projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: support FALLOC_FL_COLLAPSE_RANGE
[deliverable/linux.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
24 #include "f2fs.h"
25 #include "node.h"
26 #include "segment.h"
27 #include "xattr.h"
28 #include "acl.h"
29 #include "trace.h"
30 #include <trace/events/f2fs.h>
31
32 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
33 struct vm_fault *vmf)
34 {
35 struct page *page = vmf->page;
36 struct inode *inode = file_inode(vma->vm_file);
37 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
38 struct dnode_of_data dn;
39 int err;
40
41 f2fs_balance_fs(sbi);
42
43 sb_start_pagefault(inode->i_sb);
44
45 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
46
47 /* block allocation */
48 f2fs_lock_op(sbi);
49 set_new_dnode(&dn, inode, NULL, NULL, 0);
50 err = f2fs_reserve_block(&dn, page->index);
51 if (err) {
52 f2fs_unlock_op(sbi);
53 goto out;
54 }
55 f2fs_put_dnode(&dn);
56 f2fs_unlock_op(sbi);
57
58 file_update_time(vma->vm_file);
59 lock_page(page);
60 if (unlikely(page->mapping != inode->i_mapping ||
61 page_offset(page) > i_size_read(inode) ||
62 !PageUptodate(page))) {
63 unlock_page(page);
64 err = -EFAULT;
65 goto out;
66 }
67
68 /*
69 * check to see if the page is mapped already (no holes)
70 */
71 if (PageMappedToDisk(page))
72 goto mapped;
73
74 /* page is wholly or partially inside EOF */
75 if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
76 unsigned offset;
77 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
78 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
79 }
80 set_page_dirty(page);
81 SetPageUptodate(page);
82
83 trace_f2fs_vm_page_mkwrite(page, DATA);
84 mapped:
85 /* fill the page */
86 f2fs_wait_on_page_writeback(page, DATA);
87 out:
88 sb_end_pagefault(inode->i_sb);
89 return block_page_mkwrite_return(err);
90 }
91
92 static const struct vm_operations_struct f2fs_file_vm_ops = {
93 .fault = filemap_fault,
94 .map_pages = filemap_map_pages,
95 .page_mkwrite = f2fs_vm_page_mkwrite,
96 };
97
98 static int get_parent_ino(struct inode *inode, nid_t *pino)
99 {
100 struct dentry *dentry;
101
102 inode = igrab(inode);
103 dentry = d_find_any_alias(inode);
104 iput(inode);
105 if (!dentry)
106 return 0;
107
108 if (update_dent_inode(inode, &dentry->d_name)) {
109 dput(dentry);
110 return 0;
111 }
112
113 *pino = parent_ino(dentry);
114 dput(dentry);
115 return 1;
116 }
117
118 static inline bool need_do_checkpoint(struct inode *inode)
119 {
120 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
121 bool need_cp = false;
122
123 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
124 need_cp = true;
125 else if (file_wrong_pino(inode))
126 need_cp = true;
127 else if (!space_for_roll_forward(sbi))
128 need_cp = true;
129 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
130 need_cp = true;
131 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
132 need_cp = true;
133 else if (test_opt(sbi, FASTBOOT))
134 need_cp = true;
135 else if (sbi->active_logs == 2)
136 need_cp = true;
137
138 return need_cp;
139 }
140
141 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
142 {
143 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
144 bool ret = false;
145 /* But we need to avoid that there are some inode updates */
146 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
147 ret = true;
148 f2fs_put_page(i, 0);
149 return ret;
150 }
151
152 static void try_to_fix_pino(struct inode *inode)
153 {
154 struct f2fs_inode_info *fi = F2FS_I(inode);
155 nid_t pino;
156
157 down_write(&fi->i_sem);
158 fi->xattr_ver = 0;
159 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
160 get_parent_ino(inode, &pino)) {
161 fi->i_pino = pino;
162 file_got_pino(inode);
163 up_write(&fi->i_sem);
164
165 mark_inode_dirty_sync(inode);
166 f2fs_write_inode(inode, NULL);
167 } else {
168 up_write(&fi->i_sem);
169 }
170 }
171
172 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
173 {
174 struct inode *inode = file->f_mapping->host;
175 struct f2fs_inode_info *fi = F2FS_I(inode);
176 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
177 nid_t ino = inode->i_ino;
178 int ret = 0;
179 bool need_cp = false;
180 struct writeback_control wbc = {
181 .sync_mode = WB_SYNC_ALL,
182 .nr_to_write = LONG_MAX,
183 .for_reclaim = 0,
184 };
185
186 if (unlikely(f2fs_readonly(inode->i_sb)))
187 return 0;
188
189 trace_f2fs_sync_file_enter(inode);
190
191 /* if fdatasync is triggered, let's do in-place-update */
192 if (get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
193 set_inode_flag(fi, FI_NEED_IPU);
194 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
195 clear_inode_flag(fi, FI_NEED_IPU);
196
197 if (ret) {
198 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
199 return ret;
200 }
201
202 /* if the inode is dirty, let's recover all the time */
203 if (!datasync && is_inode_flag_set(fi, FI_DIRTY_INODE)) {
204 update_inode_page(inode);
205 goto go_write;
206 }
207
208 /*
209 * if there is no written data, don't waste time to write recovery info.
