?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
vfs: allow O_PATH file descriptors for fstatfs()
[linux-2.6.git] / fs / f2fs / file.c
blob02c906971cc6311f43527a6adfb9f75ee77eae9e
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
23 #include "f2fs.h"
24 #include "node.h"
25 #include "segment.h"
26 #include "xattr.h"
27 #include "acl.h"
28 #include <trace/events/f2fs.h>
29
30 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
31 struct vm_fault *vmf)
32 {
33 struct page *page = vmf->page;
34 struct inode *inode = file_inode(vma->vm_file);
35 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
36 block_t old_blk_addr;
37 struct dnode_of_data dn;
38 int err, ilock;
39
40 f2fs_balance_fs(sbi);
41
42 sb_start_pagefault(inode->i_sb);
43
44 /* block allocation */
45 ilock = mutex_lock_op(sbi);
46 set_new_dnode(&dn, inode, NULL, NULL, 0);
47 err = get_dnode_of_data(&dn, page->index, ALLOC_NODE);
48 if (err) {
49 mutex_unlock_op(sbi, ilock);
50 goto out;
51 }
52
53 old_blk_addr = dn.data_blkaddr;
54
55 if (old_blk_addr == NULL_ADDR) {
56 err = reserve_new_block(&dn);
57 if (err) {
58 f2fs_put_dnode(&dn);
59 mutex_unlock_op(sbi, ilock);
60 goto out;
61 }
62 }
63 f2fs_put_dnode(&dn);
64 mutex_unlock_op(sbi, ilock);
65
66 file_update_time(vma->vm_file);
67 lock_page(page);
68 if (page->mapping != inode->i_mapping ||
69 page_offset(page) > i_size_read(inode) ||
70 !PageUptodate(page)) {
71 unlock_page(page);
72 err = -EFAULT;
73 goto out;
74 }
75
76 /*
77 * check to see if the page is mapped already (no holes)
78 */
79 if (PageMappedToDisk(page))
80 goto mapped;
81
82 /* page is wholly or partially inside EOF */
83 if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
84 unsigned offset;
85 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
86 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
87 }
88 set_page_dirty(page);
89 SetPageUptodate(page);
90
91 mapped:
92 /* fill the page */
93 wait_on_page_writeback(page);
94 out:
95 sb_end_pagefault(inode->i_sb);
96 return block_page_mkwrite_return(err);
97 }
98
99 static const struct vm_operations_struct f2fs_file_vm_ops = {
100 .fault = filemap_fault,
101 .page_mkwrite = f2fs_vm_page_mkwrite,
102 .remap_pages = generic_file_remap_pages,
103 };
104
105 static int get_parent_ino(struct inode *inode, nid_t *pino)
106 {
107 struct dentry *dentry;
108
109 inode = igrab(inode);
110 dentry = d_find_any_alias(inode);
111 iput(inode);
112 if (!dentry)
113 return 0;
114
115 if (update_dent_inode(inode, &dentry->d_name)) {
116 dput(dentry);
117 return 0;
118 }
119
120 *pino = parent_ino(dentry);
121 dput(dentry);
122 return 1;
123 }
124
125 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
126 {
127 struct inode *inode = file->f_mapping->host;
128 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
129 int ret = 0;
130 bool need_cp = false;
131 struct writeback_control wbc = {
132 .sync_mode = WB_SYNC_ALL,
133 .nr_to_write = LONG_MAX,
134 .for_reclaim = 0,
135 };
136
137 if (f2fs_readonly(inode->i_sb))
138 return 0;
139
140 trace_f2fs_sync_file_enter(inode);
141 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
142 if (ret) {
143 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
144 return ret;
145 }
146
147 /* guarantee free sections for fsync */
148 f2fs_balance_fs(sbi);
149
150 mutex_lock(&inode->i_mutex);
151
152 /*
153 * Both of fdatasync() and fsync() are able to be recovered from
154 * sudden-power-off.
