case SEEK_HOLE:
464 if (offset < 0)
465 return -ENXIO;
466 return f2fs_seek_block(file, offset, whence);
467 }
468
469 return -EINVAL;
470}
471
472static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
473{
474 struct inode *inode = file_inode(file);
475 int err;
476
477 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
478 return -EIO;
479
480 /* we don't need to use inline_data strictly */
481 err = f2fs_convert_inline_inode(inode);
482 if (err)
483 return err;
484
485 file_accessed(file);
486 vma->vm_ops = &f2fs_file_vm_ops;
487 return 0;
488}
489
490static int f2fs_file_open(struct inode *inode, struct file *filp)
491{
492 int err = fscrypt_file_open(inode, filp);
493
494 if (err)
495 return err;
496
497 filp->f_mode |= FMODE_NOWAIT;
498
499 return dquot_file_open(inode, filp);
500}
501
502void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
503{
504 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
505 struct f2fs_node *raw_node;
506 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
507 __le32 *addr;
508 int base = 0;
509
510 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
511 base = get_extra_isize(dn->inode);
512
513 raw_node = F2FS_NODE(dn->node_page);
514 addr = blkaddr_in_node(raw_node) + base + ofs;
515
516 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
517 block_t blkaddr = le32_to_cpu(*addr);
518
519 if (blkaddr == NULL_ADDR)
520 continue;
521
522 dn->data_blkaddr = NULL_ADDR;
523 f2fs_set_data_blkaddr(dn);
524
525 if (__is_valid_data_blkaddr(blkaddr) &&
526 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
527 continue;
528
529 f2fs_invalidate_blocks(sbi, blkaddr);
530 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
531 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
532 nr_free++;
533 }
534
535 if (nr_free) {
536 pgoff_t fofs;
537 /*
538 * once we invalidate valid blkaddr in range [ofs, ofs + count],
539 * we will invalidate all blkaddr in the whole range.
540 */
541 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
542 dn->inode) + ofs;
543 f2fs_update_extent_cache_range(dn, fofs, 0, len);
544 dec_valid_block_count(sbi, dn->inode, nr_free);
545 }
546 dn->ofs_in_node = ofs;
547
548 f2fs_update_time(sbi, REQ_TIME);
549 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
550 dn->ofs_in_node, nr_free);
551}
552
553void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
554{
555 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
556}
557
558static int truncate_partial_data_page(struct inode *inode, u64 from,
559 bool cache_only)
560{
561 loff_t offset = from & (PAGE_SIZE - 1);
562 pgoff_t index = from >> PAGE_SHIFT;
563 struct address_space *mapping = inode->i_mapping;
564 struct page *page;
565
566 if (!offset && !cache_only)
567 return 0;
568
569 if (cache_only) {
570 page = find_lock_page(mapping, index);
571 if (page && PageUptodate(page))
572 goto truncate_out;
573 f2fs_put_page(page, 1);
574 return 0;
575 }
576
577 page = f2fs_get_lock_data_page(inode, index, true);
578 if (IS_ERR(page))
579 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
580truncate_out:
581 f2fs_wait_on_page_writeback(page, DATA, true, true);
582 zero_user(page, offset, PAGE_SIZE - offset);
583
584 /* An encrypted inode should have a key and truncate the last page. */
585 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
586 if (!cache_only)
587 set_page_dirty(page);
588 f2fs_put_page(page, 1);
589 return 0;
590}
591
592int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
593{
594 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
595 struct dnode_of_data dn;
596 pgoff_t free_from;
597 int count = 0, err = 0;
598 struct page *ipage;
599 bool truncate_page = false;
600
601 trace_f2fs_truncate_blocks_enter(inode, from);
602
603 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
604
605 if (free_from >= sbi->max_file_blocks)
606 goto free_partial;
607
608 if (lock)
609 f2fs_lock_op(sbi);
610
611 ipage = f2fs_get_node_page(sbi, inode->i_ino);
612 if (IS_ERR(ipage)) {
613 err = PTR_ERR(ipage);
614 goto out;
615 }
616
617 if (f2fs_has_inline_data(inode)) {
618 f2fs_truncate_inline_inode(inode, ipage, from);
619 f2fs_put_page(ipage, 1);
620 truncate_page = true;
621 goto out;
622 }
623
624 set_new_dnode(&dn, inode, ipage, NULL, 0);
625 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
626 if (err) {
627 if (err == -ENOENT)
628 goto free_next;
629 goto out;
630 }
631
632 count = ADDRS_PER_PAGE(dn.node_page, inode);
633
634 count -= dn.ofs_in_node;
635 f2fs_bug_on(sbi, count < 0);
636
637 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
638 f2fs_truncate_data_blocks_range(&dn, count);
639 free_from += count;
640 }
641
642 f2fs_put_dnode(&dn);
643free_next:
644 err = f2fs_truncate_inode_blocks(inode, free_from);
645out:
646 if (lock)
647 f2fs_unlock_op(sbi);
648free_partial:
649 /* lastly zero out the first data page */
650 if (!err)
651 err = truncate_partial_data_page(inode, from, truncate_page);
652
653 trace_f2fs_truncate_blocks_exit(inode, err);
654 return err;
655}
656
657int f2fs_truncate(struct inode *inode)
658{
659 int err;
660
661 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
662 return -EIO;
663
664 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
665 S_ISLNK(inode->i_mode)))
666 return 0;
667
668 trace_f2fs_truncate(inode);
669
670 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
671 f2fs_show_injection_info(FAULT_TRUNCATE);
672 return -EIO;
673 }
674
675 /* we should check inline_data size */
676 if (!f2fs_may_inline_data(inode)) {
677 err = f2fs_convert_inline_inode(inode);
678 if (err)
679 return err;
680 }
681
682 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
683 if (err)
684 return err;
685
686 inode->i_mtime = inode->i_ctime = current_time(inode);
687 f2fs_mark_inode_dirty_sync(inode, false);
688 return 0;
689}
690
691int f2fs_getattr(const struct path *path, struct kstat *stat,
692 u32 request_mask, unsigned int query_flags)
693{
694 struct inode *inode = d_inode(path->dentry);
695 struct f2fs_inode_info *fi = F2FS_I(inode);
696 struct f2fs_inode *ri;
697 unsigned int flags;
698
699 if (f2fs_has_extra_attr(inode) &&
700 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
701 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
702 stat->result_mask |= STATX_BTIME;
703 stat->btime.tv_sec = fi->i_crtime.tv_sec;
704 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
705 }
706
707 flags = fi->i_flags & F2FS_FL_USER_VISIBLE;
708 if (flags & F2FS_APPEND_FL)
709 stat->attributes |= STATX_ATTR_APPEND;
710 if (flags & F2FS_COMPR_FL)
711 stat->attributes |= STATX_ATTR_COMPRESSED;
712 if (f2fs_encrypted_inode(inode))
713 stat->attributes |= STATX_ATTR_ENCRYPTED;
714 if (flags & F2FS_IMMUTABLE_FL)
715 stat->attributes |= STATX_ATTR_IMMUTABLE;
716 if (flags & F2FS_NODUMP_FL)
717 stat->attributes |= STATX_ATTR_NODUMP;
718
719 stat->attributes_mask |= (STATX_ATTR_APPEND |
720 STATX_ATTR_COMPRESSED |
721 STATX_ATTR_ENCRYPTED |
722 STATX_ATTR_IMMUTABLE |
723 STATX_ATTR_NODUMP);
724
725 generic_fillattr(inode, stat);
726
727 /* we need to show initial sectors used for inline_data/dentries */
728 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
729 f2fs_has_inline_dentry(inode))
730 stat->blocks += (stat->size + 511) >> 9;
731
732 return 0;
733}
734
735#ifdef CONFIG_F2FS_FS_POSIX_ACL
736static void __setattr_copy(struct inode *inode, const struct iattr *attr)
737{
738 unsigned int ia_valid = attr->ia_valid;
739
740 if (ia_valid & ATTR_UID)
741 inode->i_uid = attr->ia_uid;
742 if (ia_valid & ATTR_GID)
743 inode->i_gid = attr->ia_gid;
744 if (ia_valid & ATTR_ATIME)
745 inode->i_atime = timespec64_trunc(attr->ia_atime,
746 inode->i_sb->s_time_gran);
747 if (ia_valid & ATTR_MTIME)
748 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
749 inode->i_sb->s_time_gran);
750 if (ia_valid & ATTR_CTIME)
751 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
752 inode->i_sb->s_time_gran);
753 if (ia_valid & ATTR_MODE) {
754 umode_t mode = attr->ia_mode;
755
756 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
757 mode &= ~S_ISGID;
758 set_acl_inode(inode, mode);
759 }
760}
761#else
762#define __setattr_copy setattr_copy
763#endif
764
765int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
766{
767 struct inode *inode = d_inode(dentry);
768 int err;
769
770 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
771 return -EIO;
772
773 err = setattr_prepare(dentry, attr);
774 if (err)
775 return err;
776
777 err = fscrypt_prepare_setattr(dentry, attr);
778 if (err)
779 return err;
780
781 if (is_quota_modification(inode, attr)) {
782 err = dquot_initialize(inode);
783 if (err)
784 return err;
785 }
786 if ((attr->ia_valid & ATTR_UID &&
787 !uid_eq(attr->ia_uid, inode->i_uid)) ||
788 (attr->ia_valid & ATTR_GID &&
789 !gid_eq(attr->ia_gid, inode->i_gid))) {
790 f2fs_lock_op(F2FS_I_SB(inode));
791 err = dquot_transfer(inode, attr);
792 if (err) {
793 set_sbi_flag(F2FS_I_SB(inode),
794 SBI_QUOTA_NEED_REPAIR);
795 f2fs_unlock_op(F2FS_I_SB(inode));
796 return err;
797 }
798 /*
799 * update uid/gid under lock_op(), so that dquot and inode can
800 * be updated atomically.
