projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
mm: implement ->map_pages for page cache
[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
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 struct dnode_of_data dn;
37 int err;
38
39 f2fs_balance_fs(sbi);
40
41 sb_start_pagefault(inode->i_sb);
42
43 /* block allocation */
44 f2fs_lock_op(sbi);
45 set_new_dnode(&dn, inode, NULL, NULL, 0);
46 err = f2fs_reserve_block(&dn, page->index);
47 f2fs_unlock_op(sbi);
48 if (err)
49 goto out;
50
51 file_update_time(vma->vm_file);
52 lock_page(page);
53 if (unlikely(page->mapping != inode->i_mapping ||
54 page_offset(page) > i_size_read(inode) ||
55 !PageUptodate(page))) {
56 unlock_page(page);
57 err = -EFAULT;
58 goto out;
59 }
60
61 /*
62 * check to see if the page is mapped already (no holes)
63 */
64 if (PageMappedToDisk(page))
65 goto mapped;
66
67 /* page is wholly or partially inside EOF */
68 if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
69 unsigned offset;
70 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
71 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
72 }
73 set_page_dirty(page);
74 SetPageUptodate(page);
75
76 trace_f2fs_vm_page_mkwrite(page, DATA);
77 mapped:
78 /* fill the page */
79 wait_on_page_writeback(page);
80 out:
81 sb_end_pagefault(inode->i_sb);
82 return block_page_mkwrite_return(err);
83 }
84
85 static const struct vm_operations_struct f2fs_file_vm_ops = {
86 .fault = filemap_fault,
87 .map_pages = filemap_map_pages,
88 .page_mkwrite = f2fs_vm_page_mkwrite,
89 .remap_pages = generic_file_remap_pages,
90 };
91
92 static int get_parent_ino(struct inode *inode, nid_t *pino)
93 {
94 struct dentry *dentry;
95
96 inode = igrab(inode);
97 dentry = d_find_any_alias(inode);
98 iput(inode);
99 if (!dentry)
100 return 0;
101
102 if (update_dent_inode(inode, &dentry->d_name)) {
103 dput(dentry);
104 return 0;
105 }
106
107 *pino = parent_ino(dentry);
108 dput(dentry);
109 return 1;
110 }
111
112 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
113 {
114 struct inode *inode = file->f_mapping->host;
115 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
116 int ret = 0;
117 bool need_cp = false;
118 struct writeback_control wbc = {
119 .sync_mode = WB_SYNC_NONE,
120 .nr_to_write = LONG_MAX,
121 .for_reclaim = 0,
122 };
123
124 if (unlikely(f2fs_readonly(inode->i_sb)))
125 return 0;
126
127 trace_f2fs_sync_file_enter(inode);
128 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
129 if (ret) {
130 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
131 return ret;
132 }
133
134 /* guarantee free sections for fsync */
135 f2fs_balance_fs(sbi);
136
137 mutex_lock(&inode->i_mutex);
138
139 /*
140 * Both of fdatasync() and fsync() are able to be recovered from
141 * sudden-power-off.
142 */
143 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
144 need_cp = true;
145 else if (file_wrong_pino(inode))
146 need_cp = true;
147 else if (!space_for_roll_forward(sbi))
148 need_cp = true;
149 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
150 need_cp = true;
151 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
152 need_cp = true;
153
154 if (need_cp) {
155 nid_t pino;
156
157 F2FS_I(inode)->xattr_ver = 0;
158
159 /* all the dirty node pages should be flushed for POR */
160 ret = f2fs_sync_fs(inode->i_sb, 1);
161 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
162 get_parent_ino(inode, &pino)) {
163 F2FS_I(inode)->i_pino = pino;
164 file_got_pino(inode);
165 mark_inode_dirty_sync(inode);
166 ret = f2fs_write_inode(inode, NULL);
167 if (ret)
168 goto out;
169 }
170 } else {
171 /* if there is no written node page, write its inode page */
172 while (!sync_node_pages(sbi, inode->i_ino, &wbc)) {
173 mark_inode_dirty_sync(inode);
174 ret = f2fs_write_inode(inode, NULL);
175 if (ret)
176 goto out;
177 }
178 ret = wait_on_node_pages_writeback(sbi, inode->i_ino);
179 if (ret)
180 goto out;
181 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
182 }
183 out:
184 mutex_unlock(&inode->i_mutex);
185 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
186 return ret;
187 }
188
189 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
190 {
191 file_accessed(file);
192 vma->vm_ops = &f2fs_file_vm_ops;
