projects / GitHub / MotorolaMobilityLLC / kernel-slsi.git / blob
? search:


23cbc5019f06f5e76d76895fc65bf6051ee62afe
[GitHub/MotorolaMobilityLLC/kernel-slsi.git] / fs / f2fs / file.c
1 /*
2 * fs/f2fs/file.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11 #include <linux/fs.h>
12 #include <linux/f2fs_fs.h>
13 #include <linux/stat.h>
14 #include <linux/buffer_head.h>
15 #include <linux/writeback.h>
16 #include <linux/blkdev.h>
17 #include <linux/falloc.h>
18 #include <linux/types.h>
19 #include <linux/compat.h>
20 #include <linux/uaccess.h>
21 #include <linux/mount.h>
22 #include <linux/pagevec.h>
23 #include <linux/uio.h>
24 #include <linux/uuid.h>
25 #include <linux/file.h>
26
27 #include "f2fs.h"
28 #include "node.h"
29 #include "segment.h"
30 #include "xattr.h"
31 #include "acl.h"
32 #include "gc.h"
33 #include "trace.h"
34 #include <trace/events/f2fs.h>
35
36 static int f2fs_filemap_fault(struct vm_fault *vmf)
37 {
38 struct inode *inode = file_inode(vmf->vma->vm_file);
39 int err;
40
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 err = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
44
45 return err;
46 }
47
48 static int f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 {
50 struct page *page = vmf->page;
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 struct dnode_of_data dn;
54 int err;
55
56 if (unlikely(f2fs_cp_error(sbi))) {
57 err = -EIO;
58 goto err;
59 }
60
61 sb_start_pagefault(inode->i_sb);
62
63 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
64
65 /* block allocation */
66 f2fs_lock_op(sbi);
67 set_new_dnode(&dn, inode, NULL, NULL, 0);
68 err = f2fs_reserve_block(&dn, page->index);
69 if (err) {
70 f2fs_unlock_op(sbi);
71 goto out;
72 }
73 f2fs_put_dnode(&dn);
74 f2fs_unlock_op(sbi);
75
76 f2fs_balance_fs(sbi, dn.node_changed);
77
78 file_update_time(vmf->vma->vm_file);
79 down_read(&F2FS_I(inode)->i_mmap_sem);
80 lock_page(page);
81 if (unlikely(page->mapping != inode->i_mapping ||
82 page_offset(page) > i_size_read(inode) ||
83 !PageUptodate(page))) {
84 unlock_page(page);
85 err = -EFAULT;
86 goto out_sem;
87 }
88
89 /*
90 * check to see if the page is mapped already (no holes)
91 */
92 if (PageMappedToDisk(page))
93 goto mapped;
94
95 /* page is wholly or partially inside EOF */
96 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
97 i_size_read(inode)) {
98 unsigned offset;
99 offset = i_size_read(inode) & ~PAGE_MASK;
100 zero_user_segment(page, offset, PAGE_SIZE);
101 }
102 set_page_dirty(page);
103 if (!PageUptodate(page))
104 SetPageUptodate(page);
105
106 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
107
108 trace_f2fs_vm_page_mkwrite(page, DATA);
109 mapped:
110 /* fill the page */
111 f2fs_wait_on_page_writeback(page, DATA, false);
112
113 /* wait for GCed page writeback via META_MAPPING */
114 if (f2fs_post_read_required(inode))
115 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
116
117 out_sem:
118 up_read(&F2FS_I(inode)->i_mmap_sem);
119 out:
120 sb_end_pagefault(inode->i_sb);
121 f2fs_update_time(sbi, REQ_TIME);
122 err:
123 return block_page_mkwrite_return(err);
124 }
125
126 static const struct vm_operations_struct f2fs_file_vm_ops = {
127 .fault = f2fs_filemap_fault,
128 .map_pages = filemap_map_pages,
129 .page_mkwrite = f2fs_vm_page_mkwrite,
130 };
131
132 static int get_parent_ino(struct inode *inode, nid_t *pino)
133 {
134 struct dentry *dentry;
135
136 inode = igrab(inode);
137 dentry = d_find_any_alias(inode);
138 iput(inode);
139 if (!dentry)
140 return 0;
141
142 *pino = parent_ino(dentry);
143 dput(dentry);
144 return 1;
145 }
146
147 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
148 {
149 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
150 enum cp_reason_type cp_reason = CP_NO_NEEDED;
151
152 if (!S_ISREG(inode->i_mode))
153 cp_reason = CP_NON_REGULAR;
154 else if (inode->i_nlink != 1)
155 cp_reason = CP_HARDLINK;
156 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
157 cp_reason = CP_SB_NEED_CP;
158 else if (file_wrong_pino(inode))
159 cp_reason = CP_WRONG_PINO;
160 else if (!space_for_roll_forward(sbi))
161 cp_reason = CP_NO_SPC_ROLL;
162 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
163 cp_reason = CP_NODE_NEED_CP;
164 else if (test_opt(sbi, FASTBOOT))
165 cp_reason = CP_FASTBOOT_MODE;
166 else if (F2FS_OPTION(sbi).active_logs == 2)
167 cp_reason = CP_SPEC_LOG_NUM;
168 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
169 need_dentry_mark(sbi, inode->i_ino) &&
170 exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
171 cp_reason = CP_RECOVER_DIR;
172
173 return cp_reason;
174 }
175
176 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
177 {
178 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
179 bool ret = false;
180 /* But we need to avoid that there are some inode updates */
181 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
182 ret = true;
183 f2fs_put_page(i, 0);
184 return ret;
185 }
186
187 static void try_to_fix_pino(struct inode *inode)
188 {
189 struct f2fs_inode_info *fi = F2FS_I(inode);
190 nid_t pino;
191
192 down_write(&fi->i_sem);
193 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
194 get_parent_ino(inode, &pino)) {
195 f2fs_i_pino_write(inode, pino);
196 file_got_pino(inode);
197 }
198 up_write(&fi->i_sem);
199 }
200
201 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
202 int datasync, bool atomic)
203 {
204 struct inode *inode = file->f_mapping->host;
205 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
206 nid_t ino = inode->i_ino;
207 int ret = 0;
208 enum cp_reason_type cp_reason = 0;
209 struct writeback_control wbc = {
210 .sync_mode = WB_SYNC_ALL,
211 .nr_to_write = LONG_MAX,
212 .for_reclaim = 0,
213 };
214
215 if (unlikely(f2fs_readonly(inode->i_sb)))
216 return 0;
217
218 trace_f2fs_sync_file_enter(inode);
219
220 if (S_ISDIR(inode->i_mode))
221 goto go_write;
222
223 /* if fdatasync is triggered, let's do in-place-update */
224 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
225 set_inode_flag(inode, FI_NEED_IPU);
226 ret = file_write_and_wait_range(file, start, end);
227 clear_inode_flag(inode, FI_NEED_IPU);
228
229 if (ret) {
230 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
231 return ret;
232 }
233
234 /* if the inode is dirty, let's recover all the time */
235 if (!f2fs_skip_inode_update(inode, datasync)) {
236 f2fs_write_inode(inode, NULL);
237 goto go_write;
238 }
239
240 /*
241 * if there is no written data, don't waste time to write recovery info.
242 */
243 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
244 !exist_written_data(sbi, ino, APPEND_INO)) {
245
246 /* it may call write_inode just prior to fsync */
247 if (need_inode_page_update(sbi, ino))
248 goto go_write;
249
250 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
251 exist_written_data(sbi, ino, UPDATE_INO))
252 goto flush_out;
253 goto out;
254 }
255 go_write:
256 /*
257 * Both of fdatasync() and fsync() are able to be recovered from
258 * sudden-power-off.
259 */
260 down_read(&F2FS_I(inode)->i_sem);
261 cp_reason = need_do_checkpoint(inode);
262 up_read(&F2FS_I(inode)->i_sem);
263
264 if (cp_reason) {
265 /* all the dirty node pages should be flushed for POR */
266 ret = f2fs_sync_fs(inode->i_sb, 1);
267
268 /*
269 * We've secured consistency through sync_fs. Following pino
270 * will be used only for fsynced inodes after checkpoint.
271 */
272 try_to_fix_pino(inode);
273 clear_inode_flag(inode, FI_APPEND_WRITE);
274 clear_inode_flag(inode, FI_UPDATE_WRITE);
275 goto out;
276 }
277 sync_nodes:
278 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
279 if (ret)
280 goto out;
281
282 /* if cp_error was enabled, we should avoid infinite loop */
283 if (unlikely(f2fs_cp_error(sbi))) {
284 ret = -EIO;
285 goto out;
286 }
287
288 if (need_inode_block_update(sbi, ino)) {
289 f2fs_mark_inode_dirty_sync(inode, true);
290 f2fs_write_inode(inode, NULL);
291 goto sync_nodes;
292 }
293
294 /*
295 * If it's atomic_write, it's just fine to keep write ordering. So
296 * here we don't need to wait for node write completion, since we use
297 * node chain which serializes node blocks. If one of node writes are
298 * reordered, we can see simply broken chain, resulting in stopping
299 * roll-forward recovery. It means we'll recover all or none node blocks
300 * given fsync mark.