155 */
156 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
157 need_cp = true;
158 else if (file_wrong_pino(inode))
159 need_cp = true;
160 else if (!space_for_roll_forward(sbi))
161 need_cp = true;
162 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
163 need_cp = true;
164 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
165 need_cp = true;
166
167 if (need_cp) {
168 nid_t pino;
169
170 F2FS_I(inode)->xattr_ver = 0;
171
172 /* all the dirty node pages should be flushed for POR */
173 ret = f2fs_sync_fs(inode->i_sb, 1);
174 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
175 get_parent_ino(inode, &pino)) {
176 F2FS_I(inode)->i_pino = pino;
177 file_got_pino(inode);
178 mark_inode_dirty_sync(inode);
179 ret = f2fs_write_inode(inode, NULL);
180 if (ret)
181 goto out;
182 }
183 } else {
184 /* if there is no written node page, write its inode page */
185 while (!sync_node_pages(sbi, inode->i_ino, &wbc)) {
186 mark_inode_dirty_sync(inode);
187 ret = f2fs_write_inode(inode, NULL);
188 if (ret)
189 goto out;
190 }
191 filemap_fdatawait_range(sbi->node_inode->i_mapping,
192 0, LONG_MAX);
193 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
194 }
195 out:
196 mutex_unlock(&inode->i_mutex);
197 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
198 return ret;
199 }
200
201 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
202 {
203 file_accessed(file);
204 vma->vm_ops = &f2fs_file_vm_ops;
205 return 0;
206 }
207
208 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
209 {
210 int nr_free = 0, ofs = dn->ofs_in_node;
211 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
212 struct f2fs_node *raw_node;
213 __le32 *addr;
214
215 raw_node = F2FS_NODE(dn->node_page);
216 addr = blkaddr_in_node(raw_node) + ofs;
217
218 for ( ; count > 0; count--, addr++, dn->ofs_in_node++) {
219 block_t blkaddr = le32_to_cpu(*addr);
220 if (blkaddr == NULL_ADDR)
221 continue;
222
223 update_extent_cache(NULL_ADDR, dn);
224 invalidate_blocks(sbi, blkaddr);
225 nr_free++;
226 }
227 if (nr_free) {
228 dec_valid_block_count(sbi, dn->inode, nr_free);
229 set_page_dirty(dn->node_page);
230 sync_inode_page(dn);
231 }
232 dn->ofs_in_node = ofs;
233
234 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
235 dn->ofs_in_node, nr_free);
236 return nr_free;
237 }
238
239 void truncate_data_blocks(struct dnode_of_data *dn)
240 {
241 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
242 }
243
244 static void truncate_partial_data_page(struct inode *inode, u64 from)
245 {
246 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
247 struct page *page;
248
249 if (!offset)
250 return;
251
252 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false);
253 if (IS_ERR(page))
254 return;
255
256 lock_page(page);
257 if (page->mapping != inode->i_mapping) {
258 f2fs_put_page(page, 1);
259 return;
260 }
261 wait_on_page_writeback(page);
262 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
263 set_page_dirty(page);
264 f2fs_put_page(page, 1);
265 }
266
267 static int truncate_blocks(struct inode *inode, u64 from)
268 {
269 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
270 unsigned int blocksize = inode->i_sb->s_blocksize;
271 struct dnode_of_data dn;
272 pgoff_t free_from;
273 int count = 0, ilock = -1;
274 int err;
275
276 trace_f2fs_truncate_blocks_enter(inode, from);
277
278 free_from = (pgoff_t)
279 ((from + blocksize - 1) >> (sbi->log_blocksize));
280
281 ilock = mutex_lock_op(sbi);
282 set_new_dnode(&dn, inode, NULL, NULL, 0);
283 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