801 */
802 if (attr->ia_valid & ATTR_UID)
803 inode->i_uid = attr->ia_uid;
804 if (attr->ia_valid & ATTR_GID)
805 inode->i_gid = attr->ia_gid;
806 f2fs_mark_inode_dirty_sync(inode, true);
807 f2fs_unlock_op(F2FS_I_SB(inode));
808 }
809
810 if (attr->ia_valid & ATTR_SIZE) {
811 bool to_smaller = (attr->ia_size <= i_size_read(inode));
812
813 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
814 down_write(&F2FS_I(inode)->i_mmap_sem);
815
816 truncate_setsize(inode, attr->ia_size);
817
818 if (to_smaller)
819 err = f2fs_truncate(inode);
820 /*
821 * do not trim all blocks after i_size if target size is
822 * larger than i_size.
823 */
824 up_write(&F2FS_I(inode)->i_mmap_sem);
825 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
826
827 if (err)
828 return err;
829
830 if (!to_smaller) {
831 /* should convert inline inode here */
832 if (!f2fs_may_inline_data(inode)) {
833 err = f2fs_convert_inline_inode(inode);
834 if (err)
835 return err;
836 }
837 inode->i_mtime = inode->i_ctime = current_time(inode);
838 }
839
840 down_write(&F2FS_I(inode)->i_sem);
841 F2FS_I(inode)->last_disk_size = i_size_read(inode);
842 up_write(&F2FS_I(inode)->i_sem);
843 }
844
845 __setattr_copy(inode, attr);
846
847 if (attr->ia_valid & ATTR_MODE) {
848 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
849 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
850 inode->i_mode = F2FS_I(inode)->i_acl_mode;
851 clear_inode_flag(inode, FI_ACL_MODE);
852 }
853 }
854
855 /* file size may changed here */
856 f2fs_mark_inode_dirty_sync(inode, true);
857
858 /* inode change will produce dirty node pages flushed by checkpoint */
859 f2fs_balance_fs(F2FS_I_SB(inode), true);
860
861 return err;
862}
863
864const struct inode_operations f2fs_file_inode_operations = {
865 .getattr = f2fs_getattr,
866 .setattr = f2fs_setattr,
867 .get_acl = f2fs_get_acl,
868 .set_acl = f2fs_set_acl,
869#ifdef CONFIG_F2FS_FS_XATTR
870 .listxattr = f2fs_listxattr,
871#endif
872 .fiemap = f2fs_fiemap,
873};
874
875static int fill_zero(struct inode *inode, pgoff_t index,
876 loff_t start, loff_t len)
877{
878 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
879 struct page *page;
880
881 if (!len)
882 return 0;
883
884 f2fs_balance_fs(sbi, true);
885
886 f2fs_lock_op(sbi);
887 page = f2fs_get_new_data_page(inode, NULL, index, false);
888 f2fs_unlock_op(sbi);
889
890 if (IS_ERR(page))
891 return PTR_ERR(page);
892
893 f2fs_wait_on_page_writeback(page, DATA, true, true);
894 zero_user(page, start, len);
895 set_page_dirty(page);
896 f2fs_put_page(page, 1);
897 return 0;
898}
899
900int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
901{
902 int err;
903
904 while (pg_start < pg_end) {
905 struct dnode_of_data dn;
906 pgoff_t end_offset, count;
907
908 set_new_dnode(&dn, inode, NULL, NULL, 0);
909 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
910 if (err) {
911 if (err == -ENOENT) {
912 pg_start = f2fs_get_next_page_offset(&dn,
913 pg_start);
914 continue;
915 }
916 return err;
917 }
918
919 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
920 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
921
922 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
923
924 f2fs_truncate_data_blocks_range(&dn, count);
925 f2fs_put_dnode(&dn);
926
927 pg_start += count;
928 }
929 return 0;
930}
931
932static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
933{
934 pgoff_t pg_start, pg_end;
935 loff_t off_start, off_end;
936 int ret;
937
938 ret = f2fs_convert_inline_inode(inode);
939 if (ret)
940 return ret;
941
942 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
943 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
944
945 off_start = offset & (PAGE_SIZE - 1);
946 off_end = (offset + len) & (PAGE_SIZE - 1);
947
948 if (pg_start == pg_end) {
949 ret = fill_zero(inode, pg_start, off_start,
950 off_end - off_start);
951 if (ret)
952 return ret;
953 } else {
954 if (off_start) {
955 ret = fill_zero(inode, pg_start++, off_start,
956 PAGE_SIZE - off_start);
957 if (ret)
958 return ret;
959 }
960 if (off_end) {
961 ret = fill_zero(inode, pg_end, 0, off_end);
962 if (ret)
963 return ret;
964 }
965
966 if (pg_start < pg_end) {
967 struct address_space *mapping = inode->i_mapping;
968 loff_t blk_start, blk_end;
969 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
970
971 f2fs_balance_fs(sbi, true);
972
973 blk_start = (loff_t)pg_start << PAGE_SHIFT;
974 blk_end = (loff_t)pg_end << PAGE_SHIFT;
975
976 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
977 down_write(&F2FS_I(inode)->i_mmap_sem);
978
979 truncate_inode_pages_range(mapping, blk_start,
980 blk_end - 1);
981
982 f2fs_lock_op(sbi);
983 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
984 f2fs_unlock_op(sbi);
985
986 up_write(&F2FS_I(inode)->i_mmap_sem);
987 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
988 }
989 }
990
991 return ret;
992}
993
994static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
995 int *do_replace, pgoff_t off, pgoff_t len)
996{
997 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
998 struct dnode_of_data dn;
999 int ret, done, i;
1000
1001next_dnode:
1002 set_new_dnode(&dn, inode, NULL, NULL, 0);
1003 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1004 if (ret && ret != -ENOENT) {
1005 return ret;
1006 } else if (ret == -ENOENT) {
1007 if (dn.max_level == 0)
1008 return -ENOENT;
1009 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
1010 blkaddr += done;
1011 do_replace += done;
1012 goto next;
1013 }
1014
1015 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1016 dn.ofs_in_node, len);
1017 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1018 *blkaddr = datablock_addr(dn.inode,
1019 dn.node_page, dn.ofs_in_node);
1020 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1021
1022 if (test_opt(sbi, LFS)) {
1023 f2fs_put_dnode(&dn);
1024 return -ENOTSUPP;
1025 }
1026
1027 /* do not invalidate this block address */
1028 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1029 *do_replace = 1;
1030 }
1031 }
1032 f2fs_put_dnode(&dn);
1033next:
1034 len -= done;
1035 off += done;
1036 if (len)
1037 goto next_dnode;
1038 return 0;
1039}
1040
1041static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1042 int *do_replace, pgoff_t off, int len)
1043{
1044 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1045 struct dnode_of_data dn;
1046 int ret, i;
1047
1048 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1049 if (*do_replace == 0)
1050 continue;
1051
1052 set_new_dnode(&dn, inode, NULL, NULL, 0);
1053 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1054 if (ret) {
1055 dec_valid_block_count(sbi, inode, 1);
1056 f2fs_invalidate_blocks(sbi, *blkaddr);
1057 } else {
1058 f2fs_update_data_blkaddr(&dn, *blkaddr);
1059 }
1060 f2fs_put_dnode(&dn);
1061 }
1062 return 0;
1063}
1064
1065static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1066 block_t *blkaddr, int *do_replace,
1067 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1068{
1069 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1070 pgoff_t i = 0;
1071 int ret;
1072
1073 while (i < len) {
1074 if (blkaddr[i] == NULL_ADDR && !full) {
1075 i++;
1076 continue;
1077 }
1078
1079 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1080 struct dnode_of_data dn;
1081 struct node_info ni;
1082 size_t new_size;
1083 pgoff_t ilen;
1084
1085 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1086 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1087 if (ret)
1088 return ret;
1089
1090 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1091 if (ret) {
1092 f2fs_put_dnode(&dn);
1093 return ret;
1094 }
1095
1096 ilen = min((pgoff_t)
1097 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1098 dn.ofs_in_node, len - i);
1099 do {
1100 dn.data_blkaddr = datablock_addr(dn.inode,
1101 dn.node_page, dn.ofs_in_node);
1102 f2fs_truncate_data_blocks_range(&dn, 1);
1103
1104 if (do_replace[i]) {
1105 f2fs_i_blocks_write(src_inode,
1106 1, false, false);
1107 f2fs_i_blocks_write(dst_inode,
1108 1, true, false);
1109 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1110 blkaddr[i], ni.version, true, false);
1111
1112 do_replace[i] = 0;
1113 }
1114 dn.ofs_in_node++;
1115 i++;
1116 new_size = (dst + i) << PAGE_SHIFT;
1117 if (dst_inode->i_size < new_size)
1118 f2fs_i_size_write(dst_inode, new_size);
1119 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1120
1121 f2fs_put_dnode(&dn);
1122 } else {
1123 struct page *psrc, *pdst;
1124
1125 psrc = f2fs_get_lock_data_page(src_inode,
1126 src + i, true);
1127 if (IS_ERR(psrc))