301 */
302 if (!atomic) {
303 ret = wait_on_node_pages_writeback(sbi, ino);
304 if (ret)
305 goto out;
306 }
307
308 /* once recovery info is written, don't need to tack this */
309 remove_ino_entry(sbi, ino, APPEND_INO);
310 clear_inode_flag(inode, FI_APPEND_WRITE);
311 flush_out:
312 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
313 ret = f2fs_issue_flush(sbi, inode->i_ino);
314 if (!ret) {
315 remove_ino_entry(sbi, ino, UPDATE_INO);
316 clear_inode_flag(inode, FI_UPDATE_WRITE);
317 remove_ino_entry(sbi, ino, FLUSH_INO);
318 }
319 f2fs_update_time(sbi, REQ_TIME);
320 out:
321 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
322 f2fs_trace_ios(NULL, 1);
323 return ret;
324 }
325
326 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
327 {
328 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
329 return -EIO;
330 return f2fs_do_sync_file(file, start, end, datasync, false);
331 }
332
333 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
334 pgoff_t pgofs, int whence)
335 {
336 struct pagevec pvec;
337 int nr_pages;
338
339 if (whence != SEEK_DATA)
340 return 0;
341
342 /* find first dirty page index */
343 pagevec_init(&pvec, 0);
344 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
345 PAGECACHE_TAG_DIRTY, 1);
346 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
347 pagevec_release(&pvec);
348 return pgofs;
349 }
350
351 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
352 pgoff_t dirty, pgoff_t pgofs, int whence)
353 {
354 switch (whence) {
355 case SEEK_DATA:
356 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
357 is_valid_data_blkaddr(sbi, blkaddr))
358 return true;
359 break;
360 case SEEK_HOLE:
361 if (blkaddr == NULL_ADDR)
362 return true;
363 break;
364 }
365 return false;
366 }
367
368 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
369 {
370 struct inode *inode = file->f_mapping->host;
371 loff_t maxbytes = inode->i_sb->s_maxbytes;
372 struct dnode_of_data dn;
373 pgoff_t pgofs, end_offset, dirty;
374 loff_t data_ofs = offset;
375 loff_t isize;
376 int err = 0;
377
378 inode_lock(inode);
379
380 isize = i_size_read(inode);
381 if (offset >= isize)
382 goto fail;
383
384 /* handle inline data case */
385 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
386 if (whence == SEEK_HOLE)
387 data_ofs = isize;
388 goto found;
389 }
390
391 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
392
393 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
394
395 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
396 set_new_dnode(&dn, inode, NULL, NULL, 0);
397 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
398 if (err && err != -ENOENT) {
399 goto fail;
400 } else if (err == -ENOENT) {
401 /* direct node does not exists */
402 if (whence == SEEK_DATA) {
403 pgofs = get_next_page_offset(&dn, pgofs);
404 continue;
405 } else {
406 goto found;
407 }
408 }
409
410 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
411
412 /* find data/hole in dnode block */
413 for (; dn.ofs_in_node < end_offset;
414 dn.ofs_in_node++, pgofs++,
415 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
416 block_t blkaddr;
417 blkaddr = datablock_addr(dn.inode,
418 dn.node_page, dn.ofs_in_node);
419
420 if (__is_valid_data_blkaddr(blkaddr) &&
421 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
422 blkaddr, DATA_GENERIC)) {
423 f2fs_put_dnode(&dn);
424 goto fail;
425 }
426
427 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
428 pgofs, whence)) {
429 f2fs_put_dnode(&dn);
430 goto found;
431 }
432 }
433 f2fs_put_dnode(&dn);
434 }
435
436 if (whence == SEEK_DATA)
437 goto fail;
438 found:
439 if (whence == SEEK_HOLE && data_ofs > isize)
440 data_ofs = isize;
441 inode_unlock(inode);
442 return vfs_setpos(file, data_ofs, maxbytes);
443 fail:
444 inode_unlock(inode);
445 return -ENXIO;
446 }
447
448 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
449 {
450 struct inode *inode = file->f_mapping->host;
451 loff_t maxbytes = inode->i_sb->s_maxbytes;
452
453 switch (whence) {
454 case SEEK_SET:
455 case SEEK_CUR:
456 case SEEK_END:
457 return generic_file_llseek_size(file, offset, whence,
458 maxbytes, i_size_read(inode));
459 case SEEK_DATA:
460 case SEEK_HOLE:
461 if (offset < 0)
462 return -ENXIO;
463 return f2fs_seek_block(file, offset, whence);
464 }
465
466 return -EINVAL;
467 }
468
469 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
470 {
471 struct inode *inode = file_inode(file);
472 int err;
473
474 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
475 return -EIO;
476
477 /* we don't need to use inline_data strictly */
478 err = f2fs_convert_inline_inode(inode);
479 if (err)
480 return err;
481
482 file_accessed(file);
483 vma->vm_ops = &f2fs_file_vm_ops;
484 return 0;
485 }
486
487 static int f2fs_file_open(struct inode *inode, struct file *filp)
488 {
489 int err = fscrypt_file_open(inode, filp);
490
491 if (err)
492 return err;
493
494 filp->f_mode |= FMODE_NOWAIT;
495
496 return dquot_file_open(inode, filp);
497 }
498
499 void truncate_data_blocks_range(struct dnode_of_data *dn, int count)
500 {
501 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
502 struct f2fs_node *raw_node;
503 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
504 __le32 *addr;
505 int base = 0;
506
507 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
508 base = get_extra_isize(dn->inode);
509
510 raw_node = F2FS_NODE(dn->node_page);
511 addr = blkaddr_in_node(raw_node) + base + ofs;
512
513 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
514 block_t blkaddr = le32_to_cpu(*addr);
515 if (blkaddr == NULL_ADDR)
516 continue;
517
518 dn->data_blkaddr = NULL_ADDR;
519 set_data_blkaddr(dn);
520
521 if (__is_valid_data_blkaddr(blkaddr) &&
522 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
523 continue;
524
525 invalidate_blocks(sbi, blkaddr);
526 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
527 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
528 nr_free++;
529 }
530
531 if (nr_free) {
532 pgoff_t fofs;
533 /*
534 * once we invalidate valid blkaddr in range [ofs, ofs + count],
535 * we will invalidate all blkaddr in the whole range.
536 */
537 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
538 dn->inode) + ofs;
539 f2fs_update_extent_cache_range(dn, fofs, 0, len);
540 dec_valid_block_count(sbi, dn->inode, nr_free);
541 }
542 dn->ofs_in_node = ofs;
543
544 f2fs_update_time(sbi, REQ_TIME);
545 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
546 dn->ofs_in_node, nr_free);
547 }
548
549 void truncate_data_blocks(struct dnode_of_data *dn)
550 {
551 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
552 }
553
554 static int truncate_partial_data_page(struct inode *inode, u64 from,
555 bool cache_only)
556 {
557 unsigned offset = from & (PAGE_SIZE - 1);
558 pgoff_t index = from >> PAGE_SHIFT;
559 struct address_space *mapping = inode->i_mapping;
560 struct page *page;
561
562 if (!offset && !cache_only)
563 return 0;
564
565 if (cache_only) {
566 page = find_lock_page(mapping, index);
567 if (page && PageUptodate(page))
568 goto truncate_out;
569 f2fs_put_page(page, 1);
570 return 0;
571 }
572
573 page = get_lock_data_page(inode, index, true);
574 if (IS_ERR(page))
575 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
576 truncate_out:
577 f2fs_wait_on_page_writeback(page, DATA, true);
578 zero_user(page, offset, PAGE_SIZE - offset);
579
580 /* An encrypted inode should have a key and truncate the last page. */
581 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
582 if (!cache_only)
583 set_page_dirty(page);
584 f2fs_put_page(page, 1);
585 return 0;
586 }
587
588 int truncate_blocks(struct inode *inode, u64 from, bool lock)
589 {
590 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
591 struct dnode_of_data dn;
592 pgoff_t free_from;
593 int count = 0, err = 0;
594 struct page *ipage;
595 bool truncate_page = false;
596
597 trace_f2fs_truncate_blocks_enter(inode, from);
598
599 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
600
601 if (free_from >= sbi->max_file_blocks)
602 goto free_partial;
603
604 if (lock)
605 f2fs_lock_op(sbi);
606
607 ipage = get_node_page(sbi, inode->i_ino);
608 if (IS_ERR(ipage)) {
609 err = PTR_ERR(ipage);
610 goto out;
611 }
612
613 if (f2fs_has_inline_data(inode)) {
614 truncate_inline_inode(inode, ipage, from);
615 f2fs_put_page(ipage, 1);
616 truncate_page = true;
617 goto out;
618 }
619
620 set_new_dnode(&dn, inode, ipage, NULL, 0);
621 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
622 if (err) {
623 if (err == -ENOENT)
624 goto free_next;
625 goto out;
626 }
627
628 count = ADDRS_PER_PAGE(dn.node_page, inode);
629
630 count -= dn.ofs_in_node;
631 f2fs_bug_on(sbi, count < 0);
632
633 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
634 truncate_data_blocks_range(&dn, count);
635 free_from += count;
636 }
637
638 f2fs_put_dnode(&dn);
639 free_next:
640 err = truncate_inode_blocks(inode, free_from);
641 out:
642 if (lock)
643 f2fs_unlock_op(sbi);
644 free_partial:
645 /* lastly zero out the first data page */
646 if (!err)
647 err = truncate_partial_data_page(inode, from, truncate_page);
648
649 trace_f2fs_truncate_blocks_exit(inode, err);
650 return err;
651 }
652
653 int f2fs_truncate(struct inode *inode)
654 {
655 int err;
656
657 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
658 return -EIO;
659
660 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
661 S_ISLNK(inode->i_mode)))
662 return 0;
663
664 trace_f2fs_truncate(inode);
665
666 #ifdef CONFIG_F2FS_FAULT_INJECTION
667 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
668 f2fs_show_injection_info(FAULT_TRUNCATE);
669 return -EIO;
670 }
671 #endif
672 /* we should check inline_data size */
673 if (!f2fs_may_inline_data(inode)) {
674 err = f2fs_convert_inline_inode(inode);
675 if (err)
676 return err;
677 }
678
679 err = truncate_blocks(inode, i_size_read(inode), true);
680 if (err)
681 return err;
682
683 inode->i_mtime = inode->i_ctime = current_time(inode);
684 f2fs_mark_inode_dirty_sync(inode, false);
685 return 0;
686 }
687
688 int f2fs_getattr(const struct path *path, struct kstat *stat,
689 u32 request_mask, unsigned int query_flags)
690 {
691 struct inode *inode = d_inode(path->dentry);
692 struct f2fs_inode_info *fi = F2FS_I(inode);
693 struct f2fs_inode *ri;
694 unsigned int flags;
695
696 if (f2fs_has_extra_attr(inode) &&
697 f2fs_sb_has_inode_crtime(inode->i_sb) &&
698 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
699 stat->result_mask |= STATX_BTIME;
700 stat->btime.tv_sec = fi->i_crtime.tv_sec;
701 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
702 }
703
704 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
705 if (flags & FS_APPEND_FL)
706 stat->attributes |= STATX_ATTR_APPEND;
707 if (flags & FS_COMPR_FL)
708 stat->attributes |= STATX_ATTR_COMPRESSED;
709 if (f2fs_encrypted_inode(inode))
710 stat->attributes |= STATX_ATTR_ENCRYPTED;
711 if (flags & FS_IMMUTABLE_FL)
712 stat->attributes |= STATX_ATTR_IMMUTABLE;
713 if (flags & FS_NODUMP_FL)
714 stat->attributes |= STATX_ATTR_NODUMP;
715
716 stat->attributes_mask |= (STATX_ATTR_APPEND |
717 STATX_ATTR_COMPRESSED |
718 STATX_ATTR_ENCRYPTED |
719 STATX_ATTR_IMMUTABLE |
720 STATX_ATTR_NODUMP);
721
722 generic_fillattr(inode, stat);
723
724 /* we need to show initial sectors used for inline_data/dentries */
725 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
726 f2fs_has_inline_dentry(inode))
727 stat->blocks += (stat->size + 511) >> 9;
728
729 return 0;
730 }
731
732 #ifdef CONFIG_F2FS_FS_POSIX_ACL
733 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
734 {
735 unsigned int ia_valid = attr->ia_valid;
736
737 if (ia_valid & ATTR_UID)
738 inode->i_uid = attr->ia_uid;
739 if (ia_valid & ATTR_GID)
740 inode->i_gid = attr->ia_gid;
741 if (ia_valid & ATTR_ATIME)
742 inode->i_atime = timespec_trunc(attr->ia_atime,
743 inode->i_sb->s_time_gran);
744 if (ia_valid & ATTR_MTIME)
745 inode->i_mtime = timespec_trunc(attr->ia_mtime,
746 inode->i_sb->s_time_gran);
747 if (ia_valid & ATTR_CTIME)
748 inode->i_ctime = timespec_trunc(attr->ia_ctime,
749 inode->i_sb->s_time_gran);
750 if (ia_valid & ATTR_MODE) {
751 umode_t mode = attr->ia_mode;
752
753 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
754 mode &= ~S_ISGID;
755 set_acl_inode(inode, mode);
756 }
757 }
758 #else
759 #define __setattr_copy setattr_copy
760 #endif
761
762 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
763 {
764 struct inode *inode = d_inode(dentry);
765 int err;
766 bool size_changed = false;
767
768 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
769 return -EIO;
770
771 err = setattr_prepare(dentry, attr);
772 if (err)
773 return err;
774
775 err = fscrypt_prepare_setattr(dentry, attr);
776 if (err)
777 return err;
778
779 if (is_quota_modification(inode, attr)) {
780 err = dquot_initialize(inode);
781 if (err)
782 return err;
783 }
784 if ((attr->ia_valid & ATTR_UID &&
785 !uid_eq(attr->ia_uid, inode->i_uid)) ||
786 (attr->ia_valid & ATTR_GID &&
787 !gid_eq(attr->ia_gid, inode->i_gid))) {
788 err = dquot_transfer(inode, attr);
789 if (err)
790 return err;
791 }
792
793 if (attr->ia_valid & ATTR_SIZE) {
794 if (attr->ia_size <= i_size_read(inode)) {
795 down_write(&F2FS_I(inode)->i_mmap_sem);
796 truncate_setsize(inode, attr->ia_size);
797 err = f2fs_truncate(inode);
798 up_write(&F2FS_I(inode)->i_mmap_sem);
799 if (err)
800 return err;
801 } else {
802 /*
803 * do not trim all blocks after i_size if target size is
804 * larger than i_size.