193 return 0;
194 }
195
196 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
197 {
198 int nr_free = 0, ofs = dn->ofs_in_node;
199 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
200 struct f2fs_node *raw_node;
201 __le32 *addr;
202
203 raw_node = F2FS_NODE(dn->node_page);
204 addr = blkaddr_in_node(raw_node) + ofs;
205
206 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
207 block_t blkaddr = le32_to_cpu(*addr);
208 if (blkaddr == NULL_ADDR)
209 continue;
210
211 update_extent_cache(NULL_ADDR, dn);
212 invalidate_blocks(sbi, blkaddr);
213 nr_free++;
214 }
215 if (nr_free) {
216 dec_valid_block_count(sbi, dn->inode, nr_free);
217 set_page_dirty(dn->node_page);
218 sync_inode_page(dn);
219 }
220 dn->ofs_in_node = ofs;
221
222 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
223 dn->ofs_in_node, nr_free);
224 return nr_free;
225 }
226
227 void truncate_data_blocks(struct dnode_of_data *dn)
228 {
229 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
230 }
231
232 static void truncate_partial_data_page(struct inode *inode, u64 from)
233 {
234 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
235 struct page *page;
236
237 if (!offset)
238 return;
239
240 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false);
241 if (IS_ERR(page))
242 return;
243
244 lock_page(page);
245 if (unlikely(page->mapping != inode->i_mapping)) {
246 f2fs_put_page(page, 1);
247 return;
248 }
249 wait_on_page_writeback(page);
250 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
251 set_page_dirty(page);
252 f2fs_put_page(page, 1);
253 }
254
255 int truncate_blocks(struct inode *inode, u64 from)
256 {
257 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
258 unsigned int blocksize = inode->i_sb->s_blocksize;
259 struct dnode_of_data dn;
260 pgoff_t free_from;
261 int count = 0, err = 0;
262
263 trace_f2fs_truncate_blocks_enter(inode, from);
264
265 if (f2fs_has_inline_data(inode))
266 goto done;
267
268 free_from = (pgoff_t)
269 ((from + blocksize - 1) >> (sbi->log_blocksize));
270
271 f2fs_lock_op(sbi);
272
273 set_new_dnode(&dn, inode, NULL, NULL, 0);
274 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
275 if (err) {
276 if (err == -ENOENT)
277 goto free_next;
278 f2fs_unlock_op(sbi);
279 trace_f2fs_truncate_blocks_exit(inode, err);
280 return err;
281 }
282
283 if (IS_INODE(dn.node_page))
284 count = ADDRS_PER_INODE(F2FS_I(inode));
285 else
286 count = ADDRS_PER_BLOCK;
287
288 count -= dn.ofs_in_node;
289 f2fs_bug_on(count < 0);
290
291 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
292 truncate_data_blocks_range(&dn, count);
293 free_from += count;
294 }
295
296 f2fs_put_dnode(&dn);
297 free_next:
298 err = truncate_inode_blocks(inode, free_from);
299 f2fs_unlock_op(sbi);
300 done:
301 /* lastly zero out the first data page */
302 truncate_partial_data_page(inode, from);
303
304 trace_f2fs_truncate_blocks_exit(inode, err);
305 return err;
306 }
307
308 void f2fs_truncate(struct inode *inode)
309 {
310 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
311 S_ISLNK(inode->i_mode)))
312 return;
313
314 trace_f2fs_truncate(inode);
315
316 if (!truncate_blocks(inode, i_size_read(inode))) {
317 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
318 mark_inode_dirty(inode);
319 }
320 }
321
322 int f2fs_getattr(struct vfsmount *mnt,
323 struct dentry *dentry, struct kstat *stat)
324 {
325 struct inode *inode = dentry->d_inode;
326 generic_fillattr(inode, stat);
327 stat->blocks <<= 3;
328 return 0;
329 }
330
331 #ifdef CONFIG_F2FS_FS_POSIX_ACL
332 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
333 {
334 struct f2fs_inode_info *fi = F2FS_I(inode);
335 unsigned int ia_valid = attr->ia_valid;
336
337 if (ia_valid & ATTR_UID)
338 inode->i_uid = attr->ia_uid;
339 if (ia_valid & ATTR_GID)
340 inode->i_gid = attr->ia_gid;
341 if (ia_valid & ATTR_ATIME)
342 inode->i_atime = timespec_trunc(attr->ia_atime,
343 inode->i_sb->s_time_gran);
344 if (ia_valid & ATTR_MTIME)
345 inode->i_mtime = timespec_trunc(attr->ia_mtime,
346 inode->i_sb->s_time_gran);
347 if (ia_valid & ATTR_CTIME)
348 inode->i_ctime = timespec_trunc(attr->ia_ctime,
349 inode->i_sb->s_time_gran);
350 if (ia_valid & ATTR_MODE) {