210 */
211 if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
212 !exist_written_data(sbi, ino, APPEND_INO)) {
213
214 /* it may call write_inode just prior to fsync */
215 if (need_inode_page_update(sbi, ino))
216 goto go_write;
217
218 if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
219 exist_written_data(sbi, ino, UPDATE_INO))
220 goto flush_out;
221 goto out;
222 }
223 go_write:
224 /* guarantee free sections for fsync */
225 f2fs_balance_fs(sbi);
226
227 /*
228 * Both of fdatasync() and fsync() are able to be recovered from
229 * sudden-power-off.
230 */
231 down_read(&fi->i_sem);
232 need_cp = need_do_checkpoint(inode);
233 up_read(&fi->i_sem);
234
235 if (need_cp) {
236 /* all the dirty node pages should be flushed for POR */
237 ret = f2fs_sync_fs(inode->i_sb, 1);
238
239 /*
240 * We've secured consistency through sync_fs. Following pino
241 * will be used only for fsynced inodes after checkpoint.
242 */
243 try_to_fix_pino(inode);
244 clear_inode_flag(fi, FI_APPEND_WRITE);
245 clear_inode_flag(fi, FI_UPDATE_WRITE);
246 goto out;
247 }
248 sync_nodes:
249 sync_node_pages(sbi, ino, &wbc);
250
251 /* if cp_error was enabled, we should avoid infinite loop */
252 if (unlikely(f2fs_cp_error(sbi)))
253 goto out;
254
255 if (need_inode_block_update(sbi, ino)) {
256 mark_inode_dirty_sync(inode);
257 f2fs_write_inode(inode, NULL);
258 goto sync_nodes;
259 }
260
261 ret = wait_on_node_pages_writeback(sbi, ino);
262 if (ret)
263 goto out;
264
265 /* once recovery info is written, don't need to tack this */
266 remove_dirty_inode(sbi, ino, APPEND_INO);
267 clear_inode_flag(fi, FI_APPEND_WRITE);
268 flush_out:
269 remove_dirty_inode(sbi, ino, UPDATE_INO);
270 clear_inode_flag(fi, FI_UPDATE_WRITE);
271 ret = f2fs_issue_flush(sbi);
272 out:
273 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
274 f2fs_trace_ios(NULL, 1);
275 return ret;
276 }
277
278 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
279 pgoff_t pgofs, int whence)
280 {
281 struct pagevec pvec;
282 int nr_pages;
283
284 if (whence != SEEK_DATA)
285 return 0;
286
287 /* find first dirty page index */
288 pagevec_init(&pvec, 0);
289 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
290 PAGECACHE_TAG_DIRTY, 1);
291 pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
292 pagevec_release(&pvec);
293 return pgofs;
294 }
295
296 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
297 int whence)
298 {
299 switch (whence) {
300 case SEEK_DATA:
301 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
302 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
303 return true;
304 break;
305 case SEEK_HOLE:
306 if (blkaddr == NULL_ADDR)
307 return true;
308 break;
309 }
310 return false;
311 }
312
313 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
314 {
315 struct inode *inode = file->f_mapping->host;
316 loff_t maxbytes = inode->i_sb->s_maxbytes;
317 struct dnode_of_data dn;
318 pgoff_t pgofs, end_offset, dirty;
319 loff_t data_ofs = offset;
320 loff_t isize;
321 int err = 0;
322
323 mutex_lock(&inode->i_mutex);
324
325 isize = i_size_read(inode);
326 if (offset >= isize)
327 goto fail;
328
329 /* handle inline data case */
330 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
331 if (whence == SEEK_HOLE)
332 data_ofs = isize;
333 goto found;
334 }
335
336 pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT);
337
338 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
339
340 for (; data_ofs < isize; data_ofs = pgofs << PAGE_CACHE_SHIFT) {
341 set_new_dnode(&dn, inode, NULL, NULL, 0);
342 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
343 if (err && err != -ENOENT) {
344 goto fail;
345 } else if (err == -ENOENT) {
346 /* direct node does not exists */
347 if (whence == SEEK_DATA) {