805 */
806 down_write(&F2FS_I(inode)->i_mmap_sem);
807 truncate_setsize(inode, attr->ia_size);
808 up_write(&F2FS_I(inode)->i_mmap_sem);
809
810 /* should convert inline inode here */
811 if (!f2fs_may_inline_data(inode)) {
812 err = f2fs_convert_inline_inode(inode);
813 if (err)
814 return err;
815 }
816 inode->i_mtime = inode->i_ctime = current_time(inode);
817 }
818
819 down_write(&F2FS_I(inode)->i_sem);
820 F2FS_I(inode)->last_disk_size = i_size_read(inode);
821 up_write(&F2FS_I(inode)->i_sem);
822
823 size_changed = true;
824 }
825
826 __setattr_copy(inode, attr);
827
828 if (attr->ia_valid & ATTR_MODE) {
829 err = posix_acl_chmod(inode, get_inode_mode(inode));
830 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
831 inode->i_mode = F2FS_I(inode)->i_acl_mode;
832 clear_inode_flag(inode, FI_ACL_MODE);
833 }
834 }
835
836 /* file size may changed here */
837 f2fs_mark_inode_dirty_sync(inode, size_changed);
838
839 /* inode change will produce dirty node pages flushed by checkpoint */
840 f2fs_balance_fs(F2FS_I_SB(inode), true);
841
842 return err;
843 }
844
845 const struct inode_operations f2fs_file_inode_operations = {
846 .getattr = f2fs_getattr,
847 .setattr = f2fs_setattr,
848 .get_acl = f2fs_get_acl,
849 .set_acl = f2fs_set_acl,
850 #ifdef CONFIG_F2FS_FS_XATTR
851 .listxattr = f2fs_listxattr,
852 #endif
853 .fiemap = f2fs_fiemap,
854 };
855
856 static int fill_zero(struct inode *inode, pgoff_t index,
857 loff_t start, loff_t len)
858 {
859 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
860 struct page *page;
861
862 if (!len)
863 return 0;
864
865 f2fs_balance_fs(sbi, true);
866
867 f2fs_lock_op(sbi);
868 page = get_new_data_page(inode, NULL, index, false);
869 f2fs_unlock_op(sbi);
870
871 if (IS_ERR(page))
872 return PTR_ERR(page);
873
874 f2fs_wait_on_page_writeback(page, DATA, true);
875 zero_user(page, start, len);
876 set_page_dirty(page);
877 f2fs_put_page(page, 1);
878 return 0;
879 }
880
881 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
882 {
883 int err;
884
885 while (pg_start < pg_end) {
886 struct dnode_of_data dn;
887 pgoff_t end_offset, count;
888
889 set_new_dnode(&dn, inode, NULL, NULL, 0);
890 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
891 if (err) {
892 if (err == -ENOENT) {
893 pg_start = get_next_page_offset(&dn, pg_start);
894 continue;
895 }
896 return err;
897 }
898
899 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
900 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
901
902 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
903
904 truncate_data_blocks_range(&dn, count);
905 f2fs_put_dnode(&dn);
906
907 pg_start += count;
908 }
909 return 0;
910 }
911
912 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
913 {
914 pgoff_t pg_start, pg_end;
915 loff_t off_start, off_end;
916 int ret;
917
918 ret = f2fs_convert_inline_inode(inode);
919 if (ret)
920 return ret;
921
922 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
923 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
924
925 off_start = offset & (PAGE_SIZE - 1);
926 off_end = (offset + len) & (PAGE_SIZE - 1);
927
928 if (pg_start == pg_end) {
929 ret = fill_zero(inode, pg_start, off_start,
930 off_end - off_start);
931 if (ret)
932 return ret;
933 } else {
934 if (off_start) {
935 ret = fill_zero(inode, pg_start++, off_start,
936 PAGE_SIZE - off_start);
937 if (ret)
938 return ret;
939 }
940 if (off_end) {
941 ret = fill_zero(inode, pg_end, 0, off_end);
942 if (ret)
943 return ret;
944 }
945
946 if (pg_start < pg_end) {
947 struct address_space *mapping = inode->i_mapping;
948 loff_t blk_start, blk_end;
949 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
950
951 f2fs_balance_fs(sbi, true);
952
953 blk_start = (loff_t)pg_start << PAGE_SHIFT;
954 blk_end = (loff_t)pg_end << PAGE_SHIFT;
955 down_write(&F2FS_I(inode)->i_mmap_sem);
956 truncate_inode_pages_range(mapping, blk_start,
957 blk_end - 1);
958
959 f2fs_lock_op(sbi);
960 ret = truncate_hole(inode, pg_start, pg_end);
961 f2fs_unlock_op(sbi);
962 up_write(&F2FS_I(inode)->i_mmap_sem);
963 }
964 }
965
966 return ret;
967 }
968
969 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
970 int *do_replace, pgoff_t off, pgoff_t len)
971 {
972 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
973 struct dnode_of_data dn;
974 int ret, done, i;
975
976 next_dnode:
977 set_new_dnode(&dn, inode, NULL, NULL, 0);
978 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
979 if (ret && ret != -ENOENT) {
980 return ret;
981 } else if (ret == -ENOENT) {
982 if (dn.max_level == 0)
983 return -ENOENT;
984 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
985 blkaddr += done;
986 do_replace += done;
987 goto next;
988 }
989
990 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
991 dn.ofs_in_node, len);
992 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
993 *blkaddr = datablock_addr(dn.inode,
994 dn.node_page, dn.ofs_in_node);
995 if (!is_checkpointed_data(sbi, *blkaddr)) {
996
997 if (test_opt(sbi, LFS)) {
998 f2fs_put_dnode(&dn);
999 return -ENOTSUPP;
1000 }
1001
1002 /* do not invalidate this block address */
1003 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1004 *do_replace = 1;
1005 }
1006 }
1007 f2fs_put_dnode(&dn);
1008 next:
1009 len -= done;
1010 off += done;
1011 if (len)
1012 goto next_dnode;
1013 return 0;
1014 }
1015
1016 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1017 int *do_replace, pgoff_t off, int len)
1018 {
1019 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1020 struct dnode_of_data dn;
1021 int ret, i;
1022
1023 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1024 if (*do_replace == 0)
1025 continue;
1026
1027 set_new_dnode(&dn, inode, NULL, NULL, 0);
1028 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1029 if (ret) {
1030 dec_valid_block_count(sbi, inode, 1);
1031 invalidate_blocks(sbi, *blkaddr);
1032 } else {
1033 f2fs_update_data_blkaddr(&dn, *blkaddr);
1034 }
1035 f2fs_put_dnode(&dn);
1036 }
1037 return 0;
1038 }
1039
1040 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1041 block_t *blkaddr, int *do_replace,
1042 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1043 {
1044 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1045 pgoff_t i = 0;
1046 int ret;
1047
1048 while (i < len) {
1049 if (blkaddr[i] == NULL_ADDR && !full) {
1050 i++;
1051 continue;
1052 }
1053
1054 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1055 struct dnode_of_data dn;
1056 struct node_info ni;
1057 size_t new_size;
1058 pgoff_t ilen;
1059
1060 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1061 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1062 if (ret)
1063 return ret;
1064
1065 get_node_info(sbi, dn.nid, &ni);
1066 ilen = min((pgoff_t)
1067 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1068 dn.ofs_in_node, len - i);
1069 do {
1070 dn.data_blkaddr = datablock_addr(dn.inode,
1071 dn.node_page, dn.ofs_in_node);
1072 truncate_data_blocks_range(&dn, 1);
1073
1074 if (do_replace[i]) {
1075 f2fs_i_blocks_write(src_inode,
1076 1, false, false);
1077 f2fs_i_blocks_write(dst_inode,
1078 1, true, false);
1079 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1080 blkaddr[i], ni.version, true, false);
1081
1082 do_replace[i] = 0;
1083 }
1084 dn.ofs_in_node++;
1085 i++;
1086 new_size = (dst + i) << PAGE_SHIFT;
1087 if (dst_inode->i_size < new_size)
1088 f2fs_i_size_write(dst_inode, new_size);
1089 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1090
1091 f2fs_put_dnode(&dn);
1092 } else {
1093 struct page *psrc, *pdst;
1094
1095 psrc = get_lock_data_page(src_inode, src + i, true);
1096 if (IS_ERR(psrc))
1097 return PTR_ERR(psrc);
1098 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1099 true);
1100 if (IS_ERR(pdst)) {
1101 f2fs_put_page(psrc, 1);
1102 return PTR_ERR(pdst);
1103 }
1104 f2fs_copy_page(psrc, pdst);
1105 set_page_dirty(pdst);
1106 f2fs_put_page(pdst, 1);
1107 f2fs_put_page(psrc, 1);
1108
1109 ret = truncate_hole(src_inode, src + i, src + i + 1);
1110 if (ret)
1111 return ret;
1112 i++;
1113 }
1114 }
1115 return 0;
1116 }
1117
1118 static int __exchange_data_block(struct inode *src_inode,
1119 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1120 pgoff_t len, bool full)
1121 {
1122 block_t *src_blkaddr;
1123 int *do_replace;
1124 pgoff_t olen;
1125 int ret;
1126
1127 while (len) {
1128 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1129
1130 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1131 sizeof(block_t) * olen, GFP_KERNEL);
1132 if (!src_blkaddr)
1133 return -ENOMEM;
1134
1135 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1136 sizeof(int) * olen, GFP_KERNEL);
1137 if (!do_replace) {
1138 kvfree(src_blkaddr);
1139 return -ENOMEM;
1140 }
1141
1142 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1143 do_replace, src, olen);
1144 if (ret)
1145 goto roll_back;
1146
1147 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1148 do_replace, src, dst, olen, full);
1149 if (ret)
1150 goto roll_back;
1151
1152 src += olen;
1153 dst += olen;
1154 len -= olen;
1155
1156 kvfree(src_blkaddr);
1157 kvfree(do_replace);
1158 }
1159 return 0;
1160
1161 roll_back:
1162 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1163 kvfree(src_blkaddr);
1164 kvfree(do_replace);
1165 return ret;
1166 }
1167
1168 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1169 {
1170 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1171 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1172 int ret;
1173
1174 f2fs_balance_fs(sbi, true);
1175 f2fs_lock_op(sbi);
1176
1177 f2fs_drop_extent_tree(inode);
1178
1179 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1180 f2fs_unlock_op(sbi);
1181 return ret;
1182 }
1183
1184 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1185 {
1186 pgoff_t pg_start, pg_end;
1187 loff_t new_size;
1188 int ret;
1189
1190 if (offset + len >= i_size_read(inode))
1191 return -EINVAL;
1192
1193 /* collapse range should be aligned to block size of f2fs. */
1194 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1195 return -EINVAL;
1196