351 umode_t mode = attr->ia_mode;
352
353 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
354 mode &= ~S_ISGID;
355 set_acl_inode(fi, mode);
356 }
357 }
358 #else
359 #define __setattr_copy setattr_copy
360 #endif
361
362 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
363 {
364 struct inode *inode = dentry->d_inode;
365 struct f2fs_inode_info *fi = F2FS_I(inode);
366 int err;
367
368 err = inode_change_ok(inode, attr);
369 if (err)
370 return err;
371
372 if ((attr->ia_valid & ATTR_SIZE) &&
373 attr->ia_size != i_size_read(inode)) {
374 err = f2fs_convert_inline_data(inode, attr->ia_size);
375 if (err)
376 return err;
377
378 truncate_setsize(inode, attr->ia_size);
379 f2fs_truncate(inode);
380 f2fs_balance_fs(F2FS_SB(inode->i_sb));
381 }
382
383 __setattr_copy(inode, attr);
384
385 if (attr->ia_valid & ATTR_MODE) {
386 err = posix_acl_chmod(inode, get_inode_mode(inode));
387 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
388 inode->i_mode = fi->i_acl_mode;
389 clear_inode_flag(fi, FI_ACL_MODE);
390 }
391 }
392
393 mark_inode_dirty(inode);
394 return err;
395 }
396
397 const struct inode_operations f2fs_file_inode_operations = {
398 .getattr = f2fs_getattr,
399 .setattr = f2fs_setattr,
400 .get_acl = f2fs_get_acl,
401 .set_acl = f2fs_set_acl,
402 #ifdef CONFIG_F2FS_FS_XATTR
403 .setxattr = generic_setxattr,
404 .getxattr = generic_getxattr,
405 .listxattr = f2fs_listxattr,
406 .removexattr = generic_removexattr,
407 #endif
408 };
409
410 static void fill_zero(struct inode *inode, pgoff_t index,
411 loff_t start, loff_t len)
412 {
413 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
414 struct page *page;
415
416 if (!len)
417 return;
418
419 f2fs_balance_fs(sbi);
420
421 f2fs_lock_op(sbi);
422 page = get_new_data_page(inode, NULL, index, false);
423 f2fs_unlock_op(sbi);
424
425 if (!IS_ERR(page)) {
426 wait_on_page_writeback(page);
427 zero_user(page, start, len);
428 set_page_dirty(page);
429 f2fs_put_page(page, 1);
430 }
431 }
432
433 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
434 {
435 pgoff_t index;
436 int err;
437
438 for (index = pg_start; index < pg_end; index++) {
439 struct dnode_of_data dn;
440
441 set_new_dnode(&dn, inode, NULL, NULL, 0);
442 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
443 if (err) {
444 if (err == -ENOENT)
445 continue;
446 return err;
447 }
448
449 if (dn.data_blkaddr != NULL_ADDR)
450 truncate_data_blocks_range(&dn, 1);
451 f2fs_put_dnode(&dn);
452 }
453 return 0;
454 }
455
456 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
457 {
458 pgoff_t pg_start, pg_end;
459 loff_t off_start, off_end;
460 int ret = 0;
461
462 ret = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1);
463 if (ret)
464 return ret;
465
466 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
467 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
468
469 off_start = offset & (PAGE_CACHE_SIZE - 1);
470 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
471
472 if (pg_start == pg_end) {
473 fill_zero(inode, pg_start, off_start,
474 off_end - off_start);
475 } else {
476 if (off_start)
477 fill_zero(inode, pg_start++, off_start,
478 PAGE_CACHE_SIZE - off_start);
479 if (off_end)
480 fill_zero(inode, pg_end, 0, off_end);
481
482 if (pg_start < pg_end) {
483 struct address_space *mapping = inode->i_mapping;
484 loff_t blk_start, blk_end;
485 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
486
487 f2fs_balance_fs(sbi);
488
489 blk_start = pg_start << PAGE_CACHE_SHIFT;
490 blk_end = pg_end << PAGE_CACHE_SHIFT;
491 truncate_inode_pages_range(mapping, blk_start,
492 blk_end - 1);
493
494 f2fs_lock_op(sbi);
495 ret = truncate_hole(inode, pg_start, pg_end);
496 f2fs_unlock_op(sbi);
497 }
498 }
499
500 return ret;
501 }
502
503 static int expand_inode_data(struct inode *inode, loff_t offset,
504 loff_t len, int mode)
505 {
506 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
507 pgoff_t index, pg_start, pg_end;
508 loff_t new_size = i_size_read(inode);
509 loff_t off_start, off_end;
510 int ret = 0;
511
512 ret = inode_newsize_ok(inode, (len + offset));
513 if (ret)
514 return ret;
515
516 ret = f2fs_convert_inline_data(inode, offset + len);
517 if (ret)
518 return ret;
519