1128 return PTR_ERR(psrc);
1129 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1130 true);
1131 if (IS_ERR(pdst)) {
1132 f2fs_put_page(psrc, 1);
1133 return PTR_ERR(pdst);
1134 }
1135 f2fs_copy_page(psrc, pdst);
1136 set_page_dirty(pdst);
1137 f2fs_put_page(pdst, 1);
1138 f2fs_put_page(psrc, 1);
1139
1140 ret = f2fs_truncate_hole(src_inode,
1141 src + i, src + i + 1);
1142 if (ret)
1143 return ret;
1144 i++;
1145 }
1146 }
1147 return 0;
1148}
1149
1150static int __exchange_data_block(struct inode *src_inode,
1151 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1152 pgoff_t len, bool full)
1153{
1154 block_t *src_blkaddr;
1155 int *do_replace;
1156 pgoff_t olen;
1157 int ret;
1158
1159 while (len) {
1160 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1161
1162 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1163 array_size(olen, sizeof(block_t)),
1164 GFP_KERNEL);
1165 if (!src_blkaddr)
1166 return -ENOMEM;
1167
1168 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1169 array_size(olen, sizeof(int)),
1170 GFP_KERNEL);
1171 if (!do_replace) {
1172 kvfree(src_blkaddr);
1173 return -ENOMEM;
1174 }
1175
1176 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1177 do_replace, src, olen);
1178 if (ret)
1179 goto roll_back;
1180
1181 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1182 do_replace, src, dst, olen, full);
1183 if (ret)
1184 goto roll_back;
1185
1186 src += olen;
1187 dst += olen;
1188 len -= olen;
1189
1190 kvfree(src_blkaddr);
1191 kvfree(do_replace);
1192 }
1193 return 0;
1194
1195roll_back:
1196 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1197 kvfree(src_blkaddr);
1198 kvfree(do_replace);
1199 return ret;
1200}
1201
1202static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1203{
1204 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1205 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1206 pgoff_t start = offset >> PAGE_SHIFT;
1207 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1208 int ret;
1209
1210 f2fs_balance_fs(sbi, true);
1211
1212 /* avoid gc operation during block exchange */
1213 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1214 down_write(&F2FS_I(inode)->i_mmap_sem);
1215
1216 f2fs_lock_op(sbi);
1217 f2fs_drop_extent_tree(inode);
1218 truncate_pagecache(inode, offset);
1219 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1220 f2fs_unlock_op(sbi);
1221
1222 up_write(&F2FS_I(inode)->i_mmap_sem);
1223 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1224 return ret;
1225}
1226
1227static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1228{
1229 loff_t new_size;
1230 int ret;
1231
1232 if (offset + len >= i_size_read(inode))
1233 return -EINVAL;
1234
1235 /* collapse range should be aligned to block size of f2fs. */
1236 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1237 return -EINVAL;
1238
1239 ret = f2fs_convert_inline_inode(inode);
1240 if (ret)
1241 return ret;
1242
1243 /* write out all dirty pages from offset */
1244 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1245 if (ret)
1246 return ret;
1247
1248 ret = f2fs_do_collapse(inode, offset, len);
1249 if (ret)
1250 return ret;
1251
1252 /* write out all moved pages, if possible */
1253 down_write(&F2FS_I(inode)->i_mmap_sem);
1254 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1255 truncate_pagecache(inode, offset);
1256
1257 new_size = i_size_read(inode) - len;
1258 truncate_pagecache(inode, new_size);
1259
1260 ret = f2fs_truncate_blocks(inode, new_size, true);
1261 up_write(&F2FS_I(inode)->i_mmap_sem);
1262 if (!ret)
1263 f2fs_i_size_write(inode, new_size);
1264 return ret;
1265}
1266
1267static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1268 pgoff_t end)
1269{
1270 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1271 pgoff_t index = start;
1272 unsigned int ofs_in_node = dn->ofs_in_node;
1273 blkcnt_t count = 0;
1274 int ret;
1275
1276 for (; index < end; index++, dn->ofs_in_node++) {
1277 if (datablock_addr(dn->inode, dn->node_page,
1278 dn->ofs_in_node) == NULL_ADDR)
1279 count++;
1280 }
1281
1282 dn->ofs_in_node = ofs_in_node;
1283 ret = f2fs_reserve_new_blocks(dn, count);
1284 if (ret)
1285 return ret;
1286
1287 dn->ofs_in_node = ofs_in_node;
1288 for (index = start; index < end; index++, dn->ofs_in_node++) {
1289 dn->data_blkaddr = datablock_addr(dn->inode,
1290 dn->node_page, dn->ofs_in_node);
1291 /*
1292 * f2fs_reserve_new_blocks will not guarantee entire block
1293 * allocation.
1294 */
1295 if (dn->data_blkaddr == NULL_ADDR) {
1296 ret = -ENOSPC;
1297 break;
1298 }
1299 if (dn->data_blkaddr != NEW_ADDR) {
1300 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1301 dn->data_blkaddr = NEW_ADDR;
1302 f2fs_set_data_blkaddr(dn);
1303 }
1304 }
1305
1306 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1307
1308 return ret;
1309}
1310
1311static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1312 int mode)
1313{
1314 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1315 struct address_space *mapping = inode->i_mapping;
1316 pgoff_t index, pg_start, pg_end;
1317 loff_t new_size = i_size_read(inode);
1318 loff_t off_start, off_end;
1319 int ret = 0;
1320
1321 ret = inode_newsize_ok(inode, (len + offset));
1322 if (ret)
1323 return ret;
1324
1325 ret = f2fs_convert_inline_inode(inode);
1326 if (ret)
1327 return ret;
1328
1329 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1330 if (ret)
1331 return ret;
1332
1333 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1334 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1335
1336 off_start = offset & (PAGE_SIZE - 1);
1337 off_end = (offset + len) & (PAGE_SIZE - 1);
1338
1339 if (pg_start == pg_end) {
1340 ret = fill_zero(inode, pg_start, off_start,
1341 off_end - off_start);
1342 if (ret)
1343 return ret;
1344
1345 new_size = max_t(loff_t, new_size, offset + len);
1346 } else {
1347 if (off_start) {
1348 ret = fill_zero(inode, pg_start++, off_start,
1349 PAGE_SIZE - off_start);
1350 if (ret)
1351 return ret;
1352
1353 new_size = max_t(loff_t, new_size,
1354 (loff_t)pg_start << PAGE_SHIFT);
1355 }
1356
1357 for (index = pg_start; index < pg_end;) {
1358 struct dnode_of_data dn;
1359 unsigned int end_offset;
1360 pgoff_t end;
1361
1362 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1363 down_write(&F2FS_I(inode)->i_mmap_sem);
1364
1365 truncate_pagecache_range(inode,
1366 (loff_t)index << PAGE_SHIFT,
1367 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1368
1369 f2fs_lock_op(sbi);
1370
1371 set_new_dnode(&dn, inode, NULL, NULL, 0);
1372 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1373 if (ret) {
1374 f2fs_unlock_op(sbi);
1375 up_write(&F2FS_I(inode)->i_mmap_sem);
1376 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1377 goto out;
1378 }
1379
1380 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1381 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1382
1383 ret = f2fs_do_zero_range(&dn, index, end);
1384 f2fs_put_dnode(&dn);
1385
1386 f2fs_unlock_op(sbi);
1387 up_write(&F2FS_I(inode)->i_mmap_sem);
1388 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1389
1390 f2fs_balance_fs(sbi, dn.node_changed);
1391
1392 if (ret)
1393 goto out;
1394
1395 index = end;
1396 new_size = max_t(loff_t, new_size,
1397 (loff_t)index << PAGE_SHIFT);
1398 }
1399
1400 if (off_end) {
1401 ret = fill_zero(inode, pg_end, 0, off_end);
1402 if (ret)
1403 goto out;
1404
1405 new_size = max_t(loff_t, new_size, offset + len);
1406 }
1407 }
1408
1409out:
1410 if (new_size > i_size_read(inode)) {
1411 if (mode & FALLOC_FL_KEEP_SIZE)
1412 file_set_keep_isize(inode);
1413 else
1414 f2fs_i_size_write(inode, new_size);
1415 }
1416 return ret;
1417}
1418
1419static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1420{
1421 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1422 pgoff_t nr, pg_start, pg_end, delta, idx;
1423 loff_t new_size;
1424 int ret = 0;
1425
1426 new_size = i_size_read(inode) + len;
1427 ret = inode_newsize_ok(inode, new_size);
1428 if (ret)
1429 return ret;
1430
1431 if (offset >= i_size_read(inode))
1432 return -EINVAL;
1433
1434 /* insert range should be aligned to block size of f2fs. */
1435 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1436 return -EINVAL;
1437
1438 ret = f2fs_convert_inline_inode(inode);
1439 if (ret)
1440 return ret;
1441
1442 f2fs_balance_fs(sbi, true);
1443
1444 down_write(&F2FS_I(inode)->i_mmap_sem);
1445 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1446 up_write(&F2FS_I(inode)->i_mmap_sem);
1447 if (ret)
1448 return ret;
1449
1450 /* write out all dirty pages from offset */
