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 if (fsync_mark_done(sbi, inode->i_ino))
181 mark_inode_dirty_sync(inode);
182 ret = f2fs_write_inode(inode, NULL);
186 ret = wait_on_node_pages_writeback(sbi, inode->i_ino);
189 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
192 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
196 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
199 vma->vm_ops = &f2fs_file_vm_ops;
203 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
205 int nr_free = 0, ofs = dn->ofs_in_node;
206 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
207 struct f2fs_node *raw_node;
210 raw_node = F2FS_NODE(dn->node_page);
211 addr = blkaddr_in_node(raw_node) + ofs;
213 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
214 block_t blkaddr = le32_to_cpu(*addr);
215 if (blkaddr == NULL_ADDR)
218 update_extent_cache(NULL_ADDR, dn);
219 invalidate_blocks(sbi, blkaddr);
223 dec_valid_block_count(sbi, dn->inode, nr_free);
224 set_page_dirty(dn->node_page);
227 dn->ofs_in_node = ofs;
229 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
230 dn->ofs_in_node, nr_free);
234 void truncate_data_blocks(struct dnode_of_data *dn)
236 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
239 static void truncate_partial_data_page(struct inode *inode, u64 from)
241 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
247 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false);
252 if (unlikely(page->mapping != inode->i_mapping)) {
253 f2fs_put_page(page, 1);
256 f2fs_wait_on_page_writeback(page, DATA);
257 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
258 set_page_dirty(page);
259 f2fs_put_page(page, 1);
262 int truncate_blocks(struct inode *inode, u64 from)
264 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
265 unsigned int blocksize = inode->i_sb->s_blocksize;
266 struct dnode_of_data dn;
268 int count = 0, err = 0;
270 trace_f2fs_truncate_blocks_enter(inode, from);
272 if (f2fs_has_inline_data(inode))
275 free_from = (pgoff_t)
276 ((from + blocksize - 1) >> (sbi->log_blocksize));
280 set_new_dnode(&dn, inode, NULL, NULL, 0);
281 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
286 trace_f2fs_truncate_blocks_exit(inode, err);
290 if (IS_INODE(dn.node_page))
291 count = ADDRS_PER_INODE(F2FS_I(inode));
293 count = ADDRS_PER_BLOCK;
295 count -= dn.ofs_in_node;
296 f2fs_bug_on(count < 0);
298 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
299 truncate_data_blocks_range(&dn, count);
305 err = truncate_inode_blocks(inode, free_from);
308 /* lastly zero out the first data page */
309 truncate_partial_data_page(inode, from);
311 trace_f2fs_truncate_blocks_exit(inode, err);
315 void f2fs_truncate(struct inode *inode)
317 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
318 S_ISLNK(inode->i_mode)))
321 trace_f2fs_truncate(inode);
323 if (!truncate_blocks(inode, i_size_read(inode))) {
324 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
325 mark_inode_dirty(inode);
329 int f2fs_getattr(struct vfsmount *mnt,
330 struct dentry *dentry, struct kstat *stat)
332 struct inode *inode = dentry->d_inode;
333 generic_fillattr(inode, stat);
338 #ifdef CONFIG_F2FS_FS_POSIX_ACL
339 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
341 struct f2fs_inode_info *fi = F2FS_I(inode);
342 unsigned int ia_valid = attr->ia_valid;
344 if (ia_valid & ATTR_UID)
345 inode->i_uid = attr->ia_uid;
346 if (ia_valid & ATTR_GID)
347 inode->i_gid = attr->ia_gid;
348 if (ia_valid & ATTR_ATIME)
349 inode->i_atime = timespec_trunc(attr->ia_atime,
350 inode->i_sb->s_time_gran);
351 if (ia_valid & ATTR_MTIME)
352 inode->i_mtime = timespec_trunc(attr->ia_mtime,
353 inode->i_sb->s_time_gran);
354 if (ia_valid & ATTR_CTIME)
355 inode->i_ctime = timespec_trunc(attr->ia_ctime,
356 inode->i_sb->s_time_gran);
357 if (ia_valid & ATTR_MODE) {
358 umode_t mode = attr->ia_mode;
360 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
362 set_acl_inode(fi, mode);
366 #define __setattr_copy setattr_copy
369 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
371 struct inode *inode = dentry->d_inode;
372 struct f2fs_inode_info *fi = F2FS_I(inode);
375 err = inode_change_ok(inode, attr);
379 if ((attr->ia_valid & ATTR_SIZE) &&
380 attr->ia_size != i_size_read(inode)) {
381 err = f2fs_convert_inline_data(inode, attr->ia_size);
385 truncate_setsize(inode, attr->ia_size);
386 f2fs_truncate(inode);
387 f2fs_balance_fs(F2FS_SB(inode->i_sb));
390 __setattr_copy(inode, attr);
392 if (attr->ia_valid & ATTR_MODE) {
393 err = posix_acl_chmod(inode, get_inode_mode(inode));
394 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
395 inode->i_mode = fi->i_acl_mode;
396 clear_inode_flag(fi, FI_ACL_MODE);
400 mark_inode_dirty(inode);
404 const struct inode_operations f2fs_file_inode_operations = {
405 .getattr = f2fs_getattr,
406 .setattr = f2fs_setattr,
407 .get_acl = f2fs_get_acl,
408 .set_acl = f2fs_set_acl,
409 #ifdef CONFIG_F2FS_FS_XATTR
410 .setxattr = generic_setxattr,
411 .getxattr = generic_getxattr,
412 .listxattr = f2fs_listxattr,
413 .removexattr = generic_removexattr,
417 static void fill_zero(struct inode *inode, pgoff_t index,
418 loff_t start, loff_t len)
420 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
426 f2fs_balance_fs(sbi);
