4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
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.
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>
28 #include <trace/events/f2fs.h>
30 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
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;
41 sb_start_pagefault(inode->i_sb);
43 /* block allocation */
45 set_new_dnode(&dn, inode, NULL, NULL, 0);
46 err = f2fs_reserve_block(&dn, page->index);
51 file_update_time(vma->vm_file);
53 if (unlikely(page->mapping != inode->i_mapping ||
54 page_offset(page) > i_size_read(inode) ||
55 !PageUptodate(page))) {
62 * check to see if the page is mapped already (no holes)
64 if (PageMappedToDisk(page))
67 /* page is wholly or partially inside EOF */
68 if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
70 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
71 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
74 SetPageUptodate(page);
76 trace_f2fs_vm_page_mkwrite(page, DATA);
79 f2fs_wait_on_page_writeback(page, DATA);
81 sb_end_pagefault(inode->i_sb);
82 return block_page_mkwrite_return(err);
85 static const struct vm_operations_struct f2fs_file_vm_ops = {
86 .fault = filemap_fault,
87 .page_mkwrite = f2fs_vm_page_mkwrite,
88 .remap_pages = generic_file_remap_pages,
91 static int get_parent_ino(struct inode *inode, nid_t *pino)
93 struct dentry *dentry;
96 dentry = d_find_any_alias(inode);
101 if (update_dent_inode(inode, &dentry->d_name)) {
106 *pino = parent_ino(dentry);
111 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
113 struct inode *inode = file->f_mapping->host;
114 struct f2fs_inode_info *fi = F2FS_I(inode);
115 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
117 bool need_cp = false;
118 struct writeback_control wbc = {
119 .sync_mode = WB_SYNC_ALL,
120 .nr_to_write = LONG_MAX,
124 if (unlikely(f2fs_readonly(inode->i_sb)))
127 trace_f2fs_sync_file_enter(inode);
128 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
130 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
134 /* guarantee free sections for fsync */
135 f2fs_balance_fs(sbi);
137 down_read(&fi->i_sem);
140 * Both of fdatasync() and fsync() are able to be recovered from
143 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
145 else if (file_wrong_pino(inode))
147 else if (!space_for_roll_forward(sbi))
149 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
151 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
159 /* all the dirty node pages should be flushed for POR */
160 ret = f2fs_sync_fs(inode->i_sb, 1);
162 down_write(&fi->i_sem);
163 F2FS_I(inode)->xattr_ver = 0;
164 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
165 get_parent_ino(inode, &pino)) {
166 F2FS_I(inode)->i_pino = pino;
167 file_got_pino(inode);
168 up_write(&fi->i_sem);
169 mark_inode_dirty_sync(inode);
170 ret = f2fs_write_inode(inode, NULL);
174 up_write(&fi->i_sem);
177 /* if there is no written node page, write its inode page */
178 while (!sync_node_pages(sbi, inode->i_ino, &wbc)) {
179 mark_inode_dirty_sync(inode);
180 ret = f2fs_write_inode(inode, NULL);
184 ret = wait_on_node_pages_writeback(sbi, inode->i_ino);
187 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
190 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
194 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
197 vma->vm_ops = &f2fs_file_vm_ops;
201 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
203 int nr_free = 0, ofs = dn->ofs_in_node;
204 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
205 struct f2fs_node *raw_node;
208 raw_node = F2FS_NODE(dn->node_page);
209 addr = blkaddr_in_node(raw_node) + ofs;
211 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
212 block_t blkaddr = le32_to_cpu(*addr);
213 if (blkaddr == NULL_ADDR)
216 update_extent_cache(NULL_ADDR, dn);
217 invalidate_blocks(sbi, blkaddr);
221 dec_valid_block_count(sbi, dn->inode, nr_free);
222 set_page_dirty(dn->node_page);
225 dn->ofs_in_node = ofs;
227 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
228 dn->ofs_in_node, nr_free);
232 void truncate_data_blocks(struct dnode_of_data *dn)
234 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
237 static void truncate_partial_data_page(struct inode *inode, u64 from)
239 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
245 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false);
250 if (unlikely(page->mapping != inode->i_mapping)) {
251 f2fs_put_page(page, 1);