348 pgofs = PGOFS_OF_NEXT_DNODE(pgofs,
349 F2FS_I(inode));
350 continue;
351 } else {
352 goto found;
353 }
354 }
355
356 end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
357
358 /* find data/hole in dnode block */
359 for (; dn.ofs_in_node < end_offset;
360 dn.ofs_in_node++, pgofs++,
361 data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
362 block_t blkaddr;
363 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
364
365 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
366 f2fs_put_dnode(&dn);
367 goto found;
368 }
369 }
370 f2fs_put_dnode(&dn);
371 }
372
373 if (whence == SEEK_DATA)
374 goto fail;
375 found:
376 if (whence == SEEK_HOLE && data_ofs > isize)
377 data_ofs = isize;
378 mutex_unlock(&inode->i_mutex);
379 return vfs_setpos(file, data_ofs, maxbytes);
380 fail:
381 mutex_unlock(&inode->i_mutex);
382 return -ENXIO;
383 }
384
385 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
386 {
387 struct inode *inode = file->f_mapping->host;
388 loff_t maxbytes = inode->i_sb->s_maxbytes;
389
390 switch (whence) {
391 case SEEK_SET:
392 case SEEK_CUR:
393 case SEEK_END:
394 return generic_file_llseek_size(file, offset, whence,
395 maxbytes, i_size_read(inode));
396 case SEEK_DATA:
397 case SEEK_HOLE:
398 if (offset < 0)
399 return -ENXIO;
400 return f2fs_seek_block(file, offset, whence);
401 }
402
403 return -EINVAL;
404 }
405
406 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
407 {
408 struct inode *inode = file_inode(file);
409
410 /* we don't need to use inline_data strictly */
411 if (f2fs_has_inline_data(inode)) {
412 int err = f2fs_convert_inline_inode(inode);
413 if (err)
414 return err;
415 }
416
417 file_accessed(file);
418 vma->vm_ops = &f2fs_file_vm_ops;
419 return 0;
420 }
421
422 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
423 {
424 int nr_free = 0, ofs = dn->ofs_in_node;
425 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
426 struct f2fs_node *raw_node;
427 __le32 *addr;
428
429 raw_node = F2FS_NODE(dn->node_page);
430 addr = blkaddr_in_node(raw_node) + ofs;
431
432 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
433 block_t blkaddr = le32_to_cpu(*addr);
434 if (blkaddr == NULL_ADDR)
435 continue;
436
437 dn->data_blkaddr = NULL_ADDR;
438 set_data_blkaddr(dn);
439 f2fs_update_extent_cache(dn);
440 invalidate_blocks(sbi, blkaddr);
441 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
442 clear_inode_flag(F2FS_I(dn->inode),
443 FI_FIRST_BLOCK_WRITTEN);
444 nr_free++;
445 }
446 if (nr_free) {
447 dec_valid_block_count(sbi, dn->inode, nr_free);
448 set_page_dirty(dn->node_page);
449 sync_inode_page(dn);
450 }
451 dn->ofs_in_node = ofs;
452
453 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
454 dn->ofs_in_node, nr_free);
455 return nr_free;
456 }
457
458 void truncate_data_blocks(struct dnode_of_data *dn)
459 {
460 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
461 }
462
463 static int truncate_partial_data_page(struct inode *inode, u64 from,
464 bool cache_only)
465 {
466 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
467 pgoff_t index = from >> PAGE_CACHE_SHIFT;
468 struct address_space *mapping = inode->i_mapping;
469 struct page *page;
470
471 if (!offset && !cache_only)
472 return 0;
473
474 if (cache_only) {
475 page = grab_cache_page(mapping, index);
476 if (page && PageUptodate(page))
477 goto truncate_out;
478 f2fs_put_page(page, 1);
479 return 0;
480 }
481
482 page = get_lock_data_page(inode, index);
483 if (IS_ERR(page))
484 return 0;
485 truncate_out:
486 f2fs_wait_on_page_writeback(page, DATA);
487 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
488 if (!cache_only)
489 set_page_dirty(page);
490 f2fs_put_page(page, 1);
491 return 0;
492 }
493
494 int truncate_blocks(struct inode *inode, u64 from, bool lock)