1451 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1452 if (ret)
1453 return ret;
1454
1455 pg_start = offset >> PAGE_SHIFT;
1456 pg_end = (offset + len) >> PAGE_SHIFT;
1457 delta = pg_end - pg_start;
1458 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1459
1460 /* avoid gc operation during block exchange */
1461 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1462 down_write(&F2FS_I(inode)->i_mmap_sem);
1463 truncate_pagecache(inode, offset);
1464
1465 while (!ret && idx > pg_start) {
1466 nr = idx - pg_start;
1467 if (nr > delta)
1468 nr = delta;
1469 idx -= nr;
1470
1471 f2fs_lock_op(sbi);
1472 f2fs_drop_extent_tree(inode);
1473
1474 ret = __exchange_data_block(inode, inode, idx,
1475 idx + delta, nr, false);
1476 f2fs_unlock_op(sbi);
1477 }
1478 up_write(&F2FS_I(inode)->i_mmap_sem);
1479 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1480
1481 /* write out all moved pages, if possible */
1482 down_write(&F2FS_I(inode)->i_mmap_sem);
1483 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1484 truncate_pagecache(inode, offset);
1485 up_write(&F2FS_I(inode)->i_mmap_sem);
1486
1487 if (!ret)
1488 f2fs_i_size_write(inode, new_size);
1489 return ret;
1490}
1491
1492static int expand_inode_data(struct inode *inode, loff_t offset,
1493 loff_t len, int mode)
1494{
1495 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1496 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1497 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1498 .m_may_create = true };
1499 pgoff_t pg_end;
1500 loff_t new_size = i_size_read(inode);
1501 loff_t off_end;
1502 int err;
1503
1504 err = inode_newsize_ok(inode, (len + offset));
1505 if (err)
1506 return err;
1507
1508 err = f2fs_convert_inline_inode(inode);
1509 if (err)
1510 return err;
1511
1512 f2fs_balance_fs(sbi, true);
1513
1514 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1515 off_end = (offset + len) & (PAGE_SIZE - 1);
1516
1517 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1518 map.m_len = pg_end - map.m_lblk;
1519 if (off_end)
1520 map.m_len++;
1521
1522 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1523 if (err) {
1524 pgoff_t last_off;
1525
1526 if (!map.m_len)
1527 return err;
1528
1529 last_off = map.m_lblk + map.m_len - 1;
1530
1531 /* update new size to the failed position */
1532 new_size = (last_off == pg_end) ? offset + len :
1533 (loff_t)(last_off + 1) << PAGE_SHIFT;
1534 } else {
1535 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1536 }
1537
1538 if (new_size > i_size_read(inode)) {
1539 if (mode & FALLOC_FL_KEEP_SIZE)
1540 file_set_keep_isize(inode);
1541 else
1542 f2fs_i_size_write(inode, new_size);
1543 }
1544
1545 return err;
1546}
1547
1548static long f2fs_fallocate(struct file *file, int mode,
1549 loff_t offset, loff_t len)
1550{
1551 struct inode *inode = file_inode(file);
1552 long ret = 0;
1553
1554 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1555 return -EIO;
1556
1557 /* f2fs only support ->fallocate for regular file */
1558 if (!S_ISREG(inode->i_mode))
1559 return -EINVAL;
1560
1561 if (f2fs_encrypted_inode(inode) &&
1562 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1563 return -EOPNOTSUPP;
1564
1565 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1566 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1567 FALLOC_FL_INSERT_RANGE))
1568 return -EOPNOTSUPP;
1569
1570 inode_lock(inode);
1571
1572 if (mode & FALLOC_FL_PUNCH_HOLE) {
1573 if (offset >= inode->i_size)
1574 goto out;
1575
1576 ret = punch_hole(inode, offset, len);
1577 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1578 ret = f2fs_collapse_range(inode, offset, len);
1579 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1580 ret = f2fs_zero_range(inode, offset, len, mode);
1581 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1582 ret = f2fs_insert_range(inode, offset, len);
1583 } else {
1584 ret = expand_inode_data(inode, offset, len, mode);
1585 }
1586
1587 if (!ret) {
1588 inode->i_mtime = inode->i_ctime = current_time(inode);
1589 f2fs_mark_inode_dirty_sync(inode, false);
1590 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1591 }
1592
1593out:
1594 inode_unlock(inode);
1595
1596 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1597 return ret;
1598}
1599
1600static int f2fs_release_file(struct inode *inode, struct file *filp)
1601{
1602 /*
1603 * f2fs_relase_file is called at every close calls. So we should
1604 * not drop any inmemory pages by close called by other process.
1605 */
1606 if (!(filp->f_mode & FMODE_WRITE) ||
1607 atomic_read(&inode->i_writecount) != 1)
1608 return 0;
1609
1610 /* some remained atomic pages should discarded */
1611 if (f2fs_is_atomic_file(inode))
1612 f2fs_drop_inmem_pages(inode);
1613 if (f2fs_is_volatile_file(inode)) {
1614 set_inode_flag(inode, FI_DROP_CACHE);
1615 filemap_fdatawrite(inode->i_mapping);
1616 clear_inode_flag(inode, FI_DROP_CACHE);
1617 clear_inode_flag(inode, FI_VOLATILE_FILE);
1618 stat_dec_volatile_write(inode);
1619 }
1620 return 0;
1621}
1622
1623static int f2fs_file_flush(struct file *file, fl_owner_t id)
1624{
1625 struct inode *inode = file_inode(file);
1626
1627 /*
1628 * If the process doing a transaction is crashed, we should do
1629 * roll-back. Otherwise, other reader/write can see corrupted database
1630 * until all the writers close its file. Since this should be done
1631 * before dropping file lock, it needs to do in ->flush.
1632 */
1633 if (f2fs_is_atomic_file(inode) &&
1634 F2FS_I(inode)->inmem_task == current)
1635 f2fs_drop_inmem_pages(inode);
1636 return 0;
1637}
1638
1639static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1640{
1641 struct inode *inode = file_inode(filp);
1642 struct f2fs_inode_info *fi = F2FS_I(inode);
1643 unsigned int flags = fi->i_flags;
1644
1645 if (f2fs_encrypted_inode(inode))
1646 flags |= F2FS_ENCRYPT_FL;
1647 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1648 flags |= F2FS_INLINE_DATA_FL;
1649 if (is_inode_flag_set(inode, FI_PIN_FILE))
1650 flags |= F2FS_NOCOW_FL;
1651
1652 flags &= F2FS_FL_USER_VISIBLE;
1653
1654 return put_user(flags, (int __user *)arg);
1655}
1656
1657static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1658{
1659 struct f2fs_inode_info *fi = F2FS_I(inode);
1660 unsigned int oldflags;
1661
1662 /* Is it quota file? Do not allow user to mess with it */
1663 if (IS_NOQUOTA(inode))
1664 return -EPERM;
1665
1666 flags = f2fs_mask_flags(inode->i_mode, flags);
1667
1668 oldflags = fi->i_flags;
1669
1670 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL))
1671 if (!capable(CAP_LINUX_IMMUTABLE))
1672 return -EPERM;
1673
1674 flags = flags & F2FS_FL_USER_MODIFIABLE;
1675 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE;
1676 fi->i_flags = flags;
1677
1678 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1679 set_inode_flag(inode, FI_PROJ_INHERIT);
1680 else
1681 clear_inode_flag(inode, FI_PROJ_INHERIT);
1682
1683 inode->i_ctime = current_time(inode);
1684 f2fs_set_inode_flags(inode);
1685 f2fs_mark_inode_dirty_sync(inode, true);
1686 return 0;
1687}
1688
1689static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1690{
1691 struct inode *inode = file_inode(filp);
1692 unsigned int flags;
1693 int ret;
1694
1695 if (!inode_owner_or_capable(inode))
1696 return -EACCES;
1697
1698 if (get_user(flags, (int __user *)arg))
1699 return -EFAULT;
1700
1701 ret = mnt_want_write_file(filp);
1702 if (ret)
1703 return ret;
1704
1705 inode_lock(inode);
1706
1707 ret = __f2fs_ioc_setflags(inode, flags);
1708
1709 inode_unlock(inode);
1710 mnt_drop_write_file(filp);
1711 return ret;
1712}
1713
1714static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1715{
1716 struct inode *inode = file_inode(filp);
1717
1718 return put_user(inode->i_generation, (int __user *)arg);
1719}
1720
1721static int f2fs_ioc_start_atomic_write(struct file *filp)
1722{
1723 struct inode *inode = file_inode(filp);
1724 int ret;
1725
1726 if (!inode_owner_or_capable(inode))
1727 return -EACCES;
1728
1729 if (!S_ISREG(inode->i_mode))
1730 return -EINVAL;
1731
1732 ret = mnt_want_write_file(filp);
1733 if (ret)
1734 return ret;
1735
1736 inode_lock(inode);
1737
1738 if (f2fs_is_atomic_file(inode)) {
1739 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1740 ret = -EINVAL;
1741 goto out;
1742 }
1743
1744 ret = f2fs_convert_inline_inode(inode);
1745 if (ret)
1746 goto out;
1747
1748 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1749
1750 /*
1751 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1752 * f2fs_is_atomic_file.