284 if (err) {
285 if (err == -ENOENT)
286 goto free_next;
287 mutex_unlock_op(sbi, ilock);
288 trace_f2fs_truncate_blocks_exit(inode, err);
289 return err;
290 }
291
292 if (IS_INODE(dn.node_page))
293 count = ADDRS_PER_INODE(F2FS_I(inode));
294 else
295 count = ADDRS_PER_BLOCK;
296
297 count -= dn.ofs_in_node;
298 BUG_ON(count < 0);
299
300 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
301 truncate_data_blocks_range(&dn, count);
302 free_from += count;
303 }
304
305 f2fs_put_dnode(&dn);
306 free_next:
307 err = truncate_inode_blocks(inode, free_from);
308 mutex_unlock_op(sbi, ilock);
309
310 /* lastly zero out the first data page */
311 truncate_partial_data_page(inode, from);
312
313 trace_f2fs_truncate_blocks_exit(inode, err);
314 return err;
315 }
316
317 void f2fs_truncate(struct inode *inode)
318 {
319 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
320 S_ISLNK(inode->i_mode)))
321 return;
322
323 trace_f2fs_truncate(inode);
324
325 if (!truncate_blocks(inode, i_size_read(inode))) {
326 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
327 mark_inode_dirty(inode);
328 }
329 }
330
331 int f2fs_getattr(struct vfsmount *mnt,
332 struct dentry *dentry, struct kstat *stat)
333 {
334 struct inode *inode = dentry->d_inode;
335 generic_fillattr(inode, stat);
336 stat->blocks <<= 3;
337 return 0;
338 }
339
340 #ifdef CONFIG_F2FS_FS_POSIX_ACL
341 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
342 {
343 struct f2fs_inode_info *fi = F2FS_I(inode);
344 unsigned int ia_valid = attr->ia_valid;
345
346 if (ia_valid & ATTR_UID)
347 inode->i_uid = attr->ia_uid;
348 if (ia_valid & ATTR_GID)
349 inode->i_gid = attr->ia_gid;
350 if (ia_valid & ATTR_ATIME)
351 inode->i_atime = timespec_trunc(attr->ia_atime,
352 inode->i_sb->s_time_gran);
353 if (ia_valid & ATTR_MTIME)
354 inode->i_mtime = timespec_trunc(attr->ia_mtime,
355 inode->i_sb->s_time_gran);
356 if (ia_valid & ATTR_CTIME)
357 inode->i_ctime = timespec_trunc(attr->ia_ctime,
358 inode->i_sb->s_time_gran);
359 if (ia_valid & ATTR_MODE) {
360 umode_t mode = attr->ia_mode;
361
362 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
363 mode &= ~S_ISGID;
364 set_acl_inode(fi, mode);
365 }
366 }
367 #else
368 #define __setattr_copy setattr_copy
369 #endif
370
371 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
372 {
373 struct inode *inode = dentry->d_inode;
374 struct f2fs_inode_info *fi = F2FS_I(inode);
375 int err;
376
377 err = inode_change_ok(inode, attr);
378 if (err)
379 return err;
380
381 if ((attr->ia_valid & ATTR_SIZE) &&
382 attr->ia_size != i_size_read(inode)) {
383 truncate_setsize(inode, attr->ia_size);
384 f2fs_truncate(inode);
385 f2fs_balance_fs(F2FS_SB(inode->i_sb));
386 }
387
388 __setattr_copy(inode, attr);
389
390 if (attr->ia_valid & ATTR_MODE) {
391 err = f2fs_acl_chmod(inode);
392 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
393 inode->i_mode = fi->i_acl_mode;
394 clear_inode_flag(fi, FI_ACL_MODE);
395 }
396 }
397
398 mark_inode_dirty(inode);
399 return err;
400 }
401
402 const struct inode_operations f2fs_file_inode_operations = {
403 .getattr = f2fs_getattr,
404 .setattr = f2fs_setattr,
405 .get_acl = f2fs_get_acl,
406 #ifdef CONFIG_F2FS_FS_XATTR
407 .setxattr = generic_setxattr,
408 .getxattr = generic_getxattr,
409 .listxattr = f2fs_listxattr,
410 .removexattr = generic_removexattr,
411 #endif
412 };
413
414 static void fill_zero(struct inode *inode, pgoff_t index,
415 loff_t start, loff_t len)
416 {
417 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
418 struct page *page;