1197 ret = f2fs_convert_inline_inode(inode);
1198 if (ret)
1199 return ret;
1200
1201 pg_start = offset >> PAGE_SHIFT;
1202 pg_end = (offset + len) >> PAGE_SHIFT;
1203
1204 /* avoid gc operation during block exchange */
1205 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1206
1207 down_write(&F2FS_I(inode)->i_mmap_sem);
1208 /* write out all dirty pages from offset */
1209 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1210 if (ret)
1211 goto out_unlock;
1212
1213 truncate_pagecache(inode, offset);
1214
1215 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1216 if (ret)
1217 goto out_unlock;
1218
1219 /* write out all moved pages, if possible */
1220 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1221 truncate_pagecache(inode, offset);
1222
1223 new_size = i_size_read(inode) - len;
1224 truncate_pagecache(inode, new_size);
1225
1226 ret = truncate_blocks(inode, new_size, true);
1227 if (!ret)
1228 f2fs_i_size_write(inode, new_size);
1229 out_unlock:
1230 up_write(&F2FS_I(inode)->i_mmap_sem);
1231 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1232 return ret;
1233 }
1234
1235 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1236 pgoff_t end)
1237 {
1238 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1239 pgoff_t index = start;
1240 unsigned int ofs_in_node = dn->ofs_in_node;
1241 blkcnt_t count = 0;
1242 int ret;
1243
1244 for (; index < end; index++, dn->ofs_in_node++) {
1245 if (datablock_addr(dn->inode, dn->node_page,
1246 dn->ofs_in_node) == NULL_ADDR)
1247 count++;
1248 }
1249
1250 dn->ofs_in_node = ofs_in_node;
1251 ret = reserve_new_blocks(dn, count);
1252 if (ret)
1253 return ret;
1254
1255 dn->ofs_in_node = ofs_in_node;
1256 for (index = start; index < end; index++, dn->ofs_in_node++) {
1257 dn->data_blkaddr = datablock_addr(dn->inode,
1258 dn->node_page, dn->ofs_in_node);
1259 /*
1260 * reserve_new_blocks will not guarantee entire block
1261 * allocation.
1262 */
1263 if (dn->data_blkaddr == NULL_ADDR) {
1264 ret = -ENOSPC;
1265 break;
1266 }
1267 if (dn->data_blkaddr != NEW_ADDR) {
1268 invalidate_blocks(sbi, dn->data_blkaddr);
1269 dn->data_blkaddr = NEW_ADDR;
1270 set_data_blkaddr(dn);
1271 }
1272 }
1273
1274 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1275
1276 return ret;
1277 }
1278
1279 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1280 int mode)
1281 {
1282 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1283 struct address_space *mapping = inode->i_mapping;
1284 pgoff_t index, pg_start, pg_end;
1285 loff_t new_size = i_size_read(inode);
1286 loff_t off_start, off_end;
1287 int ret = 0;
1288
1289 ret = inode_newsize_ok(inode, (len + offset));
1290 if (ret)
1291 return ret;
1292
1293 ret = f2fs_convert_inline_inode(inode);
1294 if (ret)
1295 return ret;
1296
1297 down_write(&F2FS_I(inode)->i_mmap_sem);
1298 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1299 if (ret)
1300 goto out_sem;
1301
1302 truncate_pagecache_range(inode, offset, offset + len - 1);
1303
1304 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1305 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1306
1307 off_start = offset & (PAGE_SIZE - 1);
1308 off_end = (offset + len) & (PAGE_SIZE - 1);
1309
1310 if (pg_start == pg_end) {
1311 ret = fill_zero(inode, pg_start, off_start,
1312 off_end - off_start);
1313 if (ret)
1314 goto out_sem;
1315
1316 new_size = max_t(loff_t, new_size, offset + len);
1317 } else {
1318 if (off_start) {
1319 ret = fill_zero(inode, pg_start++, off_start,
1320 PAGE_SIZE - off_start);
1321 if (ret)
1322 goto out_sem;
1323
1324 new_size = max_t(loff_t, new_size,
1325 (loff_t)pg_start << PAGE_SHIFT);
1326 }
1327
1328 for (index = pg_start; index < pg_end;) {
1329 struct dnode_of_data dn;
1330 unsigned int end_offset;
1331 pgoff_t end;
1332
1333 f2fs_lock_op(sbi);
1334
1335 set_new_dnode(&dn, inode, NULL, NULL, 0);
1336 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1337 if (ret) {
1338 f2fs_unlock_op(sbi);
1339 goto out;
1340 }
1341
1342 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1343 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1344
1345 ret = f2fs_do_zero_range(&dn, index, end);
1346 f2fs_put_dnode(&dn);
1347 f2fs_unlock_op(sbi);
1348
1349 f2fs_balance_fs(sbi, dn.node_changed);
1350
1351 if (ret)
1352 goto out;
1353
1354 index = end;
1355 new_size = max_t(loff_t, new_size,
1356 (loff_t)index << PAGE_SHIFT);
1357 }
1358
1359 if (off_end) {
1360 ret = fill_zero(inode, pg_end, 0, off_end);
1361 if (ret)
1362 goto out;
1363
1364 new_size = max_t(loff_t, new_size, offset + len);
1365 }
1366 }
1367
1368 out:
1369 if (new_size > i_size_read(inode)) {
1370 if (mode & FALLOC_FL_KEEP_SIZE)
1371 file_set_keep_isize(inode);
1372 else
1373 f2fs_i_size_write(inode, new_size);
1374 }
1375 out_sem:
1376 up_write(&F2FS_I(inode)->i_mmap_sem);
1377
1378 return ret;
1379 }
1380
1381 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1382 {
1383 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1384 pgoff_t nr, pg_start, pg_end, delta, idx;
1385 loff_t new_size;
1386 int ret = 0;
1387
1388 new_size = i_size_read(inode) + len;
1389 ret = inode_newsize_ok(inode, new_size);
1390 if (ret)
1391 return ret;
1392
1393 if (offset >= i_size_read(inode))
1394 return -EINVAL;
1395
1396 /* insert range should be aligned to block size of f2fs. */
1397 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1398 return -EINVAL;
1399
1400 ret = f2fs_convert_inline_inode(inode);
1401 if (ret)
1402 return ret;
1403
1404 f2fs_balance_fs(sbi, true);
1405
1406 /* avoid gc operation during block exchange */
1407 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1408
1409 down_write(&F2FS_I(inode)->i_mmap_sem);
1410 ret = truncate_blocks(inode, i_size_read(inode), true);
1411 if (ret)
1412 goto out;
1413
1414 /* write out all dirty pages from offset */
1415 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1416 if (ret)
1417 goto out;
1418
1419 truncate_pagecache(inode, offset);
1420
1421 pg_start = offset >> PAGE_SHIFT;
1422 pg_end = (offset + len) >> PAGE_SHIFT;
1423 delta = pg_end - pg_start;
1424 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1425
1426 while (!ret && idx > pg_start) {
1427 nr = idx - pg_start;
1428 if (nr > delta)
1429 nr = delta;
1430 idx -= nr;
1431
1432 f2fs_lock_op(sbi);
1433 f2fs_drop_extent_tree(inode);
1434
1435 ret = __exchange_data_block(inode, inode, idx,
1436 idx + delta, nr, false);
1437 f2fs_unlock_op(sbi);
1438 }
1439
1440 /* write out all moved pages, if possible */
1441 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1442 truncate_pagecache(inode, offset);
1443
1444 if (!ret)
1445 f2fs_i_size_write(inode, new_size);
1446 out:
1447 up_write(&F2FS_I(inode)->i_mmap_sem);
1448 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1449 return ret;
1450 }
1451
1452 static int expand_inode_data(struct inode *inode, loff_t offset,
1453 loff_t len, int mode)
1454 {
1455 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1456 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1457 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1458 pgoff_t pg_end;
1459 loff_t new_size = i_size_read(inode);
1460 loff_t off_end;
1461 int err;
1462
1463 err = inode_newsize_ok(inode, (len + offset));
1464 if (err)
1465 return err;
1466
1467 err = f2fs_convert_inline_inode(inode);
1468 if (err)
1469 return err;
1470
1471 f2fs_balance_fs(sbi, true);
1472
1473 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1474 off_end = (offset + len) & (PAGE_SIZE - 1);
1475
1476 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1477 map.m_len = pg_end - map.m_lblk;
1478 if (off_end)
1479 map.m_len++;
1480
1481 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1482 if (err) {
1483 pgoff_t last_off;
1484
1485 if (!map.m_len)
1486 return err;
1487
1488 last_off = map.m_lblk + map.m_len - 1;
1489
1490 /* update new size to the failed position */
1491 new_size = (last_off == pg_end) ? offset + len:
1492 (loff_t)(last_off + 1) << PAGE_SHIFT;
1493 } else {
1494 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1495 }
1496
1497 if (new_size > i_size_read(inode)) {
1498 if (mode & FALLOC_FL_KEEP_SIZE)
1499 file_set_keep_isize(inode);
1500 else
1501 f2fs_i_size_write(inode, new_size);
1502 }
1503
1504 return err;
1505 }
1506
1507 static long f2fs_fallocate(struct file *file, int mode,
1508 loff_t offset, loff_t len)
1509 {
1510 struct inode *inode = file_inode(file);
1511 long ret = 0;
1512
1513 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1514 return -EIO;
1515
1516 /* f2fs only support ->fallocate for regular file */
1517 if (!S_ISREG(inode->i_mode))
1518 return -EINVAL;
1519
1520 if (f2fs_encrypted_inode(inode) &&
1521 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1522 return -EOPNOTSUPP;
1523
1524 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1525 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1526 FALLOC_FL_INSERT_RANGE))
1527 return -EOPNOTSUPP;
1528
1529 inode_lock(inode);
1530
1531 if (mode & FALLOC_FL_PUNCH_HOLE) {
1532 if (offset >= inode->i_size)
1533 goto out;
1534
1535 ret = punch_hole(inode, offset, len);
1536 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1537 ret = f2fs_collapse_range(inode, offset, len);
1538 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1539 ret = f2fs_zero_range(inode, offset, len, mode);
1540 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1541 ret = f2fs_insert_range(inode, offset, len);
1542 } else {
1543 ret = expand_inode_data(inode, offset, len, mode);
1544 }
1545
1546 if (!ret) {
1547 inode->i_mtime = inode->i_ctime = current_time(inode);
1548 f2fs_mark_inode_dirty_sync(inode, false);
1549 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1550 }
1551
1552 out:
1553 inode_unlock(inode);
1554
1555 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1556 return ret;
1557 }
1558
1559 static int f2fs_release_file(struct inode *inode, struct file *filp)
1560 {
1561 /*
1562 * f2fs_relase_file is called at every close calls. So we should
1563 * not drop any inmemory pages by close called by other process.