1753 */
1754 if (get_dirty_pages(inode))
1755 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1756 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1757 inode->i_ino, get_dirty_pages(inode));
1758 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1759 if (ret) {
1760 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1761 goto out;
1762 }
1763
1764 set_inode_flag(inode, FI_ATOMIC_FILE);
1765 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1766 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1767
1768 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1769 F2FS_I(inode)->inmem_task = current;
1770 stat_inc_atomic_write(inode);
1771 stat_update_max_atomic_write(inode);
1772out:
1773 inode_unlock(inode);
1774 mnt_drop_write_file(filp);
1775 return ret;
1776}
1777
1778static int f2fs_ioc_commit_atomic_write(struct file *filp)
1779{
1780 struct inode *inode = file_inode(filp);
1781 int ret;
1782
1783 if (!inode_owner_or_capable(inode))
1784 return -EACCES;
1785
1786 ret = mnt_want_write_file(filp);
1787 if (ret)
1788 return ret;
1789
1790 f2fs_balance_fs(F2FS_I_SB(inode), true);
1791
1792 inode_lock(inode);
1793
1794 if (f2fs_is_volatile_file(inode)) {
1795 ret = -EINVAL;
1796 goto err_out;
1797 }
1798
1799 if (f2fs_is_atomic_file(inode)) {
1800 ret = f2fs_commit_inmem_pages(inode);
1801 if (ret)
1802 goto err_out;
1803
1804 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1805 if (!ret) {
1806 clear_inode_flag(inode, FI_ATOMIC_FILE);
1807 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1808 stat_dec_atomic_write(inode);
1809 }
1810 } else {
1811 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1812 }
1813err_out:
1814 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1815 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1816 ret = -EINVAL;
1817 }
1818 inode_unlock(inode);
1819 mnt_drop_write_file(filp);
1820 return ret;
1821}
1822
1823static int f2fs_ioc_start_volatile_write(struct file *filp)
1824{
1825 struct inode *inode = file_inode(filp);
1826 int ret;
1827
1828 if (!inode_owner_or_capable(inode))
1829 return -EACCES;
1830
1831 if (!S_ISREG(inode->i_mode))
1832 return -EINVAL;
1833
1834 ret = mnt_want_write_file(filp);
1835 if (ret)
1836 return ret;
1837
1838 inode_lock(inode);
1839
1840 if (f2fs_is_volatile_file(inode))
1841 goto out;
1842
1843 ret = f2fs_convert_inline_inode(inode);
1844 if (ret)
1845 goto out;
1846
1847 stat_inc_volatile_write(inode);
1848 stat_update_max_volatile_write(inode);
1849
1850 set_inode_flag(inode, FI_VOLATILE_FILE);
1851 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1852out:
1853 inode_unlock(inode);
1854 mnt_drop_write_file(filp);
1855 return ret;
1856}
1857
1858static int f2fs_ioc_release_volatile_write(struct file *filp)
1859{
1860 struct inode *inode = file_inode(filp);
1861 int ret;
1862
1863 if (!inode_owner_or_capable(inode))
1864 return -EACCES;
1865
1866 ret = mnt_want_write_file(filp);
1867 if (ret)
1868 return ret;
1869
1870 inode_lock(inode);
1871
1872 if (!f2fs_is_volatile_file(inode))
1873 goto out;
1874
1875 if (!f2fs_is_first_block_written(inode)) {
1876 ret = truncate_partial_data_page(inode, 0, true);
1877 goto out;
1878 }
1879
1880 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1881out:
1882 inode_unlock(inode);
1883 mnt_drop_write_file(filp);
1884 return ret;
1885}
1886
1887static int f2fs_ioc_abort_volatile_write(struct file *filp)
1888{
1889 struct inode *inode = file_inode(filp);
1890 int ret;
1891
1892 if (!inode_owner_or_capable(inode))
1893 return -EACCES;
1894
1895 ret = mnt_want_write_file(filp);
1896 if (ret)
1897 return ret;
1898
1899 inode_lock(inode);
1900
1901 if (f2fs_is_atomic_file(inode))
1902 f2fs_drop_inmem_pages(inode);
1903 if (f2fs_is_volatile_file(inode)) {
1904 clear_inode_flag(inode, FI_VOLATILE_FILE);
1905 stat_dec_volatile_write(inode);
1906 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1907 }
1908
1909 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1910
1911 inode_unlock(inode);
1912
1913 mnt_drop_write_file(filp);
1914 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1915 return ret;
1916}
1917
1918static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1919{
1920 struct inode *inode = file_inode(filp);
1921 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1922 struct super_block *sb = sbi->sb;
1923 __u32 in;
1924 int ret = 0;
1925
1926 if (!capable(CAP_SYS_ADMIN))
1927 return -EPERM;
1928
1929 if (get_user(in, (__u32 __user *)arg))
1930 return -EFAULT;
1931
1932 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1933 ret = mnt_want_write_file(filp);
1934 if (ret)
1935 return ret;
1936 }
1937
1938 switch (in) {
1939 case F2FS_GOING_DOWN_FULLSYNC:
1940 sb = freeze_bdev(sb->s_bdev);
1941 if (IS_ERR(sb)) {
1942 ret = PTR_ERR(sb);
1943 goto out;
1944 }
1945 if (sb) {
1946 f2fs_stop_checkpoint(sbi, false);
1947 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1948 thaw_bdev(sb->s_bdev, sb);
1949 }
1950 break;
1951 case F2FS_GOING_DOWN_METASYNC:
1952 /* do checkpoint only */
1953 ret = f2fs_sync_fs(sb, 1);
1954 if (ret)
1955 goto out;
1956 f2fs_stop_checkpoint(sbi, false);
1957 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1958 break;
1959 case F2FS_GOING_DOWN_NOSYNC:
1960 f2fs_stop_checkpoint(sbi, false);
1961 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1962 break;
1963 case F2FS_GOING_DOWN_METAFLUSH:
1964 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1965 f2fs_stop_checkpoint(sbi, false);
1966 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1967 break;
1968 case F2FS_GOING_DOWN_NEED_FSCK:
1969 set_sbi_flag(sbi, SBI_NEED_FSCK);
1970 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
1971 set_sbi_flag(sbi, SBI_IS_DIRTY);
1972 /* do checkpoint only */
1973 ret = f2fs_sync_fs(sb, 1);
1974 goto out;
1975 default:
1976 ret = -EINVAL;
1977 goto out;
1978 }
1979
1980 f2fs_stop_gc_thread(sbi);
1981 f2fs_stop_discard_thread(sbi);
1982
1983 f2fs_drop_discard_cmd(sbi);
1984 clear_opt(sbi, DISCARD);
1985
1986 f2fs_update_time(sbi, REQ_TIME);
1987out:
1988 if (in != F2FS_GOING_DOWN_FULLSYNC)
1989 mnt_drop_write_file(filp);
1990 return ret;
1991}
1992
1993static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1994{
1995 struct inode *inode = file_inode(filp);
1996 struct super_block *sb = inode->i_sb;
1997 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1998 struct fstrim_range range;
1999 int ret;
2000
2001 if (!capable(CAP_SYS_ADMIN))
2002 return -EPERM;
2003
2004 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2005 return -EOPNOTSUPP;
2006
2007 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2008 sizeof(range)))
2009 return -EFAULT;
2010
2011 ret = mnt_want_write_file(filp);
2012 if (ret)
2013 return ret;
2014
2015 range.minlen = max((unsigned int)range.minlen,
2016 q->limits.discard_granularity);
2017 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2018 mnt_drop_write_file(filp);
2019 if (ret < 0)
2020 return ret;
2021
2022 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2023 sizeof(range)))
2024 return -EFAULT;
2025 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2026 return 0;
2027}
2028
2029static bool uuid_is_nonzero(__u8 u[16])
2030{
2031 int i;
2032
2033 for (i = 0; i < 16; i++)
2034 if (u[i])
2035 return true;
2036 return false;
2037}
2038
2039static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2040{
2041 struct inode *inode = file_inode(filp);
2042
2043 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2044 return -EOPNOTSUPP;
2045
2046 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2047
2048 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2049}
2050
2051static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2052{
2053 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2054 return -EOPNOTSUPP;
2055 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2056}
2057
2058static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2059{
2060 struct inode *inode = file_inode(filp);
2061 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2062 int err;
2063
2064 if (!f2fs_sb_has_encrypt(sbi))
2065 return -EOPNOTSUPP;
2066
2067 err = mnt_want_write_file(filp);
2068 if (err)
2069 return err;
2070
2071 down_write(&sbi->sb_lock);
2072
2073 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2074 goto got_it;
2075
2076 /* update superblock with uuid */
2077 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2078
2079 err = f2fs_commit_super(sbi, false);
2080 if (err) {
2081 /* undo new data */
2082 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2083 goto out_err;
2084 }
2085got_it:
2086 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2087 16))
2088 err = -EFAULT;
2089out_err:
2090 up_write(&sbi->sb_lock);
2091 mnt_drop_write_file(filp);
2092 return err;
2093}
2094
2095static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2096{
2097 struct inode *inode = file_inode(filp);
2098 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2099 __u32 sync;
2100 int ret;
2101
2102 if (!capable(CAP_SYS_ADMIN))
2103 return -EPERM;
2104
2105 if (get_user(sync, (__u32 __user *)arg))
2106 return -EFAULT;
2107
2108 if (f2fs_readonly(sbi->sb))
2109 return -EROFS;
2110
2111 ret = mnt_want_write_file(filp);
2112 if (ret)
2113 return ret;
2114
2115 if (!sync) {
2116 if (!mutex_trylock(&sbi->gc_mutex)) {
2117 ret = -EBUSY;
2118 goto out;
2119 }
2120 } else {
2121 mutex_lock(&sbi->gc_mutex);
2122 }
2123
2124 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2125out:
2126 mnt_drop_write_file(filp);
2127 return ret;
2128}
2129
2130static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2131{
2132 struct inode *inode = file_inode(filp);
2133 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2134 struct f2fs_gc_range range;
2135 u64 end;
2136 int ret;
2137
2138 if (!capable(CAP_SYS_ADMIN))
2139 return -EPERM;
2140
2141 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2142 sizeof(range)))
2143 return -EFAULT;
2144
2145 if (f2fs_readonly(sbi->sb))
2146 return -EROFS;
2147
2148 end = range.start + range.len;
2149 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2150 return -EINVAL;
2151 }
2152
2153 ret = mnt_want_write_file(filp);
2154 if (ret)
2155 return ret;
2156
2157do_more:
2158 if (!range.sync) {
2159 if (!mutex_trylock(&sbi->gc_mutex)) {
2160 ret = -EBUSY;
2161 goto out;
2162 }
2163 } else {
2164 mutex_lock(&sbi->gc_mutex);
2165 }
2166
2167 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2168 range.start += BLKS_PER_SEC(sbi);
2169 if (range.start <= end)
2170 goto do_more;
2171out:
2172 mnt_drop_write_file(filp);
2173 return ret;
2174}
2175
2176static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2177{
2178 struct inode *inode = file_inode(filp);
2179 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2180 int ret;
2181
2182 if (!capable(CAP_SYS_ADMIN))
2183 return -EPERM;
2184
2185 if (f2fs_readonly(sbi->sb))
2186 return -EROFS;
2187
2188 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2189 f2fs_msg(sbi->sb, KERN_INFO,
2190 "Skipping Checkpoint. Checkpoints currently disabled.");
2191 return -EINVAL;
2192 }
2193
2194 ret = mnt_want_write_file(filp);
2195 if (ret)
2196 return ret;
2197
2198 ret = f2fs_sync_fs(sbi->sb, 1);
2199
2200 mnt_drop_write_file(filp);
2201 return ret;
2202}
2203
2204static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2205 struct file *filp,
2206 struct f2fs_defragment *range)
2207{
2208 struct inode *inode = file_inode(filp);
2209 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2210 .m_seg_type = NO_CHECK_TYPE ,
2211 .m_may_create = false };
2212 struct extent_info ei = {0, 0, 0};
2213 pgoff_t pg_start, pg_end, next_pgofs;
2214 unsigned int blk_per_seg = sbi->blocks_per_seg;
2215 unsigned int total = 0, sec_num;
2216 block_t blk_end = 0;
2217 bool fragmented = false;
2218 int err;
2219
2220 /* if in-place-update policy is enabled, don't waste time here */
2221 if (f2fs_should_update_inplace(inode, NULL))
2222 return -EINVAL;
2223
2224 pg_start = range->start >> PAGE_SHIFT;
2225 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2226
2227 f2fs_balance_fs(sbi, true);
2228
2229 inode_lock(inode);
2230
2231 /* writeback all dirty pages in the range */
2232 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2233 range->start + range->len - 1);
2234 if (err)
2235 goto out;
2236
2237 /*
2238 * lookup mapping info in extent cache, skip defragmenting if physical
2239 * block addresses are continuous.