254 f2fs_wait_on_page_writeback(page, DATA);
255 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
256 set_page_dirty(page);
257 f2fs_put_page(page, 1);
260 int truncate_blocks(struct inode *inode, u64 from)
262 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
263 unsigned int blocksize = inode->i_sb->s_blocksize;
264 struct dnode_of_data dn;
266 int count = 0, err = 0;
268 trace_f2fs_truncate_blocks_enter(inode, from);
270 if (f2fs_has_inline_data(inode))
273 free_from = (pgoff_t)
274 ((from + blocksize - 1) >> (sbi->log_blocksize));
278 set_new_dnode(&dn, inode, NULL, NULL, 0);
279 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
284 trace_f2fs_truncate_blocks_exit(inode, err);
288 if (IS_INODE(dn.node_page))
289 count = ADDRS_PER_INODE(F2FS_I(inode));
291 count = ADDRS_PER_BLOCK;
293 count -= dn.ofs_in_node;
294 f2fs_bug_on(count < 0);
296 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
297 truncate_data_blocks_range(&dn, count);
303 err = truncate_inode_blocks(inode, free_from);
306 /* lastly zero out the first data page */
307 truncate_partial_data_page(inode, from);
309 trace_f2fs_truncate_blocks_exit(inode, err);
313 void f2fs_truncate(struct inode *inode)
315 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
316 S_ISLNK(inode->i_mode)))
319 trace_f2fs_truncate(inode);
321 if (!truncate_blocks(inode, i_size_read(inode))) {
322 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
323 mark_inode_dirty(inode);
327 int f2fs_getattr(struct vfsmount *mnt,
328 struct dentry *dentry, struct kstat *stat)
330 struct inode *inode = dentry->d_inode;
331 generic_fillattr(inode, stat);
336 #ifdef CONFIG_F2FS_FS_POSIX_ACL
337 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
339 struct f2fs_inode_info *fi = F2FS_I(inode);
340 unsigned int ia_valid = attr->ia_valid;
342 if (ia_valid & ATTR_UID)
343 inode->i_uid = attr->ia_uid;
344 if (ia_valid & ATTR_GID)
345 inode->i_gid = attr->ia_gid;
346 if (ia_valid & ATTR_ATIME)
347 inode->i_atime = timespec_trunc(attr->ia_atime,
348 inode->i_sb->s_time_gran);
349 if (ia_valid & ATTR_MTIME)
350 inode->i_mtime = timespec_trunc(attr->ia_mtime,
351 inode->i_sb->s_time_gran);
352 if (ia_valid & ATTR_CTIME)
353 inode->i_ctime = timespec_trunc(attr->ia_ctime,
354 inode->i_sb->s_time_gran);
355 if (ia_valid & ATTR_MODE) {
356 umode_t mode = attr->ia_mode;
358 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
360 set_acl_inode(fi, mode);
364 #define __setattr_copy setattr_copy
367 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
369 struct inode *inode = dentry->d_inode;
370 struct f2fs_inode_info *fi = F2FS_I(inode);
373 err = inode_change_ok(inode, attr);
377 if ((attr->ia_valid & ATTR_SIZE) &&
378 attr->ia_size != i_size_read(inode)) {
379 err = f2fs_convert_inline_data(inode, attr->ia_size);
383 truncate_setsize(inode, attr->ia_size);
384 f2fs_truncate(inode);
385 f2fs_balance_fs(F2FS_SB(inode->i_sb));
388 __setattr_copy(inode, attr);
390 if (attr->ia_valid & ATTR_MODE) {
391 err = posix_acl_chmod(inode, get_inode_mode(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);
398 mark_inode_dirty(inode);
402 const struct inode_operations f2fs_file_inode_operations = {
403 .getattr = f2fs_getattr,
404 .setattr = f2fs_setattr,
405 .get_acl = f2fs_get_acl,
406 .set_acl = f2fs_set_acl,
407 #ifdef CONFIG_F2FS_FS_XATTR
408 .setxattr = generic_setxattr,
409 .getxattr = generic_getxattr,
410 .listxattr = f2fs_listxattr,
411 .removexattr = generic_removexattr,
415 static void fill_zero(struct inode *inode, pgoff_t index,
416 loff_t start, loff_t len)
418 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
424 f2fs_balance_fs(sbi);
427 page = get_new_data_page(inode, NULL, index, false);
431 f2fs_wait_on_page_writeback(page, DATA);
432 zero_user(page, start, len);
433 set_page_dirty(page);
434 f2fs_put_page(page, 1);
438 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
443 for (index = pg_start; index < pg_end; index++) {
444 struct dnode_of_data dn;
446 set_new_dnode(&dn, inode, NULL, NULL, 0);
447 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
454 if (dn.data_blkaddr != NULL_ADDR)
455 truncate_data_blocks_range(&dn, 1);
461 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
463 pgoff_t pg_start, pg_end;
464 loff_t off_start, off_end;
467 ret = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1);
471 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
472 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