495 {
496 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
497 unsigned int blocksize = inode->i_sb->s_blocksize;
498 struct dnode_of_data dn;
499 pgoff_t free_from;
500 int count = 0, err = 0;
501 struct page *ipage;
502 bool truncate_page = false;
503
504 trace_f2fs_truncate_blocks_enter(inode, from);
505
506 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
507
508 if (lock)
509 f2fs_lock_op(sbi);
510
511 ipage = get_node_page(sbi, inode->i_ino);
512 if (IS_ERR(ipage)) {
513 err = PTR_ERR(ipage);
514 goto out;
515 }
516
517 if (f2fs_has_inline_data(inode)) {
518 if (truncate_inline_inode(ipage, from))
519 set_page_dirty(ipage);
520 f2fs_put_page(ipage, 1);
521 truncate_page = true;
522 goto out;
523 }
524
525 set_new_dnode(&dn, inode, ipage, NULL, 0);
526 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
527 if (err) {
528 if (err == -ENOENT)
529 goto free_next;
530 goto out;
531 }
532
533 count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
534
535 count -= dn.ofs_in_node;
536 f2fs_bug_on(sbi, count < 0);
537
538 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
539 truncate_data_blocks_range(&dn, count);
540 free_from += count;
541 }
542
543 f2fs_put_dnode(&dn);
544 free_next:
545 err = truncate_inode_blocks(inode, free_from);
546 out:
547 if (lock)
548 f2fs_unlock_op(sbi);
549
550 /* lastly zero out the first data page */
551 if (!err)
552 err = truncate_partial_data_page(inode, from, truncate_page);
553
554 trace_f2fs_truncate_blocks_exit(inode, err);
555 return err;
556 }
557
558 void f2fs_truncate(struct inode *inode)
559 {
560 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
561 S_ISLNK(inode->i_mode)))
562 return;
563
564 trace_f2fs_truncate(inode);
565
566 /* we should check inline_data size */
567 if (f2fs_has_inline_data(inode) && !f2fs_may_inline_data(inode)) {
568 if (f2fs_convert_inline_inode(inode))
569 return;
570 }
571
572 if (!truncate_blocks(inode, i_size_read(inode), true)) {
573 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
574 mark_inode_dirty(inode);
575 }
576 }
577
578 int f2fs_getattr(struct vfsmount *mnt,
579 struct dentry *dentry, struct kstat *stat)
580 {
581 struct inode *inode = d_inode(dentry);
582 generic_fillattr(inode, stat);
583 stat->blocks <<= 3;
584 return 0;
585 }
586
587 #ifdef CONFIG_F2FS_FS_POSIX_ACL
588 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
589 {
590 struct f2fs_inode_info *fi = F2FS_I(inode);
591 unsigned int ia_valid = attr->ia_valid;
592
593 if (ia_valid & ATTR_UID)
594 inode->i_uid = attr->ia_uid;
595 if (ia_valid & ATTR_GID)
596 inode->i_gid = attr->ia_gid;
597 if (ia_valid & ATTR_ATIME)
598 inode->i_atime = timespec_trunc(attr->ia_atime,
599 inode->i_sb->s_time_gran);
600 if (ia_valid & ATTR_MTIME)
601 inode->i_mtime = timespec_trunc(attr->ia_mtime,
602 inode->i_sb->s_time_gran);
603 if (ia_valid & ATTR_CTIME)
604 inode->i_ctime = timespec_trunc(attr->ia_ctime,
605 inode->i_sb->s_time_gran);
606 if (ia_valid & ATTR_MODE) {
607 umode_t mode = attr->ia_mode;
608
609 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
610 mode &= ~S_ISGID;
611 set_acl_inode(fi, mode);
612 }
613 }
614 #else
615 #define __setattr_copy setattr_copy
616 #endif
617
618 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
619 {
620 struct inode *inode = d_inode(dentry);
621 struct f2fs_inode_info *fi = F2FS_I(inode);
622 int err;
623
624 err = inode_change_ok(inode, attr);
625 if (err)
626 return err;
627
628 if (attr->ia_valid & ATTR_SIZE) {
629 if (attr->ia_size != i_size_read(inode)) {
630 truncate_setsize(inode, attr->ia_size);
631 f2fs_truncate(inode);
632 f2fs_balance_fs(F2FS_I_SB(inode));
633 } else {
634 /*
635 * giving a chance to truncate blocks past EOF which
636 * are fallocated with FALLOC_FL_KEEP_SIZE.