1564 */
1565 if (!(filp->f_mode & FMODE_WRITE) ||
1566 atomic_read(&inode->i_writecount) != 1)
1567 return 0;
1568
1569 /* some remained atomic pages should discarded */
1570 if (f2fs_is_atomic_file(inode))
1571 drop_inmem_pages(inode);
1572 if (f2fs_is_volatile_file(inode)) {
1573 clear_inode_flag(inode, FI_VOLATILE_FILE);
1574 stat_dec_volatile_write(inode);
1575 set_inode_flag(inode, FI_DROP_CACHE);
1576 filemap_fdatawrite(inode->i_mapping);
1577 clear_inode_flag(inode, FI_DROP_CACHE);
1578 }
1579 return 0;
1580 }
1581
1582 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1583 {
1584 struct inode *inode = file_inode(file);
1585
1586 /*
1587 * If the process doing a transaction is crashed, we should do
1588 * roll-back. Otherwise, other reader/write can see corrupted database
1589 * until all the writers close its file. Since this should be done
1590 * before dropping file lock, it needs to do in ->flush.
1591 */
1592 if (f2fs_is_atomic_file(inode) &&
1593 F2FS_I(inode)->inmem_task == current)
1594 drop_inmem_pages(inode);
1595 return 0;
1596 }
1597
1598 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1599 {
1600 struct inode *inode = file_inode(filp);
1601 struct f2fs_inode_info *fi = F2FS_I(inode);
1602 unsigned int flags = fi->i_flags &
1603 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1604 return put_user(flags, (int __user *)arg);
1605 }
1606
1607 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1608 {
1609 struct f2fs_inode_info *fi = F2FS_I(inode);
1610 unsigned int oldflags;
1611
1612 /* Is it quota file? Do not allow user to mess with it */
1613 if (IS_NOQUOTA(inode))
1614 return -EPERM;
1615
1616 flags = f2fs_mask_flags(inode->i_mode, flags);
1617
1618 oldflags = fi->i_flags;
1619
1620 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1621 if (!capable(CAP_LINUX_IMMUTABLE))
1622 return -EPERM;
1623
1624 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1625 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1626 fi->i_flags = flags;
1627
1628 if (fi->i_flags & FS_PROJINHERIT_FL)
1629 set_inode_flag(inode, FI_PROJ_INHERIT);
1630 else
1631 clear_inode_flag(inode, FI_PROJ_INHERIT);
1632
1633 inode->i_ctime = current_time(inode);
1634 f2fs_set_inode_flags(inode);
1635 f2fs_mark_inode_dirty_sync(inode, false);
1636 return 0;
1637 }
1638
1639 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1640 {
1641 struct inode *inode = file_inode(filp);
1642 unsigned int flags;
1643 int ret;
1644
1645 if (!inode_owner_or_capable(inode))
1646 return -EACCES;
1647
1648 if (get_user(flags, (int __user *)arg))
1649 return -EFAULT;
1650
1651 ret = mnt_want_write_file(filp);
1652 if (ret)
1653 return ret;
1654
1655 inode_lock(inode);
1656
1657 ret = __f2fs_ioc_setflags(inode, flags);
1658
1659 inode_unlock(inode);
1660 mnt_drop_write_file(filp);
1661 return ret;
1662 }
1663
1664 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1665 {
1666 struct inode *inode = file_inode(filp);
1667
1668 return put_user(inode->i_generation, (int __user *)arg);
1669 }
1670
1671 static int f2fs_ioc_start_atomic_write(struct file *filp)
1672 {
1673 struct inode *inode = file_inode(filp);
1674 int ret;
1675
1676 if (!inode_owner_or_capable(inode))
1677 return -EACCES;
1678
1679 if (!S_ISREG(inode->i_mode))
1680 return -EINVAL;
1681
1682 ret = mnt_want_write_file(filp);
1683 if (ret)
1684 return ret;
1685
1686 inode_lock(inode);
1687
1688 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1689
1690 if (f2fs_is_atomic_file(inode))
1691 goto out;
1692
1693 ret = f2fs_convert_inline_inode(inode);
1694 if (ret)
1695 goto out;
1696
1697 set_inode_flag(inode, FI_ATOMIC_FILE);
1698 set_inode_flag(inode, FI_HOT_DATA);
1699 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1700
1701 if (!get_dirty_pages(inode))
1702 goto inc_stat;
1703
1704 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1705 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1706 inode->i_ino, get_dirty_pages(inode));
1707 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1708 if (ret) {
1709 clear_inode_flag(inode, FI_ATOMIC_FILE);
1710 clear_inode_flag(inode, FI_HOT_DATA);
1711 goto out;
1712 }
1713
1714 inc_stat:
1715 F2FS_I(inode)->inmem_task = current;
1716 stat_inc_atomic_write(inode);
1717 stat_update_max_atomic_write(inode);
1718 out:
1719 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1720 inode_unlock(inode);
1721 mnt_drop_write_file(filp);
1722 return ret;
1723 }
1724
1725 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1726 {
1727 struct inode *inode = file_inode(filp);
1728 int ret;
1729
1730 if (!inode_owner_or_capable(inode))
1731 return -EACCES;
1732
1733 ret = mnt_want_write_file(filp);
1734 if (ret)
1735 return ret;
1736
1737 inode_lock(inode);
1738
1739 down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1740
1741 if (f2fs_is_volatile_file(inode))
1742 goto err_out;
1743
1744 if (f2fs_is_atomic_file(inode)) {
1745 ret = commit_inmem_pages(inode);
1746 if (ret)
1747 goto err_out;
1748
1749 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1750 if (!ret) {
1751 clear_inode_flag(inode, FI_ATOMIC_FILE);
1752 clear_inode_flag(inode, FI_HOT_DATA);
1753 stat_dec_atomic_write(inode);
1754 }
1755 } else {
1756 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1757 }
1758 err_out:
1759 up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1760 inode_unlock(inode);
1761 mnt_drop_write_file(filp);
1762 return ret;
1763 }
1764
1765 static int f2fs_ioc_start_volatile_write(struct file *filp)
1766 {
1767 struct inode *inode = file_inode(filp);
1768 int ret;
1769
1770 if (!inode_owner_or_capable(inode))
1771 return -EACCES;
1772
1773 if (!S_ISREG(inode->i_mode))
1774 return -EINVAL;
1775
1776 ret = mnt_want_write_file(filp);
1777 if (ret)
1778 return ret;
1779
1780 inode_lock(inode);
1781
1782 if (f2fs_is_volatile_file(inode))
1783 goto out;
1784
1785 ret = f2fs_convert_inline_inode(inode);
1786 if (ret)
1787 goto out;
1788
1789 stat_inc_volatile_write(inode);
1790 stat_update_max_volatile_write(inode);
1791
1792 set_inode_flag(inode, FI_VOLATILE_FILE);
1793 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1794 out:
1795 inode_unlock(inode);
1796 mnt_drop_write_file(filp);
1797 return ret;
1798 }
1799
1800 static int f2fs_ioc_release_volatile_write(struct file *filp)
1801 {
1802 struct inode *inode = file_inode(filp);
1803 int ret;
1804
1805 if (!inode_owner_or_capable(inode))
1806 return -EACCES;
1807
1808 ret = mnt_want_write_file(filp);
1809 if (ret)
1810 return ret;
1811
1812 inode_lock(inode);
1813
1814 if (!f2fs_is_volatile_file(inode))
1815 goto out;
1816
1817 if (!f2fs_is_first_block_written(inode)) {
1818 ret = truncate_partial_data_page(inode, 0, true);
1819 goto out;
1820 }
1821
1822 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1823 out:
1824 inode_unlock(inode);
1825 mnt_drop_write_file(filp);
1826 return ret;
1827 }
1828
1829 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1830 {
1831 struct inode *inode = file_inode(filp);
1832 int ret;
1833
1834 if (!inode_owner_or_capable(inode))
1835 return -EACCES;
1836
1837 ret = mnt_want_write_file(filp);
1838 if (ret)
1839 return ret;
1840
1841 inode_lock(inode);
1842
1843 if (f2fs_is_atomic_file(inode))
1844 drop_inmem_pages(inode);
1845 if (f2fs_is_volatile_file(inode)) {
1846 clear_inode_flag(inode, FI_VOLATILE_FILE);
1847 stat_dec_volatile_write(inode);
1848 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1849 }
1850
1851 inode_unlock(inode);
1852
1853 mnt_drop_write_file(filp);
1854 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1855 return ret;
1856 }
1857
1858 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1859 {
1860 struct inode *inode = file_inode(filp);
1861 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1862 struct super_block *sb = sbi->sb;
1863 __u32 in;
1864 int ret = 0;
1865
1866 if (!capable(CAP_SYS_ADMIN))
1867 return -EPERM;
1868
1869 if (get_user(in, (__u32 __user *)arg))
1870 return -EFAULT;
1871
1872 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1873 ret = mnt_want_write_file(filp);
1874 if (ret)
1875 return ret;
1876 }
1877
1878 switch (in) {
1879 case F2FS_GOING_DOWN_FULLSYNC:
1880 sb = freeze_bdev(sb->s_bdev);
1881 if (IS_ERR(sb)) {
1882 ret = PTR_ERR(sb);
1883 goto out;
1884 }
1885 if (sb) {
1886 f2fs_stop_checkpoint(sbi, false);
1887 thaw_bdev(sb->s_bdev, sb);
1888 }
1889 break;
1890 case F2FS_GOING_DOWN_METASYNC:
1891 /* do checkpoint only */
1892 ret = f2fs_sync_fs(sb, 1);
1893 if (ret)
1894 goto out;
1895 f2fs_stop_checkpoint(sbi, false);
1896 break;
1897 case F2FS_GOING_DOWN_NOSYNC:
1898 f2fs_stop_checkpoint(sbi, false);
1899 break;
1900 case F2FS_GOING_DOWN_METAFLUSH:
1901 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1902 f2fs_stop_checkpoint(sbi, false);
1903 break;
1904 default:
1905 ret = -EINVAL;
1906 goto out;
1907 }
1908
1909 stop_gc_thread(sbi);
1910 stop_discard_thread(sbi);
1911
1912 drop_discard_cmd(sbi);
1913 clear_opt(sbi, DISCARD);
1914
1915 f2fs_update_time(sbi, REQ_TIME);
1916 out:
1917 if (in != F2FS_GOING_DOWN_FULLSYNC)
1918 mnt_drop_write_file(filp);
1919 return ret;
1920 }
1921
1922 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1923 {
1924 struct inode *inode = file_inode(filp);
1925 struct super_block *sb = inode->i_sb;
1926 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1927 struct fstrim_range range;
1928 int ret;
1929
1930 if (!capable(CAP_SYS_ADMIN))
1931 return -EPERM;
1932
1933 if (!blk_queue_discard(q))
1934 return -EOPNOTSUPP;
1935
1936 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1937 sizeof(range)))
1938 return -EFAULT;
1939
1940 ret = mnt_want_write_file(filp);
1941 if (ret)
1942 return ret;
1943
1944 range.minlen = max((unsigned int)range.minlen,
1945 q->limits.discard_granularity);
1946 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1947 mnt_drop_write_file(filp);