2240 */
2241 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2242 if (ei.fofs + ei.len >= pg_end)
2243 goto out;
2244 }
2245
2246 map.m_lblk = pg_start;
2247 map.m_next_pgofs = &next_pgofs;
2248
2249 /*
2250 * lookup mapping info in dnode page cache, skip defragmenting if all
2251 * physical block addresses are continuous even if there are hole(s)
2252 * in logical blocks.
2253 */
2254 while (map.m_lblk < pg_end) {
2255 map.m_len = pg_end - map.m_lblk;
2256 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2257 if (err)
2258 goto out;
2259
2260 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2261 map.m_lblk = next_pgofs;
2262 continue;
2263 }
2264
2265 if (blk_end && blk_end != map.m_pblk)
2266 fragmented = true;
2267
2268 /* record total count of block that we're going to move */
2269 total += map.m_len;
2270
2271 blk_end = map.m_pblk + map.m_len;
2272
2273 map.m_lblk += map.m_len;
2274 }
2275
2276 if (!fragmented)
2277 goto out;
2278
2279 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2280
2281 /*
2282 * make sure there are enough free section for LFS allocation, this can
2283 * avoid defragment running in SSR mode when free section are allocated
2284 * intensively
2285 */
2286 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2287 err = -EAGAIN;
2288 goto out;
2289 }
2290
2291 map.m_lblk = pg_start;
2292 map.m_len = pg_end - pg_start;
2293 total = 0;
2294
2295 while (map.m_lblk < pg_end) {
2296 pgoff_t idx;
2297 int cnt = 0;
2298
2299do_map:
2300 map.m_len = pg_end - map.m_lblk;
2301 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2302 if (err)
2303 goto clear_out;
2304
2305 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2306 map.m_lblk = next_pgofs;
2307 continue;
2308 }
2309
2310 set_inode_flag(inode, FI_DO_DEFRAG);
2311
2312 idx = map.m_lblk;
2313 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2314 struct page *page;
2315
2316 page = f2fs_get_lock_data_page(inode, idx, true);
2317 if (IS_ERR(page)) {
2318 err = PTR_ERR(page);
2319 goto clear_out;
2320 }
2321
2322 set_page_dirty(page);
2323 f2fs_put_page(page, 1);
2324
2325 idx++;
2326 cnt++;
2327 total++;
2328 }
2329
2330 map.m_lblk = idx;
2331
2332 if (idx < pg_end && cnt < blk_per_seg)
2333 goto do_map;
2334
2335 clear_inode_flag(inode, FI_DO_DEFRAG);
2336
2337 err = filemap_fdatawrite(inode->i_mapping);
2338 if (err)
2339 goto out;
2340 }
2341clear_out:
2342 clear_inode_flag(inode, FI_DO_DEFRAG);
2343out:
2344 inode_unlock(inode);
2345 if (!err)
2346 range->len = (u64)total << PAGE_SHIFT;
2347 return err;
2348}
2349
2350static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2351{
2352 struct inode *inode = file_inode(filp);
2353 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2354 struct f2fs_defragment range;
2355 int err;
2356
2357 if (!capable(CAP_SYS_ADMIN))
2358 return -EPERM;
2359
2360 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2361 return -EINVAL;
2362
2363 if (f2fs_readonly(sbi->sb))
2364 return -EROFS;
2365
2366 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2367 sizeof(range)))
2368 return -EFAULT;
2369
2370 /* verify alignment of offset & size */
2371 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2372 return -EINVAL;
2373
2374 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2375 sbi->max_file_blocks))
2376 return -EINVAL;
2377
2378 err = mnt_want_write_file(filp);
2379 if (err)
2380 return err;
2381
2382 err = f2fs_defragment_range(sbi, filp, &range);
2383 mnt_drop_write_file(filp);
2384
2385 f2fs_update_time(sbi, REQ_TIME);
2386 if (err < 0)
2387 return err;
2388
2389 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2390 sizeof(range)))
2391 return -EFAULT;
2392
2393 return 0;
2394}
2395
2396static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2397 struct file *file_out, loff_t pos_out, size_t len)
2398{
2399 struct inode *src = file_inode(file_in);
2400 struct inode *dst = file_inode(file_out);
2401 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2402 size_t olen = len, dst_max_i_size = 0;
2403 size_t dst_osize;
2404 int ret;
2405
2406 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2407 src->i_sb != dst->i_sb)
2408 return -EXDEV;
2409
2410 if (unlikely(f2fs_readonly(src->i_sb)))
2411 return -EROFS;
2412
2413 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2414 return -EINVAL;
2415
2416 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2417 return -EOPNOTSUPP;
2418
2419 if (src == dst) {
2420 if (pos_in == pos_out)
2421 return 0;
2422 if (pos_out > pos_in && pos_out < pos_in + len)
2423 return -EINVAL;
2424 }
2425
2426 inode_lock(src);
2427 if (src != dst) {
2428 ret = -EBUSY;
2429 if (!inode_trylock(dst))
2430 goto out;
2431 }
2432
2433 ret = -EINVAL;
2434 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2435 goto out_unlock;
2436 if (len == 0)
2437 olen = len = src->i_size - pos_in;
2438 if (pos_in + len == src->i_size)
2439 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2440 if (len == 0) {
2441 ret = 0;
2442 goto out_unlock;
2443 }
2444
2445 dst_osize = dst->i_size;
2446 if (pos_out + olen > dst->i_size)
2447 dst_max_i_size = pos_out + olen;
2448
2449 /* verify the end result is block aligned */
2450 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2451 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2452 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2453 goto out_unlock;
2454
2455 ret = f2fs_convert_inline_inode(src);
2456 if (ret)
2457 goto out_unlock;
2458
2459 ret = f2fs_convert_inline_inode(dst);
2460 if (ret)
2461 goto out_unlock;
2462
2463 /* write out all dirty pages from offset */
2464 ret = filemap_write_and_wait_range(src->i_mapping,
2465 pos_in, pos_in + len);
2466 if (ret)
2467 goto out_unlock;
2468
2469 ret = filemap_write_and_wait_range(dst->i_mapping,
2470 pos_out, pos_out + len);
2471 if (ret)
2472 goto out_unlock;
2473
2474 f2fs_balance_fs(sbi, true);
2475
2476 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2477 if (src != dst) {
2478 ret = -EBUSY;
2479 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2480 goto out_src;
2481 }
2482
2483 f2fs_lock_op(sbi);
2484 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2485 pos_out >> F2FS_BLKSIZE_BITS,
2486 len >> F2FS_BLKSIZE_BITS, false);
2487
2488 if (!ret) {
2489 if (dst_max_i_size)
2490 f2fs_i_size_write(dst, dst_max_i_size);
2491 else if (dst_osize != dst->i_size)
2492 f2fs_i_size_write(dst, dst_osize);
2493 }
2494 f2fs_unlock_op(sbi);
2495
2496 if (src != dst)
2497 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2498out_src:
2499 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2500out_unlock:
2501 if (src != dst)
2502 inode_unlock(dst);
2503out:
2504 inode_unlock(src);
2505 return ret;
2506}
2507
2508static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2509{
2510 struct f2fs_move_range range;
2511 struct fd dst;
2512 int err;
2513
2514 if (!(filp->f_mode & FMODE_READ) ||
2515 !(filp->f_mode & FMODE_WRITE))
2516 return -EBADF;
2517
2518 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2519 sizeof(range)))
2520 return -EFAULT;
2521
2522 dst = fdget(range.dst_fd);
2523 if (!dst.file)
2524 return -EBADF;
2525
2526 if (!(dst.file->f_mode & FMODE_WRITE)) {
2527 err = -EBADF;
2528 goto err_out;
2529 }