429 page = get_new_data_page(inode, NULL, index, false);
433 f2fs_wait_on_page_writeback(page, DATA);
434 zero_user(page, start, len);
435 set_page_dirty(page);
436 f2fs_put_page(page, 1);
440 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
445 for (index = pg_start; index < pg_end; index++) {
446 struct dnode_of_data dn;
448 set_new_dnode(&dn, inode, NULL, NULL, 0);
449 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
456 if (dn.data_blkaddr != NULL_ADDR)
457 truncate_data_blocks_range(&dn, 1);
463 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
465 pgoff_t pg_start, pg_end;
466 loff_t off_start, off_end;
469 ret = f2fs_convert_inline_data(inode, MAX_INLINE_DATA + 1);
473 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
474 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
476 off_start = offset & (PAGE_CACHE_SIZE - 1);
477 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
479 if (pg_start == pg_end) {
480 fill_zero(inode, pg_start, off_start,
481 off_end - off_start);
484 fill_zero(inode, pg_start++, off_start,
485 PAGE_CACHE_SIZE - off_start);
487 fill_zero(inode, pg_end, 0, off_end);
489 if (pg_start < pg_end) {
490 struct address_space *mapping = inode->i_mapping;
491 loff_t blk_start, blk_end;
492 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
494 f2fs_balance_fs(sbi);
496 blk_start = pg_start << PAGE_CACHE_SHIFT;
497 blk_end = pg_end << PAGE_CACHE_SHIFT;
498 truncate_inode_pages_range(mapping, blk_start,
502 ret = truncate_hole(inode, pg_start, pg_end);
510 static int expand_inode_data(struct inode *inode, loff_t offset,
511 loff_t len, int mode)
513 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
514 pgoff_t index, pg_start, pg_end;
515 loff_t new_size = i_size_read(inode);
516 loff_t off_start, off_end;
519 ret = inode_newsize_ok(inode, (len + offset));
523 ret = f2fs_convert_inline_data(inode, offset + len);
527 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
528 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
530 off_start = offset & (PAGE_CACHE_SIZE - 1);
531 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
533 for (index = pg_start; index <= pg_end; index++) {
534 struct dnode_of_data dn;
537 set_new_dnode(&dn, inode, NULL, NULL, 0);
538 ret = f2fs_reserve_block(&dn, index);
543 if (pg_start == pg_end)
544 new_size = offset + len;
545 else if (index == pg_start && off_start)
546 new_size = (index + 1) << PAGE_CACHE_SHIFT;
547 else if (index == pg_end)
548 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
550 new_size += PAGE_CACHE_SIZE;
553 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
554 i_size_read(inode) < new_size) {
555 i_size_write(inode, new_size);
556 mark_inode_dirty(inode);
562 static long f2fs_fallocate(struct file *file, int mode,
563 loff_t offset, loff_t len)
565 struct inode *inode = file_inode(file);
568 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
571 mutex_lock(&inode->i_mutex);
573 if (mode & FALLOC_FL_PUNCH_HOLE)
574 ret = punch_hole(inode, offset, len);
576 ret = expand_inode_data(inode, offset, len, mode);
579 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
580 mark_inode_dirty(inode);
583 mutex_unlock(&inode->i_mutex);
585 trace_f2fs_fallocate(inode, mode, offset, len, ret);
589 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
590 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
592 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
596 else if (S_ISREG(mode))
597 return flags & F2FS_REG_FLMASK;
599 return flags & F2FS_OTHER_FLMASK;
602 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
604 struct inode *inode = file_inode(filp);
605 struct f2fs_inode_info *fi = F2FS_I(inode);
610 case F2FS_IOC_GETFLAGS:
611 flags = fi->i_flags & FS_FL_USER_VISIBLE;
612 return put_user(flags, (int __user *) arg);
613 case F2FS_IOC_SETFLAGS:
615 unsigned int oldflags;
617 ret = mnt_want_write_file(filp);
621 if (!inode_owner_or_capable(inode)) {
626 if (get_user(flags, (int __user *) arg)) {
631 flags = f2fs_mask_flags(inode->i_mode, flags);
633 mutex_lock(&inode->i_mutex);
635 oldflags = fi->i_flags;
637 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
638 if (!capable(CAP_LINUX_IMMUTABLE)) {
639 mutex_unlock(&inode->i_mutex);
645 flags = flags & FS_FL_USER_MODIFIABLE;
646 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
648 mutex_unlock(&inode->i_mutex);
650 f2fs_set_inode_flags(inode);
651 inode->i_ctime = CURRENT_TIME;
652 mark_inode_dirty(inode);
654 mnt_drop_write_file(filp);
663 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
666 case F2FS_IOC32_GETFLAGS:
667 cmd = F2FS_IOC_GETFLAGS;
669 case F2FS_IOC32_SETFLAGS:
670 cmd = F2FS_IOC_SETFLAGS;
675 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
679 const struct file_operations f2fs_file_operations = {
680 .llseek = generic_file_llseek,
681 .read = do_sync_read,
682 .write = do_sync_write,
683 .aio_read = generic_file_aio_read,
684 .aio_write = generic_file_aio_write,
685 .open = generic_file_open,
686 .mmap = f2fs_file_mmap,
687 .fsync = f2fs_sync_file,
688 .fallocate = f2fs_fallocate,
689 .unlocked_ioctl = f2fs_ioctl,
691 .compat_ioctl = f2fs_compat_ioctl,
693 .splice_read = generic_file_splice_read,
694 .splice_write = generic_file_splice_write,