1948 if (ret < 0)
1949 return ret;
1950
1951 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1952 sizeof(range)))
1953 return -EFAULT;
1954 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1955 return 0;
1956 }
1957
1958 static bool uuid_is_nonzero(__u8 u[16])
1959 {
1960 int i;
1961
1962 for (i = 0; i < 16; i++)
1963 if (u[i])
1964 return true;
1965 return false;
1966 }
1967
1968 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1969 {
1970 struct inode *inode = file_inode(filp);
1971
1972 if (!f2fs_sb_has_encrypt(inode->i_sb))
1973 return -EOPNOTSUPP;
1974
1975 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1976
1977 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1978 }
1979
1980 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1981 {
1982 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
1983 return -EOPNOTSUPP;
1984 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1985 }
1986
1987 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1988 {
1989 struct inode *inode = file_inode(filp);
1990 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1991 int err;
1992
1993 if (!f2fs_sb_has_encrypt(inode->i_sb))
1994 return -EOPNOTSUPP;
1995
1996 err = mnt_want_write_file(filp);
1997 if (err)
1998 return err;
1999
2000 down_write(&sbi->sb_lock);
2001
2002 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2003 goto got_it;
2004
2005 /* update superblock with uuid */
2006 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2007
2008 err = f2fs_commit_super(sbi, false);
2009 if (err) {
2010 /* undo new data */
2011 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2012 goto out_err;
2013 }
2014 got_it:
2015 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2016 16))
2017 err = -EFAULT;
2018 out_err:
2019 up_write(&sbi->sb_lock);
2020 mnt_drop_write_file(filp);
2021 return err;
2022 }
2023
2024 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2025 {
2026 struct inode *inode = file_inode(filp);
2027 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2028 __u32 sync;
2029 int ret;
2030
2031 if (!capable(CAP_SYS_ADMIN))
2032 return -EPERM;
2033
2034 if (get_user(sync, (__u32 __user *)arg))
2035 return -EFAULT;
2036
2037 if (f2fs_readonly(sbi->sb))
2038 return -EROFS;
2039
2040 ret = mnt_want_write_file(filp);
2041 if (ret)
2042 return ret;
2043
2044 if (!sync) {
2045 if (!mutex_trylock(&sbi->gc_mutex)) {
2046 ret = -EBUSY;
2047 goto out;
2048 }
2049 } else {
2050 mutex_lock(&sbi->gc_mutex);
2051 }
2052
2053 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2054 out:
2055 mnt_drop_write_file(filp);
2056 return ret;
2057 }
2058
2059 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2060 {
2061 struct inode *inode = file_inode(filp);
2062 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2063 struct f2fs_gc_range range;
2064 u64 end;
2065 int ret;
2066
2067 if (!capable(CAP_SYS_ADMIN))
2068 return -EPERM;
2069
2070 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2071 sizeof(range)))
2072 return -EFAULT;
2073
2074 if (f2fs_readonly(sbi->sb))
2075 return -EROFS;
2076
2077 ret = mnt_want_write_file(filp);
2078 if (ret)
2079 return ret;
2080
2081 end = range.start + range.len;
2082 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2083 ret = -EINVAL;
2084 goto out;
2085 }
2086 do_more:
2087 if (!range.sync) {
2088 if (!mutex_trylock(&sbi->gc_mutex)) {
2089 ret = -EBUSY;
2090 goto out;
2091 }
2092 } else {
2093 mutex_lock(&sbi->gc_mutex);
2094 }
2095
2096 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2097 range.start += sbi->blocks_per_seg;
2098 if (range.start <= end)
2099 goto do_more;
2100 out:
2101 mnt_drop_write_file(filp);
2102 return ret;
2103 }
2104
2105 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2106 {
2107 struct inode *inode = file_inode(filp);
2108 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2109 int ret;
2110
2111 if (!capable(CAP_SYS_ADMIN))
2112 return -EPERM;
2113
2114 if (f2fs_readonly(sbi->sb))
2115 return -EROFS;
2116
2117 ret = mnt_want_write_file(filp);
2118 if (ret)
2119 return ret;
2120
2121 ret = f2fs_sync_fs(sbi->sb, 1);
2122
2123 mnt_drop_write_file(filp);
2124 return ret;
2125 }
2126
2127 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2128 struct file *filp,
2129 struct f2fs_defragment *range)
2130 {
2131 struct inode *inode = file_inode(filp);
2132 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2133 .m_seg_type = NO_CHECK_TYPE };
2134 struct extent_info ei = {0,0,0};
2135 pgoff_t pg_start, pg_end, next_pgofs;
2136 unsigned int blk_per_seg = sbi->blocks_per_seg;
2137 unsigned int total = 0, sec_num;
2138 block_t blk_end = 0;
2139 bool fragmented = false;
2140 int err;
2141
2142 /* if in-place-update policy is enabled, don't waste time here */
2143 if (should_update_inplace(inode, NULL))
2144 return -EINVAL;
2145
2146 pg_start = range->start >> PAGE_SHIFT;
2147 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2148
2149 f2fs_balance_fs(sbi, true);
2150
2151 inode_lock(inode);
2152
2153 /* writeback all dirty pages in the range */
2154 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2155 range->start + range->len - 1);
2156 if (err)
2157 goto out;
2158
2159 /*
2160 * lookup mapping info in extent cache, skip defragmenting if physical
2161 * block addresses are continuous.
2162 */
2163 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2164 if (ei.fofs + ei.len >= pg_end)
2165 goto out;
2166 }
2167
2168 map.m_lblk = pg_start;
2169 map.m_next_pgofs = &next_pgofs;
2170
2171 /*
2172 * lookup mapping info in dnode page cache, skip defragmenting if all
2173 * physical block addresses are continuous even if there are hole(s)
2174 * in logical blocks.
2175 */
2176 while (map.m_lblk < pg_end) {
2177 map.m_len = pg_end - map.m_lblk;
2178 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2179 if (err)
2180 goto out;
2181
2182 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2183 map.m_lblk = next_pgofs;
2184 continue;
2185 }
2186
2187 if (blk_end && blk_end != map.m_pblk)
2188 fragmented = true;
2189
2190 /* record total count of block that we're going to move */
2191 total += map.m_len;
2192
2193 blk_end = map.m_pblk + map.m_len;
2194
2195 map.m_lblk += map.m_len;
2196 }
2197
2198 if (!fragmented)
2199 goto out;
2200
2201 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2202
2203 /*
2204 * make sure there are enough free section for LFS allocation, this can
2205 * avoid defragment running in SSR mode when free section are allocated
2206 * intensively
2207 */
2208 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2209 err = -EAGAIN;
2210 goto out;
2211 }
2212
2213 map.m_lblk = pg_start;
2214 map.m_len = pg_end - pg_start;
2215 total = 0;
2216
2217 while (map.m_lblk < pg_end) {
2218 pgoff_t idx;
2219 int cnt = 0;
2220
2221 do_map:
2222 map.m_len = pg_end - map.m_lblk;
2223 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2224 if (err)
2225 goto clear_out;
2226
2227 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2228 map.m_lblk = next_pgofs;
2229 continue;
2230 }
2231
2232 set_inode_flag(inode, FI_DO_DEFRAG);
2233
2234 idx = map.m_lblk;
2235 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2236 struct page *page;
2237
2238 page = get_lock_data_page(inode, idx, true);
2239 if (IS_ERR(page)) {
2240 err = PTR_ERR(page);
2241 goto clear_out;
2242 }
2243
2244 set_page_dirty(page);
2245 f2fs_put_page(page, 1);
2246
2247 idx++;
2248 cnt++;
2249 total++;
2250 }
2251
2252 map.m_lblk = idx;
2253
2254 if (idx < pg_end && cnt < blk_per_seg)
2255 goto do_map;
2256
2257 clear_inode_flag(inode, FI_DO_DEFRAG);
2258
2259 err = filemap_fdatawrite(inode->i_mapping);
2260 if (err)
2261 goto out;
2262 }
2263 clear_out:
2264 clear_inode_flag(inode, FI_DO_DEFRAG);
2265 out:
2266 inode_unlock(inode);
2267 if (!err)
2268 range->len = (u64)total << PAGE_SHIFT;
2269 return err;
2270 }
2271
2272 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2273 {
2274 struct inode *inode = file_inode(filp);
2275 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2276 struct f2fs_defragment range;
2277 int err;
2278
2279 if (!capable(CAP_SYS_ADMIN))
2280 return -EPERM;
2281
2282 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2283 return -EINVAL;
2284
2285 if (f2fs_readonly(sbi->sb))
2286 return -EROFS;
2287
2288 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2289 sizeof(range)))
2290 return -EFAULT;
2291
2292 /* verify alignment of offset & size */
2293 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2294 return -EINVAL;
2295
2296 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2297 sbi->max_file_blocks))
2298 return -EINVAL;
2299
2300 err = mnt_want_write_file(filp);
2301 if (err)
2302 return err;
2303
2304 err = f2fs_defragment_range(sbi, filp, &range);
2305 mnt_drop_write_file(filp);
2306
2307 f2fs_update_time(sbi, REQ_TIME);
2308 if (err < 0)
2309 return err;
2310
2311 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2312 sizeof(range)))
2313 return -EFAULT;
2314
2315 return 0;
2316 }
2317
2318 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2319 struct file *file_out, loff_t pos_out, size_t len)
2320 {
2321 struct inode *src = file_inode(file_in);
2322 struct inode *dst = file_inode(file_out);
2323 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2324 size_t olen = len, dst_max_i_size = 0;
2325 size_t dst_osize;
2326 int ret;
2327
2328 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2329 src->i_sb != dst->i_sb)
2330 return -EXDEV;
2331
2332 if (unlikely(f2fs_readonly(src->i_sb)))
2333 return -EROFS;
2334
2335 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2336 return -EINVAL;
2337
2338 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2339 return -EOPNOTSUPP;
2340
2341 if (src == dst) {
2342 if (pos_in == pos_out)
2343 return 0;
2344 if (pos_out > pos_in && pos_out < pos_in + len)