2530
2531 err = mnt_want_write_file(filp);
2532 if (err)
2533 goto err_out;
2534
2535 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2536 range.pos_out, range.len);
2537
2538 mnt_drop_write_file(filp);
2539 if (err)
2540 goto err_out;
2541
2542 if (copy_to_user((struct f2fs_move_range __user *)arg,
2543 &range, sizeof(range)))
2544 err = -EFAULT;
2545err_out:
2546 fdput(dst);
2547 return err;
2548}
2549
2550static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2551{
2552 struct inode *inode = file_inode(filp);
2553 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2554 struct sit_info *sm = SIT_I(sbi);
2555 unsigned int start_segno = 0, end_segno = 0;
2556 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2557 struct f2fs_flush_device range;
2558 int ret;
2559
2560 if (!capable(CAP_SYS_ADMIN))
2561 return -EPERM;
2562
2563 if (f2fs_readonly(sbi->sb))
2564 return -EROFS;
2565
2566 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2567 return -EINVAL;
2568
2569 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2570 sizeof(range)))
2571 return -EFAULT;
2572
2573 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2574 __is_large_section(sbi)) {
2575 f2fs_msg(sbi->sb, KERN_WARNING,
2576 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2577 range.dev_num, sbi->s_ndevs,
2578 sbi->segs_per_sec);
2579 return -EINVAL;
2580 }
2581
2582 ret = mnt_want_write_file(filp);
2583 if (ret)
2584 return ret;
2585
2586 if (range.dev_num != 0)
2587 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2588 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2589
2590 start_segno = sm->last_victim[FLUSH_DEVICE];
2591 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2592 start_segno = dev_start_segno;
2593 end_segno = min(start_segno + range.segments, dev_end_segno);
2594
2595 while (start_segno < end_segno) {
2596 if (!mutex_trylock(&sbi->gc_mutex)) {
2597 ret = -EBUSY;
2598 goto out;
2599 }
2600 sm->last_victim[GC_CB] = end_segno + 1;
2601 sm->last_victim[GC_GREEDY] = end_segno + 1;
2602 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2603 ret = f2fs_gc(sbi, true, true, start_segno);
2604 if (ret == -EAGAIN)
2605 ret = 0;
2606 else if (ret < 0)
2607 break;
2608 start_segno++;
2609 }
2610out:
2611 mnt_drop_write_file(filp);
2612 return ret;
2613}
2614
2615static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2616{
2617 struct inode *inode = file_inode(filp);
2618 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2619
2620 /* Must validate to set it with SQLite behavior in Android. */
2621 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2622
2623 return put_user(sb_feature, (u32 __user *)arg);
2624}
2625
2626#ifdef CONFIG_QUOTA
2627int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2628{
2629 struct dquot *transfer_to[MAXQUOTAS] = {};
2630 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2631 struct super_block *sb = sbi->sb;
2632 int err = 0;
2633
2634 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2635 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2636 err = __dquot_transfer(inode, transfer_to);
2637 if (err)
2638 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2639 dqput(transfer_to[PRJQUOTA]);
2640 }
2641 return err;
2642}
2643
2644static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2645{
2646 struct inode *inode = file_inode(filp);
2647 struct f2fs_inode_info *fi = F2FS_I(inode);
2648 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2649 struct page *ipage;
2650 kprojid_t kprojid;
2651 int err;
2652
2653 if (!f2fs_sb_has_project_quota(sbi)) {
2654 if (projid != F2FS_DEF_PROJID)
2655 return -EOPNOTSUPP;
2656 else
2657 return 0;
2658 }
2659
2660 if (!f2fs_has_extra_attr(inode))
2661 return -EOPNOTSUPP;
2662
2663 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2664
2665 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2666 return 0;
2667
2668 err = -EPERM;
2669 /* Is it quota file? Do not allow user to mess with it */
2670 if (IS_NOQUOTA(inode))
2671 return err;
2672
2673 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2674 if (IS_ERR(ipage))
2675 return PTR_ERR(ipage);
2676
2677 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2678 i_projid)) {
2679 err = -EOVERFLOW;
2680 f2fs_put_page(ipage, 1);
2681 return err;
2682 }
2683 f2fs_put_page(ipage, 1);
2684
2685 err = dquot_initialize(inode);
2686 if (err)
2687 return err;
2688
2689 f2fs_lock_op(sbi);
2690 err = f2fs_transfer_project_quota(inode, kprojid);
2691 if (err)
2692 goto out_unlock;
2693
2694 F2FS_I(inode)->i_projid = kprojid;
2695 inode->i_ctime = current_time(inode);
2696 f2fs_mark_inode_dirty_sync(inode, true);
2697out_unlock:
2698 f2fs_unlock_op(sbi);
2699 return err;
2700}
2701#else
2702int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2703{
2704 return 0;
2705}
2706
2707static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2708{
2709 if (projid != F2FS_DEF_PROJID)
2710 return -EOPNOTSUPP;
2711 return 0;
2712}
2713#endif
2714
2715/* Transfer internal flags to xflags */
2716static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2717{
2718 __u32 xflags = 0;
2719
2720 if (iflags & F2FS_SYNC_FL)
2721 xflags |= FS_XFLAG_SYNC;
2722 if (iflags & F2FS_IMMUTABLE_FL)
2723 xflags |= FS_XFLAG_IMMUTABLE;
2724 if (iflags & F2FS_APPEND_FL)
2725 xflags |= FS_XFLAG_APPEND;
2726 if (iflags & F2FS_NODUMP_FL)
2727 xflags |= FS_XFLAG_NODUMP;
2728 if (iflags & F2FS_NOATIME_FL)
2729 xflags |= FS_XFLAG_NOATIME;
2730 if (iflags & F2FS_PROJINHERIT_FL)
2731 xflags |= FS_XFLAG_PROJINHERIT;
2732 return xflags;
2733}
2734
2735#define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2736 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2737 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2738
2739/* Transfer xflags flags to internal */
2740static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2741{
2742 unsigned long iflags = 0;
2743
2744 if (xflags & FS_XFLAG_SYNC)
2745 iflags |= F2FS_SYNC_FL;
2746 if (xflags & FS_XFLAG_IMMUTABLE)
2747 iflags |= F2FS_IMMUTABLE_FL;
2748 if (xflags & FS_XFLAG_APPEND)
2749 iflags |= F2FS_APPEND_FL;
2750 if (xflags & FS_XFLAG_NODUMP)
2751 iflags |= F2FS_NODUMP_FL;
2752 if (xflags & FS_XFLAG_NOATIME)
2753 iflags |= F2FS_NOATIME_FL;
2754 if (xflags & FS_XFLAG_PROJINHERIT)
2755 iflags |= F2FS_PROJINHERIT_FL;
2756
2757 return iflags;
2758}
2759
2760static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2761{
2762 struct inode *inode = file_inode(filp);
2763 struct f2fs_inode_info *fi = F2FS_I(inode);
2764 struct fsxattr fa;
2765
2766 memset(&fa, 0, sizeof(struct fsxattr));
2767 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2768 F2FS_FL_USER_VISIBLE);
2769
2770 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2771 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2772 fi->i_projid);
2773
2774 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2775 return -EFAULT;
2776 return 0;
2777}
2778
2779static int f2fs_ioctl_check_project(struct inode *inode, struct fsxattr *fa)
2780{
2781 /*
2782 * Project Quota ID state is only allowed to change from within the init
2783 * namespace. Enforce that restriction only if we are trying to change
2784 * the quota ID state. Everything else is allowed in user namespaces.