419 int ilock;
420
421 if (!len)
422 return;
423
424 f2fs_balance_fs(sbi);
425
426 ilock = mutex_lock_op(sbi);
427 page = get_new_data_page(inode, NULL, index, false);
428 mutex_unlock_op(sbi, ilock);
429
430 if (!IS_ERR(page)) {
431 wait_on_page_writeback(page);
432 zero_user(page, start, len);
433 set_page_dirty(page);
434 f2fs_put_page(page, 1);
435 }
436 }
437
438 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
439 {
440 pgoff_t index;
441 int err;
442
443 for (index = pg_start; index < pg_end; index++) {
444 struct dnode_of_data dn;
445
446 set_new_dnode(&dn, inode, NULL, NULL, 0);
447 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
448 if (err) {
449 if (err == -ENOENT)
450 continue;
451 return err;
452 }
453
454 if (dn.data_blkaddr != NULL_ADDR)
455 truncate_data_blocks_range(&dn, 1);
456 f2fs_put_dnode(&dn);
457 }
458 return 0;
459 }
460
461 static int punch_hole(struct inode *inode, loff_t offset, loff_t len, int mode)
462 {
463 pgoff_t pg_start, pg_end;
464 loff_t off_start, off_end;
465 int ret = 0;
466
467 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
468 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
469
470 off_start = offset & (PAGE_CACHE_SIZE - 1);
471 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
472
473 if (pg_start == pg_end) {
474 fill_zero(inode, pg_start, off_start,
475 off_end - off_start);
476 } else {
477 if (off_start)
478 fill_zero(inode, pg_start++, off_start,
479 PAGE_CACHE_SIZE - off_start);
480 if (off_end)
481 fill_zero(inode, pg_end, 0, off_end);
482
483 if (pg_start < pg_end) {
484 struct address_space *mapping = inode->i_mapping;
485 loff_t blk_start, blk_end;
486 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
487 int ilock;
488
489 f2fs_balance_fs(sbi);
490
491 blk_start = pg_start << PAGE_CACHE_SHIFT;
492 blk_end = pg_end << PAGE_CACHE_SHIFT;
493 truncate_inode_pages_range(mapping, blk_start,
494 blk_end - 1);
495
496 ilock = mutex_lock_op(sbi);
497 ret = truncate_hole(inode, pg_start, pg_end);
498 mutex_unlock_op(sbi, ilock);
499 }
500 }
501
502 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
503 i_size_read(inode) <= (offset + len)) {
504 i_size_write(inode, offset);
505 mark_inode_dirty(inode);
506 }
507
508 return ret;
509 }
510
511 static int expand_inode_data(struct inode *inode, loff_t offset,
512 loff_t len, int mode)
513 {
514 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
515 pgoff_t index, pg_start, pg_end;
516 loff_t new_size = i_size_read(inode);
517 loff_t off_start, off_end;
518 int ret = 0;
519
520 ret = inode_newsize_ok(inode, (len + offset));
521 if (ret)
522 return ret;
523
524 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
525 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
526
527 off_start = offset & (PAGE_CACHE_SIZE - 1);
528 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
529
530 for (index = pg_start; index <= pg_end; index++) {
531 struct dnode_of_data dn;
532 int ilock;
533
534 ilock = mutex_lock_op(sbi);
535 set_new_dnode(&dn, inode, NULL, NULL, 0);
536 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
537 if (ret) {
538 mutex_unlock_op(sbi, ilock);
539 break;
540 }
541
542 if (dn.data_blkaddr == NULL_ADDR) {
543 ret = reserve_new_block(&dn);
544 if (ret) {
545 f2fs_put_dnode(&dn);
546 mutex_unlock_op(sbi, ilock);
547 break;
548 }
549 }
550 f2fs_put_dnode(&dn);
551 mutex_unlock_op(sbi, ilock);
552
553 if (pg_start == pg_end)
554 new_size = offset + len;
555 else if (index == pg_start && off_start)
556 new_size = (index + 1) << PAGE_CACHE_SHIFT;