637 */
638 f2fs_truncate(inode);
639 }
640 }
641
642 __setattr_copy(inode, attr);
643
644 if (attr->ia_valid & ATTR_MODE) {
645 err = posix_acl_chmod(inode, get_inode_mode(inode));
646 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
647 inode->i_mode = fi->i_acl_mode;
648 clear_inode_flag(fi, FI_ACL_MODE);
649 }
650 }
651
652 mark_inode_dirty(inode);
653 return err;
654 }
655
656 const struct inode_operations f2fs_file_inode_operations = {
657 .getattr = f2fs_getattr,
658 .setattr = f2fs_setattr,
659 .get_acl = f2fs_get_acl,
660 .set_acl = f2fs_set_acl,
661 #ifdef CONFIG_F2FS_FS_XATTR
662 .setxattr = generic_setxattr,
663 .getxattr = generic_getxattr,
664 .listxattr = f2fs_listxattr,
665 .removexattr = generic_removexattr,
666 #endif
667 .fiemap = f2fs_fiemap,
668 };
669
670 static void fill_zero(struct inode *inode, pgoff_t index,
671 loff_t start, loff_t len)
672 {
673 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
674 struct page *page;
675
676 if (!len)
677 return;
678
679 f2fs_balance_fs(sbi);
680
681 f2fs_lock_op(sbi);
682 page = get_new_data_page(inode, NULL, index, false);
683 f2fs_unlock_op(sbi);
684
685 if (!IS_ERR(page)) {
686 f2fs_wait_on_page_writeback(page, DATA);
687 zero_user(page, start, len);
688 set_page_dirty(page);
689 f2fs_put_page(page, 1);
690 }
691 }
692
693 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
694 {
695 pgoff_t index;
696 int err;
697
698 for (index = pg_start; index < pg_end; index++) {
699 struct dnode_of_data dn;
700
701 set_new_dnode(&dn, inode, NULL, NULL, 0);
702 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
703 if (err) {
704 if (err == -ENOENT)
705 continue;
706 return err;
707 }
708
709 if (dn.data_blkaddr != NULL_ADDR)
710 truncate_data_blocks_range(&dn, 1);
711 f2fs_put_dnode(&dn);
712 }
713 return 0;
714 }
715
716 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
717 {
718 pgoff_t pg_start, pg_end;
719 loff_t off_start, off_end;
720 int ret = 0;
721
722 if (!S_ISREG(inode->i_mode))
723 return -EOPNOTSUPP;
724
725 if (f2fs_has_inline_data(inode)) {
726 ret = f2fs_convert_inline_inode(inode);
727 if (ret)
728 return ret;
729 }
730
731 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
732 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
733
734 off_start = offset & (PAGE_CACHE_SIZE - 1);
735 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
736
737 if (pg_start == pg_end) {
738 fill_zero(inode, pg_start, off_start,
739 off_end - off_start);
740 } else {
741 if (off_start)
742 fill_zero(inode, pg_start++, off_start,
743 PAGE_CACHE_SIZE - off_start);
744 if (off_end)
745 fill_zero(inode, pg_end, 0, off_end);
746
747 if (pg_start < pg_end) {
748 struct address_space *mapping = inode->i_mapping;
749 loff_t blk_start, blk_end;
750 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
751
752 f2fs_balance_fs(sbi);
753
754 blk_start = pg_start << PAGE_CACHE_SHIFT;
755 blk_end = pg_end << PAGE_CACHE_SHIFT;
756 truncate_inode_pages_range(mapping, blk_start,
757 blk_end - 1);
758
759 f2fs_lock_op(sbi);
760 ret = truncate_hole(inode, pg_start, pg_end);
761 f2fs_unlock_op(sbi);
762 }
763 }
764
765 return ret;
766 }
767
768 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
769 {
770 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
771 struct dnode_of_data dn;
772 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
773 int ret = 0;
774
775 f2fs_lock_op(sbi);
776
777 for (; end < nrpages; start++, end++) {
778 block_t new_addr, old_addr;
779
780 set_new_dnode(&dn, inode, NULL, NULL, 0);
781 ret = get_dnode_of_data(&dn, end, LOOKUP_NODE_RA);
782 if (ret && ret != -ENOENT) {
783 goto out;
784 } else if (ret == -ENOENT) {
785 new_addr = NULL_ADDR;
786 } else {
787 new_addr = dn.data_blkaddr;
788 truncate_data_blocks_range(&dn, 1);
789 f2fs_put_dnode(&dn);
790 }
791
792 if (new_addr == NULL_ADDR) {
793 set_new_dnode(&dn, inode, NULL, NULL, 0);
794 ret = get_dnode_of_data(&dn, start, LOOKUP_NODE_RA);
795 if (ret && ret != -ENOENT)
796 goto out;
797 else if (ret == -ENOENT)
798 continue;
799
800 if (dn.data_blkaddr == NULL_ADDR) {
801 f2fs_put_dnode(&dn);
802 continue;
803 } else {
804 truncate_data_blocks_range(&dn, 1);
805 }
806
807 f2fs_put_dnode(&dn);
808 } else {
809 struct page *ipage;
810
811 ipage = get_node_page(sbi, inode->i_ino);
812 if (IS_ERR(ipage)) {