474 off_start = offset & (PAGE_CACHE_SIZE - 1);
475 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
477 if (pg_start == pg_end) {
478 fill_zero(inode, pg_start, off_start,
479 off_end - off_start);
482 fill_zero(inode, pg_start++, off_start,
483 PAGE_CACHE_SIZE - off_start);
485 fill_zero(inode, pg_end, 0, off_end);
487 if (pg_start < pg_end) {
488 struct address_space *mapping = inode->i_mapping;
489 loff_t blk_start, blk_end;
490 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
492 f2fs_balance_fs(sbi);
494 blk_start = pg_start << PAGE_CACHE_SHIFT;
495 blk_end = pg_end << PAGE_CACHE_SHIFT;
496 truncate_inode_pages_range(mapping, blk_start,
500 ret = truncate_hole(inode, pg_start, pg_end);
508 static int expand_inode_data(struct inode *inode, loff_t offset,
509 loff_t len, int mode)
511 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
512 pgoff_t index, pg_start, pg_end;
513 loff_t new_size = i_size_read(inode);
514 loff_t off_start, off_end;
517 ret = inode_newsize_ok(inode, (len + offset));
521 ret = f2fs_convert_inline_data(inode, offset + len);
525 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
526 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
528 off_start = offset & (PAGE_CACHE_SIZE - 1);
529 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
531 for (index = pg_start; index <= pg_end; index++) {
532 struct dnode_of_data dn;
535 set_new_dnode(&dn, inode, NULL, NULL, 0);
536 ret = f2fs_reserve_block(&dn, index);
541 if (pg_start == pg_end)
542 new_size = offset + len;
543 else if (index == pg_start && off_start)
544 new_size = (index + 1) << PAGE_CACHE_SHIFT;
545 else if (index == pg_end)
546 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
548 new_size += PAGE_CACHE_SIZE;
551 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
552 i_size_read(inode) < new_size) {
553 i_size_write(inode, new_size);
554 mark_inode_dirty(inode);
560 static long f2fs_fallocate(struct file *file, int mode,
561 loff_t offset, loff_t len)
563 struct inode *inode = file_inode(file);
566 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
569 mutex_lock(&inode->i_mutex);
571 if (mode & FALLOC_FL_PUNCH_HOLE)
572 ret = punch_hole(inode, offset, len);
574 ret = expand_inode_data(inode, offset, len, mode);
577 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
578 mark_inode_dirty(inode);
581 mutex_unlock(&inode->i_mutex);
583 trace_f2fs_fallocate(inode, mode, offset, len, ret);
587 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
588 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
590 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
594 else if (S_ISREG(mode))
595 return flags & F2FS_REG_FLMASK;
597 return flags & F2FS_OTHER_FLMASK;
600 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
602 struct inode *inode = file_inode(filp);
603 struct f2fs_inode_info *fi = F2FS_I(inode);
608 case F2FS_IOC_GETFLAGS:
609 flags = fi->i_flags & FS_FL_USER_VISIBLE;
610 return put_user(flags, (int __user *) arg);
611 case F2FS_IOC_SETFLAGS:
613 unsigned int oldflags;
615 ret = mnt_want_write_file(filp);
619 if (!inode_owner_or_capable(inode)) {
624 if (get_user(flags, (int __user *) arg)) {
629 flags = f2fs_mask_flags(inode->i_mode, flags);
631 mutex_lock(&inode->i_mutex);
633 oldflags = fi->i_flags;
635 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
636 if (!capable(CAP_LINUX_IMMUTABLE)) {
637 mutex_unlock(&inode->i_mutex);
643 flags = flags & FS_FL_USER_MODIFIABLE;
644 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
646 mutex_unlock(&inode->i_mutex);
648 f2fs_set_inode_flags(inode);
649 inode->i_ctime = CURRENT_TIME;
650 mark_inode_dirty(inode);
652 mnt_drop_write_file(filp);
661 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
664 case F2FS_IOC32_GETFLAGS:
665 cmd = F2FS_IOC_GETFLAGS;
667 case F2FS_IOC32_SETFLAGS:
668 cmd = F2FS_IOC_SETFLAGS;
673 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
677 const struct file_operations f2fs_file_operations = {
678 .llseek = generic_file_llseek,
679 .read = do_sync_read,
680 .write = do_sync_write,
681 .aio_read = generic_file_aio_read,
682 .aio_write = generic_file_aio_write,
683 .open = generic_file_open,
684 .mmap = f2fs_file_mmap,
685 .fsync = f2fs_sync_file,
686 .fallocate = f2fs_fallocate,
687 .unlocked_ioctl = f2fs_ioctl,
689 .compat_ioctl = f2fs_compat_ioctl,
691 .splice_read = generic_file_splice_read,
692 .splice_write = generic_file_splice_write,