2345 return -EINVAL;
2346 }
2347
2348 inode_lock(src);
2349 down_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2350 if (src != dst) {
2351 ret = -EBUSY;
2352 if (!inode_trylock(dst))
2353 goto out;
2354 if (!down_write_trylock(&F2FS_I(dst)->dio_rwsem[WRITE])) {
2355 inode_unlock(dst);
2356 goto out;
2357 }
2358 }
2359
2360 ret = -EINVAL;
2361 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2362 goto out_unlock;
2363 if (len == 0)
2364 olen = len = src->i_size - pos_in;
2365 if (pos_in + len == src->i_size)
2366 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2367 if (len == 0) {
2368 ret = 0;
2369 goto out_unlock;
2370 }
2371
2372 dst_osize = dst->i_size;
2373 if (pos_out + olen > dst->i_size)
2374 dst_max_i_size = pos_out + olen;
2375
2376 /* verify the end result is block aligned */
2377 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2378 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2379 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2380 goto out_unlock;
2381
2382 ret = f2fs_convert_inline_inode(src);
2383 if (ret)
2384 goto out_unlock;
2385
2386 ret = f2fs_convert_inline_inode(dst);
2387 if (ret)
2388 goto out_unlock;
2389
2390 /* write out all dirty pages from offset */
2391 ret = filemap_write_and_wait_range(src->i_mapping,
2392 pos_in, pos_in + len);
2393 if (ret)
2394 goto out_unlock;
2395
2396 ret = filemap_write_and_wait_range(dst->i_mapping,
2397 pos_out, pos_out + len);
2398 if (ret)
2399 goto out_unlock;
2400
2401 f2fs_balance_fs(sbi, true);
2402 f2fs_lock_op(sbi);
2403 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2404 pos_out >> F2FS_BLKSIZE_BITS,
2405 len >> F2FS_BLKSIZE_BITS, false);
2406
2407 if (!ret) {
2408 if (dst_max_i_size)
2409 f2fs_i_size_write(dst, dst_max_i_size);
2410 else if (dst_osize != dst->i_size)
2411 f2fs_i_size_write(dst, dst_osize);
2412 }
2413 f2fs_unlock_op(sbi);
2414 out_unlock:
2415 if (src != dst) {
2416 up_write(&F2FS_I(dst)->dio_rwsem[WRITE]);
2417 inode_unlock(dst);
2418 }
2419 out:
2420 up_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2421 inode_unlock(src);
2422 return ret;
2423 }
2424
2425 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2426 {
2427 struct f2fs_move_range range;
2428 struct fd dst;
2429 int err;
2430
2431 if (!(filp->f_mode & FMODE_READ) ||
2432 !(filp->f_mode & FMODE_WRITE))
2433 return -EBADF;
2434
2435 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2436 sizeof(range)))
2437 return -EFAULT;
2438
2439 dst = fdget(range.dst_fd);
2440 if (!dst.file)
2441 return -EBADF;
2442
2443 if (!(dst.file->f_mode & FMODE_WRITE)) {
2444 err = -EBADF;
2445 goto err_out;
2446 }
2447
2448 err = mnt_want_write_file(filp);
2449 if (err)
2450 goto err_out;
2451
2452 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2453 range.pos_out, range.len);
2454
2455 mnt_drop_write_file(filp);
2456 if (err)
2457 goto err_out;
2458
2459 if (copy_to_user((struct f2fs_move_range __user *)arg,
2460 &range, sizeof(range)))
2461 err = -EFAULT;
2462 err_out:
2463 fdput(dst);
2464 return err;
2465 }
2466
2467 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2468 {
2469 struct inode *inode = file_inode(filp);
2470 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2471 struct sit_info *sm = SIT_I(sbi);
2472 unsigned int start_segno = 0, end_segno = 0;
2473 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2474 struct f2fs_flush_device range;
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 (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2484 sizeof(range)))
2485 return -EFAULT;
2486
2487 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2488 sbi->segs_per_sec != 1) {
2489 f2fs_msg(sbi->sb, KERN_WARNING,
2490 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2491 range.dev_num, sbi->s_ndevs,
2492 sbi->segs_per_sec);
2493 return -EINVAL;
2494 }
2495
2496 ret = mnt_want_write_file(filp);
2497 if (ret)
2498 return ret;
2499
2500 if (range.dev_num != 0)
2501 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2502 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2503
2504 start_segno = sm->last_victim[FLUSH_DEVICE];
2505 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2506 start_segno = dev_start_segno;
2507 end_segno = min(start_segno + range.segments, dev_end_segno);
2508
2509 while (start_segno < end_segno) {
2510 if (!mutex_trylock(&sbi->gc_mutex)) {
2511 ret = -EBUSY;
2512 goto out;
2513 }
2514 sm->last_victim[GC_CB] = end_segno + 1;
2515 sm->last_victim[GC_GREEDY] = end_segno + 1;
2516 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2517 ret = f2fs_gc(sbi, true, true, start_segno);
2518 if (ret == -EAGAIN)
2519 ret = 0;
2520 else if (ret < 0)
2521 break;
2522 start_segno++;
2523 }
2524 out:
2525 mnt_drop_write_file(filp);
2526 return ret;
2527 }
2528
2529 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2530 {
2531 struct inode *inode = file_inode(filp);
2532 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2533
2534 /* Must validate to set it with SQLite behavior in Android. */
2535 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2536
2537 return put_user(sb_feature, (u32 __user *)arg);
2538 }
2539
2540 #ifdef CONFIG_QUOTA
2541 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2542 {
2543 struct inode *inode = file_inode(filp);
2544 struct f2fs_inode_info *fi = F2FS_I(inode);
2545 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2546 struct super_block *sb = sbi->sb;
2547 struct dquot *transfer_to[MAXQUOTAS] = {};
2548 struct page *ipage;
2549 kprojid_t kprojid;
2550 int err;
2551
2552 if (!f2fs_sb_has_project_quota(sb)) {
2553 if (projid != F2FS_DEF_PROJID)
2554 return -EOPNOTSUPP;
2555 else
2556 return 0;
2557 }
2558
2559 if (!f2fs_has_extra_attr(inode))
2560 return -EOPNOTSUPP;
2561
2562 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2563
2564 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2565 return 0;
2566
2567 err = mnt_want_write_file(filp);
2568 if (err)
2569 return err;
2570
2571 err = -EPERM;
2572 inode_lock(inode);
2573
2574 /* Is it quota file? Do not allow user to mess with it */
2575 if (IS_NOQUOTA(inode))
2576 goto out_unlock;
2577
2578 ipage = get_node_page(sbi, inode->i_ino);
2579 if (IS_ERR(ipage)) {
2580 err = PTR_ERR(ipage);
2581 goto out_unlock;
2582 }
2583
2584 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2585 i_projid)) {
2586 err = -EOVERFLOW;
2587 f2fs_put_page(ipage, 1);
2588 goto out_unlock;
2589 }
2590 f2fs_put_page(ipage, 1);
2591
2592 err = dquot_initialize(inode);
2593 if (err)
2594 goto out_unlock;
2595
2596 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2597 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2598 err = __dquot_transfer(inode, transfer_to);
2599 dqput(transfer_to[PRJQUOTA]);
2600 if (err)
2601 goto out_dirty;
2602 }
2603
2604 F2FS_I(inode)->i_projid = kprojid;
2605 inode->i_ctime = current_time(inode);
2606 out_dirty:
2607 f2fs_mark_inode_dirty_sync(inode, true);
2608 out_unlock:
2609 inode_unlock(inode);
2610 mnt_drop_write_file(filp);
2611 return err;
2612 }
2613 #else
2614 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2615 {
2616 if (projid != F2FS_DEF_PROJID)
2617 return -EOPNOTSUPP;
2618 return 0;
2619 }
2620 #endif
2621
2622 /* Transfer internal flags to xflags */
2623 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2624 {
2625 __u32 xflags = 0;
2626
2627 if (iflags & FS_SYNC_FL)
2628 xflags |= FS_XFLAG_SYNC;
2629 if (iflags & FS_IMMUTABLE_FL)
2630 xflags |= FS_XFLAG_IMMUTABLE;
2631 if (iflags & FS_APPEND_FL)
2632 xflags |= FS_XFLAG_APPEND;
2633 if (iflags & FS_NODUMP_FL)
2634 xflags |= FS_XFLAG_NODUMP;
2635 if (iflags & FS_NOATIME_FL)
2636 xflags |= FS_XFLAG_NOATIME;
2637 if (iflags & FS_PROJINHERIT_FL)
2638 xflags |= FS_XFLAG_PROJINHERIT;
2639 return xflags;
2640 }
2641
2642 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2643 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2644 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2645
2646 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2647 #define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2648 FS_IMMUTABLE_FL | \
2649 FS_APPEND_FL | \
2650 FS_NODUMP_FL | \
2651 FS_NOATIME_FL | \
2652 FS_PROJINHERIT_FL)
2653
2654 /* Transfer xflags flags to internal */
2655 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2656 {
2657 unsigned long iflags = 0;
2658
2659 if (xflags & FS_XFLAG_SYNC)
2660 iflags |= FS_SYNC_FL;
2661 if (xflags & FS_XFLAG_IMMUTABLE)
2662 iflags |= FS_IMMUTABLE_FL;
2663 if (xflags & FS_XFLAG_APPEND)
2664 iflags |= FS_APPEND_FL;
2665 if (xflags & FS_XFLAG_NODUMP)
2666 iflags |= FS_NODUMP_FL;
2667 if (xflags & FS_XFLAG_NOATIME)
2668 iflags |= FS_NOATIME_FL;
2669 if (xflags & FS_XFLAG_PROJINHERIT)
2670 iflags |= FS_PROJINHERIT_FL;
2671
2672 return iflags;
2673 }
2674
2675 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2676 {
2677 struct inode *inode = file_inode(filp);
2678 struct f2fs_inode_info *fi = F2FS_I(inode);
2679 struct fsxattr fa;
2680
2681 memset(&fa, 0, sizeof(struct fsxattr));
2682 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2683 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2684
2685 if (f2fs_sb_has_project_quota(inode->i_sb))
2686 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2687 fi->i_projid);
2688
2689 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2690 return -EFAULT;
2691 return 0;
2692 }
2693
2694 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2695 {
2696 struct inode *inode = file_inode(filp);
2697 struct f2fs_inode_info *fi = F2FS_I(inode);
2698 struct fsxattr fa;
2699 unsigned int flags;
2700 int err;
2701
2702 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2703 return -EFAULT;
2704
2705 /* Make sure caller has proper permission */
2706 if (!inode_owner_or_capable(inode))
2707 return -EACCES;
2708