520 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
521 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
522
523 off_start = offset & (PAGE_CACHE_SIZE - 1);
524 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
525
526 for (index = pg_start; index <= pg_end; index++) {
527 struct dnode_of_data dn;
528
529 f2fs_lock_op(sbi);
530 set_new_dnode(&dn, inode, NULL, NULL, 0);
531 ret = f2fs_reserve_block(&dn, index);
532 f2fs_unlock_op(sbi);
533 if (ret)
534 break;
535
536 if (pg_start == pg_end)
537 new_size = offset + len;
538 else if (index == pg_start && off_start)
539 new_size = (index + 1) << PAGE_CACHE_SHIFT;
540 else if (index == pg_end)
541 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
542 else
543 new_size += PAGE_CACHE_SIZE;
544 }
545
546 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
547 i_size_read(inode) < new_size) {
548 i_size_write(inode, new_size);
549 mark_inode_dirty(inode);
550 }
551
552 return ret;
553 }
554
555 static long f2fs_fallocate(struct file *file, int mode,
556 loff_t offset, loff_t len)
557 {
558 struct inode *inode = file_inode(file);
559 long ret;
560
561 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
562 return -EOPNOTSUPP;
563
564 if (mode & FALLOC_FL_PUNCH_HOLE)
565 ret = punch_hole(inode, offset, len);
566 else
567 ret = expand_inode_data(inode, offset, len, mode);
568
569 if (!ret) {
570 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
571 mark_inode_dirty(inode);
572 }
573 trace_f2fs_fallocate(inode, mode, offset, len, ret);
574 return ret;
575 }
576
577 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
578 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
579
580 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
581 {
582 if (S_ISDIR(mode))
583 return flags;
584 else if (S_ISREG(mode))
585 return flags & F2FS_REG_FLMASK;
586 else
587 return flags & F2FS_OTHER_FLMASK;
588 }
589
590 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
591 {
592 struct inode *inode = file_inode(filp);
593 struct f2fs_inode_info *fi = F2FS_I(inode);
594 unsigned int flags;
595 int ret;
596
597 switch (cmd) {
598 case F2FS_IOC_GETFLAGS:
599 flags = fi->i_flags & FS_FL_USER_VISIBLE;
600 return put_user(flags, (int __user *) arg);
601 case F2FS_IOC_SETFLAGS:
602 {
603 unsigned int oldflags;
604
605 ret = mnt_want_write_file(filp);
606 if (ret)
607 return ret;
608
609 if (!inode_owner_or_capable(inode)) {
610 ret = -EACCES;
611 goto out;
612 }
613
614 if (get_user(flags, (int __user *) arg)) {
615 ret = -EFAULT;
616 goto out;
617 }
618
619 flags = f2fs_mask_flags(inode->i_mode, flags);
620
621 mutex_lock(&inode->i_mutex);
622
623 oldflags = fi->i_flags;
624
625 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
626 if (!capable(CAP_LINUX_IMMUTABLE)) {
627 mutex_unlock(&inode->i_mutex);
628 ret = -EPERM;
629 goto out;
630 }
631 }
632
633 flags = flags & FS_FL_USER_MODIFIABLE;
634 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
635 fi->i_flags = flags;
636 mutex_unlock(&inode->i_mutex);
637
638 f2fs_set_inode_flags(inode);
639 inode->i_ctime = CURRENT_TIME;
640 mark_inode_dirty(inode);
641 out:
642 mnt_drop_write_file(filp);
643 return ret;
644 }
645 default:
646 return -ENOTTY;
647 }
648 }
649
650 #ifdef CONFIG_COMPAT
651 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
652 {
653 switch (cmd) {
654 case F2FS_IOC32_GETFLAGS:
655 cmd = F2FS_IOC_GETFLAGS;
656 break;
657 case F2FS_IOC32_SETFLAGS:
658 cmd = F2FS_IOC_SETFLAGS;
659 break;
660 default:
661 return -ENOIOCTLCMD;
662 }
663 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
664 }
665 #endif
666
667 const struct file_operations f2fs_file_operations = {
668 .llseek = generic_file_llseek,
669 .read = do_sync_read,
670 .write = do_sync_write,
671 .aio_read = generic_file_aio_read,
672 .aio_write = generic_file_aio_write,
673 .open = generic_file_open,
674 .mmap = f2fs_file_mmap,
675 .fsync = f2fs_sync_file,
676 .fallocate = f2fs_fallocate,
677 .unlocked_ioctl = f2fs_ioctl,
678 #ifdef CONFIG_COMPAT
679 .compat_ioctl = f2fs_compat_ioctl,
680 #endif
681 .splice_read = generic_file_splice_read,
682 .splice_write = generic_file_splice_write,
683 };
Linux kernel - Deliverables
This page took 0.043158 seconds and 5 git commands to generate.