813 ret = PTR_ERR(ipage);
814 goto out;
815 }
816
817 set_new_dnode(&dn, inode, ipage, NULL, 0);
818 ret = f2fs_reserve_block(&dn, start);
819 if (ret)
820 goto out;
821
822 old_addr = dn.data_blkaddr;
823 if (old_addr != NEW_ADDR && new_addr == NEW_ADDR) {
824 dn.data_blkaddr = NULL_ADDR;
825 f2fs_update_extent_cache(&dn);
826 invalidate_blocks(sbi, old_addr);
827
828 dn.data_blkaddr = new_addr;
829 set_data_blkaddr(&dn);
830 } else if (new_addr != NEW_ADDR) {
831 struct node_info ni;
832 struct f2fs_summary sum;
833
834 get_node_info(sbi, dn.nid, &ni);
835 set_summary(&sum, dn.nid, dn.ofs_in_node,
836 ni.version);
837
838 f2fs_replace_block(sbi, &sum, old_addr,
839 new_addr, true);
840
841 dn.data_blkaddr = new_addr;
842 set_data_blkaddr(&dn);
843 f2fs_update_extent_cache(&dn);
844 }
845
846 f2fs_put_dnode(&dn);
847 }
848 }
849 ret = 0;
850 out:
851 f2fs_unlock_op(sbi);
852 return ret;
853 }
854
855 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
856 {
857 pgoff_t pg_start, pg_end;
858 loff_t new_size;
859 int ret;
860
861 if (!S_ISREG(inode->i_mode))
862 return -EINVAL;
863
864 if (offset + len >= i_size_read(inode))
865 return -EINVAL;
866
867 /* collapse range should be aligned to block size of f2fs. */
868 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
869 return -EINVAL;
870
871 pg_start = offset >> PAGE_CACHE_SHIFT;
872 pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
873
874 /* write out all dirty pages from offset */
875 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
876 if (ret)
877 return ret;
878
879 truncate_pagecache(inode, offset);
880
881 ret = f2fs_do_collapse(inode, pg_start, pg_end);
882 if (ret)
883 return ret;
884
885 new_size = i_size_read(inode) - len;
886
887 ret = truncate_blocks(inode, new_size, true);
888 if (!ret)
889 i_size_write(inode, new_size);
890
891 return ret;
892 }
893
894 static int expand_inode_data(struct inode *inode, loff_t offset,
895 loff_t len, int mode)
896 {
897 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
898 pgoff_t index, pg_start, pg_end;
899 loff_t new_size = i_size_read(inode);
900 loff_t off_start, off_end;
901 int ret = 0;
902
903 f2fs_balance_fs(sbi);
904
905 ret = inode_newsize_ok(inode, (len + offset));
906 if (ret)
907 return ret;
908
909 if (f2fs_has_inline_data(inode)) {
910 ret = f2fs_convert_inline_inode(inode);
911 if (ret)
912 return ret;
913 }
914
915 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
916 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
917
918 off_start = offset & (PAGE_CACHE_SIZE - 1);
919 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
920
921 f2fs_lock_op(sbi);
922
923 for (index = pg_start; index <= pg_end; index++) {
924 struct dnode_of_data dn;
925
926 if (index == pg_end && !off_end)
927 goto noalloc;
928
929 set_new_dnode(&dn, inode, NULL, NULL, 0);
930 ret = f2fs_reserve_block(&dn, index);
931 if (ret)
932 break;
933 noalloc:
934 if (pg_start == pg_end)
935 new_size = offset + len;
936 else if (index == pg_start && off_start)
937 new_size = (index + 1) << PAGE_CACHE_SHIFT;
938 else if (index == pg_end)
939 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
940 else
941 new_size += PAGE_CACHE_SIZE;
942 }
943
944 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
945 i_size_read(inode) < new_size) {
946 i_size_write(inode, new_size);
947 mark_inode_dirty(inode);
948 update_inode_page(inode);
949 }
950 f2fs_unlock_op(sbi);
951
952 return ret;
953 }
954
955 static long f2fs_fallocate(struct file *file, int mode,
956 loff_t offset, loff_t len)
957 {
958 struct inode *inode = file_inode(file);
959 long ret = 0;
960
961 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
962 FALLOC_FL_COLLAPSE_RANGE))
963 return -EOPNOTSUPP;
964
965 mutex_lock(&inode->i_mutex);
966
967 if (mode & FALLOC_FL_PUNCH_HOLE) {
968 if (offset >= inode->i_size)
969 goto out;
970
971 ret = punch_hole(inode, offset, len);
972 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
973 ret = f2fs_collapse_range(inode, offset, len);
974 } else {
975 ret = expand_inode_data(inode, offset, len, mode);
976 }
977
978 if (!ret) {
979 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
980 mark_inode_dirty(inode);
981 }
982
983 out:
984 mutex_unlock(&inode->i_mutex);
985
986 trace_f2fs_fallocate(inode, mode, offset, len, ret);