2785 */
2786 if (current_user_ns() == &init_user_ns)
2787 return 0;
2788
2789 if (__kprojid_val(F2FS_I(inode)->i_projid) != fa->fsx_projid)
2790 return -EINVAL;
2791
2792 if (F2FS_I(inode)->i_flags & F2FS_PROJINHERIT_FL) {
2793 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT))
2794 return -EINVAL;
2795 } else {
2796 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
2797 return -EINVAL;
2798 }
2799
2800 return 0;
2801}
2802
2803static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2804{
2805 struct inode *inode = file_inode(filp);
2806 struct f2fs_inode_info *fi = F2FS_I(inode);
2807 struct fsxattr fa;
2808 unsigned int flags;
2809 int err;
2810
2811 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2812 return -EFAULT;
2813
2814 /* Make sure caller has proper permission */
2815 if (!inode_owner_or_capable(inode))
2816 return -EACCES;
2817
2818 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2819 return -EOPNOTSUPP;
2820
2821 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2822 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2823 return -EOPNOTSUPP;
2824
2825 err = mnt_want_write_file(filp);
2826 if (err)
2827 return err;
2828
2829 inode_lock(inode);
2830 err = f2fs_ioctl_check_project(inode, &fa);
2831 if (err)
2832 goto out;
2833 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2834 (flags & F2FS_FL_XFLAG_VISIBLE);
2835 err = __f2fs_ioc_setflags(inode, flags);
2836 if (err)
2837 goto out;
2838
2839 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2840out:
2841 inode_unlock(inode);
2842 mnt_drop_write_file(filp);
2843 return err;
2844}
2845
2846int f2fs_pin_file_control(struct inode *inode, bool inc)
2847{
2848 struct f2fs_inode_info *fi = F2FS_I(inode);
2849 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2850
2851 /* Use i_gc_failures for normal file as a risk signal. */
2852 if (inc)
2853 f2fs_i_gc_failures_write(inode,
2854 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2855
2856 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2857 f2fs_msg(sbi->sb, KERN_WARNING,
2858 "%s: Enable GC = ino %lx after %x GC trials\n",
2859 __func__, inode->i_ino,
2860 fi->i_gc_failures[GC_FAILURE_PIN]);
2861 clear_inode_flag(inode, FI_PIN_FILE);
2862 return -EAGAIN;
2863 }
2864 return 0;
2865}
2866
2867static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2868{
2869 struct inode *inode = file_inode(filp);
2870 __u32 pin;
2871 int ret = 0;
2872
2873 if (!inode_owner_or_capable(inode))
2874 return -EACCES;
2875
2876 if (get_user(pin, (__u32 __user *)arg))
2877 return -EFAULT;
2878
2879 if (!S_ISREG(inode->i_mode))
2880 return -EINVAL;
2881
2882 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2883 return -EROFS;
2884
2885 ret = mnt_want_write_file(filp);
2886 if (ret)
2887 return ret;
2888
2889 inode_lock(inode);
2890
2891 if (f2fs_should_update_outplace(inode, NULL)) {
2892 ret = -EINVAL;
2893 goto out;
2894 }
2895
2896 if (!pin) {
2897 clear_inode_flag(inode, FI_PIN_FILE);
2898 f2fs_i_gc_failures_write(inode, 0);
2899 goto done;
2900 }
2901
2902 if (f2fs_pin_file_control(inode, false)) {
2903 ret = -EAGAIN;
2904 goto out;
2905 }
2906 ret = f2fs_convert_inline_inode(inode);
2907 if (ret)
2908 goto out;
2909
2910 set_inode_flag(inode, FI_PIN_FILE);
2911 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2912done:
2913 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2914out:
2915 inode_unlock(inode);
2916 mnt_drop_write_file(filp);
2917 return ret;
2918}
2919
2920static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2921{
2922 struct inode *inode = file_inode(filp);
2923 __u32 pin = 0;
2924
2925 if (is_inode_flag_set(inode, FI_PIN_FILE))
2926 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2927 return put_user(pin, (u32 __user *)arg);
2928}
2929
2930int f2fs_precache_extents(struct inode *inode)
2931{
2932 struct f2fs_inode_info *fi = F2FS_I(inode);
2933 struct f2fs_map_blocks map;
2934 pgoff_t m_next_extent;
2935 loff_t end;
2936 int err;
2937
2938 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2939 return -EOPNOTSUPP;
2940
2941 map.m_lblk = 0;
2942 map.m_next_pgofs = NULL;
2943 map.m_next_extent = &m_next_extent;
2944 map.m_seg_type = NO_CHECK_TYPE;
2945 map.m_may_create = false;
2946 end = F2FS_I_SB(inode)->max_file_blocks;
2947
2948 while (map.m_lblk < end) {
2949 map.m_len = end - map.m_lblk;
2950
2951 down_write(&fi->i_gc_rwsem[WRITE]);
2952 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2953 up_write(&fi->i_gc_rwsem[WRITE]);
2954 if (err)
2955 return err;
2956
2957 map.m_lblk = m_next_extent;
2958 }
2959
2960 return err;
2961}
2962
2963static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2964{
2965 return f2fs_precache_extents(file_inode(filp));
2966}
2967
2968long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2969{
2970 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2971 return -EIO;
2972
2973 switch (cmd) {
2974 case F2FS_IOC_GETFLAGS:
2975 return f2fs_ioc_getflags(filp, arg);
2976 case F2FS_IOC_SETFLAGS:
2977 return f2fs_ioc_setflags(filp, arg);
2978 case F2FS_IOC_GETVERSION:
2979 return f2fs_ioc_getversion(filp, arg);
2980 case F2FS_IOC_START_ATOMIC_WRITE:
2981 return f2fs_ioc_start_atomic_write(filp);
2982 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2983 return f2fs_ioc_commit_atomic_write(filp);
2984 case F2FS_IOC_START_VOLATILE_WRITE:
2985 return f2fs_ioc_start_volatile_write(filp);
2986 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2987 return f2fs_ioc_release_volatile_write(filp);
2988 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2989 return f2fs_ioc_abort_volatile_write(filp);
2990 case F2FS_IOC_SHUTDOWN:
2991 return f2fs_ioc_shutdown(filp, arg);
2992 case FITRIM:
2993 return f2fs_ioc_fitrim(filp, arg);
2994 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2995 return f2fs_ioc_set_encryption_policy(filp, arg);
2996 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2997 return f2fs_ioc_get_encryption_policy(filp, arg);
2998 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2999 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3000 case F2FS_IOC_GARBAGE_COLLECT:
3001 return f2fs_ioc_gc(filp, arg);
3002 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3003 return f2fs_ioc_gc_range(filp, arg);
3004 case F2FS_IOC_WRITE_CHECKPOINT:
3005 return f2fs_ioc_write_checkpoint(filp, arg);
3006 case F2FS_IOC_DEFRAGMENT:
3007 return f2fs_ioc_defragment(filp, arg);
3008 case F2FS_IOC_MOVE_RANGE:
3009 return f2fs_ioc_move_range(filp, arg);
3010 case F2FS_IOC_FLUSH_DEVICE:
3011 return f2fs_ioc_flush_device(filp, arg);
3012 case F2FS_IOC_GET_FEATURES:
3013 return f2fs_ioc_get_features(filp, arg);
3014 case F2FS_IOC_FSGETXATTR:
3015 return f2fs_ioc_fsgetxattr(filp, arg);
3016 case F2FS_IOC_FSSETXATTR:
3017 return f2fs_ioc_fssetxattr(filp, arg);
3018 case F2FS_IOC_GET_PIN_FILE:
3019 return f2fs_ioc_get_pin_file(filp, arg);
3020 case F2FS_IOC_SET_PIN_FILE:
3021 return f2fs_ioc_set_pin_file(filp, arg);
3022 case F2FS_IOC_PRECACHE_EXTENTS:
3023 return f2fs_ioc_precache_extents(filp, arg);
3024 default:
3025 return -ENOTTY;
3026 }
3027}
3028
3029static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3030{
3031 struct file *file = iocb->ki_filp;
3032 struct inode *inode = file_inode(file);
3033 ssize_t ret;
3034
3035 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3036 return -EIO;
3037
3038 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
3039 return -EINVAL;
3040
3041 if (!inode_trylock(inode)) {
3042 if (iocb->ki_flags & IOCB_NOWAIT)
3043 return -EAGAIN;
3044 inode_lock(inode);
3045 }
3046
3047 ret = generic_write_checks(iocb, from);
3048 if (ret > 0) {
3049 bool preallocated = false;
3050 size_t target_size = 0;
3051 int err;
3052
3053 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3054 set_inode_flag(inode, FI_NO_PREALLOC);
3055
3056 if ((iocb->ki_flags & IOCB_NOWAIT) &&
3057 (iocb->ki_flags & IOCB_DIRECT)) {
3058 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3059 iov_iter_count(from)) ||
3060 f2fs_has_inline_data(inode) ||
3061 f2fs_force_buffered_io(inode,
3062 iocb, from)) {
3063 clear_inode_flag(inode,
3064 FI_NO_PREALLOC);
3065 inode_unlock(inode);
3066 return -EAGAIN;
3067 }
3068
3069 } else {
3070 preallocated = true;
3071 target_size = iocb->ki_pos + iov_iter_count(from);
3072
3073 err = f2fs_preallocate_blocks(iocb, from);
3074 if (err) {
3075 clear_inode_flag(inode, FI_NO_PREALLOC);
3076 inode_unlock(inode);
3077 return err;
3078 }
3079 }
3080 ret = __generic_file_write_iter(iocb, from);
3081 clear_inode_flag(inode, FI_NO_PREALLOC);
3082
3083 /* if we couldn't write data, we should deallocate blocks. */
3084 if (preallocated && i_size_read(inode) < target_size)
3085 f2fs_truncate(inode);
3086
3087 if (ret > 0)
3088 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3089 }
3090 inode_unlock(inode);
3091
3092 if (ret > 0)
3093 ret = generic_write_sync(iocb, ret);
3094 return ret;
3095}
3096
3097#ifdef CONFIG_COMPAT
3098long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3099{
3100 switch (cmd) {
3101 case F2FS_IOC32_GETFLAGS:
3102 cmd = F2FS_IOC_GETFLAGS;
3103 break;
3104 case F2FS_IOC32_SETFLAGS:
3105 cmd = F2FS_IOC_SETFLAGS;
3106 break;
3107 case F2FS_IOC32_GETVERSION:
3108 cmd = F2FS_IOC_GETVERSION;
3109 break;
3110 case F2FS_IOC_START_ATOMIC_WRITE:
3111 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3112 case F2FS_IOC_START_VOLATILE_WRITE:
3113 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3114 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3115 case F2FS_IOC_SHUTDOWN:
3116 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3117 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3118 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3119 case F2FS_IOC_GARBAGE_COLLECT:
3120 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3121 case F2FS_IOC_WRITE_CHECKPOINT:
3122 case F2FS_IOC_DEFRAGMENT:
3123 case F2FS_IOC_MOVE_RANGE:
3124 case F2FS_IOC_FLUSH_DEVICE:
3125 case F2FS_IOC_GET_FEATURES:
3126 case F2FS_IOC_FSGETXATTR:
3127 case F2FS_IOC_FSSETXATTR:
3128 case F2FS_IOC_GET_PIN_FILE:
3129 case F2FS_IOC_SET_PIN_FILE:
3130 case F2FS_IOC_PRECACHE_EXTENTS:
3131 break;
3132 default:
3133 return -ENOIOCTLCMD;
3134 }
3135 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3136}
3137#endif
3138
3139const struct file_operations f2fs_file_operations = {
3140 .llseek = f2fs_llseek,
3141 .read_iter = generic_file_read_iter,
3142 .write_iter = f2fs_file_write_iter,
3143 .open = f2fs_file_open,
3144 .release = f2fs_release_file,
3145 .mmap = f2fs_file_mmap,
3146 .flush = f2fs_file_flush,
3147 .fsync = f2fs_sync_file,
3148 .fallocate = f2fs_fallocate,
3149 .unlocked_ioctl = f2fs_ioctl,
3150#ifdef CONFIG_COMPAT
3151 .compat_ioctl = f2fs_compat_ioctl,
3152#endif
3153 .splice_read = generic_file_splice_read,
3154 .splice_write = iter_file_splice_write,
3155};
Linux 6.x block layer and io_uring tree(s)