557 else if (index == pg_end)
558 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
559 else
560 new_size += PAGE_CACHE_SIZE;
561 }
562
563 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
564 i_size_read(inode) < new_size) {
565 i_size_write(inode, new_size);
566 mark_inode_dirty(inode);
567 }
568
569 return ret;
570 }
571
572 static long f2fs_fallocate(struct file *file, int mode,
573 loff_t offset, loff_t len)
574 {
575 struct inode *inode = file_inode(file);
576 long ret;
577
578 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
579 return -EOPNOTSUPP;
580
581 if (mode & FALLOC_FL_PUNCH_HOLE)
582 ret = punch_hole(inode, offset, len, mode);
583 else
584 ret = expand_inode_data(inode, offset, len, mode);
585
586 if (!ret) {
587 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
588 mark_inode_dirty(inode);
589 }
590 trace_f2fs_fallocate(inode, mode, offset, len, ret);
591 return ret;
592 }
593
594 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
595 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
596
597 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
598 {
599 if (S_ISDIR(mode))
600 return flags;
601 else if (S_ISREG(mode))
602 return flags & F2FS_REG_FLMASK;
603 else
604 return flags & F2FS_OTHER_FLMASK;
605 }
606
607 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
608 {
609 struct inode *inode = file_inode(filp);
610 struct f2fs_inode_info *fi = F2FS_I(inode);
611 unsigned int flags;
612 int ret;
613
614 switch (cmd) {
615 case F2FS_IOC_GETFLAGS:
616 flags = fi->i_flags & FS_FL_USER_VISIBLE;
617 return put_user(flags, (int __user *) arg);
618 case F2FS_IOC_SETFLAGS:
619 {
620 unsigned int oldflags;
621
622 ret = mnt_want_write_file(filp);
623 if (ret)
624 return ret;
625
626 if (!inode_owner_or_capable(inode)) {
627 ret = -EACCES;
628 goto out;
629 }
630
631 if (get_user(flags, (int __user *) arg)) {
632 ret = -EFAULT;
633 goto out;
634 }
635
636 flags = f2fs_mask_flags(inode->i_mode, flags);
637
638 mutex_lock(&inode->i_mutex);
639
640 oldflags = fi->i_flags;
641
642 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
643 if (!capable(CAP_LINUX_IMMUTABLE)) {
644 mutex_unlock(&inode->i_mutex);
645 ret = -EPERM;
646 goto out;
647 }
648 }
649
650 flags = flags & FS_FL_USER_MODIFIABLE;
651 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
652 fi->i_flags = flags;
653 mutex_unlock(&inode->i_mutex);
654
655 f2fs_set_inode_flags(inode);
656 inode->i_ctime = CURRENT_TIME;
657 mark_inode_dirty(inode);
658 out:
659 mnt_drop_write_file(filp);
660 return ret;
661 }
662 default:
663 return -ENOTTY;
664 }
665 }
666
667 #ifdef CONFIG_COMPAT
668 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
669 {
670 switch (cmd) {
671 case F2FS_IOC32_GETFLAGS:
672 cmd = F2FS_IOC_GETFLAGS;
673 break;
674 case F2FS_IOC32_SETFLAGS:
675 cmd = F2FS_IOC_SETFLAGS;
676 break;
677 default:
678 return -ENOIOCTLCMD;
679 }
680 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
681 }
682 #endif
683
684 const struct file_operations f2fs_file_operations = {
685 .llseek = generic_file_llseek,
686 .read = do_sync_read,
687 .write = do_sync_write,
688 .aio_read = generic_file_aio_read,
689 .aio_write = generic_file_aio_write,
690 .open = generic_file_open,
691 .mmap = f2fs_file_mmap,
692 .fsync = f2fs_sync_file,
693 .fallocate = f2fs_fallocate,
694 .unlocked_ioctl = f2fs_ioctl,
695 #ifdef CONFIG_COMPAT
696 .compat_ioctl = f2fs_compat_ioctl,
697 #endif
698 .splice_read = generic_file_splice_read,
699 .splice_write = generic_file_splice_write,
700 };
Linus' kernel tree