987 return ret;
988 }
989
990 static int f2fs_release_file(struct inode *inode, struct file *filp)
991 {
992 /* some remained atomic pages should discarded */
993 if (f2fs_is_atomic_file(inode))
994 commit_inmem_pages(inode, true);
995 if (f2fs_is_volatile_file(inode)) {
996 set_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
997 filemap_fdatawrite(inode->i_mapping);
998 clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE);
999 }
1000 return 0;
1001 }
1002
1003 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1004 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
1005
1006 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1007 {
1008 if (S_ISDIR(mode))
1009 return flags;
1010 else if (S_ISREG(mode))
1011 return flags & F2FS_REG_FLMASK;
1012 else
1013 return flags & F2FS_OTHER_FLMASK;
1014 }
1015
1016 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1017 {
1018 struct inode *inode = file_inode(filp);
1019 struct f2fs_inode_info *fi = F2FS_I(inode);
1020 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1021 return put_user(flags, (int __user *)arg);
1022 }
1023
1024 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1025 {
1026 struct inode *inode = file_inode(filp);
1027 struct f2fs_inode_info *fi = F2FS_I(inode);
1028 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1029 unsigned int oldflags;
1030 int ret;
1031
1032 ret = mnt_want_write_file(filp);
1033 if (ret)
1034 return ret;
1035
1036 if (!inode_owner_or_capable(inode)) {
1037 ret = -EACCES;
1038 goto out;
1039 }
1040
1041 if (get_user(flags, (int __user *)arg)) {
1042 ret = -EFAULT;
1043 goto out;
1044 }
1045
1046 flags = f2fs_mask_flags(inode->i_mode, flags);
1047
1048 mutex_lock(&inode->i_mutex);
1049
1050 oldflags = fi->i_flags;
1051
1052 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1053 if (!capable(CAP_LINUX_IMMUTABLE)) {
1054 mutex_unlock(&inode->i_mutex);
1055 ret = -EPERM;
1056 goto out;
1057 }
1058 }
1059
1060 flags = flags & FS_FL_USER_MODIFIABLE;
1061 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1062 fi->i_flags = flags;
1063 mutex_unlock(&inode->i_mutex);
1064
1065 f2fs_set_inode_flags(inode);
1066 inode->i_ctime = CURRENT_TIME;
1067 mark_inode_dirty(inode);
1068 out:
1069 mnt_drop_write_file(filp);
1070 return ret;
1071 }
1072
1073 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1074 {
1075 struct inode *inode = file_inode(filp);
1076
1077 return put_user(inode->i_generation, (int __user *)arg);
1078 }
1079
1080 static int f2fs_ioc_start_atomic_write(struct file *filp)
1081 {
1082 struct inode *inode = file_inode(filp);
1083
1084 if (!inode_owner_or_capable(inode))
1085 return -EACCES;
1086
1087 f2fs_balance_fs(F2FS_I_SB(inode));
1088
1089 if (f2fs_is_atomic_file(inode))
1090 return 0;
1091
1092 set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1093
1094 return f2fs_convert_inline_inode(inode);
1095 }
1096
1097 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1098 {
1099 struct inode *inode = file_inode(filp);
1100 int ret;
1101
1102 if (!inode_owner_or_capable(inode))
1103 return -EACCES;
1104
1105 if (f2fs_is_volatile_file(inode))
1106 return 0;
1107
1108 ret = mnt_want_write_file(filp);
1109 if (ret)
1110 return ret;
1111
1112 if (f2fs_is_atomic_file(inode))
1113 commit_inmem_pages(inode, false);
1114
1115 ret = f2fs_sync_file(filp, 0, LONG_MAX, 0);
1116 mnt_drop_write_file(filp);
1117 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1118 return ret;
1119 }
1120
1121 static int f2fs_ioc_start_volatile_write(struct file *filp)
1122 {
1123 struct inode *inode = file_inode(filp);
1124
1125 if (!inode_owner_or_capable(inode))
1126 return -EACCES;
1127
1128 if (f2fs_is_volatile_file(inode))
1129 return 0;
1130
1131 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1132
1133 return f2fs_convert_inline_inode(inode);
1134 }
1135
1136 static int f2fs_ioc_release_volatile_write(struct file *filp)
1137 {
1138 struct inode *inode = file_inode(filp);
1139
1140 if (!inode_owner_or_capable(inode))
1141 return -EACCES;
1142
1143 if (!f2fs_is_volatile_file(inode))
1144 return 0;
1145
1146 if (!f2fs_is_first_block_written(inode))
1147 return truncate_partial_data_page(inode, 0, true);
1148
1149 punch_hole(inode, 0, F2FS_BLKSIZE);
1150 return 0;
1151 }
1152
1153 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1154 {