2709 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2710 return -EOPNOTSUPP;
2711
2712 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2713 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2714 return -EOPNOTSUPP;
2715
2716 err = mnt_want_write_file(filp);
2717 if (err)
2718 return err;
2719
2720 inode_lock(inode);
2721 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2722 (flags & F2FS_FL_XFLAG_VISIBLE);
2723 err = __f2fs_ioc_setflags(inode, flags);
2724 inode_unlock(inode);
2725 mnt_drop_write_file(filp);
2726 if (err)
2727 return err;
2728
2729 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2730 if (err)
2731 return err;
2732
2733 return 0;
2734 }
2735
2736 int f2fs_pin_file_control(struct inode *inode, bool inc)
2737 {
2738 struct f2fs_inode_info *fi = F2FS_I(inode);
2739 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2740
2741 /* Use i_gc_failures for normal file as a risk signal. */
2742 if (inc)
2743 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
2744
2745 if (fi->i_gc_failures > sbi->gc_pin_file_threshold) {
2746 f2fs_msg(sbi->sb, KERN_WARNING,
2747 "%s: Enable GC = ino %lx after %x GC trials\n",
2748 __func__, inode->i_ino, fi->i_gc_failures);
2749 clear_inode_flag(inode, FI_PIN_FILE);
2750 return -EAGAIN;
2751 }
2752 return 0;
2753 }
2754
2755 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2756 {
2757 struct inode *inode = file_inode(filp);
2758 __u32 pin;
2759 int ret = 0;
2760
2761 if (!inode_owner_or_capable(inode))
2762 return -EACCES;
2763
2764 if (get_user(pin, (__u32 __user *)arg))
2765 return -EFAULT;
2766
2767 if (!S_ISREG(inode->i_mode))
2768 return -EINVAL;
2769
2770 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2771 return -EROFS;
2772
2773 ret = mnt_want_write_file(filp);
2774 if (ret)
2775 return ret;
2776
2777 inode_lock(inode);
2778
2779 if (should_update_outplace(inode, NULL)) {
2780 ret = -EINVAL;
2781 goto out;
2782 }
2783
2784 if (!pin) {
2785 clear_inode_flag(inode, FI_PIN_FILE);
2786 F2FS_I(inode)->i_gc_failures = 1;
2787 goto done;
2788 }
2789
2790 if (f2fs_pin_file_control(inode, false)) {
2791 ret = -EAGAIN;
2792 goto out;
2793 }
2794 ret = f2fs_convert_inline_inode(inode);
2795 if (ret)
2796 goto out;
2797
2798 set_inode_flag(inode, FI_PIN_FILE);
2799 ret = F2FS_I(inode)->i_gc_failures;
2800 done:
2801 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2802 out:
2803 inode_unlock(inode);
2804 mnt_drop_write_file(filp);
2805 return ret;
2806 }
2807
2808 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2809 {
2810 struct inode *inode = file_inode(filp);
2811 __u32 pin = 0;
2812
2813 if (is_inode_flag_set(inode, FI_PIN_FILE))
2814 pin = F2FS_I(inode)->i_gc_failures;
2815 return put_user(pin, (u32 __user *)arg);
2816 }
2817
2818 int f2fs_precache_extents(struct inode *inode)
2819 {
2820 struct f2fs_inode_info *fi = F2FS_I(inode);
2821 struct f2fs_map_blocks map;
2822 pgoff_t m_next_extent;
2823 loff_t end;
2824 int err;
2825
2826 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2827 return -EOPNOTSUPP;
2828
2829 map.m_lblk = 0;
2830 map.m_next_pgofs = NULL;
2831 map.m_next_extent = &m_next_extent;
2832 map.m_seg_type = NO_CHECK_TYPE;
2833 end = F2FS_I_SB(inode)->max_file_blocks;
2834
2835 while (map.m_lblk < end) {
2836 map.m_len = end - map.m_lblk;
2837
2838 down_write(&fi->dio_rwsem[WRITE]);
2839 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2840 up_write(&fi->dio_rwsem[WRITE]);
2841 if (err)
2842 return err;
2843
2844 map.m_lblk = m_next_extent;
2845 }
2846
2847 return err;
2848 }
2849
2850 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2851 {
2852 return f2fs_precache_extents(file_inode(filp));
2853 }
2854
2855 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2856 {
2857 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2858 return -EIO;
2859
2860 switch (cmd) {
2861 case F2FS_IOC_GETFLAGS:
2862 return f2fs_ioc_getflags(filp, arg);
2863 case F2FS_IOC_SETFLAGS:
2864 return f2fs_ioc_setflags(filp, arg);
2865 case F2FS_IOC_GETVERSION:
2866 return f2fs_ioc_getversion(filp, arg);
2867 case F2FS_IOC_START_ATOMIC_WRITE:
2868 return f2fs_ioc_start_atomic_write(filp);
2869 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2870 return f2fs_ioc_commit_atomic_write(filp);
2871 case F2FS_IOC_START_VOLATILE_WRITE:
2872 return f2fs_ioc_start_volatile_write(filp);
2873 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2874 return f2fs_ioc_release_volatile_write(filp);
2875 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2876 return f2fs_ioc_abort_volatile_write(filp);
2877 case F2FS_IOC_SHUTDOWN:
2878 return f2fs_ioc_shutdown(filp, arg);
2879 case FITRIM:
2880 return f2fs_ioc_fitrim(filp, arg);
2881 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2882 return f2fs_ioc_set_encryption_policy(filp, arg);
2883 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2884 return f2fs_ioc_get_encryption_policy(filp, arg);
2885 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2886 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2887 case F2FS_IOC_GARBAGE_COLLECT:
2888 return f2fs_ioc_gc(filp, arg);
2889 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2890 return f2fs_ioc_gc_range(filp, arg);
2891 case F2FS_IOC_WRITE_CHECKPOINT:
2892 return f2fs_ioc_write_checkpoint(filp, arg);
2893 case F2FS_IOC_DEFRAGMENT:
2894 return f2fs_ioc_defragment(filp, arg);
2895 case F2FS_IOC_MOVE_RANGE:
2896 return f2fs_ioc_move_range(filp, arg);
2897 case F2FS_IOC_FLUSH_DEVICE:
2898 return f2fs_ioc_flush_device(filp, arg);
2899 case F2FS_IOC_GET_FEATURES:
2900 return f2fs_ioc_get_features(filp, arg);
2901 case F2FS_IOC_FSGETXATTR:
2902 return f2fs_ioc_fsgetxattr(filp, arg);
2903 case F2FS_IOC_FSSETXATTR:
2904 return f2fs_ioc_fssetxattr(filp, arg);
2905 case F2FS_IOC_GET_PIN_FILE:
2906 return f2fs_ioc_get_pin_file(filp, arg);
2907 case F2FS_IOC_SET_PIN_FILE:
2908 return f2fs_ioc_set_pin_file(filp, arg);
2909 case F2FS_IOC_PRECACHE_EXTENTS:
2910 return f2fs_ioc_precache_extents(filp, arg);
2911 default:
2912 return -ENOTTY;
2913 }
2914 }
2915
2916 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2917 {
2918 struct file *file = iocb->ki_filp;
2919 struct inode *inode = file_inode(file);
2920 struct blk_plug plug;
2921 ssize_t ret;
2922
2923 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2924 return -EIO;
2925
2926 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
2927 return -EINVAL;
2928
2929 if (!inode_trylock(inode)) {
2930 if (iocb->ki_flags & IOCB_NOWAIT)
2931 return -EAGAIN;
2932 inode_lock(inode);
2933 }
2934
2935 ret = generic_write_checks(iocb, from);
2936 if (ret > 0) {
2937 bool preallocated = false;
2938 size_t target_size = 0;
2939 int err;
2940
2941 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2942 set_inode_flag(inode, FI_NO_PREALLOC);
2943
2944 if ((iocb->ki_flags & IOCB_NOWAIT) &&
2945 (iocb->ki_flags & IOCB_DIRECT)) {
2946 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
2947 iov_iter_count(from)) ||
2948 f2fs_has_inline_data(inode) ||
2949 f2fs_force_buffered_io(inode, WRITE)) {
2950 inode_unlock(inode);
2951 return -EAGAIN;
2952 }
2953
2954 } else {
2955 preallocated = true;
2956 target_size = iocb->ki_pos + iov_iter_count(from);
2957
2958 err = f2fs_preallocate_blocks(iocb, from);
2959 if (err) {
2960 clear_inode_flag(inode, FI_NO_PREALLOC);
2961 inode_unlock(inode);
2962 return err;
2963 }
2964 }
2965 blk_start_plug(&plug);
2966 ret = __generic_file_write_iter(iocb, from);
2967 blk_finish_plug(&plug);
2968 clear_inode_flag(inode, FI_NO_PREALLOC);
2969
2970 /* if we couldn't write data, we should deallocate blocks. */
2971 if (preallocated && i_size_read(inode) < target_size)
2972 f2fs_truncate(inode);
2973
2974 if (ret > 0)
2975 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2976 }
2977 inode_unlock(inode);
2978
2979 if (ret > 0)
2980 ret = generic_write_sync(iocb, ret);
2981 return ret;
2982 }
2983
2984 #ifdef CONFIG_COMPAT
2985 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2986 {
2987 switch (cmd) {
2988 case F2FS_IOC32_GETFLAGS:
2989 cmd = F2FS_IOC_GETFLAGS;
2990 break;
2991 case F2FS_IOC32_SETFLAGS:
2992 cmd = F2FS_IOC_SETFLAGS;
2993 break;
2994 case F2FS_IOC32_GETVERSION:
2995 cmd = F2FS_IOC_GETVERSION;
2996 break;
2997 case F2FS_IOC_START_ATOMIC_WRITE:
2998 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2999 case F2FS_IOC_START_VOLATILE_WRITE:
3000 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3001 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3002 case F2FS_IOC_SHUTDOWN:
3003 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3004 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3005 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3006 case F2FS_IOC_GARBAGE_COLLECT:
3007 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3008 case F2FS_IOC_WRITE_CHECKPOINT:
3009 case F2FS_IOC_DEFRAGMENT:
3010 case F2FS_IOC_MOVE_RANGE:
3011 case F2FS_IOC_FLUSH_DEVICE:
3012 case F2FS_IOC_GET_FEATURES:
3013 case F2FS_IOC_FSGETXATTR:
3014 case F2FS_IOC_FSSETXATTR:
3015 case F2FS_IOC_GET_PIN_FILE:
3016 case F2FS_IOC_SET_PIN_FILE:
3017 case F2FS_IOC_PRECACHE_EXTENTS:
3018 break;
3019 default:
3020 return -ENOIOCTLCMD;
3021 }
3022 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3023 }
3024 #endif
3025
3026 const struct file_operations f2fs_file_operations = {
3027 .llseek = f2fs_llseek,
3028 .read_iter = generic_file_read_iter,
3029 .write_iter = f2fs_file_write_iter,
3030 .open = f2fs_file_open,
3031 .release = f2fs_release_file,
3032 .mmap = f2fs_file_mmap,
3033 .flush = f2fs_file_flush,
3034 .fsync = f2fs_sync_file,
3035 .fallocate = f2fs_fallocate,
3036 .unlocked_ioctl = f2fs_ioctl,
3037 #ifdef CONFIG_COMPAT
3038 .compat_ioctl = f2fs_compat_ioctl,
3039 #endif
3040 .splice_read = generic_file_splice_read,
3041 .splice_write = iter_file_splice_write,
3042 };
Motorola kernel-slsi