1155 struct inode *inode = file_inode(filp);
1156 int ret;
1157
1158 if (!inode_owner_or_capable(inode))
1159 return -EACCES;
1160
1161 ret = mnt_want_write_file(filp);
1162 if (ret)
1163 return ret;
1164
1165 f2fs_balance_fs(F2FS_I_SB(inode));
1166
1167 if (f2fs_is_atomic_file(inode)) {
1168 commit_inmem_pages(inode, false);
1169 clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
1170 }
1171
1172 if (f2fs_is_volatile_file(inode)) {
1173 clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1174 filemap_fdatawrite(inode->i_mapping);
1175 set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
1176 }
1177 mnt_drop_write_file(filp);
1178 return ret;
1179 }
1180
1181 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1182 {
1183 struct inode *inode = file_inode(filp);
1184 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1185 struct super_block *sb = sbi->sb;
1186 __u32 in;
1187
1188 if (!capable(CAP_SYS_ADMIN))
1189 return -EPERM;
1190
1191 if (get_user(in, (__u32 __user *)arg))
1192 return -EFAULT;
1193
1194 switch (in) {
1195 case F2FS_GOING_DOWN_FULLSYNC:
1196 sb = freeze_bdev(sb->s_bdev);
1197 if (sb && !IS_ERR(sb)) {
1198 f2fs_stop_checkpoint(sbi);
1199 thaw_bdev(sb->s_bdev, sb);
1200 }
1201 break;
1202 case F2FS_GOING_DOWN_METASYNC:
1203 /* do checkpoint only */
1204 f2fs_sync_fs(sb, 1);
1205 f2fs_stop_checkpoint(sbi);
1206 break;
1207 case F2FS_GOING_DOWN_NOSYNC:
1208 f2fs_stop_checkpoint(sbi);
1209 break;
1210 default:
1211 return -EINVAL;
1212 }
1213 return 0;
1214 }
1215
1216 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1217 {
1218 struct inode *inode = file_inode(filp);
1219 struct super_block *sb = inode->i_sb;
1220 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1221 struct fstrim_range range;
1222 int ret;
1223
1224 if (!capable(CAP_SYS_ADMIN))
1225 return -EPERM;
1226
1227 if (!blk_queue_discard(q))
1228 return -EOPNOTSUPP;
1229
1230 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1231 sizeof(range)))
1232 return -EFAULT;
1233
1234 range.minlen = max((unsigned int)range.minlen,
1235 q->limits.discard_granularity);
1236 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1237 if (ret < 0)
1238 return ret;
1239
1240 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1241 sizeof(range)))
1242 return -EFAULT;
1243 return 0;
1244 }
1245
1246 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1247 {
1248 switch (cmd) {
1249 case F2FS_IOC_GETFLAGS:
1250 return f2fs_ioc_getflags(filp, arg);
1251 case F2FS_IOC_SETFLAGS:
1252 return f2fs_ioc_setflags(filp, arg);
1253 case F2FS_IOC_GETVERSION:
1254 return f2fs_ioc_getversion(filp, arg);
1255 case F2FS_IOC_START_ATOMIC_WRITE:
1256 return f2fs_ioc_start_atomic_write(filp);
1257 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
1258 return f2fs_ioc_commit_atomic_write(filp);
1259 case F2FS_IOC_START_VOLATILE_WRITE:
1260 return f2fs_ioc_start_volatile_write(filp);
1261 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1262 return f2fs_ioc_release_volatile_write(filp);
1263 case F2FS_IOC_ABORT_VOLATILE_WRITE:
1264 return f2fs_ioc_abort_volatile_write(filp);
1265 case F2FS_IOC_SHUTDOWN:
1266 return f2fs_ioc_shutdown(filp, arg);
1267 case FITRIM:
1268 return f2fs_ioc_fitrim(filp, arg);
1269 default:
1270 return -ENOTTY;
1271 }
1272 }
1273
1274 #ifdef CONFIG_COMPAT
1275 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1276 {
1277 switch (cmd) {
1278 case F2FS_IOC32_GETFLAGS:
1279 cmd = F2FS_IOC_GETFLAGS;
1280 break;
1281 case F2FS_IOC32_SETFLAGS:
1282 cmd = F2FS_IOC_SETFLAGS;
1283 break;
1284 default:
1285 return -ENOIOCTLCMD;
1286 }
1287 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1288 }
1289 #endif
1290
1291 const struct file_operations f2fs_file_operations = {
1292 .llseek = f2fs_llseek,
1293 .read_iter = generic_file_read_iter,
1294 .write_iter = generic_file_write_iter,
1295 .open = generic_file_open,
1296 .release = f2fs_release_file,
1297 .mmap = f2fs_file_mmap,
1298 .fsync = f2fs_sync_file,
1299 .fallocate = f2fs_fallocate,
1300 .unlocked_ioctl = f2fs_ioctl,
1301 #ifdef CONFIG_COMPAT
1302 .compat_ioctl = f2fs_compat_ioctl,
1303 #endif
1304 .splice_read = generic_file_splice_read,
1305 .splice_write = iter_file_splice_write,
1306 };
Linux kernel - Deliverables
This page took 0.105998 seconds and 6 git commands to generate.